Introduction

During the zebrafish CNS development, the neural plate undergoes secondary neurulation and is converted into the neural keel, which subsequently develops into neural tube by cell detachment in the center (Schmitz et al. 1993; Kimmel et al. 1994). The anterior part of the neural tube further expands and swells mediolaterally, generating the forbrain, midbrain, and hindbrain. Later the hindbrain will be further subdivided into seven reiterated units, termed rhombomeres. Each rhombomere has its unique identity by expressing distinct gene profile, such as different combination of Hox genes. Rhombomeres have been shown to be involved in patterning neural crest cells and directing their migration to proper branchial arches (reviewed by Trainor and Krumlauf 2000). The interface between two rhombomeres is termed rhombomere boundary, which is formed by internal subdivision of the hindbrain according to a species-specific sequence. In zebrafish, the first rhombomere boundaries (r3/4 and r4/5) appear between five to seven somite stage (ss) and the last one appears at about 16 ss (Moens et al. 1998). Initially, segregation of rhombomeres results from different affinity between odd and even rhombomeres, which eventually acquire distinct identity (Guthrie and Lumsden 1991). Later, a finer scale cell sorting is accomplished by interactions between Eph receptor-expressing cells and Ephrin ligand-expressing cells (Xu et al. 1999). Recent studies have suggested that EphA4-dependent adhesion plays a redundant role with EphA4-dependent repulsion in rhombomere boundary formation (Cooke et al. 2005).

In Drosophila, it has been well studied that Notch signaling is involved in the dorsal–ventral boundary formation of wing disk (reviewed by Tepass et al. 2002). Fringe modulates the Notch sensitivity to different ligands and hence ensures that Notch activity is limited to the boundary cells. Recently, several studies have converged to provide evidence that Notch activation plays a similar role in the formation of rhombomere boundaries in a teleost, zebrafish. Firstly, large-scale genetic screen has identified several mutants that displayed rhombomere boundary defects, including mib mutants (Jiang et al. 1996). Positional cloning has revealed that mib encodes an E3 ligase required for efficient Notch activation in the neighboring cells by regulating ligand endocytosis (Itoh et al. 2003). Secondly, in deltaA dominant-negative mutants and mib mutants, strong neurogenic phenotype is observed and the rhombomere boundaries are disrupted (Riley et al. 2004; Appel et al. 1999; Cheng et al. 2004). Thirdly, further mosaic analysis has provided evidence that Notch activation is required for the regulation of specific rhombomere boundary cell movement: cells with high activation of Notch receptor or expressing a dominant-active form of Su(H) are prone to segregate into the boundary region, whereas cells expressing a dominant-negative form of Su(H) are expelled away from the boundary region (Cheng et al. 2004). In addition, rfng is expressed in the rhombomere boundaries and acts as a modulator involving in regulating neurogenesis (Qiu et al. 2004; Amoyel et al. 2005). Despite the mild neurogenic phenotype, no rhombomere boundary defect was observed in zebrafish notch1a mutants, suggesting a redundancy of multiple Notch homologs in the zebrafish (Gray et al. 2001). Recent study in mice has demonstrated that persistent and high levels of Hes1 expression repress Mash1 expression in boundary regions and hence ensure the generation of neuron-free zones in boundaries (Baek et al. 2006). Loss of Hes1 together with Hes5 and Hes3 resulted in ectopic neurogenesis and hence the disruption of rhombomere boundaries. On the other hand, Wnt signaling has been proved to interplay with Notch signaling in the rhombomere boundary formation (Riley et al. 2004; Amoyel et al. 2005). Forced expression of wnt1 partially rescues hindbrain patterning in mib mutants. Furthermore, similar to the regulatory network in Drosophila, Wnt1 regulates delta and proneural gene expression in non-boundary cells, and hence prevents them from acquiring boundary cell fate.

In this study we observed progressive changes in cell morphology and premature neuron differentiation in the hindbrain of mib mutants. Inhibition of neurogenesis in mib mutants leads to the recovery of rhombomere boundaries. We have also confirmed and further demonstrated that the disruption of rhombomere boundary is Su(H)- and dosage-dependent. In addition, we established that Notch1a and Notch3 play a redundant role and their activation is required in boundary maintenance before 15 ss.

Materials and methods

Fish maintenance and mutant identification

Zebrafish maintenance and breeding were carried out as previously described (Kimmel et al. 1995). Embryos were raised at 28.5°C and the approximate stage is determined under the dissecting microscope.

hadc1 hi1618 mutants (Golling et al. 2002) used for characterization were genotyped by using two pairs of primers, one pair derived from the lacZ gene (P1: 5′-ATCCTCTAGACTGCCATGG-3′; P2: 5′-ATCGTAACCGTGCATCTG-3′) harbored by the viral vector for confirming insertion and the other from the genomic sequence flanking the exon 1 (P1: 5′-CCTACAGTGATGGAACCTGT-3′; P2: 5′-CGGTCCACAGTATGAAGCTA-3′) for confirming the hdac1 gene. Homozygous mib mutants were distinguished from siblings by their abnormal trunk morphology. Identification of mib ta52b ;hadc1 hi1618 double mutants in Fig. 3 was further sequenced by using intron 21 and intron 22 of the mib genomic sequence (P1: 5′-AGCTTGACAGGCGTAGCAACA-3′; P2: 5′-ACGATTGAACGCTACGTCACACA-3′), where the T to G transversion leads to a change of Met (ATG) to Arg (AGG).

Heat shock induction

Embryos were given heat shock for 1 h at 39°C and later incubated at 28.5°C for further development. The embryos were fixed after they reached the desired stage and processed for in situ hybridization or immunohistochemistry. Using this heat shock method, no hindbrain boundary defects in any batch of the wild-type siblings were observed.

Live embryo imaging

Bodipy ceramide (Molecular Probes) was dissolved in dimethyl sulfoxide (DMSO) at a concentration of 5 mM for stock solution. Dechorinated embryos were soaked into 200 μM bodipy ceramide solution for 30 min in the dark. The embryos were then washed with egg water 3 × 10 min each and mounted in wells in 1% low melting agarose for confocal microscopy. Confocal imaging was performed using a Zeiss LSM510 laser-scanning microscope (Cooper et al. 1999) and its images were analyzed using LSM software (Zeiss) and Photoshop 6.0 (Adobe).

Whole-mount in situ hybridization and immunohistochemical staining

In situ hybridization was performed as described (Qiu et al. 2004). For immunohistochemical staining, embryos between 12 and 48 hpf were fixed in either 4% formaldehyde buffered with phosphate-buffered saline (PBS) or 2% trichloroacetic acid (TCA) (Sigma-Aldrich) for 4 h at room temperature (RT). After fixation, embryos were rinsed with PBS and then with distilled water, permeabilized with acetone treatment at −20°C for 5 to 10 min, rinsed again with distilled water and then PBS, and then blocked in PBDT (PBS with 0.1% Triton X-100 and 1% DMSO) with 10% goat serum for 2 h at RT. After incubation in primary antibody overnight at 4°C, embryos were then washed intensively and incubated in biotin-labeled secondary antibody for 2 h followed by soaking in a peroxidase-conjugated avidin–biotin complex (ABC kit; Pierce) for 1 h. The following antibody and dilution were used: anti-phospho-histone3 (Cell Signaling), 1:200 and anti-Hu (Molecular Probe), 1:1,000. A brown precipitate was formed by incubating embryos with 0.8 mg/ml of diaminobenzidine (DAB) (Sigma-Aldrich) and 0.001% H2O2. For whole-mount immunofluorescent staining, after the primary antibody incubation and intensive wash, embryos were incubated in PBDT with goat anti-mouse IgG Alexa Fluo 488 (Molecular Probe) at 1:400 dilution at RT for 2 h. Embryos were then rinsed with PBDT and analyzed using a Zeiss fluorescence microscope. Images were obtained using a Zeiss 510 confocal microscope.

Morpholino microinjections

Morpholinos were designed to target the translation or block the intron–exon splicing following the rule recommended by Gene Tools. The morpholinos were dissolved in 1× Danieu’s buffer [58 mM NaCl, 0.7 mM KCl, 0.4 mM MgSO4, 0.6 mM Ca(NO3)2, 5.0 mM HEPES (pH 7.0)] to give a final stock concentration of 5.0 mM as previously described (Nasevicius and Ekker 2000). For injection, the morpholinos were diluted in 1× Danieu’s buffer at concentrations from 0.2 to 1.0 mM. One to two nanoliters of MO was injected into each embryo. Sequence for su(h)-MO: 5′-CAAACTTCCCTGTCACAACAGGCGC-3′ (Sieger et al. 2003), sequence for notch1b exon 27 donor-MO: 5′-AATCTCAAACTGACCTCAAACCGAC-3′ (Milan et al. 2006), sequence for notch3-ATG-MO: 5′-ATATCCAAAGGCTGTAATTCCCCAT-3′ (Lorent et al. 2004), sequence for notch3-3-MO: 5′-ATCAGTCATCTTACCTTCGCTGTTG-3′, notch3-utr-MO: 5′-ACATCCTTTAAGAAATGAATCGGCG-3′ (Ma and Jiang 2007). Injected embryos of each individual notch3-MO displayed similar phenotypes and marker expression changes.

In vitro transcription and translation reactions

N-terminal notch3 cDNA was amplified by reverse transcription polymerase chain reaction (RT-PCR) using the forward primer 5′-ATCGATGGGGAATTACAGCCTTTGG-3′ and the reverse primer 5′-ATCGATGCATGAGACACTCGTTGAC-3′, based on the published notch3 cDNA sequence (NM_131549). After digested with ClaI, the PCR product was ligated to the pCS2+-myc vector. Six hundred nanograms of pCS2+-notch3-myc plasmid was added to TNT quick coupled transcription/translation reactions (Promega). Different amounts of notch3-ATG-MO were added into the reaction as indicated. Following incubation, the reaction products were run in a 10% acrylamide gel and detected by Western blot using antibody against Myc.

Results

Different mib mutants display a progressive defect in maintaining rhombomere boundaries with varying severity

Zebrafish mib gene encodes an E3 ligase that is an essential component of Notch signaling (Itoh et al. 2003). Previous studies have revealed that mib ta52b mutants display rhombomere boundary defects by 26 hpf (Jiang et al. 1996) and that the neurogenic phenotype of mib tfi91 is much weaker than that of mib ta52b (Zhang et al. 2007b). To address whether neurogenic defects affect the severity of boundary phenotype, we investigated the severity of rhombomere boundary defects in different mib alleles. In situ hybridization analysis of rfng expression revealed that all rhombomere boundaries were affected in these two alleles (Fig. 1a–c). In wild-type embryos, rfng expression was restricted to the rhombomere boundaries (Fig. 1a), whereas in mib ta52b mutants, the expression level of rfng was down-regulated and almost abolished in r4/5 and r5/6 boundaries (Fig. 1b). Different from mib ta52b mutants, only lateral part of rfng expression was lost in mib tfi91 mutants (Fig. 1c). As it has been demonstrated that rhombomeres are maintained as lineage-restricted compartments (Xu et al. 1999) and that distinctive cellular morphology can be observed in the successive rhombomere boundaries (Moens et al. 1996), we next examined the cellular changes in hindbrain region in different mib mutants. Bodipy ceramide labeling outlined cell contours and revealed the elongated epithelial structure of cells in the wild-type hindbrain (Fig. 1d, g). In mib ta52b mutants, neuroepithelial cell organization was abnormal and cells were frequently rounder and also presented in aggregates separated by large intercellular spaces. These aggregates were often organized into rosettes that appear to cavitate and enclose a central lumen (Fig. 1e, h). In mib tfi91 mutants, while the cellular shape appeared normal in the medial part of the hindbrain, cells became rounder in the lateral part and also had lost their epithelialized shape (Fig. 1f, i). These observations indicated that the disruption of rhombomere boundary in mib ta52b mutants was more severe than that in mib tfi91 mutants.

Fig. 1
figure 1

Disruption of hindbrain rhombomere boundaries in mib mutants. a–c Representative rfng expression. a rfng is strongly expressed at rhombomere boundaries in wild-type embryos at 24 hpf. b In mib ta52b mutants, almost no rfng transcripts can be detected at the r4/5 or r5/6 boundary and rfng is present at lower levels in the remaining boundaries. c In mib tfi91 mutants, rfng expression remains intact in the medial part of the boundaries and disappears in the lateral part. d–m Confocal microscopic images at hindbrain levels by using bodipy ceramide staining technique: wild-type siblings at d 24 hpf, g 32 hpf, j 17 hpf, and l 22 hpf; mib ta52b mutants at e 24 hpf, h 32 hpf, k 17 hpf, and m 22 hpf; and mib tfi91 mutants at f 24 hpf and i 32 hpf. White arrows indicate ectopic lumens in dorsal rosette-like structures in mib mutants. Rhombomere interfaces are marked with yellow arrowheads. Note that in mib mutants the cellular shapes are irregular, while in wild-type siblings cells are polarized and elongated. k The majority of cells in mib ta52b mutants display elongated epithelial structure at 17 hpf and the interfaces between rhombomeres are visible. m Rosette-like structure can be found in mib ta52b mutants at 22 hpf and the boundary interface between r4 and r5 is hard to be identified at this stage. n, o Anti-Hu antibody, which recognizes the pan-neuronal marker Hu, was used to identify Hu-positive cells in n wild-type and o mib ta52b mutants at 17 hpf. Note that in mib ta52b mutants the number of Hu-positive cells is greatly increased in the r5 region. p In wild-type embryos at 17 hpf, foxb1.2 is expressed at high level in the boundary cells. q Despite the slight down-regulation of foxb1.2 in the posterior hindbrain of mib ta52b mutants at 17 hpf, high expression level of foxb1.2 still can be observed in the boundary cells. r Expression level of rfng is gradually reduced in embryos at 30 hpf. s, t Expression of n-cad was greatly reduced in t mib ta52b mutants, when compared to s wild-type embryos at 15 ss. All images are dorsal views with anterior to the left. ot, otic vesicle

Full size image

N-cadherin (N-cad) is a cell adhesion molecule and was expressed at high level in the whole hindbrain in wild-type embryos at 15 ss (Fig. 1s). In mib ta52b mutants, the n-cad level was greatly down-regulated (Fig. 1t). In contrast, we did not detect obvious n-cad down-regulation in mib tfi91 mutants even at 22 hpf (data not shown). Therefore, it implied that changes in the expression level of cell adhesion molecule could be one of the reasons accounting for the different degree of cellular changes between mib ta52b and mib tfi91 mutants. However, in mib ta52b mutants at 15 ss, high level of n-cad expression in the boundaries flanking r5 was observed, suggesting that other adhesion molecules might be also required for the cell morphological change in mib mutants.

Recent studies have demonstrated that Notch activation is not sufficient for hindbrain boundary cell specification and loss of Mib function does not affect the initiation of rhombomere boundaries (Cheng et al. 2004). We provided further evidence to confirm that in mib ta52b embryos the boundaries were initiated despite the blockade of Notch activation. In mib ta52b mutants at 17 hpf, elongated epithelial shape could be clearly identified and no obvious morphological difference between wild-type and mutant embryos could be detected in the hindbrain (Fig. 1k; compared to the control in j). Consistently, we still observed high level of foxb1.2 expression in the boundary cells of mib ta52b mutants at corresponding stage, despite its slight down-regulation in posterior boundaries (Fig. 1q; compared to the control in p). However, a few round cells were present and boundary cells could not be clearly distinguished from non-boundary cells in mib ta52b mutants by 22 hpf (Fig. 1m; compared to the control in l). Furthermore, the cellular change was most severe in r5. We also found an increase of Hu-positive cells in the r5 region of mib ta52b mutants, suggesting that a loss of r4/r5 and r5/r6 boundaries is correlated with the strong neurogenic phenotype in the r5 region (Fig. 1n, o).

We have also checked whether cell proliferation and/or cell death attributes to the observed hindbrain phenotype in mib ta52b mutants. Fewer dividing cells were observed in mib ta52b mutants when compared to their wild-type siblings. However, reduction of cell proliferation in wild-type embryos by aphidicolin treatment did not result in boundary disruption (Supplementary Fig. 1). Similarly, no obvious difference in apoptosis could be detected in wild-type and mib ta52b embryos by using TUNEL staining (data not shown). These results indicated that decreased cell proliferation and apoptosis are unlikely responsible for the disruption of rhombomere boundaries in mib ta52b mutants.

Maintenance of rhombomere boundaries is Su(H)- and dosage-dependent

The preceding analysis of boundary cell phenotype in mib tfi91 and mib ta52b mutants suggested that different level of compromise in Notch signaling has graded consequence in boundary formation and cell morphology. To address this issue more quantitatively, we injected different amount of su(h)-MO into wild-type embryos and checked their hindbrain development. In addition to the loss of all somites posterior to the first five to seven somites (Sieger et al. 2003), irregular hindbrain morphology like mib ta52b was observed (data not shown). Moreover, in 0.2 pmol su(h)-MO-injected embryos (n = 12/13), rfng expression was reduced and most obviously at r4/r5 and r5/r6 boundaries (Fig. 2b; compared to the control in a). Using her4 as a Notch activation readout, we have observed that the expression level of her4 was greatly reduced in su(h)-MO-injected embryos at 24 hpf when compared to the wild-type control embryos (n = 22/22; Fig. 2c, d). This analysis confirmed that Notch activation in rhombomere boundary maintenance is via Su(H) (Cheng et al. 2004).

Fig. 2
figure 2

Lack of Su(H) by injecting su(h)-MO causes a dosage-dependent rhombomere boundary defect. Representative rfng expression in a uninjected embryos or b embryos injected with 0.2 pmol su(h)-MO at 24 hpf. In su(h) morphants, rfng expression is significantly reduced, compared to wild-type uninjected embryos. c, d her4 expression changes in c uninjected embryos and d 0.2 pmol su(h)-MO-injected morphants at 24 hpf. Note the reduction of her4 expression level in injected embryos. e–h Confocal microscopic images at hindbrain levels by using bodipy ceramide staining technique. In embryos injected either with e 0.1 pmol or f 0.5 pmol su(h)-MO and observed at 17 hpf, the interfaces between rhombomeres can be easily identified and the cellular structure is elongated as wild-type. Yellow arrowheads indicate the interfaces between two adjacent rhombomeres. By 22 hpf, cells start to lose their epithelial structure in embryos injected with g 0.1 pmol su(h)-MO. More severe effect is observed in the embryos injected with h 0.5 pmol su(h)-MO. Note that in the r4 and r5 region, the cellular structure is irregular and the arrangement is disorganized. i I–k III Representative progressive disruption of the rhombomere boundaries in su(h)-MO-injected embryos. Same batch of embryos injected with 0.1 pmol (affected embryo number shown in blue) or 0.5 pmol (affected embryo number shown in red) and fixed at different time points, i I–III 21 hpf, j I–III 24 hpf, and k I–III 26 hpf, for foxb1.2 in situ hybridization analysis in hindbrain. Variable foxb1.2 expression changes in su(h)-MO-injected embryos can be classified into three classes: class I shows minor changes; class II displays less severe changes; class III displays the most severe changes. Elevated expression of foxb1.2 can be observed in the rhombomere boundaries of l wild-type embryos. foxb1.2 expression in m mib tfi91 and n mib ta52b mutants at 24 hpf. Note that foxb1.2 transcripts can be detected at the medial part of the boundary in mib tfi91 mutants, while in mib ta52b mutants the expression of foxb1.2 is uniformly at low level and no boundary expression is visible. her6 expression in o wild-type, p mib tfi91, and q mib ta52b embryos at 24 hpf. Note that her6 expression level is dramatically reduced in mib ta52b as compared to that in mib tfi91 mutants. her9 expression in r wild-type, s mib tfi91, and t mib ta52b embryos at 24 hpf. Note that her9 expression level is decreased to a greater extent in mib ta52b as compared to that in mib tfi91 mutants. All panels are dorsal views with anterior to the left

Full size image

We further injected 0.1 pmol and 0.5 pmol su(h)-MO into wild-type embryos and compared their effects on the disruption of rhombomere boundaries. Under both conditions, interfaces between rhombomeres could be easily identified at 17 hpf (Fig. 2e, f; n = 4), suggesting that the initiation of rhombomere boundaries were not affected in embryos injected with either dosage of su(h)-MO. However, by 22 hpf obvious cellular shape change was observed in r4 and r5 region and the interstitial space increased in the lateral part of hindbrain (Fig. 2g, h). Furthermore, the change in 0.5 pmol su(h)-MO-injected embryos was more severe than that in 0.1 pmol su(h)-MO-injected embryos. Next we fixed MO-injected embryos at different time points and processed for foxb1.2 in situ hybridization (Fig. 2i–k). While more than half the number of 0.1 pmol su(h)-MO-injected embryos displayed wild-type like foxb1.2 expression pattern at 21 hpf, all the embryos injected with 0.5 pmol su(h)-MO showed a change in foxb1.2 expression, though to a different degree. As we expected, by 26 hpf the expression of foxb1.2 was down-regulated in embryos injected with either 0.1 pmol or 0.5 pmol su(h)-MO. Moreover, the severity of the foxb1.2 expression changes in these injected embryos at 26 hpf was stronger than that of mib tfi91 mutants and was comparable to that of mib ta52b mutants (Fig. 2m, n). Thus, these observations suggested that the severity of the disruption in rhombomere boundaries of su(h)-MO-injected embryos is dosage-dependent.

Early studies have shown that Notch activation is repressed in hindbrain to a greater extent in mib ta52b mutants than in mib tfi91 mutants by using her4 as a Notch activation readout (Zhang et al. 2007b). her6, mouse Hes1 ortholog, is expressed in hindbrain and regulated by Notch signaling (Pasini et al. 2004). We used it to specifically test whether Notch is differentially compromised in hindbrain of different mib alleles. While her6 was moderately reduced in mib tfi91 mutants (Fig. 2o, p), it was greatly reduced in mib ta52b mutants (Fig. 2q), demonstrating that Notch activation in hindbrain is decreased to a lesser degree in mib tfi91 mutants when compared to mib ta52b mutants, which is consistent with the data from the su(h)-MO knockdown. her9, another Hes homologue, is expressed at high level in rhombomere boundaries and regulated by Notch signaling in pronephros (Ma and Jiang 2007). Similarly, her9 expression was reduced to a greater extent in mib ta52b mutants when compared to mib tfi91 mutants (Fig. 2r–t).

Partial rescue of rhombomere boundaries in hdac1 hi1618;mib ta52b double mutants

Previous studies have shown that hdac1 is required for the promotion of neurogenesis through repression of Notch-activated target genes (Cunliffe 2004). We next questioned whether the blockade of neurogenesis by hdac1 deficiency in mib ta52b mutants can restore the boundaries. Identification of hdac1 hi1618 mutants, mib ta52b mutants, and hdac1 hi1618;mib ta52b double mutants was confirmed both by PCR analysis of viral insertion and sequencing PCR products flanking the exon 21 of mib gene (Fig. 3a, b). At 24 hpf, expression level of huC, a pan-neural marker, was dramatically increased in mib ta52b mutants (Fig. 3c (2)). As HuC antibody staining has revealed that r5 is a region where neurons mature much earlier than other rhombomeres (Fig. 1o), thus, it was not surprising to discover that the expression of neuronal markers, such as deltaA and ngn1, was enhanced in the hindbrain but not r5 region of mib ta52b mutants (Fig. 3d (2), e (2)). Similar to rfng, expression of another boundary marker, foxb1.2, was greatly reduced in mib ta52b mutants (Fig. 3f (2), g (2)). In hdac1 hi1618 mutants, expression levels of huC, deltaA and ngn1 were decreased to a great extent when compared to wild-type embryos (Fig. 3c (3), d (3), e (3)). However, the expression level of boundary markers, rfng and foxb1.2, was slightly reduced, though the pattern remained unperturbed, in hdac1 hi1618 mutants (Fig. 3f (3), g (3)). In hdac1 hi1618;mib ta52b mutants, the expression level of huC was restored as compared to mib ta52b mutants (Fig. 3c (1), c (2), c (4)). Similarly, the expression level and segmental pattern of deltaA and ngn1 were also restored, though not completely, in double mutants when compared to mib ta52b mutants (Fig. 3d (1, 2, and 4), e (1, 2, and 4)). To determine whether the inhibition of neurogenesis in hdac1 hi1618;mib ta52b double mutants could rescue the boundary disruption in mib ta52b mutants, expression of rfng and foxb1.2 was checked. The expression of rfng and foxb1.2 was increased and high expression level was observed in the middle region of rhombomere boundaries (Fig. 3f (4), g (4)). These results suggested that upon inhibition of neurogenesis, boundary cells maintain their undifferentiated state.

Fig. 3
figure 3

Loss of Hdac1 function in the mib ta52b mutants results in partial rescue of the rhombomere boundary defects. a Genotyping hdac1 hi1618, mib ta52b, and hdac1 hi1618;mib ta52b embryos. Schematic drawing shows the insertion site of the viral vector in hdac1 gene. The positions of the two pairs of primers used for genotyping the mutant embryos are indicated and their PCR products are shown in the gel image. Lanes 1, 3, 5, and 7 use primer set indicated in blue arrowheads while lanes 2, 4, 6, and 8 use primer set indicated in red arrowheads. Red arrowheads represent the primer pair for checking the viral insertion in homozygous hdac1 hi1618 mutants and hdac1 hi1618;mib ta52b double mutants; blue arrowheads for checking the wild-type hdac1 gene in the wild-type and mib ta52b embryos. b In mib ta52b mutants and hdac1 hi1618;mib ta52b double mutants, sequence analysis reveals that a T to G transversion caused an amino acid change from methionine to arginine. c–g Representative expression changes of marker in the mib ta52b, hdac1 hi1618, and mib ta52b;hdac1 hi1618 mutants. In hdac1 hi1618 mutants, reduction in the expression level of c huC, d deltaA, and e ngn1 indicated the inhibition of the neurogenesis when compared to wild-type embryos. Note that in mib ta52b mutants, neuronal hyperplasia resulted in the severe disruption of rhombomere boundaries as indicated by the expression of boundary markers f rfng or g foxb1.2. Moreover, the boundary disruption in double mutants was milder than that of the mib ta52b mutants. Note that all embryos were at 24 hpf stage and anterior to the left

Full size image

Heat-shock-induced dn-XSu(H) expression beginning before 8 s stage results in rhombomere boundary disruption

Rhombomere boundaries are transient structures during the hindbrain development. Previously, our results have shown that the expression of rfng is high at 24 hpf (Qiu et al. 2004) and it will gradually disappear. We only observed low level of rfng expression by 30 hpf (Fig. 1r) and no rfng expression was detected by 36 hpf. Next, to determine the time period required for Notch activation in rhombomere boundary maintenance, we used Tg(hsp70:XdnSu(H)myc) vu21 transgenic line (Latimer et al. 2005) to conditionally block Notch signaling. Heat shock beginning at tail bud (tb) to15 ss led to transgenic embryos with neurogenic phenotypes similar to the Notch deficient embryos by checking huC expression level (n = 16/16, 10/10, and 9/9 for tb, 8 ss and 15 ss, respectively; Fig. 4e, h, k; compared to the control in b). However, upon further investigation, we found that embryos heat-shocked at tb (n = 8/8) and 8 ss (n = 6/6) show a more significant reduction in her4 expression than those heat-shocked at 15 ss (n = 9/9; Fig. 4f, i, l; compared to the control in c). These results suggested that Notch activation was blocked to a greater extent in embryos heat-shocked at tb and 8 ss. Consistently, the rfng expression was attenuated significantly in embryos heat-shocked at both stages, indicating a blockade of Notch activation starting from tb (n = 10/10) to 8 ss (n = 11/11) was adequate to disrupt the boundaries by 24 hpf (Fig. 4d, g). Embryos heat-shocked after 15 ss (n = 23/23) only showed neurogenic phenotype but the boundary marker was robustly expressed as wild-type embryos at 24 hpf (Fig. 4j). We thus conclude that hindbrain boundary cells lose sensitivity to a blockade of Notch activation between 8 ss to 15 ss.

Fig. 4
figure 4

Temporal requirement of Notch signaling for the maintenance of rhombomere boundaries at segmentation stage. a–c Wild-type and d–l Tg(hsp70:XdnSu(H)myc) vu21 embryos heat-shocked at d–f tail bud stage, g–i 8 ss, and j–l 15 ss. Note that in transgenic embryos heat-shocked at d tail bud stage and g 8 ss, the expression of boundary marker rfng is decreased considerably compared to a wild-type heat-shocked embryos, whereas the expression level of rfng is almost unaffected in the transgenic embryos heat-shocked at j 15 ss. In all Tg(hsp70:XdnSu(H)myc) vu21 heat-shocked embryos; e, h, k huC expression level is greatly increased, whereas her4 expression is down-regulated to various extent in f, i, l, indicating that the neuron overpopulation in heat-shocked embryos results from the loss of Notch activation. Note that her4 expression in the posterior part of hindbrain is diminished to a greater extent compared to that in the anterior part of all the f, i, l heat-shocked Tg(hsp70:XdnSu(H)myc) vu21 embryos. All embryos are at 24 hpf and all panels are dorsal views with anterior to the left

Full size image

Knockdown of Notch3 function in notch1a mutants results in a loss of rhombomere boundary cells and neuronal hyperplasia

Previous studies have shown that notch1a, notch1b, and notch3 are expressed in the hindbrain (Westin and Lardelli 1997). Therefore, it is of interest to identify the receptors involved in the maintenance of rhombomere boundary cell fate. To determine this, we injected published notch1b or notch3 morpholino into notch1a th35b mutants. The specificity of notch3-ATG-MO has been confirmed as previously published (Fig. 5a; Lorent et al. 2004). Interestingly, notch1b and notch3 morphants, similar to notch1a (deadly seven) mutants, did not show a rhombomere boundary phenotype. Similar to mib ta52b mutants, we found that cells aggregate and organize into rosette-like structures in the posterior hindbrain in notch3-MO-injected notch1a th35b mutants (Fig. 5b, c; n = 4). Furthermore, in notch3-MO-injected notch1a th35b mutants, we observed a great reduction of boundary cells by using rfng or foxb1.2 as rhombomere boundary markers (Fig. 5d–g; n = 18/20 for e, n = 22/23 for g). By contrast, we did not observe a loss of rhombomere boundary cells in the notch1b-MO-injected notch1a th35b mutants (data not shown). Consistent with the loss of Notch activation, we observed that her4 was obviously down-regulated in notch3-MO-injected notch1a th35b mutants at 15 hpf (n = 12/12, Fig. 5j, k). In addition to the boundary disruption, we also noted that the neural keel was irregular and the hindbrain appeared highly disorganized in notch3-MO-injected notch1a th35b mutants (data not shown). These studies implied that Notch activation through two redundant receptors, Notch1a and Notch3, is required for the fate maintenance of rhombomere boundary cells by regulating their differentiation timing. Thus we further analyzed the neuron formation in notch3-MO-injected notch1a th35b mutants and found that cells in the hindbrain are differentiated into neurons as mib ta52b mutants (Fig. 5i; n = 18/18). The dramatic increase in huC-positive cells further confirmed the neurogenic phenotype in the notch3-MO-injected notch1a th35b mutants (Fig. 5 m; n = 31/31). These results suggested that Notch1a and Notch3 play a redundant role in the late differentiation of rhombomere boundary cells.

Fig. 5
figure 5

Loss of Notch3 function in notch1a th35b mutants results in rhombomere boundary disruption and neuronal hyperplasia. a The notch3 ATG-MO can specifically block translation of notch3 expression constructs (containing ATG site) in vitro by TNT reactions in a dose-dependent manner. b, c Confocal microscopic images at hindbrain levels by using bodipy ceramide staining technique at 24 hpf. Note that interfaces between rhombomeres (yellow arrowheads) can be clearly identified in uninjected embryos, whereas the interface between r4 and r5 becomes obscure in notch3-MO-injected notch1a th35b mutants. Also the irregular cellular shape changes can be easily found in r5 region (white arrow). d–i In situ hybridization for boundary marker d, e rfng and f, g foxb1.2, and for proneural marker h, i ngn1 at 24 hpf. Note that the expression level of rfng or foxb1.2 is greatly reduced while the expression level of ngn1 is greatly increased, indicating that the loss of boundary cells in notch3-MO-injected notch1a th35b mutants results from early differentiation of boundary cells. In situ hybridization for a Notch activation readout marker j, k her4 and postmitotic neuronal marker l, m huC. Note that her4 was down-regulated in notch3-MO-injected notch1a th35b mutants after loss of Notch activation at 15 hpf. Also note that more huC-positive cells were present, indicating the neurogenic phenotype in notch3-MO-injected notch1a th35b mutants at 24 hpf. All panels are dorsal views with anterior to the left

Full size image

Discussion

Su(H)-dependent Notch activation is required for the maintenance but not initiation of rhombomere boundaries by suppressing premature neuronal differentiation

Rhombomere boundary cells can be easily identified at 17 hpf in su(h)-MO-injected embryos or mib mutants. Therefore it is unlikely that Su(H)-dependent Notch activation is required for the initiation of hindbrain boundary. Consistent with this, weak rfng expression was detected in the rhombomere boundaries in mib ta52b mutants at 16.5 hpf (Cheng et al. 2004). It has been shown that some residual Notch activation could be observed in mib mutants or su(h) morphants by using her4 as a Notch-activated target gene (Itoh et al. 2003; Zhang et al. 2007b). This observation suggests that the residual Notch activation might be through a Su(H)-independent signaling that requires a cytoplasmic RING finger protein, Deltex (Le Borgne et al. 2005). However, knockdown of a hindbrain-expressing Deltex, deltex1 (DQ497599), did not lead to detectable hindbrain defects. Though the su(h)-MO we used targeting two known su(h) genes (Echeverri and Oates 2007; Zhang et al. 2007b), it still remains possible that maintenance of rhombomere boundaries is entirely through the Su(H)-dependent Notch activation, as morpholino injection may be unable to completely knock down su(h).

How does the boundary get lost? One possibility is that cell intermixing between rhombomeres is mis-regulated and hence the boundary cells cannot be distinguished. However, the expression of krox-20 is still very prominent at 20 hpf and is only greatly reduced by 24 hpf (Bingham et al. 2003). Yet we observed that the boundary cells become indiscernible from 17 hpf in mib ta52b mutants. In mib tfi91 mutants, krox-20 expression is normal at 24 hpf and therefore cell intermixing is less likely accounted for the disruption of lateral rhombomere boundaries. We also observed reduced cell proliferation in the hindbrain region from 20 ss in mib ta52b mutants. However, when we blocked the cell proliferation in wild-type embryos we found that the expression of boundary marker was normal. Therefore, reduced cell proliferation is unlikely to be a cause of disruption of hindbrain boundaries.

An alternative possibility is the premature differentiation of boundary cells into neurons. Detection of a pan-neuronal marker huC revealed that ectopic neurons in 24 hpf embryos fills almost the whole hindbrain (Fig. 3c (2); Cheng et al. 2004). Also, we found that Hu-positive cells were increased to a great extent in r5 region at 17 hpf in mib ta52b mutants. This correlates well with the earliest loss of r4/5 and r5/r6 boundaries. It has been shown that CBF1, a mammalian Su(H) ortholog, binds to the HDAC1-containing SMRT corepressor complex before its Notch-dependent switch from a transcriptional repressor to an activator (Kao et al. 1998). Studies in zebrafish further revealed that hadc1 is required for repressing Notch activation and the abundance of ash1b and ngn1 transcripts is dramatically reduced at 26 hpf in the hindbrain of hdac1 morphants (Cunliffe 2004). Indeed, we observed that boundary cells were partially rescued in hdac1 hi1618;mib ta52b double mutants.

There are several interpretations for the different severity between mib ta52b and mib tfi91 mutants. Recent reports have shown that Mib can interact with Mib2 and injecting mib2-MO can further enhance the mib tfi91 phenotype (Zhang et al. 2007a). Additionally, we observed that wnt1 expression was disorganized in mib ta52b mutants while in mib tfi91 mutants the wnt1 expression pattern was still similar to that of wild-type embryos (data not shown). Thus we speculated that Wnt signaling was affected to a great extent in mib ta52b mutants. Loss/reduction of Notch activation together with the reduction of Wnt signaling results in the severe disruption in rhombomere boundaries in mib ta52b mutants. Consistent with this interpretation, heat shock-driven wnt1 expression in mib ta52b mutants partially rescued the hindbrain boundary phenotype (Riley et al. 2004). Moreover, we observed that n-cad was significantly reduced in mib ta52b mutants and this might partially explain why cellular shape changes in mib ta52b were more severe when compared to the mib tfi91 mutants. However, it remains unanswered why the medial part of rhombomere boundary is more robust than in the lateral part.

Hindbrain boundary formation displays a low sensitivity to the level of Notch activation

notch1a/des, deltaD/aei, and deltaC/bea mutants show somite defect and mild neurogenic phenotype (Gray et al. 2001; Holley et al. 2000; Jülich et al. 2005), whereas mib mutants exhibit somite abnormality, strong neurogenic phenotype, and disruption of rhombomere boundaries. Previous reports have indicated that her4 can be used as a readout of Notch activation and we found that her4 expression is normal in notch1a mutants, while its expression is greatly reduced in mib ta52b mutants (Takke et al. 1999; Zhang et al. 2007b). In addition, apart from her4 expression level, expression level of her6 and her9 was also higher in mib tfi91 mutants than that in mib ta52b mutants. Consistent with this, the boundary disruption in mib tfi91 is milder than that in mib ta52b mutants. Furthermore, the higher the su(h)-MO dosage was injected, the more severe the boundary defect was observed. These results suggest that hindbrain boundary has a lower functional threshold requirement for Notch activation than somite segmentation and neurogenesis, since only the embryos with severely compromised Notch activation show all the phenotypes, including rhombomere defects. Previously, it has been shown in mice that a low threshold for Notch activity is required to maintain proper somite segmentation (Huppert et al. 2005). Taken together, it seems that hindbrain boundary formation has the lowest threshold among all the known Notch-dependent processes.

While Tg(hsp70:XdnSu(H)myc) vu21 embryos heat-shocked after 15 ss maintained rfng expression at 24 hpf, embryos heat-shocked at 8 ss impaired its expression at hindbrain boundaries. Therefore, we conclude that boundary cells lose sensitivity to reduced Notch activation between 8 to 15 ss. Previous studies have suggested that boundary markers are up-regulated at 16–17 hpf, which is later than the appearance of morphological boundaries. It is also observed that the neurogenic phenotype of Tg(hsp70:XdnSu(H)myc) vu21 embryos heat-shocked after 15 ss is comparable to that of notch3-MO-injected notch1a th35b mutants. Therefore, it is unlikely that the presence of boundaries in these embryos heat-shocked after 15 ss is due to insufficient reduction in Notch signaling. Thus, our results suggested that the rhombomere boundaries became stable only after the Notch controlling timing of cell differentiation in boundary cells is properly executed.

Notch1a and Notch3 function redundantly in maintaining rhombomere boundary cell fates

notch1a, notch1b, and notch3 were expressed in the entire hindbrain and the expression level was reduced in mib ta52b mutant embryos. Previous studies have shown that both Notch1 and Notch3 have the same effects on astroglial development and HES-1 expression (Tanigaki et al. 2001). Overexpression of activated murine Notch1 and Notch3 in transgenic mice blocks mammary gland development and induces mammary tumors (Hu et al. 2006). However, the defects of Notch1 signaling accelerate the differentiation of pancreatic endocrine cells, while overexpression of the intracellular region of Notch3 induces the same phenotypes (Apelqvist et al. 1999). Biochemical analysis has further revealed that Notch3IC can compete with Notch1IC for binding to the RBP-Jκ and thus acts as a repressor of Notch1IC-mediated HES activation (Beatus et al. 1999). These observations seem controversial but can be reconciled by the explanation that different tissues/organs have their specific multi-potent progenitors or have their specific repertoire of co-activators or co-repressors. In zebrafish hindbrain, our observation that notch3-MO-injected notch1a th35b mutants displayed strong neurogenic phenotypes (Fig. 5i, m), suggesting that notch1a and notch3 play a redundant role in the neuron differentiation. Therefore, these results indicate that both Notch1a and Notch3 activation are required for the maintenance of boundary cells in undifferentiating state at segmentation stage. Our recent studies on zebrafish kidney and ionocyte formation demonstrated that Notch1a and Notch3 play a redundant role in these organs as well (Ma and Jiang 2007; Hsiao et al. 2007), suggesting that a similar mechanism is utilized in the hindbrain. However, it remains unclear what is the exact function of Notch1b. One possible scenario is that its activation is involved in other developmental events rather than neurogenesis.

deltaA, deltaB, deltaD, and jagged1a are segmentally expressed in hindbrain (Haddon et al. 1998; Appel and Eisen 1998; Dornseifer et al. 1997; Zecchin et al. 2005). However, deltaA insertional mutants (hi781 and hi840), deltaD mutants (tr233), and their double mutants have normal rfng expression, which suggests that deltaA and deltaD are not involved in boundary cell differentiation or that the failure of exhibiting a phenotype is due to the redundancy of deltaB and/or jagged1a.

It has been shown that mouse Hes1 is expressed at high level in boundary regions and at low level in non-boundary regions (Baek et al. 2006). Similarly, we identified a Hairy and Enhancer of split homolog in zebrafish, her9, and found that it also shows persistent high expression level in the hindbrain boundary. However, knockdown of her9 alone did not result in the loss of boundary cells. Although Hes1-null mice did not exhibit defects in nervous system, the inter-rhombomere boundaries are initiated normally at E9.5 but become ambiguous at E10.5 in Hes1;Hes5 double-null embryos. In Hes1;Hes3;Hes5 triple-null mice the boundary defects become more significant, indicating that Hes1, Hes3, and Hes5 are all involved in regulating the boundary formation (Hatakeyama et al. 2004). In agreement with this, her3 (mouse Hes3 ortholog), her4, her6 (mouse Hes1 ortholog), her8a, her9, and her12 (mouse Hes5 homolog) have been identified to be expressed in the hindbrain (Hans et al. 2004; Takke et al. 1999; Pasini et al. 2001; Leve et al. 2001; Gajewski et al. 2006). Therefore, we speculate that similar to mice some her genes alone or in combination may play a role in maintaining boundary cells in an undifferentiated state. This warrants future study.