The role of E-cadherin and integrins in mesoderm ... · obtained by manual dissection with glass needles, as shown in Fig. 1. Separate epiblast and mesodermal tissues were isolated

INTRODUCTION

Adhesive interactions among cells and between cells and

their surrounding extracellular matrix (ECM) are importantduring morphogenetic processes, including the cell migra-tions occurring during embryonic development (Edelman,

829Development 118, 829-844 (1993)Printed in Great Britain © The Company of Biologists Limited 1993

We have examined the role of cell-cell and cell-extracel-lular matrix (ECM) interactions during mesodermdifferentiation and migration at the primitive streak ofthe mouse embryo with the use of function-perturbingantibodies. Explants of epiblast or mesoderm tissuedissected from the primitive streak of 7.5- to 7.8-daymouse embryos were cultured on a fibronectin substra-tum in serum-free, chemically defined medium. After 16-24 hours in culture, cells in explants of epiblast exhibitedthe typical close-packed morphology of epithelia, and thetissue remained as a coherent patch of cells that wereshown to express transcripts of the cytokeratin Endo Bby in situ analysis. In contrast, cells in explants ofprimitive streak mesoderm exhibited a greatly flattened,fibroblastic morphology, did not express Endo B tran-scripts, and migrated away from the center of theexplant.

As epiblast cells in vivo undergo the epithelial-mes-enchymal transition at the primitive streak, they ceaseexpressing the prominent calcium-sensitive cell adhesionmolecule E-cadherin (uvomorulin, Cell-CAM 120/80).We asked whether the loss of E-cadherin expression wasa passive result of differentiation or if it might play amore causative role in mesoderm differentiation andmigration. Culture with function-perturbing antibodiesagainst E-cadherin caused cells within epiblast explantsto lose cell-cell contacts, to flatten, and to assume a mes-enchymal morphology; they were also induced tomigrate. Anti-E-cadherin antibodies had no effect onexplants of primitive streak mesoderm. In immunofluo-rescence studies, anti-E-cadherin-treated epiblast cellsceased to express SSEA-1, a carbohydrate moiety that islost as mesoderm differentiates from the epiblast in vivo,and they also ceased to express E-cadherin itself. Incontrast, these cells began to express the intermediatefilament protein vimentin, a cytoskeletal protein charac-teristic of the primitive streak mesoderm at this stage ofdevelopment.

As epiblast cells differentiate into mesoderm, their

predominant adhesive interactions change from cell-cellto cell-substratum. Therefore, we also investigated theadhesive interactions between primitive streak tissuesand extracellular matrix (ECM) components. Epiblastexplants adhered well to fibronectin, more poorly tolaminin and type IV collagen, and not at all to vit-ronectin. In contrast, mesoderm explants attached wellto all these proteins. Furthermore, epiblast, but notmesoderm, displayed an anchorage-dependent viabilityin culture. After anti-E-cadherin treatment, epiblastcells that had assumed the mesenchymal morphology didattach to vitronectin, another characteristic shared withprimitive streak mesoderm. Adhesion of epiblast tofibronectin and of primitive streak mesoderm tofibronectin, vitronectin, laminin, and type IV collagenwas completely blocked by incubation with a broad-spectrum polyclonal serum, anti-ECM receptor (anti-ECMR) antiserum, which recognizes 1 and 3integrins. Anti-ECMR-treated mesodermal explantsrecovered and attached, spread, and migrated normallyafter antibodies were removed. In addition, an antibodyspecific for 6 1 integrin, which mediates adhesion tolaminin, selectively blocked attachment of mesoderm tolaminin but not to fibronectin, indicating that 6 1 is amajor laminin receptor for these cells.

We conclude that disruption of E-cadherin function inmammalian epiblast cells at the primitive streak in vitrocauses them to acquire a phenotype characteristic ofmesoderm, and we propose that similar mechanisms actduring mesoderm differentiation in the intact embryo.Our results also show that the cell-substratum adhesionof primitive streak tissues is mediated by the integrinsuperfamily of receptors and that developmentallyregulated changes in cell-ECM adhesion accompany theepithelial-mesenchymal transition at the mammalianprimitive streak.

Key words: cell-cell adhesion, cell migration, mouse gastrulation,extracellular matrix

SUMMARY

The role of E-cadherin and integrins in mesoderm differentiation and

migration at the mammalian primitive streak

Carol A. Burdsal1, Caroline H. Damsky2 and Roger A. Pedersen1,3

1Laboratory of Radiobiology and Environmental Health, 2Departments of Stomatology and Anatomy, and 3Departments ofAnatomy, Radiology, and Obstetrics, Gynecology and Reproductive Sciences, University of California, San Francisco, CA94143, USA

830

1988; Takeichi, 1991; Hynes and Lander, 1992). Changes incell-cell and cell-matrix adhesion accompany the ingressionof primary mesenchyme cells during gastrulation in the seaurchin (Fink and McClay, 1985; Burdsal et al., 1991). Inavian embryos, microinjection of antibodies against thefibronectin receptor, or of peptides mimicking the cell-binding domain of fibronectin, perturbs neural crest migration(Boucaut et al., 1984; Bronner-Fraser, 1985). In addition, cellinteractions with an intact basal lamina containingcrosslinked collagen are necessary for the completion of thecell movements of gastrulation in the sea urchin embryo(Wessel and McClay, 1987). In urodele amphibian embryos,a fibrillar ECM composed of fibronectin and laminin presenton the basal surface of the blastocoel roof directs mesoder-mal cell migration during gastrulation (reviewed by Johnsonet al., 1992). Also in amphibian embryos, microinjection ofcell-binding domain peptides inhibits cellular movementsduring gastrulation (Boucaut et al., 1984; Darribere et al.,1988). These studies demonstrate the vital role that changesin cell-cell adhesion and cell-ECM interactions play in mor-phogenetic processes during development.

At 7 days of development in the mouse, cells present inthe posterior epiblast (embryonic ectoderm) lose contactswith other cells in that region and form the primitive streak,the site of gastrulation. This loss of cell-cell contacts isaccompanied by the loss of expression of cell-cell adhesionmolecules. For example, in the mouse embryo, E-cadherin(uvomorulin, Cell-CAM 120/80) is expressed in the epiblastbut is not expressed in primitive streak mesoderm; in thechick embryo, N-CAM and E-cadherin (L-CAM) arepresent in the epiblast but disappear from the primitivestreak as mesoderm ingresses (Cunningham and Edelman,1990; Damjanov et al., 1986).

As gastrulation continues, mesoderm cells activelymigrate through and away from the primitive streak(Nakatsuji et al., 1986) and spread into embryonic regionsanteriorly as two sheets, or mesodermal wings; mesodermalcells also migrate posteriorly into extraembryonic regionswhere they will contribute to extraembryonic membranes(Snell and Stevens, 1966; Lawson et al., 1991; reviewed byCruz and Pedersen, 1991; and Lawson and Pedersen, 1992).Mesodermal migration takes place in the extracellular spacebetween the epiblast and the visceral endoderm, wherefibronectin, laminin, type IV collagen, and heparan sulfateproteoglycan have been detected by immunocytochemicalstudies (Wartiovaara et al., 1979; Adamson and Ayers;1979; Leivo et al., 1980; Leivo, 1983). On the basis of theseobservations, in addition to functional studies in otherembryonic systems, it appears that mesoderm cells use theseECM components as a migratory substratum during gastru-lation. However, despite numerous descriptive studies(Spiegelman and Bennett, 1974; Solursh and Revel, 1978;Franke et al., 1983; Nakatsuji et al., 1986; Hashimoto et al.,1987; Hashimoto and Nakatsuji, 1989), the molecular mech-anisms of gastrulation in the mouse embryo, including thefactors that control ingression and migration, have yet to beidentified.

To address these questions directly, we have developed asystem for the culture of mouse primitive streak tissues andhave used it to characterize the adhesive interactions ofepiblast and primitive streak mesoderm. Making use of

function-perturbing antibodies, we have studied the role ofE-cadherin function during mesoderm differentiation. Wehave also examined the cell-substratum interactions ofepiblast and mesoderm during gastrulation. The integrinsuperfamily of receptors (Hynes and Lander, 1992) arecandidate molecules for mediating adhesion of mesoderm toECM proteins; therefore, we have examined the attachmentof explants of primitive streak tissues to purified ECM com-ponents in vitro and have made use of function-perturbingantibodies against integrins to determine whether thesemolecules are functional receptors in this system.

MATERIALS AND METHODS

MiceMice used in this study were Dub:ICR (originally obtained fromDominion Laboratories, Inc., Dublin, VA) and CD1:ICR (origi-nally obtained from Charles River Laboratories, Wilmington MA).Both stocks were maintained as closed colonies in our animalfacility. For all morphological results presented, tissues from Dubembryos were used, unless otherwise indicated. Animals weremaintained on a cycle of 14 hours light/10 hours dark (light period,6.00 a.m. to 8.00 p.m.). All embryos were obtained after naturalmating. Noon of the day the vaginal plug was detected was desig-nated 0.5 day of gestation.

Dissection and culture of explantsPrimitive streak tissues were dissected from 7.5- to 7.8-dayembryos. Mesoderm was dissected from embryos in which theprimitive streak had advanced to approximately the distal tip of theegg cylinder (and mesoderm had migrated to the midline andpartially into the anterior half of the egg cylinder in the mesoder-mal wings). Embryos that had developed to the late streak-stage(i.e., those in which head folds were present) were not used. Theposterior halves of the embryonic region of each embryo wereobtained by manual dissection with glass needles, as shown in Fig.1. Separate epiblast and mesodermal tissues were isolatedaccording to the protocol of Svajger and Levak-Svajger (1975).Briefly, the dissected posterior halves were placed in a solution of0.5% trypsin (Calbiochem, San Diego, CA) and 2.5% pancreatin(Sigma Chemical Co., St. Louis, MO) in Ca2+- and Mg2+-freephosphate-buffered saline (PBS, 140 mM NaCl, 1.5 mM KH2PO4,3 mM KCl, and 8 mM Na2HPO4) for 15 minutes at 4°C. For furtherdissection and stopping of enzymatic activity, the egg cylinderswere transferred into medium containing serum (FM-II, Spindle1980). The visceral endoderm was removed from the egg cylindersand discarded. Epiblast and mesoderm were teased apart with glassneedles. Under a stereomicroscope the sheet of loosely connectedmesodermal cells was easily distinguished from the thicker, moretightly coherent sheet of epiblast. Small tissue explants were trans-ferred to prepared wells in 16-chamber slides (Nunc, Inc.,Naperville, IL).

Separate wells in 16-chamber slides were incubated withsolutions of fibronectin (20 µg/ml), laminin (35 µg/ml), vitronectin(20 µg/ml), or collagen type IV (35 µg/ml) in PBS for 2 hours.Human plasma fibronectin, mouse type IV collagen, and lamininwere obtained from Collaborative Research, Inc. (Bedford, MA).Vitronectin was obtained from Telios Pharmaceuticals (San Diego,CA). In a control experiment, wells were coated with Cell-tak™(Collaborative Research, Inc.), a formulation of the repeatingdecapeptide in the adhesive protein of a marine mussel that is L-dopa- and hydroxyproline-rich and does not contain the tripeptideArg-Gly-Asp (RGD; Waite and Tanzer, 1981; Waite, 1983). Wellswere then washed three times with PBS and explants were trans-ferred to the prepared wells and cultured in approximately 130 µl

C. A. Burdsal and others

831Adhesive interactions at the primitive streak

of a serum-free, chemically defined medium. The defined medium(T + 2× AA; Spindle, 1980) was supplemented with 0.8 mg/mladenosine, 0.85 mg/ml guanosine, 0.73 mg/ml cytidine, 0.73mg/ml uridine, 0.24 mg/ml thymidine, 0.2 mg/ml insulin, Mito+serum extender (which contains defined factors includingepidermal growth factor, transferrin, insulin, endothelial cellgrowth supplement, triiodothyronine, hydrocortisone, proges-terone, testosterone, estradiol-17B, selenium, and O-phospho-rylethanolamine [Collaborative Research Inc.]), 100 µg/ml strep-tomycin, and 100 Units/ml penicillin G; 4 mg/ml of bovine serumalbumin was added just before use. Explants were cultured at 37°Cin an atmosphere of 5% CO2 and 95% air.

AntibodiesAnti-E-cadherin antibodies (anti-Cell-CAM 120/80 antibodies),which block the activity of E-cadherin (Damsky et al., 1983; Richaet al., 1985), were purified from a rabbit polyclonal serum. Pre-immune and nonimmune antibodies were used at the same con-centrations. Anti-ECMR antibodies are function-perturbing goatpolyclonal antibodies that recognize a broad range of integrins,including the β1 and β3 families (Knudsen et al., 1981; Damskyet al., 1982; Sutherland et al., 1988). Anti-ECMR antibodies werepurified on immobilized protein-G (Pierce, Rockford, IL).Antibody concentrations were determined at OD280, using anextinction coefficient of 1.4 for a 1 mg/ml solution. Nonimmunegoat antibodies were purified in an identical fashion. GoH3, anantibody known to inhibit the function of α6β1 integrins, wasobtained from the Central Blood Transfusion Laboratory,Amsterdam, The Netherlands (rat monoclonal antibody againstCDW49F; Sonnenberg et al., 1987, 1988, 1990). Other antibodiesused in immunofluorescence studies were anti-vimentin (Sigma)and anti-SSEA-1 monoclonal antibodies (gift of D. Solter, WistarInstitute, Philadelphia, PA).

Antibody perturbation studiesTwo or three explants were transferred to each prepared well of the16-chamber slides. The percentage of explants that attached andspread on the various ECM components was determined bycounting explants viewed at 16× magnification on an invertedmicroscope (Zeiss Inc., Germany), and explants were pho-tographed under differential interference contrast (DIC) optics. Todetermine the function of E-cadherin in maintaining epiblast

integrity, we cultured epiblast explants for 16 hours on afibronectin substratum and then added anti-E-cadherin antibodiesto a final concentration of 90 µg/ml in culture medium. Explantswere cultured with anti-E-cadherin anitbodies for 4-6 hours andwere then washed 3 times with PBS. Explants were cultured for anadditional 12-30 hours (depending on the experiment) to ensurethat anti-E-cadherin antibodies were no longer present on the cellsurface. In control experiments, mesoderm explants were culturedwith anti-E-cadherin antibodies, and epiblast explants werecultured with pre-immune or non-immune antibodies.

To determine the function of integrins in epiblast and mesodermattachment, we added function-perturbing anti-ECMR antibodiesat the beginning of the culture period. Mesoderm explants wereexamined after 4 hours with 75 µg/ml anti-ECMR antibodies. Atthis time explants were washed 3 times with PBS and furthercultured for 12-16 hours in the absence of antibodies. The effectof the anti-ECMR antibodies on epiblast attachment was examinedafter 12-16 hours, because of the slower kinetics of epiblast attach-ment. In control experiments, epiblast and mesoderm explants werecultured in an equal concentration of nonimmune goat antibodies.

To test the ability of cells from anti-E-cadherin-treated epiblastexplants to reattach to vitronectin and fibronectin, we culturedepiblast explants on fibronectin with anti-E-cadherin antibodies for4-6 hours and then for an additional 12 hours without antibodies.To remove cells from the fibronectin substratum, we cultured anti-E-cadherin-treated epiblast explants with 100 µg/ml anti-ECMRantibodies, which caused more than 80% of the cells to round upin 2 hours and allowed the gentle detachment of cells. Cells weredislodged with a gentle stream of medium from a pipette tip andwere transferred to fibronectin- or vitronectin-coated wells. Cellswere cultured for 4 hours without antibodies and the fraction ofcells that had attached and spread on fibronectin and vitronectinwas determined by cell counts. In control experiments, mesodermcells that had attached and spread on fibronectin were also treatedwith the anti-ECMR antibodies, as described for epiblast, and weretransferred to fibronectin- or vitronectin-coated wells.

Labelling of probes and in situ hybridizationSense and antisense constructs specific for Endo B were the kindgift of R. Oshima (La Jolla Cancer Research Center, La Jolla, CA).Probes were labeled with digoxigenin via in vitro transcription withSP6 RNA polymerase (Boehringer Mannheim, Indianapolis, IN)and the (SP6/T7) Dig-RNA Labeling kit (Boehringer Mannheim).The labeled mRNAs were quantified on gels by comparison withknown standards, and the sensitivity of antibody detection ofmRNAs was determined and normalized on blots.

In situ hybridization was performed on explanted tissuesaccording to a protocol modified from that of Conlon and Rossant(1992). Explants were fixed for 1 hour in 4% paraformaldehyde anddehydrated by sequential incubation with 25%, 50%, and 75%methanol diluted in Ca2 +- M g2 +-free-PBS containing 0.1% Tween-20 (PBT). Final dehydration was in 2 changes of 100% methanol.Explants were stored overnight in 100% methanol at −2 0 ° C .Samples were rehydrated through the methanol:PBT series to PBT,treated with 6% hydrogen peroxide in PBT and then detergentextracted in RIPA buffer (150 mM NaCl, 1% NP-40, 0.5% sodiumdeoxycholate, 0.1% sodium dodecyl sulfate [SDS], 1 mM EDTA,50 mM Tris). After extraction, samples were refixed in 0.2% glu-taraldehyde and 4% paraformaldehyde in PBT for 15 minutes.Further RIPA buffer treatments were followed by washes in PBT.Explants were then incubated for 1 hour at 70°C in prehybridiza-tion buffer (50% formamide [Molecular Biology Grade; Fisher Sci-e n t i fic, Santa Clara, CA], 5× SSC [1× SSC=0.15 M sodiumchloride, 0.015 M sodium citrate] pH 5, 50 µg/ml yeast RNA[Sigma], 1% SDS, and 50 µg/ml heparin [Sigma]). Digoxigenin-labeled probes were added to the hybridization buffer at a concen-tration of approximately 1 mg/ml. Explants were incubated with

Fig. 1. Dissection of egg cylinder. Dashed lines represent cutsmade with glass needles under the dissecting microscope. The firsthorizontal cut separates extraembryonic regions, from embryonicregions and the second vertical cut separates the anteriorembryonic region from the posterior region, which contains theprimitive streak. EPC, ectoplacental cone; M, primitive streakmesoderm; VE, visceral endoderm. Modified from Tam (1990).Bar, 100 µm.

832

sense and antisense digoxigenin probes at a concentration of 1 µg / m lovernight at 70°C. Unbound probe was removed by a sequence ofwashes at 70°C in a solution of 50% formamide, 5× SSC (pH 5),and 1% SDS, followed by a 70°C wash in a 1:1 mixture of the 50%formamide, 5× SSC (pH 5), and 1% SDS solution and a solution of0.5 mM NaCl, 10 mM Tris-HCl (pH 7.5), and 0.1% Tween-20.Further washes included a wash at room temperature in 0.5 mMNaCl, 10 mM Tris-HCl (pH 7.5), and 0.1% Tween-20; a wash at37°C in the same solution plus RNase A (Sigma); and another washat room temperature in 0.5 mM NaCl, 10 mM Tris-HCl (pH 7.5),and 0.1% Tween-20. Finally, explants were washed at room tem-perature in 50% formamide and 2× SSC (pH 5), followed by a washat 65°C in the same solution. To block nonspecific binding of anti-bodies, we incubated explants at room temperature in 10% heat-inac-tivated sheep serum (Sigma) in TBST buffer (140 mM NaCl, 2.7mM KCl, 25 mM Tris-HCl, pH 7.5, and 0.1% Tween-20). Alkalinephosphatase-conjugated anti-digoxigenin antibodies (BoehringerMannheim) were preabsorbed with powder prepared from 12.5- to13.5-day embryos before use. Explants were incubated with a 1:2000dilution of antibodies for 4 hours at room temperature. To removeunbound antibodies we washed samples overnight at room temper-ature in TBST buffer with the addition of 2 mM levamisole, followedby incubation in NTMT buffer (100 mM NaCl, 100 mM Tris-HCl,pH 9.5, 50 mM MgCl 2, and 0.1% Tween-20) made fresh each dayof use. Alkaline phosphatase activity was visualized by incubationwith a chromogenic substrate consisting of 4.5 µl/ml NBT solution(75 mg/ml nitroblue tetrazolium salt [Boehringer Mannheim] in 70%dimethylformamide) and 3.5 µl/ml BCIP solution (50 mg/ml 5-bromo-4-chloro-3-indolyl phosphate [Boehringer Mannheim] indimethylformamide) in NTMT buffer. Color development wasstopped by extensive washing with PBT.

Immunofluorescence and cytoskeletal stainingFor immunocytochemistry, explants to be stained with anti-vimentin were simultaneously fixed and permeabilized in 1:1methanol:acetone at −20°C for 5 minutes, rinsed 3 times with PBS,and incubated for 1 hour with a 1:30 dilution of the anti-vimentinantiserum (mouse IgM). Primary antibodies were removed bywashing three times with PBS, and explants were then incubatedfor 45 minutes at room temperature with a 1:200 dilution ofbiotinylated anti-mouse IgM (Vector Laboratories, Inc.,Burlingame, CA). Biotinylated antibodies were washed away, andsamples were incubated in a 1:100 dilution of fluoresceinated strep-tavidin (Pierce) for 30-45 minutes at room temperature. Explantsto be stained with anti-SSEA-1 (mouse IgM) or anti-α6 (GoH3)antibodies were fixed and permeabilized in 100% methanol at−20°C for 5 minutes. Samples were rinsed three times andincubated for 1 hour at room temperature with anti-SSEA-1 or wereincubated overnight at 4°C with anti-α6 antibodies. For anti-SSEA-1 staining, biotinylated antibodies and fluoresceinated strep-tavidin were used as described for anti-vimentin staining. For anti-α6 integrin staining, biotinylated anti-rat secondary antibodies(Sigma) were used at a 1:100 dilution followed by streptavidin asdescribed above. The actin cytoskeleton in mesoderm and epiblastexplants was visualized in the following manner. Explants werefixed and permeabilized in 3.7% formaldehyde containing 100µg/ml lysophosphatidyl choline, and actin stress fibers were stainedwith rhodamine-labeled phalloidin (Molecular Probes, Inc.,Eugene, OR). Immunofluorescent staining was observed and pho-tographed with a Zeiss microscope equipped with epifluorescence.

RESULTS

Characterization of in vitro explants of epiblastand mesodermFragments of epiblast and mesoderm were dissected from

the primitive streak region of 7.5- to 7.8-day embryos (theposterior half of the egg cylinder as shown in Fig. 1).Explants composed of 50-200 cells were then cultured onfibronectin-coated chamber slides in serum-free, chemicallydefined medium. After 16-24 hours in culture, epiblast andmesodermal tissues exhibited very different morphologicalfeatures (Fig. 2). Cells in the epiblast explants had numerouscell-cell contacts and the close-packed morphology charac-teristic of epithelial cells in culture (Fig. 2A). In addition,almost all epiblast explants (95%±4% s.d., n=50 explants)maintained a coherent border as shown in the smallerexplant in Fig. 2B. This is in contrast to explants of primitivestreak mesoderm, in which cells had migrated away fromthe center of the explant (and each other) and had the char-acteristic appearance of fibroblastic cells in culture (Fig.2C). Mesoderm cells were greatly flattened compared withepiblast, and mesoderm cells often exhibited an elongatedshape, with a lamellipodium at the leading edge and retrac-tion fibers at the rear of the cell.

The purity of the dissected tissues was assayed by in situhybridization by using probes against the epiblast-specificcytokeratin Endo B (Jackson et al., 1981). Endo B mRNAwas detected only when epiblast was probed with antisenseRNAs (Fig. 3A, E). No transcripts were detected inmesoderm explants probed with antisense RNAs (Fig. 3B,F) and none were observed when either tissue was probedwith sense RNAs (Fig. 3C, D, G, H). In three experiments,all epiblast explants (n=9) were positive for Endo B mRNAswhile no mesoderm explants were positive (n=8). In boththe in situ hybridization analysis and in the immunofluores-cent studies described below, we found that 100% purity oftissues was routinely obtained.

Characteristic differences in epiblast and mesodermalexplants were also observed when the organization of theactin cytoskeleton was examined. When explants werecultured on glass coverslips and stained with rhodamine-labeled phalloidin, cultured explants of both tissuesexhibited prominently staining actin stress fibers (Fig. 4).However, patches of epiblast displayed strong bands ofstress fibers that ran parallel to the coherent edge of theexplant, indicating that the architecture of junctionalcomplexes was maintained (Fig. 4A). In contrast, mesoder-mal cells migrating away from the edges of explantsdisplayed stress fibers that ran parallel to the direction ofmigration, and that had no relation to other cells, thus reflect-ing the migratory status of the mesoderm cells (Fig. 4B).

Anti-E-cadherin antibodies disrupt cell-cellcontacts in epiblast explantsE-cadherin is expressed from the onset of development inthe mouse and antibodies against E-cadherin (anti-CellCAM 120/80 antibodies) reversibly block compaction at the8- to 16-cell stage of development (Hyafil et al., 1983;Damsky et al., 1983; Vestweber and Kemler, 1984). At laterstages, E-cadherin is prominently expressed in the epiblastbut not in the mesoderm of the primitive streak-stageembryo (Damjanov et al., 1986). To determine whether thisloss of E-cadherin expression was a passive result of differ-entiation or if it might play a more causative role inmesoderm differentiation and migration at the primitivestreak, we made use of function-perturbing antibodies



specific for E-cadherin. Epiblast explants were cultured for12-16 hours on fibronectin, and anti-E-cadherin antibodieswere then added at a concentration of 90 µg/ml. After 4hours in culture with anti-E-cadherin antibodies, cellsthroughout the epiblast explants had lost cell-cell contactsand begun to flatten to the substratum (Fig. 5A). After 16hours in culture with anti-E-cadherin antibodies, cellsdisplayed a more flattened and elongated morphology, andgreater spaces separated cells within an explant (Fig. 5B).In any one experiment, 5-8% of cells in treated epiblastexplants were completely rounded (i.e., did not spread on

the substratum) after antibody treatment and died. While thismay account for a small percentage of the spaces betweenpreviously contiguous cells in the center of an explant, afteranti-E-cadherin treatment (and an additional 16 hours inculture) cells had migrated, as was evidenced by the fact thatcells now occupied an area greater than that of the originalexplant (Fig. 5C). In the explant shown in Fig. 5C, theoriginal diameter of the explant was approximately 200 µmand in this instance, some cells had migrated an additional80-120 µm from the edge of the explant after anti-E-cadherin treatment and culture. Pre-immune or nonimmuneserum, at the same concentration, had no effect on epiblastexplants; they remained a coherent epithelial patch (Fig.5D). In contrast, mesoderm explants were not affected byanti-E-cadherin treatment (Fig. 5E), although in 2 of 12experiments the mesoderm cells had migrated as usual butwere somewhat less flattened on the substratum comparedwith controls. The effects of anti-E-cadherin treatment onepiblast cells were apparent after 2-4 hours and were notreversible. After 4 hours of culture with anti-E-cadherinantibodies, explants were washed in PBS and culturedwithout antibodies for up to 30 hours (to ensure that anti-bodies were no longer present on the cell surface). In allcases, treated epiblast cells did not revert to an epithelialorganization but continued to exhibit a fibroblastic mor-phology. The induced change in morphology was observedin every experiment with the anti-E-cadherin antibodies(consisting of 2 or 3 epiblast explants dissected from eachof 1 to 3 littermates treated with anti-E-cadherin antibodiesin each of 20 experiments).

Cells from epiblast explants treated with anti-E-cadherin antibodies cease expressing markers ofepiblast and begin to express vimentinTreatment of epiblast explants with anti-E-cadherin anti-bodies induced a change in treated cells that was typical ofthe epithelial-mesenchymal transition during gastrulation invivo. We next asked whether this change in morphology wasaccompanied by changes in gene expression, which weassayed in a series of immunofluorescence experiments.Antibodies against SSEA-1 recognize a carbohydratemoiety that is present on epiblast cells but is lost as tissues,including the primitive streak mesoderm, differentiate invivo (Solter and Knowles, 1978; Skreb et al., 1991). Exceptfor a punctate spot of fluorescence on approximately 10%of cells, anti-E-cadherin-treated epiblast cells did not stainwith anti-SSEA-1 monoclonal antibodies (Fig. 6A, n=6explants). In contrast, all cells in control epiblast explantsstained strongly for this antigen (Fig. 6B), and cells inexplants of primitive streak mesoderm never stained posi-tively for SSEA-1 (Fig. 6C).

We also examined the expression of vimentin, an inter-mediate filament protein that is expressed by mesenchymalcells and is newly expressed in primitive streak mesodermas it traverses the streak (Jackson et al., 1981; Franke et al.,1983). Almost all cells (94%±3% s.d., n=6 explants) in anti-E-cadherin-treated epiblast explants stained strongly forintermediate filaments containing vimentin (Fig. 6D). Incontrast, treated control epiblast explants never stained pos-itively for vimentin (Fig. 6E). In control explants ofprimitive streak mesoderm 98% (±2% s.d., n=6 explants) of

Fig. 2. Cultured explants of epiblast (A,B) and mesoderm (C)isolated from primitive streak-stage embryos. Explants weretransferred to fibronectin-coated wells of 16-chamber glass slides,cultured for 16-24 hours in serum-free, chemically definedmedium, and photographed under DIC optics. Bar, 40 µm.

834 C. A. Burdsal and others

Fig. 3. In situ hybridization of cytokeratin (Endo B) transcripts in dissected primitive streak tissues. Digoxigenin-labeled sense andantisense probes for Endo B were generated for nonradioactive detection of Endo B mRNAs. (A-D) Bright-field microscopy. (E-H) Samefield viewed under DIC optics. (A, E) Epiblast explants probed with antisense RNAs. (B, F) Mesoderm explants probed with anti-senseRNAs. (C, G) Epiblast explants probed with sense RNAs. (D, H) Mesoderm explants probed with sense RNAs. Bar, 40 µm.


cells stained strongly with anti-vimentin antibodies (Fig.6F).

When explants treated with anti-E-cadherin antibodieswere cultured for 12-30 hours after antibodies were washedaway, the cells within the treated explants maintained theirmesenchymal morphology. Most cells (86%±3% s.d., n=4explants) within anti-E-cadherin-treated epiblast explantsthat were cultured for an additional 24-30 hours withoutantibodies no longer stained with anti-E-cadherin antibod-ies (Fig. 7A). In contrast, cells within control epiblastexplants continued to stain positively for E-cadherin (Fig.7B).

Taken together, the morphological and immunocyto-chemical data suggest that disruption of E-cadherin functionin epiblast is sufficient to trigger an irreversible transition toa mesodermal phenotype.

In vivo differentiation of epiblast to mesoderm andthe acquisition of a migratory phenotype areaccompanied by changes in cell-ECM interactionsIn vivo, the epiblast is organized on a basal lamina con-taining fibronectin, laminin, and type IV collagen (Wartio-vaara et al., 1979; Adamson and Ayers; 1979; Leivo et al.,1980; Leivo, 1983). As mesoderm differentiates, it interactswith these ECM components as it migrates between theepiblast and the visceral endoderm. Because of these devel-

opmentally regulated changes in cell-ECM interactions, weexamined cell-substratum interactions of primitive streaktissues in our culture system. Interestingly, there werequantitative differences between the attachment of epiblastto fibronectin and to the other ECM components, and therewere also quantitative differences between epiblast andmesoderm in their attachment to ECM components (Fig.8A). The percentage of epiblast explants that attached tofibronectin was far greater than the percentage that attachedto laminin or type IV collagen (83%±13% s.d. versus31%±18% s.d. and 24%±19% s.d., respectively). Epiblastexplants spread on fibronectin (Fig. 8B) had the same mor-phological appearance as when they attached and spread onlaminin or type IV collagen (data not shown). In strikingcontrast to their behavior on fibronectin (and to a lesserextent on laminin and type IV collagen), epiblast explantsfailed to attach to vitronectin (n=25 explants; Fig. 8C).Interestingly, epiblast, but not mesoderm, displayedanchorage-dependent viability upon culture; epiblastexplants that had not attached and begun to spread by 4-6hours failed to do so with overnight culture, and after 16hours, these explants were composed of vesiculated, deadcells (Fig. 8C).

In contrast to epiblast, when primitive streak mesodermwas dissected from the embryo and its attachment assayedin vitro, the percentage of mesoderm explants that attachedto fibronectin, vitronectin, laminin, and type IV collagenwas routinely 97-100% (±1-3% s.d.; Fig. 8A; the morphol-ogy of adherent mesoderm explants is shown in Figs 9 and10). Therefore, differentiation of mesoderm is accompaniedby changes in the cell’s adhesive phenotype.

Adhesion to ECM glycoproteins is mediated primarily bythe integrin superfamily of receptors (Hynes and Lander,1992). Anti-ECMR antibodies, which recognize a broadrange of integrins, have been shown to block the outgrowthof mouse trophectoderm cells from blastocysts in vitro(Richa et al., 1985; Sutherland et al., 1988). In the followingexperiments, these function-perturbing antibodies were usedto determine whether integrins are active in epiblast andmesoderm adhesion. The effect of anti-ECMR antibodies onepiblast adhesion was tested only on attachment tofibronectin because this substratum was the only one thatsupported significant levels of adhesion. The effect of anti-ECMR antibodies on adhesion of primitive streak mesodermto all four ECM glycoproteins was examined.

All epiblast cells cultured on fibronectin with 75 µg/mlanti-ECMR antibodies failed to attach (n=15; Fig. 9A),whereas 77% (±3% s.d.) of epiblast explants cultured in thesame concentration of control antibodies attached andspread on the fibronectin substratum (Fig. 9B). Attachmentof mesoderm to both fibronectin and vitronectin was com-pletely blocked by anti-ECMR antibodies, and the explantsremained as round, floating aggregates (Fig. 9C,F).Mesoderm explants attached quickly (within 1-4 hours) tofibronectin and vitronectin in the presence of nonimmuneantibodies (Fig. 9D,G). The effect of the anti-ECMR anti-bodies on mesoderm attachment was reversible. Afterremoval of the anti-ECMR antibodies (by washing theexplants in PBS), the treated mesoderm explants attached,spread, and migrated normally after additional culturewithout antibodies (Fig. 9E,H). Not unexpectedly, the anti-

Fig. 4. Phalloidin staining of the actin-containing cytoskeleton incultured explants of epiblast (A) and mesoderm (B) isolated fromprimitive streak-stage embryos. Explants were fixed,permeabilized, and stained with rhodamine-labeled phalloidin tovisualize the organization of the cytoskeleton. Bar, 20 µm.

836

ECMR-treated epiblast explants did not recover (data notshown), probably because of their anchorage dependence forviability.

Because mesoderm explants attached more slowly tolaminin and type IV collagen (within 6-8 hours), the effectof the anti-ECMR antibodies was scored after 16 hours inculture. At this time, mesoderm explants remained asrounded clumps of cells which were unattached to thelaminin or type IV collagen substratum (Fig. 10A,D).

Nonimmune antibodies had no effect on attachment (Fig.10B,E). After the removal of the anti-ECMR antibodies andan additional 24 hours in culture without antibodies,mesoderm explants attached, spread, and migrated normallyon laminin and type IV collagen (Fig. 10C, F).

To determine whether the effect of the anti-ECMR anti-bodies was due to a specific interference between integrinsand their ligands, we studied the ability of the antibodies toblock adhesion to a substratum that is not mediated by


Fig. 5. Effect of anti-E-cadherin antibodies on epiblast andmesoderm explants. (A) Epiblast explant cultured for 4 hourswith anti-E-cadherin antibodies. (B) Epiblast explant after 16hours of culture with anti-E-cadherin antibodies. (C) Anti-E-cadherin-treated epiblast explant in which cells havesignificantly migrated from the edge of the explant afterantibody treatment. The original center of the explant isindicated by the arrowhead. (D) Control epiblast cultured for16 hours in pre-immune antibodies (D). Primitive streakmesoderm explants cultured for 16 hours in the presence ofanti-E-cadherin antibodies. Bar, 40 µm.


integrins. Cell-tak™ is a nonspecific attachment factor fortissue culture cells that does not contain the RGD tripeptiderecognized by integrins (Waite and Tanzer, 1981; Waite,1983). When mesoderm was cultured on Cell-tak™ withanti-ECMR antibodies at the same concentrations as used inthe previous experiments, the antibodies had no effect.Attachment occurred within 1-4 hours (Fig. 10G).Therefore, the anti-ECMR-mediated inhibition of attach-ment to the ECM glycoproteins was not attributable to sterichindrance caused by antibody binding to the cells but rather

appeared to be due to specific interference with the interac-tion of integrin receptors and their ligands.

Primitive streak mesoderm cells use 6 1 as alaminin receptorSeveral integrin receptors mediate adhesion to laminin(Hynes and Lander, 1992). To determine more specificallywhich integrins play a role in the attachment of primitivestreak mesoderm to laminin, we examined the expressionand function of α6β1 integrin, a well-characterized laminin

Fig. 6. Expression of SSEA-1 and vimentin in control and anti-E-cadherin-treated explants. (A) Anti-SSEA-1 staining of an anti-E-cadherin-treated epiblast explant. (B) Anti-SSEA-1 staining of a control epiblast explant. (C) Anti-SSEA-1 staining of a controlmesoderm explant. (D) Anti-vimentin staining of an anti-E-cadherin-treated epiblast explant. (E) Anti-vimentin staining of a controlepiblast explant. (F) Anti-vimentin staining of a control mesoderm explant. Bar, 20 µm.

838

receptor (Sonnenberg et al., 1988). GoH3 monoclonal anti-bodies, which recognize α6β1 integrin and block thefunction of this receptor (Sonnenberg et al., 1987, 1988,1990) were used in immunofluorescence studies and in func-tional assays. Mesoderm cells stained prominently for α6β1(Fig. 11A) and when explanted into wells coated withfibronectin or laminin and incubated with 75 µg/ml GoH3antibodies, the cell’s attachment to laminin but not tofibronectin was specifically blocked (Fig. 11B,C). After theremoval of the GoH3 antibodies, the mesoderm explantsattached, spread, and migrated normally on laminin as well(Fig. 11D). These results confirm the role of the α6β1integrin as a major receptor for laminin on primitive streakmesoderm cells.

Cells from epiblast explants treated with anti-E-cadherin antibodies acquire the ability to adhereto vitronectinBecause the anti-E-cadherin-treated epiblast cells hadadopted a mesodermal phenotype with respect to morphol-ogy and protein expression, we asked whether they had alsoacquired adhesive behavior that was characteristic ofmesoderm. Epiblast explants did not attach to vitronectin,whereas mesoderm explants attached and spread readily on

it (Fig. 8A). We exploited this difference in the next set ofexperiments. Epiblast explants plated on fibronectin weretreated with anti-E-cadherin antibodies which produced afield of elongated and flattened cells as described above (Fig.


Fig. 7. Expression of E-cadherin after anti-E-cadherin treatment.An epiblast explant was cultured with anti-E-cadherin antibodiesfor 4-6 hours and antibodies were then washed away. The explantwas cultured for an additional 24 hours without antibodies andthen stained with anti-E cadherin antibodies (A). The small arrowindicates a cell which stained faintly for E-cadherin on itsperiphery. (B) Anti-E-cadherin staining of a control epiblastexplant. Bar, 20 µm.

Fig. 8. Attachment of epiblast and primitive streak mesodermcells to various ECM proteins. Explants were isolated andtransferred to wells of chamber slides coated with fibronectin,vitronectin, laminin, or type IV collagen. (A) Percentage ofexplants that attached and spread normally on each substratum.Error bars indicate s.d. (B) Appearance of attached and spreadepiblast explant on fibronectin. (C) Unattached explant of epiblastcultured on vitronectin. Bar, 40 µm.


12A). The cells were then incubated for 2 hours with anti-ECMR antibodies, which caused the cells to round up andlose contact with the fibronectin substratum and allowedtheir transfer to new wells (Fig. 12B). After transfer tofibronectin- or vitronectin-coated wells and 4 hours inculture (in the absence of antibodies), the percentage of cellsthat attached and spread to both substrata was determinedby cell counts (Fig. 12C-E, appearance of treated epiblastcells attached to vitronectin). When control explants ofepiblast were incubated with anti-ECMR antibodies, only asmall percentage of cells (approximately 5%) on the edge ofthe explant rounded from the substratum (probably due to alack of antibody accessibility in the interior of the explant).In addition, when control epiblast explants were trypsinizedand the resulting small clumps of cells were transferred tonew wells, no cells survived; therefore, the attachment ofanti-E-cadherin-treated cells to vitronectin was compared tothe attachment of primitive streak mesoderm cells whichwere removed from contact with fibronectin by anti-ECMRantibodies. The results of three experiments are presentedin Fig. 12F. Primitive steak mesoderm cells reattachedto fibronectin and vitronectin at high levels (73%±6% s.d.and 71%±8% s.d., respectively). Anti-E-cadherin-treatedepiblast cells also readily reattached to fibronectin (71%±8%s.d.) and, when compared with untreated epiblast (Fig. 8A),

anti-E-cadherin-treated cells now displayed a significantlevel of attachment to vitronectin (42%±10% s.d.).

DISCUSSION

In this study we used function-perturbing antibodies in a cellculture system to demonstrate the role of E-cadherin duringthe epithelial-mesenchymal transition that occurs at theprimitive streak during mouse gastrulation. Perturbation ofE-cadherin-mediated cell adhesion in explants of epiblastinduced a more flattened, fibroblastic morphology and ledto the expression of vimentin, an intermediate filamentprotein characteristic of mesenchymal cells and primitivestreak mesoderm cells at this stage of development in vivo.In addition, when antibodies were washed away, treatedepiblast cells did not resume their epithelial organization andno longer expressed E-cadherin or SSEA-1, two productsthat are lost as mesoderm differentiates in vivo (reviewedby Richa and Solter, 1986; Takeichi, 1991). We concludethat the loss of E-cadherin-mediated cell contacts in vitrotriggered steps of differentiation in the treated cells that werecharacteristic of the differentiation pathway of primitivestreak mesoderm.

E-cadherin-mediated cell adhesion is important for anumber of morphogenetic events during development,

Fig. 9. Inhibition of attachment of epiblast tofibronectin and of primitive streak mesoderm tofibronectin and vitronectin by anti-ECMRantibodies. (A) Unattached epiblast explantcultured on fibronectin with anti-ECMRantibodies. (B) Epiblast explant attached andspread on fibronectin when cultured withcontrol antibodies. (C) Unattached primitivestreak mesoderm explant cultured on fibronectin with anti-ECMR antibodies. (D) Primitive streak mesoderm explant attached and spreadon fibronectin when cultured with control antibodies. Mesoderm explant attached and spread on fibronectin after anti-ECMR antibodieswere washed away (E). (F) Unattached primitive streak mesoderm explant cultured on vitronectin with anti-ECMR antibodies. (G)Primitive streak mesoderm explant attached and spread on vitronectin when cultured with control antibodies. Mesoderm explant attachedand spread on vitronectin after anti-ECMR antibodies were washed away (H). Bar, 40 µm.

840 C. A. Burdsal and others

Fig. 10. Inhibition of attachment of primitive streak mesoderm to laminin and type IV collagen. (A) Unattached primitive streakmesoderm explant cultured on laminin with anti-ECMR antibodies. (B) Primitive streak mesoderm explant attached and spread onlaminin when cultured with control antibodies. Mesoderm explant attached and spread on laminin after anti-ECMR antibodies werewashed away (C). (D) Unattached primitive streak mesoderm explant cultured on type IV collagen with anti-ECMR antibodies. (E)Primitive streak mesoderm explant attached and spread on type IV collagen when cultured with control antibodies. Mesoderm explantattached and spread on type IV collagen after anti-ECMR antibodies were washed away (F). (G) Primitive streak mesoderm explantattached and spread on a substratum of Cell-tak™ when cultured with anti-ECMR antibodies. Bar, 40 µm.


including compaction and primitive endoderm segregation(Hyafil et al., 1983; Damsky et al., 1983; Vestweber andKemler, 1984; Richa et al., 1985). Transfection andexpression of recombinant cadherins lead to the sorting outof transfected cells in mixed cell populations, including pop-ulations of embryonic lung epithelia and mesenchyme (Noseet al., 1988; Friedlander et al., 1989). This implies that dif-ferential expression of cadherins and other cell adhesionmolecules could play a critical role in tissue segregationduring organogenesis. In another epithelial-mesenchymaltransition, the formation of the secondary palate, cellsdestined to transform into mesenchyme turn off expressionof E-cadherin (L-CAM) and syndecan. However, there is noevidence that the changes in E-cadherin (or syndecan)expression play a causative role in secondary palateformation or in the differentiation of tissues at the secondarypalate (Hay, 1991), as we have demonstrated for the differ-entiation of tissues at the primitive streak.

What do the results obtained in vitro imply about theprocess of gastrulation in the intact embryo? One possibil-

ity is that the loss of E-cadherin-mediated cell contacts isone of the first steps in a synchronized pathway of differen-tiation at the primitive streak leading to the ingression ofnewly formed mesodermal cells. The antibody-induced per-turbations of E-cadherin-mediated cell contacts in vitro mayactuate a pathway that occurs naturally during the processof streak formation and gastrulation in vivo. Such a pathwaymight include the localized activity of proteinases, espe-cially metalloproteinases, which are known to be promi-nently expressed during invasive processes such as theingression of epiblast cells through the primitive streak(Sanders and Prasad, 1989; Alexander and Werb, 1991).Alternatively, the expression of a number of molecules thatcould initiate or transduce the signal for primitive streakformation has been described in the gastrulating mouseembryo. These include brachyury, goosecoid (which acts asan organizer during Xenopus gastrulation), and the tran-scripts for a number of genes such as FGF-4 and otherfibroblast growth factor family members such as Wnt-2(Wilkinson et al., 1988, 1990; Cho et al., 1991; Blum et al.,

Fig. 11. Expression of α6β1 integrin in mesoderm and functional blockade by anti-α6β1 (GoH3) antibodies. GoH3 monoclonalantibodies stained a filamentous network of α6β1 integrins in the mesodermal cells (A). (B) Unattached primitive streak mesodermexplant cultured on laminin with GoH3 antibodies. (C) Primitive streak mesoderm explant attached and spread on fibronectin whencultured with GoH3 antibodies. Mesoderm explants attached and spread normally on laminin when antibodies were washed away (D).Bars: A, 20 µm; B-D, 40 µm.

842

1992; and reviewed by Latimer and Pedersen 1993). Theloss of E-cadherin expression may or may not be the catalystfor the organized cell movements of gastrulation in vivo;however, our data demonstrate that perturbation of E-cadherin function alone can induce a pathway of mesoderm-like differentiation. Furthermore, the effect of anti-E-cadherin treatment on epiblast differentiation was notreversible, as was the effect of anti-E-cadherin on the com-pacting 8- to 16-cell stage embryo (Damsky et al., 1983,Richa et al, 1985).

We also tested the ability of primitive streak tissues toattach to various ECM ligands and demonstrated that theintegrin superfamily of receptors mediate adhesion in thisembryonic system. The viability of epiblast, but notmesoderm, explants was anchorage dependent, and in vitroassays demonstrated that developmentally regulatedchanges in the adhesion of cells to vitronectin, laminin, andtype IV collagen occur as epiblast differentiates intomesoderm at the primitive streak. In addition, epiblast cellsacquired the ability to adhere to vitronectin, a characteristic


F

Fig. 12. Attachment of anti-E-cadherin-treated epiblast to vitronectin.(A) Epiblast cells treated with anti-E-cadherin antibodies and culturedfor an additional 12-16 hours without antibodies. (B) Roundedappearance of treated epiblast cells after 2 hours of incubation withanti-ECMR antibodies. (C-E) Anti-E-cadherin-treated epiblast cellsattached and spread on a vitronectin substratum. Bar, 40 µm.(F) Percentage of mesoderm and anti-E-cadherin-treated epiblast cellsthat attached and spread on vitronectin. Data are presented as themean fraction of cells attached to fibronectin and vitronectin in threeexperiments (n =250 for mesoderm cells; n=225 for anti-E-cadherin-treated epiblast cells). Error bars indicate s.d.


of the mesoderm cells in vitro, after anti-E-cadherin-induceddifferentiation. The differences in the number of epiblastversus mesoderm explants that attached to vitronectin,laminin, and collagen type IV imply that quantitative orqualitative changes in integrin expression occur withmesoderm differentiation at the primitive streak. Changes inintegrin expression may provide useful markers for themigratory mesoderm in mammals when more reagents thatrecognize mouse integrins become available to identifyspecific integrins functionally.

Because the events of gastrulation are staggering in theircomplexity, a simplified culture system, such as the onedescribed in this report, provides a powerful tool foraddressing questions concerning control mechanisms duringvertebrate gastrulation. This culture system can be used tofurther characterize the differentiation of anti-E-cadherin-treated epiblast tissues by in situ hybridization with addi-tional primitive streak mesoderm-specific transcripts, suchas PDGFα and Wnt-2 (Wilkinson et al., 1988; Mercola etal., 1990; Schatteman et al., 1992). Other mechanisticquestions can also be addressed; for example, do peptidegrowth factors play a role in inducing the loss of E-cadherinexpression in the posterior epiblast? Or, what role do peptidegrowth factors play in axial pattern formation and specifi-cation of mesodermal fate in the mouse, as has been previ-ously investigated in the amphibian system (reviewed bySlack, 1990; Jessell and Melton, 1992)? The knowledgeobtained by investigating these questions in vitro can thenbe used to improve experimental design when returning tothe intact embryo.

The authors are grateful to Robert Oshima and Davor Solter forthe kind gifts of Endo B constructs and anti-SSEA-1 antibodies,respectively. We also thank Ann Sutherland and Akiko Spindle fortheir generosity and for the development of the serum-free, chem-ically defined medium, Margaret Flannery for assistance, and AnnSutherland, Zena Werb, Jean J. Latimer, Stephen G. Grant, andMary McKenney for critical reading of this manuscript. This workwas supported by a National Institutes of Health National ResearchService Award 5 T32 ESO7106 from the National Institute of Envi-ronmental Health, by NIH Program Project Grant HD26732, andby the Office of Health and Environmental Research, US Depart-ment of Energy contract no. DE-AC03-76-SF01012.

REFERENCES

Adamson, E. D. and Ayers, S. E. (1979). The localization and synthesis ofsome collagen types in developing mouse embryos. Cell 16, 953-965.

Alexander, C. M. and Werb, Z. (1991). Extracellular matrix degradation.In Cell Biology of Extracellular Matrix ed. 2 (ed. E. D. Hay) pp. 255-302.New York: Plenum Press.

Blum, M., Gaunt, S. J., Cho, K. W. Y., Steinbeisser, H., Blumberg, B.,Bittner, D. and De Robertis, E. M. (1992). Gastrulation in the mouse:The role of the homeobox gene goosecoid. Cell 69, 1097-1106.

Boucaut, J. C., Darribere, T., Poole, T. J., Aoyama, H., Yamada, K. M.and Thiery, J.-P. (1984). Biologically active synthetic peptides as probesof embryonic development: A competitive peptide inhibitor of fibronectinfunction inhibits gastrulation in amphibian embryos and neural crest cellmigration in avian embryos. J. Cell Biol. 99, 1822-1830.

Bronner-Fraser, M. (1985). Alterations in neural crest migration by amonoclonal antibody that affects cell adhesion. J. Cell Biol. 101, 610-617.

Burdsal, C. A., Alliegro, M. C. and McClay, D. R. (1991). Tissue-specific,temporal changes in cell adhesion to echinonectin in the sea urchinembryo. Dev. Biol. 144, 327-334.

Cho, K. W. Y., Blumberg, B., Steinbeisser, H. and De Robertis, E. M.(1991). Molecular nature of Spemann’s organizer: the role of the Xenopushomeobox gene goosecoid. Cell 67, 1111-1120.

Conlon, R. A. and Rossant, J. (1992). Exogenous retinoic acid rapidlyinduces anterior ectopic expression of murine Hox-2 genes in vivo.Development 116, 357-368.

Cruz, Y. P. and Pedersen, R. A. (1991). Origin of embryonic andextraembryonic cell lineages in mammalian embryos, In AnimalApplications of Research in Mammalian Development (ed. R. A.Pedersen, A. McLaren, and N. L. First). pp. 147-204. Cold Spring Harbor,New York; Cold Spring Harbor Laboratory.

Cunningham, B. A. and Edelman, G. M. (1990). Structure, expression,and cell surface modulation of cell adhesion molecules. InMorphoregulatory Molecules (ed. G. M. Edelman, B. A. Cunningham,and J.-P. Thiery). pp. 9-40. New York:Wiley.

Damjanov, I., Damjanov, A. and Damsky, C. H. (1986). Developmentallyregulated expression of the cell-cell adhesion glycoprotein Cell-CAM120/80 in peri-implantation mouse embryos and extraembryonicmembranes. Dev. Biol. 116, 194-202.

Damsky, C. H., Knudsen, K. A. and Buck, C. A. (1982). Integralmembrane glycoproteins related to cell-substratum adhesion inmammalian cells. J. Cell Biochem. 18, 1-13.

Damsky, C. H., Richa, J., Solter, D., Knudsen, K. and Buck, C. A.(1983). Identification and purification of a cell surface glycoproteinmediating intercellular adhesion in embryonic and adult tissue. Cell 34,455-56.

Darribere, T., Yamada, K. M., Johnson, K. E. and Boucaut, J.-C.(1988). The 140-kDa fibronectin receptor complex is required formesodermal cell adhesion during gastrulation in the amphibianPleurodeles waltlii. Dev.Biol. 126, 182-194.

Edelman, G. M. (1988). Morphoregulatory molecules. Biochemistry 27,3533-3543.

Fink, R. D. and McClay, D. R. (1985). Three cell recognition changesaccompany the ingression of sea urchin primary mesenchyme cells. Dev.Biol. 107, 66-74.

Franke, W. W., Grund, C., Jackson, B. W. and Illmensee, K. (1983).Formation of cytoskeletal elements during mouse embryogenesis. IV.Ultrastructure of primary mesenchyme cells and their cell-cellinteractions. Differentiation 25, 121-141.

Friedlander, D. R., Mege, R. M., Cunningham, B. A. and Edelman, G.M. (1989). Cell sorting-out is modulated by both the specificity andamount of different cell adhesion molecules (CAMs) expressed on cellsurfaces. Proc. Natl. Acad. Sci. USA 86, 7043-7047.

Hashimoto, K. and Nakatsuji, N. (1989). Formation of the primitive streakand mesoderm cells in mouse embryos. Detailed scanning electronmicroscopical study. Dev. Growth Differ. 31, 209-218.

Hashimoto, K., Fujimoto, H. and Nakatsuji, N. (1987). An ECMsubstratum allows mouse mesodermal cells isolated from the primitivestreak to exhibit motility similar to that inside the embryo and reveals adeficiency in the T/T mutant cells. Development 100, 587-598.

Hay, E. D. (1991). Collagen and other matrix glycoproteins inembryogenesis. In Cell Biology of Extracellular Matrix ed. 2 (ed. E. D.Hay) pp. 419-462. New York: Plenum Press.

Hyafil, F., Babinet, C., Huet, C. and Jacob, F. (1983). Uvomorulin andcompaction. In Teratocarcinoma Stem Cells (ed. L. M. Silver, G. M.Martin,and S. Strickland) pp. 197-207. Cold Spring Harbor, New York:Cold Spring Harbor Laboratory.

Hynes, R. O. and Lander, A. D. (1992). Contact and adhesive specificitiesin the associations, migrations, and targeting of cells and axons. Cell 68,303-322.

Jackson, B. W., Grund, C., Winter, S., Franke, W. W. and Illmensee, K.(1981). Formation of cytoskeletal elements during mouse embryogenesis.II. Epithelial differentiation and intermediate-sized filaments in earlypostimplantation embryos. Differentiation 20, 203-216.

Jessell, T. M. and Melton, D. A. (1992). Diffusible factors in vertebrateembryonic induction. Cell 68, 257-270.

Johnson, K. E., Boucaut, J.-C. and DeSimone, D. W. (1992). The role ofthe extracellular matrix in amphibian gastrulation. In Current Topics inDevelopmental Biology Vol. 27 (ed. R. A. Pedersen and D. R. McClay)pp. 92-127. New York: Academic Press.

Knudsen, K. A., Rao, P., Damsky, C. H. and Buck, C. A. (1981).Membrane glycoproteins involved in cell-substratum adhesion. Proc.Natl. Acad. Sci. USA78, 6071-6076.

Latimer, J. J. and Pedersen, R. A. (1993). Epigenetic interactions and

844

gene expression in peri-implantation mouse embryo development. InGenes in Mammalian Reproduction (ed. R. B. L. Gwatkin) pp. 131-171.New York: Wiley-Liss.

Lawson, K. A. and Pedersen, R. A. (1992). Clonal analysis of cell fateduring gastrulation and early neurulation in the mouse. InPostimplantation Development in the Mouse (ed. D. J. Chadwick and J.Marsh) pp. 3-26. New York:Wiley.

Lawson, K. A., Meneses, J. J. and Pedersen, R. A. (1991). Clonal analysisof epiblast fate during germ layer formation in the mouse embryo.Development 113, 891-911.

Leivo, I. (1983). Structure and composition of early basement membranes:Studies with early embryos and teratocarcinoma cells. Med. Biol. 61, 1-30.

Leivo, I., Vaheri, A., Timpl, R. and Wartiovaara, J. (1980). Appearanceand distribution of collagens and laminin in the early mouse embryo. Dev.Biol. 76, 100-114.

Mercola, M., Wang, C., Kelly, J., Brownlee, C., Jackson-Grusby, L.,Stiles, C. and Bowen-Pope, D. F. (1990). Selective expression of PDGFA and its receptor during early mouse embryogenesis. Dev. Biol. 138,114-122.

Nakatsuji, N., Snow, M. H. L. and Wylie, C. C. (1986). Cinemicrographicstudy of the cell movement in primitive-streak stage mouse embryo. J.Embryol. Exp. Morph.96, 99-109.

Nose, A., Nagafuchi, A. and Takeichi, M. (1988). Expressed recombinantcadherins mediate cell sorting in model systems. Cell 54, 993-1001.

Richa, J. and Solter, D. (1986). Role of cell surface molecules in earlymammalian development. In Experimental Approaches to MammalianEmbryonic Development (ed. J. Rossant and R. A. Pedersen) pp. 293-320.Cambridge:Cambridge University Press.

Richa, J., Damsky, C. H., Buck, C. A., Knowles, B. B. and Solter, D.(1985). Cell surface glycoproteins mediate compaction, trophoblastattachment, and endoderm formation during early mouse development.Dev. Biol. 108, 513-521.

Sanders, E. J. and Prasad, S. (1989). Invasion of a basement membranematrix by chick embryo primitive streak cells in vitro. J. Cell Sci. 92, 497-504.

Schatteman, G. C., Morrison-Graham, K., Van Koppen, A., Weston, J.A. and Bowen-Pope, D. F. (1992). Regulation and role of PDGF receptorα-subunit expression during embryogenesis. Development 115, 123-131.

Skreb, N., Solter, D. and Damjanov, I. (1991). Developmental biology ofthe murine egg cylinder. Int. J. Dev. Biol.35, 161-176.

Slack, J. M. (1990). Growth factors as inducing agents in early Xenopusdevelopment. J. Cell Sci. 13Supplement, 119-130.

Snell, G. D. and Stevens, L. C. (1966). Early embryology. In Biology of theLaboratory Mouse (ed. E. L. Green) pp. 205-245. New York: McGraw-Hill.

Solter, D. and Knowles, B. B. (1978). Monoclonal antibody defining astage specific mouse embryonic antigen (SSEA-1). Proc. Natl. Acad. Sci.USA 75, 5565-5569.

Solursh, M. and Revel, J.-P. (1978). A scanning electron microscope study

of cell shape and cell appendages in the primitive streak region of the ratand chick embryo. Differentiation 11, 185-190.

Sonnenberg, A., Janssen, H., Hogervorst, F., Calafat, J. and Hilgers J.(1987). A complex of platelet glycoproteins Ic and IIa identified by a ratmonoclonal antibody. J. Biol. Chem. 262, 10376-10383.

Sonnenberg, A., Modderman, P. W. and Hogervorst, F. (1988). Lamininreceptor on platelets is the integrin VLA-6. Nature 336, 487-489.

Sonnenberg, A., Linders, C. J. T., Modderman, P. W., Damsky, C. H.,Aumailley, M. and Timpl, R. (1990). Integrin recognition of differentcell-binding fragments of laminin (P1, E3, E8) and evidence that α6β1but not α6β4 functions as a major receptor for fragment E8. J. Cell Biol.110, 2145-2155.

Spiegelman, M. and Bennett, D. (1974). Fine structural study of cellmigration in the early mesoderm of normal and mutant mouse embryos(T-locus: t9/t9). J. Embryol. Exp. Morph.32, 723-738.

Spindle, A. (1980). An improved culture medium for mouse blastocysts. InVitro 16, 669-674.

Sutherland, A. E., Calarco, P. G. and Damsky, C. H. (1988). Expressionand function of cell surface receptors in mouse blastocyst attachment andoutgrowth. J. Cell Biol. 106, 1331-1348.

Svajger, A. and Levak-Svajger, B. (1975). Technique of separation ofgerm layers in rat embryonic shields. Roux Arch. Dev. Biol. 178, 303-308.

Takeichi, M. (1991). Cadherin cell adhesion receptors as a morphogeneticregulator. Science 251, 1451-1455.

Tam, P. P. L. (1990). Studying development in embryo fragments. InPostimplantation Mammalian Embryos. A Practical Approach (ed. A. J.Copp and D. L. Cockroft) pp. 317-337. New York: IRL Press.

Vestweber, D. and Kemler, R. (1984). Rabbit antiserum against a purifiedsurface glycoprotein decompacts mouse preimplantation embryos andreacts with specific adult tissues. Exp. Cell Res. 152, 169-178.

Waite, J. H. (1983). Evidence for a repeating 3,4-dihydroxyphenylalanine-and hydroxyproline-containing decapeptide in the adhesive protein of themussel, Mytilus edulis L. J. Biol. Chem.258, 2911-2915.

Waite, J. H. and Tanzer, M. L. (1981). Polyphenolic substance of Mytilusedulis: Novel adhesive containing L-dopa and hydroxyproline. Science212, 1038-1040.

Wartiovaara, I., Leivo, I. and Vaheri, A. (1979). Expression of the cellsurface-associated glycoprotein fibronectin in the early mouse embryo.Dev. Biol. 69, 247-257.

Wessel, G. M. and McClay, D. R. (1987). Gastrulation in the sea urchinembryo requires the deposition of crosslinked collagen within theextracellular matrix. Dev. Biol. 121, 149-165.

Wilkinson, D. G., Peters, G., Dickson, C. and McMahon, A. P. (1988).Expression of the FGF-related proto-oncogene int-2 during gastrulationand neurulation in the mouse. EMBO J. 7, 691-695.

Wilkinson, D. G., Bhatt, S. and Hermann, B. G. (1990). Expressionpattern of the mouse T gene and its role in mesoderm formation. Nature343, 657-659.

(Accepted 18 March 1993)


Documents

The role of E-cadherin and integrins in mesoderm ... · obtained by manual dissection with glass needles, as shown in Fig. 1. Separate epiblast and mesodermal tissues were isolated