EMILIN1 represents a major stromal element determining human trophoblast invasion of the uterine wall.

The detection of EMILIN1, a connective tissue glycoprotein associated with elastic fibers, at the level of the ectoplacental cone and trophoblast giant cells of developing mouse embryos (Braghetta et al., 2002) favored the idea of a structural as well as a functional role for this protein in the process of placentation. During the establishment of human placenta, a highly migratory subpopulation of extravillous trophoblasts (EVT), originating from anchoring chorionic villi, penetrate and invade the uterine wall. In this study we show that EMILIN1, produced by decidual stromal and smooth muscle uterine cells, is expressed in the stroma and in some instances as a gradient of increasing concentration in the perivascular region of modified vessels. This distribution pattern is consistent with the haptotactic directional migration observed in in vitro functional studies of freshly isolated EVT and of the immortalized HTR-8/SVneo cell line of trophoblasts. Function-blocking monoclonal antibodies against α4-integrin chain and against EMILIN1 as well as the use of EMILIN1-specific short interfering RNA confirmed that trophoblasts interact with EMILIN1 and/or its functional gC1q1 domain via α4β1 integrin. Finally, membrane type I-matrix metalloproteinase (MT1-MMP) and MMP-2 were upregulated in co-cultures of trophoblast cells and stromal cells, suggesting a contributing role in the haptotactic process towards EMILIN1.


Introduction
Placenta is a newly formed tissue organized at the fetomaternal interface during pregnancy with the contribution of both the fetus and the mother. The fetal counterpart is represented by trophoblasts that cover the chorionic villi and which comprise two layers of cells. The inner layer is formed by cytotrophoblasts (CTBs) that adhere to the basal membrane of the villi. These cells mature into multinucleated syncytiotrophoblasts localized in the outer surface of the villi in direct contact with the maternal blood circulating in the intervillous space. The villous trophoblasts constitute a continuous physical barrier that allows a selective passage of nutrients and protective factors from the mother to the fetus.
Anchoring villi insert into the maternal decidua, which contributes to establish a physical interaction between the fetus and the mother. From the tip of the anchoring villi, cytotrophoblasts migrate as extravillous trophoblasts (EVTs) into the decidua. Endovascular trophoblasts represent a special subpopulation of EVTs that colonize the spiral arteries moving against the blood flow and partially replacing the endothelial cells. Zhou et al. (Zhou et al., 1997) and Bulla et al. (Bulla et al., 2005) have shown that endovascular trophoblasts replace the endothelium via a non-destructive process leading the trophoblasts to adhere to the extravascular side of endothelial cells and to migrate through the endothelium using vascular endothelial (VE)-cadherin for cell-cell interaction. Other EVTs invade the interstitium of the decidua, reaching the inner-third of the myometrium. These cells tend to accumulate around the spiral arteries and contribute to the vascular remodeling characterized by progressive loss of smooth muscle cells and deposition of fibrinoid material on the arterial wall. This process leads to an increased diameter and decreased resistance of the spiral arteries and provides a means of steady perfusion of the placental sinusoids with maternal blood (Graham and Lala, 1992;Kaufmann and Castellucci, 1997;Aplin et al., 2000).
Studies directed at elucidating mechanisms that regulate EVT cell proliferation, migration and invasion have indicated that regulation is provided by a variety of factors present in the EVT cell microenvironment, including growth factors, growth The detection of EMILIN1, a connective tissue glycoprotein associated with elastic fibers, at the level of the ectoplacental cone and trophoblast giant cells of developing mouse embryos (Braghetta et al., 2002) favored the idea of a structural as well as a functional role for this protein in the process of placentation. During the establishment of human placenta, a highly migratory subpopulation of extravillous trophoblasts (EVT), originating from anchoring chorionic villi, penetrate and invade the uterine wall. In this study we show that EMILIN1, produced by decidual stromal and smooth muscle uterine cells, is expressed in the stroma and in some instances as a gradient of increasing concentration in the perivascular region of modified vessels. This distribution pattern is consistent with the haptotactic directional migration observed in in vitro functional studies of freshly isolated EVT and of the immortalized HTR-8/SVneo cell line of trophoblasts. Function-blocking monoclonal antibodies against ␣ ␣4integrin chain and against EMILIN1 as well as the use of EMILIN1-specific short interfering RNA confirmed that trophoblasts interact with EMILIN1 and/or its functional gC1q1 domain via ␣ ␣4␤ ␤1 integrin. Finally, membrane type I-matrix metalloproteinase (MT1-MMP) and MMP-2 were upregulated in co-cultures of trophoblast cells and stromal cells, suggesting a contributing role in the haptotactic process towards EMILIN1.
During invasion and migration of tissues and vessels, EVTs encounter obstacles similar to those that tumor cells must overcome to invade surrounding and distant tissues during the metastatic process. EVT invasiveness, while remaining largely confined to the endometrium-myometrium junction, depends on mechanisms similar to those of invasive tumor cells. Invasion in situ is regulated by locally derived factors as well as by interactions of EVT with certain ECM components through cell surface integrins . For instance, fibronectin (FN) is abundant at sites of anchoring villus formation in vivo and both maternal and trophoblastderived FNs are present in the ECM encountered by EVT in their pathway within the decidual tissue. Thus, it was suggested that FN acts as a bridging ligand mediating anchorage and/or migratory activity following the interaction with the cognate ␣5␤1 integrin receptor of EVTs (Ilic et al., 2001).
The aim of the present investigation was to define the role of EMILIN1, a connective tissue glycoprotein associated with elastic fibers (Doliana et al., 1999), in the process of interstitial (stromal) migration and/or invasion of EVTs. The study was prompted by a previous finding that abundant EMILIN1 mRNA is present at the level of the ectoplacental cone and trophoblast giant cells of developing mouse embryos (Braghetta et al., 2002). EMILIN1 belongs to a new family of proteins of the ECM characterized by a unique arrangement of structural domains, including a unique cysteine-rich sequence of approximately 80 amino acids at the amino-terminus, the EMI domain, an ␣-helical domain with high probability for coiled-coil structure formation in the central part, and a region homologous to the globular domain of C1q (gC1q domain) at the carboxyl-terminal end (Doliana et al., 1999;. EMILIN1 is often observed in vivo closely adjacent to the surface of cells; it is recognized at its gC1q1 domain by the ␣4␤1 integrin and it is able to promote tumor cell migration (Spessotto et al., 2003a).
In the present study, by using ex vivo cells and an in vitro cellular model system of trophoblast, we have identified EMILIN1 as a candidate molecule exerting a crucial role in promoting EVT migration and/or invasion from the anchoring columns into the decidualized and perivascular stroma.

Immunolocalization of EMILIN1 in human placenta
The presence of EMILIN1 was documented both at the level of the chorionic villi and in decidua (Fig. 1A). No staining was (A) Cryostat sections of first-trimester placenta were stained with antibodies specific for EMILIN1 followed by secondary HRPlabeled antibodies and developed with DAB. The sections were counterstained with hemallum. (B) This section was immunostained for EMILIN1 (in brown) and for the presence of vWF, visualized with an alkaline-phosphatase-labeled secondary antibody, and it was not counterstained. Asterisk indicates a modified vessel; lower panel: triple staining for EMILIN1 (in brown), and cytokeratin (in green fluorescence) and vWF (in red) to visualize EVT and endothelial cells, respectively. (C) Cryostat sections of first-trimester placenta were immunostained for EMILIN1 and for the presence of vWF. Upper panel: EMILIN1 (in brown) and vWF (in pink); lower panel: EMILIN1 (in pink) and vWF (in brown). Arrowheads indicate spiral arteries and asterisks large modified vessels where the gradient distribution of EMILIN1 seems more evident. Bars, 50 m. seen in the sincytiotrophoblast layer and in the underlying cytotrophoblast stem cells (Fig. 1A, upper panel). The analysis of a section of an immature mesenchymal villous revealed an intensive staining in large polygonal stromal cells with vescicular nuclei and abundant cytoplasm (Fig. 1A, upper panel). In decidua EMILIN1 was expressed at low levels in proximity to apparently unmodified spiral arteries (Fig. 1A, lower panel; Fig. 1C, upper panel; supplementary material Fig.  S1). Interestingly, the staining of EMILIN1 in certain areas of the decidual stroma displayed a gradient of increasing intensity towards 'modified' vessels ( Fig. 1B; Fig. 1C, lower panel). This pattern was more evident in modified vessels of larger size. The EMILIN1 gradient stopped abruptly at the level of perivascular EVTs (Fig. 1B). The use of double and/or triple immunostaining better illustrated the relationship among blood vessels, EVT, stromal cells and EMILIN1 ( Fig. 1B; supplementary material Fig. S1).
At the level of unmodified vessels (supplementary material Fig. S1A) EMILIN1 is distributed more evenly in the stroma as well as in the vessel media, as shown by serial z-sections (supplementary material Fig. S1B).

Cell adhesion of trophoblast cells to ECM components
The villous ECM sustains the overlaying trophoblast and supports major placental morphogenetic processes, such as remodeling and vascular development. Several ECM components are expressed in placental stroma and few were suggested to play major functional roles (Ilic et al., 2001;Damsky et al., 1992;Korhonen et al., 1991;Kilburn et al., 2000). To evaluate the pro-adhesive properties of ECM components, we used an extensive panel of purified ECM molecules and observed that HTR-8/SVneo cells Journal of Cell Science 119 (21) bound weakly (i.e. from 15% to 25% of cell adhesion) to collagen (Col) III, Col V, fibrinogen, matrigel, tenascin, moderately well (i.e. from 26% to 49% of cell adhesion) to laminin-1 (LN-1), perlecan, Col IV, Col VI, and strongly attached (i.e. above 50% cell adhesion) to Col I, FN, vitronectin (VN), hyaluronan (HA), EMILIN1 ( Fig. 2A). Several molecules that showed a weak ability to support HTR-8/SVneo adhesion were not altered or devoid of adhesive function since they actively sustained attachment of other cell types (data not shown). To investigate whether the expression of EMILIN1 in placenta could exert a specific and functional role, we examined the ability of fresh EVT cells derived from first-trimester human placentas to interact with this ECM molecule. Although with some variability, fresh EVT cells displayed a strong reactivity for EMILIN1 and FN comparable to HTR-8/SVneo cells (Fig. 2B).
gC1q1, the functional C-terminal domain of EMILIN1 (Spessotto et al., 2003a), was required and sufficient to promote cell adhesion (Fig. 2C). The specificity of the interaction was supported by the finding that a complete inhibition of HTR-8/SVneo cell attachment was obtained in the presence of the P1H4 function-blocking anti-␣4-integrin antibody (Fig. 2C). The adhesion was also blocked using the monoclonal antibody (mAb) 1H2G8 mapping to the gClq1 domain, indicating that trophoblast ␣4␤1 integrin interacts with EMILIN1 through the gC1q1 domain of the protein. FACS analyses confirmed the presence of the ␣4␤1 integrin receptor on freshly isolated and Ck7-positive EVT cells and provided for HTR-8/SVneo cells an integrin profile corresponding to their adhesive pattern (data not shown).
In agreement with the ␣4␤1 usage in cell adhesion to EMILIN1, there was a pronounced qualitative difference in the   S2A). Phalloidin staining of cells spread on FN and VN revealed a high number of actin-containing stress fibers. In addition, paxillin localized to large focal contacts at the tips of stress fibers (supplementary material Fig. S2A, upper panel). Accordingly, focal adhesion kinase (FAK) also localized to focal contacts (supplementary material Fig. S2A, lower panel). By contrast, phalloidin staining of cells attached to EMILIN1 indicated that actin was mainly present along the cell periphery at the level of extended cell protrusions, reflecting a different organization status of the actin cytoskeleton compared with the cells on FN or VN; both paxillin and FAK were evenly distributed in the cell cytoplasm without any apparent focal contact formation (supplementary material Fig. S2). Focal contacts were not visible even after 3 hours of adhesion on EMILIN1 (supplementary material Fig. S2), suggesting that the different morphology initially detected at 30 minutes of adhesion was not dependent upon an incomplete occupancy of the integrin.

Trophoblast cell movement towards EMILIN1
These findings prompted us to ask whether EMILIN1 promoted efficient migration of trophoblast cells. Accordingly, as a first approach, we compared migration towards EMILIN1 versus a selected series of ECM molecules and used HTR-8/SVneo cells to determine quantitatively the substrate preference. HTR-8/SVneo cells did not migrate above background levels towards Col IV and moved poorly towards LN-1 and Col I (Fig. 3A). Instead, a significant number of cells migrated effectively as early as 5 hours towards FN-, VN-and EMILIN1-coated filters (Fig. 3A).
To obtain further insight into the motility of HTR-8/SVneo, cells were allowed to migrate under two different experimental conditions. Chamber filters were either coated with ECM molecules on their lower side to enable cells to engage in haptotactic directional migration towards the substratumbound density gradient of adsorbed protein (McCarthy and Furcht, 1984), or coated on the upper side to evaluate the intrinsic motility of the cells in the absence of an 'ECM gradient' (a cell movement defined as 'migration'). HTR-8/SVneo cells displayed polarized migration to a similar extent (approximately 20% of migrated cells) irrespective of which side of the filter was coated with FN ( Fig. 3B). In sharp contrast, EMILIN1 was able to significantly enhance directional motility when filters were coated on the lower surface (35% migration) compared with coating on the upper side (18% migration), suggesting that this protein promotes haptotaxis more efficiently than FN. The observed migratory preference displayed by HTR-8/SVneo cells was also reproduced with EVT cells that moved towards EMILIN1 with a significantly higher efficiency than towards FN (Fig. 3C). Thus, this similarity suggests the potential biological  significance of EMILIN1 in placentation and allowed us to consider HTR-8/SVneo similar to primary cells. The haptotactic motility towards EMILIN1 or gC1q1 was to a large extent inhibited by the addition of function-blocking ␣4integrin subunit mAbs (Fig. 3D), indicating that this cellular receptor plays a crucial role.
In vitro production of EMILIN1 Based on the finding that EMILIN1 appeared to be a preferential substrate for trophoblast directional migration, we searched for the cellular source of EMILIN1. Decidual stromal cells (DSC), human umbilical vein endothelial cells (HUVEC), Journal of Cell Science 119 (21) uterine smooth muscle cells (UtSMC) and HTR-8/SVneo cells were stained with anti-FN and anti-EMILIN1 antibodies. All cell types stained for FN, with DSC and UtSMC showing stronger positivity (Fig. 4A). By contrast, EMILIN1 was detected only in DSC and UtSMC cultures. Immunoblot analysis with specific antibodies confirmed the pattern of ECM expression (supplementary material Fig. S3A). In accordance with the results in situ on cell layers, FN was detected in several cell extracts and was absent in HTR-8/SVneo cells. EMILIN1 was only detected in cell extracts of DSC and UtSMC. Immunofluorescence staining on decidua specimens supported these data: EMILIN1 was clearly present very close and associated with vimentin-positive cells (stromal cells), indicating that these cells probably represent the principal source of EMILIN1 in vivo (Fig. 4B, supplementary material Fig.  S3B). The increasing gradient of expression of EMILIN1 towards CD31-positive cells of a large vessel was also evident (supplementary material Fig. S3B). At higher magnification it was clear that staining for EMILIN1 was not detectable in the trophoblast layer and around clumps of trophoblast cells within the stroma (Fig. 4B). Taken together, these results suggest that the strong expression of EMILIN1 in areas surrounding blood vessels in vivo might be because of a prevalent secretion and deposition by DSC and in part by muscular cells when still present in the vessel walls.

Transmigration of trophoblasts through a stromal cell layer
To support our immunohistochemical and preliminary functional observations, we investigated whether EMILIN1 expressed in vitro by DSC also represented a migratory substrate. For this purpose, stromal cells were plated on the surface of the inserts and grown to confluence for 72 hours. The inserts were then placed in the wells and HTR- Nuclei are visualized with Sytox Green and pseudocoloured in blue, EMILIN1 is detected with Alexa Fluor ® 633 and pseudocolored in green, trophoblast cells are stained with anti-cytokeratin 7 (red, upper panel), and stromal cells with anti-vimentin (red, lower panel). EMILIN1 is deposited in proximity to stromal cells, whereas it seems absent near to trophoblast cells (magnification). Images were acquired with a Leica TCS SP2 confocal system (Leica Microsystems Heidelberg). EMILIN1 and trophoblast migration 8/SVneo cells were added to the upper chamber and allowed to migrate for 5 hours. DSC did not migrate across the filter pores when either plated alone or when HTR-8/SVneo cells were also added (Fig. 5A). By contrast, HTR-8/SVneo cells displayed a high directional motility when stromal cells were grown on the lower surface of the membranes (P<0.002; a representative field of the upper and lower side of the filter is shown in Fig. 5B). The quantitative analysis indicated that nearly 50% of HTR-8/SVneo cells were able to transmigrate under these experimental conditions (Fig. 5B). The presence of specific function-blocking antibodies was able to reduce the migratory process: the inhibition was up to 45% with anti-␣4 integrin chain mAb and up to 65% with the anti-␤1 integrin chain mAb (Fig. 5C). The use of anti-EMILIN1 mAb 1H2G8 (Spessotto et al., 2003a) produced a reduction of 65% and the simultaneous addition of anti-␣4 and anti-EMILIN1 mAbs was responsible for an inhibition as high as 82%. Freshly isolated EVT displayed a similar behaviour: the use of anti-␣4 chain and anti-EMILIN1 reduced their migratory ability. Since DSC were shown to produce both FN and EMILIN1 (Fig. 4, supplementary material Fig. S3A), we cannot formally exclude that FN also contributed in part to the haptotactic migration of HTR-8/SVneo cells. Gene silencing was then used to confirm that EMILIN1 was important for trophoblast cell migration. The efficiency of EMILIN1 gene silencing was partial and evident only at 48 hours in our targeted stromal cells (Fig. 5D, upper panel). ECM proteins usually take approximately 48-72 hours to be deposited and adequately organized and we noticed a decrease in migrated cells in areas where EMILIN1 was produced in lower amounts, as shown in the images reported in Fig. 5D (lower panel). Overall, the inhibitory data with the combination of mAbs and the use of siRNA suggested that EMILIN1 plays a crucial role among ECM molecules in directional movement of trophoblast, and indicates that this motility is largely dependent on ligation of deposited EMILIN1 interacting with the ␣4␤1 integrin expressed on trophoblast.

MMP involvement in trophoblast migration
The role of MMPs in trophoblast invasion has been well documented, with the suggestion that they could affect migratory processes of these cells (Nardo et al., 2003). Among the multiple MMPs produced by human placenta, MMP-2 and MMP-9 have been assigned a key role in promoting the invasive capacity of cytotrophoblasts (Campbell et al., 2003). To further elucidate the physiological meaning of our findings we also evaluated a possible function of MMPs in the contest of the haptotactic process involving EMILIN1 and trophoblast cells. Under normal resting culture conditions HTR-8/SVneo cells secreted MMP-9, whereas DSC secreted MMP-2 (supplementary material Fig. S4A, lane1). Both types of cells were able to express MMP-1 (supplementary material Fig.  S4A, middle panel, lane 1) and membrane type I (MT1)-MMP (supplementary material Fig. S4A, lower panel, lane 1). The secretion of tissue inhibitor of metalloproteinase-1 (TIMP-1) or TIMP-2 was almost undetectable in both types of cells (data not shown).
Since the cells could modulate the expression of MMPs in the presence of soluble stimulating agents [i.e. 12-O-Tetradecanoylphorbol 13 acetate (TPA) or tumor necrosis factor (TNF)-␣] such as MMP-3 and MMP-11 produced by DSC and HTR-8/SVneo cells, respectively (supplementary material Fig. S4A, middle panel, lanes 2 and 3), we asked whether such changes may also be observed in particular culture conditions resembling the in vivo situation. MT1-MMP is the well-known activator of MMP-2 (Seiki, 1999). When HTR-8/SVneo cells were plated on a DSC monolayer, a significant increase (P=0.002) in the activity of MT1-MMP in co-cultured cells was found (Fig. 6A, lower panel). Accordingly, the membrane-bound active form of MMP-2 increased, whereas the enzyme in the supernatants decreased (Fig. 6A, upper panel).
To verify whether MMPs played any role in the context of trophoblast migration, first haptotaxis of HTR-8/SVneo cells towards FN or EMILIN1 was performed in the presence of a pan-MMP inhibitor, GM6001. Migration towards FN or gC1q1 was only slightly affected by the presence of GM6001, being more evident at the beginning and less effective during the migratory process towards gC1q1, whereas it was more effective at later times in cells migrating towards FN (supplementary material Fig. S4B). The migration of DiIlabeled HTR-8/SVneo cells towards FN or gC1q1 was then evaluated in the presence of an equal number of DSC (Fig. 6B). As demonstrated for the upregulation of MT1-MMP activity in co-cultured cells (Fig. 6A), the migration of HTR-8/SVneo cells was strongly and significantly reduced (P<0.01) by GM6001 only towards gC1q1 but not towards FN. That the presence of DSC is a determining factor was demonstrated by the fact that reducing their number to one-third rendered the GM6001-inhibitory effect less evident (data not shown).

Discussion
An adequate perfusion of the placenta by maternal blood is dependent upon optimal migration and invasion of EVT cells into the decidua, followed by a remodeling of the uteroplacental arteries. The present study used ex vivo EVT and an immortalized human EVT cell line with phenotypic and functional characteristics of EVT to investigate the nature of the receptor-ligand interaction underlying this migratory function of trophoblast cells. To elucidate the role of ECM constituents in the trophoblast-invasion phenomenon, we have: (a) mapped the distribution of ECM constituents in several representative human placenta; (b) examined the ability of these proteins to promote attachment and migration of the HTR-8/SVneo human EVT cell line and of freshly purified EVT cells; (c) used function-perturbing antibodies and siRNA in conjunction with in vitro models of invasion to determine whether EMILIN1 is responsible for mediating migration and invasion. The present study extends a recent finding of a geneexpression study in which the presence of EMILIN1 mRNA was demonstrated in placental tissue (Chen and Aplin, 2003) and demonstrates deposition of EMILIN1 in the decidual stroma. Despite the fact that several substrates were expressed in the stroma, only EMILIN1 displayed gradients with increasing expression towards some unmodified vessels. EMILIN1 strongly promoted cell adhesion and migration of trophoblast cells. Strikingly, directional haptotaxis was much more efficient towards EMILIN1 compared with FN. This finding, along with the observation that DSC produce EMILIN1 in vitro led us to hypothesize that deposited EMILIN1 in the uterine tissue favors the invasion of the uterine wall by trophoblast cells. The important role of DSC inferred by our immunohistochemical data was strongly corroborated by the quantitative functional data on directional transmigration of HRT-8/SVneo cells in vitro. The strong inhibition of trophoblast cells' movements towards the gradient of EMILIN1 deposited by DSC obtained with the functionblocking anti-EMILIN1 mAb supports the notion that EMILIN1 is largely responsible for the directional motility of trophoblast cells. The formation of a gradient of deposited EMILIN1 by DSC on the inserts is only intuitive since it was not formally demonstrated. However, the nearly threefold higher transmigration of HTR-8/SVneo cells when DSC were grown on the lower surface of the membrane of the inserts compared with the migration obtained when DSC formed a layer on the upper surface is in accord with our hypothesis.
Our data point to a relevant role of the ␣4␤1 integrin receptor in this EMILIN1-dependent trophoblast migration. Since ␣4␤1 exhibits a predominantly leukocyte expression pattern (Rose et al., 2002), the finding that trophoblast cells (HTR-8/SVneo and EVT) could attach to and very efficiently move towards EMILIN1 through ␣4␤1 without any prior artificial cellular activation was unexpected. It has been reported that EVT cells migrating out of chorionic villi selectively express ␣5␤1 integrin and that treatment of EVT cells with an anti-␣5␤1-integrin-blocking antibody (MacPhee et al., 2001) caused an almost complete inhibition of migration. This finding implied that FN and ␣5␤1 might represent the prevalent ligand-receptor pair involved in the initial anchoring of trophoblast cells in the columns and in the subsequent migration within the uterine wall. ␣5 and ␤1 integrins are concentrated at the local outgrowths, as is FN, which is abundant at corresponding sites in vivo (Chen and Aplin, 2003) (present report). Other FN receptors, including those of the ␣v integrin subfamily and ␣4␤1 were examined, but levels were very low and considered not relevant in the invasion process (Miyamoto et al., 1998). Our results are not in disagreement with the reported migration-stimulating effects of FN. However, EMILIN1, which is not produced by EVT cells, seems to provide a more efficient substrate than FN for EVTs, and it is used by these cells for haptotaxis through the ␣4␤1 integrin receptor.
Cooperation of MMPs with integrins in order to better direct cellular migration could play a contributing role (Deryugina et al., 2001). The ligation of ␣4␤1 integrin upregulated MT1-MMP and conferred a migratory phenotype to mesenchymal cells (Pender et al., 2000). MT1-MMP is a type I membraneanchored MMP that has been shown to play a crucial role in the pericellular degradation of connective tissue matrix by directly hydrolyzing native collagen (Hotary et al., 2003) and indirectly by initiating a cascade of zymogen activation at the cell surface, which leads to the generation of active MMP-2, MMP-13 and MMP-9 (Strongin et al., 1995;Cowell et al., 1998). A strong upregulation of MT1-MMP was detected only when HTR-8/SVneo cells were co-cultured with DSC. When HTR-8/SVneo cells were present alone in the system we could demonstrate only a partial effect of the general MMP inhibitor GM6001 on cell haptotaxis. Interestingly, this inhibition is more evident for EMILIN1 at one-hour migration, whereas it is stronger when cells move towards FN at 4 hours, suggesting that both ligands might be involved in the invasion process but with a sequential role. However, when labeled HTR-8/SVneo cells migrated in the presence of DSC cells a significant reduction in haptotaxis by GM6001 was detected only towards gC1q1 but not towards FN, linking ␣4-integrin activation, EMILIN1 ligands as well as MMPs as a part of an interrelated phenomenon that facilitates trophoblast motility.
Our working hypothesis is depicted in Fig. 7: the upregulated and newly released MT1-MMP (60 kDa form) is directed towards MMP-2 produced by DSC in order to activate this enzyme. The active MMP-2 does not cleave EMILIN1 that is processed by other enzymes present in placental tissue, such as plasmin (C.D. et al., unpublished). MMP-2 might instead unmask cryptic sites within the molecule and in this way The conditioned media were collected after 24 hours and analysed by zymography. The cells were lysed and loaded for zymography or assayed for MT1-MMP activity. Quantitative determination of total MT1-MMP was performed using the Biotrak MT1-MMP activity assay system (Amersham Biosciences Europe, Orsay, France) according to the manufacturer's instructions. Equal amounts of HTR-8/SVneo and DSC cells were cultivated alone and then processed for zymography and MT1-MMP activity. The increase in MT1-MMP activity in co-cultures was significantly higher (P=0.002) than the summed activity of the enzyme produced by DSC and HTR-8/SVneo cells. (B) Inhibition of haptotaxis. GM6001 (10 M) was added to block migration of DiI-labeled HTR-8/SVneo cells in the presence of an equal number of DSC. HTS FluoroBlok TM inserts were previously coated underside with gC1q1 or FN. The migratory process was evaluated at different times (2, 4, 6 and 18 hours). P values were determined versus control samples. Open squares, P>0.05; asterisks, P<0.01.
facilitate the migration process. Alternatively, it might involve cell membrane receptors of the integrin and/or of a different kind in order to create a suitable environment for trophoblast migration. This mechanism has to be validated by further studies to better define the targets of MMP proteolysis and the unique crosstalk with the ␣4␤1 receptor in EMILIN1-mediated migration.
FN was purchased from Calbiochem-Novabiochem (La Jolla, CA). VN was purified from human plasma. Human skin Col III, placental Col IV, placental Col V, human plasma fibrinogen and high M r HA were obtained from Sigma; rat tail Col I and matrigel were purchased from BD Biosciences (Bedford, MA). Native LN-1-nidogen complex from Engelbreth-Holm-Swarm (EHS) mouse tumor was obtained as previously described (Engvall et al., 1983). Chick Col VI tetramers were purified as previously described (Mazzucato et al., 1999). Tenascin and human perlecan were kindly provided by Luciano Zardi and Anders Malmström, respectively.
For the production of EMILIN1 293 cells, constitutively expressing EBNA-1 protein (293-EBNA) and secreting EMILIN1 were used. Briefly, polypeptides secreted in the cell supernatant were purified by ion-exchange chromatography on a DEAE-cellulose column (Amersham Pharmacia Biotech) and size-exclusion chromatography using Sepharose CL-4B (1.0ϫ90.0 cm column, Amersham Pharmacia Biotech) as described . The preparation was checked for the presence of LN and FN by immunoblotting with specific antibodies and proved to be nearly devoid of these contaminants. The C-terminal 6His-tagged recombinant fragment of EMILIN1 (gC1q1) was purified by affinity chromatography on Ni-NTA resin (Qiagen GmbH) under native conditions as described (Spessotto et al., 2003a).

Isolation of fresh EVT and decidual stromal cells
Placenta and decidual specimens were obtained from women undergoing elective termination of pregnancy at 8-12 weeks of gestation. The study was approved by the institutional review board of The Maternal-Children Hospital (IRCCS 'Burlo Garofolo', Trieste, Italy) and informed consent was obtained from all women providing the tissue specimens. In this study we have investigated seven placentas. Fragments of decidual tissue of approximately 1 cm 3 containing some floating and anchoring villi were used for the study of the implantation side. EVTs were purified from placental tissue incubated with Hanks balanced salt solution (HBSS; Invitrogen) containing 0.25% trypsin and 0.2 mg/ml Dnase type I (Roche, Mannheim, Germany) for 20 minutes at 37°C, following a previously published procedure (Tedesco et al., 1993;Bulla et al., 1999). After Percoll gradient fractionation, the contaminating leukocytes were removed by immunomagnetic beads coated with mAb to Dynal,Oxoid,Milan,Italy).
DSCs were isolated from decidual specimens. Briefly, finely minced decidual tissue was digested with collagenase type I (1 mg/ml) (Worthington Biochemical Corporation, DBA, Milan, Italy) for 2 hours at 37°C, filtered through a 100 m pore filter (BD Biosciences) and further digested with 0.1% trypsin (Sigma) and 0.05% Dnase type I (Roche) for 7 minutes. The cells were grown in RPMI with 10% fetal calf serum (FCS). HUVEC were isolated from three to five normal umbilical cord veins by collagenase digestion as previously reported (Tedesco et al., 1997;Lorenzon et al., 1998). UtSMC were provided with their respective culture media and additives by Cambrex Bio Science (Rockland, ME).

Immunohistochemistry and immunofluorescence staining
First-trimester healthy human placenta tissues were obtained from patients undergoing elective termination of pregnancy at 10-12 weeks of gestation. Fragments of placental tissue of approximately 1 cm 3 were embedded in OCT (Miles Laboratory, Milan, Italy), snap-frozen in liquid nitrogen and kept at -80°C until use. Cryostat sections of approximately 6 m were air dried, fixed in acetone for 10 minutes at room temperature and either used immediately or kept at -80°C wrapped in aluminum foil. Immunoperoxidase staining was performed on placental sections incubated with the primary antibodies followed by peroxidase-labeled secondary antibodies (EnVision TM+ kit, DakoCytomation). Double labeling was performed using Dako EnVision Doublestain System (DakoCytomation). The Fig. 7. Schematic representation for the proposed mechanism in the migratory process involving trophoblast cells. The interaction of the gC1q1 domain of EMILIN1, produced by decidual stromal cells, with ␣4␤1 promotes trophoblast migration towards the vessels. The engagement of ␣4␤1 also promotes the release from trophoblast cells of MT1-MMP (60 kDa form) that in turn activates MMP-2 on the membrane of stromal cells. The activated enzyme could digest the ECM component of blood vessels (i.e. laminins), exert an improved efficiency in binding capability or signaling functions of the ␣4␤1 receptor, or unmask EMILIN1 cryptic sites in order to facilitate the migratory phenotype of trophoblast cells. The potential role of FN and other ECM molecules and of integrins other than ␣4␤1 is not represented for clarity. enzymatic reactions were developed with diaminobenzidine (DAB) for immunoperoxidase and Fast Red for alkaline phosphatase (ALP). The reaction product of ALP cromogen produce a brilliant red fluorescence that is visible by fluorescence microscopy using a fluorescein filter (Murdoch et al., 1990). Triple labeling was performed adding fluorescein isothiocyanate (FITC)-labeled mAb anticytokeratin 7 on the same sections. In some cryostat sections nuclei were visualized with Sytox Green and other structures were detected with Alexa Fluor ® 546-and Alexa Fluor ® 633-labeled secondary antibodies (Molecular Probes, Eugene, OR) to obtain multiple staining. Images were acquired with a Leica TCS SP2 confocal system (Leica Microsystems Heidelberg, Mannheim, Germany), using the Leica Confocal Software (LCS) and a 40ϫ fluorescence objective on a Leica DM IRE2 microscope.

EMILIN1 gene silencing
For the EMILIN1 expression knockdown the Dharmacon siRNA-specific SMARTpool (Dharmacon, Chicago, IL) was used. The transfection of decidual stromal cells was performed according to the manufacturer's instructions. Steadystate RNA expression levels were analyzed at different cell culture times by semiquantitative RT/PCR using the SV RNA purification kit (Promega Italia) to isolate total RNA and the access kit (Promega) for RNA retrotranscription and cDNA amplification. To perform PCR, primers designed on distinct exons to distinguish for possible DNA-derived amplification products were used, and gel analysis was done after 25, 30 and 40 amplification cycles to select a cycle number in which the amplification process was linear. ␤-actin was co-amplified for normalization. Negative controls included scrambled RNA oligos (Dharmacon).

Cell adhesion assay
The quantitative cell adhesion assay used in this study is based on centrifugation and has previously been described Spessotto et al., 2000). Briefly, six-well strips of flexible polyvinyl chloride-denoted centrifugal assay for fluorescence-based cell adhesion (CAFCA) miniplates, covered with double-sided tape (bottom units), were coated with the different substrates. Cells were labeled with the vital fluorochrome calcein AM (Molecular Probes) for 15 minutes at 37°C and then aliquoted into the bottom CAFCA miniplates, which were centrifuged to synchronize the contact of the cells with the substrate. The miniplates were then incubated for 20 minutes at 37°C and subsequently mounted together with a similar CAFCA miniplate to create communicating chambers for subsequent reverse centrifugation. The relative number of cells bound to the substrate (i.e. remaining in the wells of the bottom miniplates) and cells that fail to bind to the substrate (i.e. remaining in the wells of the top miniplates) was estimated by top/bottom fluorescence detection in a computer-interfaced SPECTRAFluor Plus microplate fluorometer (Tecan, Rapperswit, Switzerland). In experiments aimed at examining the effects of blocking antibodies, the various antibodies were added directly to the wells, just before plating the cells.

Cell motility assay
Haptotactic-like migration of HTR-8/SVneo or ETV cells in response to ECM substrates was assessed by fluorescence-assisted transmigration invasion and motility assay (FATIMA) , Spessotto et al., 2003b. The procedure is based on the use of Transwell-like inserts carrying fluorescenceshielding porous polyethylene terephthalate (PET) membranes (polycarbonate-like material with 8 m pores), HTS FluoroBlok TM inserts (Becton-Dickinson, Falcon, Milan, Italy). Membranes were coated (20-50 g/ml in 50 l) on either the upperor underside with the various ECM molecules in bicarbonate buffer at 4°C overnight and blocked with 1% BSA for 1 hour at room temperature. Cells were fluorescently tagged with the lipophilic dye DiI (Molecular Probes) used at a final concentration of 5 g/ml for 10-15 minutes at 37°C. The cells were washed, resuspended in RPMI with 0.1% BSA and then added to the upperside of the inserts (2ϫ10 5 cells/insert). In some instances, cells were pre-incubated with blocking antibodies or non-blocking control antibodies and added to the upperside of the FluoroBlok inserts. Migratory behavior of the cells was then monitored at different time-intervals by independent fluorescence detection from the top (corresponding to non-transmigrated cells) and bottom (corresponding to transmigrated cells) side of the membrane using the computer-interfaced SPECTRAFluor Plus instrument (Tecan).

Western blotting
For MT1-MMP, MMP-1, MMP-3, and MMP-11, detection-standardized supernatants and/or cell lysates were loaded onto a sodium dodecyl sulfate (SDS)acrylamide gel (4-15% polyacrylamide), subjected to electrophoresis under reducing conditions and transferred to a nitrocellulose membrane. For ECM production, different adherent cell-types at confluence were treated with 0.1 M NH 4 OH for 5 minutes, then extensively washed and the ECM molecules eventually deposited were extracted with SDS sample buffer. These extracts were loaded onto SDS-acrylamide-gels (8% polyacrylamide) under reducing conditions and blotted onto nitrocellulose filters. The membranes were saturated with Tris-buffered saline (TBS) buffer (containing 20 mM TRIS and 0.15 M NaCl) containing 0.1% Tween 20 (TBST) and 5% non-fat dry milk for 2 hours at room temperature and then incubated with primary antibodies at 4°C overnight. After extensive washing in TBST, the membranes were incubated with horseradish peroxidase (HRP)conjugated appropriate secondary antibodies (Amersham) and then revealed with the ECL Plus chemiluminescence kit (Amersham).

Gelatin zymography
The presence of MMP-2 and MMP-9 in the media was demonstrated by gelatin zymography. The culture media were standardized according to the protein content of cell lysates and were loaded onto SDS-acrylamide-gelatin gels (8% polyacrylamide, 0.1% gelatin). The gels were then washed twice for 30 minutes with 50 mM Tris-HCl (pH 7.4) containing 2.5% Triton X-100, incubated overnight at 37°C in 50 mM Tris-HCl (pH 7.4), containing 150 mM NaCl and 10 mM CaCl 2 , stained for 30 minutes with 30% methanol/10% acetic acid containing 0.5% Coomassie Brilliant Blue R-250, destained, and finally photographed.

Statistical evaluation
Statistical significance of the results was determined by using the Student's t-test. A value of P<0.05 was considered to be significant. Mean ± s.d. values of three to five experiments are presented. For some experiments, data were subjected to repeated-measures analysis of variance (ANOVA), followed when appropriate by a post-hoc Dunnet's test.