Expression and function of matrix metalloproteinases and their inhibitors at the maternal-embryonic boundary during mouse embryo implantation

The process of mouse embryo implantation starts 5.0 days post coitum (p.c.), when regularly spaced blastocysts lodge in clefts in the convoluted epithelial surface of the antimesometrial side of the asymmetric uterine wall (reviewed by Cross et al., 1994). Expansion of the stroma closes down the uterine lumen (Weitlauf, 1994). The embryos adhere via the mural trophoblast cell surface, and then the uterine epithelium sloughs off, starting closest to the embryo and spreading to the uterine glands (El-Shershaby and Hinchliffe, 1975; Schlafke and Enders, 1975; Welsh and Enders, 1991a). Differentiation of the uterine mesenchymal cells closest to the implantation site during decidualization is accompanied by a transition of basic cellular characteristics from stromal to para-epithelial: Cells acquire extensive cell-cell contacts (including gap junctions), swell up to 10 times their stromal cell volume, become polyploid and lay down a basement membrane-like extracellular matrix (ECM) around each cell. Decidual cells are cohesive; each implantation site becomes a separable mass of decidual tissue without an epithelium-lined lumen, but with an embryo at the core. The extraembryonic tissues of the mouse blastocyst attach and ingress into the stromal/ decidual mass, dividing, expanding and differentiating. At 6.5 days p.c., local blood vessels lose patency, and blood drains into the lumen that surrounds the ectoplacental cone (El-Shershaby and Hinchliffe, 1975; Welsh and Enders, 1991b). At 7.5 days p.c., the integrity of the circulation is restored. Large blood sinuses form, and active angiogenesis is apparent at the polar (mesometrial) end of the embryo. The embryo grows, remodeling the surrounding yolk sac with a thick basement membrane (Reichert’s membrane). Gradually, the supporting decidual cells atrophy to become residual at 12 days p.c. (Abrahamsohn and Zorn, 1993).

During these migratory, invasive and remodeling reactions, there is clearly a requirement for ECM metabolism and modulation. We have studied the expression of several matrix metalloproteinases (MMPs) and their inhibitors during implantation to test the hypothesis that this class of enzyme is functionally involved. Previously, we showed that the expression of gelatinase B is upregulated in parallel with the differentiation of trophoblast in cultures of mouse blastocysts (Behrendtsen et al., 1992). We assayed the lytic properties of giant cells in culture and found that both the MMP inhibitor TIMP-1 and an anti-gelatinase B antibody inhibited the clearing of subjacent matrix by trophoblast cells (Behrendtsen et al., 1992). Gelatinase B has a wide substrate specificity (Birkedal-Hansen et al., 1993), and the murine enzyme has a high specific activity and has been correlated with the invasive behavior of a number of sarcoma, melanoma and carcinoma cell lines (Bernhard et al., 1990).

Matrix metalloproteinases are clearly implicated in invasive, erosive and remodeling reactions by three factors. First, TIMP-1 and TIMP-2 produce significant alteration in the invasive properties of various normal and pathologic cell types (Alexander and Werb, 1992; Davies et al., 1993; Khokha, 1994; Montgomery et al., 1994). Second, their expression correlates with the expression of erosive pathologies such as arthritic conditions (Singer et al., 1995). Third, their substrate specificities in vitro suggest that these enzymes can cleave many ECM target proteins (Birkedal-Hansen et al., 1993). MMPs and their inhibitors tend to be co-expressed, leading to the hypothesis that the net extracellular active proteinase concentration determines the invasive or lytic properties of a cell. However, the normal function of this class of enzymes in vivo remains undetermined. We have studied the expression and function of MMPs and TIMPs at the maternal-fetal interface during mouse embryo implantation.

Outbred CF-1 and CD-1 mice were obtained from Charles River (Wilmington, MA). UltraSpec RNA purification reagent (Biotecx, San Antonio, TX), Duralon membranes and QuikHyb (Stratagene, San Diego, CA) were used to prepare RNA blots. The MMP inhibitor, 3-(N-hydroxycarbamoyl)-2(R)-isobutylpropionyl-L-tryptophan methylamide (MPI; Grobelny et al., 1992) and an MMP inhibitor control (N-(tert-butyloxycarbonyl)-L-leucine-L-tryptophan methylamide; MIC) were gifts from Glycomed Corp., Alameda, CA (courtesy of R. Galardy). The derivation of the rabbit antibody to gelatinase B was described by Behrendtsen et al. (1992). IgG fractions of high-titer rabbit bleeds were purified by protein A-Sepharose chromatography. Rabbit antiserum to mouse placental lactogen-1 was kindly provided by Dr Frank Talamantes (Colosi et al., 1987).

Female mice were mated and checked for vaginal plugs in the morning (midnight = 0 days p.c.). They were killed by cervical dislocation, and uteri were removed at the time points indicated. The uterus was cut open, and deciduae were teased out. Subfractions of decidua, for evaluation of mRNA or protein content, were obtained by pinning the mesometrium downward, cutting out the top third of the exposed antimesometrial decidua and removing it to liquid nitrogen. The opposite, mesometrial third of the decidua was also saved. After all the decidual tissue was removed, the residue was called the uterine sheath. Deciduae were processed for morphologic evaluation or in situ hybridization by fixing in 4% paraformaldehyde for 4-16 hours and paraffin embedding.

Paraformaldehyde-fixed deciduae were embedded in paraffin, and 6-mm sections were rehydrated, incubated in 0.1 M glycine for 3 minutes and incubated in blocking solutions (5% nonfat milk powder,1mg/ml ovalbumin and 5% sheep serum in phosphate-buffered saline; PBS) for 30 minutes each. Blocking serum was removed, rabbit polyclonal antibody to gelatinase B or mouse placental lactogen-1 (diluted to 50 mg/ml in 0.1% bovine serum albumin) was added, and sections were incubated for 1 hour at ambient temperature. Slides were washed in PBS and incubated for 30 minutes at ambient temperature in 5% sheep serum. Sections were incubated with biotinylated sheep anti-rabbit secondary antibody (Sigma #B9140), diluted 1:100 in 0.1% bovine serum albumin for 1 hour at ambient temperature, washed in PBS and then incubated with streptavidin/alkaline phosphatase (Vector Labs SA-5100), diluted 1:100, for 1 hour at ambient temperature. The alkaline phosphatase was developed in substrate from Vector Labs kit SK-5100. Controls used in place of the specific primary antiserum were preimmune serum or normal rabbit serum (Sigma Chemical Co., St. Louis, MO). Slides were counterstained with methyl green.

Paraffin sections (8 mm) were hybridized with ³⁵S cRNA probes as described by Frohman et al. (1990), with a few modifications. The Probe On capillary gap system with Probe On Plus slides was used for hybridization incubations. To decrease solution viscosity, we included 0.1% Brij in all aqueous solutions and 2% chondroitin sulfate and 1% dextran sulfate in the hybridization solution. Hybridization was performed at 55°C for all probes except TIMP-1 (50°C). Near-adjacent sections were probed with at least three different anti-sense cRNAs. Unique expression of each probe served as a positive control for nonspecific signal. In some cases, near-adjacent sections were probed with sense cRNAs confirming the positive signal. A non-specific signal, with sense and anti-sense probes, was frequently seen to be associated with erythrocytes, which are ineffectively removed from implantation sites, even with perfusion fixation. Therefore, all silver grains were examined at high magnification to confirm their association with cells other than erythrocytes. Each slide had at least two adjacent sections, and duplicate slides were used for each probe. Slides were coated with nuclear track emulsion (NTB2 Kodak or K.5D Ilford), and duplicate slides were exposed for two time points (4 days to 3 weeks). After development, slides were counterstained with hematoxylin and eosin or the intercalating DNA dye, Hoechst 33258 (Grevin et al., 1993). The following mouse anti-sense probes were used: gelatinase B (pSP65 92b; Reponen et al., 1994), spanning nt 1917-2240, transcribed with SP6 polymerase; gelatinase A (pSP65 72; Reponen et al., 1992), spanning nt 604-1165, transcribed with SP6 polymerase; stromelysin-1 (pTRM11; Hammani et al., 1992), spanning nt 3115-4051, transcribed with T7 polymerase; stromelysin-3 (H. Luk and Z. Werb, unpublished data), a 383-bp probe generated by polymerase chain reaction (PCR) of uterus RNA with redundant oligos spanning nt 1094-1476, transcribed with T7 polymerase; TIMP-1 (pTIMP-8; Gewert et al., 1987), encoding the entire cDNA, transcribed with T3 polymerase; TIMP-2 (Alexander and Werb, 1992), a 360-bp probe spanning amino acid sequence 73-194, generated by PCR of mouse 3T3 fibroblast RNA, transcribed with T3 polymerase; TIMP-3 (Leco et al., 1994), a 2.4-kb cDNA including 160 bp of the 5′ noncoding region, together with the complete coding region and 3′ noncoding region, transcribed with T7 and hydrolyzed to 300 nt.

DNA in nuclei of cells in 8-mm paraffin sections was labeled at free 3′OH termini by using digoxigenin-labeled dUTP and terminal deoxynucleotidyl transferase (TUNEL) according to the manufacturer’s instructions (Oncor, Gaithersburg, MD). Free 3′OH termini are present at strand breaks produced by endonuclease cleavage during apoptosis. Sections were incubated in fluorescein isothiocyanate-labeled anti-digoxigenin antibody and visualized using fluorescence microscopy. Nuclei were counterstained with propidium iodide. The number of positive nuclei per unit area was determined on a Macintosh 7100/80 AV computer using the public domain NIH Image program (U.S. National Institutes of Health; available from the Internet by anonymous FTP from zippy.nimb.nih.gov). Unit area was designated as number of square pixels, and images were scanned at 300 pixels/inch.

Tissue was homogenized in UltraSpec reagent according to the manufacturer’s instructions with an Omnigene Polytron; RNA was separated on 1.2% gels and transferred to Duralon membranes. RNA blots were hybridized by standard techniques.

Lysates of decidual fractions were prepared by homogenizing tissue pieces in RIPA buffer (1:4 (wt:vol) in 150 mM NaCl, 1.0% NP40, 0.5% deoxycholate, 0.1% sodium dodecyl sulfate (SDS), 50 mM Tris-HCl buffer, pH 8.0) at 4°C. Lysates were centrifuged at 14,000 g for 15 minutes, and the insoluble, ECM-enriched fractions were washed in RIPA buffer, aliquoted, and stored at −80°C or immediately processed. No difference between fresh samples and once-thawed samples was seen. Pellets were resuspended in substrate gel sample buffer for 10 minutes at 37°C. Zymography was described by Behrendtsen et al. (1992), and reverse zymography was described by Staskus et al. (1991). Briefly, gelatin (a general proteinase substrate) was included in the SDS-polyacrylamide gel mixture at 1 mg/ml. Samples were not denatured with heat or reducing agents, but were mixed with sample buffer containing SDS and loaded on 10% or 15% separating gels. After separation, proteins were renatured by incubating the gel in 2.5% Triton X-100 and then incubating in Tris-HCl buffer, pH 8.0, containing Ca²⁺ at 37°C to allow gelatinolysis for 16-48 hours. Gels were stained with Coomassie Blue. Clearing of back-ground gelatin by proteinases was revealed as clear bands. For reverse zymography, a crude mixture of gelatinases from medium conditioned by Rous sarcoma virus-transformed chick embryo fibroblasts at an empirically derived concentration was copolymerized with the gel mixture. Inhibitor activities were revealed as dark bands against the partially cleared background. SDS-polyacrylamide gel electrophoresis of the same extracts revealed which bands were protein bands with no inhibitor activity.

Pseudopregnant females were generated by mating with vasectomized males and were induced to decidualize by the transfer of 30 ml of peanut oil with a 25-gauge needle on a 1-ml syringe into the uterus of anesthetized females at 4.25 days p.c.

The control compound (MIC) or the proteinase inhibitor (MPI) was injected intraperitoneally at 100 mg/kg body weight as a 20 mg/ml slurry in 4% carboxymethyl cellulose in 0.9% saline. Injections began on the evening of day 3 p.c. (late day 3.5) and were administered every 12 hours thereafter until the mice were killed at 6.5 days p.c.

The derivation of transgenic mice expressing human TIMP-1, driven by the β-actin promoter, has been described by C.M. Alexander, E.W. Howard, M.J. Bissell and Z. Werb (unpublished data). Homozygous mice on the CD-1 background were used in the present study. Timed matings, decidual harvest and tissue lysate methods for control CD-1 mice were as described earlier. The β-actin promoter directed the ubiquitous expression of TIMP-1 in all organs, and serum levels of about 50 ng/ml. No gross changes were induced by expression of this transgene, and fertility of these mice was comparable to that of controls.

Gelatinase B synthesis is induced in cultures of differentiating trophoblast cells (Behrendtsen et al., 1992). To show that this enzyme is also expressed in vivo during implantation, we localized mRNA and protein in sections of decidua by using in situ hybridization and immunocytochemistry with an immunopurified antibody, respectively (Fig. 1). This antigelatinase B antibody was found to be function-perturbing in our previous study (Behrendtsen et al., 1992), specifically inhibiting the lysis of a subjacent ECM by cultured trophoblast cells. Primary giant cells around the implantation site at 5.5 days p.c. (Fig. 1A) stained positively with anti-gelatinase B antiserum. At this stage, trophoblasts emigrate out of the primitive trophectoderm surrounding the embryo, displace uterine epithelial cells, cross the underlying basement membrane, and differentiate to form a shell of primary giant cells embedded in maternal decidua. Accumulation of gelatinase B protein was paralleled by synthesis of gelatinase B mRNA in these cells; in situ hybridization of sections from 6.5-day-p.c. implantation sites (Fig. 1B) showed that every trophoblast giant cell, identified by morphologic criteria, was positive for gelatinase B mRNA expression. Immunostaining of 7.5- and 8.5-day-p.c. deciduae (Fig. 1C,E) showed that, as the ectoplacental cone grew, the population of trophoblast giant cells synthesizing gelatinase B increased, all around the margins of the expanding ectoplacental cone. Placental lactogen-1, a marker of trophoblast differentiation (Cross et al., 1994, 1995), was also present in the giant cells (data not shown). The ingression, growth and differentiation of the ectoplacental cone into the decidua mostly occurs after the uterine epithelium has undergone apoptosis and after the underlying basement membrane has been remodeled and penetrated by reactive endothelial cells (Welsh and Enders, 1991a). Therefore, in contrast to primary trophoblast giant cells, secondary giant cells may not be functionally dependent on the expression of invasive, lytic properties. Immature cells in the ectoplacental cone did not stain positively for gelatinase B (Fig. 1C), confirming that expression of this enzyme is a marker of terminal differentiation of trophoblast giant cells.

To determine which MMPs and TIMPs were poised to contribute functionally to the remodeling processes characteristic of the implantation site, we probed RNA extracted from a timed series of deciduae (and embryos) and residual uterine tissue for their expression (Fig. 2). TIMP-1 increased steadily throughout the time course under study (5.5 to 9.5 days p.c.), whereas TIMP-2 expression was approximately constant (Fig. 2A,B). TIMP-3 mRNA showed a striking temporal regulation, highly induced by 6.5 days p.c., decreasing at 7.5 days p.c., and ceasing at 8.5 days p.c. (Fig. 2A,C). PECAM-1, a marker of endothelial cells (Baldwin et al., 1994) and therefore of vascularization, was high and constant in decidual tissue and little expressed in the fibrous uterine sheath. Gelatinase A expression declined by 6.5 days p.c. (Fig. 2A,B), but gelatinase B (like TIMP-1) increased; it was low but detectable at 5.5 and 6.5 days p.c., increased at 7.5 days and was highest at 9.5 days p.c. (Fig. 2A,C). Gelatinase B expression paralleled the increasing number of differentiated trophoblast giant cells. Interestingly, the peak expression of TIMP-3 preceded the accumulation of gelatinase B and the maximal invasive activity of the secondary trophoblast giant cells after 8.5 days p.c. Stromelysin-1 and stromelysin-3 showed low and constant expression (data not shown).

We then determined gelatinolytic enzyme and inhibitor activities by zymography. Gelatinase A was found in the decidual extracts in both precursor (72×10³Mr) and activated (62×10³Mr) forms (Fig. 3A). Its activity was highest in 5.5-day-p.c. samples and declined thereafter; this pattern corresponds with mRNA expression for this enzyme (Fig. 2A). The proportion of activated enzyme was constant throughout. A 25×10³Mr gelatinase was expressed in uterus and in undifferentiated decidual tissue (5.5 days p.c.) but disappeared upon differentiation of the decidua between 5.5 and 6.5 days p.c., making it an effective marker of this transition of stromal cells. This 25×10³Mr enzyme was not an MMP, because it was not inhibited by 1,10-phenanthroline (Fig. 3B).

Surprisingly, when normalized to wet weight of tissue, gelatinase B activity was similar in decidua and uterus throughout the time course. In fact, extracts of oil-induced deciduoma, which does not contain embryonic trophoblast, also yielded similar quantities of gelatinase B activity (Fig. 3F). The gelatinase B activity seen by zymography therefore did not correlate with data from RNA blots or with the embryonic trophoblast-specific signal observed by immunostaining or by in situ hybridization (Figs 1, 2). The discrepancy between expression patterns of mRNA and activity may be due to the release of preformed gelatinase B protein from the intracellular granules of infiltrating neutrophils or monocytes (Shapiro et al., 1995) or to its immobilization on cell surfaces or on the ECM surrounding cells in the decidua and uterus (Menashi et al., 1995). However, the implantation reaction was accompanied by the specific activation of gelatinase B (Fig. 3A). Distribution of the enzyme between inactive and active forms in extracts of whole decidua embryo varied; the proportion of activated gelatinase peaked at 7.5 days p.c. and declined until the last time point examined (9.5 days p.c.). Activated gelatinase B was not present in the non-decidual uterine tissue (Fig. 3D). Dissection of the decidua into mesometrial and antimesometrial regions (see below for diagram) showed that gelatinase B activation was specific to the antimesometrial peri-implantation domain and not to the mesometrial domain (Fig. 3D). Deciduomal tissue lysates typically showed less activation of gelatinase B than did corresponding decidual lysates (Fig. 3F), except when deciduomas had bloody cores. In that case, the amount and activation of gelatinase B were greatly increased. Reverse zymography of inhibitor activities confirmed that TIMP-3 was the major MMP inhibitor present in decidual extracts at 5.5-8.5 days p.c. (Fig. 3C). This is not surprising, given the fact that TIMP-3 binds to ECM much more avidly than do TIMP-1 and TIMP-2, increasing its effective half-life. Low levels of TIMP-1 and TIMP-2 were also detected, as well as some higher molecular mass complexes, which are likely to be TIMP-3 dimers or complexes of MMPs with TIMPs. Between 5.5and 6.5 days p.c., TIMP-3 activity was highly upregulated. By 8.5 days p.c., TIMP-3 disappeared from the extracts, closely following the mRNA expression time course. Analysis of the dissected antimesometrial peri-implantation zone showed that it was relatively enriched in TIMP-3, whereas the uterus and mesometrium contained higher levels of TIMP-2 (Fig. 3D).

Deciduomal tissue showed the same pattern of localization of TIMP-3 protein expression by zymography. The innermost tissue around the lumen showed loss of expression of the 25×10³Mr enzyme activity and high TIMP-3 activity (Fig. 3E).

We used in situ hybridization to precisely locate cells expressing TIMP-3 and gelatinase B, which were shown by microdis-section and zymography of decidual extracts to be enriched in the peri-implantation zone. In 6.5-day-p.c. (data not shown) and 7.5-day-p.c. decidua (Fig. 4), gelatinase B-positive giant cells at the margins of the trophoblast were bounded by primary decidual cells expressing high levels of TIMP-3 mRNA. Using enlargement of stromal cells as a morphologic marker of decidualization, we found that expression of TIMP-3 at this stage of development was limited to differentiating decidual cells, which peak at 6.5 days p.c. and are gone by 8 days. Similarly, TIMP-3 mRNA responded in parallel with a wave of decidual differentiation that was initiated at the antimesometrial pole of the implantation chamber and spread toward the mesometrial pole and outward through the spiny layer and the zone of vascularization (data not shown).

Mature decidual cells are present transiently during implantation, undergoing differentiation followed by apoptosis to allow expansion of the growing embryo and placenta. To offer an explanation for the transient induction of TIMP-3 during terminal decidual differentiation, we examined the relationship of TIMP-3 expression to the initiation of an apoptotic program, determined by staining for DNA fragmentation in situ. Several cross-sections of deciduae, 6.5 to 8.5 days p.c., stained by the TUNEL method, were captured and overlaid to show the accumulated distribution (Fig. 5). The distribution of apoptotic cells in single sections, compared with the TIMP-3 mRNA distribution seen by in situ hybridization on near-adjacent sections, is shown in Fig. 5A,B. The decidual stroma at 6.5 days p.c. showed a low rate of apoptosis (1.03±0.24 cells/unit area) and the TIMP-3-positive zone had an even lower rate (0.63±0.07).

Inside the decidual cells was a ring of dying uterine epithelial cells, just behind the shell of primary giant trophoblast cells. The apoptotic index for this peri-embryonic zone was 12.8±7.2. At 7.5 days p.c., apoptosis was low in the remaining region of TIMP-3 and in the decidual stroma outside this zone. However, apoptosis in the uterine layer of cells next to the embryo, consisting of dead and dying primary decidual cells, had increased (45±5.6). By 8.5 days p.c., when most of the TIMP-3-expressing decidual cells had disappeared, the peri-embryonic apoptotic embryonic rate had decreased, except around the ectoplacental cone, and the decidual rate was negligible. Thus, cells that synthesized TIMP-3 showed low rates of cell death, whereas cells no longer expressing TIMP-3 in proximity to gelatinase B synthesis showed high rates of cell death.

The biochemical data shown here indicate that TIMP-3 and gelatinase B are only two of a number of contributors to the proteolytic environment of the implantation site. Accordingly, we used in situ hybridization to determine the spatial localization of other MMPs and TIMPs at 7.5 days p.c. Gelatinase A was expressed in the zones of the decidua that were not differentiated (Fig. 6A), in striking contrast to the embryonic trophoblast-specific expression of gelatinase B (Fig. 6B). The low level of gelatinase B mRNA in decidua did not localize to any specific site. Because the number of undifferentiated decidual cells declines as implantation progresses, the localization of gelatinase A explains the decrease in total activity observed by zymographic analysis of decidual extracts during the implantation period. Stromelysin-1 (Fig. 6F) and stromelysin-3 (Fig. 6E) were also expressed in the progressively smaller zone of undifferentiated cells toward the outer margin of the decidua. TIMP-1 (Fig. 6C) was expressed in undifferentiated zones, and TIMP-2 (Fig. 6H) was expressed in a similar, but relatively larger, domain. A summary diagram (Fig. 7) illustrates the bat-wing shape of the zone of differentiating maternal cells in sections of the 7.5-day decidua. We conclude that the expression of the enzymes gelatinase A, stromelysin-1, stromelysin-3 and the inhibitors TIMP-1 and TIMP-2 described here was typical of undifferentiated stroma and was excluded from the bat-wing domain.

To probe the function of MMPs during the implantation reaction, we inhibited MMP activity by one of two methods. Either we injected the MMP inhibitor MPI intraperitoneally twice daily for 3 days, using MIC for control experiments, or we overexpressed TIMP-1 under the control of a β-actin promoter by transgenic means.

Deciduae from mice injected with the control compound MIC were indistinguishable from those from uninjected mice. Normal deciduae had an elongated egg shape, and the embryos were displaced toward the antimesometrial end (Fig. 8D). The position of the mural tip of the egg cylinder was offset to 65% of the total decidual length (Table 1). In contrast, deciduae of mice injected with the inhibitor MPI were difficult to dissect out. They adhered to the uterine wall and tended to split into two halves, spilling out the embryos. After the deciduae were teased out of the uterine sheath, parts of uterine tissue were left adhering to them (indicated by dots outlining the decidua proper in Fig. 8A-C). We conclude that the boundary that usually separates the decidua from the uterine myometrium did not form effectively in MPI-treated mice. The deciduae were round instead of egg-shaped, giving a higher width-to-length ratio than controls (Table 1). The width of MPI-treated deciduae was unaffected, but they were shorter (83% of control), and their cross-sectional area, measured through the midplane, was reduced to 82% of normal (Table 1). Embryos were more centrally located within the decidual mass; on average, the mural tip of the embryo was at 60% of total length. These features suggest that the mesometrial decidua fails to undergo typical extension, growth and differentiation in the presence of MPI. The embryos inside MPI-treated deciduae tended to be mis-oriented (Fig. 8A-C), giving cross-sections, or glancing sections, through the egg cylinder instead of longitudinal sections. There was little difference in the size or developmental stage of embryos in MPI-treated deciduae or in the number of implantation sites.

The concentration of MPI is likely to be highest close to blood vessels. This distribution may limit the inhibitor’s effectiveness at the initial uterinetrophoblast interface. Therefore, we used transgenic mice that overexpressed TIMP-1 both maternally and embryonically under the control of the β-actin promoter to generate pericellular inhibitor even in avascular domains, as well as on the trophoblast side. Unlike MPI, which inhibits a number of metalloproteinases (Grobelny et al., 1992), TIMP-1 is specific for MMPs. Transgenic mice overexpressing TIMP-1 had a phenotype similar to that of MPI-treated mice, but less severe, confirming that MPI was acting primarily to inhibit MMPs. Deciduae in transgenic mice were reduced in size (84%), and length was reduced significantly (92%) but not as much as with MPI treatment (Table 1). Embryo orientation did not appear to be affected. TIMP-1 overexpression did not significantly affect placental function or embryo viability. The tissue and serum levels of human TIMP-1 induced by transgenic expression were approximately 2 nM (unpublished observations), whereas injection of MPI at 100 mg/kg produced TIMP-1 serum levels of 10-20 mM. This difference is likely to explain the less extreme phenotype shown in TIMP-1-overexpressing mice.

The data obtained with the MPI-treated and TIMP-1-overexpressing mice suggest that decidualization is the process most affected by the presence of ectopic MMP inhibitors during implantation. To test this hypothesis, we studied the effect of MMP inhibitors on the decidual reaction that is artificially induced by peanut oil treatment. The formation of deciduomas was compared in the presence and absence of MPI. In this case, there were no embryo- or trophoblast-derived MMPs. Because it is difficult to assess shape and size changes in these irregular decidual tissues, we used biochemical criteria to assess the effect of treatment. A zymo-graphic profile typical of less differentiated decidua, with low expression of TIMP-3 and high expression of the 25×10³Mr MMP (Fig. 9), was typical of MPI-treated deciduomas. We conclude that the decidual reaction itself is slowed by inhibition of MMPs.

We tested the hypothesis that gelatinase B mediates the ingression of the trophoblast into the maternal decidua. The maternal decidua has been ascribed an unusual ability to limit the invasion of embryonic trophoblast (Abrahamsohn and Zorn, 1993). Undecidualized uterine stroma allows penetration of the fetus to the level of the myometrium. If invasion is mediated by an MMP, then expression of an MMP inhibitor by decidual cells could serve to limit outgrowth. We have shown that differentiating decidual cells do indeed express high concentrations of the MMP inhibitor TIMP-3. However, our data suggest that this simplified explanation of enzyme and inhibitor function may be incorrect.

The first wave of trophoblast giant cells, emerging from the mural side of the coherent primary trophoblast layer surrounding the late blastocyst, invades the antimesometrial zone of the forming decidua to form a meshwork shell of interlinked embryonic cells within the endometrium (Welsh and Enders, 1987; Hoffman and Wooding, 1993). The antimesometrial zone is the site of initiation of the decidualization reaction. Both mRNA and protein for gelatinase B were expressed by trophoblast giant cells in vivo, similar to the expression pattern for Hxt, a transcription factor that regulates trophoblast giant cell development, and placental lactogen-1, a marker of trophoblast differentiation (Cross et al., 1994, 1995). Differentiation of maternal stromal cells to decidual cells was accompanied by a striking induction of TIMP-3 synthesis, shown as expression of specific mRNA and an inhibitory zymographic activity. Thus, expression of gelatinase B by primary giant trophoblast cells was met by a high and probably saturating synthesis of the MMP inhibitor TIMP-3 in differentiating maternal decidual cells. This pattern of expression of gelatinase B and TIMP-3 was similar to the mRNA localization reported in studies by Harvey et al. (1995) and Reponen et al. (1995) that appeared while this study was in progress.

The secondary wave of trophoblast growth, namely, the expansion and maintenance of decidual cells and ingression of the ectoplacental cone into the decidua around 7.5 days p.c., occurs when the expression of TIMP-3 mRNA has declined to an undetectable level. This suggests that this inhibitor is not the primary barrier to embryonic trophoblast outgrowth but may function in other, coincident processes, including differentiation and assembly of the maternal-fetal blood interface. For example, there are ECM-associated components that help to suppress inflammation and rejection of embryos by inhibiting macrophage function (McKay et al., 1992).

Expression of TIMP-3 by decidual cells did not require the fetal trophoblast; maternal cells that were artificially induced to decidualize (deciduomas) also expressed TIMP-3. The morphology of deciduomas is similar to that of decidual tissue (O’Shea et al., 1983); hence, analysis of deciduomas effectively identifies the relative contribution of mother and fetus. More surprising than the induction of TIMP-3 was the observation that deciduomas contained the same level of gelatinase B enzyme activity as deciduae, despite the absence of fetal trophoblast cells. Either the gelatinase B protein was being carried into the decidual site, without mRNA, by inflammatory cells, or there was a low level of mRNA directing synthesis of the enzyme in a large number of cells. We conclude that the enzyme expressed by trophoblast giant cells is present in a background of uterine gelatinase B of undetermined origin. During the process of implantation, but not decidualization, gelatinase B was activated specifically, suggesting that trophoblasts may express a gelatinase B activator (Omura et al., 1994; Sato et al., 1994).

In the present study, expression of TIMP-3 by decidual cells was transient, and the decidua showed negligible synthesis of TIMP-3 after 8 days p.c. We propose that TIMP-3 is a robust marker for differentiating uterine decidual cells during the early implantation period. The regulation of TIMP-3 synthesis by differentiating decidual cells may be related to their passage through their last cell cycle (Wick et al., 1994). Primary decidual cells involute as placentation proceeds (Katz and Abrahamsohn, 1987; Abrahamsohn and Zorn, 1993); the wave of decidual cell death mimics the earlier wave of differentiation and is accompanied by the induction of tissue-type transglutaminase (Piacentini and Autuori, 1994) and DNA laddering (Gu et al., 1994). Because overt cell death has not been reported in morphologic studies of mouse decidua before day 9 p.c., it was important to ascertain the temporal relationship between TIMP-3 expression and the initiation of programmed cell death.

TIMP-3 expression was lost before the onset of programmed cell death. Although we predicted that TIMP-3 protein would be long-lived in the extracellular space, since it is bound to the ECM (Blenis and Hawkes, 1984; Staskus et al., 1991; Leco et al., 1994), it was removed with the same kinetics as the transcriptional decline. Decidual cells actively edit and remodel ECM during this stage of their growth and development; notably, collagen (Martello and Abrahamsohn, 1986) and fibronectin (Rider et al., 1992) are degraded. TIMP-3 may protect key structural or signaling components of ECM and act as a survival factor for decidual cells, which may be sensitive to the induction of apoptosis by gelatinase B and other MMPs from both decidua and trophoblast. Interestingly, during an analogous process of stromal differentiation, the corpus luteal cells of the ovary synthesize very high levels of TIMP-1 after ovulation (Nomura et al., 1989; antibody staining for TIMP-1 protein, unpublished observations). TIMP induction in steroid-sensitive stromal cells could be a general phenomenon associated with the differentiation of mesenchymal cells to paraepithelial secretory cells, serving a common purpose.

ECM has been shown to be a survival factor for several cell types, including endothelial cells, both in vitro (Meredith et al., 1993) and in vivo (Brooks et al., 1994). Overexpression of the MMP stromelysin-1 initiates apoptosis of mammary cells both in vitro and in vivo (Boudreau et al., 1995). ECM rescues ectodermal cells from apoptosis-inducing signals secreted by endodermal cells in cavitating mouse embryoid bodies (Coucouvanis and Martin, 1995). Depending on the repertoire of integrins and ECM molecules expressed, MMP processing could change the cell-ECM interface to generate an apoptotic signal.

Morphologically, the decidua has clearly specialized domains (Welsh and Enders, 1985). Although only TIMP-3 and gelatinase B were expressed at high levels in the proximal peri-implantation zone, other enzymes and inhibitors were expressed in distal domains of the decidua. Stromelysin-1, stromelysin-3, gelatinase A, TIMP-1 and TIMP-2 were all expressed to a greater or lesser extent in the undifferentiated component of the decidua, outlining a bat-wing-shaped core around the implantation site. In fact, no decidual domain was devoid of synthesis of at least one inhibitor or MMP. Several other phenotypic markers associated with decidualization have been described. Inhibin is localized to the decidual zone, whereas follistatin is expressed in the outer, undifferentiated domain (Albano and Smith, 1994). Tenascin (Julian et al., 1994), transforming growth factor-α (Han et al., 1987), laminin B1 (Farrar and Carson, 1992) and desmin (Glasser and Julian, 1986) are expressed by differentiated decidua. α1-Acid glycoprotein is expressed in the antimesometrial zone and α2-macroglobulin in the mesometrial area (Thomas, 1993). c-fms is expressed transiently in peri-implantation decidua (Regenstreif and Rossant, 1989) and later in spongiotrophoblasts, as is TIMP-3 (Apte et al., 1994).

Invasion and implantation of trophoblasts occurs simultaneously with other processes. The undifferentiated stroma in the implantation chamber proliferates and remodels, separating from the surrounding myometrium. There is active angiogenesis in the mesometrial zone, and this is also likely to have some proteinase/inhibitor participation: early dilation of a subepithelial capillary plexus is followed by general activation and hypertrophy of endothelial cells, which migrate to form sinusoids in contact with the implantation chamber (Welsh and Enders, 1991b). Synthesis of gelatinase A, stromelysin-3 and TIMP-1 has also been observed for proliferating human endometrium (Rodgers et al., 1994).

Because inhibitors and enzymes are continuously synthesized in decidua, we artificially disturbed the proteinase:inhibitor ratio. Either injection of MPI, a peptide hydroxamic acid metalloproteinase inhibitor that inhibits trophoblast invasion in vitro (Librach et al., 1991), or overexpression of TIMP-1 in transgenic mice produced similar results. MPI treatment induced a more extreme phenotype that is probably explained by the higher inhibitor concentration achieved: 2 nM for transgenic TIMP-1 overexpression and 10-20 mM for MPI administration. Despite expression of this transgene in the trophoblast giant cells, trophoblast invasion proceeded relatively efficiently. The orientation of implanting embryos is normally strictly regulated; this was relaxed after MPI treatment. Since nothing is known about the molecular determinants that control orientation, it is difficult to project the role of MMPs. Deciduae from inhibitor-treated mice were reduced in length and cross-sectional area when compared to control mice, so that they were not egg-shaped but spherical. The reduced size resulted from a delay in decidual differentiation with a lack of expansion in the mesometrial zone of the decidua, because the embryo was not asymmetrically located in the antimesometrial region, but was central in a smaller, round decidua. During their dissection, instead of splitting along the decidual-uterine junction, MPI-treated deciduae would split along the luminal axis and spill out the embryo. Thus, the formation of a shear zone that separates the decidual mass from the uterus was inhibited in the presence of MPI. We suggest that the target for these inhibitors may be the MMPs expressed in the undifferentiated domains of the decidua. The mesometrial pole is relatively undifferentiated and would be predicted to be affected more by treatment with inhibitor. Furthermore, administration of MPI to mice forming deciduomas also reduced the development and morphogenesis of deciduomal tissue. These data lead us to suggest that MMP activity facilitates decidual morphogenesis and development.

Taken together, our data provide evidence for a critical role of MMPs and their inhibitors during decidualization. Our study still leaves several important questions unanswered. Is there a decidual barrier to trophoblast invasion? What controls the invasive program? How critical is MMP expression by trophoblasts during the murine implantation process? A definitive resolution of these issues awaits completion of targeted mutations in specific MMP genes.

We thank J. Cross and A. MacAuley for stimulating discussion about many aspects of this work. This work was supported by an Investigator Award from the Arthritis Foundation (C. M. A.), grants from the National Institutes of Health (HD 26732 to Z. W., and CA 39919 to S. H.) and a contract from the Office of Health and Environmental Research, U.S. Department of Energy (DE-AC03-76-SF01012).