In Vitro Development of Mouse Embryonic Stem Cells Lacking JNK/Stress-activated Protein Kinase-associated Protein 1 (JSAP1) Scaffold Protein Revealed Its Requirement during Early Embryonic Neurogenesis*

The Jsap1 gene encodes a scaffold protein for c-Jun N-terminal kinase cascades. We established c-Jun N-terminal kinase (JNK)/stress-activated protein kinase-associated protein 1 (JSAP1)-null mouse embryonic stem cell lines by homologous recombination. The JSAP1-null embryonic stem cells were viable, however, exhibited hyperplasia of the ectoderm during embryoid body formation, and spontaneously differentiated into neurons more efficiently than did wild type. The expression of components of c-Jun N-terminal kinase cascades and a subset of marker mRNAs during early embryogenesis was altered in the JSAP1-null mutants. Retinoic acid dramatically increased the expression of JSAP1 and JNK3, which were co-precipitated with anti-JNK3 in the neuroectoderm of wild type but not JSAP1-null embryoid bodies. In the neurons differentiated from the wild type embryoid bodies, JSAP1 was localized in the soma, neurites, and growth cone-like structure of the neurites, and neurite outgrowth from the JSAP1-null embryoid bodies was apparently less efficient than from wild type. JSAP1 and c-Jun N-terminal kinase 3 were coexpressed in the embryonic ectoderm of E7.5 mouse embryo, whereas Wnt1 and Pax2 were coexpressed with JSAP1 at the midbrain-hindbrain junction in E12.5 mouse embryo, thus suggesting that JSAP1 is required for early embryonic neurogenesis.

The JNK pathway of MAPKs is primarily activated in response to extracellular stimuli such as cytokines, heat shock, or radiation and mediates the regulations of cell proliferation, apoptosis, tumorigenesis, and embryonic morphogenesis. Three genes encoding JNK were identified in mammals; the Jnk1 and Jnk2 genes are ubiquitously expressed, and the Jnk3 gene is specifically expressed in the brain, heart, and testis in adults. Mice lacking only one of three genes or Jnk1 and Jnk3 or Jnk2 and Jnk3 develop normally and are viable; however, the double-knockout mice of Jnk1 and Jnk2 are lethal during early embryogenesis at mid-gestation with a defect of neural tube closure (5), thus demonstrating a functional redundancy of the JNKs during embryogenesis. In both Drosophila and Xenopus, the JNK signal pathway has also been reported to mediate embryonic morphogenesis through regulating the epithelial movement and planar cell polarity during gastrulation (6,7).
It has been shown that the MEKK1 and MLKs, as MAPK kinase kinases, selectively activate the JNK cascade (8) through the phosphorylation of MAPK kinases such as MAPK kinase-4 (MKK4, also known as SEK1) and MKK7, which specifically phosphorylate JNKs. Once activated, the JNKs phosphorylate several transcription factors such as c-Jun, Jun-D, ATF2, and Elk1, which in turn regulate the transcription activity of many target genes. As well as JNK1/2-deficient mice, mice deficient in MKK4 or c-Jun are lethal at mid-gestation with a severe abnormality of liver formation (9,10). MKK7 is also required for embryonic viability (11). A disruption of MEKK1 gene results in defective eyelid closure (12). It has, thus, been established that the JNK-signaling pathways play a critical role in embryogenesis.
In addition to MEKK1 and MLKs, other MAPK kinase kinases such as MEKK4, TAK1, and ASK1 can also activate the JNK and p38 cascades through activation of MKK4 and MKK7 for JNKs and MKK3 and MKK6 for p38s, whereas MEKK3 and TPL-2 can activate the extracellular signal-regulated kinase, p38, and the fourth extracellular signal-regulated kinase 5 cascades as well as the JNK cascade. These MAPK kinase kinases are, thus, highly promiscuous for the selection of MAPK kinase (3,4). To maintain the signaling specificity, efficiency, and integrity among these MAPK cascades, protein kinase components of each signaling pathway have to be tightly organized both spatially and temporally.
Studies of yeast have established that the protein kinase components of the mating MAPK pathway interact with the scaffold protein Ste5p and that this interaction is essential for the formation of a functional signaling module, thus suggesting that such a scaffold protein may be essential for any MAPK pathway (13). Indeed, recent studies of the JNK signal transduction pathway also have led to the identification of potential scaffold proteins such as JNK-interacting protein 1 and 2, which are also known as islet brain 1 and 2, and JSAP1 (JNK/ stress-activated protein kinase-associated protein 1, also known as JNK-interacting protein 3) (14 -16). We and others have shown that JSAP1 may function as a scaffold protein for JNK cascades, in which the components were reconstituted by their overexpression in cultured cells (15)(16)(17).
JSAP1 is one of the known clustered genes in the t complex of the mouse chromosome 17 that exhibit random monoallelic expression (18). The t complex contains several critical loci affecting embryonic development, male fertility, and male transmission ratio distortion (19), such as Brachyury, an essential gene for the mesoderm differentiation of vertebrates (20,21). It is, thus, likely that JSAP1 may play an important role during early embryogenesis. In the present study, we established JSAP1-null embryonic stem cell lines by homologous recombination to elucidate their in vitro differentiation capacity and obtained evidence suggesting that JSAP1 plays an important role in early embryogenesis, especially in neurogenesis.
Generation of Jsap1-deficient Embryonic Stem Cell Clones-Genomic fragments of 2.8 and 5.45 kilobases flanking the exon 1 were used to generate a targeting construct in which a 0.5-kilobase region of the exon 1, including the initiation codon and 0.4-kilobase region of the intron 1, were replaced by a neomycin-resistant cassette, pol II-neo-poly(A) cassette. To increase the frequency of the gene targeting, a pair of the herpes simplex virus-1 and -2 thymidine kinase cassettes were placed flanking the Jsap1 genomic sequence in the targeting vector (23). CCE embryonic stem (ES) cells were electroporated with the SalI-linearized targeting vector as described (24). Colonies doubly resistant to G418 (250 g/ml) (Sigma-Aldrich) and ganciclovir (5 M) (Japan Syntechs, Tokyo, Japan) were selected and homologous recombinants (Jsap1 ϩ/Ϫ ) were identified by a Southern blot analysis. The targeted allele of Jsap1 is designated as Jsap1 tm1Yun or Jsap1 Ϫ . Subsequently, homozygous Jsap1-deficient clones (Jsap1 Ϫ/Ϫ ) were obtained by selection in the presence of higher concentration of G418 (1.5 and 2.0 mg/ml).
In Vitro Differentiation of ES Cells-ES cells were subjected to an 8-day induction procedure that consisted of 4 days of culture as aggregates without retinoic acid (RA) (4 Ϫ ) followed by 4 days of culture with (4 Ϫ /4 ϩ ) or without (4 Ϫ /4 Ϫ ) RA according to the protocols described by Bain et al. (25,26). After the 8-day induction period, aggregates were transferred to gelatin-coated tissue culture wells to provide a substrate for cell attachment. About 100 aggregates were seeded into a 35-mm well with 2 ml of standard medium lacking leukemia inhibitory factor and ␤-mercaptoethanol. The culturing of these aggregates was continued as indicated in the legend for Fig. 2A.
Laser-scanning Fluorescence Microscopy-The cells were plated in a two-well gelatin-coated chamber slide (Nalge, Nunc) and cultured as described in the legend for Fig. 2A. The cells were washed in PBS and fixed with 4% paraformaldehyde for 60 min at room temperature. The slides were washed in PBS and incubated in 95% methanol containing 2.5% acetic acid at Ϫ30°for 5 min. Next, the slides were washed in PBS and incubated in 1% skimmed milk in PBS for 1 h at room temperature. Thereafter, the slides were incubated with primary antibodies overnight at 4°C and incubated with proper second antibodies conjugated with Alexa Fluor 488 or 594 for 1 h at 37°C. Washed slides were mounted and observed under Eclipse TE300 (Nikon Co.) equipped with a Radiance 2100 laser-scanning fluorescence microscope System (Bio-Rad). Digitized images were obtained and processed for publication using Adobe Photoshop 5.5J.
Immunohistochemistry-Embryoid bodies (EBs) differentiated from ES cells were washed by PBS and fixed with 4% paraformaldehyde at 4°C for overnight. After washing in PBS, the EBs were serially dehydrated in 25, 50, and 75% ethanol in PBS for 30 min, respectively, and 100% ethanol for each 1 h at room temperature. Next, the EBs were incubated in xylene for 1 h at room temperature, 3 times, and in paraffin each for 1 h at 65°C three times. Finally, the EBs were embedded in paraffin. Mouse embryos were also prepared as described above. Immunohistochemistry of EBs and embryos for JSAP1, JNK3, ␤-tubulin III was performed on 4-m paraffin-embedded sections by the indirect immunoperoxidase method (28).
Terminal dUTP Nick-end Labeling Assay-To detect the DNA strand breaks in the apoptotic cells, 4-m paraffin-embedded sections of EBs were prepared as described above and subjected to a terminal dUTP nick-end labeling reaction according to the recommended protocol (In Situ Cell Death Detection Kit, peroxidase; Roche Applied Science; catalog number 1684817).
Western Analysis-Protein samples were separated by SDS-PAGE and transferred onto Immobilon-P membrane (Millipore), and Western blotting analyses were performed as previously described (22).
RT-PCR-For the RT-PCR analyses, total RNA was prepared from ES cells or embryoid bodies using ISOGEN (Nippon Gene Co. Ltd., Toyama, Japan). First strand cDNA was synthesized by a first strand cDNA synthesis kit (Amersham Biosciences) using random primers. The primers used to amplify the specific cDNAs are as shown in Table I.
Statistical Analysis-The data are expressed as the mean Ϯ S.D. All data were compared using an unpaired t test for data with gaussian distribution, the Welch test for data with non-gaussian distribution. Statistical significance between wild type and JSAP1-null mutant was accepted at a level of p Ͻ 0.0001.

Isolation and Characterization of JSAP1-null ES Cell Line-
To obtain experimental material for investigating JSAP1 function during the early embryogenesis, targeted disruption of the Jsap1 gene by homologous recombination in mouse ES cells was achieved by using a targeting vector designed to replace a region including the initiation codon in the exon 1 and a part of intron 1 with pol II-neo cassette (Fig. 1A). Three of 53 G418/ gancyclovir-resistant ES clones were identified as correctly targeted clones by Southern blot analyses (Fig. 1B).
To examine whether or not JSAP1 is essential for the viability of ES cells, Jsap1 ϩ/Ϫ ES cells (clones 16 and 50) were exposed to a high concentration of G418, and more than 70 resistant colonies were isolated. About 30 and 50% of them were found to be homozygous (Jsap1 Ϫ/Ϫ ) (Fig. 1B). There was neither any detectable JSAP1 mRNA nor JSAP1 protein in the homozygous mutants, whereas the levels of JSAP1 mRNA and protein in the heterozygotes were ϳ50% that of the levels observed in wild type ES cells (Fig. 1, C and D).
Two independent clones of JSAP1-null (Jsap1 Ϫ/Ϫ ) ES cells grew faster than did either wild type or parental Jsap1 ϩ/Ϫ ES cells (Fig. 1E). The homozygous mutants were morphologically indistinguishable from the wild type and Jsap1 ϩ/Ϫ ES cells and expressed a normal level of cell surface stage-specific embryonic antigen-1, which is a phenotypic marker of undifferentiated ES cells (Fig. 1F) (29). Immunofluorescent microscopy with anti-JSAP1 revealed that all of wild type ES cells examined expressed a substantial level of JSAP1; however, about 15% of the Jsap1 ϩ/Ϫ ES cells expressed no detectable JSAP1 (Fig. 1G), thus supporting the random monoallelic expression of Jsap1 gene (18). Again, there was no detectable JSAP1 in the JSAP1-null ES cells.
Poor Embryoid Body Formation and Altered Neurite Outgrowth from the JSAP1-null Mutant-To explore whether or not JSAP1-null ES cells are still totipotent, we examined their differentiated phenotypes in vitro in comparison with those of wild type ES cells using an in vitro differentiation protocol (25, 30) ( Fig. 2A). JSAP1-null (Jsap1 Ϫ/Ϫ ) ES cells formed smaller EBs than wild type in the absence of RA (4 Ϫ /4 Ϫ ), and the treatment of JSAP1-null EBs with RA (4 Ϫ /4 ϩ ) resulted in the formation of much smaller EBs, determined by measuring the area of a cross-section for each EB (Fig. 2, B and C). The capacities of neural differentiation of each EB were also examined. From wild type EBs numerous neurites elongated especially when EBs were exposed to RA for 4 days (4 Ϫ /4 ϩ ), whereas significantly poor neurite elongation or synapse formation was seen in the JSAP1-null EBs even with RA treatment. In contrast, JSAP1-null EBs tend to elongate longer neurites more efficiently than wild type in the absence of RA (4 Ϫ /4 Ϫ ) (Fig. 2, D and E).
Altered Expression of Differentiation Markers during the in Vitro Differentiation of JSAP1-null Mutant-Because the JSAP1-null EBs underwent abnormal in vitro differentiation, we next investigated the temporal expression pattern of various differentiation markers by semi-quantitative RT-PCR (Fig.   3A). The expression patterns of Gata4 and HNF1, which are essential for visceral endoderm differentiation (31,32), were essentially similar during the differentiation of wild type and JSAP1-null ES cells with or without RA. However, the expression of transferrin mRNA, a target for Hnf3␥ involved in the differentiation of definitive endoderm (33), was consistently higher in the late stage of JSAP1-null EBs than in wild type regardless of RA treatment.
The expression of Brachyury, a marker for the mesoderm (20), was significantly altered in JSAP1-null EBs. The expression of Brachyury was highly induced in the wild type EBs (4 Ϫ ) and was diminished by RA treatment (4 Ϫ /4 ϩ ). In contrast, in the early JSAP1-null EBs (4 Ϫ ), there was no detectable Brachyury mRNA, but the level tended to increased later with or without RA treatment (4 Ϫ /4 ϩ , 4 Ϫ /4 Ϫ ). The expression of mRNAs for BMP4 and Nodal, which are members of transforming growth factor-␤ superfamily and involved in mesoderm induction and gastrulation (34), were not apparently altered throughout the in vitro differentiation both in the wild type and JSAP1-null mutants.
Keratin 17 is a marker for epidermal cells (35), and its expression level was higher in the JSAP1-null mutants than wild type in all stages examined. The expression of both Emx2, which is essential for forebrain development (36), and Otx1, whose expression begins at the 1-3 somite stage in the anterior neuroectoderm of mouse embryo (37), was low in both undifferentiated wild type and JSAP1-null ES cells. Closely similar levels of Emx2 and Otx1 mRNA were detected in the early EBs (4 Ϫ ), which were formed from the two ES cell lines; however, the levels only highly increased in the late JSAP1-null EBs (4 Ϫ/Ϫ ) in the absence of RA. RA treatment, in contrast, increased the expression levels of Emx2 and Otx1 only in the wild type EBs. The expression of En2, Wnt1, Pax2, and Pax5, which are known to be involved in the regulation of mesencephalon and metencephalon development (38), was further examined. In the undifferentiated JSAP1-null but not wild type ES cells, only high level expression of Pax2 was detected among the four genes. In the early EBs (4 Ϫ ), the expression of Pax5 was increased in wild type EBs but not JSAP1-null EBs. During the formation of the late EBs in the absence of RA, the expression levels of Wnt1 and Pax5 were highly increased in the JSAP1null EBs (4 Ϫ/Ϫ ), and RA treatment increased the expression levels of the four genes in the wild type but not in the JSAP1null EBs. High levels of Wnt1 protein in the late EBs (4 Ϫ /4 ϩ ), which were formed from wild type ES cells in the presence of RA, or in the late EBs, which were formed from the JSAP1-null mutants in the absence of RA, were confirmed by Western TGCAGCTTCTGAGAGAACAAGG GTAGTGTTGGGAAGCACTGAGG blotting (Fig. 3B, lanes 3 and 5), whereas the expression of Smad4 protein, which mediates BMP signals into the nucleus (39), was not apparently altered by a loss of JSAP1. Altered Expression of Components of JNK Cascades during the in Vitro Differentiation of the JSAP1-null Mutant-A substantial level of JSAP1 protein (180-kDa, p180) was detected in undifferentiated wild type ES cells by Western blotting (Fig.  3B, lane 1). During EB formation, the expression level of JSAP1 apparently decreased in the early (4 Ϫ ) and late EBs (4 Ϫ /4 Ϫ ) or differentiated cells formed in the absence of RA (Fig.  3B, lanes 2, 5, and 6); however, RA treatment significantly increased the level of JSAP1 accompanied with two bands (p190, p150) in the wild type EBs (4 Ϫ /4 ϩ ); thereafter, their levels were reduced in neurons (Fig. 3B, lanes 3 and 4). No appropriate band was seen in the JSAP1-null mutants of all stages examined (data not shown).
Because in vitro experiments indicated that JSAP1 acts as a scaffold protein for JNK cascades (15,16), the expression levels of various components for JNK cascades were examined by Western blotting (Fig. 3). The expression level of MEKK1was high in undifferentiated JSAP1-null ES cells as well as in wild type ES cells (Fig. 3B, lane 1); however, the expression level of MEKK1 in late EBs (4 Ϫ /4 Ϫ ) formed from JSAP-null mutants in the absence of RA was much less than that in wild type EBs (Fig. 3B, lane 5). On the other hand, the expression level of MLK3 was almost constant throughout all stages examined both in wild type and JSAP1-null mutants.
The levels of MKK4 and MKK7 increased during EB formation from wild type but not JSAP1-null ES cells in the presence of RA (Fig. 3B, lanes 3 and 4). In contrast, higher levels of MKK4 and MKK7 were detected in the late EBs formed from JSAP1-null mutants in the absence of RA than in those from wild type (Fig. 3B, lanes 5).
A substantial level of JNK1 and/or JNK2 (JNK1/2) was expressed in the undifferentiated wild type ES cells and neurons differentiated from the RA-treated EBs. JNK1/2 were barely detected in the undifferentiated JSAP1-null ES cells, whereas the highest level of JNK1/2 was detected in neurons differentiated from JSAP1-null EBs (Fig. 3B, lanes 1 and 4). In contrast, expression of JNK3 in wild type EBs is highly induced after RA treatment but not in JSAP1-null EBs (4 Ϫ /4 ϩ ) (Fig. 3B, lanes 3).
In an immunocomplex precipitated from the extracts of wild type EBs (4 Ϫ /4 ϩ ) with anti-JNK3 but not normal rabbit IgG, the larger band, 190-kDa JASP1 polypeptide, was specifically detected (Fig. 3C), thus demonstrating that the authentic  JNK3 and JSAP1 co-exist in a complex in vivo. We further attempted to detect JSAP1 in the immunocomplexes precipitated with anti-MEKK1, MLK3, MKK4, and MKK7 but failed to detect it (data not shown).
c-Jun and ATF2 were preferentially expressed in undifferentiated wild type ES cells, and the levels were significantly reduced in undifferentiated JSAP1-null ES cells (Fig. 3D, lanes  1). The expression level of Elk1 was not significantly altered in JSAP1-null ES cells and EBs in comparison to wild type. The phosphorylation status of JNK and c-Jun was examined with anti-phospho-JNK (Thr-183 and Tyr-185) and anti-phospho-c-Jun (Ser-63) antibodies. Substantial levels of phosphorylated JNKs were detected in the early EBs (4 Ϫ ), which were formed from wild type but not JSAP1-null mutants, and also in the differentiated cells from RA-treated JSAP1-null EBs as well as wild type EBs (Fig. 3B, lanes 2 and 4). Much lower levels of phosphorylated c-Jun were detected in the JSAP1-null ES cells or their early EBs than in wild type. In contrast, an increased level of phosphorylated c-Jun was detected in the JSAP1-null EBs (4 Ϫ /4 Ϫ ), which were formed in the absence of RA, in comparison with the level seen in wild type EBs (Fig.  3B, lanes 5).
As shown in Fig. 3D, Jsap1, c-jun, Elk1, and Atf2 but not Jnk3 mRNAs were highly expressed in the undifferentiated wild type ES cells, and those mRNAs other than Jsap1 were also expressed in the JSAP1-null mutant. In the wild type a higher level of Jsap1 mRNA was expressed in the late EBs (4 Ϫ /4 ϩ ) and neurons differentiated from them as well as in undifferentiated ES cells (Fig. 3D, lanes 1, 3, and 4), in comparison with EBs and differentiated cells in the absence of RA (Fig. 3D, lanes 2, 5, and 6). The level of Jnk3 mRNA gradually increased throughout in vitro differentiation especially with RA treatment, both in wild type and JSAP1-null mutant. In contrast, the c-jun mRNA level apparently decreased in the early EBs (4 Ϫ ) (Fig. 3D, lane 2) and then later recovered in both the wild type and JSAP1-null mutants. mRNAs for Elk1 and Atf2 were constantly expressed throughout the experiments in both the wild type and JSAP1-null mutants. Levels of Jnk3, c-jun, Elk1, and Atf2 mRNAs were not altered by loss of JSAP1.
Abnormal Ectoderm Formation in the JSAP1-null EBs-To examine the morphological alteration in EBs formed from wild type and JSAP1-null ES cells, sections prepared from EBs embedded in paraffin were processed for hematoxylin-eosin staining and indirect immunohistochemistry for JSAP1, JNK3, ␤-tubulin III, and Wnt1, as shown in Fig. 4. Seventy-six percent of JSAP1-null EBs (76/100 EBs) formed in the absence of RA (4 Ϫ /4 Ϫ ) exhibited hyperplastic proliferation of ectodermal epithelium, thus forming a multitubular structure of varying sizes, whereas less than 5% of wild type EBs exhibited such a multitubular structure (Fig. 4, A1 and B1). JSAP1 mainly expressed in the ectoderm and outer layer of endoderm of wild type but not JSAP1-null EBs (Fig. 4, A2 and B2). JNK3 expression in wild type EBs was spatially similar to that of JSAP1, and a much lower expression of JNK3 was detected in the hyperplastic ectodermal layers of JSAP1-null EBs (Fig. 4, A3  and B3). A much higher level of ␤-tubulin III, a marker for immature neurons (25), was detected in the hyperplastic ectodermal layers of JSAP1-null EBs than in the ectoderm of wild type (Fig. 4, A4 and B4). Similarly, a part of hyperplastic ectodermal layers of JSAP1-null EBs also expressed a much higher level of Wnt1 than in the wild type EBs (Fig. 4, A5 and B5).
JSAP1-null EBs formed in the presence of RA (4 Ϫ /4 ϩ ) were apparently smaller in size with a smaller cavity than wild type EBs (Fig. 4, C1 and D1). JNK3 as well as JSAP1 was highly expressed in the outer layer of endoderm and the ectodermal layer of the wild type (Fig. 4, C2 and C3), whereas JNK3 was expressed only in the outer layer of the endoderm of JSAP1null EBs (Fig. 4, D3). Most ectodermal and endodermal cells in the wild type EBs exhibited a high level expression of ␤-tubulin III as well as Wnt1 (Fig. 4, C4 and C5), and both expressions were barely detectable in JSAP1-null EBs (Fig. 4, D4 and D5). These results correlated with the results from an RT-PCR analysis shown in Fig. 3A.
Expression of JSAP1 in the E7.5 and E12.5 Developing Mouse Embryos-The expression of JSAP1 in the developing mouse embryo was examined to evaluate its involvement in mouse embryogenesis. As shown in Fig. 5A, JSAP1 and JNK3 were simultaneously expressed in embryonic and extra-embryonic ectoderm and also in the extra-embryonic endoderm in E7.5 developing mouse embryos, whereas a high level of Wnt1 expression was seen only in the embryonic ectoderm. Lower but substantial levels of JSAP1, JNK3, and Wnt1 were also detected in the embryonic mesoderm. ␤-Tubulin III, JNK3, and Wnt1 were likely to be coexpressed with JSAP1, especially in the embryonic ectoderm, namely the neuroectoderm, from where developmental neurogenesis occurs (40,41) (Fig. 5B). A Western blotting analysis of the whole embryo also demonstrated a high level expression of JSAP1 (p180) and JNK3 in the embryos (E6.5 and E7.5) (Fig. 5C).
A high level of JSAP1 expression was further detected at the midbrain-hindbrain junction in E12.5 embryos, and where Pax2 and Wnt1, markers for definition of the midbrain-hindbrain junction, were coexpressed (42). The expression of JSAP1 was also detected in the ependymal layer of the forebrain and spinal cord.
Expression of JSAP1 in the Neurons Differentiated from Wild Type EBs and Altered Neuronal Differentiation from JSAP1-null EBs-In the wild type neurons differentiated from EBs formed in the presence of RA (4 Ϫ /4 ϩ ) JSAP1 was localized in the neurites, growth cones, and to some extent in the soma of neurons (Fig. 6, A and B; Fig. 7, A-C). GFAPpositive glia-like cells which express JSAP1 (Fig. 6C) were also efficiently differentiated from wild type EBs but much less efficiently from JSAP1-null EBs (Fig. 6D). About 10 -20% of neurofilament-positive neurites were tyrosine hydroxylase-positive in the neurons differentiated from both wild type and JSAP1-null EBs treated with RA, although neurofilament-positive neurites were poorly elongated from JSAP1-null EBs in comparison with wild type (Fig. 6E). On the other hand JSAP1-null EBs, which were formed in the absence of RA (4 Ϫ /4 Ϫ ), elongated neurites expressing neurofilament, GAP43, and tyrosine hydroxylase or differentiated into GFAP-positive glia-like cells more efficiently than did wild type EBs (Fig. 6, F and G).
In JSAP1-null Neurons, Synaptophysin Accumulated in the Soma-In the wild type neurons differentiated from RAtreated EBs, synaptophysin as well as JSAP1 was detected evenly from the soma to the neurite, thus indicating that the synaptic vesicles were functionally transported from the soma to the terminal of each neurite (Fig. 7, A-D). On the other hand, in JSAP1-null mutant neurons, a much stronger signal for synaptophysin was detected in the soma than in neurites (Fig.  7, E and F). To elucidate any quantitative differences in the synaptophysin distribution between the wild type and mutant neurons, fluorescence intensity was digitized using NIH image software (Fig. 7G). The accumulation of synaptophysin in the soma of the JSAP1-null mutant neurons was more than three times greater than that observed in the soma of wild type neurons.

DISCUSSION
In the present study we obtained evidence that JSAP1-null ES cells exhibit large alterations in levels and/or functions of multiple components of JNK-signaling cascades, resulting in substantial alterations in the process of differentiation of cells within EBs; however, our results should be replicated with mouse embryos generated from the JSAP1-null ES cells. The JSAP1 mutation appears to have strong effects on the size of EBs, suggesting the possibility that multiple changes in ES cell differentiation observed in the JSAP1-null ES cells may result as an indirect effect of the altered EB size. Because we observed altered expression of various components of JNK cascades even in undifferentiated JSAP1-null ES cells and because an altered expression pattern of various differentiation markers during formation of early EB (4 Ϫ ) from the JSAP1-null ES cells was evident before the effects on the size of EB became apparent,  A and B) of RA were fixed and embedded in paraffin. Each five serial sections (4 m) was prepared and subjected to hematoxylin-eosin staining (A1, B1, C1, and D1) and indirect immunohistochemical staining for JSAP1 (A2, B2, C2, and D2), JNK3 (A3, B3, C3, and D3), ␤-tubulin III (A4, B4, C4, and D4) and Wnt1 (A5, B5, C5, and D5). The independent EBs (E and F) were subjected to a terminal dUTP nick-end labeling (TUNEL) assay. Scale bar: 35 m.
we concluded that most of the alterations observed in EBs derived from the JSAP1-null ES cells were not due to an indirect effect of the size, thus delineating the roles of JSAP1 during the early embryogenesis.
Expression of JSAP1 during in Vitro Differentiation of ES Cells and Early Embryogenesis-The Jsap1 mRNA level was substantially high in the undifferentiated ES cells and was slightly altered during EB formation but apparently decreased in the differentiated cells from EBs that were formed in the absence of RA, thus suggesting that the transcription of the Jsap1 gene may be partly regulated during cell differentiation. Furthermore, the JSAP1 protein levels were more dramatically altered during EB formation, thus suggesting that the stability of JSAP1 protein is also regulated during cell differentiation as well. In the RA-treated EBs and neurons differentiated from them, expression of two or three polypeptides (p190/p180 and p150) reactive with anti-JSAP1 were strongly induced, and only the p180 was detected in undifferentiated ES cells or EBs formed in the absence of RA. Multiple polypeptides of JSAP1 detected by Western blotting may represent variants encoded by alternatively spliced forms of JSAP1 transcripts (43) or those modified in a post-translational manner (15). Accordingly, we detected several splicing variants of Jsap1 mRNA during the in vitro differentiation. 2 It is noteworthy that only the p190 JSAP1 was co-precipitated with anti-JNK3 from RAtreated EBs, thus indicating that alternative splicing or posttranslational modification of JSAP1 may regulate its interaction with JNKs.
The expression of JSAP1 in the RA-treated EBs (4 Ϫ /4 ϩ ) was very high and mostly detected in the ␤-tubulin III-positive neuroectoderm and where JNK3 and Wnt1 were also coexpressed. In developing mouse embryos the neuroectoderm formation initiates during stages E7.0 -7.5 (40) and expression levels of JSAP1, JNK3, and Wnt1 were as high in the ␤-tubulin III-positive neuroectoderm of the E7.5 embryo as in the RAtreated EBs (4 Ϫ /4 ϩ ), thus indicating that the neuroectoderm formed in RA-treated EBs (4 Ϫ /4 ϩ ) may represent the neuroectoderm in E7.5 mouse embryos. Furthermore, a high level of JASP1 expression was detected at the midbrain-hindbrain junction of E12.5 embryo with coexpression of Pax2 and Wnt1, which are considered as important factors regulating isthmic organization of the midbrain-hindbrain junction (42,44), thus suggesting that JSAP1 may contribute to the formation of central nervous system as well as the early embryonic neurogenesis.
JSAP1 Negatively Regulates Growth of Undifferentiated ES Cells-JSAP1-null ES cells apparently grew faster than did parental CCE cells, probably reflecting the loss of either JNK1/ 2-c-Jun-or JNK1/2-ATF2-signaling cascades. It has been established that unphosphorylated c-Jun and ATF2 are subjected to proteolysis by proteasomes through ubiquitinylation (45,46). In our results the level of phosphorylated c-Jun decreased in JSAP1-null ES cells in undifferentiated ES cells. We hypothesize that a loss of functional JSAP1 results in the destabilization of JNKs, and their substrates, such as c-Jun and ATF2, remain unphosphorylated and then are degraded. It is, thus, likely that JSAP1 scaffold protein is required to maintain the stability of components for particular JNK signaling modules, at least in the undifferentiated ES cells. We could not rule out the possibility that the accelerated proliferation of the JSAP1null ES cells may partly be attributed to an extracellular signal-regulated kinase cascade, since JSAP1 has been shown to have a capacity to suppress the extracellular signal-regulated kinase-signaling pathway through its interaction with MEK1 and Raf-1 (47), and the expression level of Elk1, which is a main target of extracellular signal-regulated kinase 1/2, in undifferentiated JSAP1-null ES cells was apparently higher than that in the wild type ES cells. ture accompanied with an increased expression of neural specific genes, such as Emx2, Otx1, Pax5, and Wnt1, and resulted in a more efficient differentiation of neurons and glias in comparison with wild type EBs.
In vertebrates the patterning of neuroectoderm along the anteroposterior axis is initiated during gastrulation and mainly regulated by Wnts and fibroblast growth factor signal pathways (44,48). It has been established that an appropriate level of Wnt activity may specify posterior-to-anterior fates within the neural plate that arise from the neuroectoderm during the head formation (49,50). Wnt1 knockout mice have a severe phenotype of most loss of the midbrain and cerebellum (51). However, in mice lacking the Six3 gene, which is essential for forebrain development, the prosencephalon was severely truncated, and expression of Wnt1 was rostrally expanded, thus suggesting that the suppression of Wnt signal is essential for gradient-specific posterior fates in the anterior neural plate (52). These facts strongly indicate that an increased expression of Wnt1 in the JSAP1-null EBs may be the cause of their abnormal hyperplastic proliferation of the neuroectoderm.
Furthermore, a disruption of Pax5, Otx1, or Emx2 in mice also causes severe abnormalities during pattern formation in the central nervous system (36,37,53). Because a loss of JSAP1 also resulted in an up-regulation of Emx2, Otx1, and Pax5 as well as Wnt1, some of which may constitute positive feedback loop with fibroblast growth factor 8 (54), JSAP1 may provide an appropriate negative signaling for the loop.
In the JSAP1-null EBs (4 Ϫ /4 Ϫ ) the phosphorylation of c-Jun was very high, whereas the levels of JNK1/2, JNK3, and phosphorylated JNKs was very low. In contrast, the expression levels of MKK4, MKK7, and MLK3, but not MEKK1, were apparently higher than in wild type EBs (4 Ϫ /4 Ϫ ). These results indicate that loss of JSAP1 resulted in a down-regulation/ destabilization or up-regulation/stabilization of certain components of JNK or MAPK cascades, respectively. We and others show that JSAP1 may function as a scaffold protein for MEKK1-MKK4-JNKs, MLK3-MKK7-JNKs, and ASK1-MKK4/ 7-JNKs signaling modules (15)(16)(17), thus promoting signal transduction through the modules. Because we detected JSAP1 protein in a complex precipitated with only anti-JNK3 but not anti-MEKK1, MLK3, MKK4, and MKK7 in the wild type EBs, JSAP1 may temporally interact with these components and alter their stability or functions. The loss of JSAP1, thus, may disturb such a temporal regulation of components in the JNK cascades. In the JSAP1-null EBs (4 Ϫ /4 Ϫ ), the stabilization of MKK4 and MKK7 with MLK3 may activate JNKs, thus enhancing the phosphorylation of c-Jun. Because the levels of JNKs or phosphorylated JNKs were very low in the JSAP1-null EBs (4 Ϫ /4 Ϫ ), it is still not clear as to how such a high level of c-Jun phosphorylation was achieved at this time. Thus, loss of JSAP1 might lead to a destabilization and/or inactivation of a phosphatase(s), which may result in a high level of c-Jun phosphorylation. We also cannot rule out the possibility that the high level of c-Jun phosphorylation occurs independently from the JNK cascades.
JSAP1 Positively Regulates Apoptosis in Association with JNK3 during Neuroectoderm Formation Induced by RA-RA significantly increased the expression of both JSAP1 and JNK3 in wild type but not JSAP1-null EBs, and apoptosis was highly induced only in the former along with processing of caspase 3 and PARP. The JNK3 has been reported to positively regulate apoptosis especially in neurons (55). Interaction of JSAP1 with JNK3 in the 4 Ϫ /4 ϩ EBs indicates that JSAP1 may positively regulate JNK3 function or its stability as a scaffold protein for the JNK cascade, as proposed previously (15)(16)(17), at least during neuroectoderm formation in the EBs, thus inducing a prominent degree of apoptosis through the caspase 3 pathway.
We may suggest that JSAP1 plays an important role in association with JNK3 during neuroectoderm formation in the mouse embryo, since JSAP1 and JNK3 are coexpressed in the ␤-tubulin III-positive neuroectoderm of E7.5 embryo. Jnk1/ Jnk3 and Jnk2/Jnk3 double knockout but not Jnk1/Jnk2 double knockout mice develop normally; therefore, it is likely that JNK1 and JNK2, which mainly regulate apoptosis at midgestation during embryogenesis (5), can compensate for any JNK3 deficiency during neuroectoderm formation.
Functional Roles of JSAP1 in Regulation of Vesicle Transport in Neurons-From the JSAP1-null EBs formed in the presence of RA, the apparent outgrowth of neurites was poor, thus suggesting that JSAP1 is also involved in the late stage of neural maturation. In Drosophila, a JSAP1 homolog has been identified as Sunday Driver (SYD). which mediates kinesin-dependent axonal transport (56), and UNC-16, a JSAP1 homolog in Caenorhabditis elegans was reported to regulate the localization of the vesicular cargo by integrating JNK signaling and kinesin-1 transport (57).
In the JSAP1-null neurons, synaptophysin, which is a component of the synaptic vesicle and is usually localized in axon (58), was mostly localized in the soma and much less in neurites, thus suggesting that a loss of JSAP1 partly retarded the axonal transport in mouse neurons, thus resulting in a poor outgrowth of neurites. In the wild type neurons, JSAP1 was also detected in the terminal of the neurites, namely growth cone, as previously reported (16,59). Our results, thus, also provide evidence that JSAP1, a mammalian scaffold protein for JNKs, is involved in the regulation of vesicle transport, as are SYD and UNC-16.