Search

There has been considerable progress in the development of computational methods for designing protein-protein interactions, but engineering high-affinity binders without extensive screening and maturation remains challenging. Here, we test a protein design pipeline that uses iterative rounds of deep learning (DL)-based structure prediction (AlphaFold2) and sequence optimization (ProteinMPNN) to design autoinhibitory domains (AiDs) for a PD-L1 antagonist. Inspired by recent advances in therapeutic design, we sought to create autoinhibited (or masked) forms of the antagonist that can be conditionally activated by proteases. Twenty-threede novodesigned AiDs, varying in length and topology, were fused to the antagonist with a protease sensitive linker, and binding to PD-L1 was tested with and without protease treatment. Nine of the fusion proteins demonstrated conditional binding to PD-L1 and the top performing AiDs were selected for further characterization as single domain proteins. Without any experimental affinity maturation, four of the AiDs bind to the PD-L1 antagonist with equilibrium dissociation constants (KDs) below 150 nM, with the lowest KDequal to 0.9 nM. Our study demonstrates that DL-based protein modeling can be used to rapidly generate high affinity protein binders.Significance statementProtein-protein interactions are critical to most processes in biology, and improved methods for designing protein binders will enable the creation of new research reagents, diagnostics, and therapeutics. In this study, we show that a deep learning-based method for protein design can create high-affinity protein binders without the need for extensive screening or affinity maturation.

Supplementary Figures and Legends

Chimeric antigen receptor (CAR) tonic signaling, defined as spontaneous activation and release of proinflammatory cytokines by CAR-T cells, is considered a negative attribute because it leads to impaired antitumor effects. Here, we report that CAR tonic signaling is caused by the intrinsic instability of the mAb single-chain variable fragment (scFv) to promote self-aggregation and signaling via the CD3ζ chain incorporated into the CAR construct. This phenomenon was detected in a CAR encoding either CD28 or 4-1BB costimulatory endodomains. Instability of the scFv was caused by specific amino acids within the framework regions (FWR) that can be identified by computational modeling. Substitutions of the amino acids causing instability, or humanization of the FWRs, corrected tonic signaling of the CAR, without modifying antigen specificity, and enhanced the antitumor effects of CAR-T cells. Overall, we demonstrated that tonic signaling of CAR-T cells is determined by the molecular instability of the scFv and that computational analyses of the scFv can be implemented to correct the scFv instability in CAR-T cells with either CD28 or 4-1BB costimulation.

Many protein therapeutics are competitive inhibitors that function by binding to endogenous proteins and preventing them from interacting with native partners. One effective strategy for engineering competitive inhibitors is to graft structural motifs from a native partner into a host protein. Here, we develop and experimentally test a computational protocol for embedding binding motifs in de novo designed proteins. The protocol uses an “inside-out” approach: Starting with a structural model of the binding motif docked against the target protein, the de novo protein is built by growing new structural elements off the termini of the binding motif. During backbone assembly, a score function favors backbones that introduce new tertiary contacts within the designed protein and do not introduce clashes with the target binding partner. Final sequences are designed and optimized using the molecular modeling program Rosetta. To test our protocol, we designed small helical proteins to inhibit the interaction between Gαqand its effector PLC-β isozymes. Several of the designed proteins remain folded above 90°C and bind to Gαqwith equilibrium dissociation constants tighter than 80 nM. In cellular assays with oncogenic variants of Gαq, the designed proteins inhibit activation of PLC-β isozymes and Dbl-family RhoGEFs. Our results demonstrate that computational protein design, in combination with motif grafting, can be used to directly generate potent inhibitors without further optimization via high throughput screening or selection.statement for broader audienceEngineered proteins that bind to specific target proteins are useful as research reagents, diagnostics, and therapeutics. We used computational protein design to engineer de novo proteins that bind and competitively inhibit the G protein, Gαq, which is an oncogene for uveal melanomas. This computational method is a general approach that should be useful for designing competitive inhibitors against other proteins of interest.

Many proteins have low thermodynamic stability, which can lead to low expression yields and limit functionality in research, industrial and clinical settings. This article introduces two, web-based tools that use the high-resolution structure of a protein along with the Rosetta molecular modeling program to predict stabilizing mutations. The protocols were recently applied to three genetically and structurally distinct proteins and successfully predicted mutations that improved thermal stability and/or protein yield. In all three cases, combining the stabilizing mutations raised the protein unfolding temperatures by more than 20°C. The first protocol evaluates point mutations and can generate a site saturation mutagenesis heatmap. The second identifies mutation clusters around user-defined positions. Both applications only require a protein structure and are particularly valuable when a deep multiple sequence alignment is not available. These tools were created to simplify protein engineering and enable research that would otherwise be infeasible due to poor expression and stability of the native molecule.

Using the molecular modeling program Rosetta, we designed a de novo protein, called SEWN0.1, which binds the heterotrimeric G protein Gα(q.) The design is helical, well‐folded, and primarily monomeric in solution at a concentration of 10 μM. However, when we solved the crystal structure of SEWN0.1 at 1.9 Å, we observed a dimer in a conformation incompatible with binding Gα(q). Unintentionally, we had designed a protein that adopts alternate conformations depending on its oligomeric state. Recently, there has been tremendous progress in the field of protein structure prediction as new methods in artificial intelligence have been used to predict structures with high accuracy. We were curious if the structure prediction method AlphaFold could predict the structure of SEWN0.1 and if the prediction depended on oligomeric state. When AlphaFold was used to predict the structure of monomeric SEWN0.1, it produced a model that resembles the Rosetta design model and is compatible with binding Gα(q), but when used to predict the structure of a dimer, it predicted a conformation that closely resembles the SEWN0.1 crystal structure. AlphaFold's ability to predict multiple conformations for a single protein sequence should be useful for engineering protein switches.

Promiscuous enzymes that modify peptides and proteins are powerful tools for labeling biomolecules; however, directing these modifications to desired substrates can be challenging. Here, we use computational interface design to install a substrate recognition domain adjacent to the active site of a promiscuous enzyme, catechol O-methyltransferase. This design approach effectively decouples substrate recognition from the site of catalysis and promotes modification of peptides recognized by the recruitment domain. We determined the crystal structure of this novel multidomain enzyme, SH3-588, which shows that it closely matches our design. SH3-588 methylates directed peptides with catalytic efficiencies exceeding the wild-type enzyme by over 1,000-fold, whereas peptides lacking the directing recognition sequence do not display enhanced efficiencies. In competition experiments, the designer enzyme preferentially modifies directed substrates over undirected substrates, suggesting that we can use designed recruitment domains to direct post-translational modifications to specific sequence motifs on target proteins in complex multisubstrate environments.

Understanding protein folding is crucial for protein sciences. The conformational spaces and energy landscapes of cold (unfolded) protein states, as well as the associated transitions, are hardly explored. Furthermore, it is not known how structure relates to the cooperativity of cold transitions, if cold and heat unfolded states are thermodynamically similar, and if cold states play important roles for protein function. We created the cold unfolding 4-helix bundle DCUB1 with a de novo designed bipartite hydrophilic/hydrophobic core featuring a hydrogen bond network which extends across the bundle in order to study the relative importance of hydrophobic versus hydrophilic protein-water interactions for cold unfolding. Structural and thermodynamic characterization resulted in the discovery of a complex energy landscape for cold transitions, while the heat unfolded state is a random coil. Below ∼0 °C, the core of DCUB1 disintegrates in a largely cooperative manner, while a near-native helical content is retained. The resulting cold core-unfolded state is compact and features extensive internal dynamics. Below -5 °C, two additional cold transitions are seen, that is, (i) the formation of a water-mediated, compact, and highly dynamic dimer, and (ii) the onset of cold helix unfolding decoupled from cold core unfolding. Our results suggest that cold unfolding is initiated by the intrusion of water into the hydrophilic core network and that cooperativity can be tuned by varying the number of core hydrogen bond networks. Protein design has proven to be invaluable to explore the energy landscapes of cold states and to robustly test related theories.

Using the molecular modeling program Rosetta, we designed a de novo protein, called SEWN0.1, that binds the heterotrimeric G protein Gαq. The design is helical, well-folded, and primarily monomeric in solution at a concentration of 10 uM. However, when we solved the crystal structure of SEWN0.1, we observed a dimer in a conformation incompatible with binding Gαq. Unintentionally, we had designed a protein that adopts alternate conformations depending on its oligomeric state. Recently, there has been tremendous progress in the field of protein structure prediction as new methods in artificial intelligence have been used to predict structures with high accuracy. We were curious if the structure prediction method AlphaFold could predict the structure of SEWN0.1 and if the prediction depended on oligomeric state. When AlphaFold was used to predict the structure of monomeric SEWN0.1, it produced a model that resembles the Rosetta design model and is compatible with binding Gαq, but when used to predict the structure of a dimer, it predicted a conformation that closely resembles the SEWN0.1 crystal structure. AlphaFold’s ability to predict multiple conformations for a single protein sequence should be useful for engineering protein switches.

We report the measurement and analysis of sulfonium-π, thioether-π, and ammonium-π interactions in a β-hairpin peptide model system, coupled with computational investigation and PDB analysis. These studies indicated that the sulfonium-π interaction is the strongest and that polarizability contributes to the stronger interaction with sulfonium relative to ammonium. Computational studies demonstrate that differences in solvation of the trimethylsulfonium versus the trimethylammonium group also contribute to the stronger sulfonium-π interaction. In comparing sulfonium-π versus sulfur-π interactions in proteins, analysis of SAM- and SAH-bound enzymes in the PDB suggests that aromatic residues are enriched in close proximity to the sulfur of both SAM and SAH, but the populations of aromatic interactions of the two cofactors are not significantly different, with the exception of the Me-π interactions in SAM, which are the most prevalent interaction in SAM but are not possible for SAH. This suggests that the weaker interaction energies due to loss of the cation-π interaction in going from SAM to SAH may contribute to turnover of the cofactor.

Protein engineering has contributed to successes in the field of T cell-based immunotherapy, including chimeric antigen receptor (CAR) T cell therapy. CAR T cell therapy has become a pillar of cancer immunotherapy, demonstrating clinical effectiveness against B cell malignancies by targeting the B cell antigen CD19. Current gene editing techniques have limited safety controls over CAR T cell activity, which presents a hurdle for control of CAR T cells in patients. Alternatively, CAR T cell activity can be controlled by engineering CARs to bind soluble adapter molecules that direct the interaction between the CAR T cell and target cell. The flexibility in this adapter-mediated approach overcomes the rigid specificity of traditional CAR T cells to allow targeting of multiple cell types. Here we describe adapter CAR T technologies and how these methods emphasize the growing role of protein engineering in the design of programmable tools for T cell therapies.

Dengue viruses (DENV serotypes 1-4) and Zika virus (ZIKV) are related flaviviruses that continue to be a public health concern, infecting hundreds of millions of people annually. The traditional live-attenuated virus vaccine approach has been challenging for the four DENV serotypes because of the need to achieve balanced replication of four independent vaccine components. Subunit vaccines represent an alternative approach that may circumvent problems inherent with live-attenuated DENV vaccines. In mature virus particles, the envelope (E) protein forms a homodimer that covers the surface of the virus and is the major target of neutralizing antibodies. Many neutralizing antibodies bind to quaternary epitopes that span across both E proteins in the homodimer. For soluble E (sE) protein to be a viable subunit vaccine, the antigens should be easy to produce and retain quaternary epitopes recognized by neutralizing antibodies. However, WT sE proteins are primarily monomeric at conditions relevant for vaccination and exhibit low expression yields. Previously, we identified amino acid mutations that stabilize the sE homodimer from DENV2 and dramatically raise expression yields. Here, we tested whether these same mutations raise the stability of sE from other DENV serotypes and ZIKV. We show that the mutations raise thermostability for sE from all the viruses, increase production yields from 4-fold to 250-fold, stabilize the homodimer, and promote binding to dimer-specific neutralizing antibodies. Our findings suggest that these sE variants could be valuable resources in the efforts to develop effective subunit vaccines for DENV serotypes 1 to 4 and ZIKV.

Engineered, light-sensitive protein switches are used to interrogate a broad variety of biological processes. These switches are typically constructed by genetically fusing naturally occurring light-responsive protein domains with functional domains from other proteins. Protein activity can be controlled using a variety of mechanisms including light-induced colocalization, caging, and allosteric regulation. Protein design efforts have focused on reducing background signaling, maximizing the change in activity upon light stimulation, and perturbing the kinetics of switching. It is common to combine structure-based modeling with experimental screening to identify ideal fusion points between domains and discover point mutations that optimize switching. Here, we introduce commonly used light-sensitive domains and summarize recent progress in using them to regulate protein activity.