The production, design and application of antimicrobial peptides

Abstract

With the spread of drug-resistant bacteria, existing antibiotics are losing their potency. Antimicrobial peptides (AmPs) represent an exciting class of drug candidates, particularly because their mechanism of action is unlikely to induce drug resistance. If resistance to AmPs were also slower to emerge in the clinic, they would have longer useful lifetimes than existing antibiotics. Nevertheless, a number of limitations exist for AmPs in the clinic. The high cost of peptide manufacture requires that highly potent sequences are created. Additionally, AmP selectivity must be improved if effective systemic doses are to be given without hemolytic activity or other toxicity. Improved high-throughput methods for AmP design or discovery could enable the achievement of both of these goals. To this end, we developed an approach based on the discovery of semi-conserved motifs across natural AmPs, which we demonstrated are associated with antimicrobial activity. Additionally, we created novel AmP formulations that may bypass some of these clinical limitations. In order to evaluate AmP design approaches, a high-throughput production and assay platform was created using in vitro translation. This technology may produce peptides that would be toxic to recombinant hosts and synthesize peptides of arbitrary length.

(cont.) The cost per peptide was minimized through a series of process improvements. First, we created methods to construct oligonucleotides that mimicked our motif-based design of AmPs. This approach allowed the reuse of primers for many peptides, reducing cost and enabling the study of pattern synergy. Additionally, we found peptide translation was enhanced by co-translating a fusion partner in frame with the AmP. The AmP could be freed from the fusion partner after translation using enterokinase digestion. Further, we increased yield 3-fold by optimizing the length of fusion partner. The partner was made as short as possible to limit the translational resources required to synthesize the fusion partner, while being long enough to ensure stability from proteases. The solubility of the fusion partner-AmP construct was also improved through the selection of a highly soluble partner of the optimal length. Finally, we developed a purification scheme to ensure that the in vitro translation extract would not impact measurement of antimicrobial activity. We also developed and evaluated the design of AmPs using semi-conserved motifs. We used a database of over 500 natural AmPs as a training set for pattern discovery.

(cont.) The resulting motifs were exhaustively recombined to create all 20 amino acid sequences that were entirely covered by these patterns. These sequences were clustered, and 42 diverse members selected for characterization using representative Gram negative and Gram positive bacteria. Approximately 50% of the designed AmPs were active against at least one of the bacteria at 256 ug/ml. Control peptides were created in which the amino acids in the designed peptides were rearranged such that they were not homologous to any antimicrobial motifs. Thus, these controls had the same bulk physiochemical properties frequently associated with antimicrobial activity as the designed sequences, but we hypothesized they would not be active because they did not match the antimicrobial motifs. In fact, only 5% of the control sequences had activity at 256 ug/ml, indicating that the antimicrobial motifs give a 10-fold enrichment in activity. Further, two highly active designed peptides had MICs of 16 ug/ml against Bacillus cereus and 64 ug/ml against Escherichia coli. Additionally, AmPs active against B. cereus were all active against the hospital pathogen Staphylococcus aureus, and the bioterror agent, Bacillus anthracis.

(cont.) Our motif-based design may be most effective as the first stage of a two-stage design tool. In the first stage, highly diverse leads with novel profiles are created and evaluated. Promising leads could then be optimized using a variety of techniques. By creating just 44 variants of one lead, we designed an AmP with broad spectrum activity that had MICs of 16 ug/ml against E. coli and 8 ug/ml against B. cereus and 4 ug/ml against S. aureus. Another approach to build on our design tool would be to incorporate activity and toxicity characteristics of members of the training set into the design or scoring of new sequences. In order to begin assembling this data using a standardized method, a representative set of 100 natural, linear AmPs was chosen through clustering. Their antimicrobial activity against E. coli, S. epidermidis, and S. aureus were evaluated, along with hemolytic activity. When further supplemented, this information may enable an improved scoring metric to be created. Additionally, we systematically demonstrated that amidating the c-terminus of natural AmPs improves both antimicrobial activity and therapeutic index. Finally, we recognized that AmP's mechanism of action would allow activity to be retained when they are permanently tethered to medical device surfaces.

(cont.) Unlike existing coatings which rely on the slow release of silver or other antibiotics, a permanently tethered approach could have a longer lifetime and reduced systemic toxicity concerns. A versatile chemistry was developed to create immobilized AmP coatings. These formulations had broad spectrum antimicrobial activity without significant hemolytic activity. Further, the coatings were effective through multiple bacterial challenges. The combination of the AmP design tool along with localized formulations represent a significant advance in the process of moving AmPs to the clinic to combat drug-resistant infections.