Developmental expression of the α receptor for platelet-derived growth factor, which is deleted in the embryonic lethal Patch mutation

PDGF (platelet-derived growth factor) is a major mitogen for connective tissue and glial cells in vitro (reviewed by Ross et al., 1986 and by Heldin and Westermark, 1989). It stimulates chemotaxis, it is possibly active in wound healing and it is thought to be involved in the development of atherosclerotic plaques (Rubin et al., 1988; Fems et al., 1991). Abnormal regulation of PDGF, due to mutation or overexpression, can contribute to malignant growth and both the B and A chains of PDGF can cause malignant transformation (Beckmann et al., 1988). A derivative of PDGFB, the v-sis oncogene, transforms cells that express the PDGF receptor (Downward et al., 1984; Heldin and Westermark, 1989). The role of PDGF in cell differentiation was demonstrated in the glial cells of the optic nerve, where PDGF is produced by type 1 astrocytes, acts on bipotential glial progenitor cells, and is required for the timing of their differentiation into oligodendrocytes (Noble et al., 1988; Raff et al., 1988).

PDGF is also expressed in embryonic carcinoma cells (Gudas et al., 1983), m the extraembryonic membranes and in the placenta (Goustin et al., 1985), suggesting that it may have a role in embryogenesis. Numerous other polypeptide growth factors, such as those belonging to the FGF and the TGFβ gene family, are active in mesoderm induction, as detected by the Xenopus animal cap model. PDGF however has no detectable effect in this system (Smith, 1989; Whitman and Melton, 1989), which may be one reason why we know comparatively little about its role in early development.

Biochemical analysis revealed that PDGF and its receptor (PDGF-R) each have two isoforms, designated A and B for the growth factor, α and β for the receptor. PDGF and PDGF-R isoforms function as dimers, and both homo- and heterodimers are formed. The α receptor (Pdgfra) recognizes the A and B chains, whereas the β receptor (Pdgfrb) is specific for PDGF-B. The PDGF AA homodimer is recognized exclusively by the o-o-receptor homodimer (Seifert et al., 1989; Heldin et al., 1988).

Recent evidence drew attention to the involvement of Pdgfra in early development. Mercóla et al. (1989) showed that Pdgfra and PDGF-A are expressed prior to Pdgfrb and PDGF-B in the 7.5- to 8.5-day-old mouse embryo, suggesting that Pdgfra and PDGF-A may have special roles at the primitive streak stage. RFLP experiments by Stephenson et al. (1991) with interspecies crosses revealed that Pdgfra is deleted in Patch (Ph), an embryonic lethal mutation of the mouse. We report here similar results with DNA from homozygous Ph embryos. The genetic data are connected with an analysis of the developmental expression of Pdgfra and this is compared with the phenotype of homozygous Ph embryos. We demonstrate that Pdgfra is expressed in derivatives of the primitive endoderm and mesoderm of early mouse embryos and that its expression follows the differentiation of the two cell layers throughout embryogenesis. We also show that the mutant defects of the Ph homozygote correlates well with the wild-type expression pattern of Pdgfra. Nevertheless the mutant phenotype of our C57BL/6J congenie Ph stock manifested earlier, and the cranio-facial “cleft face” phenotype occurred much less frequently, than in Gnineberg and Truslove’s (1960) original CBA stock, or in BALB/c-based stocks of others (Weston, personal communication). Requirement for Pdgfra receptors during early development and the effect of the genetic background on the Ph phenotype will be discussed.

Breeding nuclei of C57BL/6J-Ph/+ and congenie littermate C57BL/6J mice (4 registered males and 4 females from two litters) were purchased from the Jackson Laboratories (Bar Harbor, Maine). They were propagated by brother-sister mating, since 1989. For experiments they were kept under a 14 hour light, 10 hour dark (from 7.30 p.m. to 5.30 a.m.) regime. The time of pregnancy was established the morning following mating. This was regarded as day 0.5 post coitum (p.c.). Early embryos (6.5 to 8.5 days pc) were prepared for sectioning within the decidua. Later embryos were carefully dissected from their embryonic membranes. For in situ hybridization, the embryos were prefixed in 4% paraformaldehyde at 4°C. For histological investigations, Bouin fixative was used. The viability of the embryos after the age of 8.5 days (p.c.), was ascertained by their beating heart.

For the isolation of the mouse Pdgfra cDNA a c-fms probe was used to screen a mouse (BALB/c) brain λgtl0 cDNA library (a gift from Dr. Yoav Citn of the Weizmann Institute). Our full length murine Pdgfra cDNA is 6576 bp long (GenBank accession N° 84607). This is in good agreement with a single 6.5 kb band detected by RNA blotting in liver, metastatic murine sarcoma and embryo RNA (Do et aL, unpublished data). The murine Pdgfra sequence displays high degree of homology to its human homologue (Claesson-Welsh et al., 1989; Matsui et al., 1989). Amino acid homology between the highly conserved tyrosine kinase domains, is 99%, whereas between the kinase insertion domains, which belong to the least conserved areas of receptor tyrosine kinases, it is 97%. The full length murine Pdgfra sequence of Do et al. (unpublished data) was also compared to the partial cDNA sequence of murine (from F9 tumor cells) Pdgfra, recently communicated by Wang and Stiles (GenBank, accession N° M57683). Nucleotide sequence homology was 99.2% over 3565 bp. The 0.8% difference caused either no change or conservative changes in the amino acid sequence. It was concluded therefore that the cloned cDNA used in the present study is indeed a transcript of the murine Pdgfra locus.

Probes used for in situ hybridization: a 359 bp long cDNA fragment (corresponding to residue 2178 to residue 2536 of the human sequence of Claesson-Welsh et al., 1989) isolated from the interkinase region served as the Pdgfra probe for in situ hybridization. As positive control, c-kit (Orr-Urtreger et al 1990) and Hox-4.1 probes (from the area 3’ to the homeobox; Lonai et al., 1987) cDNA fragments were used. All probes were cloned into Bluescnptll KS+ T3 and T7 polymerasecatalyzed transcripts were synthesized in the presence of ³⁵S-UTP for in situ hybridization. Both sense and anti-sense transcripts were used. No significant signals were obtained with the sense (control) transcript.

Genomic Southern blot hybridization was performed at high stringency (hybridization at 42°C in 50% formamide, post-hybridization wash at 67°C in 0.l×SSC) according to Mamatis et al. (1982). Probes were isolated from the full length murine cDNA (Do et al. unpublished data), they are identified here according to the respective residue number in the human cDNA sequence (Claesson-Welsh et al., 1989) and their position is indicated in Fig. 8A. Probe A: 558–1859; probe B: 1885–4036; probe C: 4037–5200, D: 5191–6576 (note, that the 3’ noncoding regions of the two sequences are highly divergent).

Embedding, cryostat sectioning, post-fixation, hybridization and washing was as previously described (Orr-Urtreger et al., 1990).

Fixation in Bouin fixative and embedding in paraffin was as described by Kaufman (1990). Serial sections of 4 to 7 μm thickness were stained with hematoxylin-eosin or with light green.

In contrast to c-kit (Orr-Urtreger et al., 1990) and similarly to two FGF receptors, bek and flg, (Orr-Urtreger et al., 1991) the expression of Pdgfra was not detectable in ovarian oocytes (not shown). At the egg-cylinder stage, Pdgfra was expressed in the parietal and visceral endoderm, but not in the primitive (embryonic) ectoderm or trophectoderm (Fig. 1A-F). Fig. 1 shows a 6.5-day- (p.c.) old embryo before mesoderm induction and before the separation of the proamniotic cavity into the three cavities characteristic of the late primitive streak stage. It can be seen (Fig. 1B,F) that the extraembryonic visceral endoderm accumulated more Pdgfra transcripts than the embryonic part, which however is definitely positive, as it is apparent also in sections across the embryonic region (Fig. 1D). Pdgfra is expressed in the visceral endoderm-like derivatives of activated embryonic carcinoma and embryonic stem cell cultures (Schatteman et al., 1992 and our unpublished results). In addition to the visceral endoderm, Pdgfra was also expressed in the parietal endoderm (Fig. 1B,D), a closely related derivative of the primitive endoderm (Gardner, 1983).

The mesoderm first appears in the posterior portion of the egg cylinder, it exfoliates through the primitive streak and invades both its embryonic and extraembryonic halves. Fig. 2A-D demonstrate that Pdgfra is expressed in the embryonic and extraembryonic mesoderm and also in the visceral and parietal endoderm of the 7.5 day p.c. embryo. Pdgfra expression in the embryonic part of the visceral endoderm was greatly diminished by this stage and no hybridization was detectable in the neural plate-pnmitive ectoderm.

At around 8-8.5 days p.c., the neural folds start to close in the occipital region and a distinct head fold develops. At the same time along the longitudinal body axis paraxial mesoderm cells aggregate into pairs of cell blocks, the somites, and, ventrally from the head fold, the heart anlage forms. Parasagittal sections of 8.0-day-old embryos show that Pdgfra is strongly expressed both in the early, non-segmented and in the more mature, segmented paraxial mesoderm (Fig. 2E-H). No hybridization signals were detectable in the definitive endoderm of the foregut or in the neuroectoderm (Fig. 2F,H). Pdgfra at this stage was however transcribed in the extraembryonic mesoderm, in the blood islands in the head mesenchyme and in the heart anlage.

Starting with the formation of the first somites (Fig. 2E,F) and then throughout development, Pdgfra expression follows certain aspects of somite differentiation, as illustrated in Fig. 3. First the homogeneous somitic blocks separate into dermomyotome and sclerotome. The myotome will form skeletal muscle, whereas vertebrae and other chondrogenic elements of the skeleton develop from the sclerotome. Pdgfra is expressed in all or most cells of the undifferentiated somitic block (Figs 2F and 3B). Later, however, Pdgfra transcription concentrates in the dermatome and sclerotome, and no hybridization is observable in the adjacent myotome (Fig. 3C,D).

During advanced organogenesis, the sclerotomes condense into prevertebrae and the chondrification of the vertebral body begins. Differentiation of these axial structures follows an anteroposterior direction. Thus in the tail of a 14.5-day-old embryo shown in Fig. 3E-J, the somites still appear as homogeneous blocks and they display strong accumulation of Pdgfra transcripts (Fig. 3I,J). At anterior positions, where axial development is more advanced vertebrae and intervertebral discs form at this stage. Fig. 3G,H show that the intervertebral discs accumulate high levels of Pdgfra transcripts, whereas in the body of the differentiating vertebrae, as in prospective ossification centers at other locations, their amount is greatly diminished.

Another differentiated somitic derivative that expresses Pdgfra, the dermatome, gives rise to the dermis, the mesodermal component of the skin. It can be seen in Fig. 3K that Pdgfra expression concentrates to the dermis and no transcripts are seen in the tail epidermis of a 14.5-day embryo.

At 9.5 days of development (Figs. 3A,B, 4A–D, 5A,B) Pdgfra transcripts accumulated in the lateral and splanchnic mesoderm. This signal was more prominent in the vicinity of the forelimb bud and in the anterior part of the lateral plate (Figs 4B, 5B). Pdgfra transcripts were absent from the neuroectoderm. An exception at this stage was a weak hybridization signal in the dorsal region of the posterior neural tube (see small arrows in Figs 3B and 4B) and in the yet unclosed, caudal neural folds (Fig. 4C,D). This may suggest involvement of Pdgfra in early neural crest development and migration. Additional examples for this possibility are the Pdgfra transcripts observed along the ventral side of the neural tube (Figs 4B,H, 5B), that co-localize with the ventral pathway of neural crest migration (Rickmann, 1985). Sagittal and transverse sections of 10.5-day-old embryos (Fig. 4E-H) demonstrate the expression of Pdgfra in the mesenchyme of the face and branchial arches, a destination for neural-crest-derived cells, which adds further support to the above suggestion.

In the early forelimb bud (9.5 days p.c.), Pdgfra transcription was unevenly localized (Fig. 5A,B). Concentration of the transcripts increased proximally, towards the body wall. Similar observations were made also in the limb bud of 10.5-day-old embryos. Later, at 11.5 days p.c. (Fig. 5C,D), Pdgfra expression significantly decreased at the site of mesenchymal condensation, which precedes chondrification in the center of the limb. Later the anlagen of the limb bone undergo chondrogenesis leading to osteogenesis. As can be seen in Figs 5G-I and 6E,F, Pdgfra is not expressed in these chondrification (carpals and digits) and ossification centers (humerus, ulna and radius). Whereas the two FGF receptors, bek and flg, occupy characteristic positions at the same sites during this stage of development (Orr-Urtreger et al., 1991). Pdgfra in contrast is detected in the mesenchymal tissue around the bone anlage and also in the joints between the prospective bones (Fig. 5H). Absence of Pdgfra expression in the centers of bone formation was also detected in the vertebral column and sternum (Figs 3H, 6D,G). This suggests that the PDGF system is less involved in bone development than the FGF system. It rather seems to contribute to the development of structures, which connect and surround prospective bones, like muscles and joints. Pdgfra transcripts were detected also in the apical ectodermal ridge (AER) of the forelimb (Fig. 5C-F) as early as 9.5 days p.c. and one day later in the hind limb (not shown), showing further involvement of Pdgfra in limb development.

Sagittal and transverse sections of 12.5- (Fig. 6A,B) and 14.5- (Fig. 6C-F) day-old embryos demonstrate the expression of Pdgfra in the mesenchymal components of numerous organs. Transcripts can be detected in the facial mesenchyme with a clear border of expression in the epithelium of the nose and mouth (Fig. 6), and in the mesenchyme of the lung, bladder, kidney and genital ridge (Fig. 6B,D). Cross-sections of the bronchi and midgut loops (Fig. 7A-D) show transcripts accumulating between the epithelium and the outer layer, in the supporting mesenchyme of these tubular structures. This pattern is consistent with Pdgfra expression in the early precursors of these structures. Figs 3B and 4B show that the mesenchyme of the hindgut copiously transcribes this receptor. Another example of Pdgfra activity in mesenchymal components is its expression in the supporting capsules of the kidney, adrenal gland and thymus and in the central part of the diaphragm (Fig. 6B,D).

In the heart, Pdgfra expression was detected from the earliest stages of its development. At 8 days, Pdgfra hybridization was detectable in mesenchymal cells around, but not in, the heart tube epithelium (Fig. 2F,H). At later stages (10.5 days p.c.: Fig. 4F,H; 14.5 days p.c.: Figs 6D,F, 7F) transcripts were detected mostly in the pericardium. They were consistently detectable in all four heart valves of 12.5- and 14.5-day-old embryos (Figs 6E,F, 7E,F). This specific expression in the epithelial elements of the heart indicates that Pdgfra may have a role in cardiac valve development. Similarly it was expressed in the wall of the major blood vessels (dorsal aorta, umbilical and branchial arch arteries) of 9.5- and 10.5-day-old embryos (Fig. 4). The PDGF system was implicated in the pathogenesis of atherosclerosis (Ross et al., 1986). Experimental evidence demonstrates selective induction of PDGF-A synthesis during the regeneration of arterial injury (Rubin et al., 1988; Fems et al., 1991). Transcription of the PDGF α receptor in the inner wall of developing major arteries and in the heart may have bearing on these observations.

As has been mentioned earlier, very little or no Pdgfra expression was detectable in the neuroectoderm before 12.5 days p.c. An exception to this was the expression of Pdgfra in the neural fold. By 12.5 and 14.5 days, Pdgfra expression did became detectable in the central nervous system (CNS). Sporadic signals could be seen both in the vertebral column and in the fore-, mid- and hindbrain (Fig. 6). At higher magnification, the signals were found in the cell body surrounding some, but not all nuclei of the developing central nervous system (Fig. 6G-I). Strong hybridization was observed in the differentiating choroid plexus (Fig. 6H,I), which develops from the roof of the fourth ventricle and expressed Pdgfra already at day 10.5 p.c. (Fig. 4H).

Homozygous Ph mutants were obtained from C57BL/6J-Ph/+-females crossed to C57BL/6J-Ph/+ males. More than 1000 embryos were investigated between days 6.5 and 12.5 of gestation (Table 1). Most died by 11.5 days p.c. and hemorrhagic ‘deciduomata’ were found in their place (Table 1, lanes 1 and 2). Earlier, up to 10.5 days p.c., approximately 1/4 to 1/3 of the embryos were distinguishable by their small size, enlarged, fluid-filled pericardium, subepidermal blisters and thick and convoluted yolk sac (Fig. 10). Gr üneberg and Truslove (1960) in their original report also described a later phenotype, which was distinguished by craniofacial and neural tube defects. Among 647 embryos investigated between days 11.5 and 12.5 p.c., only four survived over 11.5 days p.c. These four were the only ones that corresponded to Gr üneberg and Truslove’s (1960) description of the “cleft face” embryo (Table 1, line 2). The frequency of cleft face embryos in our population was 0.6%, which was more than fivefold less than the normal frequency of fetal defects in the wild type at the same gestational age (see line 2 in Table 1).

Southern blot hybridization of DNA from 9.5- and 10.5-days-old homozygous Ph embryos showed the presence of a genomic deletion, which included most or all of the Pdgfra locus (Fig. 8). Four cDNA fragments were used as probes. Fig. 8B and C demonstrate hybridization with the 5’ coding region probe, A, and with the 3’ non-coding region probe C. It can be seen that both were deleted in Ph/Ph DNA. However, the same DNA digests gave normal RFLP with other gene-probes, like c-kit and Hox-4.1. Approximately 10 kb of genomic DNA upstream from the initiator AUG was also used for hybridization. Probes from this source did not reveal rearrangement in heterozygous Ph/+ DNA, indicating that the upstream deletion breakpoint may be farther from Pdgfra than 10 kb. Additional evidence for the deletion came from PCR assays of individual 9.5- to 10.5-day-old Ph/Ph embryos. Embryos distinguished by their small size, enlarged pericardium, wavy neural tube and subepidermal blebs consistently failed to contain Pdgfra sequences amplifiable with primers from the interkinase region, in contrast to their normal littermates (data not shown).

In agreement with a genomic deletion encompassing most of the Pdgfra locus, homozygous Ph mutants failed to transcribe this receptor. Figs 9H-J and 10F-H demonstrate examples of the in situ hybridization analysis of 7.75- and 10.5-day-old Ph/Ph embryos, respectively. It can be seen that these embryos displayed no hybridization with the Pdgfra probe. In contrast, the adjacent maternal decidua (Fig. 9I) transcribed Pdgfra and the Ph mutant embryo expressed the bek receptor (Fig. 9J; Orr-Urtreger et al., 1991). Taken together, our data lend support to those of Stephenson et al. (1991) and demonstrate that the early Ph embryo is associated with a null mutation of Pdgfra.

Experiments with homozygous mutant embryos sectioned in the decidua, also allowed us to distinguish zygotic and maternal Pdgfra transcription. Hybridization in the wild type (Fig. 9G) demonstrated strong, aggregated signals in the area of mural trophoblastic giant cells. Much less signal could however be seen in the same area of the homozygous mutant (Fig. 9I), suggesting that most of the mural giant cells are of embryonic origin. In addition the same figures show increased maternal Pdgfra transcription in the decidua distal to the mutant embryo, as compared to wild type. This could be a compensatory reaction connected to interactions between trophectoderm and maternal decidua (Goustin et al., 1985).

To analyze the Ph phenotype, serial sections were prepared from 165 Ph/+ × Ph/+ embryos collected between 6.5 and 11.5 days p.c. Table 2 demonstrates that the mutant phenotype, distinguished by convoluted visceral endoderm, was first visible at 6.5 days, when its frequency was significantly less than 1/4. By 7.5 days p.c., a quarter of the embryos had this phenotype (see also Fig. 9B-D) and between days 8.5 and 10.5 several disorganized partially autolytic embryos were seen. This indicates that, after 7.5 days p.c., the Ph homozygote in our crosses start to lose viability and only a minority survives beyond 10.5 days p.c.

One of the major morphological characteristics of these embryos was their relatively hypertrophic and convoluted visceral endoderm. The exocoelom and the ectoplacental cavity were separated into numerous compartments, and the amniotic cavity, surrounded by the primitive ectoderm and mesoderm, was compressed (Fig. 9B-D). A change in cell morphology could also be observed in the visceral endoderm. The columnar vesiculated cells of the extraembryonic visceral endoderm descended more towards the embryonic region of the egg-cylinder in the mutant embryo, where they in part replaced the flattened squamous cells of the wildtype embryonic visceral endoderm (Dziadek, 1978). This distinction could be also seen in transverse sections (Fig. 9E,F). They also revealed abnormalities in the mesoderm; in some Ph embryos the mesoderm was irregularly shaped, whereas in others, like in Fig. 9F, it failed to develop altogether.

Next 9.5- and 10.5-day-old homozygous Ph embryos were studied. Most striking was their very small size (Fig. 10A) and that they were found within hypertrophic, cauliflower-like yolk sacs (Fig. 10E). They also had a greatly extended fluid-filled pericardium, characteristic subepidermal blisters, retarded limb buds and branchial arches, open abdominal walls with enlarged umbilical cord and wavy neural tube (Fig. 10A-D). These mutant embryos, like the earlier phenotype failed to transcribe Pdgfra (Fig. 10G), whereas the unrelated Hox-4.1 gene was transcribed normally (Fig. 10H). Histological sections revealed small, irregular somites in the homozygous mutant, which did not differentiate into myotome, dermatome and sclerotome (Fig. 10D). The interstitial spaces contained few and comparatively sparsely spaced mesenchymal cells and the waviness of the neural tube extended to its entire length (Fig. 10C).

The developmental localization of Pdgfra has general characteristics common to other receptor tyrosine kinases, like c-kit (Orr-Urtreger et al., 1990) and the two FGF receptors, bek and flg (Orr-Urtreger et al., 1991). All four are expressed in more than one major cell lineage soon after implantation and they stay active throughout development (and sometimes also during adult life) in some, but not all, differentiated derivatives of the lineages where they were first detected.

The Pdgfra receptor is first expressed in derivatives of the primitive endoderm, and its transcripts are detectable in the parietal and visceral yolk sac also during later embryogenesis. Pdgfra is not expressed in the primitive ectoderm or in the definitive endoderm. However, it is copiously expressed in the nascent mesoderm, and later in the paraxial and lateral mesoderm, though apparently not in the notochord. During mesoderm differentiation this receptor is expressed in all cells of the early mesenchymal somitic block, but later only in the sclerotome and in the dermatome of the ensuing dermomyotome. With the formation of vertebrae, Pdgfra concentrates to the intervertebral disc and it is not detectable in the chondrification centers of the developing vertebral body. An additional characteristic is its localization in the parietal, and visceral endoderm of the primitive streak embryo. The functional importance of this localization gams support from the abnormal development of the yolk sac in homozygous mutants surviving into midgestation.

During the first half of gestation no Pdgfra expression could be detected in the central nervous system. A probable exception was its expression in the dorsal part of the neural tube and in the caudal neural folds, which may point to involvement with early neural crest development. This is consistent with Pdgfra’s later expression in the branchial arches and in facial structures of neural crest origin. Because Pdgfra was not expressed in the peripheral nervous system (dorsal ganglia), it is possible that during early stages of neural crest differentiation, this receptor could be preferentially expressed in non-neuronal derivatives of the neural crest, as was suggested before by Weston and his colleagues (Weston, 1991; Morrison-Graham et al. 1992).

We have observed sporadic Pdgfra expression in cells of the spinal cord and brain at the later stages investigated in this study (12.5, 14.5 days p.c.). A decisive part of mammalian CNS development does not start until midgestation and the definitive brain structure is built largely during the prenatal and early postnatal period. Our findings may indicate the involvement of Pdgfra in the beginning of this definitive period of CNS maturation. Recent reports describe the involvement of both PDGF isotypes (A and B) in multiple neuron and glial cells of the developing and mature CNS (Yeh et al., 1991; Sasahara et al., 1991). Additional studies are required for a comprehensive evaluation of PDGF and PDGF-R expression in the development of the brain and spinal cord.

The expression pattern of Pdgfra is distinctly different from that of the related FGF receptors. This could be seen already at the egg-cylinder stage, bek and flg were first expressed in the primitive ectoderm and they retained more or less extensive ectodermal and neuroectodermal expression throughout development (Orr-Urtreger et al., 1991). In contrast, Pdgfra was expressed in the visceral and parietal endoderm and later in the nascent mesoderm. Mesodermal expression of bek and flg was initiated only later, just prior to mesoderm segmentation. If in addition we take into account that FGFs are involved in mesoderm induction, whereas no such indications were found for PDGF (Whitman and Melton, 1989), we may conclude that the respective contribution of the FGF and PDGF systems to early development is essentially different. The expression pattern of Pdgfra and the two FGF receptors was also distinguishable during limb development and osteogenesis. Pdgfra was expressed in the mesenchyme and in the AER of the early limb bud, but not in the chondrification centers, where transcripts of the two FGF receptors could be found (Orr-Urtreger et al., 1991).

The developmental mechanism served by the PDGF system is not known. A few clues however may have emerged from our data. Visceral endoderm differentiation was specifically altered in the early Ph mutant. Its extraembryonic cellular phenotype prevailed in areas where the embryonic phenotype is normally found (Fig. 9). Previous evidence suggests that the alternative extraembryonic or embryonic cell fate of the visceral endoderm depends on its interaction with the trophectoderm and the primitive ectoderm, the cell layers with which it comes into contact (Dziadek, 1978). Another example for the effect of Pdgfra on cell fate may be its expression in the AER and its absence from the chondrification centers of the developing limb. This localization may indicate the contribution of Pdgfra in the AER’s inhibitory effect on cartilage differentiation (Solursh and Reiter, 1988). Both examples suggest that Pdgfra may be involved in interactions by which one cell layer influences the differentiation of another (Saunders, 1972). Support for this interpretation was found when the localized expression of Pdgfra and its ligand, PDGF-A, were compared. We found that the two are localized in separate but adjacent cell layers during primitive streak stage development, axial skeleton development and organogenesis. Hence their localization suggested that this hormone-receptor pair may serve interactions across cell layers (unpublished data).

Our studies on the genomic structure of the Ph allele support and extend those of Stephenson et al. (1991). They studied interspecies crosses. Our results were obtained with DNA from homozygous Ph embryos. They suggest that most or all of the Pdgfra gene is deleted in this mutation. Using the full-length Pdgfra cDNA as probe, Stephenson et al. (1991) observed laboratory mouse specific RFLP-s in Ph/Mus spretus heterozygote DNA, which was interpreted to suggest that the deletion, or at least one breakpoint of it, is within the Pdgfra locus. We could not confirm this observation, because no upstream breakpoint was found within 10 kb 5’ of the Pdgfra initiator methionine and we have not been able to use the full-length cDNA for RFLP analysis, because of a 900 bp long reiterated Bl type repeat in its 3’ end. Hence at present the extent of the Ph deletion is not known, therefore, in a strict sense, we cannot rule out the possibility that the deletion may involve other genes besides Pdgfra.

The homozygous Ph phenotype was investigated here in considerable detail. The genotype of the embryos was ascertained by in situ hybridization and by PCR applied to individual embryos. Diagnostic for the early mutant phenotype was the abnormality of the visceral endoderm and mesoderm at the egg-cylinder stage. Embryos with abnormal visceral endoderm could be detected in our Ph/+ × Ph/+ material as early as 6.5 days p.c. and the full complement of the Ph phenotype became apparent by 7.5 days p.c.; however, by 10-11 days of gestation, most homozygous Ph embryos died. Those that survived suffered from defects in the yolk sac and had functionally connected defects in somite and neural tube development (Watterson et al., 1955; Van Straaten et al., 1989), all of which correlate with Pdgfra expression during early development. Thus, there was a good correlation, both in time and space, between the deletion phenotype and the developmental expression of Pdgfra. Hence we conclude that the normal Ph allele is required for visceral endoderm and mesoderm differentiation.

One of our observations is at variance with Grüneberg and Truslove’s (1960) original description of the Ph mutation. In our hands, the ‘cleft face’ phenotype was considerably less frequent. We think that this was due to differences in the genetic backgrounds used to maintain the Ph allele by the various research groups. We used the C57BL/6J-Ph/+ strain from the Jackson Laboratories (Bar Harbor, Maine). Although the Griineberg and Truslove (1960) stock originated in C57BL, most of their original studies of the Ph mutation were performed on the CBA backgound. In another Ph stock, which was crossed to BALB/c mice over a long time and many generations, cleft face embryos with Pdgfra deletion were obtained regularly (Weston and Bowen-Pope, personal communication). It appears, therefore, that the Ph phenotype may be sensitive to the effect of unlinked alleles residing in various inbred backgrounds.

We thank Dr Jim Weston and Dr Dan Bowen-Pope for sharing their results with us before publication. Our thanks are due to Dr Jean-Louis Guenet (Institut Pasteur) for discussion and advice. We are gratefull to Mr David Barkan for the photographic work. This research was supported by the Raphael and Regina Levy Fund, by the Pasteur-Weizmann Joint Research Programme and by the Centers for the Biology of Aging, Molecular Genetics and Neurobiology of the Weizmann Institute