Allosteric inhibitors of membrane type 1 matrix metalloproteinase MT1-MMP (MMP14)

Allosteric Inhibitors of Membrane Type I Matrix Metalloproteinase (MT1-MMP/MMP14)

Metastatic cancer

       Metastatic cancer is a cancer that has spread from the part of the body where the primary tumor was located to other parts of the body. This process of cancer cell spreading is referred to as metastasis. Metastasis is what makes cancer such an efficient killer. To eradicate possible metastatic tumors in different part of the body cancer patients have to through debilitating course of radio- and/or chemotherapy after the primary tumor was surgically removed. Contemporary chemotherapeutic agents are essentially poisons that are more harmful to cancer cells because of their high rate of metabolism and cellular division. However, many normal body functions require rapid cellular division of healthy cells. Immune system failure, bone marrow death, hair loss and many other side effects of chemotherapy are caused by sensitivity of specific healthy cell types toward chemotherapeutic agents.

       Cancer cells, unlike healthy cells, can quickly adapt to aggressive environment through an evolution-like process during the clonal expansion. A chemotherapeutic agent applies a selection pressure on a population of cancer cells, which eventually leads to appearance of cancer cells that are no longer sensitive to this particular drug. The development of the resistance to chemotherapeutic drugs leads to complete failure of a chemotherapy campaign because a patient cannot tolerate ever increasing doses of a drug. 

To summarize, we desperately need chemotherapeutic agents able to control cancer metastasis both lacking damage to healthy cells and selection pressure on cancer cells. 
Role of MT1-MMP/MMP14 in metastasis

       In order for cancer cells to start spreading through the body, they have to be able to activate a complex molecular machinery necessary for migration through the surrounding extracellular matrix (ECM) and invasion of blood and lymphatic vessels. The ECM provides all tissues and organs with form, mechanical strength and physical properties. The ECM remodeling enzymes are essential for cancer cells to break through dense protein mesh of the ECM. In particular, matrix metalloproteinases (MMP) are considered as key players in tumor progression because of their ability to remodel the ECM and cleave/activate membrane-bound and matrix molecules. Because of their intimate involvement in the metastasis, MMPs have been attracting significant attention as drug targets. However, broad inhibitors of MMP catalytic activity have been unsuccessful in clinic. It is now known that some MMPs can have both stimulatory and inhibitory effects on tumor progression. Attempts to develop specific inhibitors of the active site of an MMP fail because of high degree of structural homology among different MMPs.


     Membrane type 1 matrix metalloproteinase (MT1-MMP/MMP14), a cell surface bound MMP, plays a significant role in cancer cell migration and tumor progression. MT1-MMP cleaves a wide variety of extracellular matrix components, such as collagens I, II, and III, laminins 1 and 5, aggrecan core protein, fibrin and fibronectin, vitronectin, and lumican. Thus, MT1-MMP is essential for remodeling of fibrous structural components of ECM.

MT1-MMP as a drug target

                As mentioned earlier, direct inhibition of MMPs catalytic activity with small molecules is counter-productive. The multi domain organization of MT1-MMP offers alternative strategy of specific interference with its function using allosteric small molecule antagonists. MT1-MMP is composed of several structurally distinct domains including the catalytic domain and the hemopexin domain. The hemopexin (HPX14) domain is not required for the catalytic activity of MT1-MMP but is necessary for MT1-MMP to be able to cleave fibrous insoluble substrates such as collagens I, II, III, and some other types of substrates.

                HPX14 performs its function by forming a homodimer, thus causing the catalytic domain to dimerize too. This dimerization is believed to be required for proper positioning of the MT1-MMP catalytic domain on a relatively static fibrous protein substrate. Additionally, the HPX14 domain itself performs functions of a helicase perturbing the secondary structure of collagen and unwinding collagen triple helix (1). This triple helicase activity is an independent event and separable from the proteolytic activity. Importantly, helicase activity is absolutely required for efficient native collagen cleavage. Thus, one can hypothesize that interference with the ability of HPX14 to dimerize and/or bind native collagen will specifically shutdown the ECM remodeling function of MT1-MMP.

Feasibility of designing allosteric inhibitors of MT1-MMP

         Rationally designing small molecule modulators of protein-protein interactions is a challenge. Aside of ability to correctly take care of protein flexibility, there are often insurmountable thermodynamic constraints limiting application of small molecules as antagonists of protein-protein interactions. The main thermodynamic bottleneck is to balance excess of free energy released when large areas of a protein-protein interface are destroyed. This excess in free energy must be compensated in order for stable binding of a small molecule to occur. This energy compensation must come either from alternative protein-protein interface formation or from an internal conformational rearrangement of a target protein.

          Recently, the atomic resolution X-ray structure of HPX14 domain in two alternative homodimeric forms has become available (PDB 3C7X) (2) (Figure 1A). Structurally, HPX14 domain belongs to the topological family of b-propeller domains, which encompasses proteins with 4-8 blades of b-strands (4 blades in this case). The vast interface of biologically relevant symmetric homodimer is problematic for targeting by a small molecule simply because of its area. However, targeting the opposite side of the HPX14 monomer is feasible for two reasons: first, it is an opening into a pocket-like channel; second, b-propeller domains are highly flexible, and as such, can be stabilized by binding of a ligand (Figure 1B).

Figure 1.  A. Hpx14 homodimer (PDB 3C7X). Subunits are colored differently to highlight protein-protein interface. B.  Definition of the docking site (red patch). Note channel-like pocket. 
Successful application of Q-MOL virtual ligand screening to HPX14 hemopexin domain of MT1-MMP

                The Q-MOL protein-ligand docking methodology in the core of Q-MOL virtual ligand screening (VLS) is perfectly suited for targeting highly flexible allosteric sites of proteins of known structure. Q-MOL protein-ligand docking achieves its prediction fidelity by exploiting protein flexibility. The more flexible the protein at a particular site, the higher the success rate of Q-MOL VLS. During protein-ligand docking simulation Q-MOL evaluates all most probable conformations of a target protein. These protein conformations are taken from protein folding energy landscape (also known as folding funnel) (3, 4) of a target protein using singular crystal structure of a target protein as a reference point. The cumulative Gaussian nature of Q-MOL docking curves (Figure 2A) reflects Gaussian distribution of protein conformations within protein folding funnel. Thus, each dot on a docking curve corresponds to not only a predicted hit but also to a particular protein conformation this ligand fits the best. Importantly, the conformational diversity of a target protein leads to the structural (chemical) diversity of the predicted hits.

Figure 2.  A. The output of Q-MOL VLS.  The docking curve represents 125 best predicted hits out of over 200,000 NCI DTP compounds.  Each dot corresponds to an individual compound, rank of an in vivo active ligand is indicated by a red dot.  E, relative binding energy.  Inset:  distribution of ligands molecular weight (MW) versus rank.  B.  Ligand NSC405020 was docked into Hpx14 subunit (red) using full atom fully flexible protein-ligand docking protocol as implemented in Q-MOL program.  Note that binding of the ligand leads to clashes at the protein-protein interface (native Hpx14 subunit is shown in yellow).             

           The application of Q-MOL VLS to the complete NCI DTP ligand database (over 200,000 entries) resulted in identification of 125 top predicted hits (Figure 2A) out of which only 39 were selected after visual inspection. From 39 selected hits only 19 were available for ordering from NCI DTP ( About half of these ligands inhibited enzymatic activity of the MT1-MMP catalytic domain. This cross-reactivity with the catalytic domain is an indicator of successful targeting strategy because both catalytic and HPX14 domains bind collagen.  

        One molecule (NSC405020) performed adequately in all enzyme-based, cell-based and animal model experiments. This ligand binds and stabilizes particular HPX14 conformation which is not able to dimerize (Figure 2B). The ligand interfered with processing of native collagen by MT1-MMP, and thus, inhibited tumor growth by cocooning tumor in ECM, and blocking tumor cells from access to nutrients and oxygen. 

Importantly, this small molecule is not toxic to both healthy and cancer cells. The cancer cells are affected only when they form sizable tumor mass. These results represent potential implementation of long sought-after paradigm of non-toxic antimetastatic therapy.

What is next?

           Q-MOL protein-ligand docking technology exploits protein flexibility to find specific and potent small molecule ligands. These ligands target specifically different protein conformations. The chemical structure of a ligand is what drives selection of a protein conformation at a particular docking site. The discovery of ligands with desired biological properties for a particular docking site validates this docking site (proof of targeting strategy).

          Because ligand chemical space modulates protein conformational changes, the chemical diversity of small molecule binders is limited only by protein flexibility at a validated docking site. Thus, flexible enough protein is able to specifically interact at a validated docking site with a significant number of chemically diverse small molecules. In this context, a validated docking site can be used as a structure activity relationship (SAR) optimization “device” to drive drug discovery efforts. 

          Immediate application of this approach is the docking of a library of already FDA-approved clinical drugs into the biologically validated docking site of HPX14 domain (Figure 3). The most successful outcome of this strategy is the discovery of new application (antimetastatic in this case) for a drug which is already in clinics (drug repositioning). In the worst case, new scaffolds are excellent leads for further development because they are sampled from already “humanized” chemical space (minimum interference with human biology).

Figure 3. Docking of a library of FDA-approved drugs into the Hpx14 domain.  Each dot corresponds to an individual drug.

Q-MOL L.L.C. is currently accepting collaboration/partnership proposals toward the fast-track development of new antimetastatic drugs.


1.            Tam EM, Moore TR, Butler GS, & Overall CM (2004) Characterization of the distinct collagen binding, helicase and cleavage mechanisms of matrix metalloproteinase 2 and 14 (gelatinase A and MT1-MMP): the differential roles of the MMP hemopexin c domains and the MMP-2 fibronectin type II modules in collagen triple helicase activities. (Translated from eng) J Biol Chem 279(41):43336-43344 (in eng).
2.            Tochowicz A, et al. (2011) The dimer interface of the membrane type 1 matrix metalloproteinase hemopexin domain: crystal structure and biological functions. (Translated from eng) J Biol Chem 286(9):7587-7600 (in eng).
3.            Onuchic JN, Luthey-Schulten Z, & Wolynes PG (1997) Theory of protein folding: the energy landscape perspective. (Translated from eng) Annu Rev Phys Chem 48:545-600 (in eng).
4.            Wolynes PG (1997) Folding funnels and energy landscapes of larger proteins within the capillarity approximation. (Translated from eng) Proc Natl Acad Sci U S A 94(12):6170-6175 (in eng).