Tissue Inhibitors of Metalloproteinases are proteolytic targets of Matrix Metalloproteinase 9

Abstract

Extracellular proteolysis and turnover are core processes of tissue homeostasis. The predominant matrix-degrading enzymes are members of the Matrix Metalloproteinase (MMP) family. MMPs extensively degrade core matrix components in addition to processing a range of other factors in the extracellular, plasma membrane, and intracellular compartments. The proteolytic activity of MMPs is modulated by the Tissue Inhibitors of Metalloproteinases (TIMPs), a family of four multi-functional matrisome proteins with extensively characterized MMP inhibitory functions. Thus, a well-regulated balance between MMP activity and TIMP levels has been described as critical for healthy tissue homeostasis, and this balance can be chronically disturbed in pathological processes. The relationship between MMPs and TIMPs is complex and lacks the constraints of a typical enzyme-inhibitor relationship due to secondary interactions between various MMPs (specifically gelatinases) and TIMP family members. We illustrate a new complexity in this system by describing how MMP9 can cleave members of the TIMP family when in molar excess. Proteolytic processing of TIMPs can generate functionally altered peptides with potentially novel attributes. We demonstrate here that all TIMPs are cleaved at their C-terminal tails by a molar excess of MMP9. This processing removes the N-glycosylation site for TIMP3 and prevents the TIMP2 interaction with latent proMMP2, a prerequisite for cell surface MMP14-mediated activation of proMMP2. TIMP2/4 are further cleaved producing ~14kDa N-terminal proteins linked to a smaller C-terminal domain through residual disulfide bridges. These cleaved TIMP2/4 complexes show perturbed MMP inhibitory activity, illustrating that MMP9 may bear a particularly prominent influence upon the TIMP:MMP balance in tissues.

Introduction

The activity of Matrix Metalloproteinases (MMPs), key mediators of extracellular proteolysis and matrix turnover, has wide-ranging implications within the tissue microenvironment. There are at least 23 different human MMPs that display both redundancy and exclusivity in their proteolytic, and non-proteolytic activities [1–3]. MMPs have highly conserved sequences within their zinc-binding active sites. As a result, the specificity for their proteolytic targets appears to be dictated by peripheral regions/domains (known as ectodomains) within the active protein [3].

The activity of MMPs can be regulated through various avenues, and their expression/activation in healthy adult tissues is largely restricted. MMPs are mainly produced in latent forms with an inhibitory prodomain that masks the proteolytic active site. Activation is stimulated under conditions such as inflammation and oxidative stress [4]. Once activated, the predominant antagonistic pressure within tissues comes from the Tissue Inhibitors of Metalloproteinases (TIMPs). TIMPs are a family of 4 conserved proteins that, in addition to MMP-independent functions, inhibit and control the activity of MMPs [5, 6]. Excessive MMP activity contributes to the pathophysiology of multiple diseases [7–11], and MMP targeting of factors such as cytokines and cytokine receptors can have impactful consequences within the tissue microenvironment [12]. Historically, MMPs are grouped based on their substrate recognition, and these groups are maintained in sequence-based phylogenetic analysis [13], broadly categorized as the collagenases and gelatinases (archetypal MMPs), matrilysins, and membrane-type MMPs. The gelatinases consist of MMP2 and MMP9, originally known as gelatinase A and B, both of which display a unique relationship with TIMPs due to a secondary mode of interaction between the hemopexin domain of the pro-gelatinase enzymes and the C-terminal domains of certain TIMPs [14].

In this manuscript, we reveal a new complexity in the relationship between gelatinases and TIMPs. We show that at molar excess concentrations, activated MMP9 can cleave full-length TIMP proteins at their C-terminal tails which limits the interaction of TIMPs with pro-gelatinases. Furthermore, TIMP2 and TIMP4 are targets for additional cleavages that generate 14kDa N-terminal proteins which display a profound loss of biological capabilities. Our observations extend upon the existing paradigm that TIMPs regulate MMP activity, supporting the notion that the relationship between these two families is a complicated balancing act between the proteinases and their inhibitors.

Results

Active MMP9 targets TIMP1/2/3/4

We first observed the appearance of a 14kDa band in TIMP2 immunoblots following incubation of TIMP2 with MMP9 in the presence of 1mM APMA (to promote activation of MMP9, referred to as activated MMP9 or aMMP9), at physiologic pH, ionic strength, and temperature. Further experimentation revealed that aMMP9 cleaved TIMP2 when in a molar excess (Figure 1A), although increasing molar ratios did not correspond with a proportional increase in cleavage efficiency. Recombinant N-terminal TIMP2 (corresponding to residues C1-E127, plus a 6x His-tag), which also runs at 14kDa in immunoblots, could not be cleaved with a 10-fold molar excess of aMMP9 (Figure 1B), indicating that cleavage occurs within the C-domain of TIMP2.

Figure 1. Molar excess of active MMP9 (aMMP9) cleaves TIMPs.

(A) Combining TIMP2, MMP9 and 1mM APMA produces a 14kDa TIMP2 fragment when analyzed by immunoblot, only when MMP9 is in excess. (B) A protein corresponding to the N-terminal domain of TIMP2 does not exhibit cleavage via immunoblotting. (C) Cleavage of TIMP2 can be blocked by 200uM BB94 (Batimastat) and is not inhibited by a protease inhibitor cocktail of 10.4mM AEBSF, 8uM aprotonin, 0.4mM bestatin, 140uM E-64, 0.2mM leupeptin, 150uM pepstatin A. (D) The catalytic domain of MMP9 (Annaspec, # AS-55576–1) is not sufficient to support TIMP2 cleavage. (E) Alpha fold predicted structure of full length MMP9, the catalytic domain (pale purple), PEX domain (blue), hinge domain (grey), and pro domain (magenta) are highlighted. (F) aMMP9 can also target TIMP1/3/4, as determined by immunoblotting. (G) A 10-fold molar excess of active MMP2 does not cleave TIMPs.

Ala+TIMP2 is a TIMP2 analog containing an N-terminal appended alanine residue that prevents TIMP2-mediated inhibition of MMP proteolytic activity [15]. We tested the ability of aMMP9 to cleave C-terminally His-tagged Ala+TIMP2 (Ala+TIMP2-His) under the same conditions as previous experiments, allowing us to chemically target MMP9 activity without the additional influence of TIMP2 intrinsic MMP inhibitory activity. In these experiments with Ala+TIMP2, we observed a more efficient cleavage compared to that of TIMP2 (Figure 1C). We also show that the cleavage of Ala+TIMP2 is inhibited by the broad-spectrum metalloproteinase inhibitor BB94, but is not significantly inhibited by an excess of protease inhibitor cocktail (targeting aminopeptidases and serine/cysteine/acid proteases) (Figure 1C), clarifying that cleavage is the result of metalloproteinase activity. The protease inhibitor cocktail we utilized shows an appreciable inhibitory effect against cleavage through the activity of Bestatin, classically referred to as an aminopeptidase inhibitor that displays MMP inhibitory characteristics in the micromolar range. [16]. Furthermore, the selective MMP2/9 inhibitor (MMP-2/MMP-9 Inhibitor I, Millipore Sigma; catalog number 444241) also prevents cleavage (Figure S1A). Additionally, we show that independently sourced MMP9 (Sino Biological, catalog number 10327-HNAH) could cleave Ala+TIMP2-His at ratios as low as 1:1 in a concentration-dependent fashion (Figure S1B), and this was not significantly inhibited by an excess of protease inhibitor cocktail (Figure 1C, S1C). These findings imply that cleavage can occur under physiological conditions. The aMMP9 cleavage of Ala+TIMP2-His generates several protein products, one of around 20kDa and a subtle doublet near 14kDa (Figure 1C). Furthermore, independently sourced MMP9 catalytic domain did not induce cleavage of TIMP2, suggesting that the hemopexin (PEX) domain of MMP9 is required for cleavage to occur (Figure 1D). The AlphaFold predicted domain structure of MMP9 is highlighted in Figure 1E, illustrating the proximity between the PEX and catalytic domain of MMP9.

Human TIMP proteins display 40–50% sequence similarity. We show that aMMP9 can also cleave other TIMP family members: As with TIMP2, aMMP9 processing of TIMP4 generates at least two fragments, a C-terminal cleaved protein of 20kDa that is easy to resolve from full-length TIMP4 due to a large C-terminal His-tag (TGH(8)GGQ) (BioLegend, # 767502), and a secondary cleavage site that produces a ~14kDa protein (Figure 1F). aMMP9 cleavage of TIMP1 and TIMP3 each produced a single fragment of 32kDa and 20kDa, respectively (Figure 1F). In the case of TIMP1, the cleaved fragment is poorly resolved from the full-length protein on SDS-PAGE but cleavage is verified by loss of the C-terminal His-tag epitope (Figure 1F). The TIMP1/2/3 antibodies utilized in these experiments are directed toward epitopes located in the N-termini, supporting the idea that cleavage occurs within the C-terminal domains of these TIMP proteins. This is further evidenced by the use of C-terminal epitope-targeted TIMP4 antibodies, which show a clear loss of antigenicity towards 20kDa and 14kDa cleaved TIMP4 products (Figure S2A/B). We determined that the TIMP2 antibody (Cell Signaling Technology, #5738) also detects TIMP4, so this N-terminal targeting antibody was used to detect aMMP9 cleavage products of TIMP4 (Figure 1F). Cleavage of TIMP4 also occurred with independently sourced MMP9 (Figure S2C). In agreement with earlier observations, cleavage of TIMP1/3/4 was BB94-sensitive and not blocked by protease inhibitor cocktail (Figure S2D).

Like MMP9, MMP2 is an endogenous gelatinase with 46% identity to human MMP9 (NCBI BLAST). Unlike aMMP9, a 10-fold molar excess of active MMP2 (aMMP2) did not cleave TIMPs (Figure 1G). However, the addition of aMMP2 contributes a protective effect against TIMP proteins that is apparent in these experiments as an enhanced signal for TIMP proteins in lanes shared with aMMP2. We speculate that this effect may be due to decreased non-specific absorption of free TIMP proteins in the presence of excess MMP2.

By comparing the aMMP9 proteolytic activity towards Ala+TIMP2–6xHis and TIMP2, we illustrate the variation in cleavage efficiency between experimental replicates, which could be quite striking across the large number of preliminary experiments performed (Figure 2A/B). Despite the variable reaction efficiency, the observation that Ala+TIMP2–6xHis is more efficiently cleaved than TIMP2 provides further evidence that the cleavage of TIMPs by aMMP9 is dependent on its metalloproteinase activity. The use of 4-aminophenylmercuric acetate (APMA) is an effective tool in the activation of MMPs, promoting a stepwise autoactivation through cleavage of the prodomain from the metalloproteinase. Experimental evidence suggests this effect is a result of disruption of the interactions between a conserved cysteine residue of the MMP prodomain and active site [17]. We show that TIMP2 can inhibit this organomercurial-mediated activation cascade (Figure 2C & S3) and this observation could contribute to the high levels of variability in cleavage efficiency. Both the efficiency and reproducibility of TIMP cleavage reactions can be significantly improved by pre-activating MMP9 for 6–16h, followed by desalting the aMMP9 using Bio-Gel P-6, and incubating the aMMP9 with TIMP2 for up to 24 hours (Figure 2D). Addition of freshly prepared APMA to cleavage reactions with desalted aMMP9 reveals no change in cleavage efficiency, suggesting that a factor other than APMA is interfering with MMP9 cleavage capabilities (Figure 2E). Using this new method, we observe the anticipated dose-dependency of cleavage reaction efficiency (Figure 2F). Time-course experiments describing TIMP1/3/4 cleavage by MMP9 reveal a similar trend to that of TIMP2 (Figure S4).

Figure 2. Cleavage efficiency can be great improved by gel filtration of activated MMP9.

(A) The variation in cleavage efficiency can be noted by comparing immunoblots of the ala+TIMP2-His and TIMP2 cleavage experiments. (B) SDS-PAGE and Coomassie staining reveals that TIMP2 inhibits APMA-induced MMP9 activation. (C) Schema describing the workflow for subsequent cleavage experiments. (D) Gel filtration of activated MMP9 using Bio-Gel P-6 greatly supports aMMP9 cleavage of TIMP2. (E) Using the new workflow, cleavage of TIMP2 by aMMP9 shows greater dose-dependency.

Cleaved TIMP2/4 display perturbed MMP2/9 inhibition

MMP2/9 reverse zymography shows that processed TIMPs retain MMP2 and MMP9 inhibitory capability (Figures 3A & 3B). TIMP1–6xHis (33kDa) displays a slight drop in molecular weight to 32kDa following incubation with aMMP9 (Figure 1D). Furthermore wild-type untagged TIMP1 (Abcam, # ab280945) also shows a subtle drop in molecular weight following cleavage by aMMP9, indicating that the 6xHis-tag is not the target of aMMP9 (Figure S5). Cleaved TIMP3 presents as a lower molecular weight doublet, unique to reverse zymography, that may represent resolution of two subtle nicking sites or post-translationally modified isoforms of TIMP3. In contrast to TIMP1, there is a distinct drop in the observed molecular weight of aMMP9-treated TIMP3 from 30kDa to 20kDa, suggesting that the larger immunoreactive fragment lacks the solitary N-glycosylation site at residue N184. On the contrary, TIMP1 has two N-glycosylation sites that reside within its N-terminal domain at N30 and N78. The reverse zymography data from TIMP2/4 demonstrate unexpected results that do not correspond with results from chemically reduced immunoblotting experiments. This can be explained by the maintenance of disulfide linkages between the cleavage site and the C-terminus, resulting in a disulfide-linked cleavage product migrating as a dimer of two cleavage fragments (Figure 3C). Performance of reverse zymography requires that samples are not reduced or alkylated prior to electrophoresis in order to maintain MMP inhibitory activity. What is evidenced in these reverse zymograms is that the 14kDa fragments of TIMP2 and, more prominently, TIMP4 display diminished MMP2/9 inhibitory capabilities when compared with corresponding immunoblots from the same reaction (Figure 3A/B/E/F). By comparing with a 14kDa recombinant N-terminal domain of TIMP2 (N-T2; highlighted in red in Figure 3D) we show, via a reducing immunoblot, that aMMP9 cleavage produces an approximately 14kDa peptide (Figure 3E). When the same immunoblot is performed under non-reducing conditions, identical to those used for reverse zymography, a slightly higher molecular weight cleavage product (~17 kDa) is observed that displays greatly reduced MMP inhibitory activity in the reverse zymography assays (Figure 3E). Further experiments illustrate that TIMP1/3 cleaved products are not altered in their apparent molecular weights under reducing conditions. In contrast, TIMP4 behaves similarly to TIMP2 as reduction reveals the presence of a disulfide-linked cleavage fragment that is resolved under reducing conditions. In conclusion, the aMMP9 cleavage products of TIMP1 and TIMP3 do not maintain adjoining disulfide linkages, whereas the cleaved TIMP2 and TIMP4 fragments do maintain these linkages (Figure 3F).

Figure 3. Cleavage impacts the MMP-inhibitory capabilities of TIMP2/4.

(A/B) TIMP proteins were incubated alone or with aMMP9 overnight, followed by loading on a (A) MMP2 reverse zymogram or (B) MMP9 reverse zymogram to reveal the MMP-inhibitory capabilities of cleaved TIMP fragments. (C) Schema highlighting the N-terminal domain of TIMP2 in red, generated from Protein Data Bank structure ID 1BR9. (D) Comparison reduced and non-reduced immunoblots, and reverse zymography (non-reduced) of cleaved TIMP2 in comparison to the N-terminal domain fragment of TIMP2 reveals that residual disulfides maintain cleaved peptides as a heterodimer. (E) TIMP4-His cleavage proteins are also maintained as a heterodimer due to residual disulfide linkages.

Identification of cleavage regions

Initial observations using an N-terminal domain peptide of TIMP2 suggest that cleavage occurs within its C-terminal domain (Figure 1B), and to support this premise we utilized an antibody that specifically detects amino acids at positions 30–40 of the mature TIMP2 protein (without signal peptide) (Clone 3A4, Santa Cruz Biotechnology, # sc-21735) to illustrate detection of a 14kDa protein corresponding to the same 14kDa fragment detected in previous immunoblots (Figure S6). To identify the MMP9 cleavage sites in TIMPs, we performed sequence analysis through mass spectrometry (MS). A series of experiments were performed that included direct analysis of cleavage reaction mixtures, and cleavage fragment enrichment using SDS-PAGE. These experiments revealed an unanticipated difficulty with the identification of specific cleavage sites across all TIMP members, but consistent with the possibility of cleavage at multiple loci within their C-terminal domains. To corroborate that the 14kDa cleaved TIMP2 peptide retains an intact N-terminus, using mass spectrometry we sequenced the 14kDa peptide that was separated by reduced SDS-PAGE and confirmed that this peptide fragment had an intact N-terminal sequence (CSCSPVH) which is the same as full-length TIMP2 (Table S1A). Surprisingly, using this method we could not identify an expected ~6kDa peptide representing the residual C-domain of TIMP2 (Figure S7), suggesting that there could be multiple additional cleavage sites within the C-terminal domain or that the ~6kDa C-domain fragment is particularly unstable, possibly due to extensive exposure of hydrophobic amino acids that usually reside at the interface between the N- and C-domains of TIMPs (Figure S8). Additional experiments were performed to identify the cleavage sites of TIMP2, yet these experiments could only identify the sites that occur at TIMP2’s C-terminal tail. The location of a cleavage site corresponding to the 14kDa protein could not be identified, although the data from Table S1A supports that the 14kDa TIMP2 remains intact until E127. Interestingly, multiple cleavage sites could be identified within TIMP2’s C-terminal tail at the carboxyl side of negatively charged E187/192, D190/193, or the structurally similar polar residue Q186. The predominant product of aMMP9 processing of TIMP2 produces a truncated TIMP2 peptide ending at Q186 (Tables S1B–D). A particularly striking consequence of fully processed TIMP2 (TIMP2_Q186) is the loss of 4 charged residues (two D resides and two E), producing a predicted overall change in isoelectric point from 6.48 to 8.23 (Expasy - Compute pI/Mw tool).

The predominant product of TIMP3 cleavage produced TIMP3_K180, a truncation of 8 C-terminal amino acids (Table S2), although peptides truncated at D179 and S181 were also detected, suggesting some fragmentation at this cleavage site. This data also confirms the loss of N184, the sole glycosylation site in TIMP3. The C-terminal tail cleavage site of TIMP4 is predicted by sequence comparison with TIMP2/3, based on Clustal Omega Multiple Sequence Alignment (https://www.ebi.ac.uk/) (Figure S8). Two experiments using LysC or GluC digestion of cleavage reactions could not successfully identify unique cleavage residues in TIMP4. Due to a lack of a comparable C-terminal tail, the TIMP1 cleavage site could not be predicted using sequence alignment, however our analyses indicate the removal of up to four C-terminal residues in TIMP1 by aMMP9 (Table S3). All of the findings and conclusions from this series of experiments, including the putative and predicted cleavage sites, are summarized in Figure 4A.

Figure 4. Summary of findings from mass spectrometry experiments investigating the sequence of cleaved TIMP proteins.

(A) Linear structure of TIMP proteins highlighting their N-terminal domains in grey, and C-terminal domains with identified and predicted cleavage sites. The cleavage site for TIMP1 is predicted to reside within the final four residues. For TIMP2, the predominant detected P1 site is Q186, with other less abundant sites at E187, D190 E192, and D193. The cleavage site within the 14kDa TIMP2 could not be identified, but is following C128 and, based on the observed molecular weight, likely within the following 8 residues. The most abundant P1 site for TIMP3 was identified as K180, with less abundant detection at D179 and S181. No specific cleavage sites were detected for TIMP4 but based on sequence comparison using Clustal Omega Multiple Sequence Alignment, the cleavage sites and regions for the 20kDa and 14kDa TIMP4 are predicted. (B) Alpha fold predicted structures for the TIMP family proteins, depicting a largely conserved tertiary structure. Highlighted are the C-terminal tail regions, which TIMP1 is lacking.

Using Alpha-Fold to predict the structure of TIMP proteins, which performs accurately when comparing with the defined structure of TIMP2 (PDB ID 1BR9; structure truncated at Ala182), we compare the structures of the TIMP family [18, 19] . The structure predictions highlight the much smaller C-terminal tail domain of TIMP1 compared with that of the other TIMP family members. This shortened tail in TIMP1 may explain the lack of significant C-terminal cleavage by aMMP9 (Figure 4B, S9).

Altered biological functions of cleaved TIMPs

The high-affinity interactions between TIMP2 and MMP2 are well characterized, and the expression of these two proteins is strongly correlated within tissues [20]. The importance of the acidic C-terminal tail of TIMP2 (QEFLDIEDP) for interaction with the C-domain (hemopexin (PEX) domain) of MMP2 (C-MMP2) has been described elsewhere [21]. To assess the affinity of 20kDa cleaved TIMP2 (His+Ala+TIMP2Q186) with C-MMP2, we performed a nickel affinity pulldown assay and surface plasmon resonance to confirm that 20kDa TIMP2 displays a striking loss in affinity for C-MMP2 (Figure 5A & 5B). TIMP2 (His+Ala+TIMP2) displays a sub-nanomolar affinity for C-MMP2 (KD = 0.76 nM), whereas recombinant TIMP2_Q186 demonstrated a much lower KD of 1.9 μM (Figure 5B), validating the apparent loss of MMP2 binding following cleavage observed in the pulldown assay in Figure 5A. Considering the strong affinity of this interaction, and the significant correlation between TIMP2 and MMP2 expression, it is reasonable to expect that TIMP2 may predominantly reside as a heterodimer with MMP2 in tissues. To assess whether the TIMP2:MMP2 heterodimer is protected from cleavage by aMMP9, we performed a similar cleavage reaction to previous experiments, this time utilizing complexes of TIMP2:MMP2 that were purified by gelatin-affinity chromatography from A2058 melanoma cells [22]. These results describe how TIMP2 is largely protected from aMMP9 cleavage when residing as a complex with both pro- and activated MMP2 (Figure 5C, the crystal structure of the complex illustrated in Figure S10 [23]). Furthermore, we find that activation of MMP2 in this complex using APMA does not remove protection, suggesting that the protective interaction with the MMP2-PEX domain can still occur between TIMP2 and active MMP2 (Figure 5C). Additionally, by combining TIMP2 with proMMP2 or active MMP2 for 2 hours, we show that TIMP2 is still protected (Figure 5D). Using only the catalytic domain of MMP2 (Enzo Life Sciences, # BML-SE237-0010) we show that the protective effect is lost, illustrating that the PEX domain of MMP2 is required to protect TIMP2 from MMP9-mediated cleavage (Figure 5E). All TIMPs can bind to MMP2 through the formation of an inhibitory complex, and TIMP2/3/4 can also bind to the PEX domain of MMP2 through their C-termini. We assessed whether proMMP2 or active MMP2 could protect the other TIMP family members from cleavage. Our results show that the presence of pro- or active MMP2 exerted significant protection for TIMP3 against cleavage by aMMP9. In contrast, MMP2 exerted minimal protection against TIMP4 cleavage, although the C-terminal TGH(8)GGQ tag of TIMP4 likely interferes with PEX domain binding. TIMP1 was not protected in the presence of either form of MMP2 (Figure S11).

Figure 5. Exploration of the properties of cleaved TIMPs.

(A) Recombinant His-Ala-TIMP2Q186, corresponding to 20kDa cleaved TIMP2, loses affinity for proMMP2 assayed via nickel pulldown. (B) The loss of His-Ala-TIMP2Q186 affinity with the MMP2 C-domain is also described by surface plasmon resonance. (C) TIMP2:MMP2 complex is protected from cleavage by aMMP9. (D) Active MMP2 can also protect TIMP2 against aMMP9 cleavage. (E) Gelatin sepharose pulldown of MMP9/Ala+TIMP1/EDTA cleavage samples followed by analysis via (i) immunoblotting and (ii) gelatin zymography. (F) 20kDa cleaved TIMP3 retains heparin-agarose binding capabilities, TIMP1 also displays an affinity for heparin-agarose and TIMP2 displays poor affinity for heparin-agarose.

To determine whether the cleavage of TIMP1 resulted in an analogous loss of MMP9 hemopexin C-domain binding, we performed a cleavage experiment using Ala+TIMP1-His followed by gelatin-sepharose pulldown. The N-terminal alanine of Ala+TIMP1-His precludes TIMP1 interaction with the MMP9 catalytic domain, thereby preventing TIMP1 co-precipitation through interactions with the MMP9 catalytic domain. Immunoblot analysis of the output and elution samples reveals the presence of a unique cleavage product of Ala+TIMP1-His at a molecular weight of approximately 18kDa (Figure 5F(i)). Assessment of these same samples using zymography reveals that, as expected, gelatin sepharose could pulldown pro/active MMP9 very efficiently, but Ala+TIMP1-His was only bound to proMMP9. The ~18kDa Ala+TIMP1 cleavage fragment did not interact with either pro- or aMMP9 (Figure 5F(ii)). This data suggests that, unlike the TIMP2:aMMP2 C-terminal (hemopexin domain) interaction that confers a protective effect against MMP9 cleavage, TIMP1 does not bind to the C-terminal domain of active MMP9. Indeed, it has been previously reported that the PEX domain of aMMP9 has diminished affinity for components of the ECM, versus proMMP9 [24].

TIMP3 is unique amongst the family due to a strong association with components of the extracellular matrix (mainly glycosaminoglycans). To assess whether loss of the C-terminal tail and glycosylation site could mediate any gain/loss in this ability, we performed a heparinaffinity pulldown with full-length and 20kDa cleaved TIMP3 (Figure 5G). The results demonstrate how nicking does not significantly alter the affinity of TIMP3 to binding to heparin, although there is potentially a slight decrease in affinity that is suggested by the detection of 20kDa TIMP3 in the output sample (Figure 5G(i)). This finding is consistent with prior reports that identified numerous critical basic amino acid residues (K26, K27, K30, K76, R163, K165) mediating TIMP3 binding to extracellular matrix component such as heparin [25]. To confirm the specificity of the pulldown experiment, TIMP1 and TIMP2 control pulldowns were also performed illustrating an appreciable affinity for TIMP1 with heparin, and a low affinity of TIMP2 for heparin-binding as previously reported [26] (Figure 5G(ii)).

Discussion

The balance between MMPs and TIMPs has long been touted as a key indicator of tissue health that plays a key role in ECM turnover in healthy and diseased tissues [4, 27, 28]. However, direct evidence of this generally accepted theory is limited, and sometimes conflicting, due to complex regulatory mechanisms that include; a large family of Metzincin proteinases, transcriptional regulation, post-translational modifications, chemical/physical activation/inactivation, and the existence of numerous endogenous inhibitors of MMP proteolytic activity. The latter point is further complicated by the role that endogenous MMP inhibitors also play in the activation of MMPs [29, 30]. These variables make it difficult to directly evaluate the balance between MMP activity and the effectiveness of their endogenous inhibitors. In this report, we reveal an additional layer of complexity to consider when assessing metalloproteinase activity in tissues.

After discovering that active MMP9 can cleave TIMP2, we show that all TIMPs are proteolytic targets for this activated gelatinase. We also show the specificity for this interaction in that the related gelatinase, MMP2, does not cleave TIMP proteins. The C-terminal tail of TIMP1 is truncated compared with that of other TIMPs, and the cleavage that occurs in this region relieves TIMP1 of up to four amino acids. Considering that the C-terminal tail of TIMP1 is shorter and has a distinct sequence compared to that of TIMP2, our data suggest further that the mode of interaction between TIMP1 and latent MMP9 involves unique interactions when compared to the TIMP2:proMMP2 complex. While the TIMP2:proMMP interaction demonstrates a clear dependence on acidic charged residues at the C-terminus of TIMP2 [18], the shorter C-terminal tail of TIMP1 does not possess similarly charged amino acid residues [31]. We also show that aMMP9 can produce a unique, lower molecular weight (~18kDa) cleavage product within the Ala+TIMP1-His analog of TIMP1. Since the 18kDa cleavage product is unique to Ala+TIMP1-His, we did not investigate it further. Furthermore, because we found that TIMP1 does not interact with the C-domain (PEX domain) of active MMP9, we did not assess whether ~32kDa cleaved TIMP1 retained binding capacity with the MMP9 PEX domain. TIMP3 is cleaved at its C-terminal tail, which is prominently revealed by immunoblotting largely due to loss of the N-glycosylation site at N184. Although TIMP3 can bind to both proMMP2 and proMMP9, it does so with slightly lower affinity than TIMP2 and TIMP1, respectively. The conditions that may support TIMP3:MMP2/9 complex formation in vivo are poorly understood but are likely influenced by C-terminal cleavage of TIMP3. Unlike the other family members, TIMP3 is anchored to the extracellular matrix through interactions with sulfated glycosaminoglycans such as chondroitin sulfate, heparan sulfate and heparin [32]. We show that the ability of TIMP3 to bind to heparin is retained by 20kDa cleaved TIMP3, which is consistent with the localization of the TIMP3 amino acid residues involved in glycosaminoglycan binding of TIMP3 in the N-terminal domain [32]. Furthermore, our experimentation also demonstrated that TIMP1–6xHis exhibited strong binding with heparin. Whether this observation is related to the His-tag, or the source of the heparin (porcine, type I. Sigma Aldrich #H6508) was not investigated. Interestingly, deglycosylation has been associated with increased peptide aggregation and angiogenic properties of TIMP3, a phenomenon that may contribute to the pathogenesis of Sorsby fundus dystrophy [33], although it is unclear whether MMP9 activity plays a role in the etiology of this disease.

TIMP2 and TIMP4 are both cleaved by aMMP9 at two or more regions, producing predominantly 20kDa and 14kDa peptides that are detected by an N-terminal targeting antibody for TIMP2/4 in reduced immunoblots. Both cleavage products (20kDa and 14kDa products) have significant effects on the biological functions of TIMP2 and TIMP4. TIMP2 transcript expression displays a strong correlation with that of MMP2 in tissues [20], and considering the high affinity of their interaction in formation of the proMMP2:TIMP2 complex, it is reasonable to predict that the bulk of secreted MMP2 exists as a heterodimer with TIMP2. Furthermore, this predicted complex formation is consistent with studies in which proMMP2 was originally isolated from human A2058 melanoma cell conditioned media as a 1:1 molar complex with TIMP2 [34]. We show that cleavage of approximately 8 residues from TIMP2’s C-terminal tail to produce the 20kDa TIMP2_Q186 causes a dramatic loss in TIMP2 affinity for the C-domain of MMP2, as expected [18]. Conversely, the TIMP2:MMP2 complex itself is protected from aMMP9 cleavage, revealing that this level of regulation is only relevant for MMP2-free TIMP2. The factors that govern binding/dissociation of the TIMP2:MMP2 complex remain poorly characterized, although recent reports highlight a potential role for phosphorylation of TIMP2, and interaction with extracellular HSP90, that could represent key regulatory steps in this system [35–37]. TIMP4 has also been described to bind with latent MMP2, although this complex is of lower affinity and is unable to mediate activation of MMP2 via trimolecular complex formation with MMP14 [38]. The TIMP4 we utilized contains a C-terminal TGH(8)GGQ tag that we propose will interfere with PEX binding, however it is expected that endogenous TIMP4 will receive at least partial protection against cleavage through interaction with MMP2. Due to the highly restricted expression of TIMP4 [39], we did not assess this further. Interestingly, our data suggests that TIMP2 retains affinity for the PEX domain of active MMP2, however, the same is not true for TIMP1 and the C-domain of active MMP9. This provides further evidence for the unique modes of interaction between TIMPs and gelatinases. The interaction between TIMPs and the PEX domain of MMP2 requires penetration of the TIMP C-terminal tail between beta-propeller blades III and IV of the PEX domain, forming three salt bridges [19]; this extended tail is absent from TIMP1. The interaction between TIMPs and proMMP9 requires more extensive interactions [40], which likely include crucial stabilizing interfaces with residues of the MMP9 pro-domain, and may explain why Ala+TIMP1-His did not pulldown with aMMP9. Considering that the protective effect of MMP2 likely occurs through steric hindrance, it is possible that other interactions may provide an analogous protective effect, such as that between TIMP1 and CD63 that is proposed to require the C-terminus of TIMP1 [41]. Furthermore, it would be of great interest to determine whether cleaved TIMP1 loses affinity for CD63.

The cleavage of TIMP2 and TIMP4 by aMMP9 produces a 14kDa N-terminal protein that, in addition to the loss of their C-domains and latent proMMP2 affinity, demonstrate a clear diminution of MMP inhibitory activity as shown by reverse zymography. These observations were surprising, considering the inhibitory N-terminal domain remains intact. Peripheral interactions with MMPs that extend into the C-domain of TIMPs exist, and these can support the unique affinities of the different TIMP members for certain MMP family members [42–44]. This can be appreciated in the reverse zymogram in figure 3E where the 50ng N-terminal TIMP2 protein translates to a 3.7-fold molar excess when compared to 20ng TIMP2, yet the inhibitory abilities are comparable. The pronounced decrease in MMP inhibitory capability of cleaved TIMP2/4 may be explained by the probability that the cleavage products are maintained as a larger structure due to intact disulfide bridges that are not broken following cleavage, potentially creating an awkward configuration that may perturb the formation of an inhibitory complex. This raises the question as to whether there is an endogenous secreted reductase that can mediate the reduction of the retained disulfides, or whether the cleaved TIMP2/4 is subject to further degradation.

Considering the long research history between TIMPs and MMP9, it is surprising that the proteolytic targeting of TIMPs by MMP9 remained elusive. This could be explained by the variation we experienced in our experimental results. The inconsistency in cleavage could be a result of the complex enzyme:inhibitor relationship between MMPs and TIMPs, the relative instability of APMA, and the requirement to remove low molecular weight contaminants to generate efficient cleavage. The presence of TIMPs can perturb the APMA-induced autoactivation of MMPs, adding additional variables to compensate for when studying this occurrence. Furthermore, it has been previously reported that a sequence corresponding to the pro-domain of MMP2 can independently inhibit the activity of active MMP2 [45]. We provide evidence that this may also be true for MMP9, whereby the cleavage of TIMPs is much more efficient following the gel filtration of activated MMP9 using Bio-Gel P-6.

Excess MMP activity has been implicated in a wide range of pathologies, observations that led to the development of broad-spectrum MMP inhibitors that ultimately failed in clinical trials [4, 46–48]. Despite this clinical set back, aberrant MMP activity continues to be linked with the pathophysiology and poor prognosis of many human diseases. In cancer, clear patterns of TIMP expression have begun to emerge that correlate with disease progression and morbidity/mortality, with enhanced TIMP1 expression being strongly associated with poor prognoses in various cancers [49]. TIMP2/3 RNA levels, on the other hand, are often unchanged or reduced, and numerous studies have reported the anti-tumor/tissue normalizing capabilities of these two abundant TIMP family members [50, 51]. Our report describing the targeting of TIMPs by MMP9 raises important questions as to whether this has an important role in normal physiology, and whether it is associated with the pathophysiology of the various diseases associated with high MMP9 activity. If true, excess MMP9 activity could spawn regions of unchecked metalloproteinase activity within the afflicted tissue. The conditions that support a molar excess of activated MMP9 over TIMP proteins are unknown, although likely culprits include myeloid cell types such as neutrophils and macrophages. Neutrophils are the most abundant human leukocytes and are often described as first responders to tissue injury and inflammation. They also play key roles in chronic inflammatory conditions, to where they are constantly recruited [52]. Furthermore, neutrophil-derived TIMP1-free MMP9 has been reported to undergo efficient activation in vivo [53]. This may represent a seeding event for the establishment of a molar excess of active MMP9 over TIMP family members who may already be consumed with regulating other inflammation-associated metalloproteinases. MMP9 targeting of TIMPs may not be strictly limited to pathology, with well-regulated MMP9 activity being described as critical for central nervous system development and plasticity [54]. These findings also have implications for the larger metalloproteinase family that includes A Disintegrin and Metalloproteinases (ADAMs) and ADAMs with Thrombospondin Type 1 Motifs (ADAMTS), whose activities can also be modulated by members of the TIMP family.

In vivo, MMP9 exists in monomeric, multimeric, and heterocomplex forms [55], as well as exhibiting a range of post-translational modifications [56], some of which can be appreciated in our 2D gel electrophoresis (Figure S3). Whether these different forms of MMP9 display differing activities towards TIMP cleavage is of great interest, but beyond the goals of this investigation. Proteolytic targeting of TIMPs is not restricted to MMP9, with others reporting TIMP2 can be degraded in cell lines in a manner that was inhibited by synthetic metalloproteinase inhibitors [57]. However, this report also described how these inhibitors prevent TIMP2 cell binding, and that TIMP2 degradation was lysosomal. Additionally, human neutrophil elastase (HNE) has been long known to directly cleave TIMP1 between Val⁶⁹-Cys⁷⁰ (without signal peptide), destroying the MMP-inhibitory capabilities of TIMP1, and that this activity is accelerated by heparin [58–60]. Whether HNE targets other TIMP family members is unknown but there is some conservation of these residues across the family, with TIMP2 being identical and TIMP3/4 substituting Valine with the similar aliphatic hydrophobic amino acid leucine.

To conclude, we show that TIMP proteins are proteolytic targets for active MMP9 in an interaction that likely occurs with lower affinity to that of the inhibitory complex between TIMPs and MMP9. For this reason, aMMP9 needs to be in a molar excess for cleavage to occur. These observations shed new light on the complex relationship between metalloproteinases and their inhibitors and reveals that MMP9 may play a prominent role in dictating the metzincin:TIMP balance within tissue microenvironments.

Methods

Recombinant proteins

Human recombinant proteins utilized in this study are summarized in Table 1. In-house produced his-tagged proteins were generated by HEK-293-F cells and purified using a two-step chromatographic procedure [61]. TIMP2:MMP2 complex was purified by gelatin-affinity chromatography from A2058 cell conditioned media [22, 34]. Recombinant MMP2/MMP9 were purified by gelatin-affinity chromatography of the conditioned media of BSC-1 cells infected with vaccinia virus constructs encoding full length MMP2/MMP9 cDNA [22, 62]. Recombinant C-domain of MMP2 (C-MMP2) was produced in E. coli cultures as described previously [63]. Untagged recombinant TIMP2 was produced by the Protein Expression Laboratory, National Cancer Institute (Frederick, MD) using preparative size-exclusion chromatography on conditioned media from Expi293F cells (ThermoFisher) transfected with full length human TIMP2 cDNA.

Table 1.

Protein	Tag	Source	Notes
TIMPl-His	H(6)	In-house
TIMP1		Abcam (ab280945)
Ala+TIMPl-His	H(6)	In-house
TIMP2-His	H(6)	In-house
TIMP2		In-house	Produced at the Protein Expression Laboratory, Frederick National Laboratory, NCI.
ala+TIMP2-His	H(6)	In-house
H+A+TIMP2	H(6) (N-terminus)	Genscript (contract production)
H+A+TIMP2Q186	H(6) (N-terminus)	Genscript (contract production)
N-TIMP2-His	H(6)	In-house
TIMP3
TIMP4-His	TGH(8)GGQ	BioLegend (767502)
MMP2		In-house
TIMP2:MMP2 Complex		In-house
C-MMP2	H(6) (N-terminus)	In-house
MMP2(cat)		Enzo Life Sciences (BML-SE237–0010)
MMP9		In-house
MMP9		Sino Biological (10327-HNAH)
MMP9 catalytic domain		Annaspec #AS-55576–1		Corresponding to residues G112-P445)

In vitro cleavage reactions

Cleavage reactions between MMP9 and TIMP were incubated with 1mM 4-Aminophenylmercuric acetate (APMA) at 37°C for 16 hours in enzyme buffer (EB; 50mM Tris, pH 7.5, 200mM NaCl. 5mM CaCl2, 0.02% Brij-35). Alternatively, MMP2/9 were activated by 1mM APMA for 1 hour (MMP2) or 6–16 hours (MMP9) at 37 °C, with MMP2 and MMP9 activation concentrations being 4uM and 8uM, respectively. Activation conditions were preassessed by zymography to confirm full/majority activation to 62kDa and 82kDa proteoforms for MMP2 and MMP9. Activated MMPs were desalted into EB using Bio-Gel P-6 packed centrifugal filter devices (Nanosep centrifugal devices, Pall Life Sciences) by centrifugation at 1000G for 2 minutes. Cleavage reactions were allowed to incubate at 37°C for 16–24 hours. Reactions were stopped by freezing or mixing with Laemmli sample buffer + 350mM 2-mercaptoethanol.

SDS-PAGE Immunoblotting

Protein samples were mixed with Laemmli sample buffer (Bio-Rad). If reducing conditions were required, then the samples were supplemented with 350mM 2-mercaptoethanol and heated at 95 °C for 5 minutes. Equivalent to 20ng of TIMP or cleaved peptide was loaded into a 4–20% polyacrylamide gel (Bio-Rad) or hand-cast 15% polyacrylamide gel and run at 150V for 40–90 minutes, until acceptable resolution of the PageRuler Prestained Protein Ladder (ThermoFisher) was achieved. Gels were transferred to a nitrocellulose membrane using a TransBlot Turbo System (Bio-Rad). Nitrocellulose membranes were briefly washed with Tris-buffered Saline 0.1% Tween-20 (TBST), then blocked in 5% milk TBST for 30 minutes. Antibody staining was performed for 2 hours at RT or overnight 4°C in 2.5% milk TBST, washed 3x with TBST (5 minutes per wash) and incubated with the appropriate secondary antibody conjugated to horse radish peroxidase in 2.5% milk for 45 minutes. Blots were washed 3x with TBST, then imaged on a ChemiDoc MP imaging system (Bio-Rad), using either SuperSignal^™ West Pico PLUS Chemiluminescent Substrate (ThermoFisher) or Radiance Plus (Azure Biosystems).

2-Dimensional Gel Electrophoresis (2DGE)

7uM of MMP9 and 1.4uM TIMP2 was combined with 1mM APMA in a 60uL reaction and incubated for 16h. 55uL of the samples were desalted using Amicon ultra 10K filter and diluted to a final total volume of 50 μl in 1:1 diluted SDS Boiling Buffer: Urea Sample Buffer with reducing agents before loading all of each. Two-dimensional electrophoresis was performed according to the carrier ampholine method of isoelectric focusing (IEF) [64] by Kendrick Labs, Inc. (Madison, WI). IEF was carried out in a glass tube of inner diameter 2.3 mm using 2.0% pH 3–10 Isodalt Servalytes (Serva, Heidelberg, Germany) for 9,600 volt-hrs. One μg of an IEF internal standard, tropomyosin, was added to each sample. This protein migrates as a doublet with lower polypeptide spot of MW 33,000 and pI 5.2; an arrow on the stained gels marks its position. The enclosed tube gel pH gradient plot for this set of ampholines was determined with a surface pH electrode. After equilibration for 10 min in Buffer ‘O’ (10% glycerol, 50 mM dithiothreitol, 2.3% SDS and 0.0625 M tris, pH 6.8), each tube gel was sealed to the top of a stacking gel that overlaid a 10% acrylamide slab gel (0.75 mm thick). SDS slab gel electrophoresis was carried out for about 4 hours at 15 mA/gel. The following proteins (Millipore Sigma) were used as molecular weight standards: myosin (220,000), phosphorylase A (94,000), catalase (60,000), actin (43,000), carbonic anhydrase (29,000), and lysozyme (14,000). These standards appear as bands at the basic edge of the Coomassie brilliant Blue R250-stained 10% acrylamide slab gels. The gels were dried between sheets of cellophane paper with the acid edge to the left.

Surface Plasmon Resonance

Interactions between TIMP2 and C-MMP2 were analyzed by surface plasmon resonance using a Biacore T200 system (GE Healthcare). A CM5 chip was utilized to capture ~2000RU of TIMP2 and TIMP2Q186. C-MMP2 was prepared in degassed, filtered HBS-P+ (GE Healthcare) buffer. Single cycle kinetic experiments were carried out using five injections (30 μL/min) of increasing concentration of protein (10–500 nM) passed over the sensor chip for 180 seconds association followed by a 420 second dissociation. Following buffer and reference subtraction kinetic constants and binding affinities were determined utilizing the Biacore T200 evaluation software (GE Healthcare).

Nickel Pulldown

Using a final volume of 100uL, 200nM MMP2 was combined with 100nM H+A+TIMP2 or H+A+TIMP2Q186 in enzyme buffer (EB; 50mM Tris, pH 7.5, 200mM NaCl. 5mM CaCl2, 0.02% Brij-35) and incubated at 4°C for 2 hours. 100uL Nickel-NTA resin beads (Thermo Scientific; 88221) were blocked for 1 hour with 100uL DPBS 1% BSA at 4°C. 10% of “input” sample was retained for testing, then protein mixtures were incubated with the blocked and washed Nickel-NTA resin for 2 hours with rotation. The tubes were centrifuged at 1250G to pellet the resin, and supernatant retained as “output”. The resin was washed 3x with 300uL EB + 25mM imidazole, then eluted with 100uL EB + 250mM imidazole for 10 minutes with gentle vortexing every 60 seconds.

Heparin Pulldown

100uL heparin-agarose beads (Sigma Aldrich; H6508–5ML) were blocked overnight with 100uL Dulbecco’s phosphate buffered saline (DPBS) 1% BSA at 4°C. The following morning, the beads were washed with DPBS and centrifuged at 800G for 1 minute. Following the removal of the supernatant, the beads were split between two 1.5mL tubes. 18uL control (0.35uM TIMP3) or cleavage (0.35uM TIMP3 + 7uM aMMP9) samples were quenched with 20uM of the MMP inhibitor BB94, added to the blocked heparin-agarose beads, and incubated for 4 hours with shaking at 4°C. Samples were centrifuged at 800 x g for 1-minute, washed 4x with DPBS, and eluted with 40uL 2x Laemmli sample buffer at 95°C for five minutes.

Gelatin Pulldown

150uL gelatin sepharose 4B (Cytiva; 17095601) per sample were washed twice with 10 volumes enzyme buffer, then incubated with 150uL cleavage experiment samples (0.7uM MMP9:0.07uM Ala+TIMP1-His) for 2 hours at 4°C while rotating. Gelatin sepharose was pelleted at 1000G for 1 minute, and supernatant retained as output. Gelatin sepharose was washed 3 times with 400uL enzyme buffer, and then eluted in 150uL enzyme buffer + 15% DMSO for 10 minutes with rotation. 20uL/1uL of sample was used for immunoblotting/zymography, respectively.

Liquid Chromatography and Tandem Mass Spectrometry

Liquid chromatography tandem mass spectrometry (LC-MS/MS) was performed by Poochon Scientific (Frederick, MD). If gel/membrane-captured proteins were processed, these were eluted into solution according to Poochon Scientific standard operating procedures (SOPs). In-solution protein samples were subjected to 10 volumes cold acetone precipitation for 4–16 hours, followed by centrifugation at 13,000 x g for 15 minutes. The supernatant was removed, and pellet allowed to air dry. The protein pellet was resuspended in 96uL ammonium bicarbonate and digested with trypsin/LysC, LysC, Chymotrypsin, or GluC according to Poochon Scientific SOPs. The digested peptide mixture was then concentrated and desalted using a C18 resin tip and eluted in 15uL of 0.1% formic acid. 12uL of peptides for each sample were analyzed by a 60-minute LC-MS/MS run using an Orbitrap Exploris 240 Mass Spectrometer and a Dionex UltiMate 3000 RSLCnano System (Thermo Scientific). Digested peptides were loaded onto a peptide trap cartridge at a flow rate of 5 μL/min. The trapped peptides were eluted onto a reversed-phase Easy-Spray Column PepMap RSLC, C18, 2 μM, 100A, 75 μm × 250 mm (Thermo Scientific) using a linear gradient of acetonitrile (3–36%) in 0.1% formic acid. The elution duration was 110 min at a flow rate of 0.3 μl/min. Eluted peptides from the Easy-Spray column were ionized and sprayed into the mass spectrometer, using a Nano Easy- Spray Ion Source (Thermo) under the following settings: spray voltage, 1.6 kV, Capillary temperature, 275°C. Other settings were empirically determined. Raw data files were searched against human TIMP protein sequences using the Proteome Discoverer 2.4 software (Thermo Scientific) based on the SEQUEST algorithm. Carbamidomethylation (+57.021 Da) of cysteines was set as fixed modification, and Oxidation / +15.995 Da (M), and Deamidated / +0.984 Da (N, Q), Phospho / +79.966 Da (S, T, Y), Acetylation /+42.011 Da (K), Ubiquitination / +114.043 Da (K) were set as dynamic modifications. The minimum peptide length was specified to be five amino acids. The precursor mass tolerance was set to 15 ppm, whereas fragment mass tolerance was set to 0.05 Da. The maximum false peptide discovery rate was specified as 0.01.

Supplementary Material

1

Figure S1. MMP9 cleavage of TIMP2. (A) MMP2/9 Inhibitor (Sigma Aldrich, # 444241) prevents MMP9 cleavage of ala+TIMP2. (B) A unique source of MMP9 (Sino Biological, # 10327-HNAH) also cleaves ala+TIMP2. (C) Protease inhibitor cocktail (PIC; 10.4mM AEBSF, 8uM aprotonin, 0.4mM bestatin, 140uM E-64, 0.2mM leupeptin, 150uM pepstatin A) does not inhibit cleavage.

2

Figure S2. MMP9 cleaves other members of the TIMP family. (A–C) Immunoblotting of TIMP4 cleavage samples using different antibodies. Identical samples corresponding to ~20ng TIMP4 were immunoblotted using different TIMP4 antibody clones to illustrate C-domain cleavage of TIMP4. (D) All members of the TIMP family are cleaved by aMMP9, and cleavage is inhibited by the MMP inhibitor BB94, and not inhibited by protease inhibitor cocktail.

3

Figure S3. 2D gel electrophoresis (2DGE) of Active MMP9:TIMP2 cleavage samples. (A) Approximately 1.6ug TIMP2 +/− 29.2ug MMP9 and a tropomyosin control were loaded for 2DGE and stained with Coomassie brilliant Blue R250. (B) Corresponding immunoblot with the same samples.

4

Figure S4. Time-course experiments illustrating the time-dependency of TIMP cleavage by aMMP9. Each experiment was performed using 2uM aMMP9 and 0.2uM TIMP.

5

Figure S5. Reverse zymography of a cleavage experiment utilizing TIMP1 without an epitope tag. Epitope tag-free TIMP1 was purchased from Abcam (Cambridge, UK), product ID AB280945. The ladder is labeled but is not an accurate depiction of sample molecular weight in reverse zymograms.

6

Figure S6. Cleaved TIMP2 detected with a N-terminal targeted antibody provides evidence for an intact N-terminal domain. An antibody that specifically detects amino acids at positions 30–40 of the mature TIMP2 protein (without signal peptide) (Clone 3A4, Santa Cruz Biotechnology, # sc-21735) illustrates detection of a 14kDa protein corresponding to the same 14kDa fragment detected in previous immunoblots.

7

Figure S7. Coomassie stained PVDF membrane reveals prominent detection of 14kDa cleaved TIMP2. Smaller fragments are not detected.

8

Figure S8. Alpha Fold structure of TIMP2, with residues colored by hydrophobicity. (A) Full length TIMP2 versus (B) TIMP2 C-domain, revealing exposure of numerous hydrophobic residues within the C-domain that originally interface with the N-domain.

9

Figure S9. Clustal Omega Multiple Sequence Alignment of human TIMP proteins, with the identified/predicted cleavage sites highlighted.

10

Figure S10. Crystal structure of TIMP2 in complex with proMMP2 (PDB # 1GXD).

11

Figure S11. MMP2 can protect TIMP3 from aMMP9 cleavage. TIMP1-His and TIMP4-His are not protected against cleavage with the presence of MMP2.

Highlights.

• Active MMP9 targets TIMPs when in molar excess.

• All TIMPs are cleaved at their C-terminus.

• TIMP2/4 are processed further, producing 14kDa N-terminal peptides.

• Cleavage has functional consequences on TIMP function.

Abbreviations

MMP

Matrix Metalloproteinase

aMMP

Active Matrix Metalloproteinase

TIMP

Tissue Inhibitor of Metalloproteinase

ECM

Extracellular Matrix