Mesodermal defects and cranial neural crest apoptosis in α5 integrin-null embryos

Transcription

1 Development 124, (1997) Printed in Great Britain The Company of Biologists Limited 1997 DEV Mesodermal defects and cranial neural crest apoptosis in α5 integrin-null embryos Keow Lin Goh, Joy T. Yang and Richard O. Hynes* Howard Hughes Medical Institute, Center for Cancer Research and Department of Biology, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139, USA *Author for correspondence SUMMARY α5β1 integrin is a cell surface receptor that mediates cellextracellular matrix adhesions by interacting with fibronectin. α5 subunit-deficient mice die early in gestation and display mesodermal defects; most notably, embryos have a truncated posterior and fail to produce posterior somites. In this study, we report on the in vivo effects of the α5-null mutation on cell proliferation and survival, and on mesodermal development. We found no significant differences in the numbers of apoptotic cells or in cell proliferation in the mesoderm of α5-null embryos compared to wildtype controls. These results suggest that changes in overall cell death or cell proliferation rates are unlikely to be responsible for the mesodermal deficits seen in the α5-null embryos. No increases in cell death were seen in α5-null embryonic yolk sac, amnion and allantois compared with wild-type, indicating that the mutant phenotype is not due to changes in apoptosis rates in these extraembryonic tissues. Increased numbers of dying cells were, however, seen in migrating cranial neural crest cells of the hyoid arch and in endodermal cells surrounding the omphalomesenteric artery in α5-null embryos, indicating that these subpopulations of cells are dependent on α5 integrin function for their survival. Mesodermal markers mox-1, Notch-1, Brachyury (T) and Sonic hedgehog (Shh) were expressed in the mutant embryos in a regionally appropriate fashion. Both T and Shh, however, showed discontinuous expression in the notochords of α5-null embryos due to (1) degeneration of the notochordal tissue structure, and (2) non-maintenance of gene expression. Consistent with the disorganization of notochordal signals in the α5-null embryos, reduced Pax-1 expression and misexpression of Pax-3 were observed. Anteriorly expressed HoxB genes were expressed normally in the α5-null embryos. However, expression of the posteriormost HoxB gene, Hoxb-9, was reduced in α5- null embryos. These results suggest that α5β1-fibronectin interactions are not essential for the initial commitment of mesodermal cells, but are crucial for maintenance of mesodermal derivatives during postgastrulation stages and also for the survival of some neural crest cells. Key words: α5-integrin, apoptosis, cell proliferation, mesoderm, neural crest, mouse INTRODUCTION Cell-cell and cell-extracellular matrix (ECM) interactions have long been thought to play critical roles in developmental processes (Damsky and Werb, 1992; Adams and Watt, 1993). One class of cell surface receptors that mediate such interactions are the integrins (Hynes, 1992). Integrins are heterodimeric transmembrane glycoproteins consisting of noncovalently bound α and β subunits. ECM molecules recognized by integrins include collagens, laminins and fibronectin (FN). Many integrins localize to focal adhesions, sites of cell-ecm adhesion where integrin cytoplasmic domains bind to components of the actin cytoskeleton. These integrin-cytoskeletal networks are crucial in regulating changes in cell adhesion and shape that occur during cell migration and spreading, and also form the foundation for the construction of signalling complexes (Damsky and Werb, 1992; Clark and Brugge, 1995; Schwartz et al., 1995). The α5β1 integrin is a major fibronectin receptor expressed at high levels in early developing Xenopus and chicken embryos, but is later down-regulated in adult tissues (Muschler and Horwitz, 1991; DeSimone, 1994). It is found in mesenchymal and connective tissues, and in all muscle types in the chicken embryo. Studies have shown that α5β1-fn binding regulates a variety of cellular responses including gene induction (Huhtala et al., 1995), oncogenic transformation (Giancotti and Ruoslahti, 1990; Dedhar, 1995), FN assembly (Fogerty et al., 1990), differentiation (Adams and Watt, 1990; Boukamp and Fusenig, 1993), adhesion and migration (Akiyama et al., 1989; Giancotti and Ruoslahti, 1990), and proliferation and cell survival (Varner et al., 1995; Zhang et al., 1995). The targeted disruption of the α5 integrin subunit in mice has shown that α5β1 plays a crucial role in postgastrulation development (Yang et al., 1993). Mice lacking α5 integrin subunit (α5-nulls) fail to complete the turning process, and die at embryonic gestation day (E) apparently due to vascular defects. Most strikingly, the α5-null embryos have

2 4310 K. L. Goh, J. T. Yang and R. O. Hynes normal anterior structures, forming heart, brain, otic and optic vesicles, but beyond the axial level of somites 7-10, possess distorted trunk regions, have a kinked neural tube and fail to form posterior somites. At E8-8.5, mutant embryos show a slightly shortened anterior-posterior axis but have normal numbers of somites. As the mutant embryos develop, their posterior defects become more pronounced and their tail regions fail to increase in size relative to the embryos anterior regions. At E9.5, mutant embryos are significantly delayed, equivalent in size to E9 wild-type embryos. In vitro studies demonstrated that embryonic cells derived from the α5-null mice remain capable of FN matrix assembly and migration on a FN substratum (Yang et al., 1993). Our aim in this study was to investigate the basis of the mesodermal deficit seen in α5-null embryos. The deficit might be due to defects in the migration, determination or differentiation of cells that make up the posterior trunk region of the embryo. Alternatively, a decrease in cell proliferation and/or increase in cell death might be responsible for the mutant phenotype. Previous studies, mentioned above, that have implicated α5β1- FN interactions in all these processes, were based on experiments performed in cell culture systems, on studies of expression patterns and, in some cases, on the injection of inhibitory agents into embryos. In contrast, in this study, we have used in situ assays to analyze the in vivo functions of α5 integrin in early embryogenesis. We have examined the effects of the loss of α5 integrin on embryonic cell proliferation and cell death. Using markers, we have investigated the ability of mesodermal precursor cells to undergo determination and differentiation into axial and paraxial structures, and examined the extent of axis determination in α5-null embryos. To examine further the effects of α5 on the cell cycle, we investigated cell cycle status of embryonic fibroblasts derived from α5-null and wild-type embryos. Our results suggest that proliferation and survival of mesodermal cells are independent of the presence of α5β1 integrin, and that absence of α5 does not increase cell deaths in the visceral yolk sac, amnion and allantois. Survival of subpopulations of cranial neural crest and of posterior endodermal cells, however, was dependent on α5β1 integrin. Absence of α5 integrin resulted in alterations in the expression patterns of several mesodermal markers and posteriorly expressed HoxB genes. Most strikingly T and Shh expression were not maintained in the notochordal tissues. Concurrently, notochordal tissues underwent degeneration in the α5-null embryos. MATERIALS AND METHODS Genotyping of embryos E mouse embryos were dissected out as described by Cockroft (1990). Embryos were treated as described in the protocols below. Yolk sacs were removed and lysed in buffer containing 50 mm Tris- HCl ph 8.0, 10 mm EDTA, 100 mm NaCl, 0.1% SDS, 1.2 mg/ml Proteinase K. DNA was subjected to PCR with Taq polymerase (Perkin-Elmer Cetus, Norwalk, CT), using α5-specific primers as well as primers for the neo R gene (Yang et al., 1995), under the following conditions: 94 C, 1 minute; 60 C, 1 minute 30 seconds and 72 C, 1 minute 30 seconds for 33 cycles. Immunohistochemistry 8 µm frozen sections from wild-type embryos were made as described by Yang et al. (1995). Sections were fixed in acetone for 5 minutes at 20 C, air dried for 1 hour and stored at 80 C. Sections were washed with PBS/0.1% bovine serum albumin (BSA), blocked in blocking buffer (5% normal goat serum in PBS/0.1% BSA) for 30 minutes, and incubated for 30 minutes with rat anti-mouse α 5 integrin antibody (anti- CD49e: Pharmingen, San Diego, CA) diluted 1:500 in blocking buffer. Control slides were treated with antibody isotype control, rat IgG 2a (Zymed, San Francisco, CA). Sections were then washed with PBS and incubated with biotinylated goat anti-rat IgG secondary antibody (Vector Labs, Burlingame, CA). After 3 washes in PBS, sections were incubated with 0.3% H 2O 2 in methanol for 30 minutes, followed by Endoblock (Biomeda, Foster City, CA) for 5 minutes. Sections were then washed in PBS and incubated with ABC reagent mix (Vector Labs) for 30 minutes, followed by 3,3 -diaminobenzidine (DAB)/H 2O 2 mix (DAB substrate kit, Vector Labs) for 2 minutes. Sections were counterstained with hematoxylin for 10 seconds, dehydrated in ethanol and xylene, mounted with Permount (Fisher, Chicago, IL), and then photographed on an Axiophot microscope (Carl Zeiss, Thornwood, NY). In situ hybridization and histology Whole-mount RNA in situ hybridization was performed essentially as described by Wilkinson (1992). After hybridization procedures, embryos were fixed in 0.1% glutaraldehyde/4% paraformaldehyde, cleared in 50% glycerol/pbs, then 80% glycerol/pbs overnight. Whole-mount embryos were mounted in 80% glycerol/pbs and photographed using bright-field or Nomarski optics. For histology, fixed whole-mount embryos were dehydrated in ethanol, cleared in xylene and embedded in paraffin. 4 µm sections were counterstained with eosin, examined and photographed. In situ probes used were gifts from the following researchers: Brachyury, Bernhard Herrmannn (Herrmannn, 1991); Notch-1, Tom Gridley (Del Amo et al., 1992); HNF3β, Siew-Lan Ang; BMP-4, Brigid Hogan; Wnt3a and Shh, Andrew McMahon; Pax-1 (Deutsch et al., 1988) and Pax-3, Peter Gruss; Hoxb-1 (#866), Hoxb-4 (#486), Hoxb-5 (#267) and Hoxb-9 (#803) Robert Krumlauf;CRABP-1 (Stoner and Gudas, 1989) and Mox-1 (Candia et al., 1992), Pierre Chambon. Immunohistochemical analysis of cell proliferation DNA synthesis was determined using immunohistochemical detection of 5 bromodeoxyuridine (BrdU) incorporation adapted from Morganbesser et al. (1995). Briefly, pregnant mice bearing E8.5-E9 embryos were injected intraperitoneally, 2 hours prior to killing, with a solution of 25 mg/ml BrdU (Sigma, St Louis, MO) in PBS at a final concentration of 250 mg BrdU per gram of mouse body weight. Embryos and sections of the mothers large intestine, as positive controls for mothers uptake of BrdU, were fixed in 10% formalin for 2 hours, paraffin-embedded and sectioned. Sections were then treated as described by Morganbesser et al. (1995). After exposure to BrdUspecific primary antibody (Becton-Dickinson, Bedford, MA) and biotinylated secondary antibody (Vector Labs), sections were stained with avidin/biotin complex and DAB (ABC and DAB substrate kits, Vector Labs). Photomicrographs of anterior, midsection and posterior regions of nine mutant and eight wild-type embryos were taken at 40 magnification and labelling was scored without knowledge of the genotype. All mesenchymal cells in a cm field of the photomicrographs were counted and scored as labelled or unlabelled. Paired groups of data were analyzed using the Mann-Whitney one-tailed test. Results were designated to be significant at P<0.05. In vitro cell cycle and growth analysis Embryonic fibroblasts were cultured from E9.25 embryos as described previously (Yang et al., 1996). Passage 3-6 cells were used in all experiments. For FACS analysis, cells were stained using the rapid Triton-X method for propidium iodide staining of unfixed cells (Bauer, 1992). In brief, cells were incubated in stain solution (3% PEG 6000, 2.5 µg/ml propidium iodide, 180 U/ml RNAse, 0.1% Triton-X, 4 mm citrate buffer ph 7.8) at cells per ml for 20 minutes at

3 Defects in α5-integrin null embryos C. Volume of stain and cell mixture was then doubled using salt solution (3% PEG 6000, 2.5 µg/ml propidium iodide, 0.1% Triton-X, 0.4 M NaCl), giving a final concentration of 10 6 cells/ml. Cells were then stored in the dark at 4 C overnight prior to flow cytometry. Percentage of cells in cell cycle phases were determined using ModFit LT (Verity Software House, Topsham, ME) FACS-based program for analysis of DNA content (n=5 for wild-type; n=3 for α 5-null; where n=number of embryos the cells were derived from). Data are given as percentage of cells ± s.d. For determining cell growth curves, cells were plated in duplicate on tissue culture plastic at cells/35 mm dish and samples were taken daily. Cells at each timepoint were trypsinized, subjected to trypan blue exclusion and counted. Cell counts included attached and unattached cells. Cell death determination Apoptotic cells were detected using terminal transferase biotinylateddutp nick-end labelling (TUNEL) on sectioned and whole-mount embryos (Gavrieli, 1992). TUNEL on paraffin sections was performed as described by Morganbesser et al. (1995), with the addition of coverslipping of sections after addition of solutions. Whole-mount TUNEL was performed essentially as described by Conlon et al. (1995), using digoxigenin-conjugated dutp [Boehringer Mannheim, Indianapolis, IN] in place of biotinylated-dutp. Digoxigenin epitope was detected as described by Conlon and Rossant (1992) with the following modifications. TBST washes, after addition of anti-digoxigenin antibody, were done overnight to reduce background signal. Substrate concentration was modified to 6.75 ml/2 ml NBT (75 mg/ml nitroblue tetrazolium salt in 70% dimethylformamide) and 5.25 ml/2 ml BCIP (50 mg/ml 5-bromo- 4-chloro-3-indolyl phosphate toluidine salt in 100% dimethylformamide). After color development, embryos were washed several times with TBST, and cleared in 50% and 80% glycerol/pbs. Photography and histology were performed as described for whole-mount in situs. Photomicrographs of head and ventral tail regions of 13 mutant and 21 wild-type littermates were taken at 40 magnification. All apoptotic cells in the branchial arch and ventral tail regions in photomicrographs were scored and counted without knowledge of genotypes. The resulting data were analysed statistically as described earlier for the BrdU data. Double labellings were done using a protocol adapted from White et al. (1994). In brief, embryos were labelled by TUNEL using biotinylated-dutp, washed after the terminal transferase step and taken through the in situ hybridization protocol as described above. Embryos were incubated overnight at 4 C with absorbed FITC-conjugated anti-digoxigenin antibody (1:10). Embryos were washed, incubated for 1 hour at room temperature with Cy3-conjugated avidin (1:100), washed again and mounted in gelvatol. Signals were visualized by confocal microscopy. RESULTS α5 integrin protein is expressed throughout the developing mouse embryo The truncated posterior and grossly normal anterior of the α5- null embryos could in principle be due to the presence of α5 solely in the posterior regions of the embryos. To examine this possibility, we investigated the expression of α5 protein in mouse embryos from E7.5-E11.5. At all stages observed, α5 protein was expressed anteriorly and posteriorly throughout the mouse embryonic mesoderm. α5 protein was expressed strongly in mesenchymal and connective tissues, with lowered expression in dermomyotome relative to sclerotome. Protein was undetectable in neural tissues and epithelial tissues lining the pleural cavities (Fig. 1). These results suggest that the mesodermal defects of α5-null embryos are restricted to the posterior for reasons other than limitations on protein expression (see Discussion). Mesodermal cell growth and proliferation is not dependent on α5 integrin Mesodermal defects in α5-null embryos might be due to decreased proliferation and/or increased apoptosis in cells that Fig. 1. α5 integrin protein expression. (A) Photomicrograph of transverse frozen section from an E9.5 normal mouse embryo incubated with anti-α5 monoclonal antibody, stained with DAB (brown) and counterstained with hematoxylin (blue). α5 is expressed in head (h), thoracic midsection (m) and posterior tail regions (t). (B) Serial section to A incubated with control antibody. (C) Higher magnification of thoracic section shown in A. Note lack of α5 expression in epithelium (e) of the pleural cavities (pc) and neural tube (nt), while extensive α5 expression may be seen in mesencyhme, heart (ht), blood vessel (bv) lining and somitic sclerotome (s). Decreased α5 expression relative to sclerotome is seen in the dermomyotome (dm). Scale bars: (A,B) 133 µm; (C) 100 µm.

4 4312 K. L. Goh, J. T. Yang and R. O. Hynes give rise to the posterior structures of the embryo. To address these possibilities, we compared cell proliferation and survival in α5-null and wild-type embryos at E8.5-E9. These stages were chosen for analysis since embryonic defects in the α5- null embryos are visible, though not yet as severe as in later stages, reducing the possibility that differences in proliferation or cell death are due to secondary effects of the α5 mutation. Relative levels of mesodermal cell proliferation were determined using 5 bromo-deoxyuridine (BrdU) incorporation, while relative numbers of apoptotic cell deaths were assayed using terminal transferase biotinylated-dutp nick-end labelling (TUNEL) on sectioned and whole-mount embryos. The results of the BrdU assay indicated no significant differences in cell proliferation between α5-null and wild-type embryos as measured at three different axial levels (Fig. 2). Due to the limitations of the BrdU assay, we cannot, however, completely rule out changes in cell proliferation rates in the α 5 -null embryos (see Discussion). To examine further the effects of the loss of α5 on mesodermal cell cycle, we analyzed percentages of wild-type versus α 5 -null mouse embryonic fibroblasts (MEFs) in G 0 /G 1, G 2 /M and S phases using FACS analysis of DNA content. We observed no significant differences in percentage of α5-null cells in each of the above phases compared to wild-type MEFs (G 0 /G 1 : α5-null, 55.7±2.6%; wild-type, 62.8±7.0%; G 2 /M: α5- null, 18.0±2.6%; wild-type, 18.1±4.8%; S: α5-null, 26.2±2.2%; wild-type, 19.0±4.0%). Cell growth assays (see Materials and Methods), moreover, showed no significant differences between null and wild-type MEFs growth curves (data not shown). These results suggest that α5 integrin does not affect growth or proliferation rates of mesodermally derived cells in culture. null (Fig. 3B,C) versus wild-type embryos (Fig. 3A). Similarly, no significant increases in cell deaths were seen in anterior mesodermal tissues in the mutant relative to wild-type embryos (data not shown). Closer examination of the primitive streak region of the α5-null embryos, moreover, showed no increase in cell death over wild type (Fig. 3C). These results indicate that changes in cell death are not responsible for the mesodermal deficits of the α5-null embryos. Cell deaths in extraembryonic tissues are not responsible for α5-null vascular defects As vascular defects were implicated in causing early death of the α5-null embryos (Yang et al., 1993), we specifically examined apoptosis in the allantois (compare Fig. 3A and B), amnion (Fig. 3D), and embryonic yolk sac (compare Fig. 3E and F). No increases in cell death relative to wild-type embryos were seen in any of these extraembryonic tissues. These results indicate that changes in apoptosis rates are unlikely to be factors in the appearance of the vascular defects in the mutants. Absence of α5 integrin increases cranial neural crest and endodermal cell death Using the whole-mount TUNEL assay, we did observe an increase in the number of apoptotic cells in the branchial arches, in particular the second (hyoid) arch (compare Fig. 4A and B), and in the ventral face of the tail in E8.75 α 5 -null Embryonic mesodermal cell survival is independent of α5 integrin function TUNEL assays showed no significant increase in the low level of apoptosis seen in the posterior mesodermal tissues of α5- % BrdU labeled nuclei/field Head Midsection Tail Region of Embryo wildtype mutant Fig. 2. Quantitation of BrdU labelling. Cell proliferation in three different embryonic regions, as measured by BrdU assay, is similar in α5 null (n=9) and wild-type (n=8) littermates at E8.75-E9. Total numbers of nuclei per section were equivalent between mutant and wild-type embryos. Error bars indicate s.e. Fig. 3. Lack of gross mesodermal cell death in α5-null embyros. Bright-field photomicrogaphs of sagittal sections of E8.75 α5-wildtype (A,E) and α5-null (B-D,F) littermate embryos labelled with TUNEL, counterstained with methyl green. Dark nuclei indicate apoptotic cells (arrowheads). (A,B) Sections of the tail region (m, mesoderm; al, allantois). (C) Primitive streak region (a, artery; g, hind gut; m, mesoderm; np, neural plate). (D) Branchial arch region of α5-null embryo (am, amnion; b1, branchial arch1; b2, branchial arch2; ot, otic pit). Note cell deaths in both superficial and subectodermal regions (arrowheads). (E,F) Sections through embryonic yolk sacs (e, endoderm; m, mesoderm). Note absence of apoptoic cells in both cases. Scale bar, 100 µm, except in B, 90 µm.

5 Defects in α5-integrin null embryos 4313 embryos (Fig. 4C). Wild-type littermates never exhibited apoptotic cells in the ventral tail regions (Fig. 4D and data not shown). Sectioning of the whole-mount TUNEL-stained α5- null embryos indicated that cell death took place specifically in the periphery of the hyoid arch (Fig. 4E) and in endodermal cells surrounding the omphalomesenteric artery in the α5-null embryos (Fig. 4F). TUNEL analysis on embryonic sections confirmed that dying cells in the mutant lay subectodermally and ranged to superficial tissue layers (Fig. 3D). The cells dying in the branchial arches of the α5-null embryos are thus in the appropriate position and time period to be migrating cranial neural crest cells (Hogan et al., 1994). Moreover, double-labelling of TUNEL and neural crest marker CRABP- 1 confirmed the existence of a population of apoptotic neural crest cells in the branchial arches of the α5-null embryos (Fig. 5). Normal mouse embryos do exhibit a basal level of dying neural crest cells in their branchial arches that is age-dependent (Fig. 4; Jeffs et al., 1992; Graham et al., 1993). We therefore compared appropriate developmental stages of α5-null embryos and their wildtype littermates in a statistical study of branchial arch apoptosis. It is, however, difficult to compare the stages of α5-null embryos with those of wild-type littermates by simply using somite counts, as the mutant embryos fail to form the usual number of posterior somites. This staging difficulty was taken into consideration by including all stages of α5-null embryos and their wild-type littermates and doing comparisons of various groupings in the analysis (Fig. 6). The analysis confirmed the statistically significant increase in cell death in the hyoid arches of the α5-null embryos relative to that of wild-type littermates. Our results suggest that increased numbers of cranial neural crest cells undergo apoptosis along their migration pathways in the α5-null embryo, particularly in the hyoid arch region. Survival of endodermal cells of the omphalomesenteric artery is also dependent on the presence of α5 integrin receptor. examined at E8.5-E8.75 and E9.5 stages in α5-null embryos and comparable wild-type littermates. Axial mesoderm determination and development Somitogenesis per se is not blocked in the absence of α5 integrin since anterior somites do condense from presomitic mesoderm in α 5 -null embryos. The lack of posterior somites seen later in the α5-null embryos could be due to a later (temporal) absence of notochordal signals, or to a temporal failure in activation of genes required for presomitic mesoderm determination and/or differentiation (Tam and Trainor, 1994). To address the first point, we examined the expression of notochordal markers, T and Shh, in α5-null embryos. To address the latter point, we examined expression of markers implicated Mesodermal differentiation and development in α5-null embryos An inability of precursor cells to undergo determination and/or differentiation into posterior structures could account for the mesodermal defects, exemplified by the lack of posterior somites, seen in the α5-null embryos. To investigate this, we examined markers for axial mesoderm (T, Shh, HNF-3β) and paraxial mesoderm (mox-1, mouse Notch- 1, pax-1, pax-3), and primitive streak (BMP- 4, Wnt3a) in α5-null embryos, using RNA in situ hybridization. Similarly, to investigate the extent of axis determination in α5-null embryos, we analyzed expression of the HoxB genes. Expression of markers in all cases was Fig. 4. Cell death in the second branchial arch and endoderm of α5-null embryos. Whole-mount TUNEL assay on E8.75 α5-null (A,C) and wild-type (B,D) embryos. Black nuclei indicate apoptotic cells. (A,B) Head regions of the embryos. Arrowheads indicate dying cells in branchial arches. b1, branchial arch 1; b2, branchial arch 2. (C,D) Tail regions of the embryos. Arrowheads indicate equivalent regions in the tails, ventral view. Note the increase of apoptotic cells in the second branchial arch, and in the ventral face of the tail, in the α5-nulls relative to wild-type embryos. (E,F) Histological sections through whole-mount TUNEL-stained E8.75 α5-null embryo, counterstained with eosin. (E) Section through branchial arches. Arrowheads indicate apoptotic cells both in superficial layers (grey arrowheads) and subectodermally (black arrowheads). (F) Transverse section through tail region. a, omphalomesenteric artery; b1, branchial arch 1; b2, branchial arch 2; hg, hindgut; nt, neural tube; s, somite. Scale bars: (A-D) 200 µm; (E,F) 25 µm.

6 4314 K. L. Goh, J. T. Yang and R. O. Hynes n=7 wildtype mutant # apoptotic nuclei n=18 n=8 n=3 n=5 n= # somites Fig. 5. Apoptotic cranial neural crest cells in α5-null embryos. Double labelling of TUNEL (red) and CRABP-1 (green) in cells in the second branchial arch of E8.75 α5-mutant embryo. Overlap of the two signals is seen in yellow. Dashed line outlines otic pit; d, distal and p, proximal regions of arch. Bar, 20 µm. in presomitic mesoderm determination and morphogenesis: homeobox gene Mox-1 and Notch-1. Gene products of both T and Shh are known to play major roles in somitogenesis (Beddington et al., 1992; Chang et al., Fig. 6. Quantitation of neural crest cell death in branchial arches. Significant increase in neural crest cell deaths in second branchial arch region in E8.75 α5-null embryos compared to wild-type embryos. n = number of embryos analyzed. Error bars indicate s.e. 1994). In α5-null embryos,t and Shh were expressed in regions comparable to those in wild-type embryos. T was expressed in the primitive streak and migrating notochordal precursors (Fig. 7A, note the expression of T in anterior notochord in the mutant), while Shh was expressed in notochord, neural floorplate and gut regions (Fig. 7C). As early as E8.5 stage, however, we observed abnormal discontinuities in T (Fig. 7B) Fig. 7. Expression of axial mesodermal markers in α5-null embryos. Whole-mount RNA in situ analysis of (A,B) Brachyury (T), early E8.5; (C-F) Sonic hedgehog (Shh), late E8.75. Note discontinuous T expression in notochord of α5-null embryo seen in (A). The discontinuity can be seen more clearly at higher magnification (B). (C) Ventral and (D) dorsal view of same embryos show discontinuous expression (arrowheads) of Shh in notochord of α5-null embryos. (E) Higher magnification of one such interruption (arrowheads) of Shh pattern in α5-null. (F) Shh expression in notochord of wild-type embryo showing no interruptions, (G) HNF3β expression in posterior of E9.5 α5-null embryo, again showing interruptions (arrowheads). fb, forebrain; h, heart; hg, hindgut; n, notochord; nt, neural tube; ps, primitive streak; /, α5-null embryo; wt, wild-type embryo. Scale bar, 200 µm.

7 Defects in α5-integrin null embryos 4315 Fig. 8. Disorganization of the notochordal structure and gene expression. (A-C) Histological sections of E8.5 α5- null embryo shown in Fig. 7B, counterstained with eosin. (D,E) Sections of α 5-null embryo shown in Fig. 7C. (A) Note normal T expression (blue) in notochord. (B) Serial section to (A) showing complete absence of notochord structure and lack of T staining. Arrowheads in A and B indicate equivalent regions. (C) Section showing presence of notochord tissue but lack of T staining. g, gut; n, notochord; nf, neural fold; nt, neural tube. (D) Note strong Shh expression in notochord, floorplate and gut. (E) Serial section to D showing absence of notochord but presence of floorplate (arrows) and gut staining (grey arrowheads). Black arrowheads in D and E indicate equivalent regions. (F) More caudal section showing presence of notochord but lack of shh expression. g, gut; n, notochord; nt, neural tube; fp, floorplate. Scale bars: (A-C) 50 µm; (D,E) 150 µm; (F) 100 µm. and Shh (Fig. 7C-E) expression in the notochord of the α5 mutants. HNF-3β, whose expression precedes and overlaps that of Shh, also showed discontinuous expression patterns in the notochord (Fig. 7G). Sectioning through embryos labelled in situ for T, we observed that the discontinuities were due both to a loss of the integrity of the notochordal structure (compare Fig. 8A and B), and to a failure in maintenance of gene expression (Fig. 8C, note absence of T staining in intact notochordal structure). Similarly, sections through Shh-labelled embryos also indicate loss of notochord structure (Fig. 8D,E) and failure to maintain expression in intact notochord in more caudal regions (Fig. 8F). Interestingly, despite loss of notochord expression of Shh, floorplate expression of Shh remains present though at a lower level (compare Fig. 8D and E, and data not shown). Expressions of wnt3a in the tail bud, and BMP-4 in primitive streak and lateral mesoderm were Fig. 9. Expression of paraxial mesodermal markers. Wholemount RNA in situ expression of (A) mox-1, E8.5; (B,C) Notch-1, E8.5 (arrowheads indicate expression in presomitic mesoderm); (D,E) pax-1, E8.75 (arrowheads indicate somite pairs with pax-1 expression); (F,G) pax- 3, E9.5; Arrow labelled a indicates anterior direction. d, dorsal; v, ventral. (G) Arrows indicate misexpression of pax-3 in oldest somites; (F,G) arrowheads indicate normal restriction of Pax- 3 to the caudal ventrolateral edge in mature somites. (H,I) Sections of embryos stained for Pax-1. Note normal expression of Pax-1 in the sclerotome (H) is absent in mutant embryo somites (I). (J,K) Sections of embryos stained for Pax-3. Note misexpression of Pax-3 in mutant sclerotome in K as compared to J. Brackets in H-K mark somite positions; dm, dermomyotome; nt, neural tube; s, sclerotome. Scale bars: (A) 400 µm; (B-E) 200 µm; (F,G) 133 µm; (H-K) 33 µm.

8 4316 K. L. Goh, J. T. Yang and R. O. Hynes normal at E8.5 (data not shown). Taken together, these results indicate that notochord formation, and at least one inductive signal from the notochord, initially occur normally in the α5 mutants. As the α5-null embryos developed, however, a progressive degeneration of notochord structure and a concurrent failure to maintain notochordal gene expression occurred. Presomitic mesoderm specification and differentiation The mouse homolog of Drosophila Notch, Notch-1, is expressed in embryonic presomitic mesoderm, but not in condensed somites in the mouse. Mox-1 is expressed in both presomitic and somitic mesoderm. Studies have shown the involvement of Notch-1 in coordination of somite organization (Del Amo et al., 1992; Conlon et al., 1995), while lack of Mox- 1 expression has been correlated with a lack of posterior somites in thewnt3a mutant embryos (Candia et al., 1992). Presomitic and somitic mesoderm of the α5-null embryos, however, expressed a normal pattern of mox-1 (Fig. 9A) and Notch-1 at all stages observed (compare Fig. 9B and C). This suggested that failure of presomitic tissue determination or organization was unlikely to account for the failure to produce posterior somites in the α5-null embryo. Anterior somite development Knowing that somitogenesis is dependent on signals from the notochord, such as Shh, and since Shh expression was disrupted in the notochord in α5-null embryos, we predicted that differentiation of the anterior somites would also be affected in α5- null embryos. We therefore examined the expression of somitic differentiation genes, Pax-1 and Pax-3. Pax-1 is a sclerotomespecific marker whose expression is known to be induced by Shh expression (Johnson et al., 1994; Fan and Tessier-Lavigne, 1994). Pax-3, on the other hand, is commonly viewed as a marker for dermomyotome (Goulding et al., 1991; Bober et al., 1994). Absence of Pax-1 expression or Shh signalling, has been correlated with the expansion of Pax-3 expression (Johnson et al., 1994; Fan and Tessier-Lavigne, 1994). Conversely, ectopic Shh expression increases Pax-1 expression while decreasing that of Pax-3. Consistent with the initially normal expression of T and Shh, we confirmed that Pax-1 (Fig. 9D-E) and Pax-3 (compare Fig. 9F to G, arrowheads) were expressed initially in appropriate regions in anterior somites in the α5-null embryos. As early as E8.75, however, Pax-1 began to show variable expression across somite pairs in the α5 mutants (Fig. 9E) compared with wild-type littermates (Fig. 9D). At E9.5, α5 mutants showed little or no expression of Pax-1 in their somites compared with wild-type controls (compare Fig. 9H and I). Pax-3 expression is normally evenly distributed over newly formed somites while, at later stages of somite development, its expression is restricted to the most caudal ventrolateral edge in each somite and, even later, to the dermomyotome (Goulding et al., 1991; Bober et al., 1994; Fig. 9F). Anterior somites of E9.5 α5 mutant embryos initially exhibited the characteristic caudal ventrolateral restriction of Pax-3 (Fig. 9G, arrowheads) The oldest somites in the mutants, however, showed misexpression of Pax-3 (Fig. 9G, arrows). Sections through α5-null embryos stained for Pax-3, showed ectopic Pax-3 expression in the sclerotome (compare Fig. 9J and K). These results suggest that somites initially develop normally in the α5-null embryos; differentiation or maintenance of the differentiated state of the somites, however, is affected later. Anterior-posterior patterning Hox genes are thought to be mediators of anterior-posterior positional information. Anterior boundaries and extents of expression of these genes along the murine axis as well as their involvement in positional identity of various structures are well documented (Holland and Hogan, 1988; Kessel and Gruss, 1990). The anteriorly expressed Hox genes, Hoxb-1 (Fig. 10A) and Hoxb-4 (data not shown) showed normal expression patterns in the α5-null embryos. α5-null embryos showed the characteristic Hoxb-1 banding pattern at rhombomere 4, and expression in foregut epithelium and posterior neural tube and lateral mesoderm (Conlon and Rossant, 1992; Fig. 10A). Hoxb-1 expression extended normally to the posterior edge of the second branchial arch. Similarly, Hoxb-4 and Hoxb-5 whose anterior boundaries are posterior to that of Hoxb-1, were expressed in appropriate regions in the α5-null embryos (data not shown). The most posteriorly expressed HoxB gene, Hoxb-9, is normally expressed in the neural tube, posterior to the third somite pair level, and in the lateral and somitic mesoderm, posterior to the level of the fifth somite pair (Conlon and Rossant, 1992; Fig. 10B). Anterior boundaries of Hoxb-9 are normal in the α5-null Fig. 10. HoxB gene expression. Whole-mount RNA in situ expression of (A) Hoxb-1, E9.5; (B) Hoxb-9, E9.5. (C) Posterior of E8.5 α5-null embryo expressing Hoxb-9. (D) Posterior of E8.5 wild-type embryo expressing Hoxb-9. Note the lowered expression of Hoxb-9 at the posteriormost region, in neural folds (arrowheads) and lateral mesoderm, of the α5-null embryos relative to wild-type embryos. Scale bars: (A,B) 307 µm; (C,D) 160 µm.

9 Defects in α5-integrin null embryos 4317 embryos. Analysis of several embryos, however, indicated that the overall expression of Hoxb-9 in the α5-null embryos was reduced compared with that seen in wild-type embryos at both E8.5 and E9.5 stages. This reduction in expression of Hoxb-9 was most readily apparent in the posteriormost regions of the α5-null embryos (compare Fig. 10C and D). DISCUSSION We have analyzed the in vivo functions of the α5 integrin subunit through a number of in situ assays on murine embryos deficient for α5 integrin. Our aim was to characterize the basis for the α5-integrin null phenotype by examining the cellular events affected and determining the extent of mesodermal development, in order to address the basis for the mesodermal defects seen in these embryos. We have also examined the possiblity that changes in cell death in extraembryonic tissues were responsible for vascular defects, and hence the retarded growth and death of the null embryos. We have shown that the α5 protein is found throughout the mouse embryonic mesoderm at stages E7.5-E11.5. This suggests that restriction of the α5 mesodermal defects to the embryonic posterior does not reflect restricted expression of α5 in normal mesoderm. Instead, α5β1 may be functionally active only in the posterior regions of the embryo, or (an)other integrin(s) may overlap with or compensate for α 5 function in the anterior of the embryo. Indeed, studies by Yang and Hynes (1996), using embryonic cells deficient in α5β1, have shown that αv integrins can compensate for the functions of α5 in cell adhesion and spreading on FN. Mesodermal cell proliferation and survival We have found that α5 is not essential for the survival or normal proliferation of mesodermal cells. Our measurements of apoptosis in situ contrast with in vitro studies of Zhang et al. (1995), which suggest that α5 protects against cell death in cultured chinese hamster ovary (CHO) cells, perhaps reflecting differences in cell types and cellular environment present in the embryo. Results of the BrdU assay, however, cannot completely rule out changes in cell cycle occurring in the α5-null embryos. BrdU labelling measures the proportion of mesodermal cells in S phase at a given period of time. The assay does not measure changes in the length of the cell cycle, which could cause the lowered mesodermal amounts seen in the α5-null embryos; for example, assuming a cycle time of 10 hours, just a 10% difference in cell cycle length could lead to a two-fold difference in cell numbers over 2 days of embryonic growth. However, results from FACS analysis of cell cycle and cell growth assays using MEFs derived from α5-null embryos indicate that α5 is not essential for normal proliferation and growth of mesodermal cells in culture, supporting the BrdU results. Keeping in mind that the proliferation asaays used here may not detect small differences in a critical subpopulation of proliferative cells, the in situ and in vitro data argue that the mesodermal defects arise for reasons other than gross general changes in cell proliferation or surivival. Cell death in extraembryonic tissues We have investigated the extent of cell deaths in visceral yolk sac, allantois and amnion, as defects in these tissues could lead to retarded growth of the embryo and early lethality due to defective blood circulation. We observed no increase in cell deaths in these tissues in the α5-null embryos, indicating that the phenotype of the mutant is due to reasons other than changes in cell death in these extraembryonic tissues. Alterations in HoxB, axial and paraxial gene expression To determine whether the α5 mutant phenotype was due to defects in the determination and differentiation of mesodermal tissues, we used whole-mount, in situ hybridization techniques to investigate the RNA expression patterns of a number of genes implicated in mesodermal determination and differentiation. Specifically, we looked at expression patterns of HoxB genes, and axial and paraxial mesoderm and primitive streak markers (summary given in Table 1). As the anterior of the α5 mutants looks largely normal, we would expect anterior Hox gene expression patterns to be normal, while more posterior Hox genes would be affected. Our results follow this prediction. Anterior axis markers Hoxb- 1, Hoxb-4 and Hoxb-5 were expressed in appropriate regions. The posteriormost HoxB gene, Hoxb-9, showed reduced expression in both neural folds and in lateral mesoderm. This reduction of expression was most evident in the posteriormost regions of the α5-null embryos. Changes in expression of posterior patterning genes such as Hoxb-9, could plausibly affect the development of posteriormost tissues in the mutants, contributing to the defects seen in the absence of α5. There are several possible explanations for the reduction of Hoxb-9 expression: α5β1 could be regulating the Hox gene s expression; or cells expressing Hoxb-9 may be decreased in α5-null embryos. The latter possibility may be due to the failure of precursor cells to proliferate and/or differentiate to the stage of expressing Hoxb-9 At present, it is not possible to distinguish among these possibilities. We found that a selection of mesodermal markers were expressed in appropriate regions in α 5 -null embryos at E8.5-E9.5. Paraxial markers Notch-1 and mox-1 were expressed normally at all stages observed, suggesting that presomitic determination is normal in the α5-null embryos. T and BMP-4 genes, both implicated in mesodermal determination, were each expressed Table 1. Summary of in situ markers* # / embryos (n) Expression pattern results Axial/primitive streak (ps): bmp-4 n=4 normal in ps and mesoderm Shh n=5 interrupted notochord (nd) T n=7 interrupted nd; normal ps HNF-3β n=2 interrupted nd wnt3a n=4 normal Paraxial: Mox-1 n=4 normal Notch-1 n=2 normal pax-1 n=9 variable/lowered in somites pax-3 n=7 variable/ectopic dorsalization Anterior/posterior axis: hoxb-1 n=3 normal hoxb-4 n=5 normal hoxb-5 n=6 variable hox-9 n=7 reduced *See text for details

10 4318 K. L. Goh, J. T. Yang and R. O. Hynes normally in primitive streak regions in the α5-null embryos, indicating that initial mesoderm specification proceeds in the absence of α5. T staining indicated that notochordal precursors were able to migrate normally in α5-null embryos. T and Shh, however, had interrupted expression patterns in axial mesoderm due to degeneration of notochordal structure and failure of maintenance of gene expression. In α5-null embryos, notochordal Shh can still induce floorplate transcription of Shh prior to notochordal disintegration. However, in the absence of notochord Shh signal, floorplate expression is reduced. As it has been suggested that floorplate Shh expression can feed back to maintain itself (Placzek et al., 1993), perhaps in the absence of sufficient level of notochord Shh signal, the floorplate does not reach Shh expression levels at which it begins selfmaintenance of signal. These notochord marker results, moreover, were consistent with variations seen in somitic differentiation genes, Pax-1 and Pax-3. Pax-1 expression was variable and decreased in older mutant embryos relative to wild type, while Pax-3 expression showed an altered pattern of expression. As discussed earlier, expression patterns of these genes are known to be dependent on the levels of Shh expression. Decreased Shh expression in the α5-null embryos could plausibly lead to the decrease of Pax-1 and the increased extent of Pax-3 expression. Taken together these results support the possibility of a temporal failure of notochordal signalling in the α5-null embryos, contributing to the posterior mesodermal defects seen in the α5 mutants. The expression patterns for Notch-1, mox-1, T and Shh in α5-null embryos are similar to those seen in FN-deficient mice (Georges-Labouesse et al., 1996). FN mutants die at an earlier stage (E9-10) and exhibit more serious mesodermal defects at E8 (George et al., 1993) than do α5-null embryos. They have shortened anterior-posterior axis and a kinked neural tube, fail to form somites and notochord, and display variable defects in the heart and embryonic vessels. Similar to our conclusions concerning α5 function in maintenance of mesodermal derivatives, marker analyses of the FN mutant mice have shown that specification of notochordal and somitic mesodermal lineages is FN-independent but that correct morphogenesis of these structures is FN-dependent (Georges-Labouesse et al., 1996). Basis for the mesodermal deficit in the α5-null embryos In summary, the posterior mesodermal defects in α5-null embryos are unlikely to be caused by a general failure of these mesodermal cells to proliferate or by an increase in cell death, as shown by BrdU, TUNEL and cell culture assays. Our data are consistent with the hypothesis that failure to maintain continuous and/or correct levels of notochordal signalling (Shh, HNF3β), or changes in transcription of T or posterior patterning genes (hoxb- 9), contribute to the posterior mesodermal defects seen in the α5 mutants. One model for the α5 mutant phenotype is that at early stages of α5-null embryonic development, the notochord initially develops normally allowing anterior structures to form; at later stages, the notochordal structure and/or signalling degenerates contributing to the failure of development of posterior structures and non-maintenance of anterior structures. Changes in T expression seen in α5-null embryos are interesting as the α5 mutant phenotype is somewhat reminiscent of that seen in mice lacking T (Beddington et al., 1992). T-deficient mice fail to form somites and notochord, have defects in allantois and mesoderm and die at E10.5. Cell proliferation and death in T/T mice have been shown to be normal, while T/T nascent mesodermal cells have been shown to have an autonomous defect in migration (Wilson et al., 1993). The similarity of α5-null and Brachyury phenotypes may be indicative of common pathways involving their gene products, though α5-null embryos do not appear to show a migration defect in their notochordal precursors (Fig. 7A). However, the causal relationships between notochordal disruptions of T and Shh and the posterior mesodermal defects seen in the α5-null embryos have not been proven definitively. Cranial neural crest and endodermal cell death Our results do suggest that increased numbers of cranial neural crest cells are undergoing apoptosis along their migration pathways in α5-null embryos, particularly in the hyoid arch region. Cell death could be due to failure of these neural crest cells to migrate to their correct destinations. Alternatively, the dying cells may have reached their proper destinations, but subsequently failed to undergo differentiation. The increase in cranial neural crest cell death in absence of α 5 integrin is consistent with results of Beauvais et al. (1995) that show α5β1 integrins are involved in vivo in control of cell migration in avian neural crest pathways. α5β1 has been shown to be expressed in cranial neural crest cells (Wang and Burke, 1995). A number of studies done in vivo in the chicken and in cell cultures have shown that the ligand of α5β1, FN, is a major matrix molecule present in neural crest cell pathways and plays a role in neural crest motility (Erickson and Perris, 1993). Studies are currently underway to study further the effects of lack of α5 on neural crest cell migration, differentiation, development and survival. We also observed cell death in endoderm surrounding the omphalomesenteric artery. This artery leads to the vitelline artery and to the yolk sac vasculature of the embryo (Rugh, 1990). The significance of the cell death in this region is as yet unkown. We can speculate that dying cells could weaken the support of the artery, exacerbating the circulation problems in the embryo, perhaps contributing to the early embryonic death. These results on apoptosis indicate that the dependence of the survival of different cells on specific integrins vary. Some neural crest and endodermal cells are dependent on α5β1 for survival, whereas other cells are not, despite the fact that many of them also express this integrin. It will be of considerable interest to investigate the differing and potentially overlapping roles of different integrins in the survival of specific embryonic cell types. We thank the following for gifts of in situ probes: Pierre Chambon, Bernhard Herrmann, Tom Gridley, Siew-Lan Ang, Peter Gruss, Brigid Hogan, Robert Krumlauf and Andrew McMahon. We are grateful to Kim Mercer and Denise Crowley for histology, Glenn Paradis for FACS analysis, Steve Cornwall for technical assistance and Shona Chattarji for confocal help. We thank members of the lab for helpful discussions, and Laird Bloom, Mike DiPersio, Herbert Haack and Bernhard Bader for critical reading of the manuscript. This work was supported by the Howard Hughes Medical Insitute. K. L. G was supported by predoctoral fellowship from HHMI and R. O. H. is an Investigator of the HHMI. REFERENCES Adams, J. C. and Watt, F. M. (1990). Changes in keratinocyte adhesion during terminal differentiation: reduction in fibronectin binding precedes α5β1 loss from the cell surface. Cell 63, Adams, J. C. and Watt, F. M. (1993). Regulation of development and differentiation by the extracelullar matrix. Development 117,