Radisky, Evette S.

Abstract Deep mutational scanning (DMS) has proven effective for mapping protein–protein interactions (PPIs), but it cannot provide complete coverage of the mutation landscape, particularly for multi‐mutant variants. To address this limitation, we trained machine‐learning (ML) models on previously generated DMS data for a stabilized amyloid precursor protein inhibitor (APPI) binding to either of two serine proteases, mesotrypsin and kallikrein‐6 (KLK6), which are implicated in various human disorders. We combined the models to accurately predict the binding selectivity of APPI variants, including double‐mutant variants, for the two serine proteases. We achieved a Pearson correlation of 0.937 between predicted log 2 selectivity enrichment ratios and DMS‐derived values. We further validated the predictions of our combined model by yeast‐surface‐display measurements and inhibition assays of purified APPI variants and revealed epistatic interactions that shape protease selectivity. Guided by binding selectivity predictions, we identified highly selective APPI variants, including the most selective mesotrypsin inhibitor reported to date. Together, these findings support the use of DMS and ML as a framework for predicting PPI selectivity and prioritizing selective therapeutic protein variants.

Abstract Matrix metalloproteinases (MMPs) are key regulators of tumor progression, invasion, and metastasis; however, clinical efforts to inhibit MMPs have largely failed due to a lack of selectivity and unintended suppression of protective family members. In triple-negative breast cancer (TNBC), MMP-9 (gelatinase B) is strongly associated with tumor aggressiveness, extracellular matrix remodeling, and metastatic dissemination. In contrast, MMP-8 has been shown to exert anti-tumorigenic effects in breast cancer by limiting inflammation and suppressing metastasis. This functional divergence underscores the importance of selectively targeting MMP-9 while preserving MMP-8 activity as a rational therapeutic strategy in TNBC. To address this challenge, we engineered variants of the endogenous MMP inhibitor tissue inhibitor of metalloproteinases-1 (TIMP-1) with enhanced selectivity for MMP-9 over MMP-8. A yeast surface display library of human TIMP-1 was constructed and screened using fluorescence-activated cell sorting through iterative rounds of positive selection for recombinant MMP-9, followed by competitive counter-selection against MMP-8 to eliminate off-target binders. This strategy enabled enrichment of TIMP-1 variants with improved specificity for MMP-9 compared with wild-type TIMP-1. Several TIMP-1 variants exhibited enhanced binding and inhibitory selectivity toward MMP-9, with TIMP-1-C6.2 emerging as the most promising candidate. TIMP-1-C6.2 preferentially inhibited MMP-9 with a Ki of 0.3 ± 0.04 nM, while demonstrating substantially weaker inhibition of MMP-8 (Ki = 6.8 ± 0.27 nM), corresponding to an approximately 23-fold selectivity for MMP-9. Importantly, functional evaluation in Matrigel invasion assays using MDA-MB-231 TNBC cells revealed that TIMP-1-C6.2 significantly enhanced inhibition of tumor cell invasion compared with wild-type TIMP-1 in a dose-dependent manner (p < 0.0001). This improved anti-invasive activity reflects refined target selectivity rather than a generalized increase in inhibitory potency. Collectively, these findings demonstrate that selective engineering of TIMP-based inhibitors can overcome a central limitation of MMP-targeted therapies by suppressing pro-tumorigenic MMP-9 while sparing anti-tumorigenic MMP-8. This work provides a strong foundation for the development of selective MMP-9 inhibitors with improved safety and therapeutic potential for TNBC and other MMP-driven cancers. Citation Format: Alireza Shoari, Alexandra Hockla, Mathew A. Coban, Ella E. Aitchison, Alexandra M. Dimesa, Evette S. Radisky. Protein engineering of selective MMP-9 inhibitors reveals a therapeutic strategy for triple-negative breast cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 2 (Late-Breaking, Clinical Trial, and Invited Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(8_Suppl):Abstract nr LB233.

Matrix metalloproteinases (MMPs) are secreted and cell membrane‐associated enzymes that remodel the extracellular matrix (ECM) and cleave extracellular proteins to promote tumor invasion, angiogenesis, immune evasion, and many other aspects of cancer progression. Among this family, the gelatinases MMP‐2 and MMP‐9, specialized for cleaving collagen IV, are tightly linked to metastatic progression in breast cancer and adverse clinical outcomes. Here, we summarize gelatinase biochemistry and regulation, including zymogen activation, TIMP control, cell‐surface trafficking and tethering, and receptor‐mediated clearance, and explain how crosstalk between ECM and proteases amplifies invasion and metastatic seeding. We then review therapeutic strategies in two parts: direct inhibition and indirect pathway modulation. Direct approaches emphasize selective small molecules (thiirane mechanism‐based inhibitors, allosteric blockers of pro‐MMP‐9 activation) and protein biologics (monoclonal antibodies, nanobodies, engineered TIMPs). Indirect strategies target upstream signals that drive MMP‐2/‐9, including MAPK/AP‐1, PI3K/Akt/mTOR, NF‐κB, EGFR/JAK/STAT, and nuclear receptor and Nrf2/HO‐1 programs, with agents that curb invasion, angiogenesis, and metastasis in breast cancer models. Together, these advances define a maturing toolkit for precise gelatinase control and support prospective evaluation in rational combinations to restrain metastatic progression in high‐risk breast cancer.

Selective inhibitors are essential for targeted therapeutics and for probing enzyme functions in various biological systems. The two main challenges in identifying such protein‐based inhibitors lie in the extensive experimental effort required, including the generation of large libraries, and in tailoring the selectivity of inhibitors to enzymes with homologous structures. To address these challenges, machine learning (ML) is being used to improve protein design by training on targeted libraries and identifying key interface mutations that enhance affinity and specificity. However, such ML‐based methods are limited by inaccurate energy calculations and difficulties in predicting the structural impacts of multiple mutations. Here, we present an ML‐based method that leverages HTS data to streamline the design of selective protease inhibitors. To demonstrate its utility, we applied our new method to find inhibitors of matrix metalloproteinases (MMPs), a family of homologous proteases involved in both physiological and pathological processes. By training ML models on binding data for three MMPs (MMP‐1, MMP‐3, and MMP‐9), we successfully designed a novel N‐TIMP2 variant with a differential specificity profile, namely, high affinity for MMP‐9, moderate affinity for MMP‐3, and low affinity for MMP‐1. Our experimental validation showed that this novel variant exhibited a significant specificity shift and enhanced selectivity compared with wild‐type N‐TIMP2. Through molecular modeling and energy minimization, we obtained structural insights into the variant's enhanced selectivity. Our findings highlight the power of ML‐based methods to reduce experimental workloads, facilitate the rational design of selective inhibitors, and advance the understanding of specific inhibitor–enzyme interactions in homologous enzyme systems.

Selective inhibition of the more than 100 S1 family serine proteases is a long-standing challenge due to their active site similarity. Mesotrypsin, implicated in cancer progression, exemplifies these difficulties; no current inhibitors achieve selectivity over other human trypsins. We found an unexpected autoinhibited conformation of mesotrypsin via x-ray crystallography, revealing a cryptic pocket adjacent to the active site. Using high-throughput virtual screening targeting this cryptic pocket, we identified a conformationally selective small-molecule inhibitor that stabilizes the inactive state of mesotrypsin. This inhibitor demonstrates selectivity for mesotrypsin over other trypsins. Our findings challenge the accepted view of digestive trypsins as constitutively active enzymes lacking potential for allosteric regulation. Furthermore, analyses of other structures suggest that dynamic sampling of closed states with analogous allosteric cryptic pockets appears widespread among S1 serine proteases. These observations point to a potentially generalizable strategy to achieve selective inhibition, offering broad implications for drug development targeting serine proteases in cancer and other diseases.

Abstract Tissue inhibitors of metalloproteinases (TIMPs) are endogenous inhibitors that regulate the function and activity of matrix metalloproteinases (MMPs), enzymes that are known for their involvement in cancer progression. MMP-9, also known as gelatinase B, plays a crucial role in tumorigenesis of several cancers including triple negative breast cancer (TNBC); tumor cell-produced MMP-9 promotes malignant progression and metastasis of basal-like TNBC. Here, we have used full-length TIMP-1 as a scaffold for engineering inhibitors that bind with higher selectivity to MMP-9, for potential therapeutic applications. We previously developed a yeast surface display library of human TIMP-1, with diversity incorporated at 17 residues within five loops; mutation burden was tuned to 3-5 mutations per clone. Here, we have screened the TIMP-1 library for affinity toward recombinant MMP-9 by using fluorescence-activated cell sorting (FACS). Yeast cells were incubated with an incrementally decreasing concentration of biotinylated MMP-9 (400 nM to 25 nM) in successive rounds of sorting. Samples were analyzed using flow cytometry and highest affinity populations were selected using FACS. Among the TIMP-1 variants with highest enrichment, we identified the TIMP-1-C15 variant that showed 12-fold enhanced binding with improved selectivity toward MMP-9 compared with wild type TIMP-1 (WT TIMP-1). We next expressed and purified soluble recombinant TIMP-1-C15 and conducted inhibition studies using a fluorogenic substrate to evaluate inhibitory potency towards MMP-9 activity. These experiments showed that TIMP-1-C15 inhibits MMP-9 activity with an inhibition constant (Ki) of 0.61 nM, more than 5-fold stronger than WT TIMP-1; these studies also corroborated the improved selectivity towards MMP-9 over other MMPs. Additionally, we further evaluated the functional impact of the TIMP-1-C15 variant on cancer cell invasiveness using TNBC cell lines in cellular invasion assays. The TIMP-1-C15 variant markedly reduced the invasive capabilities of TNBC cells to a greater extent than WT TIMP-1. Currently, we are working to isolate additional TIMP-1 variants from our library that exhibit enhanced selectivity for MMP-9, with MMP-8 included in the sorting process as a competitive target to ensure specificity. Selective inhibition of MMP-9 over MMP-8 is particularly important in breast cancer therapy, as MMP-8 has been shown to have anti-tumorigenic properties, and its inhibition could counteract therapeutic efforts aimed at reducing cancer invasiveness. Our findings suggest that engineered TIMP-1 variants could serve as potent inhibitors of MMP-9, offering a novel approach to reducing the invasiveness of breast cancer cells and potentially limiting the progression of metastatic disease. Citation Format: Alireza Shoari, Alexandra Hockla, Mathew A. Coban, Alexandra M. Dimesa, Derek C. Radisky, Evette S. Radisky. Engineered TIMP-1 for selective MMP-9 inhibition: A novel strategy to suppress breast cancer invasiveness [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 5115.

Matrix metalloproteinases (MMPs) are significant drivers of many diseases, including cancer, and are established targets for drug development. Tissue inhibitors of metalloproteinases (TIMPs) are endogenous MMP inhibitors and are being pursued for the development of anti-MMP therapeutics. TIMPs possess many attractive properties for drug candidates, such as complete MMP inhibition, low toxicity, low immunogenicity, and high tissue permeability. However, a major challenge with TIMPs is their rapid clearance from the bloodstream due to their small size. This study explores a method for extending the plasma half-life of the N-terminal domain of TIMP2 (N-TIMP2) by appending it with a long, intrinsically unfolded tail containing Pro, Ala, and Thr (PATylation). We designed and produced two PATylated N-TIMP2 constructs with tail lengths of 100 and 200 amino acids (N-TIMP2-PAT100 and N-TIMP2-PAT200). Both constructs demonstrated higher apparent molecular weights and retained high inhibitory activity against MMP-9. N-TIMP2-PAT200 significantly increased plasma half-life in mice compared to the non-PATylated variant, enhancing its therapeutic potential. PATylation offers distinct advantages for half-life extension, such as fully genetic encoding, monodispersion, and biodegradability. It can be easily applied to N-TIMP2 variants engineered for high affinity and selectivity toward individual MMPs, creating promising candidates for drug development against MMP-related diseases.

Ovarian cancer remains a devastating disease with 5 year survival rate in advanced stage of only 27%. Advanced stage tumors metastasize through continuous shedding of tumor cells into the peritoneal fluid, aided by the resistance of these cells to anoikis, a form of apoptosis normally triggered by loss of cell‐matrix interactions. We have previously shown that the serine protease inhibitor SPINK1 activates anoikis resistance in ovarian cancer cells via its protease inhibitory function; however, the specific relevant protease target(s) remain unknown. To find these targets, we have synthesized an activity‐based protein probe with a diphenylphosphonate warhead that covalently modifies serine proteases, with greatest specificity towards trypsin‐like proteases. We have used this probe to identify ovarian cancer cell secreted and membrane‐bound serine proteases possessing differential activity in the presence versus absence of SPINK1, using a proteomics approach. We are evaluating candidate proteases for involvement in anoikis resistance in ovarian cancer models. Our studies implicate a serine protease network in regulation of anoikis in ovarian cancer cells. By identifying critical proteases that control anchorage‐independent viability in ovarian cancer metastasis, we aim to elucidate important downstream pathways and ultimately develop therapeutic approaches that can intervene in mechanisms of anoikis resistance. Support or Funding Information Mayo Clinic Graduate School, Rochester MN

Enzyme catalytic power and specificity derive from dynamic interactions between enzymes and their substrates. While considerable effort has been devoted to understanding how enzyme conformational dynamics can affect catalysis, the importance of conformational dynamics within protein substrates has been overlooked. Proteases are ubiquitous enzymes that catalyze the hydrolysis of peptide bonds within protein substrates; they have served as key model enzymes for studying the molecular basis for catalysis and specificity. Recent work from our research group, focusing on trypsins and chymotrypsins as model proteolytic enzymes, reveals that differing conformational dynamics in otherwise similar protein substrates can profoundly impact proteolysis. In this talk, I will present some of our recent findings, integrating evidence from crystal structures, molecular dynamics simulations, and enzyme kinetics studies, which implicate flexibility and conformational motions of protease substrates as critical determinants of proteolysis. Our studies suggest that specific substrate conformational dynamics have evolved to regulate biological proteolytic events. More generally, our work suggests that we must look beyond enzymes, and examine substrates also, to fully appreciate the importance of protein dynamics in enzyme catalysis. Support or Funding Information Supported by NIH grant R01CA154387.

Proteases are ubiquitous enzymes that catalyze the hydrolysis of peptide bonds within protein substrates; they have served as key model enzymes for studying the molecular basis for catalytic power and specificity. Protease substrate specificity is most often defined in terms of linear sequence motifs that flank the cleavage site; however, the natural substrates of proteases are proteins with 3‐dimensional shapes and complex conformational dynamics that are not well represented by 1‐dimensional sequence alone. These structural and dynamical properties can impact recognition and binding of substrates by proteases, as well as the efficiency of catalysis itself. In this study, we explore the importance of substrate structure and dynamics for proteolysis using as our model the cleavage of the Kunitz‐BPTI family of canonical serine protease inhibitors by mesotrypsin. Bovine pancreatic trypsin inhibitor (BPTI), an archetypal serine protease inhibitor of the Kunitz family, has a high affinity substrate‐like interaction with trypsin, yet is cleaved many orders of magnitude slower than other peptide substrates. Mesotrypsin, a trypsin variant, has been shown to hydrolyze Kunitz family inhibitors at accelerated rates; this is especially true of human Kunitz domain inhibitors. Amyloid precursor protein inhibitor (APPI) and amyloid precursor like protein‐2 Kunitz domain (APLP2‐KD), two human Kunitz domain family members, are hydrolyzed by mesotrypsin several hundred times faster than BPTI. Here, we present a new crystal structure of a cleavage intermediate of APLP2‐KD bound to mesotrypsin, refined to 1.4 Å resolution. This newly solved structure reveals a dramatic substrate conformational change that we hypothesize is required during Kunitz domain cleavage. Using this structure along with published APPI and BPTI complexes, we have modeled acyl‐enzyme intermediates of mesotrypsin and we have carried out molecular dynamic simulations to explore the conformational transformation that allows the progression of the hydrolysis reaction. We identify dynamic features of the inhibitor scaffolds, located far from the enzyme binding site, that correlate with observed inhibitor cleavage rates and plausibly explain their differential propensity to undergo both conformational transformation and cleavage. Our findings implicate global scaffold flexibility as a determinant of proteolytic rate, as rigid substrates possessing stabilizing features that render them highly resistant to conformational change are proteolyzed more slowly than more flexible substrates of similar structure. Support or Funding Information NIH R01 CA154387 Mayo Graduate school