Zebrafish husbandry and fish lines

Zebrafish (Danio rerio) were maintained on a 14-h light/10-h dark cycle at 28.5 °C. Embryos were obtained from natural paired and/or grouped spawnings of wild-type (WT) (AB, TL or AB/TL hybrid) fish or identified heterozygous lmx1bb ^jj410 , lmx1ba ^mw80 , evx1 ⁱ²³² ;evx2 ^sa140 or smoothened ^b641 mutant fish or Tg(slc17a6:EGFP) [formerly called Tg(vGlut2a:EGFP)] [33] or Tg(evx1:EGFP) ^SU1 [11] transgenic fish or lmx1bb ^jj410 crossed into the background of either Tg(slc17a6b(vglut2a):loxP-DsRed-loxP-GFP) ^nns14 [41, 42] or Tg(evx1:EGFP) ^SU1 fish respectively. Embryos were reared at 28.5 °C and staged by hours post fertilization (h) and/or days post fertilization (dpf). Most embryos were treated with 0.2 mM 1-phenyl 2-thiourea (PTU) at 24 h to inhibit melanogenesis [34–36].

The evx1 ⁱ²³² , evx2 ^sa140 and lmx1bb ^jj410 mutants have been previously described [11, 37–39]. All three of these mutations are single base pair changes that lead to premature stop codons before the homeobox. Therefore, if any of these RNAs are not degraded by nonsense mediated decay, the resulting proteins will lack the DNA binding domain. lmx1ba ^mw80 mutant zebrafish were generated using TALENs constructs that target the sequences TCAAGTAGACATGCTGGACG and TCCGCTCCTGTCCTGAACTG within the first exon of lmx1ba. Constructs were made using steps 1–38 outlined in [40]. To generate mRNA encoding the TALENs, approximately 5 μg of plasmid DNA was digested with ApoI and purified via the Invitrogen PureLink PCR Purification Kit (ThermoFisher, K310001). RNA was synthesized using the Ambion mMessage mMachine T7 kit (ThermoFisher, AM1344) with a poly(A) tail added from the Poly(A) Tailing Kit (Ambion, AM1350) and purified with the Megaclear Kit (Ambion, AM1908). 100 pg of RNA for each TALEN was co-injected into 1-cell WT embryos. The lmx1ba ^mw80 allele was recovered and identified as a single base pair deletion 20 bp into the coding sequence. This results in a frameshift after the first six amino acids and a premature stop codon 11 amino acids later. This stop codon is upstream of both the Lim and homeobox domains, suggesting that this allele is likely to be a complete loss of function.

Genotyping

DNA for genotyping was isolated from both anesthetized adults and fixed embryos via fin biopsy or head dissections respectively. Fin biopsy and evx1 and evx2 genotyping of adults were performed as previously described [11, 37]. KASP assays, designed by LGC Genomics LLC, using DNA extracted from head dissections, were used to identify embryos carrying the evx1 ⁱ²³² and evx2 ^sa140 mutations. These assays use allele-specific PCR primers which differentially bind fluorescent dyes that we quantified with a BioRad CFX96 real-time PCR machine to distinguish genotypes. The proprietary primers used are: Evx1_y32_i232 and Evx2_sa140.

Heads of fixed embryos were dissected in 80 % glycerol/20 % phosphate-buffered saline (PBS) with insect pins. Embryo trunks were stored in 70 % glycerol/30 % PBS at 4 °C for later analysis. DNA was extracted via the HotSHOT method [41] using 20 μL of NaOH and 2 μL of Tris-HCl (pH-7.5).

The lmx1ba ^mw80 and lmx1bb ^jj410 alleles were identified by restriction enzyme digestion assays as both of these mutations disrupt endogenous restriction enzyme sites. For lmx1ba ^mw80 , a 540 bp amplicon encompassing the mutation site was generated with the following primers:

Forward GATCCTCAAGAGGAGCTCATACACA and Reverse CATGCACATTTAACTATGATCTGAGCCGTG.

This amplicon was digested with MluCI to yield 311 bp and 142 bp and 87 bp (WT), 453 bp and 87 bp (homozygous mutant), or 453 bp and 311 bp and 142 bp and 87 bp (heterozygous mutant) products. Similarly, for lmx1bb ^jj410 , a 264 bp amplicon encompassing the mutation site was generated with the following primers: Forward GAAGGCTCGTCTCTGCTGTGTGGTG and Reverse CGTTATGGATGCGCTGAGACTGAATACC. This amplicon was digested with BfaI to yield 211 bp and 53 bp (WT), 264 bp (homozygous mutant), or 264 bp and 211 bp and 53 bp (heterozygous mutant) products.

Expression profiling V0v neurons & microarray design

To identify transcription factors expressed by V0v neurons, V0v spinal neurons, all spinal cord neurons and all cells within the trunk were isolated from live transgenic zebrafish embryos at 27 h using fluorescence activated cell-sorting (FACS). Prior to FACS, embryos were prim-staged, de-yolked, dissected and dissociated as in [42, 43]. Heads and tails were removed from all samples to ensure that only trunk or spinal cord cells were collected. Trunk samples correspond to FAC-sorted trunk cells (spinal cord and other tissues). All neuron samples are EGFP-positive cells from Tg(elav13:EGFP) trunks [44]. V0v neurons are EGFP-positive cells from Tg(evx1:EGFP) ^SU1 trunks [11]. Total RNA was extracted using an RNeasy Micro Kit (Qiagen, 74004). The quality of RNA was assessed via an Agilent 2100 Bioanalyser (RNA 6000 Pico Kit, Agilent, 5067–1513) before being converted to fluorescently-labeled cDNA (Ovation Pico WTA System V2, Pico, 3302) and hybridized to a custom-designed Agilent microarray (Agilent #027382). Data pre-processing and normalization was performed using Bioconductor software (https://www.bioconductor.org/). A three-class ANOVA analysis was performed using GEPAS software [45, 46]. Relative expression levels were subjected to a Z-transformation normalization and are presented as Z scores where mean = 0 and standard deviation = +1 (red) to -1 (blue) [47–49]. All reported statistics were corrected for multiple testing [50].

To generate the custom-designed Agilent microarray (Agilent #027382) we first performed comprehensive bioinformatic searches for proteins that contain at least one of the 483 InterPro domains identified in [51] as being specific to transcriptional regulators. These domains comprise three functional classes: DNA binding, chromatin remodeling and general transcription machinery. We identified 3192 potential transcription factors. 2644 of these proteins were identified in Zv8 (Ensembl release 54) of the zebrafish genome and a further 548 non-overlapping transcription factors were identified in the zebrafish Unigene dataset (release 117). Our custom arrays contain 33784 probes corresponding to eight distinct 60-mer probes for each of the transcripts associated with these 3192 proteins. We also included 170 housekeeping genes (five copies of eight probes each), 23 positive controls, such as neurotransmitter markers (two copies of eight probes each) and 49 negative controls (Arabidopsis sequences; multiple copies of eight probes each) on the arrays. Four biological replicates were performed per sample type. Microarray data are deposited at NCBI GEO entry number GSE83723.

in situ hybridization

Embryos were fixed in 4 % paraformaldehyde and single and double in situ hybridization experiments were performed as previously described [52, 53]. Probes for in situ hybridization experiments were prepared using the following templates: evx1 [30], evx2 [29], lbx1a [54] and lmx1ba [24]. A probe for lmx1bb was generated from cDNA as previously described [11, 43] with the following primers: forward CTGGATATCAAGCCGGAGAA; reverse AATTAACCCTCACTAAAGGGATCCGAACATCACATTTCAACA. The lmx1bb probe sequence was selected to avoid cross-hybridization with lmx1ba and other lmx1 family members.

To try and improve signal strength of the lmx1ba probe, we also hydrolyzed the full length lmx1ba probe described above [24] to approximately 200 bp fragments as outlined in [55] and tested two additional lmx1ba probes. The second probe was synthesized from a plasmid containing the last 584 bp of the coding sequence of lmx1ba. The third probe, which recognizes the 3’ coding sequence and UTR of lmx1ba, was generated from cDNA, as previously described [11, 43], with the following primers: forward CGCATGCGTTGGTATCTATG; reverse AATTAACCCTCACTAAAGGGAAAGCATCCTCCACAATGTCC. As these probes did not improve the signal quality when compared to the first probe described above [24], results from these in situ hybridization experiments are not included in this paper.

To determine neurotransmitter phenotypes, we used in situ probes for genes that function as transporters of neurotransmitters or that synthesize specific neurotransmitters as these are some of the most specific molecular markers of these cell fates (e.g. see [56] and references therein). A mixture of probes to slc17a6a and slc17a6b, which encode glutamate transporters, was used to label glutamatergic neurons [56, 57]. To label inhibitory cells we used slc32a1, which encodes a vesicular inhibitory amino acid transporter [33]. To label glycinergic cells a mixture of probes (glyt2a and glyt2b) for the gene slc6a5 were used [56, 57]. The slc6a5 gene encodes for a glycine transporter necessary for glycine reuptake and transport across the plasma membrane. GABAergic neurons were labeled by a mixture of probes to gad1b and gad2 genes (probes previously called gad67a, gad67b and gad65) [56, 57]. The gad1b and gad2 genes encode for glutamic acid decarboxylases, which are necessary for the synthesis of GABA from glutamate.

Immunohistochemistry

Embryos were fixed in 4 % paraformaldehyde and stored in PBS with 0.1 % tween20. To permeabilize embryos they were treated with acetone at -20 °C for 30 min (36 h or younger), 1 h (48 h) or 3 h (7 dpf) and then processed as previously described [11]. Primary antibodies were: mouse anti-GFP (Roche Applied Science, 11814460001, 1:500), rabbit anti-DsRed (Clontech, 632496, 1:200) or rabbit anti-activated caspase-3 (Fisher Scientific/BD, BDB559565, 1:500). Secondary antibodies were: Alexa Fluor 568 goat anti-rabbit (Molecular Probes, A11036, 1:500), Alexa Fluor 488 goat anti-mouse (Molecular Probes, A11029, 1:500) or Alexa Fluor 488 goat anti-rabbit (Molecular Probes, A11034, 1:500).

Double stains

Both double in situ hybridization and immunohistochemistry plus in situ hybridization double labeling experiments were performed as in [52].

Acridine orange treatment

A stock acridine orange base (Sigma-Aldrich, 235474) solution of 2.5 mg/mL in dimethyl sulfoxide (DMSO) was made and stored at -20 °C until used. At 24 h, 36 h and 48 h acridine orange stock solution was added to embryo medium (5 mM NaCl, 0.17 mM KCl, 0.33 mM CaCl₂ · 2H₂O and 0.33 mM MgSO₄ · 7H₂O in water) to make a final concentration of 5 μg/mL. Embryos were bathed in the acridine orange / embryo medium solution in the dark at 28.5 °C for 28 min. Embryos were then washed five times in embryo medium for 5 min each and analyzed using fluorescent microscopy on a Zeiss Axio Imager M1 compound microscope and Olympus SZX16 dissecting microscope.

Imaging

Embryos were mounted in 70 % glycerol, 30 % PBS and differential interference contrast (DIC) pictures were taken using an AxioCam MRc5 camera mounted on a Zeiss Axio Imager M1 compound microscope. Embryos from acridine orange experiments and anti-activated caspase-3 experiments were mounted in 2 % 1,4-diazabicyclo[2.2.2] octane (DABCO) and imaged in the same way. Zeiss LSM 710 and LSM 780 confocal microscopes were used to image embryos mounted in DABCO from fluorescent in situ hybridization and immunohistochemistry experiments. Images were processed using Adobe Photoshop software (Adobe, Inc), GNU Image Manipulations Program (GIMP 2.6.10, http://gimp.org) and Image J software (Abramoff et al. [58]). In some cases, different focal planes were merged to show labeled cells at different medial-lateral positions in the spinal cord.

Cell counts and statistics

For acridine orange staining and activated caspase-3 immunohistochemistry experiments, cells were counted along both sides of the entire rostral-caudal axis of the spinal cord. For all other experiments, we identified somites 6–10 in each embryo and counted the number of labeled cells in that stretch of the spinal cord. In all cases, embryos were mounted laterally with the somite boundaries on each side of the embryo exactly aligned and the apex of the somite over the middle of the notochord. This ensures that the spinal cord is straight along its dorsal-ventral axis and that cells in the same dorsal/ventral position on opposite sides of the spinal cord will be directly above and below each other. Cell counts for fluorescently-labeled cells were performed by analyzing all focal planes in a confocal stack of the appropriate region(s) of the spinal cord. Labeled cells in embryos analyzed by DIC were counted while examining embryos on the Zeiss Axio Imager M1 compound microscope. We adjusted the focal plane as we examined the embryo to count cells at all medial/lateral positions (both sides of the spinal cord; also see [7, 11, 52, 59]). Values are reported as the mean +/- the standard error of the mean. Results were analyzed using the student’s t-test.

lmx1ba and lmx1bb are expressed by zebrafish dI5 and V0v neurons

To identify transcription factors that may play a role in V0v neuron specification and/or maintenance, we expression-profiled V0v neurons and compared them to all post-mitotic neurons and all trunk cells (see methods; NCBI GEO GSE83723; [43]). These analyses identified lmx1ba and lmx1bb, zebrafish ohnologs of Lmx1b (Fig. 1a), as two transcription factor genes potentially expressed in V0v neurons. Prior to this study, the only report of lmx1b expression in the zebrafish spinal cord established that lmx1bb is expressed in at least some rostral spinal neurons at 24 h [24]. However, it was unclear if lmx1bb expression was restricted to the rostral spinal cord and these earlier studies did not detect lmx1ba expression in the spinal cord [24]. Therefore, to further confirm our microarray data, we examined the spinal cord expression of lmx1ba and lmx1bb in more detail (Fig. 1).

In contrast, at 27 h, lmx1bb spinal cord expression already extends along the entire rostral-caudal axis in a narrow dorsal-ventral domain (Fig. 1c). Like lmx1ba, lmx1bb is also expressed in the roof plate and floor plate at this stage. As development progresses, more spinal cord neurons express lmx1bb and roof plate expression becomes more prominent while floor plate expression is lost by 36 h (Fig. 1c, e and g; Table 1). By 48 h, lmx1ba and lmx1bb are expressed in presumably overlapping domains, although, as all lmx1ba in situ probes tested produced very weak staining (see methods), it was not possible to confirm this with co-labeling experiments.

As mentioned above, our expression profiling of V0v neurons suggested that zebrafish lmx1b genes may also be expressed by these cells (Fig. 1a). To confirm these results we performed EGFP immunohistochemistry and lmx1bb in situ hybridization in Tg(evx1:EGFP) ^SU1 embryos that express EGFP in V0v neurons [11]. These experiments showed that at 30 h at least 38 % of lmx1bb-expressing neurons are V0v neurons (Fig. 1i; Table 2).

Both V0v cells and dI5 cells are glutamatergic [8, 11, 16, 33, 34]. Moreover, Lmx1b-expressing neurons are glutamatergic in the amniote spinal cord [8, 16, 32]. Therefore, to further confirm the identity of zebrafish lmx1bb-expressing spinal neurons we performed double-labeling experiments. Double in situ hybridization for lmx1bb and glutamatergic markers slc17a6a + slc17a6b (a mixture of probes for both genes, referred to here as slc17a6; see methods), showed that at 30 h at least 79 % of lmx1bb-expressing cells co-express slc17a6 (Fig. 1j; Table 2). To further confirm that most lmx1bb-expressing neurons are glutamatergic, we also performed double staining for EGFP and lmx1bb in 30 h Tg(slc17a6:EGFP) embryos in which many glutamatergic neurons express EGFP [33, 65–67]. In these embryos, we found that approximately 70 % of lmx1bb-expressing neurons also express EGFP (Fig. 1k; Table 2). In contrast, double in situ hybridization with lmx1bb and slc32a1, which labels all spinal cord inhibitory neurons [33, 68], revealed that only 10 % of lmx1bb neurons are inhibitory (Fig. 1l; Table 2). Taken together, these data suggest that the vast majority of zebrafish lmx1bb-expressing cells are glutamatergic and that these glutamatergic cells correspond to dI5 and V0v neurons.

lmx1bb is required for glutamatergic neurotransmitter phenotypes at later developmental stages but does not repress inhibitory neurotransmitter phenotypes

To investigate the functions of lmx1ba and lmx1bb in the zebrafish spinal cord we used mutations in each of these genes (see methods). We consider that both of these mutant alleles are likely to cause a complete loss of function as they result in premature stop codons before the homeobox (lmx1bb) or before both the homeobox and the lim domains (lmx1ba) (see methods). In fact, if the mutated lmx1ba RNA is translated, it would consist of only six amino acids of WT sequence followed by 11 altered amino acids. To test if the RNAs are degraded by nonsense mediated decay, we performed in situ hybridization for each gene in the respective mutant. For lmx1ba, we do not see any obvious changes in lmx1ba RNA (Fig. 2b). In contrast, we see a loss of lmx1bb RNA in the spinal cord of lmx1bb homozygous mutants (Fig. 2f), although some, potentially weaker than normal, expression remains in other regions of the embryo. This suggests that at least Lmx1bb function is completely lost from the spinal cord.

Since we see a loss of lmx1bb spinal cord expression in lmx1bb mutants and lmx1bb is expressed by more spinal interneurons at an earlier developmental time point than lmx1ba, we first examined the function of lmx1bb. As lmx1bb is expressed predominantly by glutamatergic neurons in the spinal cord, we assessed the expression of the glutamatergic marker slc17a6 at 27, 36, and 48 h [18, 32]. At 27 h there was no statistically significant difference in the number of glutamatergic neurons in the spinal cord (p = 0.41, Fig. 3a, b and g; Table 3). However, at 36 h there was a statistically significant reduction in the number of glutamatergic neurons in lmx1bb mutants compared to WT siblings (p < 0.001, Fig. 3c, d and g; Table 3) and this reduction became more pronounced by 48 h (p < 0.001, Fig. 3e-g; Table 3). Taken together, these results suggest that lmx1bb is required either to maintain the glutamatergic phenotype of a subset of excitatory spinal neurons or to specify the glutamatergic phenotype of a later-forming subset of neurons.

		27 h		36 h		48 h
Marker		WT	lmx1bb -/-	WT	lmx1bb -/-	WT	lmx1bb -/-
slc17a6	Mean	121.6	127.2	137.2	123.6	211	175
	SEM	5.3	4.1	2.5	2.4	5.5	2.9
	n	10	10	13	17	10	13
	p value	0.411		<0.001		<0.001
slc32a1	Mean	149.7	148	173.2	169.7	210.5	202
	SEM	2.7	5.8	10.9	17.1	4	10.8
	n	10	6	14	7	12	6
	p value	0.77		0.85		0.48

Mean number of slc17a6a + slc17a6b (slc17a6) or slc32a1-expressing cells counted in the spinal cord region adjacent to somites 6–10 in 27 h, 36 h and 48 h embryos. SEM is the standard error of the mean. n is the number of embryos analyzed for each data set. p value is from a student’s paired t-test comparing WT and lmx1bb mutant embryos. Statistically significant p values are indicated in bold

lmx1ba and lmx1bb single mutants and lmx1ba;lmx1bb double mutants have the same spinal cord phenotype

As shown above (Fig. 1d-g), lmx1ba and lmx1bb are expressed in potentially overlapping domains within the zebrafish spinal cord during the developmental time points that we detected neurotransmitter phenotypes in lmx1bb mutants. This suggested that these two ohnologs might function redundantly in the spinal cord. Therefore, we examined spinal cord neurotransmitter phenotypes in lmx1ba single and lmx1ba;lmx1bb double mutants.

When we examined lmx1ba single mutants at 48 h, we found that the number of glutamatergic neurons were statistically significantly reduced (p < 0.001) compared to WT siblings (Fig. 4e, e’ and h; Table 5). Interestingly, the number of glutamatergic neurons lost in the lmx1ba mutant was not statistically significantly different from the number of glutamatergic neurons lost in the lmx1bb mutant (p = 0.7; Fig. 4f, f’ and h; Table 5). More surprisingly, we also found that the number of spinal cord glutamatergic neurons lost in lmx1ba;lmx1bb double mutants, was not statistically significantly different from either lmx1ba single mutants (p = 0.78) or lmx1bb single mutants (p = 0.45; Fig. 4g, g’ and h; Table 5).

	48 h	lmx1ba +/+ lmx1bb +/+	lmx1ba +/- lmx1bb +/+	lmx1ba +/+ lmx1bb +/-	lmx1ba +/- lmx1bb +/-	lmx1ba -/- lmx1bb +/+	lmx1ba +/+ lmx1bb -/-	lmx1ba -/- lmx1bb -/-
slc17a6	Mean	246.3	250.2	233.8	196.4	199.4	204.5	193.1
	SEM	7.4	5.9	4.5	5.5	10.4	10	6.2
	n	8	5	4	8	11	9	13
	p value 1	n/a	0.72	0.3	0.001	0.001	0.003	0.001
	p value 2	0.3	0.08	n/a	0.005	0.03	0.05	0.004
	p value 3	0.001	0.001	0.005	n/a	0.66	0.38	0.78
	p value 4	0.003	0.004	0.05	0.38	0.7	n/a	0.45
	p value 5	0.001	0.001	0.03	0.66	n/a	0.7	0.78
slc32a1	Mean	179.8						188.5
	SEM	12.9						14.1
	n	6						6
	p value 1	n/a						0.94

Mean number of slc17a6a + slc17a6b (slc17a6) or slc32a1-expressing cells detected in the spinal cord region adjacent to somites 6–10 in 48 h embryos. SEM is the standard error of the mean. n is the number of embryos analyzed for each data set. p values are from student’s paired t-test. Statistically significant p values are indicated in bold. p value 1 is from comparing WT (lmx1ba ^+/+ ;lmx1bb ^+/+ ) embryos pairwise with all other genotypes, p value 2 is from comparing lmx1ba ^+/+ ;lmx1bb ^+/- embryos pairwise with all other genotypes, p value 3 is from comparing lmx1ba ^+/- ;lmx1bb ^+/- embryos pairwise with all other genotypes, p value 4 is from comparing lmx1ba ^+/+ ;lmx1bb ^-/- embryos pairwise with all other genotypes and p value 5 is from comparing lmx1ba ^-/- ; lmx1bb ^+/+ embryos pairwise with all other genotypes

Given the similarity of the phenotypes in lmx1ba and lmx1bb single and double mutants, we tested whether lmx1bb is required for lmx1ba spinal cord expression or vice versa. However, when we analyzed expression of lmx1ba in lmx1bb mutants and expression of lmx1bb in lmx1ba mutants we saw no obvious differences between WT and mutant embryos (Fig. 2c and e). This suggests that the phenotypic similarities between the mutants are not due to cross-regulation of these two lmx1b genes.

Together, these results suggest that lmx1ba and lmx1bb function partially redundantly in the spinal cord and that the presence of two or more mutant alleles (regardless of whether the mutation is in lmx1ba or lmx1bb) is sufficient to cause a reduction in the number of glutamatergic cells in the spinal cord. To test this, we examined the number of glutamatergic spinal neurons in lmx1ba;lmx1bb double heterozygous embryos and both lmx1ba and lmx1bb single heterozygous embryos. Consistent with our hypothesis, the reduction in the number of glutamatergic neurons in lmx1ba;lmx1bb double heterozygous embryos was not statistically significantly different from the reduction in lmx1ba mutants (p = 0.66), lmx1bb mutants (p = 0.38) or lmx1ba;lmx1bb double mutants (p = 0.78; Fig. 4d, d’ and h; Table 5). In contrast, neither lmx1ba nor lmx1bb single heterozygous embryos had a statistically significant reduction in the number of glutamatergic neurons when compared to WT siblings (p = 0.72 and p = 0.3 respectively; Fig. 4b-c’ and h; Table 5).

To test the possibility that lmx1ba might compensate for the loss of lmx1bb in the repression of inhibitory neurotransmitter phenotypes, we also analyzed the expression of slc32a1 in lmx1ba;lmx1bb double mutants. However, like the lmx1bb single mutant results, the lmx1ba;lmx1bb double mutants had no statistically significant change (p = 0.94) in the number of spinal inhibitory neurons (Fig. 4i-k; Table 5). These data suggest that lmx1ba and lmx1bb are not required to repress (or specify) inhibitory neurotransmitter phenotypes and that the reduction in spinal cord glutamatergic cells in these mutants does not correlate with an increase in inhibitory cells.

The reduction in spinal glutamatergic cells is not due to cell death

To test whether the reduction in glutamatergic neurons might be an indirect consequence of increased cell death, we used both acridine orange (AO) and an activated caspase-3 antibody [76, 77]. As the glutamatergic phenotype is comparable among lmx1ba;lmx1bb double mutants and both single mutants, we used lmx1bb single mutants for these and all subsequent experiments.

AO is a vital dye that labels apoptotic cells in live zebrafish embryos [76, 78–81], as demonstrated in our positive control, smoothened mutant embryos, where many cells undergo apoptosis [76] (Fig. 5a-b’). We performed AO staining in lmx1bb mutants at 36 h, when we first observe a reduction in the number of glutamatergic spinal cells, and at 48 h, when the loss of glutamatergic spinal cord cells is more pronounced. At both of these time points there were no obvious differences in spinal cord AO staining in any of the live embryos derived from incrosses of heterozygous lmx1bb mutants (Fig. 5c-f’). Following imaging and analysis, a subset of embryos were genotyped to confirm that we had analyzed both WT and lmx1bb homozygous mutant embryos. These analyses demonstrated that there was no apparent correlation between the amount of AO staining and embryo genotype.

V0v and dI5 cells form in normal numbers in lmx1bb mutants

Since lmx1bb is co-expressed by both lbx1a-expressing dI5 cells and evx-expressing V0v neurons (Fig. 1h and i), we also examined whether lbx1a or evx1 + evx2 (a mixture of probes for both genes, referred to here as evx) expression was altered in lmx1bb mutants. At 48 h there was no statistically significant difference in the number of cells expressing lbx1a or evx in lmx1bb mutants compared to WT siblings (p = 0.93 and p = 0.78 respectively; Fig. 6a-d and g; Table 4). Additionally, there was no statistically significant change in the number of EGFP-labeled V0v neurons in lmx1bb mutants expressing the Tg(evx1:EGFP) ^SU1 transgene when compared to WT siblings (p = 0.63; Fig. 6e-g; Table 4). These data suggest that Lmx1bb function is not required for either lbx1a or evx expression and that V0v and dI5 cells form in normal numbers in lmx1bb mutants. This is consistent with our previously described apoptosis assays which suggest that spinal neurons are not dying in these mutants (Fig. 5). Furthermore, these findings suggest that there is no effect on V0v or dI5 cell proliferation and that these cells are not transfating into different cell types in the lmx1bb mutants, they are just losing or do not develop their glutamatergic fates.

lmx1bb is required for the glutamatergic phenotype of at least a subset of V0v interneurons

As described above (Fig. 1i), we have shown for the first time in any animal that at least a subset of V0v neurons express lmx1bb. To test whether V0v cells in particular are affected in lmx1bb mutants we performed double labels for EGFP and slc17a6 in WT and lmx1bb mutant Tg(evx1:EGFP) ^SU1 embryos. We found that at 48 h, there is a significant reduction in the number of glutamatergic double-labeled V0v cells in mutant embryos compared to their WT siblings (p < 0.001; Fig. 7a, b and e; Table 7). This suggests that at least some of the cells that are losing their glutamatergic phenotypes in lmx1bb mutants are V0v neurons.

	slc17a6 (glutamatergic)		EGFP (V0v)		co-labeled
48 h	WT	lmx1bb -/-	WT	lmx1bb -/-	WT	lmx1bb -/-
Mean	155	142.2	92.3	93.4	44.3	32.4
SEM	1.54	2.22	2.72	2.58	0.9	1.9
n	6	5	6	5	6	5
p value	<0.001		0.76		<0.001
	DsRed (glutamatergic)		EGFP (V0v)		co-labeled
7 dpf	WT	lmx1bb -/-	WT	lmx1bb -/-	WT	lmx1bb -/-
Mean	132.6	105.3	100.1	91.3	42.9	21.8
SEM	5.5	5.2	1.4	2.1	4.9	2.7
n	9	12	9	12	9	12
p value	<0.005		0.002		<0.001

Mean number of cells expressing slc17a6, EGFP or both in the spinal cord region adjacent to somites 6–10 in 48 h embryos and the mean number of cells expressing DsRed, EGFP or both in the spinal cord region adjacent to somites 6–10 in 7 dpf embryos. SEM is the standard error of the mean. n is the number of embryos analyzed for each data set. p value is from a student’s paired t-test comparing WT embryos and lmx1bb homozygous mutants. Statistically significant values are indicated in bold

We were also interested in establishing if the reduction of glutamatergic cells in general and/or the reduction in the number of glutamatergic V0v cells persists at later stages of development. As our slc17a6 RNA probe does not label cells effectively in double staining experiments at later stages of development, we created fish transgenic for both Tg(slc17a6b:loxP-DsRed-loxP-GFP) ^nns14 and Tg(evx1:EGFP) ^SU1 and heterozygous for the lmx1bb mutation. Embryos from these parents express DsRed in glutamatergic neurons and EGFP in V0v neurons [11, 85, 86]. We then counted single and double-labeled cells in WT and lmx1bb mutant embryos at 7 dpf to determine if the number of excitatory cells and excitatory V0v neurons in particular were reduced in the lmx1bb mutants. We again observed a statistically significant (p < 0.005) reduction in the total number of glutamatergic, DsRed-labeled neurons (Fig. 7d and f; Table 7) as well as a statistically significant reduction in the number of glutamatergic DsRed-labeled V0v neurons (p < 0.001; Fig. 7d and f; Table 7) in lmx1bb mutant embryos compared to WT siblings. More surprisingly, we also observed a very small but statistically significant (p = 0.002) reduction in the number of V0v (EGFP-labeled) neurons at 7 dpf in lmx1bb mutants (Fig. 7d’” and f; Table 7). However, this slight reduction was substantially less than the reduction in double-labeled glutamatergic V0v neurons, demonstrating that the second result cannot be explained by the first. Taken together, these results suggest that lmx1bb is required for the glutamatergic neurotransmitter phenotype of at least a subset of V0v neurons at later developmental stages.

lmx1ba and lmx1bb expression requires evx1 and evx2

Evx1 and Evx2 function partially redundantly to specify the glutamatergic neurotransmitter phenotype of V0v neurons [11]. As demonstrated above (Fig. 7), lmx1bb is required at later developmental stages for the glutamatergic neurotransmitter phenotype of at least a subset of V0v neurons and evx1 and evx2 spinal cord expression is normal in lmx1bb mutants (Fig. 6c-g), suggesting that Evx1 and Evx2 do not act downstream of Lmx1b in V0v cells. To determine whether Lmx1b acts downstream of Evx1 and Evx2 or in a parallel pathway we examined lmx1b expression in evx1;evx2 double mutants. At 30 h we found that there was a statistically significant reduction in the number of lmx1ba (p = 0.029) and lmx1bb (p < 0.001) expressing spinal cord cells in evx1;evx2 double mutants when compared to WT siblings (Fig. 8a-e; Table 8). This suggests that these two genes require Evx function for their expression in V0v cells and that they are downstream of evx1 and evx2 in these cells (Fig. 8f).

Marker	30 h	WT	evx1 -/- ;evx2 -/-
lmx1ba	Mean	8.6	4.8
	SEM	0.81	1.1
	n	5	6
	p value	0.029
lmx1bb	Mean	35.9	19.7
	SEM	6.4	5.3
	n	12	6
	p value	<0.001

Mean number of lmx1ba or lmx1bb-expressing cells in the spinal cord region adjacent to somites 6–10 in 30 h embryos. SEM is the standard error of the mean. n is the number of embryos analyzed for each data set. p value is from a student’s paired t-test comparing WT and lmx1bb mutant embryos. Statistically significant p values are indicated in bold