Site-Specific Conjugation of Metal-Chelating Polymers to Anti-Frizzled-2 Antibodies via Microbial Transglutaminase.

Metal-chelating polymer-based radioimmunoconjugates (RICs) are effective agents for radioimmunotherapy but are currently limited by nonspecific binding and off-target organ uptake. Nonspecific binding appears after conjugation of the polymer to the antibody and may be related to random lysine conjugation since the polymers themselves do not bind to cells. To investigate the role of conjugation sites on nonspecific binding of polymer RICs, we developed a microbial transglutaminase reaction to prepare site-specific antibody-polymer conjugates. The reaction was enabled by introducing a Q-tag (i.e., 7M48) into antibody (i.e., Fab) fragments and synthesizing a polyglutamide-based metal-chelating polymer with a PEG amine block to yield substrates. Mass spectrometric analyses confirmed that the microbial transglutaminase conjugation reaction was site-specific. For comparison, random lysine conjugation analogs with an average of one polymer per Fab were prepared by bis-aryl hydrazone conjugation. Conjugates were prepared from an anti-frizzled-2 Fab to target the Wnt pathway, along with a nonbinding specificity control, anti-Luciferase Fab. Fabs were engineered from a trastuzumab-based IgG1 framework and lack lysines in the antigen-binding site. Conjugates were analyzed for thermal conformational stability by differential scanning fluorimetry, which showed that the site-specific conjugate had a similar melting temperature to the parent Fab. Binding assays by biolayer interferometry showed that the site-specific anti-frizzled-2 conjugate maintained high affinity to the antigen, while the random conjugate showed a 10-fold decrease in affinity, which was largely due to changes in association rates. Radioligand cell-binding assays on frizzled-2+ PANC-1 cells and frizzled-2- CHO cells showed that the site-specific anti-frizzled-2 conjugate had ca. 4-fold lower nonspecific binding compared to the random conjugate. Site-specific conjugation appeared to reduce nonspecific binding associated with random conjugation of the polymer in polymer RICs.

Structural and functional validation of a highly specific Smurf2 inhibitor.

Smurf1 and Smurf2 are two closely related member of the HECT (homologous to E6AP carboxy terminus) E3 ubiquitin ligase family and play important roles in the regulation of various cellular processes. Both were initially identified to regulate transforming growth factor-ß and bone morphogenetic protein signaling pathways through regulating Smad protein stability and are now implicated in various pathological processes. Generally, E3 ligases, of which over 800 exist in humans, are ideal targets for inhibition as they determine substrate specificity; however, there are few inhibitors with the ability to precisely target a particular E3 ligase of interest. In this work, we explored a panel of ubiquitin variants (UbVs) that were previously identified to bind Smurf1 or Smurf2. In vitro binding and ubiquitination assays identified a highly specific Smurf2 inhibitor, UbV S2.4, which was able to inhibit ligase activity with high potency in the low nanomolar range. Orthologous cellular assays further demonstrated high specificity of UbV S2.4 toward Smurf2 and no cross-reactivity toward Smurf1. Structural analysis of UbV S2.4 in complex with Smurf2 revealed its mechanism of inhibition was through targeting the E2 binding site. In summary, we investigated several protein-based inhibitors of Smurf1 and Smurf2 and identified a highly specific Smurf2 inhibitor that disrupts the E2-E3 protein interaction interface.

Potent Neutralization of Staphylococcal Enterotoxin B In Vivo by Antibodies that Block Binding to the T-Cell Receptor.

To develop an antibody (Ab) therapeutic against staphylococcal enterotoxin B (SEB), a potential incapacitating bioterrorism agent and a major cause of food poisoning, we developed a "class T" anti-SEB neutralizing Ab (GC132) targeting an epitope on SEB distinct from that of previously developed "class M" Abs. A systematic engineering approach was applied to affinity-mature Ab GC132 to yield an optimized therapeutic candidate (GC132a) with sub-nanomolar binding affinity. Mapping of the binding interface by hydrogen-deuterium exchange coupled to mass spectrometry revealed that the class T epitope on SEB overlapped with the T-cell receptor binding site, whereas other evidence suggested that the class M epitope overlapped with the binding site for the major histocompatibility complex. In the IgG format, GC132a showed ∼50-fold more potent toxin-neutralizing efficacy than the best class M Ab in vitro, and fully protected mice from lethal challenge in a toxic shock post-exposure model. We also engineered bispecific Abs (bsAbs) that bound tetravalently by utilizing two class M binding sites and two class T binding sites. The bsAbs displayed enhanced toxin neutralization efficacy compared with the respective monospecific Ab subunits as well as a mixture of the two, indicating that enhanced efficacy was due to heterotypic tetravalent binding to two non-overlapping epitopes on SEB. Together, these results suggest that class T anti-SEB Ab GC132a is an excellent candidate for clinical development and for bsAb engineering.

Tailored tetravalent antibodies potently and specifically activate Wnt/Frizzled pathways in cells, organoids and mice.

Secreted Wnt proteins regulate development and adult tissue homeostasis by binding and activating cell-surface Frizzled receptors and co-receptors including LRP5/6. The hydrophobicity of Wnt proteins has complicated their purification and limited their use in basic research and as therapeutics. We describe modular tetravalent antibodies that can recruit Frizzled and LRP5/6 in a manner that phenocopies the activities of Wnts both in vitro and in vivo. The modular nature of these synthetic Frizzled and LRP5/6 Agonists, called FLAgs, enables tailored engineering of specificity for one, two or multiple members of the Frizzled family. We show that FLAgs underlie differentiation of pluripotent stem cells, sustain organoid growth, and activate stem cells in vivo. Activation of Wnt signaling circuits with tailored FLAgs will enable precise delineation of functional outcomes directed by distinct receptor combinations and could provide a new class of therapeutics to unlock the promise of regenerative medicine.

A synthetic anti-Frizzled antibody engineered for broadened specificity exhibits enhanced anti-tumor properties.

Secreted Wnt ligands play a major role in the development and progression of many cancers by modulating signaling through cell-surface Frizzled receptors (FZDs). In order to achieve maximal effect on Wnt signaling by targeting the cell surface, we developed a synthetic antibody targeting six of the 10 human FZDs. We first identified an anti-FZD antagonist antibody (F2) with a specificity profile matching that of OMP-18R5, a monoclonal antibody that inhibits growth of many cancers by targeting FZD7, FZD1, FZD2, FZD5 and FZD8. We then used combinatorial antibody engineering by phage display to develop a variant antibody F2.A with specificity broadened to include FZD4. We confirmed that F2.A blocked binding of Wnt ligands, but not binding of Norrin, a ligand that also activates FZD4. Importantly, F2.A proved to be much more efficacious than either OMP-18R5 or F2 in inhibiting the growth of multiple RNF43-mutant pancreatic ductal adenocarcinoma cell lines, including patient-derived cells.

A Highly Diverse and Functional Naïve Ubiquitin Variant Library for Generation of Intracellular Affinity Reagents.

We report the design, construction, and validation of a highly diverse phage-displayed naïve ubiquitin variant (Ubv) library. We first conducted a mutation tolerance scan of 27 residues and confirmed that 24 of these could be substituted by chemically diverse amino acids without compromising the display of Ubvs on phage. Subsequently, we constructed a library containing 6.8×1010 unique members, in which these 24 positions were diversified with a degenerate codon that encodes for 6 aa that are prevalent in protein interaction sites. To ensure the optimal structural stability of the Ubvs, we constructed the library in a two-step process, whereby 12 positions were randomized first, and following the selection for displayed Ubvs, the resulting pool was further diversified at the other 12 positions. The resulting library was validated by conducting binding selections against a panel of 40 diverse protein antigens and was found to be as functional as a highly validated synthetic antibody library, yielding binders against 30 of the antigens. Detailed characterization of an Ubv that bound to the cell-surface receptor human epidermal growth factor receptor 3 revealed tight binding in the single-digit nanomolar range. Moreover, Ubvs that bound to two distinct sites on the intracellular adapter Grb2 could be combined to generate a potent inhibitor that functioned in cells. These results validate ubiquitin as a robust scaffold for the construction of naïve libraries that can be used to generate Ubvs that target signaling networks both outside and inside the cells.

Systematic chemical-genetic and chemical-chemical interaction datasets for prediction of compound synergism.

The network structure of biological systems suggests that effective therapeutic intervention may require combinations of agents that act synergistically. However, a dearth of systematic chemical combination datasets have limited the development of predictive algorithms for chemical synergism. Here, we report two large datasets of linked chemical-genetic and chemical-chemical interactions in the budding yeast Saccharomyces cerevisiae. We screened 5,518 unique compounds against 242 diverse yeast gene deletion strains to generate an extended chemical-genetic matrix (CGM) of 492,126 chemical-gene interaction measurements. This CGM dataset contained 1,434 genotype-specific inhibitors, termed cryptagens. We selected 128 structurally diverse cryptagens and tested all pairwise combinations to generate a benchmark dataset of 8,128 pairwise chemical-chemical interaction tests for synergy prediction, termed the cryptagen matrix (CM). An accompanying database resource called ChemGRID was developed to enable analysis, visualisation and downloads of all data. The CGM and CM datasets will facilitate the benchmarking of computational approaches for synergy prediction, as well as chemical structure-activity relationship models for anti-fungal drug discovery.

Prediction of Synergism from Chemical-Genetic Interactions by Machine Learning.

The structure of genetic interaction networks predicts that, analogous to synthetic lethal interactions between non-essential genes, combinations of compounds with latent activities may exhibit potent synergism. To test this hypothesis, we generated a chemical-genetic matrix of 195 diverse yeast deletion strains treated with 4,915 compounds. This approach uncovered 1,221 genotype-specific inhibitors, which we termed cryptagens. Synergism between 8,128 structurally disparate cryptagen pairs was assessed experimentally and used to benchmark predictive algorithms. A model based on the chemical-genetic matrix and the genetic interaction network failed to accurately predict synergism. However, a combined random forest and Naive Bayesian learner that associated chemical structural features with genotype-specific growth inhibition had strong predictive power. This approach identified previously unknown compound combinations that exhibited species-selective toxicity toward human fungal pathogens. This work demonstrates that machine learning methods trained on unbiased chemical-genetic interaction data may be widely applicable for the discovery of synergistic combinations in different species.

Selection of recombinant anti-SH3 domain antibodies by high-throughput phage display.

Antibodies are indispensable tools in biochemical research and play an expanding role as therapeutics. While hybridoma technology is the dominant method for antibody production, phage display is an emerging technology. Here, we developed and employed a high-throughput pipeline that enables selection of antibodies against hundreds of antigens in parallel. Binding selections using a phage-displayed synthetic antigen-binding fragment (Fab) library against 110 human SH3 domains yielded hundreds of Fabs targeting 58 antigens. Affinity assays demonstrated that representative Fabs bind tightly and specifically to their targets. Furthermore, we developed an efficient affinity maturation strategy adaptable to high-throughput, which increased affinity dramatically but did not compromise specificity. Finally, we tested Fabs in common cell biology applications and confirmed recognition of the full-length antigen in immunoprecipitation, immunoblotting and immunofluorescence assays. In summary, we have established a rapid and robust high-throughput methodology that can be applied to generate highly functional and renewable antibodies targeting protein domains on a proteome-wide scale.

Prediction and experimental characterization of nsSNPs altering human PDZ-binding motifs.

Single nucleotide polymorphisms (SNPs) are a major contributor to genetic and phenotypic variation within populations. Non-synonymous SNPs (nsSNPs) modify the sequence of proteins and can affect their folding or binding properties. Experimental analysis of all nsSNPs is currently unfeasible and therefore computational predictions of the molecular effect of nsSNPs are helpful to guide experimental investigations. While some nsSNPs can be accurately characterized, for instance if they fall into strongly conserved or well annotated regions, the molecular consequences of many others are more challenging to predict. In particular, nsSNPs affecting less structured, and often less conserved regions, are difficult to characterize. Binding sites that mediate protein-protein or other protein interactions are an important class of functional sites on proteins and can be used to help interpret nsSNPs. Binding sites targeted by the PDZ modular peptide recognition domain have recently been characterized. Here we use this data to show that it is possible to computationally identify nsSNPs in PDZ binding motifs that modify or prevent binding to the proteins containing the motifs. We confirm these predictions by experimentally validating a selected subset with ELISA. Our work also highlights the importance of better characterizing linear motifs in proteins as many of these can be affected by genetic variations.

A short DNA aptamer that recognizes TNFα and blocks its activity in vitro.

Tumor necrosis factor-alpha (TNFα) is a pivotal component of the cytokine network linked to inflammatory diseases. Protein-based, TNFα inhibitors have proven to be clinically valuable. Here, we report the identification of short, single-stranded DNA aptamers that bind specifically to human TNFα. One such 25-base long aptamer, termed VR11, was shown to inhibit TNFα signaling as measured using NF-κB luciferase reporter assays. This aptamer bound specifically to TNFα with a dissociation constant of 7.0 ± 2.1 nM as measured by surface plasmon resonance (SPR) and showed no binding to TNFß. Aptamer VR11 was also able to prevent TNFα-induced apoptosis as well as reduce nitric oxide (NO) production in cultured cells for up to 24 h. As well, VR11, which contains a GC rich region, did not raise an immune response when injected intraperitoneally into C57BL/6 mice when compared to a CpG oligodeoxynucleotide (ODN) control, a known TLR9 ligand. These studies suggest that VR11 may represent a simpler, synthetic scaffold than antibodies or protein domains upon which to derive nonimmunogenic oligonucleotide-based inhibitors of TNFα.

Charged and hydrophobic surfaces on the a chain of shiga-like toxin 1 recognize the C-terminal domain of ribosomal stalk proteins.

Shiga-like toxins are ribosome-inactivating proteins (RIP) produced by pathogenic E. coli strains that are responsible for hemorrhagic colitis and hemolytic uremic syndrome. The catalytic A(1) chain of Shiga-like toxin 1 (SLT-1), a representative RIP, first docks onto a conserved peptide SD[D/E]DMGFGLFD located at the C-terminus of all three eukaryotic ribosomal stalk proteins and halts protein synthesis through the depurination of an adenine base in the sarcin-ricin loop of 28S rRNA. Here, we report that the A(1) chain of SLT-1 rapidly binds to and dissociates from the C-terminal peptide with a monomeric dissociation constant of 13 µM. An alanine scan performed on the conserved peptide revealed that the SLT-1 A(1) chain interacts with the anionic tripeptide DDD and the hydrophobic tetrapeptide motif FGLF within its sequence. Based on these 2 peptide motifs, SLT-1 A(1) variants were generated that displayed decreased affinities for the stalk protein C-terminus and also correlated with reduced ribosome-inactivating activities in relation to the wild-type A(1) chain. The toxin-peptide interaction and subsequent toxicity were shown to be mediated by cationic and hydrophobic docking surfaces on the SLT-1 catalytic domain. These docking surfaces are located on the opposite face of the catalytic cleft and suggest that the docking of the A(1) chain to SDDDMGFGLFD may reorient its catalytic domain to face its RNA substrate. More importantly, both the delineated A(1) chain ribosomal docking surfaces and the ribosomal peptide itself represent a target and a scaffold, respectively, for the design of generic inhibitors to block the action of RIPs.