Loss of Eph-receptor expression correlates with loss of cell adhesion and chondrogenic capacity in Hoxa13 mutant limbs

The condensation of mesenchymal cells is critical to the formation of organs and skeletal elements in vertebrates (Hall and Miyake, 1992; Hall and Miyake, 1995). In the developing limb bud, the mesenchymal condensations that give rise to the skeletal elements form from an undifferentiated mass of mesoderm derived from the progress zone. Genetic analysis has demonstrated that members of the HoxA and HoxD linkage groups are essential for this process (Dollé et al., 1993; Davis et al., 1995; Kondo and Duboule, 1999). For example, mice homozygous for a loss-of-function mutation in Hoxa13 show major perturbations in the formation of the autopod skeletal elements and mice doubly mutant for both Hoxa13 and Hoxd13 fail to form autopod skeletal elements (Fromental-Ramain et al., 1996). Complementing these loss-of-function studies in mice, Yokouchi et al. (Yokouchi et al., 1995) demonstrated that misexpression of Hoxa13 in the chick proximal limb alters homophilic cell-cell interactions of mesenchyme derived from these limb buds in vitro and formation of the zeugopod skeletal elements in vivo. More recently, Wada et al. (Wada et al., 1998) identified a role for glycosylphosphatidylinositol-linked (GPI) ephrin ligands and their receptor, EphA4, in mediating self-sorting of undifferentiated mesenchyme into distinct proximal and distal limb skeletal structures. Hox genes belonging to the 3′-end of the HoxA and HoxB linkage groups appear to regulate expression of EphA receptors that are involved in hindbrain rhombomere boundary formation and axon guidance in this region (Taneja et al., 1996; Chen and Ruley, 1998; Dottori et al., 1998; Studer et al., 1998). These receptor genes are also expressed at high levels in the developing autopod (Ciossek et al., 1995; Flenniken et al., 1996; Wada et al., 1998). Taken together, these data suggest that EphA receptors may be targets of Hoxa13 and provide a molecular mechanism for how boundaries may be established during the formation of autopod mesenchymal condensations, as well as during the formation of the umbilical arteries (UAs). To test this hypothesis, we have examined how loss of Hoxa13 function affects patterning of the mesenchyme in the autopod and UAs, and specifically how EphA ligand/receptor expression and function are affected by this mutation.

The Hoxa13 replacement vector was constructed from 10 kb SphI-EcoRI genomic fragment containing the Hoxa13 locus. The GFP loxP neo cassette (Godwin et al., 1998) was cloned into a unique EcoRV site located in the second exon of Hoxa13. Continuity of the reading frame across the Hoxa13/GFP junction was verified by DNA sequencing. Finally the herpes simplex virus thymidine kinase gene was inserted 3′ of the Hoxa13 genomic fragment for negative selection (Mansour et al., 1988).

The targeting vector was electroporated into R1 embryonic stem cells (Nagy et al., 1993) and the cells subjected to positive and negative selection (Mansour et al., 1988). Genomic DNA from 144 colonies was digested with SphI and analyzed by Southern transfer analysis to identify two homologous recombinants. Cells from both recombinant clones were microinjected into C57BL/6 blastocysts to generate chimeric mice. Chimeric males were crossed to C57BL/6 females and DNA from Agouti progeny was tested by Southern blotting and PCR analysis to confirm germline transmission of the mutant Hoxa13^GFPneo allele (Fig. 1C). Embryos from heterozygous intercrosses were genotyped by PCR using yolk-sac DNA. PCR conditions consisted of 35 cycles of 94°C for 15 seconds, 62°C for 15 seconds and 72°C for 35 seconds. Mutant (378 bp) and wild-type (177 bp) amplification products (Fig. 1D) were produced using the Hoxa13 forward primer (GTCGTCTCCCATCCTTCAGAC), the GFP reverse primer (GCACTGCACGCCGTAGGTCA) and the Hoxa13 reverse primer (TGTTCTGGAACCAGATTGTGAC; Fig. 1B). The neomycin resistance cassette was removed by crossing Hoxa13^GFP heterozygous females to Cre deleter mice (Schwenk et al., 1995).

Embryos derived from heterozygous intercrosses were harvested at gestational ages ranging from embryonic day (E) 9.5 to 15.5 and placed in Leibovitz’s L-15 media lacking Phenol Red (Gibco/BRL). Imaging of dissected limbs, tissue culture samples and intact embryos used either Sylgard plates or depression slides, depending on sample size. GFP, CY5, TO-PRO-3 and Texas red fluorescence were recorded using a BioRad MRC 1024 laser scanning confocal imaging system fitted to a Leitz Aristoplan microscope using filters supplied by the manufacturer. A digital Kalman averaging filter was used with the GFP and Texas Red fluorescence to reduce background fluorescence.

Staged limb buds (Wanek et al., 1989) from mutant and heterozygous embryos at gestational age E11.5 were collected in 4°C Ca²⁺- and Mg²⁺-free phosphate-buffered saline (PBS; Gibco/BRL). Enrichment for Hoxa13^GFP expressing cells was achieved either by microdissection of the fluorescent regions of the limb bud using fine tungsten needles and a fluorescence equipped Leica MZ12 stereoscope or by cell sorting using a Becton Dickinson FACS ADVANTAGE cell sorter. The two enrichment methods produced similar results. Dissected tissues were dissociated at 37°C for 10 minutes in Ca²⁺- and Mg²⁺-free PBS (Gibco/BRL) containing 0.1% trypsin and 0.1% collagenase as described (Owens and Solursh, 1982), followed by pipetting three to five times through a sterile 70 μm cell basket (Costar). The cell suspension was washed twice using 5 ml of serum free Dulbecco’s MEM media (Gibco/BRL) and counted on a hemocytometer. The cell suspensions were diluted to a final concentration of 2×10⁷ cells/ml with Dulbecco’s MEM media containing 15% FBS supplemented with non-essential amino acids, 50 U/ml penicillin and 25 ug/ml streptomycin. Cell suspensions were inoculated onto 60 mm Falcon tissue culture dishes as described (Ahrens et al., 1977) and placed in a 37°C incubator with a 6% CO₂ atmosphere for 1 hour to allow for cell attachment. After 1 hour the plates were flooded with 8 ml of media which was changed daily. Mesenchyme aggregations were recorded 19 hours after plate inoculations, whereas cartilaginous nodules were analyzed 3-4 days later.

Hoxa13^GFP-expressing tissues from the autopods of E12.5 embryos were dissected and dissociated with 0.05% trypsin-EDTA and 0.05% collagenase followed by passage through 70 μm Netwell filters (Costar) to generate single-cell suspensions. The single cell suspensions were plated at a density of 2×10⁷ cells/ml in 20 μl aliquots in 60 mm tissue culture dishes. A second 20 μl aliquot containing either 4 μg of a rabbit EphA7 polyclonal antibody dialyzed for 4 hours in 1× PBS (Santa Cruz Biotechnology Cat. No. SC-917), pre-immune rabbit serum, or DMEM media was immediately added to the micromass cells which were then incubated at 37°C for 1 hour. After incubation, 7 ml of DMEM media supplemented with 10% fetal calf serum and 50 U/ml penicillin, and 25 μg/ml streptomycin was added to each dish. A second aliquot of the EphA7 antibody, or rabbit serum was added to the respective dishes in conjunction with the 7 ml of media. The cells were grown in a standard 37°C incubator at 5% CO₂. Micromass cultures were photographed 48 hours after plating under phase contrast using a Leica DMIL microscope fitted with a Nikon Coolpix 990 digital camera.

Proximal limb bud tissues (400-500 μm from the AER) from stage matched E11.5 Swiss Webster embryos were dissected and processed for micromass culture as described above. At the same time, the fluorescent regions of stage matched E11.5 limb buds from Hoxa13^GFP mutant and heterozygous embryos were dissected and processed for micromass culture. Stock suspensions of wild-type Swiss Webster cells, Hoxa13^GFP mutant cells and heterozygous cells were adjusted to equivalent concentrations of 4×10⁷ cells/ml. From these stock suspensions, micromass plates were inoculated with equal concentrations of wild-type and either mutant or heterozygous cells as described (Ahrens et al., 1977; Owens and Solursh, 1982; Ide et al., 1994). Nineteen hours after inoculation, the attached cells were stained with TO-PRO-3 iodide (Molecular Probes) and analyzed for proximal-distal sorting of wild-type and Hoxa13^GFP-expressing cells using the previously described BioRad Confocal Imaging System.

E11.5 limbs were fixed at room temperature in Carnoy’s fixative for 2 hours followed by rinses in 1× PBS containing 0.5% Triton X-100 (PBX) for an additional 2 hours. Tissues were then placed in a PBX blocking solution containing 4% skim milk and 2% donkey serum, and rocked slowly at 4°C for 4 hours. NCAM (5B8) and collagen type II (II-II6B3) mouse antibodies (Developmental Hybridoma Bank, University of Iowa, Iowa City, IA) were diluted in PBX containing 2% skim milk and 2% donkey serum, and incubated with the limb tissue overnight at 4°C. The primary antibody was removed and the limbs were washed for 3 hours in PBX at room temperature. After washing, the limbs were incubated overnight at 4°C with a donkey anti-mouse secondary antibody conjugated with either Texas Red or Cy5 (Jackson Immunological). Confocal analysis was performed as described above.

For micromass cultures, the media was removed and the anchored cells were washed briefly in 1× PBS followed by fixation in 4% formaldehyde for 10 minutes. After fixing, the cells were again rinsed in 1× PBS for 20 minutes, followed by permeabilization and blocking in PBX containing 2% skim milk and 2% donkey serum for 1 hour. After blocking, primary antibodies, including EphA2, EphA4 and EphA7 (Santa Cruz Biotechnology), ephrin A1, ephrin A2 and ephrin A3 (Santa Cruz Biotechnology), N-Cadherin (Transduction Labs), β-catenin (Transduction Labs), Paxillin (Transduction Labs), E-Cadherin (Transduction Labs), antiphosphohistone H3 (Upstate Biotechnology), and D1C4 (a kind gift from Anthony Capehart), were incubated with the micromass cultures in blocking solution overnight at 4°C. The next day, the primary antibody was removed, and the micromass cultures were washed for 3 hours at room temperature in 1× PBS, followed by an overnight incubation at 4°C in blocking solution with a species specific Cy5-conjugated donkey secondary antibody.

For cryosections, embryos were fixed for 2 hours in 4% paraformaldehyde and briefly rinsed in 1× PBS for 10 minutes. Cryoprotection of the embryos was achieved using a sequential series of 10, 20 and 30% sucrose/PBS solutions. The embryos were oriented in OCT (Tissue Tek) filled molds and rapidly frozen. The embryos were sectioned on a Zeiss Cryostat at a thickness of 20-30 μm. Sections were mounted on Superfrost plus slides (Fisher) for microscopic and immunohistochemical analysis.

Plasmids containing the EphA7 (Mdk1) and Msx1 genes generously provided by Drs T. Mori and Y. Chen, respectively, were used to produce antisense riboprobes. Embryo preparation, hybridization and analysis were performed as previously described (Manley and Capecchi, 1995). BM-Purple (BMB-Roche) was used for the alkaline phosphatase color reactions, which were extended for 17 hours at room temperature for the EphA7 riboprobe.

Terminal UTP nick end labeling (TUNEL) of DNA was performed as a modification of the technique described by Maden et al. (Maden et al., 1997). Limbs from E11.5-13.5 embryos were fixed at room temperature for 2 hours in 4% paraformaldehyde. After fixing, the limbs were washed for 3 hours at room temperature with several changes of 1× PBS containing 1% TritonX-100. The limbs were placed in 1.5 ml microfuge tubes and pre-incubated at 37°C for 30 minutes in PBS containing 1× terminal transferase buffer (Boehringer Mannheim Biochemical), 1% Triton X-100 and 2.5 mM CoCl₂.. The preincubation buffer was replaced with a 1× PBS solution containing 1× terminal transferase buffer, 10 μM dUTP (2:1 dUTP:dUTP-biotin), 2.5 mM CoCl₂, 1% Triton X-100 and 0.5 U terminal transferase/μl buffer, and incubated at 37°C for 3 hours. The limbs were then washed for 3 hours in PBS and incubated overnight at 4°C with streptavidin conjugated with Texas Red (Jackson Immunological). After washing for 1 hour in PBS, confocal analysis of the limbs was performed as described above.

The loss-of-function allele, Hoxa13^GFPneo, was targeted into the Hoxa13 locus of mouse ES cells by homologous recombination (Capecchi, 1994) (Fig. 1A-D). The floxed neomycin resistance cassette was removed from the germline of animals generated from these ES cells by breeding to a Cre- deleter mouse (Schwenk et al., 1995), generating mice with the Hoxa13^GFP allele. Mice carrying either Hoxa13 mutant allele showed similar phenotypes in heterozygous and homozygous form. The studies reported herein predominantly used mice and embryos carrying the Hoxa13^GFP allele.

Hoxa13^GFP expression was characterized in E10.5-E13.5 embryos to confirm accurate recapitulation of the published Hoxa13 mRNA expression pattern (Haack and Gruss, 1993; Warot et al., 1997). As early as E10.5, strong expression can be seen in the distal forelimb mesenchyme, but not in the overlying apical ectodermal ridge (AER) (Fig. 2A). By E 11.5, Hoxa13^GFP expression is readily detected in the progress zone (Fig. 2B) of the elongating limb bud. By E13.5 limb expression of Hoxa13^GFP appears to be restricted to the condensing digit mesenchyme, at the sites where many of the chondrogenic defects occur in Hoxa13 homozygous mutant mice (Fig. 2C,D) (Fromental-Ramain et al., 1996). Finally, the GFP-tagged allele of Hoxa13 identifies the population of mesenchymal cells expressing Hoxa13 in the developing umbilical arteries (Fig. 2E,G,H) and genitourinary regions, including the genital ridge and tissues surrounding the urogenital sinus (Fig. 2F). Our Hoxa13^GFP-null homozygotes die in utero between E11.5 and E15.5, exhibiting limb and umbilical vascular defects similar to those in mice homozygous for the Hoxa13 mutant alleles as described previously (Fromenthal-Ramain et al., 1996; Warot et al., 1997). Examination of umbilical arteries from homozygous mutants (Fig. 2H) revealed nearly a complete loss of mesenchymal/endothelial cell layer stratification relative to heterozygous littermate controls (Fig. 2G).

Enriched populations (∼95% purity) of Hoxa13^GFP-expressing mesenchymal cells, either heterozygous or homozygous for the Hoxa13^GFP mutation, were isolated by FACS (see Materials and Methods) from E11.5-E12.5 limb buds. These cells were compared in vitro for phenotypes attributable to the loss of Hoxa13 function. The first property examined was their ability to form condensations in culture, using the mass culture assay developed by Solursh and colleagues. This assay accurately recapitulates prechondrogenic and chondrogenic events associated with normal limb development (Ahrens et al., 1977; Owens and Solursh, 1982).

Although care was taken to plate identical numbers of Hoxa13^GFP heterozygous and homozygous mutant cells, it is apparent that after 12 hours in culture, fewer mutant homozygous cells adhere to the culture dish and initiate aggregation, compared with Hoxa13^GPF heterozygous cells (Fig. 3A,B). Eighteen to 24 hours after plating, the Hoxa13 heterozygous mesenchymal cells aggregate into multiple dense precartilaginous condensations (Fig. 3C). Seventy-two hours later, these aggregates differentiate into cartilaginous nodules (Fig. 3I) that stain positively with Alcian Blue (Fig. 3I, inset) and express collagen type II (Fig. 3J). With further in vitro culture, the Hoxa13-expressing cells become restricted to the periphery of the condensate, forming digit-like structures (Fig. 3K). This restriction in Hoxa13^GFP expression to the periphery of the digit-like elements in vitro is noteworthy, because during the later stages of normal chondrogenesis in the developing limb bud, 5′-Hox gene expression becomes restricted to the perichondrium, a layer of cells that surrounds each condensate and controls its pattern of growth (Davis and Capecchi, 1994). The same sequence of differentiation is observed with unsorted mesenchymal cells isolated from wild-type embryonic limbs. However, approximately 24 hours of additional culture time is needed with the wild-type, unsorted mesenchymal cells isolated from limb buds to reach the equivalent state of differentiation. The precocious behavior of the Hoxa13^GFP heterozygous cells relative to wild-type cells may be due to the former cells being a more purified and distinct (i.e. FACS-purified and Hoxa13-expressing) population of mesenchymal cells.

In sharp contrast to either Hoxa13^GFP heterozygous or wild-type mesenchymal cells derived from embryonic autopods, Hoxa13^GFP mutant homozygous mesenchymal cells form fewer precartilaginous aggregates (i.e. approximately eightfold less), and those that do form are often smaller and less robust than those prepared from control mesenchyme (Fig. 3D).

To determine whether Hoxa13 homozygous mutant cells can participate in the production of condensates in the presence of wild-type cells, equal numbers of dissociated wild-type cells (GFP negative) and homozygous mutant cells (GFP positive) were combined and plated onto culture dishes. The combined cells produced large mesenchymal condensates containing fluorescent (–/–) and nonfluorescent (+/+) cells (Fig. 3E,F). Examination of these mixed aggregates revealed that the mutant (–/–) cells contributed to the condensing mesenchyme by attaching to anchored wild-type cells (Fig. 3G,H). Both mutant and wild-type cells within these aggregates subsequently express collagen II. Thus, in the presence of wild-type cells, Hoxa13^–/– cells are capable of forming aggregates and differentiating along the chondrogenic pathway. This suggests that a major deficit of Hoxa13^–/– mesenchymal cells is in their inability to self-aggregate and adhere to the culture dish.

In homozygous mutant and wild-type limbs, no quantitative differences in the expression of the following cell surface and pro-adhesion molecules were detected at E11.5-E13.5: paxillin, β-Catenin, E-cadherin, EphA2, EphA4, ephrin A1, ephrin A2, D1C4, desmoglein, NCAM, fibronectin, catenin, collagen Type II and integrin B1 (data not shown).

In E13.5 limbs, EphA7 mRNA is localized to the condensing digit and carpal/tarsal mesenchyme (Fig. 4A-D). In mutant fore- and hindlimbs, levels of EphA7 expression were consistently lower in all digits, with even lower levels of expression in the regions where digit I and digit V are derived. In the forelimb, an additional region of EphA7 expression was also present in the dorsalmost aspect of each digit, delineating the region where tendons differentiate (Fig. 4A). However, in mutant forelimbs, only digit II appeared to demonstrate EphA7 expression in the dorsal tendon region (Fig. 4B).

The reduction of EphA7 expression is also apparent from immunohistochemical analysis of frozen sections of mutant versus control autopods (Fig. 4E-J). Interestingly, the very low level of EphA7 expression in the mutant autopods is not restricted to the mesenchyme, where Hoxa13 is normally expressed, but is also observed in the overlying epidermis. From analysis of these sections, it is apparent that the mesenchyme in the mutant homozygotes is disorganized and that the boundaries between the mesoderm and epidermis are poorly defined. We consistently observe that the epidermis in these mutant animals is thicker than normal and distorted (Fig. 4L). By contrast, the distal digit mesenchyme in age-matched heterozygous controls forms a distinct boundary of Hoxa13^GFP-expressing cells immediately adjacent to the overlying epidermis (Fig. 4K). Mesenchyme condensations also appear more organized in heterozygous digits, forming proximal to the boundary layer of Hoxa13^GFP-expressing cells (Fig. 4E-H).

Consistent with the above observations, the marked reduction in EphA7 expression is also observed in vitro, comparing micromass mesenchymal cultures prepared from mutant homozygous and heterozygous control embryonic autopods (Fig. 4M-P).

The distribution of the ephrin ligand A3, which binds with high affinity to the EphA7 receptor (Janis et al., 1999), was examined in the developing autopod of homozygous Hoxa13^GFP mutant and heterozygous embryos. By E13.5 the expression of ephrin A3 and Hoxa13^GFP in heterozygotes is restricted to the perichondrial region (Fig. 5A,C,E). However, in mutant limbs, this perichondrial restriction is not apparent, as both HoxaA13^GFP and ephrin A3 are expressed throughout the developing digit and interdigit regions (Fig. 5B,D,F). Notably, the cells contributing to the digit condensations were discernable in the mutant limb sections under bright field microcopy (Fig. 5G,H), and the cells defining the perichondrial boundary were not visible, suggesting that Hoxa13 is involved in establishing the boundary between the perichondrium and the adjacent condensing mesenchyme.

A blocking antibody specific for the extracellular domain of EphA7 was used to correlate how reductions in EphA7 expression may contribute to the changes in cell adhesion and mesenchymal condensation exhibited by Hoxa13^GFP mutant (–/–) mesenchymal cells. Incubations of heterozygous (+/–) limb mesenchymal cells with either pre-immune rabbit sera or DMEM media, had no affect on the capacity of these cells to adhere and aggregate into chondrogenic nodules in vitro (Fig. 6A,B,D,E). By contrast, the addition of the anti-EphA7 antibody dramatically reduced both cell adhesion and chondrogenic nodule formation of Hoxa13^GFP heterozygous cells. This reduction in cell adhesion and aggregate formation in the antibody treated samples of heterozygous cells closely resembles the aggregation and cell adhesion efficiencies of Hoxa13^GFP mutant (–/–) cells, which, because of their low levels of EphA7 expression, were unaffected by the blocking antibody (compare Fig. 6G,H with 6C,F,I).

Because EphA receptors and their ligands may regulate mesenchymal cell sorting along the proximal-distal axis of the autopod (Wada et al., 1998), we asked whether the loss of Hoxa13 affects the ability of distal cells to sort themselves in vitro from condensates formed from more proximal wild-type mesenchyme. In micromass cell cultures prepared from equal numbers of dissociated mesenchymal cells of distal, Hoxa13^GFP heterozygous cells (green) and more proximal wild-type cells, the GFP-labeled cells formed tight aggregates that excluded wild-type cells (Fig. 7A,C,E). By contrast, in similar aggregates prepared from mixtures of wild-type cells and Hoxa13^GFP mutant homozygous cells, the GFP-labeled cells were not able to sort themselves from the wild-type cell aggregates (Fig. 7B,D,F).

In 37% of adult Hoxa13^GFP heterozygotes (n=106) there are soft tissue fusions between hindlimb digits II and III (Fig. 8A,B). In mutant homozygotes, no interdigital separation was apparent between any of the forelimb or hindlimb digits of living embryos at stages E13.5 and E14.5 (data not shown). TUNEL assays for apoptosis in E13.5 heterozygous and mutant homozygous embryos revealed reduced and undetectable levels of apoptosis between digit II and III of Hoxa13^GFP mutant heterozygous and homozygous embryos, respectively, compared with wild-type controls (Fig. 8D-F). Confocal microscopic examination of sectioned hindlimb tissues revealed the ectopic presence of Hoxa13^GFP -expressing cells in the interdigital region of Hoxa13^GFP heterozygous and homozygous animals, respectively (Fig. 8G,H). The presence of ectopic Hoxa13^GFP expressing cells in the interdigital region of the mutant embryos is suggestive of a failure in cell sorting and maintenance of appropriate cellular boundaries.

Detection of normal expression levels of Msx-1 in the mutant limb interdigital tissues (Fig. 8K,L) suggests that the loss of programmed cell death (PCD) in these tissues is not the result of delayed growth and development. More likely, the loss of PCD in Hoxa13^GFP mutant limbs reflects changes in either the capacity of these tissues to respond to apoptotic signals or in the levels of pro-apoptotic factors required for interdigital PCD.

Hoxa13^GFP homozygous mutant embryos demonstrate a consistent narrowing of the umbilical arteries as early as E10.5, with complete ablation of the lumen of one of the vessels occurring by E13.5 in all (4/4) mutant embryos examined (data not shown). Confocal microscopy of sections containing the umbilical arteries revealed a high level of Hoxa13^GFP expression in the condensing mesenchyme and differentiating endothelial layers (Fig. 2E,G,H). At higher magnification, cells forming the vascular wall appear highly disorganized in Hoxa13^GFP mutant homozygotes (Fig. 2H). Indeed, the stratification of vascular mesenchyme and endothelium was absent in mutants, whereas heterozygous littermate controls demonstrated normal mesenchymal and endothelial layer formation (Fig. 2G).

Immunohistochemical examination of proteins involved in vasculogenesis revealed a significant reduction in EphA4 and EphA7 expression in the condensing vascular mesenchyme and endothelium of Hoxa13^GFP mutant homozygotes (Fig. 9). These data demonstrate the levels of EphA4 and EphA7 expression are reduced in the mutant tissue, and again that the abnormal cellular organization of the mesenchyme and endothelium that form these vessels. It is important to note the changes in morphology exhibited in the mutant umbilical arteries does not create a generalized reduction in UA marker expression as even the ligand for EphA4 and EphA7, ephrin A3, is normally expressed in the mutant vessels (Fig. 2H, inset). Interestingly, both EphA4 and EphA7 are still expressed normally in the gut, ventral neural tube and dorsal root ganglia in these same mutant homozygous embryos (data not shown).

Cell sorting, adhesion and aggregation represent some of the earliest stages in patterning undifferentiated mesenchyme (Takeichi, 1991; Gumbiner, 1996). During formation of the limb bones, this follows a stereotypic progression of condensation, branching and segmentation generating a pattern of prechondrogenic condensations, which subsequently give rise to each of the bony elements (Shubin and Alberch, 1986). In Hoxa13^GFP mutant homozygotes, the condensations needed to form the entire first digit of the fore- and hindlimbs are absent, and the prechondrogenic condensations needed to form carpal, metacarpal, tarsal and metatarsal bones are not resolved. The residual capacity to form some of the autopod condensations in the Hoxa13^GFP mutant homozygotes is explained by the compensating expression of Hoxd13, as in embryos homozygous mutant for both genes, no autopod condensations are formed (Fromental-Ramain et al., 1996).

We have shown that Hoxa13^GFP -expressing cells, which lack functional Hoxa13 protein, are unable to attach efficiently to culture dishes and to form large condensates. They remain detached, cease to divide and die after a few days in culture. By contrast, Hoxa13^GFP -expressing cells that retain Hoxa13 function (i.e. are heterozygous for the Hoxa13 mutation) form robust prechondrogenic condensations, which in culture progress along the chondrogenic differentiation pathway.

Hoxa13^GFP -null mesenchymal cells can form mixed prechondrogenic condensates with wild-type mesenchymal cells and then undergo chondrogenic differentiation. However, when Hoxa13^GFP -null cells are plated in the presence of wild-type mesenchymal cells isolated from a more proximal region of the limb bud, they do not sort themselves from the proximal cells. Thus, Hoxa13 may regulate genetic pathways that control the adhesive properties of the distal limb mesenchyme and in the process, confer positional value along the proximodistal axis to this population of cells. Previous studies in chick (Yokouchi et al., 1995) support this conclusion, as misexpression of Hoxa13 in the proximal limb bud alters proximal mesenchymal cell identity in the opposite manner, appearing to confer more distal limb character onto cells that normally do not express Hoxa13.

In the condensing digit mesenchyme, EphA7 expression closely parallels the expression of Hoxa13^GFP . At E11.5 both Hoxa13^GFP and EphA7 are expressed strongly throughout the dorsal mesenchyme (this work) (Araujo et al., 1998). By E13.5, the expression of Hoxa13^GFP and EphA7 remains coincident in the perichondrium of the condensing forelimb digits. Interestingly, ephrin A3 expression also overlaps with Hoxa13^GFP and EphA7 in both E11.5 and E13.5 limbs, with expression finally restricted to the perichondrial region (Fig. 5C). This co-localization suggests a role for Hoxa13 in mediating perichondrial boundary formation by positively regulating EphA7 expression, which in turn interacts with ephrin A3 to delineate the perichondrial boundary. The Eph receptors and their membrane-bound ligands are ideally suited to establish such boundaries, as their interactions provide the repulsive signals necessary to restrict intermingling between heterogeneous mesenchymal cell populations (Ciossek et al., 1995; Valenzuela et al., 1995; Wada et al., 1998). A similar role for this interaction may be evident in the reinforcement of the boundaries between rhombomeres of the vertebrate hindbrain (Taneja, 1996; Mellitzer et al., 1999; Xu et al., 1999).

In the autopod, reduced EphA7 expression in conjunction with alterations in the expression of ephrin A3 provides a molecular mechanism to explain both aberrant cell adhesion and defects in limb patterning associated with loss of Hoxa13 function. In the distal limb, differential cell sorting is required to organize the undifferentiated mesenchyme into morphologically distinct musculoskeletal domains. Finally, antibody blocking of EphA7 demonstrates that this receptor has the capacity to regulate autopod mesenchymal cell aggregation and chondrogenic nodule formation in vitro, providing a mechanistic link between the regions affected by loss of Hoxa13 function and the reduction of EphA7 expression.

Because Hox genes have the capacity to regulate genes at multiple levels of a developmental cascade (Weatherbee et al., 1998), it is possible that other pro-adhesion molecules involved in autopod patterning are regulated by Hoxa13. Although we did not see a reduction in EphA4 expression in the mutant autopod, other EphA family members may be involved in this process. The residual capacity of homozygous mutant cells to attach and form some chondrogenic nodules, as well as our observation that EphA7 expression is not completely lost in intact Hoxa13^GFP homozygous mutant embryos, leaves open the possibility that additional proteins, such as Hoxd13, function in parallel with Hoxa13 to pattern the autopod. Finally, our observation that EphA7 expression is reduced in homozygous mutant limbs may suggest an indirect role for Hoxa13 in mediating EphA7 expression. However, our in vitro characterization of EphA7 expression in FACS-enriched homozygous mutant cells, indicates that EphA7 expression is absent. This finding, in conjunction with normal levels of Msx-1 expression in intact mutant limbs, suggests that changes in EphA7 expression in the mutant autopod are more likely the result of direct regulation in specific tissues.

TUNEL assays of homozygous mutant limbs at E11.5 and 12.5 did not reveal significant increases in programmed cell death relative to control embryos (data not shown) which indicates that the reduced capacity to form the appropriate autopod prechondrogenic condensations observed in Hoxa13 mutant homozygotes is not explained by increased apoptosis. On the contrary, this study revealed a loss of normal apoptosis in the interdigital regions of the autopod in Hoxa13 mutant homozygotes and reduced apoptosis between digits II and III of embryos heterozygous for the Hoxa13 mutation. The reduced level of apoptosis provides an explanation for the persistence of interdigital tissue in heterozygous and homozygous Hoxa13 mutants. Normally, Hoxa13 expression becomes restricted to the digit perichondrium by E13.5. However, in both heterozygous and homozygous Hoxa13 mutant embryos, Hoxa13-GFP labeled cells are observed in the interdigital tissue (Fig. 7G,H). This observation provides another example of the reduced ability of Hoxa13 mutant cells to sort themselves properly, resulting in the failure to establish an appropriate boundary between Hoxa13-expressing and nonexpressing tissues. The reduction or absence of normal apoptosis in the interdigital region of these mutant embryos might in turn follow from a failure of the ectopic Hoxa13-expressing cells to respond to apoptotic signals. Alternatively, the persistence of interdigital tissues could also result from changes in cell proliferation. An examination of mitotic cells in interdigital tissue using a mitosis-specific marker, anti-phosphohistone H3, did not reveal differences in the number of mitotic cells (data not shown).

The high level of Hoxa13 expression in the condensing mesenchyme and endothelial layer of the UA suggests a major role for Hoxa13 in the formation of these vessels. Narrowing of the UA is observed as early as E10.5 in Hoxa13^GFP homozygous embryos, with complete unilateral closure being seen in all of the E13.5 embryos examined. It is important to note that during murine gestation, the exiting UAs fuse into a single vessel outside the embryo proper (Kaufman and Bard, 1999), whereas humans typically maintain two UAs during gestation. This difference might explain why stenosis of a single UA may not be life threatening in midgestation to humans (Pavlopoulos et al., 1998), whereas stenosis of the UA in mice would severely impact the volume of blood that reaches the placenta in mice, particularly if the occlusion occurred proximal to or outside of the umbilical ring.

Analysis of the vessel walls of E13.5 embryos reveals distinct mesenchymal and endothelial cell types in heterozygous and wildtype littermates. The UA walls of Hoxa13 mutant homozygotes, on the other hand, are disorganized with no stratification into morphologically distinct cell types. The loss of EphA4 and EphA7 expression in the developing UA wall of Hoxa13 mutant homozygotes by E11.5 may well explain the loss of cellular organization in the formation and maintenance of the boundaries between vascular mesenchyme and endothelium. Numerous studies have suggested the interaction of ephrins and their receptors during vasculogenesis and angiogenesis (Pandey et al., 1995; McBride and Ruiz, 1998; Wang et al., 1998; Adams et al., 1999, Ogawa et al., 2000).

In summary, a prominent feature common to both the limb and umbilical artery defects observed in Hoxa13 mutant homozygous embryos is a failure to organize and properly pattern the mesenchyme within these developing tissues. We have provided evidence that mesenchymal dysfunction, in part, be understood as defects in mesenchymal cell aggregation and adhesion. A loss of ephrin receptor expression was common to both tissues. These observations suggest that Hoxa13 directly or indirectly controls the aggregation, adhesion and sorting capacity of the mesenchymal cells within the autopod in part by regulating the ephrin/Ephrin receptor signaling system. Modulation of this signaling system would ensure that the appropriate cellular boundaries are formed and maintained during morphogenesis of these tissues. This role of Hoxa13, though not necessarily its only role in the formation of the distal limb skeletal elements and umbilical vasculature, nevertheless begins to explain many of the cellular defects observed in these mutant animals.

We thank M. Allen, S. Barnett, C. Lenz, G. Peterson, K. Lustig, S. Nguyen and V. Scott for expert technical assistance. In addition, we thank Rebecca Reiter for helpful suggestions regarding the establishment of the Hoxa13^GFP micromass cultures as well as L. Oswald for help in preparation of this manuscript. Part of this research was supported by grants to HSS from the March of Dimes (Basil O’Connor Award) and the NIH (1R01DK59150-01).