Zebrafish Lmx1b.1 and Lmx1b.2 are required for maintenance of the isthmic organizer

The cerebellum and tectum develop in response to signals emanating from an organizer located at the isthmus, a constriction of the neural tube that forms between the mesencephalic and metencephalic vesicles(Joyner, 1996; Puelles et al., 1996; Wassef and Joyner 1997; Liu and Joyner, 2001). The organizing activity of the isthmus has been demonstrated by numerous classical embryonic manipulations (Nakamura et al.,1986; Alvarado-Mallart et al.,1990; Gardner and Barald,1991; Itasaki et al.,1991; Martinez et al.,1991; Bally-Cuif et al.,1992; Marin and Puelles,1994; Martinez et al.,1995). Several important signaling molecules and transcription factors have been shown to mediate isthmic organizer (IsO) function. These molecules coordinate the development of cells within the mes-metencephalic region (MMR), as well as provide necessary autocrine functions to maintain the IsO. Defining the signaling pathways that direct MMR cell fates and determining the mechanisms by which these pathways are regulated remain long-term goals for understanding IsO development.

The IsO is positioned at the juxtaposition of the otx2 and gbx2 expression domains (Simeone et al., 1992; Millet et al.,1996; Wassarman et al.,1997; Niss and Leutz,1998; Shamim and Mason,1998; Li and Joyner,2001). Coincident with this border is the expression boundary of two major signaling molecules, wnt1 and fgf8. During IsO maintenance, wnt1 is expressed at the caudal edge of the mesencephalic vesicle (Wilkinson et al.,1987; Bally-Cuif et al.,1992; Kelly and Moon,1995; Hidalgo-Sanchez et al.,1999), and wnt1 mutant mice fail to maintain a number of mesencephalic and metencephalic structures(McMahon and Bradley, 1990; Thomas and Capecchi, 1990). Although Wnt1 signaling is necessary for MMR development, ectopic Wnt1 does not appear to be sufficient to globally change the fate of MMR cells(Adams et al., 2000). By contrast, isthmic fgf8 expression is refined to the rostral edge of the metencephalic vesicle (Heikinheimo et al., 1994; Crossley and Martin, 1995; Riefers et al., 1998), and is necessary and sufficient to mediate IsO function (Brand et al., 1996; Meyers et al.,1998; Riefers et al., 1998). In fact, ectopic Fgf8 induces changes in gene expression and morphology strikingly similar to transplantation of isthmic tissue (Crossley et al.,1996; Funahashi et al.,1999; Martinez et al.,1999; Shamim et al.,1999). IsO regulation also requires a number of transcription factors that work in a coordinated fashion, including members of the Pax family (Brand et al., 1996; Favor et al., 1996; Torres et al., 1996; Lun and Brand, 1998; Pfeffer et al., 1998), the Engrailed family (Millen et al.,1994; Wurst et al.,1994) and Lmx1b.

Lmx1b is a LIM-homeodomain protein whose role in IsO patterning has only been addressed in gain-of-function studies. Originally identified as a regulator in dorsoventral limb patterning(Riddle et al., 1995; Vogel et al., 1995; Chen et al., 1998), Lmx1b has recently been shown to be required for dopaminergic and serotonergic neuron development in vertebrates (Smidt et al.,2000; Cheng et al.,2003). We originally reported that Lmx1b was expressed in the chick MMR and, using a retroviral approach, demonstrated that it was sufficient to maintain the expression of Wnt1 in the mesencephalon(Adams et al., 2000). More recently, Matsunaga et al. (Matsunaga et al., 2002) used an electroporation approach in the chick to demonstrate that Lmx1b induced wnt1 cell-autonomously and fgf8 non-cell-autonomously. However, direct evidence of a requirement for Lmx1b in IsO function has been lacking.

To further elucidate transcriptional regulation of the IsO, we have extended these studies to the zebrafish. The zebrafish provides a powerful means of studying the genetic basis of IsO formation and function. IsO regulation appears largely conserved among vertebrates, and the relative ease of gain- and loss-of-function experiments in zebrafish allows for a number of developmental studies not possible in the chick. Mutants of several major IsO genes are available in zebrafish, including pax2.1 (no isthmus) (Lun and Brand,1998) and fgf8 (acerebellar) (Riefers et al.,1998). Study of these mutants and other developmental studies in fish have contributed to our understanding of a regulatory feedback loop that maintains IsO patterning function.

We report the isolation and functional analysis of lmx1b.1 and lmx1b.2, two zebrafish orthologs of lmx1b. Loss- and gain-of-function studies indicate that these two closely related transcription factors have redundant functions in maintaining gene expression and cell fate at the IsO, and that these genes are necessary and sufficient for maintenance of wnt1 and fgf8 expression. Pax2.1 is required for maintenance of lmx1b.1 and lmx1b.2 at the IsO, and Lmx1b.1 and Lmx1b.2 are required for pax8 maintenance. We propose a model in which Lmx1b.1 and Lmx1b.2 cooperate with Pax, Wnt and Fgf genes to maintain the IsO.

Zebrafish were raised under standard laboratory conditions as described before (Mullins et al., 1994). All injections were performed with wild-type TL embryos. Developmental stage was determined according to Kimmel et al.(Kimmel et al., 1995). No isthmus (noi) heterozygous mutants were verified by random intercrosses and homozygous mutant embryos were produced by crossing heterozygous parents.

Based on the published sequence of hamster(German et al., 1992) and human (German et al., 1994)Lmx1a, and human (Iannotti et al.,1997), mouse (Chen et al.,1998) and chick (Tsuchida et al., 1994) Lmx1b, a single pair of degenerate PCR primers was designed using the amino acid sequences indicated in Fig. 1A, that was predicted to amplify both lmx1a and lmx1b orthologs. Fragments resulting from PCR amplification of 24 hpf zebrafish cDNA were subcloned and two distinct species were isolated multiple times. Each insert was used to screen a 22-26 hpf zebrafish lambda zap cDNA library and a total of 38 purified positives was identified following screening of 1 million plaques. Multiple isolates of two distinct cDNAs were obtained, each ∼2 kb in length and containing 300-500 bp of untranslated sequence (data not shown). Although this degenerate PCR strategy was predicted to identify zebrafish orthologs of lmx1a and lmx1b, both cDNAs isolated were closely related to lmx1b (Fig. 1B,C). The sequences for Lmx1b.1 (AY894989) and Lmx1b.2 (AY894990) have been submitted to GenBank. We have subsequently identified a putative lmx1a ortholog in the zebrafish genome database, so the failure to identify this sequence in our screen suggests that expression levels at 22-26 hpf were too low. ClustalW alignments and phylogenetic analyses were performed using MacVector(Accelrys), and the phylogenetic tree was produced using the Neighbor Joining Method with bootstrapping (1000 replicates).

Digoxigenin- or fluorescein-labeled probes were generated from linearized templates using an RNA labeling and detection kit (Roche). Hybridization and detection with alkaline phosphatase-conjugated antibodies is described elsewhere (Odenthal and Nüsslein-Vollhard, 1998). Stained embryos were cleared in benzyl alcohol:benzyl benzoate (2:1), mounted in Canada balsam, and photographed using a Magnafire SP digital camera (Optronics). Alternatively,embryos were photographed in PBS using a Micropublisher digital camera(Qimaging). To obtain histological sections, embryos were embedded in JB-4 plastic resin (Polysciences). Sections (5 μm) were prepared with glass knives using a Leica RM2155 microtome. Sections were affixed to glass slides and stained with Methylene Blue/Azure II.

Antisense morpholino oligonucleotides were obtained from Gene Tools(Corvallis, OR). Morpholino sequences were CTTCGATTTTTATACCGTCCAACAT for lmx1b.1-MO (B1-MO) and CCTCAATTTTGATTCCGTCCAGCAT for lmx1b.2-MO (B2-MO). Mismatch oligonucleotides were designed for use as negative controls: CGTCTATTTTCATGCCGTCCATCAT for lmx1bN-MO (BN-MO)and CATCCATTTTAATCCCGTCCACCAT for lmx1bX-MO (BX-MO). An unrelated control oligonucleotide (CON-MO) was provided by the manufacturer. Morpholino microinjections were performed on 1- to 4-cell embryos. Morpholino solutions(4 mg/ml in Danieu buffer) were back-loaded into glass needles before injection, and 1.5-2.5 nl of morpholino solution were delivered to the yolk of each embryo. To verify morpholino specificity, in vitro translation analysis was performed using the TNT Coupled Transcription/Translation System (Ambion)with 0.5 μg of plasmid DNA and 0.5 μg of morpholino oligonucleotide. ³⁵S-methionine was incorporated and labeled proteins were resolved by 10% SDS-PAGE and visualized by autoradiography.

For immunohistochemistry, a mixture of two mouse monoclonal anti-GFP antibodies was used (Roche) and detected with biotinylated anti-mouse secondary antibody. Phosphohistone H3 protein was detected with a rabbit polyclonal antibody (Upstate) and biotinylated anti-rabbit antibody. Color reaction was carried out using Vectastain ABC Kit (Vector) and FAST DAB(Sigma). For TUNEL staining, the ApopTag Peroxidase Detection Kit (Chemicon)was used, but anti-digoxigenin-AP was substituted for anti-digoxigenin-HRP. Color reaction was carried out using NBT/BCIP.

A previously described transgenic line allows for the visualization of cranial motoneurons that express GFP under control of the Islet1 promotor(Higashijima et al., 2000). Transgenic embryos were injected with both morpholino oligos, and expression of GFP was observed in live embryos at 12-hour intervals beginning at 24 hpf. For images shown in Fig. 7,embryos were fixed at 48 hpf and processed for immunohistochemical visualization of GFP.

To block FGF signaling, tailbud stage embryos were incubated in an 8 μM solution of SU5402 (Calbiochem) in E3/DMSO for 4 hours. Control embryos were incubated in E3/DMSO alone.

To misexpress Lmx1b.1 and Lmx1b.2, cDNAs containing the complete coding regions were cloned into pBluescript SK^– plasmids downstream of the hsp70 promoter. Injection solutions were prepared at a concentration of 10 ng/μl in 0.2 M KCl and Phenol Red, and 1-3 nl of solution was injected directly into the cytoplasm of one-cell stage embryos. Embryos were then reared until early somitogenesis (1-5 somites). At that time, embryos were heat shocked at 37°C for 45 minutes. After heat shock,embryos were allowed to further develop for 2-29 hours before fixation.

Degenerate PCR primers were designed based on Lmx1a and Lmx1b sequence information from a number of species (see Materials and methods), and two distinct zebrafish orthologs of Lmx1b were isolated(Fig. 1A). Comparison of the putative amino acid sequences of these cDNAs, designated lmx1b.1 and lmx1b.2, with other Lmx proteins in the published databases(Fig. 1A,B) shows that both zebrafish cDNAs encode proteins that are most similar to chicken Lmx1b(Lmx1b.1 88% identical, and Lmx1b.2 80% identical). These sequences also exhibit a high degree of identity with mouse and human Lmx1b proteins. Comparison of the two zebrafish Lmx1b genes reveal a relatively high degree of divergence (Fig. 1B). Although mammalian Lmx1b orthologs are very highly conserved (97-99% identical),zebrafish Lmx1b.1 and Lmx1b.2 share only 77% identity. Somewhat lower identity is observed when comparing Lmx1a and Lmx1b sequences (64-68% identity),whether within a species or between distantly related species.

Although our cloning strategy did not yield a Lmx1a homolog, we have identified a putative zebrafish Lmx1a by performing a BLAST search of available zebrafish genomic sequences. Phylogenetic analysis supports the grouping of this sequence with other known Lmx1a proteins, while both zebrafish sequences obtained in our screen cluster with Lmx1b proteins(Fig. 1C).

We next determined when and where each gene was expressed. Although the focus of these experiments is on the developing CNS, both genes are also expressed elsewhere, including the developing limb field (data not shown). lmx1b.2 expression initiates first at the margin of 90% epiboly embryos (Fig. 2A). Soon thereafter, lmx1b.1 and lmx1b.2 are both expressed at the tailbud stage in chevron-shaped domains corresponding to the presumptive MMR(Fig. 2B,C). Both genes are also expressed in distinct, overlapping domains at the midline, but only lmx1b.1 is expressed in the presumptive otic anlage. Expression of lmx1b.1 and lmx1b.2 in the MMR becomes refined to a narrow ring by the five-somite stage (Fig. 2D,E). By the 13-somite stage, both lmx1b.1 and lmx1b.2 are strongly expressed in the MMR and in the ventral diencephalon (Fig. 2F,G). Additionally, lmx1b.1 is expressed in the developing otic placodes and the dorsal diencephalon, whereas lmx1b.2 is expressed in the dorsal midline of the hindbrain. The expression patterns of lmx1b.1and lmx1b.2 are thus dynamic and different in some aspects, but both are strongly expressed in the developing MMR.

Early expression of lmx1b.1 and lmx1b.2 overlaps extensively with pax2.1, a marker and regulator of the IsO. By the tailbud stage, the expression of lmx1b.1 at the presumptive MMR is contained entirely within the expression domain of Pax2.1(Fig. 3A). lmx1b.2 has an identical expression domain at the MMR (data not shown). By the 13-somite stage, pax2.1 expression at the IsO becomes refined to a ring that appears to encompass the expression domains of lmx1b.1 and lmx1b.2 (Fig. 3B,C),and extend further caudally.

At 24 hpf, lmx1b.1 and lmx1b.2 are expressed at the caudal limit of the mesencephalon that abuts the isthmic constriction in a pattern similar to the wnt1 expression domain(Fig. 3D-F). All three genes are also expressed in the dorsal midline of the mesencephalon and metencephalon. During IsO patterning, the MMR expression of lmx1b.1and wnt1 is contained within the anterior domain of pax2.1expression (Fig. 3G,H), as is lmx1b.2 (data not shown).

To determine what developmental events require Lmx1b.1 and Lmx1b.2, we performed a series of morpholino knockdown experiments. Antisense morpholino oligos were designed to specifically block translation of Lmx1b.1 (B1-MO) or Lmx1b.2 (B2-MO). An in vitro translation assay to asses oligo efficacy and specificity demonstrated that B1-MO effectively blocked only Lmx1b.1 translation, and B2-MO blocked only Lmx1b.2 translation. Mismatch control oligos (see Materials and methods) did not block translation of either(Fig. 4A). Injection of embryos with single morpholinos directed against either Lmx1b.1 or Lmx1b.2 does not produce a discernible morphological phenotype (data not shown), nor does injection of any of the three control morpholinos, although one control,BN-MO, causes a small degree of nonspecific toxicity. Embryos injected with morpholinos targeting both Lmx1b.1 and Lmx1b.2 (B1B2-MO) fail to maintain the isthmic constriction or cerebellum (Fig. 4). At 24 hpf, experimental and control embryos are morphologically similar, but the loss of the isthmic constriction and cerebellum in B1B2-MO-injected embryos is easily observed by 30 hpf (92%, n=60). The affected embryos subsequently develop circulatory problems and severe cardiac edema by 48 hpf and die soon thereafter.

Several genes have been identified that are expressed at the MMR and are necessary for MMR development. To determine if Lmx1b.1 and Lmx1b.2 loss of function affects gene expression in the MMR, we injected B1B2-MO and then determined the effects on the temporal and spatial expression of MMR genes. Based on the previously reported relationship between chick Lmx1b and Wnt1(Adams et al., 2000; Matsunaga et al., 2002), we first examined wnt1 regulation. Initiation and early expression of wnt1 is unaffected by knockdown of Lmx1b.1 and Lmx1b.2(n=18, Fig. 5A,B). However, wnt1 expression at the MMR is lost in B1B2-MO-injected embryos at the 15- to 18-somite stage (100%, n=37). By contrast, wnt1 expression persists in the midline and forebrain(Fig. 5C,D). Embryos injected with single morpholinos against Lmx1b.1 or Lmx1b.2 exhibit normal wnt1 expression as late as 24 hpf (data not shown). Knockdown of Lmx1b.1/2 also affects wnt3a and wnt10b expression (data not shown). At the 18-somite stage, wnt3a is normally expressed in the dorsal midline of the midbrain, with ventral extensions into the MMR. In embryos injected with B1B2-MO, the ventral extensions of wnt3aexpression at the MMR are missing at the 18-somite stage (100%, n=12). Wnt3a expression is normal in embryos injected with BX-MO (100%, n=13). Wnt10b is expressed at the MMR and the dorsal midline of the mesencephalon, hindbrain and spinal cord at the 18-somite stage, and its expression strongly resembles that of wnt1. Injection with B1B2-MO results in loss of wnt10b expression at the MMR at the 18-somite stage (100%, n=9), and injection of BX-MO does not affect expression of wnt10b (n=8).

The effect of Lmx1b.1 and Lmx1b.2 knockdown on Pax gene expression at the MMR was also examined. At the 15- to 18-somite stage, knockdown of Lmx1b.1 and Lmx1b.2 results in complete loss of pax8 expression (100%, n=10, Fig. 5E,F). Expression of pax2.1, pax2.2 and pax5 is unaffected until 24 hpf. Fgf8 expression was unaffected by morpholino injection at the 18-somite stage. However, knockdown of both Lmx1b.1 and Lmx1b.2 causes fgf8 expression at the MMR to fade by the 19- to 22-somite stage(93%, n=14, Fig. 5G,H). The loss of fgf8 expression in the MMR proceeds from ventral to dorsal. At the 19- to 22-somite stage in knockdown embryos, fgf8 is lost ventrally and consistently retained dorsally, but by 24 hpf, this dorsal expression is almost completely absent (data not shown).

The MMR deletion caused by knockdown of Lmx1b.1/2 affects cells at the isthmus and cerebellum, but not the tectum or ventral rhombomere 1 (R1). At 24 hpf, pax5 expression marks the caudal limit of the mesencephalon, the isthmic constriction and the cerebellar anlage(Pfeffer et al., 1998). Knockdown of Lmx1b.1/2 causes a strong reduction of pax5 expression at the MMR (100%, n=10), while injection of BX-MO has no effect(Fig. 6A,B). pax2.1(100%, n=36), pax2.2 (100%, n=12) and pax8(100%, n=10), markers of the isthmus but not the cerebellum, are strongly reduced by 24 hpf at the MMR following knockdown of Lmx1b.1/2 (data not shown). otx2 marks the mesencephalon, and its expression is neither deleted nor expanded in Lmx1b.1/2 knockdown embryos at 24 hpf(n=28) (Fig. 6C,D). Nor is an affect on otx2 expression observed after 36 hpf(n=18, data not shown). mab21l2 is an additional marker of the mesencephalon that is unaffected by knockdown of Lmx1b.1/2 (n=20)(Fig. 6E,F). epha4aexpression marks several regions of the developing CNS, including the region of R1 caudal and ventral to the cerebellum. epha4a expression is unaffected in Lmx1b.1/2 knockdown embryos at 24 hpf (n=20)(Fig. 6G,H). en3expression marks both the tectum and the isthmocerebellar region. Knockdown of Lmx1b.1/2 results in a partial deletion of en3 expression limited to the ventral isthmocerebellar region (50%, n=18)(Fig. 6I,J), while expression in the tectum and the dorsal isthmocerebellar region is unaffected. Therefore,Lmx1b.1/2 knockdown results in a loss of gene expression at the isthmocerebellar region, but not in the adjacent mecencephalon or rhombomeres. This suggests that anteroposterior patterning and specification is not altered in the absence of Lmx1b.1/2, but that cells fated to form the IsO and cerebellum are not maintained.

To determine the consequences of altered IsO gene expression that cause the deletion of recognizable isthmic and cerebellar structures, we examined cell death and cell proliferation. We performed TUNEL staining to visualize apoptotic cells. At 24 hpf, uninjected embryos contain a low level of apoptotic cells distributed uniformly throughout the CNS (100%, n=11)(Fig. 7A). Injection of the BX-MO control results in a low and variable increase in cell death throughout the embryo (100%, n=19) (Fig. 7B). By contrast, injection of B1B2-MO results in a dramatic increase in the number of apoptotic cells in the MMR (83%, n=24)(Fig. 7C). Interestingly, the region of increased apoptosis is centered caudal to the lmx1b.1/2expression domain. Conversely, in noi mutant embryos, cell death is visible at the same stages, but is centered more rostral to the isthmus (20%, n=20) (Fig. 7D). At 18 hpf (18 somites), embryos injected with B1B2-MO do not exhibit increased apoptosis around the isthmus, and therefore, the onset of increased cell death occurs between 18 and 24 hpf. In addition to examining cell death, we used antibodies against phosphohistone H3 (pH3) to determine if knockdown of Lmx1b.1/2 affected mitosis at the MMR. When embryos injected with B1B2-MO(n=35) were compared with embryos injected with BX-MO (n=20)at 24 hpf, no significant difference in pH3-positive cells at the MMR was observed (data not shown). Therefore, loss of isthmic and cerebellar identity in Lmx1b.1/2 knockdown embryos is associated with increased levels of apoptosis, but not altered cell proliferation.

We next examined the consequences of Lmx1b.1 and Lmx1b.2 loss of function on neuronal differentiation. Cranial motoneurons (CMNs) III and IV flank the isthmus, and thus serve as markers of neuronal differentiation at the MMR. A transgenic islet1-GFP line allows for the visualization of Islet1-expressing CMNs in the developing zebrafish brain(Higashijima et al., 2000). GFP expression in the cell bodies of these neurons was observed as early as 36 hpf. Live B1B2-MO-injected embryos were examined through 54 hpf, and CMNIII and CMNIV were never observed. CMNs were also visualized in fixed embryos using an antibody against GFP, which yielded better images of CMNs, and these data are presented in Fig. 7E,F. Islet1-GFP embryos were injected with BX-MO or B1B2-MO and fixed at 36, 48 and 72 hpf. Antibody staining for GFP revealed that while CMNs developed normally in BX-MO-injected embryos, CMNIII and CMNIV were not detected in B1B2-MO-injected embryos at any stage examined.

To test the requirement for Pax2.1 in lmx1b.1 and lmx1b.2expression, we performed whole-mount in situ hybridization on noimutant embryos (Table 1). At the one-somite stage, all embryos expressed lmx1b.1 (n=17)and lmx1b.2 (n=17) normally. However, at subsequent developmental stages (5, 7, 12 and 18 somites) lmx1b.1 and lmx1b.2 expression at the MMR was absent in ∼25% of embryos(59/234 and 64/214, respectively). This corresponds to the predicted one quarter of embryos homozygous for the pax2.1 mutation. Significantly,this loss of lmx1b.1 and lmx1b.2 expression occurs earlier than the loss of wnt1 (first affected at six somites) and fgf8 (first affected at nine somites) in noi mutant embryos(Lun and Brand, 1998). Therefore, Pax2.1 is not required for the initiation of lmx1b.1 or lmx1b.2 expression, but soon thereafter is required for their maintenance at the IsO in a manner independent of Wnt1 and Fgf8.

To test the dependence of lmx1b.1 and lmx1b.2 expression on FGF signaling, we treated embryos with SU5402, a drug that inhibits the function of the FGFR1-4 receptors, and effectively eliminates the signaling of FGF8 and other FGFs (Mohammadi et al.,1997). Exposure of embryos to SU5402 in a DMSO solution at tailbud stage resulted in a loss of lmx1b.1 and lmx1b.2 expression at the MMR within 4 hours, while expression at the midline was retained(n=15) (Fig. 8A,B). Expression of lmx1b.1 and lmx1b.2 was not affected in embryos treated with DMSO alone (n=15)(Fig. 8C,D). Therefore,maintenance of lmx1b.1 and lmx1b.2 expression at the MMR requires FGF signaling.

To determine which IsO genes are regulated by Lmx1b.1 and Lmx1b.2, we induced ectopic expression using constructs containing the lmx1b.1-or lmx1b.2-coding sequence under the control of the hsp70promoter. Embryos were injected with plasmid at the one-cell stage and were heat-shocked during early somitogenesis to induce expression of lmx1b.1 (Fig. 9A,B) or lmx1b.2 (data not shown). Ectopic expression was highly variable, and the number of expressing clones varied from one to hundreds. The Lmx-positive clones were distributed randomly throughout most tissues of heat-shocked embryos, but the cells surrounding the yolk seemed most sensitive to the treatment. Wild-type embryos were heat-shocked, resulting in no ectopic induction of lmx1b.1 (n=26) or lmx1b.2(n=20). Control embryos injected with hsp70-lmx1b.1, but not heat shocked, did not express ectopic lmx1b.1 (n=13) or wnt1 (n=12). Embryos injected with lmx1b.2-hsp70and not heat-shocked did not express ectopic lmx1b.2 (n=10)or wnt1 (n=19).

Both Lmx1b.1 and Lmx1b.2 induced ectopic expression of wnt1(Fig. 9C,D). Hsp70-Lmx1b.1 induced mosaic ectopic expression of wnt1 in 36% of embryos(n=141), and Hsp70-Lmx1b.2 induced wnt1 in 71% of embryos(n=17). Wnt1 expression was induced in all regions of the embryo. Both Lmx1b.1 and Lmx1b.2 also induced fgf8(Fig. 9E,F). Hsp70-Lmx1b.1 induced fgf8 in 81% of embryos (n=69) and Hsp70-Lmx1b.2 induced fgf8 in 96% (n=52).

As the Hsp70-Lmx1b.2 construct was more effective, we used it to test Lmx1b regulation of other IsO genes. We injected one-cell stage embryos with hsp70-lmx1b.2, heat shocked during early somitogenesis, and fixed embryos 2, 13 or 29 hours after treatment. Embryos were then examined for wnt1, fgf8, pax2.1 and pax8 expression. At each time point,ectopic expression of wnt1 (9/19, 4/14, 13/33) and fgf8(8/20, 8/21, 2/29) was observed. Although strong endogenous expression was detected, no ectopic expression of pax2.1 (0/21, 0/20, 0/20) or pax8 (0/20, 0/18, 0/26) was detected. Therefore, whereas Lmx1b.1 and Lmx1b.2 are sufficient to induce wnt1 and fgf8, Lmx1b.2 is not sufficient to induce pax2.1 or pax8.

In this study, we have identified two zebrafish orthologs of lmx1b,lmx1b.1 and lmx1b.2, and show that both genes are expressed at the isthmus and are necessary for maintenance of the MMR. The functions of Lmx1b.1 and Lmx1b.2 at the isthmus are redundant: both proteins must be knocked down to disrupt MMR development, and either gene is sufficient to induce wnt1 and fgf8. The requirement for Lmx1b.1 and Lmx1b.2 at the isthmus is likely a result of their roles in maintaining expression of wnt1, wnt3a, wnt10b, pax8 and fgf8. Here, we discuss the interaction of Lmx1b.1 and Lmx1b.2 with other transcription factors and secreted factors to maintain the function of the IsO(Fig. 10).

Loss of Lmx1b.1 and Lmx1b.2 function causes the degeneration of the isthmus and cerebellum by 30 hpf. This morphological effect is preceded by a substantial increase in cell death at the MMR between 18 and 24 hpf. It is possible that the primary role of Lmx1b.1 and Lmx1b.2 at the isthmus is to maintain cell survival in the isthmocerebellar region. Alternatively, Lmx1b.1 and Lmx1b.2 may play an indirect role in the cell death; loss of Lmx1b.1 and Lmx1b.2 may result in a failure of IsO autoregulation, which leads to loss of maintenance of downstream trophic factors necessary for cell survival.

The requirement of Lmx1b.1 and Lmx1b.2 for maintenance of cell survival is probably tied to their regulation of Wnt1 and Fgf8. Wnt1 has been shown to have a proliferative role in the CNS(Dickinson et al., 1994; Matsunaga et al., 2002; Megason and McMahon, 2002; Panhuysen et al., 2004), and Wnt1^–/– mice have ectopic cell death at the MMR (Serbedzija et al., 1996; Chi et al., 2003). Simultaneous loss of signaling from Wnt1, Wnt10b and Wnt3a results in increased apoptosis at the MMR (Buckles et al.,2004). Similarly, knockout of MMR fgf8 results in ectopic apoptosis in mice (Chi et al.,2003). In zebrafish, ace mutants lacking functional Fgf8 exhibit increased cell death concentrated above the rostral metencephalon from midsomitogenesis onwards, and a marked decrease in mitosis at the MMR at 57 hpf (Jászai et al.,2003). Cell death observed in the MMR of B1B2-MO-injected embryos could thus be attributed to the loss of wnt1/wnt3a/wnt10b and fgf8, and cell death may be initiated when cells fail to receive these essential proliferative signals.

A close regulatory relationship between Lmx proteins and Wnt1 at the IsO is conserved among vertebrates. Lmx1b can induce expression of wnt1 in the chick (Adams et al., 2000; Matsunaga et al., 2002). The gain-of-function experiments in the present study demonstrate that this regulatory relation is conserved in zebrafish, with wnt1 being regulated by both Lmx1b.1 and Lmx1b.2. Morpholino knockdown experiments demonstrate that Lmx1b.1 and Lmx1b.2 are required for maintenance of wnt1, but only at the IsO and only after the 18-somite stage. This function is preserved in the mouse, as Lmx1b-null mice fail to maintain wnt1 expression at the IsO beyond E9.5, and display a subsequent failure of MMR development (R. Johnson, personal communication).

As in the chick, the expression of lmx1b.1 and lmx1b.2 at the isthmus most closely resembles the expression of wnt1/10b during IsO development. Combined with our loss- and gain-of-function results, we conclude that Wnt gene expression at the IsO is fundamentally linked to Lmx1b.1 and Lmx1b.2 activity during this developmental period. Whether this transcriptional control is direct requires further study. Interestingly,knockdown of Wnt1 results in only a slight reduction of gene expression at the isthmus in zebrafish (Lekven et al., 2001), while simultaneous loss of Wnt1,Wnt10b and Wnt3a results in a complete loss of the isthmic constriction(Buckles et al., 2004). Consistent with these observations, Lmx1b.1 and Lmx1b.2 are required to maintain wnt1, wnt3a and wnt10b in the MMR and may regulate other Wnt genes.

A clear positive feedback relationship exists between Lmx and Pax genes at the IsO, but we are only beginning to understand the details of these interactions. pax2.1, pax2.2, pax5 and pax8 are not maintained at the isthmus in the absence of functional Lmx1b.1/2 activity, and early maintenance of lmx1b.1 and lmx1b.2 requires Pax2.1. The failure of pax2.1, pax2.2 and pax5 maintenance past 24 hpf in B1B2-MO-injected embryos is consistent with the hypothesis that loss of Lmx1b.1/2 function results in a breakdown of IsO signaling, and precipitates programmed cell death at the isthmus. By contrast, the loss of pax8expression by the 15-somite stage may reveal a more direct interaction between Pax8 and Lmx1b.1/2.

During IsO initiation, Pax, Fgf and Lmx gene expression is not yet interdependent, but by the mid- to late-somitogenesis stages, the IsO is regulated by a positive feedback loop in which these genes are mutually dependent for maintenance of expression. Regulation of pax8 by Lmx1b.1 and Lmx1b.2 may represent an intermediate stage in IsO development,between initiation and maintenance. As fgf8 is still expressed at the 19- to 22-somite stage in B1B2-MO-injected embryos, regulation of pax8 by Lmx1b.1 and Lmx1b.2 must occur via an Fgf8-indepentent mechanism that is yet to be defined. The early requirement for Pax2.1 in lmx1b.1/2 expression is further evidence of a Fgf8-independent regulatory pathway required for IsO development. Further supporting this idea,in noi mutant embryos lmx1b.1 and lmx1b.2 become dependent on Pax2.1 by the five-somite stage, while fgf8 is maintained until the nine-somite stage(Lun and Brand, 1998).

Although Lmx1b.1 and Lmx1b.2 are necessary for expression of multiple Pax genes, they do not appear to be sufficient. Ectopic Lmx1b.2 induced wnt1 and fgf8, but not pax2.1 or pax8. This result is surprising given that fgf8 is induced in this assay. As Fgf8 gain of function was sufficient to induce pax2.1 and other IsO genes in chick (Crossley et al.,1996), we expected that misexpression of Lmx1b.2 would result in induction of pax2.1, either directly or through Fgf8 induction. Failure to detect ectopic pax2.1 suggests that either the level of Fgf8 induced in this experiment was insufficient to subsequently induce pax2.1, or that, unlike in the chick, Fgf8 is not sufficient to induce pax2.1 in the zebrafish.

The results presented are consistent with a hypothesis that ascribes a polarizing activity to the IsO (Lee et al., 1997; Riefers et al., 1998; Picker et al., 1999). We propose that Lmx1b.1 and Lmx1b.2 participate by maintaining cell survival in the region just posterior to the lmx1b.1/2 expression zone. During the maintenance phase of the IsO, MMR gene expression is co-dependent. Loss-of-function for a number of individual transcription factors and secreted factors results in loss of expression of the other regulators, increased cell death and the degeneration of the MMR. However, loss-of-function for individual IsO regulators has differing consequences for anteroposterior patterning of the MMR. Although lmx1b.1 and lmx1b.2 are expressed more rostral to the isthmic constriction, knockdown of Lmx1b.1 and Lmx1b.2 results in increased apoptosis caudal to the constriction. However, pax2.1 is expressed in a domain that flanks the isthmus, but apoptosis in the noi homozygous mutants is concentrated rostrally and extends well into the midbrain. This indicates that both Lmx1b.1/2 and Pax2.1 have asymmetric, non-cell-autonomous effects on cell survival. This is probably because of their regulation of trophic factors. Although a number of factors may be involved, we hypothesize that maintenance of the MMR depends on asymmetric responses to signaling from the expression boundary of Fgf8 and Wnt1/3a/10b, and that Lmx1b.1/2 mediate this regulation by maintaining Wnt1. Electroporation studies in chick suggest that Lmx1b.1/2 may have an additional role. Matsunaga et al. (Matsunaga et al.,2002) have proposed that fgf8 was cell-autonomously repressed by Lmx1b, but induced non-cell-autonomously by Lmx1b through Wnt1/3a/10b. However, we have not observed a repressive effect of Lmx1b.1 or Lmx1b.2 on fgf8 in zebrafish.

Our results also indicate a fine degree of specificity in dorsoventral patterning at the IsO. Morpholino knockdown of Lmx1b.1 and Lmx1b.2 affects ventral expression of IsO genes more than dorsal expression. pax2.1,pax2.2 and fgf8 expression was often lost throughout the MMR,but in some cases, a variable amount of dorsal expression was retained. The observed pattern of dorsal retention of IsO gene expression is also seen in the fgf8 mutant acerebellar(Pfeffer et al., 1998), while ventral retention is seen in the pax2.1 mutant noi(Lun and Brand, 1998). These results indicate that maintenance of the IsO is controlled differently in dorsal and ventral domains, although there is as yet no known mechanism that regulates dorsoventral patterning of the MMR. Further studies of Lmx1b.1 and Lmx1b.2 function at the IsO may help answer this and other questions concerning patterning and cell fate decisions in the MMR.

We thank Drs G. Bellipanni and E. Weinberg for providing noiheterozygotes; Dr M. Granato for the cDNA library; Drs A. Lekven and K. L. Chow for providing plasmids; Drs J. Golden, P. Klein, and V. Hatini for helpful comments on the manuscript; and Dr R. Johnson for sharing unpublished results. A. Sebastian and S. Wong aided in preliminary expression and in vitro translation studies. This work was supported by a grant from the National Institutes of Health (NS041005).