Deficiency of β1 Integrins in Teratoma Interferes with Basement Membrane Assembly and Laminin-1 Expression

EXPERIMENTAL CELL RESEARCH ARTICLE NO.

238, 70–81 (1998)

EX973837

Deficiency of b1 Integrins in Teratoma Interferes with Basement Membrane Assembly and Laminin-1 Expression Takako Sasaki,* Erik Forsberg,* Wilhelm Bloch,† Klaus Addicks,† Reinhard Fa¨ssler,* and Rupert Timpl*,1 *Max-Planck-Institut fu¨r Biochemie, 82152 Martinsried, Germany; and †Institute for Anatomy, University of Cologne, 50931 Ko¨ln, Germany

dent networks, one containing collagen IV and the other laminins, and they also contain some connecting elements such as nidogen (entactin) as well as several proteoglycans and other proteins [2]. Apart from their involvement in supramolecular matrix interactions, several basement membrane proteins, in particular most of the laminin isoforms, collagen IV and perlecan, also bind strongly to integrin receptors, mainly of the b1 chain subfamily [3]. Some of these receptors are dual collagen/laminin receptors (a1b1, a2b1) and some are specific for various laminin isoforms (a3b1, a6b1, a7b1, a9b1, a6b4). Laminin also interacts with dystroglycan receptors, which, like integrins, provide connections to the actin cytoskeleton [4]. It is not known how basement formation is initiated but nucleation may occur through cellular receptors as shown for fibronectin microfibrils [5, 6]. The importance of such interactions has been shown in embryonic organ cultures by antibody perturbation of either a6b1 or a-dystroglycan binding to laminin-1 [7, 8] or the laminin-nidogen interaction [9, 10]. These treatments interfered with the formation of novel basement membranes and resulted in massive cell necrosis. Several of the major basement membrane proteins are produced in very early embryonic stages [11, 12] and are constitutively expressed in cultured embryonic stem (ES) cells. ES cells form benign teratomas when injected ectopically into syngeneic recipient mice. Teratoma growth progresses by differentiation along several cell lineages and is accompanied by the formation of abundant basement membranes [13]. This makes ES cells another suitable model for studying the essential features of basement membrane assembly. A complete knock-out of the b1 integrin chain gene has recently been described for ES cells [14] and a teratocarcinoma cell line [15], and some aberrant features were observed when these cells were examined in culture. Furthermore, mice carrying a null mutation in the b1 integrin gene showed an early lethality at around a stage when the first basement membranes form in the inner cell mass [16, 17]. Yet studies with chimeric mice also demonstrated that integrin b1-negative ES cells can still survive and participate in organ development [16]. Such cells are, however, deficient in several respects,

Subcutaneous injection of b1 integrin-deficient embryonic stem cells in mice causes the formation of teratomas although they occur with a lower frequency and are smaller than wild-type cells. Immunofluorescence analysis of these deficient tumors indicates a disorganized deposition of several basement membrane proteins. This was confirmed by electron microscopy which demonstrated frequent gaps in cell-associated basement membranes or loss of close contacts to the cells. Further aberrant features were multilaminar structures and amorphous deposits, indicating a strong impairment of correct basement membrane assembly. Quantitative radioimmunoassays were used to determine the levels of specific proteins in successive tissue extracts with neutral buffer in the absence and presence of EDTA and with 6 M guanidine. This demonstrated a more than 90% decrease in the content of laminin-1 (a1b1g1) and a 70% decrease in nidogen in the b1 integrin-deficient teratomas. No significant changes were detected for other matrix proteins (perlecan, fibronectin, fibulins). This selective change impaired the formation of laminin-nidogen complex and enhanced nidogen degradation. Northern blots also demonstrated a distinctly reduced expression of laminin a1, b1, and g1 chains. Similar reductions were also observed in cultured embryonic stem cells prior to any differentiation. No or only smaller changes were observed for laminin a2 and b2 chain, nidogen, and perlecan mRNA. These data emphasize a distinct role of b1 integrins in the correct assembly of basement membranes which may occur through direct ligand binding and/or regulatory events at the transcriptional level. q 1998 Academic Press

INTRODUCTION

Basement membranes are thin extracellular matrices deposited in close contact with cells and are involved in the maintenance and regulation of cellular phenotypes [1]. They are constituted from two indepen1 To whom correspondence and reprint requests should be addressed at Max-Planck-Institut fu¨r Biochemie, D-82152 Martinsried, Germany. Fax: 089/8578 2422.

70

0014-4827/98 $25.00 Copyright q 1998 by Academic Press All rights of reproduction in any form reserved.

AID

ECR 3837

/

6i2d$$$$$1

12-16-97 13:55:15

ecal

INTEGRIN DEFICIENCY AND BASEMENT MEMBRANE ASSEMBLY

as shown for immune cell precursors which could potentially differentiate but failed to migrate to appropriate developmental places [18], heart muscle cells which formed sarcomers that became increasingly unstable [19], and keratinocytes that failed to differentiate in embryoid bodies [20]. A more recent study showed that b1 integrin-deficient ES cells were still capable of forming teratomas but with a much lower efficiency than wild-type cells. This is apparently due to a low level of angiogenesis since in both cases proliferation capacity and rate of apoptosis were similar [21]. Here we use this model for studying basement membrane assembly by electron microscopy and biochemical analyses and have demonstrated an aberrant basement membrane morphology and a selective down-regulation of laminin-1 expression. These two observations are consistent with one another and also indicate that b1 integrin deficiency may cause defects at different molecular and/or cellular levels. MATERIALS AND METHODS Teratoma induction. For the production of teratomas the wildtype ES cell line D3 [22] and the b1 integrin-deficient ES cell line G201 were used [14]. Both ES cell lines were routinely cultured in the absence of a fibroblast feeder layer in Dulbecco’s minimal essential medium supplemented with 20% heat-inactivated fetal calf serum (GIBCO BRL, Gaithersburg, MD), 0.1 mM b-mercaptoethanol (Sigma Chemical Co., St. Louis, MO), 11 nonessential amino acids (GIBCO BRL), and 1000 U/ml recombinant leukemia inhibiting factor (GIBCO BRL). ES cells (1 1 107) were trypsinized, washed two times, suspended in 100 ml PBS, and injected subcutaneously on the back of syngeneic 129/Sv male mice. After 21 days tumors were surgically removed and frozen in ice-cold isopentane. Some of the teratomas were fixed in 4% paraformaldehyde and stained with hematoxylin-eosin following a previous protocol [16]. Electron microscopy. Teratoma tissue was obtained from 6-weekold animals which were killed by cervical dislocation and subsequently transcardially perfused with a 0.1 M phosphate-buffered saline (PBS), pH 7.3, containing 4% paraformaldehyde and 2% glutaraldehyde at a perfusion pressure of 60 cm H2O. Teratomas were removed and fixed for an additional 4 h in the same fixative. Afterward tumors were cut into small pieces and fixed in 0.1 M PBS containing 2% osmium tetroxide for 2 h at 47C, rinsed three times in PBS, block-stained for 8 h in 70% ethanol, and embedded in araldite. Ultrathin sections (30–60 nm) obtained with a diamond knife on a Reichert ultramicrotome were placed on copper grids and examined with a Zeiss EM 902A electron microscope. Tissue extraction and protein analysis. Teratoma tissue was obtained from several animals and extracted at 47C in the presence of protease inhibitors following a previously established protocol [23]. First, samples were homogenized twice in 0.05 M Tris-HCl, pH 7.4, 0.15 M NaCl (TBS) using 10 ml/g wet tissue and immediately centrifuged. This was followed by two extractions with TBS containing 10 mM EDTA (5 ml/g) for 1 and 24 h, respectively, followed by two extractions in 6 M guanidine (5 ml) for 24 h each. The residue was washed twice with water and suspended in 0.1 M acetic acid to allow quantitation of its protein content. Protein concentrations were determined from hydrolyzed aliquots (6 M HCl, 16 h, 1107C) on a LC 3000 amino acid analyzer. SDS gel electrophoresis in 5–15% polyacrylamide gradient gels followed established protocols. Molecular sieve chromatography on Sepharose Cl-6B [23] or on Superose 6 (HR16/50) was done in TBS, 10 mM EDTA. Recovery of laminin and nidogen was about 80%, as determined by radioimmunoassay.

AID

ECR 3837

/

6i2d$$$$$2

12-16-97 13:55:15

71

Northern blotting. Total RNA from teratoma tissue was isolated by acid guanidium thiocyanate-phenol-chloroform extraction [24]. Total RNA (20–25 mg) was separated by electrophoresis on a 1% agarose gel, transferred to a Zeta Probe membrane (Bio-Rad Laboratories), and UV cross-linked. The membrane was hybridized with 32 P-labeled DNA probes prepared using a random primer labeling kit (Stratagene, La Jolla) in rapid hybridization solution (Stratagene) following the supplier’s instructions. The following DNA probes were used: mouse laminin a1 cDNA (nt 5903–6645) [25], mouse laminin b1 cDNA (nt 4883–5544) [26], mouse laminin g1 cDNA (nt 4355– 5020) [27], mouse laminin a2 cDNA (nt 6345–6839) [28], mouse laminin b2 cDNA (29 bp exon 29–exon 33) [29], mouse nidogen cDNA (170 bp exon 3–370 bp exon 4) [30], mouse perlecan cDNA (nt 4392– 5583) [31], mouse fibulin-2 cDNA (nt 2081–3181) [32], and rat GAPDH cDNA [33]. Filters were then exposed to X-ray film in order that strength of the individual signals could be judged. The relative radioactivity of bands was, however, quantified with a Bio-Imaging Analyzer BAS1000 (Fuji Photo Film Co., Japan). Quantitative densitometry was done here when necessary at different exposure times in order to assure linearity of the Northern signals. Northern blots of ES cells were those used previously [14]. Immunological assay. Radioimmuno-inhibition assays with specific rabbit antisera were used for the quantitation of individual proteins and established protocols were followed [34]. All samples were analyzed at three to four different dilutions and the average values showed a standard deviation of õ20%. The assay for laminin-1 was based on labeled laminin-1 fragment P1 and an antiserum against P1 and has previously been shown to detect epitopes of the laminin a1, b1, and g1 chains in a similar fashion [35]. Since this assay is based on the laminin fragment most stable against proteolysis it would also detect laminin fragments. The assays for nidogen [36], fibulin-1C [37], and fibulin-2 [32] were those described for these recombinant mouse proteins. The assay for mouse perlecan was based on its recombinant fragment III-3 and was calibrated with tissuederived perlecan as a reference inhibitor [38]. An assay with 125Ilabelled mouse fibronectin and an antiserum against human fibronectin was used for the quantitation of fibronectin. Commercially available mouse fibronectin (Chemicon International Inc.) was used as reference inhibitor. Immunoblotting of proteins followed a previously described procedure [39]. For immunofluorescence tumors were cut into 6-mm sections and incubated with specific rabbit antisera (1 h, room temperature) diluted in PBS, 1% bovine serum albumin. Sections were then washed three times with PBS and incubated (1 h, room temperature) with fluorescein-labeled goat anti-rabbit IgG at dilutions recommended by the manufacturer (Jackson Immunoresearch Lab.). After a further three washes with PBS, specimens were mounted using elvanol and examined with an Axiophot fluorescence microscope (Zeiss). Antibodies against b1 integrin and collagen III were the same as those used in previous studies [14, 34]. Antibodies specific for mouse laminin a1 chain were raised against recombinant LG4 module of fragment E3 (Z. Andac and R. Timpl, unpublished).

RESULTS

Deposition of Basement Membrane Proteins Is Altered in Integrin b1-Deficient Teratomas Subcutaneous injection of either wild-type or b1 integrin-deficient ES cells caused the formation of teratomas but in the latter case the average weight was about 90% lower. Both teratomas showed the same characteristic differentiation into several cell lineages after 15– 21 days, as revealed by histology [13]. This included differentiation of cells derived from all three cell layers such as chondrocytes, muscle cells, and epithelial and neuronal cells which could be clearly identified in both

ecal

72

SASAKI ET AL.

FIG. 1. Morphological comparison of wild-type (A) and b1 integrin-deficient (B) teratomas by hematoxylin-eosin staining. Well-formed multilayer or glandular epithelium is marked by arrows. Arrowheads show irregular epithelial regions which were exclusively found in the mutant teratoma. Bar: 50 mm.

teratomas (data not shown). Major differences, however, were noticed in the arrangement of several epithelial cells (Fig. 1). In the wild-type teratoma they formed multilayered nonkeratinizing epithelial sheets as well as glandular structures composed of cuboidal and columnar epithelial cords (Fig. 1A). Epithelial cells are still abundant in b1 integrin-deficient teratomas but have lost their cuboidal shape and are arranged in irregular layers (Fig. 1B). Immunofluorescence analysis of b1 integrins demonstrated a high expression through all different compartments of a wild-type teratoma section while a much more restricted, spot-like staining was observed in the b1 integrin-deficient teratoma (Fig. 2A). These latter b1 integrin-positive regions are due to invading host cells [21]. Staining for the basement membrane proteins laminin (Figs. 2B and C), nidogen (Fig. 1D), and perlecan (Fig. 1E) revealed in the wild-type teratoma mainly broad cord-like deposits which, although abundant, did not cover the whole tissue section. This was changed to a much more diffuse and less continuous distribution in the b1 integrin-deficient teratoma. In addition, the fluorescence intensity of laminin a1 staining was distinctly lower in the mutant compared to the

wild-type teratoma (Fig. 2B). A more similar intensity was, however, observed for laminin g1 chain staining (Fig. 2C). Staining for collagen III as a marker of interstitial connective tissue was, however, more similar in both teratomas (Fig. 1F). Basement Membranes Are Disorganized in Integrin b1-Deficient Teratomas To test whether the ultrastructural morphology of basement membranes is affected by the absence of b1 integrins, electron microscopy was performed with tissues derived from normal and mutant tumors. In normal tumors the basement membrane was continuously present along the basal surface of epithelial cells. The width of the lamina lucida was never greater than 30 nm (Fig. 3A). In contrast, all epithelial cells in mutant teratomas were covered by a basement membrane with an enlarged lamina lucida of irregular width (Fig. 3B). In several areas the basement membrane was partially detached from the cell surface which led to the formation of loops (Fig. 3C). In other areas the detached basement membranes formed a multilayered lamina (Fig. 3E) or produced condensed structures which had lost

FIG. 2. Indirect immunofluorescence staining of wild-type (///) and b1 integrin-deficient (0/0) teratomas. Staining was with antibodies against b1 integrin subunit (A), against laminin a1 chain (B), against laminin g1 chain (C), against nidogen (D), against perlecan (E), and against collagen III (F). Inoculation with wild-type cells was done with a 10-fold smaller number (106 cells/site) in order to obtain about equal sizes for both tumors. Bar: 50 mm.

AID

ECR 3837

/

6i2d$$$$$3

12-16-97 13:55:15

ecal

INTEGRIN DEFICIENCY AND BASEMENT MEMBRANE ASSEMBLY

AID

ECR 3837

/

6i2d$$3837

12-16-97 13:55:15

ecal

73

74

SASAKI ET AL.

FIG. 3. Electron microscopical analysis of basement membrane structure of wild-type (A) and b1 integrin-deficient teratomas (B–E). Note in the wild type a smooth basement membrane strictly located in close vicinity to the cell surface. Many abnormalities of basement membranes were observed in the mutant teratoma including partial detachment from cells (B, arrowheads), gaps (B, arrows), involutions without cell contact (C, arrowheads), large amorphous matrix deposits (D, arrowheads), and multilayer deposits close to cells (E, arrowheads). Bar: 150 nm.

AID

ECR 3837

/

6i2d$$3837

12-16-97 13:55:15

ecal

75

INTEGRIN DEFICIENCY AND BASEMENT MEMBRANE ASSEMBLY

TABLE 1 Distribution of Total Proteins and 4-Hydroxyproline in Successive Extracts of Teratomas Derived from b1 Integrin /// or 0/0 Embryonic Stem Cellsa Extracts Parameter analyzed

b1 type, (content)

Total protein

/// 0/0 /// 0/0 /// 0/0 /// 0/0

4-Hydroxyproline

(mg/g) (mg/g) (%) (%) (mg/g) (mg/g) (%) (%)

TBS

TBS/E

Gu

Insoluble residue

Sum of all

44.0 52.9 68 93 167 202 38 51

2.2 1.4 3 2.7 õ1 9 õ1 2

19.0 2.4 29 4 267 174 62 45

õ0.02 0.2 õ0.1 0.3 õ1 8 õ1 2

65.2 56.9 100 100 434 393 100 100

a Both parameters were determined in triplicate hydrolysates on the amino acid analyzer and are expressed as mg or mg per g wet tissue and as % of the total amount recovered. The extracts were obtained with neutral buffer (TBS), TBS containing EDTA (TBS/E) and 6 M guanidine (Gu).

contact with cells (Fig. 3D). None of these abnormali- brane and other extracellular matrix proteins by radioties were observed in normal teratomas. In addition, immunoassays (Table 2). This demonstrated that the similar basement membrane detachments were also content of laminin-1 was about 1% of total proteins (comobserved in mutant teratomas for muscle cells and nests of neuronal cells (data not shown). This indicates a general basement membrane defect rather than being TABLE 2 restricted to certain subtypes of cells. Quantitative Radioimmunoassay Analysis of Individual Selective Deficiency in the Synthesis and Formation of the Laminin-Nidogen Complex Because of the apparent differences in basement membrane morphology between wild-type and b1 integrin-deficient teratomas we used previously established extraction protocols to analyze the amounts and distribution of basement membrane proteins. After homogenization in neutral buffer, which removed most cellular and easily soluble matrix proteins, EDTA was used to extract the laminin-nidogen complex [23] followed by 6 M guanidine for the solubilization of the proteoglycan perlecan [40] and non-cross-linked collagens. Amino acid analysis of these extracts demonstrated similar amounts of total protein in both teratomas and their efficient extraction (ú99%) by the protocol used (Table 1). Yet the protein fraction requiring 6 M guanidine for solubilization was much greater in the wild-type teratomas, which could reflect a stronger association of their extracellular matrix. Analysis of 4-hydroxyproline was used as a measure of total collagen content and showed similar amounts and distributions in both teratoma extracts (Table 1). Assuming an average 4-hydroxyproline content of 100/1000 residues in collagens, the data demonstrate that collagen represents about 8% of the total protein in the teratoma tissues. Since the tissue extracts contained almost all of the teratoma proteins, they were sufficiently representative to compare the contents of individual basement mem-

AID

ECR 3837

/

6i2d$$$$$3

12-16-97 13:55:15

Extracellular Matrix Proteins in Teratoma Extracts Derived from b1 Integrin /// or 0/0 Embryonic Stem Cellsa Extracts Protein analyzed Laminin-1

Nidogen

Perlecan

Fibulin-1

Fibulin-2

Fibronectin

b1 type, (content)

TBS

TBS/E

Gu

Sum of all

(mg/g) (mg/g) (%) (%) (mg/g) (mg/g) (%) (%) (mg/g) (mg/g) (%) (%) (mg/g) (mg/g) (%) (%) (mg/g) (mg/g) (%) (%) (mg/g) (mg/g) (%) (%)

40.6 19.5 6 48 11.3 13.3 9 51 16.2 41.3 32 80 8.7 13.0 79 93 1.36 2.03 41 67 20.9 22.8 86 98

625.2 20.0 93 49 103.5 12.1 87 46 25.6 9.5 48 18 2.1 1.1 19 7 1.92 0.96 58 32 0.8 õ0.05 3 õ1

7.8 1.4 1 3 4.6 0.7 4 3 10.8 0.8 20 2 0.2 õ0.01 2 õ1 0.02 0.03 1 1 2.7 0.5 11 2

673.6 40.9 100 100 119.4 26.0 100 100 52.6 51.6 100 100 11 14.1 100 100 3.30 3.02 100 100 24.4 23.3 100 100

///, 0/0, ///, 0/0, ///, 0/0, ///, 0/0, ///, 0/0, ///, 0/0, ///, 0/0, ///, 0/0, ///, 0/0, ///, 0/0, ///, 0/0, ///, 0/0,

a The extracts were those used in Table 1 and amounts expressed as mg/g wet tissue and as % of the total amounts recovered. The standard deviation of values was õ20%.

ecal

76

SASAKI ET AL.

pare Tables 1 and 2) and that it was preferentially extracted by EDTA-containing buffer in wild-type teratomas. Integrin b1-deficient teratoma, however, showed a more than 90% reduction in laminin-1 content and a clear shift to a higher, EDTA-independent solubility. A similar shift but a somewhat smaller reduction in total content was observed for nidogen. However, other basement membrane constituents such as perlecan, fibulin1, fibulin-2, and fibronectin showed no significant differences in total amounts but some shift to higher solubility, particularly for perlecan (Table 2). The changes in content and solubility recorded in Table 2 were significant based on the fact that mixtures of teratomas from three to five animals were used for each extraction and the low variability of the quantitative data (õ20%) obtained by radioimmunoassay. In order to assure reproducibility, we repeated the inoculation, extraction, and important radioimmunoassays in a second experiment. The data demonstrated a similar reduction in laminin-1 and nidogen contents and no change in perlecan (Table 3). The shifts in solubility were also comparable (data not shown). The large change in laminin-1 content could also be confirmed by SDS gel electrophoresis under reducing conditions which demonstrated the characteristic major bands of about 400 kDa (a1 chain) and 200 kDa (b1/g1 chains) in the EDTA extract of wild-type teratoma. Both bands were strongly reduced in the b1 integrin-deficient teratoma and not replaced by similar bands which would indicate complementation by other laminin isoforms (Fig. 4A). All these bands appeared to be assembled into disulfide-linked laminin heterotrimers, as shown by electrophoresis under nonreducing conditions, indicating that intracellular laminin chain assembly is not impaired by the b1 integrin deficiency. Electrophoresis also demonstrated a reduction in the 150-kDa band of intact nidogen in b1 integrindeficient teratoma (Fig. 4A), consistent with the radioimmunoassay data. Nidogen is also known to be sensitive to degradation by various tissue proteases when not complexed to laminins [41]. Enhanced nidogen deg-

TABLE 3 Variability in the Relative Amounts of Basement Membrane Proteins in Wild-Type and b1 Integrin-Deficient Teratomasa Amount (%) in mutant teratomas Protein analyzed

Experiment 1

Experiment 2

Laminin-1 Nidogen Perlecan

6 22 98

8 37 116

FIG. 4. Comparison of EDTA extracts from wild-type (///) and b1 integrin-deficient (0/0) teratomas by SDS gel electrophoresis and immunoblotting for nidogen. (A) Analysis by protein staining with lanes loaded with 6 mg total protein from b1/// (lanes 1, 3) and b10/ 0 (lanes 2, 4) tissues. The extracts were used in either nonreduced (lanes 1, 2) or reduced (lanes 3, 4) form. Migration positions of laminin (Ln), its a1, b1, and g1 chains, and of nidogen (Nd) are indicated in the left margin and were determined by comparison with EHS laminin-1nidogen complex [23]. (B) Immunoblotting with anti-mouse nidogen of b1/// (lane 5) and b10/0 integrin (lane 6) teratoma extracts loaded with the same amount of nidogen (5 ng) as determined by radioimmunoassay. The sizes of intact nidogen (150 kDa) and major fragments (100, 45 kDa) are indicated in the right margin.

radation to a major 100-kDa band and a weaker 45-kDa band could in fact be shown for b1 integrin-deficient teratoma by immunoblotting of the EDTA extracts (Fig. 4B). The total amounts of laminin-1 (850 kDa) and nidogen (150 kDa) determined by radioimmunoassay in wild-type teratoma (Table 2) indicated the formation of a complex with a 1:1 stoichiometry. This was shown by molecular sieve chromatography of the EDTA extracts, in which about 90% of the nidogen coeluted with laminin (data not shown, but see [23]). Extracts of b1 integrin-deficient teratoma, however, showed a more than threefold molar excess of nidogen over laminin-1 (Table 2). About 80% of the nidogen eluted after laminin in molecular sieve chromatography, demonstrating a severe impairment in the formation of the lamininnidogen complex. This may in part also be caused by proteolytic degradation of nidogen. Down-Regulation of Laminin-1 mRNA

a Data are from two different inoculations where tissues were extracted and analyzed by radioimmunoassays as in Table 2. The amounts recovered in the b1 mutant are expressed as % of the wildtype control (set at 100%).

AID

ECR 3837

/

6i2d$$$$$3

12-16-97 13:55:15

Northern blots were used to examine whether the selective deficiency of laminin-1 and nidogen in b1 integrin-deficient teratomas was due to reduced transcription or proteolysis (Fig. 5). There was a significant decrease in the mRNA for all laminin-1 chains, a1, b1, and g1, of about 40–60% as shown by densitometry

ecal

INTEGRIN DEFICIENCY AND BASEMENT MEMBRANE ASSEMBLY

77

and possibly nidogen (Fig. 6D). No or smaller changes were observed for g1 chains and perlecan (Figs. 6C and E). This also demonstrated a decent expression of major basement membrane components in wild-type ES cells prior to any differentiation. DISCUSSION

FIG. 5. Northern blots of wild-type (///) and b1 integrin-deficient (0/0) teratomas for basement membrane components. mRNAs were detected at their expected positions for nidogen (A), laminin a1(B), b1(C), and g1(D) chain, perlecan (E), laminin b2(F) and a2(G) chain, and fibulin-2 (H). A GAPDH probe was used as control (I) and was the same for each vertical row in a set of four different filters. All lanes were loaded with equal amounts of total RNA (20–25 mg). Exposure time to X-ray film was 2 days except for a2 and b2 (7 days).

(Table 4). However, only marginal effects were observed for nidogen, perlecan, and fibulin-2 mRNA. Other laminin mRNAs, for example those for a2 and b2 chains, were only expressed at very low levels in both teratomas and did not show a distinct up-regulation in the b1 integrin-negative tumor (Fig. 5). This indicated a distinct down-regulation of laminin-1 mRNA which is not compensated for by an increased expression of other laminin isoforms containing a2 and b2 chains. Since down-regulation of mRNA expression of some integrin a subunits was also previously observed for cultured b1 integrin-deficient ES cells [14] we now examined this possibility for several basement membrane proteins (Fig. 6). When compared to the actin control (Fig. 6F) a clear down-regulation of mRNA was indicated for laminin a1 (Fig. 6A) and b1 chain (Fig. 6B)

At least six receptors of the b1 integrin subfamily (a1b1, a2b1, a3b1, a6b1, a7b1, and a9b1) have been shown to be involved in the recognition of various laminins and collagen IV in in vitro cell adhesion, spreading and migration assays, and direct receptor binding studies [3, 42]. These observations strongly suggested that the same receptors are responsible for the firm anchorage of cells to basement membranes but so far no clear in vivo evidence has been obtained for this proposal. The deletion of the b1 integrin gene in transgenic mice which caused peri-implantation lethality [16, 17] was, however, in strong favor of this concept. Through such studies embryonic stem cells with a double knock-out of the b1 integrin gene [14] became available and were used to study b1 integrin dependence of differentiation and migration of various cell lines in chimeric mice [18–20]. Here we used these mutant cells to form subcutaneous teratomas. They were much less efficient than wild-type ES cells in inducing tumor growth [21] but nevertheless were still capable of secreting and depositing a large variety of extracellular matrix components including collagens and basement membrane proteins. Major differences observed with the b1 integrin-deficient tumors were, however, an extensive ultrastructural disorganization of basement membranes and, very surprisingly, a selective down-regulation of laminin-1. In teratomas, wild-type embryonic stem cells develop along different cell lineages, many of them being sur-

TABLE 4 Ratios of mRNA Levels between Wilde-Type and b1 Integrin-Deficient Teratomas as Determined by Densitometry of Northern Blotsa mRNA

Ratio

Laminin a1 chain Laminin b1 chain Laminin g1 chain Nidogen Perlecan Fibulin-2

0.40 { 0.15 0.49 { 0.10 0.65 { 0.03 0.88 { 0.02 0.88 1.15

a All values were corrected for the same amount of GAPDH mRNA and, except for perlecan and fibulin-2, represent means from two experiments.

AID

ECR 3837

/

6i2d$$$$$3

12-16-97 13:55:15

FIG. 6. Northern blots of wild-type (///) and b1 integrin-deficient (0/0) ES cells for basement membrane components. The same filter [14] was successively used for mRNA detection of laminin a1 chain (A), b1 chain (B), g1 chain (C), nidogen (D), perlecan (E), and actin (F) as internal standard.

ecal

78

SASAKI ET AL.

rounded by a basement membrane. Electron microscopy shows this to have, depending on cell type, a rather uniform breadth (range 40–200 nm) and to split into a distinct lamina lucida and lamina densa [43]. b1 integrin-deficient tumors frequently show a considerable broadening of the lamina lucida region and also some increase in the lamina densa. Both structures are also occasionally disrupted along small sections of the cell surface. These observations strongly indicate loss of molecular contacts with cellular receptors. More unusual features include multilaminar structures and rounded amorphous matrix deposits. The latter are reminiscent of basement membrane deposits, although this will still need confirmation by immunoelectron microscopy. Together, the data demonstrate that basement membranes can still form, but with remarkable ultrastructural differences. Most basement membranes are considered to possess two intermingling networks composed of collagen IV and laminins, respectively, which require connecting components such as nidogen for stabilization [2]. The low level of laminin1 found in b1 integrin-deficient teratomas therefore provides an additional explanation for the aberrant ultrastructural features. Our data on deficient interactions between cells and basement membranes are in accordance with those reported for transgenic mice lacking a more restricted repertoire of integrin receptors. Mice with a null mutation of the a3 integrin gene die neonatally because of abnormal kidney and lung development [44]. Glomerular basement membranes show a fragmented and disorganized structure and are deficient in the connection to podocyte foot processes. This mice show in addition skin blistering, indicating that a3b1 integrin has a role in the maintenance of basement membrane integrity [45]. Much more severe neonatal skin blisters are caused by a6 [46] and b4 [47] integrin deficiencies and mice die around birth. No overt tissue phenotype was, however, observed in mice lacking a1 integrins but cells derived from such mice were deficient in spreading and migration on laminin and collagen IV [48]. All these data demonstrate that even the elimination of individual receptors for laminins and collagens IV produces distinct but tissue-restricted phenotypes. It is likely that such phenotypes should be more severe when combined and we think that the teratoma model described here supports the prediction. The down-regulation of laminin-1 (chain composition a1b1g1) in the mutant teratoma could be shown both by mRNA analysis for the three chains and by a radioimmunoassay which is dependent on the assembly of the chains into intact laminin [35]. Disulfide-mediated assembly was also demonstrated by electrophoresis of tissue extracts (Fig. 4A), indicating that the major defect lay in the synthesis of individual laminin subunits but not despite their low concentration in their potential for assembly. Yet reduction at the mRNA level (40–

AID

ECR 3837

/

6i2d$$$$$3

12-16-97 13:55:15

60%) was less than at the protein level (90%). This could probably be explained by the need for heterotrimeric laminin assembly forms for secretion [49], which at the level of individual cells implies that the lowest amount of a single laminin chain will become rate-limiting. A molecular explanation of this selective laminin1 down-regulation, which is not found for several other basement membrane proteins, still remains tentative. It could include the control of laminin-1 synthesis at the transcriptional level by laminin-specific b1 integrin receptors. This has already been discussed for the decreased levels of a3 and a6 integrin mRNA observed for b1 integrin-deficient ES cells [14]. A similar and selective reduction in mRNA of basement membrane components could now be observed for b1 integrin-deficient ES cells (Fig. 6), indicating that this defect is independent of differentiation. Alternatively, other mechanisms including control of mRNA stability may operate and remain to be elucidated. Teratomas differentiate along various cell lineages including epithelial cells derived from endoderm and ectoderm [13]. We can therefore not exclude the possibility that insufficient differentiation of certain epithelial precursor cells may occur in the mutant teratoma and at least in part account for the low levels of laminin-1. Yet no deficiency has been observed for keratinocyte precursors [20] and muscle cell development [21]. A specific deficit seems, however, to exist for b1 integrin-deficient endothelial and other endoderm-derived cells which could not be identified in the teratoma [21] or in chimeric mice [16]. Endothelial cells are known to differentiate in wild-type embryoid bodies [50]. They could be also detected in b1 integrin-deficient embryoid bodies but have apparently a deficit in forming branched vessels [21]. A different explanation may apply for the reduction of the nidogen content in mutant tumors by about 70% which is not reflected by a comparable decrease in mRNA. Here degradation of nidogen by tissue proteases is most likely responsible, as demonstrated by large amounts of a 100-kDa nidogen fragment (Fig. 4B). Nidogen is known to be rather sensitive to various tissue proteases [41]. Major cleavage sites include a link region between the globular domains G1 and G2, which would generate a 100-kDa fragment, and the laminin-binding domain G3 if it is not occupied by laminin. Thus the substoichiometric amounts of laminin-1 in b1 integrin-deficient teratomas could be the major cause of reduced nidogen levels and not necessarily an enhanced proteolytic activity. A similar explanation may exist for developing salivary glands, where antibody blockage of laminin-nidogen interaction prevents basement formation and probably enhances nidogen degradation as shown by immunofluorescence [10]. Based on these data, we speculate that the defects described for b1 integrin-deficient teratomas may also cause the very early lethality of transgenic mice lacking

ecal

INTEGRIN DEFICIENCY AND BASEMENT MEMBRANE ASSEMBLY

b1 integrins [16, 17]. This occurs around the time when the blastocysts implant into the uterus. Shortly before implantation, part of the inner cell mass differentiates into a layer of primitive endodermal cells and becomes separated from the remaining cells by a distinct basement membrane which contains laminin. b1 integrindeficient blastocysts were shown to produce some laminin but do not deposit it into a linear basement membrane structure [17]. Visceral endoderm derived from differentiated b1 integrin-deficient F9 teratocarcinoma cells apparently shows a similar defect [15]. Furthermore, studies of the fate of b1 integrin-deficient embryonic stem cells in chimeric mice demonstrated their absence from all endoderm-derived tissues such as liver, gut, and lung epithelium [16]. These observations are consistent with the possibility that the failure to form the first endodermal basement membrane due to lack of b1 integrins is followed by cell necrosis and death of the embryo. Electron microscopy of late blastocysts is still required, however, to show whether the same morphological changes can be found as described here for basement membranes of the mutant tumors. Even though teratomas have become a popular model for studying the differentiation of embryonic stem cells [13], little has been known about the composition of their extracellular matrix. Our data show that about 8% of the total proteins in the wild-type teratomas are collagens and 1% are laminin-1. The latter exists mainly as a stoichiometric 1:1 complex with nidogen as previously shown for the Engelbreth-Holm-Swarm (EHS) mouse tumor [23]. In embryonic organogenesis, the formation of this complex is known to depend on cellular cooperation between epithelium-derived laminin and mesenchymal cells which provide the nidogen [9, 51, 52]. Perturbation of this extracellular interaction by antibodies which block the nidogen-binding site of laminins was shown to interfere with proper basement membrane assembly and to inhibit organogenesis in embryonic kidney, lung, and salivary glands [9, 10]. This was interpreted as indicating that the connection of laminins to the collagen IV network and perlecan through nidogen [36] is instrumental for basement membrane stability. Undifferentiated ES cells are, however, able to produce laminin and nidogen, and perlecan and cell cooperativity during normal organogenesis may therefore develop differently in teratomas. Lack of b1 integrins is obviously another means of destabilizing basement membranes in teratomas, but here the mechanisms would include either disruption of cellular interactions or down-regulation of laminin1. Cells derived from such teratomas may therefore become important tools for studying other factors that contribute to basement membrane formation and stability. The value of this model has already been shown for the in-culture polymerization of fibronectin in a previous study [53].

AID

ECR 3837

/

6i2d$$$$$4

12-16-97 13:55:15

79

We thank Dr. Yoshi Yamada for providing several cDNA probes and are grateful for the excellent technical assistance of Vera van Delden, Mischa Reiter, and Stefan Benkert. This study was supported by the Deutsche Forschungsgemeinschaft (SFB 266, Fa¨ 296/ 1-2) and by an EMBO fellowship (to E.F.). R.F. was supported by the Hermann and Lilly Schilling Foundation.

REFERENCES 1. Streuli, C. (1996). Basement membrane as a differentiation and survival factor. In ‘‘The Laminins’’ (P. Ekblom and R. Timpl, Eds.), pp. 217–233, Harwood Academic, Reading. 2. Timpl, R., and Brown, J. C. (1996). Supramolecular assembly of basement membranes. BioEssays 18, 123–132. 3. Aumailley, M., Gimond, C., and Rousselle, P. (1996). Integrinmediated cellular interactions with laminins. In ‘‘The Laminins’’ (P. Ekblom and R. Timpl, Eds.), pp. 127–158, Harwood Academic, Reading. 4. Henry, M. D., and Campbell, K. P. (1996). Dystroglycan: An extracellular matrix receptor linked to the cytoskeleton. Curr. Opin. Cell Biol. 8, 625–631. 5. Peters, D. M. P., and Mosher, D. F. (1987). Localization of cell surface sites involved in fibronectin fibrillogenesis. J. Cell Biol. 104, 121–130. 6. Wu, C., Bauer, J. S., Juliano, R. L., and McDonald, J. A. (1993). The a5b1 integrin fibronectin receptor, but not the a5 cytoplasmic domain, functions in an early and essential step in fibronectin matrix assembly. J. Biol. Chem. 268, 21883–21888. 7. Sorokin, L., Sonnenberg, A., Aumailley, M., Timpl, R., and Ekblom, P. (1990). Recognition of the laminin E8 cell-binding site by an integrin possessing the a6 subunit is essential for epithelial polarization in developing kidney tubules. J. Cell Biol. 111, 1265–1273. 8. Durbeej, M., Larsson, E., Ibraghimov-Beskrovnaja, O., Roberds, S. L., Campbell, K. P., and Ekblom, P. (1995). Non-muscle adystroglycan is involved in epithelial development. J. Cell Biol. 130, 79–91. 9. Ekblom, P., Ekblom, M., Fecker, L., Klein, G., Zhang, H.-Y., Kadoya, Y., Chu, M.-L., Mayer, U., and Timpl, R. (1994). Role of mesenchymal nidogen for epithelial morphogenesis in vitro. Development 120, 2003–2014. 10. Kadoya, Y., Salmivirta, K., Talts, J. F., Kadoya, K., Mayer, U., Timpl, R., and Ekblom, P. (1997). Nidogen binding to laminin g1 chain is essential for epithelial-mesenchymal interactions in developing submandibular gland. Development 124, 683–691. 11. Dziadek, M., and Timpl, R. (1985). Expression of nidogen and laminin in basement membranes during mouse embryogenesis and in teratocarcinoma cells. Dev. Biol. 111, 372–382. 12. Dziadek, M., Fujiwara, S., Paulsson, M., and Timpl, R. (1985). Immunological characterization of basement membrane types of heparan sulfate proteoglycan. EMBO J. 4, 905–912. 13. Damjanov, I., Damjanov, A., and Solter, D. (1987). Production of teratocarcinomas from embryos transplanted to extra-uterine sites. In ‘‘Teratocarcinomas and Embryonic Stem Cells, a Practical Approach’’ (E. Robertson, Ed.), pp. 1–17, IRL Press, Oxford. 14. Fa¨ssler, R., Pfaff, M., Murphy, J., Noegel, A. A., Johansson, S., Timpl, R., and Albrecht, R. (1995). Lack of b1 integrin in embryonic stem cells affects morphology, adhesion and migration but not integration into the inner cell mass of blastocysts. J. Cell Biol. 128, 979–988. 15. Stephens, L. E., Sonne, J. E., Fitzgerald, L., and Damsky, C. H. (1993). Targeted deletion of b1 integrins in F9 embryonic carcinoma cell affects morphological differentiation but not tissuespecific gene expression. J. Cell Biol. 123, 1607–1620.

ecal

80

SASAKI ET AL.

16. Fa¨ssler, R., and Meyer, M. (1995). Consequences of lack of b1 integrin gene expression in mice. Genes Dev. 9, 1896–1908. 17. Stephens, L. E., Sutherland, A. E., Klimanskaya, I. V., Andrieux, A., Meneses, J., Pedersen, R. A., and Damsky, C. H. (1995). Deletion of b1 integrins in mice results in inner cell mass failure and peri-implantation lethality. Genes Dev. 9, 1883–1895. 18. Hirsch, E., Iglesias, A., Potocnik, A. J., Hartmann, U., and Fa¨ssler, R. (1996). Impaired migration but not differentiation of haemotopoetic stem cells in the absence of b1 integrins. Nature 380, 171–175. 19. Fa¨ssler, R., Rohwedel, J., Meltsev, V., Addicks, K., Hescheler, J., and Wobus, A. M. (1996). Loss of b1 integrin function results in abnormal specification of cardiac cells during mouse ES cellderived cardiogenesis. J. Cell Sci. 109, 2989–2999. 20. Bagutti, C., Wobus, A., Fa¨ssler, R., and Watt, F. M. (1996). In the absence of b1 integrin differentiation of embryonic stem cells into keratinocytes is inhibited in vitro but not in vivo. Dev. Biol. 179, 184–196. 21. Bloch, W., Forsberg, E., Lentini, S., Brakebusch, C., Martin, K., Krell, H. W., Weidle, U. H., Addicks, K., and Fa¨ssler, R. (1997). b1 integrin is essential for teratoma growth and angiogenesis. J. Cell Biol. 38, 265–278. 22. Doetschman, T. C., Eistetter, H., Katz, M., Schmidt, W., and Kemler, R. (1985). The in vitro development of blastocyst-derived embryonic stem cell lines: Formation of visceral yolk sac, blood islands and myocardium. J. Embryol. Exp. Morphol. 87, 27–45. 23. Paulsson, M., Aumailley, M., Deutzmann, R., Timpl, R., Beck, K., and Engel, J. (1987). Laminin-nidogen complex: Extraction with chelating agents and structural characterization. Eur. J. Biochem. 166, 11–19. 24. Chomczynski, P., and Sacchi, N. (1987). Single-step method of RNA isolation by acid guanidinium thiocyanate-phenol-chloroform extraction. Anal. Biochem. 162, 156–159. 25. Sasaki, M., Kleinman, H. K., Huber, H., Deutzmann, R., and Yamada, Y. (1988). Laminin, a multidomain protein: The A chain has a unique globular domain and homology with the basement membrane proteoglycan and the laminin B chains. J. Biol. Chem. 263, 16536–16544. 26. Sasaki, M., Kato, S., Kohno, K., Martin, G. R., and Yamada, Y. (1987). Sequence of cDNA encoding the laminin B1 chain reveals a multidomain protein containing cysteine-rich repeats. Proc. Natl. Acad. Sci. USA 84, 935–939. 27. Sasaki, M., and Yamada, Y. (1987). Structure of the laminin B2 chain shows multidomain structures homologous to the B1 chain. J. Biol. Chem. 262, 17111–17117. 28. Bernier, S. M., Utani, A., Sugiyama, S., Doi, T., Polistina, C., and Yamada, Y. (1994). Cloning and expression of laminin a2 chain (M-chain) in the mouse. Matrix Biol. 14, 447–455. 29. Durkin, M. E., Gautam, M., Loechel, F., Sanes, J. R., Merlie, J. P., Albrechtsen, R., and Wewer, U. M. (1996). Structural organization of the human and mouse laminin b2 chain genes, and alternative splicing at the 5* end of the human transcript. J. Biol. Chem. 271, 13407–13416. 30. Mann, K., Deutzmann, R., Aumailley, M., Timpl, R., Raimondi, L., Yamada, Y., Pan, T., Conway, D., and Chu, M.-L. (1989). Amino acid sequence of mouse nidogen, a multidomain basement membrane protein with binding activity for laminin, collagen IV and cells. EMBO J. 8, 65–72. 31. Noonan, D. M., Fulle, A., Valente, P., Cai, S., Horigan, E., Sasaki, M., Yamada, Y., and Hassell, J. R. (1991). The complete sequence of perlecan, a basement membrane heparan sulfate proteoglycan, reveals extensive similarity with laminin A chain, LDL-receptor and N-CAM. J. Biol. Chem. 266, 22939–22947.

AID

ECR 3837

/

6i2d$$$$$4

12-16-97 13:55:15

32. Pan, T.-C., Sasaki, T., Zhang, R.-Z., Fa¨ssler, R., Timpl, R., and Chu, M.-L. (1993). Structure and expression of fibulin-2, a novel extracellular matrix protein with multiple EGF-like repeats and consensus motifs for calcium-binding. J. Cell Biol. 123, 1269–1277. 33. Fort, P., Marty, L., Piechaczyk, S., el Sabrouty, S., Dani, C., Jeanteur, P., and Blanchard, J. M. (1985). Various rat adult tissues express only one major mRNA species from the glyceraldehyde-3-phosphate-dehydrogenase multigenic family. Nucleic Acids Res. 13, 1431–1442. 34. Timpl, R., and Risteli, L. (1982). Radioimmunoassays in studies of connective tissue proteins. In ‘‘Immunochemistry of the Extracellular Matrix’’ (H. Furthmayr, Ed.), Vol. I, pp. 199–235, CRC Press, Boca Raton, FL. 35. Gerl, M., Mann, K., Aumailley, M., and Timpl, R. (1991). Localization of a major nidogen-binding site to domain III of laminin B2 chain. Eur. J. Biochem. 202, 167–174. 36. Fox, J. W., Mayer, U., Nischt, R., Aumailley, M., Reinhardt, D., Wiedemann, H., Mann, K., Timpl, R., Krieg, T., Engel, J., and Chu, M.-L. (1991). Recombinant nidogen consists of three globular domains and mediates binding of laminin to collagen type IV. EMBO J. 10, 3137–3146. 37. Sasaki, T., Kostka, G., Go¨hring, W., Wiedemann, H., Mann, K., Chu, M.-L., and Timpl, R. (1995). Structural characterization of two variants of fibulin-1 that differ in nidogen affinity. J. Mol. Biol. 245, 241–250. 38. Schulze, B., Mann, K., Battistutta, R., Wiedemann, H., and Timpl, R. (1995). Structural properties of recombinant domain III-3 of perlecan containing a globular domain inserted into an epidermal-growth-factor-like motif. Eur. J. Biochem. 231, 551– 556. 39. Sasaki, T., Wiedemann, H., Matzner, M., Chu, M.-L., and Timpl, R. (1996). Expression of fibulin-2 by fibroblasts and deposition with fibronectin into a fibrillar matrix. J. Cell Sci. 109, 2895– 2904. 40. Paulsson, M., Yurchenco, P. D., Ruben, G. C., Engel, J., and Timpl, R. (1987). Structure of low density heparan sulfate proteoglycan isolated from a mouse tumor basement membrane. J. Mol. Biol. 197, 297–313. 41. Mayer, U., Mann, K., Timpl, R., and Murphy, G. (1993). Sites of nidogen cleavage by proteases involved in tissue homeostasis and remodelling. Eur. J. Biochem. 217, 877–884. 42. Vandenberg, P., Kern, A., Ries, A., Luckenbill-Edds, L., Mann, K., and Ku¨hn, K. (1991). Characterization of a type IV collagen major cell binding site with affinity to the a1b1 and a2b1 integrins. J. Cell Biol. 113, 1475–1483. 43. Inoue, S., and Leblond, C. P. (1988). Three-dimensional network of cords: the main components of basement membranes. Am. J. Anat. 118, 341–358. 44. Kreidberg, J. A., Donovan, M. J., Goldstein, S. L., Rennke, H., Shepherd, K., Jones, R. C., and Jaenisch, R. (1996). a3b1 integrin has a crucial role in kidney and lung organogenesis. Development 122, 3537–3547. 45. DiPersio, C. M., Hodivila-Dilke, K. M., Jaenisch, R., Kreidberg, J. A., and Hynes, R. O. (1997) a3b1 integrin is required for normal development of the epidermal basement membrane. J. Cell Biol. 137, 729–742. 46. Georges-Labouesse, E., Mesaddeq, N., Yehia, G., Cadalbert, L., Dierich, A., and Le Meur, M. (1996). Absence of integrin a6 leads to epidermolysis bullosa and neonatal death in mice. Nature Genet. 13, 370–373. 47. van der Nent, R., Krimpenfort, P., Calafat, J., Niessen, C. M., and Sonnenberg, A. (1996). Epithelial detachment due to absence of hemidesmosomes in integrin b4 null mice. Nature Genet. 13, 366–369.

ecal

INTEGRIN DEFICIENCY AND BASEMENT MEMBRANE ASSEMBLY 48. Gardner, H., Kreidberg, J., Koteliansky, V., and Jaenisch, R. (1996). Deletion of integrin a1 by homologous recombination permits normal murine development but gives rise to a specific deficit in cell adhesion. Dev. Biol. 175, 301–313. 49. Timpl, R. (1996). Macromolecular assembly of basement membranes. Curr. Opin. Cell Biol. 8, 618–624. 50. Risau, W., Sariola, H., Zerwes, H. G., Sasse, J., Ekblom, P., Kemler, R., and Doetschman, T. C. (1988). Vasculogenesis and angiogenesis in embryonic stem cell-derived embryoid bodies. Development 102, 471–478. 51. Thomas, T., and Dziadek, M. (1993). Genes coding for basement

membrane glycoproteins laminin, nidogen and collagen IV are differentially expressed in the nervous system and by epithelial, endothelial and mesenchymal cells of the mouse embryo. Exp. Cell Res. 208, 54–67. 52. Dziadek, M. (1995). Role of laminin-nidogen complexes in basement membrane formation during embryonic development. Experientia 51, 901–913. 53. Wennerberg, K., Lohikangas, L., Gullberg, D., Pfaff, M., Johansson, S., and Fa¨ssler, R. (1996). b1 integrin-dependent and -independent polymerization of fibronectin. J. Cell Biol. 132, 227–238.

Received June 27, 1997 Revised version received September 26, 1997

AID

ECR 3837

/

6i2d$$$$$4

12-16-97 13:55:15

81

ecal