The Hβ58 transgenic mouse line carries a recessive insertional mutation that results in developmental abnormalities beginning at day 7.5 p.c. and embryonic arrest at about day 9.5. In this paper, we describe the characterization of a novel gene encoded at the Hβ58locus, whose disruption appears to be responsible for the mutant phenotype. The wild-type Hβ58 gene encodes a single 2.7 kb mRNA during embryonic and fetal development, and in many adult somatic tissues. In the mutant locus, this transcription unit is split by the transgene insertion, and one of its coding exons is deleted. Consistent with the physical disruption of the gene, the level of the Hβ58 mRNA in heterozygous mutant mouse tissues was half the normal level, indicating that the mutant allele fails to encode a stable mRNA. In situ hybridization studies revealed that expression of the wild-type Hβ58 gene begins in the oocyte, and continues throughout pre- and post-implantation embryogenesis, despite the fact that homozygous mutant embryos develop successfully through the egg cylinder stage (day 6.5 p.c.). In the early post-implantation embryo, expression of the normal Hβ58 gene is relatively low in the embryonic ectoderm, the tissue displaying the earliest phenotypic effects of the mutation, and highest in the visceral endoderm. We therefore propose that the effects of the mutation on the embryonic ectoderm may be exerted indirectly, via the visceral endoderm. Sequence analysis of Hβ58 cDNA clones revealed no homology between the 38 ×103MrHβ58 protein and other known proteins. However, the Hβ58 gene displayed extremely strong conservation between mammals and birds (>96% amino acid identity), although it appeared less conserved in amphibians and invertebrates.
The normal course of mammalian embryogenesis depends on the correct temporal and spatial regulation of a large number of genes, as well as on complex regulative interactions between tissues. One strategy to identify genes that are expressed in the early embryo and, concurrently, to gain insight into their function in embryogenesis is to generate and characterize mutations that interfere with early murine development (reviewed by Magnuson, 1986; Gridley et al., 1987; Rossant and Joyner, 1989; Reith and Bernstein, 1991; Gridley, 1991). When retroviruses or transgenes are introduced into the mouse germ line, they appear to integrate in a largely random manner and they occasionally disrupt a host gene at the locus of insertion. Consequently, 5 –10% of retroviral or trans-gene insertions are associated with recessive mutations with discernable developmental consequences (Pal-miter and Brinster, 1986; Gridley et al., 1987). Although many such mutations have been identified, only a handful so far have been successfully characterized at a molecular level, leading to the identification of the mutant gene (Mov-13, Schnieke et al., 1983; limb deformity, Woychik et al., 1985,1990; Maas et al., 1990; Mov-34, Soriano et al., 1987; Gridley et al., 1991; Mpv-17, Weiher et al., 1990; dilute, Mercer et al., 1991). With the exception of limb deformity, those cases in which a mutant gene has been identified involved a retroviral insertion event (dilute, Mov 13, Mov-34, Mpv-17).
We have recently described a recessive lethal mutation in transgenic Une Hβ58 (Costantini et al., 1985, 1989), which was caused by the integration of 10 –20 tandem copies of an 8 kb transgene and an accompanying 2 –3 kb deletion of host DNA on mouse chromosome 10 (Radice et al., 1991). Homozygous mutant H(558 embryos developed normally through the egg cylinder stage, day 6.5postcoitum (p.c.). However, at day 7.5 p.c., the primitive streak stage, the homozygotes began to display reproducible abnormalities, the earliest of which was a severe reduction in the size of the embryonic ectoderm, the progenitor of all future embryonic tissues. Homozygotes subsequently became increasingly abnormal and, while forming recognizable derivatives of all germ layers, they never exceeded the size of a normal embryo at an early somite stage (8.5 days p.c.).
Among the extraembryonic lineages, the amnion and chorion of 7.5 day homozygous mutant embryos sometimes displayed abnormal folding, suggesting over-proliferation of these tissues, although at 8.5 or 9.5 days the amnion appeared normal. The allantois developed normally through day 8.5, but it failed to fuse with the chorion to form the chorioallantoic placenta. The visceral endoderm also appeared normal from its inception through the subsequent development of the visceral yolk sac (Radice et al., 1991). Thus, based on the mutant phenotype, we concluded that the gene(s) inactivated by the Hβ58 insertion was not required for development prior to the egg cylinder stage, and that its earliest function was likely to involve the development of the embryonic ectoderm.
Using molecular probes derived from this locus, we detected the presence of a transcription unit active during embryonic development (Radice et al., 1991). In this paper, we describe the identification of a previously unknown gene that has been physically disrupted by the Hβ58 insertion, and we characterize the normal pattern of expression of this gene during embryogenesis, as well as its evolutionary conservation. We discuss the possible relationship between the developmental expression pattern of the Hβ58 gene and the phenotype of mutant embryos lacking the gene.
Materials and methods
Isolation and analysis of cDNA clones
Genomic probe C (Fig. 1c) was used to isolate the cDNA clone λE30.1 from an 8.5 day p c mouse embryo library (Fahmer et al., 1987) The cDNA insert was subcloned in the plasmid pBluescript KS(+) (Stratagene), to yield clone pE30.1. Clone pE30.1 was used as a probe to screen a chick dl3 embryo brain cDNA library (a gift of Barbara Ranscht, La Jolla Cancer Research Foundation), and the longest chicken cDNA, pCK4.4 (2493 nt), was subcloned in pBlue-script KS(+).
Preparation of probes
Southern and Northern blots were hybridized with the insert DNA from cDNA clone pE30.1, which was isolated from agarose gels by electroelution and 32P-labelled by nick translation to specific activity ∼6 × 108 disnts/min/ μg For molecular titration assays, the 1.2 kb EcoRI fragment of cDNA clone pE30.1, containing the 3’ untranslated region, was subcloned into pBluescript KS(+) to produce clone pE1.2. This DNA was linearized with Hzndlll and transcribed using T7 RNA polymerase (Lee and Costlow, 1987), yielding a 32P-labelled RNA probe of ∼6 × 108 disintÿmin/gg In situ hybridizations were performed using 35S-labelled RNA probes (∼2 × 109 disints/min/ μg) derived from cDNA clone pE30.1. The probes were synthesized using T7 or T3 RNA polymerase (Pharmacia) from templates linearized with HindIII (anti-sense probe) or BamHI (sense probe), and were reduced in size to 100 –200 nucleotides by limited alkaline hydrolysis (Cox et al, 1984).
Northern blot analysis
Northern blot analyses were performed essentially as described by Posakony et al. (1983). Briefly, 10 pg of poly(A)+ RNA was fractionated by gel electrophoresis on 1 2% agarose gels containing formaldehyde, and transferred to Gene-Screen-plus (NEN-Dupont). The filters were hybridized in a buffer containing 50% formamide, 5 × SET, l × Denhardts, 20 mM PB, 1% SDS, 100 μg/ml salmon sperm DNA, and 3 –5 ng/ml of nick-translated probe, at 42°C overnight. The filters were washed at 65°C, first for 3 × 20 min in 4 × SET, 0.2% SDS, then for 3 × 20 min in 2x SET, 0 2% SDS, and finally for 3 × 20 mm in 0.2 × SET, 0.2% SDS. Hybridization was visualized by autoradiography using Kodak AR-5 X-ray film.
Molecular titration of Hβ58 transcripts
Transcript prevalence in tissue RNAs from wild-type and Hβ58 heterozygous mice was determined using a molecular titration assay with an RNA probe (see above), according to the procedures described by Lee and Costlow (1987). Briefly, hybridizations were carried out at 50°C, 0.40 M [Na+] in 50% formamide. The quantity of RNA-RNA hybrid in each sample was assayed by digestion with RNase A and RNase Ti in 0.375 M [Na+] followed by precipitation in 5% trichloroacetic acid.
DNA sequence analysis
DNA fragments derived from the murine cDNA pE30.1, and from chicken cDNA pCK4.4, were subcloned in pBluescript KS(+) Single- or double-stranded clones were employed in sequencing reactions (Sanger et al., 1977) using Sequenase (US Biochemical). The sequence data were analyzed using the program PRONUC (Devereux et al., 1984).
Sample preparation for in situ hybridization
Adult ovaries and testes were obtained from C57BL/6J mice, and embryos from natural matings between inbred FVB/N mice. Midday on the day of the vaginal plug was considered day 0 5 p.c Embryos or tissue samples were fixed at 4°C overnight in freshly prepared 4% paraformaldehyde-1 × PBS. The fixed samples were dehydrated in an ethanol series, cleared with xylene, and embedded in paraffin. Sections were cut at 6 –8 μm, collected on TESPA-treated slides (Berger, 1986), air dried, and stored at 4°C. Blastocysts, too small to be processed independently, were inserted into the ampulla dissected from a pregnant female (0.5 day p.c.) before fixation and embedding. Pretreatment, hybridization, and post-hybndization washes of the slides were performed as described (Toth et al., 1987; Wilkinson and Green, 1990). For autoradiography, the slides were coated with Kodak N IB 2 emulsion, exposed for 1 –4 days at 4°C, developed in Kodak D19 and stained with H and E.
Cross-species genomic blot hybridization
Genomic DNAs were digested with EcoRI, fractionated on a 0.7% agarose gel, and transferred to nitrocellulose (Schleicher and Schuell). The blot was hybridized overnight with 10 ng/ml of probe, at 37°C, in a buffer containing 43% formamide, 5 × SSC, 20 mM phosphate buffer, l × Denhardts, 0.2% SDS, 10% dextran sulphate, 100 μg/ml sonicated, denatured salmon sperm DNA (McGinnis et al., 1984). The blot was washed at 50°C for 4 × 30 min in 2 × SSC, 0 2% SDS.
cDNA cloning of a mRNA encoded at the Hβ58 locus
In order to identify and characterize the gene disrupted by the Hβ58 insertion event (Fig. 1b), we used a genomic fragment from the locus (probe C, Fig. 1c) to screen a cDNA library prepared from 8.5 day p.c. mouse embryos (Fahrner et al., 1987). This probe had been previously shown to hybridize to a 2.7 kb mRNA in 8.5 day p.c. embryos, while four other probes from the locus, (A,B,D,E in Fig. 1c) failed to hybridize (Radice et al., 1991). Several overlapping cDNA clones were isolated, the longest of which was the nearly fulllength (2.6 kb) clone, pE30.1. Hybridization of clone pE30.1 to restriction digests of genomic λ-clones from this region, as well as sequence analysis of the cDNA and genomic clones, indicated that the cDNA included sequences both 5’ and 3’ to the inserted transgene array, as well as within the deleted region (Fig. 1d, data not shown). Therefore, the cDNA represented a gene that had been structurally disrupted by the Hβ58mutation, which we will refer to as the Hβ58gene.
To examine initially the expression of this gene during normal murine embryogenesis, we used cDNA pE30.1 to probe Northern blots containing poly(A)+ RNA from whole, normal mouse embryos at days 8.5, 10.5, 12.5 and 16.5 p.c., and several adult tissues. A single band of 2.7 kb was observed, whose absolute level steadily decreased during fetal development (Fig. 2), but which continued to be present in all somatic tissues of adult mice (Fig. 2 and data not shown).
A two-fold reduction in the level of Hβ58transcripts in mice heterozygous for the insertional mutation To confirm at the RNA level that the expression of the Hβ58 gene was blocked by the insertional mutation, we measured the abundance of the mRNA in several tissues of heterozygous Hβ58 mice compared to wild-type mice. We expected that the prevalence of Hβ58transcripts in tissues of heterozygous mutant mice would be approximately half the wild-type level, in the absence of compensation for the loss of one allele. Probe excess hybridizations (Lee et al., 1986) were carried out using an anti-sense RNA probe derived from a subclone of the cDNA pE30.1, and varying amounts of total RNA from adult kidney or testis. The data are plotted in Fig. 3 as RNase-resistant probe cts/min vs. input RNA, where the slope of each line is proportional to the abundance of the Hβ58 transcripts. For each tissue, the H β58 RNA was approximately two-fold less abundant in heterozygous than in wild-type mice (testis: 1.9; kidney: 2.1). This quantitative reduction in the level of Hβ58 transcripts implied that the structural alterations at the Hβ58 locus either blocked transcription or interfered with normal post-transcriptional processing, resulting in the failure of the mutant allele to produce significant levels of the mature mRNA.
Sequence of embryonic cDNA pE30.1 and its potential protein
Having concluded that cDNA clone pE30.1 represented a gene that had been disrupted by the Hβ58 insertional mutation, we determined the nucleotide sequence of the 2600 bp cDNA insert (Fig. 4). The sequence contained an open reading frame of 1073 nt (327 amino acids) ending with a COOH-terminal methionine, and followed by multiple stop codons in all reading frames. To verify that we had identified the correct open reading frame, we synthesized in vitro a sense-strand RNA, translated it in vitro and analyzed the products by SDS-PAGE. The size of the major polypeptide produced was consistent with the predicted size of 38 ×103Mr (data not shown).
Secondary structure and hydropathicity analysis of the prospective protein showed no identifiable structures such as signal peptides or transmembrane regions. Homology searches of the Genebank and NBRF data bases with the nucleotide and predicted amino acid sequences of the pE30.1 cDNA clone did not reveal any significant similarities with known genes. In addition, searches for specific conserved motifs characteristic of known classes of proteins (e.g. DNA-binding domains, hgand binding sites, etc.) were negative. We concluded that the protein potentially encoded by the embryonic Hβ58 mRNA was a novel protein, lacking recognizable structures that might provide insight into its biochemical properties or subcellular location.
Detection of Hβ58 Panscripts in ovarian oocytes, eggs and pre-implantation embryos
While Northern blot analysis (Fig. 2) established that the Hβ58 gene was active as early as 8.5 days p.c., these data did not indicate when the gene was first expressed, or determine its spatial pattern of expression in the embryo. To address these issues, we performed in situ hybridization studies, using an 35S-labelled probe representing the complete cDNA pE30.1 (Fig. 5A). Analysis of adult ovaries (Fig. 5B-C) revealed that the Hβ58 mRNA was present at high levels in oocytes of all stages, while the ovarian stroma and follicle cells contained only low, uniform levels of the mRNA. The RNA was still present at high levels in the mature oocyte (Fig. 5C), as well as in the ovulated egg (Fig. 5D). In the 3.5 day p.c. blastocyst (Fig. 5E-F), Hβ58 transcripts were readily detected in both the inner cell mass and trophectoderm. While we cannot rule out the possibility that Hβ58 transcripts in the blastocyst include maternal mRNA inherited from the egg, they most likely represent newly synthesized mRNA (see Discussion). If true, this would indicate that the Hβ58gene is already transcriptionally active during preimplantation development.
Spatial distribution of Hβ58 transcripts in post-implantation embryos
Figs 6 and 7 illustrate the accumulation of H β58 transcripts in the embryonic and extraembryonic lineages during post-implantation stages. At the egg cylinder stage (day 6.0 and day 6.5 p.c., Fig. 6A–D), the embryonic ectoderm contained a low but detectable level of H β58 RNA, and the extraembryonic ectoderm displayed a slightly higher level. The RNA level in the embryonic ectoderm remained low and increased only slightly during the gastrulation stage (day 7.0 and 7.5, Fig. 6E-H). However, progressively higher levels of H β58 mRNA were seen in several extraembryonic tissues, including the amniotic fold at day 7.0, the chorion and the ectopiacental cone at day 7.0 –7.5, and most dramatically in the visceral endoderm from day 6.5 onward.
At the early somite stage (day 8.5 p.c.), all three germ layers of the embryo proper displayed a relatively low and spatially uniform level of Hβ58 mRNA (Fig. 7A-B), as did most extraembryonic tissues including the allantois (Fig. 7A–B), amnion (Fig. 7C–D), and blood islands (Fig. 7C–D). However, the visceral yolk sac (previously the visceral endoderm) and the ectopiacental cone continued to accumulate relatively high levels of Hβ58 transcripts (Fig. 7A–D).
During organogenesis (day 9.5 p.c), distinct quantitative modulations of Hβ58 expression became visible in both embryonic and extraembryonic lineages (Fig. 7E-F). In addition to the visceral yolk sac, the region of the chorioallantoic placenta immediately adjacent to the proliferating allantois displayed a very high level of Hβ58 transcripts. The embryonic tissues displayed an increase in the average level of the mRNA, and quantitative regional differences could be seen for the first time.
At a later stage of fetal development (day 15.5 p.c., Fig. 7G–H), the Hβ58gene appeared to be expressed ubiquitously, although the mRNA level varied considerably between different tissues. The highest levels were seen in several organs derived from diverse embryonic cell lineages, including the telencephalon, cranial ganglia, tongue, liver, and gut.
The H β58 protein is highly conserved between mammals and birds
To assess the degree to which the Hp58 gene had been conserved in evolution, we used the munne cDNA as a probe in low-stringency Southern hybridization experiments with genomic DNAs of diverse species, representing four classes of vertebrata as well as invertebrate and single-cell species. As shown in Fig. 8, strong cross-hybridization was detected with DNA from all mammals examined (human, hamster, and rat), as well as the avian (chicken). However, the hybridization signal fell considerably in the amphibian, Xenopus laevis, and no reproducible signal was observed with any of the invertebrate or single-cell species. To further define the pattern of sequence conservation between mammals and birds, we cloned and sequenced a cDNA representing a chicken homolog of the murine H β58 gene. The predicted chicken amino acid sequence was remarkably similar to the murine sequence, with identity at 314 out of 326 residues, and conservative substitutions at most other residues (Fig. 9).
Mouse embryos homozygous for the Hβ58 transgene insertion display a characteristic set of abnormalities beginning at the primitive streak stage (day 7.5 p.c.), they undergo limited development to the early somite stage, and die shortly thereafter. We have proposed that these developmental defects are due to the disruption of an essential gene at the locus of the insertion on chromosome 10 (Radice et al., 1991). In this paper, we have described the identification of a gene at the Hβ58 locus, which is normally expressed during embryogenesis but has been physically and functionally disrupted by the transgene insertion. Sequence analysis of cDNA clones revealed that the wild-type Hβ58 gene encodes a previously unidentified 38 ×103Mr protein that is not obviously related to any known protem, but which has been strongly conserved in evolution among mammals and birds. In order to compare the normal expression pattern of the Hβ58 gene with the phenotypic defects seen in mutant embryos, we have characterized the spatial and temporal pattern of Hβ58 mRNA accumulation during normal mouse development. Although the earliest phenotypic effects of the mutation are seen in the embryonic ectoderm of the early post-implantation embryo, in situ hybridization analysis showed that the normal Hβ58 gene is expressed throughout pre-as well as post-implantation development, and that the highest levels of the mRNA are found not in the embryonic ectoderm but in the visceral endoderm and other extraembryonic lineages.
Several lines of evidence estabfish that the gene that we identified has been inactivated by the transgene insertion. First, the transcription unit encoding the Hβ58 mRNA spans the insertion site, with its 5’ untranslated region located upstream of the insertion, its first coding exon deleted from the mutant chromo-some, and its remaining exons located downstream from the insertion site (Fig. 1; Lee et al., manuscript in preparation). Thus, the mutant Hβ58gene is incapable of encoding the normal 2.7 kb mRNA and 38 ×103Mr protein product. To confirm the disruption of the H β;58 gene at the RNA level, we measured the mRNA abundance in tissues of heterozygous mutant mice (which are phenotypically normal) compared with those of wild-type animals, and found the expected two-fold reduction. In addition, we have observed that embryonic stem cell lines homozygous for the Hβ58mutation (Robertson et al., 1992) contain no detectable transcripts from the Hβ58gene, although the 2.7 kb mRNA is abundant in wild-type ES cells (unpublished data). We therefore conclude that the synthesis of stable mRNA from the Hβ58gene has been extinguished by the insertional mutation, and that this gene is likely to be responsible for the mutant phenotype.
The Hβ58gene was found to be expressed in normal embryos, at least at the RNA level, well before any abnormalities become apparent in homozygous mutant embryos, where the gene product is absent. The Hβ58mRNA detected in normal blastocysts and egg cylinder stage embryos is unlikely to represent maternal mRNA, most of which is destroyed at the two-cell stage (Piko and Clegg, 1982). While the possibility cannot yet be excluded that the normal development of homozygous mutant embryos for the first 6 – 7 days of embryogenesis is dependent on Hβ58mRNA or protein inherited from the oocyte, such maternal rescue seems unlikely due to relatively large size and cell number of the 6.5 day embryo (∼ 1000 cells; Snow, 1977) compared to the oocyte. Therefore, our results may imply that the Hβ58gene product is redundant with other gene products (zygotic or maternal in origin) at early stages, but unique and therefore essential after the egg cylinder stage.
As the earliest defect seen in the homozygous mutant embryo was a retardation in growth of the embryonic ectoderm beginning at day 7.5 p.c., it was surprising to find that the normal Hβ58gene was expressed at much lower levels in this cell lineage than in several extraembryonic lineages, from day 6.5 onward. It also appeared paradoxical that the visceral endoderm/visceral yolk sac lineage, which in normal 6.5 – 8.5 day embryos contained the highest levels of Hβ58mRNA, was unaffected morphologically by the mutation (Radice et al., 1991). However, given the established nutritive role, and possible inductive role, of the visceral endoderm (Beck et al., 1967; Azar and Eyal-Giladi, 1981; Mitrani et al., 1983), it is interesting to speculate that the normal expression of the Hβ58gene product in this extraembryonic tissue may be required for proper development of the embryonic ectoderm. This would be easiest to imagine if the Hβ58gene product were a secreted protein, although it lacks a recognizable signal sequence. Alternatively, the Hβ58protein may function at an earlier step in the synthesis, transport, or secretion of some other molecule, and thus indirectly affect the ability of the visceral endoderm to support embryonic growth.
While the lack of correspondence between the major embryonic sites of Hβ58gene expression and the earliest phenotypic effects of the mutation was surprising, there have recently been several reports of highly tissue-specific phenotypes resulting from mutations in genes that are rather widely expressed (e.g., c-src, Soriano et al., 1991; c-abl, Schwartzberg et al., 1989; Wnt-1, McMahon and Bradley, 1990). Nevertheless, it is important to consider the evidence that the Hβ58gene is a single copy gene, and that the hybridization probes that we have used are specific for transcripts from the Hβ58locus. First, Southern hybridization experiments using genomic probes (Radice et al., 1991) or cDNA probes (Fig. 8 and unpublished data) provided no indication that there are closely related sequences elsewhere in the genome. Similarly, genetic mapping studies using a cDNA fragment as a probe indicated a unique locus on chromosome 10 of Mus musculus (Justice et al., 1990). Second, Northern blot experiments detected only a single 2.7 kb mRNA in adult somatic tissues and in whole embryos (Fig. 2 and unpublished data), and this transcript was absent in homozygous mutant embryonic stem cells (unpublished data). Therefore, it seems highly unlikely that the pattern of expression defined by our in situ hybridization experiments includes transcripts from a different but related gene.
While the widespread expression of the Hβ58gene in many differentiated tissues of fetal and adult mice argues that the gene does not function solely in early embryogenesis, neither is the data consistent with a “housekeeping” role for the gene. Homozygous mutant embryos survive for 10 – 11 days p.c., developing a variety of organized tissues, and several terminally differentiated cell types (Radice et al., 1991). Further-more, embryonic stem cell lines derived from homozygous Hβ58embryos grow well in culture, and can contribute extensively to the tissues of chimeric fetuses (Robertson et al., 1992). Therefore, the Hβ58gene product is apparently not required for indefinite cell profiferation in vitro or extensive cell differentiation in vivo. We have also observed that the Hβ58gene is highly expressed in male germ cells, where it uses an alternative promoter and alternative splicing to produce a larger mRNA, potentially encoding a 42 × 103Mr protein (Lee et al., manuscript in preparation). Therefore, this alternate form of the Hβ58gene product may play a role in spermatogenesis.
Cross-hybridization studies revealed sequences closely related to the murine Hβ58gene in the DNA of other mammals as well as the chicken. The hybridization signal, however, decreased considerably using DNA from the amphibian X. laevis, and was not detectable in invertebrates. The analysis of a homologous chicken cDNA revealed a remarkable degree of conservation (>96% identity) between the murine and chicken amino acid sequences. In contrast, most proteins show only 60–80% similarity between mouse and chicken (R. Doolittle, personal communication). High levels of sequence conservation, however, are typically seen in structural proteins such as actins (Vandekerckhove and Weber, 1984), tubulins (Alexandrak1 and Ruderman, 1983) and histones (Isenberg, 1979). Whereas the extreme conservation of structural proteins extends throughout the animal kingdom, the Hβ58protein is likely to be far less conserved in the amphibian X. laevis based on the low extent of DNA cross-hybridization with the mouse cDNA. It is of interest that both avians and mammals are amniotes, whose embryos utilize extraembryonic membranes to interact with the external environment (Gauthier et al., 1988), while amphibians are non-amniotes, whose embryos interact directly with their environment. Given the observation that the highest levels of expression of the Hβ58gene were in the extraembryonic tissues of the 6.5 to 8.5 day p.c. conceptus, as well as the abnormalities of the amnion and chorion of several homozygous mutant 7.5 day embryos (Radice et al., 1991), we suggest that the strong evolutionary conservation of the gene among amniotes may reflect a role of the protem m the functions of extraembryonic membranes.
We thank E. Robertson and C Perez-Stable for their comments on the manuscript. This work was supported by a Postdoctoral Fellowship (to J J L ) and a Faculty Research Award (to F. C.) from the American Cancer Society, and by grants from the National Foundation March of Dimes and the NIH (to F. C.).