Cytokeratins 8 and 19 in the Mouse Placental Development

Cytokeratin-intermediate filaments are encoded by a multigene family of ∼20 related polypeptides. They are expressed mainly in epithelial cells depending on the extent of differentiation (Moll et al. 1982). Cytokeratin proteins are classified into two types: smaller and acidic type I, and larger and more basic type II (Steinert and Roop 1988). In simple epithelia, they form cytokeratin filaments through obligate heteropolymerization of the type I and type II proteins in particular combinations. For example, cytokeratin 8 (K8) and cytokeratin 18 (K18); K8, K18, and K19; or cytokeratin 7 (K7), K8, K18 and K19 (Moll et al. 1982; Lane et al. 1983; Sun et al. 1984; Quinlan et al. 1985). The smallest type I cytokeratin, K19 (Mr; 40 kD) was first recognized as a distinct form in squamous cell carcinoma lines (Wu and Rheinwald 1981). Cytokeratins share a common structural organization: a central α-helical rod domain flanked by head and tail domains. Although K19 has a highly conserved central α-helical rod essential for filament formation (Steinert and Roop 1988) and the head domain, it is distinguished from other cytokeratins by lacking a tail domain (Bader et al. 1986; Ichinose et al. 1989; Lussier et al. 1989; Stasiak et al. 1989). The gene encoding mouse K19 (Krt1-19) is located on chromosome 11, in the immediate neighborhood of the K15 gene (Krt1-15), separated by only 4 kb (Lussier et al. 1990; Nozaki et al. 1994). The human K19 gene (KRT19) is located on chromosome 17 (Bader et al. 1988). While K19 and K15 are coexpressed in certain cell types, they are not in others (Ichinose et al. 1989; Nozaki et al. 1994). K19 is one of the most representative proteins of epithelial cells such as those in the intestine, kidney collecting ducts, gallbladder, mesothelium, and secretory glands (Moll et al. 1982; Sun et al. 1984; Quinlan et al. 1985; Bosch et al. 1988). In the mouse blastocyst trophectoderm, K19 is induced after K8 and K18 upon implantation (Jackson et al. 1980; Nozaki et al. 1988; Takemoto et al. 1991; Tamai et al. 1991). In the mouse embryo proper, K19 mRNA becomes detectable by 9.5 days post coitum (dpc) (Lussier et al. 1989), and is probably expressed earlier (Jackson et al. 1981; Franke et al. 1982). Differentially expressed individual cytokeratins, including K19, are remarkably stable markers in various types of carcinomas even when other markers have been lost (Kasper et al. 1987; Pujol et al. 1993; Moll 1994). Accordingly, cytokeratins are used for carcinoma subtyping regarding the extent of differentiation and their origins, especially for metastatic foci. Progress has been made in understanding the functions of cytokeratins in the skin. Point mutations in cytokeratin genes (Fuchs 1994; McLean and Lane 1995) and transgenic (Vassar et al. 1991) or knockout mice have indicated a structural function for epidermal keratins (for a review, see Magin 1997). Homozygous mutant embryos for the simple epithelia type K8 gene (Krt2-8) are retarded in growth and suffer from internal bleeding, with an abnormal accumulation of erythrocytes in the fetal liver (Baribault et al. 1993, Baribault et al. 1994). These alterations are consistent with a placental or extraembryonic deficiency. While this mutation shows a 94% penetrance in the C57BL/6 background, the proportion of viable homozygotes increases in the FVB/N background, and colorectal hyperplasia and inflammation are observed in adults that escape embryonic lethality. Recently, K8 and K18 have been shown to confer resistance to tumor necrosis factor–induced apoptosis (Caulin et al. 2000). Based on its structure and tissue distribution, K19 has been suggested to counterbalance type II keratins (Ecker 1988; Stasiak et al. 1989). To determine the expression pattern and function of K19, we have constructed a Krt1-19 null mutant allele by knocking in a lacZ cassette at the translation initiation site. We also generated compound homozygotes of Krt1-19 and Krt2-8 mutations and found that homozygous embryos are not viable due to defective placental tissues, a phenotype not observed in the simple homozygous mutants.

A λ recombinant clone containing the mouse Krt1-19 was isolated from a 129/Sv genomic library using a 1.4-kb full-length K19 cDNA fragment as a probe as previously characterized (Ichinose et al. 1989; Nozaki et al. 1994). A 6.5-kb SalI–KpnI fragment containing a part of exon 1 and a 641-bp KpnI–BglII fragment downstream from a KpnI site in exon 1 were excised from this recombinant phage and subcloned in the pUC19 vector. The ATG translation initiation codon in exon 1 was converted to a KpnI site. A 365-bp KpnI fragment in exon 1 was replaced with a cassette containing a lacZ reporter and PGKneobpA (Soriano et al. 1991). Thus, the targeting vector pJBX carried a long homology arm of 6.1 kb.

D3a2 embryonic stem (ES) cells (Shull et al. 1992) were cultured on neomycin-resistant (neo^r) mouse embryonic fibroblasts (Oshima et al. 1995). 50 μg of the targeting vector were linearized at the unique Sse8387I site and electroporated into 10⁷ ES cells at 250 V and 500 μF. 24 h later, 150 μg/ml (titer) of G418 (Geneticin; Sigma-Aldrich) was added. 7 d later, single colonies were isolated and cultured in duplicate. G418-resistant colonies were screened for homologous recombination by PCR using oligonucleotide primers: one complementary to neo^r; PGK^r (5′-CTA AAG CGC ATG CTC CAG ACT-3′) and the other located 92-bp downstream from the 3′ BglII site in Krt1-19; K19-R2G (5′-AAG AGC TCC CTG ACT AGA TTC AAG TTA ACT-3′). 35 cycles of amplification were performed (denaturation, 1 min at 94°C; hybridization, 2 min at 60°C; and elongation, 2 min at 72°C) in 50 μl reaction mixtures containing 50 pmol of each oligonucleotide, 1 U Taq polymerase (TAKARA), 0.5 U Perfect Match^® (Stratagene), and 1 μl of the crude cell lysate in each mixture. As shown in Fig. 1 (below), only homologous recombinants showed a 739-bp band. Candidates for homologous recombinants were expanded and verified by Southern hybridization with probes either internal (a 630-bp PstI–XbaI fragment containing the neo^r segment of pPGKneobpA) or external (a 556-bp fragment located downstream from the 3′ BglII site) to the targeting vector sequence.

Chimeras were generated by injection of the ES cells into C57BL/6 (B6) blastocysts, followed by transfers to MCH (CLEA) foster mothers. Chimeric mice were backcrossed to B6 mice, and germ line transmission was determined by the presence of the agouti coat color. Heterozygous offsprings were identified by PCR and Southern analysis of tail DNA samples. Tail tips were incubated in lysis buffer of SepaGene^® (Sankojunyaku), in the presence of 1 mg/ml proteinase K overnight at 55°C. DNA was extracted according to the manufacturer's protocol. The wild-type allele was identified using two primers sandwiching the deleted portion of exon 1 (sense primer, endo-F: 5′-CCT GAC TAG ATT CAA GTT AAC TG-3′, and antisense primer, endo-R: 5′-TGG CGG AGT CCG CGG TGG AAG TT-3′). 35 cycles (denaturation, 1 min at 94°C; annealing, 2 min at 60°C; and elongation, 2 min at 72°C) were performed to amplify the 958-bp fragment. The Krt2-8 alleles were identified as described before (Baribault et al. 1994).

For the timing of embryos, the morning of the vaginal plug was considered as 0.5 dpc. When removed from the decidua, embryos were staged according to Kaufman 1992 and genotyped as described above using DNA extracted from the yolk sac or from the embryo itself. To detect the β-galactosidase activity, the samples were incubated at 30°C for 30 min to 3 h, depending on their sizes in a staining solution (5 mM potassium ferricyanide, 5 mM potassium ferrocyanide, 1 mg/ml X-gal in PBS). For paraffin sectioning, samples were fixed in 4% paraformaldehyde according to the standard procedures.

For immunofluorescence staining, tissues were fixed in 100% methanol at 4°C for 10 min, rinsed in PBS, and embedded in the O.C.T. compound. The frozen sections were prepared at 20-μm thickness, followed by incubation in 0.3% H₂O₂/PBS for 30 min at room temperature. Sections of embryos and adult tissues were incubated with a blocking solution (3% BSA, 5% goat serum, and 0.2% Triton X-100 in PBS) for 1 h at room temperature. The primary antibodies TROMA-1, -2, and -3 raised against K8, K18, and K19, respectively, were obtained from Dr. Kemler (Max-Planck Institute of Immunology, Freiburg, Germany; Kemler et al. 1981; Boller et al. 1987), and were used at a dilution of 1:10 and incubated with the specimens for 1 h at room temperature. Antibodies against K7 and Kip2 were purchased from Euro-Diagnostica and Santa Cruz Biotechnology, Inc., respectively. Sections were washed, and then incubated with FITC-conjugated secondary antibodies (Amersham, UK). For peroxidase-coupled reactions, tissues were fixed in 4% paraformaldehyde or 70% ethanol and embedded in paraffin. Sections were incubated with peroxidase-conjugated secondary antibodies (Amersham Pharmacia Biotech) using a substrate kit (Vector Laboratories) according to the manufacturer's protocol.

Whole mount in situ hybridizations of the 9.0–9.5-dpc embryos were carried out as described previously (Saga et al. 1996) using digoxigenin-UTP–labeled single-strand RNA probe. The K19 probe was transcribed from a K19 cDNA fragment (Ichinose et al. 1989).

To construct Krt1-19 knock out mice, one of the Krt1-19 alleles was inactivated by homologous recombination in ES cells. The translation initiation site of Krt1-19 was disrupted so that a bacterial lacZ gene was placed under the control of the Krt1-19 promoter (Fig. 1 a). 200 G418-resistant clones were analyzed, and three independent homologous recombinant candidates were isolated. Homologous recombination in these ES clones were verified further by PCR and Southern analyses (Fig. 1 b). All three recombinant ES lines were injected into C57BL/6 (B6) blastocysts and two germ line chimeras were generated. About half of their agouti offspring carried the knockout allele. The heterozygous mutant mice were viable, fertile, and appeared normal.

Upon intercrosses of heterozygous mutants, the distribution of the wild type (+/+), heterozygous (+/−), and homozygous (−/−) mice were not significantly different from the expected Mendelian ratio in the F₁ and B6 backcross mice (Table). In the FBV/N backcross offspring, however, the homozygote numbers were smaller than expected (P < 0.05). While the reason for this phenomenon remains to be investigated, the FVB/N K19 mutant was maintained thereafter as homozygotes because they survived and reproduced without affecting the litter size or the sex ratio (data not shown). The homozygous Krt1-19 mutants did not show any particular phenotypic changes, and appeared identical to their heterozygous and wild-type littermates in their gross anatomy, histology, and behavior. Up to the age of 20 mo, we have not observed any difference in changes associated with aging.

To investigate K19 expression in the mutant mice, RNA samples from the small intestine and colon of the homozygous, heterozygous, and wild-type mice were analyzed by Northern hybridization (Fig. 1 c). The knockout mutation abolished the K19 mRNA in the small intestine and colon. The levels of the mRNAs for K8 that complexes with K19, and for K18, another type-I cytokeratin, were essentially unaffected in both heterozygous and homozygous mutants (data not shown). As expected, no K19 protein was detected in the homozygous tissues by Western immunoblot analysis with specific antibody TROMA-3 (data not shown). As shown in Fig. 2, immunofluorescence stainings of the intestinal sections demonstrated no K19 filaments in the homozygous tissues, whereas expression of K8, K18, and K7 proteins remained unaffected. It is worth noting that the K7 staining was observed in goblet cells in both genotypes.

To monitor K19 expression in various tissues, the targeting vector was designed so that a lacZ reporter gene was driven by the Krt1-19 promoter of the knocked-out allele. We determined the β-galactosidase reporter activity by whole mount in situ staining of various organs and embryos. As expected, simple epithelia of the small intestine, cecum, and colon were stained. In the intestines, not only the villus epithelium, but also the crypts were stained. The squamous epithelium of the esophagus, the hair follicle cells, and some gland cells in the skin also expressed β-galactosidase (data not shown). These results are consistent with the in situ mRNA analysis and immunostaining data published previously (Bosch et al. 1988).

In the preimplantation embryo, the trophectoderm cells of the blastocyst were stained only very weakly with X-gal. In the 6.0-dpc embryos, β-galactosidase was expressed only in the ectoplacental cone. In the 7.0-dpc embryo, the enzyme activity was detected mainly at the ectoplacental cone (data not shown). In the late primitive streak stage, β-galactosidase was expressed in the embryo proper, at the notochordal plate, and the node (Fig. 3, a and b). Shortly after (7.5–8.5 dpc), an arch-shaped staining appeared in the posterior embryo (Fig. 3 c). The diameter of this arch became smaller thereafter and the staining was localized in the hindgut pocket. When the foregut and hindgut invaginations were formed, the endodermal cells lining them expressed β-galactosidase (Fig. 3d and Fig. e). Later (9.0–9.5 dpc), β-galactosidase expression was limited to the primitive gut and the notochord (Fig. 3, f–h). In vivo, cytokeratins are not stable unless type I and type II polypeptides are copolymerized (Kulesh et al. 1989). To confirm the expression patterns monitored by the β-galactosidase activity, the K19 mRNA and protein were analyzed in the embryos. Expression of the endogenous Krt1-19 gene essentially corresponded to the β-galactosidase activity, although there were subtle differences in the signal intensities (data not shown).

To determine the expression patterns of K8 and K19, B6 embryos (6.5–7.5 dpc) were analyzed by immunohistochemistry with TROMA-1 (anti–K8) and TROMA-3 (anti–K19), respectively. As a result, both K8 and K19 proteins turned out to be expressed mostly in the extraembryonic tissues, in almost identical patterns. At 6.5 dpc, both K8 and K19 were expressed mainly in the ectoplacental cone and trophoblasts, although the level of K19 was lower than that of K8 (data not shown). These staining patterns were more obvious and stronger at 7.5 dpc. In addition to ectoplacental cone and giant trophoblasts, glycogen cells, parietal and visceral endoderm cells expressed both K8 and K19 at similar levels (Fig. 4, a–d). On the other hand, Kip2, a marker for the ectoplacental cone and trophoblastic giant cells (Riley et al. 1998), was expressed in the nuclei of these cells, as expected (Fig. 4e and Fig. f). Accordingly, compound homozygous mutants were constructed in the FVB/N background to determine whether K8 and K19 have compensatory functions in vivo. The FVB/N background was chosen because more than half of the K8 gene (Krt2-8) homozygotes survive to adults in the FVB/N background (Baribault et al. 1994), although the homozygous null mutation in the B6 background causes an embryonic lethality around 12 dpc (Baribault et al. 1993). First, the F₁ mice from the respective homozygous parents were obtained. These compound heterozygotes [Krt2-8 (+/−):Krt1-19 (+/−)] were viable and fertile. When they were intercrossed, however, no compound homozygous pups [Krt2-8 (−/−):Krt1-19 (−/−)] were found, although [Krt2-8 (+/−):Krt1-19 (−/−)] and [Krt2-8 (−/−):Krt1-19 (+/−)] were viable and fertile (data not shown). To determine the stage when the compound homozygous embryos became lethal, the concepti were examined chronologically from the intercrosses of mice that were heterozygous for the K8 mutation, but homozygous for K19 [Krt2-8 (+/−):Krt1-19 (−/−)] (Table). DNA extracted from fetal membranes or portions of embryos were used for genotyping. As a result, the embryos from 7.5–9.25 dpc showed the Mendelian distribution, and no difference was observed among the [Krt2-8 (+/+):Krt1-19 (−/−)], [Krt2-8 (+/−):Krt1-19 (−/−)], and [Krt2-8 (−/−):Krt1-19 (−/−)] embryos. After 9.5 dpc, however, the number of the compound homozygous embryos decreased gradually, with increasing numbers of necrotic remnants and resorption sites. At 9.75–10.5 dpc, some of the compound homozygous embryos were still alive, but they were easily distinguishable by the smaller size of their embryo-derived portion of the placenta and varying degrees of growth retardation compared with their littermates (Fig. 5).

The most prominent characteristic under a dissection microscope was much more blood at the implantation sites. Histological examination of serial sections of the trophoblastic tissues at 10.5 dpc revealed lesions filled with unnucleated maternal erythrocytes in the placenta of the compound homozygous embryos (Fig. 6, a, c, and e compared with b, d, and f, respectively). Upon closer examination at higher magnifications, the lesions also contained some large and nucleated embryonic erythrocytes as well (e.g., Fig. 6 g, arrows). Compared with the [Krt2-8 (+/−):Krt1-19(−/−)] or [Krt2-8 (+/+):Krt1-19 (−/−)] concepti, the placentas of the compound homozygotes contained multiple layers of giant cells without the normal architecture of the labyrinthine trophoblasts or spongiotrophoblasts. The allantois was also formed very poorly (Fig. 6, compare g with h).

To verify the histological findings at 10.5 dpc, and to investigate earlier changes in the compound homozygous placenta, we then investigated the tissues at 9.5 dpc. A trophoblast differentiation marker Kip2 was expressed in the ectoplacental cone, trophoblast giant cells, and diploid trophoblastic cells in the wild-type embryos (Fig. 4e and Fig. f). An immunohistochemical analysis in the compound homozygous placenta showed a marked increase in the number of the secondary giant cells already at 9.5 dpc (Fig. 7, a, c, and e compared with b, d and f, respectively). Moreover, these giant trophoblasts were with much larger cell bodies and nuclei, and they were pulled apart but not tightly attached to each other as in the simple K19 homozygotes. In contrast, the number of the labyrinthine trophoblast and spongiotrophoblast cells were decreased in cell number, and they were poorly organized. This apparently caused flooding of maternal blood directly into the embryonic tissues where these trophoblasts normally separate embryonic blood from the maternal circulation. Therefore, this lesion was not hemorrhaged as shown by lack of overt clotting (i.e, fibrin network formation), but there was intermixing of the maternal blood with the embryonic blood in the placenta due to the inability of the placenta to keep the circulation systems separate. The embryonic placenta, including the vascular structures, was also deteriorated, and the maternal and embryonic blood cells showed signs of degeneration by 10.5 dpc (data not shown).

On the other hand, staining for another type II cytokeratin K7 in the compound homozygous placenta showed a much weaker immunofluorescence than in the simple K19 homozygotes, and was found essentially in the trophoblasts, but not in the allantoic cells. At a higher magnification, the distribution of K7 in the compound homozygotes was more diffuse than in the K19 homozygotes (Fig. 7g compared with h). At the same time, a very weak K18 staining was observed in the K8 (−/−)/K19 (−/−) placenta (data not shown). Additional experiments will be needed to confirm the residual K18 and its presumptive partner, which may be K7 or an additional keratin (see Discussion).

Our results suggest that the K19 promoter is active in the gut epithelium both in embryos and adults. At ∼8.0 dpc, β-galactosidase was expressed in an arch pattern in the posterior embryo. And this expression was soon restricted to the inner wall of the hindgut pocket. If K19 is expressed in the stem cells of the endoderm, this arch-shaped cluster of cells is likely to be the precursor of the hindgut endoderm. Use of the K19 expression as a marker may help understand the gut differentiation and development in embryos and adults. Taking advantage of this finding, we recently constructed a gene knockin mouse strain in which a bacteriophage P1-derived recombinase gene cre was placed under the control of the K19 gene promoter (Krt1-19^cre). This mouse strain has turned out to be very useful in introducing conditional mutations in a gut-specific manner, because a floxed β-catenin–stabilizing mutation caused thousands of polyps in the intestines of the offspring when crossed with the Krt1-19^cre mice (Harada et al. 1999).

Type I cytokeratins, K18 and K19 can copolymerize with type II cytokeratins such as K8 and/or K7. Our results suggest that depletion of K19 per se does not interfere with embryonic development or postnatal life in the mouse, even though K19 is expressed at substantial levels in the developing embryo and in the adult. Given the evolutionary conservation of K19 across the vertebrates, it is reasonable to assume that K19 plays an important role that confers a selective advantage. It remains to be investigated whether K19 null mice show changes in such characteristics as tumor development or wound healing. Defects may become apparent in the K19 null mice when they are exposed to a less-protected environment rather than laboratory conditions. The absence of an overt phenotype may be due to compensatory functions of K18, which is coexpressed in the extraembryonic tissues, and K20, which is also coexpressed in the intestine (Moll et al. 1990; Calnek and Quaroni 1993). If K19 and K8 contribute to the same function, the two mutant alleles should interact genetically. Our results confirmed this prediction: the compound homozygous embryos have a more severe phenotype than embryos deficient in only one of the two genes. In the FVB/N background, >50% of the homozygous K8 null embryos survive to adulthood (Baribault et al. 1993). However, no compound homozygotes were found among >200 newborn pups. This result indicates a marked increase in the penetrance of the lethal phenotype to 100% in the compound homozygotes, possibly at an earlier stage than K8 homozygotes. The compound homozygous embryos have defects in the placenta, the major site of gas and metabolite exchanges between the maternal and fetal blood. The mouse placenta consists of three morphologically distinct trophoblast layers; namely, the labyrinthine trophoblast, spongiotrophoblast, and giant cell layers (Rossant 1995; Rossant and Croy 1985). In the compound homozygous placenta, giant trophoblasts were increased in number, whereas the spongiotrophoblasts and labyrinthine trophoblasts were decreased. Failure to properly form the chorioallantoic placenta results in embryonic lethality at midgestation (Cross et al. 1994; Copp 1995). Recent gene knockout results show that the placental development is controlled by such genes as basic helix-loop-helix transcription factors Mash2 (Guillemot et al. 1994) and Hand1 (Riley et al. 1998), orphan nuclear receptor Errb (Luo et al. 1997), and Ets2 (Yamamoto et al. 1998). In FVB/N K8 knockout mice that escape embryonic lethality, partial hepatectomy results in gross structural alterations and explanted hepatocyte death (Loranger et al. 1997). This may reflect hypersensitivity to tumor necrosis factor (TNF), which is moderated by K8 and K18 (Caulin et al. 2000). The primary cause of early embryonic lethality of K8 homozygous mutants is likely due to placental malfunction because, in the B6 background, K8-deficient embryos are rescued by aggregation with tetraploid embryos (Kupriyanov and Baribault 1998; Baribault, unpublished results). This leads to an intriguing consideration that TNF family cytokines may be involved in the placental abnormalities of both B6 K8-deficient embryos and the FVB/N compound homozygotes described here, because TNF is implicated in trophoblast function (Rasmussen et al. 1999).

Our results indicate that compound K8 and K19 homozygous mutants, in a permissive genetic background for K8 homozygotes, are defective in the development of placental tissues. K8 deficiency results in the complete loss of both K18 and K19 filaments in the liver and yolk sac (Baribault and Oshima 1991; Baribault et al. 1993). Why is the phenotype of the compound homozygotes more severe than the simple K8 homozygotes? Perhaps K8 is not the only type II keratin functional in the midgestational placenta. Indeed, low levels of K7 are detectable in embryonic placenta in addition to K18 and K19 (Fig. 7). The removal of K8 results in the degradation of K18 and most K19. However, the residual K19 that is polymerized with K7 would remain in the K8 homozygotes, whereas K19 is also removed in the compound homozygotes. The residual weak signal for K7 in the compound homozygotes may be due to K7/K8 heteropolymers.

It is worth noting that K18 null mice are viable, fertile, and show a normal life span. This may be due to compensation by K19. Old K18 null mice, however, develop a distinctive liver pathology with abnormal hepatocytes containing K8 aggregates (Mallory bodies). Additionally, in the K18 null mice, K7 was absent or markedly reduced in the uterus, where K7, 8, 18, and 19 are to be coexpressed. This observation indicates that cytokeratins have distinct assembly characteristics particular to the specific tissues (Magin et al. 1998).

In conclusion, we have constructed a K19 gene knockout mouse strain, the homozygotes of which were viable, fertile, and appeared normal. When the mutation was introduced into a viable K8 gene knockout mouse strain, however, the compound homozygous embryos died in utero due to defects in the placenta. These results indicate that K19 and K8 cooperate in ensuring the normal development of placental tissues.

We thank M. Oshima and Y. Saga for helpful discussions, and R. Kemler for the gift of the TROMA-1, -2, and -3. We also thank N. Sugimoto, R. Nakajima, and H. Oshima for technical assistance.

This work was supported in part by the Joint Research Fund between The University of Tokyo and Banyu Pharmaceutical Co., grants from Monbusho (MESSC) and the Organization for Pharmaceutical Safety and Research of Japan, and grant CA42302 from the National Cancer Institute.