Role of metalloproteinases at the onset of liver development


*Author to whom all correspondence should be addressed.


At the onset of liver development, the hepatic precursor cells, namely, the hepatoblasts, derive from the ventral foregut endoderm and form a bud surrounded by a basement membrane (BM). To initiate liver growth, the hepatoblasts migrate across the BM and invade the neighboring septum transversum mesenchyme. In the present study, carried out in the mouse embryo, we searched for effectors involved in this process and we examined the role of matrix metalloproteinases (MMPs). We found expression of a broad range of MMPs, among which MMP-2 was predominantly expressed in the septum transversum and MMP-14 in the hepatoblasts. Using a new liver explant culture system we showed that inhibition of MMP activity represses migration of the hepatoblasts. We conclude that MMPs are required to initiate expansion of the liver during development and that our culture system provides a new model to study hepatoblast migration.


Liver development in mammals starts with the expression of liver-specific proteins in a region of the ventral foregut endoderm. Shortly thereafter, an epithelial liver bud, consisting of liver progenitors (hepatoblasts), forms as an outgrowth of the endoderm. This bud is delineated by a laminin- and collagen IV-rich basement membrane (BM) and is closely associated with endothelial cells, which control hepatoblast proliferation (Zaret 2002; Zhao & Duncan 2005). Then, the hepatoblasts invade the neighboring septum transversum mesenchyme (STM) where they intermingle with mesenchymal and endothelial cells. To invade the STM, the hepatoblasts transit from an epithelial state to a migratory phenotype. This transition is associated with a reduction in the expression of E-cadherin, and is controlled by a gene network that involves the transcription factors prospero-related homeobox 1 (Prox1), hepatocyte nuclear factor-6/Onecut-1 (HNF-6/OC-1) and OC-2 (Sosa-Pineda et al. 2000; Margagliotti et al. 2007). Beyond the stage of STM invasion, liver development further proceeds by proliferation of the hepatoblasts and their differentiation into hepatocytes and biliary cells, and by the formation of hepatocyte cords and bile ducts, which become intimately associated with the liver vasculature (Lemaigre & Zaret 2004; Zhao & Duncan 2005).

How the hepatoblats leave the endodermal bud to invade the STM is not described, and little is known about effectors that directly mediate BM and extracellular matrix remodeling at this stage of liver development. Members of the matrix metalloproteinase (MMP) family are key effectors in extracellular matrix remodeling and tissue invasion. MMPs degrade extracellular matrix components and modulate the function of molecules that mediate cell–cell and cell–matrix interactions. They also activate growth factors and their receptors (Sternlich & Werb 2001; Mott & Werb 2004). Twenty-five MMPs, subdivided into two subgroups, have been described in vertebrates. Secreted MMPs are released into the extracellular medium as latent zymogens (proMMPs), and membrane-type MMPs (MT-MMPs) are expressed in active configuration on the cell surface as transmembrane proteins or as glycosylphosphatidylinositol-anchored proteins (Sternlich & Werb 2001). ProMMPs are activated by proteolytic release of their pro-peptide. This proteolysis is mainly catalyzed by MT-MMPs (Strongin et al. 1995; Itoh et al. 2001), but other nonspecific serine-proteinases and activated MMPs can also be involved (Nagase & Woessner 1999). Moreover, MMP synthesis and activity are modulated by several regulators, which include the extracellular matrix metalloproteinase inducer EMMPRIN (basigin/CD147), a highly glycosylated cell surface transmembrane protein that stimulates MMP synthesis in neighboring cells (Gabison et al. 2005). Finally, MMP activity is reversibly repressed by four tissue inhibitors of metalloproteinases (TIMPs 1–4) (Sternlich & Werb 2001; Lambert et al. 2004).

The study of liver development serves as a paradigm for understanding the development of other organs as well (Zaret 2002). Moreover, liver diseases such as invasion of tissues by hepatocellular carcinoma cells, often recapitulate processes that occur during development. Therefore, in this work we investigated the role of MMPs in the migration of mouse liver bud cells in the STM. We examined which MMPs are present at the onset of liver development, and designed a new culture system to analyze their function.

Materials and methods


CD1 mice were raised in our animal facilities and were treated according to the principles of laboratory animal care of the local Animal Welfare Committee.

Reverse transcription – polymerase chain reaction

Total RNA preparation and reverse transcription were as described [4]. For polymerase chain reaction (PCR) amplification, 3 µL of the reverse transcription mix were incubated in a volume of 50 µL containing Tris-Cl 10 mm (pH 9.0), KCl 50 mm, Triton X-100 0.1% (v/v), MgCl2 1.5 mm, dNTPs 40 µm, 10 pmol of each primer, and 2.5 units of Taq polymerase (Promega Corporation, Madison, WI, USA). Primer sequences were 5′-AGATCTTCTTCTTCAAG GACCGGTT-3′ and 5′-GGCTGGTCAGTGGCTTGGG GTA-3′ for MMP-2, 5′-GATCTCTTCATTTTGGCCATCT CTTC-3′ and 5′-CTCCAGTATTTGTCCTCTACAAAGAA-3′ for MMP-3, 5′-CCCACATTTGACGTCCAGAGAAGAA-3′ and 5′-GTTTTTGATGCTATTGCTGAGATCCA-3′ for MMP-9, 5′-TCCCGAGCCTGAATTTCAT-3′ and 5′-AG CCTCATAGGCAGCATCTAA-3′ for MMP-10, 5′-ATTTG GTTCTTCCAAGGTGCTCAGT-3′ and 5′-CCTCGGAA GAAGTAGATCTTGTTCT-3′ for MMP-11, 5′-ATGATCT TTAAAGACAGATTCTTCTGG-3′ and 5′-TGGGATAAC CTTCCAGAATGTCATAA-3′ for MMP-13, 5′-GGATAC CCAATGCCCATTGGCCA-3′ and 5′-CCATTGGGCAT CCAGAAGAGAGC-3′ for MMP-14, 5′-GGTACATGT GAAAGCCAACCT-3′ and 5′-GTACCAGCCCAGCT TCTCAG-3′ for MMP-15, 5′-GTAACTCCAAAAGTTGG AGATCCTG-3′ and 5′-TCTAATTCACTGTAGGGAACT TCTTCAA-3′ for MMP-16, 5′-GCCGGGATACTGTGCGT- 3′ and 5′-CTACCTCGTGGAAGTTCAAGG-3′ for MMP-17, 5′-TGGGCCACTGGAGAAAGAAG-3′ and 5′-TCA GCCCAACCAGCTTTCAC-3′ for MMP-19, 5′-TTCCC CATTCAGTTTCCGTG-3′ and 5′-AAGAAAGCGTGGG CCAGTT-3′ for MMP-23, 5′-AGGCTATTCGTCAGGCT TTC-3′ and 5′-TTTTGATCTCATGGTATGGCA-3′ for MMP-24, 5′-GGCATCCTCTTGTTGCTATCACTG-3′ and 5′-GTCATCTTGATCTCATAACGCTGG-3′ for TIMP-1, 5′-CTCGCTGGACGTTGGAGGAAAGAA-3′ and 5′-AGC CCATCTGGTACCTGTGGTTCA-3′ for TIMP-2, 5′-CTT CTGCAACTCCGACATCGTGAT-3′ and 5′-CAGCAGG TACTGGTACTTGTTGAC-3′ for TIMP-3, and 5′-ACTT GCTATGCAGTGCCATG-3′ and 5′-TCGGTACCAGCT GCAGATG-3′ for TIMP-4 (Maquoi et al. 2002).


Tissue blocks comprising the liver bud and STM were manually dissected from three embryonic day (E)9 mouse embryos and were pooled and homogenized in a solution containing Tris-Cl 50 mm (pH 7.5), CaCl2 10 mm, NaCl 150 mm, ZnCl2 1 mm, NaN3 3 mm and Triton X-100 0.1%. The homogenates were diluted in nonreducing sodium dodecyl sulfate (SDS) 2% (w/v) sample buffer and electrophoresed on a 8% polyacrylamide SDS gel containing 0.05% (w/v) of gelatin. After electrophoresis, the gel was washed at room temperature three times for 20 min in Triton X-100 2.5% (v/v) to remove SDS and incubated overnight at 37°C in Tris-Cl 50 mm buffer (pH 7.5), containing NaN3 3.5 mm, CaCl2 5 mm, ZnCl2 3.5 mm and Triton X-100 1% (v/v). After incubation, the gel was stained for 30 min with G-250 Coomassie Blue 0.1% (w/v) in methanol 45% (v/v), acetic acid 10% (v/v), and destained in the same solution without dye.

MMP-2 and MMP-14 activity assay

MMP-2 and MMP-14 activities were measured in manually dissected tissue blocks (liver bud plus STM), using the MMP-2 and the MMP-14 activity assay systems Biotrack (Amersham, Little Chalfont, United Kingdom, RPN 2631 and RPN 2637). Each tested sample contained a pool of three blocks. MMP activities were converted to MMP concentrations according to calibration curves.

In situ hybridization

Embryos were fixed overnight at 4°C in ethanol 60%, formaldehyde 30% and acetic acid 10%, paraffin embedded and sectioned (16 µm). Biotin-labeled antisense RNA probe for EMMPRIN (mRNA sequence 890–1322 nt, GenBank accession number NM_009768.6) was produced by RT–PCR followed by in vitro transcription. Dilution of antibiotin antibody coupled with alkaline phosphatase was 1:250 (Fab fragment, Roche R & D, Basel, Switzerland). Labeling was detected with 4-nitroblue tetrazolium chloride/5-bromo-4-chloro-3-indolyl-phosphate (NBT/BCIP) (Roche) (Jacquemin et al. 2003) and the section was counterstained with aqueous eosin 1% (w/v).

Culture of hepatic buds

Embryos at E9 were soaked in phosphate-buffered saline (PBS) and a tissue block comprising the liver bud, the STM and a portion of the foregut was dissected using tungsten needles with a dissection microscope. The tissue blocks were cultured in rolling bottles at 37°C, in an atmosphere containing O2 5%, CO2 5% and N2 90% (Pierreux et al. 2004), and in Dulbecco's Modified Eagle's Medium (DMEM) (4500 mg/L glucose) supplied with fetal bovine serum 10% (BioWhittaker, Walkersville, MD, USA), bovine albumin 5% (BD Biosciences, Erembodegem, Belgium), L-Glutamine 2 mm (Gibco Invitrogen, Merelbeke, Belgium), penicillin 50 Units/mL, streptomycin 50 mg/mL, and fungizone 2.5 mg/mL (Gibco Invitrogen, Merelbeke, Belgium). The cultures were treated with the specific MMP-2 Inhibitor I (Calbiochem-EMD Chemicals, Gibbstown, NJ, USA) at 6,7 µm, with the broad-spectrum MMP inhibitor GM6001 (Calbiochem) at 20 µm, or with dimethylsulfoxide (DMSO) 0.1% (v/v) as control.


Immunodetection was carried out as described by Margagliotti et al. (2007). Primary antibodies and dilutions were: goat anti-HNF-4α at 1:100 (Santa Cruz Biotechnology, Santa Cruz, CA, USA), rabbit antilaminin at 1:100 (Sigma-Aldrich, Bornem, Belgium), rat anti-PECAM at 1:50 (PharMingen–EMD Chemicals, Gibbstown, NJ, USA), goat anti-GATA-4 at 1:50 (Santa Cruz), rabbit anti-Prox1 at 1:500 (Covance, Princeton, NJ, USA), rabbit anti-Hex at 1:200 (gift from C. Bogue), rabbit anti-MMP-2 at 1:100 (Chemicon Millipore, Brussels, Belgium) and mouse anti-MMP-14 at 1:50 (Calbiochem). Secondary antibodies and dilutions were donkey antigoat/AlexaFluor 594 at 1: 2000, donkey antirabbit/AlexaFluor 594 at 1:3000, donkey antirabbit/AlexaFluor 488 at 1:1000, and antimouse horse radish peroxidase (HRP) at 1:100 (Molecular Probes-Invitrogen, Merelbeke, Belgium). The antimouse HRP was detected with the Tyramide Signal Amplification (TSA) kit (Molecular Probes). Quantification of migrating hepatoblasts was carried out on three sections for each explant, by calculating the ratio of HNF-4α positive nuclei having invaded the STM over total HNF-4α-positive nuclei.


Disruption of the basement membrane surrounding the liver bud

To investigate the migration of hepatoblasts from the liver bud into the STM, we first determined the timing of disruption of the BM surrounding the liver bud. We analyzed transverse sections of mouse embryos at the 16, 18, 21 and 23-somite stages (i.e. from E9 to E9.5) for the presence of two constituents of the BM, namely laminin (Fig. 1 and Shiojiri & Shugiyama 2004) and collagen IV (not shown). HNF-4α was used as a hepatoblast marker (Duncan et al. 1994; Watt et al. 2003). The results showed that BM disruption starts at the 18-somite stage. Hepatoblast migration into the STM starts at the 21-somite stage, and this is initiated at the lateral edges of the liver bud before extending to the ventral edge at the 23-somite stage.

Figure 1.

Hepatoblast migration and basement membrane disruption at the onset of liver development. (A–D) The hepatoblasts, detected by staining for hepatocyte nuclear factor-4α (HNF-4α), invade the septum transversum mesenchyme (STM) starting at the 21-somite stage. (A′–D′) The basement membrane (BM), detected by staining for laminin, starts to be degraded at the 18-somite stage (arrows). Bars, 50 µm. lb, liver bud; *indicates liver bud lumen.

MMP expression and activity in the developing liver bud

We have recently studied the expression of genes related to extracellular biology at the onset of liver development, by resorting to an array analysis (Margagliotti et al. 2007). This approach showed that, among the 113 genes tested, MMP-2 and MMP-14 (also known as MT1-MMP) were the only matrix degrading enzyme detectable at the mRNA level at that stage of liver development. Thus, we examined the possibility that these two MMPs are involved in BM degradation and hepatoblast migration.

We first verified if MMP-2 and MMP-14 activity could be detected. Three 20-somite stage dissected livers were pooled and protein extracts were analyzed by in gel gelatin zymography. Strong proMMP-2 activity was visualized (Fig. 2A). Since the active form of MMP-2 was not detected, an enzyme linked immunosorbent assay (ELISA) designed to measure the activity of endogenously activated MMP-2 and total MMP-2, was used (Fig. 2B). The activity of proMMP-2 was calculated as the difference between total MMP-2 and activated MMP-2. The results (Fig. 2B) showed that active MMP-2 is detected, but at lower concentrations than proMMP-2. Moreover, since proteolytic activation of MMP-2 is usually catalyzed by the membrane-type MMP-14 (Strongin et al. 1995; Butler et al. 1998; Deryugina et al. 2001), and since MMP-14 mRNA was detected in our earlier array analysis, we also measured the activity of MMP-14 by ELISA. MMP-14 was detected, with the highest concentration at the early stages (17-somite stage), in effect, at the period correlating with the onset of BM degradation (see Fig. 1).

Figure 2.

Matrix metalloproteinase-2 (MMP-2) and MMP-14 activity at the onset of liver development. (A) Detection by gelatin zymography of MMP-2 activity in liver bud and septum transversum mesenchyme (STM) at the 20-somite stage (E9). Only proMMP-2 is detectable in the tested sample (t), as compared with a standard reference (r). (B) Detection and quantification by enzyme linked immunosorbent assay (ELISA) of MMP-2 and MMP-14 activities at the 17-, 25- and 33-somite stages (S) in liver bud and STM. Data are means ± SEM (n = 3). MMP-14, as well as pro-MMP-2 and active MMP-2 are detectable by ELISA.

The experiments described above demonstrated the presence of MMP activity in dissected tissues comprising both the liver bud and the STM. To identify the cell types expressing MMP-2 and MMP-14, we localized total MMP-2 and MMP-14 by immunofluorescence staining. As shown in Figure 3, MMP-2 was predominantly expressed in the STM, whereas HNF-4α-positive hepatoblasts expressed low levels of this MMP (Fig. 3A–C). In contrast, MMP-14 was absent from the STM, but was present in hepatoblasts, which were identified in this experiment by the expression of the transcription factor Prox1 (Sosa-Pineda et al. 2000) (Fig. 3D–F).

Figure 3.

Expression of matrix metalloproteinases (MMPs) and EMMPRIN in developing liver at the 20- to 24-somite stages. (A–C) hepatocyte nuclear factor-4α (HNF-4α) marks the hepatoblasts and MMP-2 is predominantly expressed in the septum transversum mesenchyme (STM), as shown by immunofluorescent staining of liver bud sections. (D–F) MMP-14 is expressed in hepatoblasts, which are identified by staining for Prox1. (G) EMMPRIN mRNA is detected by in situ hybridization predominantly in hepatoblasts at the 24-somite stage (E9.5); a weaker expression is also detected in the adjacent mesenchyme. (H) Reverse transcription–polymerase chain reaction (RT–PCR) detection of MMP and tissue inhibitors of metalloproteinase (TIMP) transcripts in liver bud and STM at the 20-somite stage (E9). Bars, 50 µm. d, duodenum; lb, liver bud; *indicates liver bud lumen.

Finally, since EMMPRIN is known to stimulate the production of both MMP-2 and MMP-14 (Kanekura et al. 2002; Taylor et al. 2002), we also analyzed its expression on liver bud sections. Using in situ hybridization, we found that EMMPRIN is strongly expressed in hepatoblasts (Fig. 3G), and more weakly in the adjacent mesenchyme.

From this set of experiments we concluded that MMP-2 and MMP-14 are active at the onset of liver development, when the BM surrounding the liver bud becomes degraded and when hepatoblasts migrate into the STM. Predominant expression of MMP-2 is found in the STM, and MMP-2 can be activated by MMP-14, which is expressed by the hepatoblasts. The production of MMP-2 and MMP-14 could be promoted by EMMPRIN.

MMPs are necessary for hepatoblasts migration

The previous results led to the assumption that MMP-2 and MMP-14 are required for extracellular matrix remodeling and migration of the hepatoblasts. To address this possibility, an ex vivo model of liver bud culture in which hepatoblasts migrate out of the liver bud and invade the STM was set up. Embryos were dissected and a tissue block that includes a portion of the foregut, the liver bud and the STM was isolated. The embryos were collected at the 16- to 17-somite stage, before the BM starts to be degraded. The dissected tissue was then cultured in liquid medium in a rolling bottle. To validate the culture conditions, cell migration and BM degradation in the explants were analyzed by immunostaining after 12 h and 20 h of culture (Fig. 4). Sections were stained with antibodies against HNF-4α, Prox-1 or Hex to identify the hepatoblasts (Duncan et al. 1994; Bogue et al. 2000; Sosa-Pineda et al. 2000; Watt et al. 2003; Hunter et al. 2007), and the BM was detected using antilaminin antibodies. Disruption of the BM and invasion of the mesenchyme by the hepatoblasts were detected after 20 h of culture (Fig. 4A–B). Moreover, GATA-4 and PECAM expression, which, respectively, mark the STM and the endothelial cells surrounding the liver bud (Matsumoto et al. 2001; Zhao et al. 2005), remains detectable during the culture period, indicating that the culture system maintains the integrity of the tissues and closely mimics in vivo liver development (Fig. 4C–D).

Figure 4.

Matrix metalloproteinases (MMPs) are required for hepatoblast migration. (A, B) Analysis of the cultured liver bud morphology by immunofluorescent labeling of laminin (BM) and hepatocyte nuclear factor-4α (HNF-4α) (hepatoblasts). Hepatoblasts have started to invade the septum transversum mesenchyme (STM) after 20 h of culture, but not after 12 h (C, D) Labeling for GATA-4 and PECAM demonstrated that the STM and the endothelial cells are maintained. The transcription factors Prox1 and Hex are used as hepatoblast markers. (E–F) The broad spectrum MMP inhibitor GM6001 inhibit hepatoblast migration. The analysis of the liver bud morphology was carried out by immunofluorescent labeling of laminin (E–F, E′–F′) and HNF-4α (E–F). Bars, 50 mm. lb, liver bud; *indicates liver bud lumen.

The availability of the liver bud culture system allowed us to investigate the role of MMPs at the onset of liver development. To this end, dissected liver buds were cultivated for 20 h in the presence of the specific MMP-2 inhibitor I. This did not result in inhibition of hepatoblasts migration (not shown); suggesting that other MMPs may act redundantly with MMP-2. Our earlier work (Margagliotti et al. 2007) had only detected MMP-2 and MMP-14 by an array approach. Therefore we used a more sensitive RT–PCR approach. In addition to MMP-2 and MMP–14, we could now detect by RT–PCR, transcripts for the secreted MMP-11 and MMP-19, for the membrane-type MMP-15, MMP-16, MMP-17 (also known as MT2-, MT3- and MT4-MMP, respectively), MMP-23 and MMP-24, as well as for TIMP-2 and TIMP-4 (Fig. 3H). These transcripts had not been detected in our earlier work, due to the lower sensitivity of the array analysis as compared with RT–PCR. Therefore, the broad-spectrum MMP inhibitor GM6001 was added to cultured liver buds. The degradation of the BM was not affected in the presence of GM6001 as compared with controls. In contrast, this treatment resulted in inhibition of hepatoblast migration from the liver bud to the STM (Fig. 4E-F,E′-F′). Hepatoblast migration was quantified by calculating the percentage of hepatoblasts that have invaded the STM after 20 h of culture. We found that this ratio was significantly decreased in the presence of GM6001 (36 ± 7%) as compared with controls (76 ± 4%; mean ± SEM; n = 4; P = 0.0036, Student's t-test). Taken together, these results suggest that MMPs play important roles in the migration of hepatoblasts from the liver bud to the STM.


The migration of hepatoblasts out of the hepatic bud allows the organ to expand and to initiate interactions with mesenchymal or endothelial cells (Zaret 2002; Zhao & Duncan 2005). Work by others in chick embryos suggests the need for as yet unidentified chemoattractants (Tatsumi et al. 2007). In the present study, we investigated the role of MMPs in this process. Our analyses established that MMP-2 is expressed predominantly in the STM and that MMP-14 is present at the hepatoblast membrane. In addition, expression of MMP regulators, such as TIMP-2, TIMP-4, and EMMPRIN, was detected.

There is little information available about MMP expression and function at early stages of liver development. Nuttal et al. (2004) analyzed the RNA levels of metalloproteinases in developing liver, starting at E15.5, in effect, at a later stage than our analysis. Their expression profile broadly overlaps with that described in the present paper. However, our work now provides functional evidence for a role of MMPs in early liver expansion.

Interestingly, mice deficient in the transcription factor Prox-1 show a lack of migration of hepatoblasts in the STM (Sosa-Pineda et al. 2000) and Prox-1 can stimulate MMP-2 expression in cultured hepatoblasts (Papoutsi et al. 2007). These data suggest that MMP-2 contributes to extracellular matrix remodeling during liver development and is a member of the gene network that controls hepatoblasts migration (Sosa-Pineda et al. 2000; Margagliotti et al. 2007). We note, however, that MMP-2 expression in hepatoblasts is weaker than in the STM, suggesting that the predominant role of MMP-2 is played by STM-produced MMP-2.

The inability of the MMP-2-specific inhibitor to suppress hepatoblasts migration ex vivo suggests that other MMPs are involved. This is in line with the observation that inactivation of MMP genes only induces mild defects (Vu & Werb 2000; Page-McCaw et al. 2007). This is also consistent with the expression of several MMPs during liver development, and with our finding that a broad-range MMP inhibitor GM6001 represses hepatoblast migration. The addition of broad-range MMP inhibitors repressed hepatoblast migration, but did not significantly affect BM degradation. This suggests that other matrix-degrading enzymes play a role in the latter process.

The hepatoblast-restricted distribution of MMP-14 suggests that it controls extracellular matrix remodeling in the near vicinity of the hepatic cells. MMP-14 could exert this effect directly or via activation of proMMP-2 (Stetler-Stevenson et al. 1993; Seiki et al. 2003; Itoh & Seiki 2006). Moreover, TIMP-2, a MMP-14 co-activator involved in MMP-2 activation, was also detected by RT–PCR at the onset of liver development (Fig. 3H). Once hepatoblasts have invaded the STM, MMP-14 may activate proMMP-2 produced by the STM cells, thereby allowing further extracellular matrix remodeling and hepatoblast migration. During this process, EMMPRIN, which is a stimulator of MMP production, may be involved.

Finally, MMP-2 and MMP-14 expression in hepatocellular carcinoma is high and correlates with tissue invasion and metastasis (Théret et al. 1998; Koshikawa et al. 2000; Giannelli et al. 2001; Ogasawa et al. 2005). In this work we developed a new culture system in which hepatic epithelial cells, mesenchymal cells and vascular cells interact to allow the hepatoblasts to migrate through the BM and invade the STM. This culture system provides a new tool to study developmental processes and carcinogenesis, and for development of pharmacological compounds.


The authors thank K. Zaret for discussions at the initiation of this project, and A. Berton and D. Delvaux for help. The work was supported by grants from the Interuniversity Attraction Poles Program (Belgian Science Policy), from the D.G. Higher Education and Scientific Research of the French Community of Belgium, and from the Fund for Scientific Medical Research to F.L. S.M. held a fellowship from the Fonds pour la formation à la Recherche dans l’Industrie et dans l’Agriculture, and from the Université Catholique de Louvain. F.C., C.P. and P.H. are Research Associates of the F.R.S.-FNRS.