Fragment-Growing Strategy Targets Elusive S1′ Pocket to Unlock Selective MMP-9 Cancer Inhibitors

Matrix metalloproteinase-9, or MMP-9, has long tempted cancer researchers as a drug target and long frustrated them in equal measure. The enzyme sits at the heart of extracellular matrix remodelling, the process by which tumours break down surrounding tissue, carve out routes for invasion, seed metastases and recruit the new blood vessels they need to grow. Decades of biological evidence have tied elevated MMP-9 activity to poor prognosis in cancers ranging from colon to breast, making the enzyme an obvious candidate for pharmacological blockade. Yet despite this compelling rationale, no potent and selective MMP-9 inhibitor has survived the journey to the clinic. A new computational study from researchers at the National Institute of Pharmaceutical Education and Research in Hyderabad, India, published in Molecular Diversity, now describes a fragment-based design strategy that aims squarely at the structural feature that has historically doomed previous attempts: the remarkable similarity of MMP enzymes to one another.

The core problem is homology. The catalytic domain of MMP-9, where zinc-mediated peptide cleavage takes place, is architecturally almost indistinguishable from the catalytic domains of its more than twenty family members. Inhibitors designed to wedge into the catalytic cleft of one isoform tend to bind the others with nearly equal enthusiasm, and it is this promiscuity that is widely blamed for the musculoskeletal side effects that halted broad-spectrum matrix metalloproteinase inhibitors such as batimastat and marimastat in late-stage clinical trials during the 1990s. Blocking a family of enzymes that also performs essential physiological tissue remodelling, the reasoning goes, produces collateral damage that patients cannot tolerate. The field’s response has been a search for isoform-specific features, and one such feature has emerged as the most promising: the S1′ specificity pocket, the subsite that accommodates the side chain of a substrate’s substrate-position-one-prime residue.

Unlike the conserved catalytic machinery, the S1′ pocket varies substantially across MMP isoforms in size, shape and residue composition. In MMP-9, crystal structures reveal a pocket whose geometry differs from that of related gelatinases and other family members, offering a toehold for selectivity. Structural work on MMP-9 in complex with hydroxamate inhibitors, and later studies highlighting the contribution of the flexible Arg424 side chain, established that ligands filling the S1′ pocket can engage residues that simply do not exist in the same conformation elsewhere in the family. What had not been done, according to the Indian team, was a systematic fragment-based drug design campaign explicitly targeted at this pocket. Fragment-based approaches start from small, low-molecular-weight molecules that bind weakly but efficiently, and grow or link them into larger, higher-affinity ligands, a philosophy that has produced several approved drugs and that lends itself naturally to exploiting small, subtle binding cavities.

The study’s workflow began with fragment-based virtual screening of commercially available fragment libraries from suppliers including Enamine, ChemBridge, FCH Group and Otava Chemicals. The researchers prepared the MMP-9 receptor structure carefully, following established protein and ligand preparation protocols known to influence the enrichment quality of virtual screens, and docked the fragment collection into the S1′ pocket using high-throughput virtual screening followed by progressively more rigorous docking tiers. This triage surfaced low-molecular-weight fragments with moderate predicted binding affinity toward the pocket. Among them, one promising fragment hit was selected for optimisation using a receptor cavity-guided fragment-growing strategy, in which chemical groups are appended to the parent fragment so that they extend into neighbouring space, in this case reaching toward residues lining the adjacent S2′ pocket.

Growing a fragment is only as good as the poses it produces, so the team turned to induced-fit docking, a method that permits flexibility in both the ligand and the binding-site side chains. This step identified several optimised molecules whose predicted orientations within the S1′ pocket were favourable and whose binding affinities exceeded those of the parent fragment. Crucially, docking scores alone are notoriously optimistic, so the investigators subjected the protein-ligand complexes to molecular dynamics simulations of 200 nanoseconds each, tracking root mean square deviation, root mean square fluctuation, solvent accessible surface area, radius of gyration and secondary structure stability throughout. Four of the optimised molecules maintained stable binding interactions over the full simulation, their key contacts with the S1′ pocket persisting as the protein breathed and flexed around them.

Selectivity, the whole point of the exercise, was then interrogated directly. The team redocked the four stable hits against a panel of MMP isoforms and compared binding free energies calculated with the molecular mechanics/generalized born surface area, or MM-GBSA, method. The analysis demonstrated a preferential affinity of the hit molecules for MMP-9, with favourable energetic profiles and orientations driven specifically by their occupancy of the non-conserved S1′ pocket. While computational selectivity is not a substitute for measured enzyme inhibition, the concordance of binding orientation, free energy and pocket engagement across independent calculations suggests a genuine structural basis for discrimination between MMP-9 and its close relatives rather than a numerical artefact of any single scoring function.

Drug-likeness was assessed in parallel. In silico ADME profiling of all four stable hits indicated favourable pharmacokinetic properties, an encouraging sign for molecules descended from fragments, which typically begin life with excellent solubility and ligand efficiency. The team also applied density functional theory calculations, mapping electrostatic potential and the energies of the highest occupied and lowest unoccupied molecular orbitals to understand the electronic distribution and chemical reactivity of each candidate. These quantum chemical descriptors provide medicinal chemists with a preview of how the molecules might behave in biological systems, flagging reactive sites and electronic features that correlate with metabolic stability and target engagement. Collectively, the computational evidence supports the four hits as credible starting points for synthetic optimisation.

The significance of the work lies less in any single molecule than in the framework it establishes. By demonstrating that an integrated pipeline of fragment screening, cavity-guided growing, induced-fit docking, long molecular dynamics simulation, cross-isoform MM-GBSA comparison, ADME prediction and quantum chemical analysis can converge on molecules that discriminate MMP-9 from its family members, the study offers a reproducible template for one of medicinal chemistry’s most stubborn selectivity problems. It also reinforces a lesson from the broader fragment-based literature: fragments, precisely because they are small and efficient, are exquisitely sensitive to the fine geometric differences between protein pockets, making them ideal probes for pockets as subtly differentiated as the S1′ site. The authors acknowledge that the findings are computational and that supporting data are available on request; experimental validation of binding and enzyme inhibition will be the necessary next step.

If the hits translate, the implications extend beyond oncology. MMP-9 has been implicated in inflammatory, cardiovascular and neurodegenerative conditions, and a selective inhibitor scaffold that spares other matrix metalloproteinases could reopen a therapeutic avenue that clinical failures closed a generation ago. For cancer patients, the prospect is an anti-metastatic agent that disarms the tumour’s remodelling machinery without destabilising the normal tissue turnover on which health depends. The Hyderabad team’s S1′-targeted fragment strategy provides the molecular blueprint for pursuing that goal rationally, transforming a protein family once written off as undruggable-without-toxicity into a target whose selectivity problem now has a concrete, structurally grounded path forward.

Part of what makes the S1′ pocket such an attractive lever for selectivity is its functional role in catalysis. The MMP active site centres on a catalytic zinc ion coordinated by three histidines, and substrate cleavage depends on how the residue immediately N-terminal to the scissile bond slots into the S1′ subsite. Because different isoforms prefer different P1′ side chains, the depth and hydrophobic character of this pocket directly shape each enzyme’s substrate repertoire, and MMP-9’s preference for bulky, hydrophobic P1′ residues is reflected in a pocket architecture that diverges measurably from even its closest gelatinase relative, MMP-2.

The fragment-based philosophy the study adopts has a well-documented pedigree. Since the landmark SAR-by-NMR work of the late 1990s, fragment approaches have delivered approved drugs across multiple target classes, largely because small starting points leave ample chemical space for optimisation while maintaining favourable ligand efficiency. Commercial fragment libraries from vendors such as Enamine and ChemBridge typically enforce strict rules on molecular weight, hydrogen-bond donors and acceptors, and calculated lipophilicity, ensuring that hits remain tractable for the growing stage that follows.

It is also worth noting that MMP-9 offers domains beyond the catalytic site as selectivity handles. Its C-terminal hemopexin domain, organised into four propeller blades, mediates substrate interactions distinct from zinc-dependent cleavage, and earlier work showed that antibodies or ligands engaging this domain can block gelatin degradation without touching the catalytic machinery. The present study’s focus on the S1′ pocket complements such allosteric and exosite strategies, giving medicinal chemists multiple, mechanistically different entry points toward the same long-elusive goal of an isoform-selective MMP-9 inhibitor.

Subject of Research: Fragment-based computational design of selective MMP-9 inhibitors targeting the S1′ pocket

Article Title: Addressing the selectivity challenge in matrix metalloproteinases: fragment-based design of MMP-9 inhibitors targeting the S1′ pocket

Article References: Kumar, A., Md Sameer, Shaikh, A. S., Chauhan, Y., & Kaki, V. R. (2026). Addressing the selectivity challenge in matrix metalloproteinases: fragment-based design of MMP-9 inhibitors targeting the S1′ pocket. Molecular Diversity. https://doi.org/10.1007/s11030-026-11723-y

Image Credits: AI Generated

DOI: 10.1007/s11030-026-11723-y

Keywords: MMP-9, fragment-based drug design, S1' pocket, selectivity, induced-fit docking, molecular dynamics, MM-GBSA, anticancer drug discovery, ADME, density functional theory, extracellular matrix, matrix metalloproteinase inhibitors