TIMP-3 facilitates binding of target metalloproteinases to the endocytic receptor LRP-1 and promotes scavenging of MMP-1

Matrix metalloproteinases (MMPs) and the related families of disintegrin metalloproteinases (ADAMs) and ADAMs with thrombospondin repeats (ADAMTSs) play a crucial role in extracellular matrix (ECM) turnover and shedding of cell-surface molecules. The proteolytic activity of metalloproteinases is post-translationally regulated by their endogenous inhibitors, known as tissue inhibitors of metalloproteinases (TIMPs). Several MMPs, ADAMTSs and TIMPs have been reported to be endocytosed by the low-density lipoprotein receptor-related protein-1 (LRP-1). Different binding affinities of these proteins for the endocytic receptor correlate with different turnover rates which, together with differences in their mRNA expression, determines their nett extracellular levels. In this study, we used surface plasmon resonance to evaluate the affinity between LRP-1 and a number of MMPs, ADAMs, ADAMTSs, TIMPs and metalloproteinase/TIMP complexes. This identified MMP-1 as a new LRP-1 ligand. Among the proteins analyzed, TIMP-3 bound to LRP-1 with highest affinity (KD = 1.68 nM). Additionally, we found that TIMP-3 can facilitate the clearance of its target metalloproteinases by bridging their binding to LRP-1. For example, the free form of MMP-1 was found to have a KD of 34.6 nM for LRP-1, while the MMP-1/TIMP-3 complex had a sevenfold higher affinity (KD = 4.96 nM) for the receptor. TIMP-3 similarly bridged binding of MMP-13 and MMP-14 to LRP-1. TIMP-1 and TIMP-2 were also found to increase the affinity of target metalloproteinases for LRP-1, albeit to a lesser extent. This suggests that LRP-1 scavenging of TIMP/metalloproteinase complexes may be a general mechanism by which inhibited metalloproteinases are removed from the extracellular environment.

Ki values of TIMP-3 for various metalloproteinases were determined under equilibrium kinetic conditions 32 , as previously described in Ref. 21 . Final enzyme concentrations for Ki determinations were as follows: MMP-1ΔC and MMP-3ΔC were used at 1 nM; MMP-13 at 0.1 nM, and ADAMTS-4 and -5 were used at 0.5 nM; ADAM17 was used at 1 nM. Enzymes were pre-incubated (1 h, 37 °C) with TIMP-3 (0.5-100 nM) and equilibrium rate of substrate hydrolysis was determined (1-18 h, 37 °C). Prism software (GraphPad, La Jolla, CA, USA) was used to fit the data to the tight-binding equation 32 .
Surface plasmon resonance analysis. The affinity of various ligands to LRP-1 was evaluated using a BIAcore T200 (GE, Amersham, UK). LRP-1 was immobilized on a CM5 sensor chip by amine coupling using N-hydroxysuccinimide (NHS), according to the manufacturer's instructions. Immobilization was performed at 10 μl/min until reaching a target ligand concentration of 3000 RUs on the chip. The instrument was maintained Scientific RepoRtS | (2020) 10:12067 | https://doi.org/10.1038/s41598-020-69008-9 www.nature.com/scientificreports/ at 25 °C. SPR analysis was performed in 150 mM NaCl, 10 mM CaCl2, 50 mM Tris-HCl buffer, 0.01% Tween-20, pH 7.5, at a flow rate of 30 μl/min. After each cycle, the sensor chip was regenerated by injecting 30 μl of 10 mM glycine-HCl buffer, pH 2.5. All ligands were injected over the immobilized LRP-1 at 6 concentrations ranging from 0 to 80 nM (except BSA which was used as a negative control and injected at up to 2 μM; MMP-1ΔC that was injected at up to 2 μM; and MMP-14, ADAM-10 and ADAM-17 ectodomain that were injected at up to 160 nM). Metalloproteinase/TIMP complexes were pre-formed in vitro by incubating equimolar concentrations of enzyme and inhibitor for 1 h at 37 °C. Complexes were injected at 5 different concentrations (5 nM, 10 nM, 20 nM, 40 nM and 80 nM). The K i of MMP-1 for TIMP-1, TIMP-2 and TIMP-3 is 0.38 nM, 1.03 nM and 1.1 nM respectively 17,22 , therefore metalloproteinase/TIMP complexes should be stable at all the concentrations used for injection. Each determination was performed in triplicate. BIAevaluation software was used to analyze the resulting sensorgrams and determine the association (k a ) and dissociation (k d ) rate constants, and the constant of dissociation (K D ). Binding data were fitted in a one-site binding model.

TIMP-3 forms stable complexes with selected metalloproteinases in the low nanomolar range.
In order to analyze the binding of TIMP-3/MMP complexes to LRP-1, we first investigated the inhibition of selected metalloproteinases by TIMP-3 in the low nanomolar range. This allowed us to evaluate the lowest concentration at which these metalloproteinases and TIMP-3 form a stable complex. Generally, enzyme kinetic theory and experimental observation indicate that incomplete inhibition is seen at concentrations near the K i value, while complete inhibition is observed at enzyme and inhibitor concentrations higher than 5 times the K i value 32 . We determined a K i(app) for TIMP-3 inhibition of MMP-1ΔC of 1.86 nM (Table 1), in agreement with previous reports 21,33 , meaning that around 50% of enzyme activity was blocked by equimolar amounts of enzyme. However, when enzyme concentrations were increased to 5 nM, 90% of MMP-1 activity was blocked by an equimolar amount of TIMP-3, indicating that the large majority of the enzyme is in a stable complex with the inhibitor (Fig. 1A). TIMP-3 bound to MMP-3ΔC and MMP-13 even more tightly, with 5 nM TIMP-3 completely blocking the activity of 1 nM MMP-3ΔC and 0.1 nM MMP-13 ( Fig. 1B-C). For MMP-3ΔC, we calculated a K i(app) value of 0.55 nM (Table 1) in agreement with previous reports 21, 33 , but for MMP-13, the K i value was too low for accurate determination by this method, as has been previously described 34 .
In conclusion, these enzymatic assays demonstrated that TIMP-3 and a number of its target metalloproteinases form stable complexes in vitro at concentrations in the very low nanomolar range. To analyze binding to LRP-1 and endocytosis of the metalloproteinase/TIMP-3 complexes, we thus selected to incubate the metalloproteinases with TIMP-3 at concentrations of 5 nM and above to ensure stable complex formation.

TIMP-3 bridges the binding of metalloproteinases to LRP-1 in vitro.
We have previously shown that TIMP-3 directly binds to LRP-1 using ELISA 17 . In order to understand this interaction in greater detail, we used surface plasmon resonance to determine the affinity of TIMP-3 for LRP-1 immobilised on a BIAcore sensor chip. Receptor-associated protein (RAP), which is a well-established LRP-1 ligand 35,36 , was used as positive control for binding. We measured a K D of 0.6 nM ( Table 2), in good agreement with the previous determination of Lee and colleagues, who reported a K D of 1.4 nM for this interaction 37 . The negative control, BSA, did not show any binding at concentrations up to 2 μM. TIMP-3 bound to LRP-1 with high affinity (K D = 1.68 nM, Table 2, Fig. 2A). The N-terminal domain of TIMP-3 (N-TIMP-3) bound to LRP-1 with a K D of 5.65 nM ( Table 2), confirming that the minimal binding determinants reside in the N-terminal domain of the inhibitor. This is in line with our previous finding that N-TIMP-3 is endocytosed by chondrosarcoma cells in an LRP-1-dependent manner with similar kinetics to that of TIMP-3 17 . TIMP-1 and TIMP-2 also bound to LRP-1, albeit with 18-fold and 16-fold lower affinities respectively (Table 2).
MMP-1 is a new LRP-1 ligand-Because MMP-1 undergoes autocatalysis at room temperature, we used a catalytically inactive form of MMP-1 [MMP-1 (E200A)] to investigate binding of this enzyme to LRP-1. SPR Table 2. Kinetics of ligand binding to LRP-1 measured by SPR. LRP-1 was immobilized on a CM5 sensor chip and the binding constant of different analytes were determined. Each determination was performed in triplicate at six concentration for each analyte [0-80 nM except ADAMTS4-2 (0-180 nM), BSA, MMP-1ΔC and MMP-14 ectodomain (0-2 μM)]. Binding constants were calculated using the BIOevaluation software.  www.nature.com/scientificreports/ analysis showed that MMP-1 (E200A) bound to LRP-1 with a K D of 34.6 nM ( Table 2). The MMP-1 catalytic domain (MMP-1ΔC) showed no binding to LRP-1 even at micromolar concentrations ( Table 2), indicating that the hemopexin domain is required for binding of MMP-1 to the receptor. This is in agreement with previous studies showing that the hemopexin domain of MMP-9 mediates the majority of MMP-9 interaction with LRP-1 38 .
TIMPs facilitate binding of MMP-1 and MMP-13 to LRP-1 in vitro-Emonard et al. reported that the pro-MMP-2/TIMP-2 complex has a higher affinity for LRP-1 than either pro-MMP-2 or TIMP-2 alone 8 . We thus investigated whether interaction with TIMP-1, TIMP-2 or TIMP-3 could increase MMP-1 binding to LRP-1. MMP-1/TIMP complexes were made in vitro by incubating equimolar concentrations of MMP-1 and TIMPs (5-80 nM) for 1 h at 37 °C before addition to immobilised LRP-1. Previously reported K i values for MMP-1 inhibition by TIMP-1 and TIMP-2, which are 0.38 nM and 1.03 nM respectively, are similar to that of TIMP-3, indicating that complexes formed by MMP-1 with all 3 of these TIMPs remain stable at the concentrations used, since these are all well above the K i values 17,22 . Interaction with TIMP-3 increased MMP-1 affinity for LRP-1 by sevenfold, to a K D of 4.96 nM (Table 2, Fig. 2B). MMP-1ΔC had no detectable affinity for LRP-1 on its own, but when in complex with TIMP-3, a K D of 6.42 nM was detected ( Table 2), indicating that TIMP-3 is largely responsible for the binding of this complex to LRP-1. To further prove that formation of a complex with TIMP-3 is necessary for its binding to LRP-1, we exogenously added MMP-1ΔC either alone or in a complex with TIMP-3 to HTB94 chondrosarcoma-like cells and analyzed its uptake from the conditioned media after 12 h. In agreement with the SPR results, MMP-1ΔC, which had no affinity for LRP-1, was not taken up by cells (Fig. 2C). In contrast, when the enzyme was complexed to TIMP-3, both MMP-1ΔC and TIMP-3 were internalized by cells. The LRP-1 inhibitor RAP inhibited this process. Altogether, these results indicate that MMP-1ΔC can only bind to and be internalized by LRP-1 when in a complex with TIMP-3.
Interaction with TIMP-1 and TIMP-2 also increased MMP-1 affinity for LRP-1, by 2-and 4.5-fold respectively. Unlike for TIMP-3, the affinities of the complexes are higher than that for TIMP-1 or TIMP-2 alone (Table 2), www.nature.com/scientificreports/ suggesting that MMP-1 either contributes to the binding, or that it stabilises a TIMP-1/-2 conformation with increased LRP-1 affinity. TIMPs were also able to promote binding of other MMPs to LRP-1. To study MMP-13, which undergoes autodegradation at room temperature, we used a catalytically inactive mutant, namely MMP-13 (E204A). This collagenase had a K D value of 25.6 nM for LRP-1 (Table 2). Interaction with TIMP-3 increased its affinity for LRP-1 by 7.6-fold, while interaction with TIMP-1 or TIMP-2 increased its affinity by 2.4-and 4.6-fold respectively. ( Table 2). The ectodomain of MMP-14 showed no detectable binding to LRP-1 even at micromolar concentrations. Incubation with TIMP-3 facilitated binding with a K D of 2.95 nM, while incubation with TIMP-1 or TIMP-2 did not detectably improve affinity (Table 2).

TIMP-3 increases the affinity of ADAMTS-4 and deletion mutants of ADAMTS-5 for LRP-1-
The aggrecanase ADAMTS-5 is an LRP-1 ligand, with the LRP-1-binding determinants localized in the ancillary domains of the enzyme. Affinity for LRP-1 thus decreases with consecutive deletion of the non-catalytic C-terminal domains 13,14 .
TIMP-3 also improved affinity of ADAM10 and ADAM17 for LRP-1-Binding of LRP-1 was undetectable for the free form of either ADAM10 or ADAM17, while binding to TIMP-3 improved their affinity markedly, with K D values of 8.67 for ADAM10/TIMP-3 and 6.84 for ADAM17/TIMP-3. This suggests that TIMP-3 can facilitate the binding of these enzymes to LRP-1.
In conclusion, a systematic SPR analysis showed that TIMP-3 binds to LRP-1 with the highest affinity among TIMPs. TIMPs generally increased affinity of the binding of specific collagenases (MMP-1 and -13) to LRP-1, with TIMP-3 mediating this process to the greatest extent.  (Fig. 3A). Similar results were obtained when [ 35 S]TIMP-3 was in complex with MMP-1ΔC, which had no affinity at all for LRP-1 in vitro (Fig. 3A). In addition to full-length MMP-1 and MMP-1ΔC, clearance of [ 35 S]TIMP-3 was not affected when the inhibitor was in complex with MMP-3ΔC, MMP-9 or MMP-14 ectodomain (Fig. 3B). Then K D of TIMP-3 for LRP-1 when in complex with ADAMTS5-2 or ADAMTS4-2 was similar to that of TIMP-3 alone. In agreement, complex formation with ADAMTS5-2 or ADAMTS4-2 did not significantly affect the internalization kinetics and clearance of [ 35 S]TIMP-3 (Fig. 3C, D).

Complex formation with metalloproteinases does not affect the kinetics of TIMP-3 endocytosis. TIMP-3 is rapidly internalized by various cell types via
In conclusion, the internalization rate of TIMP-3 is not affected by complex formation with metalloproteinases, suggesting that the inhibitor can drive the internalization of TIMP-3/metalloproteinase complexes.

Discussion
Two classes of metalloproteinases, MMPs and ADAMTSs, play a central role in ECM turnover, together with the ADAMs that are majorly involved in ectodomain shedding of cell-surface components. The activity of these enzymes is finely regulated by their endogenous inhibitors, known as TIMPs. LRP-1 is an endocytic receptor that has been reported to mediate the internalization and degradation of a number of MMPs, ADAMTSs and TIMPs, thus functioning as a master regulator of ECM turnover 8,9,11,[13][14][15]17 . Levels of these proteins in the ECM can tightly correlate to their affinity for the receptor. For example, TIMP-3, which was reported to bind to LRP-1 with high affinity, is hardly found extracellularly unless its interaction with LRP-1 and subsequent endocytosis are prevented 17,39 . Investigating affinities for LRP-1 of different metalloproteinases, TIMPs and metalloproteinase/TIMP complexes can provide useful information about turnover of these proteins and potential effects on ECM turnover. www.nature.com/scientificreports/ In the current study, we used SPR to systematically analyze LRP-1 interaction with a number of MMPs, ADAMs, ADAMTSs, TIMPs and metalloproteinase/TIMP complexes. We found that, among these proteins, TIMP-3 bound to LRP-1 with highest affinity. Furthermore, TIMP-3 increases the affinity for LRP-1 of target proteinases that bind to LRP-1 by themselves, such as MMP-1, MMP-13 and ADAMTS-4, but also that of metalloproteinases, including MMP-14, ADAM10 and ADAM17, which have no affinity for the receptor. ADAMTS-5 binds to and is rapidly endocytosed by LRP-1 14 , and we found that the affinity of ADAMTS-5 lacking the C-terminal thrombospondin domain (ADAMTS5-2) for the receptor was similar to that of TIMP-3. After secretion, ADAMTS-5 is extracellularly processed to truncated forms that acquire distinct features from the fulllength counterpart 40,41 , including different LRP-1 affinities and endocytosis rates 13 . Expression and purification of recombinant full-length ADAMTS-5 is inefficient, so we did not analyze the full-length enzyme in this study. While TIMP-3 had little effects on the affinity of ADAMTS-5 for LRP-1, it doubled the affinity of a truncated form of ADAMTS-5 lacking of the cysteine-rich and spacer domain (ADAMTS5-4) for the receptor, suggesting TIMP-3 may promote scavenging of smaller, degraded forms of the enzyme.
Similarly to TIMP-3, TIMP-1 and TIMP-2 directly bound to LRP-1, although with lower affinities. Both TIMP-1 and TIMP-2 increased the affinities of MMPs for LRP-1, but not to the same extent as TIMP-3. This indicates that there are at least two modes by which inhibitors can increase proteinase affinity for LRP-1. In the simplest case, binding may be mediated primarily through residues on the inhibitor, such that the inhibitor/ proteinase complex has similar affinity for LRP-1 as the free inhibitor alone. Such behaviour was observed for TIMP-3, with TIMP-3/MMP-13, TIMP-3/MMP-1 and TIMP-3/MMP-14 complexes having similar affinity for LRP-1 as free TIMP-3. We demonstrated that this mode of binding is able to support biologically relevant bridging of the MMP to LRP-1, as TIMP-3 was able to promote MMP-1 endocytosis by HTB94 chondrosarcoma cells.
Alternatively, the proteinase-inhibitor complex may bind with higher affinity than either protein alone. We observed that TIMP-1 and TIMP-2 complexes with both MMP-1 and MMP-13 had ~ 2-fivefold higher affinity for LRP-1 than any of the four proteins alone. This is in line with what has been reported for the proMMP-2/ TIMP-2 complex, which has ~ tenfold higher affinity for LRP-1 than either component alone 8 . This may be due to a conformational change induced by proteinase-inhibitor interaction, as is the case with α 2 -macroglobulin, which has low affinity for LRP-1 in its free form, but which undergoes a conformational change upon proteinase binding that increases affinity of the complex for LRP-1. Alternatively, both enzyme and inhibitor may directly participate in LRP-1 binding, each contributing a lysine residue to form the minimal LRP-1 binding unit 35 . We consider this the more likely mechanism, as TIMPs are considered to be conformationally stable, with only minor structural rearrangements upon interaction with MMPs. However, this mode of binding appears to be insufficient to support bridging to LRP-1 in a complex biological environment, as we found that neither TIMP-1 nor TIMP-2 could increase the rate of MMP-1 endocytosis by HTB94 chondrosarcoma cells. Thus, our kinetic data indicate that TIMP-3 bridges MMP binding to LRP-1 though a different mechanism to that of TIMP-1 and TIMP-2, and that only TIMP-3 can promote scavenging of these target enzymes from the extracellular environment.
The observation that TIMP-3/metalloproteinase complexes and free TIMP-3 have similar affinity for LRP-1 to that of TIMP-3 alone suggests that TIMP-3 interacts with LRP-1 through a site distinct from its inhibitory ridge. Studies of a number of LRP-1 ligands suggest that a pair of basic residues is responsible for binding to acidic pockets on the complement-type repeats (CRs) of the receptor 42,43 . For example, extensive mutagenesis and crystallography studies indicate that RAP binds to LRP-1 via two lysine residues, K256 and K270, which are located 22 Å apart on the surface of the molecule 35 . TIMP-3 contains a cluster of lysine and arginine residues situated on the opposite side to the inhibitory ridge. This basic region mediates binding to sulfated proteoglycans of the extracellular matrix 44 . We previously found that heparin blocks TIMP-3 binding to LRP-1 17 , suggesting that the heparin-binding and LRP-1-binding sites of TIMP-3 overlap, and supporting the postulate that the metalloproteinase-inhibitory region of TIMP-3 is spatially separated from the LRP-1-binding and overlapping ECM-binding basic region.
There is emerging evidence that metalloproteinases have functions other than proteolytic cleavage, and that these functions are often related to their non-catalytic domains. For example, a non-catalytic C-terminal fragment of MMP-12 has been shown to have antimicrobial activity 45 . ADAMTS-1 and ADAMTS-4 display anti-angiogenic properties through their ability to bind VEGF and sequester it from VEGFR 46,47 . Fragments of ADAMTS-5 derived from its ancillary domains have been shown to be anti-tumorigenic both in vitro and in vivo 48 . ADAMTS-12 modulates neutrophil apoptosis, likely by interacting with CD36 and CD47 through its thrombospondin domains 49 . Although it has not been explicitly demonstrated, it is likely that at least some of these non-catalytic functions are maintained when metalloproteinases interact with TIMPs. In addition to its role in inhibiting the proteolytic activity of target metalloproteinases, TIMP-3 may thus also inhibit non-proteolytic activities of target proteinases by facilitating their LRP-1-mediated endocytosis. This would position TIMP-3 as a central regulator of metalloproteinase levels and activity in the extracellular environment.
It is interesting to speculate that competition between different TIMPs for metalloproteinases could lead to diverse biological responses. TIMP-1 and TIMP-2 would serve primarily to inhibit the catalytic activity of the proteinases, but would not facilitate their endocytic clearance. TIMP-3, on the other hand, is uniquely able to address the enzymes to intracellular degradation and therefore regulate both their proteolytic and non-proteolytic functions. We found that TIMP-3 also bridges binding of members of the ADAM and ADAMTS families to LRP-1. The ectodomains of ADAM10 and ADAM17 do not interact with the receptor unless they are in complex with TIMP-3. This may explain why TIMP-3 decreases cell surface levels of ADAM10 50 .
In conclusion, this work shows that TIMP-3 can promote LRP-1-mediated endocytosis of target metalloproteinases by facilitating their binding to LRP-1. As such, TIMP-3 can inhibit both proteolytic-dependent and -independent activities of metalloproteinases in the extracellular environment.