Accurate top protein variant discovery via low-N pick-and-validate machine learning.

A strategy to obtain the greatest number of best-performing variants with least amount of experimental effort over the vast combinatorial mutational landscape would have enormous utility in boosting resource producibility for protein engineering. Toward this goal, we present a simple and effective machine learning-based strategy that outperforms other state-of-the-art methods. Our strategy integrates zero-shot prediction and multi-round sampling to direct active learning via experimenting with only a few predicted top variants. We find that four rounds of low-N pick-and-validate sampling of 12 variants for machine learning yielded the best accuracy of up to 92.6% in selecting the true top 1% variants in combinatorial mutant libraries, whereas two rounds of 24 variants can also be used. We demonstrate our strategy in successfully discovering high-performance protein variants from diverse families including the CRISPR-based genome editors, supporting its generalizable application for solving protein engineering tasks. A record of this paper's transparent peer review process is included in the supplemental information.

Engineering an autonomous VH domain to modulate intracellular pathways and to interrogate the eIF4F complex.

An attractive approach to target intracellular macromolecular interfaces and to model putative drug interactions is to design small high-affinity proteins. Variable domains of the immunoglobulin heavy chain (VH domains) are ideal miniproteins, but their development has been restricted by poor intracellular stability and expression. Here we show that an autonomous and disufhide-free VH domain is suitable for intracellular studies and use it to construct a high-diversity phage display library. Using this library and affinity maturation techniques we identify VH domains with picomolar affinity against eIF4E, a protein commonly hyper-activated in cancer. We demonstrate that these molecules interact with eIF4E at the eIF4G binding site via a distinct structural pose. Intracellular overexpression of these miniproteins reduce cellular proliferation and expression of malignancy-related proteins in cancer cell lines. The linkage of high-diversity in vitro libraries with an intracellularly expressible miniprotein scaffold will facilitate the discovery of VH domains suitable for intracellular applications.

Machine learning-coupled combinatorial mutagenesis enables resource-efficient engineering of CRISPR-Cas9 genome editor activities.

The genome-editing Cas9 protein uses multiple amino-acid residues to bind the target DNA. Considering only the residues in proximity to the target DNA as potential sites to optimise Cas9's activity, the number of combinatorial variants to screen through is too massive for a wet-lab experiment. Here we generate and cross-validate ten in silico and experimental datasets of multi-domain combinatorial mutagenesis libraries for Cas9 engineering, and demonstrate that a machine learning-coupled engineering approach reduces the experimental screening burden by as high as 95% while enriching top-performing variants by ∼7.5-fold in comparison to the null model. Using this approach and followed by structure-guided engineering, we identify the N888R/A889Q variant conferring increased editing activity on the protospacer adjacent motif-relaxed KKH variant of Cas9 nuclease from Staphylococcus aureus (KKH-SaCas9) and its derived base editor in human cells. Our work validates a readily applicable workflow to enable resource-efficient high-throughput engineering of genome editor's activity.

High-fidelity KKH variant of Staphylococcus aureus Cas9 nucleases with improved base mismatch discrimination.

The Cas9 nuclease from Staphylococcus aureus (SaCas9) holds great potential for use in gene therapy, and variants with increased fidelity have been engineered. However, we find that existing variants have not reached the greatest accuracy to discriminate base mismatches and exhibited much reduced activity when their mutations were grafted onto the KKH mutant of SaCas9 for editing an expanded set of DNA targets. We performed structure-guided combinatorial mutagenesis to re-engineer KKH-SaCas9 with enhanced accuracy. We uncover that introducing a Y239H mutation on KKH-SaCas9's REC domain substantially reduces off-target edits while retaining high on-target activity when added to a set of mutations on REC and RuvC domains that lessen its interactions with the target DNA strand. The Y239H mutation is modelled to have removed an interaction from the REC domain with the guide RNA backbone in the guide RNA-DNA heteroduplex structure. We further confirmed the greatly improved genome-wide editing accuracy and single-base mismatch discrimination of our engineered variants, named KKH-SaCas9-SAV1 and SAV2, in human cells. In addition to generating broadly useful KKH-SaCas9 variants with unprecedented accuracy, our findings demonstrate the feasibility for multi-domain combinatorial mutagenesis on SaCas9's DNA- and guide RNA- interacting residues to optimize its editing fidelity.

Liposome Click Membrane Permeability Assay for Identifying Permeable Peptides.

PURPOSE: To develop a novel, target agnostic liposome click membrane permeability assay (LCMPA) using liposome encapsulating copper free click reagent dibenzo cyclooctyne biotin (DBCO-Biotin) to conjugate azido modified peptides that may effectively translocate from extravesicular space into the liposome lumen. METHOD: DBCO-Biotin liposomes were prepared with egg phosphatidylcholine and cholesterol by lipid film rehydration, freeze/thaw followed by extrusion. Size of DBCO-Biotin liposomes were characterized with dynamic light scattering. RESULTS: The permeable peptides representing energy independent mechanism of permeability showed higher biotinylation in LCMPA. Individual peptide permeability results from LCMPA correlated well with shifts in potency in cellular versus biochemical assays (i.e., cellular/ biochemical ratio) demonstrating quantitative correlation to intracellular barrier in intact cells. CONCLUSION: The study provides a novel membrane permeability assay that has potential to evaluate energy independent transport of diverse peptides.

NanoClick: A High Throughput, Target-Agnostic Peptide Cell Permeability Assay.

Macrocyclic peptides open new opportunities to target intracellular protein-protein interactions (PPIs) that are often considered nondruggable by traditional small molecules. However, engineering sufficient membrane permeability into these molecules is a central challenge for identifying clinical candidates. Currently, there is a lack of high-throughput assays to assess peptide permeability, which limits our capacity to engineer this property into macrocyclic peptides for advancement through drug discovery pipelines. Accordingly, we developed a high throughput and target-agnostic cell permeability assay that measures the relative cumulative cytosolic exposure of a peptide in a concentration-dependent manner. The assay was named NanoClick as it combines in-cell Click chemistry with an intracellular NanoBRET signal. We validated the approach using known cell penetrating peptides and further demonstrated a correlation to cellular activity using a p53/MDM2 model system. With minimal change to the peptide sequence, NanoClick enables the ability to measure uptake of molecules that enter the cell via different mechanisms such as endocytosis, membrane translocation, or passive permeability. Overall, the NanoClick assay can serve as a screening tool to uncover predictive design rules to guide structure-activity-permeability relationships in the optimization of functionally active molecules.

Rigorous Computational and Experimental Investigations on MDM2/MDMX-Targeted Linear and Macrocyclic Peptides.

There is interest in peptide drug design, especially for targeting intracellular protein-protein interactions. Therefore, the experimental validation of a computational platform for enabling peptide drug design is of interest. Here, we describe our peptide drug design platform (CMDInventus) and demonstrate its use in modeling and predicting the structural and binding aspects of diverse peptides that interact with oncology targets MDM2/MDMX in comparison to both retrospective (pre-prediction) and prospective (post-prediction) data. In the retrospective study, CMDInventus modules (CMDpeptide, CMDboltzmann, CMDescore and CMDyscore) were used to accurately reproduce structural and binding data across multiple MDM2/MDMX data sets. In the prospective study, CMDescore, CMDyscore and CMDboltzmann were used to accurately predict binding affinities for an Ala-scan of the stapled α-helical peptide ATSP-7041. Remarkably, CMDboltzmann was used to accurately predict the results of a novel D-amino acid scan of ATSP-7041. Our investigations rigorously validate CMDInventus and support its utility for enabling peptide drug design.

Publisher Correction: Combinatorial mutagenesis en masse optimizes the genome editing activities of SpCas9.

An amendment to this paper has been published and can be accessed via a link at the top of the paper.

Stereoisomerism of stapled peptide inhibitors of the p53-Mdm2 interaction: an assessment of synthetic strategies and activity profiles.

All-hydrocarbon, i, i+7 stapled peptide inhibitors of the p53-Mdm2 interaction have emerged as promising new leads for cancer therapy. Typical chemical synthesis via olefin metathesis results in the formation of both E- and Z-isomers, an observation that is rarely disclosed but may be of importance in targeting PPI. In this study, we evaluated the effect of staple geometry on the biological activity of five p53-reactivating peptides. We also present strategies for the modulation of the E/Z ratio and attainment of the hydrogenated adduct through repurposing of the metathesis catalyst.

Combinatorial mutagenesis en masse optimizes the genome editing activities of SpCas9.

The combined effect of multiple mutations on protein function is hard to predict; thus, the ability to functionally assess a vast number of protein sequence variants would be practically useful for protein engineering. Here we present a high-throughput platform that enables scalable assembly and parallel characterization of barcoded protein variants with combinatorial modifications. We demonstrate this platform, which we name CombiSEAL, by systematically characterizing a library of 948 combination mutants of the widely used Streptococcus pyogenes Cas9 (SpCas9) nuclease to optimize its genome-editing activity in human cells. The ease with which the editing activities of the pool of SpCas9 variants can be assessed at multiple on- and off-target sites accelerates the identification of optimized variants and facilitates the study of mutational epistasis. We successfully identify Opti-SpCas9, which possesses enhanced editing specificity without sacrificing potency and broad targeting range. This platform is broadly applicable for engineering proteins through combinatorial modifications en masse.

Incorporation of Putative Helix-Breaking Amino Acids in the Design of Novel Stapled Peptides: Exploring Biophysical and Cellular Permeability Properties.

Stapled α-helical peptides represent an emerging superclass of macrocyclic molecules with drug-like properties, including high-affinity target binding, protease resistance, and membrane permeability. As a model system for probing the chemical space available for optimizing these properties, we focused on dual Mdm2/MdmX antagonist stapled peptides related to the p53 N-terminus. Specifically, we first generated a library of ATSP-7041 (Chang et al., 2013) analogs iteratively modified by L-Ala and D-amino acids. Single L-Ala substitutions beyond the Mdm2/(X) binding interfacial residues (i.e., Phe3, Trp7, and Cba10) had minimal effects on target binding, α-helical content, and cellular activity. Similar binding affinities and cellular activities were noted at non-interfacial positions when the template residues were substituted with their d-amino acid counterparts, despite the fact that d-amino acid residues typically 'break' right-handed α-helices. d-amino acid substitutions at the interfacial residues Phe3 and Cba10 resulted in the expected decreases in binding affinity and cellular activity. Surprisingly, substitution at the remaining interfacial position with its d-amino acid equivalent (i.e., Trp7 to d-Trp7) was fully tolerated, both in terms of its binding affinity and cellular activity. An X-ray structure of the d-Trp7-modified peptide was determined and revealed that the indole side chain was able to interact optimally with its Mdm2 binding site by a slight global re-orientation of the stapled peptide. To further investigate the comparative effects of d-amino acid substitutions we used linear analogs of ATSP-7041, where we replaced the stapling amino acids by Aib (i.e., R84 to Aib4 and S511 to Aib11) to retain the helix-inducing properties of α-methylation. The resultant analog sequence Ac-Leu-Thr-Phe-Aib-Glu-Tyr-Trp-Gln-Leu-Cba-Aib-Ser-Ala-Ala-NH2 exhibited high-affinity target binding (Mdm2 Kd = 43 nM) and significant α-helicity in circular dichroism studies. Relative to this linear ATSP-7041 analog, several d-amino acid substitutions at Mdm2(X) non-binding residues (e.g., d-Glu5, d-Gln8, and d-Leu9) demonstrated decreased binding and α-helicity. Importantly, circular dichroism (CD) spectroscopy showed that although helicity was indeed disrupted by d-amino acids in linear versions of our template sequence, stapled molecules tolerated these residues well. Further studies on stapled peptides incorporating N-methylated amino acids, l-Pro, or Gly substitutions showed that despite some positional dependence, these helix-breaking residues were also generally tolerated in terms of secondary structure, binding affinity, and cellular activity. Overall, macrocyclization by hydrocarbon stapling appears to overcome the destabilization of α-helicity by helix breaking residues and, in the specific case of d-Trp7-modification, a highly potent ATSP-7041 analog (Mdm2 Kd = 30 nM; cellular EC50 = 600 nM) was identified. Our findings provide incentive for future studies to expand the chemical diversity of macrocyclic α-helical peptides (e.g., d-amino acid modifications) to explore their biophysical properties and cellular permeability. Indeed, using the library of 50 peptides generated in this study, a good correlation between cellular permeability and lipophilicity was observed.

Structure-activity studies of Mdm2/Mdm4-binding stapled peptides comprising non-natural amino acids.

As primary p53 antagonists, Mdm2 and the closely related Mdm4 are relevant cancer therapeutic targets. We have previously described a series of cell-permeable stapled peptides that bind to Mdm2 with high affinity, resulting in activation of the p53 tumour suppressor. Within this series, highest affinity was obtained by modification of an obligate tryptophan residue to the non-natural L-6-chlorotryptophan. To understand the structural basis for improved affinity we have solved the crystal structure of this stapled peptide (M011) bound to Mdm2 (residues 6-125) at 1.66 Å resolution. Surprisingly, near identity to the structure of a related peptide (M06) without the 6-chloro modification is observed. Further analysis of linear and stapled peptides comprising 6-Me-tryptophan provides mechanistic insight into dual Mdm2/Mdm4 antagonism and confirms L98 of Mdm4 as a mutable steric gate. The results also highlight a possible role of the flexible hinge region in determining Mdm2/Mdm4 plasticity.

Benzene Probes in Molecular Dynamics Simulations Reveal Novel Binding Sites for Ligand Design.

Protein flexibility poses a major challenge in binding site identification. Several computational pocket detection methods that utilize small-molecule probes in molecular dynamics (MD) simulations have been developed to address this issue. Although they have proven hugely successful at reproducing experimental structural data, their ability to predict new binding sites that are yet to be identified and characterized has not been demonstrated. Here, we report the use of benzenes as probe molecules in ligand-mapping MD (LMMD) simulations to predict the existence of two novel binding sites on the surface of the oncoprotein MDM2. One of them was serendipitously confirmed by biophysical assays and X-ray crystallography to be important for the binding of a new family of hydrocarbon stapled peptides that were specifically designed to target the other putative site. These results highlight the predictive power of LMMD and suggest that predictions derived from LMMD simulations can serve as a reliable basis for the identification of novel ligand binding sites in structure-based drug design.