Hemato-vascular specification requires arnt1 and arnt2 genes in zebrafish embryos

Endothelial and hematopoietic cells are derived from a common mesodermal progenitor (Huber et al., 2004; Kataoka et al., 2011; Liao et al., 2000). In mice and zebrafish, fully multipotent hematopoietic stem cells (HSCs) are derived from specialized endothelial cells, called hemogenic endothelium (Hirschi, 2012; Kobayashi et al., 2014; Ramírez-Bergeron et al., 2004) The first hemato-vascular progenitors were traced back to the lateral plate mesoderm of zebrafish and in the extra-embryonic yolk sac mesoderm of mice (Lee et al., 1994; Palis et al., 1999; Stainier et al., 1993). The complete mechanism of the specification of these hemato-vascular progenitors from the multipotent mesoderm remains elusive in mammals.

Many genes involved in embryonic hematopoiesis and vascular development are evolutionally conserved between zebrafish and mammals like mice and humans (Begley et al., 1989; Casie Chetty et al., 2017; Dooley et al., 2005; Ferrara et al., 1996; Herbert and Stainier, 2011; Kalev-Zylinska et al., 2002; Kataoka et al., 2011; Okuda et al., 1996; Wang et al., 1996). In zebrafish, the neuronal PAS4-like gene (npas4l), formerly known as cloche, is the earliest known transcription factor to drive the specification of hemato-vascular progenitors (Reischauer et al., 2016). npas4l mutants have a distinctive phenotype as they lack most endothelial and all hematopoietic cells (Stainier et al., 1995; Thompson et al., 1998).

Npas4l is a basic helix-loop-helix Per-Arnt-Sim (bHLH-PAS) transcription factor and is thought to be the master regulator of hemato-vascular specification (Reischauer et al., 2016). bHLH-PAS transcription factors are classified as Class I or as Class II, which must dimerize to regulate gene transcription or repression (Fig. 1A) (Card et al., 2005; Edwards and Gorelick, 2022; Pongratz et al., 1998). Npas4l is a Class I bHLH-PAS protein, but its Class II dimerization partner during hemato-vascular progenitor specification is not known. We hypothesized that Arnt1 or Arnt2 is the Class II dimerization partner of Npas4l in zebrafish because arnt1 and arnt2 genes are expressed at the same stages of embryo development as hemato-vascular specification (Baranasic et al., 2022; Prasch et al., 2006; Reischauer et al., 2016; Tanguay et al., 2000).

To test this hypothesis, we examined hemato-vascular development in zebrafish with predicted loss-of-function mutations in arnt1 and arnt2 genes. We find that Arnt1 and Arnt2 are essential for hemato-vascular specification. Double homozygous arnt1;arnt2 zebrafish embryos (herein called arnt1/2 mutants) phenocopy npas4l mutants and show lack of circulating blood, absence of hematopoietic stem and progenitor cells, and dramatic reduction in the number of vascular endothelial cells. We find that expression of npas4l-target genes requires npas4l and arnt1 or arnt2. Additionally, we show that zebrafish Npas4l protein interacts with both zebrafish Arnt1 and zebrafish Arnt2 proteins in vitro. Together, our results suggest that Arnt1 and Arnt2 act redundantly and form a transcriptional complex with Npas4l that sits atop the hemato-vascular specification cascade.

To determine whether arnt1 or arnt2 is required for hemato-vascular development, we generated heritable mutant alleles for arnt1 (arnt1^bcm1 and arnt1^bcm2) and arnt2 (arnt2^bcm3) using CRISPR/Cas9. For arnt1^bcm1, we generated zebrafish with a 5 bp insertion in exon 5, which results in a frameshift at amino acid (AA) 126 and a premature stop codon in the first PAS domain at AA139. For arnt1^bcm2, we generated zebrafish with an 11 bp deletion in exon 10, which results in a frameshift at AA230 and a premature stop codon between the two PAS domains at AA250 (Fig. 1B). We did not observe any phenotypic differences between homozygous arnt1^bcm/bcm1 and arnt1^bcm2/bcm2 mutant embryos or larvae; these fish will hereafter be called arnt1 mutants. During embryogenesis and larval development, arnt1 mutants appear to have a normal phenotype and are indistinguishable from age-matched wild-type larvae (Fig. 1D,E). arnt1 mutants are viable to adulthood and are fertile, consistent with previously published arnt1 mutants (Marchi et al., 2020). Maternal zygotic arnt1 mutants did not show a developmental phenotype and were indistinguishable from wild-type or zygotic arnt1 mutants. For arnt2^bcm3, we generated zebrafish with a 3 bp deletion and 1 bp insertion in exon 9, which resulted in a missense mutation at AA286 and a premature stop codon between the two PAS domains at AA287 (Fig. 1C). During embryogenesis and larval development, arnt2^bcm3/bcm3 embryos, hereafter called arnt2 mutants, were indistinguishable from wild-type embryos (Fig. 1D-F). arnt2 mutants were viable at 5 days post fertilization (dpf) but did not survive to adulthood, consistent with other independently generated arnt2 mutant zebrafish lines (Hill et al., 2009; Löhr et al., 2009).

Because Arnt1 and Arnt2 can interact with multiple Class I bHLH-PAS proteins (Fig. 1A), we examined whether embryos that lacked both arnt1 and arnt2 had a hemato-vascular phenotype. To generate these embryos, we crossed arnt1^bcm2/+;arnt2^bcm3/+ adult fish with each other. We observed all genotypes at the expected Mendelian ratios in offspring at 24 to 48 h post-fertilization (Table 1). However, arnt1^bcm2/bcm2; arnt2^bcm3/bcm3 embryos, hereafter called arnt1/2 mutants, showed a clear phenotype (Fig. 1G). These embryos lack all circulating blood (Fig. 1H-L, and Movies 3, 4, 7 and 8). All other intermediate arnt1;arnt2 genotypes, e.g. arnt1^bcm2/+;arnt2^bcm3/+, appeared phenotypically wild-type (Fig. 1H,J,K, and Movies 1, 2, 5 and 6). These results suggest that arnt1 and arnt2 function redundantly because only the arnt1/2 double homozygous mutants show a phenotype. Additionally, the arnt1/2 mutant phenotype appears identical to the previously described npas4l mutant phenotype, also known as cloche (Reischauer et al., 2016; Stainier et al., 1995). Like npas4l mutants, arnt1/2 mutants exhibit pericardial edema and have an enlarged heart atrium, likely a secondary effect of the lack of blood and endocardial cells (Fig. 1G, Movies 3, 7). Taken together, these data suggest that arnt1 and arnt2 have a similar function to npas4l.

Next, we examined the frequencies of the genotypes in adult fish from arnt1^bcm2/+; arnt2^bcm3/+ parents. Given that arnt2 mutants are not viable as adults, these adults can be one of six possible genotypes. Interestingly, we did not observe the expected Mendelian ratios of alleles in these adult fish [Table 2; χ²(d.f.=5, n=138)=45.57, P<0.0001, n=138 offspring from two different clutches]. There is a reduction in the number of observed arnt1^bcm2/bcm2;arnt2^bcm3/+ heterozygous adults compared with the expected number. We hypothesize that the number of arnt1 and arnt2 wild-type alleles correlates with the likelihood of juvenile fish surviving to adulthood.

We hypothesized that the Arnt proteins are dimerization partners with Npas4l and that arnt1/2 mutants would exhibit the same phenotypes as npas4l mutants. Therefore, we examined arnt1/2 mutant embryos for additional phenotypes found in npas4l mutants. Previous studies determined that npas4l mutants lack almost all endothelial cells, except a population of endothelial cells located in the lower trunk and caudal tail of the larvae (Liao et al., 1997; Reischauer et al., 2016; Stainier et al., 1995). To determine whether our arnt1/2 mutants showed a similar reduction in endothelial cells, we performed whole-mount in situ hybridization on 2 dpf embryos for kdrl (flk-1). kdrl, a widely used marker of endothelial cells in blood vessels, is essential for the development of blood vessels in mice and zebrafish (Dumont et al., 1995; Liao et al., 1997; Shalaby et al., 1997). We found that arnt1 mutants, arnt2 mutants and wild-type embryos showed clear expression of kdrl in blood vessels at 2 dpf (Fig. 2A-C, n=2-5 biological replicates per genotype per probe, 10-100 embryos examined per biological replicate). In contrast, arnt1/2 mutants had a reduced number of kdrl-positive cells at 2 dpf (Fig. 2D). The kdrl-positive cells present in the arnt1/2 mutant were in the caudal tail and likely made up part of the caudal vein (Fig. 2D′). To determine whether this population of cells was present earlier in development, we bred arnt1^bcm2/+; arnt2^bcm3/+ onto a Tg(fli1:egfp^y1) transgenic background, which labels all blood vessels and the pharyngeal arches of the developing embryos. At 1 dpf, arnt1 mutants, arnt2 mutants and wild-type embryos had GPF-positive vasculature structures (Fig. 2I-K, n=2 biological replicates, with 21-131 embryos per clutch, 232 embryos total). Within the tails of these embryos, the dorsal aorta, caudal vein and intersegmental vessels are clear and present (Fig. 2I′-K′). In contrast, arnt1/2 mutants had reduced GFP-positive vascular structures (Fig. 2L). Within these embryos, only the extreme caudal vein was fluorescent (Fig. 2L′). This population of endothelial cells is also seen in npas4l mutants (Liao et al., 1997). Together, these results suggest that arnt1 or arnt2 is required for npas4l-dependent blood vessel development.

In addition to a clear reduction of endothelial cells, npas4l mutants show a reduction in the number of hematopoietic stem cells (HSCs) (Stainier et al., 1995; Thompson et al., 1998). To determine whether arnt1/2 mutants contain HSCs, we examined runx1 expression by whole-mount in situ hybridization. runx1 is a marker of HSCs and is required for definitive hematopoiesis in zebrafish (Kalev-Zylinska et al., 2002). At 48 hpf, runx1-positive HSCs are known to localize to the area between the dorsal aorta and the posterior caudal vein (Lam et al., 2009; North et al., 2009). At 2 dpf, arnt1 mutants, arnt2 mutants and wild-type embryos all contain runx1-positive cells near the dorsal aorta (Fig. 2E-G). However, arnt1/2 mutants have little or no runx1 labeling in this region (Fig. 2H). This suggests that arnt1 or arnt2 is necessary for hematopoietic cell development during embryogenesis, just like npas4l (Liao et al., 1997, 1998; Stainier et al., 1995; Thompson et al., 1998).

Hemato-vascular progenitor specification from the multipotent mesoderm begins in mid-gastrula, around shield stage, and peaks during segmentation (Vogeli et al., 2006). Npas4l is thought to directly upregulate target genes that cause cells to differentiate into endothelial cells and into hematopoietic stem and progenitor cells. If Npas4l, Arnt1 and Arnt2 act in tandem to regulate target gene expression, then we expect that Npas4l-target genes would be downregulated in arnt1/2 mutants. Therefore, we hypothesized that Npas4l-target genes would be downregulated in arnt1/2 mutants. We focused on four putative npas4l target genes known to be important for the hemato-vascular lineage: etsrp (etv2), tal1, lmo2 (scl) and egfl7 (Marass et al., 2019; Sumanas and Lin, 2006; Sumanas et al., 2005). We bred arnt1^bcm2/+; arnt2^bcm3/+ adults with each other to generate embryos that were arnt1/2 mutant, arnt1 mutant, arnt2 mutant, wild type and all other intermediate genotypes (referred to as arnt1/2 siblings). At the 2- to 6-somite stages (about 10-14 hpf), npas4l mutant embryos show a reduction of expression in etsrp, tal1 and lmo2 (Reischauer et al., 2016; Thompson et al., 1998). We examined the expression of these genes in arnt1/2 mutants and arnt1/2 siblings between the 4- and 6-somite stage (Fig. 3). We found that 7% of embryos from arnt1^bcm2/+;arnt2^bcm3/+ parents showed reduced expression of etsrp compared with siblings (Fig. 3A,B, n=4 biological replicates, 80-250 embryos per clutch, 604 embryos total). This did not differ from the expected Mendelian ratio of arnt1/2 mutants (6.25%; binomial test, P>0.05). To confirm that these embryos were in fact arnt1/2 mutants, after whole-mount in situ hybridization we genotyped 24 embryos with reduced expression of etsrp and found that 100% were arnt1/2 mutants. We also genotyped 26 embryos with normal expression of etsrp, only two of which were arnt1/2 mutants (Table S2). To examine whether there is an association between the number of mutant alleles and the observed phenotype, we performed a Spearman Rank correlation test (see Materials and Methods for details) and found that there is a statistically significant correlation between the number of mutant alleles and the observed phenotype (Spearman correlation, r=−0.9, P<0.001). We conclude that the expression of etsrp is reduced in arnt1/2 mutants compared with wild type.

We found that 7% of embryos from arnt1^bcm2/+; arnt2^bcm3/+ parents showed reduced expression of tal1 compared with arnt1/2 siblings (Fig. 3C,D, n=6 biological replicates, 60-300 embryos per clutch, 681 embryos total). This did not differ significantly from the expected Mendelian ratio of arnt1/2 mutants (binomial test, P>0.05). To confirm these embryos were in fact arnt1/2 mutants, we genotyped 18 embryos with reduced expression and found that 83% (15 of 18 embryos) of these embryos were arnt1/2 mutants (Table S3). We genotyped 11 embryos with normal expression of tal1, none of which were arnt1/2 mutants (Table S3). To examine whether there is a relationship between the number of mutant alleles and the observed phenotype, we performed a Spearman Rank correlation test and found that there is a statistically significant correlation between the number of mutant alleles and the observed phenotype (Spearman correlation, r=−0.6, P<0.005). We conclude that the expression of tal1 is reduced in arnt1/2 mutants compared with wild type.

We found that 11% of embryos from arnt1^bcm2/+;arnt2^bcm3/+ parents showed reduced expression of lmo2 compared with siblings (Fig. 3D-E, n=2 biological replicates, 95-175 embryos per clutch, 270 embryos total). This did significantly differ from the expected 6.25% Mendelian ratio of arnt1/2 mutants (binomial test, P<0.01). To test for a phenotype-genotype correlation, we genotyped 185 embryos after labeling for lmo2: 160 embryos had normal lmo2 expression and 25 embryos had reduced lmo2 expression. Of the 25 embryos with reduced expression of lmo2, we observed eight embryos that were arnt1/2 mutants, eight embryos with three mutant alleles, four embryos with two mutant alleles, three embryos with one mutant allele, and two wild-type embryos (Table S4). Of the 160 embryos with normal expression of lmo2, five embryos were arnt1/2 mutants, 36 embryos had three mutant alleles, 64 embryos had two mutant alleles, 46 embryos had one mutant allele, and nine were wild-type embryos (Table S4). To examine whether there is a relationship between the number of mutant alleles and the observed phenotype, we performed a Spearman rank correlation and found that there is a statistically significant correlation between the number of mutant alleles and the observed phenotype (Spearman correlation, r=−0.2, P<0.01). We conclude that the expression of lmo2 is reduced in arnt1/2 mutants compared with wild type.

We examined the expression of egfl7 in 10-somite stage embryos from arnt1^bcm2/+; arnt2^bcm3/+ parents. We observed egfl7 expression in regions of the ALPM and PLPM in wild type and in 95% of the embryos from arnt1^bcm2/+; arnt2^bcm3/+ parents, but no expression of egfl7 in the anterior of 5% of embryos from arnt1^bcm2/+; arnt2^bcm3/+ parents, n=11 out of 214 embryos (Fig. S1, n=2 biological replicates, 94-120 embryos per clutch, 214 total embryos, and age-matched wild-type embryos, n=1 biological replicate, 28 embryos). This did not differ from the expected Mendelian ratio of arnt1/2 mutants (6.25%; binomial test, P>0.05). To examine whether there is a relationship between the number of mutant alleles and the observed phenotype, we performed a Spearman Rank correlation test and found that there was a statistically significant correlation between the number of mutant alleles and the observed phenotype (Spearman correlation, r=−0.5, P<0.001) (Table S5). We conclude that the expression of egfl7 is reduced in arnt1/2 mutants compared with wild type.

These results support the idea that npas4l, arnt1 and arnt2 regulate identical target genes for hemato-vascular specification. arnt1 and arnt2 could interact with npas4l at the same level of the pathway, consistent with bHLH-PAS transcription factors acting as heterodimers. Alternatively, it is possible that arnt1 and arnt2 act upstream of npas4l. To distinguish among these ideas, we tested whether npas4l expression was reduced in arnt1/2 mutants. At the 2- to 4-somite stage, arnt1/2 mutants exhibit normal expression of npas4l in the lateral plate mesoderm, indistinguishable from age-matched wild-type embryos (Fig. S2, Table S6, n=2 biological replicates, 68-100 embryos per clutch, 168 total embryos). This result indicates that npas4l expression occurs independently of arnt1 and arnt2. We conclude that npas4l does not act downstream of arnt1 or arnt2. Taken together, our results suggest that arnt1 and arnt2 act in conjunction with npas4l during hemato-vascular specification.

If arnt1, arnt2 and npas4l act together to specify hemato-vascular progenitor cells, then all three genes should be expressed in the same cells in the lateral plate mesoderm. To determine whether arnt1, arnt2 and npas4l are expressed in the same cells, we performed whole-mount in situ hybridization chain reaction (WIHCR) on wild-type zebrafish embryos at the 4-somite stage (∼10.5 hpf). At this stage, npas4l is robustly expressed in the LPM (Reischauer et al., 2016). We found that arnt1 and arnt2 are expressed ubiquitously throughout the embryo at this stage, although both genes appear to be expressed at low levels (Fig. 4A and Fig. S3A, n=3 biological replicates, 5-9 embryos per replicate, 22 embryos total).

Because no previous publications have examined the spatial distribution of arnt1 or arnt2 mRNA in embryos younger than 24 hpf (Löhr et al., 2009), we tested whether background or autofluorescence influenced our results. Samples probed for arnt1, arnt2 and npas4l were compared with negative control embryos, embryos from the same clutch that underwent the same WIHCR protocol, but in the absence of anti-sense RNA probe sets (Fig. 4B and Fig. S3B, n=3 biological replicates, 6-12 embryos per replicate, 28 embryos total). We verified that the observed npas4l, arnt1 and arnt2 expression was specific using two different image analysis tools. First, we analyzed representative images in three dimensions using the ImageJ plug-in ‘3D Count Objects’ (Bolte and Cordelières, 2006). The number of objects above a set threshold was measured for each channel and then normalized to background by subtracting the average number of objects detected in the corresponding channel of the negative control embryos. npas4l-, arnt1- and arnt2-positive objects were all detected in the posterior of the embryos (Fig. 4C-E). Second, we analyzed the expression of arnt1 and arnt2 in the presumptive LPM by drawing regions of interest (ROIs) encompassing the observed npas4l signal, and then measured the percent area of expression and the mean gray intensity above the threshold for each channel. The integrated intensity was calculated by multiplying the percent area by the mean grey intensity, thus taking both brightness and area of signal into consideration (Fig. 4F-H, Fig. S3C-E). We found that npas4l, arnt1 and arnt2 all showed statistically significantly greater expression in experimental embryos compared with the negative control, as measured by both number of objects and integrated intensity of objects in the LPM (Fig. 4F-H, npas4l Welsh's t-test, P<0.0001; arnt2 Welsh's t-test, P<0.0001; arnt1 Mann–Whitney test, P<0.0001).

We asked whether the expression of npas4l, arnt1 and arnt2 was the same in the anterior LPM (ALPM) and posterior LPM (PLPM). npas4l and its associated downstream targets exhibit more expression in the PLPM compared with the ALPM in embryos throughout segmentation (Liao et al., 1997, 2000; Marass et al., 2019; Mattonet et al., 2022; Qian et al., 2005; Stainier et al., 1995). In 4- to 6-somite stage embryos, we found that expression of npas4l and arnt2 was greater in the PLPM than in the ALPM, but the expression of arnt1 was similar in both regions (Fig. S3C,E, npas4l Welsh's t-test, P<0.0001; arnt2 Welsh's t-test, P<0.005). We conclude that arnt1 and arnt2 are expressed within the npas4l-positive LPM, consistent with our hypothesis that all three genes together regulate the specification of hemato-vascular progenitor cells.

npas4l is not expressed until tailbud stage (Reischauer et al., 2016). A previous study demonstrated that ectopic expression of npas4l in shield-stage embryos, before the onset of expression of endogenous npas4l, resulted in upregulation of and ectopic expression of early hemato-vascular genes etsrp and tal1 (Reischauer et al., 2016). We suspect that upregulation of npas4l target genes occurred because arnt1 and arnt2 are ubiquitously expressed at shield stage. Single cell RNA-sequencing studies detected arnt1 and arnt2 mRNA throughout gastrulation, which encompasses shield stage, in zebrafish embryos (Baranasic et al., 2022). To test whether expression of npas4l-target genes requires npas4l, arnt1 and arnt2, we injected wild-type embryos at the one-cell stage with either npas4l mRNA or arnt1 and arnt2 mRNA. The embryos were fixed at shield stage, and assayed for expression of tal1 or lmo2 using whole-mount in situ hybridization. Wild-type embryos injected with npas4l mRNA exhibited robust expression of tal1 compared with uninjected control embryos (Fig. 5A,B). In contrast, embryos injected with arnt1 and arnt2 mRNA showed little or no expression of tal1, similar to uninjected embryos (Fig. 5C). This suggests that neither arnt1 nor arnt2 can drive expression of tal1 in the absence of npas4l. To directly test whether arnt1 and arnt2 can drive expression of tal1 in the absence of npas4l, we injected npas4l mutant embryos with npas4l RNA or with arnt1 and arnt2 RNA (Fig. 5D-F). Embryos injected with npas4l mRNA exhibited robust expression of tal1 (Fig. 5E), consistent with previously published results (Reischauer et al., 2016). However, all the embryos (derived from npas4l^s5/+ parents crossed to each other) injected with arnt1 and arnt2 mRNA showed little to no expression of tal1 (Fig. 5F). As 25% of these embryos are npas4l^s5/s5, we conclude that arnt1 and arnt2 cannot drive tal1 expression in the absence of npas4l. Finally, to test whether npas4l upregulates target genes in the absence of arnt1 and arnt2, we injected arnt1/2 mutant embryos with npas4l mRNA. We found that arnt1/2 siblings showed upregulation of the npas4l-target gene lmo2, but arnt1/2 mutants did not show upregulation of lmo2 (Fig. 5G,H). From these results, we conclude that expression of npas4l and either arnt1 or arnt2 is required for upregulation of npas4l-target genes.

Npas4l is a Class I bHLH-PAS domain transcription factor (Reischauer et al., 2016) and, as such, must heterodimerize with a Class II bHLH-PAS transcription factor, such as Arnt1 and Arnt2, to regulate gene transcription. Considering that (1) bHLH-PAS transcription factors function as dimers, (2) arnt1 and arnt2 fail to upregulate hemato-vascular genes in the absence of npas4l, and (3) npas4l fails to upregulate hemato-vascular genes in arnt1/2 mutants, we hypothesize that Arnt1 and Arnt2 proteins form a transcription complex with Npas4l protein to regulate gene expression. To determine whether Npas4l interacts with Arnt1 or Arnt2, we performed a co-immunoprecipitation experiment. Owing to the absence of commercially available antibodies that recognize zebrafish Npas4l, Arnt1 or Arnt2 proteins, we co-expressed tagged versions of zebrafish npas4l (Npas4l-V5), arnt1 (Arnt1-myc) and arnt2 (Arnt2-HA) in HEK293T cells. We performed co-immunoprecipitation studies and found that Arnt1-myc interacts with Npas4l-V5 and that Arnt2-HA interacts with Npas4l-V5 (Fig. 6, n=2 biological replicates per condition). These results suggest that Npas4l uses either Arnt1 or Arnt2, or both, as binding partners. Furthermore, these data indicate that Arnt1 and Arnt2 may be functionally interchangeable and redundant, at least in the case of Npas4l.

Our results demonstrate that Arnt1 and Arnt2 are functionally interchangeable and redundant when interacting with Npas4l in the context of hemato-vascular development. It is unclear whether Arnt1 and Arnt2 are functionally interchangeable with all Class I bHLH-PAS domain transcription factors or just with Npas4l. To determine whether arnt1/arnt2 redundancy is unique to Npas4l, we examined whether arnt1 and arnt2 are interchangeable with a different class I bHLH-PAS transcription factor: the aryl hydrocarbon receptor 2 (Ahr2). Previous studies demonstrate that Ahr2 can bind DNA in the presence of either Arnt1 or Arnt2 in vitro (Andreasen et al., 2002; Hirose et al., 1996; Tanguay et al., 2000), but whether Arnt1 and Arnt2 bind Ahr2 interchangeably in vivo is not known. Ahr2 is a ligand-dependent transcription factor. A known ligand of Ahr2 is 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD), which we have previously shown induces cardiotoxicity in zebrafish larvae and requires ahr2 (Souder and Gorelick, 2019). We asked whether TCDD causes cardiotoxicity in arnt1 or arnt2 single mutants, reasoning that if Ahr2 interacts with Arnt1 or Arnt2 interchangeably, then only arnt1/2 double mutants, but not single mutants, would be resistant to TCDD exposure. We treated clutches of larvae from wild-type parents (n=3 biological replicates, 10-60 larvae per clutch per drug treatment, 157 larvae total), arnt1 heterozygous parents (n=3 or 4 biological replicates per allele, 21-36 larvae per clutch per drug treatment, 533 larvae total, 248 larvae from allele bcm1 and 285 larvae allele from bcm2) or arnt2 heterozygous parents (n=3 biological replicates, 40-45 embryos per clutch per drug treatment, 253 larvae total) with either 0.1% DMSO (vehicle) or with 10 ng/ml TCDD from 1 dpf to 3 dpf (Fig. S4). Wild-type larvae exposed to TCDD exhibited pericardial edema and abnormal heart looping, consistent with previous results (Fig. 7A,D). arnt2 mutants treated with TCDD exhibited a phenotype similar to wild type (Fig. 7C,F). In contrast, arnt1 mutants treated with TCDD showed normal heart looping and no pericardial edema, similar to vehicle-treated larvae (Fig. 7B,E). This demonstrates that arnt1, but not arnt2, is required for Ahr2-dependent TCDD toxicity. We conclude that arnt1 and arnt2 are not always functionally redundant in vivo. Certain Class I bHLH-PAS transcription factors may prefer to function with a specific Class II bHLH-PAS transcription faction.

We find that Arnt1 and Arnt2 are obligate co-regulators of Npas4l. Together, npas4l, arnt1 and arnt2 genes regulate specification of hemato-vascular cells. npas4l, arnt1 and arnt2 are each necessary, but on their own not sufficient, for hemato-vascular specification. Npas4l serves as a master regulator, with Arnt proteins acting as permissive co-regulators. Arnt1 and Arnt2 are functionally redundant in the case of Npas4l but are not functionally redundant in the case of Ahr2, where there is a preference for Ahr2-Arnt1 complex.

Class I bHLH-PAS proteins must dimerize with Class II bHLH-PAS proteins to transcribe their target genes (Card et al., 2005; Ema et al., 1996; Hirose et al., 1996; Jiang et al., 1996; Lees and Whitelaw, 1999; Michaud et al., 2000; Wu and Rastinejad, 2017). It has previously been established that Npas4l, a Class I bHLH-PAS transcription factor, is required for hemato-vascular development in zebrafish (Reischauer et al., 2016). To determine the Class II bHLH-PAS binding partner of Npas4l, we generated arnt1 mutant and arnt2 mutant zebrafish, and expected one of these mutants to mimic the npas4l mutant phenotype. However, both arnt1 mutant and arnt2 single mutant larvae were viable to 5 dpf and were grossly indistinguishable from wild-type embryos. Only arnt1/2 double mutant embryos exhibited a cardiovascular phenotype. Our results demonstrate that double homozygous arnt1/2 mutant embryos have a similar phenotype to npas4l mutants (Reischauer et al., 2016; Stainier et al., 1995; Thompson et al., 1998) and that Npas4l can function with either Arnt1 or Arnt2. Thus, Arnt1 and Arnt2 act redundantly to specify hemato-vascular progenitor cells in zebrafish.

Are arnt1 and arnt2 functionally redundant with all Class I bHLH-PAS transcription factors? We argue not. In the case of the class I transcription factor Ahr2, both Arnt1 and Arnt2 bind Ahr2 in vitro (Lanham et al., 2011; Tanguay et al., 2000). However, we find that only Arnt1, and not Arnt2, is required for Ahr2-dependent cardiotoxicity in zebrafish embryos. This argues that, in vivo, Ahr2 preferentially interacts with Arnt1 over Arnt2. Another class I protein, Sim1a, appears to preferentially interact with Arnt2 in vivo. arnt2 mutant zebrafish embryos have fewer dopaminergic neurons in the brain compared with wild-type embryos (Löhr et al., 2009). The arnt2 mutant phenotype is similar to the phenotype when Sim1a expression is knocked down (Löhr et al., 2009). These results suggest that, in the case of the growth and/or survival of dopaminergic neurons, Sim1a preferentially interacts with Arnt2 versus Arnt1. To what degree this is due to differences in expression levels, versus an intrinsic structural preference for a particular protein-protein interaction, requires further study. Such interactions may also depend on the cell type and developmental stage where the bHLH-PAS proteins function. Arnt1 and Arnt2 appear to be expressed ubiquitously during embryonic development, which suggests that such interactions may be due to factors other than expression levels.

Endothelial cell development and hematopoietic cell development are closely related during embryogenesis. Both cell types primarily arise from the lateral plate mesoderm (LPM) and share a common progenitor (Qian et al., 2005; Reischauer et al., 2016; Stainier et al., 1995; Thompson et al., 1998). Additionally, during definitive hematopoiesis, some HSCs are derived from specialized endothelial cells, called hemogenic endothelium (Boisset et al., 2010; Kissa and Herbomel, 2010). Both npas4l and arnt1/2 mutants lack all blood cells but do have a small population of endothelial cells, which is restricted to the most posterior region of the animal (Fig. 2; Reischauer et al., 2016). In zebrafish, there is a population of endothelial cells in the caudal region of the embryo that are not derived from the LPM and differentiate in the absence of npas4l expression (Pak et al., 2020). We hypothesize that arnt1 and arnt2 are required for the differentiation of endothelial cells from the LPM but are not required for the differentiation of endothelial cells, such as the caudal population, from outside the LPM. The transcriptomic signature of the caudal population of endothelial cells was distinct from endothelial cells that require npas4l for differentiation, with genes for somitogenesis and neurogenesis highly enriched compared with LPM-derived endothelial cells (Pak et al., 2020). Based on this transcriptome analysis, we speculate that the caudal population of endothelial cells fails to express genes required for hemogenic endothelium and that no blood cells are derived from the caudal population of endothelial cells. This is consistent with the observation that the caudal populations of endothelial cells, but not blood cells, are present in npas4l or arnt1/2 mutants.

Arnt proteins interact with multiple class I bHLH-PAS domain proteins to regulate cardiovascular development. Our results focused on the role of Arnt proteins in hemato-vascular specification; however Arnt proteins are likely involved in later steps of endothelial cell development. Differentiation of hemogenic endothelium in zebrafish and in mice is dependent upon several Class I bHLH-PAS transcription factors: Hypoxia inducible factor 1 alpha (mouse Hif1a, zebrafish genes hif1aa and hif1ab) and hypoxia inducible factor 2 alpha (mouse Hif2a, zebrafish genes epas1a and epas1b) (Gerri et al., 2018; Ramírez-Bergeron et al., 2004). Given that these HIF proteins are Class I bHLH-PAS transcription factors, it is likely that one or both Arnt proteins are required for HIF-dependent differentiation of hemogenic endothelium.

The function of Arnt proteins in zebrafish cardiovascular development is likely conserved in mice. Arnt1 is essential for the viability of adult hematopoietic stem cells in mice and is required for the survival of hematopoietic progenitor cells in the fetal mouse liver (Krock et al., 2015). Homozygous Arnt1 mutant mice are not viable beyond E10.5 and show severe defects in vascularization, suggesting that Arnt1 is essential for normal vascular development (Kozak et al., 1997; Maltepe et al., 1997). The degree to which Arnt2 contributes to hemato-vascular development in mice is not well understood. Arnt2 has primarily been studied within the context of neural development (Hao et al., 2013; Hosoya et al., 2001; Jain et al., 1998); however, there is some evidence to suggest that Arnt1 and Arnt2 may exhibit functional redundancy during development of non-neural tissues (Keith et al., 2001; Sekine et al., 2006). We observed lower than expected numbers of arnt1^bcm2/bcm2; arnt2^bcm3/+ adult zebrafish derived from arnt1^bcm2/+; arnt2^bcm3/+ parents. This suggests that arnt1 and arnt2 have a dose-dependent effect on the survival or fitness of the animal, where anything less than four wild-type alleles (two arnt1 alleles, two arnt2 alleles) is deleterious. A similar dose-dependent effect on survival can be seen in mouse embryos from Arnt1^+/−; Arnt2^+/− parents (Keith et al., 2001). The mechanisms behind these dose-dependent gene effects are not known.

Although zebrafish arnt1 and arnt2 genes are conserved in mammals, there is no known equivalent of zebrafish npas4l in mammalian genomes. Zebrafish Npas4l shows the highest sequence homology to mouse Npas4 (Reischauer et al., 2016). Although Npas4 is known to dimerize with Arnt1 and Arnt2 in vivo (Brigidi et al., 2019), mammalian Npas4 is unlikely to be the functional equivalent to zebrafish Npas4l because homozygous Npas4 mutant mice are viable to adulthood and do not show any overt vascular or hematopoietic defects (Bloodgood et al., 2013). Given our results that npas4l requires arnt1 or arnt2 to act as co-regulators of hemato-vascular specification, and the abundance of conserved genes that regulate vascular development in zebrafish and humans (Hogan and Schulte-Merker, 2017), we speculate that the functional mammalian equivalent of npas4l is a Class I bHLH-PAS transcription factor, or multiple such factors, that function together with Arnt1 or Arnt2.

Adult zebrafish were raised at 28.5°C on a 14 h light, 10 h dark cycle in the BCM Zebrafish Research Facility in an Aquaneering recirculating water system (San Diego, CA, USA) and a Tecniplast recirculating water system (West Chester, PA, USA). Wild-type zebrafish were AB strain (Westerfield, 2000), and arnt1 and arnt2 mutant lines were generated on the AB strain. Once the mutant lines were established, they were bred onto the Tg(Fli1:EGFP)^y1 transgenic background (Lawson and Weinstein, 2002). All procedures were performed in accordance with and approved by the BCM Institutional Animal Care and Use Committee.

Adult zebrafish were allowed to spawn naturally in pairs or in groups. Embryos were collected in intervals of 20 min to ensure precise developmental timing or staged after collection, placed in 60 cm² Petri dishes at a density of no more than 100 per dish in E3B media (60× E3B: 17.2 g NaCl, 0.76 g KCl, 2.9 g CaCl₂-2H₂O and 2.39 g MgSO₄ dissolved in 1 l Milli-Q water; diluted to 1× in 9 l Milli-Q water plus 100 μl 0.02% Methylene Blue), and then stored in an incubator at 28.5°C on a 14 h light, 10 h dark cycle.

Genomic DNA was isolated from tail biopsies from individual adult zebrafish and incubated in 50 μl ELB [10 mM Tris (pH 8.3), 50 mM KCl and 0.3% Tween 20] with 0.5 μl proteinase K (800 U/ml, NEB) in 96-well plates, one sample per well, at 55°C for 8 h. Proteinase K was inactivated by incubation at 98°C for 10 min and DNA was stored at −20°C. Genotyping was performed by PCR and high-resolution melting curve analysis (HRMA) as described previously (Parant et al., 2009; Romano et al., 2017). All melting curves were generated with a Bio-Rad CFX96 or CFX Opus 96 Real-Time System over a 70-95°C range and analyzed with the Bio-Rad CFX Manager 3.1 or the Bio-Rad CFX Maestro 4.1 software. All mutations were confirmed by TA cloning and sequencing. All primers used for HRMA can be found in Table S1.

Cas9 mRNA and gRNAs for arnt1 and arnt2 mutants were generated as previously described (Romano et al., 2017; Souder and Gorelick, 2019). Cas9 mRNA was transcribed from a linearized pT3TS-nCas9n plasmid (Addgene 46757; Jao et al., 2013). The target sequences were identified using CHOPCHOP (Labun et al., 2016; Labun et al., 2019). Oligonucleotides were annealed to each other and cloned into a pT7-gRNA plasmid (Addgene 46759; Jao et al., 2013). Oligonucleotides used are shown in Table S1. gRNAs were synthesized from plasmids using a MEGAshortscript T7 Transcription Kit (Invitrogen AM1354) and purified.

One-cell-stage embryos were injected using glass needles pulled on a Sutter Instruments Fleming/Brown Micropipette Puller, model P-97 and a regulated air-pressure micro-injector (Harvard Apparatus, PL1–90). Each embryo was injected with a 1 nl solution containing one gRNA for arnt1 and one gRNA for arnt2 (30 ng/μl per target), Cas9 mRNA (150 ng/μl), and 0.1% Phenol Red. Mixtures were injected into the yolk of each embryo. Approximately 100 injected embryos per gRNA pair were raised to adulthood and crossed to AB zebrafish to generate F1 embryos. F1 offspring with heritable mutations were sequenced to identify mutations predicted to cause loss of function.

Embryos were imaged with a Nikon SMZ25 microscope equipped with a Hamamatsu ORCA-Flash4.0 digital CMOS camera or with a Nikon SMZ18 microscope equipped with a Nikon DS-Fi3 camera. Images were equally adjusted for brightness and contrast in Adobe Photoshop CC 2020. Embryos and larvae were anesthetized with 0.04% tricaine and imaged in Petri dishes containing E3B. All embryos were genotyped after imaging.

We used the following DNA plasmids to generate RNA probes: egfl7, lmo2 (Sun et al., 2018), etsrp (a gift from Dr Nathan Lawson, University of Massachusetts, USA; Addgene 49005; Moore et al., 2013), tal1 (a gift from Dr Arne Lekven, University of Houston, TX, USA), npas4l (a gift from Dr Leonard Zon, Harvard University, Cambridge, MA, USA) (Reischauer et al., 2016), and runx1 and kdrl (gifts from Dr Mary Goll, University of Georgia, Athens, GA, USA) (Li et al., 2015). For lmo2 and egfl7, RNA was extracted from 10 pooled 1 dpf wild-type embryos using TRIzol Reagent (ThermoFisher Scientific, 15596026) and purified using RNA clean & concentration kit (Zymo Research, R1019). cDNA was synthesized using iScript cDNA synthesis (BioRad, 1708890). lmo2 and egfl7 were PCR amplified and Topo TA cloned into pCRII vector (Invitrogen, K4600-01). Primers used to amplify lmo2 and egfl7 can be found in Table S1. All DNA clones were verified by sequencing. Digoxigenin-labeled antisense RNA probes were synthesized using T7 or SP6. Colorimetric in situ hybridization was performed on fixed zebrafish embryos and larvae as previously described, with 5% dextran in the hybridization buffer (Lauter et al., 2011; Thisse and Thisse, 2008). Following in situ hybridization, embryos were washed several times in 1×PBST (phosphate-buffered saline, 0.01% Tween), cleared with glycerol or mounted in 3% methylcellulose on a glass coverslip, and imaged on either a Nikon SMZ18 stereoscopic microscope or an inverted Nikon ECLIPSE Ti2-E microscope. Embryos were then genotyped. If embryos were older than 24 hpf, genomic DNA was extracted in 20 μl ELB [10 mM Tris (pH 8.3), 50 mM KCl and 0.3% Tween 20] with 0.2 μl proteinase K (800 U/ml, NEB) in 96-well plates, one sample per well, at 55°C for 2 h. If the embryos were 24 hpf or younger, they were first de-crosslinked in 300 mM NaCl for 4 h at 65°C and held at 4-8°C until DNA could be extracted. Genomic DNA was extracted using the HotSHOT protocol (Dobrzycki et al., 2018). Briefly, embryos were suspended in 10-20 μl of lysis buffer (50 mM NaOH) and incubated at 95°C for 30 min, cooled to 4°C, then one-tenth the volume of neutralization buffer (1 M Tris-HCl at pH 8.0) was added. Embryo lysate was incubated at 4°C overnight before starting PCR reactions. DNA regions of interest were amplified using one of two methods. The genes of interest were PCR amplified using JumpStart REDTaq ReadyMix (Sigma-Millipore, P0982) according to the manufacturer's protocols. Alternatively, the DNA regions of interest were PCR amplified in two successive reactions using HS Taq (Takara, R007). The first PCR amplified a larger fragment of arnt1 (about 700 bp) or arnt2 (about 970 bp). The product of the larger fragment was then used as a template to amplify a smaller fragment of arnt1 (about 400 bp) or arnt2 (about 650 bp). After PCR amplification, the PCR products were sequenced to determine the genotype. All primer sequences can be found in Table S1.

HCR Antisense RNA probe sets for arnt1, arnt2 and npas4l, and HCR amplifiers and HCR buffers were obtained from Molecular Instruments. Each gene had a probe set size of 20. Whole-mount in situ hybridization chain reaction (WIHCR) was performed on fixed AB embryos between 4 and 6 somites, as previously described (Choi et al., 2018; Ibarra-García-Padilla et al., 2021). Following HCR, embryos were washed several times in 1×PBST and cleared in glycerol. All embryos were mounted in 75% glycerol and imaged on a Nikon Ti2-E inverted microscope with Yokogawa CSU-W1 spinning disk confocal and Photometrics Prime95b sCMOS camera.

All images used in processing and analysis were captured on the same day, using identical acquisition controls. We selected only images where a clear, bilateral view of the presumptive lateral plate mesoderm (LPM) and presumptive notochord was available (n=6 embryos from a single clutch and n=4 negative control embryos from the same clutch). Images of anterior and posterior views of the embryos were collected. To calculate the number of 3D objects in a stack, we used Image J, FIJI version 2.3.0/1.53q (Schindelin et al., 2012) and the 3D-OC plug-in (Bolte and Cordelières, 2006). A custom threshold for each channel was applied to each image to filter out background autofluorescence seen in the negative control embryos. The number of detected objects for probed samples was then normalized to a negative control by subtracting the average number of detected objects from control embryos from the number of detected objects in probed embryos. We used ImageJ to calculate the adjusted integrated intensity of probed and control embryos (Schindelin et al., 2012). First, maximum intensity projections (MIP) of the z-stacks were created, then the region of interest (ROI) was drawn around the areas with the highest concentration of npas4l expression (presumptive LPM) or around the corresponding region of the presumptive LPM in the negative control embryos. The percent area of pixels and the mean gray value for each channel were calculated and limited to the threshold. The adjusted integrated density was then calculated by multiplying the percent area and the mean gray value.

Zebrafish embryos were stained at ∼72 h post fertilization. Embryos were dechorionated using 0.2% pronase in E3B media for 5-10 min and were gently triturated with a 1000 µl pipette. Embryos were transferred to glass vials and submerged in o-dianisidine working solution, prepared as described previously (Kafina and Paw, 2018). Vials were kept in the dark for 5 min during staining. Samples were then fixed in 4% formaldehyde for 30 min at room temperature. Embryos were washed in PBS-Tween 0.1% before mounting for imaging in 3% methylcellulose on a glass dish or coverslip. Embryos were genotyped after imaging.

Cloning of plasmids for expression in cultured cells was performed as follows. For npas4l-V5, we took pCS2+npas4lORF (a gift from Dr Leonard Zon) that contains the open reading frame (ORF) of zebrafish npas4l downstream of the CMV promoter. This plasmid was modified to include a C-terminal V5 tag using Q5 Site-Directed Mutagenesis Kit (New England Biolabs, E0554S). For arnt1-myc, we commercially synthesized a plasmid containing CMV promoter driving a codon-optimized zebrafish arnt1 ORF (ENSDART00000081852.5) with a C-terminal myc tag (Twist Bioscience, San Francisco, CA, USA). For arnt2-myc, we commercially synthesized a gene block of the zebrafish arnt2 ORF (ENSDART00000158162.2, Integrated DNA Technologies). The gene block was integrated into a plasmid containing a CMV promoter using pcDNA3.3-TOPO TA cloning kit (Invitrogen, K8300-01) in accordance with manufacturer's instructions. A C-terminal HA tag was added using Q5 Site-Directed Mutagenesis Kit (New England Biolabs, E0554S). For microinjection, an SP6 promoter was inserted upstream of the transcriptional start using Q5 Site-Directed Mutagenesis Kit (New England Biolabs, E0554S). All plasmids were confirmed by sequencing.

HEK293T cells (a gift from Dr Charles Foulds, Baylor College of Medicine, Houston, TX, USA), were thawed and maintained at 37°C in humidified 5% CO₂ atmosphere incubator. Cells were cultured in DMEM supplemented with 10% fetal bovine serum (FBS; Gibco, 2062564) and 1% penicillin-streptomycin (Gibco, 15140-122). After 24 h of passaging in a 100 mm plate at a density of 2.2×10⁶ cells per dish and reaching at least 70% confluency, the media was changed to DMEM supplemented with 1% dialyzed FBS. Cells were transiently transfected with 12 μg of total DNA and Lipofectamine 2000 according to the manufacturer's instructions (Invitrogen, 2400439) with npas4l-v5 together with either arnt1-myc or arnt2-HA plasmid DNA.

Transiently transfected cells were washed with ice-cold PBS and lysed in 500 μl of 1×RIPA buffer containing Pierce Protease (ThermoScientific, A32955). Cells were sonicated six times for 30 s at a time using a Bioruptor 300 Machine (Diagenode), then centrifuged for 10 min at 10,000 g at 4°C. The lysate (supernatant) was removed from the pellet and placed in fresh 1.5 ml Eppendorf tube. Cell lysate was precleared with Protein A beads (Invitrogen 159 18-014) for 30 min at 4°C, incubated with either 5 μg of V5 antibody (ThermoFisher Scientific R960-25) or myc antibody (SantaCruz, SC-40), diluted 1:200 for 1 h, then immunoprecipitated with 20 μl of Protein A beads for 1 h at 4°C. The beads were washed three times with RIPA buffer and once with PBS, and finally suspended in 30 μl of PBS. Equal volume of 2×Laemmli Sample Buffer (BioRad, 1610737) was added to each sample and boiled for 5 min at 95°C then placed on ice. Samples were run on 4-20% Mini-PROTEAN TGX gels (BioRad, 4561093) and transferred onto PVDF membrane using Trans-Blot Turbo Transfer System (BioRad) using the pre-defined Mixed MW program. The membranes were washed several times with 1×Tris-buffered saline, 0.1% Tween 20 (TBST) (GBiosciences, R043) and incubated for 1 h with agitation in 5% milk in 1×TBST. Membranes were probed with 1:1000 V5-HRP (ThermoFisher Scientific, R960-25), 1:1000 myc-HRP (SantaCruz, SC-40) or 1:200 HA (Cell Signaling, 3724) antibodies by rocking overnight at 4°C. Membranes were washed five times with 1×TBST on orbital rocker for 15 min at room temperature. Membranes probed with the anti-HA antibody were then probed with 1:10,000 IRDye 800CW donkey anti-rabbit (Licor, 926-32213) in 2.5-5% milk in 1×TBST for 1 h and then washed multiple times with 1×TBST on orbital rocker for 5 min before imaging. Membranes probed with HRP-conjugated antibodies were developed using Clarity Western ECL Substrate (BioRad, 170-5061). Membranes were imaged using BioRad ChemiDoc MP Imaging System.

Embryo injections were performed as described previously (Westerfield, 2000) with the following modifications. Plasmids containing either npas4l, arnt1 or arnt2 were linearized by restriction enzyme digestion. mRNA was transcribed using the mMessage mMachine kit (Invitrogen, AM1340) and purified with the RNA clean & concentration kit (Zymo Research, R1019). Embryos were microinjected at the one-cell stage with 1 nl total volume containing either npas4l mRNA (50 pg mRNA) or a mixture of arnt1 and arnt2 mRNA (50 pg each for 100 pg total mRNA). All solutions were injected into the yolk.

At 1-day post fertilization (dpf), embryos were exposed to 10 ng/ml 2,3,7,8-tetrachlorodibenzodioxin (TCDD, AccuStandard, D404N) or vehicle (0.1% DMSO). All treatments were performed on a single clutch of at least 20 embryos. Embryos treated with vehicle were from the same clutches as embryos treated with TCDD. Chemical exposures were performed on at least two clutches per genotype. Embryos were raised at 28.5°C on a 14 h light, 10 h dark cycle until 3 dpf, when they were imaged and then genotyped.

All statistical analyses were performed using Prism version 9.4.1 (GraphPad Software).

For the Chi-squared analysis of embryo and adult genotype ratios, we eliminated all 0 values from the analysis.

For whole-mount in situ hybridization images of arnt1/2 mutants, we used a binomial test to compare the observed distribution of embryos with the reduced expression phenotype to the calculated expected number of arnt1/2 mutants. Additionally, we performed a Spearman rank correlation analysis on genotyped embryos to determine whether there was a correlation between the number of mutant alleles and the observed phenotype. Each genotype was assigned a number 0-4: with 0 indicating no mutant alleles (wild-type siblings), 1 indicating one mutant allele (arnt1 heterozygous and arnt2 heterozygous siblings), 2 indicating two mutant alleles (arnt1 homozygous, arnt2 homozygous or arnt1/2 heterozygous siblings), 3 indicating three mutant alleles (arnt1 homozygous; arnt2 heterozygous and arnt1 heterozygous; arnt2 homozygous siblings) and 4 indicating four mutant alleles (arnt1/2 mutants). The observed phenotypes were assigned a number value: normal expression was coded as a 1 and reduced expression was coded as 0.

For WIHCR image analysis, the variance was not assumed for npas4l and arnt2 samples. arnt1 samples did not have a normal distribution, thus a non-parametric Mann–Whitney test was performed. We were able to identify an outlier in the arnt1 integrated intensity data using Grubb's test (α=0.05); however, when this outlier was removed from analysis there was no change in significance or Mann–Whitney U value.

For TCDD treated embryos, unpaired Welch's t-tests were performed to compare the average number of embryos with cardiotoxicity with the average number of embryos with a normal phenotype from a given genotype.

We thank Lauren Pandolfo and the staff of the Baylor College of Medicine aquatics facility for taking care of our zebrafish colony, Rosa Uribe and Adam Howard from Rice University for help with WIHCR, and Paula Pimienta-Ramirez and Yunping Lei in the CPEH DNA sequencing core at BCM for help with sequencing embryos following in situ hybridization.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.200500.reviewer-comments.pdf.