Mrj encodes a DnaJ-related co-chaperone that is essential for murine placental development

Implantation and formation of the placenta are critical for embryonic survival in eutherian mammals. Indeed much of early embryonic development is devoted to establishing extraembryonic cell types which make up the placenta (Copp, 1995; Cross et al., 1994; Rossant, 1995). A critical point in gestation occurs when simple diffusion of gases and nutrients from the mother is no longer sufficient to maintain embryo viability and a transformation in placental structure must occur (Copp, 1995; Cross et al., 1994). In the mouse, this occurs at mid-gestation with the formation of the labyrinth, a vascularized placenta. The labyrinth is a ‘chorioallantoic placenta’ in that it forms after attachment of the allantois to the chorionic plate (chorioallantoic fusion). Thereafter, extensive morphogenesis produces the three-dimensional labyrinthine structure which consists of narrow maternal blood sinuses lined by trophoblast cells. In this way trophoblast cells act as a barrier between the maternal and fetal blood compartments.

A number of genes that are essential for development and early morphogenesis of the chorioallantoic placenta have been identified (Copp, 1995; Cross et al., 1994; Rinkenberger et al., 1997). These include transcription factor genes that are essential for formation and/or maintenance of different trophoblast cell subtypes; Err2 of the chorion (Luo et al., 1997), Mash2 of the spongiotrophoblast (Guillemot et al., 1994; Tanaka et al., 1997) and Hand1 (formerly called Hxt/eHAND) of trophoblast giant cells (Riley et al., 1998). Err2 mouse mutants fail to form a chorioallantoic placenta because they lack chorionic trophoblast cells. Lack of chorioallantoic placentae can also be due to primary defects in the allantois. This is observed in conceptuses with mutations in genes such as brachyury (Glueksohn-Shoenheimer, 1944), DNA methyltransferase (Li et al., 1992), Lim1 (Shawlot and Behringer, 1995) and Csk1 (Thomas et al., 1995). The attachment of the allantois to the chorion depends on specific cell adhesion molecules. Vascular cell adhesion molecule-1 (VCAM1) is expressed on the distal tip of the allantois in anticipation of binding to its receptor, α4 integrin, which is expressed on the basal surface of the chorion (Gurtner et al., 1995; Kwee et al., 1995; Yang et al., 1995). Deficiencies in either VCAM1 or α4 integrin result in failure of chorioallantoic fusion in mice (Gurtner et al., 1995; Kwee et al., 1995; Yang et al., 1995). However, this phenotype occurs in only a portion of mutant conceptuses indicating that the VCAM1/α4 integrin interaction is not the only mechanism mediating chorioallantoic fusion. FGF signaling also plays a role in placental development since a hypomorphic mutation in the FGFR2 gene causes either defects in chorioallantoic fusion or labyrinthine morphogenesis (Xu et al., 1998).

Gene trapping in murine embryonic stem (ES) cells has been widely used to identify new developmentally important genes. We have made use of the ROSAβgeo retroviral vector (Friedrich and Soriano, 1991) which contains a promoterless βgeo gene, a fusion of β-galactosidase and neomycin resistance genes, flanked by a splice acceptor at the 5′ end and a polyadenylation signal at the 3′ end. If the βgeo cassette inserts into a transcriptionally active gene, the βgeo protein will be expressed, thus conferring neomycin resistance. In addition, the expression pattern of the trapped gene can be observed by staining specimens for β-galactosidase activity. In about 30% of the cases, the vector insertion disrupts gene function thus producing a mutant phenotype (Friedrich and Soriano, 1991). In our screen, expression patterns of ‘trapped’ genes were studied by in situ hybridization using probes from endogenous sequences that were cloned by 5′ RACE (Baker et al., 1997). One ES cell line (6AD1) was selected for further study and subsequent analysis revealed that this line carries the βgeo insertion in a novel gene, named Mrj, that we show here is essential for chorioallantoic fusion. Mrj is a member of a large gene family related to the DnaJ gene in E. coli. DnaJ-related proteins in other organisms function as adaptors and activators for HSP70-type chaperones (Hartl, 1996). The specific nature of the Mrj mutant phenotype, despite the fact that several other DnaJ-related genes are expressed in the placenta, suggests that these proteins do not have redundant functions.

The 6AD1 cell line was identified in a previously described gene trap screen (Baker et al., 1997) using the ROSAβgeo retrovirus (Friedrich and Soriano, 1991) to infect R1 ES cells. The allele created by the proviral integration was called 6AD1βgeo. ES cells were grown with or without STO feeder cells (ATCC) in medium consisting of 15% fetal bovine serum (HyClone), 0.1 mM β-mercaptoethanol, 2 mM L-glutamine, 0.1 mM MEM nonessential amino acids, 150 μg/ml G418 and 1000 U/ml leukemia inhibitory factor in DMEM. mRNA was harvested from the 6AD1 cell line and 5′ rapid amplification of cDNA ends (RACE) was performed as previously described (Baker et al., 1997) in order to clone DNA adjacent to the βgeo insertion. A 90 bp 5′ RACE product, which represented the 5′ end of the Mrj cDNA, was cloned and used to screen a λgt10 cDNA library made from E8.5 embryos (kindly provided by Dr Brigid Hogan). A cDNA of approximately 1.6 kb was recovered and was cloned as two EcoRI fragments of approximately 400 (pC400) and 1200 bp (pC1200) encompassing the entire mRNA. cDNA fragments were ligated into pBluescript (KS−) (Stratagene) and sequenced (GenBank accession no. AF035962).

To clone genomic DNA 5′ to the βgeo insertion site (from intron one), inverse PCR was performed as previously described (Jonsson et al., 1996) using primers, oriented in divergent directions, which anneal to sequences in the 3′ LTR (5′-TGGGAGGGTCTCCTCTGAGT-3′) and β-galactosidase (5′-CACATGGCTGAATATCGACGGTT-3′) regions of the βgeo insertion. To prepare the template DNA, genomic DNA isolated from 6AD1 cells was digested with EcoRI, diluted, ligated to form circular DNAs and then linearized with EcoRV. A single band of approximately 700 bp was produced after PCR amplification. It was ligated into pBluescript to produce the plasmid pE5G and sequenced. Southern blots made from 6AD1 cell genomic DNA confirmed linkage between the cloned intronic DNA and βgeo.

Aggregation chimeras were generated with 6AD1 ES cells using wild-type CD-1 morulae as previously described (Nagy et al., 1993). Two founder male chimeras were backcrossed to wild-type 129Sv and outcrossed to CD-1 females to produce progeny which were heterozygous for the 6AD1βgeo allele. Heterozygous mice were intercrossed to produce homozygotes.

Southern blot analysis of genomic DNA isolated from tail samples, yolk sacs or embryos (Riley et al., 1998) and northern blot analysis of tissue total RNA (Cross et al., 1995) was performed as previously described. The EcoRI fragment of pE5G (genomic sequence from intron one) was used as a probe for genotyping specimens by Southern blot analysis because it detects a polymorphism in the Mrj locus caused by the insertion of βgeo (see Fig. 5). Since the Mrj coding region is similar to several other DnaJ-related genes in mice, we generated a 3′ untranslated region probe that was Mrj-specific. The pC1200 plasmid containing part of the Mrj cDNA was digested with Eco01091 and re-closed to produce the plasmid pE3 which contained only the distal 3′ untranslated region. This fragment was use to generate probes for northern blot and in situ hybridization experiments. The XhoI/EcoRV fragment of pSAβgeo (Friedrich and Soriano, 1991) was used as a probe to detect βgeo sequences. The mouse GAPDH cDNA (Piechaczyk et al., 1984) was used as a probe to show loaded amounts of RNA (Fig. 2).

Conceptuses were dissected at various gestational ages: noon of the day that a vaginal plug was detected was defined as embryonic day (E) 0.5. For routine histology, conceptuses were fixed in 4% paraformaldehyde and paraffin embedded. For X-gal staining, specimens were fixed for 15 to 30 minutes in 1% formaldehyde, 0.2% glutaraldehyde, 0.02% NP-40, 5 mM EGTA, 2 mM MgCl₂, 0.1 M sodium phosphate (pH 7.3). Specimens were stained whole, or as cryosections, for 4-24 hours at 37°C in 0.1% 4-chloro-5-bromo-3-idolyl-β-D-galactopyranoside (X-gal; Nova Biochem), 5 mM K₃Fe(CN)₆, 5 mM K₄Fe(CN)₆ in buffer (0.02% NP-40, 0.01% deoxycholate, 2 mM MgCl₂, 0.1 M sodium phosphate (pH 7.3). Some X-gal-stained conceptuses were paraffin embedded and cut into 8 μm histological sections. For cryosections, fixed conceptuses were equilibrated in 15% sucrose followed by 30% sucrose in PBS for 12 hours each at 4°C. Conceptuses were then embedded in OCT medium (Miles) and stored at −80°C prior to cutting into 10 μm sections with a cryostat. Following X-gal staining, sections were counterstained with eosin (Sigma).

In situ hybridization data presented in Fig. 1 were prepared as described by Baker et al. (1997). Otherwise, conceptuses were fixed in 4% paraformaldehyde, 0.02% glutaraldehyde in PBS and paraffin embedded. Serial histological sections (5 μm) were either stained with Harris’ haematoxylin and eosin (Sigma) or subjected to in situ hybridization (Millen and Hui, 1996). Antisense ³³P-labeled riboprobes were prepared using an RNA transcription kit (Stratagene). Probes specific for Gcm1 (Altshuller et al., 1996), Err2 (Pettersson et al., 1996), 4311 (Lescisin et al., 1988), and Pl1 (Colosi et al., 1987) have been described previously. A Mrj-specific riboprobe was prepared from pE3 which was linearized with Asp718. A βgeo riboprobe was prepared from pSAβgeo which was linearized with PstI. After development, the sections were counterstained with Carazzi’s haematoxylin.

Conceptuses were fixed in 2% paraformaldehyde for 2 hours at 4°C, equilibrated in 8% sucrose followed by 18% sucrose in PBS for 4 and 12 hours, respectively, at 4°C and finally embedded in OCT medium (Miles) and stored at −80°C. Cryosections were air-dried and post-fixed in acetone at −20°C for 5 minutes. They were then subjected to immunoperoxidase staining for VCAM1 using the MK-2 monoclonal antibody (Gurtner et al., 1995) (generously provided by Dr Myron Cybulsky), α4 integrin using the PS-2 monoclonal antibody (Yang et al., 1995) (Chemicon), and E-cadherin using the DECMA-1 monoclonal antibody (Sigma). Horseradish peroxidase-conjugated secondary antibodies (Amersham) were used at a 1:50 dilution.

A gene trap screen was previously performed by infecting R1 ES cells with the ROSAβgeo retrovirus vector (Baker et al., 1997). The 6AD1 ES cell line, from which the Mrj gene was identified, was selected for further study. The expression of Mrj during mouse development was studied by following β-galactosidase expression in conceptuses carrying the 6AD1βgeo allele. Mice carrying this allele were produced by first generating chimeric males from the 6AD1 ES cells. The chimeras were bred to wild-type females in order to produce progeny that were heterozygous for the 6AD1βgeo allele (+/6AD1βgeo). Conceptuses were dissected at embryonic days (E) 7.5 to 15.5 and stained with X-gal to detect β-galactosidase activity. Positive (blue) staining was observed in approximately half of the embryos. In this strain, embryos that genotyped as +/+ failed to show any blue staining except for occasional staining in visceral endoderm after prolonged incubation (Fig. 3E; data not shown).

Among the β-galactosidase-positive embryos, activity was detected in the embryo proper at all stages (Fig. 1A-C). Positive staining was observed in the egg cylinder at E7.5, and was fairly widespread in the embryo at subsequent stages. In situ hybridization experiments using conceptuses aged between E8.5 and 17.5 revealed that the β-galactosidase activity in the embryo essentially replicated the pattern of the wild-type Mrj gene (Fig. 1, compare D with E, G with H). Although Mrj appeared to be widely expressed at low levels throughout most of the embryo, it was expressed at notably higher levels in a few tissues. Specifically, Mrj expression was detected in the ganglion neural layer of the developing retina (Fig. 1C,F). Beginning at E12.5, Mrj expression in the brain was consistently higher in the trigeminal ganglia, diencephalon and midbrain (Fig. 1G,H). Other prominently expressing tissues included the dorsal root ganglia (Fig. 1D,E), thymus (Fig. 1D,E), nasal epithelium (Fig. 1G,H) and testis (not shown). Expression of Mrj and βgeo in adult organs of +/6AD1βgeo mice was assessed by northern blot analysis. Mrj mRNA was readily detected in the testis, uterus, liver and brain with somewhat weaker expression in the eye, heart and gut. The mRNA was not detected in muscle or kidney (Fig. 2). The βgeo transcript showed a similar tissue distribution except that expression was not detected in the eye and heart.

At all embryonic stages examined, the highest β-galactosidase activity in the conceptus was observed in trophoblast giant cells of the placenta (Fig. 3). Secondary giant cells (which form around the ectoplacental cone) showed weaker staining before E9.5 compared to primary giant cells. β-galactosidase activity was also evident in trophoblast cells of the chorion but not the ectoplacental cone at E8.5, and in the labyrinth but not spongiotrophoblast at E10.5 (Fig. 3A,C,F). In situ hybridization was performed on histological sections from placentae of similar stages. Nine sections for each tissue and probe combination were examined and typical results shown (Fig. 3G,H). Mrj and βgeo mRNAs were detectable in trophoblast cells of the chorion and ectoplacental cone, and in giant cells. However, a subset of giant cells expressed these transcripts at strikingly higher levels. The latter result differed from the β-galactosidase staining which was uniformly high in all giant cells. Another difference between enzyme activity and transcript levels was apparent in the ectoplacental cone and spongiotrophoblast layer. Both Mrj and βgeo mRNAs were detected by in situ hybridization whereas β-galactosidase activity was never observed (Fig. 3A-C). β-galactosidase was extremely weak or undetectable in the allantois and the mesothelial cell layer of the chorion (Fig. 3C), tissues which provide mesodermally derived components of the chorioallantoic placenta. Likewise, Mrj and βgeo mRNAs were undetectable by in situ hybridization (Fig. 3G,H; data not shown). After chorioallantoic fusion and subsequent formation of the labyrinth, β-galactosidase activity was detected in the trophoblast component of the labyrinth (Fig. 3F), a pattern which resembled the Mrj mRNA expression (data not shown).

The Mrj cDNA was cloned from an E8.5 mouse embryo cDNA library using a 5′ RACE product as the initial probe. A 1.6 kb cDNA was recovered which was similar to the predicted size of the full length mRNA based on northern blot analysis of mouse placental mRNA (Fig. 5D). The cDNA sequence predicted an open reading frame encoding a 242-amino acid protein (Fig. 4A). Several cDNAs were identified in the NCBI database of expressed sequence tags (ESTs) that together represent the complete human Mrj cDNA. The open reading frames of the mouse and human cDNAs were 96% and 90% identical in nucleotide and amino acid sequence, respectively (Fig. 4A).

Although the MRJ protein was unique when compared to sequences in GenBank, the N-terminal 74 amino acids were similar to the J domain present in the E. coli DnaJ protein, as well as in several proteins in yeast, Drosophila, C. elegans and mammals. DnaJ-related proteins interact with HSP70 chaperones via the J domain and stimulate their ATPase activity (Hartl, 1996). In searching GenBank and EST databases, we found more than 40 unique cDNA sequences which encode J domain proteins in both humans and mice. Each of these sequences was predicted to encode the His-Pro-Asp tripeptide within the J domain, which is essential for interaction with HSP70, as well as flanking α-helices which are conserved among family members. Twenty cDNAs encoded J domains which shared greater than 50% amino acid sequence identity with MRJ (Fig. 4B shows a dendrogram). Within this group were five previously identified proteins called MSJ1 (Berruti et al., 1998), HSJ1 (Cheetham et al., 1992), HSP40 (HDJ1) (Ohtsuka, 1993; Raabe and Manley, 1991), HDJ2 (Chellaiah et al., 1993) and MTJ1 (Brightman et al., 1995). C-terminal to the J domain, the MRJ protein had three other regions of sequence similarity to three of these family members (Fig. 4A, regions II-IV). Region II is a Gly- or Gly/Phe-rich sequence which is also present in E. coli DnaJ. The significance of regions III and IV which are conserved in MRJ, MSJ1, HSJ1 and HSP40 is unknown.

Southern blot analysis of genomic DNA extracted from 6AD1 cells indicated that a complete copy of ROSAβgeo cassette, including full length LTR sequences, had inserted into the Mrj locus. Since the cDNA cloned by 5′ RACE represented the first 90 bases of 5′ untranslated region in the full length mRNA, this region was assumed to be exon 1. Therefore, βgeo had inserted either into exon one or downstream within an intron. To distinguish these possibilities, Southern blot analysis was used to generate a restriction map around exon 1 and the βgeo insertion (data summarized in Fig. 5A). There was no overlap between the restriction maps around exon 1 and the βgeo insertion indicating that the two were separated by some distance (>12 kb). Because of this distance, we were unable to detect any restriction enzyme polymorphisms on Southern blots caused by the insertion of βgeo into the Mrj locus when using exon 1 as the probe (Fig. 5B). To determine if the insertion had disrupted exon 2 or sequence further 3′, we probed Southern blots using distal 3′ cDNA probes (plasmids pC400 and pC1200). However, we were unable to detect restriction enzyme polymorphisms (data not shown). We concluded, therefore, that Mrj exon sequences were not disrupted by the insertion and that βgeo had inserted into intron one. In order to detect polymorphisms associated with the 6AD1βgeo allele that were required to genotype mice by Southern blotting, we cloned a fragment of genomic DNA flanking the 5′ end of βgeo by using inverse PCR. The sequence of this fragment was unique and, when used as a probe, revealed restriction site polymorphisms between DNA from wild-type and +/6AD1βgeo mice (Fig. 5C).

The mapping data indicated that the βgeo insertion had not disrupted the Mrj coding sequence. To determine if the βgeo insertion had disrupted the function of the Mrj gene, we examined mice carrying the mutant gene for an abnormal phenotype. Heterozygous mice appeared normal and transmitted the 6AD1βgeo allele at roughly the predicted Mendelian frequency of 50% (Table 1), but in intercrosses of heterozygous animals, no homozygotes were detected among the progeny at birth (Table 1). Progeny from heterozygous matings were then dissected at E8.5 to E14.5. Conceptuses that were homozygous for the 6AD1βgeo allele were viable only up to about E11.5 (Table 1). The matings summarized in Table 1 represent mice produced by outcrossing the founder chimeras to an outbred background. However, the same phenotype was observed on a 129Sv inbred background.

To investigate the embryonic lethal phenotype of conceptuses that were homozygous for the 6AD1βgeo allele, we determined if Mrj mRNA expression was reduced. Northern blots of E10.5 placental RNA from conceptuses of wild-type (+/+), heterozygous (+/6AD1βgeo) and homozygous (6AD1βgeo/6AD1βgeo) mutant genotypes were probed with a fragment of Mrj gene which lies downstream of the βgeo insertion (3′ UTR fragment, see Materials and Methods). We were unable to detect any Mrj mRNA in homozygous mutant placentae (Fig. 5D). Furthermore, Mrj transcript levels appeared to be reduced (by about one-half) in samples from heterozygous conceptuses. To confirm the northern blot results, mRNA in situ hybridization analysis was performed on E8.5 conceptuses using a riboprobe generated from the same 3′ fragment. Fourteen conceptuses produced from a heterozygous mating were analyzed, in which nine histological sections from each conceptus were assessed. We failed to observe Mrj hybridization signals above background in trophoblast cells of the three homozygous mutant embryos (Fig. 5E). Based on these data, the 6AD1 mutation appeared to be a null allele of Mrj.

At E8.25, all embryos produced from crosses between heterozygous mice had reached the three somite stage and were indistinguishable (Fig. 6A). At E8.5, the homozygous mutant embryos appeared to be of normal size and had developed 6-7 somites, similar to their wild-type and heterozygous littermates. However, chorioallantoic fusion was never observed in the homozygous mutants in contrast to the wild-type and heterozygous littermates (Table 2; Fig. 6). For other mouse strains, chorioallantoic fusion occurs during a narrow window between E8.25 and 8.5 when embryos are at the 5-6 somite stage (Downs and Gardner, 1995). Dissection of embryos between E8.25 and 8.5 revealed that this is also true of the 129/CD-1 background as well (data not shown). It is significant therefore that while development of the embryo proper occurred with normal timing in homozygous mutants up to E8.5, the process of chorioallantoic fusion did not. For embryos dissected at E9.5, although they had turned, the homozygous mutant embryos were smaller than their littermates and had arrested at the 18 somite stage (E9.25). The allantois remained unattached to the chorion in the vast majority of homozygous mutant conceptuses at E9.5 and later (Fig. 6; Table 2). However, at E9.5, 10.5 and 11.5, we found a few conceptuses in which the allantois had formed a loose attachment to the chorion (Table 2). Notably, placental labyrinth morphogenesis never proceeded in these cases and all homozygous mutants were undergoing resorption by E12.5.

We looked for abnormalities in placental histology in homozygous Mrj mutants. At E8.25-8.5, no differences in size or histological appearance of the allantois were apparent in mutants except that it remained unattached. The mesothelium lining the basal surface of the chorion (Fig. 6B) and all trophoblast cell subtypes appeared morphologically normal in the homozygous mutant conceptuses (Fig. 6B). As development proceeded, the chorionic plate remained intact in the placentae of Mrj mutants (Fig. 6B) although, starting at E9.5, vacuolated cells and pyknotic nuclei were observed at high magnification (data not shown). Marker analysis was performed to detect changes in gene expression of trophoblast cell types, using serial histological sections. Probes were hybridized to nine sections of 3-4 mutants at each developmental stage and typical results are shown in Fig. 7. Expression patterns for most markers appeared to be normal. For example, Pl1, a trophoblast giant cell-specific gene (Colosi et al., 1987), and 4311, an ectoplacental cone and spongiotrophoblast-specific gene (Lescisin et al., 1988), were both expressed normally in Mrj mutant placentae (Fig. 7). E-cadherin is expressed by basal cells in the chorionic plate prior to allantoic fusion (Reuss et al., 1996), a pattern which was unaltered in Mrj mutants (data not shown). However, in situ hybridization analysis showed that signals for Gcm1 (Altshuller et al., 1996), a gene whose expression is restricted to the chorion at E8.5 and the trophoblast component of the labyrinth at later stages (J. C. C., unpublished data), were dramatically reduced and difficult to discern above background at E8.5 coincident, therefore, with the first observed defects in Mrj mutants. Hybridization signals for another chorionic trophoblast marker, Err2 (Pettersson et al., 1996), were also reduced in mutants compared to wild-type and heterozygous conceptuses (data not shown).

To investigate the molecular basis of the chorioallantoic fusion defect, we looked for abnormalities in expression of cell adhesion molecules which are known to be involved. Around the time of chorioallantoic fusion (E8.25-8.5), α4 integrin is normally expressed on the surface of the chorion (Yang et al., 1995) and VCAM1 is expressed on the distal two thirds of the allantois (Gurtner et al., 1995; Kwee et al., 1995). The expression of these proteins was assessed using immunohistochemistry on serial histological sections (10 sections for each antibody) of three homozygous mutant conceptuses that had been dissected at E8.5 (5-6 somite stage). We found that both VCAM1 and α4 integrin were expressed in the mutants similar to wild-type and heterozygous conceptuses (Fig. 8; data not shown). The timing of receptivity for chorioallantoic fusion is thought to be tightly regulated (Downs, 1998). It was notable, therefore, that we saw persistent expression of α4 integrin and VCAM1 to at least E9.5 in Mrj mutants in which chorioallantoic fusion had not occurred (Fig. 8).

Chaperone and co-chaperone proteins have been highly conserved throughout evolution and are implicated in a number of cellular processes. We have described here the identification of Mrj, a gene encoding a new member of the DnaJ-related co-chaperone family, and its expression pattern during development. We found that Mrj is expressed in the placenta, several regions of the embryo and in some tissues into adulthood. Mrj is essential for formation of the chorioallantoic placenta, implying a previously unsuspected requirement for chaperone activity in this developmental process.

Mrj expression occurs broadly in several organs during development and into postnatal life. We studied its expression during placental development in detail because the phenotype of Mrj-deficient conceptuses indicated an essential function in development of the chorioallantoic placenta. We readily detected Mrj expression in trophoblast cells of the chorion but not in the chorionic mesothelium or the allantois. The trophoblast lineage arises first as the trophectoderm at the blastocyst stage (E3.5 in mice) (Cross et al., 1994; Rossant, 1995). By the early postimplantation period (E6.5-7.5), three anatomically and functionally distinct trophoblast cell types are apparent. Chorionic trophoblasts (also called extraembryonic ectoderm) lie next to the embryo; ectoplacental cone trophoblasts sit as a cap of tissue between the chorion and the outer layer of trophoblast giant cells. Chorionic trophoblast cells, in addition to contributing to the labyrinth after contact with the allantois, are thought to be the proliferating trophoblast stem cells (Rossant, 1995; Rossant and Ofer, 1977). In culture, chorionic trophoblast cells differentiate first into ectoplacental cone-type and subsequently to trophoblast giant cells (Carney et al., 1993), suggesting that these three cell types represent stages in a differentiation pathway. Mrj mRNA is, therefore, expressed throughout the trophoblast lineage since we detected it in chorion, ectoplacental cone and giant cells. Nonetheless, we have observed a mutant phenotype associated with only the chorion of Mrj-deficient conceptuses. Expression studies revealed some potentially interesting features of Mrj regulation. First, there were differences between Mrj and βgeo transcript levels in some tissues. For example, Mrj transcripts were detectable in the heart and eye, albeit at low levels, but βgeo transcripts were not. It is possible that the βgeo insertion disrupted intronic sequences which regulate tissue-specific transcription or splicing. However, in all other tissues we observed a good correlation between Mrj and βgeo mRNA expression. Two other interesting features of expression were apparent in the trophoblast lineage. In the trophoblast giant cell population, while expression was detectable in all cells by mRNA in situ hybridization, a much higher expression level was observed in a subset of cells. The same pattern was observed when using the βgeo probe. These strongly expressing cells were randomly distributed around the conceptus in a pattern unlike any other gene expression pattern in giant cells that is known to us. Notably, the variable expression level was not apparent from the β-galactosidase staining, which was uniformly strong in every giant cell. An explanation for this difference is that the βgeo protein is stable and, therefore, persists in the cell even though Mrj mRNA expression may be variable. Another difference between Mrj and βgeo mRNA expression and β-galactosidase enzymatic activity was apparent in the spongiotrophoblast layer and its precursor, the ectoplacental cone. β-galactosidase activity was never observed in these trophoblast cells despite the presence of Mrj transcripts. Importantly, we detected βgeo transcripts in these cells indicating that splicing to the βgeo cassette occurred properly. It is unlikely that protein instability accounts for the absence of β-galactosidase enzymatic activity, since it can be detected in the ectoplacental cone and spongiotrophoblast of ROSA26 conceptuses (Tanaka et al., 1997). It is possible, though, that a βgeo transcript with the Mrj 5′ untranslated region is not efficiently translated in ectoplacental cone or spongiotrophoblast. It will be important to study MRJ protein expression in order to clarify these issues.

Although the βgeo insertion in the 6AD1 cell line did not disrupt coding exons, it apparently created a null allele of the Mrj gene because we were unable to detect Mrj mRNA in conceptuses that were homozygous for the 6AD1βgeo allele. This probably resulted from failure to splice around the βgeo cassette and truncation of the transcript by the polyadenylation signal at the 3′ end of the βgeo sequence. We also noted that there was a reduction in Mrj mRNA levels in heterozygotes indicating that there was no compensation for loss of one allele. Despite this, heterozygotes had no obvious phenotypic defects though have not been examined in detail.

It seems likely that the phenotype of Mrj-deficient conceptuses is due to the defect in placentation. The formation of a chorioallantoic placenta is a critical ‘checkpoint’ that must be achieved by mid-gestation (Copp, 1995; Cross et al., 1994). We observed no obvious developmental defects in the embryo proper of Mrj mutants up to E9.25 as they were able to develop to the 18 somite stage on schedule. Arrest of the mutant embryos at the 18 somite stage (approximately E9.25) is also observed in mouse embryos which are deficient for VCAM1 (Gurtner et al., 1995; Kwee et al., 1995) and α? integrin (Yang et al., 1995). For several reasons we favour the hypothesis that the failure of chorioallantoic fusion is specifically due to defects in chorionic trophoblast function. First, Mrj expression appeared to be limited to the chorionic trophoblast cells and was not appreciably detected in the allantois or the chorionic mesothelium. Second, at E8.25-8.5 (the period when chorioallantoic fusion normally occurs) the allantois of Mrj mutants appeared to be of normal size, showed no histological abnormalities and expressed the cell adhesion molecule VCAM1. Third, we detected changes in two chorionic trophoblast-specific transcription factor genes; downregulation of Err2 and an apparent absence of Gcm1 expression. Err2-deficient mouse mutants lack chorionic structures, thus implicating Err2 in chorion cell proliferation (Luo et al., 1997) but the function of Gcm1 is unknown. Whether the reduction of Err2 and Gcm1 expression in Mrj mutants is a primary cause of the phenotype or is secondary to other events is not clear. The fact that the chorion formed at the normal time and persisted in Mrj mutants implies that the failure probably resulted from a lack of receptivity of the chorion. An alternative hypothesis is that chorionic trophoblast cells produce something which affects either the mesothelium or the allantois. The precise receptivity mechanism which is affected in Mrj mutants is unresolved since the expression of α? and VCAM1 was normal.

Chorioallantoic fusion was never observed in Mrj homozygous mutants dissected at E8.5, in contrast to nearly all of the wild-type and heterozygous littermates. Examination of conceptuses between E9.5 and 11.5 revealed that a loose chorioallantoic attachment had occurred in a few homozygous mutants (8/50) (Table 2). Although the numbers of embryos are too few to make a firm conclusion, there was a trend for the frequency to increase with gestational age. Therefore, the most conservative interpretation of these data is that Mrj- deficiency results in a significant delay in chorioallantoic fusion. It is clear though that even if chorioallantoic fusion does occur by E9.5 or 10.5 in Mrj mutants, a functional chorioallantoic placenta does not form and the mutants die at a similar time in development. FGFR2, VCAM1 and α? integrin are the only other molecules which have been implicated, to date, in the process of chorioallantoic fusion (Gurtner et al., 1995; Kwee et al., 1995; Xu et al., 1998; Yang et al., 1995). Notably, mouse mutants for all of these factors show only variably penetrant defects in chorioallantoic fusion (Gurtner et al., 1995; Kwee et al., 1995; Xu et al., 1998; Yang et al., 1995). Chorioallantoic fusion was observed in two-thirds of FGFR2 mutants (Xu et al., 1998), one-half of α4 integrin mutants (Yang et al., 1995) and 9-50% of VCAM1 mutants (depending on genetic background) (Gurtner et al., 1995; Kwee et al., 1995). Given the incomplete penetrance, these studies suggest that there may be independent pathways for chorioallantoic attachment. It would be interesting to determine if inactivating both the FGF and α4 integrin/VCAM1 pathways produces a more penetrant phenotype.

The only recognizable motif in the MRJ sequence was the J domain at the N terminus of the protein. Based on the conserved function of J domains, it is likely, though not yet proved, that MRJ functions as a co-chaperone for an HSP70. The J domain of DnaJ-like proteins binds to HSP70s and thereby stimulates their ATPase activity (Burston and Clarke, 1995; Caplan et al., 1993). ATP hydrolysis allows conformational changes in the HSP70 necessary for the binding and release of unfolded proteins (Hartl, 1996). Distal to the J domain of MRJ, there are 6 Gly residues which are conserved with E. coli DnaJ; in other DnaJ-related proteins, a Gly/Phe-rich region occurs at the same position. This region may form a flexible linker between the J domain and the rest of the protein (Pellecchia et al., 1996; Qian et al., 1996). The remainder of MRJ protein differs from E. coli DnaJ but shares regions of similarity with three mammalian DnaJ-like proteins, MSJ1, HSJ1 and HSP40. Chaperone function has been implicated in a number of cellular processes including protein folding and re-folding after cell stress (e.g., the heat shock response), intracellular protein trafficking and protein-protein interactions (Hartl, 1996).

In contrast to prokaryotes, eukaryotic genomes appear to encode multiple Hsp70- and DnaJ-related proteins. This raises the question as to whether the multiple members have unique or overlapping functions. In the budding yeast S. cerevisiae, there are eight DnaJ-like genes and mutations in each produce distinct phenotypes (Cyr et al., 1994), indicating that they have non-overlapping functions. The J domain protein specificity is thought to reflect a restricted interaction with different HSP70s, of which there are 14 in yeast (James et al., 1997), as well as different substrate binding abilities. In mammals, 11 Hsp70-related genes have been identified thus far (Tavaria et al., 1996) compared to over 40 different J domain proteins (this study). There are only a few examples in higher eukaryotes of chaperones whose essential functions have been identified by loss-of-function gene mutations. In Drosophila, the lethal(2) tumorous imaginal discs (l(2)tid) gene encodes a J domain protein which is involved in imaginal disc cell differentiation (Kurzik-Dumke et al., 1995). In mice, Hsp70-2 deficiency results in arrest of spermatogenesis due to failure of cdc2 to associate with cyclin B1 despite the fact that other Hsp70s are expressed in testis (Zhu et al., 1997). It is clear from these examples that individual chaperones and co-chaperones can have very specific functions during development, similar to our observations with Mrj mutants. A role for chaperone activity during development of the placenta was previously unsuspected. In addition to our insights from Mrj mutants though, it has recently been discovered that the chaperone Hsp90β is also required for placental development in mice (A.KVoss, T. Thomas and P. Gruss, personal communication). Importantly by searching UniGene and EST databases, we have found that several DnaJ-related genes other than Mrj are expressed in the developing placenta: ESTs representing several DnaJ-related genes have been identified in human placental libraries (7 different genes) as well as murine placental (1 gene) and ectoplacental cone libraries (2 genes) (J.C. C., unpublished data). The occurrence of the placentation defect in Mrj-deficient conceptuses, despite the fact that other DnaJ-related proteins are expressed in the placenta, supports the idea that individual members of this gene family may have non-overlapping cellular and molecular functions.

We thank M. Cybulsky, B. Hogan, D. Linzer, J. Rossant, P. Soriano for reagents, Z. Chen and R. Han for technical advice, A. Voss and P. Gruss for sharing information prior to publication, C. C. Hui, J. Rossant and members of the Cross lab for discussions, and J. Copeman and P. Riley for comments on the manuscript. The work was supported by a grants from the MRC of Canada (to J. C. C.) and the NIH (HD29471 to G. E. L. and GMO 7507-18 to M. A. H.). P. H. was supported by a studentship from the Genesis Foundation. G. E.Lis an Established Investigator of the American Heart Association and J. C. C. is an MRC Scholar.