A novel conserved enhancer at zebrafish zic3 and zic6 loci drives neural expression

Abstract Background Identifying enhancers and deciphering their putative roles represent a major step to better understand the mechanism of metazoan gene regulation, development, and the role of regulatory elements in disease. Comparative genomics and transgenic assays have been used with some success to identify critical regions that are involved in regulating the spatiotemporal expression of genes during embryogenesis. Results We identified two novel tetrapod‐teleost conserved noncoding elements within the vicinity of the zic3 and zic6 loci in the zebrafish genome and demonstrated their ability to drive tissue‐specific expression in a transgenic zebrafish assay. The syntenic analysis and robust green fluorescent expression in the developing habenula in the stable transgenic line were correlated with known sites of endogenous zic3 and zic6 expression. Conclusion This transgenic line that expresses green fluorescent protein in the habenula is a valuable resource for studying a specific population of cells in the zebrafish central nervous system. Our observations indicate that a genomic sequence that is conserved between humans and zebrafish acts as an enhancer that likely controls zic3 and zic6 expression.


| INTRODUCTION
Gene regulation is a highly complex process that requires the presence of intact coding sequences and the proper functioning of cis-regulatory elements. The precise spatiotemporal activity of a gene involves the coordination of RNA polymerase, multiple transcription factors, and various cis-regulatory elements, including enhancers, promoters, silencers, insulators, and locus control regions. 1,2 cis-regulatory elements are often located upstream, downstream, within an intron, or hundreds of kilobases (kb) away from the genes they control. [2][3][4] A gene might thus be surrounded by multiple cis-regulatory elements, each of which combinatorially contributes to its overall spatial and temporal expression pattern. cis-regulatory elements and gene expression also play a significant role in the evolution of vertebrate complexity and diversity during development. 5,6 The identification and functional characterization of cis-regulatory elements are notable challenges in the postgenomic era and are needed to gain a better understanding of the language, syntax, and grammar that are encoded in regulatory DNA. Conserved noncoding elements (CNEs) from human to fish (evolutionary distance of 450 million years [Myr]) remain an important landmark to annotate the regulatory landscape around developmental regulators. 4,[7][8][9] These noncoding genomic intervals can be highlighted by comparing the genomes of distantly related species. The CNEs that have been identified harbor multiple transcription factor binding sites (TFBSs). Mutations of these critical regions can alter gene expression and lead to anomalies and developmental defects. 1,2,7,10 Therefore, they are rarely lost during the evolution of distantly related species. 11,12 The vertebrate Zic family evolved by gene duplication events in higher vertebrates (zic1, zic2, zic3, zic4, zic5), with two additional zic family members in the zebrafish genome (zic2b and zic6 13,14 ). Mutations of zic genes lead to developmental anomalies of the cardiac, neural tube, skeleton, and muscles. 15,16 Zic genes are expressed in an overlapping fashion during early embryogenesis, particularly in the dorsal part of the neural tube and somites. 17 However, differences in spatiotemporal expression patterns and different combinations of zic genes have been shown to be expressed in the limbs, eyes, and tail buds. 18 The role of the Zic gene family has been extensively studied over the past two decades, but the regulatory elements that are responsible for the complex expression of Zic genes have not yet been elucidated in detail. The present study sought to identify novel functional enhancers around loci of the zic3 and zic6 genes.
Zebrafish zic3 is paired with zic6 in an opposite orientation in one of three extant loci that resulted from chromosomal duplication early in the evolution of vertebrates. Previously, two CNEs (E1 and E2) were reported to be nearby zic3 and zic6 loci, each of which was linked to both genes. 19 The zic3-associated enhancer was previously shown to drive expression in the neural tube and neural plate border. 19 The present study employed comparative genomics and orthology mapping approaches to highlight putative enhancer regions around zic3/zic6 and characterize their ability to drive the expression patterns of an enhanced green fluorescent protein (EGFP) reporter. Using UCSC Zv9/danRer7 genome assembly, we identified three noncoding genomic intervals that are close to zic3/zic6 and have been under strong evolutionary pressure and conserved during the course of 450 Myr from human to fish. We identified and characterized a novel CNE that encompasses 453 bp and is a neural enhancer that regulates gene expression in the habenula during early embryogenesis that correlates with the expression pattern of zic3 and zic6. Two other enhancers that we identified have partially overlapping sequences with previously reported zic3/zic6-associated neural enhancers. 19 Our results show that the novel intragenic CNE around the zic3 and zic6 loci acts as a neural-specific enhancer and possibly drives the expression of zic3 and zic6.
2 | RESULTS 2.1 | Tetrapod-teleost CNEs maintain a conserved association with the zic3 gene Multispecies alignment of the DNA sequence of the zebrafish zic3/zic6 locus with orthologous genomic sequences from human, mouse, medaka, and fugu revealed three CNEs ( Figure 1A) that maintain at least 80% sequence identity over a 60 bp window across all species (Table 1). For simplicity, these CNEs were named according to the zCNEv1 and UCSC genome browsers (CNE12001, CNE12030, and CNE12032). In the zebrafish genome, two of the CNEs (ie, CNE12030 [novel] and CNE12032) are located 93 kb upstream of zic3, and the third CNE (CNE12001) is located 28 kb upstream of zic3 ( Figure 1A, Table 1). To confirm whether these CNEs are truly associated with zic3 and zic6 in silico, a gene synteny was drawn for zic3/zic6 and its flanking genes between tetrapod and teleost orthologous loci ( Figure 1A). This comparison revealed that these CNEs maintained a syntenic relationship only with the zic3 gene because the zic6 gene is retained only in teleosts. These CNEs harbored motifs for important transcription factors, such as Tcf3, Oct1, Nr2f1, Pax, Pitx2, Tbx5, and Sox ( Figure 1B, Table 2).

| Green fluorescent protein expression is driven by full-length CNE12030 and CNE12032 in the transient assay
To investigate the potential role of these CNEs in detail, we initially performed transient transgenesis assays based on a Tol2 vector with a c-fos minimal promoter 20 and tested two different lengths of each CNE ( Figure 1C): (a) an extended or longer sequence and (b) a core conserved sequence between human and fish ( Table 1). The injected embryos were screened at different developmental stages (26-33 hours postfertilization [hpf], 48-56 hpf, and 72 hpf) for GFP signals. Transient reporter gene activity was induced by the extended and core conserved regions of CNE12030 and CNE12032. However, both the extended and short lengths of CNE12001 failed to induce reproducible EGFP expression in the transient assays. We also injected a Tol2-c-fos-EGFP backbone without a CNE in a transient assay as a negative control ( Figure 1D). At both the 48 and 72 hpf stages, 69% (44/63) of the embryos that were injected with the Tol2-c-fos-EGFP backbone without a CNE showed background expression in muscle cells as previously reported. 20 F I G U R E 1 A, Comparative syntenic analysis of human, mouse, medaka, fugu, and zebrafish orthologous loci, depicting the conserved presence of two genes, fgf13a and zic3, in the nearest vicinity of three CNEs (red, green, and purple vertical lines). Genes are color-coded. The arrow direction depicts the direction of gene transcription. The vertical line depicts the position of the CNE. The horizontal blue line depicts the scale. B, Human and zebrafish alignment of CNE12030 and CNE12032, highlighting the highly similar sequences between them and graphical representation of the transcription factor binding motif that was identified by MEME. Each of the motifs was further screened against the TRANSFAC library to mark binding sites for multiple developmentally important transcription factors. C, Schematic representation of the reporter gene cassette that was used for the in vivo characterization of CNEs. Brown vertical boxes represent different lengths of noncoding DNA that were selected for the transgenic assays. D, Embryos injected with Tol2-c-fos-EGFP backbone without a CNE (left side embryos, negative) and with a CNE12032-containing reporter construct (right side embryos, positive) at 48 hpf in the transient transgenic assay. Both pictures are lateral views with the dorsal side toward the right. Scale bar = 500 μm. Chr, chromosome; CNE, conserved noncoding element; EGFP, enhanced green fluorescent protein; MEME, multiple EM for motif elicitation Based on our transient Tol2 transgenesis results, we subcategorized GFP expression in several developmental domains as indicated in Figure 2A. The prominent domains that were observed in the transient assays for both CNEs were in the central nervous system (CNS; forebrain, midbrain, hindbrain), eyes, heart, and notochord.

| CNE12030 drives GFP expression in the habenula
To generate stable transgenic lines, approximately 50 fish were injected with each CNE construct and raised to adulthood. After 3 months, founder fish were outcrossed with wild-type fish for germline transmission screening in F1 ( Figure 2B). Among the F1 offspring of founders that were injected with either CNE12030 or CN12032 core conserved regions, no reporter gene expression was observed. This suggested that no germline transmission occurred to the F1 generation. Therefore, we were unable to validate the activity of these shorter versions of the CNE. Germline transmission was identified in two F0 females (4%) that carried the  F I G U R E 2 Sites of GFP signals that were recorded in zebrafish embryos that were transiently transfected with a full-length construct. The reporter genes that were induced by individual CNEs (indicated by the name) are depicted in schematic representations of day 2 (48 hpf) or day 3 (72 hpf) embryos. Categories of a cell type that was positive for a given element are color-coded, with a GFP-expressing cell. The bar graphs (yaxis) show the percentage of GFP-expressing embryos that presented expression in each tissue category for a given element. The percentage of GFPexpressing embryos per CNE is indicated beneath each schematic (EE = %). The bar graphs use the same color code as the schematics for each cell type. B, The injected fish were raised to adulthood and crossed with wild-type fish to screen the germ-line transmission of enhancers. Each GFPpositive fish was further studied in situ to compliment the GFP signals and anti-GFP to confirm GFP expression. CNE, conserved noncoding element; EE, GFP-expressing embryos; GFP, green fluorescent protein; hpf, hours postfertilization longer form of CNE12030. The expression domain for both founders was the habenula, with 15 to 20 GFP-positive embryos per 100 F1 embryos. Comparisons of GFP signals and intensity revealed that both founders had negligible differences in expression patterns. However, one had higher levels of GFP expression in the habenula ( Figure 3A-G). GFP expression in the habenula in stable transgenics persisted from early embryogenesis (ie, day 1-3). To characterize other weaker signals in more detail, RNA whole-mount in situ hybridization was performed using an antisense probe against GFP ( Figure 3F,H-K). Whole-mount anti-GFP immunohistochemistry was also performed to search for domains of GFP expression other than the habenula, but no other additional domains were identified ( Figure 3G). GFP transcripts were detected in the same region where GFP signals were present at 30 hpf ( Figure 3H,I) and 48 hpf ( Figure 3J,K). To confirm that the structure that expressed GFP was indeed the habenula, RNA whole-mount in situ hybridization was performed using an antisense probe against a known habenula marker, pou4f1 21 ; Figure 3L,M). Whole-mount in situ hybridization for pou4f1 revealed expression in the same territory of the brain as that of GFP, thus confirming that GFP-expressing cells were in the habenula ( Figure 3H-M). Moreover, RNA in situ hybridization using an antisense probe against zic3 ( Figure 3N,O) showed that CNE12030 transgene expression coincided with known sites of zic3 expression. 22,23 Therefore, the extended version of CNE12030 was shown to drive expression in the habenula, which coincided with the zic3 expression domain.
2.4 | CNE12032 drives expression predominantly in the optic tectum, olfactory pit, spinal cord, fin, and heart CNE12032 that is located 72 kb downstream of fgf13a and 95 kb downstream of zic3 is conserved from human to fish. The 466-bp genomic segment that spanned the humanfish conserved core sequence (101 bp) drove GFP expression in founder fish, although we identified only one male individual with germline transmission (ie, 2% of the total number of fish that were injected). However, GFP expression was robust and widespread in several brain regions compared with the other CNE. This expression pattern was similar to previously reported domains of this CNE in the neural tube, 19 with some additional domains that have not been reported previously, including olfactory neurons, the eyes, the pectoral fin, and the heart (Figure 4, Video S1). To confirm GFP expression in olfactory neurons, cells that expressed GFP in olfactory receptor neurons were stained with the lipophilic fluorescent membrane tracer dye DiI and simultaneously imaged for GFP and dye fluorescence using a confocal microscope. We observed the colocalization of GFP and DiI fluorescence in olfactory neurons ( Figure 4B). GFP expression in the optic tectum, eyes, pectoral fin, olfactory pit, and heart persisted from 24 to 72 hpf. These results suggest that this enhancer may be responsible for driving gene expression not only in the neural tube but also in the pectoral fin, the heart, and sensory organs. This expression pattern that was driven by CNE12032 corresponded to distinct domains of zic3 expression compared to that driven by CNE12030.
Overall, this in vivo analysis of stable zebrafish transgenics identified two different enhancers in the vicinity of zic3/zic6 loci in zebrafish with different tissue specificity. These two CNEs may thus drive expression in a subset of tissues that express Zic3.

| DISCUSSION
The Zic gene family contributes to the patterning and growth of many organs, including the CNS, the heart, retinal ganglion cells, somites, and limbs, during vertebrate embryogenesis. Among the Zic family, Zic3 plays a key role in development and disease. Studies have shown that the loss of Zic3 function leads to congenital developmental defects, including heterotaxy and multiple types of congenital heart defects. 24,25 Comparative genomics-based approaches have emerged as reliable tools for predicting noncoding genomic features that harbor transcriptional regulatory elements, even in the absence of knowledge about the detailed characteristics of individual cis-regulatory elements. 3,4,26-28 Two critical steps have been suggested to be necessary to ascertain a functional CNE: (a) carefully selecting the species that are compared and (b) choosing the appropriate length of genomic intervals. 29 Depending on the alignment tools, different levels of stringency are applied to identify a number of putative enhancers, without an a priori biological rationale. 30 Conventional comparative criteria, such as 70% sequence identity over at least 100 bp between human and mouse, have disadvantages because these two species share a relatively short evolutionary distance (65 Myr). The most obvious solution to identify the functional module in the genome is to compare more distantly related species, such as humans and fish (450 Myr), using suitable identification criteria. The functional validation of CNEs should include additional steps beyond steps (a) and (b) above, such as (c) analyzing the ability of the CNE to drive expression in transient transgenics and (d) identifying the ability of the CNE to drive expression in stable transgenics. Furthermore, random integration of the reporter construct (positional effect) and the lack of a positive control for transgenesis pose challenges to defining the efficacy of integration events in F0 embryos and in nonexpressing F1 embryos. 31 In this case, positive stable expression in F1, together with mapping In the present study, we initially selected 32 CNEs that were predicted by the Bejerano group and available with the custom track Bej zCNEv1 in the UCSC genome browser in the 158 kb intragenic region between zic3 and fgf13a. 32 Using the relatively stringent criteria of 80% sequence identity within the 60-bp window, we ascertained three of 32 elements that are highly conserved from human to fish. The gene environment of the loci that harbored the subset of CNEs was calculated further to associate the selected CNEs with their target gene body using an orthology mapping approach. Zebrafish zic3 and zic6 loci have been compared with orthologous loci from genomic sequences of Homo sapiens (mammal), Mus musculus (rodent), Oryzias latipes (teleost fish), and Takifugu rubripes (teleost fish; Figure 1). In teleosts, CNE12001, CNE12030, and CNE12032 maintained a conserved physical linkage with fgf13a, zic6, and zic3. Closer inspection of our syntenic analysis and previously reported studies showed that Zic6 was deleted during mammalian evolution, whereas Zic3 maintained its association with these CNEs. 33 Three lines of evidence corroborate the CNEs that were identified as zic3/zic6 enhancers in the present study. First, the in vivo reporter expression pattern that was driven by CNE12030 and CNE12032 coincided with the pattern that was previously reported for zic3 in the CNS. 13,33 Second, the presence of conserved binding sites for recognized developmental regulators based on rigorous criteria in both zebrafishhuman tracks suggested that these CNEs may be regulated by transcription factors ( Table 2) that are coregulated with zic3 during early embryonic development of the CNS, such as members of the Wnt and Nodal pathways. 34,35 Third, a survey of expression patterns of all genes within 1.5 Mb up-and downstream of the zic3/zic6 cluster did not identify other candidates with clear expression in the habenula (ZFIN database), strongly suggesting that zic3 and zic6 are the only obvious targets within the vicinity of these enhancers.
The habenula is part of a highly conserved limbic system conduction pathway in vertebrates and has been shown to GFP, green fluorescent protein; l, lens; op, olfactory pit; ot, optic tectum; pf, pectoral fin; sc, spinal cord play a crucial role in a diverse set of behavioral systems, including brain stimulation, endocrine function, mating, ingestion, fear responses, and olfaction. [36][37][38] The habenula in zebrafish is subdivided into the dorsal and ventral habenula, which are homologous to lateral and medial parts in mammals. 39 Within the CNS, zic3 is known to be expressed in a restricted manner during neurulation. 17 In the present study, CNE12030 was shown to drive GFP expression in the habenula ( Figure 3A-C,H-K) where zic3 is expressed ( Figure 3N,O). Additionally, the habenula-specific marker pou4f1 validated that the CNE12030-driven expression domain represents the right and left habenula. Altogether, for zic3 and pou4f1, the novel CNE12030 might play important roles in the development of the habenula. This possibility could be confirmed by deleting this CNE or mutating binding sites for multiple transcription factors. The zebrafish is an ideal model for elucidating neurocircuitry mechanisms that underlie animal behavior. 40 Further investigations of this transgenic line may open new avenues for studying the vertebrate brain limbic system. Moreover, this novel CNE will further our understanding of the genetic basis of human birth defects that cannot be attributed to a mutation in the coding sequence of Zic3.
Garnett et al 19 previously showed that the enhancer E2 that is evolutionarily conserved between humans and fish autonomously controlled individual aspects of zic3 expression in the developing neural tube in zebrafish embryos. CNE12032 is located 74 kb downstream of fgf13a and 93 kb upstream of zic3, which overlaps by an additional 112 bp with E2. Interestingly, in the stable transgenic line, the overall expression pattern that was driven by CNE12032 was similar to E2, particularly in the CNS, with some additional domains that were not previously reported. These novel domains of prominent GFP signals that were observed at 48 and 72 hpf included cardiac cells, the pectoral fin, and olfactory sensory neurons. This largely fits with the known expression patterns of the Zic family in mice, chick embryos, and zebrafish, which are expressed in a partially overlapping manner within the neural tube, somites, and other ectoderm-and mesodermderived structures in the future head and trunk. 14,17,41 The transgenic line ET33-mi59B carries an enhancer trap that inserted 155 kb upstream of zic3 and 140 kb downstream of zic6 and exhibits GFP expression in the cardiac conduction system of the heart. 42 However, none of our tested CNEs that were located 93 kb upstream of zic3 drove GFP expression in cardiac conduction system cells at 48 to 72 hpf. The enhancer that is responsible for cardiac conduction system-specific expression may not be conserved from humans to fish as many enhancers were previously reported to be nonconserved. 43 Alternatively, additional flanking sequences that were omitted in the present analysis may contribute to a more comprehensive enhancer element that may control specific expression in an additional subset of tissues. 29 This is a particularly interesting possibility because of the presence of the fgf13a gene, which has been implicated in heart development, 42 adjacent to the zic3 locus (167 kb upstream).

| CONCLUSION
The present study used a comparative genomics approach to highlight putative enhancers around the zic3 and zic6 loci. Like other members of Zic family genes, zic3 and zic6 might share common regulatory elements. The novel enhancer CNE12030 drove GFP expression in the habenula, which is a key part of the CNS (diencephalon) and is highly conserved from teleost fish to mammals. The present findings suggest that the newly identified CNE12030 may be involved in habenula development and spatially control zic3 expression. This transgenic line can be a useful tool for studying zebrafish habenula development and function.

| EXPERIMENTAL PROCEDURES
5.1 | Selection of putative cis-regulatory elements around zic3-zic6 for functional analysis CNEs around zic3 and zic6 loci were selected for functional studies from annotated regions 32 that are available at the UCSC Zv9/danRer7 genome browser (http://genome.ucsc. edu/) with a custom track of Bej zCNEv1. The subset of predicted CNEs was further subjected to BLASTN with human, mouse, medaka, and fugu genomes using the Ensembl genome browser, applying stringent conservation criteria of at least a 60 bp length and 80% sequence identity. CNEs that were only conserved in teleost fish and did not meet our selection criteria were discarded. Three CNEs upstream of zic3 and downstream of zic6 (CNE12001, CNE12030, and CNE12032) were selected for further functional studies.

| In silico analysis of CNEs from mammals to teleost fish to identify conserved putative TFBSs
To identify putative conserved TFBSs for each CNE, orthologous sequences from mammals to teleost fish were retrieved from the Ensembl genome browser using a BLASTN-based similarity search as previously described. 3 Each of the CNEs with orthologous sequences was evaluated using the Multiple Em for Motif Elicitation (MEME) algorithm. 44 According to the expected length of DNA sequences that are recognized by transcription factors, the criteria for the minimum length was set to 6 to 12 bp. 45,46 The identified motifs of each CNE were further characterized using the STAMP online tool 47 to determine known transcription factors against the TRANSFAC (v. 11.3) database. 48 The gene expression patterns of each of the specified transcription factors were checked in the ZFIN database (https://zfin.org) and are listed in Table 2.

| Zebrafish transgenic assays and screening
Wild-type and transgenic zebrafish were maintained in the zebrafish facility of the International Institute of Molecular and Cell Biology in Warsaw (license no. PL14656251) according to standard procedures and ethical practices. Embryos were grown in embryo medium at 28 C and staged according to standard morphological criteria. 49 The transgenic zebrafish experiments were performed according to standard protocols that were established by the Polish Laboratory Animal Science Association. Putative enhancer regions were amplified by polymerase chain reaction (PCR) from wild-type zebrafish genomic DNA using primers with the BamHI site ( Table 1). The PCR products were then purified using the MN DNA purification kit (Macherey-Nagel) according to the manufacturer's instructions and subsequently cloned into the BamHI site of a reporter system that consisted of EGFP under the control of the mouse c-fos minimal promoter and Tol2 sites. Each of the plasmid constructs that contained the CNEs was then sequenced to control for the presence of any point mutations that were generated during PCR amplification. The reporter constructs that contained the Tol2-c-fos-EGFP promoter-reporter cassette 50 were combined with 0.5% phenol red (Sigma), which was used as a tracer dye as described previously. 3,51 Injection mix (1 nL) was injected into 1 to 2 cell-stage zebrafish embryos using the PV820 pressure injection system (WPI) under an SZX16 stereo microscope (Olympus). The Tol2-cfos-EGFP plasmid without CNEs was also injected as a negative control. The injected embryos were raised at 28.5 C in 1X embryo medium. The zebrafish embryos were dechorionated manually using fine forceps at 2 days postfertilization (dpf) and anesthetized using tricaine. 52 The transient transgenic embryos were screened for GFP signals under an M165 FC inverted fluorescent microscope (Leica). Imaging was performed using the Leica application suite (v. 4.0) and Zeiss Axio Imager 2 using Axiocam MRm. The schematics for the location and tissue-specific expression of each CNE were generated using Adobe Photoshop software. Approximately 50 embryos with GFP signals were raised to adulthood to screen germ line transmission.

| Imaging using light-sheet fluorescence and confocal microscopy
Embryos were anesthetized at the desired developmental stages and embedded in a 1 mm inner-diameter glass capillary that was filled with 1.5% low-melting agarose (LMA) in embryo medium (0.03% Instant Ocean salt into doubledistilled water). After the complete polymerization of agarose, the capillary was inserted in the sample holder and placed in the microscope chamber that was filled with embryo medium (0.02% tricaine), and the embryo was pushed out of the capillary for imaging. The temperature of the sample chamber was maintained at 28 C during imaging. Image acquisition was performed using a Zeiss Z.1 light-sheet microscope with a W Plan-Apochromat 20×/1.0 objective. Z-stacks (3.83 μm thickness, 60 ms exposure time) were saved in LSM format and then processed using ZEN software (Zeiss). For each Z-stack, maximum-intensity projections were generated. For confocal microscopy, embryos were mounted in 1.5% LMA in glassbottom dishes after the complete polymerization of LMA.

| Whole-mount zebrafish larvae immunofluorescence
Zebrafish embryos were fixed with 4% paraformaldehyde at 4 C overnight. Embryos were permeabilized in methanol, rehydrated in a series of MeOH/PBST (Phosphate Buffer Saline with Tween 20) concentrations (75%, 50%, and 25%), and washed with 1X PBST. Treatment with Proteinase K (10 μg/mL) for 30 minutes was followed by postfixation in 4% paraformaldehyde for 30 minutes and several washes in PBST. Embryos were blocked with blocking solution (10% sheep serum and 1% dimethylsulfoxide [DMSO] in PBST) for 4 hours to reduce nonspecific antibody binding and then incubated overnight at 4 C with rabbit polyclonal anti-GFP antibody (Gentex) at a dilution of 1:500. After incubation, several washes were performed, and the primary antibody was detected using Alexa Fluor 488 Donkey anti-Rabbit IgG secondary antibody (Invitrogen) at a dilution of 1:200. Embryos were stored in PBST at 4 C. Images were taken using a Zeiss LSM 800 microscope after mounting in 1.5% LMA.

| Whole-mount in situ hybridization
Whole-mount in situ hybridization of zebrafish embryos at 30 hpf was performed using a digoxigenin-labeled antisense RNA probe. The pou4f1 cLone was provided by Steve Wilson UCL, UK. The probe that was specific for zic3 was synthesized from 1990 bp fragment cDNA using the DIG-RNA labeling kit (Roche) according to the manufacturer's instructions. Whole-mount in situ hybridization was performed as described previously 53 with minor modifications. After rehydration, permeabilization, and blocking, the RNA probe was hybridized to target mRNA overnight at 68 C. Antidigoxigenin-AP Fab fragments (Roche) were used for target recognition. The color reaction was performed using NBT/ BCIP Solution (Roche).

| DiI staining to label olfactory pits
DiI stain (1 μL) was dissolved in 100% DMSO (final concentration of 2.5 mg per 1 mL of DMSO) and used for 1 mL of embryo media. The embryos were then incubated for 15 to 30 minutes at 28 C. To remove residual DiI stain, the embryos were washed three times with embryo medium. Olfactory pits were observed with red fluorescence at an excitation wavelength of 549 nm and emission wavelength of 565 nm.

ACKNOWLEDGMENTS
We are grateful to zebrafish core facility IIMCB Warsaw for excellent fish care. We also thank Smijin Soman and Justyna Jędrychowska for useful discussions during this project. We thank Steve Wilson UCL, UK for the kind gift of pou4f1 clone. The authors would like to thank Michael Arends for proofreading the manuscript. R.M. has received funding from the European Union's Horizon 2020 research and innovation program under the Marie Skłodowska-Curie grant agreement no. 665778, the fellowship registration number: "2015/19/ P/NZ3/03613" supported by (National Science Centre [NCN], Poland). The Opus Grant (2016/21/B/NZ3/00354) from the National Center of Science, Poland supports V.K. The project no. POIR.04.04.00-00-1AF0/16-00/ is carried out within the First TEAM programme of the Foundation for Polish Science cofinanced by the European Union under the European Regional Development Fund.