Investigation of Islet2a function in zebrafish embryos: Mutants and morphants differ in morphologic phenotypes and gene expression

Zebrafish primary motor neurons differ from each other with respect to morphology, muscle targets and electrophysiological properties. For example, CaP has 2-3-fold larger densities of both inward and outward currents than do other motor neurons. We tested whether the transcription factor Islet2a, uniquely expressed in CaP, but not other primary motor neurons, plays a role in specifying its stereotypic electrophysiological properties. We used both TALEN-based gene editing and antisense morpholino approaches to disrupt Islet2a function. Our electrophysiology results do not support a specific role for Islet2a in determining CaP’s unique electrical properties. However, we also found that the morphological phenotypes of CaP and a later-born motor neuron differed between islet2a mutants and morphants. Using microarrays, we tested whether the gene expression profiles of whole embryo morphants, mutants and controls also differed. Morphants had 174 and 201 genes that were differentially expressed compared to mutants and controls, respectively. Further, islet2a was identified as a differentially expressed gene. To examine how mutation of islet2a affected islet gene expression specifically in CaPs, we performed RNA in situ hybridization. We detected no obvious differences in expression of islet1, islet2a, or islet2b in CaPs of mutant versus sibling control embryos. However, immunolabeling studies revealed that an Islet protein persisted in CaPs of mutants, albeit at a reduced level compared to controls. While we cannot exclude requirement for some Islet protein, we conclude that differentiation of the CaP’s stereotypic large inward and outward currents does not have a specific requirement for Islet2a.


Introduction
Mammalian spinal motor neurons comprise a heterogeneous population, as evidenced by their different morphological and functional properties [1] [2]. Several lines of evidence support the view that different combinations of LIM-homeodomain (LIM-HD) transcription factors direct specification of the diverse set of mammalian motor neuron subtypes [3] [4]. PLOS  In the zebrafish spinal cord, early born primary motor neurons (PMNs) also are heterogeneous and express LIM-HD transcription factors in a combinatorial manner [5][6][7][8][9][10]. Each hemisegment has three different PMNs-RoP, MiP, CaP; in some hemisegments, there is also a variably-present PMN, VaP, a CaP duplicate that dies early [11]. PMNs differ from each other with respect to gene expression, soma position, axonal trajectory and/or electrical membrane properties [5,8 12]. At the time of axon genesis, CaP expresses islet2a, whereas other PMNs do not. These findings raise the possibility that Islet2a plays a role in specifying CaP's motor neuron subtype-specific properties.
In addition to islet2a, zebrafish embryos express islet1 and islet2b [9,10] [8] [13] [14]. All PMNs initially express islet1 [8]. In zebrafish, knock-down of Islet1 results in failure of presumptive motor neurons to extend peripheral axons and innervate muscle targets, a fundamental requirement for motor function [15]. Further, upon disruption of Islet1 function, presumptive motor neurons differentiate novel non-motor neuron-like membrane electrical membrane properties [16]. These results implicate an essential role for Islet1 in general specification of motor neurons.
A few hours after the initial expression of islet1, CaPs begin to express islet2a and simultaneously downregulate islet1 [8]. The exclusive expression of islet2a in CaP vs. other PMNs raises the possibility that it may selectively specify this motor neuron subtype. Consistent with this, disruption of Islet2a function, either by overexpression of Islet2a-LIM domains or morpholino (MO) knock-down, leads to defects in outgrowth of CaP axons [17] [15]. In addition, in prdm14 mutants, CaPs lack expression of islet2a and have axons with stunted growth [18].
In Drosophila larvae, loss-of-function of islet, the fly orthologue of zebrafish islet genes, leads to changes in morphology as well as electrophysiological properties of one motor neuron subtype [19,20]. CaP's electrophysiological properties also distinguish it from MiP [12]. Specifically, the densities of inward and outward currents are 2.9 and 2.5-fold greater, respectively, in CaPs compared to MiPs. On this basis, we tested whether Islet2a determines CaP's distinct electrophysiological as well as morphological properties. The results do not support a specific, non-redundant role for Islet2a in determining CaP's stereotypic large inward and outward current densities.

Zebrafish transgenic lines
All animal procedures used in this study have been approved by the University of Colorado Denver Animal Care and Use Committee (Approval number-74810(04)1D). Embryos and larvae were sacrificed by tricaine overdose accomplished by prolonged immersion in 0.02% tricaine methane sulfonate (Sigma-Aldrich, St. Louis, MO).
We isolated two different mutations, 105(CO7002) and 102(CO7003). Both had frameshifts introduced within exon 2 leading to a downstream premature STOP codon ( Fig 1A) The 105 allele consisted of a 13-nucleotide deletion within exon 2, leading to a frame shift and premature STOP codon. The second allele, CO7003, also had a frameshift at the same position and a premature stop codon following 30 non-conserved amino acids. For both alleles, homozygous mutants were viable and to adult stages. Here, we report results for the 105 allele.

Morpholino knockdown
Antisense MO oligonucleotides (Gene Tools, Philomath, OR) were injected into one-cell stage zebrafish embryos. We used a previously published translation blocking antisense, T-MO [15], The sequence of homozygous islet2a 105 genomic DNA revealed a 13-nuc deletion (at position Ã ) that removed a restriction enzyme site and introduced a frameshift. (B) Schematic of wildtype and mutant Islet2a proteins. The islet2a 105 sequence predicts a truncated Islet2a protein product lacking the homeobox (HD) and LIM2 (L2) domains and containing only a portion (~70%) of the LIM1 (L1) domain. The Isl11/2 monoclonal antibody used in this study was raised against a carboxy-terminus sequence not present in the predicted mutant protein. (C-F) Expression of islet2a mRNA (C, D) and Isl1/2 immunoreactivity (E, F) in wildtype (wt) and homozygous mutant (mut) islet2a 28 hpf embryos. Images are lateral views of embryos, with dorsal up and anterior to the left. Bright field and fluorescent images have been merged. (C, D) A tissue that expresses islet2a, but not islet1 or islet 2b, was used to assess the efficacy of the TALEN mutation. The proctodeum (P), just caudal to the posterior end of the yolk sac extension (yse) emerges during embryonic stages and develops later into the anal passage. In both wildtype (n = 6) and mutant (n = 8) embryos, the proctodeum expressed islet2a mRNA. (E, F) The Isl1/2 antibody immunolabeled the proctodeum in wildtype (E; n = 10) but not mutant (F; n = 7) 28 hpf embryos. Scale bar in F, for C-F: 25 μm.
https://doi.org/10.1371/journal.pone.0199233.g001 directed to the start codon of the islet2a mRNA sequence (5'-GGATGCGGTAGAATATCCA CCATAC-3') at a concentration of 5 mg/ml. This T-MO differs from but partially overlaps with another T-MO previously used to perturb Islet2a function [17]. We used a control MO (Ctl MO) with 5-base pair mismatches compared to the T-MO (5'-GaATGCGcTAcAATA TCCAgCAaAC-3') at a concentration of 5 mg/ml. In addition, a splice blocking MO, Sp-MO (10 mg/ml) was designed to a sequence overlapping the splice junction between intron 1-2 and exon 2 (5'-CAGACTTCTCTGGATATGGAAAGCA-3'; S1 Fig). Comparisons of the MO and islet gene sequences supports specificity of the MOs targeting islet2a (Table 1 and S1  Table) [28,29]. The T-MO and Sp-MO produced similar motor neuron morphological results and we report results obtained with T-MO.

RNA in situ hybridization
RNA in situ hybridization was carried out on fixed whole-mount 24, 28, 30 and 48 hpf embryos, as described previously [30]. Briefly, embryos were fixed in 4% paraformaldehyde, followed by dehydration with increasing concentrations of methanol and overnight incubation in 100% methanol (-20˚C). Digoxigenin-labeled sense and antisense islet1, islet2a and islet2b RNA probes [8,31] were synthesized and hybridized to whole-mount preparations [32,33]. Hybridized RNA probes were detected with anti-digoxigenin Fab fragments coupled to alkaline phosphatase, followed by reaction with the Fast Red chromogen (Sigma-Aldrich, St. Louis, MO). In most cases, RNA in situ hybridization was followed by immunocytochemistry.

Immunocytochemistry
For whole mount immunohistochemistry, 24-72 hpf embryos and larvae were fixed in 4% paraformaldehyde in PBS. Antibodies were diluted in 10% heat inactivated goat or fetal bovine serum in phosphate-buffered saline (PBS) containing 0.2% Tween. When immunochemistry followed in situ hybridization, 2% bovine serum albumin (Sigma-Aldrich, St. Louis, MO) was also included.
To reveal SMN axons, we used the neurolin monoclonal, zn8 (1:200; DSHB). The following Alexa-conjugated secondary antibodies (Thermo Fisher Scientific) were used at 1:1000: goat For islet1 and islet2b, the regions corresponding to the islet2a sequences targeted by the MO were identified and compared. We include isl1l, identified as the duplicate of islet1 on the basis of syntenic relationships [28,29].
However, zebrafish embryos express little to no isl1l mRNA (zfin.org). Further, the predicted Isl1l amino acid sequence shares only 47% identity with islet1. S1 anti-rabbit IgG Alexa-488; goat anti-mouse IgG 2b Alexa-546; goat anti-mouse IgG 2a Alexa-488; goat anti-mouse IgG 1 Alexa-488 or -546. The number of dorsally-projecting SMNs (dSMNs) were counted in 72 hpf tg(islet1:gfp) larvae in 4 hemisegments above the junction between the yolk sac and yolk sac extension. For purposes of comparison, we normalized values to the average number in uninjected larvae. Data were analyzed for statistical analysis by ANOVA followed by post hoc Bonferroni to correct for multiple comparisons.

Confocal imaging
For imaging of processed embryos, we used a Marianas spinning disk confocal microscope (Intelligent Imaging Innovations, Denver, CO). Images were taken in the region comprising the~3-4 hemisegments caudal and rostral to the junction between the yolk sac and yolk sac extension, except when examining the more-caudally located proctodeum. The same settings (e.g., gain, laser intensity) were used for imaging all samples in an experiment. Z-stack projections were made using open source software (FIJI version of Image J) [34,35]. Sample size information is provided in the figure legends.
The holding potential was set to -80 mV and trials consisting of 16 50 msec depolarizing steps; each successive step varied by +10 mV starting at -40 mV and extending to 110 mV. For each step, before returning the membrane potential to -80 mV, a 20 msec step to -40 mV was applied to record tail currents. For presentation of data, exemplar traces for steps to +20 mV are shown.
For calculation of current density, current amplitudes were normalized to cell size by dividing by the cell capacitance, a measure of the cell's surface area. For inward and outward current density calculations, we used the peak inward current amplitude recorded during the entire trial of 16 steps and the steady-state outward current recorded during the final 20 msec of the step to +40 mV, respectively. Sample size information is provided in the figure legends.

Whole embryo RNA isolation
RNA was isolated from 30-60 pooled 48 hpf whole embryos of the following conditions: islet2a sibling wildtype, islet2a mutant; sibling wt uninjected, wildtype injected with T-MO or Ctl MO. For microarray analyses, wildtype and uninjected embryo RNAs were each collected in duplicate, and RNA of mutants and morphants in triplicate. For quantitative polymerase chain reaction (qPCR) studies, all samples were collected in triplicate. RNA was isolated by column purification using the RNeasy Mini Kit (Qiagen, Venlo, Netherlands). RNA concentration and integrity were assessed using an Agilent 4200 Tape Station (Agilent Technologies, Santa Clara, CA).

Microarray analysis of gene expression
RNA was delivered to the University of Colorado Denver Genomics and Microarray Core for analysis of gene expression using RNA microarrays (GeneChip™ Zebrafish Gene 1.0 ST Array, Thermo Fisher Scientific). A hybridization cocktail was prepared using 100 ng total RNA and the GeneChip WT PLUS Reagent Kit (Thermo Fisher Scientific). Samples were hybridized to the arrays for 16 hrs at 45˚C (GeneChip™ Hybridization Oven 645, Thermo Fisher Scientific). Arrays were washed, stained (GeneChip™ Fluidics Station 450, Thermo Fisher Scientific), and scanned using a GeneChip Scanner 3000 (Thermo Fisher Scientific). Expression data from the resulting CEL files (see Supplementary information) were extracted and the OLIGO package in R was used to perform RMA (robust multichip average) normalization (S1 File). From this, pairwise ANOVAs in R were performed to compare the morphant vs. mutant, mutant vs. control and morph vs. control datasets; the two wildtype and two uninjected wildtype samples comprised the control group.

cDNA synthesis and quantitative polymerase chain reaction
RNA samples were treated with Amplification Grade DNase (Thermo Fisher Scientific) prior to synthesis of cDNA. cDNA synthesis and qPCR were performed using the EXPRESS One- Step Superscript qRT-PCR Kit and a 7500 Fast Instrument (Thermo Fisher Scientific). Table 2 provides the assay IDs for the genes studied: islet1, islet2a, islet2b, nrp1a, and plexinA3. The assays targeted a region that flanked an intron, to allow distinction between amplification of mRNA vs. genomic DNA. Gene expression levels were normalized to that of the housekeeping gene eef1a1a.
Cycling conditions were adjusted to those specified by the kit manufacturer (Thermo Fisher Scientific). The biological triplicates were each run in technical triplicate. The data were analyzed with the ABI 7500 Software Version 2.0.6 using all default parameters except that the threshold for the C t standard deviation was changed from >0.5 to >0.3. The relative standard curve method was used with gene expression normalized to that of eef1a1a. The validity of eef1a1a as the endogenous control was tested by measuring the standard deviation of the threshold cycle (C t ) of all the samples of equal concentrations. The standard deviation was consistently below 0.5, indicating that eef1a1a did not vary significantly between samples. For analysis of statistical significance, the three technical replicates for each biological replicate were first averaged. Then, the means of the biological triplicates for each condition (wildtype, islet2a homozygous mutants, embryos injected with T-MO or Ctl MO) were calculated; sibling wildtype and uninjected wildtype were pooled as one group, referred to as wildtype. Statistical comparisons were done using ANOVA followed by post hoc Bonferroni to correct for multiple comparisons.

Disruption of the islet2a gene
Using TALEN based methods, we introduced a 13-nucleotide deletion into exon 2 of islet2a, leading to a frame shift and premature STOP codon (105 allele; Fig 1A). The predicted truncated protein lacks the majority of the LIM1 domain and all of the LIM2 and homeobox DNAbinding domains (Fig 1B), regions required for function as well as recognition by the Isl1/2 antibody.
Many cells coexpress islet2a with other islet1 genes [14,31]. A notable exception is the proctodeum that expresses islet2a but not islet1 [18]. RNA in situ hybridization demonstrated that both wildtype and mutant embryos expressed islet2a mRNA at comparable levels in the proctodeum as well as the spinal cord (Fig 1C and 1D). This result indicates that the TALENinduced mutation in exon 2 of islet2a did not trigger nonsense mediated decay, a mechanism that targets and degrades many, but not all, mRNAs containing premature stop codons [37].
Analysis of Islet protein expression in zebrafish tissues typically involves use of a monoclonal antibody, Isl1/2, that recognizes both Islet1 and Islet2 proteins. The proctodeum of wildtype embryos showed positive Isl1/2 immunoreactivity as did cells within the spinal cord ( Fig  1E), consistent with islet2a expression in the proctodeum. In islet2a mutants, however, the proctodeum lacked Isl1/2 immunoreactivity (Fig 1F), as predicted upon loss of full-length wildtype Islet2a protein. These data indicate that the TALEN-induced mutation of islet2a led to loss of expression of full-length Islet2a protein.
In addition, we used a previously reported translation blocking antisense MO (T-MO) to investigate the role of islet2a in the zebrafish embryo [15]. As controls for the T-MO, we used a MO (Ctl MO) with 5 mismatches compared to T-MO and a splice-blocking MO (Sp-MO) that targets the junction between intron 1-2 and exon 2 of the islet2a gene (S1 Fig). For the Sp-MO, RT-PCR analysis revealed the presence of two novel bands (S1B Fig). The additional transcripts corresponded to forms predicted by retention of intron 1-2 (S1A Fig, a) or skipping of exon 2 (S1A Fig, b), as confirmed by DNA sequencing, as expected upon targeting of the junction between intron 1-2 and exon 2 by the Sp-MO.
Effects of gene-editing and MO targeting of islet2a on spinal neuron electrical membrane properties. Our previous work has shown that the electrical membrane properties of CaP differ significantly from another PMN: in 24 hpf embryos, compared to MiPs, CaPs had 2-3 fold larger densities of both inward and outward currents [12]. If the unique expression of islet2a in CaPs determines its stereotypic larger inward and outward current densities, loss of Islet2a would be expected to result in obvious decreases in CaP current densities.
In 24 hpf tg(mnx1:gfp) embryos, CaPs were identified on the basis of gfp expression, soma position and a ventrally-projecting axon. We obtained whole cell voltage clamp recordings from CaPs in 24 hpf islet2a mutant and sibling wildtype embryos (Fig 2A). In contrast to our prediction, we did not detect statistically significant differences in either inward or outward densities for data obtained from CaPs in islet2a wild type, heterozygous or homozygous mutant embryos (Fig 2B and 2C). However, even though the mean values of outward current densities were not significantly different statistically, there was a small difference in the means of CaP outward current density recorded from wildtype compared to mutant embryos (wildtype: 3.5±0.5 pA/μm 2 ; heterozygote: 2.8±0.3 pA/μm 2 ; mutant: 2.7±0.3 pA/μm 2 ). It is possible that with a larger sample size, we might have detected a statistically significant difference between wildtype and mutants. However, on the basis of our results, the difference would not fully account for the previously reported~3-fold difference in outward current densities of CaPs compared to MiPs [12].
Whole cell voltage clamp recordings were also obtained from islet2a morphant and control embryos. Similarly, no significant differences were noted between CaPs in morphant and control embryos for either inward or outward current densities (Fig 3).
Other spinal neurons, e.g., vSMN and RB, also express islet2a [8,14,31]  We also recorded from RBs, given the widespread expression islet2a in this population [31]. However, we detected no significant differences in the densities of inward or outward current of RBs in uninjected vs. morphant embryos, at either 24 or 48 hpf (S2C and S2D Fig).
Overall, these findings do not support a unique role for Islet2a in specifying electrophysiological properties of vSMNs or RBs. However, islet2a is not expressed in all vSMNs (S2 Fig). In addition, RBs co-express several islet genes [14,31] and Islet1 can substitute for Islet2a [15]. Thus, we cannot rule out the possibility that Islet2a normally plays a role in differentiation of RB's electrical membrane properties but in its absence, other Islet proteins effectively compensate. In contrast, at 24 hpf, only islet2a is detected in CaPs, making the possibility of compensation by another normally co-expressed islet gene unlikely.

Effects of MO knock-down vs. Talen mutagenesis of islet2a on CaP morphology
While performing the electrophysiological experiments, we observed that CaPs in morphants often had abnormal, truncated axons (Fig 4A and 4B), as reported previously when MOs or dominant-negative overexpression was used to perturb Islet2a function [15,17]. In contrast, we did not observe truncated axons in live islet2a wildtype, heterozygous or mutant embryos (Fig 4C-4E). In mutants, CaP axons appeared normal on the basis of gfp expression in live tg (mnx1:gfp) 24 hpf embryos (Fig 4C-4E) or immunolabeling with axonal markers in fixed embryos non-transgenic (Fig 4F and 4G).
Islet2a morphants have also been previously observed to have a loss of dorsally-projecting SMNs (dSMNs) [15]. We examined dSMN axons in tg(islet1:gfp) 3 dpf larvae on the basis of gfp expression as well as zn8 (neurolin) immunoreactivity (Fig 5A-5C). As found previously, injection of the T-MO resulted in fewer dSMN axons in the periphery (Fig 5C vs. Fig 5A and  5B). In addition, the number of dSMN cell bodies was also reduced in morphant larvae ( Fig  5D).
https://doi.org/10.1371/journal.pone.0199233.g002 mutants showed a small reduction in the number of dSMN cell bodies, the difference was not statistically significant compared to control but statistically non-significant reduction in the number of dSMN cell bodies (Fig 5H). Overall, islet2a mutants had neither the CaP axon nor the dSMN phenotypes that were present upon injection of the T-MO (Figs 4 and 5).

Comparison of gene expression in morphant and mutant embryos
The different motor neuron phenotypes of islet2a mutants vs. morphants raised questions about the extent to which mutants and morphants differed. Accordingly, we compared their transcriptomes using microarrays. For these analyses, we chose a single developmental stage, 48 hpf, that was intermediate between the two stages at which we observed differences in motor neuron morphologies between mutants and morphants.
Principal component analysis of gene expression in control, mutant and morphant transcriptomes segregated them into different quadrants (Fig 6A). This provided an initial indication that the gene expression profiles of mutant and morphant transcriptomes differed from each other. the two groups did not differ significantly (unpaired two-tailed Student t-test: p value, 0.65). (C) Similar to outward current densities, the densities of peak inward currents recorded from CaPs in the control and morphant embryos were not statistically different (unpaired two-tailed Student t-test: p value, 0.51).
https://doi.org/10.1371/journal.pone.0199233.g003 We identified 174 genes that were differentially expressed in morphants compared to mutants (Q<0.05; fold change <-2 or >2; S2 File). Including control values, a heat map revealed that these 174 genes were often regulated in opposing manners in morphants and mutants compared to controls (Fig 6B; S3 File). Compared to control, genes that were upregulated in the morphant (red) were often downregulated in the mutant transcriptomes (blue or less white) and vice versa.
Mutants did not have any genes that were differentially expressed compared to controls. In contrast, we identified 201 differentially expressed genes that were differentially expressed in morphants compared to control (Supplementary file S4 File). The heat map of expression levels of the 201 genes shows that gene expression in morphants and mutants, compared to controls, was often regulated opposing directions (Fig 6C; S5 File). There). Overall, there were 64 genes that were differentially expressed in morphants compared to both mutants and controls  Neither pathway analysis (IPA 1 ) nor the identities of differentially expressed genes in morphants compared to mutants, controls or both provided any strong insights into potential mechanisms underlying the different motor neuron morphologies of mutants vs. morphants. For example, neither nrp1a nor plexinA3 were identified as differentially expressed genes, even though they have been implicated in regulation of CaP axon morphology [39] [40]. To confirm the microarray results, we carried out focused analysis using qPCR and found that neither nrp1a nor plexinA3 showed any differences in expression levels between wildtype, mutant and morphants embryos (S4 Fig). One gene that was differentially expressed between morphants and controls was islet2a (ENSDART0000012862). Interestingly, islet2a levels were upregulated in morphants. This led us to question whether other islet genes had different expression levels in morphants and/or mutants compared to controls. For this focused analysis, we performed qPCR for islet1, islet2a and islet2b using RNA isolated from wildtype, mutant, morphant, and Ctl MO-injected embryos. As predicted by the microarray results, the level of islet2a in morphants was increased almost 4-fold compared to either wildtype or morphants (Fig 7). expression levels for purposes of comparison. Many of the 174 genes had expression altered in opposing directions in mutants and morphants compared to controls. (C) 201 genes were differentially expressed in morphants compared to controls. For purposes of comparison, the heat map also included mutant gene expression levels.
https://doi.org/10.1371/journal.pone.0199233.g006 Fig 7. Relative levels of islet and downstream gene transcripts in morphants vs. mutants. qPCR was used to compare expression levels of islet1, islet2a and islet2b between wildtype (wt), mutant (mut), morphant (morph) and Ctl MO-injected (Ctl MO) embryos. In islet2a mutants, islet1 levels were significantly increased by~25%; no changes in the levels of other islet transcripts were detected in mutants. Both morphants and Ctl MO-injected embryos had significantly decreased levels of islet1. The largest change detected was a~4-fold increase in islet2a levels in morphant compared to wildtype or mutant. ÃÃÃ , p<0.001; ÃÃ , p<0.003; Ã , p<0.005; ANOVA with post hoc Bonferroni. To facilitate comparisons, for each gene, expression levels were normalized to that of the control group.
https://doi.org/10.1371/journal.pone.0199233.g007 qPCR analyses also detected significantly different levels of islet1 in mutant and morphant vs. control transcriptomes (Fig 7). Compared to control, islet1 expression was increased slightly (~25%) in mutants but decreased to~67% and 33% of the wildtype level in Ctl MOinjected and morphant embryos, respectively (Fig 7). We did not detect any changes in expression levels of islet2b between any of the conditions.
In summary, the microarray results support the view that mutants and morphants differed not only in motor neuron morphology but also in gene expression profiles. Further, focused examination of islet gene expression using qPCR revealed that MOs targeting islet2a and geneediting of islet2a led to different changes in expression of islet genes.

islet expression in CaPs of islet2a wildtype and mutant embryos
The microarray and qPCR analyses revealed changes in islet gene expression in 48 hpf whole embryo RNA. However, our electrophysiological studies focused on CaPs in 24-28 hpf embryos. To identify potential changes in islet gene expression that might have occurred specifically in 24-28 hpf CaPs, we preformed RNA in situ hybridization for islet1, islet2a and islet2b. In contrast to the microarray and qPCR analyses, RNA in situ hybridization did not detect any obvious changes in expression of any islet gene in CaPs of wildtype vs. mutant embryos (Fig 8). However, the change in islet1 expression detected by qPCR was <2-fold (Fig  7), and RNA in situ hybridization is not a quantitative method. Further, as mentioned, RNA in situ hybridization focused on CaPs in 24-28 hpf embryos while the microarray and qPCR studies used RNA isolated from whole 48 hpf embryos. Overall, for CaPs, the RNA in situ hybridization results did not reveal any changes in islet2a expression or novel, ectopic expression of either islet1 or islet2b.
Similar to prdm14 mutants [18], CaPs in islet2a mutants were positive for Isl1/2 immunolabeling at a reduced level compared to wildtypes (Fig 9A, 9B and 9E vs. Fig 9C, 9D and 9F). This result differs from what has been reported for islet2a morphants [15]. While inefficient knock-out of Islet2a protein in the mutant is a potential possibility in principle, we did not detect any Isl1/2 immunolabeling in the proctodeum, a tissue that expresses islet2a but not islet1 (Fig 1E vs. Fig 1F). Despite the lack of novel expression of islet1 or islet2b in mutant CaPs (Fig 8), the results suggest that a non-Islet2a Islet protein was present in CaPs of 28 hpf mutants.

Discussion
Motor neuron subtypes differ not only with respect to peripheral axon trajectories and muscle targets but also electrical membrane properties [12,20]. However, little is known about the mechanisms that direct differentiation of vertebrate motor neuron subtype-specific electrical membrane properties. In contrast, several studies implicate LIM-HD transcription factors in specification of motor neuron subtype-specific morphological properties (for review, [2,3,41,42]. In Drosophila, islet (orthologous to zebrafish islet1 genes) is expressed in ventral, but not dorsal, motor neurons [19]. Further, loss of islet results in loss of dorsal motor neuron's distinguishing axonal morphology and larger outward current densities [20,[43][44][45]. These studies suggest that the same transcription factor code might specify motor neuron subtype-specific differentiation of electrical membrane properties as well as morphological differentiation, a possibility not yet tested in vertebrates.
Prior studies have tested for effects of Islet2a functional disruption on specification of CaP axon morphology [15,17]. Overexpression of a dominant-negative Islet2a construct, consisting of only the Islet2a LIM domains, led to disruption of CaP axon morphology [17]. Even though the LIM domains are conserved across the different Islet proteins, the effect could be rescued by overexpression of Islet2a but not Islet1or Islet2b, suggesting a specific requirement for Islet2a. These authors also showed that an islet2a T-MO had similar but less severe effects on CaP axon morphology. Similarly, Hutchinson and Eisen [15] found that a different islet2a T-MO (the same one used here) impaired CaP axon morphology to a lesser extent than did overexpression of dominant-negative Isl2 constructs [17]. However, in contrast to the results of Segawa et al. [17], Hutchinson and Eisen [15] rescued effects of the islet2a T-MO on CaP axon morphology by overexpression of either islet1 or islet2a mRNA. Murine islet2 and hb9 mutants both have a decreased number of one the the visceral motor neuron subtype, the visceral motor neuron [46][47][48]. Given that islet2 and hb9 mutants have in common a reduction in islet1 levels, Thaler et al. [47] proposed that the requirement for Islet2 reflects a need for a specific total concentration of all Islet proteins, regardless of identity as Islet1 or Islet2. Thus, Segawa et al.'s [17] findings support a specific non-redundant role for islet2a in differentiation of motor neuron subtype-specific properties, while those of Hutchinson and Eisen [15] and Thaler et al. [47] do not.
Here, we tested whether Islet2a plays a role in differentiation of another motor neuron subtype-specific characteristic, electrical membrane properties. Our prior work had demonstrated that CaPs have 2.9 and 2.5-fold larger densities of both inward and outward currents, respectively, than do other PMNs [12]. Subsequent work showed that MOs targeting islet1 had significant effects on the electrophysiological properties of several spinal neuron types [16]. However, in either islet2a mutants or morphants, we do not detect any significant differences in CaP inward or outward current densities compared to controls (Figs 2 and 3). Thus, these studies provide no evidence to support a specific role for Islet2a in specifying the >2-fold larger voltage-dependent inward and outward current densities that distinguish CaP from other PMNs.
However, while islet2a morphants had motor neuron phenotypes (Figs 4 and 5) similar to those reported previously [15] [17], CaPs and dSMNs in islet2a mutants had normal morphologies. It is possible that the islet2a T-MO was not sufficiently selective for islet2a. However, the primary sequences of the islet genes support preferential targeting of islet2a over islet1 or islet2b (Table 1; S1 Table). On the other hand, the islet2a Ctl MO also had a mild effect on SMN properties (Fig 5).
Discrepant mutant and morphant phenotypes have been reported for an ever-increasing number of genes, raising several caveats about the use of MOs [49] [56]. Conversely, gene mutation can lead to compensatory changes in expression [57] [58]. For LIM homeodomain proteins, another consideration is that their stoichiometry in transcriptional complexes influences regulation of downstream genes [59].
Mutants and morphants also differed with respect to gene expression profiles (Fig 6). However, none of the differentially-expressed genes that we identified provided insights into the different motor neuron phenotypes present in mutants compared to morphants. The different developmental stages at which we examined gene expression (48 hpf) and motor neuron morphologies (24 and 72 hpf) may have preempted identification of such genes. It is possible that if we had examined gene expression in 1 dpf embryos and/or just motor neurons, a more informative result would have been obtained.
Our manipulations targeted one member of the Islet gene family, but prior studies suggest that Islet proteins may have redundant functions [15,47]. We tested whether targeting of islet2a either by MOs or gene-editing led to genetic compensation in islet gene expression. Consistent with other differentially expressed genes (Fig 6), islet1 had slightly increased levels in mutants but decreased >2-fold in morphants (Fig 7). The small increase in islet1 expression detected in 48 hpf mutant transcriptomes might be sufficient to provide genetic compensation, at least at this stage, given that zebrafish Islet1 can substitute for Islet2a and vice versa [15].
In contrast, morphants had >2-fold decreased expression levels of islet1 (Fig 7). The mechanism underlying the changes in islet gene expression induced by the T-MO is not obvious. One possibility is that the T-MO may have targeted islet1 in such a way as to affect expression levels. Arguing against this possibility is the fact that the sequence of the T-MO would be expected to have higher specificity for islet2a vs. islet1 or islet2b (Table 1 and S1 Table). However, the Ctl MO also led to a small decrease in islet1 expression (e.g. , Fig 7), suggesting that the T-and Ctl MOs may have had effects not due to targeting of islet2a. Regardless of the mechanism(s) underlying the different changes in islet gene expression in mutants compared to morphants, the results suggest that targeting islet2a, either by MO or gene-editing, leads to changes in expression of islet1. Moreover, the effects on islet1 expression produced by MO compared to gene-editing opposed each other.
An unexpected result was the persistence of reduced, but not abolished, Isl1/2 immunoreactivity in CaPs of 24-28 hpf mutant embryos (Fig 9). Given that we did not detect Isl1/2 expression in the proctodeum (a tissue that expresses islet2a but not islet1 ; Fig 1), Islet1 or Islet2b are likely candidates for the Isl1/2 immunoreactive protein detected in CaPs of mutants. In mutants, CaPs may have upregulated either of these genes leading to Isl1/2 immunoreactivity. However, this possibility is not supported by the RNA in situ hybridization studies (Fig 8) that did not detect any novel expression of either islet1 or islet2b in CaPs. Segawa et al. [17]also found that overexpression of a dominant-negative Isl2a construct did not lead to ectopic expression of islet1 in CaPs ((17)- Fig 5B and 5D therein). Another possibility is that there was novel expression of either islet1 or islet2b in CaPs, but at levels lower than can be detected using RNA in situ hybridization.
An alternative explanation recalls that all motor neurons normally express Islet1 protein prior to differentiation as different PMN subtypes [8,10,60]. It is possible that the turn-over rate of Islet1 protein was sufficiently slow in mutant CaPs to allow persistence of the protein until at least 28 hpf. Further, Hutchinson and Eisen [15] demonstrated that overexpression of islet1 mRNA rescued spinal neuron phenotypes present in islet2a morphants, providing evidence that zebrafish Islet proteins are functionally redundant in the embryonic spinal cord. Similarly, although murine Islet2 mutants lack visceral motor neurons, Thaler et al. [47] proposed that the requirement for Islet2 reflected a need for a specific total concentration of all Islet protein, regardless of identity as Islet1 or Islet2. Given these prior results indicating that Islet1 and Islet2/a can substitute for each other, persistence of an Islet protein in mutant CaPs could have compensated for loss of Islet2a. Distinguishing between these possibilities would need tools not presently available, e.g., antibodies that distinguish between zebrafish Islet1, Islet2a and Islet2b proteins.
In summary, we find no evidence to support a specific, non-redundant role for Islet2a in differentiation of the large inward and outward conductances that distinguish CaP from other PMNs. However, in islet2a mutants, CaPs had persistent, albeit reduced, Isl1/2 immunoreactivity. In addition, zebrafish Islet1 and Islet2a proteins can functionally substitute for each other [15]. Thus, we cannot rule out the possibility that function of another (non-Islet2a) Islet protein, during the time when islet2a is normally expressed, suffices to specify the large current densities that distinguish CaPs from other PMNs. morphant (black trace) embryos. No obvious differences were noted in the amplitudes of peak inward or steady-state outward currents between conditions. The voltage protocol used to elicit currents was as described for Fig 3A. (B) The average current densities for net outward and peak inward currents recorded from vSMNs in control (n = 5 cells from 3 embryos) vs. morphant (n = 5 cells from 2 embryos) 24 hpf embryos did not differ. (C) Recordings were obtained from RB neurons in 24 (left) and 48 (right) hpf control (grey traces) and morphant (black traces) islet2a embryos. No obvious differences were noted in the amplitudes of peak inward or outward currents between conditions. The voltage protocol used to elicit currents was as described for Fig 3A. (D) The average densities for net outward and peak inward currents recorded from RBs in wildtype vs. morphant embryos did not differ at either 24 (left) Table. MO target and corresponding sequences in islet1, islet2a and islet2b. The sense sequences corresponding to each MO are shown in the top line. For each gene, the intended (islet2a) or potential (islet1, islet2b, isl1l) targets corresponding to each MO are shown. (DOCX) S1 File. RMA file. The excel file presents the RMA normalized data for all 10 samples and all genes. (XLSX) S2 File. Significant genes for morph vs. mut. The excel file presents the expression levels for 174 genes that were differentially expressed between morph and mut at Q<0.05 and with fold changes that were <-2 or >2. (XLSX) S3 File. Heat map-Morph vs mut significant genes. The file presents the expression levels for the 174 differentially expressed genes of S2 File and control values as formatted for the heat map of Fig 6B. (XLSX) S4 File. Significant genes for morph vs. ctl. The excel file presents the expression levels for 201 genes that were differentially expressed between morph and ctl at Q<0.05 and with fold changes that were <-2 or >2. (XLSX) S5 File. Heat map-Morph vs. ctl significant genes. The file presents the expression levels for the 201 differentially expressed genes of S2 File and control values as formatted for the heat map of Fig 6C. (XLSX) S6 File. Intersection of morph vs mut and morph vs ctl. The file presents the expression levels for the 64 genes that were differentially expressed between morph and mut as well as between morph and ctl at Q<0.05 and with fold changes that were <-2 or >2. (XLSX)