Zebrafish GADD45β genes are involved in somite segmentation

Abstract

Somites in vertebrates are periodic segmented structures that give rise to the vertebrae and muscles of body. Somites are generated from presomitic mesoderm (PSM), but it is not fully understood how cellular differentiation and segment formation are achieved in the anterior PSM. We report here that zebrafish gadd45β1 and gadd45β2 genes are periodically expressed as paired stripes adjacent to the neural tube in the anterior PSM region where presomitic cells mature. In mammals, it is known that GADD45 (growth arrest and DNA damage) family proteins play a role in cell-cycle control. We found that both knockdown and overexpression of gadd45β genes caused somite defects with different consequences for marker gene expression. Knockdown of gadd45β genes with antisense morpholino oligonucleotides caused a broad expansion of mesp-a in the PSM, and both cyclic expression of her1 and segmented expression of MyoD were disorganized. On the other hand, injection of gadd45β1 or gadd45β2 suppressed expression of mesp-a and her1 in anterior PSM and MyoD in paraxial mesoderm. These results indicate that regulated expression of gadd45β genes in the anterior PSM is required for somite segmentation.

In vertebrates, somites are formed at regular intervals as repeated segments that subsequently give rise to skeletal muscle, vertebrae, and dermis (1). Somites are generated from the mesenchymal presomitic mesoderm (PSM) along the anterioposterior axis with spatially and temporally coordinated periodicity known as the segmentation clock (2, 3). Thus, it is thought that somite segmentation involves a molecular oscillator in the PSM. Although it is largely unknown how the periodicity of somite segmentation is generated, recent evidence indicates that the Notch/Delta signaling pathway plays an essential role in this process. In all vertebrates studied, such as mouse, chick, frog, and zebrafish, the existence of a molecular oscillator is indicated by the observation that the Notch ligand Delta and Notch targets related to the Hairy family of basic helix–loop–helix transcription factors display periodic expression patterns in the PSM (1). In zebrafish, it has been proposed that oscillations of deltaC activate the Notch pathway, leading to the cyclic expression of Hairy-related basic helix–loop–helix repressors Her1 and Her7 (4–6). An important point is that her1 expression is an intrinsic, cell-autonomous property of the PSM and does not depend on cell movement. Subsequently, both Her1 and Her7 proteins negatively regulate their own expression to establish synchronized oscillations.

In zebrafish, several somite mutants (aei, after eight; des, deadly seven; bea, beamter; fss, fused somites; and mib, mind bomb) were isolated in genetic screens (2). Among these, notch1a and deltaD are disrupted in des and aei mutants, respectively (5, 7). Further, mib encodes a RING E3 ligase that modifies Delta, affecting Notch signaling (8). Cyclic expression of genes such as her1 and deltaC in the posterior PSM is disorganized in these Notch pathway mutants. In the mutant fss that encodes the T-box factor 24 (Tbx24) (9), most of the anterior her1 strip is missing, but her1 oscillations in the posterior PSM occur normally. Thus, fss may not be required to generate cyclic gene expression but may be required to stabilize such expression in the anterior PSM. Other signaling pathways, including the Wnt, FGF, and retinoic acid (RA) pathways, contribute to the regulation of somite segmentation (10–12). Despite this progress in understanding somitogenesis, it remains unclear how presomitic cells transit from an immature to a mature state in the anterior PSM.

The growth arrest and DNA damage 45 (GADD45) family proteins, including GADD45α, GADD45β, and GADD45γ, play important roles in cell-cycle control by interaction with cell-cycle regulators. GADD45 proteins can associate with proliferating cell nuclear antigen and a cyclin-dependent kinase inhibitor, p21 (13, 14). In addition, GADD45 proteins also interact with cdc2 kinase and inhibit its activity (15). Further, GADD45 proteins activate the p38 and/or c-jun N-terminal kinase pathway by direct binding to MAP three kinase 1 (MTK1)/MAP/ERK kinase kinase 4 (MEKK4) in response to environmental stress (16). These findings suggest that GADD45 members regulate certain signaling pathways, including cell-cycle regulators, but the developmental functions of these proteins are still unknown. In this paper, we report that zebrafish gadd45β genes, but not gadd45α, are expressed as a single bilateral stripe in the anterior PSM immediately posterior to the position where the first somite forms. Both gain-of-function and knockdown analyses indicate that gadd45β genes are required for somite segmentation in the anterior PSM.

Materials and Methods

In Situ-Based Screening and Isolation of gadd45β Genes. Total RNA was isolated from zebrafish shield-stage embryos that had been treated for 10 min with 0.3 M LiCl at the 64/256-cell stage. A unidirectional cDNA library was prepared by using the SMART cDNA library construction kit (Clontech) in the pCS2-SfiI expression vector. Clone 9886 is specifically expressed as a paired signal in the PSM during somitogenesis stages. Sequence analysis showed that clone 9886 encodes the zebrafish counterpart of mammalian GADD45β gene. Full-length gadd45β1 (pCS2-GADD45β1) was isolated by PCR amplification using GADD45β1-S, 5′-GGGATTCTTACTACTGCCACACAACATC-3′, and GADD45β1-AS, 5′-GGAATTCCATAAACATGGAGGC-3′. We found additional zebrafish EST clones with similarity to gadd45β1 (GenBank accession no. AB180735), representing a distinct gene. We name this gene gadd45β2 (GenBank accession no. AB180736). Full-length gadd45β2 (pCS2-GADD45β2) was PCR-amplified by using GADD45β2-S, 5′-CGGGATCCGAACACTATACACACTTT-3′, and GADD45β2-AS, 5′-GCTCTAGATTACCATGCGCCACAGTTCC-3′. For the construct of FLAG-tagged MTK1-N, the N-terminal fragment of MTK1 was amplified from pcDNA-MTK1 (a gift from H. Saito, University of Tokyo) by using the following primer sets: 5′-CGGGATCCATGGAGGAGCCGCCGCCA-3′ and 5′-AAGGCCTACCTCTGGCGTTGGAGATGC-3′.

Microinjection of Synthetic RNA or Morpholino Oligonucleotides (MOs). Control and GADD45β MOs were obtained from Gene Tools (Philomath, OR) as follows: control MO, 5′-CCTCTTACCTCAGTTACAATTTATA-3′; GADD45β1-MO, 5′-ATCCAACAACCTCCTCCAGAGTCAT-3′; and GADD45β2-MO, 5′-ATCCAACGACTTCTTCCAGGGTCAT-3′. The nucleotides complementary to the initiation sites in the zebrafish gadd45β1 and gadd45β2 mRNA are underlined. MOs or synthetic RNAs were injected together with EGFP RNA as tracer into one blastomere of two-cell-stage zebrafish embryos.

Transfection and Western Blotting. Human 293T cells (3 × 10⁶) were transfected with expression vectors (total of 4 μg) by using the Lipofectamine reagent (Invitrogen) according to the manufacturer's instruction. Western blotting was performed as described in ref. 17. Anti-phospho-p38 and anti-p38 were obtained from Cell Signaling Technology (Beverly, MA), anti-FLAG from Sigma, and anti-HA from Roche Diagnostics.

Detection of Mitotic and Apoptotic Cells. For detection of mitotic cells, immunostaining with anti-phosphohistone H3 antibody that recognizes cells in late G₂ and M phase was performed, and for the detection of apoptotic cells, TUNEL assay was performed, both as described in ref. 18.

Results

Isolation of Zebrafish gadd45β Genes. Spatially and temporally restricted genes often function as regulatory molecules in the formation and differentiation of organs in vertebrate embryogenesis. To identify such genes, we performed in situ-based screening using probes from a unidirectional full-length enriched cDNA library (see Materials and Methods). Clone 9886 showed a unique expression profile highly restricted to a single bilateral stripe in the anterior mesenchymal PSM during somitogenesis stages. Because ectopic expression of sense RNA for this clone caused severe defects in the somite segmentation (see below), we further investigated the function of this clone in somite development. Sequence analysis revealed that clone 9886 contains a full-length cDNA for a zebrafish homolog of human and mouse GADD45β (16, 19); we designated this gene gadd45β1. The GADD45 family, which includes GADD45α, GADD45β, and GADD45γ, plays an important role in cell-cycle regulation and DNA repair. From EST database searches, we found an additional zebrafish gadd45β gene and isolated a full-length cDNA for this gene, which we named gadd45β2. GADD45β1 and GADD45β2, which are 77% identical in amino acid sequence, are very similar to mammalian GADD45β proteins and less similar to GADD45α and GADD45γ (Fig. 1 and data not shown). Mouse GADD45β can associate with multiple cellular proteins such as cdc2 kinase and MTK1/MEKK4 kinase, and the high level of sequence conservation (67–71% identity) suggests that zebrafish GADD45β proteins can undergo the same associations.

Fig. 1.

Primary structures of zebrafish gadd45β genes. Predicted amino acid sequences of Gadd45β1 and Gadd45β2 and sequence alignment of zebrafish (z), human (h), and mouse (m) GADD45β are shown. Conserved residues are boxed. Dashes indicate gaps introduced to optimize the sequence alignment.

Examination of gadd45β Gene Expression. Because there is no report about the expression pattern of GADD45β in early mammalian embryogenesis, we examined the developmental expression profile of the zebrafish gadd45β genes by whole-mount in situ hybridization. Neither gadd45β1 nor gadd45β2 is supplied maternally (data not shown). At the beginning of gastrulation, shield stage, gadd45β2 is specifically detected in the involuting mesoderm at the dorsal side (Fig. 2 J and K; red arrowhead), whereas gadd45β1 is expressed in the yolk syncytial layer (YSL) (Fig. 2 A; green arrowhead). The expression of gadd45β1 in the YSL is gradually decreased during gastrulation stages, whereas mesodermal expression of gadd45β2 was maintained throughout gastrulation and then disappeared during somitogenesis (Fig. 2 L and N; red arrowhead). Around the end of gastrulation (bud stage), a new region of specific expression of both gadd45β1 and gadd45β2 is detected in the form of a bilateral stripe adjacent to the neural tube in the anterior PSM (Fig. 2 B, L, and M). Further, gadd45β2 is expressed in the tail-bud domain (Fig. 2M; yellow arrowhead). In contrast, we never observed such restricted expression of gadd45α in the PSM (data not shown). As somite segmentation proceeds, we always observed one bilateral stripe for both gadd45β genes in the PSM (Fig. 2 C–G and N–P), suggesting that these genes are activated precisely at the maturation stage of presomitic cells. gadd45β1 and gadd45β2 are expressed at the same position in the anterior PSM as shown in Fig. 2S (black arrowhead), whereas weak gadd45β2 expression also is detected in the newly formed somites (Fig. 2P). Thus, the stripe of gadd45β expression shifts caudally with a constant interval. On the other hand, the Notch/Delta target her1 is expressed as an oscillating wave of usually three to four stripes sweeping once across the entire PSM during each cycle of somite formation (7). The her1 expression domains are moving in a caudal-to-rostral direction. Two-color in situ hybridization with gadd45β1 and her1 demonstrated that the stripe of gadd45β1 is located anterior of the second her1 expression domain with some overlap (Fig. 2T). This observation is consistent with the fact that gadd45β1 expression is positioned posterior to the MyoD expression domain, leaving a small gap (Fig. 2U); it is known that most of the posterior MyoD expression domain corresponds to the somite that is just forming (20). Expression of gadd45β genes in PSM appears linked to somite segmentation, as it is no longer detected in posterior mesoderm at the end of segmentation (Fig. 2 I and R). At the 24-h postfertilization stage, gadd45β1, but not gadd45β2, is expressed in the lens of eyes (Fig. 2H; blue arrowhead). Thus, expression of gadd45β1 is partly overlapping and partly distinct from that of gadd45β2 during early embryogenesis, and the uniquely restricted expression of both genes in the anterior PSM suggests that gadd45β genes contribute to segmentation in this region where the somite primordia mature.

Fig. 2.

Expression patterns of zebrafish gadd45β genes. Whole-mount in situ hybridization with probes for gadd45β1 (A–I) and gadd45β2 (J–R) is shown. Stages are indicated at the lower left of each micrograph. S, somite; hpf, hours postfertilization. Lateral view, dorsal is right (A and J) and anterior is left (B–D, F, H, I, L, N, and P–S). Dorsal view, anterior is left (E and O) and anterior is up (U). Anterior view, dorsal is right (K). Posterior view, dorsal is up (G, M, and T). Expression of gadd45β1 in the YSL at the shield stage is shown in A (green arrowhead). gadd45β2 expression in the shield and involuting mesoderm is shown in J–L (red arrowhead). At bud stage, one bilateral stripe arises for both gadd45β1 and gadd45β2 at the anterior edge of presomitic mesoderm (C, L, and M; black arrowhead); weak expression of gadd45β2 is observed in the tail-bud mesoderm (L and M; yellow arrowhead). Expression in anterior presomitic mesoderm is maintained throughout somitogenesis (C–G and L–P; black arrowhead). Weak expression of gadd45β2, but not gadd45β1, is maintained in the somites. At 24 hpf, gadd45β1 expression is newly detected in the lens of the eye (H, blue arrowhead). (S) Whole-mount in situ hybridization with both gadd45β1 and gadd45β2 probes shows coexpression in the anterior PSM (black arrowhead). (T) Double in situ hybridization shows that gadd45β1 (purple and black arrowhead) is located just anterior of the second her1 expression domain (red and red asterisks) in the anterior PSM. (U) Double in situ hybridization shows that gadd45β1 is positioned posterior to the MyoD expression domain (red and red bracket), leaving a gap between the domains.

Knockdown Analysis of gadd45β Genes in Somite Segmentation. To examine a possible requirement for GADD45β proteins in somitogenesis, we interfered with their synthesis with the aid of antisense MOs. Although gadd45β1 and gadd45β2 are expressed in the YSL and in the involuting mesoderm at the shield stage, respectively, double MO injection did not affect the expression of goosecoid/gsc (organizer gene) (21), vega1 (complementary expression to gsc) (22), and no tail/ntl (mesodermal marker) (23) at the shield stage, and epiboly movements during gastrulation were normal (data not shown). However, somite segmentation of embryos injected with both MOs (GADD45β1-MO; 7.5 ng + GADD45β2-MO; 7.5 ng) was severely inhibited (85%; n = 71), whereas uninjected (minor somite defects, 3%; n = 100), control MO-injected (15 ng, 5%; n = 92), single GADD45β1-MO-injected (7.5 ng, 11%; n = 53), or GADD45β2-MO-injected embryos (7.5 ng, 12%; n = 57) were normally segmented (Fig. 3 A–C and data not shown). These results suggest that there is redundancy of the gadd45β genes in somite segmentation. To characterize the molecular events underlying these somite defects, we examined the expression of several genes implicated in somitogenesis. The genes encoding Mesp family transcriptional factors, mesp-a and mesp-b, are segmentally expressed in the anterior PSM, and both genes are required for the maturation of somite primordia (24, 25). We observed expansion of mesp-a into a broad band in the anterior PSM of double MO-injected embryos, whereas mesp-a expression in control MO-injected embryos was normal (Fig. 3 D–F). Oscillation of her1 in double MO-injected embryos became obscure as this gene was expressed in a broad domain rather than in the three stripes seen in control embr yos (Fig. 3 G–I). Further, in double GADD45β-MO-injected embryos, MyoD and fgf8 expression in somites was present but disorganized (Fig. 3 J–O). Interestingly, fgf8 expression in the PSM was anteriorly expanded in double GADD45β-MO-injected embryos. Thus, we suggest that expression of gadd45β2 in the tail bud may have a role in regulating fgf8 expression in the PSM. In contrast, injection of both GADD45β-MOs did not affect the expression of ntl in the notochord or of tbx6, tbx24, and wnt3l (previously named wnt3a) in the PSM (Fig. 7, which is published as supporting information on the PNAS web site). Thus, gadd45β genes are required for proper patterning of genes involved in somite development.

Fig. 3.

Effect of gadd45β MOs on somite segmentation. MOs (control-MO, 15 ng; GADD45β1-MO + GADD45β2-MO, 7.5 ng each) were injected into one- or two-cell-stage embryos. Dorsal view, anterior is up (A–C and J–L). Posterior view, dorsal is up (D–I). Lateral view, anterior is left (M–O). (A–C) Live embryos at 8- or 10-somite stages. Injection of GADD45β1-MO plus GADD45β2-MO, but not control-MO, caused severe defects in somite segmentation (A–C). (D–O) Whole-mount in situ hybridization with probes are shown at the upper right of each micrograph. Expression of mesp-a and her1 in the anterior PSM and segmental expression of MyoD and fgf8 (red asterisk) in the somites were disorganized in the GADD45β1-MO + GADD45β2-MO-injected embryos (D–O). Expression of fgf8 in the PSM (red bracket) was expanded in the double GADD45β MO-injected embryos (M–O).

Overexpression of gadd45β Genes Affects Somite Segmentation. The function of gadd45β genes in somitogenesis was also examined by overexpression experiments. Synthetic GADD45β RNAs were injected together with EGFP RNA as tracer into one blastomere of two-cell-stage embryos; we could judge the injected area by GFP expression. Injection of gadd45β1 or gadd45β2 RNA (75 pg) did not affect the expression patterns of gsc, vega1, and ntl at the shield stage, and epiboly proceeded normally during gastrulation (data not shown). Subsequently, somite segmentation was severely inhibited on the side that was injected with gadd45β1 (86%; n = 103) or gadd45β2 (59%; n = 111) RNA (Fig. 4 and data not shown), whereas somites were normally segmented in uninjected or EGFP RNA-injected embryos (minor somite defects, 4%; n = 27). Essentially, the same segmentation defects were evident in embryos injected with mouse GADD45β RNA (81%; n = 37) (data not shown). In contrast, such defects were not observed in the embryos injected with a mutant gadd45β1 construct (gadd45β1S N-terminal fragment, amino acids 1–60, minor somite defects, 4%; n = 98) in which both middle- and C-terminal domains of GADD45β1 are deleted. These somite segmentation defects are similar to those in double MO-injected embryos. Therefore, we compared the effect of gadd45β RNA injections on marker gene expression during somite formation. Strikingly, MyoD expression was strongly reduced on the side into which gadd45β1 or gadd45β2 RNA had been injected (Fig. 4 J–L). Paraxial protocadherin (PAPC) is segmentally expressed in the anterior PSM (26, 27). In gadd45β1 or gadd45β2 RNA-injected embryos, both mesp-a and PAPC induction in anterior PSM was severely suppressed (Fig. 4 M–O and data not shown). Interestingly, expression of the two anterior her1 stripes was strongly suppressed in the anterior PSM in gadd45β1 or gadd45β2 RNA-injected embryos, whereas the most posterior her1 signal near the tail-bud region remained (Fig. 4 G–I). As expected, the expression of her1, MyoD, and mesp-a, in addition to ntl, was not altered in EGFP RNA-injected embryos (data not shown). Thus, the gain-of-function and knockdown results lead to similar somite segmentation defects, but the effect on the expression of segmentation genes is different. Whereas overexpression of GADD45β factors inhibits segmentation genes expression, loss of function of GADD45β destroys their pattern.

Fig. 4.

Overexpression of gadd45β genes affects somite segmentation. Synthetic RNA (gadd45β, 75 pg) together with EGFP RNA (300 pg; tracer) was injected into one blastomere of two-cell-stage embryos. The injection side was identified by EGFP before in situ hybridization. Dorsal view, anterior is up (A–F, J–L, and P–R). Posterior view, dorsal is up (G–I and M–O). (A–C) Morphology of developing somites in live embryos at 8- or 10-somite stages. (D–F) GFP expression at the corresponding embryos. Injection of zebrafish gadd45β1, but not a gadd45β1 mutant lacking middle- and C-terminal domain, caused defects in somite segmentation (A–C). (G–R) Whole-mount in situ hybridization with probes shown at the upper right. Expression of segmentation genes (mesp-a, her1, and MyoD) was severely reduced in the injected side (G–O), whereas ntl expression was not changed (P–R).

Next, we examined the effect of forced expression of gadd45β genes on cell proliferation and cell death. In the injected half of gadd45β1 or gadd45β2 RNA-injected embryos, we observed a slightly decreased number of mitotic cells (late G₂ and M phases) as visualized with anti-phosphohistone H3 antibody (Fig. 5 A–C). In contrast, the number of apoptotic cells detected by TUNEL assay was somewhat increased in the gadd45β1 or gadd45β2 RNA-injected side (Fig. 5 D–F). Thus, overexpression of gadd45β genes weakly affects cell death and mitosis, but these defects appear mild, compared with the suppression of segmentation genes under the same conditions.

Fig. 5.

Injection of gadd45β1 RNA weakly affects cell proliferation and apoptosis. Synthetic RNA (gadd45β1, 75 pg) together with EGFP RNA (300 pg; tracer) was injected into one blastomere of two-cell-stage embryos. Dorsal view, anterior is up (A–F). (A–C) Detection of mitotic cells by immunostaining of anti-phosphohistone H3 antibody. The number of mitotic cells is slightly decreased in the injected side of gadd45β1 or gadd45β2 RNA-injected embryos. (D–F) Detection of apoptotic cells by TUNEL assay. A slightly increased number of apoptotic cells was observed at the injected side of gadd45β1 or gadd45β2 RNA-injected embryos.

Functional Interaction of gadd45β Genes and the MTK1/MEKK4 Signaling Pathway. It has been shown in mammals that GADD45β can directly associate with MTK1/MEKK4 and activate MTK1/MEKK4 kinase activity (16). Activated MTK1/MEKK4 leads to the activation of the p38 kinase and/or c-jun N-terminal kinase pathways. To investigate the biochemical activity of zebrafish GADD45β, we examined whether this protein can interact with MTK1/MEKK4. As reported in ref. 16, the N-terminal domain of human MTK1/MEKK4 binds to the human GADD45β protein. FLAG-tagged human MTK1 (FLAG-MTK1-N; N-terminal fragment, amino acids 22–348) and hemagglutinin (HA)-tagged zebrafish GADD45β1 (HA-GADD45β; full-length) were transfected into human 293T cells. As shown in Fig. 6A, FLAG-MTK1-N protein was coimmunoprecipitated with HA-GADD45β1, demonstrating that zebrafish GADD45β1 can interact with human MTK1/MEKK4. Next, we examined whether GADD45β1 cooperates with MTK1 in the activation of the p38 signaling pathway. To measure p38 activity, we used an antibody that specifically recognizes the activated double-phosphorylated form of p38 (28). Potent phosphorylation of p38 was observed when MTK1 was cotransfected with HA-GADD45β (Fig. 6B), suggesting that the induction of zebrafish GADD45β would lead to the activation of p38 signaling pathway. In addition, we found that injection of human MTK1 RNA (300 pg) plus GFP (300 pg) into one blastomere at the two-cell stage caused severe somite defects (59%; n = 39) and strong suppression of mesp-a, her1, and MyoD expression (Figs. 6C and 7 M–R). These effects are similar to those seen in gadd45β1-injected embryos, suggesting that expression of gadd45β genes results in the activation of the MTK1/MEKK4 signaling pathway in the zebrafish PSM.

Fig. 6.

Biochemical activities of zebrafish GADD45β1. Indicated expression constructs (4 μg total: HA-GADD45β1, 2 μg; FLAG-MTK1-N, 2 μg; MTK1, 2 μg; pCS2 vector, 2 μg) were transfected into human 293T cells. Cell lysates were prepared after 24 h. (A) Cell lysates were immunoprecipitated with anti-FLAG gel and analyzed by Western blotting using anti-HA antibody (Top). Whole-cell lysates were used as controls for expression levels. (B) Phosphorylation state of p38 in the transfected cells. Cell lysates were analyzed by Western blotting using antibodies recognizing phospho-p38 and total p38. (C) Live embryos at the 10-somite stage. Synthetic RNA (human MTK1, 300 pg + EGFP, 300 pg) was injected into one blastomere of two-cell-stage embryos, leading to severe somite defects in the injection side.

Discussion

Somites are produced from the PSM as repeatedly segmented structures of the paraxial mesoderm (1). Periodicity of somite segmentation is coordinately mediated by the segmentation clock (3). Oscillation of Notch signaling components (such as her1 and deltaC in zebrafish) in the PSM functions as the basic mechanism of the clock (5), whereas FGF, Wnt, and RA contribute to the stabilization of oscillating prepatterns and determine the spatial position and maturation state of presomitic primordia (10–12). Further, the transcription factor fss/tbx24 is required for the induction of segmentation genes such as mesp-a and PAPC, as well as for the stabilization of oscillating genes (9). The important question remains how presomitic cells differentiate from the immature to the mature state. In this paper, we provide evidence that zebrafish GADD45β factors are periodically expressed in the anterior PSM where presomitic cells mature and play a role in somite segmentation.

In mammals, GADD45-family proteins (including GADD45α, GADD45β, and GADD45γ) are structurally conserved and modulate the activities of various types of cell-cycle regulators (29). However, the developmental expression and function of this protein family has remained obscure. We found that zebrafish gadd45β1 and gadd45β2 genes are periodically expressed as single bilateral stripes in the anterior PSM (Fig. 2). This expression is distinct for gadd45α in zebrafish and gadd45γ in medaka (30), both of which are not expressed in the anterior PSM during somitogenesis. In the PSM, Notch activation drives the rhythmic expression of her1 as a wave of three stripes that moves from the posterior to the anterior. Subsequently, Her1 proteins negatively regulate their own expression and control the expression of other cyclic genes. This oscillation plays an important role to determine the timing of somite boundary formation. The stripe of gadd45β gene expression is located immediately anterior to the second her1 domain where presomitic cells undergo maturation. Consistent with this positional relationship, a gap exists between the most posterior MyoD domain at the first forming somite and the gadd45β gene expression domain. Expression of gadd45β genes is reduced and displayed a diffused pattern at the 18-somite stage in fss/tbx24 and mib mutants (data not shown), suggesting that gadd45β genes are regulated by the segmentation clock in the maturation domain of successive somite primordia.

Knockdown of both gadd45β1 and gadd45β2 by double, but not single, MO injection, caused defects in somite segmentation (Fig. 3). In double MO-injected embryos, somite segmentation genes such as her1, mesp-a, PAPC, and MyoD were expressed, but their patterns were disorganized. Expression of mesp-a and fgf8 in PSM was expanded, whereas segmental expression of MyoD and cyclic expression of her1 became indistinct. Further, we found that RNA injection of zebrafish gadd45β1, gadd45β2, or mouse GADD45β into zebrafish embryos causes severe segmentation defects (Fig. 4) that are superficially similar to those obtained by double MO injection (Fig. 3). Despite this similarity in phenotype, there are differences in the behavior of marker genes between the knockdown and gain-of-function experiments. Expression of mesp-a is lost in the anterior PSM in gadd45β1-or gadd45β2-injected embryos, and so is expression of her1 in the anterior two stripes but not in its most posterior domain. One possible interpretation is that loss of her1 and mesp-a expression in the anterior PSM by ectopic expression of gadd45β1 is due to abnormal inhibition of somite primordium maturation, resulting in the suppression of MyoD activation. These results indicate that a precise level of gadd45β gene expression at the right time and place is required for somite boundary formation in the anterior PSM.

The high similarity between zebrafish and human GADD45β protein (16) suggests conservation of their biological activities. In mammals, GADD45β is involved in cell-cycle control by interaction with regulatory factors, including MTK1/MEKK4 (14). Interaction of GADD45β with the N-terminal region of MTK1/MEKK4 leads to a conformational change that releases MTK1 autoinhibition (31); activated MTK1 induces the phosphorylation and activation of MKK6, followed by the activation of p38 and/or c-jun N-terminal kinase signaling pathways. We found that zebrafish GADD45β1 can bind to the N-terminal domain of human MTK1 in human 293T cells (Fig. 6). In addition, cotransfection of zebrafish GADD45β1 and a full-length human MTK1 enhances the accumulation of the activated form of p38, supporting the view that zebrafish and mammalian GADD45β1 are functional homologues. Human MTK1/MEKK4 overexpression in zebrafish embryos caused severe segmentation defects with strong suppression of mesp-a, her1, and MyoD; these phenotypes are very similar to those induced by gadd45β injection. Whereas different models are possible for the role of GADD45β in somite segmentation, we favor the view that activation of gadd45β genes in the anterior PSM induces the localized modulation of downstream signaling molecules including MTK1/MEKK4 by direct binding to GADD45β, and that these regulatory interactions are required for somite segmentation.