nlz Gene family is required for hindbrain patterning in the zebrafish
Whitehead Institute for Biomedical Research, Nine Cambridge Center, Cambridge, Massachusetts
Department Anatomy and Developmental Biology, University College London, Gower St., London WC1E 6BT, UK and Department of Pediatrics and Child Health, Royal Free and University College Medical School, The Rayne Institute, 5 University Street, London WC1E 5JJ, UK
The embryonic vertebrate hindbrain ultimately gives rise to the cerebellum, pons, medulla, and cranial nerves. In both zebrafish and Xenopus, ontogeny of the hindbrain and the entire nervous system begins at the onset of gastrulation, when neural determination and patterning have begun (reviewed in Gamse and Sive, 2000). By mid-gastrula, a broad posterior domain of gene expression demarcates the presumptive hindbrain (Kolm and Sive, 1997; Sagerström et al., 2001), and at this stage, transplantation assays indicate that hindbrain determination has begun (Woo and Fraser, 1998). Further regionalization of the hindbrain along the anteroposterior (A/P) axis during late gastrula and early somitogenesis stages establish lineage-restricted domains that correspond to the future rhombomeres, seven transient bulges in the hindbrain that are manifested at late somitogenesis stages. Correct specification of rhombomeres is critical for development of segment-specific neurons and normal brain function.
Multiple genes and signaling systems are required to establish rhombomere identity (reviewed in Moens and Prince, 2002). Retinoic acid and its receptors are required for transformation of the caudal hindbrain from r4 identity into more posterior identities (White et al., 2000; Gavalas and Krumlauf, 2000; Dupe and Lumsden, 2001). In zebrafish, FGF signaling, and the Variant hepatocyte nuclear factor-1 transcription factor together promote formation of r5–r7 fates, while the Pou2 transcription factor is required for normal development of all rhombomeres (Sun and Hopkins, 2001; Hauptmann et al., 2002; Maves et al., 2002; Walshe et al., 2002; Wiellette and Sive, 2003). A trimeric complex formed by the homeodomain transcription factors Meis3, Hoxblb, and Pbx4 potentiates caudal hindbrain fates (Vlachakis et al., 2001; Choe et al., 2002), while Pbx proteins are required for specification of r2–r6 (Waskiewicz et al., 2002). Conversely, the Gbx2 transcription factor is required for formation of rostral but not caudal hindbrain in mice (Wassarman et al., 1997). Readout of rhombomere size and identity is dependent on activation of downstream Hox genes. For example, the Krox20 transcription factor is required for formation of r3 and r5, by altering expression of downstream Hox genes (Schneider-Maunoury et al., 1998). Rhombomere identity is also though to require sharp borders that appear progressively and, through preventing cell mixing, contribute to distinct rhombomere fate (reviewed in Cooke and Moens, 2002).
Several genes that may be involved in the early steps of zebrafish hindbrain formation were identified in a subtractive cloning screen carried out in our lab (Sagerström et al., 2001). These include the nlz1 gene (previously known as nlz or noz1) encoding a zinc finger protein expressed in a dynamic pattern from early gastrula, in both ectoderm and mesendoderm (Sagerström et al., 2001; Andreazzoli et al., 2001; Runko and Sagerström, 2003). nlz1 is related to the Drosophila nocA and elB genes, which are important in nervous system determination and trachea formation (Cheah et al., 1994; Dorfman et al., 2002). A very recent study has characterized Nlz1 as a transcriptional repressor, which when expressed ectopically results in loss of r2 and r3 gene expression (Runko and Sagerström, 2003). We present here a second zebrafish nlz gene, nlz2, and define an nlz gene family that is conserved in the vertebrates. Our data indicate that one function of the nlz gene family is to specify rhombomere 4 identity and to limit the growth of rhombomeres 3 and 5 during hindbrain formation.
nlz Gene Family in Zebrafish
Having isolated the nlz1 gene, we attempted to characterize its function by generating loss-of-function morphants by using morpholino-modified antisense oligonucleotides. However, as described below, attempts to inhibit nlz1 function alone produced unconvincing phenotypes. We hypothesized that another related gene may have redundant function with nlz1 and, therefore, searched for additional zebrafish nlz genes. A degenerate reverse transcriptase-polymerase chain reaction (RT-PCR) approach was used, resulting in isolation of a new gene similar to nlz1, which was named nlz2. Nlz2 is 40% identical to Nlz1 at the level of conceptual protein translation. Several conserved domains are present in putative Nlz proteins, including conserved stretches near the amino terminus (Fig. 1A). The carboxyl terminal halves of the two zebrafish Nlz proteins include a putative C2H2-type zinc-finger domain. These zinc fingers contain unconventional spacing of eight residues between cysteines, as opposed to the more commonly observed one to five residues (Hollemann et al., 1996).
The predicted sequences for zebrafish Nlz1 and Nlz2 proteins were used to search for homologs of nlz genes in expressed sequence tag (EST) and genome databases, including those of the zebrafish. Two nlz-like genes were found in human as well as in mouse (Fig. 1A). In addition, three different Xenopus ESTs and two chicken ESTs were also identified as nlz homologs (data not shown). Of the two nlz homologs in human, the gene located on chromosome 8 is more similar to the zebrafish nlz1. Its product, FLJ14299, has a 60% overall identity with Nlz1, and only 40% with Nlz2. The gene located on human chromosome 10 is more similar to Nlz2 (Fig. 1B). Its product, MGC2555, has more than 80% identity with Nlz2, but only 45% with Nlz1. The grouping of nlz genes into nlz1 and nlz2 categories was based on overall sequence similarity. Positionally conserved single or double residues that define the nlz1 or nlz2 assignment are scattered throughout the putative protein sequences. These conserved residues do not define obvious functional motifs, and their significance is presently unknown.
Originally nlz1 was named for its similarity to the fruit fly nocA gene (Sagerström et al., 2001). Another Drosophila gene homologous to nocA, elB (elbow), has been described recently (Dorfman et al., 2002). The putative proteins derived from vertebrate nlz genes and Drosophila nocA and elB share two features. First, these proteins include a single, unconventional C2H2-type zinc finger motif, and second, all include the tetrapeptide sequence FKRY, a putative Groucho-interacting motif (Paroush et al., 1994; Dorfman et al., 2002; Fig. 1A).
Because zebrafish have undergone genome duplication in the recent past (Prince, 2002; Smith et al., 2002), we asked how many copies similar to nlz1 or nlz2 are contained in the genome. Degenerate RT-PCR, searches in EST and genomic fragment databases all suggest that nlz1 and nlz2 are the only nlz family members in zebrafish and that these genes are not duplicated, as is the case for some other zebrafish genes (Prince, 2002). In contrast, in the teleost Fugu rubripes four nlz genes are present, two with higher overall similarity to the zebrafish nlz1 gene, two with higher similarity to nlz2 (data not shown). This finding is consistent with the finding that the Fugu genome is duplicated for other genes (Gajewski and Voolstra, 2002). Only a single nlz1 or nlz2 homolog is found in the human and mouse genomes. In summary, nlz1 and nlz2 define a gene family present in diverse vertebrates.
Comparison of nlz1 and nlz2 Expression Patterns
To understand whether nlz1 and nlz2 might have different functions, the expression pattern of each was examined by whole-mount in situ hybridization (ISH). The expression pattern of nlz2 overlaps with that of nlz1 (for nlz1 expression pattern, see Andreazzoli et al., 2001; Sagerström et al., 2001), with some significant differences. Unlike nlz1, nlz2 mRNA is maternally deposited (as demonstrated by RT-PCR and Northern blot analysis, not shown), and expressed throughout the blastula-stage embryo (Fig. 2A,B). By 40% epiboly (blastula), there is a region of intense staining in the marginal zone (Fig. 2B). Gastrula stage expression of nlz2 and nlz1 is similar, with expression exclusively posterior, and no expression in the dorsal midline (Fig. 2C,D). By 90% epiboly (late gastrula), the anterior limit of nlz2 expression is slightly more posterior than that of nlz1 (Fig. 2D). Expression is otherwise circumferential but much stronger dorsolaterally than ventrally and entirely posterior to that of otx2, the posterior extent of which marks the future midbrain–hindbrain boundary (MHB; Fig. 2E). nlz2 is expressed in both the ectoderm and the mesendoderm, similar to nlz1 (not shown; Sagerström et al., 2001). There is a gap between both otx2 and pax2a expression and nlz2 expression at 95% epiboly, suggesting that, during gastrulation, nlz2 expression is limited to the posterior hindbrain and is not expressed in the anterior hindbrain (Fig. 2E,F). In addition, at tail bud stage, nlz2 expression lies posterior to the early stripe of krox20, which marks presumptive r3 (Fig. 2G).
By tail bud stage, nlz2 expression is detectable in a region anterior to the presumptive MHB (star in Fig. 2G, marked by staining for pax2a), in marked contrast to nlz1 expression, which is present only posterior to the MHB. During early somitogenesis, two domains of nlz2 expression persist anterior to the MHB (Fig. 2H) and anterior to nlz1 expression (Fig. 2I). In contrast to nlz1, nlz2 expression is barely detectable in the anterior hindbrain or tail bud at the 5 somite stage (Fig. 2I). The most anterior domain of nlz2 expression lies in the presumptive diencephalon, posterior to the domain of six3 expression marking the presumptive telencephalon (Fig. 2J). A more posterior domain includes and extends anterior to the pax2a expression domain and, therefore, marks the midbrain and MHB (Fig. 2J). Additionally, by the eight-somite stage, nlz2 is expressed in rhombomere 3 (r3), but not r2, of the presumptive hindbrain, and in a more posterior expression domain with its anterior limit at r5 as marked by comparison with krox20 staining (Fig. 2J). Expression of nlz2 in the MHB and the anterior domains persists at the eight-somite stage (Fig. 2K), whereas the expression of nlz1 in the MHB diminishes over time (Fig. 2L). At 25 hr postfertilization (hpf), based on morphology, nlz2 is expressed in forebrain, midbrain, caudal hindbrain, and anterior spinal cord (Fig. 2M), whereas nlz1 expression extends throughout the length of the spinal cord as well as being present in forebrain regions (Fig. 2N). Thus, while expression of nlz2 and nlz1 largely overlaps suggesting some shared function, the distinct expression of these genes in certain regions suggests also that these genes may have unique functions (Fig. 2R).
To ask whether nlz gene expression patterns are phylogenetically conserved, we isolated an nlz family cDNA clone from Xenopus laevis. Because this is not a full-length clone, we were unable to assign it as an nlz1 or nlz2 homolog. However, like nlz1 and nlz2 in the zebrafish, Xnlz is expressed initially in an exclusively posterior domain, with a gap at the dorsal midline (Fig. 2O). The anterior limit of Xnlz expression at this stage (st. 14, early neurula; Nieuwkoop and Faber, 1994) is in r3, as seen by overlap with krox20 expression (not shown). Later, at tail bud stages (st. 20), this gene is expressed in both MHB and hindbrain; it is also expressed anteriorly in eyes and forebrain, similar to that of zebrafish nlz2 (Fig. 2P,Q). These data demonstrate the evolutionary conservation of nlz gene expression patterns.
Regulation of nlz1 Expression in noi Embryos
Development and function of the isthmic (MHB) organizer requires the mutually dependent functions of fgf8, eng1/2, wnt1, and pax2 (reviewed in Martinez, 2001; Rhinn and Brand, 2001). Because nlz1 is expressed in the MHB during somitogenesis, we examined its expression in no isthmus (noi) mutants, which lack pax2a expression (Lun and Brand, 1998). At the end of gastrulation and the beginning of somitogenesis, nlz1 expression in the MHB of the mutant embryos is not detectable (Fig. 3A,B). In contrast, its expression in the r3 territory is not altered. The altered nlz1 expression in noi mutants does not recover as the embryos age (Fig. 3C,D). Expression of eng3 in the MHB is dependent on pax2a expression (Lun and Brand, 1998). Because nlz1 is expressed earlier than eng3, we asked whether injected nlz1 mRNA could restore eng3 expression in noi mutants. No rescue of eng3 expression was observed (Fig. 3E–G), indicating that eng3 is not regulated by nlz1 function independently of pax2a.
Misexpression of nlz Genes Disrupts Rostral Hindbrain Development
The expression patterns of nlz1 and nlz2 suggested that these genes play a role in posterior neural patterning. To test this, mRNAs encoding Nlz1 or Nlz2 proteins were injected into two-cell stage embryos and the phenotypes of injected embryos were examined at early somitogenesis (6–10 somites) by in situ hybridization (Fig. 4A). While a phenotype was seen after misexpression of native Nlz1 protein, no phenotype was seen after wild-type Nlz2 misexpression (see Fig. 4, and not shown). One possible reason for a lack of phenotype by Nlz2 is that the protein is unstable. We therefore constructed Nlz2 fusion proteins, to green fluorescent protein (GFP) and to the ligand-binding domain of the glucocorticoid receptor (GR; Hollenberg et al., 1993; Chalfie et al., 1994; Gammill and Sive, 1997) in the hope that these would stabilize Nlz2. In the case of embryos injected with nlz2:GR, the fusion protein was activated by the addition of dexamethasone (dex) at mid-gastrula stage (see Experimental Procedures section).
As shown in Figure 4B, expression of krox20 marking the presumptive r3 territory at the six-somite stage was greatly reduced by the injection of nlz1 RNA into two cell embryos (100 pg injected, total number of embryos examined in two independent experiments n = 156, 76% affected, compare panels a and b; see also Table 1). A concomitant anterior shift of r5 was observed with loss of r3. Nlz2 fusion proteins also gave a similar phenotype when expressed: injection of nlz2:eGFP mRNA (200 pg injected, n = 48, 88% affected, compare c and d) and nlz2:GR (200 pg injected, n = 31, 58% affected, compare e and f) each result in severe reduction or loss of r3 krox20 expression. While it is possible that the fusion proteins have gained a novel activity, it is more likely that the presence of the fusion stabilized the protein and allowed it to function as Nlz2 alone would.
Table 1. Overexpression of nlz1 Disrupts Hindbrain Patterning by Early Somitogenesis
A total of 100 pg of nlz1 mRNA was injected into each embryo, along with lacz mRNA as lineage tracer. Embryos were harvested at the two- to eight-somite stages, and gene expression analyzed by in situ hybridization as described in the Experimental Procedures section (see Fig. 3). Only embryos expressing lacZ− in the appropriate A/P domain were scored for changes in neural-specific gene expression.
Given are the number of embryos expressing injected mRNA at the appropriate A/P domain. At least two independent experiments were done to assay each marker.
Hindbrain, r3 (krox20)
Hindbrain, r5 (krox20)
Embryos injected with nlz1 mRNA were studied in more detail (Fig. 4C). As judged by expression of ephrin B2a, which marks presumptive r1, r4, and r7 (Fig. 4Ca), expression in these territories was not inhibited by overexpression of nlz1 RNA, although expression of krox20 in the r3 territory was ablated. The width of the r4–r7 territories was not altered by nlz1 misexpression; however, an anterior shift of more posterior rhombomere territories was observed (Fig. 4Ca). By using hoxa2 as a marker of r2+r3, it was apparent that gene expression in the r2 territory was also reduced by nlz1 mRNA injection (Fig. 4Cb). Consistent with the data in Figure 4Ca, expression of hoxb1a, which marks r4, did not expand with r3 ablation, but was shifted anteriorly (Fig. 4Cc). Expression of the r5 marker, krox20, and the r5+6 marker valentino were not ablated under the experimental conditions generally used (Fig. 4Cd). Some of the injected embryos (28% of total embryos) showed a slight posterior expansion of the MHB pax2a expression (Table 1). However, embryos showing pax2a expansion and inhibition of r3 gene expression did not necessarily occur in the same embryos. The response to injected nlz1 mRNA was dose dependent, because at the relatively low amounts injected (100 pg), ablation of r2 and r3 was the only phenotype observed. When a higher amount (400 pg) of nlz1 mRNA was injected, krox20 expression in the r5 territory was also reduced; however, reduction of r5 krox20 expression was never observed without concurrent loss of krox20 expression in r3 (not shown). These data show that overexpression of Nlz proteins inhibit formation of presumptive rhombomeres in the developing hindbrain, with r2 and r3 most sensitive.
To investigate when hindbrain patterning can be altered by overexpressed nlz1, we constructed and tested the activity of a dex-inducible fusion protein, Nlz1:GR (Table 2). After injection of the RNA into two cell embryos, Nlz1:GR was activated by dex addition at different developmental stages. The fusion protein had some constitutive activity, since some repression of r3 krox20 was observed without dex addition. Maximal disruption of hindbrain patterning was observed when ectopic Nlz1 function was activated during gastrulation, but could occur when the protein was activated as late as tail bud stage. The above data suggest that the anterior hindbrain is sensitive to ectopic Nlz1 protein during mid- and late gastrula stages, well before krox20 expression has begun in r3. Because it is not clear whether the Nlz1:GR fusion protein persisted as late as one-somite stage, it is not possible to accurately interpret the lack of effect on r3 krox20 expression when dex is added after tail bud stage.
Table 2. Effect of Nlz1:GR Fusion Protein on the Development of Zebrafish Hindbrain
No. of embryos with r3 krox20 expression altered (% of total embryos)c
A total of 50 pg of nlz1:GR mRNA was injected into each embryo. Dexamethasone (dex) was added at various time points, and embryos were left in dex until collection.
Embryos were collected at 6- to 10-somite stages for in situ hybridization analysis. Markers used: six3, pax2a, krox20. Each number represents the result of at least two independent experiments.
Rhombomere 3 (r3) krox20 expression repressed or eliminated; in some embryos, r5 expression was affected.
No dex added
Loss of Nlz Function Disrupts Hindbrain Patterning
To test the requirement for nlz gene function in hindbrain development, morpholino-modified antisense oligonucleotides (MOs) were used to reduce the levels of functional Nlz1 and Nlz2 proteins. Initially, single MOs, directed against the translational start sites or splice donor sites of nlz1 and nlz2 were tested. Although correct splicing of nlz1 or nlz2 was specifically inhibited by at least 50% when 2 ng of MO was injected per embryo (see Experimental Procedures section), and the incorrectly spliced RNAs are not predicted to produce functional Nlz proteins, no phenotype was observed after injection of 8 ng of individual MOs (not shown). Therefore, mixtures of MOs, targeting either a single nlz gene or both nlz1 and nlz2 simultaneously, were used. Slight changes in hindbrain development were observed using a mixture of two or more MOs directed against either nlz1 or nlz2; however, substantial changes were observed only when both genes were targeted simultaneously by injection of 8 ng total of an equimolar mixture of four oligos, two targeting nlz1, and two targeting nlz2 (4MO mix). We have noted that the severity of the hindbrain phenotype appears to be dependent on the amount of MO injected, although we have not analyzed this effect extensively. MOs were injected into one-cell embryos, and embryos examined at various stages during somitogenesis.
Embryos fixed at the five-somite stage were examined for expression of krox20, which is normally expressed in presumptive r3 and r5, and for hoxB1a, which is normally expressed in presumptive r4. While these domains of expression are unaffected when a control MO is injected (Fig. 5Aa; 13 of 13 embryos), hoxB1a expression is almost entirely ablated when the nlz 4MO mix is injected (Fig. 5Ab; 20 of 21 embryos). HoxB1a expression is replaced by expanded krox20 expression such that krox20 forms a nearly consistent band from the anterior of r3 to the posterior of r5. Similar results are observed at the 18-somite stage. In two independent experiments injecting nlz 4MO mix, hoxB1a expression is nearly ablated, while krox20 expression expands into the putative r4 domain (Fig. 5Ac,d; 22 of 22 embryos). There is a small amount of hoxB1a expression remaining, which appears to be in the ventral part of the neural tube (Fig. 5Af), although hoxB1a expression is normally distributed evenly through the neural tube (Fig. 5Ae).
In addition to the loss of hoxB1a expression within the r4 domain, embryos injected with the nlz 4MO consistently show enhanced hoxB1a expression in paired domains lateral and posterior to the krox20 domain (Fig. 5Ad and f, arrows). It is not clear whether this expression is in the mesoderm or ventral-most neural tissue.
To describe more accurately some of the changes in the hindbrain in response to injection of the nlz 4MO, measurements of the length and width of regions of the hindbrain were made. Injected embryos were fixed at 18-somite stage and in situ hybridized with probes for pax2a, which labels the MHB and otic placodes, and for hoxB4, which labels posterior neural tube up to the r6/r7 boundary (Fig. 5Ba, b). Measurements were made of the anterior hindbrain, from the MHB to the otic placode, the length of the otic placode, and the posterior hindbrain from the end of the otic placode to the hoxB4 boundary, as well as the entire region, from the MHB to hoxB4 (Fig. 5Ba). While the domains anterior and posterior to the otic placodes were the same average size whether injected with the control or nlz 4MO, the length of the otic placode was significantly longer in the nlz 4MO-injected embryos (Fig. 5Be; Table 3). Measurements of the width of the hindbrain, as judged by the distance between the otic placodes, showed a broader distance in the nlz 4MO-injected embryos at the anterior edge of the otic placodes, although both nlz 4MO and control MO-injected embryos show the same width at the level of the posterior otic placode (Fig. 5Be; Table 3). Changes in otic placode length are not surprising, because otic placodes are induced in epidermal tissue overlying r5 and parts of r4 and r6, and hindbrain patterning mutants routinely cause defects in otic vesicle development (Whitfield et al., 2002). Therefore, expansion of r5 is likely to lead to expansion of the region of epidermis induced to contribute to the otic placode.
Table 3. Effect of nlz 4MO Injection on Hindbrain Sizea
Anterior HB (A)
Otic placode (B)
Posterior HB (C)
Total HB (D)
Anterior width (E)
Posterior width (F)
krox20 length (G)
krox20 width (H)
HB, hindbrain; 4MO, equimolar mixture of four morpholino-modified intisense oligonucleotides. Letters correspond to those used in Figure 5: A, hindbrain from the midbrain/hindbrain boundary to the anterior edge of the otic placode; B, length of the otic placode; C, hindbrain from the posterior edge of the otic placode to the hoxB4 (r6/r7) boundary; D, hindbrain from the midbrain/hindbrain boundary to the hoxB4 boundary; E, width between the otic placodes at the anterior edge; F, width between the otic placodes at the posterior edge; G, length from the anterior edge of r3 to the posterior edge of r5, as marked by krox20 expression; H, width across the hindbrain at about the r4 level. Measurements are given in micrometers (μm). Bold font pairs of measurements are those that showed significant change in size in response to the nlz 4MOs, as evaluated by Student's t-test using P < 0.005.
162 ± 27
75 ± 11
36 ± 11
278 ± 42
127 ± 11
130 ± 13
127 ± 9
110 ± 12
167 ± 19
115 ± 15
34 ± 10
315 ± 27
168 ± 15
132 ± 19
174 ± 21
129 ± 18
In addition, measurements of the total size of the krox20+hoxB1a expression domain were made (Fig. 5Bc), and these measurements show an increase in both A/P length and width of this central hindbrain region when nlz 4MO is injected (Fig. 5Be; Table 3). Taken together, these measurements show that, when injected with nlz 4MO, r3 and r5 expand at the expense of the r4 domain, and in addition, and possibly as a result of the identity transformation, there is overgrowth of the central hindbrain both laterally and posteriorly (Fig. 5Bd).
nlz Genes Regulate Development of r3, r4, and r5
Loss-of-function data suggest that nlz1 and nlz2 are redundantly required to specify or maintain r4 identity and to limit the size of r3 and r5, functions that are possibly interdependent. However, the gain-of-function results show that neither gene alone is sufficient to respecify any region of the hindbrain to r4 identity, although either gene is able to repress development of r2 and r3. While the nlz genes have a significant role in determining r4 identity, they also may play a direct or indirect role in restricting the size of rhombomeres: reduced nlz function results in overgrowth of the central hindbrain, whereas ectopic nlz expression results in a shortened anterior hindbrain.
Injection of MOs directed against both nlz genes results in nearly complete loss of the r4 identity by the five-somite stage, with concomitant expansion of r3 and r5, and we have observed expansion of krox20 expression in these embryos as early as the two-somite stage (data not shown). The early expression patterns of both nlz genes shows that they are expressed in the future posterior hindbrain domain throughout gastrulation, but by the beginning of somitogenesis are expressed specifically in r3 and the posterior neural tube starting at r5. Taken together, these data suggest that the role of the nlz genes is to promote r4 specification during gastrulation and to limit r3 and r5 growth during somitogenesis. Overexpression and timed activation of nlz1:GR shows that ectopic Nlz1 can inhibit r2 and r3 formation during gastrulation, and this inhibition may reflect a normal role in limiting rhombomere growth at later stages of development, when the nlz genes are expressed in r3 and r5.
Our experiments showed that ectopic expression of nlz genes was not able to transform other regions of the hindbrain toward r4 identity, a result which is in contrast to data from Runko and Sagerström (2003), who observe mild expansion of r4 markers upon injection of similar amounts of nlz1 mRNA. We have observed distortion of r4 in mid-somitogenesis stage embryos after ectopic nlz expression, but have not observed r4 expansion. This minor discrepancy may result from differences in experimental conditions, a possibility that we have not thoroughly investigated.
When interpreting the data generated by injection of MOs, it is important to note that antisense techniques do not usually lead to complete suppression of gene expression and, therefore, complete loss of function of either gene alone may uncover nonredundant activities. It is interesting that the nlz genes appear to have such a localized function, but such complex expression patterns. As well as r3 and r5, nlz1 and nlz2 are also expressed in r6 and r7, and nlz1 in r2, but it is not presently clear whether nlz loss of function affects these rhombomeres. In the MHB, despite regulation by noi, gene expression appears fairly normal after nlz loss of function; however, this region has not been carefully examined, and nlz function may play a more subtle role in MHB development at later stages. Other phenotypes may be apparent upon closer examination of nlz loss-of-function embryos. In this regard, we noted that the neural plate is wider in nlz loss-of-function embryos, suggesting that nlz gene may regulate convergent extension movements.
Our loss-of-function data generated by injection of MOs results in a phenotype distinct from that generated by injection of a dominant negative form of the Nlz1 protein (Runko and Sagerström, 2003). Expression of a dominant-negative protein (dnNlz1) drives expansion of r3 and r5, but in this case, results in overlapping expression of hoxB1a (r4) and krox20 (r3, r5). This phenotype may be a milder version of our MO-generated phenotype or may result from dnNlz1 interfering in different processes than those disrupted by the MOs.
Relationship of nlz Gene Activity to the Hierarchy of Hindbrain Patterning Genes
It is interesting to speculate how nlz1 and nlz2 may fit into the hierarchy of genes known to be involved in hindbrain pattern formation. The expression of both genes during gastrulation indicates that the nlz family may be required at this time, when other genes such as hoxB1b, hoxB1a, meis3, and vhnf1 are expressed in a similar domain (Alexandre et al., 1996; Sagerström et al., 2001; Sun and Hopkins, 2001). These genes are also required for specification and growth of individual rhombomeres in the posterior hindbrain, so it is possible that the nlz genes are functioning in parallel or in cooperation with these other posterior factors during gastrulation (McClintock et al., 2002; Wiellette and Sive, 2003). Despite their early, broad expression domain, the nlz genes have specific effects on individual rhombomeres in both gain- and loss-of-function analyses, which may indicate temporally or spatially limited roles for these genes. This may be a reflection of later nlz function when the expression domains are limited or may suggest that there are obligate cofactors involved in nlz function, such as krox20, which is first expressed early during somitogenesis and only in r3 and r5.
Mechanism of Nlz Function
The activity of Nlz proteins may correspond to their activity as transcriptional repressors. The Drosophila Nlz-related protein ElB has been shown to interact in vitro with Groucho, a transcriptional corepressor (Nibu et al., 2001; Dorfman et al., 2002; Lepourcelet and Shivdasani, 2002). Vertebrate Nlz proteins also contain the putative Groucho-interacting domain, and the ability of zebrafish Nlz proteins to inhibit rhombomere-specific gene expression is consistent with the notion that these are transcriptional repressors. Furthermore, Nlz-Gal 4 DNA binding domain fusion proteins do not increase reporter construct expression when injected into embryos, indicating that these proteins do not have transcriptional activator activity (not shown). Recent data from Runko and Sagerström (2003) show that Groucho does bind to Nlz1 and that Nlz1 appears to function as a repressor.
Nlz proteins are likely to carry out their function by interacting with other localized factors. This function was suggested by the region-specific nlz gain- and loss-of-function phenotypes, including the heightened sensitivity of r3 to gain or loss of nlz function, compared with other rhombomeres. One intriguing candidate for genetic interaction with Nlz factors is the pou2 gene, which also affects r3 and r5 development (Hauptmann et al., 2002). In contrast to nlz loss of function embryos, embryos that have mutations in pou2 have decreased r3 and r5 territories, with an anterior expansion of r4. Analysis of the functional partners of nlz genes will determine how their activity interfaces with that of other genes regulating brain development.
Fish Stock Maintenance
Wild-type AB strain and mutant Danio rerio were maintained according to established protocols (Westerfield, 1995). Embryos were kept at 28.5°C.
Degenerate RT-PCR and the Isolation of nlz2 cDNA Clones
Degenerate PCR reactions were carried out with primers ND1, 5′-CGGAATTCGARTAYYTNCARCC-3′; ND2, 5′-CCGCTCGAGGGYTTNCCDATYTG-3′; ND3, 5′-CGGAATTCACNTAYGGNTTYATG-3′; and ND4, 5′-CCGCTCGAGCKYTTRTCRCANGG-3′. The PCR program consisted of 4 cycles with 42°C annealing temperature, followed by 45 cycles with 55°C annealing temperature. PCR products were gel purified and cloned by using the TOPO TA cloning kit (Invitrogen).
Primers YPT26 (5′-GCCATTGCCATCCACCCCGG-3′) and YPT27 (5′-CACATGCGGAAGAGGGTCG- 3′) were designed according to the sequences of fragments obtained from the degenerate RT-PCR reactions. They were used to amplify a 1,078-bp fragment using late gastrula zebrafish cDNA as template. This fragment was used to generate radioactive probes, and a late gastrula cDNA library was screened. None of the several clones isolated from this library, however, contains the 3′-end of the nlz2 cDNA. Primer YPT28 (5′-GACCCTCTTCCGCATGTGTG-3′) was coupled with a vector primer to amplify the missing sequence. The entire cDNA sequence is identical to that listed in public databases (AY371081).
Isolation of Xnlz
Zebrafish nlz1 cDNA was used to screen a st. 11.5 Xenopus whole embryo cDNA library made by Catherine Nocente (Kuo et al., 1998). Three independent clones were isolated that contained fragments of the same cDNA, which we named Xnlz. The longest clone obtained was 1 kb (sequence submitted to GenBank), and this was used for in situ hybridization.
DNA and peptide sequences were analyzed and stored with both DNA Strider and DNAStar software. Homology searches were conducted on-line at www.ncbi.nlm.nih.gov/BLAST. The zebrafish genome sequence database is accessible at http://danio.mgh.harvard.edu. Unigene database is available at the NCBI Web site. Peptide line-up was performed with the MegAlign software from DNAStar, using the Clustal IV method.
Fusion Protein Construction
To construct nlz2:eGFP, first, a point mutation was introduced into the nlz2 coding region at its unique NotI site. This strategy results in an Ala to Gly missense mutation, at amino acid residue number 385. Then, the mutated nlz2 coding sequence was amplified by PCR and cloned into pCS2(+) plasmid as a ClaI–XhoI fragment. The eGFP fragment was purified from the pEGFP-1 plasmid (Clontech) as a XhoI-blunt end fragment and cloned downstream of the nlz2 coding sequence at the XhoI–SnaBI sites. The resulting plasmid pYT879 can be linearized with NotI and mRNA generated with SP6 RNA polymerase.
nlz2:GR fusion gene was constructed similarly. The GR domain (Gammill and Sive, 1997; Hollenberg et al., 1993) was amplified by PCR using JH155 (5′-GCGGCGCTCGAG- TCTGAAAATCCTGG-3′) and JH156 (5′-GAGGAGTCTAGATCACTTTTGATGAAAC-3′) primer pair.
To construct nlz1:GR, primers JH124 (5′-CGGAATTCCATGAGCGA- ACTGCCTCCTGGA-3′) and JH125 (5′-CGCCTCGAGCTGGTACCCAA- GAGCAGAAG-3′) were used to amplify the entire nlz1 opening reading frame from a cDNA clone. The PCR product was cloned into vector CS2+MT (Roth et al., 1991) as an EcoRI–XhoI fragment. The GR domain was then cloned downstream of the nlz1 open reading frame.
Antisense Morpholino-Modified Oligonucleotides
Antisense morpholino-modified oligos used were as follows. YPT-M1, 5′-CAGTTTTCATTTAGAGATATCCAGT- 3′, targeting the 5′-UTR region of nlz1; MO2, 5′-CAGTTTGCATTCAGA- AATATCTAGT-3′, similar to YPT-M1, but with four bases changed: Nlz1, 5′-AGAAGTCGTACCTCAATGCTCAC- GG-3′, targeting the splice donor site of nlz1; Nlz2ATG, 5′-ATGACCCAATTCTCATGTATTTTGT-3′, targeting the translation start site of nlz2; Nlz2SP, 5′-CATTCTTACCTCAATTGGACTGACC-3′, targeting the splice donor site of nlz2; and YPTM2, 5′-CAGTTTGCATTCAGAAATATCTAGT-3′, which has four bases modified from the YPTM1 sequence, was used as negative control.
To test the efficacy of antisense morpholino oligos targeting the splice donor site of the nlz genes, an RT-PCR approach was used. After 2 ng of either Nlz1 or Nlz2SP was injected into one- to four-cell stage zebrafish embryos, the embryos were allowed to develop to mid-gastrula stage. RNA was extracted from the injected embryos, and primer sets spanning the singular intron in nlz1 or nlz2 genes were used to amplify regions of the pre-mRNA or mature mRNA. For nlz1, primers YPT31, with the sequence 5′-ACAAGTCACATACTTCACCC-3′, and YPT32, 5′-CAGGTCTGAGCCAGCAA- TGC-3′, were used to amplify a 106-bp region on the mature mRNA or an 858-bp region on the pre-mRNA. For nlz2, primers YPT25, 5′-GGAGCATGTTTGCGCAAG-3′, and YPT26, 5′-GCCATTGCCATCCACCCCGG-3′, were used to amplify an 82-bp region on the mature mRNA, or a 474-bp on the pre-mRNA.
To establish the significance of changes in the size of the hindbrain after injection of MOs, all embryos were flat-mounted and imaged at high magnification. Prints of these images were measured, and the average size and standard deviation for each measurement were determined. A Student's t-test was used to determine the significance of the differences in the average measurements, using P < 0.005.
RNA was prepared for microinjection as previously described (Grinblat et al., 1998). For nlz1 mRNA, 100–400 pg was injected into each embryo at the two-cell stage. For nlz2:GR and nlz2:eGFP mRNA, 200 pg was injected. For nlz1:GR mRNA, 50 pg was injected. lacZ or GFP mRNA was included to ensure that injected mRNA reached the neurectoderm. After injection of mRNA from GR fusion constructs, the injected embryos were staged under a dissecting microscope, and at appropriate time points, some embryos were moved to a Petri dish containing Embryo Buffer (Westerfield, 1995) with 4 μg/ml of dexamethasone added. For morpholino oligos, a total of 8 ng was injected into each embryo.
In Situ Hybridization
ISH was performed as previously described (Grinblat et al., 1998). RNA probes for ISH were labeled with either digoxigenin or fluorescein and detected by alkaline phosphate-conjugated antibodies (Roche).
We thank members of our lab for comments on the manuscript, particularly Vince Tropepe, and Hakryul Jo. Thanks to Charles Sagerström for communicating results before publication. We also thank Amanda Dickinson for help with photography, George Bell for help with sequence alignments, Ben Pratt for fish husbandry, and Vladimir Apekin for frog husbandry. H.L.S. and E.L.W. were funded by the NIH, J.H. was funded by the Wellcome Trust Fellowship, and F.W. was funded by a Herman and Margaret Sokol Fellowship in Biomedical Research. Some of the work presented here was carried out by J.H. in the laboratory of Stephen W. Wilson, and we appreciate his support.