Utility of Glycosylated TIMP3 molecules: Inhibition of MMPs and TACE to improve cardiac function in rat myocardial infarct model

Abstract Tissue Inhibitor of Metalloproteinase 3 (TIMP3) is a secreted protein that has a great utility to inhibit elevated metalloproteinase (MMP) activity in injured tissues including infarcted cardiac tissue, inflamed vessels, and joint cartilages. An imbalance between TIMP3 and active MMP levels in the local tissue area may cause worsening of disease progression. To counter balance elevated MMP levels, exogenous administration of TIMP3 appeared to be beneficial in preclinical studies. However, the current form of WT‐TIMP3 molecule has a limitation to be a therapeutic candidate due to low production yield, short plasma half‐life, injection site retention, and difficulty in delivery, etc. We have engineered TIMP3 molecules by adding extra glycosylation sites or fusing with albumin, Fc, and antibody to improve pharmacokinetic properties. In general, the C‐terminal fusion of TIMP3 improved expression and production in mammalian cells and extended half‐lives dramatically 5‐20 folds. Of note, a site‐specific glycosylation at K22S/F34N resulted in a higher level of expression and better cardiac function compared to other fusion proteins in the context of left ventricle ejection fraction (LVEF) changes in a rat myocardial infarction model. It appeared that cardiac efficacy depends on a high ECM binding affinity, in which K22S/F34N and N‐TIMP3 showed a higher binding to the ECM compared to other engineered molecules. In conclusion, we found that the ECM binding and sustained residence of injected TIMP3 molecules are important for cardiac tissue localization and inhibition of adverse remodeling activity.

ECM turnover is a tightly regulated process that maintains a balance between matrix metalloproteinases (MMPs) and their endogenous inhibitor Tissue Inhibitor of Metalloproteinases (TIMPs) levels. It has been suggested that blocking MMP activity either by small molecule inhibitor or TIMP molecules immediately after disease-causing insults would be beneficial for various disease modifying conditions including acute myocardial infarction, acute lung injury, osteoarthritis, and inflammation. [1][2][3][4] However, small molecule MMP inhibitors have failed in a clinical heart failure trial due to intolerable side effects, particularly musculoskeletal pain syndrome, which could be due to metabolites or off-target effects. [5][6][7] There have been many trials of more specific inhibitors to avoid this side effect without resolution. 8 Cardiac tissues in response to ischemia induce the breakdown of ECM by elevating MMP levels, which induces ventricular wall thinning in the infarcted area eventually resulting in heart failure. 9,10 Postischemic loss of cardiomyocytes incites inflammatory cells, mostly neutrophils, to spike MMP levels disproportionally at the infarct site. Inadequate levels of TIMP3 to counteract protease activity at infarct site results in extensive ECM remodeling, ventricular dilation, and decline in cardiac function. 11,12 Inhibition of MMPs, specifically MMP2 and MMP9, has been shown to prevent ECM proteolysis and reduce the infarct size following myocardial injury. 13,14 All mammalian TIMPs (TIMP1-4) consist of two domains: Nterminal and C-terminal, with the N-terminal domain, designated N-TIMP exhibiting function. Each TIMP subtype shows inhibition of a distinct subset of MMPs, with TIMP3 being the broadest MMP, ADAM, and ADAM-T inhibitor. 15 TIMP3 is the most abundant subtype in the heart, and inhibits post-MI remodeling by reducing TNFα production via TACE (TNF alpha converting enzyme) inhibition as well as by decreasing endogenous MMP activity in the infarcted area. 7,16 These dynamic changes of upregulated MMPs and decreased TIMP3 levels in myocardial matrix following MI lead to myocardial ECM degradation contribute to cardiac dysfunction and adverse remodeling in the failing heart. 17,18 The level of TIMP3 was significantly reduced in post-MI animals and patients with dilated cardiomyopathy. 19,20 Thus, it is hypothesized that the exogenous administration may compensate for the loss of cardiac tissue TIMP3 after myocardial injuries.
Modification by engineering of large protein molecules is a common method to extend serum half-life. The most common modifications are fusion with albumin or Fc, which significantly extended the half-life of protein ligands, such as interferon gamma (γINF)-fused with human serum albumin. 21 PEGylation (polyethylene glycol conjugation) has a significant pharmacokinetic effect by slowing down clearance, however there seems to be some safety concerns regarding vacuolization in renal cells. 22,23 An additional modification approach is a site-specific glycosylation of protein molecules that improves pharmacokinetic (PK) property by increasing stability due to reduced protease sensitivity with bulky glycol moiety, which may reduce potential immunogenicity as well. 24,25 The serum half-life and in vivo bioactivity of glycosylated recombinant human erythropoietin (rHuEPO or darbepoetin alfa), which is used to treat anemia, was improved 3-fold upon glycosylation of two additional sites (5-N-Glyco and 1-O-Glyco) of recombinant EPO protein. 26 TIMP3 is a high affinity ECM binding protein. The ECM binding motif of TIMP3 has been characterized previously and shows that both N-and C-terminal domains interact with the ECM via positively charged Lys and Arg residues. 27 The N-terminal domain amino acid residues 1-120 is biologically active and can inhibit both MMP and TACE activity, although the inhibitory mechanisms may be distinct as revealed by mutations and crystal structures. 28,29 TIMP3 is a known substrate for the endocytic receptor called the low density lipoprotein receptor-related protein-1 (LRP-1). Blocking LRP-1 interaction may reduce clearance and extend systemic half-life. 30 In this study, we have engineered TIMP3 molecules to improve protein expression, serum half-life, and its functional consequences.

| Production of TIMP3 variants
Human TIMP3, modified TIMP3 with glycosylation mutations (v2, v82), and fusion constructs (HSA, Fc, or Ab) were expressed in a Chinese hamster ovary cell line, whereby conditioned media were concentrated by tangential flow filtration (TFF; Millipore,10 kDa MWCO) and filtered. The TIMP3 and fusion constructs were next purified through a three-column chromatography procedure, each utilizing a specific capture column; TIMP3 and glycosylation mutants with Capto MMC (GE Healthcare, Freiburg, Germany): HSA fusions with Cibacron Blue (Merck KGaA, Darmstadt, Germany): and Fc or Ab fusions with MabSelect Sure (GE Healthcare, Pittsburg, PA). Each of these was then followed by Capto Adhere and SP HP (GE Healthcare, Freiburg, Germany) chromatography steps. The TIMP3 containing SP HP fractions were finally concentrated and buffer exchanged (10 mmol/L sodium acetate pH 5.2, 9% sucrose) by TFF. N-TIMP3 (13.9 kDa N-domain; aa 1-120) was expressed as inclusion bodies in an E. coli system with a subtilisin prodomain fusion. After refolding and prodomain cleavage, the N-TIMP3 was purified by a three-column procedure using SP HP, CHT, and Butyl HP chromatography.

| Gelatin zymography
Zymography is a conventional method for detecting MMP activity.
To confirm MMP inhibition in gelatin zymography we used purified human MMP2 or MMP9 enzyme. After running the zymogram with human MMP2, the gel lanes were cut and incubated with engineered (Thermo Fisher Scientific) with or without TIMP3 molecules. The following day, gels were stained with SimplyBlue Safestain (Thermo Fisher Scientific). MMP activity was quantified using ImageJ software.

| Pharmacokinetic analysis in rats
Pharmacokinetics of TIMP3 (3 mg/kg, intravenous injection) was measured in normal male Sprague Dawley rats according to the IACUC approved protocol. TIMP3 concentrations were quantified from plasma using mass spectrometry analysis for nonfused molecules (LOQ: 10 ng/mL) and immunoassay (Gyrolab fluorescent assay, 10A7 mAb for capture and anti-HSA or Fc for detection; LOQ 100-250 ng/mL) for fused molecules. Plasma concentration data following IV administration were analyzed using either standard noncompartmental analysis or a constant IV infusion model in Phoenix version 6.4 (Pharsight, Mountain View, CA).

| Inhibition of extracellular matrix degradation
Degradation of ECM by collagenase type I (C2674 Sigma) was examined in presence of MMP inhibitors. ECM formed by 16000 HCF cultured for 7 days was used for degradation study. 100 μg of collagenase was incubated with TIMP3 or marimastat (Sigma-Aldrich) in decellularized ECM in 96-wells for 1 hour at 37°C with gentle shake. The wells were washed once with 100 μL of PBS, fixed with 4% paraformaldehyde for 15 minutes and immunostained for collagen type III.

| Fluorescent labeled TIMP3 retention in MI rat hearts
To estimate ex vivo signals of cardiac residence half-life, AF680labeled N-TIMP3 (70 μg) or AF680-labeled F-TIMP3 (270 μg) was directly injected into the infarct area of MI rats at 1 hour post-LAD ligation before closing the chest. Hearts were harvested at desired time points within 14-day postinjection, and sliced 1 mm section on slicing block before putting into microplate, which was read in Safire multimode plate reader (Tecan) at 680 nm excitation and 702 nm emission wavelength.

| Statistical analyses
For in vitro statistical analyses, data are reported as mean ± SD or SEM. Comparisons were made by paired one-tailed Student t test between control and TIMP3 treated groups for at least three experimental data sets. For in vivo statistical analyses, comparisons were performed by two-tailed Student t test between vehicle and TIMP3 treated groups. P < 0.05 was considered statistically significant.

| In vitro potency of MMP2/9 and TACE inhibition
Wild-type TIMP3 has a low expression yield and short plasma halflife due to high clearance rates compared to a recombinant antibody protein. In order to overcome these hurdles, TIMP3 was engineered with fusion, glycosylation, PEGylation, and truncation approaches. In

| Zymography of TIMP3 variants
The   respectively. As shown in Table 1, the plasma half-life of both N-TIMP3 and TIMP3v2 was <66 minutes in the rat. We hypothesized that with a specific target mediated clearance, a continuous infusion may saturate the clearance target system. Thus, a continuous intravenous infusion of TIMP3 molecules for 6-8 hours was explored in normal rats ( Figure 3C,D). The continuous infusion did not alter the fast clearance kinetics of TIMP3 at least in these rats although nonlinear PK pattern was observed between the low and high dose of TIMP3v2. A nonlinear PK was observed for TIMP3v2 between the low-and high-dose conditions. Both N-TIMP3 and TIMP3v2 displayed high clearance after systemic administration ( Table 2).

| Cardiac functional effects in MI rats
To evaluate the biological activity of TIMP3 as a heart failure target, we characterized N-TIMP3 (N-domain, 13.9 kDa), TIMP3v2 (K22S/ F34N), 5xGlyco-TIMP3 (v82), and v82 with Fc fused in the C-terminus (v82-Fc). In vitro MMP inhibitory potency was not dramatically different among modified TIMP3 molecules as these molecules potently inhibited MMP2/9 and TACE activities although they showed relatively short IV PK profiles (t 1/2 < 1 hour) except v82-Fc  (Table 3A). This is in line with the data from the pig study, in which TIMP3 inhibition of cardiac remodeling was demonstrated by reducing LVEDV and LV wall thinning in the pig MI model. 31   Cardiac tissue was homogenized in RIPA buffer with PMSF. After staining and destaining, the gel was quantified using ImageJ software (lower graph). The gel is a representative of n = 3 experiments (mean ± SEM). Graphs are averages of each ImageJ scan. *P < 0.05 (vs control), **P < 0.01 (vs control) IV delivery, which is suitable for bi-weekly injection scheme. Thus, v82-Fc was explored in the rat MI model via tail vein injection.
Unexpectedly, twice a week delivery (5 mg/kg) was not different from the untreated control group ( Figure 4D).

Time (h)
Time   Figure 5). Out of the TIMP3 variants tested, TIMP3v2 and truncated N-TIMP3 displayed significantly higher affinity for decellularized ECM compared to TIMP3-HSA and glycosylated TIMP3. Immunostaining for TIMP3 with C-terminal specific antibody ( Figure 5A) and N-terminal specific antibody ( Figure 5B (mg/heart) As shown in Figure  was also assessed in the infarct area of MI rat hearts ( Figure 6B). An intravenous delivered IR800-TIMP3v2 was detected in the infarct region as early as 30 seconds, which is technically the earliest time point possible to capture in this in vivo experimental setting. However, the signal increase was plateaued up to 10 minutes after administering through tail vein injection, which may be due to high clearance of the molecule ( Figure 6B). The plasma protein level was also detected by gel electrophoresis after blood collection (Figure 6C). This fast localization may be due to higher levels of MMP targets in the infarct area.

| Inhibition of ECM degradation by TIMP3
In order to detect ECM degradation, the decellularized ECM produced from cardiac fibroblasts was treated with type 1 collagenase. As expected, the ECM was degraded upon collagenase treatment when collagen fibril degradation was assessed by immunostaining with type 3 collagen antibody. TIMP3 displayed a concentration-dependent inhibition of collagen fibril degradation, similar to that exhibited by small molecule inhibitor marimastat ( Figure 7).  TIMP3 (µg/well) ** * * * F I G U R E 7 TIMP3 inhibition of collagen fibril degradation by collagenase: Collagenase type-I (100 μg/well) was incubated with acellular ECM in the presence of varying concentrations of TIMP3v2 (0, 1 μg, 3 μg, 10 μg, 30 μg, 100 μg), N-TIMP3 (300 μg) or MMP inhibitor Marimastat (100 μmol/L) for 1 hour at 37°C with gentle shake before fixing with 4% PFA. Collagen fibril degradation was assessed by immunostaining using anti-Collagen type III antibody. Scale bar = 200 μm. Graph is the average of n = 3 experiments with SEM. *P < 0.05 (vs control), **P < 0.01 (vs control), paired t test intracoronary catheter in a pig balloon occlusion model, and demonstrated that the intracoronary delivered TIMP3 can prevent adverse myocardial remodeling in ischemia-reperfusion injuries. 32 To identify a suitable TIMP3 form to improve biological property and function, we have generated multiple glycosylated forms of TIMP3 and fusion constructs. PEGylation with varying sizes of polyethylene glycols (2-20 kDa) was employed to many pharmaceutical molecules to improve the pharmacokinetic property as it prevents a rapid clearance due to its bulkiness with PEG conjugation. Indeed, PEGylated TIMP3v2 exhibited longer plasma half-life (>28 hours) compared to other modifications (Table 1). However, some safety concerns were raised for cellular vacuolization even though it has significant pharmacokinetic effect by slowing down clearance. 22,23 Modification by glycosylation has been well established in large molecule ligands. For example, the PK profile of darbepoetin alfa (DA), glycosylated EPO, was dramatically improved by two additional glycosylation (5-N-Glyco and 1-O-Glyco) of recombinant EPO molecule. 26 Even though TIMP3 has an endogenous glycosylation site in the C-terminus, additional glycosylation by site-specific engineering may further improve PK property by increasing stability due to reduced protease sensitivity with bulky sugar moiety, which may also reduce potential immunogenicity. 24,25 It appeared that additional glycosylation preserved MMP inhibitory activity as well as beneficial cardiac effects in MI rats upon direct myocardial injection, but its plasma half-life extension was minimal. Interestingly, myocardial half-life after direct injection into the myocardial tissue exhibited at least 30-fold longer retention than in the plasma. A direct myocardial injection of v2 and v82 variants showed no noticeable differences in terms of cardiac function in the rat MI model evaluation (Table 3). It was of interest to find if a systemic delivery with a single bolus injection is a more favorable delivery route than the myocardial injection. v82-Fc allowed us to administer intravenous delivery with extended half-life of 15 hours. However, it did not exhibit any significant improvement in the same rat MI model. We reasoned that molecule's ECM binding property may play a role. A single additional glycosylation (F34N) does not seem to interfere with ECM binding because TIMP3v2 interacted well with the ECM in the ECM binding experiment. On the contrary, v82 and v82-Fc molecules did not show binding to ECM with high affinity. It seemed the bulkiness or masking of critical amino acid residues by glycosylation or fusion proteins interferes with ECM binding. The ECM binding of TIMP3 is rapid and also accumulates in the infarcted area, which may be triggered by the There are limitations to the study. First, the variability of in vitro fluorescence-based MMP assays was high although the assay is more sensitive than the conventional zymography. It may be caused by aggregation of recombinant proteins, which can be improved further to reduce the variability. Second, there were relatively smaller animal numbers in the in vivo studies due to large consumption of proteins, especially infusion studies. A longer lasting TIMP3 variant will be required for future studies. Third, although the present study provided potential mechanistic insight regarding the ECM binding property and inhibition of ECM degradation, it remains to be unclear that this in vitro data can be directly translated into in vivo cardiac effects. More mechanism related studies are warranted. In conclusion, these data may suggest that an exogenous delivery of TIMP3 into the infarcted cardiac area can reduce worsening of cardiac function potentially by preventing ECM degradation in the rat heart.

DISCLOSURE
None declared.