Chemical and enzymatic tools to study proteins in their native
cellular environment
By
Jennifer Zhengzheng Yao
B.S. in Chemical Biology (2008)
University of California, Berkeley
MASSACHUSTS INS'Tf WE1OF TECHNOLOGY
JUL 01 2013
LIBRARIES
Submitted to the Department of Chemistry
in Partial Fulfillment of the Requirements for the
Degree of Doctor of Philosophy
at the
Massachusetts Institute of Technology
June 2013
0 2013 Massachusetts Institute of Technology
All rights reserved
Signature of the Author:
Department of Chemistry
May 24, 2013
Certified by:
f-1 Iice Y. Ting
Ellen Swallow Richards Associate Professor of Chemistry
Thesis supervisor
Accepted by:
Robert W. Field
Chairman, Departmental Committee on Graduate Students
This doctoral thesis has been examined by a committee of the Department of Chemistry
as follows:
JoAnne Stubbe
Novartis Professor of Chemistry
Professor of Biology
Thesis committee chair
Lfice Y. Ting
Ellen Swallow Richards Associate Professor of Chemistry
Thesis advisor
Alexander M. Klibanov
Firmenich Professor of Chemistry and Bioengineering
Thesis committee member
2
Chemical and enzymatic tools to study proteins in their native
cellular environment
By
Jennifer Zhengzheng Yao
Submitted to the Department of Chemistry
on May 24, 2013 in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy
ABSTRACT
A detailed understating of living systems requires methods to probe molecular
processes in cells and whole organisms. A set of technologies that combines chemical
and genetic probes have been developed to address the need for dynamic and noninvasive
assay of biological processes. In addition to be able to visualize the localization,
trafficking, and turnover of individual proteins, strategies that allow the tagging, and
imaging, and identification of entire proteomes have also offered valuable insights into
disease biology.
Since protein visualization serves as a complement to protein identification, this
thesis first describes the development of a protein labeling technique that is able to
specifically target diverse fluorophores to proteins inside live cells. The methodology
uses the E. coli lipoic acid ligase (LplA) that we have engineered to accept and ligate an
azide functional handle onto a 13-amino acid LplA acceptor peptide (LAP). Subsequent
derivatization of the azide with fluorophores functionalized with cyclooctyne via strain-
promoted azide-alkyne cycloaddition allowed us to target many bright and photostable
fluorophores that could be used in super resolution imaging. Due to the numerous
applications to which cyclooctynes are being applied, our observation of the behavior of
different cyclooctynes inside cells should also prove useful to the protein labeling
community and beyond.
For protein identification, we describe our work of engineering and using LplA to
site-specifically target a benzophenone photocrosslinker. Our observations led us to the
conclusion that although benzophenone is generally regarded as the more efficient and
specific photocrosslinker than aryl azide and diazirine, its high geometric constraint to its
proximal crosslinkable C-H bonds may decrease its crosslinking yield. Knowing the
protein structure and amino acid environment surrounding benzophenone could help in
choosing the most optimal position for the photocrosslinker. Finally, in a different
crosslinking approach, we discuss our effort towards using a promiscuous peroxidase
enzyme that generates biotin-phenol radicals to study membrane protein topology.
Thesis Advisor: Alice Y. Ting
Title: Ellen Swallow Richards Associate Professor of Chemistry
Acknowledgements
I would like to thank my advisor Professor Alice Ting for her support and inspiration
over the past five years. Her passion, relentlessness, and perseverance towards tackling
some of the most challenging problems in science have always been a source of
motivation during my PhD study. I would like to thank her for her patience and
encouragement during times when nothing was working. Her creative ideas and solutions
have always inspired and encouraged me to think outside of the box to gain different
perspectives.
I am thankful to the members of my thesis committee: Professor Alex Klibanov and
Professor JoAnne Stubbe. I've learned a lot from Alex's biological tutorial class during
my first year as a graduate student. I appreciate his encouragement during my PhD study.
As for my thesis committee chair Professor JoAnne Stubbe, her deep passion and
dedication to science are truly inspiring. I enjoyed our annual thesis meetings where she
provided me not only with insightful scientific advice on my projects, but also guidance
in my career development. During the development of this thesis, JoAnne provided very
detailed comments and careful edits. This thesis would not be possible without her
support and contribution.
I would also like to thank my undergraduate research advisor Professor Dirk Trauner
(University of Munich) and the graduate student I worked with in his lab, Dr. Matthew
Banghart. It was in Dirk's lab where I first became interested in developing chemical
tools to study biological events. It was Dirk and Matt who first instilled in me the notion
of carrying out organic chemical reactions inside a live cell or animal and what
implications it could bring. Throughout my PhD career, I was able to keep contact with
Matt who is doing a postdoc at Harvard Medical School. I enjoyed to our celebratory
dinner when we published at the same time, I, for my first paper during graduate school
and he, for his first publication during postdoc. I am thankful to Matt who is always
understanding and encouraging.
For the members of the Ting lab, past and present, I am lucky and honored to have the
opportunity to work with such a talented and diverse group of scientists. I would like to
thank Dan, for being a source of entertainment that keeps us together and for always
offering insightful advice; Tao, for training me when I first started working, for our
enjoyable collaboration on the fluorophore targeting project, and for being a wonderful
bench mate; Katie, for being the lab mother; Stephanie, for our scientific and non-
scientific conversations and for being my lab rock. To all other members of the Ting lab:
Jeff, Ken, Vicky, Kurt, Chai, Wenjing, Philipp, Jake, Carolyn, Oom, Monica, Kayvon
and Betty Lou, thank you for making lab such a special and enjoyable work place, for
supporting me through difficult times, and for sharing laughter and tears.
During the last two years of my PhD journey I feel blessed to have Philip by my side. I
would like to thank him for always being supportive and encouraging, and for keeping
me sane with his often insane ideas and fun projects. I am excited and look forward to
many adventures in the future together. At last I am deeply grateful to my parents, who
4
have supported me in every way they could and have sacrificed a lot so I can study worry
free. I know I may not have called home as often as I should, but they were the ones on
my mind when I had to work late into the night. This thesis is dedicated to my parents,
without whom I would not be the person I am today.
5
Table of Contents
Title Page ..................................................................................................................... 1
Signature Page ......................................................................................................... 2
Abstract ........................................................................................................................ 3
Acknowledgem ents.................................................................................................. 4
Table of Contents..................................................................................................... 6
List of Figures ......................................................................................................... 9
List of Tables ............................................................................................................... 14
List of Abbreviations ................................................................................................ 15
Chapter 1. Introduction: specific and non-specific enzyme-mediated small molecule
labeling m ethods .................................................................................................... 18
Introduction..................................................................................................................19
Enzymes and peptides as protein tags to target small molecule probes ................... 20
Using bioorthogonal chemistries to label proteins with small molecules................24
Site-specific targeting and crosslinking with photocrosslinkers to identify protein-protein
interactions...................................................................................................................26
Prom iscuous enzym es.............................................................................................. 28
Conclusion ................................................................................................................... 31
References....................................................................................................................32
Chapter 2. Development of intracellular protein labeling with diverse fluorophores
using enzyme-mediated azide ligation and strain-promoted cycloaddition ......... 37
Introduction..................................................................................................................38
6
Screening for an efficient LplA ligase and azide pair for site-specific protein labeling
in sid e liv e cells.............................................................................................................4 2
Characterization of an improved LplA enzyme and azide probe pair ..................... 48
Exploring strain-promoted azide-alkyne cycloaddition reaction inside live cells using
alternative source of azide ...................................................................................... 57
C on clu sion ................................................................................................................... 6 1
Experimental methods ............................................................................................. 62
R eferen ces....................................................................................................................6 7
Chapter 3. Optimization and demonstration of diverse fluorophore targeting inside
live cells ....................................................................................................................... 70
In tro d u ction .................................................................................................................. 7 1
Comparison of LplA mutant enzymes and azide pairs using a live-cell labeling assay
...................................................................................................................................... 7 2
Comparison of cyclooctyne structures.................................................................... 77
Demonstration of intracellular protein labeling with ADIBO-fluorescein diacetate... 84
Challenges and optimization for targeting large fluorophores inside live cells..... 87
Extension of site-specific two-step labeling method to diverse fluorophore structures with
labeling inside cells and on the cell surface............................................................. 97
Measurement of overall two-step ligation yield in cells.............................................. 100
C on clu sio n ................................................................................................................... 10 2
E xperim ental m ethods ................................................................................................. 107
R eferen ces....................................................................................................................1 16
Chapter 4. Development of a photocrosslinker ligase for protein-protein interaction
detection......................................................................................................................120
Introduction to methods of protein-protein interaction detection................................ 121
Classes of photocrosslinkers ........................................................................................ 126
Screening for a benzophenone ligase........................................................................... 130
Detection of ligated benzophenone via reaction with fluorophore-hydroxylamine .... 133
7
In vitro photocrosslinking of LAP-tagged FKBP and FRB proteins using benzophenone
p ro b e ............................................................................................................................ 14 2
Optimizations to improve in vitro photocrosslinking yield ......................................... 147
Synthesis and LplA(W37G) incorporation of alkyl benzophenone.............................153
In vitro photocrosslinking with new alkyl benzophenone probe ............... 157
Quantify benzophenone ligation yield inside live mammalian cells ........................... 162
Photocrosslinking inside live mammalian cells...........................................................166
LAP-actin photocrosslinking with drug treatment inside live mammalian cells.........177
C on clu sion ................................................................................................................... 18 1
E xperim ental m ethods ................................................................................................. 183
R eferen ces....................................................................................................................192
Chapter 5. Effort towards membrane protein topology mapping with an ascorbate
peroxidase enzyme ..................................................................................................... 199
Introdu ction .................................................................................................................. 200
EF-hand domain-containing protein 1 - LETM1.........................................................205
Restricted and diffusive labeling patterns of localized APEX.....................................208
Biotin-tyramide labeling patterns of APEX-fused LETMI.........................................211
Controlling expression levels of LETM 1 fusion proteins............................................218
Conclusion and future direction...................................................................................222
E xperim ental m ethods ................................................................................................. 224
R eferen ces....................................................................................................................22 8
Curriculum Vitae.......................................................................................................231
8
List of Figures
Chapter 1: Introduction: specific and non-specific enzyme-mediated small molecule
labeling methods
Chapter 2. Development of intracellular protein labeling with diverse fluorophores
using enzyme-mediated azide ligation and strain-promoted cycloaddition
Figure 2-1. Natural ligation of lipoic acid catalyzed by wild-type LplA.................38
Figure 2-2. Small and large probes to be incorporated by LplA ............................. 40
Figure 2-3. Two-step LplA ligation using functional group handle and chemoselective
bioorthogonal reaction inside live cells ................................................................. 41
Figure 2-4. Small molecule substrate binding pocket of K coli lipoic acid ligase (LplA)
w ith bound lipoyl-A M P ........................................................................................... 45
Figure 2-5. Screening to identify the best LplA mutant/azide substrate pair .......... 47
Figure 2-6. In vitro characterization of azide 9 ligation catalyzed by LplA(W371) .... 49
Figure 2-7. Fixed cell labeling on nuclear localized LAP-YFP with cy5-alkyne via
C u A A C ........................................................................................................................ 5 1
Figure 2-8. Characterization of intracellular azide ligation yields by native gel-shift assay
...................................................................................................................................... 5 3
Figure 2-9. Schematic representation of streptavidin gel-shift assay to measure azide
ligation y ield ................................................................................................................ 54
Figure 2-10. Characterization of intracellular azide ligation yields by streptavidin gel-
sh ift assay ..................................................................................................................... 5 5
Figure 2-11. Fixed cell detection of metabolically incorporated azido sugar analog azido-
GalNAc with cy3-alkyne via CuAAC .................................................................... 58
Figure 2-12. Live cell detection of metabolically incorporated azido sugar analog azido-
GalNAc with strain-promoted azide-alkyne cycloaddition ..................................... 60
Chapter 3. Optimization and demonstration of diverse fluorophore targeting inside
live cells
Figure 3-1. Determination of azide 9 washout time in live cells ............................. 72
9
Figure 3-2. Identification of the best LplA mutant/azide substrate pair for intracellular
protein labeling ......................................................................................................... 74
Figure 3-3. Quantitation of data shown in Figure 3-2 ............................................. 75
Figure 3-4. Immunofluorescence with anti-FLAG to compare labeling against expression
of all enzym es .............................................................................................................. 76
Figure 3-5. Evaluation of various cyclooctyne structures for site-specific intracellular
protein labeling ......................................................................................................... 80
Figure 3-6. Analysis of background labeling by cyclooctyne-fluorescein conjugates
...................................................................................................................................... 8 2
Figure 3-7. Two-dimensional optimization of ADIBO-fluorescein diacetate loading
concentration and washout tim e.............................................................................. 84
Figure 3-8. Two-step fluorophore targeting via strain-promoted azide-alkyne
cy cload dition ................................................................................................................ 85
Figure 3-9. ADIBO-fluorescein labeling of three localized LAP-BFP fusions, LAP-p-
actin, and LA P-M A P2 ........................................................................................... 86
Figure 3-10. Comparison of MOFO-ATTO 647N and ADIBO-ATTO 647N nuclear
labeling specificities................................................................................................. 88
Figure 3-11. Comparison of two different linker structures for MOFO-ATTO 647N
...................................................................................................................................... 8 9
Figure 3-12. Suppression of mitochondrial background labeling with FCCP.........91
Figure 3-13. Suppression of mitochondrial background labeling with QSY21.....92
Figure 3-14. Suppression of mitochondrial background labeling with FCCP and QSY21
...................................................................................................................................... 9 4
Figure 3-15. Comparison of ADIBO-ATTO 647N labeling of LAP-p-actin with and
without suppression of mitochondrial background using QSY21 quencher ........... 96
Figure 3-16. Intracellular protein labeling with diverse fluorophore structures..........97
Figure 3-17. Cell surface and intracellular protein labeling with DIBO-Alexa Fluor 647
and D IB O -biotin ....................................................................................................... 99
Figure 3-18. Streptavidin gel-shift analysis of two-step labeling yield in live cells.... 101
10
Chapter 4. Development of a photocrosslinker ligase for protein-protein interaction
detection
Figure 4-1. Three classes of photocrosslinkers............................................................126
Figure 4-2. Lipoic acid ligase (LplA) with either bound lipoic acid or a benzophenone
unnatural small molecule docked into the enzyme active site..................................... 130
Figure 4-3. Synthetic scheme of benzophenone photocrosslinkers............................. 131
Figure 4-4. Screening to identify LplA mutant/benzophenone substrate pair.............132
Figure 4-5. Unsuccessful attempt to detect in vitro ligation of benzophenone to E2p
protein using biotin-hydrazide ..................................................................................... 134
Figure 4-6. Hydrolysis of hydrazones and oximes ...................................................... 135
Figure 4-7. Reaction between benzophenone and biotin-hydroxylamine to form oxime
p ro d u ct. ........................................................................................................................ 13 7
Figure 4-8. Detection of in vitro ligated benzophenone using AF488-hydroxylamine
...................................................................................................................................... 1 3 9
Figure 4-9. Detection of benzophenone chemically conjugated to FRB-YFP using biotin-
hydroxylam ine ............................................................................................................. 14 1
Figure 4-10. Crystal structure of FKBP-rapamycin-FRB ternary complex................. 142
Figure 4-11. In vitro photocrosslinking with purified proteins FKBP-LAP and FRB-YFP
...................................................................................................................................... 14 6
Figure 4-12. Crystal structure of FRB-rapamycin-FKBP ternary complex with methionine
residues in ball-and-stick ............................................................................................. 148
Figure 4-13. In vitro photocrosslinking with methionine to alanine mutant proteins of
F K B P -L A P ................................................................................................................... 149
Figure 4-14. In vitro photocrosslinking of myc-FKBP-LAP and HA-FRB ................ 151
Figure 4-15. Scheme of benzophenone insertion into nearby C-H bond after UV
irradiation ..................................................................................................................... 153
Figure 4-16. Benzophenone and benzophenone derivatives with different functional
groups attached to the fourth position of the benzophenone core ............................... 155
Figure 4-17. Synthetic scheme of alkyl benzophenone probes with methylene unit
attached to the benzophenone core .............................................................................. 155
11
Figure 4-18. In vitro characterization of para alkyl benzophenone ligation by
L plA (W 37G )................................................................................................................ 156
Figure 4-19. In vitro photocrosslinking comparison between alkyl benzophenone and
benzophenone .............................................................................................................. 159
Figure 4-20. Reducing concentrations of FKBP-LAP(F) and FRB helps in eliminating
background photocrosslinking ..................................................................................... 161
Figure 4-21. Gel-shift analysis of benzophenone ligation yield in cells......................163
Figure 4-22. Native gel-shift assay to measure benzophenone ligation yield over time
...................................................................................................................................... 16 4
Figure 4-23. Expression of FKBP-LAP and FRB in HEK 293T cells ........................ 167
Figure 4-24. Live cell benzophenone ligation and photocrosslinking of myc-FKBP-LAP
and H A -F R B ................................................................................................................ 168
Figure 4-25. Benzophenone photocrosslinker labeling scheme of LAP-actin in live cells
...................................................................................................................................... 16 9
Figure 4-26. Benzophenone ligation and photocrosslinking on HA-LAP/LAP(F)-actin in
live m am m alian cells ................................................................................................... 170
Figure 4-27. Western blot stained with anti-FLAG antibody to detect FLAG-tagged
LplA(W37G) after live cell photocrosslinking of HA-LAP/LAP(F)-actin ................. 172
Figure 4-28. Live cell photocrosslinking of HA-LAP-actin with alkyl benzophenone
incubation tim e course ................................................................................................. 174
Figure 4-29. Benzophenone ligation and photocrosslinking performed on HA-LAP-
M A P2 in live m am m alian cells....................................................................................175
Figure 4-30. Schematic representations of different LAP-actin crosslinked complexes
with and without depolymerization drug treatment.....................................................177
Figure 4-31. Live cell photocrosslinking of LAP-actin with various drug treatments
...................................................................................................................................... 17 9
Chapter 5. Effort towards membrane protein topology mapping with an ascorbate
peroxidase enzyme
12
Figure 5-1. Scheme depicting labeling with APEX fused to either matrix or IMS side of
an IM M protein ............................................................................................................ 204
Figure 5-2. Amino acid sequence of human LETMI .................................................. 206
Figure 5-3. Restricted and diffusive biotin-tyramide labeling patterns in HEK 293T and
C O S -7 cells .................................................................................................................. 209
Figure 5-4. Constructs with APEX fused to different regions of LETMI...................211
Figure 5-5. APEX 115-LETMI fusion protein shows cleavage products on Western blot
...................................................................................................................................... 2 12
Figure 5-6. Biotin-tyramide labeling with APEX-LETM1 fusion constructs in HEK 293T
cells .............................................................................................................................. 2 17
Figure 5-7. Western blot analysis of APEX-LETMI fusion proteins expressed in HEK
2 9 3T cells.....................................................................................................................2 19
Figure 5-8. Western blot analysis of COS-7 cells lipofected (L2k) or viral infected with
LETMI or LETM1-mCherry construct ....................................................................... 220
13
List of Tables
Table 4-1. Optimized UV photocrosslinking conditions for each instrument tested... 144
14
List of Abbreviations
aa ............................................................................................ am in o acid s
A C P ............................................................................... acyl-carrier protein
A TP .......................................................................... adenosine triphosphate
azide 7...........................................................................8-azidooctanoic acid
azide 8..........................................................................9-azidononanoic acid
azide 9........................................................................1 0-azidodecanoic acid
azide 10....................................................................11 
-azidoundecanoic acid
BFP...........................................blue fluorescent protein
B SA .......................................................................... bovine serum album in
B T ......................................................................................................... biotin -tyram ide
CFD A .............................................................................. carboxyfluorescein diacetate
CH O .............................................................................. Chinese ham ster ovarian cells
CuAAC.........................copper-catalyzed azide-alkyne cycloaddition
D A B ................................................................................................. diam inobenzidine
D C M ................................................................................................... dichlorom ethane
D H FR ...................................................................................... dihydrofolate reductase
D IC ............................................................................ differential interference contrast
D IFO .................................................................................... difluorinated cyclooctyne
DIMAC..................................................6,7-dimethoxyazacyclooct-4-yne
DMEM................................................Dulbecco's modified eagle medium
D M F......................................................................................N ,N -dim ethylform am ide
D M SO .............................................................................................. dim ethylsulfoxide
D N A .......................................................................................... deoxyribonucleic acid
DPBS................................Dulbecco's phosphate buffered saline
D T T .......................................................................................................... dith iothreito l
E2p............................9 kDa hybrid lipoyl domain derived from E.coli PDH complex
E . coli..................................................................................................E scherich ia coli
EDTA...........................................................................ethylenediaminetetracetic acid
ER ............................................................................ endoplasm ic reticulum
F P .................................................................................................... fluorescent protein
E R ............................................................................................. endoplasm ic reticulum
E SI............................................................................................electrospray ionization
F -actin ................................................................................................ filam entous actin
FK B P12 .............................................................................. FK 506 binding protein 12
FlAsH................................. fluorescein arsenical hairpin binder
FRB....................FKBP-rapamycin binding domain of the mTOR protein
FRET...............................................................fluorescence resonance energy transfer
FP .................................................................................................... fl uorescent protein
G -actin .................................................................................................... globular actin
G FP ....................................................................................... green fluorescent protein
hAGT....................................................................06 
-alkylguanine-DNA alkyltransferase
HA................................human influenza hemagglutinin epitope
H alo-tag...................................................................................haloalkane dehalogenase
15
HDDA.................................................... ... . . . ....... N,N'-dimethyl- 1,6-diaminohexane
HEK 293T.............................human embryonic kidney cells 293T
HEPES.......................................4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid
H P 1......................................................................................heterochrom atin protein 1
HPLC.............................high-performance liquid chromatography
IMM.............................................................................. inner mitochondrial membrane
IM S .............................................................................................. interm em brane space
IPTG.................................Isopropyl p-D-1-thiogalactopyranoside
LAPI..................................22-amino acid rationally designed LplA acceptor peptide
LAP..............the best 13-amino acid Lp1A acceptor peptide evolved from yeast display
LAP(F).........LAP peptide with Trp downstream of catalytic Lys residue changed to Phe
LC-MS/MS...................................liquid chromatography, tandem mass spectrometry
LplA........................................................................Escherichia coli lipoic acid ligase
MAP2.......................................................................microtubule-associated protein 2
M eC N .......................................................................................................... acetonitrile
M EM ........................................................................................ m inim al eagle m edium
MOFO............................................................................monofluorinated cyclooctyne
M S ................................................................................................... m ass spectrom etry
nA v .............................................................................................................. neutravidin
N E S ............................................................................................. nuclear export signal
N H S..........................................................................................N -hydroxysuccinim ide
N LS................................................................................nuclear localization sequence
N M R ................................................................................. nuclear m agnetic resonance
OMM.............................................................................outer mitochondrial membrane
PAGE..........................................................................polyacrylamide gel electrophoresis
PA L ............................................................................ photoaffinity labeling
PALM.....................................................................photoactivated localization microscopy
PB S......................................................................................phosphate buffered saline
PCR ...................................................................................... polym erase chain reaction
P D B .................................................................................................... protein database
PFA ................................................................................................. paraform aldehyde
PK C .................................................................................................. phosphokinase C
PPTase...................................phosphopantetheine transferase
Q Y ....................................................................................................... quantum yield
R O S .......................................................................................... reactive oxygen species
S A ................................................................................................................ strep tav id in
SA-HRP..............................................streptavidin-conjugated horseradish peroxidase
s.d .................................................................................................. standard deviation
SD S...........................................................................................so dium dodecyl sulfate
STED.................................................................stimulated emission depletion microscopy
STORM...........................................................stochastic optical reconstruction microscopy
T A ........................................................................................................ transient absorption
TBS-T.................................................Tris-buffered saline supplemented with Tween-20
TBTA............................................................................tris-(benzyltriazolylmethyl)amine
T E A ......................................................................................................... triethylam ine
TLC ...................................................................................... thin-layer chrom atography
16
TM ..................................................................................................... transmembrane
TM R........................................................................................... tetramethyrhodamine
Tris.............................................................................tris(hydroxymethyl)aminomethane
UAA.....................................................................................................unnatural amino acid
W B.....................................................................................................western blotting
W CL....................................................................................................whole cell lysate
YFP.......................................................................................yellow fluorescent protein
17
Chapter 1. Introduction:
specific and non-specific enzyme-mediated small molecule labeling methods
18
Introduction
Scientists have long adapted enzymes for fundamental biochemistry research. The
unifying theme of this thesis is to utilize the natural and engineered abilities of enzymes
as robust biochemical tools that can be applied to study biological processes. This is
accomplished in parallel with the design and synthesis of organic small molecule probes,
such as fluorophores or probes baring a chemoselective functional handle, to be
recognized and incorporated by the enzymes.
Green fluorescent protein (GFP) and its multicolored variants (1) are perhaps the
most famous and widely used biochemical tools that have revolutionized cell biologists'
ability to gain insight into cellular processes in living cells. Fluorescent proteins (FPs)
serve as imaging protein reporter tags to visualize protein movement and localization.
FPs offer perfect targeting specificity through genetic fusion and are generally stable
upon chromophore maturation in all cellular compartments. Aside from routine imaging
applications, scientists have engineered FPs for various applications such as
photoactivatable FPs for super-resolution imaging (2) and split-GFP (3) to monitor
protein trafficking, interaction and maturation. However, despite the many advantages
and utilities of FPs, they carry a payload of 27 kDa when fused to the protein of interest,
and are generally less bright and photostable compare to small organic fluorophores (4).
We and others have observed interference caused by FP fusion to protein of interest (5).
For example, mCherry-tagged actin is excluded from the nucleus of the cell, likely
because the mCherry is interfering with the binding of cofilin, a nuclear importer of actin.
Small molecule organic fluorophores are brighter, more photostable, and far
smaller than FPs (4). However, they do not have the ease and perfect specificity of
genetic targeting offered by FPs. Scientists have therefore came up with strategies to
target small molecules fluorophores via the use of enzymes.
19
Enzymes and peptides as protein tags to target small molecule probes
Covalent enzymatic tags (self-modifying enzymes) and non-covalent enzymatic tags
To exploit using enzymes as protein tags to target small organic molecules, the
group of Kai Johnsson employed the ubiquitous human DNA repair enzyme O6
alkylguanine-DNA alkyltransferase (hAGT) (6) which transfers alkyl groups from the
0 6-position of guanine onto a cysteine residue in its active site in a one-turnover self-
modification reaction. The Johnsson group has derivatized the para position of the benzyl
group of 0 6-benzyl guanine with a variety of probes, like fluorophores and biotin, to
achieve specific targeting (7,8) of these molecules by fusing hAGT to protein of interest.
Further evolution based on phase display led to a more active quadruple mutant hAGT
enzyme whose activity is 20 times faster than the wild-type hAGT, which eliminated the
need for hAGT deficient cell lines (9). This more active hAGT, termed SNAP-tag, has
been used to target benzyl guanines derivatized with fluorescein, biotin, and digoxigenin
probes. Later, the Johnsson group expanded the substrate specificity of hAGT through
directed evolution to report another hAGT mutant that exhibits 100 times greater
preference for cytosine over guanine derivatives. The evolved cytosine-specific enzyme,
termed CLIP-tag (10), has been applied together with SNAP-tag to orthogonally label
two populations of proteins with different fluorophores inside the same cell.
Another example of self-modifying enzyme that has been adapted as enzymatic
tag is a mutant form of the enzyme haloalkane dehalogenase (11). The wild-type
haloalkane dehalogenase hydrolyzes haloalkanes by forming a covalent bond between the
asparagine residue in its active site and the alkyl group in an SN2 reaction. A nearby
histidine residue in the active site then hydrolyzes the covalent bond to recycle the
enzyme. In an attempt to convert this enzyme to a covalent labeling tag, researchers at
Promega installed a His289 mutation that eliminated the second hydrolysis step (12).
This mutant form of haloalkane dehalogenase is named HaloTag. Substrates for HaloTag
include haloalkane conjugated fluorophores and functionalized agarose surface resin for
protein purification (11). The HaloTag labeling technology possesses high specificity and
low background labeling due to the lack of endogenous dehalogenases in eukaryotic cells.
Recently it has been applied in purification of the protein cannabinoid receptor CB2 (13),
fused to p75 neurotrophin receptor and tubulin to assess the retrograde axonal transport
20
of the receptor (14), as well as specific labeling of xenograft tumors in living animals
(15).
In addition to covalent modification on self-labeling enzymatic tags, several
proteins with tight binding to their substrates have also been exploited as non-covalent
labeling methods. One of the first examples of non-covalent tagging is the F. coli.
dihydrofolate reductase (eDHFR), which has high affinity to its antibiotic small molecule
ligand trimethoprim (TMP) with KD as low as 1 nM (16). TMP conjugates to various
fluorescent probes can be synthesized by introducing the fluorophores to the para-
methoxy position of TMP. Modifications introduced at this position of TMP raises the KD
only slightly to ~35 nM, which is still a lot lower than the affinity between TMP and
mammalian DHFR. The eDHFR-TMP receptor-ligand pair has been demonstrated in
labeling proteins in the nucleus, plasma membrane, and cytosol in various cell types
(16,17).
Another example of non-covalent protein tag based on the high binding affinity
between enzyme and its ligand is the FK506 binding protein 12 (FKBP12), and its natural
ligand rapamycin, a system pioneered by the Schreiber group (18,19). Clarkson and
coworkers developed a nonnative FKBP12-rapamycin pair by engineering a
FKBP12(F36V) mutant which has high affinity (KD~0.094 nM) to a synthetic derivative
of rapamycin that can be further conjugated to fluorophores and other probes (20,21).
Robers et al demonstrated labeling of numerous intracellular proteins fused to
FKBP12(F36V) with different rapamycin derivative conjugates (21). However, even
though the two non-covalent enzymatic tagging strategies discussed do have very high
affinities between receptor and ligand, they are not suitable for long-term studies
compared to covalent enzymatic tags.
Overall, enzymatic protein tags, either based on covalent self-labeling or high-
affinity non-covalent interaction, provide a way to target small molecule organic
fluorophores inside live mammalian cells with high specificity. A significant drawback is
that most enzymatic tags are still fairly large. The smallest enzymatic tag that has been
demonstrated for specific probe targeting inside live cells mentioned above is the 12 kDa
FKBP12(F36V) enzyme but its non-covalent binding limits its application. In the next
21
section we review enzymatic labeling techniques that are based on small peptide tag
fusions to address this issue.
Enzymatic labeling ofpeptide tags
The ideal labeling method would combine the exquisite specificity conferred by
enzymes, the superior photophysical property of small molecule fluorophores, together
with minimal perturbation to the native protein. While specific targeting to endogenous
proteins is an extremely challenging feat, scientists have developed enzyme-based
techniques that only require fusing a small peptide to the protein of interest. Enzyme-
mediated small peptide tags are recognized by specific enzymes, which then selectively
modify the tags with small molecule substrates.
Zhou et al. used phage display evolution to evolve two orthogonal 12-amino acid
peptides Al and S6 as efficient substrates for Sfp and AcpS phosphopantetheinyl
transferases (PPTases) (22). The PPTases catalyze the covalent linkage of a coenzyme A-
derivatized probe to a serine residue in the peptide substrates. However, the charged
phosphopantetheine arm makes the probe conjugates difficult to cross the cell membranes,
limiting the method to labeling only on the cell surface (22,23).
Our lab has been a pioneer in this area of enzyme-mediated peptide tag labeling
methods. We developed probe targeting techniques based on lipoic acid ligase (LplA)
and biotin ligase (BirA) together with their respective peptide tags. We engineered the
E. coli enzyme lipoic acid ligase (LplA) though site-directed mutagenesis to incorporate
unnatural small molecules such as fluorophores (5,24,25), functional groups (26-28), and
a photocrosslinker (29) site-specifically onto the lysine residue in a 13-amino-acid LplA
acceptor peptide (LAP). To develop this technology we now call PRIME (PRobe
Incorporation Mediated by Enzymes), we explored the plasticity of the small molecule
binding pocket of LplA and also reduced the size of its protein substrate from the 9 kDa
E2p subunit of pyruvate dehydrogenase to the 13- amino acid LAP through yeast-display
evolution (30). By examining the crystal structure of E. coli LplA with bound lipoyl-
AMP, we engineered LplA to accept unnatural probes through alanine scanning of amino
acid residues in the active site (29) and identifying a key "gatekeeper" residue, Trp37,
that when mutated to smaller amino acids can enlarged the LplA active site to enable it to
22
bind to larger unnatural probes, such as different derivatives of coumarin (5,24,25), aryl
azide (29), alkyl azide (26,27), and trans-cyclooctene (28). Dr. Chayasith Uttamapinant
and coworkers demonstrated specific labeling of coumarin fluorophore by LplA(W37V)
onto many LAP-tagged proteins inside live mammalian cells (5). More importantly,
through this peptide-tagging method, they were able to detect the presence of actin in the
nucleus, which was shown to preclude actin when it was fused to a large fluorescent
protein, further highlighting and confirming the importance of a small tag.
The K coli biotin ligase (BirA) covalently attaches biotin to a lysine residue of a
15-amino acid biotin acceptor peptide (AP). A ketone derivative of biotin was
synthesized for subsequent chemoselective reaction with hydrazide or hydroxylamine-
conjugated biophysical probes in a two-step process (31). Dr. Sarah Slavoff and
coworkers also explored BirA from yeast and Pyrococcus horikoshii to label azide- or
alkyne-derivatized biotin analogs, which allowed for downstream detection using click
chemistry (32). However, one limitation of the BirA labeling method is that since
endogenous biotin is still preferred as the substrate, the labeling method either requires
biotin starvation of the cells prior to BirA intracellular labeling or is restricted to the cell
surface.
The Ploegh group harvested the enzyme sortase A from Staphylococcus aureus
which recognizes a only 5-amino acid peptide sequence LPXTG, the shortest peptide tag
available (33,34). The sortase A enzyme performs a transpeptidation reaction by first
cleaving the peptide bond between the T and G residues, and at the same time forming a
new peptide bond to a polyG-derivatized small molecule. However, this method is also
restricted to the cell surface because the polyG substrates are not cell permeable.
23
Using bioorthogonal chemistries to label proteins with small molecules
So far we have discussed using enzymes that are able to target fluorophores
functionalized with ligands they recognize - haloalkane for Halo-Tag, guanine or
cytosine for hAGT and its variant, TMP for eDHFR, so the labeling reaction is complete
in a single step by the specific enzyme. In another approach, bioorthogonal chemistries
between two functional groups that are not found inside mammalian cells have been
extensively explored as a way to target small molecules with high specificity. Chemical
labeling techniques that utilize bioorthogonal reactions (35) to target small organic
fluorophores inside live cells have seen much advancement over the years. Aside from
harnessing specificity from Nature, chemists have began to develop reactions and
engineer functional group handles that can specifically recognize and react with each
other not only in the exogenous environment of an isolated flask, but inside the complex
milieu of cells and living organisms. To develop bioorthogonal reactions as a way to
target fluorophores in a biological environment, the functional groups, or reactive
"handles" involved in the bioorthogonal reaction must stay inert until they meet to
selectively to react with each other; the reaction condition must be biocompatible in
terms of pH and temperature and nontoxic to the biological system under study; the
reaction must have fast kinetics so that the product is formed at a reasonable rate on the
scale of protein turnover and the progression of biological processes, even when the
reactant concentrations are preferably low. The bioorthogonal product formed must be
stable to proteases and not be degraded so to have a consistent readout to be useful as
chemical reporters. Excess reactants should be removed from the system to minimize
unwanted side reaction, or masked to reduce possible background signal. The versatility
of bioorthogonal reactions has been demonstrated by their ability to tag proteins and
glycans. The Staudinger ligation between azide and triarylphosphine was the first
bioorthogonal reaction used to label cell-surface azide-derivatized glycans on live cells
(36), and subsequently in live animals (37). Despite much effort in mechanistic
modifications to increase the reaction kinetics, reduce phosphine oxidation, and aza-ylide
hydrolysis, its slow second-order rate constant (10- M's 1) leaves much more to be
desired (38,39).
24
Even though the Staudinger ligation is slow, it brought to chemists' attention the
potential of organic azide as an excellent functional group handle. The small size of the
azide, its absence from living organisms, its kinetic stability inside cells, and its nontoxic
profile as demonstrated by early approved drugs like azidothymidine all hinted to the
potential of organic azide as an excellent functional group handle. The copper-free or
strain-promoted azide-alkyne cycloaddition was adapted from the Huisgen [3+2]
cycloaddition that initially used heat and pressure to drive the reaction (40), as well as the
later developed Cu(I)-catalyzed azide-alkyne cycloaddition demonstrated by Sharpless
and Meldal (41). Instead of using heat, pressure, or Cu(I) which are incompatible inside
cells, the strain-promoted azide-alkyne cycloaddition uses energy stored in the ring strain
of cyclooctyne to drive the reaction. Since 2000, numerous generations of cyclooctynes
were synthesized with improved reactivity, kinetics, hydrophilicity, and overall
biocompatibility. The highest second-order rate constant of a strain-promoted azide-
alkyne cycloaddition reaction that has been demonstrated on live cell-surface was
reported to be 0.96 M-1 s-1 (42), twenty times higher than the initial Staudinger ligation. In
the first part of this thesis, strain-promoted azide-alkyne cycloaddition is used for the
targeting of diverse fluorophores to specific proteins inside live cells.
Yet another recent addition to the toolbox of bioorthogonal reactions is the
reaction between tetrazine and trans-cyclooctene (43). Developed by the Fox group, this
inverse Diels-Alder reaction is the fastest bioorthogonal reaction known to date,
optimized to achieve an impressive rate constant of 22,000 M' s4 . Our lab has
demonstrated two-step fluorophore targeting using a lipoic acid ligase enzyme in
combination with Diels-Alder cycloaddition to perform intracellular protein labeling (28).
The very fast cycloaddition kinetics yields substantial improvements in signal to
background ratio. Furthermore, the tetrazine-fluorophore conjugate is also fluorogenic,
which is an additional benefit that masks the background caused by unreacted reagents.
25
Site-specific targeting and crosslinking with photocrosslinkers to identify protein-
protein interactions
So far we have discussed a few probe labeling techniques that capitalize on the
specificity of enzymes and bioorthogonal reactions to target small molecules to protein of
interest inside cells. We have seen many enzyme-based fluorophore or probe targeting
methods that have been demonstrated to label a specific protein among the plethora of
proteins inside live mammalian cells. High specificity is a critical requirement since the
target is often a single protein. In a different approach, scientists sought methods to label
and identify multiple proteins associated with a protein of interest or within a particular
cellular compartment. Identifying the many proteins of a certain proteome, within a
certain compartment of the cell, or under an altered cellular state requires a different
labeling approach.
One of the conventional approaches to capture groups or complexes of proteins is
affinity-based methods. Methods such as co-immunoprecipitation (44) that uses an
antibody against the protein of interest requires first lysing the cells, then immunocapture
the protein of interest together with its interacting proteins. Subsequent Western blots or
mass spectrometry (MS) analysis is used to identify proteins that are pulled down.
However, using cell lysates as starting material already introduces numerous artifacts.
First, the proteins are removed from their native environment. In cell lysate, without
cellular compartmentalization that maintains various pHs and effective concentrations of
biomolecules, complexes that form depending on those factors could dissociate or
aggregate with other proteins. Second, disruption of protein complexes during cell lysis
also presents a serious problem. Weakly bound proteins could be lost during washing, or
non-specific binding of proteins that are not endogenous interacting partners could
wrongly associate. Last but not least, the limited availability of high quality of antibodies
introduces additional constraint.
As an alternative, chemical crosslinking prior to immunocapture is employed to
covalently link interacting proteins while the proteins are in their endogenous states.
Covalent linkage allows more stringent washing conditions to remove non-specifically
interacting proteins without fear of losing weakly bound proteins (45). Traditional
chemical crosslinking reagents include formaldehyde or amine-reactive moieties such as
26
N-hydroxysuccinimide (NHS) esters. Formaldehyde can readily react with lysines,
tryptophans, or N-termini of proteins after minutes of exposure; however, it does not
have any spatial resolution or protein specificity once it permeates cell membranes. To
address this issue, photoaffinity labeling (PAL) (46) that uses site-specifically conjugated
photoreactive crosslinkers, or photocrosslinkers, enables identifying interacting proteins
specific to a protein of interest. Photocrosslinkers such as aryl azides, benzophenones,
and diazirines can be site-specifically introduced to proteins via in vitro cysteine-
maleimide chemistry or incorporated as unnatural amino acids using the unnatural amino
acid (UAA) mutagenesis technique developed by the Schultz group (47). Crosslinking
with site-specific photocrosslinkers offers high temporal and spatial resolution to capture
proteins in their native states and has been demonstrated inside live mammalian cells.
However, as discussed previously, techniques to site-specifically target these small
molecule photocrosslinkers are still limited. In Chapter 4, we introduce an enzyme-based
ligation technique for the benzophenone photocrosslinker.
27
Promiscuous enzymes
Achieving location specificity of photocrosslinkers or other reactive species to
form covalent linkage with proteins inside live cells is a challenging task. By site-
specifically attaching a photocrosslinker to a protein of interest allows us to pull down
one or multiple interacting proteins, but would not enable us to identify, for example, the
entire proteome of the mitochondrial matrix under drug treatment. To identify the
proteome of an entire cell or a specific organelle under its native environment without
contamination from cell lysate or organelle purification, scientists have yet again looked
to enzymes for solutions. One approach is to use enzymes that can convert small
molecule probes to diffusive reactive intermediates that can form covalent bonds with its
nearby proteins. Having diffusive reactive species versus site-specifically attached
crosslinkers also obviates the bottleneck of crosslinker length and shape, which limits its
range of reactivity. Several enzymes that are able to take a chemical probe and generate a
reactive intermediate have been applied in this type of promiscuous protein labeling.
Horseradish peroxidase (HRP) was first applied as a reporter enzyme for electron
microscopy (EM) (48) by reacting with diaminobenzidine (DAB) in the presence of H20 2
to form an insoluble tar-like product that is electron-opaque and can serve as an EM
contrast agent. Later, HRP was adapted by Kotani et al. to catalyze the generation of
nitrenes from aryl azides (49). In this enzyme-mediated activation of radical source
(EMARS) method, cell membrane-targeted HRP generated radicals of aryl azide-biotin
conjugate that crosslinked to nearby proteins, which can be pulled down and analyzed
using either streptavidin-coated beads or membrane protein specific antibody. EMARS
labeling has been applied to study cell-surface molecular clusters of p1-integrin,
epidermal growth factor receptor (EGFR), insulin-like growth factor-I receptor (IGF1R),
and EphA2 receptor (50,51). Immunoelectron microscopy of the EMARS-labeled HRP-
conjugated antibody showed 96% of biotinylated proteins were within 300 nm of the
antibody-coated particles (50). However, one limitation of HRP is that the enzyme is not
active in the cytosol, perhaps because its disulfide bonds were reduced under the reducing
environment.
Promiscuous biotin ligase is another example of an enzyme that can generate
reactive small molecules. Through a point mutation (RI 18G) (52) in the wild-type biotin
28
ligase which has exquisite specificity towards its endogenous protein substrate,
promiscuous biotin ligase is able to prematurely release the adenylated biotinoyl-5'-AMP
from its active site without the protein substrate present. Once released, biotinyl-5'- AMP
can react with amine functional groups on nearby proteins forming an amide bond.
Biotinylated proteins are then enriched using streptavidin binding and analyzed on
Western blot. However, the time of the reaction is especially long, requiring longer than 6
hours of biotin incubation to saturated labeling. Even though the labeling is performed in
intact live cells, but the prolonged labeling time could cause artifacts in endogenous
protein interactions. Biotinylated proteins could diffuse away leading to poor spatial
resolution or even cause subsequent unwanted cellular changes. Furthermore, the lack of
temporal resolution would not allow differentiation and detection of rapid changes in
proteomes. Labeling with promiscuous biotin ligase was demonstrated by tagging the
enzyme to nuclear lamina protein, lamin A, and the construct was expressed and labeled
in HEK 293 cells (53). Roux et al identified 122 proteins, some of which are known
proteins associated with the nuclear envelope, while others are novel, such as the protein
FAM169A (later called SLAP75) which does not have any predicted transmembrane
region or any sequence motifs that would hint to its localization. FAM169A was also
confirmed by immunostaining to be located in the nuclear envelope.
To develop an enzyme that can be specifically targeted to a region within the cell,
works in different organelle environments, and has high temporal and spatial resolution,
our lab has engineered a plant ascorbate peroxidase called APEX to catalyze the
formation of biotin-tyramide radicals in the presence of H20 2 (54). The short lifetime
(less than 1 pts) of the radical together with its membrane-impermeant nature ensure tight
labeling radius and high temporal resolution. Rhee et al. targeted APEX to the
mitochondrial matrix and the inner mitochondrial space (IMS), two regions that are
especially difficult to purify through subcellular fractionation and organelle purification
(54). Using matrix-targeted APEX, biotin-tyramide and H20 2 labeling, followed by
streptavidin enrichment and tandem mass spectrometry analysis, 495 proteins were
identified in the matrix proteome, with 464 proteins having prior mitochondrial
annotation (54). Notably, proteins such as CPOX and PPOX which were previously
assigned to the mitochondrial intermembrane space were detected in the matrix proteome,
29
leading to novel insights to their trafficking and functional roles in the heme biosynthesis
pathway. Currently APEX is being applied towards identifying the proteome of
especially challenging regions of the cell that are impossible to purify through traditional
means, such as the synaptic space or the junction between mitochondria and the
endoplasmic reticulum (ER). In the final chapter of this thesis, we describe extending
APEX labeling to probe the topology of membrane proteins.
30
Conclusion
Despite the advances in cell biology brought by the fluorescent proteins, their
large size and poor photophysical properties prompted scientists to look for alternatives.
Compared to fluorescent protein tags, chemical tags such as small molecule fluorophore
or other biophysical probes are smaller, brighter and more versatile. To target small
molecules site-specifically to a protein of interest inside the cell, scientists have utilized
the exquisite specificity of enzymes as well as developed bioorthogonal chemical
reactions that can work efficiently within the complex environment inside the cell. In
Chapter 2 and 3, we describe a fluorophore targeting method that uses only a small
peptide tag in combination with enzymatic labeling and azide-alkyne cycloaddition to
label numerous proteins inside cells with various fluorophores.
While being able to visualize the localization, movement, and interaction of a
specific protein inside its native environment has brought tremendous insights to cell
biology regarding the how a cell functions, the ability to identify with high temporal
resolution pools or complexes of proteins associated with a single protein, of a particular
cellular organelle, or even of the entire cell under a specific cellular condition has also
revolutionized the way scientists study and understand biological processes at the
molecular and cellular level. In Chapter 4, we present a method to site-specifically ligate
a photocrosslinker to identify protein-protein interactions. Aside from targeted
photocrosslinkers, scientists have developed enzymes that can generate reactive species,
such as radicals, which then react nonspecifically with proximal proteins. In Chapter 5,
we discuss current effort towards developing two parallel but distinct approaches to use
enzyme-generated radical species to study membrane protein topology.
31
References
1. Sample V, Newman RH, Zhang J. The structure and function of fluorescent proteins.
Chem Soc Rev. 2009 Oct;38(10):2852-64.
2. Wiedenmann J, Gayda S, Adam V, Oswald F, Nienhaus K, Bourgeois D, et al. From
EosFP to mIrisFP: structure-based development of advanced photoactivatable
marker proteins of the GFP-family. J Biophotonics. 2011 Jun;4(6):377-90.
3. Ozawa T. Protein reconstitution methods for visualizing biomolecular function in
living cells. Yakugaku Zasshi. 2009 Mar;129(3):289-95.
4. Lavis LD, Raines RT. Bright ideas for chemical biology. ACS Chem. Biol. 2008
Mar 20;3(3):142-55.
5. Uttamapinant C, White KA, Baruah H, Thompson S, Femindez-Suirez M,
Puthenveetil S, et al. A fluorophore ligase for site-specific protein labeling inside
living cells. PNAS. 2010 Jun 15;107(24):10914-9.
6. Keppler A, Gendreizig S, Gronemeyer T, Pick H, Vogel H, Johnsson K. A general
method for the covalent labeling of fusion proteins with small molecules in vivo.
Nat Biotech. 2003 Jan;21(1):86-9.
7. Keppler A, Pick H, Arrivoli C, Vogel H, Johnsson K. Labeling of fusion proteins
with synthetic fluorophores in live cells. Proc. Natl. Acad. Sci. U.S.A. 2004 Jul
6;101(27):9955-9.
8. Keppler A, Kindermann M, Gendreizig S, Pick H, Vogel H, Johnsson K. Labeling
of fusion proteins of 06-alkylguanine-DNA alkyltransferase with small molecules
in vivo and in vitro. Methods. 2004 Apr;32(4):437-44.
9. Juillerat A, Gronemeyer T, Keppler A, Gendreizig S, Pick H, Vogel H, et al.
Directed evolution of 06-alkylguanine-DNA alkyltransferase for efficient labeling
of fusion proteins with small molecules in vivo. Chem. Biol. 2003 Apr; 10(4):313-7.
10. Gautier A, Juillerat A, Heinis C, Correa IR Jr, Kindermann M, Beaufils F, et al. An
engineered protein tag for multiprotein labeling in living cells. Chem. Biol. 2008
Feb;15(2):128-36.
11. Los GV, Encell LP, McDougall MG, Hartzell DD, Karassina N, Zimprich C, et al.
HaloTag: a novel protein labeling technology for cell imaging and protein analysis.
ACS Chem. Biol. 2008 Jun 20;3(6):373-82.
12. Pries F, Kingma J, Krooshof GH, Jeronimus-Stratingh CM, Bruins AP, Janssen DB.
Histidine 289 is essential for hydrolysis of the alkyl-enzyme intermediate of
haloalkane dehalogenase. J. Biol. Chem. 1995 May 5;270(18):10405-11.
32
13. Locatelli-Hoops S, Sheen FC, Zoubak L, Gawrisch K, Yeliseev AA. Application of
HaloTag technology to expression and purification of cannabinoid receptor CB2.
Protein Expr. Purif. 2013 May;89(1):62-72.
14. Mok S-A, Lund K, Lapointe P, Campenot RB. A HaloTag@ method for assessing
the retrograde axonal transport of the p75 neurotrophin receptor and other proteins
in compartmented cultures of rat sympathetic neurons. J. Neurosci. Methods. 2013
Mar 30;214(1):91-104.
15. Tseng J-C, Benink HA, McDougall MG, Chico-Calero I, Kung AL. In Vivo
Fluorescent Labeling of Tumor Cells with the HaloTag@ Technology. Curr Chem
Genomics. 2012;6:48-54.
16. Miller LW, Cai Y, Sheetz MP, Cornish VW. In vivo protein labeling with
trimethoprim conjugates: a flexible chemical tag. Nat Meth. 2005 Apr;2(4):255-7.
17. Calloway NT, Choob M, Sanz A, Sheetz MP, Miller LW, Cornish VW. Optimized
fluorescent trimethoprim derivatives for in vivo protein labeling. Chembiochem.
2007 May 7;8(7):767-74.
18. Bierer BE, Somers PK, Wandless TJ, Burakoff SJ, Schreiber SL. Probing
immunosuppressant action with a nonnatural immunophilin ligand. Science. 1990
Oct 26;250(4980):556-9.
19. Spencer DM, Wandless TJ, Schreiber SL, Crabtree GR. Controlling signal
transduction with synthetic ligands. Science. 1993 Nov 12;262(5136):1019-24.
20. Clackson T, Yang W, Rozamus LW, Hatada M, Amara JF, Rollins CT, et al.
Redesigning an FKBP-ligand interface to generate chemical dimerizers with novel
specificity. Proc. Natl. Acad. Sci. U.S.A. 1998 Sep 1;95(18):10437-42.
21. Robers M, Pinson P, Leong L, Batchelor RH, Gee KR, Machleidt T. Fluorescent
labeling of proteins in living cells using the FKBP12 (F36V) tag. Cytometry A.
2009 Mar;75(3):207-24.
22. Zhou Z, Cironi P, Lin AJ, Xu Y, Hrvatin S, Golan DE, et al. Genetically encoded
short peptide tags for orthogonal protein labeling by Sfp and AcpS
phosphopantetheinyl transferases. ACS Chem. Biol. 2007 May 22;2(5):337-46.
23. Yin J, Straight PD, McLoughlin SM, Zhou Z, Lin AJ, Golan DE, et al. Genetically
encoded short peptide tag for versatile protein labeling by Sfp phosphopantetheinyl
transferase. Proc. Natl. Acad. Sci. U.S.A. 2005 Nov 1;102(44):15815-20.
24. Jin X, Uttamapinant C, Ting AY. Synthesis of 7-Aminocoumarin by Buchwald-
Hartwig Cross Coupling for Specific Protein Labeling in Living Cells.
ChemBioChem. 2011;12(1):65-70.
33
25. Cohen JD, Thompson S, Ting AY. Structure-Guided Engineering of a Pacific Blue
Fluorophore Ligase for Specific Protein Imaging in Living Cells. Biochemistry.
2011 Sep 27;50(38):8221-5.
26. Fernandez-Suairez M, Baruah H, Martinez-Hernandez L, Xie KT, Baskin JM,
Bertozzi CR, et al. Redirecting lipoic acid ligase for cell surface protein labeling
with small-molecule probes. Nature Biotechnology. 2007;25(12):1483-7.
27. Yao JZ, Uttamapinant C, Poloukhtine A, Baskin JM, Codelli JA, Sletten EM, et al.
Fluorophore Targeting to Cellular Proteins via Enzyme-Mediated Azide Ligation
and Strain-Promoted Cycloaddition. J. Am. Chem. Soc. 2012 Feb 29;134(8):3720-8.
28. Liu DS, Tangpeerachaikul A, Selvaraj R, Taylor MT, Fox JM, Ting AY. Diels-
Alder Cycloaddition for Fluorophore Targeting to Specific Proteins inside Living
Cells. J. Am. Chem. Soc. 2012 Jan 18;134(2):792-5.
29. Baruah H, Puthenveetil S, Choi Y-A, Shah S, Ting AY. An Engineered Aryl Azide
Ligase for Site-Specific Mapping of Protein-Protein Interactions through Photo-
Cross-Linking. Angewandte Chemie International Edition. 2008;47(37):7018-21.
30. Puthenveetil S, Liu DS, White KA, Thompson S, Ting AY. Yeast Display Evolution
of a Kinetically Efficient 13-Amino Acid Substrate for Lipoic Acid Ligase. J. Am.
Chem. Soc. 2009 Nov 18;131(45):16430-8.
31. McNeill E, Chen I, Ting AY. Synthesis of a Ketone Analogue of Biotin via the
Intramolecular Pauson-Khand Reaction. Org. Lett. 2006 Sep 1;8(20):4593-5.
32. Slavoff SA, Chen I, Choi Y-A, Ting AY. Expanding the Substrate Tolerance of
Biotin Ligase through Exploration of Enzymes from Diverse Species. J. Am. Chem.
Soc. 2008 Jan 1;130(4):1160-2.
33. Popp MW, Antos JM, Grotenbreg GM, Spooner E, Ploegh HL. Sortagging: a
versatile method for protein labeling. Nat. Chem. Biol. 2007 Nov;3(11):707-8.
34. Popp MW-L, Antos JM, Ploegh HL. Site-specific protein labeling via sortase-
mediated transpeptidation. Curr Protoc Protein Sci. 2009 Apr;Chapter 15:Unit 15.3.
35. Sletten EM, Bertozzi CR. Bioorthogonal Chemistry: Fishing for Selectivity in a Sea
of Functionality. Angewandte Chemie International Edition. 2009;48(38):6974-98.
36. Saxon E, Bertozzi CR. Cell surface engineering by a modified Staudinger reaction.
Science. 2000 Mar 17;287(5460):2007-10.
37. Prescher JA, Dube DH, Bertozzi CR. Chemical remodelling of cell surfaces in
living animals. Nature. 2004 Aug 19;430(7002):873-7.
34
38. Lin FL, Hoyt HM, Van Halbeek H, Bergman RG, Bertozzi CR. Mechanistic
investigation of the staudinger ligation. J. Am. Chem. Soc. 2005 Mar
2; 127(8):2686-95.
39. Agard NJ, Baskin JM, Prescher JA, Lo A, Bertozzi CR. A Comparative Study of
Bioorthogonal Reactions with Azides. ACS Chem. Biol. 2006 Nov 1;1(10):644-8.
40. Huisgen R. Centenary Lecture - 1,3-Dipolar Cycloadditions. Proc. Chem. Soc.
London. 1961;(OCT):357-96.
41. Meldal M, Tornee CW. Cu-catalyzed azide-alkyne cycloaddition. Chem. Rev. 2008
Aug;108(8):2952-3015.
42. Jewett JC, Sletten EM, Bertozzi CR. Rapid Cu-Free Click Chemistry with Readily
Synthesized Biarylazacyclooctynones. J Am Chem Soc. 2010 Mar
24;132(11):3688-90.
43. Blackman ML, Royzen M, Fox JM. The Tetrazine Ligation: Fast Bioconjugation
based on Inverse-electron-demand Diels-Alder Reactivity. J Am Chem Soc. 2008
Oct 15;130(41):13518-9.
44. Co-immunoprecipitations revisited: an update on experimental concepts and their
implementation for sensitive interactome investigations of endogenous proteins -
Springer. [cited 2013 Apr 19]; Available from:
http://link.springer.com/article/l0.1007/s00216-007-1385-x/fulltext.html
45. Phizicky EM, Fields S. Protein-protein interactions: methods for detection and
analysis. Microbiol. Rev. 1995 Mar 1;59(1):94-123.
46. Hatanaka Y, Sadakane Y. Photoaffinity Labeling in Drug Discovery and
Developments: Chemical Gateway for Entering Proteomic Frontier. Current Topics
in Medicinal Chemistry. 2002 Mar 1;2(3):271-88.
47. Liu W, Brock A, Chen S, Chen S, Schultz PG. Genetic incorporation of unnatural
amino acids into proteins in mammalian cells. Nat Meth. 2007 Mar;4(3):239-44.
48. Hopkins C, Gibson A, Stinchcombe J, Futter C. Chimeric molecules employing
horseradish peroxidase as reporter enzyme for protein localization in the electron
microscope. In: Jeremy Thorner SDE and JNA, editor. Methods in Enzymology
[Internet]. Academic Press; 2000 [cited 2013 Mar 31]. p. 35-45. Available from:
http://www.sciencedirect.com/science/article/pii/S0076687900272650
49. Kotani N, Honke K. [Novel approach for the cell surface molecular interactome
using enzyme-mediated activation of radical sources (EMARS) reaction]. Seikagaku.
2011 Aug;83(8):754-8.
50. Kotani N, Gu J, Isaji T, Udaka K, Taniguchi N, Honke K. Biochemical visualization
of cell surface molecular clustering in living cells. PNAS [Internet]. 2008 May 21
35
[cited 2013 Mar 31]; Available from:
http://www.pnas.org/content/early/2008/05/20/0710346105
51. Honke K, Kotani N. Identification of Cell-Surface Molecular Interactions under
Living Conditions by Using the Enzyme-Mediated Activation of Radical Sources
(EMARS) Method. Sensors. 2012 Nov 22;12(12):16037-45.
52. Kwon K, Streaker ED, Ruparelia S, Beckett D. Multiple disordered loops function
in corepressor-induced dimerization of the biotin repressor. J. Mol. Biol. 2000 Dec
15;304(5):821-33.
53. Roux KJ, Kim DI, Raida M, Burke B. A promiscuous biotin ligase fusion protein
identifies proximal and interacting proteins in mammalian cells. J Cell Biol. 2012
Mar 19;196(6):801-10.
54. Rhee H-W, Zou P, Udeshi ND, Martell JD, Mootha VK, Carr SA, et al. Proteomic
mapping of mitochondria in living cells via spatially restricted enzymatic tagging.
Science. 2013 Mar 15;339(6125):1328-31.
36
Chapter 2. Development of intracellular protein labeling with diverse fluorophores
using enzyme-mediated azide ligation and strain-promoted cycloaddition
Part of the work discussed in this chapter has been published in: J. Yao, C. Uttamapinant,
A. Poloukhtine, J.M. Baskin, J.A. Codelli, E.M. Sletten, C.R. Bertozzi, V.V. Popik, A.Y.
Ting, "Fluorophore targeting to cellular proteins via enzyme-mediated azide ligation and
strain-promoted cycloaddition", JAm Chem Soc 2012 (134), 3720-3728.
37
Introduction
In the previous chapter, we described various enzyme-based methods of
fluorophore targeting. We seek to develop a fluorophore targeting method that combines
high labeling specificity with small noninvasive peptide tag, works for intracellular
protein labeling inside live cells, and is generalizable towards a wide array of
fluorophores with different photophysical functionalities. The Ting lab has developed
enzyme-based fluorophore labeling techniques based on the Escherichia coli enzyme
lipoic acid ligase (LplA) (1,2). The natural function of LplA in E. coli is to catalyze the
ATP-dependent ligation of the cofactor lipoic acid onto the lysine residue in one of its
three substrates: E2p subunit of pyruvate dehydrogenase complex, E3 subunit of 2-
oxoglutarate dehydrogenase complex, and H-protein of glycine cleavage system (Figure
2-1) (1). The mechanism of lipoic acid transfer proceeds in two steps. First, the enzyme
activates the carboxylic acid of the lipoic acid cofactor using ATP and releasing
pyrophosphate in the process. In the second step, the reactive adenylate ester intermediate,
lipoyl-AMP, is attacked by the s-amino group of a lysine residue in the lipoate-acceptor
protein (e.g., E2p) to effect lipoic acid transfer.
E. Coli lipoic acid ligase (LpLA)
QNH3 H lipoic acid HNX(<)
iH + ATP
LplAelipoyl-AMP
E2p (9 kDa) lipoyl-E2p
Figure 2-1. Natural ligation of lipoic acid catalyzed by wild-type LplA.
The enzyme has high specificity towards its protein substrates, but exhibits much
more promiscuity for its small molecule substrate (2). Over the years, the Ting lab has
exploited this small molecule plasticity of LplA and engineered LplA, as well as its
protein substrate for the application of protein labeling over several stages. First, in the
work of Dr. Sujiet Puthenveetil, he used iterative cycles of rational design (3) and later
38
through yeast display evolution to evolve one of the natural acceptors of LplA, the 86
amino acid (8.9 kDa) lipoyl domain of the full-length E2p subunit of pyruvate
dehydrogenase to a small 13-amino acid peptide, which we call LplA acceptor peptide
(LAP) (4). The LAP peptide can be genetically fused to any protein of interest and
confers small tag size while maintaining high specificity. The kcat/Km of LplA for the
LAP peptide is only eight-fold lower than its kcat/Km for the H-protein (4). Second,
through site-directed mutagenesis of the lipoic acid binding pocket of LplA, our lab has
shown that the engineered LplA can incorporate a plethora of unnatural small molecules
including 7-hydroxycumarin (5,6), 7-aminocoumarin (7), Pacific Blue (8), aryl azide (9),
alkyl azide (3), and trans-cyclooctene (10) (Figure 2-2). Using this LplA enzyme-
catalyzed probe ligation platform which we call PRIME (PRobe Incorporation Mediated
by Enzymes), we have labeled LAP-fused proteins inside live cells and on the cell
surface, such as cytoskeletal proteins actin and mitogen-associated protein 2 (MAP2),
FK506 binding protein (FKBP), and neuronal adhesion protein neurexin.
The application of PRIME for intracellular fluorophore targeting was first
demonstrated by Dr. Tao Uttamapinant and co-workers (5). Through screening of many
LplA mutants, LplA(W37V) was identified as a coumarin ligase. While coumarin
labeling already offers advantages over other labeling methods in its small fusion tag and
high labeling specificity inside mammalian cells, coumarin has excitation and emission
maxima of 387 nm and 448 nm, which are not optimal for live cell imaging. The strict
restriction to only one class of blue fluorophores limits the applicability of PRIME. We
wished to extend PRIME labeling towards other more red-shifted fluorophores; to further
increase detection sensitivity compared to coumarin-PRIME via using dyes with better
photophysical properties and greater separation from cellular autofluorescence; to reduce
photodamage to cells during imaging by using longer-wavelength excitation light; and
also to provide additional colors to complement coumarin-PRIME labeling.
39
Direct ligation by LpIA Two-step ligation by LpIA via functional group handle
Fluorophores Pacifi B lue loopoe
HO HO -NH 
O O OH PcfcBu
HydraWcoumarin Aminocumarin HO n
Alexa Fluor Oregon green
Cyanine dye
Funct inal group H Hhandles HO, HON' N 4
Rhodamine Atto dye
1403"O "4 PIColyl aZide
Terminal alkyne N
Figure 2-2. Small and large probes to be incorporated by LplA. (left) Chemical probes
directly ligated by LplA mutants. (right) Large fluorophores that could be ligated by
LplA via functional group handle.
We desire a strategy for targeting to intracellular LAP-fused proteins any
fluorophore that can cross the cellular membrane and exhibit minimal nonspecific
binding to endogenous proteins. Looking at the structures of large fluorophores such as
cyanine dyes and ATTO dyes (Figure 2-2), all of them are far larger than coumarin.
Although a direct, one-step labeling approach is attractive in that it offers efficient
ligation yield and ease of use, engineering a specific LplA to accept each large
fluorophore would be very challenging and require extensive protein engineering of the
LplA active site. Examining of the crystal structure E. coli LplA with lipoate-AMP bound
in the active site shows that the dithiolane ring of lipoate-AMP is completely surrounded
by active site residues. Even though by mutagenesis of some active site residues we were
able to enlarge the lipoate binding pocket to incorporate larger, linear, bi- or tricyclic
fluorophore structures such as coumarin (5) and resorufin (Daniel Liu, unpublished), to
further mutate the binding pocket to bind to nonlinear, tetracyclic, and charged dyes like
fluorescein and tetramethylrhodamine seemed unlikely. Random mutation of active site
residues might be fruitless or hurt the folding or specificity of LplA. Instead of one-step
fluorophore targeting, we envisioned a two-step ligation process: first using LplA to
target a bioorthogonal functional group handle onto LAP, then chemoselectively
40
derivatizing the functional group handle with any fluorophore of interest (Figure 2-3). In
the first phase, we screened LplA mutants against a panel of alkyl azides to search for an
efficient azide ligase. Second, we compared the efficiency of azide ligation inside live
cells. Third, we examined five different cyclooctynes in their abilities to react with azide
inside the complex environment of live cells. Finally, we demonstrated using two-step
labeling PRIME to label a variety of LAP fusion proteins inside live cells with various
fluorophores whose emission wavelengths span the visible spectrum.
In this chapter, we describe our initial efforts to extend PRIME labeling
technology to be generalizable towards diverse fluorophores structures. We screened a
panel of LplA mutants towards the ligation of an alkyl azide small molecule. We
characterized this new LplA and azide pair in terms of its ligation kinetics, live cell
ligation yield, and ligation specificity. We also developed a method using a nuclear
localized azido-sugar to assay for the efficiency and specificity of the strain-promoted
azide-alkyne cycloaddition reaction inside live cells.
Bio-orthogonal derivatize functional group handle
functional group handle with fluorophore conjugate
LpIA L74plA OLpA
ATP HO
LP2 -
Figure 2-3. Two-step LplA ligation using functional group handle and chemoselective
bioorthogonal reaction inside live cells.
41
Screening for an efficient LplA ligase and azide pair for site-specific protein labeling
inside live cells
The first step to generalize two-step PRIME towards targeting of diverse
fluorophore structures was to engineer LplA to ligate a small molecule bioorthogonal
functional handle specifically to LAP fusion proteins inside live cells. The functional
handle can then be chemoselectively derivatized with fluorophore of interest via a
bioorthogonal reaction. At the time of the start of the project, in the fast growing area of
potentially bioorthogonal reactions, very few have been tested rigorously in a complex
biological context, let alone inside mammalian cells. Among the top candidates are the
Staudinger ligation (11-14) and the strain-promoted azide-alkyne cycloaddition (15-17).
Despite its impressive biological applications on live cell surface (18) and in mice (13,14),
the Staudinger ligation has long been plagued by phosphine oxidation, aza-ylide
hydrolysis, and slow reaction rate (15,19). The strain-promoted azide-alkyne
cycloaddition is the most promising, especially given its previous demonstrations on the
cell surface (3).
Previously, in Dr. Marta Fernandez-Suirez's work (3), two-step fluorophore
targeting via LplA azide ligation was accomplished on cell-surface proteins LDLR and
EGFR, which were fused to the 22-amino acid peptide called LAP1 that was derived
through structure comparison and site-directed mutagenesis from the 9 kDa lipoyl
domain of the full length E2p protein. The cell-surface LAP1-fused proteins were
exogenously ligated with azide using purified wild-type LplA protein. The azide
functional group on the cell surface was then chemoselectively derivatized through the
strain-promoted cycloaddition reaction with cyclooctyne that was conjugated to a cyanine
dye. Dr. Marta Femindez-Suirez screened a panel of alkyl azides with different alkyl
chain lengths against wild-type LplA in an HPLC assay. Through comparison of in vitro
labeling kinetics, 8-azido-octanoic acid, or azide 7, was most efficiently incorporated by
wild-type LplA onto the E2p substrate with a kcat of 6.66 min' and Km of 127 ptM (3).
The Km value was 75-fold increased, while the kcat value was 2-fold decreased compared
to the analogous reaction performed with lipoic acid.
While wild-type LplA and azide 7 worked well for live-cell cell surface labeling,
several problems appeared when we tried to apply the ligation inside live cells. First, after
42
azide ligation inside live HEK 293T cells expressing wild-type LplA and the 22-mer
LAP1, we fixed the cells and labeled them with alkyne functionalized cyanine dye
through CuAAC to detect ligated azides. Even though strong signal was observed only in
cells expressing both LplA and LAP1, negative control reactions with cells expressing
the Lys->Ala mutant of LAP1 or only the LplA enzyme also showed a transfection-
specific signal. Second, when labeling LAP 1 localized to only the nucleus, background in
the cytosol specific to transfected cells was also observed. Third, the washout of the
probe, monofluorinated cyclooctyne conjugated to fluorescein, was very poor or never
washed out if the probe was loaded more than 50 tM. Furthermore, labeling was
inconsistent among several cell lines tested, HEK 293T, HeLa, and COS-7 cells, with the
strongest signal in HEK 293T cells, and very weak or no signal in COS-7 cells. There are
several possible explanations for the observed intracellular labeling problems. If
unligated azides were not washed out completely and stick nonspecifically to cellular
organelles and hydrophobic surfaces, they would also be free to react with cyclooctyne-
fluorophore in the second step to give rise to unwanted background. This was
problematic not only because it would give high background, unligated azides can also
compete with specifically ligated azides to react with cyclooctyne, reducing specific
signal. With respect to the monofluorinated cyclooctyne-fluorophore, the difficulty with
probe washout could be caused by its own hydrophobic nature, nonspecific reaction with
excess azides inside cells, or reaction with cellular nucleophiles such as thiols. Although
the monofluorinated cyclooctyne has not been experimentally proven to react with thiols,
its derivative, the difluorinated cyclooctyne DIFO, showed evidence of reacting with Cys
and glutathione in vitro at the strained alkyne (20). Chang et al also reports DIFO sticks
heavily to mouse serum albumin likely because of a combination of hydrophobic and
covalent interactions (21). Therefore, the sequential nature of the protocol and various
unknown factors inside the cell caused the analysis and troubleshooting of two-step
labeling results especially complicated, making it difficult to catch the culprits that were
causing problems like nonspecific signal and high background.
Furthermore, since the development of the LAP1 peptide and the application of it
to cell-surface protein labeling (3), Dr. Sujiet Puthenveetil engineered an even smaller
13-mer peptide substrate called LAP2 (or simply LAP in this thesis) through yeast
43
display evolution (4). The LAP peptide has a kcat/Km of 0.99 pMmin-' for wild-type
LplA catalyzed lipoic acid ligation reaction, about 70-fold better than the same reaction
performed previously on the 22-mer LAP1. In light of the new LAP, as well as new
gained insights regarding LplA engineering and mutagenesis, we wished to screen for an
improve LplA ligase and azide substrate pair to use for fluorophore labeling inside live
mammalian cells.
To extend the two-step fluorophore labeling scheme inside cells, we decided to
optimize each step independently to obtain more clear results without the confounding
factors associated with the other step. In the first step of azide ligation, we needed the
LplA enzyme to specifically and efficiently ligate the alkyl azide small molecule onto
LAP-fused protein substrates with high ligation yield in the presence of other intracellular
proteins. We wished to avoid azide-dependent background possibly caused by excess
unligated azides remaining inside cells. Even though wild-type LplA can catalyze the
conjugation of azide 7 onto LAP1-fused proteins expressed on the cell surface, where the
exogenous azide probe can be loaded at high concentrations then washed away cleanly
without worrying of sticking to cellular organelles, for intracellular azide ligation, we
preferred to deliver the azide probe at lower concentrations so that less residual azide
remains after ligation reaction, and to minimize interference with the subsequent
cycloaddition reaction with cyclooctyne. We would also like the new LplA enzyme and
azide pair to have an improved kcat of LplA azide ligation to achieve efficient ligation
yield within a short labeling time. Therefore, to use lower azide concentrations without
sacrificing azide ligation yield, we needed to engineer the LplA-catalyzed azide ligation
reaction to improve its catalytic efficiency.
44
Figure 2-4. Small molecule substrate binding pocket of E. coli lipoic acid ligase (LplA)
with bound lipoyl-AMP (PDB 3A7R). Distance between the dithiolane ring of lipoyl-
AMP and the "gatekeeper" residue W37 of LplA is 3.628 A.
To screen for a new LplA azide ligase, we consulted the previous azide ligation
screen by wild-type LplA completed by Dr. Marta Femindez-Suirez as well as examined
the lipoic acid binding pocket of LplA (Figure 2-4). In Dr. Fernandez-Suilrez's work, she
observed that 8-azido-octanoic acid (azide 7) with 7 methylenes as the alkyl chain length
was the best substrate for wild-type LplA, while other azides with either shorter or longer
alkyl chains did not get ligated as well (3). This data indicated that perhaps alkyl azides
longer than azide 7 were too larger to fit in the LplA binding pocket; while alkyl azides
shorter than azide 7 were probably too small and could not form enough hydrophobic
interactions with the active site residues of wild-type LplA to enable ideal binding and
ligation. In addition, previous lab member Dr. Hemanta Baruah laid the groundwork in
identifying amino acid residues in the LplA binding pocket that are important in
modulating the enzyme's plasticity towards its small molecule substrates. Through
alanine screening mutagenesis of residues that are within 7.5 A of the dithiolane ring of
45
lipoic acid, Dr. Baruah found W37 of lipoic acid ligase to be a "gatekeeper" residue that
when mutated to smaller residues like V and S, enabled the LplA mutant to accept a
larger aryl azide substrate (9). Having the W37 residue as a lead, Dr. Chayasith
Uttamapinant and co-workers later identified W37V mutant of LplA to be an efficient
coumarin ligase (5). To find an alkyl azide ligase, we engineered the binding pocket of
LplA by mutating W37 residue to six other smaller and mainly hydrophobic amino acids
- G, A, V, I, L, S - and screened the mutants against a panel of alkyl azide substrates of
increasing alkyl chain lengths, with the shortest azide being azide 7 (Figure 2-5).
Alkyl azide small molecules were synthesized using the corresponding
bromoalkanoic acid and sodium azide as described in Experimental Methods. LplA
mutant proteins were expressed and purified from E. coli. We set up in vitro ligation
reactions using four different alkyl azide small molecules, six purified LplA proteins, and
the 13-amino acid LAP peptide as the substrate. An HPLC assay was used to determine
the percent conversion of LAP into LAP-azide conjugate under two different reaction
conditions: 500 gM probe for 10 minutes, and 20 pM probe for 20 minutes. We found
that for the shortest azide 7 probe, wild-type LplA and LplA(W37V) were the top two
ligases (Figure 2-5). As the length of the alkyl chain increases, wild-type LplA with the
large W37 "gatekeeper" residue was no longer active, instead LplA(W37V) and
LplA(W371) mutants were the best. The top four best LplA ligase/alkyl azide probe pairs
are starred in Figure 2-5.
% Conversion to product with LpIA variants
Probe
WT W37G W37A W37V W371 W37L W37S
n=8 100%
O Azd 8 7 N.D. N.D 11% 6% <3% N.D.
HO7 
n=
N.. 3% <9   
Azide 9
n=91 ..0% 9 3% ND
Ad 10 N.D. 12% 10% N.D. N.D.
0%
46
Figure 2-5. Screening to identify the best LplA mutant/azide substrate pair. The table
shows percent conversions of LAP to the LAP-azide product conjugate. Wild-type LplA
and six W37 point mutants were screened against four azidoalkanoic acid substrates of
various lengths. N.D. indicates that product was not detected. Screening was performed
with 100 nM ligase, 600 pM LAP, and 20 pM azide substrate for 20 min at 30 'C.
Conversions were measured in duplicate. Note that LplA(W37S) was active with the
natural substrate, lipoic acid (data not shown). The starred entries in the table were
further evaluated in Figure 3-2.
47
Characterization of an improved LplA enzyme and azide probe pair
Kinetics properties of LplA(W3 71) ligation of azide 9 onto LAP peptide
Even though from our in vitro HPLC assay LplA(W37V) and azide 9 pair gave
the highest azide ligation yield, we later discovered through a live-cell fluorophore
targeting imaging assay that LplA(W371) and azide 9 actually performed better inside
live cells (described later in Chapter 3; Figure 3-2). Therefore, we picked LplA(W371)
and azide 9 as our best enzyme/azide pair and proceeded to compare the kinetic
properties of LplA(W371) and azide 9 pair to the previous wild-type LplA and azide 7
pair used in cell-surface protein labeling. Figure 2-6 A shows an HPLC analysis of
LplA(W371)-catalyzed ligation of azide 9 onto LAP peptide. The identity of the LAP-
azide 9 product peak was confirmed by mass spectrometry (Figure 2-6 B). Negative
control reactions with ATP omitted or wild-type LplA in place of LplA(W371) were also
analyzed and showed no product formation. We also used HPLC to quantify product
amounts in order to measure kcat and Km values. Figure 2-6 C shows the Michaelis-
Menten plot giving a kcat of 3.62 min-' and a Km of 35 pM for azide 9 ligation catalyzed
by LplA(W371). Compared to our previously reported azide 7 ligation catalyzed by wild-
type LplA on E2p, this improved Km is 4-fold lower. The kcat is 1.8-fold reduced, giving
an overall improvement in kcat/Km of 2-fold.
48
BLpLA
C
C)
OmitATP
wTLpIA
I 1 1 i
7 11 15 19 min
I
100
11.4
00.0
4.W
2.0
10.0
[M+2H] 2+
883.7400
003 -400
ISO I 15 ,se
Si0 O00 1000 1100 1200 1300 A ~ 100 1000 17 00
m/z, amu
LAP-azide 9 conjugate
Calculated mass: 1764.16
Observed mass: 1765.48
kc 3.62 10.012 min-'
Km = 35?5 pM
0 100 200 0o 400 500 bo 00 700 600
[Azide 9], pM
Figure 2-6. In vitro characterization of azide 9 ligation catalyzed by LplA(W371). (A)
HPLC trace showing LAP conversion to LAP-azide 9 conjugate, catalyzed by
LplA(W371) (black trace). Negative controls are shown with ATP omitted (red trace) and
LplA(W371) replaced by wild-type LplA (blue trace). Reactions were performed for 2
hours with 1 pM LplA, 300 pM LAP, and 500 p.M azide 9. (B) ESI mass spectrometric
analysis of LAP-azide 9 covalent adduct (starred product peak in (A)). (C) Michaelis-
Menten curve and kinetic parameters for azide 9 ligation onto LAP, catalyzed by 100 nM
LplA(W371). Initial rates (Vo,) were measured using the HPLC assay in (A) with 600 [.M
LAP and 25-700 gM azide 9. Each V0 value was measured in triplicate. Error bars, ?1 s.d.
CuAAC detection of azide ligation in fixed cells
After finding a kinetically improved LplA enzyme and azide pair, we evaluated
the specificity of azide ligation using CuAAC as detection method. In this assay, live-cell
49
A LAP-azide 9 conjugate
LAP
C
C
-L
"
045 -
0A0
0.35-
0.30-
025
0.20-
016-
0.10-
0.06-
o oo
azide ligation is followed by fixing and permeabilizing the cells for reaction with
terminal alkyne conjugated fluorophore via CuAAC. This "live-fix" protocol provided us
with several advantages that were helpful in evaluating live-cell azide ligation.
Performing enzymatic azide ligation in live cells allowed us to critically exam the
ligation specificity and efficiency in its natural environment. Having fixed and
permeabilized the cells after azide labeling, we washed out all excess azides, except
LAP-ligated azides as well as any unspecific ligation or other background reaction
between azide and cellular molecules. Finally, detection with terminal alkyne-
fluorophore conjugate via CuAAC has high sensitivity that enabled us to uncover any
background reaction that could potentially interfere with the strain-promoted azide-
alkyne cycloaddition step.
To perform azide 9 ligation inside live cells, HEK 293T cells were transfected
with LplA(W371) and nuclear-targeted, LAP-tagged YFP. After incubating the cells with
200 ltM azide 9 for 1 hour at 37C, unligated azide 9 was washed out of cells by
incubating cells in fresh growth media for 1 hour. Following live cells azide 9 ligation
and washout, cells were fixed using paraformaldehyde, permeabilized using cold
methanol, blocked with BSA, and reacted with cy5-alkyne conjugate via CuAAC. Figure
2-7 shows cy5 signal co-localizing with LAP-YFP nuclear expression, indicating that
LplA(W371) azide 9 ligation is specific to LAP.
Cy5 nuclear LAP-YFP LpIA-mCherry Cy5+DIC
wild-type LpIA
azide 7
LpIA(W371)
azide 9
LpIA(W371)
Omit azide 9
50
Figure 2-7. Fixed cell labeling on nuclear localized LAP-YFP with cy5-alkyne via
CuAAC. HEK 293T cells expressing nuclear LAP-YFP with either wild-type LplA or
LplA(W371) were labeled with azide 7 or azide 9, respectively. After azide washout, cells
were fixed and permeabilized. Fixed cells were reacted with cy5-alkyne overnight,
washed to remove excess cy5-alkyne and imaged on the confocal microscope.
Estimating azide ligation yield in live cells
To have high labeling signal in the two-step fluorophore targeting scheme, we
needed the azide ligation step and the strain-promoted cycloaddition both to achieve
quantitative yield. A high azide ligation yield is especially instrumental to achieving this
goal. We have shown that the new LplA(W371) and azide 9 pair has superior kinetic
properties compare to wild-type LpIA and azide 7 pair. We wanted to see if the in vitro
result were transferred to better ligation efficiency inside live cells. Two methods were
developed to evaluate the yield of live-cell azide ligation. Both methods are based on the
shift in migrating position of a LAP-tagged FP protein in an acrylamide gel. The first
method is based on the difference in charge between free and azide-ligated LAP-protein.
Ali and Guest first observed that lipoic acid modified E2p moves faster than the apo E2p
in a non-denaturing native gel owing to the loss of the positively charged lysine (1). We
applied the same assay to azide-ligated LAP-protein and free LAP-protein. The loss of
the positive charge on the catalytic lysine residue in the azide-ligated LAP-protein causes
it to run faster comparing to unligated LAP-protein in native PAGE gel. Hence, we call
the method native gel-shift assay. Since azide-ligated LAP-FP protein effectively
separates itself from unmodified LAP-FP protein by its faster migration on a native gel,
the yield of azide ligation onto LAP-protein can be determine by the ratio of densitometry
of each protein band. The second method is based on the selective binding of a
streptavidin molecule onto azide ligated LAP-FP protein, causing it to migrate much
slower because of the increased molecular weight of 56 kDa than free LAP-FP protein on
an SDS-PAGE gel. We named the second method streptavidin gel-shift assay.
- Native Gel-shift Assay
To compare the live-cell azide ligation efficiencies of the newly improved
LplA(W371) and azide 9 pair and the wild-type LplA and azide 7 pair using native gel-
51
shift assay, we incubated HEK 293T cells expressing LAP-YFP and either wild-type
LplA or LplA(W371) with azide 7 or azide 9 for 30 minutes or 1 hour. Following azide
incubation, cells were washed by incubating in growth media to remove excess unligated
azides. After excess azide washout, we lysed the cells in non-detergent containing lysis
buffer and the lysates were ran on a native PAGE gel. LAP-YFP was detected by YFP in-
gel fluorescence. The results shown in Figure 2-8 demonstrated that the unmodified LAP-
YFP protein (black arrow) runs at an apparent molecular weight of approximately 42 kDa
(Figure 2-8, left, lane 2; right, lane 4). Upon azide modification, we observed that the
entire LAP-YFP-azide complex migrated faster on the native PAGE gel with the
positively charged lysine residue of LAP being converted to a neutral amide (Figure 2-8,
left, lane 1; right, lane 1-3). The apparently molecular weight of the azide-ligated LAP-
YFP (red arrow) is about 38 kDa. We further confirmed that the migration pattern is
indeed caused by the presence or absence of the positive charge on the lysine residue by
performing an additional negative control reaction with LAP (K->A)-YFP where the
lysine residue is mutated to an alanine. LAP (K-A)-YFP also migrated to around 38
kDa (Figure 2-8, left, lane 3), same as azide-ligated LAP-YFP. The improved LplA(W371)
and azide 9 pair gave nearly quantitative ligation after only 30 minutes of azide 9
incubation (Figure 2-8, right, lane 2), while wild-type LplA and azide 7 pair only gave 73%
ligation yield after 1 hour of labeling in live cells, estimated by densitometry (Figure 2-8,
right, lane 1). We were glad to see that LplA(W37I) and azide 9 can achieve quantitative
LAP ligation yield with short azide incubation time.
52
1. w371LplA + Azide 9, 1 hr 1. wTLplA + Azide 7, 1 hr
2. Omit Azide 9 control 2. W37'LplA + Azide 9, 30 min
3. LAP(K->A)-YFP control 3. W37'LplA + Azide 9, 1 hr
4. Omit Azide 9 control
1 2 3 1 2 3 1 2 3 4 1 2 3 4
230 230
150 150
100 100
80 80
60 60
30 30
25 25
kD 91% kD 73%95%94%
Coomassie YFP Coomassie YFP
fluorescence fluorescence
Figure 2-8. Characterization of intracellular azide ligation yields by native gel-shift assay.
Labeling protocol: HEK 293T cells co-expressing LplA and LAP-tagged YFP were
labeled with azide 7 or azide 9 for 30 minutes or 1 hour, washed for 1 hour. Cells were
lysed and run on a native 12% polyacrylamide gel. Labeled LAP-YFP runs faster than
unlabeled LAP-YFP due to removal of a positive charge. (Left) Gel-shift analysis of
azide ligation in cells. Lane 1 is the same as lane 3 on the right. Lane 2 shows a negative
control with azide 9 omitted. Lane 3 shows a negative control with the lysine in LAP
mutated to alanine. This mutation causes LAP-YFP to shift downward in the 12%
polyacrylamide native gel because it removes one positive charge from LAP-YFP.
Similarly, labeling of the lysine of LAP causes a downward shift because the positive
charge of lysine is converted to a neutral amide. (Right) HEK 293T cells were prepared
and labeled with either azide 7 or azide 9 as described previously. Percent conversions to
azide-LAP-YFP product are given at the bottom of the YFP fluorescence gel image. Lane
4 shows a negative control with azide 9 omitted. Estimated percent conversions to
product are given at the bottom of the YFP fluorescence gel image.
-Streptavidin gel-shift assay
53
The second approach to estimate the yield of azide ligation on LAP is based on
the attachment of a streptavidin molecule selectively onto azide-derivatized population of
LAP. As depicted in Figure 2-9, after live-cell azide ligation, cells are lysed and lysate is
derivatized with biotin-alkyne via CuAAC. Biotin-alkyne reacts specifically with azide-
ligated LAP proteins. After removal of excess biotin-alkyne, excess streptavidin is added.
Upon binding with biotin-alkyne derivatized azide-LAP protein, streptavidin molecule
causes a decrease in motility of azide-LAP protein by increasing its molecular weight by
56 kDa. However, in order to make any conclusion regarding azide ligation yield, we
have to ensure that the two subsequent steps following azide ligation - biotin-alkyne
CuAAC reaction and streptavidin binding - are all proceeding to 100% completion.
add azide
LpIA 0 (7 LpA
ATP 0 N3 1. Cell lysis
HO 0 2. CuAAC with biotin
LAP Wash out H N Na Dialysis to remove
excess azide 9 excess biotin -
LplA 1. Incubate with 
. POI-LAP+SAexcess streptavidin
*, 2 1 /biotin 2. SDS-PAGE
--- 
----- POI-LAP
mCherry
In-gel fluorescence
Figure 2-9. Schematic representation of streptavidin gel-shift assay to measure azide
ligation yield.
HEK 293T cells expressing LAP-mCherry and LplA(W371) were incubated with
azide 9. Following excess azide washout, cells were lysed and the lysate was treated with
biotin-alkyne, CuSO4 , sodium ascorbate and the tris-triazole ligand for CuAAC - tris-
(benzyltriazolylmethyl)amine (TBTA). Under CuAAC, biotin-alkyne would react with
azide-ligated LAP-mCherry proteins. After 16 hours incubation with biotin-alkyne, cell
54
lysates were dialyzed to remove excess biotin-alkyne reagent. Lysates were then
incubated with excess streptavidin for 1 hour at room temperature. Lysates were
separated and analyzed on SDS-PAGE gel. LAP-mCherry was detected by mCherry in-
gel fluorescence. Figure 2-10 shows unligated LAP-mCherry appearing to streak on gel
but nonetheless migrated to about 30 kDa. We checked in a separate experiment that the
streakiness of the bands was caused by copper in the CuAAC reaction (data not shown).
The streptavidin-bound azide-ligated LAP-mCherry appeared at around 80 kDa. Negative
control reactions with either azide 9 omitted, biotin-alkyne omitted, or streptavidin
omitted did not show increase in molecular weight to LAP-mCherry. The azide ligation
yield is high, leaving only small amount of LAP-mCherry showing weak fluorescence at
30 kDa (Figure 2-10, lane 3).
azide9acid + + + - - -
biotin-alkyne - + + - + +
Streptavidin (SA) - - + - - +
1 2 3 4 5 6 1 2 3
150 -
100-
80-
60-
50-
40-
30II
25 -
kDa
mCherry mCherry
fluorescence Fluorescence
(Higher exposure)
Figure 2-10. Characterization of intracellular azide ligation yields by streptavidin gel-
shift assay. Labeling protocol: HEK 293T cells co-expressing LAP-tagged mCherry wild-
type LplA or LplA(W37I) and were labeled with azide 9 for 1 hour, washed for 1 hour to
remove excess azide. Cells were lysed using hypotonic lysis. Click reaction was
55
; - - - - ----- - - .. . - ... _.. - --- I'll, "I'll - -" -".- - - - " -" -__'_ _-, - -- -___ --- --
performed in cell lysate with conditions as follows: Biotin-alkyne 100 pM, CuSO 4 1mM,
tris-(benzyltriazolylmethyl)amine (TBTA) ligand 100 gM, 2.5 mM sodium ascorbate in
1mM HEPES buffer. Reactions were incubated at RT for 16 hours. The lysate samples
were then dialyzed in IX PBS buffer for 4 hours to get rid of excess biotin-alkyne, and
incubated with excess streptavidin (Cf =1OuM) for 1 hour at RT. Lysates were loaded
ontol2% SDS-PAGE gel without boiling samples in order to preserve mCherry
fluorescence.
56
Exploring strain-promoted azide-alkyne cycloaddition reaction inside live cells using
alternative source of azide
To proceed systematically to develop the two-step fluorophore labeling technique,
we wished to decouple the intracellular strain-promoted cycloaddition second step from
the azide ligation step in order to have a clear assay of the cycloaddition specificity and
yield. We have to analyze the specificity and efficiency of the cycloaddition step
independent of the possible background and non-specific reaction from the azide ligation
step. In order to achieve this, we needed an isolated population of azide inside live cells
that was not from LplA azide ligation. We obtained peracetylated azido-GalNAz to use as
nuclear localized pools of azides inside live cells.
In this approach to independently study the strain-promoted cycloaddition in live
cells, we used an azide-derivatized sugar analog, peracetylated N-
azidoacetylgalactosamine, also called azido-GalNAc or GalNAz as a source of nuclear
localized azide. We acquired azido-GalNAc from the Bertozzi lab at UC Berkeley, who
was using the sugar analog to study proteins bearing O-linked glycans (14,22-24). Azido-
GalNAc is an azide-derivative of GalNAc, whose addition and removal are controlled by
two enzymes glycosyltransferase (OGT) and O-GlcNAcase (OGA), respectively (25).
GalNAc glycosylation is the most prevalent form of O-linked glycosylation, which puts
the sugar a-linked to Ser or Thr residue. There are two main pathways of GalNAc
incorporation. In the first pathway, GalNAc- 1 -kinase and UDP-GalNAc
pyrophosphorylase convert GalNAc to UDP-GalNAc, which is then recognized by the
enzyme polypeptide N-acetyl-a-galactosaminyltransferase (ppGalNAcT) to put on the
hydroxyl group of Ser or Thr (22,25). The second avenue of GalNAc addition is through
the GlcNAc salvage pathway, which converts GlcNAc to UDP-GalNAc using an
epimerase (22,25). O-GlcNAc glycosylation has been found in numerous proteins
involved in transcription, protein trafficking, cell survival, and neurodegenerative
diseases (25). Since there is no defined sequence consensus for the ppGalNAcT,
predictions of O-linked glycosylated sites have been difficult. However, studies have
shown that O-GlcNAcylation, unlike traditional glycosylation, is nearly exclusively on
cytoplasmic and nuclear proteins (26). It is possible that GalNAz could be converted to
GlcNAc by the epimerase and end up on mainly cytosolic and nuclear proteins. The
57
Bertozzi group observed that metabolic incorporation of GalNAz followed by detection
using Staudinger ligation with phosphine showed predominant presence of GalNAz
labeled proteins in the nucleus as well as weaker signal from the cytosol (personal
communication). Having evidence that GalNAz gets incorporated onto nuclear proteins,
we decided to try it as a nuclear-specific, non-enzymatic source of azide to test our strain-
promoted azide-alkyne cycloaddition reaction in live cells.
We chose CHO cells for this experiment because of its higher rate of metabolic
incorporation compare to other cell lines. CHO cells were plated on glass coverslips
coated with GalNAz or no GalNAz as a negative control. After undergoing metabolic
incorporation for three days, the cells were fixed, permeabilized, and reacted with cy3-
alkyne via CuAAC. We can observe in Figure 2-11 that cells incorporated with azido-
GalNAc gave strong nuclear cy3 signal. Both 100 RM (data not shown) and 200 pM
azido-GalNAc incubation gave similar level of signal. The same experiment was also
demonstrated using HeLa cells to show that azido-GalNAc incorporation followed by
CuAAC detection is generalizable to other cells lines.
Cy3 DIC Merge
- azido-GaNAc
0
+ azido-GaNAc
200 pM
- azido-GaINAc
+ azido-GaINAc
200 pM
58
Figure 2-11. Fixed cell detection of metabolically incorporated azido sugar analog azido-
GalNAc with cy3-alkyne via CuAAC. Metabolic incorporation of azido-GalNAc was
performed in both CHO (top) and HeLa (bottom) cells. Cells were incubated with azido-
GalNAc for three days followed. by fixation and reaction with cy3-alkyne. Strong cy3
signal was observed in the nucleus as well as weaker signal throughout the cytosol
because azido-GalNAc also weakly labels O-linked glycoproteins in the cytosol.
Negative control with no azido-GalNAc added to cells shows no cy3 retention.
The success of live cell azido-GalNAc incorporation and fixed cell cy3-alkyne
labeling via CuAAC gave us confidence that cells can successfully take up the azido
sugar analog to a detectable level and the majority of the signal localized to the nucleus.
We tried live cell strain-promoted reaction with a cyclooctyne-fluorophore conjugate
after azido-GalNAc metabolic incorporation (Figure 2-12, top scheme). After azido-
GalNAc incubation, we replaced cell culture media with media containing a cyclooctyne
derivatized fluorophore, monofluorinated cyclooctyne-carboxyfluorescein diacetate
(MOFO-CFDA, Figure 2-12). At physiological pH, fluorescein interchanges between the
monoanion and the dianion form, rendering the probe membrane impermeant. We
protected the phenolic oxygens in fluorescein with acetate groups to make MOFO-
fluorescein diacetate (MOFO-CFDA), which is nonfluorescent and can readily across the
cell membrane. Once inside the cells, the acetate esters can be cleaved by endogenous
cellular esterases to restore fluorescence. Gradually, the excess MOFO-fluorescein is
removed from cells by the action of nonspecific anion transporters. We washed out the
unreacted probe by replacing MOFO-CFDA containing media with fresh media three
times and further incubated the cells in fresh media for an hour to rely on the cell's
nonspecific anionic transporters to pump out excess probe.
Figure 2-12 shows MOFO-CFDA nuclear signal intensity and distribution at
various time points after dye incubation. Clear nuclear signal can be seen in most cells,
but signal distribution has a large variation from cell to cell, perhaps due to the different
levels of metabolic incorporation of azido-GalNAc. Metabolic incorporation of azido-
GalNAc followed by detection using cyclooctyne-fluorescein conjugate demonstrated
that strain-promoted cyclooctyne is able to react specifically with a localized population
59
of azide and the excess probe can be washed out of cells without significant background
reaction with small molecules such as thiols in the cytosol.
OAcO
AcO,,4r7%Ac, add
I nudew(&HOcytosol nudeus
Metabolic Wash out
labeling excess probe
Fluorescein DIC Merge
F '
30 min.
N,
Monofluorinated cyclooctyne
-Carboxyfluorescein-d iacetate
(MOFO-CFDA)
50 min.
after fixation
Figure 2-12. Live cell detection of metabolically incorporated azido sugar analog azido-
GalNAc with fluorescein conjugated to cyclooctyne via strain-promoted azide-alkyne
cycloaddition. (Top) Labeling scheme. Metabolic incorporation of azido-GalNAz was
performed in CHO cells following protocol previously described. Afterwards, cells were
incubated with 10 gM MOFO-CFDA (structure shown below) for 10 min. (Bottom)
Images of live cells were taken at time points (indicated on the left side of images) post
MOFO-CFDA addition. Strongest nuclear fluorescein signal was observed within the
first hour after MOFO-CFDA labeling. Cells were also imaged after fixation and
permeabilization to demonstrate the retention of nuclear signal and also to completely
wash out any remaining unreacted MOFO-CFDA.
60
Conclusion
In this chapter, we described the screening and characterization of a new LplA
ligase and azide pair, evaluation of azide ligation specificity, and performed strain-
promoted azide-alkyne cycloaddition using an azido-sugar analog inside live mammalian
cells. We developed assays to examine the two-step fluorophore targeting scheme
systematically by isolating the azide ligation step from the strain-promoted azide-alkyne
cycloaddition step so we can troubleshoot easily and identify potential problems without
many confounding and unknown factors inside the cell.
The in vitro characterization shows the LplA(W371) and azide 9 pair has an
improved kcat/Km of two-fold over the wild-type LplA and azide 7 pair. Inside live
mammalian cells, LplA(W371)/azide 9 is able to achieve quantitative azide ligation yield
over an incubation time of a brief 30 minutes, while wild-type LplA/azide 7 only had 73%
yield over an hour. Furthermore, LplA(W371)/azide 9 is also highly specific to the LAP-
tagged protein. We did not observe cytosolic background signal when labeling nuclear-
localized LAP-YFP. The strain-promoted cycloaddition reaction was tested using an
LplA-independent, nuclear-localized source of azide, azido-GalNAc, and
monofluorinated cyclooctyne conjugated to fluorescein (MOFO-CFDA). The reaction
proceeded successfully inside live cells, giving strong nuclear fluorescein signal. The
excess MOFO-CFDA probe was washed out of cells within an hour.
In the next chapter, we discuss using this optimized platform to evaluate the
cycloaddition efficiencies of different cyclooctynes and demonstrating various large
fluorophore targeting to different proteins across multiple cell lines. We also discuss
some of the challenges and problems we faced in labeling with certain fluorophores
inside live mammalian cells and ways we tried to overcome these challenges.
61
Experimental methods
General synthetic methods
All reagents were the highest grade available and purchased from Sigma-Aldrich,
Anaspec, Thermal Scientific, TCI America, Alfa Aesar, or Life Technologies and used
without further purification. Anhydrous solvents were drawn from Sigma-Aldrich
SureSeal bottles. Analytical thin layer chromatography was performed on 0.25 mm silica
gel 60 F254 plates and visualized under short or long wavelength UV light, or after
staining with bromocresol green or ninhydrin. Flash column chromatography was carried
out using silica gel (ICN SiliTech 32-63D). Mass spectrometric analysis was performed
on an Applied Biosystems 200 QTRAP mass spectrometer using electrospray ionization.
HPLC analysis and purification were performed on a Varian Prostar Instrument equipped
with a photo-diode-array detector. A reverse-phase Microsorb-MV 300 C18 column (250
x 4.6 mm dimension) was used for analytical HPLC. NMR spectra were recorded on a
Bruker AVANCE 400 MHz instrument.
Synthesis of alkyl azide probes
9 NaN 3  0
HO Br DMF HO N3
n n
n=7-10
To a solution of the corresponding bromoalkanoic acid (~1 g, 5 mmol) in 10 mL N,N-
dimethylformamide (DMF) was added sodium azide (~0.5 g, 7.5 mmol). The mixture
was allowed to stir at room temperature overnight. The progress of the reaction was
monitored by thin layer chromatography (1:2 hexanes:ethyl acetate) followed by
bromocresol green stain. Upon completion, DMF was removed under reduced pressure.
The resulting residue was re-dissolved in 15 mL of 1 M HCl and extracted with ethyl
acetate (3 x 15 mL). The organic layer was dried over magnesium sulfate, then filtered.
After removal of ethyl acetate in vacuo, the crude product was purified by silica gel
chromatography (solvent gradient 0-15% ethyl acetate in hexanes) to afford the
corresponding azidoalkanoic acid as clear or pale yellow oil. Yields ranged from 50-70%.
62
Characterization of n =7 azide (8-azidooctanoic acid). 'H NMR (CDCl 3): 11.87 (s, lH)
3.20 (t, 2H, J= 6.9), 2.28 (t, 2H, J= 7.5), 1.56 (m, 5H), 1.33 (m, 5H). ESI-MS calculated
for [M-H]F: 184.11; observed 183.66.
Characterization of n =8 azide (9-azidononanoic acid). 1H NMR (CDC13) 3.22 (t, 2H, J=
6.9), 2.30, (t, 2H, J= 7.5), 1.60 (m, 5H), 1.29 (m, 7H). ESI-MS calculated for [M-H]:
198.12; observed 198.65.
Characterization of n=9 azide (10-azidodecanoic acid). 'H NMR (CDCl 3): 3.23 (t, 2H, J
= 6.9), 2.28 (t, 2H, J= 7.5), 1.53 (m, 5H), 1.31 (m, 9H). ESI-MS calculated for [M-H]~:
212.14; observed 212.28.
Characterization of n =10 azide (I1-azidoundecanoic acid) 'H NMR (CDCl 3) 3.27 (t, 2H,
J= 7.1), 2.39, (t, 2H, J= 7.5), 1.65 (m, 5H), 1.20 (m, 11H). ESI-MS calculated for [M-
H]~: 226.16; observed 226.12.
Protein expression and purification
Wild-type LplA and its mutant variants, and E2p were expressed in BL21-DE3 bacteria
with IPTG induction as previously described (2). After Ni-NTA purification, LplA
enzymes were dialyzed and stored in 20 mM tri pH 7.5, 1 mM DTT, and 10% glycerol in
aliquots at -80'C. Protein concentrations were determined using the bicinchoninic acid
assay (Pierce).
Kinetic analysis of azide 9 ligation (Figure 2-6 C)
Reactions were set up as described in the figure legend. Aliquots were taken and
quenched before product conversion exceeded 5%. To calculate initial rates, we
determined the amount of product at each time point by generating a calibration curve
using purified LAP and LAP-azide 9 mixed at different ratios. This curve correlated the
measured ratio of integrated HPLC peak areas to the actual ratio, i.e. adjusted for any
differences in extinction coefficient of LAP vs. LAP-azide 9. Initial rates (V,) were
determined at each azide 9 concentration by plotting the amount of LAP-azide 9 product
against time. The slope of the line gives Vo. Vo values were then plotted against azide 9
concentration in Figure 2-6 C, and Origin 8.5.1 was used to fit the curve to the Michaelis-
Menten equation V0 = Vma[azide 9] / (K+ [azide 9]). From the Vmax, kcal was calculated
63
using Vmax = kcat[E]totai. Measurements of V0 values at each azide 9 concentration were
performed in triplicate.
Detection of live cell azide ligation using cy5-alkyne via CuAAC (Figure 2-7)
HEK 293T cells plated on fibronectin-coated glass coverslips in wells of a 48-well cell
culture plate (0.95 cm2 per well) were transfected with 400 ng pcDNA3-LAP-YFP-NLS
(nuclear-localized) and either 20 ng pcDNA3-wild-type LplA or 200 ng pcDNA3-
LplA(W371). 18 hours after transfection, cells were incubated in growth media
containing 200 pM azide 7 or azide 9 for 1 hour at 37C. Excess azide probe was washed
out by replacing azide-containing media with fresh media 3x over 1 hour. After azide
washout, cells were washed 3x with DPBS and fixed with 3.7% formaldehyde in DPBS
pH 7.4 for 10 min at room temperature followed by cold precipitation with methanol for
5 min at -20 "C. Cells were then blocked with 3% (w/v) BSA in DPBS for 1 hour at room
temperature. To react the fixed cells with cy5-alkyne, cells were incubated in reaction
mixture containing 50 RM cy5-alkyne, 1mM CuSO 4, 100 RM TBTA ligand, and 2.5 mM
sodium ascorbate in DPBS overnight at room temperature. Cells were then washed 3x
with DPBS and imaged.
Native gel-shift analysis of azide 7 and azide 9 ligation yields in cells (Figures 2-8)
HEK 293T cells were plated into wells of a 12-well culture plate (4 cm 2 per well) 18
hours prior to transfection and grown to 60% confluency. For azide 7 ligation, cells were
transfected with 50 ng wild-type LplA and 1000 ng pcDNA3-LAP-YFP. For azide 9
ligation, cells were transfected with 500 ng LplA(W37I) and either 1000 ng pcDNA3-
LAP-YFP or pcDNA3-LAP(K->A)-YFP using Lipofectamine 2000 (Life Technologies).
The LplA:LAP plasmid ratios are identical to the conditions used for imaging in Figures
3-2 and 3-4. 18 hr after transfection, cells were incubated in growth media (MEM
supplemented with 10% FBS) containing 200 pM azide 7 or azide 9 for 30 minutes or 1
hour at 37 *C. Excess azide probe was washed out over 1 hour. Cells were then harvested
and lysed in 500 pL hypotonic lysis buffer (1 mM HEPES pH 7.5, 5 mM MgCl 2, 1 mM
PMSF (Thermal Scientific, phenylmethanesulfonyl fluoride), 1 mM protease inhibitor
cocktail (Sigma-Aldrich)), frozen at -20 "C, thawed at room temperature, then mixed by
64
vortexing for 2 min. This freeze-thaw-vortex cycle was repeated three times. Cells were
then centrifuged at 13,000 rpm for 2 min, and the supernatant was separated and analyzed
on a 12% polyacrylamide native gel without SDS (5 ptL lysate per lane) at constant 200 V.
Prior to Coomassie staining, in-gel fluorescence of YFP was visualized on a FUJIFILM
FLA-9000 instrument using LD473 laser and Long Pass Blue (LPB) filter. A repeat of
the experiment in Figure 2-8 gave ligation yields of 67% for WTLplA (50 ng plasmid) +
azide 7, and 89% for W371LplA (500 ng plasmid) + azide 9 (data not shown).
Streptavidin gel-shift analysis of azide 9 ligation yield in cells (Figure 2-10)
HEK 293T cells were plated into wells of a 12-well culture plate (4 cm2 per well) 18
hours prior to transfection and grown to 60% confluency. For azide 9 ligation, cells were
transfected with 500 ng LplA(W371) and 1000 ng pcDNA3-LAP-mCherry using
Lipofectamine 2000 (Life Technologies). 18 hours after transfection, cells were incubated
in growth media (MEM supplemented with 10% FBS) containing 200 pM azide 9 for 1
hour at 37 "C. Excess azide probe was washed out over 1 hour. Cells were then harvested
and lysed in 500 ptL hypotonic lysis buffer (1 mM HEPES pH 7.5, 5 mM MgCl 2 , 1 mM
PMSF (Thermal Scientific, phenylmethanesulfonyl fluoride), 1 mM protease inhibitor
cocktail (Sigma-Aldrich)), frozen at -20 "C, thawed at room temperature, then mixed by
vortexing for 2 min. This freeze-thaw-vortex cycle was repeated three times. Cells were
then centrifuged at 13,000 rpm for 2 min, and the supernatant was collected. Click
reaction was performed in cell lysate supernatant with condition as follows: Biotin-
alkyne 100 pM, CuSO 4 1mM, TBTA ligand 100 pM, 2.5 mM sodium ascorbate 2.5 mM.
Reactions were incubated at RT for 16 hours. The lysate samples were then dialyzed in
IX PBS buffer for 4 hours to remove excess biotin-alkyne, and incubated with excess
streptavidin (Cf =lOuM) for 1 hour at RT. Lysates were loaded onto 12% SDS-PAGE gel
without boiling samples in order to preserve mCherry fluorescence.
Detection of metabolic incorporation of azido-GalNAc with cy3-alkyne in fixed cells
(Figure 2-11)
Glass coverslips were first coated with 100 pM or 200 pM azido-GalNAc in ethanol by
letting the ethanol evaporate inside the cell culture fume hood. CHO cells or HeLa cells
65
were then plated on azido-GalNAc-coated glass coverslips at 40% confluency in wells of
a 48-well cell culture plate (0.95 cm 2 per well) and let grown for 2-3 days. After
metabolic incorporation, cells were washed 3x with DPBS and fixed with 3.7%
formaldehyde in DPBS pH 7.4 for 10 min at room temperature followed by cold
precipitation with methanol for 5 min at -20 'C. Cells were then blocked with 3% (w/v)
BSA in DPBS for 1 hour at room temperature. To react the fixed cells with cy3-alkyne,
cells were incubated in reaction mixture containing 50 pM cy3-alkyne, ImM CuSO 4 , 100
pM TBTA ligand, and 2.5 mM sodium ascorbate in DPBS overnight at room temperature.
Cells were then washed 3x with DPBS and imaged.
Detection of metabolic incorporation of azido-GalNAc with MOFO-CFDA in live
cells (Figure 2-12)
Glass coverslips were first coated with 200 RM azido-GalNAc in ethanol by letting the
ethanol evaporate inside the cell culture fume hood. CHO cells or HeLa cells were then
plated on azido-GalNAc-coated glass coverslips at 40% confluency in wells of a 48-well
cell culture plate (0.95 cm2 per well) and let grown for 2-3 days. After metabolic
incorporation, cells were incubated in growth media containing 10 gM MOFO-CFDA for
10 minutes at 37C. Excess MOFO-CFDA was washed out of cells by replacing media
with fresh media 3x over 1 hour. Cells were imaged live at 30 minutes and 50 minutes
post MOFO-CFDA addition. Afterwards, cells were washed 3x with DPBS, fixed with
3.7% formaldehyde in DPBS pH 7.4 for 10 min at room temperature followed by cold
precipitation with methanol for 5 min at -20 *C. Fixed cells were imaged again to
illustrated that the nuclear fluorescence signal can be retained, indicating the reaction was
probably covalent.
66
References
1. Ali ST, Guest JR. Isolation and characterization of lipoylated and unlipoylated
domains of the E2p subunit of the pyruvate dehydrogenase complex of Escherichia
coli. Biochem J. 1990 Oct 1;271(1):139-45.
2. Green DE, Morris TW, Green J, Cronan JE, Guest JR. Purification and properties of
the lipoate protein ligase of Escherichia coli. Biochem J. 1995 Aug 1;309(Pt 3):853-
62.
3. Ferni'ndez-Suirez M, Baruah H, Martinez-Hernindez L, Xie KT, Baskin JM,
Bertozzi CR, et al. Redirecting lipoic acid ligase for cell surface protein labeling with
small-molecule probes. Nature Biotechnology. 2007;25(12):1483-7.
4. Puthenveetil S, Liu DS, White KA, Thompson S, Ting AY. Yeast Display Evolution
of a Kinetically Efficient 13-Amino Acid Substrate for Lipoic Acid Ligase. J. Am.
Chem. Soc. 2009 Nov 18;131(45):16430-8.
5. Uttamapinant C, White KA, Baruah H, Thompson S, Fernindez-Suarez M,
Puthenveetil S, et al. A fluorophore ligase for site-specific protein labeling inside
living cells. PNAS. 2010 Jun 15;107(24):10914-9.
6. Slavoff SA, Liu DS, Cohen JD, Ting AY. Imaging Protein-Protein Interactions inside
Living Cells via Interaction-Dependent Fluorophore Ligation. J. Am. Chem. Soc.
2011 Dec 14;133(49):19769-76.
7. Jin X, Uttamapinant C, Ting AY. Synthesis of 7-Aminocoumarin by Buchwald-
Hartwig Cross Coupling for Specific Protein Labeling in Living Cells.
ChemBioChem. 2011;12(1):65-70.
8. Cohen JD, Thompson S, Ting AY. Structure-Guided Engineering of a Pacific Blue
Fluorophore Ligase for Specific Protein Imaging in Living Cells. Biochemistry. 2011
Sep 27;50(38):8221-5.
9. Baruah H, Puthenveetil S, Choi Y-A, Shah S, Ting AY. An Engineered Aryl Azide
Ligase for Site-Specific Mapping of Protein-Protein Interactions through Photo-
Cross-Linking. Angewandte Chemie International Edition. 2008;47(37):7018-21.
10. Liu DS, Tangpeerachaikul A, Selvaraj R, Taylor MT, Fox JM, Ting AY. Diels-Alder
Cycloaddition for Fluorophore Targeting to Specific Proteins inside Living Cells. J.
Am. Chem. Soc. 2012 Jan 18;134(2):792-5.
11. Saxon E, Bertozzi CR. Cell surface engineering by a modified Staudinger reaction.
Science. 2000 Mar 17;287(5460):2007-10.
12. Kiick KL, Saxon E, Tirrell DA, Bertozzi CR. Incorporation of azides into
recombinant proteins for chemoselective modification by the Staudinger ligation.
Proc. Natl. Acad. Sci. U.S.A. 2002 Jan 8;99(1):19-24.
67
13. Prescher JA, Dube DH, Bertozzi CR. Chemical remodelling of cell surfaces in living
animals. Nature. 2004 Aug 19;430(7002):873-7.
14. Dube DH, Prescher JA, Quang CN, Bertozzi CR. Probing mucin-type O-linked
glycosylation in living animals. Proc. Natl. Acad. Sci. U.S.A. 2006 Mar
28;103(13):4819-24.
15. Agard NJ, Baskin JM, Prescher JA, Lo A, Bertozzi CR. A Comparative Study of
Bioorthogonal Reactions with Azides. ACS Chem. Biol. 2006 Nov 1;1(10):644-8.
16. Baskin JM, Prescher JA, Laughlin ST, Agard NJ, Chang PV, Miller IA, et al. Copper-
free click chemistry for dynamic in vivo imaging. PNAS. 2007 Oct
23; 104(43):16793-7.
17. Agard NJ, Prescher JA, Bertozzi CR. A Strain-Promoted [3 + 2] Azide-Alkyne
Cycloaddition for Covalent Modification of Biomolecules in Living Systems. J. Am.
Chem. Soc. 2004 Nov 1;126(46):15046-7.
18. Chang PV, Prescher JA, Hangauer MJ, Bertozzi CR. Imaging cell surface glycans
with bioorthogonal chemical reporters. J. Am. Chem. Soc. 2007 Jul
11;129(27):8400-1.
19. Lin FL, Hoyt HM, Van Halbeek H, Bergman RG, Bertozzi CR. Mechanistic
investigation of the staudinger ligation. J. Am. Chem. Soc. 2005 Mar 2;127(8):2686-
95.
20. Beatty KE, Fisk JD, Smart BP, Lu YY, Szychowski J, Hangauer MJ, et al. Live-Cell
Imaging of Cellular Proteins by a Strain-Promoted Azide-Alkyne Cycloaddition.
ChemBioChem. 2010; 11(15):2092-5.
21. Chang PV, Prescher JA, Sletten EM, Baskin JM, Miller IA, Agard NJ, et al. Copper-
free click chemistry in living animals. PNAS. 2010 Feb 2;107(5):1821-6.
22. Hang HC, Yu C, Kato DL, Bertozzi CR. A metabolic labeling approach toward
proteomic analysis of mucin-type O-linked glycosylation. Proc. Natl. Acad. Sci.
U.S.A. 2003 Dec 9;100(25):14846-51.
23. Hang HC, Yu C, Pratt MR, Bertozzi CR. Probing glycosyltransferase activities with
the Staudinger ligation. J. Am. Chem. Soc. 2004 Jan 14;126(1):6-7.
24. Laughlin ST, Baskin JM, Amacher SL, Bertozzi CR. In vivo imaging of membrane-
associated glycans in developing zebrafish. Science. 2008 May 2;320(5876):664-7.
25. Freeze, H. H., Ernst, B., Hart, G. W. & Sinay, P. Carbohydrates in Chemistry and
Biology. Carbohydrates in Chemistry and Biology. Wiley, New York; p. 3-14.
68
26. Hart GW, Slawson C, Ramirez-Correa G, Lagerlof 0. Cross talk between 0-
GlcNAcylation and phosphorylation: roles in signaling, transcription, and chronic
disease. Annu. Rev. Biochem. 2011 Jun 7;80:825-58.
69
Chapter 3. Optimization and demonstration of diverse fluorophore targeting inside
live cells
The work discussed in this chapter has been published in: J. Yao, C. Uttamapinant, A.
Poloukhtine, J.M. Baskin, J.A. Codelli, E.M. Sletten, C.R. Bertozzi, V.V. Popik, A.Y.
Ting, "Fluorophore targeting to cellular proteins via enzyme-mediated azide ligation and
strain-promoted cycloaddition", JAm Chem Soc 2012 (134), 3720-3728.
70
Introduction
In the previous chapter, we describe the in vitro screening and characterization of
the LplA mutant enzyme and azide probe, as well as assays developed to assess the live-
cell azide ligation and strain-promoted cycloaddition reaction. We have established
working conditions providing quantitative azide ligation yield, with optimized azide
concentration and incubation time. Live-cell labeling of cyclooctyne-fluorescein
conjugate with metabolically incorporated azido-GalNAc gave nuclear specific labeling
signal after excess probe washout. In this chapter, we describe further screening,
quantitative comparisons, and optimization to the final two-step labeling protocol carried
out inside live cells.
First, we performed additional intracellular labeling assays to compare the
labeling efficiency of LplA(W371) and azide 9 with three other top LplA mutants and
azide pairs from the HPLC screen (Chapter 2). We ensured for equal expression of all
enzymes by immunostaining. Second, we give a brief overview of the scientific literature
regarding cyclooctynes and their rich evolution and engineering history over the years to
improve their reaction kinetics and hydrophilicity for applications in live cells and
animals. We chose five cyclooctynes and compared their reactivity, labeling efficiency,
and background using our two-step fluorophore targeting platform. Third, we discuss the
challenges in intracellular labeling with large, hydrophobic, and positively charged
fluorophores, such as ATTO 647N, and our attempts to overcome the challenges. Finally,
we demonstrate using the optimized site-specific two-step fluorophore targeting method
to label many intracellular proteins with various fluorophores across multiple cell lines.
In summary, this chapter encloses our intracellular efforts to optimize the two-
step fluorophore labeling technique. Several of the fluorophores targeted using LplA and
strain-promoted cycloaddition are attractive alternatives to fluorescent proteins. The
methodology will be most useful as a nontoxic labeling method for abundant proteins,
whose fusion to large tags perturbs function.
71
Comparison of LplA mutant enzymes and azide pairs using a live-cell labeling assay
Knowing that the azide ligation can achieve quantitative ligation and that the
strain-promoted azide-alkyne cycloaddition can give specific labeling inside live cells,
we were ready to combine the two steps. We have shown previously that LplA(W371)
and azide 9 exhibited better in vitro ligation kinetics with 2-fold improvement in kcat/Km
over wild-type LplA and azide 7 pair. Also, native gel-shift assay showed LplA(W371)
ligation of azide 9 achieved almost complete live cell azide ligation yield in just 30
minutes, while wild-type LplA ligation of azide 7 achieved only about 73% yield in 1
hour (Figure 2-8). We wanted to further differentiate among the LplA mutant/azide pairs
by comparing LplA(W371)/azide 9 with three other top enzyme/azide pairs (starred in
Figure 2-5) from the HPLC screen under the more stringent context of live cell labeling.
We first empirically optimized the azide washout time required to fully remove
excess unligated azide, using cyclooctyne-fluorescein retention as readout. We incubated
untransfected HEK 293T cells with azide 9 for 1 hour, followed by azide washout for
either 10 minutes or 1 hour. We then added the cyclooctyne-fluorophore conjugate,
MOFO-CFDA to the cells for 10 minutes, followed by 2 hours washout. Figure 3-1
shows that 1 hour of azide washout was enough to prevent MOFO-fluorescein retention
inside cells, as the background was comparable to cells not treated with azide 9.
Image
Time - +
Azide 9 Wash MOFO- Washfluorescein
1 hr 10 min 10 min 2 hr
or
1 hr
10 min. washout 1 hr washout untreated (no azide 9)
Fluorescein
+DIC
Figure 3-1. Determination of azide 9 washout time in live cells. Following the protocol
shown at the top, untransfected HEK 293T cells were incubated with 200 pM azide 9 for
72
1 hour, then washed for either 10 min or 1 hour, before adding 10 jM MOFO-fluorescein
diacetate for 10 min. Cells were imaged after a 2 hours wash. The images show that
fluorescein background is almost as low with 1 hour azide washout, as in cells not treated
with azide at all.
Since azide 9 is two methylene units longer than azide 7 and thus more
hydrophobic, we assumed that the 1 hour washout should be sufficient for azide 7 as well.
To compare the top four LplA mutant/azide pairs using the two-step labeling method
inside live cells, we transfected HEK 293T cells with the substrate protein LAP-tagged
blue fluorescent protein (BFP) and LplA plasmids corresponding to each of the top four
ligases. After transfection, following the labeling protocol shown in Figure 3-2, we
incubated the cells with azide 7 or azide 9 for 1 hour and washed out excess azide for 1
hour. MOFO-fluorescein diacetate was added to the cells for 10 minutes to derivatize the
azide-LAPs. After 2 hours of washing to remove excess fluorophore, cells were imaged
live. Figure 3-2 shows specific labeling of LAP-BFP for all four ligase and azide
combinations, but the highest signal-to-background ratio was obtained for the
LplA(W37I) and azide 9 pair. Compared to the wild-type LplA and azide 7 pair
previously used for cell-surface protein labeling, we can see that LplA(W371) and azide 9
labeling offered substantial improvement in signal intensity, about 4-fold greater on
average.
73
Time -|
1 hr
Image
I
MOFO-
Wash fluorescein Wash
1 hr 10 min 2 hr
Fluorescein
ein LAP-BFP + DIC
Figure 3-2. Identification of the best LplA mutant/azide substrate pair for intracellular
protein labeling. (top) Labeling protocol: HEK cells co-expressing indicated LplA mutant
enzyme and LAP-tagged BFP were labeled with azide 7 or azide 9 for 1 hour, washed for
1 hour, then labeled with monofluorinated cyclooctyne conjugated to fluorescein
diacetate (MOFO-CFDA, structure in Figure 3-5) for 10 minutes. Thereafter, cells were
washed again for 2 hours to remove excess unconjugated fluorophore. (bottom) Images
of HEK cells labeled as following to top protocol, with different LplA mutant/azide probe
pairs (starred combinations in Figure 2-5). Scale bars, 10 ptm.
74
wTLpIA
+ Azide 7
W3 vLpIA
+ Azide 7
W3vLpIA
+ Azide 9
W371LpIA
+ Azide 9
50000
45000
40000 AA
30A wTLpIA, + Azide 7
30000* warv~plA, + Azide 7
25000
- 20000 M W3rvLpIA, + Azide 915 AA aA AtwariLpA, + Azide 9
2 15000 ~AA W3 lILpIA, + Azide 9o A A A2 10000 1
5000 -AA= a A
0
0 20000 40000 60000
BFP intensity (a.u.)
Figure 3-3. Quantitation of data shown in Figure 3-2. For each condition, the mean
fluorescein intensity was plotted against the mean BFP intensity, for >100 single cells.
Fluorescein ligation yield is highest for the LplA(W371)/azide 9 combination.
We further quantified the labeling signal differences in Figure 3-3, in which
fluorescein intensity is plotted against LAP-BFP expression level for >100 single cells
for each condition. Two-step labeling with LplA(W37I) and azide 9 gave about 5-fold
higher fluorescein signal than the other three ligase and azide pairs. After live cell
labeling and imaging, we fixed and permeabilized the cells to perform anti-FLAG
immunofluorescence staining to detect the enzyme expression level of FLAG-tagged
LplA in cells (Figure 3-4). We needed to ensure that the perceived higher labeling signal
was not caused simply by higher enzyme expression level of LplA(W371). Anti-FLAG
immunostaining showed that ligase expression levels are all comparable under our
experimental conditions.
75
Alexa Fluor 568 Fuoresceln
(anti-FLAG) Fluoresmin LAP-BFP + DIC
wrLpA 16000
+ Azide 7
14000
12000
W3lvLpIA C 10000
+ Azide 7
e 8000 -
4000 - ---
W3vLpA __I
+ AzIde 9 u- 20D0
0-20 200 20 200. 20 200
W371i~plA"TLplA w37vLplA "37'LplAH 600 9Plasmid Amount (ng)
Figure 3-4. Immunofluorescence with anti-FLAG to compare labeling against expression
of all enzymes. (left) A repeat of labeling in Figure 3-2 following the same labeling
protocol, except that after live cell labeling, cells were fixed and stained with anti-FLAG
antibody to visualize FLAG-tagged LplA expression levels. To obtain comparable LplA
expression levels of mutants here and in Figure 3-2, ten-fold more LplA(W371) plasmid
was introduced compared to wild-type LplA and LplA(W37V) plasmids. All scale bars,
10 gm. (right) Graph showing quantitation of LplA mutant expression levels in cells.
Mean Alexa Fluor 568 intensities (anti-FLAG staining) for > 30 single cells from each
condition were averaged and graphed here.
76
Comparison of cyclooctyne structures
After optimizing the azide ligation of the two-step fluorophore labeling method,
we proceeded to the second step of strain-promoted azide-alkyne cycloaddition.
Numerous cyclooctynes have been developed by labs all around the world. The toolbox
of cyclooctynes has been expanding rapidly over the past decade owing to the wide
application of strain-promoted [3+2] azide-alkyne cycloaddition from in vitro small
molecules (1-3) and protein functionalization (4-8), proteome profiling (9), cell surface
glycan modification (10), tagging and imaging proteins inside cells (11), to in vivo
ligation in mice (12). With a rich panel of cyclooctynes available, we needed to choose a
cyclooctyne best suited for our goal of fluorophore targeting to proteins inside live cells.
Since the first report of using cyclooctynes to label cell surface glycoproteins (13),
many research groups have engineered numerous variations of subsequent generations of
cyclooctynes (4,14-19). These structures vary in terms of ring strain, electron deficiency
and hydrophilicity, which in turn influence reactivity toward azide and other intracellular
molecules, such as thiols. The design and synthesis of these cyclooctynes revolved
around two main goals: to increase cyclooctyne's reactivity and to improve its
hydrophilicity. A cyclooctyne with fast reaction kinetics with azide can largely improve
the sensitivity of labeling which is necessary especially within an environment as
complex as inside the cell or even in animals. Two routes have been explored to promote
reactivity: first, introducing electron-withdrawing groups, like fluorine, adjacent to the
alkyne to lower the LUMO of the cyclooctyne so it has better energy orbital overlap with
the HOMO of the azide (14,15); second, imposing ring strain around the alkyne by fusing
two benzene rings on each side of cyclooctyne (4,17,18). Using these two strategies, we
have seen a second order rate constant of nearly 800-fold increase from simple
cyclooctyne (1.2x10-3 M'S 1) (14) to biarylazacyclooctynone (BARAC, 0.96 M-'S 4 ) (17).
However, increased reactivity also comes at the cost of increased hydrophobicity
especially when dibenzo groups are attached to raise ring strain (4,17,18). Although the
dibenzo rings impart a large enhancement in reactivity, the hydrophobic character of
these cyclooctynes leads to high chance of nonspecific binding to cellular membranes and
sequestration by hydrophobic pockets. Placement of di-methoxy groups and installing a
nitrogen atom within the cyclooctyne ring to break its hydrophobic surface area and to
77
reduce the extent of nonspecific hydrophobic binding to cellular hydrophobic areas have
been tried to improve the hydrophilicity of these probes (4,16). With respect to our LplA-
mediated technique of intracellular fluorophore targeting, in addition to having efficient
strain-promoted azide-alkyne cycloaddition and stability inside live cells, the ideal
cyclooctyne would strike a balance between high reactivity and low hydrophobicity.
We selected a panel of five cyclooctyne structures to test inside live cells because
it was not clear which cyclooctyne structure(s) would be the best for our purpose. In
order to compare the reactivity of all five cyclooctynes in parallel without the
compounding effect caused by the cell entry and washout properties of the fluorophore,
we derivatized each cyclooctyne with fluorescein diacetate via a stable amide bond
(Figure 3-5). We selected fluorescein for this purpose because it has excitation (?x 490
nm) and emission wavelengths (kem 514 nm) suitable for live-cell imaging and high
quantum yield (0.95). Most importantly, it is the least hydrophobic among the
fluorophores we have tested and can be washed out of cells during a relatively short
period of time (less than 2 hours) which we had seen during previous experiments using
MOFO-CFDA to label metabolically incorporated azido-GalNAc.
With the optimized protocol of LplA(W371) mediated azide 9 ligation and a panel
of cyclooctyne-fluorescein diacetate conjugates, we performed two-step intracellular
labeling on nuclear localized LAP-BFP in live HEK 293T cells following the scheme in
Figure 3-5 A (Top). After azide 9 ligation and washout, cells were treated with one of the
five cyclooctyne-fluorescein diacetate conjugates. Confocal images were taken at
multiple time points after probe incubation. It is immediately clear that the probes
ADIBO, DIBO, and MOFO showed very specific nuclear labeling with low background
(Figure 3-5 A). MOFO-fluorescein washed out the fastest among the three and specific
labeling was observed at only 1.5 hours after probe addition. However, after
normalization, it is clear that the nuclear signal of ADIBO and DIBO appeared to be
much higher than signal from MOFO. This is most likely due to the superior second order
rate constants of ADIBO (0.31 M1 s4 in methanol) (19) and DIBO (5.92x10 2 M-s~1 in
methanol) (18) compared to MOFO (4.3x10 3 M's') (14). We quantitatively analyzed
the signal to noise ratio of ADIBO-, DIBO-, and MOFO-fluorescein by plotting the
signal from the nucleus versus background in the cytosol of > 50 single cells for each
78
cyclooctyne-fluorescein (Figure 3-5 B). ADIBO- and DIBO-fluorescein both gave about
4-5 times higher nuclear signal than MOFO-fluorescein, however, their cytosolic
backgrounds are also higher. The overall signal to noise ratios (defined as nuclear signal
over cytosolic background) of all three cyclooctynes are comparable. The higher absolute
signal of ADIBO and DIBO makes them superior cyclooctynes than MOFO. However,
we note that the faster washout time of MOFO-fluorescein can be advantageous in certain
applications.
Image
Time -- +
AZxW Wash O Wash
1 hr 1 hr 10 min 1.5 hr
or
2.5 hr
Fluonscein Fluoresein LAP-BFP + DIC
H
6
0
MeO
MeO' 
-N Me Me
0- ,N 6NsMeOO 16
0
FF
1 Me Me
60
Lower contrast Higher contrast
79
0
5(6)-carboxyfluorescein diacetate
A
ADIBO
0.31 M's'
DIBO
5.9x1 02 M-sI-1
MOFO
4.3x1 0-3 -I-
DIMAC
3.0x10 -3 M'sI-
DIFO
4.2x102 M's 1
B
35000
30000 -
25000 -
20000 -
40000 15000 -
35000 10000
30000 5000
25000 0
20000 0
1K"
1000 2000 3000
MADIBO-fluorescein
*DIBO-fluorescein
AMOFO-fluorescein
0 10000 20000 30000 40000
Cytosolic fluorescein intensity (a.u.)
Figure 3-5. Evaluation of various cyclooctyne structures for site-specific intracellular
protein labeling. (A) Top: labeling protocol for HEK 293T cells co-expressing
LplA(W37I) and nuclear-localized LAP-BFP (LAP-BFP-NLS). After labeling with
azide 9 for 1 hour and washing for 1 hour, cells were treated with the indicated
cyclooctyne-fluorescein conjugate for 10 minutes. Cells were washed again for 2.5 hours
to remove excess unconjugated fluorophore, except for MOFO, which was washed for
only 1.5 hours. Bottom: images of labeled HEK 293T cells. The LAP-BFP-NLS image
is overlaid on the DIC image. Fluorescein signal intensities can be compared in the first
two columns, which show the fluorescein images at lower contrast (left) and higher
contrast (middle). Cyclooctyne structures are shown at right, and second-order rate
constants are shown at left. ADIBO, aza-dibenzocyclooctyne; DIBO, 4-
dibenzocyclooctynol; MOFO, monofluorinated cyclooctyne; DIMAC, 6,7-
dimethoxyazacyclooct-4-yne; DIFO, difluorinated cyclooctyne. All scale bars, 10 pm.
(B) Quantitation of data in (A). For the three cyclooctynes (ADIBO, DIBO, and MOFO),
the mean nuclear fluorescein intensity (representing specific labeling) was plotted against
the mean cytosolic fluorescein intensity (representing nonspecific labeling), for the same
cell. More than 50 single cells were analyzed for each cyclooctyne.
80
0 15000
:3
10000
5000
n
Nuclear labeling using DIMAC-fluorescein showed very weak signal, perhaps
due to the slow second order rate constant of DIMAC (3.0x10~3 M~'s~1 in CD 3CN) (16).
We tried to increase signal by incubating cells with up to 30 pM of DIMAC-fluorescein
without significant improvement (data not shown). Aside from low signal, we also
observed retention of the probe in certain non-transfected cells, which was unexpected
because DIMAC was noted for the hydrophilicity conferred by the di-methoxy groups.
The background fluorescence was much reduced after cell fixation and further washout
(Figure 3-6), indicating it was perhaps not caused by a covalent reaction.
For DIFO-fluorescein, we observed nuclear signal specific to transfected cells as
soon as 10 min after probe addition, however, this was accompanied by the high cytosolic
background that persisted even after cells were fixed and permeabilized (Figure 3-6), at
which point the small molecules inside cells should be thoroughly washed out. We also
tried to reduce or prevent this seemingly permanent cytosolic background by lowering the
concentration of DIFO-FDA to 1 [tM and incubation time as short as 40 seconds, but the
cytosolic background still persisted even though at lower intensity (data not shown).
Despite the fast reactivity of DIFO (7.6x10-2 in CD 3CN) (15), our observation indicates
that DIFO was retained covalently inside cells perhaps through reaction with nonspecific
nucleophiles. This hypothesis is strengthened by the result from Beatty and coworkers
that showed DIFO reacting with cysteine and glutathione at the strained-alkyne moiety in
vitro (11). With cellular glutathione concentration in the millimolar range (20), it is very
likely that the irreversible high background in intracellular DIFO labeling could be
caused by reaction with glutathione.
81
Image live Image fixed
z We 9 Washf I Wash Fix Wash
1 hr 1 hr 10 min 2.5 hr 10 min 10 min
Fluorescein Fluorescein
- 1 '
4
DIMAC-fluorescein
LAP-BFP + DIC
Live
Fixed
Lower contrast Higher contrast
Fluorescein LAP-BFP + DIC
Live
Fixed
Figure 3-6. Analysis of background labeling by cyclooctyne-fluorescein conjugates.
Samples from Figure 3-5 A labeled with DIMAC-fluorescein and DIFO-fluorescein were
re-imaged after cell fixation and permeabilization. Labeling and imaging protocol is
given at the top. Non-specific fluorescence is much reduced after fixation for DIMAC-
fluorescein but not as much for DIFO-fluorescein.
Furthermore, work from Chang et al using DIMAC and DIFO, each conjugated to
a FLAG epitope tag to detect cell surface sialic acids labeled with azides in live mouse
82
Time -+
DIFO-Ruoescain
also showed background non-specific labeling (12). In the negative control experiment,
DIMAC and DIFO seemed to bind to mouse serum albumin (MSA) in heart lysates in an
azide-independent manner. Further investigation with MSA immunoprecipitation and
probing with anti-MSA and anti-FLAG antibodies showed that the nonspecific signals
still persisted even after the samples were treated with 1 M urea, p-mercaptoethanol, and
boiled, suggesting that it might be caused by a combination of hydrophobic and covalent
interactions, perhaps with the two free cysteine residues in MSA. Even though the high
reactivity of DIFO offers the potential of high sensitivity and fast labeling, we need to be
concerned about nonspecific reactions when using this cyclooctyne.
On the basis of these results, we selected ADIBO for most of our intracellular
protein labeling experiments. However, as shown later, due to ADIBO's hydrophobicity,
MOFO is a better option when working with very hydrophobic fluorophores such as
ATTO 647N.
83
Demonstration of intracellular protein labeling with ADIBO-fluorescein diacetate
Having optimized both the azide ligase and the cyclooctyne, we proceeded to
characterize two-step fluorophore labeling inside cells, and explore its generality. We
first empirically optimized the ADIBO-fluorophore loading concentration and washout
time by testing five concentrations of the probe and imaging cells at various times post
probe incubation. As shown in Figure 3-7, washout time increased to more than 2.5 hours
at 40 1 M of ADIBO-fluorescein diacetate. We determined that 10 pLM probe, followed by
2.5 hours of washout was effective.
0 min
cc
0(U
C
Q.
0
4Z
0
0
40 min
Washout time
1.5 hr 2.5 hr 5 hr
40pM
20 pM
10 pM
5pM
2.5pM
Figure 3-7. Two-dimensional optimization of ADIBO-fluorescein diacetate loading
concentration and washout time. Various amounts of ADIBO-fluorescein were loaded
into untransfected COS-7 cells for 10 minutes at 37 'C. Various washout times were
tested, ranging from 0 to 5 hours. Fluorescein images are shown with DIC overlay. All
scale bars, 10 tm. Based on this data, we selected 10 gM ADIBO-fluorescein and 2 - 2.5
hour washout time for subsequent experiments.
84
Following the scheme shown in Figure 3-8, we labeled LAP-tagged BFP protein
localized to various compartments inside the cell. Figure 3-8 shows that only cells
expressing LAP-BFP retained fluorescein, while neighboring untransfected cells did not.
Negative controls performed in parallel showed labeling is completely abolished when
the catalytic lysine residue in LAP is mutated to alanine, when azide 9 is omitted, and
when using a catalytically inactive LplA enzyme which has residue K133 that is involved
in both lipoic acid adenylation and lipoate transfer mutated to R.
Image
Time -- +
azide 9 Wash ADIBO- Washfluorescein
1 hr 1 hr 10 min 2 hr
LAP (K->A) - azide 9 W371/K133RLpIA
Fluorescein
LAP-BFP
+ DIC --
Figure 3-8. Two-step fluorophore targeting via strain-promoted azide-alkyne
cycloaddition. Following the labeling scheme shown on top, HEK 293T cells expressing
LplA(W37I) and LAP-BFP were labeled with azide 9 and ADIBO-fluorescein diacetate.
Only transfected cells retained fluorescein signal. Negative controls with non-catalytic
LAP(K->A), omitting azide 9, and dead mutant LplA(W371, K133R) did not show
labeling.
We also demonstrated labeling of different LAP fusion proteins (Figure 3-9) and
in different cell lines. Using the two-step protocol shown in the scheme in Figure 3-8, we
successfully labeled LAP in the nucleus, cytosol, and plasma membrane, as well as LAP
85
fusions to actin and MAP2 (microtubule associated protein 2) with robust signal. These
experiments were performed in multiple mammalian cell lines, including HEK 293T,
HeLa, and COS-7, demonstrating the versatility of the method.
Eu
HEK 293T HEK 293T HeLa HeLa
Figure 3-9. ADIBO-fluorescein labeling of three localized LAP-BFP fusions, LAP-p-
actin, and LAP-MAP2 (microtubule-associated protein 2). Labeling in the cell type
indicated beneath each image was performed as in Figure 3-8, except that for LAP-p-
actin and LAP-MAP2, azide 9 was incubated for 2 hours, and washed for 1.5 hours
before fluorophore addition. NES = nuclear export sequence; CAAX = prenylation tag;
NLS = nuclear localization sequence. All scale bars, 10 pim.
86
LAP-MAP2
COS-7
Challenges and optimization for targeting large fluorophores inside live cells
Following the success of ADIBO-fluorescein diacetate labeling on many LAP-
fused proteins inside live cells, we wished to extend the method to other large
fluorophores such as rhodamine and ATTO dyes. Initial labeling experiments using
cyclooctyne conjugated to these fluorophores were challenging on two major fronts. First,
the hydrophobic nature of these fluorophores made the probes stick to hydrophobic
surfaces of proteins inside cells, causing very high background and prolonged washout
time. Second, in addition to being hydrophobic, these fluorophores are often positively
charged. The positive charge caused them to be attracted and eventually sequestered
inside the mitochondrial matrix, which has a more negative electrical potential (21). Even
by itself, without any additional conjugation, we found that ATTO 647N gave a high
level of nonspecific cell staining, primarily in the mitochondria (data not shown). In each
case, the time required to remove excess probe depended on the different physical
properties of the fluorophore. Comparing to labeling with the relatively hydrophilic probe
ADIBO-fluorescein diacetate, where excess probe in the background can be washed out
quickly to see specific labeling, we observed that labeling performed with ADIBO
conjugated to the red or far-red emitting fluorophores generally took from 3-9 hours to
remove background in order to see specific labeling signal.
To alleviate the nonspecific hydrophobic sticking background, aside from
frequent media change over a prolonged period of time, we found that using a less
hydrophobic cyclooctyne such as MOFO rather than ADIBO, especially in conjugation
with a very hydrophobic fluorophore such as ATTO 647N helped in lowering the
background. Figure 3-10 shows a comparison of LAP-BFP-NLS labeling with ADIBO
and MOFO, each conjugated to ATTO 647N. The graph on the bottom plots the specific
labeling (nuclear ATTO signal) to the nonspecific labeling (cytosolic ATTO signal)
for >50 cells for each probe. It can be seen that MOFO-ATTO 647N gives much more
specific labeling than ADIBO-ATTO 647N, likely because the total hydrophobicity of
the conjugate is reduced. This ultimately permitted us to perform ATTO 647N labeling of
LAP-p-actin in live COS-7 cells as shown in a later section.
87
MOFO-ATTO 647N
ADIBO-ATTO 647N
25000
C(1) 20000
z
*--
't- 15000
- 10000
(1) 5000
z
0
* MOFO-ATTO 647N
* ADIBO-ATTO 647N
0 2000 4000 6000
Cytosolic ATTO 647N intensity (a.u.)
Figure 3-10. Comparison of MOFO-ATTO 647N and ADIBO-ATTO 647N nuclear
labeling specificities. (Top) Nuclear labeling of LAP-BFP-NLS with MOFO- and
ADIBO-ATTO 647N in COS-7 cells. Images are shown after 8 hours of ATTO 647N
conjugates washout. (Bottom) The mean nuclear ATTO 647N intensity (representing
specific signal) was plotted against the mean cytosolic ATTO 647N intensity
(representing nonspecific signal), for the same cell, for >50 single cells for each
condition.
88
I
Aside from pairing especially hydrophobic and positively charged fluorophores
with less hydrophobic cyclooctynes, we also tested the effect of varying linker structures
between MOFO and ATTO 647N in an attempt to further reduce the nonspecific labeling
background. The conventional N, N'-dimethyl-1,6-hexanediamine (HDDA) linker that
we used for most fluorophore conjugates was replaced by the more hydrophilic PEG
linker (Figure 3-11). However, we did not observe any significant improvement in
reducing cytosolic background or shortening washout time while using MOFO-PEG-
ATTO 647N (Figure 3-11), suggesting that the degree of hydrophobicity of the
cyclooctyne and fluorophore moieties dominated probe distribution inside cells.
R I
L N+
LAP-BFP +I
R 
)O
TT AA7N I AP-Rn:P + r
2 hr wash
4 hr wash
Figure 3-11. Comparison of two different linker structures for MOFO-ATTO 647N.
Conjugates were synthesized with either an N,N'-dimethyl-1,6-hexanediamine linker
(left) or a polyethylene glycol PEG linker (right). Labeling was performed on HEK 293T
cells expressing LplA(W371) and nuclear-localized LAP-BFP as described previously.
Images are shown after the indicated fluorophore washout times. Specific labeling is
89
r-
N-. . N
+ N
0U
N
0i
ATTO 647N
2 hr wash
4 hr wash
detectable, particularly in the nucleoli of transfected cells, but the images are dominated
by the non-specific signal in these experiments.
The second major source of background fluorescence was from dyes trapped
inside the mitochondrial matrix. Since the organelle has a negative inner-electrical
potential established by the electron transport chain (21,22), positively charged
fluorophores such as tetramethylrhodamine and ATTO 647N are especially attracted to
the mitochondria. We incubated HEK 293T cells with ADIBO-ATTO 647N together
with Mitotracker Green that confirmed the mitochondrial localization of the probe (data
not shown). Compared to the hazy nonspecific background in the cytosol, the
mitochondrial background appeared as very bright dots or tubular structures depending
on the cell type. It was even more difficult to wash out because the dyes are sequestered
in the organelle as long as the inner mitochondrial potential is maintained. Low levels of
mitochondrial background fluorescence still remained after cells were fixed and
permeabilized (data not shown). Moreover, mitochondrial trapping of the probe can
reduce the effective concentration of the probe that is available elsewhere to react with
azide-ligated LAP. To improve the signal to noise ratio of two-step fluorophore targeting
using these hydrophobic and positively charged dyes, we explored two strategies to
reduce mitochondrial sequestering of the dye. The first strategy is to apply a
mitochondrial uncoupling agent that can temporarily disrupt the electrical potential across
the mitochondrial inner membrane, so trapped dye can escape and be washed out. One
routinely used uncoupling agent, p-trifluoromethoxy carbonyl cyanide phenyl hydrazone
(FCCP), is a lipid-soluble weak acid that freely diffuses in and out of the mitochondrial
matrix. It acts like a proton shuttle to disrupt the chemiosmotic gradient across the inner
membrane without interrupting the electron transport chain (23).
To test whether using FCCP to dissipate the proton gradient could help in
reducing mitochondrial trapping of positively charged fluorophore, we performed
labeling on nuclear-localized LAP-BFP with either ADIBO-X-rhodamine or ADIBO-
ATTO 647N as described in Figure 3-12. The ionophore FCCP was added 6 hours after
probe incubation. The mitochondrial background fluorescence, which initially appeared
as dots in HEK 293T cells, gradually lost its brightness and started to appear as blurred
90
nonspecific background fluorescence in the cytosol. Cells were maintained in growth
media containing FCCP for 2 hours to allow the fluorophore to escape from the
mitochondrial matrix. During this time, we relied on the nonspecific ion transporters of
the cell to pump out excess probe. Even though disrupting the mitochondrial potential
aided the removal of some of the trapped probe, further dependence on the action of
nonspecific transporters and prolonged washout time were still not sufficient to obtain a
clean background for such hydrophobic and positively charged dyes (Figure 3-12).
Image
wash wash
1 hr ihr 10 min 6 hr 2 hr
ADIBO-X-rhodamine LAP-BFP+DIC ADIBO-ATTO 647N LAP-BFP+DIC
No mitochonrial
suppression
FCCP
Figure 3-12. Suppression of mitochondrial background labeling with FCCP. Following
labeling scheme shown on top, after azide 9 incubation and washout, HEK 293T cells
expressing LplA(W371) and LAP-BFP-NLS were labeled with either ADIBO-X-
rhodamine (left) or ADIBO-ATTO 647N (right) for 10 minutes. Thereafter, cells were
washed and incubated in fresh media. After 6 hours of probe washout, media was
replaced with media containing FCCP (p-trifluoromethoxy carbonyl cyanide phenyl
hydrazone, a mitochondrial uncoupler). Live cells were imaged at 2 hours after FCCP
addition. Labeled cells with no FCCP addition were imaged at 8 hours after cyclooctyne-
dye addition for comparison.
91
The second strategy to reduce mitochondrial sequestering of the probe is to apply
small molecule quenchers. The two major factors that lead to fluorophores entering the
mitochondria are hydrophobicity and positive charge, which are also properties of some
small molecule fluorescence quencher dyes. We imagined that if the same mechanism
applies, these quenchers would also be attracted to the mitochondria, co-localizing with
the fluorophores trapped there, and efficiently quenching their fluorescence. We chose
the quencher QSY21 because its absorption wavelength overlaps well with the emission
wavelength of ATTO 647N at 660 nm. Labeling of nuclear-localized LAP-BFP in HEK
293T cells was carried out as described in Figure 3-13. At 6 hours after ADIBO-ATTO
647N washout, cell growth media was replaced with media containing QSY21. Live cell
images taken at 10 min after QSY21 addition showed much reduced mitochondrial
background fluorescence in cells incubated with QSY21.
Image
9 wash wash QSY21
1 hr ihr 10 min 6 hr 10 min
ADIBO-ATTO 647N LAP-BFP+DIC
No mitochonrial
suppression
QSY21
92
Figure 3-13. Suppression of mitochondrial background labeling with QSY2 1. Following
labeling scheme shown on top, after azide 9 incubation and washout, HEK 293T cells
expressing LplA(W371) and LAP-BFP-NLS were labeled with ADIBO-ATTO 647N for
10 minutes. Thereafter, ADIBO-ATTO 647N was removed from the cells, which were
washed and incubated in fresh media. After 6 hours of probe washout, media was
replaced with media containing 10 pM QSY21 for 10 minutes. Live cells were imaged
after QSY21 incubation. Labeled cells with no QSY21 addition were imaged at 6 hours
after ADIBO-ATTO 647N addition for comparison.
We also tested the combined effect of mitochondrial uncoupler FCCP and QSY21
quencher by first adding FCCP after ADIBO-ATTO 647N incubation to free some of the
fluorophore trapped there and allow it to wash out over 2 hours in the presence of FCCP
(Figure 3-14, top scheme). Afterwards, cell growth media containing FCCP was replaced
with fresh growth media without FCCP to allow the electrical potential to recover over 1
hour. After the mitochondrial electrical potential has recovered, QSY21 was added to
further quench the remaining fluorophore in the mitochondrial matrix. Using both FCCP
and QSY21, we were able to increase the nuclear signal to cytosolic background ratio
about two-fold (Figure 3-14 bottom).
Image
ide 9 wassh*wash FOCP wash QSY21
1 hr 1hr 10 min 6 hr 2 hr 1 hr 10 min
ATTO 647N LAP-BFP+DIC
No mitochondrial
suppression
FCCP + QSY21
93
U. .
* ADIBO-ATTO 647N
+ FCCP + QSY21
+ ADIBO-ATTO 647N
0 10000 20000 30000 40000
BFP intensity (a.u.)
Figure 3-14. Suppression of mitochondrial background labeling with FCCP and QSY21.
(Top) Following labeling scheme shown to test the combined effect of FCCP and
QSY21, after azide 9 incubation and washout, HEK 293T cells expressing LplA(W371)
and LAP-BFP-NLS were labeled with ADIBO-ATTO 647N for 10 minutes. Thereafter,
ADIBO-ATTO 647N was removed from the cells and replaced with complete media.
After 6 hours of probe washout, cells were treated with mitochondrial uncoupler FCCP
and quencher QSY21 sequentially as shown in the labeling scheme. Live cells were
imaged after QSY21 incubation. Labeled cells with no QSY21 addition were imaged at 9
hours after ADIBO-ATTO 647N addition for comparison. (Bottom) Quantitation of data
on top. For the cells labeled with ADIBO-ATTO 647N, with or without FCCP and
QSY21 application to reduce background, the ratio of the mean nuclear fluorescein
intensity (representing specific labeling) and the mean cytosolic fluorescein intensity
(representing nonspecific labeling) was plotted against the mean nuclear BFP intensity
(representing LAP-BFP-NLS expression level), for the same cell. More than 50 single
cells were analyzed for each condition.
We applied the QSY21 quenching strategy in ADIBO-ATTO 647N labeling of
LAP-tagged p-actin in live cells with the hope that we would acquire images with more
clear and distinctive actin filaments without the background fluorescence from dyes in
the cytosol. COS-7 cells transfected with LAP-p-actin and LplA(W371) were labeled with
94
5
z
4.5
3.5
3
- b 2.5
2
0 1.5
cc 1
0.5
Z 0
azide 9 and ADIBO-ATTO 647N as described previously. Following 6 hours of ADIBO-
ATTO 647N washout, cells were incubated with media containing 10 pIM of QSY21 for
10 minutes and imaged live. However, we found that even though QSY21 was supposed
to be trapped inside the mitochondria and only quench the background fluorescence
there, some amount of QSY21 in the cytosol also reduced cytosolic LAP-p-actin specific
labeling signal by about 2-fold comparing to conventional labeling with no quencher
addition (Figure 3-15). If all the images were normalized with respect to the labeling
performed without QSY21, ADIBO-ATTO 647N labeled LAP-p-actin pattern was not
visible when QSY21 was added (Figure 3-15, first column). Even though QSY21 helped
to reduce the mitochondrial background in ADIBO-ATTO 647N labeling of nuclear
LAP-BFP without decreasing nuclear signal, in the more challenging case of LAP-3-actin
labeling, the signal was very low and can be reduced by QSY21 quencher remaining in
the cytosol. Since QSY21, like other similarly large, hydrophobic, and positively charged
fluorophores, is not completely trapped in the mitochondria, trace amounts of it may still
nonspecifically stick to the cytosol to quench cytosolic ADIBO-ATTO 647N signal. We
decided this could not become a widely applicable strategy in reducing mitochondrial
background especially when it also reduced LAP-specific signal. However, if the protein
of interest were localized to the nucleus or if the LAP-specific signal were very robust,
the strategy of quenching mitochondrial background using a hydrophobic and positively
charged quencher whose absorption spectrum overlaps well with the emission spectrum
of the fluorophore in use could be applied.
95
No mitochondrial
suppression
QSY21
field of view 1
QSY21
field of view 2
ADIBO-ATTO 647N DIC ADIBO-ATTO 647N
Normalized Not normalized
Figure 3-15. Comparison of ADIBO-ATTO 647N labeling of LAP-p-actin with and
without suppression of mitochondrial background using QSY21 quencher. COS-7 cells
expressing LplA(W371) and LAP-P-actin were labeled with azide 9 and ADIBO-ATTO
647N. After 6 hours of ADIBO-ATTO 647N washout, cells were incubated with QSY21
quencher for 10 minutes. Live cells were imaged 10 min after QSY21 incubation.
Labeled cells with no QSY21 addition were imaged at 6 hours after ADIBO-ATTO 647N
incubation for comparison.
96
Extension of site-specific two-step labeling method to diverse fluorophore structures
with labeling inside cells and on the cell surface
One of the advantages of the intracellular two-step fluorophore targeting
technique is its potential to be generalizable towards any type of fluorophore as long as
the fluorophore can be conjugated to a cyclooctyne and the conjugate is cell permeable.
As a way to showcase the versatility of the method, in addition to fluorescein, we
prepared ADIBO conjugates to tetramethylrhodamine (TMR), ATTO 647N, and ATTO
655. ADIBO-TMR and ADIBO-ATTO 655 both gave specific labeling (Figure 3-16), but
ADIBO-ATTO 647N produced a high level of nonspecific binding. As shown in the
previous section, we demonstrated that conjugating the more hydrophobic fluorophore
ATTO 647N with the less hydrophobic cyclooctyne MOFO instead of ADIBO was able
to provide more specific labeling. Therefore, we conjugated the hydrophobic
fluorophores like X-rhodamine and ATTO 647N to MOFO.
Figure 3-16 shows live cell labeling of multiple LAP fusion proteins with a
diverse palette of fluorophores spanning from fluorescein to ATTO 647N. ADIBO is
used for the more hydrophilic dyes such as fluorescein, TMR and ATTO 655. DIBO,
which is structurally similar to ADIBO, is used for Oregon Green. MOFO is used for the
more hydrophobic dyes, X-rhodamine and ATTO 647N.
0-fluorescein DIBO-O Green ADIBO-TMR MOFO-X-rhodamine MOFO-ATTO 647N ADIBO-ATI
LAP-MAP2 LAP-p-actin LAP-P-actin LAP-MAP2 LAP-p-actin LAP-BF
Figure 3-16. Intracellular protein labeling with diverse fluorophore structures. COS-7
cells coexpressing LplA(W371) and the indicated LAP fusion protein (labeled across
bottom) were labeled with azide 9, followed by the indicated cyclooctyne-fluorophore
97
P-NLS
conjugate (labeled across top). MOFO was used for the more hydrophobic fluorophores
(X-rhodamine and ATTO 647N); ADIBO and DIBO were used for the others. Chemical
structures are shown in Experimental Methods. TMR = tetramethylrhodamine.
After demonstrating two-step fluorophore ligation on various intracellular
proteins, we also wished to perform labeling on cell surface proteins. We chose two cell
surface proteins, LAP-tagged LDL receptor and neuronal adhesion protein neurexin -1P.
After transfecting HEK 293T cells with each plasmid, we added purified LplA(W371)
protein, azide 9, and ATP to the cell medium for 20 minutes. Following azide ligation,
azide ligated LAP-tagged proteins were derivatized with DIBO-Alexa Fluor 647, which
is membrane-impermeant, or DIBO-biotin. Cells labeled with DIBO-biotin were
visualized by staining with streptavidin-Alexa Fluor conjugates. Figure 3-17 (top and
middle) shows both DIBO-Alexa Fluor 568 and DIBO-biotin label cell surface proteins
with specific and high signal to background ratio.
LAP-LDL receptor LAP-neurexin-1I
DIBO-Alexa Fluor 647
+ DIC
Streptavidin-Alexa Fluor
+ DIC
II
LAP-LDL receptor
I.
LAP-neurexin-1 0
98
LAP-BFP-NLS
Streptavidin-Alexa Fluor 568
LAP-BFP
+ DIC
Figure 3-17. Cell surface and intracellular protein labeling with DIBO-Alexa Fluor 647
and DIBO-biotin (both commercially available from Life Technologies). (Top and
middle) HEK 293T cells expressing LAP-LDL receptor (low density lipoprotein
receptor) or LAP-neurexin- 1p were labeled by first adding purified LplA(W371) enzyme
and azide 9 to the cell medium for 20 min. After 5 min of washing, DIBO-Alexa Fluor
647 or DIBO-biotin was added for 10 min. DIBO-biotin was detected by staining live
cells with streptavidin-Alexa Fluor conjugate. Negative controls are shown with azide 9
omitted. (Bottom) DIBO-biotin is cell permeable and its labeling could also be detected
on LAP-fused intracellular proteins. HEK 293T cells expressing LplA(W371) and
nuclear-targeted LAP-BFP-NLS were labeled live with azide 9 and then DIBO-biotin.
Thereafter, cells were fixed and permeabilized and labeled proteins were detected by
staining with streptavidin-Alexa Fluor 568 conjugate. A negative control is shown with
an alanine mutation in LAP. All scale bars, 10 pm.
Since DIBO-biotin is membrane-permeant, it can also be applied in performing
this labeling inside cells. However, unlike cell surface proteins that can be directly
visualized using streptavidin-fluorophore conjugate, biotinylated intracellular LAP-
tagged proteins can only be detected after membrane permeabilization. Figure 3-17
(bottom) shows intracellular labeling of azide-ligated LAP-BFP-NLS in HEK 293T cells
using DIBO-biotin. After DIBO-biotin addition for 10 minutes, the cells were washed,
fixed and permeabilized, then biotinylated LAP was detected with streptavidin-Alexa
Fluor 568.
99
Measurement of overall two-step ligation yield in cells
We have demonstrated previously that azide 9 ligation by the LplA(W371)
enzyme inside live cells is close to quantitative. We were curious what percentage of
azide 9-ligated LAP protein reacts with cyclooctyne inside cells. We used DIBO-biotin
labeling to measure the overall yield of the two-step labeling method in live cells. After
performing azide 9 and DIBO-biotin labeling in HEK 293T cells expressing LplA(W371)
and LAP-mCherry fusion proteins, cells were lysed, incubated with excess streptavidin
protein to bind biotinylated LAP-mCherry. The lysate was analyzed by SDS-PAGE. In-
gel mCherry fluorescence imaging in Figure 3-18 shows that LAP-mCherry runs at the
expected molecular weight of 27 kDa in negative control samples in which azide 9 or
streptavidin is omitted. However, lane 1 shows about 21% of LAP-mCherry shifted up to
approximately 80 kDa, corresponding the combined molecular weight of LAP-mCherry
(27 kDa) and streptavidin protein (56 kDa). This higher molecular weight band represents
the population of LAP-mCherry that was ligated with azide 9, which subsequently
reacted with DIBO-biotin. From the streptavidin gel-shift assay using DIBO-biotin to
label intracellular LAP-tagged protein, we conclude that under our current labeling
protocol, with azide 9 incubation for 1 hour, excess azide 9 removal for 1 hour, and
cyclooctyne-fluorophore conjugate incubation for 10 minutes, the two-step labeling yield
is approximately 20%.
100
1. W37'LplA + Azide 9 + DIBO-biotin, then streptavidin
2. Omit streptavidin control
3. Omit Azide 9 control
1 2 3 1 2 3
150 -WO
100 -
80 -- LAP-mcherry+streptavidin
60 -
50-
40-
30 -
25 - - LAP-mCherry
20 -
15 - 21%
kDa
Coomassie mCherry fluorescence
Figure 3-18. Streptavidin gel-shift analysis of two-step labeling yield in live cells. HEK
293T cells co-expressing LplA(W37I) and LAP-mCherry were labeled with azide 9 for 1
hour. Excess azide was washed out for 1 hour. Cell were then incubated with DIBO-
biotin for 10 minutes. Excess DIBO-biotin was washed out for 2 hours. After cell lysis,
excess streptavidin protein (56 kDa) was added to shift the molecular weight of
biotinylated LAP-mCherry (27 kDa). Coomassie and mCherry fluorescence visualization
are shown for lysates run on a 12% SDS-polyacrylamide gel. Lane 2 shows a negative
control with streptavidin omitted. Lane 3 shows omission of azide 9. Estimated percent
conversion to biotin-alkyne ligated LAP-mCherry product from LAP-mCherry was
approximated using densitometry and given at the bottom of the mCherry fluorescence
gel image.
101
......... . ......... ........ -- - - 1111 1- 1111, . ..... ............
Conclusion
We have developed an enzyme based labeling method that can target a diverse
panel of fluorophores to a small 13-amino acid peptide tag LAP fused to proteins of
interest inside living cells. The targeting is accomplished first by enzyme-mediated alkyl
azide ligation and then by strain-promoted cycloaddition with a fluorophore-conjugated
cyclooctyne. To develop the method, we systematically optimized the azide ligation
reaction through screening of lipoic acid ligase mutants and alkyl azide variants. We then
evaluated five different cyclooctynes structures differing in their reactivity, selectivity,
hydrophobicity, and extent of nonspecific binding to cells, using a live-cell fluorescein
targeting assay. Our final, optimized two-step labeling scheme was used to target a
diverse panel of fluorophores with emission wavelengths spanning the visible spectrum,
to a variety of LAP fusion proteins in multiple mammalian cell lines.
A few aspects that govern the choice of cyclooctyne are hydrophobicity, reactivity,
and stability. Our comparison of cyclooctynes in live cells yielded observations that
should prove useful even beyond the context of enzyme-mediated fluorophore targeting,
due to the numerous and diverse applications to which cyclooctynes are being applied. In
choosing a cyclooctyne for intracellular ligation, it is important that the probe has high
reactivity, yet remains hydrophilic and inert towards endogenous cellular nucleophiles.
One of the earliest cyclooctynes, MOFO (monofluorinated cyclooctyne) (14), performed
well inside cells, giving signal to background ratios consistently >5:1 in the context of
fluorescein targeting to nuclear LAP. However, its slow second-order rate constant
(4.3x10 3 M's') (14) was a severe drawback. DIFO (di-fluorinated cyclooctyne) (15)
was developed in attempt to increase the kinetics of cyclooctynes. However, we have
seen that in the case of DIFO, although the probe is quite reactive, its nonspecific
reactions inside cells led to a permanent background. The next generation cyclooctynes,
fused to two benzene rings, increased ring strain and hence the second-order rate
constants. Not surprisingly, we found both cyclooctynes, ADIBO (19) and DIBO (18),
gave about 4-fold higher absolute signal in cells, compared to MOFO, probably due to
increased yield of the cycloaddition product. ADIBO and DIBO also labeled the LAP-
tagged cytoskeletal proteins actin and MAP2 efficiently. However, the dibenzo rings that
confer the enhanced reactivity for such cyclooctynes also made them more hydrophobic.
102
The increase in signal was accompanied by an increase in background, likely due to
nonspecific binding of these hydrophobic probes. Consequently, the signal to background
ratios were comparable for ADIBO-, DIBO-, and MOFO-fluorescein conjugates.
When we extended the cyclooctyne comparison to other fluorophores, we found
that ADIBO and DIBO conjugates to relatively well-behaved hydrophilic fluorophores
such as fluorescein and Oregon Green gave satisfactory labeling, but when we tried to
target very hydrophobic fluorophores such as ATTO 647N, the combined hydrophobicity
of the dye and the cyclooctyne (ADIBO) precluded successful labeling, due to high
nonspecific binding. This was alleviated by using the less hydrophobic MOFO instead.
We were able to label LAP-tagged p-actin with ATTO 647N conjugated to MOFO. To
reduce the nonspecific and mitochondrial background from large, hydrophobic, and
positively charged fluorophores such as ATTO 647N, we explored strategies utilizing the
mitochondrial uncoupling agent FCCP that disrupts the chemiosmotic gradient across the
mitochondrial inner membrane, as well as the fluorescence quencher QSY21 to reduce
mitochondrial-specific background, but these approaches only led to minimal reduction
of background. In the case of LAP-p-actin labeling with ATTO 647N, QSY21 quencher
remaining in the cytosol also quenched labeling signal. Our study illustrates the need for
new cell-permeable, bright, and hydrophilic fluorophore, as well as the design and
synthesis of cyclooctyne probes that combine high reactivity and stability with low
hydrophobicity.
For some applications, however, prolonged washout of unbound probe or using a
fluorescence quencher to reduce background is sometimes impossible, or the remaining
probe can further frustrate detection. To minimize the amount of excess probe, new and
improved bioorthogonal ligation chemistries that have higher reactivity than strain-
promoted azide-alkyne cycloaddition open up alternative possibilities that offer high
labeling sensitivity and yield, even at low probe concentrations. ADIBO and DIBO have
been reported to react with nitrones at a rate ranging from 3 to 5 magnitudes faster than
with azide (24,25). Cycloaddition reaction with nitrones have been applied in protein
functionalization in vitro (25), attaching antibodies to surface of hybrid multifunctional
nanoparticles (26), and labeling EGF receptors on the cell surface of human breast cancer
103
cells (27). However, the intracellular behavior of nitrone and this fast cycloaddition
reaction have yet to be tested and characterized.
Another reaction involving the Diels-Alder cycloaddition between tetrazine and
trans-cyclootene has reported an explosive second-order rate constant of 2000 M-1s-1
(28), several magnitudes faster than strain-promoted azide-alkyne cycloaddition. Daniel
Liu and Anupong Tangpeerachaikul from our lab have demonstrated in a separate work
using LplA-based probe ligation in combination with this inverse Diels-Alder
cycloaddition to target fluorophores to intracellular and cell surface proteins in live cells
(29). The very fast cycloaddition kinetics yields substantial improvements in signal to
background ratio for protein labeling. Furthermore, the ability of tetrazine to quench
fluorescence of conjugated fluorophore until reaction with trans-cyclooctene enables the
reaction to be fluorogenic, a very desirable feature of bioorthogonal reactions. A
fluorogenic bioorthogonal reaction is essentially a background-free labeling system that
would yield a detectable signal only after a chemoselective reaction occurs. A
fluorogenic system would eliminate the time-consuming washout steps to remove excess
reagents and would allow high concentration of reagent to be used if needed without
worrying about increasing background. The inverse Diels-Alder cycloaddition has been
applied in orthogonal labeling of cell surface glycans with the click reaction between
dibenzocyclooctyne and azide (10), in vitro labeling of purified proteins that are
genetically modified through the UAA method with trans-cyclooctene (30), as well as on
the cell surface of cancer cells (31). It would be exciting to see demonstration of
fluorogenic labeling inside live cells, where the background signal caused by
hydrophobic fluorophore sticking would be quenched by tetrazine.
Another interesting advance is in copper-catalyzed click chemistry. Previously
discounted for cellular applications due to copper toxicity, new improvements in copper
ligand design and reactive oxygen species scavenging have made it possible to perform
click chemistry on live cell surfaces and even animals (32). If the toxicity can be further
reduced, while preserving the fast kinetics of ligation (currently 104-107 fold greater
than strain-promoted cycloaddition), then copper-catalyzed click chemistry will be
competitive with other methods for bioorthogonal derivatization on the cell surface.
104
Furthermore, aside from utilizing bioorthogonal chemistries to target large
fluorophores, postdoc Wenjing Wang in our lab is working towards evolving LplA to
accept and ligate various large fluorophores directly in a single step. This challenging
approach involves engineering an "exit tunnel" in LplA for large fluorophore structures
so they do not have to directly interact with LplA or fit into its binding pocket. This
method could lead to an efficient labeling technique that shortens labeling time and
possibly gives higher ligation yield than what we achieved using strain-promoted
cycloaddition. To make the enzyme recognize and ligate not a single large fluorophore
but applicable towards, for example, all rhodamine based fluorophores, yet maintain the
specificity towards LAP substrate, Wenjing is currently using yeast-evolution to evolve a
novel LplA mutant with an exit tunnel to accommodate large fluorophores.
Several fluorophores targeted using LplA and strain-promoted cycloaddition in
this study have exemplary properties that make them attractive alternatives to fluorescent
proteins. X-rhodamine is a bright and photostable fluorophore commonly used for
speckle imaging of actin (33). ATTO dyes are some of the best fluorophores in terms of
their superior photophysical properties that make them suitable for both STED
(stimulated emission depletion) (34,35) and STORM-type (36) super-resolution
microscopies. On the cell surface, we targeted Alexa Fluor 647, an excellent fluorophore
that has been used for countless ensemble and single-molecule imaging experiments. If
methods can be developed to deliver sulfonated fluorophores - which include the cyanine
dyes and Alexa Fluors - across cell membranes (37), then these too should be targetable
to specific intracellular proteins using PRIME.
In conclusion, we have extended the PRIME technology to target larger
fluorophores with superior photophysical properties than coumarin, via strain-promoted
azide-alkyne cycloaddition. Considered in the context of other protein labeling methods
(38), the disadvantages of the approach presented here are the requirement for co-
expression of the LplA labeling enzyme, the unavoidable background caused by
nonspecific binding of cyclooctyne-fluorophore conjugates (albeit low in the case of
hydrophilic fluorophores such as fluorescein and Oregon Green), and the signal which is
fundamentally limited by the kinetics of strain-promoted cycloaddition chemistry. Given
these factors, the methodology will be most useful as a nontoxic (in contrast to FlAsH
105
(39)) labeling method for abundant proteins, whose fusion to large tags (such as
fluorescent proteins, HaloTag (40), or SNAP tag (41) perturb function, such as actin. We
hope to offer insights in choosing the appropriate cyclooctynes and fluorophores for
specific applications. With the development of new ligation methods and superior
fluorophores, we anticipate the routine labeling and imaging of proteins with small-
molecule fluorescent probes inside living cells.
106
Experimental methods
General synthetic methods
All reagents were the highest grade available and purchased from Sigma-Aldrich,
Anaspec, Thermal Scientific, TCI America, Alfa Aesar, or Life Technologies and used
without further purification. Anhydrous solvents were drawn from Sigma-Aldrich
SureSeal bottles. Analytical thin layer chromatography was performed on 0.25 mm silica
gel 60 F254 plates and visualized under short or long wavelength UV light, or after
staining with bromocresol green or ninhydrin. Flash column chromatography was carried
out using silica gel (ICN SiliTech 32-63D). Mass spectrometric analysis was performed
on an Applied Biosystems 200 QTRAP mass spectrometer using electrospray ionization.
HPLC analysis and purification were performed on a Varian Prostar Instrument equipped
with a photo-diode-array detector. A reverse-phase Microsorb-MV 300 C18 column (250
x 4.6 mm dimension) was used for analytical HPLC. NMR spectra were recorded on a
Bruker AVANCE 400 MHz instrument.
Synthesis of ADIBO- and DIBO-fluorophore conjugates
HO
K0
N 5(6)-carboxyfluorescein, NHS N
N_____________ N H
O 5NH2 anhydrous DMF, Et3N O 5 CO 0
ADIBO-amine 0
O' O ADIBO-fluorescein
- 0
acetic anhydride N H
0 0E t3N O 5NO0
ADIBO-fluorescein diacetate
ADIBO-fluorescein diacetate The synthesis of aza-dibenzocyclooctyne-amine
(ADIBO-amine) has been previously described.' To a solution of ADIBO-amine (3 mg, 9
107
ptmol) in anhydrous DMF (500 piL) was added triethylamine (Et3N, 3.8 pL, 27 pmol) and
5,6-carboxyfluorescein succinimidyl ester (NHS) (9.9 ptmol, AnaSpec). The reaction was
allowed to proceed for 10 hr at room temperature. Solvent was then removed under
reduced pressure. The residue was subsequently dissolved in acetic anhydride (200 pL,
2.1 mmol) and allowed to stir for 30 min at room temperature. The color of the solution
changed from bright yellow to colorless during the course of the reaction. Excess acetic
anhydride was removed under reduced pressure. The resulting residue was purified by
silica gel chromatography using 0-5% methanol in dichloromethane to afford ADIBO-
fluorescein diacetate (Rf = 0.5 in 10% methanol in dichloromethane). Purified product
was analyzed on HPLC which showed single peak with absorbance at 210 nm. Estimated
yield for two steps is ~60%. ESI-MS for ADIBO-fluorescein diacetate: calculated for
[M+H]f: 761.24; observed 760.74. We note that the 5- and 6-regioisomers of fluorescein
derivatives could have different cell entry and washout properties, which were not
investigated in this study.
N
N. - 5(6)-carboxytetramethylrhodamine, NHS N
N ,__ _N__ _ _ _ _
H2  anhydrous DMF, DIEA 2 - N
ADIBO-amine 0
ADIBO-TMR
ADIBO-TMR To a solution of ADIBO-amine (70 mg, 0.46 mmol) in anhydrous
DMF (2 mL) was added N,N-diisopropylethylamine (DIEA, 0.12 mL, 0.69 mmol) and
5,6-carboxytetramethylrhodamine (TMR) succinimidyl ester (NHS) (100 mg, 0.23 mmol,
Sigma-Aldrich). The reaction was allowed to proceed for 12 hr at room temperature.
Solvent was then removed under reduced pressure. Conjugate was purified by silica gel
chromatography using 0-2% methanol in chloroform to provide dark red crystalline. The
purity of the product was checked by HPLC. ESI-MS for ADIBO-TMR: calculated
[M+H]*: 688.27; observed 688.8. We note that the 5- and 6-regioisomers of TMR
108
derivatives could have different cell entry and washout properties, which were not
investigated in this study.
N
N| N O N
O 5 YN O R5 N
0 0 -.
ADI BO-ATTO 647N ADI BO-ATTO 655 0-
ADIBO-ATTO 647N, ADIBO-ATTO 655 ADIBO conjugates to ATTO 647N
and ATTO 655 were synthesized in a similar manner from ADIBO-amine.1 ATTO 647
NHS ester (Sigma-Aldrich) and ATTO 655 NHS ester (Sigma Aldrich) were used.
Conjugates were purified by silica gel chromatography using 0-2% methanol in
dichloromethane. ESI-MS for ADIBO-ATTO 647N: calculated [M+H]f: 946.56;
observed 946.29. ESI-MS for ADIBO-ATTO 655: calculated [M+H]*: 827.34; observed
827.51.
O>0
/ \1 0I 0
O H -0-
O N- N / 0 0
H 0
DIBO-fluorescein diacetate
DIBO-fluorescein diacetate DIBO-fluorescein diacetate was synthesized in an
analogous manner to ADIBO-fluorescein diacetate, from commercial DIBO-amine
(Invitrogen) and fluorescein NHS ester (AnaSpec). The conjugate was purified by silica
gel chromatography using 0-5% methanol in dichloromethane. ESI-MS for DIBO-
fluorescein diacetate: calculated [M+H]*: 763.22; observed 763.86.
109
OF
4 
-' 0 F OO OF0 H 0
0 N~" N
H 0
DIBO-Oregon Green 488 diacetate
DIBO-Oregon Green diacetate DIBO-Oregon Green 488 diacetate was a gift
from Kyle Gee (Life Technologies).
Synthesis of MOFO-, DIMAC-, and DIFO-fluorophore conjugates
F 1. PFP-TFA, Et3N, F
CH 2C 2  I 5(6)-carboxyfluorescein, NHS
OH N{+NH2. N, N'-dimethyl-1,6- 6 DMF, Et3N
0 hexanediamine, Et3N, 0
MOFO acid CH 2CI2  MOFO-HDDA
HO
00
F - 0 acetic anhydride F / \ 0
N, 9 N~ /CO2  K 1 0-
N C2 NEtaN N0 0
0 000
MOFO-fluorescein MOFO-fluorescein diacetate
MOFO-fluorescein diacetate To a solution of MOFO cyclooctyne acid2 (5 mg,
19 tmol) in 500 p.L anhydrous dichloromethane was added pentafluorophenyl
trifluoroacetate (PFP-TFA, 9.8 pL, 57 pmol) and Et3N (8 p.L, 57 pmol). The reaction was
allowed to proceed for 2 hr at room temperature. N, N'-dimethyl-1,6-hexanediamine
(HDDA, 114 [tmol) was then added to the reaction mixture, which was allowed to stir for
5 hr at room temperature. Solvent was removed under reduced pressure. The reaction
mixture was purified by silica gel chromatography (10-15% methanol in
dichloromethane) to afford MOFO-N,N'-dimethyl-1,6-hexanediamine (MOFO-HDDA).
MOFO-HDDA was dissolved in anhydrous DMF (300 pL), and 5(6)-carboxyfluorescein,
succinimidyl ester (9.8 mg, 20.9 pnol) and Et3N (8 pL, 57 ptmol) were added to the
mixture, which was allowed to stir for 10 hr at room temperature. Solvent was removed
110
under reduced pressure. The residue was dissolved in a small amount of acetic anhydride
(<200 tL) and allowed to stir for 30 min at room temperature. After removal of acetic
anhydride under reduced pressure, the reaction mixture was purified by silica gel
chromatography (solvent gradient 0-5% methanol in dichloromethane) to afford MOFO-
fluorescein diacetate (Rf = 0.4 in 10% methanol in dichloromethane). Estimated overall
yield for four steps, 30-40%. ESI-MS for MOFO-fluorescein diacetate: calculated
[M+H]f: 829.34; observed 829.44.
F
0 0
MOFO-X-rhodamine
F
0
0
MOFO-ATTO 647N
MOFO-X-rhodamine, MOFO-ATTO 647N MOFO-HDDA was synthesized as
described above, then conjugated to 5(6)-X-rhodamine NHS ester (Anaspec, 5(6)-ROX,
SE) or ATTO 647N NHS ester (Sigma-Aldrich). Conjugates were purified by silica gel
chromatography using 0-5% methanol in dichloromethane for MOFO-X-rhodamine and
0-2% methanol in dichloromethane for MOFO-ATTO 647N. ESI-MS for MOFO-X-
rhodamine: calculated [M+H]*: 903.49; observed 903.72. ESI-MS for MOFO-ATTO
647N: calculated [M+H]*: 1014.66; observed 1014.42.
O
MeO 0
N0O' * . -
0N N.N< 0
0 0
DIMAC -fluorescein diacetate
N N
O O0 0
DIFO-fluorescein diacetate
111
O'
O
DIMAC-fluorescein diacetate, DIFO-fluorescein diacetate Fluorescein
diacetate conjugates to DIMAC 3 and DIFO4 were synthesized in a similar manner from
their respective acids. Conjugates were purified by silica gel chromatography using 0-5%
methanol in dichloromethane. ESI-MS for DIMAC-fluorescein diacetate: calculated [M-
H]~: 752.33; observed 752.40. ESI-MS for DIFO-fluorescein diacetate: calculated
[M+H]*: 847.33; observed 847.26.
Plasmids
For bacterial expression of LplA, we used His6-LplA in pYFJl6.5 For mammalian
expression of LplA, we used His6-FLAG-LplA in pcDNA3. 5 For mammalian expression
of LAP fusion proteins, we used LAP-p-actin and LAP-MAP2 in Clontech vector,' LAP-
LDL receptor in pcDNA4, and LAP-neurexin-l p in pNICE. LAP-BFP expression
constructs (LAP-BFP, LAP-BFP-NLS, LAP-BFP-CAAX, and LAP-BFP-NES) in
pcDNA3 and LAP-mCherry in pcDNA3 were generated from corresponding pcDNA3-
LAP-YFP plasmids5 by replacing YFP with BFP or mCherry, using the BamHI and
EcoRI restriction sites. All LplA and LAP point mutants were prepared via QuikChange
site-directed mutagenesis. Complete sequences of plasmids used in this study are
available at http://stellar.mit.edu/S/project/tinglabreagents/r02/materials.html (Ting lab
website).
Immunofluorescence staining of LplA (Figure 3-4)
After live cell imaging, cells were fixed with 3.7% formaldehyde in Dulbecco's
Phosphate Buffered Saline (DPBS) pH 7.4 for 10 min at room temperature followed by
cold precipitation with methanol for 5 min at -20 *C, then blocked with 3% BSA in
DPBS for 1 hr at room temperature. To visualize FLAG-tagged LplA, cells were
incubated with 4 gg/mL mouse monoclonal anti-FLAG antibody (Sigma-Aldrich) in 1%
BSA in DPBS for 1 hr at room temperature. Cells were further washed and incubated
with 4 pg/mL goat anti-mouse IgG antibody conjugated to Alexa Fluor 568 (Life
Technologies) in 1% BSA in DPBS for 1 hr at room temperature, then washed and
imaged.
112
Cell fixation after live cell DIMAC-fluorescein and DIFO-fluorescein labeling
(Figure 3-6)
After live cell imaging, cells were fixed with 3.7% formaldehyde in DPBS pH 7.4 for 10
min at room temperature followed by cold precipitation with methanol for 5 min at -20 *C.
Cells were then washed with DPBS several times over 10 min, before imaging.
Cell surface and intracellular labeling with commercial DIBO conjugates (Figure 3-
17)
DIBO-Alexa Fluor 647 cell surface labeling (Figure 3-17 top)
2HEK cells plated on glass coverslips in wells of a 48-well cell culture plate (0.95 cm per
well) were transfected with 100 ng pcDNA4-LAP-LDL receptor or 400 ng pNICE-LAP-
neurexin-1p using Lipofectamine 2000. At 18 hr after transfection, cells were washed
three times with MEM. Enzymatic ligation of azide 9 on the cell surface was performed
in MEM with 5 ptM wm7LplA, 500 pM azide 9, 2 mM ATP and 2 mM magnesium acetate
for 20 min at room temperature (to minimize internalization of cell-surface proteins).
After washing three times with MEM, cells were incubated with 10 pM DIBO-Alexa
Fluor 647 in MEM for 10 min at room temperature. Cells were then washed three times
with MEM and imaged live.
DIBO-biotin cell surface and intracellular labeling (Figure 3-17 middle and bottom)
DIBO-biotin cell surface labeling was performed in the same manner as DIBO-Alexa
Fluor 647 cell surface labeling, described above. After DIBO-biotin incubation, cells
were washed three times with DPBS and fixed with 3.7% formaldehyde in DPBS pH 7.4
for 10 min at room temperature, followed by cold precipitation with methanol for 5 min
at -20 "C. Fixed cells were then blocked with 1% casein in DPBS for 1 hr at room
temperature. To visualize specific labeling, cells were stained with streptavidin
conjugated to Alexa Fluor 568 or Alexa Fluor 647 in 0.5% casein in DPBS for 5 min at
room temperature, followed by washing three times with DPBS and imaging.
For DIBO-biotin intracellular labeling, HEK cells plated on glass coverslips in wells of a
48-well cell culture plate (0.95 cm2 per well) were transfected with 400 ng pcDNA3-
LAP-BFP-NLS and 200 ng pcDNA3-W37 LplA. Azide 9 labeling/washout and DIBO-
113
biotin labeling/washout were performed in the same manner as in Figure 3-18. After
DIBO-biotin washout, cells were fixed and stained with streptavidin-Alexa Fluor 568, as
described above.
Analysis of overall two-step ligation yield after strain-promoted cycloaddition in live
cells (Figure 3-18)
HEK cells plated into wells of a 12-well culture plate (4 cm 2 per well) were transfected
with 500 ng pcDNA3-W371LplA and 1000 ng pcDNA3-LAP-mCherry using
Lipofectamine 2000 (Life Technologies). Azide 9 labeling and washout were performed
in the same manner as in Figure 2-8. After excess azide 9 washout, cells were incubated
in MEM containing 10 [M DIBO-biotin (Life Technologies) for 10 min at 37 "C.
Thereafter, cells were further washed for 2.5 hr to remove excess DIBO-biotin. Cells
were then harvested and lysed in the same manner as in Figure 2-8. The cell lysate was
incubated with 5 jM of streptavidin for 1 hr at 4 *C, then analyzed on a 12% SDS-
polyacrylamide gel at constant 200 V, under conditions known to preserve biotin-
streptavidin binding as well as streptavidin's subunit association.6 Prior to Coomassie
staining, in-gel fluorescence of mCherry was visualized on FUJIFILM FLA-9000
instrument using SHG532 laser and Long Pass Green (LPG) filter.
Quantitative analysis of MOFO-fluorescein labeling of LAP-BFP using four LpIA
mutant/azide substrate pairs (Figure 3-2)
Cells with fluorescein signal at least 2-fold greater than autofluorescence from
untransfected cells, and BFP signal at least 5-fold greater than autofluorescence were
selected by hand for analysis. For each of these cells, the entire area of the cell
representing signal was circled. SlideBook was used to calculate the mean intensities in
both channels. The background-corrected mean fluorescein intensity was plotted against
the background-corrected mean BFP intensity using Excel.
Quantitative analysis of LplA mutant expression levels in cells (Figure 3-4)
Cells with Alexa Fluor 568 signal at least 1.5-fold great than background (area without
any cell) were selected by hand for analysis. For each of these cells, the entire area of the
114
cell representing signal was circled. SlideBook was used to calculate the mean intensities
in the channel. The background-corrected mean Alexa Fluor 568 intensity was plotted
using Excel.
Quantitative analysis of fluorophore-cyclooctyne labeling specificity (Figures 3-5 B
and 3-10)
Cells with signal at least 3-fold greater than autofluorescence from untransfected cells in
the cyclooctyne channel were selected by hand for analysis. For each of these cells, one
region in the cytosol (representing background) and one region in the nucleus
(representing specific signal) were manually circled. The background-corrected mean
fluorescence intensity was determined for both regions using SlideBook. Excel was used
to plot the nuclear versus cytosolic fluorescence intensity for each cell. Since ATTO
647N labeling signal was low, we selected for analysis cells with signal at least 2-fold
greater than autofluoresence from untransfected cells in the ATTO 647N channel.
Other protocols
LplA and mutants were expressed and purified as previously described.5 The 13-amino
acid LAP peptide (H2N-GFEIDKVWYDLDA-CO 2H)7 was synthesized by the Tufts
University Peptide Synthesis Core Facility and purified to >96% homogeneity.
115
References
1. Manova RK, Pujari SP, Weijers CAGM, Zuilhof H, Van Beek TA. Copper-Free
Click Biofunctionalization of Silicon Nitride Surfaces via Strain-Promoted Alkyne-
Azide Cycloaddition Reactions. Langmuir. 2012 Jun 12;28(23):8651-63.
2. Singh I, Freeman C, Madder A, Vyle JS, Heaney F. Fast RNA conjugations on solid
phase by strain-promoted cycloadditions. Org. Biomol. Chem. 2012 Aug
1;10(33):6633-9.
3. Chakrabarty R, Stang PJ. Post-assembly Functionalization of Organoplatinum(II)
Metallacycles via Copper-free Click Chemistry. J. Am. Chem. Soc. 2012 Sep
12;134(36):14738-41.
4. Debets MF, Van Berkel SS, Schoffelen S, Rutjes FPJT, Van Hest JCM, Van Delft
FL. Aza-dibenzocyclooctynes for fast and efficient enzyme PEGylation via copper-
free (3+2) cycloaddition. Chemical Communications. 2010;46(1):97.
5. Khatwani SL, Kang JS, Mullen DG, Hast MA, Beese LS, Distefano MD, et al.
Covalent protein-oligonucleotide conjugates by copper-free click reaction.
Bioorganic & Medicinal Chemistry. 2012 Jul 15;20(14):4532-9.
6. Jang S, Sachin K, Lee H, Kim DW, Lee HS. Development of a Simple Method for
Protein Conjugation by Copper-Free Click Reaction and Its Application to Antibody-
Free Western Blot Analysis. Bioconjugate Chem. 2012 Nov 21;23(11):2256-61.
7. german group, UAA into GFP OCT, and bicyclo[6.1.0]nonyne, and terminal alkyne,
1.
8. Temming RP, Eggermont L, Eldijk MB van, Hest JCM van, Delft FL van. N-terminal
dual protein functionalization by strain-promoted alkyne-nitrone cycloaddition. Org.
Biomol. Chem. 2013 Apr 3;1l(17):2772-9.
9. Hahne H, Sobotzki N, Nyberg T, Helm D, Borodkin VS, Van Aalten DMF, et al.
Proteome Wide Purification and Identification of O-GlcNAc-Modified Proteins
Using Click Chemistry and Mass Spectrometry. J. Proteome Res. 2013 Feb
1;12(2):927-36.
10. Niederwieser A, Spate A-K, Nguyen LD, Jingst C, Reutter W, Wittmann V. Two-
Color Glycan Labeling of Live Cells by a Combination of Diels-Alder and Click
Chemistry. Angewandte Chemie International Edition. 2013;52(15):4265-8.
11. Beatty KE, Fisk JD, Smart BP, Lu YY, Szychowski J, Hangauer MJ, et al. Live-Cell
Imaging of Cellular Proteins by a Strain-Promoted Azide-Alkyne Cycloaddition.
ChemBioChem. 2010; 11(15):2092-5.
12. Chang PV, Prescher JA, Sletten EM, Baskin JM, Miller IA, Agard NJ, et al. Copper-
free click chemistry in living animals. PNAS. 2010 Feb 2;107(5):1821-6.
116
13. Agard NJ, Prescher JA, Bertozzi CR. A Strain-Promoted [3 + 2] Azide-Alkyne
Cycloaddition for Covalent Modification of Biomolecules in Living Systems. J. Am.
Chem. Soc. 2004 Nov 1;126(46):15046-7.
14. Agard NJ, Baskin JM, Prescher JA, Lo A, Bertozzi CR. A Comparative Study of
Bioorthogonal Reactions with Azides. ACS Chem. Biol. 2006 Nov 1;1(10):644-8.
15. Codelli JA, Baskin JM, Agard NJ, Bertozzi CR. Second-Generation Difluorinated
Cyclooctynes for Copper-Free Click Chemistry. J. Am. Chem. Soc. 2008 Aug
1;130(34):11486-93.
16. Sletten EM, Bertozzi CR. A Hydrophilic Azacyclooctyne for Cu-Free Click
Chemistry. Org. Lett. 2008 Jul 1;10(14):3097-9.
17. Jewett JC, Sletten EM, Bertozzi CR. Rapid Cu-Free Click Chemistry with Readily
Synthesized Biarylazacyclooctynones. J Am Chem Soc. 2010 Mar 24;132(11):3688-
90.
18. Ning X, Guo J, Wolfert MA, Boons G-J. Visualizing Metabolically Labeled
Glycoconjugates of Living Cells by Copper-Free and Fast Huisgen Cycloadditions.
Angewandte Chemie International Edition. 2008;47(12):2253-5.
19. Kuzmin A, Poloukhtine A, Wolfert MA, Popik VV. Surface Functionalization Using
Catalyst-Free Azide-Alkyne Cycloaddition. Bioconjugate Chem. 2010 Nov
17;21(11):2076-85.
20. Griffith OW. Biologic and pharmacologic regulation of mammalian glutathione
synthesis. Free Radical Biology and Medicine. 1999 Nov;27(9-10):922-35.
21. Vafai SB, Mootha VK. Mitochondrial disorders as windows into an ancient
organelle. Nature. 2012 Nov 15;491(7424):374-83.
22. Poburko D, Demaurex N. Regulation of the mitochondrial proton gradient by
cytosolic Ca2+ signals. Pflugers Arch - Eur J Physiol. 2012 Jul 1;464(1):19-26.
23. Blerkom JV, Davis P, Alexander S. Inner mitochondrial membrane potential (A'Fm),
cytoplasmic ATP content and free Ca2+ levels in metaphase II mouse oocytes. Hum.
Reprod. 2003 Nov 1;18(11):2429-40.
24. McKay CS, Chigrinova M, Blake JA, Pezacki JP. Kinetics studies of rapid strain-
promoted [3 + 2]-cycloadditions of nitrones with biaryl-aza-cyclooctynone. Organic
& Biomolecular Chemistry. 2012; 10(15):3 066.
25. Ning X, Temming RP, Dommerholt J, Guo J, Ania DB, Debets MF, et al. Protein
Modification by Strain-Promoted Alkyne-Nitrone Cycloaddition. Angewandte
Chemie International Edition. 2010;49(17):3065-8.
117
26. Colombo M, Sommaruga S, Mazzucchelli S, Polito L, Verderio P, Galeffi P, et al.
Site-Specific Conjugation of ScFvs Antibodies to Nanoparticles by Bioorthogonal
Strain-Promoted Alkyne-Nitrone Cycloaddition. Angewandte Chemie International
Edition. 2012;51(2):496-9.
27. McKay CS, Blake JA, Cheng J, Danielson DC, Pezacki JP. Strain-promoted
cycloadditions of cyclic nitrones with cyclooctynes for labeling human cancer cells.
Chem. Commun. 2011 Aug 30;47(36):10040-2.
28. Blackman ML, Royzen M, Fox JM. The Tetrazine Ligation: Fast Bioconjugation
based on Inverse-electron-demand Diels-Alder Reactivity. J Am Chem Soc. 2008 Oct
15;130(41):13518-9.
29. Liu DS, Tangpeerachaikul A, Selvaraj R, Taylor MT, Fox JM, Ting AY. Diels-Alder
Cycloaddition for Fluorophore Targeting to Specific Proteins inside Living Cells. J.
Am. Chem. Soc. 2012 Jan 18;134(2):792-5.
30. Lang K, Davis L, Wallace S, Mahesh M, Cox DJ, Blackman ML, et al. Genetic
Encoding of Bicyclononynes and trans-Cyclooctenes for Site-Specific Protein
Labeling in Vitro and in Live Mammalian Cells via Rapid Fluorogenic Diels-Alder
Reactions. J. Am. Chem. Soc. 2012 Jun 27;134(25):10317-20.
31. Devaraj NK, Upadhyay R, Haun JB, Hilderbrand SA, Weissleder R. Fast and
Sensitive Pretargeted Labeling of Cancer Cells via Tetrazine/Trans-Cyclooctene
Cycloaddition. Angew Chem Int Ed Engl. 2009;48(38):7013-6.
32. Soriano del Amo D, Wang W, Jiang H, Besanceney C, Yan AC, Levy M, et al.
Biocompatible Copper(I) Catalysts for in Vivo Imaging of Glycans. J. Am. Chem.
Soc. 2010 Dec 1;132(47):16893-9.
33. Lim JI, Sabouri-Ghomi M, Machacek M, Waterman CM, Danuser G. Protrusion and
Actin Assembly are Coupled to the Organization of Lamellar Contractile Structures.
Exp Cell Res. 2010 Aug 1;316(13):2027-41.
34. Mueller V, Ringemann C, Honigmann A, Schwarzmann G, Medda R, Leutenegger
M, et al. STED Nanoscopy Reveals Molecular Details of Cholesterol- and
Cytoskeleton-Modulated Lipid Interactions in Living Cells. Biophys J. 2011 Oct
5; 101(7):1651-60.
35. Westphal V, Rizzoli SO, Lauterbach MA, Kamin D, Jahn R, Hell SW. Video-Rate
Far-Field Optical Nanoscopy Dissects Synaptic Vesicle Movement. Science. 2008
Apr 11;320(5873):246-9.
36. Dempsey GT, Vaughan JC, Chen KH, Bates M, Zhuang X. Evaluation of
fluorophores for optimal performance in localization-based super-resolution imaging.
Nat Meth. 2011 Dec;8(12):1027-36.
118
37. Pauff SM, Miller SC. Synthesis of Near-IR Fluorescent Oxazine Dyes with Esterase-
Labile Sulfonate Esters. Org. Lett. 2011 Dec 2;13(23):6196-9.
38. Wombacher R, Cornish VW. Chemical tags: Applications in live cell fluorescence
imaging. Journal of Biophotonics. 2011;4(6):391-402.
39. Adams SR, Campbell RE, Gross LA, Martin BR, Walkup GK, Yao Y, et al. New
Biarsenical Ligands and Tetracysteine Motifs for Protein Labeling in Vitro and in
Vivo: Synthesis and Biological Applications. J. Am. Chem. Soc. 2002 May
1;124(21):6063-76.
40. Los GV, Encell LP, McDougall MG, Hartzell DD, Karassina N, Zimprich C, et al.
HaloTag: A Novel Protein Labeling Technology for Cell Imaging and Protein
Analysis. ACS Chem. Biol. 2008 Jun 1;3(6):373-82.
41. Gautier A, Juillerat A, Heinis C, Correa Jr. IR, Kindermann M, Beaufils F, et al. An
Engineered Protein Tag for Multiprotein Labeling in Living Cells. Chemistry &
Biology. 2008 Feb 22;15(2):128-36.
119
Chapter 4. Development of a photocrosslinker ligase for protein-protein interaction
detection
The work discussed in this chapter is unpublished. We would like to thank Laura Gerber
(Professor Dick Schrock lab, MIT Chemistry Department) for lending us the Porta-Ray
400R photocrosslinking instrument and the Lippard Lab (MIT Chemistry Department)
for lending us Stratalinker photocrosslinking instrument for initial photocrosslinking
experiments. We would like to thank Arturo Pizano (Professor Dan Nocera lab, MIT
Chemistry Department) for his help in obtaining transient absorption spectra of
benzophenone compounds. We would also like to thank Professor Jason Chin (Molecular
Research Council, Laboratory of Molecular Biology) for helpful discussions.
120
Introduction to methods of protein-protein interaction detection
Protein-protein interactions, or the "interactome", form the basis of almost every
cellular event, from growth, migration, division, to cell-cell communication. It has been
estimated that 650,000 protein-protein interactions exist in the human interactome (1).
The interactome not only dictates the normal functioning of the organism, but also
underlies its diseased state. Mapping protein-protein interactions using biophysical,
biochemical, or genetic means have provided a wealth of information on macromolecular
protein complexes. Several major methods for mapping protein-protein interactions
include co-immunoprecipitation, the yeast two-hybrid assay, protein-fragment
complementation assay (PCA), and photoaffinity labeling (PAL).
Co-immunoprecipitation is one of the earliest methods to identify members of
interacting protein complexes. In co-immunoprecipitation, a protein of interest is
identified through affinity capture and subjected to a series of biochemical purification
methods that rely on typically conventional chromatography steps. Even though it has
undoubtedly led to new protein-protein interaction discoveries, co-immunoprecipitation
has many limitations (2). First, the technique does not allow any spatial resolution
because interacting proteins are identified in the cell lysates when compartmentalization
of various organelles has been broken down. When taken out of endogenous environment
inside the cell, the stoichiometry and concentration of interacting proteins are severely
disturbed which may lead to false loss of interacting partners. Furthermore, in a protein-
protein interaction complex, the antigen may be buried and inaccessible by the antibody.
Finally, the efficiency of co-immunoprecipitation can depend largely on the availability
and affinity of the antibodies.
One of the most widely used methods for protein-protein interaction detections is
the yeast two-hybrid system (3,4). The technology splits the two functional domains,
often the DNA binding domain and the activation domain, of a transcriptional activator
and fuses each domain to one of a pair of potentially interacting proteins, called the bait
and the prey protein. The physical interaction of the bait and prey protein brings near the
two activator domains and reconstitutes the transcriptional activator's ability to turn on a
reporter gene. The yeast two-hybrid method has not only been used to assess individual
protein-protein interactions, but also to map the interactions in the entire proteome (5).
121
Array-based automated yeast two-hybrid approaches have been instrumental for
interaction studies on a global proteomic scale. In the seminal work of Stelzl et al. by
fusing 4456 baits and 532 preys to activator domains, the authors identified more than
3000 novel protein-protein interactions including interactions proposed to form disease-
related proteins. Furthermore, yeast two-hybrid screens that focus on disease pathways
and functions have uncovered among many other remarkable studies, proteins involved in
inherited ataxias (6) and neurodegenerative Huntington's disease (7).
However, despite the high success of large-scale yeast two-hybrid approaches, the
technique is not without some serious drawbacks. First, the frequent occurrence of false
positives and false negatives imposes a challenge to meaningful data interpretation. False
positives can be caused by overexpression of the heterologous proteins as well as self-
activation of the reporter gene by the bait protein in the absence of bait-prey interaction.
Some false negatives can be caused by improper folding of the bait and prey proteins
once they are fused to their respective domains, or mislocation and the inability of the
protein partners to translocate to the nucleus, such as interactions of membrane proteins
(4). Estimations for false positives range from 25 to 45 % (8). Individual confirmation of
protein-protein interactions with alternative methods is not always possible for large-
scale yeast two-hybrid approaches. Another serious drawback of yeast two-hybrid
method is the poor time resolution. The amplification of the reporter gene signal requires
first the formation of the protein interaction under study, reconstitution of the
transcriptional activator, and finally the translocation of the transcriptional activator to
the nucleus. The entire process can take hours after the protein-protein interaction has
occurred (9). Moreover, protein-protein interactions requiring relevant cofactors that are
not present in the yeast host organism may be lost. Because of these limitations,
individual confirmation of protein-protein interactions with alternative methods, such as
co-immunoprecipitation, is often required.
Variants of the yeast two-hybrid method have been developed to address the
disadvantages and limitations of original yeast two-hybrid technique. Regarding the
cellular compartment-dependent protein-protein interactions, such as the interaction
between membrane proteins or between a membrane protein and a cytosolic protein,
systems based on split-ubiquitin and split-TEV have been created so that the reconstituted
122
protein can act to release the reporter protein, which can be either a transcription activator
or light emitting enzyme like luciferase. In the case of split-ubiquitin, reconstituted
ubiquitin is recognized and cut by ubiquitin-specific proteases, which in turn release the
transcription factor that is attached to ubiquitin. Similarly, reconstituted TEV enzyme can
directly cleave and free a reporter protein. However, the strategy is still very limited
temporally.
One of the methods derived from the idea of the yeast two-hybrid technology is
the protein-fragment complementation assay (PCA) (10). In PCA, protein-protein
interactions are measured by fusing each interacting protein of interest to a split fragment
of a reporter protein that has been rationally dissected into two fragments using protein-
engineering strategies. An improvement over yeast two-hybrid, PCA offers the direct
reconstitution of a reporter protein and avoids the long time lag from transcription
activation of the reporter gene to obtaining enough expressed protein to have a
measurable signal, thus allowing for faster detection of the dynamics of protein-protein
interaction. Split-ubiquitin is one of the first and the smallest split-protein reporters. Split-
ubiquitin offers multiple readout options, such as change in molecular weight, subcellular
localization, and protein stability. Basically any protein that changes a measureable
quality after being released from the reassembled ubiquitin can be utilized as a reporter
for the system. Other notable split-protein pairs often used include those derived from
GFP and its variants (11), dihydrofolate reductase (DHFR) (12), j-lactamase (13), and
luciferases (14). However, for the GFP-derived split-protein pairs, the irreversible nature
of the reconstituted fluorescent proteins limits their ability to report the temporal
information of the specific protein interacting pair under study.
Imaging based microscopy methods have also been applied to study protein-
protein interaction. In fluorescence resonance energy transfer (FRET), classical energy
donor and acceptor proteins, such as CFP and YFP, are fused to interacting proteins,
respectively. However, FRET experiments are only effective in assessing protein-protein
proximity due to the resolution limit and are unable to provide concrete evidence of
physical interaction. Moreover, design of protein fusions often require structural
knowledge of the protein interacting pairs under study, thus it is rather limited to well-
known, or highly proposed interacting partners. Without structural knowledge, one could
123
unknowingly block the interaction sites with protein fusions, leading to false negatives.
Additionally, the energy donor and acceptor proteins must be free of intrinsic
dimerization tendency, which would cause false positive results. The method is also
intrinsically restricted to studying only two proteins at a time.
Finally, photoaffinity labeling (PAL) is yet another method that has been used to
determine the position of a protein relative to those of other proteins within a
macromolecular complex (15). PAL requires the use of a photoactivatable but otherwise
chemically inert analogue, also called a photocrosslinker that can produce a highly
reactive species once activated by light. The site-specific incorporation of a
photocrosslinker to a protein of interest can pinpoint the site, or even the amino acid in
the vicinity of the interacting proteins. A tremendous advantage of PAL over previously
mentioned yeast two-hybrid, PCA and microscopy-based protein-protein interaction
detection methods is that it can capture interactions with endogenous proteins rather than
recombinant proteins. The proteins do not have to be modified with large split protein or
fluorescent protein tags, and the crosslinked protein complexes are stable due to the
formation of a covalent bond. Photocrosslinkers can be site-specifically introduced to
proteins through labeling of cysteine residues or using the unnatural amino acid (UAA)
mutagenesis technique (16) that has been developed to include several photocrosslinkers,
such as benzophenone, aryl azide, and diazirine as unnatural amino acids (17,18). The
high temporal resolution due to the reactive radical species of light-activated
photocrosslinkers is especially advantageous in capturing transient interactions between
proteins; some of these may be lost or not detectable by traditional co-
immunoprecipitation or split-protein techniques.
However, even though in vivo crosslinking based on the genetic code expansion
of UAA mutagenesis to site-specifically incorporate a photocrosslinker has been widely
used, examples of applying it to map protein interactions inside live mammalian cells are
still scarce (19,20). The prevalence of natural amber codons (18), low suppression
efficiency (21), and generation of truncated byproducts (an average efficiency of 20% of
full-length protein with one unnatural amino acid to truncated protein can be achieved)
(22) make UAA mutagenesis challenging to carry out. We wanted to develop a robust,
simple, and minimally invasive method to site-specifically label proteins with
124
photocrosslinkers inside live cells, so they can become a routine tool for protein-protein
interaction detection.
In this project, we describe the work taken towards the development and
application of a photocrosslinker ligase to be used for protein-protein interaction
detection inside live cells. The photocrosslinker ligase can specifically recognize and
ligate an organic small molecule photocrosslinker onto its protein substrate. We hoped
the method to be minimally invasion to endogenous proteins, has high spatial and
temporal resolution, and high efficiency of photocrosslinking. Starting from the enzyme
lipoic acid ligase (LplA) that has previously been applied to ligate an aryl azide
photocrosslinker, we first describe our enzyme engineering of LplA to make it accept the
more efficient photocrosslinker benzophenone. We then describe the effort to
demonstrate LplA ligation of benzophenone on purified proteins, in cell lysate and in live
cells. Finally, knowing that benzophenone ligation works inside live cells, we worked
toward applying benzophenone photocrosslinking to detect protein-protein interactions in
live cells.
125
Classes of photocrosslinkers
Detection of protein-protein interaction using a crosslinking probe has been used
to map protein networks over the years. The light-activated crosslinking probe, often
referred to as a photocrosslinker or a photoaffinity probe, must satisfy several
requirements for it to function ideally in a live cell environment. First of all, the
inactivated form of the photocrosslinker must be bioorthogonal to all small molecules and
functional groups inside the cell. It must remain stable in the reducing environment of the
cytosol and the oxidizing environment in cellular organelles such as the ER. The
photocrosslinker must not be metabolized or degraded by cellular enzymes. Furthermore,
the UV light activated radical of the photocrosslinker should not react or be quenched by
the cellular medium but has high specificity towards its nearby proteins. The desirable
half life of the reactive species should be long enough to give it sufficient time to
crosslink with other proteins, but not too long that it crosslinks promiscuously. Long
excitation wavelength is especially preferred for cellular applications to minimize the
damage caused by high energy UV light. Three main types of photocrosslinkers are
commonly used for photoaffmity labeling experiments: aryl azide, diazirines, and
benzophenone.
3
F
F 250-350 nm he N2
250-350 nm hu N2
F
F N
R, R3 1R H R1 H R2 R4
R2HR R4  /C%
F R1' 4 F R2 R R R3
F N F R, R 1  R4 2 F R2 RA H Z R/-R
NIcR1 R4  R 1,R2  (F F H AF H><.. R2
F ,N , F."CF3 H N CL3
AF R2 \ 'CF3
F
Aryl Azide Diazirine
R1, H
350-365 nm hv R2 'H -O CH2
Benzophenone
126
Figure 4-1. Three classes of photocrosslinkers. The structures shown at the top are the
photochemical precursors, followed by the reactive species after UV irradiation. The
nitrene radical and the carbene radical can insert into C-H bonds, heteroatom-H bonds
(i.e. N-H bond), and alkenes. Benzophenone is very selective towards C-H bond. Note
that formation of the benzophenone ketyl diradical is reversible.
Aryl azides are most efficiently activated by UV light within 250-350 nm
wavelengths depending on the substitution on the phenyl ring (Figure 4-1). The reactive
form of aryl azide is produced through a concomitant loss of N2 and nitrene formation
(23-25). Nitrene can insert into neighboring C-H bonds, heteroatom-H bonds, and form
covalent bonds through adding to alkene functional groups. However, if the reactive
nitrene did not find a suitable partner during its typical 100 ps life time, it will convert to
the less reactive but more stable 7-membered ring ketenimine, which cannot react with
proteins but only with nucleophiles (25,26). The rearrangement of the unstable nitrene to
ketenimine significantly decreases photocrosslinking yields. For this reason, perfluoro
aryl azides are often favored over nonfluorinated aryl azides because fluorine substitution
slows down the rate of ring expansion of the perfluoro aryl nitrene to ketenimine. Aryl
azide has also been utilized to study the interaction between RNA-dependent protein
kinase (PKR) and RNA molecules by site-specifically introducing the photocrosslinker to
various positions of PKR via cysteine-maleimide chemistry (27). Despite its low protein-
protein crosslinking efficiency, aryl azide can serve as a functional handle to react with
triphenyl phosphine in Staudinger ligation or with an alkyne in copper-catalyzed or
strain-promoted click chemistry (28). This chemoselective reaction serves as an ancillary
benefit of aryl azide that allows its incorporation into cellular systems to be monitored. In
our previous publication using perfluoro aryl azide to photocrosslink the FK506 binding
protein (FKBP) and the FKBP-rapamycin binding domain (FRB) (29), Dr. Hemanta
Baruah used the reaction between aryl azide and phosphine as readout of the site-specific
incorporation of the probe.
Another class of photocrosslinker is the diazirine (Figure 4-1). Diazirines form a
reactive carbene intermediate with concomitant loss of N2 upon 350 nm irradiation
(25,30). The small size of diazirine groups introduces minimal perturbation to the protein
127
under study. Its longer wavelength of activation light reduces the damage to the targeted
biological system. Carbenes generated from diazirines are highly reactive and can react
with nearby C-H or heteroatom-H bonds (28) (Figure 4-1). However, the striking reactive
carbene intermediate has a very short half-life in the nanoseconds range (31) and can
react rapidly with water, which decreases its photocrosslinking efficiency (23). Another
significant drawback of the diazirine based photocrosslinkers is that they can undergo
internal rearrangement to form diazo isomers that are long lived electrophiles and can
cause unwanted background reactions with nucleophiles (30). For this reason,
trifluoromethyl diazirines, whose diazo isomers are much less reactive, are preferred over
alkyl diazirines. Photoaffinity labeling using diazirines ranges from probing receptor and
ligand interaction, identification of drug targets, and protein-protein interaction detection
(32). In studying calcium signaling in muscles, diazirine-based dantrolene, which is a
muscle relaxant, crosslinked to a protein proposed to be an inhibitor of calcium release
(33). Diazirines have also been utilized in studying glycoproteins and carbohydrate-
mediated protein interactions, which are often poorly understood due to their
heterogenous nature (34). Diazirines are of particular interest in this application over the
other photocrosslinking probes because of their small size. Specifically, it has been
introduced into 0-linked p-N-acetylglucosamine (0-GlcNAc) by posttranslational
modification (35). This unnatural sugar was incorporated by a mutant UDP-GlcNAc
pyrophosphorylase onto O-GlcNAcylated FG-repeat nucleoporins, which was shown to
crosslink with nuclear transport factors (35).
Benzophenone is yet another class of photocrosslinker. Benzophenone is activated
by long wavelength excitation light in the range of 350 -365 nm (Figure 4-1) (36). The
activated form of benzophenone, a ketyl diradical, is stable in the biological environment
and can persist for as long as 120 pts. The excitation in the n-72* band of the a, p-
unsaturated ketone produces a diradical triplet state which makes the radical an efficient
hydrogen abstractor. Benzophenone reacts primarily and selectively with nearby C-H
bonds in preference to the stronger O-H bonds of the solvent (28,36). If an interaction
partner is not present, the ketyl diradical relaxes back to the ground state and can be
activated again. The ability to be repeatedly excited, together with long life time of the
radical and low background reaction with solvent, leads to benzophenone's relatively
128
high photocrosslinking efficiency compare to those of aryl azide and diazirine, which
rearranges to ketenimine and diazo isomer, respectively, or get quenched by water.
However, benzophenone exhibits a high geometric constraint with regard to the angle and
distance between the ketyl diradical and the nearby C-H bond. Recent work by Pavlova
and coworkers (37) showed benzophenone (Bpa) introduced by UAA mutagenesis to
various positions of an E. co/i serine protease autotransporter, EspP, crosslinked to amino
acids within 4 A of the polypeptide backbone. The Yokoyama group introduced Bpa at
Arg85 in gankyrin, a liver oncoprotein, through UAA mutagenesis approach (38). Crystal
structure of photocrosslinked complex Bpa85 gankyrin and its interacting partner C-
terminal domain of S6 proteasomal protein (S6C) revealed that the carbonyl group of
Bpa85 crosslinked exclusively to the Cy atom of Glu356 in S6C. In the 16 Bpa mutants
of gankyrin studied, the photocrosslinked products showed benzophenone exhibited
preferences for narrow range of distance from 2.36 A to 4.11 A and an angle between
104.10' and 128.29'. In addition, several groups (39-43) have discovered that
benzophenone appeared to preferentially crosslink with nearby methionine residues,
termed the "methionine magnet effect" (42). Despite its strict geometry requirement,
benzophenone has been incorporated through UAA mutagenesis (16,18) into many
proteins such as adaptor protein Grb2 (19) in live mammalian cells and glutathione S-
transferase in E. coli (44).
Our lab has previously demonstrated LplA mutant-based aryl azide labeling and
crosslinking in mammalian cell lysate and on the cell surface of live cells (29). Based on
the same method, we wanted to develop an LplA ligase that can bind and ligate
benzophenone, which has longer radical lifetime and lower nonspecific reaction with
solvent than aryl azide. More importantly, benzophenone can be repeatedly activated to
increase photocrosslinking efficiency. We hoped that the better photophysical properties
of the benzophenone photocrosslinker would allow us to detect endogenous protein
interacting partners inside live mammalian cells that we were not able to achieved
previously with aryl azide. We wished to add the LplA-mediated photocrosslinker
ligation methodology to the toolbox of protein-protein interaction mapping techniques
and extend this photocrosslinking method to inside live mammalian cells.
129
Screening for a benzophenone ligase
As we discussed in Chapter 2, our lab has engineered the enzyme lipoic acid
ligase (LplA) (45-48) to incorporate many unnatural small molecule fluorophores (49-52)
and a photocrosslinker (29). Dr. Hemanta Baruah first laid the groundwork for
engineering LplA towards accepting other organic small molecule substrates (29). Based
on the crystal structure of LplA bound with lipoic acid (47), Dr. Baruah performed
alanine scanning mutagenesis of amino acid residues that are within 7.5 A of the
dithiolane ring of lipoic acid and identified the Trp37 position to be critical for the LplA
mutant to accept a small aryl azide molecule. Further mutagenesis studies discovered that
LplA(W37V) was able to accept an aryl azide carboxylic acid with a four methylene
linker. To see if an unnatural probe as large as benzophenone could possibly fit in the
lipoic acid binding pocket of LplA, we compared crystal structures of LplA with lipoic
acid bound (Figure 4-2, left, PDB 3A7R) and LplA with a benzophenone probe docked
into the active site of the enzyme (Figure 4-2, right). The Trp37 residue is shown in blue.
We can see that one of the phenyl rings of benzophenone clashes into the Trp37 residue.
E. coli lipoic acid ligase with E. coli lipoic acid ligase with
bound lipoic acid docked benzophenone probe
Figure 4-2. Lipoic acid ligase (LplA) with (left) lipoic acid bound (PDB 3A7R) and (right)
a benzophenone unnatural small molecule docked into the enzyme active site. The
"gatekeeper" residue Trp37 is rendered in blue.
130
Even though we have seen engineered LplA being able to recognize small molecule
substrates that bear little resemblance to its natural substrate, lipoic acid, we were not
certain if simply mutating the Trp37 "gatekeeper" residue in the binding pocket of LplA
would be sufficient to make LplA bind to benzophenone, which is larger than previous
small molecule substrates of LplA. Nonetheless we decided to begin with a panel of
Trp37 LplA mutant enzymes to test their ability to ligate one of two benzophenone
carboxylic acid probes varying in linker lengths onto LAP peptide in an in vitro HPLC
assay. We picked six amino acid residues with smaller side chains than Trp - Gly, Ala,
Val, Ile, Leu, Ser. We chose mainly hydrophobic amino acids to maintain a hydrophobic
binding pocket of LplA, which is suitable for benzophenone.
0 0 0 0
NHS, DCC 0 H2 ______ H 0
a OH CH 3CN, RT N CH 3CN, TEA, RT a , N(4-.KOH
0 0 n-/
0 benzophenone (n=3,4)
Figure 4-3. Synthetic scheme of benzophenone photocrosslinkers. NHS, N-
hydroxysuccinimide; DCC, N,N'-dicyclohexylcarbodiimide; RT, room temperature; TEA,
triethylamine.
Benzophenone probes were synthesized by first activating 4-benzoylbenzoic acid
to the corresponding N-hydroxysuccinimide (NHS) ester via N,N'-
dicyclohexylcarbodiimide (DCC) activation. 4-Benzoylbenzoic-NHS ester was then
reacted with either 4-aminobutyric acid or 5-aminovaleric acid, through formation of an
amide bond (Figure 4-3, Experimental Methods). Each probe was characterized by 1H
NMR and mass spectrometry. We screened six Trp37 LplA mutants against these two
benzophenone substrates using an HPLC assay based on a shift in peak retention time to
determine the conversion of the free 13-amino acid LplA Acceptor Peptide (LAP) (53)
into LAP-benzophenone conjugate (Figure 4-4). Interestingly, only benzophenone with
the four methylene linker (n=4) was able to get incorporated onto LAP by three LplA
mutants. LplA(W37G) appeared to be the most efficient benzophenone ligase.
131
Subsequent benzophenone crosslinking experiments were performed with benzophenone
with n=4 alkyl linker and LplA(W37G).
0
O 0
I H- N
0
LpA variant
Benzophenone WT W37G W37A W37V W371 W37L W37S
n=3 N.D. N.D. N.D. N.D. N.D. N.D. N.D.
n=4 N.D. 100% N.D. 65% N.D. 73% N.D.
Figure 4-4. Screening to identify LplA mutant/benzophenone substrate pair. The table
shows percent conversions of LAP to LAP-benzophenone product conjugate. Wild-type
LplA and six W37 point mutants were screened against two benzophenone substrates of
different lengths (structures shown on left). N.D. indicates that product was not detected.
Screening was performed with 500 nM LplA enzyme, 100 [tM LAP peptide, and 300 gM
benzophenone substrate for 40 min at 30 'C.
132
Detection of ligated benzophenone via reaction with fluorophore-hydroxylamine
Unlike the aryl azide photocrosslinker that has the azide moiety to act like a
functional group handle to react with triphenyl phosphine in the Staudinger ligation or
alkyne in click chemistry to monitor its incorporation, benzophenone does not have this
ancillary benefit. For live cell application of benzophenone ligation, we needed an assay
to assess the probe's ligation efficiency and specificity. The only available functional
handle on benzophenone is the carbonyl group. Methods to detect aldehyde and ketone
functional groups in proteins were first explored as a way to monitor protein oxidation by
reactive oxygen species (54). Additionally, lipid oxidation and non-enzymatic glycation
can produce protein-bound aldehydes and carbonyl functionalities. Hydrazide was
explored as an aldehyde- and ketone-reactive probe (55,56). The reaction relies upon the
condensation between the carbonyl group and the hydrazide to yield a stable hydrazone
product. The hydrazide can be functionalized with either biotin or a fluorophore as
readout. Following the introduction of keto amino acid via the UAA mutagenesis
approach, the incorporation and site-specificity ofp-acetylphenylalanine into mammalian
GPCRs (57) and bacteriophage T4 lysozyme (58) have been monitored using hydrazide-
fluorescein conjugate under mild acidic condition. The p-acetylphenylalanine
incorporated protein can be separated by SDS-PAGE and visualized under a fluorescence
imager.
Even though the condensation reaction with hydrazide has been successfully
implemented to detect aldehyde and acetophenone, it was a challenge in the case of
benzophenone because of its very deactivated and electron-rich carbonyl group.
Regardless, we decided to try reacting benzophenone-ligated E2p (9 kDa) with biotin-
hydrazide under acidic condition (Figure 4-5). Following in vitro benzophenone ligation
onto E2p by LplA(W37G) (see Experimental Methods), the samples were dialyzed in
phosphate buffered saline (PBS) to remove excess benzophenone probe. Biotin-hydrazide
was added and the pH of the reaction was adjusted to 4.5. After 12 hours incubation with
biotin-hydrazide, reactions were separated on an SDS-PAGE gel followed by Western
blot transfer and detection with streptavidin-HRP. However, streptavidin-HRP stained
both LplA(W37G) and E2p proteins and no benzophenone-specific signal was observed
(Figure 4-5). Several explanations could be possible for the lack of benzophenone-
133
specific signal. Biotin-hydrazide could nonspecifically stick to proteins. It is also possible
that the reaction product hydrazone was hydrolyzed during subsequent analysis.
0
(NH 4
4 LpIA(W37G)
benzophenone + +
LpIA(W37G) + +
ATP +
biotin-hydrazide
pH 4.5, 12 hrs
70
50
40
30- LpIA(W37G)
20
15
10
+- EUp
5
kDa
Coomassie Westem Blot
SA-HRP
Figure 4-5. Unsuccessful attempt to detect in vitro ligation of benzophenone to E2p
protein using biotin-hydrazide. Benzophenone ligation was performed using LplA(W37G)
enzyme on purified E2p protein. Subsequently the mixture was combined with 200 pM of
biotin-hydrazide and the pH of the reaction was adjusted to 4.5 to detect the
benzophenone-E2p adduct via hydrazone formation. Blotting with streptavidin
conjugated HRP (SA-HRP) is shown to the right of the Coomassie-stained gel. Negative
control is shown with ATP omitted in the in vitro reaction (lane 2). Bands of equal
134
intensity are seen in lanes 1 and 2 in the SA-HRP blot, suggestive of non-specific binding
rather than specific signal due to successful ligation and detection of benzophenone.
In fact, studies comparing the hydrolytic stabilities of hydrazones and oximes
have shown that hydrazones are more prone to hydrolysis because of the higher
likelihood of protonation on the nitrogen atom attached to the carbon, whose
electrophilicity is enhanced as a result (59). Oximes, on the other hand, are more stable to
hydrolysis than hydrazones because of the inductive effect and the higher
electronegativity of the oxygen atom attached to the reactive nitrogen atom, which is in
turn made less likely to be protonated. Thus, the carbonyl carbon atom in oximes is less
electrophilic than the carbonyl carbon in hydrazones (Figure 4-6). Comparison of first-
order rate of hydrolysis between hydrazone and oxime shows that the half-life of methyl
oxime is 25 days (600 hours) while for methyl hydrazone, it is only 2 hours (59). Because
the ultimate application is to ligate benzophenone in live mammalian cells, where the cell
medium is hydrophilic and contains nucleophiles that could possibly react with the
hydrazone product, we decided to react benzophenone with hydroxylamine instead of
hydrazide to form a more stable oxime.
H20
H+ H+ OH
R1N 'R 2  -'2 1 R2
H H
X = NR3 hydrazone
X=O oxine
Figure 4-6. Hydrolysis of hydrazones and oximes. Hydrazones are more likely to be
hydrolyzed than oximes because its carbon atom is made more electrophilic by the higher
likelihood of protonation on the attached nitrogen (59).
Chemoselective reaction between keto groups and hydroxylamines has been
demonstrated in detecting a benzaldehyde-containing sialic acid analog that is
incorporated into cell-surface glycans through metabolic labeling (60). Furthermore, this
reaction has also been directly compared with the reaction between keto groups and
hydrazide in monitoring site-specific incorporation of p-acetylphenylalanine into
135
bacteriophage T4 lysozyme (58). The Deniz group showed that hydroxylamine-Alexa
Fluor 488 conjugate reacted with the acetophenone amino acid more effectively than
hydrazide-Alexa Fluor 488 conjugate, providing on average greater than 90% labeled
protein under optimized conditions. Hydroxylamine-Alexa Fluor 488 demonstrated more
pH flexibility, being able to detect T4 lysozyme from pH range 4 to 9, while hydrazide-
Alexa Fluor 488 only showed activity at pH 4 and 5. Reaction with hydroxylamine also
appeared to be more sensitive, detecting as low as 1.5 pM of protein, while hydrazide
only showed reaction at 100 sM protein concentration (58).
To test whether hydroxylamine derivatives can be used as a reagent to detect
benzophenone, we first tested the reaction between benzophenone and biotin-
hydroxylamine to see if they can form the oxime product (Figure 4-7). We dissolved 5
gM benzophenone and 5 jM biotin-hydroxylamine in 200 pl phosphate buffered saline
(PBS) and adjusted the pH of the reaction to 5. We chose PBS as the reaction medium
because it is compatible with later protein ligation and photocrosslinking experiments.
The reaction was incubated at room temperature for 12 hours. We were able to observe
the correct molecular weight of the oxime product by mass spectrometry at [M+H]*
639.12 (expected molecular weight: 638.25), which indicated that the reaction between
benzophenone and biotin-hydroxylamine can successfully occur.
136
00 H I +
HO)$+ N ~-. ~ . F
40
0
HN 'JNH
N N
IH pH 5 HOH 0  s
OHOAR N 'Ir40
Exact Mass: 638.25
[M+H]+
639.1200
m/z, amu
Figure 4-7. Reaction between benzophenone and biotin-hydroxylamine to form oxime
product. (Top) Scheme of chemical reaction. Each reactant (5 p.M) was combined in
phosphate buffered saline (PBS) at pH 5. Reaction was incubated at room temperature for
12 hours. (Bottom) Mass spectrometry analysis of oxime product, which was observed at
molecular weight [M+H]* 629.1200.
Subsequently, we tried to detect in vitro ligation of benzophenone to E2p protein
using hydroxylamine-Alexa Fluor 488 conjugate. After in vitro benzophenone ligation
onto E2p protein using purified enzyme LplA(W37G) (see Experimental Methods),
excess benzophenone probe was removed by dialyzing the reaction in PBS.
Hydroxylamine-Alexa Fluor 488 was added to the reaction whose pH was adjusted to 5.
137
After incubating at room temperature for 12 hours, the samples were separated on an
SDS-PAGE gel and analyzed on a fluorescence gel imager (Figure 4-8). A positive
control reaction using a ketone small molecule, 7-oxooctanoic acid that can be
incorporated by wild-type LplA onto E2p was performed side by side with benzophenone
ligation and reaction with hydroxylamine. 7-Oxooctanoic acid was chosen as a positive
control because it is a more conventional alkyl ketone whose carbonyl group is
electrophilic and not deactivated as in benzophenone. Therefore, 7-oxooctanoic acid
should react readily with hydroxylamine. As shown in Figure 4-8, strong in-gel
fluorescence of Alexa Fluor 488 was seen specifically at the position of E2p only when
benzophenone was ligated (lane 4). The negative control reaction using wild-type LpIA,
which does not accept benzophenone did not show any signal (lane 3). Compared to the
intensity of the Alexa Fluor 488 in-gel fluorescence on E2p ligated with 7-oxooctanoic
acid, E2p-benzophenone conjugate showed weaker fluorescence retention (Figure 4-8,
lane 2 and 4). The comparison of the reactivity of hydroxylamine towards the two ketone
structures indicated that the benzophenone did react with hydroxylamine to the same
extent; the reactivity was much weaker compared to a conventional alkyl ketone without
a deactivated carbonyl center.
138
0
O. N
NH4 1H
4pA 3
LpIA(W37G)
7-oxooctanoic acid +
0
SO, 503
H2N 0 NH
H H CO
N N - CO
AF 488-
hydroxylamine
pH 5, 12 hrs
+
benzophenone - + +
LpIA(W37G) - - - +
Wild-type LpIA + + + -
ATP - + + +
1 2 3 4
LpIA enzyme --
E2p -
Coomassie Alexa Fluor 488 fluorescence
Figure 4-8. Detection of in vitro ligated benzophenone using AF488-hydroxylamine. As
shown in the top scheme, benzophenone ligation performed using LplA(W37G) enzyme
on purified E2p protein was reacted with AF488-hydroxylamine at pH 5 to detect ligated
benzophenone. AF488 in-gel fluorescence is shown to the right of the Coomassie-stained
gel. Positive control reaction with wild-type LplA ligated 7-oxooctanoic acid onto E2p
was performed in parallel (lane 2). Negative controls are shown with ATP omitted (lane 1)
and wild-type LplA in place of LplA(W37G) (lane 3).
139
1 2 3 4
Following the success of detecting benzophenone ligation onto E2p with
hydroxylamine, we tested the reaction in the more complex environment of HEK 293T
cell lysate expressing LAP-tagged blue fluorescent protein (BFP). We performed in vitro
benzophenone ligation with purified enzyme LplA(W37G) in cell lysate, followed by
dialysis to remove excess benzophenone probe and incubated the cell lysate with Alexa
Fluor 488-hydroxylamine at pH 5. However, Alexa Fluor 488-hydroxylamine stuck
heavily to material in the lysate, causing high background. We were not able to see any
specific reaction with benzophenone-ligated LAP-BFP above this high background.
There are two possible causes of backgrounds: simply hydrophobic sticking of the dye, or
non-specific reactions with endogenous ketones and aldehydes in the cell lysate. To test
the second hypothesis, we first treated cell lysate with biotin-hydroxylamine to block
endogenous ketones and aldehydes, so they would not react with Alexa Fluor 488-
hydroxylamine to cause background signal. We then treated the cell lysate with
benzophenone and LplA(W37G), followed by reaction with Alexa Fluor 488-
hydroxylamine in a similar fashion as before. However, there was no significant
reduction in background (data not shown).
In order to reduce nonspecific background, we tried a similar experiment using
biotin-hydroxylamine to react with benzophenone with the hope that the more
hydrophilic biotin-hydroxylamine would stick less to proteins compared to Alexa Fluor
488-hydroxylamine. In this experiment, we first activated 7-oxooctanoic acid or
benzophenone acid to its corresponding N-hydroxysuccinimide (NHS) ester, then
chemically conjugated each probe to a purified fusion protein, FRB-YFP, to maximize
signal. 7-Oxooctanoic acid or benzophenone conjugated FRB-YFP protein was incubated
with biotin-hydroxylamine at pH 5 for 12 hours. Samples were separated on a 12% SDS-
PAGE gel and analyzed by Western blot with streptavidin (SA)-HRP staining to detect
proteins derivatized with biotin-hydroxylamine. Figure 4-9 shows strong SA-HRP signal
on 7-oxooctanoic conjugated FRB-YFP (lane 3), indicating efficient reaction between 7-
oxooctanoic acid and biotin-hydroxylamine. Benzophenone conjugated FRB-YFP shows
very minimal signal in comparison (lane 5). We subsequently tried to react biotin-
hydroxylamine in HEK 293T cell lysate containing benzophenone-ligated LAP protein.
However, we were unable to detect any specific signal (data not shown).
140
FRB-YFP-7-oxo - + + - -
FRB-YFP- 
- - + +
benzophenone
biotin-hydroxylamine - - + - +
1 2 3 4 5 1 2 3 4 5
72_
55 M52*
43
34
26
17
kDa
YFP in-gel WB SA-HRP
fluorescence
Figure 4-9. Detection of benzophenone chemically conjugated to FRB-YFP using biotin-
hydroxylamine. 7-Oxooctanoic acid or benzophenone conjugated FRB-YFP protein was
incubated with biotin-hydroxylamine at pH 5 to detect 7-oxooctanoic acid or
benzophenone. YFP in-gel fluorescence detects FRB-YFP. Immunoblot with streptavidin
(SA)-HRP detects biotinylated proteins. 7-Oxooctanoic acid conjugated FRB-YFP (lane
3) and benzophenone conjugated FRB-YFP (lane 5) both reacted with biotin-
hydroxylamine. Negative controls are shown with only FRB-YFP protein (lane 1), or
biotin-hydroxylamine omitted (lanes 2 and 4).
Although hydroxylamine was able to react with benzophenone as shown above
(Figure 4-8, lane 4; Figure 4-9, lane 5) using in vitro purified E2p ligated with
benzophenone, or benzophenone conjugated FRB-YFP protein, this detection method is
not sensitive enough to react with benzophenone-ligated protein in cell lysates.
Furthermore, reaction with hydroxylamine would only allow us to assay for whether
benzophenone ligation has occurred, but is sunable to provide any information regarding
the ligation yield. It is critical for us to have a method that enables quantitative
measurement of the benzophenone ligation yield inside live cells to ensure the ligation is
complete before proceeding to the photocrosslinking reaction. We later describe the
development of a native gel shift assay to determine benzophenone ligation yield in
section titled "Quantify benzophenone ligation yield inside live mammalian cells".
141
FRB-YFP +
In vitro photocrosslinking of LAP-tagged FKBP and FRB proteins using
benzophenone probe
Knowing that LplA(W37G) can ligate benzophenone probe onto LAP substrate in
vitro, we wanted to test photocrosslinking using a protein-protein interaction model
system as a proof-of-principle experiment. The system we chose is the ternary complex
consisting of FK506-binding protein (FKBP12 or simply FKBP), rapamycin, and the
FKBP12-rapamycin binding domain of mTOR (FRB). The interaction between FKBP
and FRB has been well characterized (61-63). FKBP is known to form a heteromeric
protein complex with FRB only in the presence of the immunosuppressive drug
rapamycin. Rapamycin binds with high affinity (Kd = 0.2 nM) to FKBP (61). The FKBP-
rapamycin complex in turn binds tightly to FRB. Previously, Dr. Hemanta Baruah
demonstrated LplA-mediated aryl azide ligation onto LAP-tagged FKBP in mammalian
cell lysate (29). Aryl azide-ligated FKBP-LAP was then photocrosslinked to its
interaction partner FRB in the presence of rapamycin in cell lysate. We would like to test
benzophenone photocrosslinking with the same interaction system as a proof-of-principle
experiment and hopefully obtain higher photocrosslinking efficiency.
FKBP
FRB C-terminus
C-terminus
FRB
C-termin s
FKBP
N-terminus
Figure 4-10. Crystal structure of FKBP-rapamycin-FRB ternary complex (PDB 2AFP)
with termini of each protein indicated.
142
Upon examination of the crystal structure of FKBP-rapamycin-FRB ternary
complex (Figure 4-10), we saw that both C-terminus of FRB and the C-terminus of
FKBP are located relatively close to the interface of the complex. We decided to try
FKBP with LAP fused to its C-terminus first because it has been demonstrated previously
in photocrosslinking using aryl azide (29). We purified myc-tagged FKBP-LAP and HA-
tagged FRB-YFP fusion proteins from E. coli. FKBP-LAP was then incubated with
purified enzyme LplA(W37G), ATP, Mg 2 , and benzophenone probe to ligate
benzophenone onto FKBP-LAP. Reactions were run on a reverse-phase HPLC column to
confirm complete conversion of FKBP-LAP to benzophenone-ligated FKBP-LAP by the
shift in peak retention time (data not shown). Excess benzophenone probe was removed
through extensive dialysis. Purified FRB-YFP protein and rapamycin were then added to
the reaction mixture. The complex was allowed to incubate with rotation at room
temperature for 1 hour. Afterwards, the reaction was exposed to 365 nm UV light.
In our preliminary experiments, we experimented with three photocrosslinking
instruments to optimize photocrosslinking conditions such as UV exposure time, distance
from UV lamp, temperature, and ease of use. First, Porta-Ray 400R from UVitron is
equipped with 400-Watt UVA enhanced arc lamp with intensity output concentrated from
350 to 400 nm wavelength. Its curing area is about 5 X 3 inches, which is ideal for the
size of a typical mammalian cell culture plate. The cells can be placed anywhere from 1
inch to 3 inches from the UV lamp. Second, handheld long wavelength UV lamp from
UVP has 4-Watt power and can be placed directly on top of cell culture plate. Even
though the power of the handheld UV lamp is low, the distance between the UV light and
cells can be as short as 1 centimeter. The third instrument is the Stratalinker from
Stratagene, which has 5 lamps with 8-Watt output per lamp. The sample can be placed
anywhere from 1 inch to 10 inches from the lamp. Since all the instruments differ in UV
lamp power, distance between cells and UV light, and the effective curing area, we tried
all three instruments with multiple UV irradiation conditions for each instrument. The
most optimized condition for each instrument is listed in Table 4-1. All three instruments
performed similarly under their optimized conditions.
143
Instrument Wavelength Power output Distance from Time of UV
samples irradiation
Porta-Ray 400R 350 - 400 nm 400-Watt 3 inches 2 minutes
Handheld long 365 nm 4-Watt 1 cm 15 minutes
wavelength UV lamp , I I I _I
Stratalinker 365 nm 5 lamps, 8-Watt/lamp 5 inches 30 minutes
Table 4-1. Optimized UV photocrosslinking conditions for each instrument tested.
We first ligated benzophenone onto FKBP-LAP using purified LplA(W37G)
protein and ATP (see Experimental Methods). After ligation was complete, we dialyzed
the samples to remove excess unligated benzophenone. Dialyzed samples were incubated
with purified FRB-YFP protein and rapamycin for 1 hour at room temperature. Using the
Stratalinker instrument, a 96-well cell culture plate containing the in vitro ligated samples
was placed on top of an ice tray that was put at 5 inches below the UV lamps. After UV
irradiation for the indicated duration from 2 to 60 minutes (Figure 4-11, lanes 1-4),
samples were separated on a 12% SDS-PAGE gel. To detect photocrosslinked products,
we utilized YFP in-gel fluorescence to detect the YFP-tagged FRB. Many similar
experiments were performed with varying UV irradiation time from 2 minutes to 15
minutes depending on the power of the instrument, and the distance between sample and
UV lamps. Figure 4-11 shows a representative result from in vitro photocrosslinking
using purified FKBP-LAP and FRB-YFP proteins. We detected an YFP-positive band
(blue arrow) at 60 kDa, corresponding to the combined molecular weights of the two
interaction partners, HA-FRB-YFP (41.2 kDa) and myc-FKBP-LAP (17.1 kDa). The
intensity of the band increased with UV irradiation time (Figure 4-11, lanes 1-4).
Negative control reactions with either UV irradiation omitted (Figure 4-11, lanes 5),
benzophenone probe omitted (lane 7), ATP omitted (lane 8), or FKBP-LAP omitted (lane
9) did not show the photocrosslinked heterodimer product.
144
However, the crosslinked product did not depend on the presence of rapamycin
(Figure 4-11, lanes 6). One explanation for this could be that the in vitro concentrations
of the two interacting proteins and the small molecule rapamycin can be difficult to
control precisely, especially when the initial concentration of benzophenone-ligated
FKBP-LAP can be slightly changed after dialysis. During in vitro photocrosslinking, the
chance of benzophenone-ligated FKBP-LAP encountering FRB in the absence of
rapamycin was high because they were the most concentrated proteins present in the test
tube. Later on, we tested lower concentrations of each binding partner and observed that
at lower concentrations, the chance of getting rapamycin-independent background
photocrosslinking was much lower (see Figure 4-20 in later section). We also observed
another YFP-positive band whose molecular weight corresponds to a homodimer product
(82.4 kDa) of FRB-YFP (Figure 4-11). We speculated that this could be caused by the
intrinsic dimerization tendency of YFP (39) in combination with UV exposure. We later
tested photocrosslinking using HA-tagged FRB without YFP fusion to eliminate this
problem.
UV irradiation + + + + - + + + +
rapamycin + + + + + - + + +
benzophenone + + + + + + - + +
ATP + + + + + + + - +
FKBP-LAP + + + + + + + + -
UV irradiation 2 15 30 60 NA 30 30 30 30time (min.)
1 2 3 4 5 6 7 8 9
95
72 FRB-YFP homodimer
55 FKBP-LAP/FRB-YFP
heterodimer
43 FRB-YFP
34
26
17
kDa
YFP fluorescence
145
Figure 4-11. In vitro photocrosslinking with purified proteins FKBP-LAP and FRB-YFP.
Benzophenone-ligated FKBP-LAP protein was incubated with FRB-YFP and rapamycin
followed by UV irradiation in Stratalinker instrument. Samples were separated on an
SDS-PAGE gel. YFP in-gel fluorescence was detected using a chemiluminescence
imager. FRB-YFP (41.2 kDa), FKBP-LAP (17.1 kDa). Photocrosslinked product of the
two proteins is indicated by blue arrow.
146
Optimizations to improve in vitro photocrosslinking yield
Since the photocrosslinking yield of FKBP-LAP and FRB-YFP was very low
despite trying different UV instruments, varying UV irradiation times, and optimizing the
distance between sample and UV light, we thought perhaps something else instead of
photocrosslinking condition was causing low yield. We tried several things to
troubleshoot the low yield of photocrosslinked product.
1. Methionine to Alanine mutation on FKBP-LAP
Benzophenone is highly sensitive to its surrounding amino acids and requires
specific geometric configurations of the nearby C-H bonds to react (36,38). Furthermore,
several studies have shown that photocrosslinking experiments with benzophenone
continued to identify methionine as the crosslinked residue at the site of interaction (39-
43). Benzophenone could perhaps photocrosslink and insert into one of the C-H bonds on
either of the methylene units of methionine. This preference of benzophenone towards
methionine has been termed the "magnet effect" of methionine (42). We suspected that in
our LplA-mediated ligation system, perhaps the methionine residues near the ligation site
of benzophenone in FKBP-LAP protein were reacting with the ligated benzophenone
before it had a chance to photocrosslink to residues of the interacting protein FRB. The
possibility of benzophenone ketyl diradical being "quenched" by nearby methionine
residues could severely handicap or completely prevent the ability of the radical to
interact with other C-H bonds. Crystal structure of the ternary complex of FRB-
rapamycin-FKBP (PDB 2FAP) (61,62) (Figure 4-12) shows a total of three methionines
on FKBP and six methionines on FRB, highlighted as ball-and-sticks. Since LAP peptide
was fused to the C-terminus of FKBP, the two methionines on FKBP that are closest to
benzophenone are Met49 and Met66.
147
FRE
C-termi
Figure 4-12. Crystal structure of FRB-rapamycin-FKBP ternary complex at 2.2 A
resolution (PDB 2FAP) with methionine residues in ball-and-stick. The termini of FRB
and FKBP are indicated by black arrows. Two methionines (Met49, Met66) most
proximal to the C-terminus of FKBP are circled in red. The LAP peptide is fused to the
C-terminus of FKBP.
To test the hypothesis of the "magnet effect" of methionine, we cloned single and
double alanine mutants of Met49 and Met66 in FKBP-LAP. Crystal structures as well as
biophysical analysis of the FKBP-rapamycin-FRB ternary complex show that these two
methionine residues are not involved in the binding of FKBP to rapamycin and there is
little interaction between FKBP and FRB (61-63). Thus these mutations should not affect
ternary complex formation. Three new FKBP-LAP constructs, FKBP(M49A)-LAP,
FKBP(M66A)-LAP, and double mutant FKBP(M49A, M66A)-LAP were cloned,
expressed and purified from E.coli. These proteins were in vitro ligated with
benzophenone probe using purified LplA(W37G) enzyme (see Experimental Methods),
and incubated with FRB-mApple (YFP was changed to mApple to reduce the amount of
homodimer formation caused by YFP as seen previously in Figure 4-11) and rapamycin
to form the ternary complex. However, photocrosslinking experiments using three
148
methionine to alanine mutant proteins of FKBP-LAP did not increase heterodimer
formation as shown in Figure 4-13. The band corresponding to FBKP-FRB heterodimer
product is indicated by a red arrow. Photocrosslinking of FKBP(M66A)-LAP and FRB-
mApple gave the highest heterodimer product signal (Figure 4-13, lane 3).
UV irradiation + + + + + +
FKBP(M49A)-LAP - + - - - - -
FKBP(M66A)-LAP - - + - - - -
FKBP(M49A, +
M66A)-LAP
FKBP-LAP + - - - - + +
benzophenone +
1I
+
2
95-
72-
55-
43 -
34-
26-
17 -
kDa
+ -
3
+
4
+
5 6
+
7
mApple fluorescence
Figure 4-13. In vitro photocrosslinking with methionine to alanine mutant proteins of
FKBP-LAP to avoid methionine quenching of benzophenone ketyl diradical. Lanes 2, 3
and 4 contain three methionine to alanine mutants of FKBP-LAP photocrosslinked to
FRB-mApple. Photocrosslinked FKBP/FRB heterodimer product is indicated by a red
arrow. Compared to the wild-type FKBP-LAP, methionine mutants of FKBP-LAP did
not help to significantly increase crosslinking efficiency with FRB.
2. Removal offluorescent proteins fused to FRB and FKBP
It is conceivable that the YFP or mApple fused to the C-terminus of FRB could
hamper movement of the proteins as they interact with each other upon rapamycin
stimulation, causing inefficient ternary complex formation. Furthermore, YFP caused
149
significant amount of homodimer formation of FRB-YFP, which might have reduced the
effective concentration of FRB-YFP that could interact with rapamycin and FKBP-LAP.
A new FRB construct with HA epitope tag fused to the N-terminus of FRB was cloned,
expressed, and purified from E coli. Similarly, benzophenone-ligated myc-FKBP-LAP
(see Experimental Methods) was incubated with HA-FRB and rapamycin, and
subsequently irradiated with UV light. Samples were separated on a 12% SDS-PAGE gel
and transferred to a nitrocellulose membrane for Western blot analysis. FKBP-LAP was
detected on Western blot with antibody against the myc tag fused to the N-terminus of
FKBP.
In Figure 4-14, photocrosslinked product of benzophenone-ligated myc-FKBP-
LAP and HA-FRB is boxed in red. The crosslinked complex is dependent on UV
irradiation, but shows a weak background under both rapamycin omitted (lane 2) and
benzophenone omitted (lane 4) conditions. We suspected that the rapamycin-independent
background was caused by the high concentration (10 pM) of proteins in the in vitro
reaction mixture. We repeated the same experiment with lower concentrations of FKBP
and FRB. Rapamycin-dependent formation of UV photocrosslinked product was
achieved if concentration of the proteins were decreased to 2 pM (see Figure 4-20 in later
section). The weak background seen in the absence of benzophenone (Figure 4-14, lane 4)
may be caused by crosslinking through Tyr2105 on FRB (PDB 2FAP), which points
directly into the interface of the ternary complex. Furthermore, even though E. coli
purified myc-FKBP-LAP appeared to be pure by Coomassie staining (data not shown),
Western blot staining with anti-myc antibody shows a nonspecific antibody staining
background (Figure 4-14, green arrow) of myc-FKBP-LAP. In addition, myc-FKBP-LAP
appeared to form higher molecular weight aggregates (Figure 4-14, blue arrow) in the
presence of the benzophenone photocrosslinker. To try to remove the anti-myc
nonspecific staining of myc-FKBP-LAP, we performed a second Ni-NTA column
purification of the His6-tagged myc-FKBP-LAP protein, but the background band still
persisted (Figure 4-14, green arrow). For the benzophenone-dependent higher molecular
weight aggregates of myc-FKBP-LAP protein (Figure 4-14, blue arrow), we tried to
remove excess benzophenone through extensive buffer exchange using minispin columns
instead of dialysis, but the background still persisted.
150
rapamycin + - + +
UV irradiation + + - +
benzophenone + + + -
1 2 3 4
100
80 .
60
40 .
30 --
25
kDa
+-myc-FKBP-LAP
WB anti-myc
Figure 4-14. In vitro photocrosslinking of myc-FKBP-LAP (17.1 kDa) and HA-FRB
(14.2 kDa). Fluorescent protein fusions on FRB were replaced by the smaller epitope tag,
human influenza hemagglutinin epitope (HA). Western blot analysis with antibody
against myc tag on FKBP detect photocrosslinked product (boxed in red), but also weak
background crosslinking (green arrow) in the absence of rapamycin (lane 2) and
benzophenone (lane 4). The myc-FKBP-LAP protein also formed higher molecular
weight complexes (blue arrow) in the presence of benzophenone (lanes 1-3).
3. LAP placement on FRB instead of FKBP
Since in vitro photocrosslinking of FRB and LAP-tagged FKBP gave very
minimal photocrosslinked product, we thought perhaps the position of the LAP peptide,
and therefore the placement of benzophenone probe, was not ideal for the
photocrosslinker to react with nearby C-H bonds on FRB. Looking at the crystal structure
of the ternary complex in Figure 4-12, we observed that fusing LAP to the C-terminus of
FRB might bring it closer to the interface of the interacting proteins, therefore providing
benzophenone with more accessibility to crosslink to C-H bonds on FKBP. We switched
the fusion site of LAP from the C-terminus of FKBP to the C-terminus of FRB to test this
151
hypothesis. We further added a 10-amino acid linker (GSGSTSGSGK) between LAP and
FRB to increase the spatial reachability of benzophenone. The new protein pairs we tried
to photocrosslink are FKBP/FRB-LAP and FKBP/FRB-(10 a.a. linker)-LAP. Following
the same protocol as previously described, we first in vitro ligated benzophenone onto
FRB-LAP or FRB-(10 a.a. linker)-LAP protein using purified LplA(W37G) protein, ATP
and Mg 2+(see Experimental Methods). Ligation product was confirmed using the HPLC
assay. We removed unligated benzophenone probe through dialysis. Afterwards, we
added FKBP protein and rapamycin to the reaction mixture to induce ternary complex
formation, then irradiated the samples with 365 nm light. After photocrosslinking,
samples were separated on a 12% SDS-PAGE gel and transferred to a nitrocellulose blot
for immunostaining analysis. However, immunostaining against the myc epitope tag on
FKBP did not show any significant increase of photocrosslinked complex as compared to
previous results (data not shown).
At this stage of the project, we have tried many photocrosslinking conditions to try
to increase the in vitro photocrosslinking yield, but none of the attempts was able to
provide high yield of the FKBP and FRB photocrosslinked product without further
increasing background. We decided to take a closer look at the design of our
benzophenone photocrosslinking probe. In the next section, we describe the synthesis of a
new benzophenone probe and the reason behind changing to a new photocrosslinker.
152
Synthesis and LpJA(W37G) incorporation of alkyl benzophenone
Since none of the previously described attempts to increase the photocrosslinking
yield of the FKBP/FRB protein-protein interaction system was successful, we thought
perhaps the problem lay with the photochemistry of the benzophenone probe. We went
back to examine the synthetic design of the benzophenone photocrosslinker. The
chemistry of benzophenone UV activation involves first absorption of a photon at 300-
365 nm, resulting in the promotion of one electron from the nonbonding n-orbital on
oxygen to an antibonding t* orbital of the carbonyl group. In this ketyl diradical state,
the half-empty n-orbital on oxygen causes oxygen to be electron-deficient and abstract a
nearby hydrogen atom from a C-H bond, leaving a carbon or alkyl radical. Subsequently,
the benzophenone ketyl radical readily combines with the alkyl radical near it to form a
C-C bond (64). A scheme of benzophenone insertion into nearby C-H bond is shown in
Figure 4-15.
H H 0
H1 T
Hi R -7
2 . R H H. HN .
0 UV I' 0 
R
300-365 nm H abstraction Com on4 Volformation
Figure 4-15. Scheme of benzophenone insertion into nearby C-H bond after UV
irradiation. Numbering of carbon positions in benzophenone is indicated in blue.
Benzophenone can be repeatedly activated to ketyl diradical via absorption of a photon at
300-365 nm. After UV activation, if a nearby C-H bond were at the optimal geometry
and distance from the ketyl diradical, the ketyl diradical would abstract a hydrogen atom
from it, leaving behind a carbon radical. If the ketyl radical and the carbon radical still
remain at the optimal interactive distance, they would combine to form a new C-C bond
(indicated in red).
Although detailed mechanisms and kinetics relating to benzophenone radical
formation and photocrosslinking remain largely unknown, we examined benzophenone-
based crosslinking reagents described in literature (64). Two main groups of crosslinkers
were used in biological experiments. The first group consists of modification of small
molecules and natural ligands as photolabile analogs used to identify their interaction
153
sites on protein receptors and macromolecular complexes. The second group includes
modifying specific amino acids of a protein of interest into photocrosslinking moieties,
such as the introduction of benzophenone through unnatural amino acid mutagenesis and
site-specific chemical conjugation of a crosslinker via cysteine-maleimide or cysteine-
iodoacetamide chemistry (65). We noticed that many benzophenone-based crosslinkers
have the benzophenone reactive core structure directly connected to a simple methylene
unit (64), which is different from our benzophenone probe that has a carbonyl functional
group directly attached to the fourth position of the benzophenone ring (Figure 4-15). We
suspected that perhaps the carbonyl group would interrupt the electronics of radical
generation by participating in the triplet excited state, or act as an electron-withdrawing
group that could cause the ketyl radical to delocalize.
In order to test this theory and confirm radical generation of the benzophenone
probe, we took transient absorption (TA) spectra of four different benzophenone
compounds (Figure 4-16) with the help of graduate student Arturo Pizano in the lab of
Professor Dan Nocera. Compound A has no attachment at the benzophenone ring;
Compounds B, the unnatural amino acid benzophenone, has a methyl group attached to
the fourth position of the benzophenone core; and Compounds C and D both have
carbonyl groups directly attached to the benzophenone ring, Compound D is our
benzophenone probe used in previous experiments. All four compounds produced TA
spectra with a sharp peak at 330 nm and a broad peak at 550 nm with a slight shoulder to
the blue region, which are all characteristics of a reduced benzophenone ketyl radical (66).
It is reassuring to know that ketyl radical can be generated. However, Compounds C and
D both showed much higher absorbance at 330 nm than Compounds A and B (data not
shown). Previous work from the Tanaka group demonstrated that having a higher 330 nm
to 550 nm absorbance ratio is an indication of greater ketyl radical character (66). Based
on our result, we can state that having a carbonyl group directly attached to the
benzophenone core made Compounds C and D have more ketyl radical character than
Compounds A and B. But further experiments would be needed to investigate the exact
effect of greater ketyl radical character on the photocrosslinking efficiency of
benzophenone crosslinkers. We decided to design and synthesize a new benzophenone
154
probe that does not have a functional group immediately linked to the benzophenone
reactive core.
0 
2
-4
A
0
OOH
C 0
0
Ko K
B H2N I-OH
0
0
D 0 0
Figure 4-16. Benzophenone (A) and benzophenone derivatives (B-D) with different
functional groups attached to the fourth position of the benzophenone core. Numbering of
carbon positions in benzophenone is indicated in blue.
Following the synthetic scheme shown in Figure 4-17, we synthesized a new
benzophenone probe with a methylene group directly attached to the benzophenone ring
through Friedel-Craft acylation of benzoyl chloride and phenyl valeric methyl ester,
followed by hydrolysis of the methyl ester to generate the corresponding carboxylic acid
(see Experimental Methods). We obtained both para and ortho products in approximately
5:1 ratio. The products were characterized by 'H NMR and mass spectrometry.
0
N Ci1 N AiC13 .
e DCM
40
0
N N 0
40
+4
0
N N 0
OH
LiOH para alkyl benzophenone
+ fUL
ortho alkyl benzophenone
Figure 4-17. Synthetic scheme of alkyl benzophenone probes with methylene unit
attached to the benzophenone core. The para and ortho products were obtained in a 5:1
ratio.
155
Aside from changing the design of the benzophenone photocrosslinker, we also
modified the LAP acceptor peptide by changing the Trp residue located two amino acids
downstream of the catalytic Lys residue to Phe residue, thus creating LAP(F)
(GFEIDKVFYDLDA). The aromatic Trp residue is well known as a mediator of electron
transfer between distant electron acceptor and donor centers in proteins (67). So we
suspected the Trp residue near the ligated benzophenone might quench the formation of
ketyl radicals by acting as an electron transfer site.
To ensure that the new alkyl benzophenone probes can be incorporated by LplA
enzyme, both para and ortho alkyl benzophenone probes were tested in in vitro ligation
reactions using purified LplA(W37G) enzyme and LAP or LAP(F) peptide. Either para or
ortho alkyl benzophenone probe was incubated with purified LplA(W37G) protein, LAP
or LAP(F) peptide, ATP and Mg 2 for 3 hours at 30 *C. Only para alkyl benzophenone
was incorporated and ligated by LplA(W37G) (Figure 4-18). Perhaps the orientation of
the benzophenone core with respect to the alkyl chain in ortho alkyl benzophenone
prevented it from fitting into the rather straight binding tunnel of LplA (47).
para benzophenone
LAP-para benzophenone
LAP(F)-para benzophenone
LAP/LAP(F) v LpIA (W37G)
omit ATP
omit ATP
8 9 10 11 12 13 14 15 16 min
Figure 4-18. In vitro characterization of para alkyl benzophenone ligation by
LplA(W37G). HPLC trace showing LAP (pink trace) or LAP(F) (red trace) peptide
conversion to the benzophenone conjugate. Negative controls are shown with ATP
omitted (black and blue trace). Reactions were performed for 3 hours with 1 pM
LplA(W37G), 100 pM LAP or LAP(F), and 300 pM para alkyl benzophenone.
156
In vitro photocrosslinking with new alkyl benzophenone probe
We wished to compare the new alkyl benzophenone photocrosslinker with our
previous benzophenone probe. In vitro we ligated either 10 ptM FKBP-LAP or FKBP-
LAP(F) protein with one of the photocrosslinkers using purified LplA(W37G) protein,
ATP, and Mg (see Experimental Methods). After ligation was complete, unligated
photocrosslinker probe was removed through dialysis. We then added 10 ptM FRB
protein and 500 nM rapamycin to the reaction mixture to form the ternary complex.
Samples were exposed to UV light using the Porta-Ray 400R instrument from UVitron
for various lengths of time from 1 to 5 minutes. Samples were kept on ice during UV
irradiation time to prevent heat-induced protein aggregation. Samples were then analyzed
by Western blot using antibodies against the myc epitope tag on FKBP-LAP and the HA
epitope tag on FRB. Coomassie stain of UV-irradiated samples showed proteins started to
smear and aggregate if UV irradiation time is longer than 2 minutes using the Porta-Ray
400R which has very powerful UV lamps (data not shown). Thus, UV irradiation time
was set to 2 minutes for subsequent experiments. Figure 4-19 (Top) shows in vitro
photocrosslinking using the new benzophenone probe, alkyl benzophenone, while Figure
4-19 (Bottom) shows same experiment as on the top but using the original benzophenone
probe.
First of all, even though both proteins appeared to be pure by Coomassie staining,
myc-FKBP-LAP and HA-FRB both showed nonspecific antibody staining (indicated by
green arrows) by mouse anti-myc and rabbit anti-HA antibodies, respectively. In addition,
myc-FKBP-LAP appeared to form higher molecular weight aggregates (indicated by blue
arrow) in the presence of the photocrosslinker. The backgrounds are the same as we have
seen before in Figure 4-14. Photocrosslinking was tested using either benzophenone-
ligated FKBP-LAP or benzophenone-ligated FKBP-LAP(F). The photocrosslinked
product stained positive for both the myc tag on FKBP and the HA tag on FRB (lanes 1-2,
boxed in red). However, we also saw background crosslinking in the absence of
rapamycin or the photocrosslinker (Figure 4-19, lanes 4 and 6), which we have also
observed before. These observations were the same for photocrosslinking carried out with
either the new alkyl benzophenone probe (Figure 4-19, top) or the original benzophenone
probe (Figure 4-19, bottom). The benzophenone-independent background (Figure 4-19,
157
lane 6) could be caused by photocrosslinking via the Tyr2105 residue on FRB (PDB
2FAP), located at the interface of the FKBP/FRB complex. We suspected that the
rapamycin-independent background (Figure 4-19, lane 4) could be caused by the high
concentration (10 gM) of proteins in the in vitro reaction mixture. We repeated the same
experiment with lower concentrations of FKBP and FRB. Rapamycin-dependent
formation of UV photocrosslinked product was achieved if concentration of the proteins
were decreased to 2 pM (Figure 4-20). In summary, alkyl benzophenone and
benzophenone, appeared to perform similarly during in vitro crosslinking of purified
proteins.
158
FKBP-LAP(F) NA + + + + +
UV irradiation + + - + + +
rapamycin + + + - + +
HA-FRB + + + + - +
+
+
4
+ +
4
WB anti-myc
NA INAI NA
+
5
NA
6
NA
FKBP-LAP(F) NA + + + + +
UV irradiation + + - + + +
rapamycin + + + - + +
HA-FRB + + + + - +
+
1
+
2
+
3
+
4
+
5 6
40..
30
25
kDa 4- myc-FKBP-LAP
+-HA-FRB
WB anti-myc WB anti-HA
Figure 4-19. In vitro photocrosslinking comparison between alkyl benzophenone and
benzophenone. (Top) Photocrosslinking using alkyl benzophenone probe and purified
proteins myc-FKBP-LAP or myc-FKBP-LAP(F), and HA-FRB. (Bottom) Same as (Top),
except the original benzophenone probe was used. Each LAP or LAP(F)-tagged FKBP
protein (10 pM), and FRB protein (10 pM) was used in this experiment. Photocrosslinked
159
alkyl
bonzophenone
FKBP-LAP
6- myc-FKBP-LAP
E- HA-FRB
WB anti-HA
benzophenone
+ INAINAFKBP-LAP NA INA NA
complex is boxed in red; nonspecific immunostaining background is indicated by green
arrows; higher molecular weight complex of FKBP-LAP or FKBP-LAP(F) is indicated
by blue arrow.
We further tried to optimize the system by lowering the concentrations of
interacting proteins to reduce background photocrosslinking. Instead of using 10 gM of
FKBP and FRB proteins, we lowered protein concentrations by five fold, to 2 pM of each
FKBP and FRB protein. With lower concentration of proteins, we were able to obtain
rapamycin-dependent and alkyl benzophenone-dependent photocrosslinking product
(Figure 4-20, lanes 3 and 5). However, we still saw a higher molecular weight
background (Figure 4-20, blue arrow) that appears to be dependent on the presence of
benzophenone. We tried to reduce this background by trying other UV lamps, decreasing
the time of UV irradiation, and keeping the temperature of the samples cool by storing
the samples on ice during UV irradiation, but the background still persisted. We were
also unable to remove the nonspecific anti-myc staining (Figure 4-20, green arrow) even
though myc-FKBP-LAP(F) appeared to be pure by Coomassie staining.
FKBP-LAP(F) + + + + +
UV irradiation + - + + +
rapamycin + + - + +
HA-FRB + + + - +
alkyl + + + + 
-
benzophenone
1 2 3 4 5
Higher molecular weight bg
Nonspecific anti-myc staining
Photocrosslinked product
WB anti-myc
160
Figure 4-20. Reducing concentrations of FKBP-LAP(F) and FRB helps in eliminating
background photocrosslinking. Alkyl benzophenone-ligated FKBP-LAP(F) (2 pM) was
incubated with 2 pM FRB and 500 nM rapamycin before UV irradiation for 1 minute
using the Porta-Ray 400R crosslinking instrument. Western blot immunostaining with
anti-myc to detect the myc tag on FKBP-LAP(F) showed crosslinked product (boxed in
red) which is dependent on UV irradiation, rapamycin, FRB, and alkyl benzophenone.
After extensive optimization of photocrosslinking conditions, we were able to
obtain photocrosslinking product of FKBP-LAP and FRB that was dependent on UV
irradiation, rapamycin, FRB and alkyl benzophenone. However, we were not able to
increase the yield of the crosslinking product without increasing background at the same
time. It is possible that the concentrations of the interacting protein partners and the small
molecule rapamycin were difficult to fine-tune to the most optimized state, where most
proteins were in the ternary complex form. Since our ultimate goal was to perform
photocrosslinking inside live mammalian cells, which may require a completely different
set of photocrosslinking conditions from the conditions for in vitro experiments, we
decided to move the FKBP/FRB photocrosslinking system to test and further optimize
inside live cells.
161
Quantify benzophenone ligation yield inside live mammalian cells
The first step in photocrosslinking inside live cells was to ensure benzophenone
ligation onto LAP-tagged protein substrate is quantitative. To determine the yield of
benzophenone ligation by LplA(W37G) inside live cells, we applied the same native gel-
shift assay first described in Chapter 2 for estimating azide ligation yield. The same
principle applies that after a small molecule probe has been ligated onto the catalytic
lysine residue in the LAP substrate, changing the positively charged lysine to a neutral
amide, the small molecule-conjugated LAP now migrates faster than the unligated LAP
on native gel. Therefore, the population of ligated LAP can effectively separate itself
from the free LAP. The assay was tested on benzophenone ligated LAP to ensure that the
same method can be applied to other small molecules aside from azide described in
Chapter 2. Different LAP substrates were also tested to make certain that the native gel
migration pattern was generalizable to all LAP conjugated fluorescent proteins and other
proteins of interest.
Three LAP substrates were tested as shown in Figure 4-21 - LAP-tagged YFP
(left), LAP-tagged mCherry (middle), and LAP-tagged mCherry-FKBP (right). In each
experiment, HEK 293T cells were transfected with the indicated LplA enzyme and LAP
substrate, incubated with benzophenone probe for 3 hours, followed by washing to
remove unligated probe for 1 hour. Cells were then lysed with hypotonic lysis buffer and
separated on 12% polyacrylamide native gel. In-gel fluorescence imaging of LAP-tagged
fluorescent proteins was used to analyze and detect the position of ligated and unligated
LAP substrates. Figure 4-21 (left, lane 1) shows benzophenone ligated LAP-YFP
migrates faster than unlabeled LAP-YFP (Figure 4-21, left, lane 3 and 4) due to removal
of a positive charge. In addition, negative control with the lysine in LAP mutated to
alanine, LAP(Ala) (Figure 4-21, left, lane 2), also ran faster on the native gel because the
positive charge has been removed. The assay is also generalizable across various LAP
fusion proteins, such as LAP-tagged mCherry (Figure 4-21, middle) and LAP-tagged
FKBP (Figure 4-21, right).
162
benzophenone + + - I +
LAP-YFP + - + +
LAP(Ala)-YFP - + - -
LpIA(W37G) + + + -
benzophenone + - +
LAP-mCherry + + +
LplA(W37G) + + -
benzophenone I + +
mCherry- +
FKBP-LAP + + +
LplA(W37G) + + -
Wild-type LpiA - - - + Wild-type LplA - - + Wild-type LpLA - - +
1 2 3 4 1 2 3 1 2 3
YFP fluorescence mCherry fluorescence mCherry fluorescence
Figure 4-21. Gel-shift analysis of benzophenone ligation yield in cells. (left)
Benzophenone ligation on LAP-YFP (lane 1). Negative controls are shown with
catalytically inactive LAP (K->A)-YFP (lane 2), benzophenone omitted (lane 3), and
wild-type LplA which does not incorporate benzophenone (lane 4). (middle)
Benzophenone ligation on LAP-mCherry (lane 1). Negative controls are shown with
benzophenone omitted (lane 2), and wild-type LplA, which does not incorporate
benzophenone (lane 3). (right) Benzophenone ligation on mCherry-FKBP-LAP (lane 1).
Negative controls are shown with benzophenone omitted (lane 2), and wild-type LplA,
which does not incorporate benzophenone (lane 3).
Having established an assay to estimate benzophenone ligation yield, we
performed a time course labeling experiment to determine the time it would take for LAP
to achieve quantitative benzophenone ligation yield, which was necessary before
proceeding to live cell photocrosslinking. We hoped to shorten the benzophenone
incubation time to reduce the chance of probe nonspecifically sticking to hydrophobic
areas inside the cell, causing high background crosslinking. HEK 293T cells expressing
LAP-YFP and LplA(W37G) were incubated with benzophenone probe from 1 hour to 3.5
hours. At the end of each ligation time point, cells were washed three times with
benzophenone-free serum-supplemented media and incubated in fresh media for an
163
additional hour before cells were pelleted and lysed. After benzophenone washout, cells
were lysed and separated on a 12% native gel with DTT-free and SDS-free protein
loading buffer to preserved native structure of the protein. Figure 4-22 shows that as
benzophenone ligation time progressed from 1 hour to 3.5 hours (lanes 1-3), we see an
increasing percentage of benzophenone-ligated LAP-YFP migrating faster than free
LAP-YFP on the native gel. At 2.5 hours of benzophenone incubation, about 52% of
LAP-YFP was converted to benzophenone-ligated LAP-YFP (lane 2). LAP-YFP was
almost fully ligated with benzophenone after about 3.5 hours (lane 4). LAP-YFP in the
negative control ligation using wild-type LplA does not appear to run faster on native gel
because wild-type LplA does not incorporate benzophenone (lane 4). Therefore, we
estimated the time to convert all free LAP-YFP to benzophenone-ligated LAP-YFP to be
between 3 hours to 3.5 hours.
benzophenone + + + +
LAP-YFP + + + +
LpIA(W37G) + + + -
Wild-type LpIA - - - +
benzophenone
incubation time (hr) 1 2 3.5 3.5
1 2 3 4
- free LAP
IW-- ligated LAP
32% 52% 94%
YFP fluorescence
Figure 4-22. Native gel-shift assay to measure benzophenone ligation yield over time.
HEK 293T cells expressing the indicated LplA enzyme and LAP-YFP were incubated
164
with 300 iM benzophenone for 1 hour, 2 hours, or 3.5 hours. Cell lysates were analyzed
on a 12% polyacrylamide native gel. Using YFP in-gel fluorescence as readout, lanes 1-3
shows an increasing percentage of benzophenone-ligated LAP-YFP. Estimated percent
conversions to product are given at the bottom of the YFP fluorescence gel image. LAP-
YFP in the negative control reaction (lane 4) with HEK 293T cells expressing wild-type
LplA does not run faster because wild-type LplA does not incorporate benzophenone.
165
Photocrosslinking inside live mammalian cells
Having established that benzophenone ligation onto LAP-tagged protein can be
quantitative inside live cells, we moved forward with performing photocrosslinking
inside live mammalian cells using both alkyl benzophenone and benzophenone probes. In
live cell ligation and photocrosslinking, we needed to ensure optimal expression of the
LplA(W37G) enzyme and the LAP-tagged protein substrate. UV irradiation time must be
ideal for the radical to form covalent bonds with nearby amino acids but not too long to
cause cell aggregation or cell death. In addition, live cells should be carefully kept on ice
during UV irradiation to prevent heat-induced protein aggregation, which could cause
false positives.
We first tried live-cell photocrosslinking on the FKBP-LAP and FRB protein-
protein interaction system. HEK 293T cells were transfected with three plasmids -
FLAG-tagged LplA(W37G), myc-tagged FKBP-LAP, and HA-tagged FRB. We checked
for the expression of each protein via immunostaining against the epitope tag on each
protein (see Experimental Methods). Figure 4-23 shows the expressions of myc-FKBP-
LAP and HA-FRB proteins. We can see that both proteins exhibited strong fluorescence
signals through the entire cell, with and without LplA(W37G) co-expression. Since the
FLAG tag on LplA(W37G) and the myc tag on FKBP-LAP share the same secondary
antibody, we can only immunostain for one of the two proteins with HA-FRB at the same
time. Separate immunostaining against the FLAG tag on LplA(W37G) confirmed that the
enzyme was also expressing well in the presence of FKBP-LAP and FRB (data not
shown).
FKBP-LAP FRB
+ LpIA(W37G)
- LpIA(W37G)
166
Figure 4-23. Expression of FKBP-LAP and FRB in HEK 293T cells. HEK 293T cells
were transfected with FKBP-LAP, FRB and LplA(W37G) plasmids. At 18 hours after
transfection, cells were fixed and immunostained with antibodies against the myc epitope
tag on FKBP-LAP and the HA epitope tag on FRB. Both proteins showed strong
expression in the presence and absence of LplA(W37G) protein.
Following transfection, live cells were incubated in cell culture media containing
either 300 pM alkyl benzophenone or benzophenone probe for up to 4 hours. Cells were
then washed with fresh media for 1 hour, so excess unligated benzophenone can be
removed from cells via the action of organic anionic transporters. Cells were further
incubated in fresh media containing 500 nM rapamycin for an additional hour to
stimulate ternary complex FKBP-rapamycin-FRB formation. Finally, cell culture media
was replaced with ice-cold DPBS buffer and the cells were placed on ice before exposure
to 365 nm UV light for either 2 minutes using the Porta-Ray 400R instrument at 2 inches
below the lamp or 12 minutes using long-wavelength handheld UV lamp with the lamp
placed directly on top of the cell culture plate. After UV irradiation, we lysed the cells
using radio-immunoprecipitation assay (RIPA) buffer. The lysates were analyzed by
Western blot with anti-myc antibody staining to detect myc-FKBP-LAP and its
crosslinked complexes (Figure 4-24). Lanes 1 and 5 in Figure 4-24 are the experimental
lanes. The negative control reactions were performed with either photocrosslinker
omitted (lanes 2 and 6), LplA(W37G) omitted (lane 3 and 7), or UV irradiation omitted
(lane 4 and 8).
167
alkyl ++ +
benzophenone + - + + - - - -
benzophenone - - - - + - + +
FKBP-LAP + + + + + + + +
LpIA(W37G) + + - + + + - +
UV irradiation + + + - + + + -
1 2 3 4 5 6 7 8
150
100 00
80
50
40
30
25
20
FKBP-LAP
15
kDa
WB anti-myc
Figure 4-24. Live cell benzophenone ligation and photocrosslinking of myc-FKBP-LAP
and HA-FRB. HEK 293T cells expressing myc-FKBP-LAP, HA-FRB, and LplA(W37G)
were labeled with alkyl benzophenone or benzophenone and incubated with rapamycin to
stimulate complex formation. Live cells were irradiated with 365 nm UV light for 2
minutes. Cells were then lysed and analyzed with anti-myc staining to detect the myc tag
on FKBP-LAP.
Live cell photocrosslinking with myc-FKBP-LAP and HA-FRB did not show any
crosslinked product, which was expected at around 40 kDa for the heterodimer. We only
detected myc-FKBP-LAP protein at 17.1 kDa on the Western blot and did not observe
any other myc-positive bands. Since we confirmed previously that benzophenone ligation
onto LAP protein substrate in live cells was quantitative under the current condition, it
was possible that ligated benzophenone did not have an ideal interaction partner nearby.
We can see from the crystal structure in Figure 4-12 that LAP tag fused to the C-terminus
of FKBP placed ligated benzophenone in a position outside of the direct interacting
168
interface between FKBP and FRB (61). Since benzophenone ketyl diradical prefers to
have nearby C-H bonds at an optimal distance of about 2 to 4 A (38) to have efficient
crosslinking, the placement of the LAP tag put the photocrosslinker at a disadvantaged
location. Another explanation is that since analysis of crystal structures of FKBP-
rapamycin-FRB shows extensive interactions between rapamycin and its two protein
partners, but relatively limited interactions between the proteins themselves, perhaps this
system is not the most ideal for testing crosslinking with benzophenone, whose diradical
species can only attack geometrically accessible C-H bonds (38,64).
Next, we decided to try a different LAP-tagged protein for live cell
photocrosslinking experiment. We chose LAP-tagged actin because of actin's abundant
interacting protein partners. In addition, actin dynamics are tightly regulated by numerous
proteins, which associate with monomeric actin and/or filamentous actin. We were
hopeful that benzophenone-labeled LAP-actin would photocrosslink to some of its
interacting proteins, such as profilin which interacts preferentially with actin monomers;
capping proteins which interact with barbed ends of actin filament assembly; and Arp2/3
complex which serve as nucleation sites for new actin filaments.
UV
W37GLplA add W GLplA
benzophenone * 1. Cell lysis
2. Western blot
wash out excess aayi
benzophenone analysis
Figure 4-25. Benzophenone photocrosslinker labeling scheme of LAP-actin in live cells.
HEK 293T cells expressing LplA(W37G) and LAP-tagged actin were incubated with
benzophenone probe. After excess unligated benzophenone is removed, cells were
irradiated with 365 nm UV light. Afterwards, cells were lysed and the lysate was
analyzed by Western blot to detect crosslinked complexes of LAP-actin.
Following the scheme in Figure 4-25, we incubated HEK 293T cells expressing
HA-LAP-actin and LplA(W37G) in cell culture media containing either 300 tM alkyl
benzophenone probe or benzophenone probe for 3 hours. After washing the cells with
169
fresh media to remove excess probe via the action of nonspecific anionic transporters, we
replaced the culture media with ice-cold DPBS buffer and put the cells on ice before
exposure to 365 nm UV light for 12 minutes using long-wavelength handheld UV lamp
that was placed directly on top of the cell culture plate, at approximately 1 centimeter
from the live cells. After UV irradiation, cells were lysed and the lysate was analyzed
with anti-HA staining to detect HA-LAP-actin and its crosslinked complexes on Western
blot (Figure 4-26).
alkyl
benzophenone + + +. +
benzophenone - - - - + + -
HA-LAP-actin + - + + + - + +
HA-LAP(F)-actin - + - - - + - -
LpLA(W37G) + + + - + + + +
UV irradiation 4. +
2
+ + + + + I-
3 4 5 6 7 8
230 -
150-
100-
80
50-
kDa -LAP/LAP(F)-actin
WB anti-HA
Figure 4-26. Benzophenone ligation and photocrosslinking on HA-LAP/LAP(F)-actin in
live mammalian cells. After live cell ligation with either alkyl benzophenone or
benzophenone probe and UV irradiation, cells were lysed and analyzed on Western blot
with anti-HA antibody staining. Negative controls are shown with either benzophenone
probe, LplA(W37G), or UV irradiation omitted.
170
We were glad to see that photocrosslinking of alkyl benzophenone ligated HA-
LAP/LAP(F)-p-actin gave two major crosslinked products at about 75 kDa and 95 kDa
indicated by red arrows and four minor products indicated by pink arrows (Figure 4-26,
lanes 1 and 2). The benzophenone probe did not yield any crosslinked product (lanes 5
and 6). The negative control experiments with either photocrosslinker omitted (lane 3),
LplA(W37G) omitted (lane 4), or UV irradiation omitted (lane 8) did not show any
crosslinked bands.
We were curious what proteins were crosslinked to LAP-actin. One of the most
obvious speculations was the LplA(W37G) enzyme. Previous LplA-based protein
labeling gave us evidence that the enzyme sometimes ligates the small molecule probe
onto itself. We stained another Western blot from the live cell alkyl benzophenone
labeling and photocrosslinking experiment with anti-FLAG antibody to detect the FLAG-
epitope tag on LplA(W37G). If any of the photocrosslinked product (Figure 4-26, red and
pink arrows) were HA-LAP-actin crosslinked to FLAG-LplA(W37G), the anti-FLAG
staining should co-localize with anti-HA staining. Figure 4-27 shows only one FLAG-
positive band in each lane, corresponding to the LplA(W37G) enzyme. Therefore, the
HA-positive, photocrosslinked products in Figure 4-26 were not simply HA-
LAP/LAP(F)-actin crosslinked to the enzyme.
171
alkyl
benzophenone + +
benzophenone - - - - + + - -
HA-LAP-actin + + + + - + +
HA-LAP(F)-actin - + - - - + - -
LpIA(W37G) + + + - + + + +
UV irradiation + + +
1 2 3
4
4
+
5
+
6 7 8
A(W37G)
WB anti-FLAG
Figure 4-27. Western blot stained with anti-FLAG antibody to detect FLAG-tagged
LplA(W37G) after live cell photocrosslinking of benzophenone-ligated HA-
LAP/LAP(F)-actin. Experiment in Figure 4-26 was repeated to stain for FLAG-tagged
LplA(W37G). No additional FLAG-positive band was detected aside from LplA(W37G)
enzyme.
Knowing that we were able to obtain photocrosslinked complex of HA-LAP-actin,
we wanted to see if we could reduce the benzophenone incubation time without
sacrificing crosslinking signal. Lowering the photocrosslinker incubation time could
decrease the chance of photocrosslinker sticking inside cells, which may cause unwanted
background. We performed an alkyl benzophenone ligation time course experiment to
determine the optimal probe incubation time to achieve highest amount of
photocrosslinked products. HEK 293T cells expressing LplA(W37G) and HA-LAP-actin
were incubated with alkyl benzophenone probe from 1 hour to 4 hours. At the end of
172
each ligation time point, cells were washed and incubated in fresh cell culture media for 1
hour to remove excess probe. We prepared cell lysate from each time point and
immunostained with anti-HA antibody to detect HA-LAP-actin. Figure 4-28 shows that 1
hour of alkyl benzophenone live-cell incubation (lane 1) was able to give comparable
signal of photocrosslinked product as 4 hours of probe incubation (lane 4). However, we
observed that negative control with alkyl benzophenone probe omitted shows a
background photocrosslinked band at around 75 kDa (Figure 4-28, lane 6, green arrow).
The intensity of the background band is weaker than that of the experiment (lanes 1-4).
However, this background band is only present in the negative control with alkyl
benzophenone omitted (lane 6, green arrow) and not in other negative controls with either
UV irradiation omitted (lane 5), LplA(W37G) omitted (lane 7), with wild-type LplA
(lane 8), or with the substrate LAP-actin omitted (lane 9). We speculated that the
background could be caused by surface-exposed Tyr residues in actin that were also
activated by the 365 nm UV light and crosslinked to nearby protein. Nonetheless, we
were glad to see that HA-LAP-actin photocrosslinked products can be obtained with only
1 hour of live cell alkyl benzophenone incubation.
UV irradiation + + + + - + + + +
alkyl benzophenone + + + + + - + + +
LpiA(W37G) + + + + + + - - +
Wild-type LpLA - - - - - - - +
LAP-actin + + + + + + + +
alkyl benzophenone
incubation time (hr) 2 3 4 4 4 4 4 4
2 3 4 5 6 7 8 9
WB anti-HA 173
Figure 4-28. Live cell photocrosslinking of HA-LAP-actin with alkyl benzophenone
incubation time course. HEK 293T cells expressing LplA(W37G) and HA-LAP-actin
were incubated with 300 pM of alkyl benzophenone probe from 1 hour to 4 hours (lanes
1-4). Cell lysates were analyzed with anti-HA staining on Western blot. Negative controls
are shown with either UV irradiation omitted (lane 5), alkyl benzophenone omitted (lane
6), LplA(W37G) omitted (lane 7), HA-LAP-actin omitted (lane 9), or with wild-type
LplA (lane 8), which does not incorporate alkyl benzophenone probe.
In another live cell photocrosslinking experiment, we performed benzophenone
ligation and photocrosslinking on the microtubule-associated protein 2 (MAP2). MAP2 is
predominantly found in neurons. Its principal function is to regulate tubulin concentration
and to maintain cell morphology by regulating the spacing of microtubules (68). We
previously demonstrated LplA-mediated small molecule labeling on MAP2 with
coumarin (49) as well as larger fluorophores like fluorescein and X-rhodamine using the
LplA-mediated two-step labeling methodology (69). Benzophenone probe ligation
followed by photocrosslinking in live HEK 293T cells also yielded photocrosslinked
products. Following a similar protocol used in LAP-actin crosslinking described
previously, we incubated HEK 293T cells expressing LplA(W37G) and HA-LAP-MAP2
in cell culture media containing either alkyl benzophenone or benzophenone
photocrosslinker probe for 3 hours. After washing the cells with fresh media to remove
excess probe, we replaced the culture media with ice-cold DPBS buffer. Cells were
placed on ice before exposure to 365 nm UV light for 12 minutes using long-wavelength
handheld UV lamp. The UV lamp was placed directly on top of the cell culture plate, at
approximately 1 centimeter from the live cells. After UV irradiation, cells were lysed and
the lysates were analyzed with anti-HA immunostaining to detect HA-LAP-MAP2 and its
crosslinked complexes on Western blot.
We observed one major photocrosslinked band that is at 120 kDa (Figure 4-29,
dark blue arrow), 40 kDa more than HA-LAP-MAP2 substrate; and one minor product at
90 kDa (Figure 4-29, light blue arrow). Photocrosslinking with the original
benzophenone probe did not produce any crosslinked product (lane 6), consistent with
our result from live cell HA-LAP-actin crosslinking. However, a weak background band
174
at 120 kDa also appeared in the omitting alkyl benzophenone negative control reaction
(Figure 4-29, lane 3, green arrow), similar to the background band we observed while
performing LAP-actin crosslinking in live cells (Figure 4-28, lane 6, green arrow).
Compared to the photocrosslinking signal (Figure 4-29, dark blue arrow), this
background signal is a lot weaker. Other negative control reactions with either UV
irradiate omitted (lane 2), LplA(W37G) omitted (lane 4), or with wild-type LplA (lane 5)
do not show any crosslinking signal.
UV irradiation + - + + + + + +
alkyl + ++ +benzophenone + + - + + - -
benzophenone - - - - - + - +
LpIA(W37G) + + + - - + + -
Wild-type LpLA - - - - + - - +
1 2 3 4 5 6 7 8
230
150
100
80 4
HA-LAP-MAP2
60
50
40
kDa WB anti-HA
Figure 4-29. Benzophenone ligation and photocrosslinking performed on HA-LAP-
MAP2 in live mammalian cells. After live cell ligation with either alkyl benzophenone or
benzophenone probe and UV irradiation, cells were analyzed on Western blot with anti-
HA antibody staining. Negative controls are shown with either UV irradiation,
benzophenone probe, or LplA(W37G) omitted (lanes 2-4, lane 7) or with wild-type LplA
(lane 5), which does not incorporate the benzophenone probe. Dark blue and light blue
arrows indicate photocrosslinked products; green arrow indicates background crosslinked
product in the absence of alkyl benzophenone.
175
In summary, live cell photocrosslinker probe ligation and crosslinking using alkyl
benzophenone on HA-LAP-actin and HA-LAP-MAP2 proteins showed crosslinked
products that were detected by immunostaining with antibody against the HA tag.
Compared to the original benzophenone probe that showed no photocrosslinked products
in either actin or MAP2 context, the alkyl benzophenone probe appeared to work better in
live cell experiments. We decided to pursue subsequent live cell photocrosslinking
experiments with the alkyl benzophenone probe.
176
LAP-actin photocrosslinking with drug treatment inside live mammalian cells
Following the success of LAP-actin and LAP-MAP2 photocrosslinking in live
mammalian cells, we wanted to use different drug stimulants to induce or prevent
different proteins from interacting with actin to change the protein-protein interaction
pattern. This experiment also served to help us in checking the validity of our
intracellular benzophenone crosslinking method. We should be able to obtain different
photocrosslinked products since different proteins are known to associate with actin
whether it is in its monomeric or filamentous state (Figure 4-30).
0 profilin capn
protein
actin monomer UV 356 nm
crosslinked complexes
without drug treatment
Add
depolymerization
drug
UV 356 nm
crosslinked complexes
with drug treatment
Figure 4-30. Schematic representations of different LAP-actin crosslinked complexes
with and without depolymerization drug treatment. At 365 nm UV light irradiation of F-
actin (a polymer of benzophenone-ligated LAP-actin monomers) causes LAP-actin to
crosslink to capping protein and also Arp2/3 complex. Addition of an actin
depolymerization drug depolymerizes a majority of F-actin into G-actin, which is then
bound by profilin. As a result, UV irradiation of G-actin leads to a different pool of
crosslinked complexes. Red bars represent covalent bond formation.
We first used various drug treatments to change the polymerization state of actin
as well as its localization. We examined a total of five stimulants. Three stimulants
changed the polymerization state of actin. First, latrunculin A is a potent macrolide that
177
binds to monomeric G-actin, forming a 1:1 complex to disrupt microfilament
polymerization (70). Second, cytochalasin B is a drug that disassembles stress fibers,
causing aggregation of actin fibers within the cytosol. Cytochalasin B is about 10- to 100-
fold less potent than latrunculin A (71). Third, CK-666 inhibits nucleation of new actin
filaments by binding at the interface of Arp2 and Arp3 proteins, locking Arp2/3 complex
in an inactive conformation (72). In addition, two stimulants increased the population of
actin in the nucleus. Leptomycin B is a fatty acid that inhibits the nuclear export of
proteins containing nuclear export signal. Previously in Dr. Chayasith Uttamapinant's
work (49), leptomycin B was used to block actin export from the nucleus, thus increasing
the amount of nuclear actin pool, so it can be ligated by coumarin ligase to provide
further evidence for the presence of nuclear actin. In addition to leptomycin B, heat-
shock induced cell stress has also been shown to increase the abundance of nuclear actin
(73). Since drug treatments and heat shock either alter actin polymerization states or its
location inside the cell dramatically, we expected to see different protein-protein
interaction complexes being crosslinked whether actin is being inhibited, disassembled,
or translocated to the nucleus.
We incubated HEK 293T cells expressing LplA(W37G) and HA-LAP-actin in
media containing the alkyl benzophenone probe. Following probe ligation, cells were
incubated in media containing each drug stimulant for 40 minutes. For heat-shock
induced cell stress, the cells was moved form the 37 'C incubator to 43 'C incubator for 1
hour. At the end of drug or heat treatment, the cell culture media was replaced with
DPBS and the cells were irradiated with 365 nm UV light for 12 minutes using the
handheld UV lamp. The cells were then lysed and analyzed on Western blot with anti-HA
immunostaining to detect HA-LAP-actin crosslinked complexes.
However, as shown in Figure 4-31, none of the stimulants was able to induce new
HA-LAP-actin crosslinked product that is different from the crosslinked complexes
already detected in the absence of drug and heat treatments. In Figure 4-31, lane 1
represents crosslinked complexes already identified without stimulant, and each lane in
lanes 4, 6, 7, 8, and 10 represents one of the five stimulants tested in this experiment. The
crosslinked products were not observed if alkyl benzophenone or UV irradiation was
omitted (lanes 2, 3, 5, 9 11). We were puzzled regarding the lack of different crosslinked
178
products. We were confident that the drug treatments were successful because we
observed drastic cell morphology changes such as retraction of the filopodia and cell
rounding, similar to literature reports (74,75).
alkyl +
benzophenone + - + + - + +
UV irradiation + + - + + + +
Leptomycin B - - - + -
Heat - - - -1- +
alkvl I I
benzophenone -I + -
Cytochalasin B + + - I -
Latrunculin A - - - I- - -I + CK-666 - + I +
1 2 3 4 5 6 7 1 2 3 4
230 230
1501
100
100
804
80
60
60
50
40-actin 50
kDa WB anti-HA WB anti-HA
Figure 4-31. Live cell photocrosslinking of LAP-actin with various drug treatments. HEK
293T cells expressing LplA(W37G) and HA-LAP-actin were incubated with 300 gM
alkyl benzophenone for 3 hours. Afterward alkyl benzophenone ligation, cells were
incubated for 40 minutes at 37 'C with 4 nM of leptomycin B, 500 nM of latrunculin A,
10 pM cytochalasin B, or 10 pM CK-666. For heat-induced stress, cells were moved
from 37 'C incubator to 43 'C incubator for 60 minutes after alkyl benzophenone ligation.
HA-LAP-actin is indicated by black arrow. Crosslinked complex of HA-LAP-actin are
indicated by red arrow.
These results led us to question the validity and the efficiency of our
photocrosslinking method. It should be obvious that different proteins interact with actin
whether actin is in its monomeric state or in its fibrous state. The fact that the
photocrosslinking pattern stayed the same under many potent drug treatments alerted us
to reexamine our system more closely. The photocrosslinked complexes of HA-LAP-
actin we identified in live cell photocrosslinking were reproducible under many
179
experiments and dependent on UV irradiation, LplA(W37G) enzyme, and alkyl
benzophenone probe. However, there are still some unresolved issues that should be
addressed in future experiments. First of all, we continued to see a weaker background
crosslinked band in the alkyl benzophenone negative control experiments in both live-cell
HA-LAP-actin photocrosslinking and HA-LAP-MAP2 photocrosslinking (Figure 4-28,
lane 6; Figure 4-29, lane 3). We do not yet understand the source of this background and
why it only appeared in the negative control with alkyl benzophenone omitted. Second,
the photocrosslinked complexes we identified in HA-LAP-actin live-cell
photocrosslinking should be altered under drug treatment, but we did not observe a
difference in the identified crosslinked bands with and without drug treatment. It is
possible that the proteins crosslinked to actin in our experiments were not changed with
different stimulants, but we find this unlikely because actin's interacting partners differ
greatly depending on its polymerization states. In the future, we could isolate the
crosslinked proteins from the gel to perform mass spectrometry analysis in order to
identify what proteins were crosslinked to HA-LAP-actin. Understanding the identity of
the crosslinked proteins would offer additional evidence and support regarding the
validity of our photocrosslinking method.
180
Conclusion
We have demonstrated in vitro photocrosslinking with the FKBP-rapamycin-FRB
interaction system using two benzophenone probes, benzophenone and alkyl
benzophenone. The two probes differ at the point of attachment to the fourth position of
the benzophenone core - the first version benzophenone has a carbonyl group attached
which we speculated might interfere with the ketyl diradical generation through electron
delocalization, therefore, alkyl benzophenone with simply a methylene group attached at
the fourth position was synthesized to address this concern. Both probes behaved
similarly under in vitro crosslinking environment with purified proteins. However, only
alkyl benzophenone produced crosslinked products once we tried photocrosslinking
inside live cells with either an alkyl benzophenone-ligated or benzophenone-ligated LAP-
actin or LAP-MAP2 protein. To further explore the nature of the crosslinked complexes,
we used various drug stimulants as well as heat-induced cellular stress in an attempt to
change the polymerization state or localization of actin, which would cause different
proteins to interact with actin. We were not able to obtain any additional or different
crosslinking signal from the ones already observed under normal state without drug
treatment.
Several aspects of the project made us reevaluate our goal of developing a robust
and easy to implement protein-protein interaction detection method. First, the placement
of the photocrosslinker often has to be optimized by screening many regions in the
protein of interest. This can be challenging in the LplA-mediated labeling technique
because it requires many different genetic fusions of the LAP peptide to the protein under
study. So far, LplA-mediated labeling technique has been demonstrated mostly on
proteins with LAP fusions at the termini (29,49,51,52,69,76,77), where LAP is likely to
be easily accessible. It is conceivable that certain regions, especially internal or buried,
would have different degrees of accessibility and probe ligation yield by LplA. Second,
prior to photocrosslinking, one would need to quantify the ligation yield for each LAP-
tagged protein using native gel-shift assay, which can be labor-intensive when many
fusion positions are tested. Third, for the protein under study, one would need to choose
the region of LAP fusion carefully in order to avoid disrupting the native structure or the
binding ability of the protein to other interacting proteins. A protein-specific functional
181
assay comparing the wild-type and LAP-tagged protein should be accomplished.
Information regarding the protein's interacting regions or a crystal structure of the protein
can help in deciding possible regions for LAP fusions. For example, we did not fuse LAP
to the internal region of FKBP because it contains many residues that form hydrogen
bonds or hydrophobic interactions with rapamycin. Ideally, we need to achieve
quantitative ligation yield of the photocrosslinker on LAP-tagged protein in order to
maximize crosslinking signal, but this requirement cannot always be guaranteed
especially if LAP is inaccessible.
Finally, we also need synthetic chemists to design novel photocrosslinkers that
could work both specifically and efficiently with high sensitivity and less background
reaction or quenching by media or solvent.
182
Experimental methods
Genetic constructs
Construct name Vector Features and variants References
FLAG-LplA pet2la FLAG: DYKDDDDK (69)
Other LplA mutants: W37G, W37A,
W37V, W371, W37L, W37S
E2p pet21a (77)
myc-FKBP-LAP pet21a myc: EQKLISEEDL (50)
LAP: GFEIDKVWYDLDA
Other FKBP mutants: M49A, M66A,
M49A and M66A
myc-FKBP-LAP(F) pet2 1 a LAP(F): GFEIDKVFYDLDA
myc-FKBP pet21a (50)
HA-FRB-YFP pet21a HA: YPYDVPDYA
YFP: yellow fluorescent protein
HA-FRB-mApple pet21a mApple: a red fluorescent protein
HA-FRB-LAP pet21a
HA-FRB-(10 a.a. pet21a 10 a.a. linker: GSGSTSGSGK
linker)-LAP
HA-FRB pet2 1 a
FLAG-LplA pcDNA3 Other LplA mutant: W37G (69)
FLAG-LplA pCAG Other LplA mutant: W37G
LAP-YFP pcDNA3 (69)
LAP (Ala)-YFP pcDNA3 LAP(Ala): GFEIDAVWYDLDA (69)
LAP-mCherry pcDNA3 mCherry: a red fluorescent protein (69)
mCherry-FKBP-LAP pcDNA3
myc-FKBP-LAP pcDNA3 (50)
HA-FRB pcDNA3
HA-LAP-actin Clontech vector Other LAP variant: LAP(F) (69)
HA-LAP-MAP2 Clontech vector (69)
General synthetic methods
All reagents were the highest grade available and purchased from Sigma-Aldrich,
Anaspec, Thermal Scientific, TCI America, Alfa Aesar, or Life Technologies and used
without further purification. Anhydrous solvents were drawn from Sigma-Aldrich
SureSeal bottles. Analytical thin layer chromatography was performed on 0.25 mm silica
gel 60 F254 plates and visualized under short or long wavelength UV light, or after
staining with 2,4-dinitrophenylhydrazine (DNPH) or bromocresol green. Flash column
chromatography was carried out using silica gel (ICN SiliTech 32-63D). Mass
spectrometric analysis was performed on an Applied Biosystems 200 QTRAP mass
183
spectrometer using electrospray ionization. HPLC analyses were performed on a Varian
Prostar Instrument equipped with a photo-diode-array detector. A reverse-phase
Microsorb-MV 300 C18 column (250 x 4.6 mm dimension) was used for analytical
HPLC. NMR spectra were recorded on a Bruker AVANCE 400 MHz instrument.
Synthesis of benzophenone probes
Synthesis of benzophenone
0 0 00
NHS, DCC O H C AOHRT HUri CHN RT OH j4 -00
~ CH 3CN, TEA, RT
0 benzophenone (n=3,4)
To a solution of 40 mg of 4-benzoylbenzoic acid (0.18 mmole, Sigma Aldrich) dissolved
in 10 ml of anhydrous acetonitrile (MeCN) was added 22 mg of N-hydroxysuccinimide
(NHS) (0.19 mmol, Sigma Aldrich) and 43 mg of N,N'-dicyclohexylcarbodiimide (DCC)
(21 mmol, TCI America). The reaction was stirred at room temperature for 10 hours.
Afterwards, MeCN was removed under reduced pressure and the crude material was
purified by silica gel column chromatography (column size: 4 inches tall, 0.8 inch in
diameter; solvent gradient from 25 to 50% ethyl acetate in hexane). Purified 4-
benzoylbenzoic-NHS ester (48 mg, 0.15 mmol) was dissolved in 4 ml of anhydrous
MeCN and reacted with either 17 mg of 4-aminobutyric acid (0.17 mmol, Sigma Aldrich)
or 20 mg of 5-aminovaleric acid (0.17 mmol, Sigma Aldrich), and 14 tl of anhydrous
triethylamine (TEA) (0.10 mmol, Sigma Aldrich) for 12 hours at room temperature.
Solvent was removed under reduced pressure and the crude mixture was purified by silica
gel column chromatography (column size: 4 inches tall, 0.8 inch in diameter; solvent
gradient from 5 to 10% methanol in dichloromethane (DCM)). Both benzophenone
products were confirmed by 'H NMR and mass spectrometry.
Characterization of benzophenone (n=3): 'H NMR (CDCl 3): 7.93 (d, 2H, J= 4.2), 7.83 (t,
1H, J= 4.8), 7.45 (in, 4H), 7.57 (m, 1H), 7.45 (in, 2H), 3.47 (in, 1H), 3.06 (q, 4H, J=
7.2), 2.41 (t, 2H). ESI-MS calculated for [M+H]*: 311.12; observed 312.03.
184
Characterization of benzophenone (n=4): 'H NMR (CDCl3): 7.92 (d, 2H, J= 4.2), 7.80
(m, 4H), 7.60 (m, 1H), 7.47 (m, 2H), 6.92 (m, 1H), 3.49 (m, 2H), 3.05 (m, 4H), 2.36 (m,
3H). ESI-MS calculated for [M+H]*: 325.13; observed 326.41.
Synthesis of alkyl benzophenone
0 0
0 .0OH
4 4
CI , o AIC13  + LiOH para alkyl benzophenone0-+ OH
0/ DCM
ortho alkyl benzophenone
To a solution of 500 mg of benzoyl chloride (3.6 mmol, Sigma Aldrich) dissolved in 10
ml of anhydrous DCM was added 142 mg of anhydrous AlCl 3 (1.1 mmol, TCI America).
The reaction was stirred at room temperature for 30 minutes, during which time the color
of the reaction changed gradually from clear yellow to red, indicating the formation of
the acylium ion. After the reaction turned red, 750 mg of 5-phenyl valeric methyl ester
(3.9 mmol, Sigma Aldrich) was added. The reaction was stirred at room temperature for 2
hours and monitored by thin layer chromatography (TLC). TLC followed by 2,4-
dinitrophenylhydrazine (DNPH) stain indicated the formation of the ketone functional
group. After 2 hours, DCM was removed under reduced pressure. The resulting residue
of the crude reaction mixture was purified by silica gel chromatography (column size: 6
inches tall, 0.8 inch in diameter; solvent gradient from 5 to 20% ethyl acetate in hexane).
Benzophenone alkyl methyl ester was hydrolyzed to the corresponding carboxylic acid
using LiOH in 5 ml of 50% tetrahydrofuran (THF) in water for 1 hour at room
temperature. Hydrolysis was monitored by TLC and bromocresol green stain to identify
the carboxylic acid. The reaction mixture was diluted into 10 ml of 1M HCL and
extracted with ethyl acetate (3 X 10 ml). The organic layer was dried over magnesium
sulfate, then filtered. Ethyl acetate was removed under reduced pressure, Even though the
para benzophenone alkyl methyl ester was favored, both para and ortho products were
185
obtained in approximately 5 to 1 ratio. Product yield for the para product was
approximately 30% and the yield for the ortho product was approximately 6%. Both
products were confirmed by 'H NMR and mass spectrometry
Characterization ofpara alkyl benzophenone: 'H NMR (CD 30D): 7.79 (m, 2H), 7.74 (m,
2H), 7.58 (m, 1H), 7.48 (m, 2H), 7.28 (m, 2H), 2.72 (t, 2H, J= 5.6), 2.40 (t, 2H, J= 5.6),
1.712 (m, 4H). ESI-MS calculated for [M+H]*: 282.13; observed 283.4.
Characterization of ortho alkyl benzophenone: 'H NMR (CD 30D): 7.83 (m, 2H), 7.61 (m,
1H), 7.46 (m, 2H), 7.30 (m, 3H), 2.70 (m, 2H), 2.31 (t, 2H, J= 5.6), 1.63 (m, 1H), 1.28
(m, 4H). ESI-MS calculated for [M+H]*: 282.13; observed 283.5.
Protein expression and purification: see Chapter 2 Experimental Methods.
HPLC screening to identify LpiA enzyme and benzophenone probe pair
For the screen in Figure 4-4, reactions containing 500 nM LplA enzyme, 300 RM
benzophenone probe, 100 p.M LAP peptide (sequence: H2N-GFEIDKVWYDLDA-
CO 2H), 2 mM ATP, and 2 mM magnesium acetate in 25 mM Na2HPO4 pH 7.2 were
incubated at 30 *C for 40 minutes. Reactions were quenched with 40 mM EDTA
(ethylenediaminetetraacetic acid, final concentration). Percent conversion to LAP-
benzophenone adduct was quantified by comparing area under the peaks on HPLC with a
C18 reverse phase column, recording absorbance at 210 nm. Elution conditions were 30-
60% acetonitrile in water with 0.1% trifluoroacetic acid over 20 min (linear gradient) at
1.0 mL/min flow rate. The percent conversion was calculated from the ratio of LAP-
benzophenone to the sum of unmodified LAP + LAP- benzophenone.
In vitro benzophenone ligation onto E2p or purified LAP-tagged protein
In vitro benzophenone ligation reactions contained 1 p.M LplA(W37G) enzyme, 500 [tM
benzophenone or alkyl benzophenone probe, 50 p.M E2p or 10 gM LAP-tagged protein
(except in Figure 4-20, where 2 pM of FKBP-LAP was used), 2 mM ATP, and 2 mM
magnesium acetate in phosphate-buffered saline (PBS) were incubated at 30 'C for 3
hours. Reactions were analyzed on HPLC using a linear gradient from 30-60%
acetonitrile in water with 0.1% trifluoroacetic acid over 20 minutes at 1.0 mL/min flow
rate to confirm quantitative benzophenone ligation on E2p or LAP-tagged proteins. After
186
ligation was complete, reactions were transferred into dialysis membrane tubing
(SnakeSkin, Cat. # 88242, Pierce) with 3,500 MWCO or dialysis cups (Cat. # 69570,
Pierce) with 10,000 MWCO to removed excess unligated benzophenone probe through
dialysis in phosphate buffered saline (PBS) buffer for a total of 10 hours. Dialysis buffer
was changed to fresh PBS buffer after 5 hours.
Detecting benzophenone-ligated E2p with biotin-hydrazide (Figure 4-5) or
hydroxylamine-Alexa Fluor 488 (Figure 4-8)
Benzophenone-ligated E2p (50 pLM) was incubated with biotin-hydrazide (200 pM) or
hydroxylamine-Alexa Fluor 488 (200 pM) at pH 4.5 or 5, respectively, for 12 hours at
room temperature. Afterwards, the samples were mixed with protein loading buffer and
separated on a 12 % SDS-PAGE gel.
For the samples reacted with biotin-hydrazide, the proteins were transferred onto a
nitrocellulose membrane at 400 mA (constant mA) for 2.5 hours. Complete transfer of
the pre-stained protein ladder was used as an indication of successful protein transfer.
The nitrocellulose membrane was blocked with 3% w/v BSA (Pierce), in TBS-T (0.1%
Tween-20 in Tris-buffered saline) buffer at room temperature for 1 hour. To detect
biotinylated proteins, membrane was immersed in a solution of 0.3 mg/mL streptavidin-
HRP (Thermo Scientific) at room temperature for 1 hour before rinsing with TBS-T 4x 5
minutes. Blots were developed with the SuperSignal West Pico reagent (Thermo
Scientific), and imaged using an Alpha Innotech digital camera.
For the samples reacted with hydroxylamine-Alexa Fluor 488, after separating the
proteins on 12% SDS-PAGE gel, in-gel fluorescence of Alex Fluor 488 was visualized on
a FUJIFILM FLA-9000 instrument.
In vitro photocrosslinking with purified FKBP and FRB proteins
Benzophenone ligation on LAP-tagged protein is described above. After ligation is
complete, reactions were transferred into dialysis membrane tubing (SnakeSkin, Cat. #
88242, Pierce) with 3,500 MWCO to removed excess unligated benzophenone probe
through dialysis in phosphate buffered saline (PBS) buffer for a total of 10 hours.
Dialysis buffer was changed to fresh PBS buffer after 5 hours. Following dialysis, 10 gM
187
of either purified FKBP or FRB protein (except in Figure 4-20, where 2 gM of FRB
protein was added) and 500 nM rapamycin (1 mM stock concentration in DMSO) were
added to the reactions. After 1 hour of incubation with rotation at room temperature for
ternary complex formation, reactions were transferred into wells in a 96-well plate, which
was placed on ice and exposed to UV light using the specified instrument for the
indicated amount of time. All samples were kept on ice during UV irradiation to prevent
heat-induced protein aggregation.
Optimized conditions for each photolysis instrument
Instrument wavelength Power output Distance from Time of UV
samples irradiation
Porta-Ray 400R 350 - 400 nm 400-Watt 3 inches 2 minutes
Handheld long 365 nm 4-Watt 1 cm 15 minutes
wavelength UV lamp
Stratalinker 365 nm 5 lamps, 8-Watt/lamp 5 inches 30 minutes
Western blot analysis of in vitro FKBP and FRB photocrosslinking
After UV irradiation, samples were combined with protein loading buffer and separated
on a 12% SDS-PAGE gel at constant 200 V. For blotting analysis, proteins in the SDS-
PAGE gel were transferred to nitrocellulose membrane. Protein transfer was run for 2.5
hours at constant 400 mA. Complete transfer of pre-stained protein ladder was used as an
indication of successful protein transfer. Afterwards, the nitrocellulose membrane was
blocked with 3% w/v BSA (Pierce), in TBS-T (0.1% Tween-20 in Tris-buffered saline)
buffer at room temperature for 1 hour. To stain with mouse anti-myc (Millipore) or rabbit
anti-HA (Invitrogen) primary antibody, the membranes were incubated with antibody
(1:1000 dilution) in TBS-T buffer at room temperature for 1 hour, then rinsed with TBS-
T buffer 4 x 5 minutes. The membranes were then incubated with goat anti-mouse or
anti-rabbit horseradish peroxidase conjugate (BioRad, 1:5000 dilution) in TBS-T buffer
at room temperature for 1 hour. Membranes were washed 4 x 5 minutes with TBS-T, then
maintained in TBS-T at room temperature for imaging. The blots were developed using
Supersignal West Pico substrate (Pierce) and visualized on an Alpha Innotech
Chemilmager 5500 instrument.
Native gel-shift analysis of benzophenone ligation yields in cells (Figures 4-21, 4-22)
188
HEK 293T cells were plated into wells of a 12-well culture plate (4 cm2 per well) 18
hours prior to transfection and grown to 60% confluency. For benzophenone ligation,
cells were transfected with 400 ng LplA(W37G) and either 800 ng LAP-YFP, LAP(Ala)-
YFP, LAP-mCherry, or mCherry-FKBP-LAP using Lipofectamine 2000 (Life
Technologies) following manufacture's protocol. At 18 hours after transfection, cells
were incubated in growth media (MEM supplemented with 10% FBS) containing 300
pM benzophenone probe for up to 3.5 hours at 37 "C. For Figure 4-22, cells were
incubated with benzophenone for either 1 hour, 2 hours, or 3.5 hours. At the end of
benzophenone incubation time, cells were then washed with fresh media for 1 hour, so
excess unligated benzophenone can be removed from cells via the action of organic
anionic transporters. Cells were then harvested using a cell scrapper and lysed in 100 pl
hypotonic lysis buffer (1 mM HEPES pH 7.5, 5 mM MgCl 2, 1 mM PMSF
(phenylmethanesulfonyl fluoride, Thermal Scientific), 1 mM protease inhibitor cocktail
(Sigma-Aldrich)) using freeze-thaw lysis method. Cell lysates were frozen at -20 *C,
thawed at room temperature, then mixed by vortexing for 2 minutes. This freeze-thaw-
vortex cycle was repeated three times. Cells were then centrifuged at 13,000 rpm for 2
minutes to pellet insoluble cellular fractions. The lysate supernatant (5 pl lysate per lane)
was combined with DTT-free and SDS-free protein loading buffer and separated and
analyzed on a 12% polyacrylamide native gel without SDS at constant 200 V. Lysate
samples were not boiled to preserve the native structure and the fluorescence of the LAP-
tagged fluorescent proteins. Prior to Coomassie staining, in-gel fluorescence of YFP or
mCherry was visualized on a FUJIFILM FLA-9000 instrument.
Immunostaining to check expressions of FKBP-LAP and FRB (Figure 4-23)
HEK 293T cells were plated on 5 mm x 5 mm glass coverslips placed within wells of a
48-well cell culture plate (0.95 cm2 per well) 12-16 hours prior to transfection. Glass
coverslips were pre-coated with 50 pg/mL fibronectin (Millipore) to increase HEK 293T
adherence. Plated cells were transfected with 100 ng FKBP-LAP in pcDNA3, 100 ng
HA-FRB in pcDNA3, and 100 ng LplA(W37G) in pCAG using lipofectamine. At 18
hours after transfection, cells were washed 3x with Dulbecco's phosphate buffered saline
(DPBS) and fixed with 3.7% paraformaldehyde in DPBS for 15 minutes at room
189
temperature. Fixed cells were washed 3x with DPBS followed by protein precipitation
with cold methanol for 6 minutes at -20 'C. After cold methanol precipitation, cells were
washed with DPBS 3x and blocked with 0.5% casein in DPBS for 1 hour at room
temperature.
To visualize myc-FKBP-LAP and HA-FRB proteins, fixed cells were immunostained by
first incubating with primary antibodies, mouse monoclonal anti-myc antibody (Millipore,
1:1000 dilution) and rabbit monoclonal anti-HA antibody (Invitrogen, 1:1000 dilution) in
1% (w/v) BSA in DPBS for 1 hour at room temperature. Cells were further washed with
DPBS 3x and incubated with secondary antibody, goat anti-mouse IgG antibody
conjugated to Alexa Fluor 488 (Life Technologies) and goat anti-rabbit IgG antibody
conjugated to Alexa Fluor 568 (Life Technologies) in 1% (w/v) BSA in DPBS for 1 hour
at room temperature. Cells were then washed 3x and imaged on a confocal microscope.
Benzophenone probe ligation and photocrosslinking on HA-LAP-actin or HA-LAP-
MAP2 in live cells (Figures 4-26, 4-27, 4-28, 4-29)
HEK 293T cells were plated into wells of a 12-well culture plate (4 cm2 per well) 18
hours prior to transfection and grown to 60% confluency. For benzophenone ligation,
cells were transfected with 400 ng LplA(W37G) and 800 ng HA-LAP-actin or HA-LAP-
MAP2 using Lipofectamine 2000 (Life Technologies) following manufacture's protocol.
18 hours after transfection, cells were incubated in growth media (MEM supplemented
with 10% FBS) containing 300 RM benzophenone or alkyl benzophenone probe 3 hours
(except in Figure 4-28, where alkyl benzophenone incubation was from 1 hour to 4 hours
as indicated in the figure) at 37 "C. Excess benzophenone probe was removed by rinsing
cells with fresh cell culture media 3 times over 1 hour and relying on the action of
organic anionic transporters. Media was replaced with cold DPBS and the cells were
placed on ice. Handheld, long wavelength UV lamp was placed directly on top of cell
culture plate, 1 centimeter from the live cells. Cells were UV irradiated for 12 minutes
and then lysed with 50 pl RIPA lysis buffer (50 mM Tris, 150 mM NaCl, 0.1% SDS, 0.5%
sodium deoxycholate, 1% Triton X-100) containing 1 mM PMSF (Thermal Scientific,
phenylmethanesulfonyl fluoride) and 1 mM protease inhibitor cocktail. Cell lysate was
separated on a 9% SDS-PAGE gel. For blotting analysis, proteins in the SDS-PAGE gel
190
were transferred to nitrocellulose membrane. Protein transfer was run for 2.5 hours at
constant 400 mA. Complete transfer of pre-stained protein ladder was used as an
indication of successful protein transfer. Afterwards, the nitrocellulose membrane was
blocked with 3% w/v BSA (Pierce), in TBS-T (0.1% Tween-20 in Tris-buffered saline)
buffer at room temperature for 1 hour.
To immunostain with rabbit anti-HA primary antibody (Invitrogen) or mouse anti-FLAG
(Sigma) primary antibody, the membranes were incubated with antibody (1:1000 dilution)
in TBS-T buffer at room temperature for 1 hour, then rinsed with TBS-T buffer 4 x 5
minutes. The membranes were then incubated with anti-rabbit horseradish peroxidase
conjugate or goat anti-mouse (BioRad, 1:5000 dilution) in TBS-T buffer at room
temperature for 1 hour. Membranes were washed 4 x 5 minutes with TBS-T, then
maintained in TBS-T at room temperature for imaging. The membranes were developed
using Supersignal West Pico substrate (Pierce) and visualized on an Alpha Innotech
Chemilmager 5500 instrument.
191
References
1. Stumpf MPH, Thorne T, Silva E de, Stewart R, An HJ, Lappe M, et al. Estimating the
size of the human interactome. PNAS. 2008 May 13;105(19):6959-64.
2. Velasco-Garcia R, Vargas-Martinez R. The study of protein-protein interactions in
bacteria. Can. J. Microbiol. 2012 Nov;58(11):1241-57.
3. Fields S, Song 0. A novel genetic system to detect protein-protein interactions. ,
Published online: 20 July 1989; doi:10.1038/340245a0. 1989 Jul 20;340(6230):245-
6.
4. Fields S. Interactive learning: Lessons from two hybrids over two decades.
PROTEOMICS. 2009;9(23):5209-13.
5. Stelzl U, Worm U, Lalowski M, Haenig C, Brembeck FH, Goehler H, et al. A Human
Protein-Protein Interaction Network: A Resource for Annotating the Proteome. Cell.
2005 Sep 23;122(6):957-68.
6. Lim J, Hao T, Shaw C, Patel AJ, Szab6 G, Rual J-F, et al. A Protein-Protein
Interaction Network for Human Inherited Ataxias and Disorders of Purkinje Cell
Degeneration. Cell. 2006 May 19;125(4):801-14.
7. Kaltenbach LS, Romero E, Becklin RR, Chettier R, Bell R, Phansalkar A, et al.
Huntingtin Interacting Proteins Are Genetic Modifiers of Neurodegeneration. PLoS
Genet. 2007 May 3(5).
8. Caufield JH, Sakhawalkar N, Uetz P. A comparison and optimization of yeast two-
hybrid systems. Methods. 2012 Dec;58(4):317-24.
9. Suter B, Kittanakom S, Stagljar I. Two-hybrid technologies in proteomics research.
Current Opinion in Biotechnology. 2008 Aug;19(4):316-23.
10. Shekhawat SS, Ghosh I. Split-protein systems: beyond binary protein-protein
interactions. Current Opinion in Chemical Biology. 2011 Dec;15(6):789-97.
11. Hu C-D, Kerppola TK. Simultaneous visualization of multiple protein interactions in
living cells using multicolor fluorescence complementation analysis. Nat Biotech.
2003 May;21(5):539-45.
12. Pelletier JN, Campbell-Valois F-X, Michnick SW. Oligomerization domain-directed
reassembly of active dihydrofolate reductase from rationally designed fragments.
PNAS. 1998 Oct 13;95(21):12141-6.
13. Galarneau A, Primeau M, Trudeau L-E, Michnick SW. p-Lactamase protein fragment
complementation assays as in vivo and in vitro sensors of protein-protein interactions.
Nat Biotech. 2002 Jun;20(6):619-22.
192
14. Remy I, Michnick SW. A highly sensitive protein-protein interaction assay based on
Gaussia luciferase. Nat Meth. 2006 Dec;3(12):977-9.
15. Hatanaka Y, Sadakane Y. Photoaffinity Labeling in Drug Discovery and
Developments: Chemical Gateway for Entering Proteomic Frontier. Current Topics
in Medicinal Chemistry. 2002 Mar 1;2(3):271-88.
16. Chin JW, Cropp TA, Anderson JC, Mukherji M, Zhang Z, Schultz PG. An Expanded
Eukaryotic Genetic Code. Science. 2003 Aug 15;301(5635):964-7.
17. Chou C, Uprety R, Davis L, Chin JW, Deiters A. Genetically encoding an aliphatic
diazirine for protein photocrosslinking. Chem. Sci. 2011 Feb 14;2(3):480-3.
18. Liu W, Brock A, Chen S, Chen S, Schultz PG. Genetic incorporation of unnatural
amino acids into proteins in mammalian cells. Nat Meth. 2007 Mar;4(3):239-44.
19. Hino N, Okazaki Y, Kobayashi T, Hayashi A, Sakamoto K, Yokoyama S. Protein
photo-cross-linking in mammalian cells by site-specific incorporation of a
photoreactive amino acid. Nat Meth. 2005 Mar;2(3):201-6.
20. Hino N, Oyama M, Sato A, Mukai T, Iraha F, Hayashi A, et al. Genetic Incorporation
of a Photo-Crosslinkable Amino Acid Reveals Novel Protein Complexes with GRB2
in Mammalian Cells. Journal of Molecular Biology. 2011 Feb 18;406(2):343-53.
21. Wang L, Xie J, Schultz PG. Expanding the Genetic Code. Annual Review of
Biophysics and Biomolecular Structure. 2006;35(1):225-49.
22. De Graaf AJ, Kooijman M, Hennink WE, Mastrobattista E. Nonnatural amino acids
for site-specific protein conjugation. Bioconjug. Chem. 2009 Jul;20(7):1281-95.
23. Knowles JR. Photogenerated reagents for biological receptor-site labeling. Acc.
Chem. Res. 1972 Apr 1;5(4):155-60.
24. Voskresenska V, Wilson RM, Panov M, Tarnovsky AN, Krause JA, Vyas S, et al.
Photoaffinity Labeling via Nitrenium Ion Chemistry: Protonation of the Nitrene
Derived from 4-Amino-3-nitrophenyl Azide to Afford Reactive Nitrenium Ion Pairs.
J. Am. Chem. Soc. 2009 Aug 19;131(32):11535-47.
25. Brunner J. New photolabeling and crosslinking methods. Annu. Rev. Biochem.
1993;62:483-514.
26. Doering W vo. E, Odum RA. Ring enlargement in the photolysis of phenyl azide.
Tetrahedron. 1966;22(1):81-93.
27. Spanggord RJ, Beal PA. Site-specific modification and RNA crosslinking of the
RNA-binding domain of PKR. Nucleic Acids Res. 2000 May 1;28(9):1899-905.
193
28. Tanaka Y, Bond MR, Kohler JJ. Photocrosslinkers illuminate interactions in living
cells. Mol. BioSyst. 2008 May 20;4(6):473-80.
29. Baruah H, Puthenveetil S, Choi Y-A, Shah S, Ting AY. An Engineered Aryl Azide
Ligase for Site-Specific Mapping of Protein-Protein Interactions through Photo-
Cross-Linking. Angewandte Chemie International Edition. 2008;47(3 7):7018-21.
30. Brunner J, Senn H, Richards FM. 3-Trifluoromethyl-3-phenyldiazirine. A new
carbene generating group for photolabeling reagents. J. Biol. Chem. 1980 Apr
25;255(8):3313-8.
31. Bayley H. Photogenerated Reagents in Biochemistry and Molecular Biology. Elsevier;
2000.
32. Dubinsky L, Krom BP, Meijler MM. Diazirine based photoaffinity labeling.
Bioorganic & Medicinal Chemistry. 2012 Jan 15;20(2):554-70.
33. Hosoya T, Hiramatsu T, Ikemoto T, Aoyama H, Ohmae T, Endo M, et al. Design of
dantrolene-derived probes for radioisotope-free photoaffinity labeling of proteins
involved in the physiological Ca2+ release from sarcoplasmic reticulum of skeletal
muscle. Bioorganic & Medicinal Chemistry Letters. 2005 Mar 1;15(5):1289-94.
34. Yu S-H, Wands AM, Kohler JJ. Photoaffinity probes for studying carbohydrate
biology. J Carbohydr Chem. 2012;31(4-6):325-52.
35. Yu S-H, Boyce M, Wands AM, Bond MR, Bertozzi CR, Kohler JJ. Metabolic
labeling enables selective photocrosslinking of O-GlcNAc-modified proteins to their
binding partners. PNAS. 2012 Mar 27;109(13):4834-9.
36. Dorman G, Prestwich GD. Benzophenone Photophores in Biochemistry.
Biochemistry. 1994 May 1;33(19):5661-73.
37. Pavlova 0, Peterson JH, Ieva R, Bernstein HD. Mechanistic link between p barrel
assembly and the initiation of autotransporter secretion. Proc. Natl. Acad. Sci. U.S.A.
2013 Mar 5;110(l0):E938-947.
38. Sato S, Mimasu S, Sato A, Hino N, Sakamoto K, Umehara T, et al. Crystallographic
Study of a Site-Specifically Cross-Linked Protein Complex with a Genetically
Incorporated Photoreactive Amino Acid,. Biochemistry. 2011 Jan 18;50(2):250-7.
39. Bremer AA, Leeman SE, Boyd ND. Evidence for Spatial Proximity of Two Distinct
Receptor Regions in the Substance P (SP)-Neurokinin-1 Receptor (NK-1R) Complex
Obtained by Photolabeling the NK-1R withp-Benzoylphenylalanine3-SP. J. Biol.
Chem. 2001 Jun 22;276(25):22857-61.
40. Lequin 0, Bolbach G, Frank F, Convert 0, Girault-Lagrange S, Chassaing G, et al.
Involvement of the Second Extracellular Loop (E2) of the Neurokinin- 1 Receptor in
194
the Binding of Substance P PHOTOAFFINITY LABELING AND MODELING
STUDIES ,. J. Biol. Chem. 2002 Jun 21;277(25):22386-94.
41. Pham V, Dong M, Wade JD, Miller LJ, Morton CJ, Ng H, et al. Insights into
Interactions between the a-Helical Region of the Salmon Calcitonin Antagonists and
the Human Calcitonin Receptor using Photoaffinity Labeling. J. Biol. Chem. 2005
Aug 5;280(31):28610-22.
42. Wittelsberger A, Thomas BE, Mierke DF, Rosenblatt M. Methionine acts as a
"magnet" in photoaffinity crosslinking experiments. FEBS Letters. 2006 Mar
20;580(7):1872-6.
43. Bisello A, Adams AE, Mierke DF, Pellegrini M, Rosenblatt M, Suva LJ, et al.
Parathyroid Hormone-Receptor Interactions Identified Directly by Photocross-linking
and Molecular Modeling Studies. J. Biol. Chem. 1998 Aug 28;273(35):22498-505.
44. Farrell IS, Toroney R, Hazen JL, Mehl RA, Chin JW. Photo-cross-linking interacting
proteins with a genetically encoded benzophenone. Nat Meth. 2005 May;2(5):377-84.
45. Ali ST, Guest JR. Isolation and characterization of lipoylated and unlipoylated
domains of the E2p subunit of the pyruvate dehydrogenase complex of Escherichia
coli. Biochem J. 1990 Oct 1;271(1):139-45.
46. Green DE, Morris TW, Green J, Cronan JE, Guest JR. Purification and properties of
the lipoate protein ligase of Escherichia coli. Biochem J. 1995 Aug 1;309(Pt 3):853-
62.
47. Fujiwara K, Toma S, Okamura-Ikeda K, Motokawa Y, Nakagawa A, Taniguchi H.
Crystal Structure of Lipoate-Protein Ligase A from Escherichia coli
DETERMINATION OF THE LIPOIC ACID-BINDING SITE. J. Biol. Chem. 2005
Sep 30;280(39):33645-51.
48. Fujiwara K, Maita N, Hosaka H, Okamura-Ikeda K, Nakagawa A, Taniguchi H.
Global Conformational Change Associated with the Two-step Reaction Catalyzed by
Escherichia coli Lipoate-Protein Ligase A. J. Biol. Chem. 2010 Mar
26;285(13):9971-80.
49. Uttamapinant C, White KA, Baruah H, Thompson S, Femndez-Sui'rez M,
Puthenveetil S, et al. A fluorophore ligase for site-specific protein labeling inside
living cells. PNAS. 2010 Jun 15;107(24):10914-9.
50. Slavoff SA, Liu DS, Cohen JD, Ting AY. Imaging Protein-Protein Interactions inside
Living Cells via Interaction-Dependent Fluorophore Ligation. J. Am. Chem. Soc.
2011 Dec 14;133(49):19769-76.
51. Jin X, Uttamapinant C, Ting AY. Synthesis of 7-Aminocoumarin by Buchwald-
Hartwig Cross Coupling for Specific Protein Labeling in Living Cells.
ChemBioChem. 201 1;12(1):65-70.
195
52. Cohen JD, Thompson S, Ting AY. Structure-Guided Engineering of a Pacific Blue
Fluorophore Ligase for Specific Protein Imaging in Living Cells. Biochemistry. 2011
Sep 27;50(38):8221-5.
53. Puthenveetil S, Liu DS, White KA, Thompson S, Ting AY. Yeast Display Evolution
of a Kinetically Efficient 13-Amino Acid Substrate for Lipoic Acid Ligase. J. Am.
Chem. Soc. 2009 Nov 18;131(45):16430-8.
54. Stadtman ER, Levine RL. Free radical-mediated oxidation of free amino acids and
amino acid residues in proteins. Amino Acids. 2003 Dec 1;25(3-4):207-18.
55. Hensley K. Detection of Protein Carbonyls by Means of Biotin Hydrazide-
Streptavidin Affinity Methods. In: Kurien BT, Scofield RH, editors. Protein Blotting
and Detection [Internet]. Humana Press; 2009 [cited 2013 Apr 15]. p. 457-62.
Available from: http://link.springer.com/protocol/10.1007/978-1-59745-542-8_46
56. Hensley K, Williamson KS. Protein Carbonyl Determination Using Biotin Hydrazide.
Methods in Biological Oxidative Stress [Internet]. New Jersey: Humana Press; [cited
2013 Apr 15]. p. 195-200. Available from:
http://www.springerprotocols.com/Abstract/doi/10.1385/1-59259-424-7:195
57. Ye S, K6hrer C, Huber T, Kazmi M, Sachdev P, Yan ECY, et al. Site-specific
Incorporation of Keto Amino Acids into Functional G Protein-coupled Receptors
Using Unnatural Amino Acid Mutagenesis. J. Biol. Chem. 2008 Jan 18;283(3):1525-
33.
58. Brustad EM, Lemke EA, Schultz PG, Deniz AA. A General and Efficient Method for
the Site-Specific Dual-Labeling of Proteins for Single Molecule Fluorescence
Resonance Energy Transfer. J. Am. Chem. Soc. 2008 Dec 31;130(52):17664-5.
59. Kalia J, Raines RT. Hydrolytic Stability of Hydrazones and Oximes. Angewandte
Chemie International Edition. 2008;47(39):7523-6.
60. Zeng Y, Ramya TNC, Dirksen A, Dawson PE, Paulson JC. High-efficiency labeling
of sialylated glycoproteins on living cells. Nat Meth. 2009 Mar;6(3):207-9.
61. Liang J, Choi J, Clardy J. Refined structure of the FKBP12-rapamycin-FRB ternary
complex at 2.2 A resolution. Acta Crystallographica Section D Biological
Crystallography. 1999 Apr 1;55(4):736-44.
62. Choi J, Chen J, Schreiber SL, Clardy J. Structure of the FKBP12-rapamycin complex
interacting with the binding domain of human FRAP. Science. 1996 Jul
12;273(5272):239-42.
63. Banaszynski LA, Liu CW, Wandless TJ. Characterization of the
FKBP.rapamycin.FRB ternary complex. J. Am. Chem. Soc. 2005 Apr
6;127(13):4715-21.
196
64. Dormin G, Prestwich GD. Benzophenone photophores in biochemistry. Biochemistry.
1994 May 17;33(19):5661-73.
65. Hassan AQ, Stubbe J. Mapping the subunit interface of ribonucleotide reductase
(RNR) using photo cross-linking. Bioorganic & Medicinal Chemistry Letters. 2008
Nov 15;18(22):5923-5.
66. Tsubomura H, Yamamoto N, Tanaka S. Transient absorption spectra of
benzophenone studied by the flash excitation. Chemical Physics Letters. 1967
Oct; 1(8):309-10.
67. Wierzchowski KL. Intramolecular electron transfer between tryptophan radical and
tyrosine in oligoproline-bridged model peptides and hen egg-white lysozyme. Acta
Biochim. Pol. 1997;44(4):627-44.
68. Kalcheva N, Albala J, O'Guin K, Rubino H, Garner C, Shafit-Zagardo B. Genomic
structure of human microtubule-associated protein 2 (MAP-2) and characterization of
additional MAP-2 isoforms. Proc Natl Acad Sci U S A. 1995 Nov 21;92(24):10894-8.
69. Yao JZ, Uttamapinant C, Poloukhtine A, Baskin JM, Codelli JA, Sletten EM, et al.
Fluorophore Targeting to Cellular Proteins via Enzyme-Mediated Azide Ligation and
Strain-Promoted Cycloaddition. J. Am. Chem. Soc. 2012 Feb 29;134(8):3720-8.
70. Ayscough KR. Coupling actin dynamics to the endocytic process in Saccharomyces
cerevisiae. Protoplasma. 2005 Oct;226(1-2):81-8.
71. Pedersen S., Hoffmann E., Mills J. The cytoskeleton and cell volume regulation.
Comparative Biochemistry and Physiology Part A: Molecular & Integrative
Physiology. 2001 Oct;130(3):385-99.
72. Nolen BJ, Tomasevic N, Russell A, Pierce DW, Jia Z, McCormick CD, et al.
Characterization of two classes of small molecule inhibitors of Arp2/3 complex.
Nature. 2009 Aug 20;460(7258):1031-4.
73. Wada A, Fukuda M, Mishima M, Nishida E. Nuclear export of actin: a novel
mechanism regulating the subcellular localization of a major cytoskeletal protein.
EMBO J. 1998 Mar 16;17(6):1635-41.
74. Effects of cytochalasin and phalloidin on actin. J Cell Biol. 1987 Oct 1;105(4):1473-
8.
75. Yarmola EG, Somasundaram T, Boring TA, Spector I, Bubb MR. Actin-latrunculin
A structure and function. Differential modulation of actin-binding protein function by
latrunculin A. J. Biol. Chem. 2000 Sep 8;275(36):28120-7.
76. Liu DS, Tangpeerachaikul A, Selvaraj R, Taylor MT, Fox JM, Ting AY. Diels-Alder
Cycloaddition for Fluorophore Targeting to Specific Proteins inside Living Cells. J.
Am. Chem. Soc. 2012 Jan 18;134(2):792-5.
197
77. Fernaindez-Suirez M, Baruah H, Martinez-Hemndez L, Xie KT, Baskin JM,
Bertozzi CR, et al. Redirecting lipoic acid ligase for cell surface protein labeling with
small-molecule probes. Nature Biotechnology. 2007;25(12):1483-7.
198
Chapter 5. Effort towards membrane protein topology mapping with an ascorbate
peroxidase enzyme
We would like to thank Dawei Jiang of the Clapham lab for providing us with LETMI-
mCherry plasmid in pEGFP-N 3 vector.
199
Introduction
Membrane proteins are important players in the cell and carry out crucial
functions from acting as small molecule transporters to taking roles in complex signaling
pathways. It has been estimated that more than half of the drugs on the market are
directed against membrane protein receptors (1). However, membrane proteins are still
very underrepresented among solved atomic protein structures, making up less than 1%
of the structures available in the Protein Data Bank (PDB) (1). The reason for the lack of
membrane protein structures lies in the technical challenges associated with expressing
them in large enough quantities for structural analysis, solubilizing them in appropriate
detergents, and crystallizing them for structure determination. Aside from solving an
entire three-dimensional structure of a membrane protein, knowing the topology - the
transmembrane regions and the in-out orientation across the membrane - of a membrane
protein can also provide important clues as to the protein's function, potential interaction
partners, and trafficking. Here, we describe our preliminary effort towards developing a
method that can be used to study the topology of membrane proteins.
Membrane proteins are difficult to mass-produce using the standard protein
production workhorse, E. coli. Because the bacterial machinery is quite different from
that of mammalian cells, researchers need to identify suitable bacterial growth conditions
that can induce proper folding of membrane proteins. Recent advances in cell-free
systems of protein translation have enabled the production of tens of milligrams of
membrane proteins (2), but the challenge is still to keep them soluble for subsequent
purification and analysis. Even when expressed at high levels, membrane proteins are
inherently difficult to purify to extremely high purity. In addition, because of the large
hydrophobic regions in membrane proteins, they are often prone to aggregation, making
it necessary to screen and test numerous stabilizing buffers and detergent solutions to
keep these proteins soluble. Solid-state NMR spectroscopy methods, which avoid the
requirement for crystallization, are rapidly advancing, but sensitivity and resolution
enhancements are still needed to make this a robust technology (3,4).
Since there are no general and reliable methods to obtain topological information
of membranes proteins, computational prediction techniques and biochemical methods
are also used to acquire structural information. Computational methods to predict
200
membrane protein topology or putative transmembrane region have been very helpful,
but the precision of these methods is limited by the amount of available structures (5,6).
Various biochemical techniques have been instrumental in elucidating membrane protein
topology and orientation. One method introduces enzymes such as alkaline phosphatase,
p-galactosidase and p-lactamase as reporters by fusing them to specific regions of the
membrane protein (7). These enzymes often have different stabilities and activities
depending on which compartment of the cell they are exposed. Alkaline phosphatase is
active upon the formation of four disulfide bonds and dimer formation in an oxidizing
environment (8). The presence of an active alkaline phosphatase, as measured by its
activity towards a chromogenic substrate, can provide evidence that a region of the
membrane protein is localized to an oxidizing environment. The enzymes B-galactosidase
and p-lactamase work in a similar and complementary fashion. The p-galactosidase
located in the cytoplasm is properly folded and active, while p-galactosidase fused
downstream of an export signal is not folded properly and therefore is inactive (9). For $-
lactamase, if the enzyme is fused to the portion of the protein that is exposed to the
outside of a bacterial cell, the enzyme has the ability to protect the cells against antibiotic
such as ampicillin, which otherwise would inactivate essential proteins anchored on the
outside of the cytoplasmic membrane that are crucial for cell wall biosynthesis. All three
enzymes are applicable to heterologously expressed eukaryotic membrane proteins.
However, they are not generalizable to all compartmental membranes in the cell. For
example, alkaline phosphatase is only applicable if the membrane separates between an
oxidizing and a reducing environment (7).
Aside from fusing enzymes to various regions of membrane proteins, researchers
have often used exogenous proteases to assay for membrane protein topology. Integral
membrane proteins cross the membrane either in a single-pass or a multi-pass fashion
that exposes their hydrophilic loops to one of the two compartments separated by the
membrane. The impermeability of the membrane allows membrane-impermeant small
molecules and proteases to be utilized to verify putative transmembrane structures. (10).
The most common practice is to externally supply promiscuous proteolytic enzymes that
cannot cross the membrane, then determine which regions of the membrane protein are
cleaved or protected from proteolysis. Regions of cleavage can be analyzed by
201
immunoblot using antibodies directed against the protein of interest. However, protease
accessibility assays may fail to identify exposed regions of the membrane protein,
especially if the region is small (11).
The goal of this project is to develop a microscopy-based imaging methodology to
study membrane protein topology. The method uses an enzymatic labeling approach in
which the enzyme is fused to different regions of a membrane protein of interest. The
enzyme is an engineered plant ascorbate peroxidase enzyme named APEX developed in
our lab that has been utilized as a genetic tag for EM imaging (12) as well as in proteomic
mapping inside live cells (13). In EM imaging, APEX oxidizes the substrate 3, 3'-
diaminobenzidine (DAB) in an H20 2-dependent manner to form DAB polymers that can
serve as an EM contrast agent after treatment with OsO4. Martell et al. utilized the
localized and membrane restricted feature of DAB staining to generate EM images of
various cellular proteins fused to APEX. Based on the location of the EM contrast,
Martell and coworkers were able to demonstrate that both termini of the mitochondrial
calcium uniporter (MCU), an inner mitochondrial membrane (IMM) protein, face the
mitochondrial matrix. Even though EM imaging provides nanometer resolution, the
instrumentation and specific skills required to perform EM imaging are not always
available or accessible to all research labs. Therefore, we wanted to develop an
alternative method that is able to use microscopy imaging of the APEX labeling patterns
to provide clues to the topology of membrane proteins.
The imaging method is based on APEX labeling of a phenol small molecule,
biotin-tyramide. In addition to using DAB as a substrate, APEX can also oxidize phenol
to phenoxyl radicals in an H20 2-dependent manner. When live mammalian cells
expressing APEX targeted to a specific cellular compartment, such as the mitochondrial
matrix, were incubated with biotin-tyramide and H20 2, endogenous proteins were
covalently modified with biotin only in the compartment where APEX was targeted (13).
In proteomic mapping, Rhee et al. targeted APEX to the mitochondrial matrix or
intermembrane space (IMS) and performed biotin-tyramide labeling by adding H20 2 for
1 minute (13). The short lifetime of phenoxyl radicals allow biotin-tyramide radicals to
only react with APEX-proximal proteins. After live cell labeling, biotinylated proteins in
202
the APEX-specific cellular compartment were enriched and analyzed by tandem mass
spectrometry.
Dr. Hyun-woo Rhee subsequently discovered that not only did DAB labeling by
APEX produce a localized signal that delineated cellular ultrastructures, APEX labeling
using biotin-tyramide followed by staining with streptavidin-fluorophore conjugate also
showed different labeling patterns depending on the location of the APEX enzyme.
Knowing from previous experiments that the biotin-tyramide radical does not cross the
membrane (13), Dr. Rhee observed that if APEX is localized in a strictly membrane-
bound compartment, such as the mitochondrial matrix, the pattern of biotin-tyramide
labeling appeared to be tight and restricted, delineating the mitochondria of the cell.
However, if APEX is in the cytosol or in places where the biotin-tyramide radical can
easily diffuse during the time of labeling, such as the mitochondrial intermembrane space
(IMS) because of the porous outer mitochondrial membrane (OMM), biotin-tyramide
labeling pattern appeared diffusive and throughout the cell. We wanted to use the
location-specific APEX biotinylation pattern as a tool to infer membrane protein
topology (Figure 5-1). APEX fused to the matrix side of an inner mitochondrial
membrane (IMM) protein would show a restricted and mitochondrial-specific labeling
pattern because matrix is enclosed by the tight IMM (Figure 5-1, left). In contrast, APEX
fused to the IMS side of an IMM protein would show a diffusive biotinylation pattern
because the outer mitochondrial membrane (OMM) is porous (Figure 5-1, right),
allowing biotin-tyramide radical to diffuse. Depending on the enzymatic fusion site of
APEX to the membrane protein of interest, the labeling pattern can appear either
restricted inside certain cellular organelle or diffusive throughout the entire cell. The
restricted or diffusive labeling patterns can provide topological clues to the location of
certain regions of a membrane protein.
203
cytosol
memrane rn
Intermembrane
space
membrane
matrix e *n pater
Restricted labeling pattern
CAPEX
9 biotin-tyramide radical
A
-1
Difusive labeling pattem
Figure 5-1. Scheme depicting labeling with APEX (green pacman) fused to either matrix
or IMS side of an IMM protein. Biotin-tyramide radical (red dot) generated in the matrix
would remain localized there because the IMM is very tight and impassable, therefore
generating what appeared to be mitochondria-localized biotinylation signal on a
conventional microscope. In contrast, biotin-tyramide radical generated in the IMS
would have the opportunity to diffuse across the porous OMM and escape into the
cytosol of the cell, leading to a diffusive or whole-cell labeling signal.
We wished to fuse APEX to different regions of a membrane protein and apply
APEX labeling of biotin-tyramide to develop a technique for membrane protein topology
prediction. The method is easy to implement and requires only a basic microscope to
observe tight or diffusive labeling pattern. In this chapter, we describe the development
and effort towards applying APEX labeling to investigate the membrane topology of a
mitochondrial inner membrane protein LETM1.
204
see
EF-hand domain-containing protein 1 - LETM1
In the work of Rhee and coworkers, mitochondrial matrix-targeted APEX was
used to label and enrich matrix localized proteins (13). By analyzing the 495 proteins
identified in the matrix proteome, we focused on the 83 kDa mitochondrial inner
membrane protein EF-hand domain-containing protein 1 (LETM1) because it was highly
enriched in the mitochondrial matrix proteome with two biotinylated peptides identified
(13). The first biotinylated peptide, KLEEGGPVYSPPAEVVVK (biotin-tyramide
modified tyrosine residue is in red) in LETM1 mapped to amino acid residues 133-150,
which corresponds to a region upstream from the putative transmembrane domain (amino
acids 206-229) (Figure 5-2). The second biotinylated peptide in LETMI
(TGEEKYVEESK) is located near the C-terminus of the protein, from amino acid
residues 593-603 (Figure 5-2). Because the biotin-tyramide radical does not cross the
membrane, both biotinylated peptides of LETMI must be located in the mitochondrial
matrix. This result is interesting because under the assumption that the N-terminus of
LETM1 is located in the matrix, there would have to be at least a second or possibly more
transmembrane domains in order for the second biotinylated peptide near the C-terminus
to be located in the matrix. The topology of LETM1 has significant implications on the
protein's proposed functions (11,14-16). Therefore, we decided to study the topology of
LETMI in the inner mitochondrial membrane (IMM).
In 1999, Endele et al. first cloned and characterize the LETMI gene, which was
identified as a chromosomally deleted gene in almost all patients with Wolf-Hirschhom
syndrome (WHS), a genetic disease characterized by growth delay, seizures, mental
retardation, and characteristic facial features (17). Computational sequence analysis of
LETM1 shows its sequence identity ranges from 31% in S. cerevisiae to 83% in mouse
(18). Putative structural features of human LETMI include a 24-amino acid
transmembrane region at the N-terminus; a highly conserved PKC phosphorylation site;
two EF-binding motifs near the C-terminus; and several coiled-coil regions (15,18)
(Figure 5-2). Studies by Schlickum et al. further demonstrated that a mitochondrial
targeting signal sequence is located at the N-terminus, without which LETMI appeared
to be mis-localized to the cytosol (18).
205
MASILLRSCRGRAPARLPPP PRYTVPRGSPGDPAHLSCAS TLGLRNCLNVPFGCCTPIHP 60
VYTSSRGDHLGCWALRPECL RIVSRAPWTSTSVGFVAVGP QCLPVRGWHSSRPVRDDSVV 120
EKSLKSLKDKNKKLEEGGPV YSPPAEVVVKKSLGQRVLDE LKHYYHGFRLLWIDTKIAAR 180
MLWRILNGHSLTRRERRQFL RICADLFRLVPFLVFVVVPF MEFLLPVAVKLFPNMLPSTF 240
ETQSLKEERLKKELRVKLEL AKFLQDTIEEMALKNKAAKG SATKDFSVFFQKIRETGERP 300
SNEEIMRFSKLFEDELTLDN LTRPQLVALCKLLELQSIGT NNFLRFQLTMRLRSIKADDK 360
LIAEEGVDSLNVKELQAACR ARGMRALGVTEDRLRGQLKQ WLDLHLHQEIPTSLLILSRA 420
MYLPDTLSPADQLKSTLQTL PEIVAKEAQVKVAEVEGEQV DNKAKLEATLQEEAAIQQEH 480
REKELQKRSEVAKDFEPERV VAAPQRPGTEPQPEMPDTVL QSETLKDTAPVLEGLKEEEI 540
TKEEIDILSDACSKLQEQKK SLTKEKEELELLKEDVQDD' IKLSKTGEEKYVE 600
ESKASKRLTKRVQQMIGQID GLISQLEMDQQAGKLAPANG MPTGENVISVAELINAMKQV 660
KHIPESKLTSLAAALD _ 1' 2 VKVIELVDKEDV HISTSQVAEIVATLEKEEKV 720
EEKEKAKEKAEKEVAEVKS 739
Figure 5-2. Amino acid sequence of human LETM1. Orange indicates conserved PKC
phosphorylation site; light blue indicates putative transmembrane region; underlining
represents two coiled-coil regions; green indicates two 13-amino acids EF-hand binding
motifs. Two biotinylated peptides labeled by matrix-targeted APEX and identified by
tandem MS in matrix proteome (13) were in dark blue with biotinylated tyrosine residues
highlighted in red.
Two yeast homologs of LETM1, Mdm38 and Ylh47, have been studied in depth
to gain insight into the role of LETM1 (15,16,18,19). They do not contain the EF-binding
motifs present in human LETMI but otherwise share high protein sequence identity
especially near the N-terminus of the proteins (15). Both yeast homolog proteins are
localized to the mitochondrial inner membrane as shown by subcellular fractionation
(19,20), protease accessibility assays (19), and alkaline extraction of membrane proteins
(15,19). Functional analysis of Mdm38 further revealed that deletion of this gene in yeast
caused elevated intramitochondrial potassium level, swollen mitochondria, and decreased
membrane potential (19). Numerous functional roles have been proposed for LETMI,
from maintaining mitochondrial volume (20) and shape (21), acting as a Ca2 /H
antiporter (14), regulating mitochondrial fragmentation (22), to aiding mitochondrial
protein export (15) and regulating mitochondrial ATP production (23). However, despite
the many functional implications of this protein, the exact correlation between LETM1
206
and WHS, as well as how it is transported into the mitochondria and its topology in the
IMM are still unclear.
Experiments investigating the topology of LETMI in the IMM revolved around
subcellular fractionation of the mitochondria and protease accessibility assay. A number
of works agree that LETMI is an IMM protein with no detectable part exposed to the
IMS (15,19,24). However, Hasegawa et al. suggest that LETMI is partially exposed to
the IMS because when the IMM was compromised by the addition of digitonin, matrix
proteins like Hsp60 and GDH were protected from trypsin digestion, but LETMI was
partially digested (20). The work of Dimmer et al. shows a proposed membrane
orientation of LETM 1 as a single transmembrane IMM protein with its N-terminus in the
IMS (24). Yet, they were unable to detect the IMS-exposed region using a proteinase K
accessibility assay. In addition to the insensitivity of most protease accessibility or
protection assays, subcellular fractionations are often prone to contamination (24), further
leading to misguided or contradictory conclusions. We hope to provide an alternative
method based on our APEX/biotin-tyramide labeling assay to infer the topology of
LETM1.
207
Restricted and diffusive labeling patterns of localized APEX
To first demonstrate the restricted and diffusive labeling patterns, we performed
biotin-tyramide labeling in live HEK 293T and COS-7 cells expressing either APEX
localized to the mitochondrial matrix or the IMS. We expected to see a tight labeling
pattern when APEX is in the mitochondrial matrix because the matrix is a tightly
membrane-bound compartment. In contrast, we expected to see a diffusive labeling
pattern with signal spread out through the cell when APEX is in the IMS, which is not
tightly bound region because the OMM is very porous. We incubated the cells in cell
culture media containing biotin-tyramide for 1 hour at 37 'C. We then added 1 mM H202
to the cells to initiate labeling. Following labeling for 1 minute, the cells were fixed with
paraformaldehyde, stained with neutravidin-Alexa Fluor 647 conjugate, and imaged on a
confocal microscope.
In HEK 293T cells, we observed tight and restricted Alexa Fluor 647 signal co-
localizing with mitochondrial green fluorescent protein (GFP) marker when labeling with
APEX in the matrix (Figure 5-3). The labeling pattern highlights the mitochondrial
network, which appears as small dots in HEK 293T cells. In contrast, labeling with
APEX localized to the mitochondrial IMS shows Alexa Fluor 647 signal spread out
throughout the cell. In COS-7 cells, we also saw tight Alexa Fluor 647 signal delineating
the mitochondria and co-localizing with mitochondrial maker. When labeling with APEX
in the IMS of COS-7 cells, Alexa Fluor 647 signal appeared to be fuzzy and spread out in
COS-7 cells (Figure 5-3).
208
matrix-APE:
matrix-APEX
8 Omit H20 2
IMS-APEX
IMS-APEX
Omit H2O2
matrix-APEX
matrix-APEX
Omit H2 02
0
IMS-APEX
IMS-APEX
Omit H2 0 2
Figure 5-3. Restricted and diffusive biotin-tyramide labeling patterns in HEK 293T and
COS-7 cells. HEK 293T and COS-7 cells expressing matrix localized APEX or IMS
localized APEX were labeled with biotin-tyramide and H20 2 . Cells were then fixed and
209
stained with neutravidin (nAv)-Alexa Fluor 647 to visualize biotinylated proteins, and
antibody to visualize v5-epitope tag on the APEX enzyme. Mito-GFP is a mitochondrial
marker. Negative control was performed with H20 2 omitted. Region of the labeled image
boxed in white, dashed box is zoomed in as a larger image on the right.
Having established that APEX localized to either the mitochondrial matrix or the
IMS gave different biotinylation patterns, we moved forward to fuse APEX to various
regions of the inner mitochondrial membrane protein LETMI to investigate which
regions of LETMI are in the matrix or in the IMS.
210
Biotin-tyramide labeling patterns of APEX-fused LETM1
Patterns and levels ofexpression of APEX-LETM fusion proteins
To apply the difference in labeling patterns in deducing LETMI topology in the
mitochondrial inner membrane, we fused APEX to various regions of LETM1 and
performed biotin-tyramide labeling in HEK 293T cells. Based on the amino acid
sequence of human LETMI in Figure 5-2, we chose regions of LETMI that do not have
previous functional designation to fuse the APEX enzyme. We cloned a total of 16
LETM1-APEX fusion constructs (Figure 5-4), all of which have v5 epitope tag at the end
of LETM1. For example, APEX115-LETM1 indicates that APEX is fused after amino
acid 115 of LETMI (Figure 5-4, top). Most of the constructs expressed well with clear
mitochondrial pattern as confirmed by anti-v5 immunostaining and co-localization with
mitochondrial GFP marker (Figure 5-6).
Construct APEXI1 5-LETM1
LETM1
1-115M
LETM1
116-739
N-terminus
Iv~
C-terminus
Construct name APEX Expression pattern/level
insertion after
APEX115-LETM1 aa 115 mitochondrial / unstable
APEX1 56-LETM1 aa 156 mitochondrial / strong
APEX180-LETM1 aa 180 mitochondrial / strong
APEX187-LETM1 aa 187 mitochondrial / strong
APEX201-LETM1 aa 201 mitochondrial / strong
APEX235-LETM1 aa 235 mitochondrial / strong
APEX285-LETM1 aa 285 mitochondrial / strong
APEX300-LETM1 aa 300 mitochondrial / strong
APEX365-LETM1 aa 365 mitochondrial / strong
APEX385-LETM1 aa 385 mitochondrial / strong
APEX444-LETM1 aa 444 mitochondrial I weak
APEX524-LETM1 aa 524 mitochondrial I strong
APEX591-LETM1 aa 591 whole cell / weak
APEX609-LETM1 aa 609 whole cell / strong
APEX657-LETM1 aa 657 whole cell / strong
LETM1-v5-APEX aa 739 mitochondrial / strong
LETM1-v5 - mitochondrial / strong
211
Figure 5-4. Constructs with APEX fused to different regions of LETMI. (Top) Schematic
representation of APEX115-LETMI fusion construct. APEX (total 249 amino acids) is
inserted after amino acid 115 of LETMI. An epitope tag v5 is fused at the C-terminus of
LETMI. (Bottom) List of APEX-LETMI fusion constructs, their expression patterns and
levels of expression. Microscopy images of each fusion construct are shown in Figure 5-6.
A few constructs did not express stably or appeared to be mistargeted. Construct
APEX 115-LETMI sometimes was undetectable by immunostaining on fixed HEK 293T
cells and showed proteolytic products on Western blot (Figure 5-5, see Experimental
methods for transfection and Western blot procedures). This is likely caused by the
APEX fusion site being immediately after the N-terminal mitochondrial targeting region
from amino acids 1 - 115 of LETM1. Schlickum et al. (18) showed that fusion of GFP
before the targeting region or deletion of the targeting region both resulted in diffusive
staining in the cytoplasm, not restricted to the mitochondria of the cell.
//d0PY 
(0
230
150
100 APEX-LETM1 fusion
80
60
50
40
30
kDa
WB anti-v5
Figure 5-5. APEX1 15-LETMI fusion protein shows cleavage products on Western blot.
Immunostaining with anti-v5 antibody detected possible cleavage products (green arrows)
of APEX115-LETMI. APEX167-LETM1 fusion protein does not show the same
212
cleavage products. APEX-LETI fusion protein (~114 kDa) is indicated by black arrow.
Red and blue arrows indicate artifacts caused by protein overexpression, which is
discussed in the following section.
Three constructs with APEX fused near the C-terminus of LETMI, APEX591-
LETMI, APEX609-LETM1, and APEX657-LETM1, did not target to the mitochondria,
but instead appeared to have whole cell expression pattern (see Figure 5-6, anti-v5 AF
568 channel). Interestingly, in all three constructs, APEX is placed between the two EF-
hand binding motifs. The first EF-hand binding motif is from amino acids 577 - 590; the
second EF-hand binding motif is from amino acids 676 - 688. This is an indication that
the region between the two EF-hand binding motifs could be critical to protein folding
and therefore does not tolerate other protein tags. It is also possible that protein fusion in
this region disrupts LETMI folding, making the mitochondrial targeting region at the N-
terminus inaccessible or cannot be recognized by transport proteins or chaperones to
bring LETM1 to the mitochondria, leading to the whole cell expression pattern we
observed with all three constructs.
Biotin-tyramide labeling patterns of APEX-LETMJ fusion proteins
We performed biotin-tyramide labeling in HEK 293T cells expressing the APEX-
LETMI fusion constructs. Since LETMI is a mitochondrial inner membrane protein
(IMM), depending on the fusion sites of the APEX, certain APEX could be in the
mitochondrial matrix, while other APEX could be in the IMS. We expected to see the
fusion proteins give either restricted or diffusive labeling patterns. Based on the labeling
patterns, we can infer the topology of LETMI in the IMM.
We transfected HEK 293T cells with each of the APEX-LETM1 fusion constructs
and a mitochondrial GFP marker using lipofectamine. At 18 hours after transfection, we
incubated the cells in cell culture media containing biotin-tyramide for 1 hour at 37 'C.
We then added 1 mM H2 0 2 to the media to initiate labeling. After 1 minute of labeling,
the cells were washed with Dulbecco's phosphate buffered saline (DPBS) and fixed using
3.7% paraformaldehyde in DPBS. We immunostained the fixed cells with anti-v5
antibody to visualize the expression of each APEX-LETM1 protein and with neutravidin-
213
Alexa Fluor 647 conjugate (nAv-AF 647) to visualize the biotin-tyramide labeled
proteins.
Figure 5-6 shows the biotin-tyramide labeling pattern (nAv-AF 647 channel) and
the expression (mouse anti-v5-AF 568 channel) of each APEX-LETMI fusion construct.
We can also visualize the mitochondrial network via the expression of a mitochondrial
marker, mito-GFP. The first nine APEX-LETMi constructs all showed tight and
restricted labeling pattern, with nAv-AF 647 signal co-localizing with both APEX-
LETMI expression (Anti-v5 AF 568 channel) and the mitochondrial GFP marker. Three
constructs with APEX fused after amino acids 385, 444, and 524 of LETMI all showed
diffusive labeling pattern, with nAv-AF 647 signal spread out throughout the cell.
214
APEX115-LETM1
APEX156-LETM1
APEX180-LETM1
APEX187-LETM1
APEX201-LETM1
APEX235-LETM1
APEX285-LETM1
APEX300-LETM1
215
APEX365-LETM1
APEX385-LETM1
APEX444-LETM1
APEX524-LETM1
APEX591 -LETM1
APEX609-LETM1
APEX657-LETM1
LETM1-v5-APEX
216
Figure 5-6. Biotin-tyramide labeling with APEX-LETMi fusion constructs in HEK 293T
cells. HEK 293T cells expressing the indicated APEX-LETM1 fusion construct and a
mitochondrial marker, mito-GFP were labeled with biotin-tyramide and H20 2. Cells were
then fixed and stained with neutravidin-Alexa Fluor 647 conjugate (nAv-AF 647) to
visualize biotinylated proteins, and with anti-v5 antibody to visualize APEX-LETM1
expression.
Our results from APEX-LETM1 fusion protein labeling provided hints to the
membrane topology of LETM1. First, if we assume the direction of the putative
transmembrane region from amino acids 206 - 229 is going from the matrix to the IMS,
then we would expect APEX235-LETM1 to be located in the IMS since APEX is fused
only 6 amino acids downstream of the transmembrane segment. However, we observed
tight and restricted biotinylation labeling pattern of APEX235-LETM1, co-localizing
with the mitochondrial GFP marker, which indicates that APEX is in the matrix (Figure
5-6). On the other hand, if we assume the direction of the transmembrane region were
going from the IMS to the matrix, then we would expect construct APEX201-LETM1
with APEX fused after amino acid 201, just 5 amino acids upstream of the
transmembrane segment, to be localized in the IMS. However, APEX201-LETM1 also
showed tight labeling pattern (Figure 5-6). This result led us to speculate that perhaps the
putative transmembrane region of LETM1 does not go through the entire IMM. Indeed,
work by Frazier et al. using subcellular fractionation of mitochondria and carbonate
extraction of membrane proteins suggested that the yeast homologs of LETM1, Mdm38
and Ylh47, could be only partially embedded in the lipid environment and may not fully
penetrate the IMM (15). This is in agreement with our imaging result showing APEX
located very near the transmembrane region (amino acids 206 - 229), either upstream in
the case of APEX201-LETM1 or downstream in the case of APEX235-LETM1, both
generated a restricted biotinylation labeling pattern. Our hypothesis is that the putative
transmembrane segment of LETMI might be embedded in IMM, perhaps passing
through only one phospholipid layer of the IMM.
217
Controlling expression levels of LETMI fusion proteins
In order to draw any conclusion regarding LETM1 membrane topology, in
addition to having correct mitochondrial localization of the APEX-LETMi fusion
construct, we must also ensure that the constructs expressed at the expected molecular
weight without artifacts such as protease cleavage. Either overexpression of the protein or
proteolysis could cause false positives or negatives of the labeling signal. Furthermore, if
APEX is cleaved from LETMI, it could diffuse away from the IMM causing false
labeling pattern that does not reflect the localization or topology of LETM1.
We performed Western blot analysis of cell lysate from HEK 293T cells
expressing different APEX fusion constructs of LETMI (Figure 5-7). We observed with
anti-LETMI immunostaining that in addition to the band at the expected molecular
weight (114 kDa for APEX-LETMI fusion protein, 83 kDa for LETMI or LETM1-v5
protein) (Figure 5-7, green and black arrows), each transfected construct also had a
smaller protein species that appeared at either 65 kDa, or 90 kDa if APEX were fused
near the C-terminus of LETMI (Figure 5-8, red arrows). Furthermore, there is also a
larger protein species appearing at about 5 kDa heavier than the expected molecular
weight of each fusion protein (Figure 5-8, blue arrows). These artifacts were also
observed in a parallel experiment performed in COS-7 cells (data not shown).
N. 1 1)
1_ _ 2 3 4 5
230
150
100 + LETM1-APEX fusion protein (-114 kDa)
80 + endogenous LETM1, LETM1-v5 (-83 kDa)
60
50
40
30
kDa
WB anti-LETM1
218
Figure 5-7. Western blot analysis of APEX-LETMi fusion proteins expressed in HEK
293T cells. HEK 293T cells were transfected with the indicated APEX-LETMi fusion
construct using lipofectamine. At 18 hours after transfection, cells were lysed and the
lysates were separated on a 9% SDS-PAGE gel and transferred onto nitrocellulose
membrane for immunostaining with anti-LETM1 antibody (see Experimental methods
for details). Black arrow indicates endogenous LETMI protein (in all lanes) and
transfected LETMl-v5 protein (lane 5) at 83 kDa (LETM1 and LETM1-v5 proteins
cannot be separated on a 9% SDS-PAGE gel). Green arrow indicates APEX-LETM1
fusion proteins (lanes 2, 3, and 4) at 114 kDa. Red arrows indicate possible protease
cleavage products. Blue arrows indicate unknown complex about 5 kDa heavier than
expected molecular weights of the expressed proteins.
Other labs have also observed larger and smaller protein species expressed in
addition to LETM1 (18,25,26). Schlickum and coworkers detected in HEK 293T cells
transfected with LETMI (83 kDa) two extra bands at 95 kDa and 60 kDa (18). Dimmer
and coworkers transfected HeLa cells with LETM1 and saw a lower molecular weight
species (25). Authors speculated that the higher molecular weight protein could be a
result of posttranslational modification, while the lower protein species may represent
protein cleavage product by an unknown protease or degradation product possibly caused
by overexpression of LETM1 (18,25). We further conducted Western blot assays of cell
lysates harvested at either 24 or 72 hours after lipofectamine transfection of APEX-
LETMI fusion constructs. We hoped that perhaps protein expression level would
decrease over time, leading to more controlled expression of LETM1. However, we
observed the same cleavage pattern as shown previously (data not shown).
In order to reduce the expression of transfected LETMI or LETMI fusion protein,
we decided to try a different transfection method. We made lentivirus encoding either
LETM1 or LETM1-mCherry fusion construct. We infected COS-7 cells with low
amounts of viruses. Thereafter, we lysed the cells and analyzed the cell lysates on
Western blot with anti-LETMI immunostaining. Figure 5-8 shows that cells transfected
with LETMI (lane 2) or LETM1-mCherry (lane 6) using lipofectamine transfection
method appeared to have cleavage products (Figure 5-8, red arrows) as well as the
219
unknown larger complexes (blue arrows) as we have seen previously in HEK 293T cells.
In the case of LETM1-mCherry, many more cleavage products were observed when the
construct was transfected with lipofectamine (lane 6, green arrows). In contrast, cells
with viral infected LETM1 (lanes 3-5) or LETM1-mCherry (lanes 7-9) appeared to
generate LETM1 or LETM1-mCherry protein at the expected molecular weight of either
83 kDa or 111 kDa, respectively, without the cleavage products. Viral infection of
LETMI gave the cleanest expression of LETMI using 10 pl of the virus (lane 3). The
upper band of the two bands we observed in Figure 5-8, lanes 4 and 5 are likely the
cytosolic precursor protein of LETM1, before the mitochondrial targeting sequence of
LETMI was cleaved. We plan to perform subsequent experiments with lentiviral
infection of LETMI fusion constructs to avoid overexpression artifacts.
L2k viral infect L2k viral infect
10 pl 50 pl 100 pl 10 pI 50 Pl 100 pI
1 2 3 4 5 6 7 8 9
230
150
100 <- LETM1-mCherry
80 < LETM1
60
50
40
30
25
kDa
WB Anti-LETM1
Figure 5-8. Western blot analysis of COS-7 cells lipofected (L2k) or viral infected with
LETMI or LETM1-mCherry construct. COS-7 cells were either transfected using
lipofectamine (L2k) (lanes 2 and 6) or viral infected using increasing amounts of virus -
10 g1, 50 g1, or 100 pl. Cells were then lysed and analyzed on Western blot. Viral
220
infected cell lysates appeared to have much cleaner expression of LETMI (lanes 3-5) or
LETM1-mCherry protein (lanes 7-9) without cleavage products or unknown larger
protein complex. Lane 1 represents endogenous LETMI protein in COS-7 cells.
221
Conclusion and future direction
In this chapter, we described our effort towards developing an enzymatic labeling
and imaging-based method for assaying membrane protein topology. The method uses
the engineered ascorbate peroxidase enzyme APEX that generates biotin tyramide
radicals in the presence of H20 2 . APEX can be fused to different regions of the
membrane protein of interest. Depending on whether APEX is located in a membrane-
bound compartment, the biotinylation pattern can appear tight and restricted to the
location of APEX or diffusive throughout the cell. The biotinylation patterns can be used
to infer the topology of a membrane protein of interest.
We fused APEX to different regions of the mitochondrial inner membrane protein,
EF-hand domain-containing protein 1, LETMI. Biotin-tyramide labeling in HEK 293T
cells expressing different APEX-LETMI constructs showed some APEX-LETMI fusion
proteins gave restricted biotinylation pattern, co-localizing with the mitochondrial GFP
marker. The tight labeling pattern is indicative of those specific regions of LETMI being
located in the mitochondrial matrix. Three APEX-LETMI proteins gave diffusive
biotinylation patterns, suggesting that those regions could be in the mitochondrial
intermembrane space (IMS). Our initial results indicated that the proposed
transmembrane region of LETM1, from amino acids 206 to 229, may not traverse the
entire phospholipid bilayer of the IMM from matrix to IMS. Instead the proposed
transmembrane region may only be embedded in one leaflet of the phospholipid bilayer,
so the regions upstream (before amino acid 206) and downstream (after amino acid 229)
of the transmembrane region of LETM 1 could be both in the matrix.
However, imaging results with APEX-LETMI fusion proteins would need to be
repeated for fear of overexpression artifacts. Each LETM 1-APEX fusion protein must be
analyzed by Western blot to ensure the expression is free of proteolytic product or other
unknown protein species. We performed lentiviral induction and lipofection transfection
of APEX-LETM 1 constructs and found viral induction expressed the fusion proteins at an
optimal level free of proteolytic products. Subsequent experiments will be performed
with viral infected plasmids to strictly control for fusion protein expression level.
Even though readout from streptavidin-fluorophore staining and imaging using a
conventional microscope would not provide the super-resolution of EM, we believed the
222
different labeling patterns of biotin-tyramide could still be exploited as a useful and
simple tool to provide hints to the localization of certain regions of a membrane protein.
In conclusion, we hope to develop a robust technique based on biotin-tyramide and
APEX labeling that can be routinely used to infer membrane protein topology.
223
Experimental methods
Genetic constructs
Construct name Vector Features and variants References
matrix-APEX pcDNA3 mitochondrial matrix targeting sequence: (13)
MLATRVFSLVGKRAISTSVCVRAH
v5: GKPIPNPLLGLDST (between matrix targeting
sequence and APEX)
IMS-APEX pcDNA3 Targeted to the IMS via the LACTB(-68) leader (13,27)
sequence:
MYRLLSSVTARAAATAGPAWDGGRRGAHRR
PGLPVLGLGWAGGLGLGLGLALGAKLVVGLR
GAVPIQS
v5: GKPIPNPLLGLDST (between LACTB
targeting sequence and APEX)
APEX115-LETMI pcDNA3 APEX located after amino acid 115 of LETMI
v5: GKPIPNPLLGLDST (at C-terminus of LETMI)
APEX156-LETM1 pcDNA3 APEX located after amino acid 156 of LETMI
v5: GKPIPNPLLGLDST (at C-terminus of LETM1)
APEX180-LETM1 pcDNA3 APEX located after amino acid 180 of LETMI
v5: GKPIPNPLLGLDST (at C-terminus of LETMI)
APEX187-LETM1 pcDNA3 APEX located after amino acid 187 of LETM1
v5: GKPIPNPLLGLDST (at C-terminus of LETMI)
APEX201-LETM1 pcDNA3 APEX located after amino acid 201 of LETMI
v5: GKPIPNPLLGLDST (at C-terminus of LETMI)
APEX235-LETM1 pcDNA3 APEX located after amino acid 235 of LETMI
v5: GKPIPNPLLGLDST (at C-terminus of LETMI)
APEX285-LETM1 pcDNA3 APEX located after amino acid 285 of LETMI
v5: GKPIPNPLLGLDST (at C-terminus of LETMI)
APEX300-LETM1 pcDNA3 APEX located after amino acid 300 of LETMI
v5: GKPIPNPLLGLDST (at C-terminus of LETMI)
APEX365-LETM1 pcDNA3 APEX located after amino acid 365 of LETMI
v5: GKPIPNPLLGLDST (at C-terminus of LETMI)
APEX385-LETM1 pcDNA3 APEX located after amino acid 385 of LETMI
v5: GKPIPNPLLGLDST (at C-terminus of LETMI)
APEX444-LETM1 pcDNA3 APEX located after amino acid 444 of LETMI
v5: GKPIPNPLLGLDST (at C-terminus of LETMI)
APEX524-LETM1 pcDNA3 APEX located after amino acid 524 of LETMI
v5: GKPIPNPLLGLDST (at C-terminus of LETMI)
APEX591-LETMI pcDNA3 APEX located after amino acid 591 of LETMI
v5: GKPIPNPLLGLDST (at C-terminus of LETMI)
APEX609-LETM1 pcDNA3 APEX located after amino acid 609 of LETMI
v5: GKPIPNPLLGLDST (at C-terminus of LETMI)
APEX657-LETM1 pcDNA3 APEX located after amino acid 657 of LETMI
v5: GKPIPNPLLGLDST (at C-terminus of LETMI)
LETM1-v5-APEX pcDNA3 APEX located after v5 tag
v5: GKPIPNPLLGLDST (at C-terminus of LETMI)
LETM1-v5 pcDNA3 v5: GKPIPNPLLGLDST (at C-terminus of LETMI)
LETM1-mCherry pEGFP- red fluorescent protein mCherry (at C-terminus of (14)
N3 LETM1)
224
Labeling and imaging of mitochondrial matrix and IMS proteins (Figure 5-3, Figure
5-6)
HEK 293T or COS-7 cells were plated on 5 mm x 5 mm glass coverslips placed within
wells of a 48-well cell culture plate (0.95 cm2 per well) 12-16 hours prior to transfection.
HEK 293T cells were plated on glass coverslips pre-coated with 50 ptg/mL fibronectin
(Millipore) to increase adherence. In general, plated cells were transfected with 100 ng
indicated APEX construct at 50-70% confluency using lipofectamine following
manufacture's protocol. At 18 hours after transfection, cells were incubated in cell culture
media (minimum essential media (MEM) + 10% FBS) containing 500 pM biotin-
tyramide for 1 hour at 37 'C. Afterwards, 1 mM H20 2 was added directly to the cell
media and the reaction was allowed to proceed for 1 minute at room temperature. Cells
were then washed 3x with Dulbecco's phosphate buffered saline (DPBS) and fixed with
3.7% paraformaldehyde in DPBS for 15 minutes at room temperature. Fixed cells were
washed 3x with DPBS followed by protein precipitation with cold methanol for 6
minutes at -20 'C. Fixed cells were washed with DPBS 3x and blocked with 0.5% casein
in DPBS for 1 hour at room temperature.
To visualize biotinylated proteins and the expression of APEX or APEX fusion proteins,
fixed cells were immunostained by first incubating with primary antibody, 4 pig/mL
mouse monoclonal anti-v5 antibody (Sigma-Aldrich) in 1% (w/v) BSA in DPBS for 1
hour at room temperature. Cells were further washed with DPBS and incubated with
secondary antibody, 4 pg/mL goat anti-mouse IgG antibody conjugated to Alexa Fluor
568 (Life Technologies) and also neutravidin-Alexa Fluor 647 conjugate (nAv-AF 647)
in 1% (w/v) BSA in DPBS for 1 hour at room temperature. Cells were then washed 3x
and imaged on a confocal microscope.
Western blot with anti-LETMI antibody to detect LETM1 and APEX-LETM1
fusion proteins (Figure 5-5, Figure 5-7, Figure 5-8)
225
HEK 293T or COS-7 cells were plated in wells of a 6-well cell culture plate (9.6 cm2
/well) and transfected with 200 ng of the indicated APEX-LETMI fusion construct
(Figure 5-5, 5-7), LETM1-v5 (Figure 5-7), LETM1 (Figure 5-8) or LETM1-mCherry
(Figure 5-8) using lipofectamine, following manufacture's protocol.
For viral infection of LETM 1 and LETM 1 -mCherry constructs (Figure 5-8), 10 l, 50 Il,
or 100 pl of the LETMI or LETM1 -mCherry virus (see below for protocol of lentivirus
generation) was added to 1 well (in 6-well plate) of HEK 293T cells in 1 ml of cell
culture media.
At 18 hours after lipofectamine transfection or 48 hours after viral infection, cells were
washed 3x with DPBS and collected in DPBS by gentle pipetting followed by
centrifugation at 200 rpm for 5 minutes at room temperature. Cell pellets were lysed with
100 g1 RIPA lysis buffer (50 mM Tris, 150 mM NaCl, 0.1% SDS, 0.5% sodium
deoxycholate, 1% Triton X-100) supplemented with lx protease cocktail (Sigma
Aldrich), 1 mM PMSF. Lysates (5 gl out of 100 pl lysate per well of a 6-well plate) were
mixed with protein loading buffer and separated on 9% SDS-PAGE gel. Proteins were
transferred to nitrocellulose membrane using transfer buffer (0.25 M Tris, 1.8 M glycine)
and transferred at constant 400 mM for 2 hours. The complete transfer of the colored
protein ladder was used as an indication of successful transfer. Thereafter, the
nitrocellulose membrane was blocked with 3% (w/v) BSA in TBS-T (0.1% Tween-20 in
Tris-buffered saline) buffer at room temperature for 1 hour. To immunostain with mouse
monoclonal anti-LETMI antibody (Abcam, Cat. ab55434), nitrocellulose membrane was
immersed in a solution of 1 ng/mL anti-LETMI antibody in TBS-T at room temperature
for 60 minutes before rinsing with TBS-T 4x 5 minutes. The membrane was then
incubated with goat anti-mouse horseradish peroxidase conjugate (BioRad, 1:5000
dilution) in TBS-T at room temperature for 60 minutes. Membrane was wash 4x 5
minutes with TBS-T, then developed with the SuperSignal West Pico reagent (Thermo
Scientific), and imaged using an Alpha Innotech digital camera.
Lentivirus generation
HEK 293T cells were plated in wells of a 6-well cell culture plate (9.6 cm2 /well) and
transfected at 50 - 70% confluency with either 1.25 gg LETMI or LETM1-mCherry in
226
FCPGW plasmid (lentiviral transfer vector), 0.125 ptg pVSVG (lentiviral packaging
vector), and 1.125 ptg A8.9 (lentiviral packaging vector) using lipofectamine (Invitrogen),
following manufacture's protocol. Transfection MEM solution was changed to 1 ml of
fresh cell culture media (MEM+10% FBS) at 4 hours after transfection. When the three
plasmids are transfected into HEK 293T cells, viral particles accumulate in the
supernatant. Cells were allowed to grow for 72 hours during which time lentivirus was
being made. To harvest the virus, cell culture media was collected and filtered through a
0.45 gm filter and stored at 4 'C. Viruses used (Figure 5-8) were no more than 5 days old.
227
References
1. Baker M. Making membrane proteins for structures: a trillion tiny tweaks. Nat Meth.
2010 Jun;7(6):429-34.
2. Park N, Um SH, Funabashi H, Xu J, Luo D. A cell-free protein-producing gel. Nat
Mater. 2009 May;8(5):432-7.
3. Verardi R, Traaseth NJ, Masterson LR, Vostrikov VV, Veglia G. Isotope Labeling for
Solution and Solid-State NMR Spectroscopy of Membrane Proteins. In: Atreya HS,
editor. Isotope labeling in Biomolecular NMR [Internet]. Springer Netherlands; 2012
[cited 2013 Apr 22]. p. 35-62. Available from:
http://link.springer.com/chapter/10.1007/978-94-007-4954-2_3
4. Ding X, Zhao X, Watts A. G-protein-coupled receptor structure, ligand binding and
activation as studied by solid-state NMR spectroscopy. Biochemical Journal. 2013
Mar 15;450(3):443-57.
5. Shimizu T, Mitsuke H, Noto K, Arai M. Internal Gene Duplication in the Evolution of
Prokaryotic Transmembrane Proteins. Journal of Molecular Biology. 2004 May
21;339(1):1-15.
6. Rapp M, Granseth E, Seppala S, Von Heijne G. Identification and evolution of dual-
topology membrane proteins. Nat Struct Mol Biol. 2006 Feb;13(2):1 12-6.
7. Geest M van, Lolkema JS. Membrane Topology and Insertion of Membrane Proteins:
Search for Topogenic Signals. Microbiol. Mol. Biol. Rev. 2000 Mar 1;64(1):13-33.
8. Akiyama Y, Ito K. Folding and assembly of bacterial alkaline phosphatase in vitro and
in vivo. J. Biol. Chem. 1993 Apr 15;268(11):8146-50.
9. Silhavy TJ, Shuman HA, Beckwith J, Schwartz M. Use of gene fusions to study outer
membrane protein localization in Escherichia coli. Proc. Natl. Acad. Sci. U.S.A. 1977
Dec;74(12):5411-5.
10. Yost-Daljev MK, Cornelissen CN. Determination of Surface-Exposed, Functional
Domains of Gonococcal Transferrin-Binding Protein A. Infect Immun. 2004
Mar;72(3):1775-85.
11. Tamai S, lida H, Yokota S, Sayano T, Kiguchiya S, Ishihara N, et al. Characterization
of the mitochondrial protein LETM 1, which maintains the mitochondrial tubular
shapes and interacts with the AAA-ATPase BCS1L. J. Cell. Sci. 2008 Aug 1;121(Pt
15):2588-600.
12. Martell JD, Deerinck TJ, Sancak Y, Poulos TL, Mootha VK, Sosinsky GE, et al.
Engineered ascorbate peroxidase as a genetically encoded reporter for electron
microscopy. Nat Biotech. 2012 Nov;30(11):1143-8.
228
13. Rhee H-W, Zou P, Udeshi ND, Martell JD, Mootha VK, Carr SA, et al. Proteomic
mapping of mitochondria in living cells via spatially restricted enzymatic tagging.
Science. 2013 Mar 15;339(6125):1328-31.
14. Jiang D, Zhao L, Clapham DE. Genome-wide RNAi screen identifies Letml as a
mitochondrial Ca2+/H+ antiporter. Science. 2009 Oct 2;326(5949):144-7.
15. Frazier AE, Taylor RD, Mick DU, Warscheid B, Stoepel N, Meyer HE, et al. Mdm38
interacts with ribosomes and is a component of the mitochondrial protein export
machinery. J Cell Biol. 2006 Feb 13;172(4):553-64.
16. Dimmer KS, Fritz S, Fuchs F, Messerschmitt M, Weinbach N, Neupert W, et al.
Genetic basis of mitochondrial function and morphology in Saccharomyces
cerevisiae. Mol. Biol. Cell. 2002 Mar;13(3):847-53.
17. Endele S, Fuhry M, Pak S-J, Zabel BU, Winterpacht A. LETMI, A Novel Gene
Encoding a Putative EF-Hand Ca2+-Binding Protein, Flanks the Wolf-Hirschhorn
Syndrome (WHS) Critical Region and Is Deleted in Most WHS Patients. Genomics.
1999 Sep 1;60(2):218-25.
18. Schlickum S, Moghekar A, Simpson JC, Steglich C, O'Brien RJ, Winterpacht A, et al.
LETMI, a gene deleted in Wolf-Hirschhorn syndrome, encodes an evolutionarily
conserved mitochondrial protein. Genomics. 2004 Feb;83(2):254-61.
19. Nowikovsky K, Froschauer EM, Zsurka G, Samaj J, Reipert S, Kolisek M, et al. The
LETM1/YOL027 Gene Family Encodes a Factor of the Mitochondrial K+
Homeostasis with a Potential Role in the Wolf-Hirschhorn Syndrome. J. Biol. Chem.
2004 Jul 16;279(29):30307-15.
20. Hasegawa A, Van der Bliek AM. Inverse correlation between expression of the
Wolfs Hirschhorn candidate gene Letml and mitochondrial volume in C. elegans and
in mammalian cells. Hum. Mol. Genet. 2007 Sep 1;16(17):2061-71.
21. Tamai S, lida H, Yokota S, Sayano T, Kiguchiya S, Ishihara N, et al. Characterization
of the mitochondrial protein LETM 1, which maintains the mitochondrial tubular
shapes and interacts with the AAA-ATPase BCS1L. J Cell Sci. 2008 Aug
1;121(15):2588-600.
22. Piao L, Li Y, Kim SJ, Sohn K-C, Yang K-J, Park KA, et al. Regulation of OPA l-
mediated mitochondrial fusion by leucine zipper/EF-hand-containing transmembrane
protein-I plays a role in apoptosis. Cell. Signal. 2009 May;21(5):767-77.
23. Piao L, Li Y, Kim SJ, Byun HS, Huang SM, Hwang S-K, et al. Association of
LETMI and MRPL36 contributes to the regulation of mitochondrial ATP production
and necrotic cell death. Cancer Res. 2009 Apr 15;69(8):3397 404.
229
24. Dimmer KS, Navoni F, Casarin A, Trevisson E, Endele S, Winterpacht A, et al.
LETM 1, deleted in Wolf-Hirschhorn syndrome is required for normal mitochondrial
morphology and cellular viability. Hum. Mol. Genet. 2008 Jan 15;17(2):201-14.
25. Dimmer KS, Navoni F, Casarin A, Trevisson E, Endele S, Winterpacht A, et al.
LETM 1, deleted in Wolf-Hirschhorn syndrome is required for normal mitochondrial
morphology and cellular viability. Hum. Mol. Genet. 2008 Jan 15;17(2):201-14.
26. Tamai S, lida H, Yokota S, Sayano T, Kiguchiya S, Ishihara N, et al. Characterization
of the mitochondrial protein LETM 1, which maintains the mitochondrial tubular
shapes and interacts with the AAA-ATPase BCS1L. J Cell Sci. 2008 Aug
1;121(15):2588-600.
27. Brown TA, Tkachuk AN, Shtengel G, Kopek BG, Bogenhagen DF, Hess HF, et al.
Superresolution fluorescence imaging of mitochondrial nucleoids reveals their spatial
range, limits, and membrane interaction. Mol. Cell. Biol. 2011 Dec;31(24):4994-
5010.
230
Jennifer (Zhengzheng) Yao
Tel: (626)758-2370 70 Pacific Street, Apt. 548C
Email: izyao(&,mit.edu Cambridge, MA 02139
EDUCATION
Ph.D. in Chemistry
Massachusetts Institute of Technology (MIT) 9/2008-expected 6/2013
B.S. in Chemical Biology
University of California at Berkeley (UC Berkeley) 9/2004-5/2008
RESEARCH EXPERIENCE
Department of Chemistry, MIT 9/2008-expected 6/2013
Advisor: Alice Y. Ting
Developed of new methods for targeting diverse fluorophores to specific proteins inside
live cells. Engineered benzophenone photocrosslinker ligase to study protein-protein
interactions.
Department of Chemistry, UC Berkeley 8/2006-5/2008
Advisor: Dirk Trauner
Designed, synthesized, and purified small molecule photoswitchable agonists of the
nicotinic acetylcholine receptor and two-photon sensitive chromophores for light-
activated ion channels.
Department of Chemical Engineering, UC Berkeley 5/2005-8/2006
Advisor: David Schaffer
Studied of regulation of proneural Mash1 and Ngn1 genes in the Wnt signaling pathway.
Department of Biological Science, California State University, LA 6/2003-8/2003
Center for Environmental Analysis-Centers of Research Excellence in Science and
Technology (CEA-CREST)
Advisor: John Gamon
Applied of optical tools to study plant ecophysiology and ecosystem gas exchange.
PUBLICATIONS
Yao, J.Z., Uttamapinant, C., Poloukhtine, A., Baskin, J.M., Codelli, J.A., Sletten, E.M.,
Bertozzi, C.R., Popik, V.V., Ting, A.Y. "Fluorophore targeting to cellular
proteins via enzyme-mediated azide ligation and strain-promoted cycloaddition",
JAm Chem Soc. 2012 (134), 3720-8.
Tochitsky, I., Banghart, M.R., Mourot, A., Yao, J.Z., Gaub, B., Kramer, R.H., Trauner, D.
"Optochemical control of genetically engineered neuronal nicotinic acetylcholine
receptors", Nat Chem. 2012 (4), 105-11.
Banghart, M.R., Mourot, A., Fortin, D.L., Yao, J.Z., Kramer, R.H., Trauner, D.
"Photochromic blockers of voltage-gated potassium channels", Angew Chem Int
Ed Engl. 2009 (48), 9097-101.
231
LABORATORY SKILLS
- Biology-related lab skills: mammalian cell and neuron culture, protein purification
from mammalian and bacterial cells, chromatography (normal and reverse phase, ion
exchange, affinity binding), molecular cloning, gel-electrophoresis,
immunoprecipitation, Western blot, microscopy, LC-MS
e Organic synthesis-related lab skills: organic synthesis and purification of small
molecules on large (>1 g) and small scales (<0.5 mg), distillation, extraction,
recrystallization, SiO 2 and C-18 column chromatography, LC-MS, NMR
TEACHING AND MENTORING EXPERIENCE
Mentor, Women In Chemistry, MIT 9/2009-Present
- Guided first-year female graduate students in Chemistry in course selections and
choosing Ph.D. research advisors.
- Introduced new students to campus resources and met with mentees monthly to offer
effective mentorship.
Teaching Assistant, Biological Chemistry 11 (5.08), MIT 2/2009-5/2009
e Taught two recitation classes with 30 students each throughout the semester.
e Developed effective recitation material working in conjunction with lecture material.
- Selected recent seminal research publications to discuss in class.
e Helped professors design problem sets and exams.
Teaching Assistant, Principles of Chemical Science (5.111), MIT 9/2008-12/2008
e HHMI teaching fellow
e Developed and implemented innovative general chemistry teaching material that
demonstrates the chemical principles behind inspiring applications in biology and
medicine.
- Taught two recitation classes with 30 students each throughout the semester.
Laboratory Instructor, General Chemistry Laboratory, UC Berkeley 1/2008-5/2008
- Conducted two general chemistry laboratory classes with 35 students each throughout
the semester.
e Guided students in developing safe and proper experimental skills and writing lab
reports.
Laboratory Instructor, Organic Chemistry Laboratory, UC Berkeley 1/2007-5/2007
- Conducted two organic chemistry laboratory classes with 35 students each throughout
the semester.
- Taught organic chemistry lab skills including handling moisture and shock-sensitive
chemicals, TLC, extraction, column purification, recrystallization, and distillation.
LEADERSHIP AND COMMUNITY SERVICE
MIT Chemistry Outreach 6/2009-Present
232
e Conducted engaging demonstrations illustrating a broad range of chemical principles
to classes of thirty to ninety students at local schools ranging from elementary to high
schools.
" Illustrated concepts relating to polymer formation, natural products, temperature,
pressure, and light through demonstrations and PowerPoint slides.
Chemistry Education and Outreach Career Panel Organizer 1/2013
- Organized and networked with college lecturers, high school teachers, and leaders in
education-related nonprofit organizations to establish the first departmental Education
and Outreach Career Panel.
HONORS AND AWARDS
MIT Presidential Fellowship 8/2008-5/2009
HHMI Teaching Assistant Fellowship 8/2008-12/2008
Berkeley Molecular Cell Biology Undergraduate Symposium Honorary Poster 3/2008
Melvin J. Heger-Horst Scholarship (College of Chemistry, UC Berkeley) 2007
Merck Undergraduate Research Scholarship (David Schaffer Lab, UC Berkeley) 2006
233