The novel protein kinase Vlk is essential for stromal function of mesenchymal cells

Gene mining has been facilitated by the recent completion of the whole-genome sequences of an increasing number of species, and by the development of high-throughput technologies, such as DNA microarrays, to detect gene expression. Among various settings for gene mining, in vitro differentiation of embryonic stem (ES) cells towards specific lineages has emerged as a promising experimental system that has marked advantages over study of the embryo itself, from which only small numbers of cells can be obtained (Hirashima et al.,2004; Keller et al.,1993; Nishikawa et al.,1998). We have been using ES cell differentiation in combination with high-throughput technologies to identify genes that are involved during early embryogenesis (Takebe et al.,2006).

Protein kinases (PKs), which constitute one of the largest eukaryotic gene families, regulate a variety of cellular processes. Mutation and dysregulation of a variety of PK genes have been implicated as causes of disease, including cancer. The way in which PKs have been studied has changed markedly with the completion of numerous whole-genome sequencing projects. A list of genes encoding PKs, including more than 500 PKs in both human(Manning et al., 2002) and mouse (Caenepeel et al., 2004)genomes, is now available(http://kinase.com). However, a proportion of these PKs remain functionally uncharacterized. Human PKs have been classified into a hierarchy of groups, families and subfamilies according to the sequence of the catalytic domain(Hanks and Hunter, 1995; Hanks et al., 1988). As each family contains well-characterized PKs, this hierarchy is being used to gain insight into the potential functions of PKs. Phylogenetic comparison is another means of investigating the functional significance of PKs. Indeed,most human PKs (510 of 518) have mouse orthologs, and the functions of many PKs are preserved across species. However, this `kinome' map also contains a few genes that cannot be fully classified into any of the characterized families. Thus, it is still important to assess their expression pattern and function on an individual basis.

In this study, we attempted to list PKs that are newly induced after differentiation to mesendoderm in ES cell culture. Among PKs that are found in mesendoderm cells but not ES cells, we found one PK, Vlk, that could not be classified into any known PK family. This previously uncharacterized PK is unusual in that it has no homologs in invertebrates, although it is highly conserved in vertebrates. We show here that expression of Vlk retards protein transport from the Golgi. Analysis of Vlk^-/- mutant mice showed that a functional environment for the organogenesis of lung, bone and palate is disturbed.

GFP and Venus proteins were detected using rabbit anti-GFP antibody(Molecular Probes). Goat anti-SP-C, anti-aquaporin 5 and anti-CC10 antibodies(Santa Cruz) were used in lung sections. Rat anti-BrdU antibody (Serotec) was used in the proliferation studies (2-hour labeling of pregnant mother). Paraffin sections and cryosections were 6 μm and 8 μm, respectively. NIH3T3 cells stably expressing Vlk-GFP or transiently expressing VSVG-GFP (constructs kindly provided by Dr Jennifer Lippincot-Schwartz, NIH) were stained with mouse anti-p115, anti-GM130,anti-Vti1a, anti-GS15 (BD Pharmingen), rabbit anti-TGN38 (Abcam) or rabbit anti-calreticulin (StressGen) antibodies. Fluorescent secondary antibodies from Molecular Probes and HRP-conjugated antibodies from Jackson ImmunoResearch were used.

In the VSVG-GFP transport assay, cells were cultured at 40°C overnight after transfection and 100 μg/ml cycloheximide was added 30 minutes before the temperature shift to 32°C. For FACS analysis of VSVG-GFP, anti-VSVG(8G5) (kindly provided by Dr Douglas S. Lyles, Wake Forest University) and anti-mouse-Alexa633 (Molecular Probes) antibodies were used.

EB5 and Gsc^Gfp+/- EB5 cell lines were maintained as described previously (Tada et al., 2005; Yasunaga et al., 2005). Mesendoderm induction was performed as described previously(Tada et al., 2005). For gene targeting, the TT2 ES cell line was maintained as described previously(Yagi et al., 1993). HEK293T cells, NIH3T3 cells, and their derivatives were maintained in DMEM containing 10% FCS.

Differentiated cells were purified by FACS at each time point. RNA extraction using Trizol (Invitrogen), cRNA probe preparation and hybridization to Affymetrix oligonucleotide arrays MGU74v2 A, B and C were performed as recommended by the manufacturers, as described previously(Takebe et al., 2006). The expression data were maintained and analyzed using the eXintegrator system(http://www.cdb.riken.jp/scb/documentation/). Probe sets representing kinase genes were identified by searching for the term`kinase' in the Ensembl (version 42) annotation for genes associated (by eXintegrator) with probe sets. These probe sets were then sorted by their similarity (Euclidean distance) to a specified profile with low or non-existent signals in ES and day-5 Gsc⁺ E-cadherin^-samples and high signals in day-4 and day-5 Gsc⁺E-cadherin⁺ samples. The comparison was carried out using an eXintegrator function that compares the individual probe-pair profiles against the specified profile and uses the mean Euclidean distance. The expression data for the resulting list of genes were then manually inspected to remove genes with expression in the ES samples and probe sets with ambiguous identities (i.e. mapping to intronic sequences). The top 36 probe sets were selected for further consideration and are listed in Table 1. Additional annotation,as well as confirmation of the probe set to gene relationships, were obtained from the Ensembl mart_42 database (ArrayExpress accession number E-MEXP-1672).

A cDNA clone representing the Vlk transcript was isolated from a lambda phage cDNA library constructed from mRNA of day-4 differentiated CCE ES cells (Nishikawa et al.,1998). The phylogenetic relationship of the Vlk sequence was determined by ClustalX (Thompson et al., 1997) followed by MEGA4(Tamura et al., 2007). Kinase subdomains in Vlk were identified by first aligning the Vlk peptide sequence to the Pfam13.0 kinase hidden Markov model (PF000069.11) using the HMMER package(http://hmmer.janelia.org/). This alignment was then visually inspected for the conserved domains reported previously (Hanks and Hunter,1995; Hanks et al.,1988).

The targeting vector was constructed according to the method described previously (Ikeya et al.,2005). The genomic DNA of G418-resistant colonies was digested with SacI and analyzed by Southern blotting. The PGK-Neo cassette was removed by transient expression of Cre or crossing of chimeric mice with EIIa-Cre mice (Jackson Laboratory). Correctly targeted clones were injected into ICR embryos. The RIKEN accession number of Vlk knock-in mice is CDB0434K.

In situ hybridization was performed as described previously(Takebe et al., 2006). The mouse Vlk probe was amplified by RT-PCR using primers 5′-TGGAGACAGCGCTACACTTG-3′ and 5′-CCCAGTTAAGACCAGCCTGA-3′. For chick Vlk,5′-GCTACACCAAGGCCGTCTAC-3′ and CCTGTCCAAGTCGTCTGGTT-3′ were used; for zebrafish Vlk, 5′-GTGGACGGAGTGTTGAAGGT-3′ and 5′-GCACCGGTCTTAAAGAGCAC-3′ were used. Type II collagen alpha I(Col2a1) and type X collagen alpha I (Col10a1) constructs were kindly provided by Dr B. R. Olsen (Harvard Medical School).

Transient transfection was performed using TransIt LT-1 (Mirus Bio). Cells were lysed with buffer [20 mM HEPES pH 7.5, 3 mM MgCl₂, 100 mM NaCl, 1 mM DTT, 100 mM Na₃VO₄, 10 mM NaF, 20 mM beta-glycerophosphate, protease inhibitor cocktail (Nacalai), 1 mM EDTA, 1%Triton X-100]. Immunoprecipitation was performed using protein G-agarose(Roche). Anti-phosphotyrosine (Upstate), anti-V5 (Invitrogen), anti-GFP(Molecular Probes) and anti-Vlk (clone 7F2, raised in our laboratory) were used. After SDS-PAGE, staining was performed with ProQ Diamond (Invitrogen) or CBB Stain One (Nacalai). For mass spectrometry analysis, protein bands were excised from CBB-stained gels, trypsinized and analyzed by LC/MS-MS.

Alizarin Red and Alcian Blue staining of P0 skeletons was performed as described previously (Kimmel and Trammell,1981).

We previously reported a defined culture condition that allows the selective differentiation of mouse ES cells into goosecoid⁺(Gsc⁺) mesendodermal cells, which subsequently differentiate into E-cadherin⁺ (E-cad; Cdh1 - Mouse Genome Informatics) definitive endoderm and E-cad^- mesoderm(Tada et al., 2005). During this process, the growth requirements of the differentiating cells change markedly from those of ES cells. In this study, we compared ES and differentiated mesendoderm cells in terms of expression of PKs. We first listed PKs that are expressed in mesendoderm but are not detectable in ES cells (Table 1). A number of these PKs have been implicated in stress responses [Erk5(Mapk7), Mapkapk3, Map3k7, Map2k4, Jnk (Mapk8), Map2k5 and Ikkb (Ikbkb)], suggesting that some of the genes are induced as a stress response to the defined culture condition.

We were also intrigued by the existence of a PK represented by the Affymetrix 104066_at probe set and identified by MGI symbol AW548124(Pkdcc), but about which nothing has been reported. One reason for this gene remaining unnoticed in the mouse might be that the genomic sequence upstream of this gene was not clearly defined, resulting in a predicted protein of 293 amino acids containing only a partial kinase domain. However,we isolated a cDNA clone containing the missing upstream sequence and demonstrated the presence of further coding sequence. We found that this gene encodes an mRNA of 2440 nt (DDBJ accession number AB381893) and a protein of 492 amino acids that contains a putative kinase domain.

Interestingly, we failed to find any homologous proteins in invertebrates by BLAST search using a stretch of sequence encoding 237 amino acids (138-374)in the kinase domain, whereas an ortholog of this protein exists in every vertebrate species for which the whole-genome sequence is available. Protein alignment reveals a high degree of conservation between the mouse, human, rat and zebrafish proteins (see Fig. S1A in the supplementary material). The human ortholog, designated SGK493, comprises a solitary branch without any other members that stems directly from the origin according to the map of the human kinome(http://kinase.com). In this tree, seven more such solitary branches, each of which has only one member, are present. We performed a TBLASTN search using a 237 amino acid stretch of the kinase domain over the nucleotide collection database (nr/nt). The results of the TBLASTN search were aligned using ClustalX(Thompson et al., 1997) and clustered using MEGA4 (Tamura et al.,2007). The phylogenetic tree of the TBLASTN results and of vertebrate Vlk proteins are shown in Fig. S1B,C in the supplementary material. The BLAST results revealed no invertebrate homologs, despite the high degree of conservation among vertebrates. Although there were a number of hits in plant and invertebrate genes, the homology was not significant. Because of this feature, we named this gene vertebrate lonesome kinase(Vlk).

During mesendoderm differentiation of ES cells, Vlk begins to be expressed in mesendodermal cells on differentiation day 4, as assessed by RT-PCR, and is then expressed only in E-cad⁺ endoderm precursor cells but not in E-cad^- cells (data not shown). We examined the correlation of this gene expression pattern with that in embryos by whole-mount in situ hybridization. Whereas Vlk expression was detected in the anterior endoderm in late gastrulation embryos(Fig. 1B), we failed to detect a signal in the organizer region of gastrulating embryos (E6.5-7.5)corresponding to the earliest mesendoderm detected in culture. As it is possible that this failure was due to insufficient detection sensitivity, we attempted to increase the sensitivity using a reporter gene, the expression of which corresponds to that of the endogenous Vlk gene. For this purpose, we established a mouse strain in which the Venus reporter gene (Nagai et al., 2002) was inserted into the Vlk allele (see Fig. S3 in the supplementary material). Consistent with our expectations, the reporter gene expression pattern indicated that Vlk begins to be expressed in the organizer region at the onset of gastrulation (∼E6.5) as well as in the anterior visceral endoderm (AVE) (Fig. 1A). At the late gastrula stage, Vlk was detectable in the node region in addition to the AVE(Fig. 1B-D). During gastrulation, the highest level of Vlk expression was found in the AVE. These results indicate that the in vitro expression pattern of Vlk is largely consistent with that observed in the embryo.

We also analyzed the expression of Vlk orthologs in gastrulae of zebrafish and chicken by in situ hybridization (see Fig. S2 in the supplementary material). In zebrafish, expression was first detected in the shield corresponding to the organizer at 6 hours post-fertilization (hpf) (see Fig. S2A in the supplementary material). At 12 hpf, expression was confined to head ectoderm and mesoderm (see Fig. S2B in the supplementary material). In the chick gastrula, Vlk expression was detectable only in the anterior endoderm (see Fig. S2C in the supplementary material). Although we failed to detect expression in the chick organizer, the overall expression pattern of Vlk appears to be largely preserved in vertebrate gastrula.

Using the Venus reporter mouse strain described above, Vlk expression was detected mainly in mesenchymal cells, but not in the E-cad⁺ populations in embryos, including the definitive endoderm and embryonic ectoderm. In E9.5 embryos, the strongest levels of expression were observed in the branchial arches and limb buds(Fig. 1E-G), tissues in which mesenchymal cells are condensed. Double staining of sections for E-cad showed that Vlk is not expressed in most of the endodermal cells, except for a small number of double-positive cells in the foregut at E9.5 (data not shown). Although most neuronal cells were negative for Vlkexpression, expression was detected in neuronal cells in the ventral part of the mid-brain (Fig. 1F).

During mid-gestation, Vlk expression in mesenchymal cells continues (Fig. 1K shows the Vlk expression pattern in an E14.5 sagittal section). Its expression was particularly conspicuous in areas where mesenchymal cells condense, such as beneath the hair bud (Fig. 1H), the lining of the developing bronchial tree(Fig. 1I), in cells aligned with the elongating choroidal plexus (Fig. 1L) and in regions of chondrogenesis and osteogenesis(Fig. 1N-P). Similarly, the mesothelium wrapping the visceral organs, such as the peritoneum, pericardium and pleura (Fig. 1M), which are derived from mesenchymal cells, showed high levels of Vlk expression,suggesting that Vlk is expressed in mesenchymal cells undergoing dynamic change.

Vlk expression in mesenchymal cells was further examined during limb bone formation (especially of the humerus) and lung organogenesis as we found defects in the formation of these tissues in Vlk-null mutant mice (as described below). Expression of Vlk was detected throughout all stages of humerus bone formation. During limb formation, condensed mesenchymal cells in the limb bud expressed Vlk. Vlk expression decreased in chondrocytes as they differentiated from condensed mesenchyme,whereas expression remained stable in the perichondrium. No Vlkexpression was found in digit chondrocytes at E12.5 (the developing paw showed no Vlk expression), whereas strong expression was observed in the primordium of the more-proximal bone (Fig. 1N). Vlk expression in chondrocytes regained strength as they became hypertrophic, and Vlk continued to be expressed during osteogenesis (Fig. 1M,P). Thus,the Vlk expression level fluctuates during the development of skeletal bones, and its expression level is high at the transition stages from mesenchymal cells to chondrocytes and from chondrocytes to osteoblasts.

Lung development in mice has been classified into four stages involving tracheal bud generation: the pseudoglandular stage (E9.5-16.5), with generation of tracheal branching; the canalicular (E16.5-17.5) and saccular stages (E17.5-P5), with generation of alveoli; and the alveolar (P5-30) stage involving the maturation of alveolar ducts and alveoli. Expression of Vlk was observed in all mesenchymal cells in the lung at the early pseudoglandular stage (E13.5, Fig. 1I). The highest level of expression at this stage was in the condensed mesenchymal cells lining the developing trachea(Fig. 1I). However, expression was rapidly lost from the mesenchyme of the lung interstitium during the mid-to-late pseudoglandular stage. By contrast, at P0 (saccular stage), the mesenchymal cells in the subpleural region and the pleural cells themselves continued to express Vlk (Fig. 1J). The mesothelial pleural cells continue to express Vlk throughout life (data not shown).

In summary, Vlk is first expressed in the mesendoderm of gastrulation stage embryos, including the early organizer and AVE cells. However, after gastrulation, its expression is found mainly in mesenchymal cells. Although most mesenchymal cells express Vlk to some degree,its expression level fluctuates. There is a clear tendency for Vlkexpression to be high at transition phases of mesenchymal cell differentiation, such as from mesenchymal cells to chondrocytes and from mesenchymal to mesothelial cells.

To determine the subcellular localization of Vlk, we constructed a GFP-tagged Vlk chimeric gene. Fig. 2A-E shows double immunostaining of NIH3T3 cells with anti-GFP and antibodies against Golgi proteins, including p115 (Uso1) [from endoplasmic reticulum (ER) exit site through ER-Golgi intermediate compartment (ERGIC) to cis-Golgi], GM130(Golga2) (ERGIC and cis-Golgi), GS15 (Bet1l) (medial-Golgi), Vti1a(trans-Golgi) and TGN38 (Tgoln1) (trans-Golgi network). The GFP signal overlapped to various degrees with those of GM130 and GS15, but segregated completely from that of TGN38.

The Vlk gene product contains at least kinase subdomains I, II,III, IV, VI, VII and VIII (partial match), but lacks subdomain IX (see Fig. S1A in the supplementary material). However, we failed to detect any kinase activity by standard in vitro kinase assays involving autophosphorylation or phosphorylation of myelin basic protein (Mbp) as a target substrate (data not shown). As its putative substrate is unknown, we next analyzed its kinase activity by comparing wild-type (WT) and mutant proteins in transfected cells.

The V5-tagged WT Vlk gene introduced into HEK293T cells was tyrosine phosphorylated, as determined by western blotting with an anti-phosphotyrosine antibody (4G10) following immunoprecipitation by an anti-V5 monoclonal (Fig. 2F). By contrast, no tyrosine phosphorylation was detected when we performed the same assay using a V5-tagged Vlk gene with a mutation in the putative ATP-binding site (K166M) (Fig. 2F). This result was confirmed by SDS-PAGE and by staining of these immunoprecipitates with ProQ Diamond, which stains phosphoproteins(Fig. 2G). Thus, in vivo tyrosine phosphorylation of Vlk is dependent on the intact kinase domain of Vlk itself, although its activity could not be detected by in vitro autophosphorylation assays. Mass spectrometry analysis identified tyrosine 148 and serine 177 as phosphorylation sites dependent on Vlk. Taken together,these observations suggest that phosphorylation of Vlk requires a phosphorylation cascade in which Vlk itself is involved as a kinase. To test this hypothesis, we investigated whether kinase-dead mutants, such as K166M-Vlk and (K166M+E208A)-Vlk, could be phosphorylated in HEK293T cells when co-expressed with WT Vlk. In this experiment, the WT Vlk gene was tagged with GFP to allow it to be distinguished from the mutant Vlkgene product. In accordance with our expectations, we detected phosphorylation of the mutant Vlk only when co-transfected with WT Vlk(Fig. 2H,I). Taken together,these observations indicate that Vlk probably has kinase activity and that Vlk autophosphorylation is dependent on this activity. The experiments shown in Fig. 2F-I were also confirmed using the NIH3T3 mesenchymal cell line (not shown).

To examine Vlk gene function in vivo, we deleted the first 213 amino acids of the N-terminus, including amino acids that form part of kinase subdomains I, II and III, and introduced the Venus gene with a stop codon in-frame with the first methionine. This construct should thus lead to disruption of endogenous gene expression. Vlk^+/- mice are healthy and fertile and we obtained Vlk^-/- mice by intercrossing Vlk^+/-. Vlk^-/- mice died soon after birth owing to respiratory failure. Contrary to our expectation from Vlk expression during gastrulation, Vlk appears to be dispensable for early embryogenesis. Vlk^-/- neonates showed growth retardation (Fig. 5A), craniofacial abnormalities(Fig. 3I), cleft palate(Fig. 3A), cyanosis and suckling defects (Fig. 4A),shortened limbs and delayed ossification (see Fig. 5B,C).

At E13.5, the developing palatal shelves are located at the lateral side of the tongue in both Vlk^+/+ and Vlk^-/-embryos (Fig. 3B,C). At E14.5,the palatal shelves change shape and position, with each palatal shelf elevating above the tongue and growing horizontally in the axial direction before finally fusing at the midline through an epithelial-to-mesenchymal transition (Fig. 3D). In Vlk^-/- mice, the process of palatal shelf elevation was severely disturbed, either unilaterally(Fig. 3E) or bilaterally(Fig. 3F). This appears to be the cause of the failure of the palate to close in all Vlk^-/- animals analyzed. Thus, Vlk plays an essential role in guiding the direction of palate elongation.

In the developing palatal shelf, stronger expression of Vlk was found in the middle part and inner side of the palatal process(Fig. 3G,H). As we could not detect a significant difference between Vlk^-/- and Vlk^+/- in the number of proliferating Vlk⁺mesenchymal cells in the upper half of the palatal shelf (dotted line in Fig. 3H, Fig. 3I) or in the number of cells undergoing apoptosis (data not shown), Vlk might be involved in the process of generating some morphogenic constraint that is required to induce the bending of the elongating process inward. Owing to this severe cleft palate, all Vlk^-/- offspring showed suckling problems as evidenced by the absence of colostrum.

In addition to the cleft palate, Vlk^-/- mice showed malformation of the craniofacial architecture. Fig. 3J shows Alizarin Red and Alcian Blue staining of head structures in neonates at P0. By comparison with controls, the nasal capsule and maxilla bone were small and severely shortened, although all bones were formed.

Vlk^-/- pups had cyanosis owing to respiratory failure(Fig. 4A). As shown in Fig. 4B,C, the neonatal lung showed enlarged air spaces with a marked decrease in the number of alveoli and an increase in the thickness of the alveolar septum. This phenotype suggests that Vlk is involved in alveolus formation by septating alveolar saccules (Massaro and Massaro,1996). Despite such severe deformity of the lung structure, most offspring survived for a few hours. This is probably because all the components required for formation of the basic architecture necessary for respiratory function were generated in the Vlk^-/- mice. No significant differences were detected in the expression of surfactant proteins between Vlk^+/+ and Vlk^-/- mice by RT-PCR (data not shown). Moreover, all types of alveolar and bronchial components were identified by both histological and immunohistological examination of E18.5 lung. The number of type II alveolar cells detected by anti-SP-C (Sftpc) staining was similar in Vlk^-/- and Vlk^+/- lung (Fig. 4D,E). Similarly, type I alveolar cells(Fig. 4F,G) and ciliated tracheal cells (Fig. 4H,I) were present in their normal positions. However, in the lungs of Vlk^-/- mice, SP-C⁺ cells that are normally localized to the epithelial lining of the alveolar lumen were instead found in both the alveolar lining and in the interstitial region(Fig. 4D,E), again suggesting a defect in the sacculation process. However, there were no notable differences in the tissue localization of CC10⁺ (Scgb1a1) tracheal epithelial cells. Taken together, Vlk is involved in the patterning of alveolar cells during the sacculation stage, whereas tracheal development is unaffected in its absence.

It should be emphasized, however, that Vlk is expressed in interstitial but not alveolar cells. Moreover, as shown in Fig. 1I, Vlk⁺mesenchymal cells that are distributed homogeneously over the lung at the pseudoglandular stage move progressively toward the distal region of the lung as the size of the lung increases. Hence, this defect in the formation of alveoli, which occurs during the late pseudoglandular to saccular stage of lung development, should be caused by events during the early pseudoglandular stage when Vlk is expressed in the mesenchymal cells adjacent to the developing alveolar cells. Interestingly, the alveolar acinus appears to be assembled within the region where Vlk is expressed at high levels. In Vlk^-/- mice, this acinar formation was significantly retarded, whereas no significant difference was observed in the number of E-cad⁺ cells (data not shown).

All offspring of Vlk^-/- mice were significantly smaller than those of Vlk^+/+ mice owing to defects in the development of the skeleton. In Vlk^-/- mice, the ossification of skeletal bone such as limbs and vertebrae, as detected by Alizarin Red staining, was suppressed (Fig. 5A-C). However, there were only minimal differences in membranous bone formation (Fig. 3J; Fig. 5D). As we described previously, a biphasic expression of Vlk is observed during limb formation; the first peak occurs during the mesenchymal condensation prior to chondrogenesis, and the second at the stage of hypertrophic chondrocyte differentiation prior to osteogenesis. No defect was detected in the process from mesenchymal condensation to chondrogenesis. However, the differentiation of hypertrophic chondrocytes was markedly retarded. Normally, at E14.5,chondrocytes at the center of the humerus become hypertrophic and lose expression of collagen II mRNA but gain collagen X expression(Fig. 5E,G). In the mutant bone, however, all chondrocytes of E14.5 embryos remained positive for collagen II and no expression of collagen X mRNA was observed(Fig. 5F,H). Differentiation of hypertrophic chondrocytes, as assessed by collagen X expression, was delayed by ∼2 days in the Vlk^-/- embryos(Fig. 5I-L). This delay might result in a subsequent delay of the ossification process, leading to a small body size. As was the case for mesenchymal cells in other regions, we did not find any significant change in BrdU incorporation(Fig. 5M) or any enhancement of apoptosis (data not shown) in proliferating chondrocytes. However, the orientation of chondrocyte growth appeared to be misdirected in the mutant because the humerus became thicker at the expense of length. Taken together,these observations indicate that Vlk is involved in the differentiation of hypertrophic chondrocytes rather than in the proliferation or survival of chondrocytes in general.

The above results strongly suggest that the morphogenetic defects of Vlk^-/- mutants are due to a failure to form a proper environment for organogenesis, rather than to a cell-autonomous failure. As Vlk is localized at the Golgi apparatus in Vlk-transduced cells(Fig. 2A-E) and is involved in stromal function of mesenchymal cells, it is plausible that Vlk is involved in the protein secretion pathway. To test this possibility, we measured vesicular transport using temperature-sensitive vesicular stomatitis virus G (VSVG)protein (ts045) fused to GFP (Hirschberg et al., 1998). In this experimental system, VSVG-GFP is first accumulated in the ER when cultured at 40°C and molecules are released from ER to plasma membrane (PM) through the Golgi apparatus once the temperature is shifted to 32°C. As NIH3T3 is a mesenchymal cell line that does not express Vlk, we chose it for assessing the effect of Vlk on vesicular transport. We introduced VSVG-GFP together with WT Vlk, kinase-dead K166M-Vlk, and ΔN-Vlk (which lacks a 30 amino acid sequence at the N-terminus, leading to Vlk mislocalization in the cytoplasm and a lack of phosphorylation; data not shown) constructs into NIH3T3 cells. When co-transfected with the control or ΔN-Vlk construct, VSVG-GFP that localized at the ER at 0 minutes moved to the Golgi apparatus 30 minutes after shifting the temperature to 32°C (Fig. 6A,D). Within 3 hours after the temperature shift, VSVG-GFP was detectable on the PM (Fig. 6A,D). In the cells in which VSVG-GFP was co-transduced with WT or K166M-Vlk, translocation of VSVG-GFP from ER to Golgi occurred normally, but its transport from Golgi to PM appeared to be retarded, as VSVG-GFP accumulated in the Golgi apparatus (Fig. 6B,C). This block from the Golgi to the PM was confirmed by flow cytometry of surface-stained cells 3 hours after the temperature shift. In control- and ΔN-Vlk-transfected cells, the GFP and extracellular VSVG signals lie on a diagonal line (Fig. 6E,H). However, the amount of surface protein relative to the total GFP signal was reduced significantly in WT and K166M-Vlk co-transfected cells (Fig. 6F,G). To clarify the causal role of Vlk in protein transport, we introduced VSVG-GFP and Vlk-mKO [monomeric Kusabira-Orange(Karasawa et al., 2004)]constructs to enable monitoring of both proteins simultaneously. Fig. 6I clearly shows that the amount of VSVG transported to the PM was inversely correlated with the amount of Vlk-mKO. This result indicates that Vlk inhibits protein transport in a dose-dependent fashion.

In this study, we first produced a list of PKs that are not expressed in ES cells but are newly induced in the differentiated Gsc⁺ mesendoderm. As many PKs occupy important positions in signal transduction pathways and a classification of all PKs has been completed in both human(Manning et al., 2002) and mouse (Caenepeel et al., 2004),we expected that this line of analysis might provide useful information on the specific signaling network that functions in early intermediate stages such as in the mesendoderm. Contrary to our expectations, excluding all PKs that are detectable in ES cells but that undergo quantitative changes during differentiation turned out not to be useful in revealing novel signals regulating mesendoderm differentiation. With respect to receptor-type PKs, we could identify only Flt1, Met, Ror1 and Ror2 that were known to be dispensable for mesendoderm differentiation. Of interest is that a group of proteins in the MAP kinase signaling pathways are newly expressed upon ES cell differentiation. Some of them, such as Map3k7, Map2k4 and Jnk2 (Mapk9), are implicated in the TGFβ and interleukin signaling pathways(Ninomiya-Tsuji et al., 1999; Su et al., 2006) and might play a role in extrinsic signaling. Alternatively, their expression is a reflection of environmental stress in the defined culture condition, as these proteins are involved in cellular stress responses.

Nearly all PKs in this list have been characterized to some extent and can be classified into one of the known PK families, but we also found one PK that was completely uncharacterized in mice. Judging from the sequence, this mouse PK is an ortholog of human SGK493. According to a previous classification of all PKs (Manning et al.,2002) and the results of our own analyses, this PK showed two intriguing features. First, no homolog was detected in invertebrates by BLAST search, whereas orthologs were found in all vertebrates examined. Second, this protein could not be classified into any of the known PK families and constitutes a solitary branch in the kinome with this PK as a sole member. Because of these two features, we designated this PK vertebrate lonesome kinase (Vlk). The human kinome includes six more PKs that constitute such solitary branches stemming directly from the origin; four of these have already been characterized to some extent: p53-related protein kinase(TP53RK) (Abe et al.,2001), haspin (GSG2)(Tanaka et al., 1999) (all species, including yeast), Slowpoke channel binding protein(Slob) (Schopperle et al.,1998) (Drosophila) and phosphoinositide-3-kinase regulatory subunit 4 (PIK3R4; VPS15)(Herman et al., 1991). Homologs for all of these exist in both vertebrates and invertebrates. By contrast, to our knowledge there have been no previous reports regarding three other genes,including Vlk. Interestingly, our BLAST search of SGK196 and SGK396 similarly failed to find any homologs in invertebrates, as was the case for Vlk. The most homologous genes identified in this search outside of the vertebrate orthologs were from plant genomes for all three PKs. Thus, it is likely that these three PKs, including Vlk, might have newly evolved in vertebrates, although we cannot exclude the possibility that genome coverage in invertebrates is incomplete. Thus, all three PKs can be designated as vertebrate lonesome kinases. Further studies on the functions of the three Vlks might provide some insight into vertebrate evolution.

The Vlk gene product has at least seven subdomains characteristic of PKs, and it is likely that Vlk is a kinase. Subdomains V, IX, X and XI were not identified, but the presence of these domains varies among kinases. Subdomain VIII plays a major role in the recognition of peptide substrates and contains a well-conserved Ala-Pro-Glu (APE) triplet, but there is no APE sequence in kinase subdomain VIII of Vlk. This APE consensus is known to be required for the autophosphorylation activity of v-Src(Bryant and Parsons, 1984). This is consistent with our finding that Vlk does not have an autophosphorylation activity. Despite our failure to obtain direct biochemical evidence of the kinase activity of Vlk, we still surmise that Vlk is a PK for the following indirect reasons. Vlk transfected into HEK293T cells is phosphorylated at tyrosine 148 and serine 177. This tyrosine phosphorylation of Vlk requires the presence of the intact kinase domain of Vlk itself, as mutation of the ATP-binding site resulted in complete loss of phosphorylation. As the mutated Vlk is tyrosine phosphorylated when co-transfected with WT Vlk into the same cell, it would appear to be a substrate of a Vlk-dependent phosphorylation cascade. Moreover, in cells co-transfected with the two forms of Vlk, the WT and kinase-dead Vlk were co-immunoprecipitated. One parsimonious explanation for these observations is that Vlk forms a multimer in the cells, with members phosphorylating each other.

We showed that Vlk is enriched preferentially in the Golgi apparatus and that Vlk overexpression inhibits VSVG transport. This phenotype is dependent on the Golgi localization of Vlk but not on its phosphorylating activity. Hence, Vlk can be classified as a member of the Golgi membrane proteins that play a role in protein transport from Golgi to PM. However, the functional significance of the kinase activity of Vlk in the Golgi remains for future study. Various proteins, including caseins(Duncan et al., 2000; Turner et al., 1993; West and Clegg, 1984),proteoglycan (Glossl et al.,1986) and osteopontin (Lasa et al., 1997), are known to be phosphorylated in the Golgi. Hence, it is possible that Vlk is involved in the phosphorylation of such proteins in the Golgi.

It should be emphasized that Vlk expression is regulated by a developmental program and that mesenchymal cell lines, such as NIH3T3 or mouse embryonic fibroblast, do not express Vlk. During embryogenesis, the Vlk expression pattern is strictly regulated in a spatiotemporal manner, but the expression per se is widespread in all germ layers, such as in the AVE, ventral part of the midbrain and mesenchymal cells. Vlkappears to be specifically expressed in areas that are actively segregating from the surroundings, such as the Gsc⁺ organizer, the AVE,condensed mesenchyme, hypertrophic chondrocytes and mesothelium. The phenotype of Vlk^-/- mice indicates that many of these processes progress normally in such embryos. Even in areas where functional involvement of Vlk is evident, the defect is in most cases incomplete. For example, in Vlk^-/- embryos endochondral ossification is delayed and the body size is small, but all processes required for chondrogenesis and subsequent osteogenesis are completed. Similarly, the elevation of the palatal shelf failed in most cases, although some proportion of palatal shelves underwent normal elevation. Finally, although the Vlk^-/-mice died owing to respiratory failure, most offspring survived for up to a few hours, suggesting the presence of partially functioning lung structures. These results suggest that Vlk has a definite functional role in some, but not all, developmental processes with which its expression is associated, but its role in each process is as a modulator rather than as a master regulator or essential component. In all processes analyzed in the present study, lack of Vlk had no effect on cell proliferation or apoptosis of normally Vlk⁺ cells. By contrast, the effect of the Vlk-null mutation was found in cells that do not normally express Vlk. In lung morphogenesis, the cells affected are Vlk^- alveolar cells, and in palate elevation Vlk appears to be involved in generating morphogenetic force. In both tissues, cells that are normally Vlk⁺are generated normally in the absence of Vlk. Thus, it is likely that Vlk is required for the functional activity of Vlk⁺ cells to support the development of cells in their vicinity. By contrast, in endochondral ossification, defects were detected in the Vlk⁺ cells themselves. Hence, Vlk can also be involved in the regulation of cell differentiation, or, alternatively, the delay of differentiation can be ascribed to defects in the surrounding environment to which differentiating chondrocytes themselves are likely to contribute.

Our cell biological analysis showing that expression of Vlk in the normally Vlk^- NIH3T3 cells suppresses VSVG-GFP transport is also consistent with the notion that Vlk is required for the formation of a proper environment for organogenesis. Of particular interest in this context is the striking similarity of Vlk^-/- mice to those null for Fgf18 or aggrecan, which encode proteins that are secreted into the extracellular space. Fgf18^-/- mice die shortly after birth owing to a failure of alveolar development and show a delay in bone formation and a cleft palate (Liu et al.,2002; Ohbayashi et al.,2002). Interestingly, the cleft palate in this strain has been suggested to involve the failure of palate elevation. Cartilage matrix deficiency (cmd), a mouse strain with a spontaneous null mutation of the aggrecan gene, shows defects in bone formation, a cleft palate and respiratory failure (Watanabe et al.,1994). Obviously, there are some differences among these strains,but these defects common to all three mutant mouse strains strongly suggest that the formation of a proper environment expressing Fgf18 and aggrecan is essential for these developmental processes. Our present results thus suggest that Vlk is also involved in the formation of a functional extracellular matrix. Given that Vlk non-specifically retards protein transport from the Golgi, we would speculate that Vlk-mediated retention of proteins plays roles in the proper processing and assembly of proteins in the Golgi, thereby quantitatively increasing the amount of functional matrix proteins.