The Mrj co-chaperone mediates keratin turnover and prevents the formation of toxic inclusion bodies in trophoblast cells of the placenta

Protein aggregation and the formation of intracellular protein inclusion bodies are associated with many degenerative disorders, including Parkinson's and Huntington's diseases, and with cirrhosis of the liver(Carrell, 2005). In many cases,the role of protein aggregates in disease pathogenesis is unclear. Efforts to understand aggregation disorders have focused mainly on mutations within the disease-associated protein. However, mounting evidence suggests that a major cause of protein inclusion body formation is failure of the protein quality-control system (McClellan et al.,2005). This surveillance mechanism recruits both molecular chaperones and proteases to safeguard against protein aggregation by maintaining the equilibrium between protein folding and degradation(Hohfeld et al., 2001).

A malfunction in the ubiquitin-proteasome degradation pathway is hypothesized to be one cause of Mallory body formation in the hepatocytes of patients with chronic liver disorders(Bardag-Gorce et al., 2003; Bardag-Gorce et al., 2004; Denk et al., 2000). Mallory bodies are inclusions composed of the keratin intermediate filaments keratin 8(K8; also known as Krt8) and keratin 18 (K18; also known as Krt18), as well as ubiquitin, the proteasome complex and molecular chaperones, including heat shock protein 70 (HSP70; HSPA1B) and heat shock protein 90 (HSP90)(Coulombe and Omary, 2002). Cells that have been treated with a chemical inhibitor of proteasome function(Bardag-Gorce et al., 2004) or that contain the UBB⁺¹ ubiquitin mutation(Bardag-Gorce et al., 2003)form Mallory bodies due to an accumulation of non-degraded keratin.

Misfolded proteins also tend to aggregate. Proper protein folding within a cell often does not occur spontaneously and, thus, molecular chaperones are required for some proteins to reach its native state efficiently(Hartl and Hayer-Hartl, 2002). Hsp70 is a ubiquitous chaperone that, together with diverse co-chaperone binding partners, facilitates protein folding and the presentation of proteins to the proteasome for degradation (Esser et al., 2004). Co-chaperones of the DnaJ (Hsp40) protein family are characterized by a conserved Jdomain that regulates substrate binding and release by activating the ATPase activity of Hsp70(Fan et al., 2003). These co-chaperones differ in their tissue and subcellular patterns of expression,and convey substrate specificity to Hsp70(Fan et al., 2003). However,little is known about the precise substrates that are regulated in this manner.

The Mrj (Mammalian relative of DnaJ; also known as Dnajb6- Mouse Genome Informatics) gene encodes a co-chaperone that is widely expressed throughout the adult mouse and during development of the embryo and placenta (Chuang et al., 2002; Dai et al., 2005; Hunter et al., 1999; Izawa et al., 2000; Seki et al., 1999). Mrj^-/- embryos die at mid-gestation due to a failure of chorioallantoic attachment during placental development(Hunter et al., 1999). Recently, a few diverse Mrj-interacting proteins have been identified,including Huntington disease (HD) protein(Chuang et al., 2002) and K18(Izawa et al., 2000). Here, we show that Mrj deficiency prevents normal keratin turnover, resulting in the formation of large keratin aggregates that are toxic to trophoblast cells, inhibiting their normal function.

Mrj^+/- mice were originally generated by 6AD1β -geo gene-trap insertion as previously described(Hunter et al., 1999). Mrj^-/- embryos were produced by mating Mrj^+/- mice. K18/K19 double-null embryos were generated by mating K18^-/- mice(Magin et al., 1998) with K19^-/- mice (Hesse et al., 2000) (provided by T. Magin, University of Bonn, Bonn,Germany) and then mating the resulting double heterozygous progeny. Mrj^-/-;K18^-/- and Mrj^-/-;K18^+/- embryos were generated by mating Mrj^+/- mice with K18^-/- mice and then mating their resulting Mrj^+/-;K18^+/- progeny together. Enhanced green fluorescent protein (Egfp) transgenic mice(Hadjantonakis et al., 1998)were maintained at heterozygosity in a CD-1 background. Experiments were performed in accordance with the Canadian Counsel on Animal Care and the University of Calgary Committee on Animal Care (Protocol no. M01025).

Mrj genotypes were determined by PCR using primers 5′-CCAATAGCAGCCAGTCCCTTCCC-3′ and 5′-TTCGGCTATGACTGGGCACAACA-3′ to detect the neomycin-resistance gene of the β-geo cassette (226 bp, mutant allele); and 5′-ACAGGGTCTTGCTACAAGTAGTGC-3′ and 5′-TTTCTCTGTCCATGAAGGACTGGG-3′ to detect the β-geogene-trap insertion site in Mrj intron 2 (463bp, wild-type allele). The PCR parameters were 30 seconds at 94°C, 30 seconds at 63°C and 30 seconds at 72°C for 30 cycles. Genotyping for K18 and K19 (Krt19) mice was performed as previously described(Hesse et al., 2000; Magin et al., 1998).

Tetraploid aggregation chimeras were generated as described previously(Nagy et al., 2003). Briefly,tetraploid embryos were formed by electrofusing two-cell Egfpembryos, which were then aggregated with eight-cell diploid embryos from Mrj heterozygous matings. Aggregated chimeric embryos were allowed to develop to the blastocyst stage and were then transferred into the uteri of pseudopregnant CD-1 females [noon of the day that the vaginal plug is detected was defined as embryonic day (E)0.5]. Embryos were dissected at E9.5. The existence of wild-type tetraploid cells within the placenta was determined by the presence or absence of enhanced GFP (EGFP) expression using a GFP filter(41017 EN GFP, Leica). The yolk sac and embryo tissue was removed for Mrj genotyping. Placentas were fixed in 4% paraformaldehyde (PFA;VWR) in 1× phosphate buffered saline (PBS) and embedded in paraffin. Tissue sections were stained with hematoxylin and Eosin (both Sigma).

Wild-type (line 6-3) and Mrj homozygous mutant (lines 1-4 and 4-1)trophoblast stem (TS) cell lines were derived from littermate blastocysts generated from Mrj heterozygote matings. Blastocysts were allowed to outgrow on embryonic feeder cells in TS cell medium(Tanaka et al., 1998). These TS cell lines and the Rs26 TS cell line (provided by J. Rossant,Hospital for Sick Children, Toronto, Canada)(Tanaka et al., 1998) were maintained in 25 ng/ml basic fibroblast growth factor (bFGF), 1 μg/ml heparin in TS cell medium, 70% of which was pre-conditioned by embryonic fibroblasts, at 37°C under 95% humidity and 5% CO₂. Cells were differentiated by withdrawing bFGF, heparin and fibroblast-conditioned medium(Tanaka et al., 1998).

Chorion-ectoplacental cones were dissected at E7.5 from Mrjheterozygous crosses in 1× PBS. Explants were placed in 0.125%trypsin/EDTA (Gibco) in PBS for 10 minutes at 37°C. Dispersed cells were then cultured on coverslips in fibroblast-conditioned TS cell medium with bFGF and heparin for 5 days at 37°C in 95% humidity, 5% CO₂, with daily medium change. Yolk sacs were removed during dissection for genotyping.

Mouse conceptuses were dissected at E8.25 and E9.5. For X-gal staining,tissue was fixed, stained as whole mount and was then paraffin embedded as previously described (Hunter et al.,1999). For antibody staining, tissue was fixed in 4% PFA and embedded in paraffin. Paraffin sections (7 μm) were rehydrated in ethanol,treated with trypsin tablets (Sigma) in PBS for 10 minutes and blocked in 1%bovine albumin serum (BSA; Sigma), 5% host serum for 1 hour. Cell cultures were fixed in 4% PFA for 10 minutes, treated with 0.1% Triton X-100 (Fisher Biotech) for 10 minutes and blocked in 1% BSA for 30 minutes. Primary antibodies and dilutions used include: anticytokeratin 18 (Ks18.04, 1:100;RDI), anti-keratin 8 (Ks8.7, 1:100; Progen), anti-keratin 19 (Troma.3, 1:100;a gift from T. Magin), anti-pan-cytokeratin (1:100; DAKO), anti-desmoplakin(DP1&2, neat; RDI), anti-20S proteasome core subunits (1:100; Calbiochem)and anti-Mrj (020417, 1:100; a gift from M. Inagaki, Aichi Cancer Centre Research Institute, Aichi, Japan) (Izawa et al., 2000). Secondary antibodies were used at a dilution of 1:300 and included: goat anti-rabbit Cy3, goat anti-rat Cy3, goat anti-mouse Cy3 and goat anti-rabbit FITC (Jackson ImmunoResearch Laboratories), and goat anti-mouse Alexa Fluor 488 (Molecular Probes). Antibody dilutions were made in 1% BSA, 5% host serum. DNA was counterstained with 1:10,000 bisbenzimide(Sigma).

E8.25 implantation sites from Mrj heterozygous matings were fixed in 2% EM-grade glutaraldehyde (Sigma), 2% PFA in 0.2 M sodium cacodylate (pH 7.4; Sigma) overnight at 4°C, and were post-fixed in 1% osmium tetroxide(EM Sciences) in 0.2 M sodium cacodylate (pH 7.4) for 2 hours at 4°C. Tissue was dehydrated in increasing concentrations of ethanol up to 70%,treated in 2% uranyl acetate (ProSciTech) in 70% ethanol, and further dehydrated in increasing ethanol concentrations (70-100%). Tissue was treated with propylene oxide (EM Sciences) and was resin embedded (EMBED 812 kit, EM Sciences). Yolk sacs were removed for genotyping prior to fixation.

Rs26 TS cells were differentiated by withdrawal of bFGF, heparin and fibroblast-conditioned media for 4 days on coverslips. Cells were exposed to 10 μM clasto-lactacystin β-lactone (Invitrogen) in differentiating TS medium overnight at 37°C under 95% humidity and 5% CO₂. Cells were washed with 1×PBS, fixed in 4% PFA and immunostained as described above.

A Leica DMR light microscope (for fluorescent and light images), Leica DMIL inverted light microscope (for cell culture images) and Leica MZ95 dissection microscope (for dissected embryo/placenta images) with Photometrics Coolsnap cf camera and Open Lab 2.2.2 imaging software program were used to obtain micrographs. Filters used were from Chroma: DAPI (31000), TRITC (31002) and FITC (31001). De-convolved images were captured using an Axioplan 2 Imaging(Zeiss) microscope and Axiovision 4.1 acquisition software. Electron micrographs were taken on a Hitachi 7000 transmission electron microscope. Minimal image processing was performed using Adobe Photoshop 7.0 and figures were constructed using Canvas X.

The chorion layer of the mouse placenta is composed of two cell types at embryonic day (E)7.5: chorionic trophoblast cells and a thin inner layer of extra-embryonic mesothelium. Normally, chorioallantoic attachment begins at approximately E8.0 when the allantois, composed of extra-embryonic mesoderm,grows from the posterior end of the embryo across the exocoelomic cavity to first contact and then firmly attach to the chorionic mesothelium(Downs, 2002). Stable attachment is required for allantoic spreading along the surface of the chorion and for the formation of trophoblastic villi (reviewed by Watson and Cross, 2005). Mrj mutant conceptuses fail to undergo chorioallantoic attachment and as a result do not form a mature placenta, dying by E10.5(Hunter et al., 1999). Mrj mutant allantoises appear to both expand and elongate normally compared to wild type at E8.0 (Hunter et al., 1999), and grow across the exocoelomic cavity to contact the chorion. However, despite the close proximity of the allantois to the chorion,subsequent firm attachment to the chorion is not observed. Our initial observations have been confirmed in careful dissections of conceptuses after E8.5, showing the inability of the allantois to stay attached to the chorion in all mutants (n=81). Presumably as a consequence of failing to attach the chorion, the allantois rapidly collapsed and shrunk back to form a bud of tissue at the posterior end of the embryo by E9.5 (see Fig. S1 in the supplementary material). This phenomenon has been observed in other mutants with defects in chorioallantoic attachment(Hildebrand and Soriano, 2002; Tetzlaff et al., 2004; Yang et al., 1995).

Failure of chorioallantoic attachment in the Mrj mutants was presumably due to a chorionic cell defect, because Mrj expression has been detected within the chorion and not the allantois(Hunter et al., 1999). Usingβ-galactosidase activity driven from the Mrj genetrap allele(Fig. 1A) and Mrj antibody staining (Fig. 1B) on E8.25 histological sections, we detected Mrj expression throughout the chorionic trophoblast. However, expression within the chorionic mesothelium was indeterminable because of the thinness of this cell layer.

To address better whether Mrj is required within the chorionic trophoblast and/or the mesothelium for chorioallantoic attachment, we used tetraploid aggregation, a technique that has previously been shown to produce chimeric conceptuses with genetically distinct chorionic cell compartments(Fig. 1C)(Tarkowski et al., 1977). Wild-type tetraploid embryos were aggregated with diploid embryos derived from Mrj heterozygous matings. The resulting conceptuses developed so that the chorionic mesothelium, embryo proper and allantois were almost exclusively composed of diploid cells (Tarkowski et al., 1977). By contrast, wild-type tetraploid cells contributed to the trophoblast lineage, resulting in a chorionic trophoblast layer that contains wild-type cells (Tarkowski et al., 1977). Therefore, we expected that, if Mrj is required only within chorionic trophoblast cells, and not in the mesothelium or allantois,then chorioallantoic attachment would be rescued in Egfp⇔Mrj^-/- (tetraploid⇔diploid)chimeric conceptuses.

Chimeric blastocysts (n=24) were transferred into the uteri of pseudopregnant female mice. At E9.5, 1 day after chorioallantoic attachment is normally complete, all 15 surviving chimeric embryos had undergone chorioallantoic attachment. Genotyping revealed that two of these embryos were Mrj mutants (Fig. 1D,E), indicating that Mrj is not required within the mesothelium or allantois for chorioallantoic attachment to occur. Furthermore, development proceeded normally up to E9.5 in Egfp⇔Mrj^-/-chimeric placentas, as indicated by the initiation of branching morphogenesis(Fig. 1H), similar to Egfp⇔Mrj^+/+ and Egfp⇔Mrj^+/- littermates(Fig. 1F and data not shown). This is in contrast to Mrj-deficient chorions, which remained flat(Fig. 1G). These data imply that Mrj function was required exclusively within the trophoblast compartment at the chorioallantoic interface.

Previous studies have demonstrated that Mrj binds to both soluble and filamentous K18 (Izawa et al.,2000). Because, in our study, Mrj function in the placenta was restricted to the trophoblast lineage, where keratin is also expressed, we explored whether Mrj mutant cells had a defective keratin cytoskeleton. Using immunofluorescence, we examined K18-containing filament organization within Mrj^-/- trophoblast cells derived from primary cultures of explanted chorion-ectoplacental cone tissues. The chorionic trophoblast and ectoplacental cones are thought to be composed of multiple progenitor populations (Simmons and Cross, 2005), yet in culture they spontaneously differentiate primarily into trophoblast giant cells. Compared with 89.7% of wild-type trophoblast giant cells (n=659), only 36.9% of Mrjdeficient cells (n=634) had a normal, dense keratin filamentous network(Fig. 2A,B). Strikingly, the majority of Mrj^-/- cells exhibited a collapsed keratin network, in which large perinuclear inclusions of keratin were present (53.9%compared with only 5.0% of wild-type cells). The remaining Mrj^-/- cells contained either sparse keratin filaments(5.8%), or K18 levels were undetectable (3.4%).

All trophectoderm-derived cells express simple epithelial type I (K18 and K19) and type II (K7 and K8) keratins (Lu et al., 2005). Each keratin filament is composed of type I and type II heterodimers. K8 can dimerize with both K18 and K19(Owens and Lane, 2003). Therefore, we also immunostained cells using K8, K19 and pan-cytokeratin antibodies. Interestingly, the results were similar to the collapsed K18 cytoskeleton or sparse K18 filaments observed(Fig. 2A,B). Ultrastructural analysis of keratin aggregates using electron microscopy in vivo revealed that they consisted of closely packed, parallel filaments(Fig. 2C). From this, we concluded that Mrj was required for the maintenance and/or turnover of the keratin cytoskeleton, but not for the assembly of filaments per se.

The keratin cytoskeleton appeared normal in approximately a third of Mrj-deficient trophoblast giant cells derived from explants(Fig. 2B). However, we noted that 37.4% of wild-type trophoblast giant cells (n=1478) had undetectable levels of Mrj at E8.25 (Fig. 2D), indicating that not all trophoblast giant cells would be expected to show a phenotype. This suggests an alternative method of keratin regulation in Mrj-negative cells. Mrj was heterogeneously expressed in cultured trophoblast giant cells, either within the nucleus, the cytoplasm or both (Fig. 2E). This may be attributed to the presence of several trophoblast giant cell subtypes that appear both in vitro and in vivo (Simmons et al., 2007).

To investigate whether Mrj regulates the keratin cytoskeleton within the chorion, we analyzed K18 expression both in vitro and in vivo. In contrast to trophoblast giant cells, which showed heterogeneous expression, Mrj was uniformly expressed throughout the chorionic trophoblast layer(Fig. 3A). Trophoblast stem(TS) cell cultures are an established model for studying chorionic trophoblast cells in vitro (Tanaka et al.,1998; Uy et al.,2002). Therefore, we derived both wild-type and Mrj-deficient TS cell lines from littermate blastocysts generated from Mrj heterozygote crosses. Approximately 80% of wild-type TS cells (n=494) had normal K18 expression(Fig. 3B), whereas the remaining 20% had undetectable levels. By contrast, 78% of Mrj^-/- TS cells (n=469) lacked a normal keratin network and contained keratin inclusions similar to, albeit smaller than,those seen within primary cultures of Mrj^-/- trophoblast giant cells (Fig. 3C). K18 expression was less-apparent in histological sections of Mrj^-/- chorionic trophoblast cells in vivo compared with wild type (Fig. 3E-H). However,electron microscopy confirmed the presence of filamentous aggregates(Fig. 3D). These data indicate that Mrj mutant chorionic trophoblast cells show the same type of keratin defects as trophoblast giant cells - both the lack of normal keratin cytoskeleton and the presence of large keratin aggregates - albeit at a higher frequency owing to the fact that a subset of giant cells do not express Mrj protein.

Next, we examined the ultrastructure of Mrj-deficient chorions to determine the effects of keratin aggregate formation on tissue integrity. In comparison to wild-type chorionic trophoblast cells(Fig. 3I,K), Mrj^-/- cells were disorganized and rounded(Fig. 3J,L). This suggested that the adhesive properties of Mrj-null cells may have been altered. Therefore, we analyzed the expression of desmoplakin, a protein that links keratin filaments to desmosomes at the cell membrane. Remarkably, both Mrj^-/- chorionic trophoblast cells and TS cells showed a significant reduction in desmoplakin immunofluorescence(Fig. 3O,P,R) compared with wild type (Fig. 3M,N,Q). Ultrastructurally, however, desmosomes appeared to be normal in Mrj^-/- chorionic trophoblast cells(Fig. 3S,T). Together, these data imply that keratin inclusion bodies disrupt the organization of chorionic trophoblast cells, which in turn may affect the ability of the allantois to attach to the chorion. Notably, although they were disorganized, the Mrj^-/- chorionic trophoblast cells showed no signs of cell death.

Keratin filaments can oligomerize without catalysts or co-factors(Owens and Lane, 2003; Quinlan et al., 1986). Accordingly, keratin inclusions both in Mrj-deficient chorionic trophoblast and in trophoblast giant cells contained filaments(Fig. 2C, Fig. 3D), indicating that Mrj was not necessary for filament assembly. Therefore, we reasoned that Mrj may have played a structural role or have been required for keratin turnover. We found that Mrj protein did not co-localize with K18-containing filaments in trophoblast giant cells (Fig. 4), implying that Mrj was not playing a direct role in the organization of the keratin filamentous network.

It was previously shown that Mrj can interact with the soluble forms of K8/K18 (Izawa et al., 2000). Therefore, we next questioned whether Mrj regulated keratin by modulating the activity of the proteasome. Immunofluorescence of Mrj and 20S proteasome core subunits revealed that these two proteins co-localized(Fig. 5A). We then inhibited the proteasome complex in wild-type trophoblast giant cells by adding lactacystin (10 μM) and found that proteasome-inhibited cells contained large, perinuclear keratin inclusions (Fig. 5B) similar to Mrj^-/- cells(Fig. 2A, Fig. 3C). Interestingly, Mrj co-localized with keratin inclusions within proteasome-inhibited cells(Fig. 5C). These data suggest that, without Mrj, keratin filaments are not properly degraded by the proteasome, leading to filament accumulation.

Along with the formation of keratin aggregates, Mrj^-/-trophoblast cells also lacked a properly organized keratin cytoskeleton. Previous studies have shown that an absence of the epithelial keratin cytoskeleton results in hemorrhaging within the trophoblast giant cell layer due to a loss of trophoblast cell integrity(Hesse et al., 2000; Jaquemar et al., 2003; Tamai et al., 2000). Interestingly, only 6.8% of Mrj-deficient embryos (4/59) developed placental site hemorrhages (see Fig. S2 in the supplementary material). No hemorrhages were observed in wild-type or heterozygous conceptuses (0/147). The rarity of placental hemorrhage in Mrj mutants indicated that the consequences of Mrj deficiency differ from those of keratin deficiency.

To investigate further whether keratin deficiency was responsible for the placental phenotype in Mrj^-/- embryos, we assessed the occurrence of chorioallantoic attachment in keratin mutant mice. Chorioallantoic attachment was normal in both K18^-/-- and K19^-/--knockout embryos(Fig. 6B,C and data not shown),resulting in mice that are viable and fertile(Hesse et al., 2000; Magin et al., 1998; Tamai et al., 2000). Although K18^-/-;K19^-/- embryos have placental hemorrhages (Hesse et al.,2000), they all exhibited chorioallantoic attachment (n=8 double mutants from five litters) (Fig. 6D), indicating that this process can occur in the absence of a keratin cytoskeleton. This raised the possibility that keratin aggregation causes the Mrj phenotype. To address this, Mrj^+/-;K18^+/- mice were mated and the resulting progeny (n=9 litters) were assessed for the presence of placental hemorrhages and chorioallantoic attachment. The incidence of placental hemorrhage in Mrj mutants was unaltered in Mrj^-/-;K18^+/- (1/13) or Mrj^-/-;K18^-/- (1/9) conceptuses. By contrast, 77% of Mrj^-/-;K18^+/- embryos(10/13) and 100% of Mrj^-/-;K18^-/-embryos (9/9) exhibited rescue of chorioallantoic attachment(Fig. 6E,F; Table 1). Histological analysis of rescued Mrj^-/-;K18^+/- and Mrj^-/-;K18^-/- placentas at E9.5 confirmed that chorioallantoic attachment had occurred and that subsequent villous formation was initiated. However, the chorionic plates remained compact with little branching or fetal blood vessel growth(Fig. 6H,I). Furthermore, fewer branchpoints were seen in Mrj^-/-;K18^+/- chorionic plates(Fig. 6H) compared with those of Mrj^-/-;K18^-/-(Fig. 6I). This contrasted with wild-type and Mrj^+/-;K18^+/+ littermate placentas (data not shown and Fig. 6G, respectively), which had extensive branching and vascularization within the labyrinth layers of their placentas.

Notably, Mrj and K18 mutant mouse strains were maintained on different genetic backgrounds (CD-1 and 129Sv, respectively). However, Mrj^-/- conceptuses in both CD-1 and 129Sv backgrounds fail to undergo chorioallantoic attachment(Hunter et al., 1999),indicating that the 129Sv genetic background, itself, is not sufficient to rescue chorioallantoic attachment in Mrj^-/-;K18^+/- or Mrj^-/-;K18^-/- embryos. Together, these results suggest that, by reducing the amount of keratin expressed, and thus aggregated, chorionic trophoblast cell function was restored and chorioallantoic attachment was rescued.

It is well-known that protein aggregation is associated with several degenerative diseases. We have shown here that both Mrj-deficient trophoblast giant cells and chorionic trophoblast cells lack a normal keratin cytoskeleton and contain large keratin inclusion bodies. Furthermore, we have implicated Mrj in the regulation of keratin-filament degradation. The inclusions resulting from Mrj deficiency are cytotoxic and prevent chorioallantoic attachment during placental development by affecting cell function but not viability.

Cells containing keratin inclusions, such as Mallory bodies, also have a diminished keratin cytoskeleton (Denk et al., 2000). To date, this has complicated our understanding of the consequences of protein aggregation. In this study, we have shown directly that keratin deficiency and keratin inclusion bodies have different effects. In the developing placenta, complete deficiency of the keratin cytoskeleton(e.g. K8 mutants or K18/19 double mutants) is associated with normal chorioallantoic attachment, whereas keratin inclusions (e.g. Mrj mutants) are associated with a failure in chorioallantoic attachment.

An essential function of keratin is to protect the cell against mechanical stresses (Owens and Lane,2003). Mice that lack K8(Jaquemar et al., 2003), both K8 and K19 (Tamai et al., 2000) or both K18 and K19(Hesse et al., 2000) fail to form keratin filaments in simple epithelial cells, such as trophoblast cells. As a result, mutant embryos die at approximately E10.5 because of compromised trophoblast cell integrity, which leads to placental site hemorrhages that are particularly associated with trophoblast giant cells. Despite a compromised keratin cytoskeleton in Mrj mutants, placental hemorrhage was rare,indicating that Mrj deficiency is distinct from keratin deficiency. One explanation for this is that Mrj is not expressed in all trophoblast giant cells and, therefore, not all Mrjdeficient trophoblast giant cells lack a normal cytoskeleton or contain aggregates. This implies the involvement of another cochaperone in the regulation of keratin degradation within these cells. As a result, trophoblast giant cells with a normal keratin cytoskeleton are probably interspersed with affected cells. This arrangement may be sufficient to maintain the integrity of the trophoblast giant cell layer surrounding the implantation site, preventing hemorrhage.

Intermediate filaments support a cell by attaching to desmosomes, cell-cell adhesion junctions at the cell membrane. Desmoplakin is not only required to attach filaments to the desmosome but it also maintains desmosomal stability(Gallicano et al., 1998). Desmoplakin mutants die at E6.5 because unstable desmosomes lead to a loss of integrity within extra-embryonic tissues(Gallicano et al., 2001; Gallicano et al., 1998). Normal localization of desmoplakin at the desmosome within K18^-/-;K19^-/- trophoblast cells(Hesse et al., 2000) suggests that, in the absence of keratin filaments, desmosomes remain relatively stable. By contrast, Mrj-deficient chorionic trophoblast cells showed a significant reduction of desmoplakin expression. As a result, desmosomes in Mrj^-/- trophoblast cells may be unstable, resulting in disrupted chorionic organization. This implies further that keratin inclusion bodies disrupt cell function independently of keratin deficiency. Presumably,the reduction of K18 expression in Mrj^-/-;K18^+/- and Mrj^-/-;K18^-/- conceptuses decreased the amount of keratin aggregated, which in turn salvaged chorionic organization and rescued chorioallantoic attachment.

During chorioallantoic attachment, the allantois attaches directly to the chorionic mesothelium (Downs,2002). However, there are a handful of genes, including Mrj, that are expressed only in the chorionic trophoblast and which,when knocked-out, result in chorioallantoic attachment defects(Hildebrand and Soriano, 2002; Parr et al., 2001; Saxton et al., 2001; Tetzlaff et al., 2004). This reveals that the mesothelium must interact with the trophoblast compartment for it to become receptive to the allantois. Conceptuses deficient for Wnt7b, a gene encoding a signaling protein secreted by chorionic trophoblast cells, fail to express α4-integrin (encoded for by Itga4), a mesothelially expressed adhesion molecule required for allantoic attachment (Parr et al.,2001; Yang et al.,1995). Interestingly, Wnt7b^-/- chorions are disorganized (Parr et al.,2001), a phenotype reminiscent of Mrj deficiency. Although it was previously shown that α4-integrin is expressed in Mrj^-/- chorionic mesothelium(Hunter et al., 1999), it is unknown whether it is properly presented and active at the cell surface. Regardless, there are probably other unidentified adhesion mechanisms that are required because the chorioallantoic attachment phenotype inα4-integrin-deficient embryos is incompletely penetrant(Yang et al., 1995). This suggests that the disorganization of the Mrj^-/- chorionic trophoblast layer may disrupt signaling between the trophoblast and mesothelium. Without this interaction, the mesothelium may not be prepared to receive the allantois, resulting in a failure of chorioallantoic attachment.

Molecular chaperones are components of the protein-folding machinery and can also be transformed into protein-degradation factors when bound to the correct regulatory proteins (Esser et al.,2004). The DnaJ co-chaperones not only regulate the ATPase activity and substrate-binding specificity of Hsp70, but also play a significant role in determining which chaperone pathway Hsp70 should pursue(Esser et al., 2004). For example, binding of Ydj1, a yeast homolog of DnaJ, to Hsp70 promotes the proper folding of its substrate (Meacham et al., 1999). By contrast, a human DnaJ homolog, HSJ1 (also known as DNAJB2 - Human Gene Nomenclature Database), and HSP70 stimulate substrate-ubiquitination and -sorting to the proteasome for degradation(Westhoff et al., 2005). MRJ can bind and activate HSP70 (Chuang et al.,2002; Izawa et al.,2000) yet, until now, determining whether this complex is involved in protein-folding or -degradation had never been pursued.

A limited number of chaperones interact with intermediate filaments, giving us few clues about how chaperone complexes regulate the keratin cytoskeleton(Izawa et al., 2000; Liao et al., 1995; Liao et al., 1997; Perng et al., 1999). The keratin network is highly dynamic, but surprisingly little is known about its regulation (Coulombe and Omary,2002). A simple keratin filament, composed of heterodimers of type I (K18 or K19) and type II (K7 or K8) keratins, can form spontaneously without catalysts (Owens and Lane,2003; Quinlan et al.,1986). Hsp70 (Izawa et al.,2000; Liao et al.,1995) and Mrj (Izawa et al.,2000) can associate with both soluble and filamentous forms of K8/K18. However, our finding that keratin filaments do form, albeit ending up in aggregates, within Mrj^-/- trophoblast cells indicates that Mrj is not required for either keratin folding or keratin-filament assembly.

With the help of various proteins, keratin filaments bundle together and attach to desmosomes at the cell membrane to become organized into a dense network (Coulombe and Omary,2002; Owens and Lane,2003). We have shown here that Mrj is diffusely expressed throughout the cytoplasm, but does not localize onto K18-containing filaments in trophoblast cells. This suggests that Mrj is not involved in playing a structural role, which contradicts previous findings that described Mrj expression as filamentous (Izawa et al.,2000). This difference may be explained by the use of different cell types or by the resolution of imaging.

Keratin filaments can be remodeled or degraded by varying the levels of phosphorylation and ubiquitination(Coulombe and Omary, 2002; Ku and Omary, 2000). Our study demonstrates the involvement of Mrj in keratin degradation. As indicated above, Mrj does not colocalize to K18-containing filaments. However, it was previously demonstrated that Mrj can interact with the soluble fraction of K8/K18 proteins (Izawa et al.,2000). Furthermore, keratin inclusions in Mrj^-/- cells are ultrastructurally similar to Mallory bodies that appear in the hepatocytes of patients with chronic liver disorders(Denk et al., 2000). A defect in the degradation mechanism of the proteasome is one cause of Mallory body formation because cells treated with inhibitors of proteasome function(Bardag-Gorce et al., 2004) or cells that contain the UBB⁺¹ ubiquitin mutation(Bardag-Gorce et al., 2003)accumulate K18 and K8. We have shown here that Mrj protein co-localizes to the proteasome in trophoblast giant cells. When we chemically inhibited proteasome function in these cells, keratin aggregates formed that were similar to those in Mrj-deficient cells, suggesting that, in the absence of Mrj, K18 is not targeted to and/or degraded by the proteasome.

Mrj deficiency was associated with effects on the entire keratin cytoskeleton, such that K8-, K18- and K19-containing filaments aggregated. Previously, Mrj was shown to specifically interact with K18 but not with K8(Izawa et al., 2000); however,whether Mrj associates with K19 is unknown. Because K18 and K19 are both type I keratins, it is plausible that Mrj also directly regulates K19-filament degradation. Alternatively, what may start as a primary effect on K18 protein aggregation may indirectly impair the activity of the ubiquitin-proteasome system, causing a vicious circle of protein aggregation(Bence et al., 2001; Esser et al., 2004). Another hypothesis suggests that there is interdependence between K8/K18 and K8/K19 filaments such that, if one type of filament aggregates, the other collapses into the inclusion body as well.

Our study clearly supports the hypothesis that the formation of keratin inclusion bodies in Mrj-deficient trophoblast cells is detrimental to cellular function rather than protective. If non-degraded keratin is sequestered to protect the cell, we would expect that the chorioallantoic attachment defect observed in Mrj^-/- conceptuses is due to a non-functioning keratin cytoskeleton. However, K18^-/-;K19^-/- embryos underwent chorioallantoic attachment. Instead, by genetically reducing or eliminating K18 expression in Mrj-deficient embryos(Mrj^-/-;K18^+/- or Mrj^-/-;K18^-/-), and thus reducing the amount of keratin aggregated, we observed that chorionic trophoblast cell function was restored and chorioallantoic attachment was rescued. This suggests that the keratin inclusions themselves are toxic to cells. Although we have not documented when keratin aggregates first appear in the trophoblast giant cells of Mrj mutants, it is remarkable that implantation occurs and that post-implantation development progresses as far as it does.

There are many mutants within the literature with defects in chorioallantoic attachment, yet several of these display incomplete penetrance(reviewed by Watson and Cross,2005). Interestingly, mutant conceptuses that achieve chorioallantoic attachment will often exhibit later defects in morphogenesis of the labyrinth layer of the placenta (reviewed by Watson and Cross, 2005). Although chorioallantoic attachment was rescued in the majority of Mrj^-/-;K18^+/- and Mrj^-/-;K18^-/- placentas, the labyrinth layer remained under-developed such that little villous branching occurred within the chorionic plate and only a small amount of fetal vascularization was apparent. Furthermore, Mrj^-/-;K18^+/- labyrinths were more-severely affected compared to those of Mrj^-/-;K18^-/-, suggesting that the continued presence of keratin aggregates, even at a reduced level, affects branchpoint morphogenesis and the subsequent development of the labyrinth. Reduced villous formation in Mrj^-/-;K18^-/- placentas compared to wild type suggests that Mrj function may be required for labyrinth development independent of its role in keratin turnover. This implies that Mrj probably interacts with another substrate during this stage in placental development.

Mrj deficiency may affect a variety of tissues that express keratins, apart from the placenta, because Mrj is widely expressed both during later embryonic development and in the adult(Chuang et al., 2002; Hunter et al., 1999; Seki et al., 1999). In addition, Mrj has been shown to interact with other substrates(Chuang et al., 2002; Dai et al., 2005), notably the HD protein (Chuang et al.,2002). In Huntington's disease, mutant HD proteins with expanded N-terminal repeats aggregate within neurons(DiFiglia et al., 1997). In vitro, the Mrj-Hsp70 complex can reduce the accumulation of mutant HD aggregates (Chuang et al.,2002). These data are consistent with a general role for Mrj in preventing intracellular inclusion bodies, probably by promoting protein degradation through the proteasome.

We gratefully acknowledge T. Magin for the K18 and K19 knockout mice and K19 antibody; M. Inagaki for the Mrj antibody; and J. Rossant for the Rs26 TS cell line. We also thank T. M. Magin, C. Schuurmans, D. R. Natale, D. G. Simmons, M. Hesse and the Cross laboratory members for critical scientific discussions. This work was supported by a grant from the Canadian Institutes of Health Research (CIHR) (to J.C.C.). E.D.W. was supported by a Queen Elizabeth II Graduate Scholarship and a studentship from the CIHR Training Program in Genetics: Child Health and Development. J.C.C. is an Investigator of the CIHR and a Scientist of the Alberta Heritage Foundation of Medical Research.