Mark J. S. Kelly

The ZF5.3 mini-protein escapes endosomes efficiently to guide proteins into the cytosol and/or nucleus. However, other than the requirement for small or unfoldable cargo and an intact HOPS complex, little is known about how ZF5.3 traverses the limiting endocytic membrane. Here we characterize the requirements for efficient endosomal escape. We confirm that ZF5.3 remains folded at high temperatures and at pH values between 5.5 and 7.5. At lower pH, ZF5.3 unfolds cooperatively upon protonation of Zn(II)-binding His side chains whose pK_a matches that of the late endolysosomal lumen. pH-induced unfolding of ZF5.3 is essential for endosomal escape, as an analogue that remains folded at low pH fails to efficiently reach the cytosol. Once unfolded, ZF5.3 interacts in a pH-dependent manner with bis(monoacylglycero)phosphate, a lipid present in the inner leaflet of late endolysosomal membranes. These data provide a biophysical model for HOPS-dependent endosomal escape.

Deep learning has advanced the design of static protein structures, but the controlled conformational changes that are hallmarks of natural signaling proteins have remained inaccessible to de novo design. Here, we describe a general deep learning–guided approach for de novo design of dynamic changes between intradomain geometries of proteins, similar to switch mechanisms prevalent in nature, with atomic-level precision. We solve four structures that validate the designed conformations, demonstrate modulation of the conformational landscape by orthosteric ligands and allosteric mutations, and show that physics-based simulations are in agreement with deep-learning predictions and experimental data. Our approach demonstrates that new modes of motion can now be realized through de novo design and provides a framework for constructing biology-inspired, tunable, and controllable protein signaling behavior de novo.

De novo protein design provides a framework to test our understanding of protein function and build proteins with cofactors and functions not found in nature. Here, we report the design of proteins designed to bind powerful photooxidants and the evaluation of the use of these proteins to generate diffusible small-molecule reactive species. Because excited-state dynamics are influenced by the dynamics and hydration of a photooxidant's environment, it was important to not only design a binding site but also to evaluate its dynamic properties. Thus, we used computational design in conjunction with molecular dynamics (MD) simulations to design a protein, designated NBP (NDI Binding Protein), that held a naphthalenediimide (NDI), a powerful photooxidant, in a programmable molecular environment. Solution NMR confirmed the structure of the complex. We evaluated two NDI cofactors in this de novo protein using ultrafast pump-probe spectroscopy to evaluate light-triggered intra- and intermolecular electron transfer function. Moreover, we demonstrated the utility of this platform to activate multiple molecular probes for protein labeling.

De novo protein design provides a framework to test our understanding of protein function and to build proteins with cofactors and functions not found in nature. Here, we report the design of proteins designed to bind powerful photooxidants and the evaluation of the use of these proteins to generate diffusible small molecule reactive species for applications in proximity labeling. Because excited state dynamics are influenced by the dynamics and hydration of a photo-oxidant’s environment, it was important to not only design a binding site, but also to evaluate its dynamic properties. Thus, we used computational design in conjunction with molecular dynamics (MD) simulations to design a protein, designated NBP ( N DI B inding P rotein) that held a naphthalenediimide (NDI), a powerful photooxidant, in a programable molecular environment. Solution NMR confirmed the structure of the complex. We evaluated two NDI cofactors in this de novo protein, using ultra-fast pump-probe spectroscopy to evaluate light-triggered intra- and intermolecular electron transfer function. Moreover, we demonstrated the utility of this platform to activate multiple molecular probes for protein proximity labeling.

Phage-encoded anti-CRISPR (Acr) proteins inhibit CRISPR-Cas systems to allow phage replication and lysogeny maintenance. Most of the Acrs characterized to date are stable stoichiometric inhibitors. While enzymatic Acrs have been characterized biochemically, little is known about their potency, specificity, and reversibility. Here, we examine AcrIF11, a widespread phage and plasmid-encoded ADP-ribosyltransferase (ART) that inhibits the Type I-F CRISPR-Cas system. We present an NMR structure of an AcrIF11 homolog that reveals chemical shift perturbations consistent with NAD (cofactor) binding. In experiments that model both lytic phage replication and MGE/lysogen stability under high targeting pressure, AcrIF11 is a highly potent CRISPR-Cas inhibitor and more robust to Cas protein level fluctuations than stoichiometric inhibitors. Furthermore, we demonstrate that AcrIF11 is remarkably specific, predominantly ADP-ribosylating Csy1 when expressed in P. aeruginosa . Given the reversible nature of ADP-ribosylation, we hypothesized that ADPr eraser enzymes (macrodomains) could remove ADPr from Csy1, a potential limitation of PTM-based CRISPR inhibition. We demonstrate that diverse macrodomains can indeed remove the modification from Csy1 in P. aeruginosa lysate. Together, these experiments connect the in vitro observations of AcrIF11’s enzymatic activity to its potent and specific effects in vivo , clarifying the advantages and drawbacks of enzymatic Acrs in the evolutionary arms race between phages and bacteria.

Deep learning has greatly advanced design of highly stable static protein structures, but the controlled conformational dynamics that are hallmarks of natural switch-like signaling proteins have remained inaccessible to de novo design. Here, we describe a general deep-learning-guided approach for de novo design of dynamic changes between intra-domain geometries of proteins, similar to switch mechanisms prevalent in nature, with atom-level precision. We solve 4 structures validating the designed conformations, show microsecond transitions between them, and demonstrate that the conformational landscape can be modulated by orthosteric ligands and allosteric mutations. Physics-based simulations are in remarkable agreement with deep-learning predictions and experimental data, reveal distinct state-dependent residue interaction networks, and predict mutations that tune the designed conformational landscape. Our approach demonstrates that new modes of motion can now be realized through de novo design and provides a framework for constructing biology-inspired, tunable and controllable protein signaling behavior de novo .

ZF5.3 is a compact, rationally designed mini-protein that escapes efficiently from the endosomes of multiple cell types. Despite its small size (27 amino acids), ZF5.3 can be isolated intact from the cytosol of treated cells and guides multiple classes of proteins into the cytosol and/or nucleus. In the best cases, delivery efficiencies reach or exceed 50% to establish nuclear or cytosolic concentrations of 500 nM or higher. But other than the requirement for unfoldable cargo and an intact HOPS complex, there is little known about how ZF5.3 traverses the limiting endocytic membrane. Here we delineate the attributes of ZF5.3 that enable efficient endosomal escape. We confirm that ZF5.3 is stable at pH values between 5.5 and 7.5, with no evidence of unfolding even at temperatures as high as 95 °C. The high-resolution NMR structure of ZF5.3 at pH 5.5, also reported here, shows a canonical p zinc-finger fold with the penta-arg motif integrated seamlessly into the C-terminal ⍺-helix. At lower pH, ZF5.3 unfolds cooperatively as judged by both circular dichroism and high-resolution NMR. Unfolding occurs upon protonation of a single Zn(II)-binding His side chain whose p K _a corresponds almost exactly to that of the late endosomal lumen. pH-induced unfolding is essential for endosomal escape, as a ZF5.3 analog that remains folded at pH 4.5 fails to efficiently reach the cytosol, despite high overall uptake. Finally, using reconstituted liposomes, we identify a high-affinity interaction of ZF5.3 with a specific lipid–BMP–that is selectively enriched in the inner leaflet of late endosomal membranes. This interaction is 10-fold stronger at low pH than neutral pH, providing a molecular picture for why escape occurs preferentially and in a HOPS-dependent manner from late endosomal compartments. The requirements for programmed endosomal escape identified here should aid and inform the design of proteins, peptidomimetics, and other macromolecules that reach cytosolic or nuclear targets intact and at therapeutically relevant concentrations.

While there has been recent success in the development of KRas^G12C inhibitors, unmet needs for selective inhibitors of KRas^G12D and the remaining oncogenic KRas proteins remain. Here, we applied trifluoromethyl-containing ligands of KRas proteins as competitive probe ligands to assay the occupancy of the switch II pocket by ¹⁹F NMR spectroscopy. Structure-activity-relationship studies of probe ligands increased the sensitivity of the assay and identified structures that differentially detected each nucleotide state of KRas^G12D. These differences in selectivity, combined with the high resolution of ¹⁹F NMR spectroscopy, enabled this method to be expanded to assay both nucleotide states of the protein simultaneously.

While there has been recent success in the development of KRas ^G12C inhibitors, unmet needs for selective inhibitors and tool compounds targeting the remaining oncogenic KRas proteins remain. Here, we applied trifluoromethyl-containing ligands of KRas proteins as competitive probe ligands to assay the occupancy of the switch II pocket by ¹⁹ F NMR spectroscopy. Structure-activity-relationship studies of probe ligands increased the sensitivity of the assay and identified structures that differentially detected each nucleotide state of KRas ^G12D . These differences in selectivity, combined with the high resolution of ¹⁹ F NMR spectroscopy, enabled this method to be expanded to assay both nucleotide states of the protein simultaneously. Abstract Figure:

PHD reader domains are chromatin binding modules often responsible for the recruitment of large protein complexes that contain histone modifying enzymes, chromatin remodelers, and DNA repair machinery. A majority of PHD domains recognize N-terminal residues of histone H3 and are sensitive to the methylation state of Lys4 in histone H3 (H3K4). Histone demethylase KDM5A, an epigenetic eraser enzyme that contains three PHD domains, is often overexpressed in various cancers, and its demethylation activity is allosterically enhanced when its PHD1 domain is bound to the H3 tail. The allosteric regulatory function of PHD1 expands roles of reader domains, suggesting unique features of this chromatin interacting module. Our previous studies determined the H3 binding site of PHD1, although it remains unclear how the H3 tail interacts with the N-terminal residues of PHD1 and how PHD1 discriminates against H3 tails with varying degrees of H3K4 methylation. Here, we have determined the solution structure of apo and H3 bound PHD1. We observe conformational changes occurring in PHD1 in order to accommodate H3, which interestingly binds in a helical conformation. We also observe differential interactions of binding residues with differently methylated H3K4 peptides (me0, me1, me2, or me3), providing a rationale for PHD1’s preference for lower methylation states of H3K4. We further assessed the contributions of various H3 interacting residues in the PHD1 domain to the binding of H3 peptides. The structural details of the H3 binding site could provide useful information to aid the development of allosteric small molecule modulators of KDM5A.