Engineering human JMJD2A tudor domains for an improved understanding of histone peptide recognition.

JMJD2A is a histone lysine demethylase which recognizes and demethylates H3K9me3 and H3K36me3 residues and is overexpressed in various cancers. It utilizes a tandem tudor domain to facilitate its own recruitment to histone sites, recognizing various di- and tri-methyl lysine residues with moderate affinity. In this study, we successfully engineered the tudor domain of JMJD2A to specifically bind to H4K20me3 with a 20-fold increase of affinity and improved selectivity. To reveal the molecular basis, we performed molecular dynamics and free energy decomposition analysis on the human JMJD2A tandem tudor domains bound to H4K20me2, H4K20me3, and H3K23me3 peptides to uncover the residues and conformational changes important for the enhanced binding affinity and selectivity toward H4K20me2/3. These analyses revealed new insights into understanding chromatin reader domains recognizing histone modifications and improving binding affinity and selectivity of these domains. Furthermore, we showed that the tight binding of JMJD2A to H4K20me2/3 is not sufficient to improve the efficiency of CRISPR-CAS9 mediated homology directed repair (HDR), suggesting a complicated relationship between JMJD2A and the DNA damage response beyond binding affinity toward the H4K20me2/3 mark.

Engineering a Histone Reader Protein by Combining Directed Evolution, Sequencing, and Neural Network Based Ordinal Regression.

Directed evolution is a powerful approach for engineering proteins with enhanced affinity or specificity for a ligand of interest but typically requires many rounds of screening/library mutagenesis to obtain mutants with desired properties. Furthermore, mutant libraries generally only cover a small fraction of the available sequence space. Here, for the first time, we use ordinal regression to model protein sequence data generated through successive rounds of sorting and amplification of a protein-ligand system. We show that the ordinal regression model trained on only two sorts successfully predicts chromodomain CBX1 mutants that would have stronger binding affinity with the H3K9me3 peptide. Furthermore, we can extract the predictive features using contextual regression, a method to interpret nonlinear models, which successfully guides identification of strong binders not even present in the original library. We have demonstrated the power of this approach by experimentally confirming that we were able to achieve the same improvement in binding affinity previously achieved through a more laborious directed evolution process. This study presents an approach that reduces the number of rounds of selection required to isolate strong binders and facilitates the identification of strong binders not present in the original library.

Discovery and Mechanism of Highly Efficient Cyclic Cell-Penetrating Peptides.

Previous cell-penetrating peptides (CPPs) generally have low cytosolic delivery efficiencies, because of inefficient endosomal escape. In this study, a family of small, amphipathic cyclic peptides was found to be highly efficient CPPs, with cytosolic delivery efficiencies of up to 120% (compared to 2.0% for Tat). These cyclic CPPs bind directly to the plasma membrane phospholipids and enter mammalian cells via endocytosis, followed by efficient release from the endosome. Their total cellular uptake efficiency correlates positively with the binding affinity for the plasma membrane, whereas their endosomal escape efficiency increases with the endosomal membrane-binding affinity. The cyclic CPPs induce membrane curvature on giant unilamellar vesicles and budding of small vesicles, which subsequently collapse into amorphous lipid/peptide aggregates. These data suggest that cyclic CPPs exit the endosome by binding to the endosomal membrane and inducing CPP-enriched lipid domains to bud off as small vesicles. Together with their high proteolytic stability, low cytotoxicity, and oral bioavailability, these cyclic CPPs should provide a powerful system for intracellular delivery of therapeutic agents and chemical probes.

Early endosomal escape of a cyclic cell-penetrating peptide allows effective cytosolic cargo delivery.

Cyclic heptapeptide cyclo(FΦRRRRQ) (cFΦR4, where Φ is l-2-naphthylalanine) was recently found to be efficiently internalized by mammalian cells. In this study, its mechanism of internalization was investigated by perturbing various endocytic events through the introduction of pharmacologic agents and genetic mutations. The results show that cFΦR4 binds directly to membrane phospholipids, is internalized into human cancer cells through endocytosis, and escapes from early endosomes into the cytoplasm. Its cargo capacity was examined with a wide variety of molecules, including small-molecule dyes, linear and cyclic peptides of various charged states, and proteins. Depending on the nature of the cargos, they may be delivered by endocyclic (insertion of cargo into the cFΦR4 ring), exocyclic (attachment of cargo to the Gln side chain), or bicyclic approaches (fusion of cFΦR4 and cyclic cargo rings). The overall delivery efficiency (i.e., delivery of cargo into the cytoplasm and nucleus) of cFΦR4 was 4-12-fold higher than those of nonaarginine, HIV Tat-derived peptide, or penetratin. The higher delivery efficiency, coupled with superior serum stability, minimal toxicity, and synthetic accessibility, renders cFΦR4 a useful transporter for intracellular cargo delivery and a suitable system for investigating the mechanism of endosomal escape.

The RESP AI model accelerates the identification of tight-binding antibodies.

High-affinity antibodies are often identified through directed evolution, which may require many iterations of mutagenesis and selection to find an optimal candidate. Deep learning techniques hold the potential to accelerate this process but the existing methods cannot provide the confidence interval or uncertainty needed to assess the reliability of the predictions. Here we present a pipeline called RESP for efficient identification of high affinity antibodies. We develop a learned representation trained on over 3 million human B-cell receptor sequences to encode antibody sequences. We then develop a variational Bayesian neural network to perform ordinal regression on a set of the directed evolution sequences binned by off-rate and quantify their likelihood to be tight binders against an antigen. Importantly, this model can assess sequences not present in the directed evolution library and thus greatly expand the search space to uncover the best sequences for experimental evaluation. We demonstrate the power of this pipeline by achieving a 17-fold improvement in the KD of the PD-L1 antibody Atezolizumab and this success illustrates the potential of RESP in facilitating general antibody development.

Distinct ligand specificity of the Tiam1 and Tiam2 PDZ domains.

Guanine nucleotide exchange factor proteins of the Tiam family are activators of the Rho GTPase Rac1 and critical for cell morphology, adhesion, migration, and polarity. These proteins are modular and contain a variety of interaction domains, including a single post-synaptic density-95/discs large/zonula occludens-1 (PDZ) domain. Previous studies suggest that the specificities of the Tiam1 and Tiam2 PDZ domains are distinct. Here, we sought to conclusively define these specificities and determine their molecular origin. Using a combinatorial peptide library, we identified a consensus binding sequence for each PDZ domain. Analysis of these consensus sequences and binding assays with peptides derived from native proteins indicated that these two PDZ domains have overlapping but distinct specificities. We also identified residues in two regions (S(0) and S(-2) pockets) of the Tiam1 PDZ domain that are important determinants of ligand specificity. Site-directed mutagenesis of four nonconserved residues in these two regions along with peptide binding analyses confirmed that these residues are crucial for ligand affinity and specificity. Furthermore, double mutant cycle analysis of each region revealed energetic couplings that were dependent on the ligand being investigated. Remarkably, a Tiam1 PDZ domain quadruple mutant had the same specificity as the Tiam2 PDZ domain. Finally, analysis of Tiam family PDZ domain sequences indicated that the PDZ domains segregate into four distinct families based on the residues studied here. Collectively, our data suggest that Tiam family proteins have highly evolved PDZ domain-ligand interfaces with distinct specificities and that they have disparate PDZ domain-dependent biological functions.

Simultaneous binding of two peptidyl ligands by a SRC homology 2 domain.

Src homology 2 (SH2) domains mediate protein-protein interactions by recognizing phosphotyrosine (pY)-containing sequences of target proteins. In all of the SH2 domain-pY peptide interactions described to date, the SH2 domain binds to a single pY peptide. Here, determination of the cocrystal structure of the N-terminal SH2 domain of phosphatase SHP-2 bound to a class IV peptide (VIpYFVP) revealed a noncanonical 1:2 (protein-peptide) complex. The first peptide binds in a canonical manner with its pY side chain inserted in the usual binding pocket, while the second pairs up with the first to form two antiparallel ß-strands that extend the central ß-sheet of the SH2 domain. This unprecedented binding mode was confirmed in the solution phase by NMR experiments and shown to be adopted by pY peptides derived from cellular proteins. Site-directed mutagenesis and surface plasmon resonance studies revealed that the binding of the first peptide is pY-dependent, but phosphorylation is not required for the second peptide. Our findings suggest a potential new function for the SH2 domain as a molecular clamp to promote dimerization of signaling proteins.

HDAC6 and Ubp-M BUZ domains recognize specific C-terminal sequences of proteins.

The BUZ/Znf-UBP domain is a protein module found in the cytoplasmic deacetylase HDAC6, E3 ubiquitin ligase BRAP2/IMP, and a subfamily of ubiquitin-specific proteases. Although several BUZ domains have been shown to bind ubiquitin with high affinity by recognizing its C-terminal sequence (RLRGG-COOH), it is currently unknown whether the interaction is sequence-specific or whether the BUZ domains are capable of binding to proteins other than ubiquitin. In this work, the BUZ domains of HDAC6 and Ubp-M were subjected to screening against a one-bead-one-compound (OBOC) peptide library that exhibited random peptide sequences with free C-termini. Sequence analysis of the selected binding peptides as well as alanine scanning studies revealed that the BUZ domains require a C-terminal Gly-Gly motif for binding. At the more N-terminal positions, the two BUZ domains have distinct sequence specificities, allowing them to bind to different peptides and/or proteins. A database search of the human proteome on the basis of the BUZ domain specificities identified 11 and 24 potential partner proteins for Ubp-M and HDAC6 BUZ domains, respectively. Peptides corresponding to the C-terminal sequences of four of the predicted binding partners (FBXO11, histone H4, PTOV1, and FAT10) were synthesized and tested for binding to the BUZ domains by fluorescence polarization. All four peptides bound to the HDAC6 BUZ domain with low micromolar K(D) values and less tightly to the Ubp-M BUZ domain. Finally, in vitro pull-down assays showed that the Ubp-M BUZ domain was capable of binding to the histone H3-histone H4 tetramer protein complex. Our results suggest that BUZ domains are sequence-specific protein-binding modules, with each BUZ domain potentially binding to a different subset of proteins.

Deciphering and engineering chromodomain-methyllysine peptide recognition.

Posttranslational modifications (PTMs) play critical roles in regulating protein functions and mediating protein-protein interactions. An important PTM is lysine methylation that orchestrates chromatin modifications and regulates functions of non-histone proteins. Methyllysine peptides are bound by modular domains, of which chromodomains are representative. Here, we conducted the first large-scale study of chromodomains in the human proteome interacting with both histone and non-histone methyllysine peptides. We observed significant degenerate binding between chromodomains and histone peptides, i.e., different histone sites can be recognized by the same set of chromodomains, and different chromodomains can share similar binding profiles to individual histone sites. Such degenerate binding is not dictated by amino acid sequence or PTM motif but rather rooted in the physiochemical properties defined by the PTMs on the histone peptides. This molecular mechanism is confirmed by the accurate prediction of the binding specificity using a computational model that captures the structural and energetic patterns of the domain-peptide interaction. To further illustrate the power and accuracy of our model, we used it to effectively engineer an exceptionally strong H3K9me3-binding chromodomain and to label H3K9me3 in live cells. This study presents a systematic approach to deciphering domain-peptide recognition and reveals a general principle by which histone modifications are interpreted by reader proteins, leading to dynamic regulation of gene expression and other biological processes.