Effect of Thrombin on Human Amnion Mesenchymal Cells, Mouse Fetal Membranes, and Preterm Birth*

Background: Bleeding during pregnancy is a risk factor for premature rupture of the fetal membranes. Results: Thrombin causes preterm birth in mice and activates PAR-1 and TLR4 to increase MMPs and COX-2. Conclusion: Thrombin acts through multiple mechanisms to increase MMPs and PGE2 in amnion. Significance: Thrombin plays a pivotal role in the pathogenesis of preterm labor and rupture of the membranes. Here, we investigated the effects of thrombin on matrix metalloproteinases (MMPs) and prostaglandin (PG) synthesis in fetal membranes. Thrombin activity was increased in human amnion from preterm deliveries. Treatment of mesenchymal, but not epithelial, cells with thrombin resulted in increased MMP-1 and MMP-9 mRNA and enzymatic activity. Thrombin also increased COX2 mRNA and PGE2 in these cells. Protease-activated receptor-1 (PAR-1) was localized to amnion mesenchymal and decidual cells. PAR-1-specific inhibitors and activating peptides indicated that thrombin-induced up-regulation of MMP-9 was mediated via PAR-1. In contrast, thrombin-induced up-regulation of MMP-1 and COX-2 was mediated through Toll-like receptor-4, possibly through thrombin-induced release of soluble fetal fibronectin. In vivo, thrombin-injected pregnant mice delivered preterm. Mmp8, Mmp9, and Mmp13, and PGE2 content was increased significantly in fetal membranes from thrombin-injected animals. These results indicate that thrombin acts through multiple mechanisms to activate MMPs and PGE2 synthesis in amnion.

Amnion and chorion, a contiguous structure of cells and extracellular matrix, comprise the fetal membranes. Preterm premature rupture of membranes (PPROM) 2 is associated with 30 -40% of preterm deliveries and occurs in ϳ1-3% of all pregnancies (1). Currently, there is no effective prevention or treatment for PROM and preterm labor.
The process of membrane rupture targets the predominant matrix component, collagen, and may involve local (e.g. intraamniotic infection) or remote factors (e.g. procoagulants). Clin-ical evidence for a role of thrombin in preterm birth includes the following: (i) thrombin-antithrombin complexes, markers of in vivo generation of thrombin, are increased in plasma (2) and amniotic fluid (3) from women in preterm labor or preterm PROM; (ii) increased plasma thrombin-antithrombin complexes in the second trimester are associated with subsequent preterm PROM (4); (iii) retroplacental hematomas in the first trimester are associated with adverse pregnancy outcomes, including preterm birth (5); and (iv) vaginal bleeding in the first or second trimester is associated with preterm birth (6,7). Several studies implicate decidual cells as the primary source and target of thrombin (8). Thrombin stimulates the production of matrix metalloproteinase-1 (MMP-1) (9) in endometrial stromal cells in culture, suggesting that matrix degradation occurs through collagenase activity. Abruption-induced thrombin generation has been associated with fetal membrane weakening and PPROM (4,10), and treatment of intact amnion explants with thrombin results in increased levels of MMP-9 and mechanical weakening (11). The precise mechanisms by which thrombin leads to preterm labor are not known.
The primary load-bearing structure of the fetal membranes is the amnion which comprises a single layer of avascular epithelial cells and an underlying layer of mesenchymal cells (12). Mesenchymal cells are the primary source of collagen and matrix support of the amnion. Interstitial collagens (types I, III, and V) maintain the mechanical integrity of the amnion. Fetal membrane rupture is preceded by degradation of collagen that is mediated primarily by MMPs in amnion. Interstitial collagenase, MMP-1, cleaves the triple helix of fibrillar collagen, which is then further degraded by the gelatinases, MMP-2 and MMP-9. MMP-1 is increased in amniotic fluid obtained from women with PPROM (13). Moreover, MMP-9 is increased in amniotic membranes obtained from women in normal labor (14) and in those from women with PPROM (15,16).
Prostaglandins (PGs) play pivotal roles in human parturition by stimulating cervical ripening, myometrial contraction, and fetal membrane rupture (17). In human pregnancy, the principle source of PGs is the amnion (18). In this study, we investigated the effects of thrombin on MMPs and PGE 2 synthesis in amnion cells in vitro and in vivo. Collectively, the data reported herein suggest novel mechanisms of action and a pivotal role of thrombin in the pathogenesis of preterm labor and PPROM.
Thrombin Activity Assay in Human Amnion-Human amnion tissues were obtained from pregnancies delivered preterm (n ϭ 8) or term (n ϭ 6). The preterm group included both PPROM and non-PPROM, and delivery mode was either vaginal or cesarean section. Term normal amnion was derived from spontaneous vaginal delivery. Tissues were rinsed with saline and, after thorough removal of blood, frozen in liquid nitrogen, stored at Ϫ80°C. Pulverized human amnion (ϳ50 mg) was homogenized in 500 l of PBS. Homogenates were centrifuged at 3000 ϫ g for 5 min at 4°C. Supernatant was collected and assayed using Sensolyte 520 Thrombin Activity Assay kit according to the manufacturer's instruction. The proteolytic activity of thrombin was quantified using fluorescence resonance energy transfer (FRET) peptide.
Isolation and Culture of Amnion Epithelial and Mesenchymal Cells-Preparation and isolation of amnion epithelial and mesenchymal cells were performed as described previously (19). Briefly, amnion tissue was separated by blunt dissection. The amnion tissue was minced, and cells were dispersed by enzymatic digestion. Isolated amnion cells were suspended in DMEM/F12 that contained fetal bovine serum (10%, v/v) and antibiotic-antimycotic solution (1%, v/v). Cells were plated in plastic culture dishes, maintained at 37°C in a humidified atmosphere of 5% CO 2 in air, and allowed to replicate in monolayer to confluence.
Isolation of Decidual Cells-Decidual cells were isolated from human decidual tissue by limited enzymatic digestion and separation on Percoll gradients according to previously described methods (20). Cells were plated in plastic plates and allowed to replicate to confluence in primary monolayer culture.
Quantitative Real-time PCR-Quantitative RT-PCR was used to determine the relative levels of gene expression as described previously (21). Primer sequences for amplifications are shown in Table 1. Taqman probe with FAM dye label was used for COX2, and SYBR Green was used for other genes. Gene expression was normalized to that of the housekeeping gene GAPDH for human amnion cells and cyclophilin A for mouse tissue.
MMP-1 Activity Assay-MMP-1 activity of the conditioned media was assayed by The Sensolyte Plus 520 MMP-1 Assay kit according to the manufacturer's instruction. Proteolytic activity of MMP-1 was quantified using FRET peptide.
Gelatin Zymography-Gelatin zymography was performed as described (22). Briefly, conditioned medium was concentrated 10-fold with Amicon Ultra-4 Centrifugal Filter Units (Millipore) at 3200 ϫ g for 1 h at 4°C. Mouse fetal membranes were homogenized with buffer (10 mM Tris, 150 mM NaCl, 10 mM CaCl 2 , and 0.1% Triton X-100, pH 7.4), centrifuged at 16,000 ϫ g for 15 min at 4°C, and the supernatant was used for assay. Concentrated media (4 g) or homogenized mouse fetal membranes (12 g) were applied to 10% gelatin polyacrylamide minigels in standard sodium dodecyl sulfate loading buffer containing 0.1% sodium dodecyl sulfate. After electrophoresis, gels were soaked in renaturing buffer (2.7% Triton X-100 in distilled water) in a shaker for 30 min with one change after 30 min to remove sodium dodecyl sulfate. Gels were soaked in assay buffer (50 mM Tris, 200 mM NaCl, 10 mM CaCl 2 , 0.05% Brij 35, pH 7.5) for 16 h at 34°C and stained with Coomassie Brilliant Blue R-250 in 50% methanol and 10% acetic acid followed by destaining with 25% methanol and 7% acetic acid. Areas of lysis were analyzed using a Fuji LAS 3000 image analysis system (Fujifilm Life Science, Tokyo, Japan). PGE 2 ELISA-PGE 2 concentration in the condition media was assayed by Parameter PGE2 immunoassay (KGE004B; R&D Systems) according to the manufacturer's instructions. Immunohistochemistry-Immunohistochemistry was performed as described previously (22). Formalin-fixed, paraffinembedded human fetal membranes from term noncomplicated deliveries were sectioned at 5 m. Slides were deparaffinized in xylene and rehydrated in graded alcohols to distilled water. Endogenous peroxidase activity was quenched for 10 min at room temperature, using 0.3% H 2 O 2 with 0.1% sodium azide. Slides were subjected to steam heat epitope retrieval in EDTA buffer (1 mM, pH 8.0) for 30 min. After rinsing in PBS, slides were incubated in primary antibody (WEDE15) for 30 min at 25°C. Negative control specimens were processed simultaneously in an identical manner, with the exception that control mouse IgG was used in place of primary antibody. After another rinse in PBS, slides were incubated with appropriate horseradish peroxidase-conjugated polymer for 30 min at 25°C. Finally, the slides were immersed for 5 min in 25°C diaminobenzidine, enhanced with 0.5% copper sulfate in PBS for 5 min at 25°C, counterstained in hematoxylin, dehydrated in graded alcohols, cleared in xylene, and covered with a coverslip.
Immunoblotting-Protein extraction from cells with urea buffer was performed as described previously (21). Briefly, after washing with PBS, cells were extracted overnight at 4°C with 6 M urea buffer (6 M urea, 16 mM potassium phosphate, pH 7.8, and 0.12 M NaCl containing a protease inhibitor mixture), centrifuged at 10,000 ϫ g for 30 min, and supernatants were used for immunoblot analysis. Conditioned media from treated amnion cells were filter-concentrated with Amicon Ultra-4 Centrifugal Filter Units at 3200 ϫ g for 50 min at 4°C. Concentrated samples were supplemented with a protease inhibitor mixture for immunoblotting. Total protein (30 g/lane) was applied to 6% polyacrylamide gels, separated by electrophoresis, and transferred to nitrocellulose membranes overnight at 4°C. Membranes were blocked with TBST containing 5% nonfat powdered milk for 1 h at 37°C. Membranes were incubated with primary antibodies of rabbit anti-human fibronectin polyclonal antibody (AB1945) or human collagen I ␣1 antibody (AF6220) overnight at 4°C. Thereafter, blots were incubated with a second antibody (goat anti-rabbit IgG-HRP conjugate or rabbit anti-sheep IgG-HRP conjugate, 1:10,000) at room temperature for 1 h. Signals were detected by chemiluminescence using a Amersham Biosciences ECL Plus Western blotting Detection kit.
Thrombin-induced Model of Preterm Labor-C57BL/6 female mice were crossed with male mice for 5 h (from 10:00 to 15:00), and pregnancy was determined by the presence of a vaginal plug. On 17 days postcoitum at 19:00, pregnant mice were anesthetized with tribromoethanol (250 g/g body weight, intraperitoneally). A ventral incision was made to visualize the uterus and thrombin (4 l, 1 unit/l), or PBS with 1% methylene blue was injected in the interface between fetal membranes and uterine lining of each fetus opposite the placental site (see Fig. 8A). Injection was limited to 6 pups. Thereafter, the maternal abdomen was closed with 5-0 silk. All procedures were approved by the Institutional Animal Care and Use Committee (IACUC) at the University of Texas Southwestern Medical Center.
Statistical Analysis-Values were expressed as means Ϯ S.D.
Data were analyzed by unpaired Student's t test in human amnion tissues. In PAR-1 or TLR4 blocking experiments, oneway analysis of variance followed by the Student-Newman-Keuls test was used. Mann-Whitney U test was used in analyzing data from the mice preterm delivery model. p values Ͻ 0.05 were regarded as statistically significant.

Thrombin Activity in Human Amnion from Pregnancies
Delivered Preterm and Term-To determine whether thrombin activity was increased at the site of membrane rupture, thrombin activity was analyzed in human amnion from pregnancies delivered preterm (33.0 Ϯ 1.3 weeks) or term (38.3 Ϯ 0.4 weeks) (Fig. 1). Thrombin activity was increased significantly in amniotic membranes from women delivered preterm regardless of sampling site. Within the group delivered preterm, differences in thrombin activity were not significant between PPROM and non-PPROM or between vaginal or cesarean delivery (data not shown).
Thrombin Dose-and Time-dependently Increased MMP Activity in Primary Amnion Mesenchymal but Not Epithelial Cells-To assess the effect of thrombin on MMP-1, MMP-2, or MMP-9, primary amnion cells were treated with thrombin (0 -2 units/ml) for 4 -48 h, and mRNA and activity levels were quantified using quantitative PCR, FRET, or quantitative zymography (Fig. 2). Whereas MMP1 mRNA levels were low and unchanged by thrombin treatment in epithelial cells, MMP1 mRNA increased dose-and time-dependently in mesenchymal cells (Fig. 2, Aa and Da). Although not regulated in epithelial cells, thrombin induced significant increases in MMP-1 activity in a dose-and time-dependent manner in mesenchymal cells (Fig. 2, C and E). Like MMP1, basal expression of MMP2 was low in epithelial compared with mesenchymal cells. Thrombin, however, did not regulate MMP-2 gene expression in either cell type (Fig. 2Ab). Thrombin treatment resulted in increased MMP9 mRNA in mesenchymal, but not epithelial, cells (Fig. 2, Ac and Db). In agreement with mRNA levels, gelatin zymography of the conditioned media revealed that proMMP-2 was not regulated by thrombin in either cell type (Fig. 2Ba). Interestingly, however, thrombin resulted in FIGURE 1. Thrombin activity in human amnion from pregnancies delivered preterm or term. Thrombin activity was quantified using the FRET system after complete removal of blood from human amnion. Preterm amnion (n ϭ 8) includes both PPROM (open symbols) and non-PPROM (filled symbols), and the delivery mode was either vaginal or cesarean section. Term normal amnion (n ϭ 6) was obtained from spontaneous vaginal delivery.
dose-and time-dependent increases in active MMP-2 in mesenchymal cells (Fig. 2, Ba and F). Incubation of conditioned media from mesenchymal cells with thrombin (4 units/ml ϫ 48 h) under cell-free conditions did not result in conversion of proMMP-2 to its active form, suggesting that the protease activity of thrombin was not directly involved in activation of MMP-2 (Fig. 2Bb). Although low and not changed in epithelial cells, proMMP9 was increased significantly in mesenchymal cells (Fig. 2, Ba and F). Thus, thrombin treatment results in increased expression of MMP-1 and MMP-9 mRNA and enzymatic activity in amnion mesenchymal, but not epithelial, cells. Further, although thrombin did not regulate MMP2 mRNA, treatment of mesenchymal cells with thrombin resulted in conversion of MMP-2 to its active form possibly through cleavage of the pro form by MMP-1 or MMP-9.
Thrombin Increases COX2 mRNA and PGE 2 in Primary Amnion Mesenchymal Cells-Prostaglandins are believed to be involved in PPROM, initiation of myometrial contractions of labor, and cervical ripening during parturition. Thus, we quan- tified the effect of thrombin on PG synthesis in primary amnion cells. In epithelial cells, COX2 mRNA was unchanged, whereas thrombin increased COX2 mRNA 10-fold in mesenchymal cells (Fig. 3, A and C). Although unchanged in the conditioned media of epithelial cells, PGE 2 was increased 10-fold in media from mesenchymal cells (Fig. 3, B and D).

Thrombin Receptor, PAR-1, Is Expressed in Human Amnion
Mesenchymal and Decidual Cells-Imunohistochemistry was used to localize expression of the major thrombin receptor, PAR-1, in human fetal membranes (Fig. 4A). PAR-1 was strongly expressed in amnion mesenchymal cells of both rupture and nonrupture sites. PAR-1 was not expressed in amnion  epithelial cells. PAR-1 was also expressed in chorion fibroblasts and interestingly, highly expressed in decidua (Fig. 4A, a and b). In primary culture, PAR-1 mRNA levels were increased 30-fold in mesenchymal cells compared with very low expression in epithelial cells (Fig. 4B), compatible with the expression pattern observed using immunohistochemistry. Because PAR-1 was highly expressed in decidua, we determined the effect of thrombin on primary decidual cells (Fig. 5). Thrombin treatment of human decidua cells in culture resulted in increased MMP1 and MMP9 mRNA levels (Fig. 5A, a and c). The magnitude of stimulation, however, was much less relative to that in amnion mesenchymal cells (Fig. 2). Similar to our findings in mesenchymal cells, proMMP-2 and MMP2 mRNA were not regulated by thrombin in decidual cells (Fig. 5, Ab and B). By gelatin zymography, however, active MMP-2 was not only more abundant in decidual cells but also increased significantly by thrombin treatment (Fig. 5B). ProMMP-9 increased dose-dependently whereas active MMP-9 was not regulated in decidual cells (Fig.  5B). Thrombin also increased MMP-1 activity in decidual cells (Fig. 5C), but the effect was modest compared with its dramatic effect on amnion mesenchymal cells. Together, these results suggest that thrombin regulates MMPs mRNA and enzymatic activity in PAR-1-expressing decidual and amnion mesenchymal cells, but the magnitude of stimulation is more dramatic in mesenchymal cells.

Up-regulation of MMP9 mRNA Is Mediated via PAR-1-To
clarify whether thrombin-induced up-regulation of MMPs and COX-2 in mesenchymal cells is mediated through PAR-1, two approaches were taken. First, we tested whether pretreatment with a PAR-1-specific inhibitor, SCH79797, blocked thrombininduced activation of MMP gene expression. Second, we determined whether a PAR-1-activating peptide (PAR-1 AP, TFLLRN) mimicked thrombin-induced effects on these cells. The PAR-1 inhibitor completely blocked thrombin-induced increases in MMP9 mRNA, and PAR-1 AP increased MMP9 mRNA levels significantly (Fig. 6, A and B). The thrombin protease inhibitor, hirudin, also partially blocked thrombin-induced up-regulation of MMP9 (Fig. 6C). Further, by gelatin zymography, thrombin-induced increases of proMMP-9 were partially blocked by the PAR-1 inhibitor (Fig. 6D), and PAR-1 AP increased proMMP9 (Fig. 6E). Zymography also indicated that increased activation of MMP-2 was not suppressed by the PAR-1 inhibitor (Fig. 6D) nor mediated by the PAR-1-activating peptide (Fig. 6E). These data indicate that activation of PAR-1 is important for thrombin-induced increases in MMP-9. In contrast, thrombin-induced increases in MMP1 and COX2 were not suppressed by the PAR-1 inhibitor nor increased by the PAR-1 AP (Fig. 7). Rather, COX2 mRNA was increased by the addition of SCH79797 (Fig. 7Aa). Inhibition of thrombin protease activity with hirudin amplified, rather than sup- pressed, the effect of thrombin on MMP1 and COX2 gene expression (Fig. 7Ac). We considered the possibility that thrombin may mediate COX2 or MMP1 gene expression through PAR-2-4. In addition to thrombin, MMPs may activate PAR-1. To determine whether MMP-1-induced activation of PAR-1 contributed to increased MMP-9 or MMP-1 mRNA in amnion mesenchymal cells, cells were treated with recombinant MMP-1 in the presence or absence of the PAR-1 inhibitor, SCH79797. MMP-1 increased MMP-9 mRNA 14-fold which was inhibited 35% by SCH79797. In contrast, MMP-1 had little effect on its own gene expression (from 1.0 Ϯ 0.06 to 2.84 Ϯ 0.2 relative units/GAPDH) which was not blocked by SCH79797. Together, these data suggest that up-regulation of MMP-9 was mediated by PAR-1 both directly through its own hirudin-sensitive protease activity and indirectly through MMP1. In contrast, thrombin-induced increases in MMP1 and COX2 were mediated by mechanisms other than its protease activity or activation of PARs.
Evidence That Up-regulation of MMP-1 and COX2 Is Mediated via Toll-like Receptor-4-Previously, we found that fetal fibronectin up-regulated MMP-1 and COX-2 mRNA and enzymatic activity in mesenchymal cells through extra domain A-mediated activation of TLR4 (23). To determine whether thrombin signals through TLR4, the effect of a TLR4 neutralizing antibody on thrombin-induced increases in MMP1 and COX2 mRNA was determined. Surprisingly, TLR4 neutralizing antibody completely blocked thrombin-induced increases in MMP1 and COX2 mRNA and partially blocked MMP9 mRNA (Fig. 8A, a-c). Furthermore, thrombin treatment (4 units/ml (36 nM) ϫ 48 h) dramatically increased soluble fibronectin in conditioned media from both amnion mesenchymal cells (TLR4-responsive) and epithelial cells which do not express functional TLR4 (Fig. 8B) (23). Further, in contrast with mesenchymal cells, thrombin-induced release of fibronectin was not accompanied by increased expression of COX2 and MMP1 in epithelial cells (data not shown). PAR-1 AP did not increase release of fibronectin (Fig. 7C). Thrombin treatment also resulted in increased degradation of collagen type I ␣1 (Fig. 8B). Specifically, thrombin purified from human plasma resulted in reduced amounts of collagen I ␣1 monomers and the appearance of a 30-kDa collagen degradation product in the media, an effect reversed by hirudin. Using purified collagen type I in vitro, thrombin resulted in 50% loss of collagen type I dimers and trimers, which was rescued by hirudin. Collagen type I ␣1 monomers were most susceptible to thrombin with 85% degradation partially rescued to 63% by hirudin. Thus, collagen degradation is mediated by thrombin protease activity, not PAR-1 nor TLR4-induced up-regulation of MMP-1. Together, these data suggest that thrombin (i) activates PAR-1 both directly and indirectly to increase MMP-9 in amnion mesenchymal cells, (ii) activates TLR4 to increase COX-2 and MMP-1 in these cells, (iii) acts to release free fetal fibronectin from both epithelial and mesenchymal cell matrix, and (iv) converts collagen type I to its degradative products through its protease activity.
To ensure that thrombin-induced activation of TLR4 was not due to contamination with either collagenases or endotoxin, the effects of recombinant thrombin (rThr) were compared with those of plasma thrombin (pThr). Thrombin activity assays revealed that plasma thrombin was 28% greater than that of rThr. rThr was much less active in terms of collagenase activity (Fig. 8D), and induction of MMP-1 gene expression (Fig. 9A) rThr was more potent that pThr in increasing MMP9 gene expression (Fig. 9B). Although a complete dose-response analysis of rThr was not completed primarily due to its costs, concentrations of 6 units/ml rThr (54 nM) did not increase MMP1 gene expression more than 4 units/ml (36 nM). To ensure that pThr effects on TLR4 were not due to endotoxin contamination, endotoxin was removed from plasma thrombin using a Detoxi-Gel Endotoxin Removing column (24,25). Endotoxin removal did not affect pThr-induced up-regulation of MMP1 or COX2 (Fig. 9, C and D). Taken together, therefore, pThrinduced increases in TLR4 targets were not mediated through LPS contamination; rather, rThr was less active than pThr in terms of collagenase activity, and activation of TLR4 was likely to due to changes in glycosylation. Silver-stained gels using 100 and 300 ng of purified pThr and rThr demonstrated remarkable changes in molecular mass due to changes in glycosylation (Fig.  9E).
Thrombin-induced Up-regulation of MMPs, PGE 2 Synthesis, and Preterm Labor in Mice-To determine the effects of thrombin in vivo, precisely timed pregnant mice were injected in the interface between fetal membranes and the uterine wall with either 4 units/fetus thrombin or PBS at 17 days postcoitum. In this model, thrombin activity increased 1.6-fold in thrombin-injected fetal membranes compared with PBS-injected controls (1.26 Ϯ 0.08 and 1.86 Ϯ 0.13 ng/g protein, respectively, n ϭ 3 in each group, p ϭ 0.013). This dose and increase in activity were comparable with that in human preterm relative to term amnion (Fig. 1) and were thereby considered physiologically relevant. All thrombin-injected mice delivered preterm (within 24 h after injection, average 18.8 Ϯ 1.5 h, Fig. 10B), and all pups either died or were cannibalized shortly after birth. All pups delivered, even in mice that were injected in only one horn. In contrast, PBS-injected mice delivered at term (59.0 Ϯ 5.8 h after injection, p Ͻ 0.01, Fig. 10B) and survived. To determine whether thrombin-induced preterm labor was accompanied by activation of MMPs and COX-2 in mouse fetal membranes, membranes were collected 14 h after control or thrombin injection. Consistent with previous reports (26), MMP1 mRNA levels were almost undetectable in murine membranes (data not shown). In mice, MMP-1 is functionally substituted by other collagenases (e.g. Mmp-8 and Mmp-13) (26 -29). Local injection of thrombin at the maternal-fetal interface resulted in increased mRNA levels of both collagenase-2 (Mmp8) and collagenase-3 (Mmp13) (Fig. 10C, a and d).
Mmp9, but not Mmp2, mRNA also increased significantly after thrombin injection (Fig. 10C, c and b), and proMMP-9 was increased in thrombin-injected fetal membranes (Fig. 10D). COX1 and COX2 mRNA tended to increase (Fig. 10E, a and b), whereas mRNA level of 15-hydroxyprostaglandin dehydrogenase, which inactivates prostaglandins, was not changed (Fig  10Ec). PGE 2 synthesis was significantly increased by thrombin (Fig. 10Ed). Thus, we confirmed that injection of thrombin at the maternal-fetal interface resulted in increased collagenase gene expression (Mmp8 and Mmp13), Mmp9 mRNA, and enzymatic activity, and PGE 2 synthesis in vivo.

DISCUSSION
In this investigation, we confirmed the effects of thrombin on MMP-1 in decidual cells and extended these findings to show that MMP-2 and MMP-9 were also activated. Relative to decidual cells, however, the effects of thrombin were more pronounced in mesenchymal cells of the amnion. We suggest, therefore, that thrombin-induced activation of MMPs in both cell types may synergistically contribute to degradation of collagen in the fetal membranes.
Thrombin activity was increased in amnion from pregnancies complicated by preterm delivery compared with term delivery. In normal uncomplicated pregnancy, amniotic fluid concentrations of thrombin-antithrombin III complexes are decreased in the mid-trimester than at term (3). Thus, the increased thrombin activity in amnions from women with preterm labor is likely the result of pathological change and suggests that thrombin may be involved in the pathophysiology of preterm labor.
Thrombin up-regulates MMPs in fetal membrane organ culture (11,30,31). In this study, we used primary amnion cells to clarify that mesenchymal, not epithelial, cells are involved in thrombin-induced up-regulation of collagenase, MMP-1, as well as the gelatinases MMP-2 and MMP-9. Thrombin-induced increases in MMP-1 and MMP-9 were associated with increased mRNA transcripts for MMP1 and MMP9, whereas MMP-2 was activated at the post-transcriptional level. Thrombin also increased COX2 mRNA as well as PGE 2 synthesis in amnion mesenchymal cells. Consistent with previous reports (32), amnion mesenchymal cells were the major source of PGE 2 in amnion. Thrombin further up-regulated PGE2 synthesis in amnion mesenchymal cells. These findings suggest that thrombin-mediated increases in PGE 2 synthesis may lead to uterine contractions and cervical ripening of preterm labor. Prostaglandin-inactivating enzymes in chorion or cervix (i.e. 15-hydroxyprostaglandin dehydrogenase) may prevent delivery of bioactive prostaglandins to these sites. Nonetheless, thrombin is a direct activator of uterine contractility (33), and our findings of thrombin-induced preterm birth in mice suggest that thrombin induces preterm birth either directly or indirectly through prostaglandin biosynthesis.
Use of a PAR-1-selective inhibitor and activating peptide indicated that thrombin-induced up-regulation of MMP-9 in mesenchymal, but not epithelial, cells was mediated by PAR-1. The results are consistent with our findings that PAR-1 receptors were localized to mesenchymal cells. Activation of PAR-1 was mediated both directly through its hirudin-sensitive protease activity and indirectly through MMP-1-induced activation of PAR-1. In contrast with MMP-9, thrombin-induced increases in MMP1 and COX2 in amnion mesenchymal cells were not mediated by activation of PAR-1. Our previous results indicated that the extra domain A of fetal fibronectin upregulated MMP-1 and COX-2 through activation of TLR4 in amnion mesenchymal cells (23). Hence, we suspected that thrombin also sent signals via TLR4. TLR4 neutralizing antibody completely blocked thrombin-induced increases in MMP1 and COX2 mRNA. Further, thrombin-induced increases in these TLR4 target genes were not mitigated by endotoxin removal, and endotoxin-free recombinant thrombin also increased MMP1 and COX2. Interestingly, we found that thrombin also stimulated release of fibronectin from its cellbound fibrillar form to its soluble form, which is known to activate TLR4 (23). Thrombin-induced release of fibronectin occurred even in epithelial cells with nonfunctioning TLR4 complexes, indicating that thrombin-induced activation of MMP1 was not required. Although multiple mechanisms may be involved, these findings suggest that solubilization of fetal fibronectin plays an important role in thrombin-induced activation of TLR4 and thereby up-regulation of MMP-1 and COX-2 in amnion mesenchymal cells. Our results indicate that glycosylated mature pThr is more active that nonglycoslyated rThr in terms of TLR4 activation and collagenase activity. Further, pThr is a more potent collagenase that rThr. The experiments, therefore, indicate that several mechanisms are involved in thrombin-induced effects on human mesenchymal cells, including activation of PAR-1 to increase MMP-9, release of fibronectin from epithelial and mesenchymal matrix, and TLR4-mediated activation of MMP1 and COX-2 in mesenchymal cells. The results are compatible with those of Puthiyachirakkal et al. in which thrombin directly weakened devitalized fetal membranes through its protease activity (34).
Thrombin caused preterm labor in pregnant mice, up-regulating MMP expression and PGE 2 synthesis in murine fetal membranes. Thus, thrombin-induced up-regulation of collagenases, gelatinase, and PGE 2 in vivo supports our data obtained in human amnion cells in vitro.
In summary, these studies indicate that thrombin, a multifunctional serine protease, plays a pathologic role in preterm labor and PPROM, particularly in amnion mesenchymal cells.
We suggest that development of drugs or strategies to block the actions of thrombin or inhibit serine (i.e. thrombin) and MMP protease activity in tissues at the maternal-fetal interface may contribute to successful prevention and treatment of preterm rupture of the fetal membranes and preterm birth.