Background
The ovariectomized (OVX) immature rat has been used extensively to study hormone-specific regulatory processes in the mammalian uterus. In this model, administration of a single physiological dose of 17β-estradiol (E
2; 40 μg/kg) induces a biphasic uterine growth response, characterized by rapid tissue hypertrophy followed by hyperplasia. The early hypertrophic growth phase peaks approximately 4 h after E
2 administration and is characterized by extensive collagen matrix remodeling of the endometrial stroma [
1]. The controlled degradation and repair of the stromal extracellular matrix is a necessary component of a variety of normal reproductive processes, including endometrial cycling [
2‐
4], implantation [
5], and pregnancy [
6], and involves regulated expression and activity of proteolytic enzymes including the matrix metalloproteinases (MMPs) by steroid hormones. Hormone-induced regulation of MMP expression is complex and appears to involve cell-type specific changes in expression of local paracrine mediators as well as direct and indirect effects on mRNA transcription via activated receptors.
Matrix metalloproteinases comprise a family of related enzymes, each of which can degrade at least one component of the extracellular matrix, consisting of numerous types of collagen, gelatin, elastin, fibronectin, and laminin, among others [reviewed in [
7]. In addition, MMPs have been shown to act on non-matrix substrates by modifying and/or activating other MMPs [
8‐
10], growth factors [
11], cytokines [
12,
13], chemokines [
14,
15], and other growth regulators that are matrix-bound or present on the cell surface [
16,
17]. Expression is controlled at the level of transcription in response to proto-oncogenes and exogenous signals, including hormones [
18,
19], cytokines [
20,
21], and growth factors [
22]. At the post-translational level, MMPs in the extracellular space are regulated by endogenous inhibitors, including a group of tissue-derived inhibitors termed TIMPs (tissue inhibitors of metalloproteinases) [
7,
23]. Furthermore, most MMPs are secreted as inactive proenzymes (proMMPs), also known as zymogens, requiring the cleavage of a specific propeptide domain for activation [
24]. Although the mechanisms governing proMMP activation are not completely understood, evidence suggests that inflammatory mediators function to regulate MMP activity in the extracellular space [
4,
9,
20,
25].
Many studies have shown that administration of physiological dose level of estrogen to the immature or OVX rodent induces an acute inflammatory-like response, including the coordinated influx of blood-born macrophages, neutrophils, and eosinophils into the uterine endometrial stroma [
26‐
31]. The pro-inflammatory and growth effects of E
2 are antagonized by co-administration of anti-inflammatory agents, including the synthetic glucocorticoid dexamethasone [
32‐
36] and the steroid hormone progesterone (P
4) [
37]. Evidence suggests that the estrogen-mediated infiltration of inflammatory cells into the endometrial stroma may lead to activation of extracellular proMMPs, which have been shown to colocalize with leukocytes in the mouse [
38] and human [
29] endometrium. Proteinases released by inflammatory cells, including tryptase, chymase, leukocyte elastase, and MMPs, have been shown to activate proMMPs in culture [
8,
9,
39,
40], while metabolites and oxidants generated by inflammatory cells may also regulate MMP activation [
41‐
43]. These studies and others suggest an important role for inflammatory mediators in the activation of proMMPs in the endometrial extracellular matrix. In turn, activated MMPs have been shown to modulate inflammatory processes, including the activity and mobilization of chemokines and cytokines [
44,
45] thereby regulating the further influx of inflammatory cells [
11‐
16]. Therefore, MMPs and inflammatory mediators are components of a complex network governing uterine growth and remodeling, as well as other normal and disease processes.
In the current study, we have developed a novel method to quantitatively analyze transmission electron micrographs to assess estrogen-induced endometrial collagen remodeling in the OVX immature rat model along a time course that includes both the hypertrophic and hyperplastic phases of uterine growth. Using matrix collagen density as a quantitative and reliable endpoint, we specifically determined that estrogen-induced stromal matrix remodeling is MMP-dependent. We further characterized hormone-regulated expression of MMP-3 (stromelysin-1), MMP-7 (matrilysin), MMP-2 (gelatinase A), and MMP-9 (gelatinase B), four metalloproteinases believed to function prominently in various reproductive processes [
6,
38], as well as the inflammatory mediators TNF-α and MCP-1 along an extended post E
2-treatment time course to assess changes in tissue protease and inflammatory potential through both the early hypertrophic and later hyperplastic growth phases. In addition, we have investigated the effect of the anti-inflammatory synthetic glucocorticoid dexamethasone on E
2-induced stromal matrix remodeling and MMP and inflammatory mediator expression to better understand the role of inflammation in the tissue pathway linking estrogen receptor activation to changes in tissue ultrastructure necessary for uterine growth and remodeling.
Methods
Animal treatments and tissue preparations
All animal experiments were performed in compliance with NIH standards as specified in the
Guide for Care and Use of Laboratory Animals (copyright 1996, National Academy of Sciences). Female Sprague-Dawley rats (Ace Animal Corp., Boyertown, PA) were ovariectomized at 21 days of age and allowed to recover for 3-5 days. For E
2 time course experiments, animals were administered a single physiological dose of E
2 (40 μg/kg in a 0.9% NaCl, 0.4% EtOH vehicle) by intraperitoneal (i.p.) injection at the indicated time intervals prior to tissue collection at necropsy. This in vivo dose level of E
2 has been shown to induce changes in uterine wet weight, tissue architecture, and gene expression characteristic of estrogen receptor activation [
1,
46‐
48]. For all other experiments, animals were i.p. administered a single 40 μg/kg bolus of E
2 4 h prior to tissue harvest, while control animals received vehicle only in all studies. Batimastat was administered i.p. at a dose level (BB-9; 40 mg/kg in a 1× PBS, 0.1% Tween-20 vehicle) shown to be effective at inhibiting MMPs in vivo [
49] 4 h prior to E
2 or saline control. Dexamethasone was administered i.p. at a dose level of 600 μg/kg in a 0.9% NaCl, 10% EtOH vehicle that has been previously used in rodent studies to assess in vivo steroid specific responses in the uterus [
46‐
48]. Animals were treated twice with dexamethasone (600 μg/kg) at 20 and 4 h before E
2 or saline control administration to elicit both early and late gene effects of the activated glucocorticoid receptor (GR) that may affect uterine responsiveness to E
2 via genomic or nongenomic mechanisms. Batimastat was generously provided by GlaxoSmithKline, while other treatment compounds were obtained from Sigma-Aldrich (USA). Dose levels of all compounds were administered at injection volumes of 1 ml/100 g body weight. Inhibitor compounds were administered i.p., with control animals receiving the corresponding vehicle solution i.p. at the same treatment times as inhibitor injections.
For uterine tissue collection at necropsy, animals were euthanized by decapitation on the same day in accordance with the specific treatment time interval and the tissue was stripped of fat and mesentery, weighed, and the horns separated. One horn from each animal was flash frozen in liquid N2 and stored at -80°C for later RNA purification. For protein analyses, two isolated horns from different animals were pooled and fresh-processed for total protein extraction. For transmission electron microscopy analyses, tissues were immediately fixed in 2.5% glutaraldehyde in 0.1 M phosphate buffer.
Transmission electron microscopy (TEM)
Individual uteri were fixed in 2.5% glutaraldehyde in 0.1 M phosphate buffer, pH 7.4, and 2 mm-cross sections of each uterine horn were cut and incubated overnight at 4°C in 2.5% glutaraldehyde in 0.1 M phosphate buffer. Glutaraldehyde buffer was removed, and the samples were rinsed three times (15' each) in 0.1 M phosphate buffer. Samples were incubated in 1% osmium in 0.1 M phosphate buffer, rinsed, and dehydrated in a series of ethanol washes. Samples were incubated twice for 5' in propylene oxide and then transferred to a rotor for 1 h at room temperature in a 1:1 mixture of propylene oxide and epon [47% Embed 812, 31% DDSA (dodenyl succinic anhydride), 19% NMA (nadic methyl anhydride), 3% BDMA (benzyldimethylamine)], followed by an overnight incubation in 1:2 propylene oxide-epon, and finally 100% epon for 2-3 hrs. Individual uterine samples were embedded in 100% epon in silicon flat embedding molds, and capsules were polymerized in a 60°C oven for over 48 hrs. Ultrathin transverse sections (70 nm) were prepared using a diamond knife (Diatome) on a MT 6000-XL ultramicrotome, captured on 300-mesh copper grids, and stained with 2% uranyl acetate. All reagents and materials were obtained from Electron Microscopy Sciences (Hatfield, PA).
A minimum of 3 blocks with tissue from different areas along the length of each uterine horn were chosen at random from a total pool of 5-6 blocks and were used to prepare ultrathin sections to provide a comprehensive survey of whole organ ultrastructure representatively. A minimum of 6 transverse sections were prepared per tissue block and viewed using a Hitachi 7600 microscope fitted with an AMT digital camera. Visual analyses and micrograph surveys were restricted to the endometrial stromal compartment at a minimum of 3 random sites within the endometrium of each section so as to avoid bias. At each random site within the endometrial stromal matrix compartment, 3-6 digital photomicrographs at 12,000× magnification were taken. For quantitative analysis of collagen density, the full area of photomicrographs was digitally fitted with an 8 × 8 square grid containing a total of 64 cross-points, and each cross-point was evaluated to determine the structure that it overlaid. Image J software (NIH) was used to quantify the total number of crossbars landing on collagen fibrils (positive hits), and the values were expressed as a ratio to the total number of crossbars to determine a percent matrix density volume for each photomicrograph. Stromal areas with a high percentage of collagen fibrils cut in transverse were selected to best ensure that quantitative crossbar analysis was not significantly affected by dramatic differences in collagen fiber orientation. A total of 36-75 micrographs were analyzed per time point or treatment condition. Values for all micrographs for each animal were averaged, and statistical differences between averages of each time point or treatment group were analyzed using one-way ANOVA. Error bars on graphs represent standard error, and a standard P value of ≤ 0.05 was considered significant.
RNA isolation and quantitative real-time PCR
For TaqMan® quantitative real-time PCR analyses of mRNA levels, total RNA was prepared individually from the uterus of each rat. Each uterus was disrupted for 2-3 minutes in 0.75 mL QIAzol lysis reagent (Qiagen, Valencia, CA) using a 5 mm bead in a TissueLyser (Qiagen-Retsch). The homogenate was subjected to a 0.2 mL chloroform extraction and centrifugation at 4°C for 15 min at 20,000 × g, and 0.35 mL aqueous phase was collected for RNA purification using an RNeasy kit (Qiagen) according to the manufacturer's protocol. Residual genomic DNA was removed by on-column RNase-free DNase treatment during RNA purification, and the RNA concentration was adjusted to 0.05 mg/mL. Messenger RNA levels were analyzed using quantitative real-time PCR on an ABI Prism 7900 Sequence Detection System and TaqMan® gene expression assays (Applied Biosystems, Foster City, CA) for MMP-2 (Rn01538174_m1), MMP-3 (Rn00591740_m1), MMP-7 (Rn00563467_m1), MMP-9 (Rn00579162_m1), TNF-α (Rn01753871_m1), MCP-1 (Rn01456716_g1) and eukaryotic 18s rRNA (Hs99999901_s1). To compare mRNA expression levels among samples, mRNA for each gene of interest was normalized to the expression of a housekeeping gene, 18s rRNA, and differences in gene expression were calculated as a fold change to the mean of the vehicle group.
Detailed methods for protein extraction, SDS-PAGE, and immunoblotting have been described previously [
46]. In brief, 2-3 uterine tissues were pooled and homogenized in 50 mM Tris pH 7.5, 150 mM NaCl, 1 mM CaCl
2, 0.05% Brij 35, 1 mM PMSF, and 10 μg/ml leupeptin using a polytron homogenizer (Brinkman Instruments, Westbury, NY). Homogenates were centrifuged at 15,000 × g for 30 minutes at 4°C and the supernatants removed and protein quantified. Extracted total protein (20 μg) samples were loaded into the wells of a pre-cast gel (4-20% Tris-HCl; BioRad, Hercules, CA) and resolved by electrophoresis in a Mini-Protein II module (BioRad) according to the manufacturer's protocol. Resolved proteins were transferred to an Immobilon-P membrane (Millipore, Billerica, MA), which was blocked (5% BSA in 1× Tris-buffered saline + 0.1% Tween-20 + 2% normal goat serum) and then incubated in a 1:2000 dilution of affinity-purified polyclonal rabbit anti-MMP-2 IgG (Novus Biologicals, Littleton, CO; #NB200-193), anti-human MMP-7 IgG (Lifespan Bioscience, Seattle, WA; #LS-B1237), anti-MMP-3 IgG (Abcam, Cambridge, MA; #ab53015), anti-MMP-9 IgG (Abcam #ab7299), anti-TNF-α IgG (Cedarlane, Burlington, NC; #CL9572AP), anti-MCP-1 IgG (Millipore, Billerica, MA; #AB1834P) or monoclonal mouse anti-β actin IgG (Abcam #ab6276) overnight in blocking solution at 4°C with gentle agitation. Membranes were washed and incubated in a 1:10,000 dilution of secondary HRP-conjugated goat α-rabbit IgG or α-mouse IgG (Jackson Immunologicals, West Grove, PA) in blocking solution and then washed and transferred to a solution of SuperSignal Substrate reagents (Pierce/Thermo Scientific, Rockford, IL) for enhanced chemiluminescent signal detection, imaged, and analyzed using the FluorChem SA system (Alpha Innotech, San Leandro, CA) and AlphaEase FC software. Band intensities were measured as integrated density volumes (IDV) and expressed as % of control lane values for comparisons between treatment groups. Specificity of bands identified using primary antibodies was confirmed by exposing sample blots to either 2% normal goat serum alone, secondary goat HRP-conjugated antibody in the absence of primary antibody, or normal rabbit IgG (Jackson Immunologicals) only prior to secondary antibody incubation.
Gelatin zymography
Extracted total protein (20 μg) samples were mixed 1:1 with zymography sample buffer (Bio-Rad) and incubated at room temperature for 10 min. Samples were loaded into the wells of a pre-cast gel (10% gelatin, BioRad) and resolved by electrophoresis in a Mini-Protein II module (BioRad) under non-reducing conditions. Active human MMP-2 (Calbiochem/EMD Biosciences, Gibbstown, NJ; #PF023) was loaded in one lane of the gel as a control. After electrophoretic separation, the gel was washed twice for 30 minutes each in 1× renaturation buffer (Bio-Rad) at room temperature (RT) with gentle agitation and then incubated in 1× development buffer (Bio-Rad) for 30 minutes at RT. Fresh development buffer was added and the gel incubated at 37°C for 18-24 hours. Developed gels were stained with Coomassie Brilliant Blue R-250 (0.5% in 30% methanol, 10% acetic acid) for 30 minutes at RT. The gel was destained in 30% methanol, 10% acetic acid until clear bands representing proteolytic areas were visualized. Gels were visualized with the FluorChem SA system (Alpha Innotech) and analyzed using AlphaEase FC software. Band intensities were measured as integrated density volumes (IDV) and expressed as % of control lane values for comparisons between treatment groups.
Statistics
Data is represented as the average ± standard error of the mean (SE). Data were analyzed initially via one-way ANOVA followed by Tukey's Honestly Significant post-comparison test to assess differences between treatment groups. P values ≤ 0.05 were considered significantly different.
Discussion
An examination of tissue responses in the E2-treated OVX rat provides an opportunity to study selective effects of the hormone that can be defined in the temporal context of the hypertrophic and hyperplastic growth responses. In this model, E2 induces uterine growth that is characterized by changes in endometrial tissue architecture and gene expression. We used TEM analysis to more specifically assess changes in stromal collagen matrix quantitatively. A significant decrease in matrix density is observed as early as 45 minutes and reaches its lowest value, with an approximate 70% reduction of the original collagen network, at 4 hours post-hormone treatment, coinciding with the peak time point of uterine hypertrophy. This indicates that activation of matrix-processing enzymes is extremely rapid and implies regulation of inherent proteases through an estrogen receptor (ER)-mediated pathway.
Pretreatment with the broad-range MMP inhibitor batimastat effectively ablates E
2-induced changes in stromal ultrastructure supporting a critical role for these enzymes in tissue remodeling. These results differ from those by Kaitu'u and colleagues [
38], who reported that batimastat reduced MMP activity but did not alter endometrial breakdown or repair as assessed through light microscopy in a mouse decidualization model. The contrasting results may reflect a difference in matrix processing during E
2-specific induction of tissue growth in the immature uterus as compared to tissue regression following progesterone withdrawal in a decidualization model. In addition, analysis of fine matrix structure and density using quantitative TEM provides a more sensitive measure of changes in collagen organization indicative of tissue remodeling as compared to qualitative assessment of tissue appearance via light microscopy [
46].
While a number of MMPs are known to be expressed in cell-type as well as cycle-specific patterns in the uterus [
2,
5,
50], E
2-specific regulation in an OVX rodent model has been studied in vivo for only a few [
19]. Using this model we show that expression of MMP-3, MMP-7, and MMP-9 is upregulated at both RNA and protein level early within the E
2-induced growth response correlating with the time point where stromal collagen matrix density reaches its lowest point (4 h). While substrates for these enzymes include fibrillar, fibril-associated, and network forming collagens [
50], in vivo substrates are unknown. Therefore, although expression correlates with maximum matrix remodeling temporally, we cannot conclude that these particular proteases are directly causative. As an epithelial cell expressed protease, MMP-7 is likely not critically involved in stromal matrix processing directly, whereas MMP-3 and MMP-9 are known to be expressed by stromal cells or infiltrating leukocytes [
50]. Expression of MMP-2 increases most significantly later in the growth response (48 h) suggesting a complex level of regulation that may involve E
2-triggered production of signal compounds which, in turn, enhance MMP-2 production. Interestingly, tissue levels of MMP-3 protein also remain elevated late in the growth response (24-72 h), indicating extended in-tissue stabilization of this enzyme. The overall elevation of MMP-2 and MMP-3 content late, as compared to MMP-7 and MMP-9 protein which peak early in the growth response, implies alternate temporal functions for tissue MMPs that may include activation of matrix processing during both tissue growth and regression phases. While providing important insight into the E
2-triggered response at the whole tissue level, our analysis of changes in MMP mRNA and protein content does not define stromal-specific changes in enzyme content and activity. Future studies to analyze MMP activity via in situ zymography and mRNA expression using laser capture microdissection and subsequent gene profiling would provide a more precise window into the stromal tissue response.
While we observed transient upregulation of MMP-2 and MMP-9 mRNA and protein expression in response to E
2 in the OVX immature rat uterus, Zhang and colleagues [
19] did not observe an effect of E
2 on MMP-2 mRNA levels in the OVX adult mouse through 24 h post-hormone treatment, though they did observe a decrease in MMP-9 mRNA, but not protein activity, 4-8 h post-treatment. Since we observe peak changes in MMP-2 mRNA expression at 48 h after E
2 administration, it is unclear if a similar response would be evident in an OVX adult model as that time point was not assessed [
19]. In contrast, our analysis of MMP-9 mRNA protein via both Western blot and zymography clearly shows upregulation by 4 h post-E
2 administration. Data inconsistencies may reflect differences in rodent model (mouse vs. rat) or in tissue responsiveness between the immature (first time hormone exposure) as opposed to an adult (hormone-primed) uterus.
A variety of inflammatory mediators [
8,
9,
40,
52,
53] have been shown to regulate expression and activity of MMPs, and, E
2 is known to induce an acute inflammatory-like response within the uterus [
54]. We show that E
2 induces potent and rapid upregulation of TNF-α, with peak increases in mRNA and protein occurring at the time point where MMP expression and stromal matrix processing are upregulated (4 h). It is difficult to understand the exact role of TNF-α expression in E
2-induced uterine matrix remodeling or in MMP gene transcription as this bioactive compound is expressed in tissues and held in an inactive state through cell surface or extracellular matrix-binding interactions. Although ADAM-17 (TACE) is the best characterized TNF-α sheddase, MMP-2, MMP-3, MMP-7, and MMP-9 have been shown to release TNF-α from the cell surface in a similar mechanism [
13,
45]. Protease-mediated shedding of TNF-α may induce signal pathway activation that precedes an overall upregulation of mRNA and protein expression for this inflammatory mediator, thus analysis of expression alone is not a full indicator of TNF-α activity in vivo. A study of E
2-induced changes in activity of the TNF-α pathway, including activation of NF-κB, will be important for a better understanding of the role of this inflammatory agent as a critical regulator of MMP expression in the estrogen response pathway.
Expression of the β-chemokine MCP-1 (CCL2), a potent chemoattractant for macrophages, T cells, basophils, and mast cells is also potently upregulated at 4 h. Enhanced production of MCP-1 is known to occur during the proliferative phase of the cycling uterus and, expression is stimulated by TNF-α [
55‐
57]. The timing of highest tissue levels of MCP-1 at 4 h correlates with the peak in expression of TNF-α, suggesting that it may be a key signal component for inducing leukocyte influx by creating in-tissue gradients of chemokines, like MCP-1, which are triggered early during the hypertrophic phase of uterine growth. MCP-1 is also a substrate for a variety of MMPs, including MMP-3 [
14], and protease-mediated processing has been shown to inhibit bioactivity of β-chemokines, thereby regulating the timing and intensity of leukocyte influx once an inflammatory response is underway [
14,
44,
45,
58].
Correlations between E
2-regulated uterine growth, inflammation, and MMP expression and activation in the immature OVX uterus led us to investigate the effects of dexamethasone, the synthetic glucocorticoid and anti-inflammatory agent. We show that dexamethasone modifies stromal matrix collagen density and, pretreatment with dexamethasone effectively inhibits E
2-induced collagen remodeling. These data suggest that the regulation of collagen density is under the influence of activated glucocorticoid-receptor (GR) and, implicates the MMP system because dexamethasone is known to inhibit MMP expression [
59]. Dexamethasone effects on MMP expression however, are complex. Of the MMPs that are upregulated 4 h post-E
2 treatment, dexamethasone pretreatment inhibits both MMP-7 and MMP-9 mRNA and protein expression but does not inhibit the E
2-induced increase in MMP-3 mRNA or protein.
In addition to effects on the MMP system, dexamethasone is also known to block the pro-inflammatory effects of E
2 in the uterus [
34]. Glucocorticoids have been shown to suppress expression of chemoattractant chemokines, such as MCP-1 [
60], and possibly, through inhibition of MMP expression, may decrease sheddase activity and ECM processing important for liberation of inherent bioactive agents, including TNF-α, that activate signal pathways to enhance inflammatory processes. Our data clearly show that dexamethasone has no effect on TNF-α or MCP-1 mRNA or protein expression. However, dexamethasone pretreatment potently inhibits E
2-induced upregulation of MCP-1 mRNA and protein expression and decreases levels of TNF-α protein observed at 4 h post-treatment, thereby supporting its effects as a suppressor of both inflammatory signaling as well as matrix remodeling.
Collectively our data show that E
2-induced endometrial matrix remodeling, production of MMPs, and inflammatory mediators correlate temporally with the early phase of uterine growth which is hypertrophic and inflammatory-like in nature. Dexamethasone as an anti-inflammatory agent inhibits these processes however, because activated GR can directly modify expression of MMPs and inflammatory agents such as MCP-1, further investigation is necessary to determine the role of inflammation itself on matrix remodeling. One model of the E
2 response would be that hormone-triggered rapid activation of inherent proteases leads to matrix processing that liberates bound inflammatory mediators like TNF-α, thereby producing inflammation and leukocyte influx. Alternatively, rapid E
2-induced inflammation triggered by ER signal pathways and production of paracrine regulators like prostaglandins would drive influx of leukocytes which, in turn, activate inherent proteases to effect changes in stromal matrix architecture necessary for tissue expansion as growth proceeds. Evidence indicates that infiltrating leukocytes contribute to tissue protease density [
8,
9,
61‐
63] and neutrophilic depletion clearly inhibits endometrial repair in the mouse [
25]. Further studies must focus upon the timing of E
2-induced changes in expression and activity of inflammatory mediators and leukocyte influx in relation to matrix processing to more clearly delineate the E
2 response pathway. In addition, the use of other kinds of anti-inflammatory agents such as COX inhibitors, that would specifically inhibit production of prostaglandins, will also provide a better understanding of the role of inflammatory reactions per se on matrix processing.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
LAR participated in the design of the study, in directing and completing all experimental analyses and in writing the manuscript. BJP participated in MMP mRNA analyses, in TEM tissue preparations and in writing the manuscript. SPT contributed significantly to generation of TEM data, mRNA analyses, and figure preparations. TDC developed the TEM quantitative method and participated in cytokine analyses to assess temporal changes in response to E2. BAO completed initial analyses of MMP expression and batimastat inhibitor studies. JAC contributed to quantitative TEM analysis for dexamethasone studies. ARS and OV participated in animal treatments and tissue preparations for MMP expression analysis. RMG participated in planning experiments and contributed to drafting of the manuscript. All authors have read and approved the final manuscript.