RESEARCH

Site-selectivity is an important requirement in bioconjugation that arises from the intrinsic complexity of biopolymers, often containing tens or hundreds of functional groups in a single molecule. An unmet need remains to perform one-step, enzyme-free, site-selective protein modifications with endogenously encoded sequences. As the only thiol-containing canonical amino acid, cysteine (Cys) is a suitable target for selective protein modification. The Cys thiol is more reactive toward electrophiles than other functional groups, enabling enzyme-free chemoselective modification.

Highly efficient macrocyclization of unprotected peptides was achieved. Novel compounds show enhanced cell penetration, increased stability to proteases, and tunable structural properties.

(A. Spokoyny et al. JACS., 2013, 135, 5946-5949)

Glutathione S-transferase efficiently catalyzed the regiospecific arylation at cysteine in peptides and proteins containing a GSH tag. This chemistry was further developed to perform macrocyclization of long, fully unprotected peptides in water.

(C. Zhang et al. Angew. Chem Int. Ed., 2013, 52, 14001–14005; C. Zhang et al. Org. Lett., 2014, 16, 3652–3655)

We applied in-solution library selection approaches to identify reactive peptide sequences that promote arylation and discovered a four-residue unit (Phe-Cys-Pro-Phe, the “π-clamp”) that accelerates cysteine perfluoroarylation in water by over 1000-fold relative to reactions of scrambled control peptides. Density functional theory calculations showed that the π-clamp promoted the reaction both kinetically (lowering the activation energy) and thermodynamically (generating a more stable product). The π-clamp is composed entirely of genetically encoded amino acids, enabling convenient insertion into an antibody heavy chain C-terminus. The antibodies expressed with C-terminal π-clamps readily conjugated to perfluoroaryl-linked probes in a single step under reducing conditions. The resulting site-specific antibody conjugates retained their target-binding affinity and showed receptor-dependent cell killing. Compared to other site-selective antibody modification methods that require unnatural amino acids, enzymes, or multiple chemical steps, the π-clamp provides an enzyme/catalyst-free method for chemo- and regioselective modification of a natural amino-acid sequence, thus generating a fundamentally new route toward next-generation homogeneous antibody-drug conjugates.

(C. Zhang et al. Nat. Chem., 2016, 8, 120-128; P. Dai et al. Sci. Rep., 2017, 7, 7954)

Performing reaction in high salt buffers greatly improves the rate of labeling by four orders of magnitude. Antibody-drug conjugates prepared maintained full biological activity. Computational studies were done collaboratively with Prof. Van Voorhis Lab.

(P. Dai et al. ACS Cent. Sci., 2016, 2, 637–646)

We determined how macrocyclization chemistry modulates the properties of Transportan-10, an established amphipathic cell-penetrating peptide. Analogs of TP10 were prepared with two Cys residues at (i, i+7) positions and treated with decafluorobiphenyl to obtain cyclic “M” variants. Upon incubation of an immortalized brain endothelial cell line, all fluorophore-labeled macrocyclic variants displayed an increase in mean fluorescence intensity (MFI) with respect to TP10 alone and the Q* linear controls. Proteolytic stability assays revealed improved resistance to Proteinase K of the macrocyclic variants, indicating that peptide macrocyclization can impart superior stability and cellular penetration.

(C.M. Fadzen et al. J. Am. Chem. Soc., 2017, 139, 15628-31)

Selenocysteine (Sec) is a structural analogue of Cys in which selenium replaces sulfur. With the Buchwald group, we developed a new approach to the arylation of selenocysteine in unprotected peptides: instead of relying on the nucleophilic character of Sec, we arylated oxidized Sec electrophiles using boronic acid reagents via a copper-mediated process. The reaction is chemoselective, enabling Sec arylation in a complex peptide. We applied selenocysteine-based arylation in collaboration with the Miller group at Yale University and the Plante group at Visterra Inc. to covalently append the unprotected RP-1 antimicrobial peptide (AMP) to the vancomycin antibiotic, obtaining the conjugate RP-1-van. Treatment of wild-type A. baumannii with RP-1-van saw a >10-fold increase in bactericidal activity vs. vancomycin alone, supporting nucleophilic selenocysteine substitution as a feasible native conjugation strategy and a potential approach to overcome Gram-negative drug resistance.

(D.T. Cohen et al. J. Am. Chem. Soc., 2015, 137, 9784-7; D.T. Cohen et al. Nat. Chem., 2019, 11, 78-85)

We identified a genetically encoded peptide tag (termed DBCO-tag) that selectively reacts with the commercially available aza-dibenzocyclooctyne (DBCO) reagents by cysteine-cyclooctyne conjugation. The seven-residue DBCO-tag (Leu-Cys-Tyr-Pro-Trp-Val-Tyr) installed at the N- or C-termini of a peptide or protein enables conjugation with various DBCO analogs. Compared to a scrambled peptide control, the DBCO-tag increases the rate of the thiol-yne reaction by 220-fold, affording site-selective conjugation to fluorescent probes, affinity tags, and cytotoxic drug molecules. Fusion of the DBCO-tag with the protein of interest enabled regioselective cysteine modification on proteins that contain multiple endogenous Cys residues, including green fluorescent protein and Trastuzumab. This study demonstrates that short peptide tags can aid in accelerating reactions that are often sluggish in water. We anticipate the DBCO-tag will become a benchmark tool for site-selective protein modification.

(C. Zhang et al. Angew. Chem. Int. Ed., 2018, 57, 6459-63)

Organometallic reagents may open important avenues of metal-mediated bioconjugation strategies. In collaboration with the Buchwald lab, we created palladium(II)-based organometallic complexes featuring biarylphosphine ligands that are efficient and versatile reagents for chemoselective cysteine arylation. The reaction proceeded at low concentrations of both the organometallic complex and the biomolecule under a wide pH range and required no heating. The initial water solubility limitation of these complexes has been overcome with water-soluble biarylphosphine ligands. These storable and air-stable arylation reagents were successfully applied to modify cysteine in unprotected peptides, bacterial toxins, antibodies and their mimics. Diverse S-aryl conjugates were synthesized with fluorescent dyes, affinity tags, and bioconjugation handles. A study of the resulting conjugates comparing maleimide and acetamide analogs showed that S-aryl conjugates are more stable in the presence of glutathione and under basic, acidic, and oxidative conditions. We extended this method to achieve Pd-mediated arylation of lysine in unprotected peptides in organic solvents. Pd-mediated arylation can be combined with other chemistries to cater to specific bioconjugation needs, as seen in our recently developed Pd-based method for cysteine-lysine cross-linking. Furthermore, we used this method for the intermolecular cross-linking between a peptide and a protein, and between two unprotected proteins. These studies demonstrate the potential of Pd-mediated chemistry to facilitate cross-linking techniques for native biomolecules.

(E. Vinogradova et al. Nature, 2015, 526, 687-691; H.G. Lee at al. Angew. Chem. Int. Ed., 2017, 56, 3177-3181; A.J. Rojas et al. Chem. Sci., 2017, 8, 4257-62; A.J. Rojas et al. Org. Lett., 2017, 19, 4263-6; K. Kubota et al. J. Am. Chem. Soc., 2018, 140, 3128-33; H.H. Dhanjee et al. J. Am. Chem. Soc., 2020, 142, 9124-9129)

A plethora of novel arylation agents allows facile lysine arylation of unprotected peptides. This chemistry was used to develop a new class of macrocyclic peptides with improved stability and excellent cell penetration.

(G. Lautrette et al. JACS, 2016, 138, 8340-8343)

Ribosomes can produce proteins in minutes and are largely constrained to proteinogenic amino acids. We aim to develop high-fidelity multistep, continuous flow chemical processes for the efficient and rapid synthesis of biopolymers. These technologies provide an alternative for producing single-domain proteins without the ribosome and support our long-term goal of constructing a “chemical factory” in which starting reagents flow in and pure biomolecules of interest flow out. We invented a novel flow-based platform for the ultra-rapid production of synthetic peptides in collaboration with Klavs Jensen’s lab. This machine, termed automated fast-flow peptide synthesizer (AFPS), allows us to increase the rate of peptide production by a factor of 60 versus commercial peptide synthesizers, enabling amide coupling reactions in 10 seconds or less and performing a full amide coupling cycle in 1 minute. We used AFPS technology for expedited and high-fidelity production of personalized cancer vaccines for clinical partners and produced the majority of human antimicrobial peptides with AFPS instrumentation. Automated flow technology not only facilitates rapid biopolymer production, but has enabled us to carry out entire D-scans of small proteins to investigate their folding and functions. We recently used this technology to achieve stepwise total chemical synthesis of protein chains nearing 200 amino acids in length that were demonstrated to retain the structure and function of native variants obtained by recombinant expression. Our future efforts will be to adapt this technology to flow synthesis of antisense oligonucleotides, harness the power of machine learning to predict and eliminate aggregation, and combine flow synthesis of protein domains with native chemical ligation and protein folding, eventually arriving at an integrated system capable of synthesizing the majority of human proteins on demand. This technology has the potential to solve the manufacturing problem for custom or on-demand proteins and life-saving personalized therapies, and is amenable to designer chemistry with the use of a variety of backbones and building blocks, including D-amino acids, glycosylated functionalities, and non-canonical residues.

(M.D. Simon et al. ChemBioChem, 2014, 15, 713-20; R.L. Policarpo et al. Angew. Chem. Int. Ed., 2014, 53, 9203-8; S.K. Mong et al. ChemBioChem, 2014, 15, 721-33; T. Luhmann et al. Org. Biomol. Chem., 2016, 14, 3345-9; M.D. Simon et al. J. Am. Chem. Soc., 2016, 138, 12099-111; A.J. Mijalis et al. Nat. Chem. Biol., 2017, 13, 464-6; N.L. Truex et al. Sci. Rep., 2020, 10, 723; J.S. Albin and B.L. Pentelute, Aust. J. Chem., 2020, 73, 380-388; N. Hartrampf et al. Science, 2020, 368, 980-987)

Total chemical synthesis of proteins remains highly labor-intensive. We addressed this problem by developing a reliable method to synthesize long peptides and protein chains using flow chemistry and optimized our AFPS technology to meet this challenge. The result of our efforts was a routine protocol that allows for stepwise chemical total synthesis of peptide chains exceeding 50 amino acids in length, with a cycle time of ~2.5 minutes per amino acid. The optimized protocol delivers products with high fidelity and of high chiral purity. Using this protocol, single domain protein chains ranging from barstar (90 amino acids) to sortase A* (164 amino acids) were synthesized in 3.5–6.5 h. To demonstrate the application to the production of functional proteins, these sequences were folded and their biophysical properties and enzymatic activities were determined to match those of recombinant samples. The timescale of chemical protein synthesis is on par with that of recombinant expression and therefore offers a practical alternative to biochemical methods, while opening up the chemical space beyond canonical amino acids. Ultimately, we intend for this protocol to serve as a blueprint for the automated flow synthesis of other artificial sequence-defined polymers.

(N. Hartrampf et al. Science, 2020, 368, 980-987)

A critical challenge in the clinical translation of large therapeutic molecules such as antisense oligonucleotides (ASOs) is their poor intracellular uptake. For example, the ASO called Eteplirsen for Duchenne muscular dystrophy (DMD) is administered at 30 mg/kg. We overcome this barrier with cell-penetrating peptides (CPPs), chemical vectors that facilitate cellular uptake and nuclear targeting of antisense cargoes. In a multi-year ongoing collaboration with Sarepta Therapeutics, we developed a peptide-based platform for enhanced cytosolic delivery and nuclear targeting of therapeutic ASOs. Our experimental results strongly support the feasibility of the CPP platform to enhance delivery of neutral phosphorodiamidate morpholino oligomers (PMOs). We will expand our studies to cover several types of therapeutically relevant ASOs, including peptide nucleic acids (PNAs) and hybrid backbones, achieve selective tissue targeting of the peptide-antisense conjugates, and develop advanced machine learning (ML) models to further predict and optimize CPPs.

Our second approach to improve cellular uptake of therapeutic macromolecules is based on harnessing a two-component delivery system evolved by Nature: protective antigen (PA) and the N-terminus of lethal factor (LFN). These components are nontoxic variants of proteins secreted by the Gram-positive pathogen Bacillus anthracis and efficiently perform protein translocation into the cytosol of cells. We combined synthetic chemistry with protein expression and enzyme-mediated bioconjugation and investigated the ability of the PA/LFN platform to transport non-natural cargo to the cytosol of mammalian cells. We found this nanomachine can deliver diverse types of non-canonical cargo, including non-natural amino acids, chemical probes, antibody mimics, antisense peptide nucleic acids, and mirror-image proteins. We showed PA and LFN components are an engineerable delivery system with therapeutic potential and investigated whether this potential can be realized by retargeting PA to specific tumor cell membrane receptors with designer antibody fragments. We redirected the anthrax protective antigen to receptors overexpressed on tumor cells and modified its lethal factor component to target cancer gene dependencies with antisense peptide nucleic acids. Our current efforts focus on adapting the anthrax platform to achieve efficient delivery of cancer vaccines into dendritic cells.

In our previous studies we showed that bicyclic arginine-rich CPPs conjugated to PMO are endowed with superior cellular uptake, stability, and gene-splicing activity compared to monocyclic variants. We further hypothesized the benefits of different CPPs can be leveraged by fusing them into chimeric constructs. We designed three constructs that combine an arginine-rich and an amphipathic CPP, linked into long linear peptides. The C-terminal peptide in all cases was Bpep, a CPP thoroughly tested for PMO delivery. For the N-terminal peptide, we chose the pVEC, Penetratin, and Melittin. These chimeras were conjugated to a reporter PMO and evaluated in HeLa cells by an enhanced green fluorescent protein (eGFP) functional readout assay. All CPP chimeras performed better than Bpep alone, with the PMO-Penetratin-Bpep chimera exhibiting ~20-fold increase in activity vs. PMO alone. PMO-Penetratin-Bpep and PMO-pVEC-Bpep displayed synergy, in which the activity of the chimera was greater than the sum of the expected activities from the individual PMO-CPP components. These studies propose chimeric CPPs as a viable strategy to improve PMO delivery and activity.

(J.M. Wolfe et al. Angew. Chem. Int. Ed., 2018, 57, 4756-59;C.M. Fadzen et al. Biochemistry, 2019, 58, 3980-3989)

We developed advanced machine learning (ML) models to predict and optimize the function of abiotic polymers. In a collaboration with Prof. Rafael Gomez-Bombarelli, we used interpretable deep learning to predict best-in-class CPPs. Using a modular peptide synthesis platform, a dataset containing the activity of 600 abiotic PMO-CPP constructs was collected. A neural network-based predictor was developed and coupled with a genetic algorithm optimizer to maximize activity. 11 peptides designed by the ML model were synthesized and shown to match activity predictions. These artificial ‘Mach’ peptides performed up to ~50-fold better than controls and outperformed all previous library and published CPP candidates for PMO delivery, while being non-toxic and effective in mice. These results highlight the power of deep learning approaches in leveraging consistent experimental data to design de novo highly-active functional polymers.

(J.M. Wolfe et al. ACS Cent. Sci., 2018, 4, 512-20; C.K. Schissel et al. BioRxiv, 2020, DOI: 10.1101/2020.1104.1110.036566)

Over the past five years, Pentelute lab demonstrated that Anthrax toxin system can be leveraged to deliver a variety of natural and unnatural macromolecules. These delivered molecules were functional inside the cytosol.

(A. Rabideau et al. ACS Chem. Bio., 2016, 11, 1490–1501)

A number of different chemical techniques enable facile conjugation of chemically diverse biomolecules to LFN; 100’s conjugates have been prepared.

(R. Policarpo et al. ACIE, 2014, 53, 9203-9208; J. Ling et al. JACS, 2012, 134, 10749-10752)

Functional antibody mimics were successfully delivered and shown to perturb cancer relevant protein-protein interactions. The system substantially outperformed established CPP techniques for delivery by at least four orders of magnitude.

(X. Liao et al. ChemBioChem, 2014, 15, 2458 – 2466)

For the first time, robust delivery into cells of D-proteins was achieved. A D-peptide was delivered into brain cancer cells and shown to perturb a cancer relevant protein-protein interaction.

(A. Rabideau et al. Chem. Sci., 2015, 6, 648 – 653)

Multiple molecules with non-peptidic backbone or peptides with unnatural side chains were successfully delivered, pushing the limits of PA pore translocation.

(A. Rabideau et al. Sci. Rep., 2015, 5:11944)

For this first time, it was shown that the simple addition of just one D-amino acid to the N-terminus of a protein significantly increased its cytosolic stability by orders of magnitude.

(A. Rabideau et al. ACS Cent. Sci., 2015, 1, 423-430)

CYCLOPS (copy-number alterations Yielding Cancer Liabilities Owing to Partial losS) are essential genes that undergo partial copy-number loss in tumors. In collaboration with Dr. Rameen Beroukhim we targeted one such gene, SF3B1, with PNA delivered by the anthrax toxin-derived LFN-PA system. We targeted this gene with two SF3B1 LFN-PNA conjugates: antisense (hybridizing) and sense (non-hybridizing control). Cell viability analysis indicated selectivity to CRISPRcopy-loss over CRISPRneutral cells. These findings were corroborated across breast and blood cancer cell lines. These results show that the PA/LFN platform efficiently delivers PNAs for gene suppression therapy.

(Z. Lu et al. ACS Chem. Biol., 2020, 15, 1358-1369)

Protein-protein interactions (PPIs) equally sustain life and drive development of disease. Historically considered “undruggable” because of the extended binding interfaces involved, PPIs are now a major drug discovery focus. Peptides are excellent candidates to disrupt PPIs by combining the binding affinities of antibodies and the cell-penetration abilities of small molecules. Unmodified canonical peptides are generally poor drug candidates, suffering from rapid proteasomal degradation and poor cellular uptake. Fully synthetic libraries containing exclusively non-natural amino acids may render stable, potent peptide binders as viable drug leads.
Affinity selection-mass spectrometry (AS-MS) is a high-throughput strategy that can meet the need for target-based de novo discovery of peptidomimetic ligands. For the past five years, the Pentelute lab developed methods for high-throughput identification of peptide binders. We leveraged high-resolution liquid chromatography-tandem mass spectrometry (LC-MS/MS) to sequence individual synthetic peptides from complex mixtures, and to increase the diversity of synthetic peptide libraries amenable to AS-MS from ~100 to ~1,000,000. This advance was used to mature the affinity of non-canonical variants of known binders by screening ‘focused’ libraries in solution. With this approach, binders with sub-nanomolar affinity (dissociation constant, Kd ≤1 nM) for the MDM2 oncogenic protein were identified using high performance size exclusion chromatography (HPSEC)-based enrichment. Using bead-based affinity selection from 10 million-member libraries, we discovered a new class of functional synthetic D-peptide binders with affinity to a variety of biomolecules. In the latest development of our technology, we perform de novo screening of fully synthetic, ultra-large (~100 million members) non-canonical libraries against soluble protein targets, isolate the target-bound hits, quantitate the target-dependent enrichment of each peptide by nano-LC-MS/MS, validate and refine the binders by iterative selections. High-affinity (Kd ≤100 nM) peptidomimetic ligands were identified with this method for the 12ca5 clone of anti-hemagglutinin antibody and the signaling protein 14-3-3. Our current work applies the AS-MS platform for discovery of non-canonical peptide disruptors of cancer PPIs that are effective, non-toxic, long-circulating, and stable in cells and animals. We will leverage the power of machine learning algorithms to classify, predict, and improve the putative binders. Finally, we are investigating affinity reagents that target specific receptors in cardiac and skeletal muscle, as well as erythrocytes in blood for superior pharmacokinetic, immunogenicity, and biodistribution profiles in vivo.

(A.A. Vinogradov et al. ACS Combi. Sci., 2017, 19, 694-701; Z.P. Gates et al. PNAS, 2018, 115, E5298-E5306; F. Touti et al. Nat. Chem. Biol., 2019, 15, 410-418; A.J. Quartararo et al. Nat. Commun., 2020, 11, 3183)

We prepared a library of non-canonical perfluoroaryl macrocyclic peptides and tested their affinity for the oncogenic MDM2 protein in a label-free affinity maturation assay. We discovered peptide P1 binds MDM2 with nanomolar affinity. Preliminary studies in mice showed accumulation of fluorophore-labeled P1 in the brain and other tissues. P1 cytotoxicity was observed in MDM2-expressing SJSA-1 and MCF-7 cells at levels comparable to a positive control peptide, ATSP-7041. Linear controls had no activity, while K562 cells which lack p53 did not respond to treatment with P1. Western blot analysis of p53 gene products and MDM2 of P1-treated SJSA-1 cell lysates showed a significant increase in MDM2 and p21 expression. Mice treated with P1 for 7 days displayed statistically significant reduction in SJSA-1 tumor xenografts over 14 days relative to vehicle and a scrambled P1 control, and at comparable levels with the ATSP-7041 inhibitor. These data support the therapeutic efficacy of P1 in MDM2-expressing tumors and highlight the utility of the AS-MS platform to discover potent non-canonical inhibitors of cancer-specific PPIs.

(F. Touti et al. Nat. Chem. Biol., 2019, 15, 410-418)