Atoh1 is required for the formation of lateral line electroreceptors and 1 hair cells, whereas Foxg1 represses an electrosensory fate 2

Abstract


Introduction
Mechanosensory hair cells in different compartments of the inner ear transduce fluid movements for hearing and balance (Moser et al., 2020;Caprara and Peng, 2022;Mukhopadhyay and Pangrsic, 2022).In fishes and aquatic-stage amphibians, hair cells are also found in lateral line neuromasts in the skin, which are stimulated by local water movement (Mogdans, 2019;Webb, 2021).Supporting cells in neuromasts differentiate into hair cells for homeostasis and repair (Kniss et al., 2016;Lush et al., 2019) and neuromasts are easily accessible in lines over the head and trunk, making the zebrafish lateral line an excellent model for hair cell development and regeneration (Kniss et al., 2016;Nicolson, 2017;Pickett and Raible, 2019).The lateral line system of zebrafish (a cyprinid teleost) is purely mechanosensory, as is the lateral line system of the other main anamniote lab model, the frog Xenopus.However, in many other vertebrates, the lateral line system also has an electrosensory division (Bullock et al., 1983;Baker et al., 2013;Crampton, 2019).In electroreceptive non-teleost jawed vertebrates, some or all of the neuromast lines on the head are flanked by fields of ampullary organs containing electroreceptors, which respond to weak cathodal stimuli such as the electric fields surrounding other animals in water (Bodznick and Montgomery, 2005;Crampton, 2019;Leitch and Julius, 2019;Chagnaud et al., 2021).Electroreception is mediated by voltage-gated calcium channels in the apical membrane (Bodznick and Montgomery, 2005;Leitch and Julius, 2019).The voltage sensor was recently identified in cartilaginous fishes as the L-type voltage-gated calcium channel Cav1.3 (Bellono et al., 2017;Bellono et al., 2018), whose pore-forming alpha subunit is encoded by Cacna1d.Cav1.3 is also required for synaptic transmission at hair-cell ribbon synapses (Moser et al., 2020;Mukhopadhyay and Pangrsic, 2022).
Fate-mapping experiments have shown that neuromasts, ampullary organs (where present) and their afferent neurons all develop from a series of pre-otic and post-otic lateral line placodes on the embryonic head (Northcutt, 1997;Piotrowski and Baker, 2014;Baker, 2019).In electroreceptive jawed vertebrates, lateral line placodes elongate to form sensory ridges that eventually fragment: neuromasts differentiate first, in a line along the centre of each ridge, and ampullary organs (if present) form later, in fields on the flanks of the ridge (Northcutt, 1997;Piotrowski and Baker, 2014;Baker, 2019).The lateral line primordia of electroreceptive vertebrates therefore provide a fascinating model for studying the formation of different sensory cell types and organs.What molecular mechanisms control the formation within the same primordium of mechanosensory neuromasts containing hair cells, versus electrosensory ampullary organs containing electroreceptors?
To gain molecular insight into electroreceptor development, we originally took a candidate-gene approach, based on genes known to be important for neuromast and/or hair cell development.This enabled us to identify a variety of genes expressed in developing ampullary organs as well as neuromasts, in embryos from the three major jawed-vertebrate groups, i.e., cartilaginous fishes (lesser-spotted catshark, Scyliorhinus canicula, and little skate, Leucoraja erinacea; O'Neill et al., 2007;Gillis et al., 2012); lobe-finned bony fishes/tetrapods (a urodele amphibian, the axolotl, Ambystoma mexicanum; Modrell and Baker, 2012); and ray-finned bony fishes (a chondrostean, the Mississippi paddlefish, Polyodon spathula; Modrell et al., 2011a;Modrell et al., 2011b;Modrell et al., 2017b).We also took an unbiased discovery approach using differential bulk RNA-seq in late-larval paddlefish, which yielded a dataset of almost 500 genes that were putatively enriched in lateral line organs (Modrell et al., 2017a).Validation by in situ hybridization of a subset of candidates from this dataset suggested that conserved molecular mechanisms were involved in hair cell and electroreceptor development, and that hair cells and electroreceptors were closely related physiologically (Modrell et al., 2017a).For example, developing ampullary organs express the key 'hair cell' transcription factor genes Atoh1 and Pou4f3 (see Roccio et al., 2020;Iyer and Groves, 2021), and genes essential for the function of hair cell ribbon synapses, including the voltage-gated calcium channel gene Cacna1d, encoding Cav1.3 (Modrell et al., 2017a).We also identified a handful of genes expressed in developing ampullary organs but not neuromasts, including two electroreceptor-specific voltage-gated potassium channel subunit genes (Kcna5 and Kcnab3) and a single transcription factor gene, Neurod4 (Modrell et al., 2017a).
Up to that point, we had reported the shared expression of fifteen transcription factor genes in both ampullary organs and neuromasts, but only one transcription factor gene with restricted expression, namely, electrosensory-restricted Neurod4 (Modrell et al., 2011a;Modrell et al., 2011b;Modrell et al., 2017b).More recently (preprint: Minařík et al., 2023), we used the late-larval paddlefish lateral line organ-enriched dataset (Modrell et al., 2017b), as well as a candidate gene approach, to identify 23 more transcription factor genes expressed within developing lateral line organs in paddlefish and/or in a related, more experimentally tractable chondrostean, the sterlet (Acipenser ruthenus, a small sturgeon; e.g., Chen et al., 2018;Baloch et al., 2019;Stundl et al., 2022).Twelve of these transcription factor genesincluding Gfi1 -were expressed in both ampullary organs and neuromasts (preprint: Minařík et al., 2023).Thus, developing ampullary organs, as well as neuromasts, express the three 'hair cell' transcription factor genes -Atoh1, Pou4f3 and Gfi1 -whose co-expression is sufficient to drive postnatal mouse cochlear supporting cells to adopt a 'hair cell-like' fate, albeit not to form fully mature hair cells (Roccio et al., 2020;Iyer and Groves, 2021;Chen et al., 2021;Iyer et al., 2022).We also identified six novel ampullary organ-restricted transcription factor genes and the first-reported mechanosensory-restricted transcription factor genes (preprint: Minařík et al., 2023).One of the five mechanosensory-restricted transcription factor genes was Foxg1, which was expressed and maintained in the central region of sensory ridges where lines of neuromasts form, although excluded from hair cells (preprint: Minařík et al., 2023).
Here, we used CRISPR/Cas9-mediated mutagenesis in F0-injected sterlet embryos to investigate the function in lateral line organ development of Atoh1, electrosensory-restricted Neurod4 and mechanosensory-restricted Foxg1 (for reference, Supplementary Figure S1 shows the normal expression patterns of these genes).We report that Atoh1 is required for the formation of electroreceptors, as well as hair cells.We did not see any phenotype after targeting ampullary organ-restricted Neurod4, potentially owing to redundancy with other Neurod family members that we found to be expressed in sterlet ampullary organs (and neuromasts).Targeting mechanosensory-restricted Foxg1 resulted in a striking phenotype: the ectopic formation of ampullary organs within neuromast lines, and in some cases the fusion of ampullary organ fields that normally develop either side of a line of neuromasts.This suggests the unexpected but intriguing hypothesis that ampullary organs may be the 'default' fate for lateral line sensory ridges, and that this is repressed by Foxg1, allowing neuromasts to form instead.

CRISPR/Cas9-mediated mutagenesis in F0-injected sterlet embryos
To test gene function during lateral line organ development, we optimised CRISPR/Cas9mediated mutagenesis in F0-injected sterlet embryos, building on established protocols for axolotl (Ambystoma mexicanum; Flowers et al., 2014;Fei et al., 2018), newt (Pleurodeles waltl; Elewa et al., 2017), and sea lamprey (Petromyzon marinus; Square et al., 2015;York et al., 2019;Square et al., 2020), whose eggs are all large (1-2 mm in diameter) and easy to microinject at the 1-2 cell stage.Ovulated sterlet eggs are very large: roughly 2.5 mm in diameter (Lenhardt et al., 2005).(Since this project started, CRISPR/Cas9-mediated mutagenesis in F0-injected sterlet embryos has been reported, including by two of us, R.F. and M.P.; Chen et al., 2018;Baloch et al., 2019;Stundl et al., 2022.)Analysis of microsatellite data had originally suggested that although a whole-genome duplication had occurred in the sterlet lineage, the sterlet was likely to be a functional diploid (Ludwig et al., 2001).Our sgRNAs were designed before the first chromosome-level sterlet genome was published (Du et al., 2020).Analysis of this genome showed that approximately 70% of ohnologs (i.e., gene paralogs originating from the whole-genome duplication) had in fact been retained, suggesting functional tetraploidy (Du et al., 2020).We comment on this in relation to our experiments at the relevant points below.
We targeted the melanin-producing enzyme tyrosinase (Tyr) as a positive control, using eight different single-guide (sg) RNAs.Table 1 shows the sgRNA target sequences (including two that were designed and recently published by Stundl et al., 2022) and the various combinations in which they were injected.Supplementary Figure S2A shows the position of each sgRNA relative to the exon structure of the tyrosinase gene.Our sgRNAs were designed before chromosome-level sterlet genomes were available (Du et al., 2020  Three different combinations of six of the eight Tyr sgRNAs, when injected at the 1-2cell stage (2-4 sgRNAs pre-complexed with Cas9), each generated at least four embryos (hereafter 'crispants') with altered pigmentation phenotypes evaluated at stage 45 (Dettlaff et al., 1993), the onset of the transition to independent feeding.We note that our genotyping results have shown that most crispants analysed, across all genes targeted, have shown some degree of targeted mutagenesis in the trunk/tail, with a range of deletion sizes.Although phenotypes from the initial spawning seasons were almost always highly mosaic, suggesting mutations occurred later in development, following optimization a proportion of embryos showed complete unilateral and occasionally bilateral phenotypes.Such phenotypes suggest that mutation occurred in one cell at the 2-cell stage (unilateral phenotype) or even as early as the 1-cell stage (bilateral phenotype).Some degree of mosaicism can be useful, however, as the unaffected tissue provides an internal control for the normal expression of the gene being examined by ISH.

Targeting Atoh1 resulted in the loss of hair cells and electroreceptors
We targeted Atoh1 for CRISPR/Cas9-mediated mutagenesis by injecting sterlet embryos at the 1-2 cell stage with Cas9 protein pre-complexed with two sgRNAs targeting Atoh1 (Table 1; Supplementary Figure S2A).Our sgRNAs were designed before chromosome-level sterlet genomes were available (Du et al., 2020 and the 2022 reference genome: https://www.ncbi.nlm.nih.gov/datasets/genome/GCF_902713425.1/).Searching the reference genome for Atoh1 showed that both Atoh1 ohnologs have been retained, on chromosomes 1 and 2, with 91.41% nucleotide identity (and 84.02% amino acid identity) in the coding sequence.The copy on chromosome 2 encodes a shorter version of the protein with a four amino acid deletion near the N-terminus (E14_G17del).Our Atoh1 sgRNA 1 (Table 1; Supplementary Figure S3A) has a one-base mismatch to the shorter Atoh1 gene on chromosome 2, in position 3 of the target sequence (PAM-distal), which is unlikely to prevent successful targeting (Wu et al., 2014).Our Atoh1 sgRNA 2 (Table 1; Supplementary Figure S3A) has a two-base mismatch to the longer Atoh1 gene on chromosome 1, in positions 1 and 2 of the target sequence (PAM-distal), which is also unlikely to prevent successful targeting.
Thus, we expect our sgRNAs to target both Atoh1 ohnologs.
The Atoh1 crispants were raised to stage 45 (the onset of independent feeding, around 14 days post-fertilization), together with Tyr-targeted siblings/half-siblings as controls (eggs were fertilized in vitro with a mix of sperm from three different males).ISH for the hair cell and electroreceptor marker Cacna1d (Modrell et al., 2017a) (also see preprint: Minařík et al., 2023), which was recently shown to be a direct Atoh1 target gene in mouse cochlear hair cells (Jen et al., 2022), revealed no obvious lateral line organ phenotype in Tyr control crispants (Figure 1A-D; n=0/17 across four independent batches; Supplementary Table 1).Even in wildtype larvae, the number of ampullary organs in individual fields varies considerably at stage 45, so ampullary organ number was not in itself scored as a phenotype.However,
We performed genotyping and ICE analysis (Conant et al., 2022)

of tails from individual
Atoh1 crispants targeted with this pair of sgRNAs (Supplementary Figure S3 shows examples).Our genotyping primers were designed before chromosome-level sterlet genomes were available; comparison with the reference genome showed two mismatches in the forward primer against the ohnolog on chromosome 2, and the genotyping data were consistent with the primers amplifying the chromosome 1 ohnolog only.Thus, we could not determine whether the chromosome 2 ohnolog was disrupted.However, the genotyping data showed successful disruption of the Atoh1 gene on chromosome 1 (Supplementary Figure S3).
Taken together, these data suggest that Atoh1 lies upstream of Pou4f3 and Gfi1 in ampullary organs as well as neuromasts, and is required for the differentiation of electroreceptors as well as hair cells.

Targeting electrosensory-restricted Neurod4 had no obvious effect on lateral line development
We previously identified Neurod4 in paddlefish as the first-reported ampullary organ-restricted transcription factor in the developing lateral line system (Modrell et al., 2017a).We confirmed that sterlet Neurod4 is similarly expressed by ampullary organs but not neuromasts (Supplementary Figure S1I,J).We targeted Neurod4 in sterlet embryos using nine different sgRNAs (Table 1; Supplementary Figure S4A), injected in eight different combinations of 1-4 different sgRNAs across 10 independent batches of 1-2 cell-stage embryos (Supplementary The lack of phenotype in Neurod4 crispants, despite successful disruption of the Neurod4 gene, suggested either that Neurod4 is not required for electroreceptor differentiation, despite its restriction to ampullary organs in both paddlefish and sterlet (Modrell et al., 2017a;this paper), or that it acts redundantly with other transcription factors.In paddlefish, Neurod1 expression was restricted to developing lateral line ganglia (Modrell et al., 2011b).We cloned sterlet Neurod1, Neurod2 and Neurod6.(Unlike Neurod4, these three Neurod family members are all direct Atoh1 targets in mouse hair cells; Jen et al., 2022.)These three genes all proved to be expressed in sterlet ampullary organs, as well as neuromasts (Supplementary Figure S5).Thus, it seems likely that the lack of effect of CRISPR/Cas9-mediated targeting of sterlet Neurod4 is due to redundancy with other Neurod family transcription factors co-expressed in ampullary organs.(Our results also show there is at least some variation in Neurod family gene expression within the developing lateral line systems of paddlefish and sterlet.)
We also investigated this phenotype by performing ISH for two ampullary organrestricted transcription factor genes, Mafa (preprint: Minařík et al., 2023) and Neurod4 (Modrell et al., 2017a;   genome showed no mismatches against either of the two ohnologs.Genotyping and ICE analysis (Conant et al., 2022) of tails from individual Foxg1 crispants showed successful disruption of the Foxg1 gene (Supplementary Figure S6 shows examples; the genotyping data were consistent with the primers amplifying both ohnologs).Overall, these data suggest that mechanosensory-restricted Foxg1 acts to repress the formation of ampullary organs and electroreceptors within neuromast lines.

Conserved molecular mechanisms underlie lateral line electroreceptor and hair cell formation
Here, we aimed to test the function in lateral line electroreceptor and/or hair cell formation of three transcription factor genes that we had previously identified as expressed in developing electrosensory ampullary organ and/or mechanosensory lateral line organs in ray-finned chondrostean fishes -paddlefish and sterlet (Butts et al., 2014;Modrell et al., 2017a;preprint: Minařík et al., 2023).The first gene we investigated was Atoh1, which is required for the formation of lateral line hair cells in zebrafish (Millimaki et al., 2007), as well as hair cells in the inner ear (Bermingham et al., 1999;Millimaki et al., 2007).In paddlefish (Butts et al., 2014;Modrell et al., 2017a) and sterlet (preprint: Minařík et al., 2023), Atoh1 is expressed in ampullary organs as well as neuromasts.Targeting Atoh1 for CRISPR/Cas9-mediated mutagenesis in F0 sterlet embryos showed that Atoh1 is required for the formation not only of Cacna1d-expressing neuromast hair cells, as expected from zebrafish (Millimaki et al., 2007), but also of Cacna1d-expressing, Kcnab3-expressing electroreceptors.These experiments also showed that Atoh1 is required for the expression of the 'hair cell' transcription factor genes Gfi1 and Pou4f3 in developing ampullary organs, as well as neuromasts.This is consistent with both of these genes being direct Atoh1 targets in mouse cochlear hair cells (Yu et al., 2021;Jen et al., 2022).
In both inner-ear hair cells and Merkel cells (epidermal mechanoreceptors found in all vertebrates; see e.g., Whitear, 1989;Brown et al., 2023), Atoh1 acts with Pou4f3 in a conserved 'feed-forward circuit', with Atoh1 directly activating Pou4f3 expression, and Pou4f3 then acting as a pioneer factor to open a significant subset of Atoh1 target enhancers (some shared and some divergent between hair cells and Merkel cells), enabling mechanosensory differentiation (Yu et al., 2021).In hair cells, Gfi1 is one of the Pou4f3-dependent Atoh1 targets (Yu et al., 2021).Together with the striking conservation of transcription factor gene expression between developing ampullary organs and neuromasts (Modrell et al., 2011b;Modrell et al., 2011a;Modrell et al., 2017b;Modrell et al., 2017a) (also preprint: Minařík et al., 2023), the phenotypes seen in Atoh1-targeted F0 sterlet crispant embryos suggest that the molecular mechanisms underlying electroreceptor formation are highly conserved with those underlying hair cell formation.Indeed, the requirement of Atoh1 for Pou4f3 and Gfi1 expression in ampullary organs, as well as neuromasts, suggests that the Atoh1-Pou4f3 'feedforward circuit' in mechanosensory cells -i.e., hair cells and epidermal Merkel cells (Yu et al., 2021) -may also be conserved, at least partly, in developing electroreceptors.Taken together, these data support the hypothesis that electroreceptors evolved as a transcriptionally related "sister cell type" to lateral line hair cells (Arendt et al., 2016;Baker and Modrell, 2018;Baker, 2019).

Electrosensory-restricted Neurod4 is likely redundant with other Neurod family members in sterlet
Paddlefish Neurod4 was the first-reported ampullary organ-restricted transcription factor gene (Modrell et al., 2017a), with conserved expression in sterlet (this study).We were unable to detect a lateral line organ phenotype in Neurod4-targeted sterlet crispants.However, we found that Neurod1, Neurod2 and Neurod6 are all expressed in sterlet ampullary organs (as well as neuromasts), suggesting that Neurod4 may act redundantly with one or more of these factors in developing ampullary organs.(In paddlefish, however, Neurod1 expression is restricted to developing lateral line ganglia; Modrell et al., 2011b.)Targeting multiple Neurod genes for CRISPR/Cas9-mediated mutagenesis in the future may shed light on the role played by this transcription factor family in ampullary organ development.

Foxg1 represses electroreceptor formation in the neuromast-forming central zone of lateral line sensory ridges
We also targeted Foxg1, a mechanosensory-restricted transcription factor gene that we recently identified in the developing lateral line system of paddlefish and sterlet (preprint: Minařík et al., 2023).Foxg1 is expressed in the central zones of lateral line sensory ridges where neuromasts form, though excluded from the central domains of neuromasts where hair cells differentiate (preprint: Minařík et al., 2023).Targeting Foxg1 for CRISPR/Cas9-mediated mutagenesis in F0 sterlet embryos led to a striking phenotype: the formation within neuromast lines of ectopic electroreceptors, often in the large clusters normally seen in ampullary organs.
In some cases, ampullary organ fields, which normally flank neuromast lines, effectively 'merged' across missing neuromast lines.This phenotype was revealed by examining expression of the electroreceptor-specific marker Kcnab3, and two ampullary organ-restricted transcription factor genes: Mafa (preprint: Minařík et al., 2023) and Neurod4 (Modrell et al., 2017a).Thus, Foxg1 seems to repress an ampullary organ fate within the central domain of lateral line sensory ridges where neuromasts form.
In the mouse inner ear, Foxg1 is expressed in the prospective cochlea and all sensory patches, in hair cell progenitors and supporting cells (Pauley et al., 2006;Tasdemir-Yilmaz et al., 2021), plus a subset of hair cells (Pauley et al., 2006).Knockout leads to a shorter cochlea with extra rows of hair cells, and to loss or reduction of vestibular end organs (Pauley et al., 2006;Hwang et al., 2009).More hair cells and fewer supporting cells were seen after conditional knockdown of Foxg1 in neonatal cochlear supporting cells, possibly via the transdifferentiation of supporting cells (Zhang et al., 2019;Zhang et al., 2020).This suggests the possibility that Foxg1 may act in the central zone of lateral line sensory ridges to maintain a proliferative progenitor state, as it does in the mouse olfactory epithelium (Kawauchi et al., 2009).
Furthermore, Fox family members can act as pioneer factors as well as transcription factors (Golson and Kaestner, 2016;Lukoseviciute et al., 2018).A pioneer factor role has been proposed for Foxi3 in otic placode development (see Singh and Groves, 2016).In the developing neural crest, Foxd3 acts early as a pioneer factor, opening enhancers and repositioning nucleosomes to prime genes controlling neural crest specification and migration and, concurrently, to repress the premature differentiation of, e.g., melanocytes (Lukoseviciute et al., 2018).Later in neural crest development, Foxd3 represses enhancers associated with mesenchymal, neuronal and melanocyte lineages (Lukoseviciute et al., 2018).In cortical progenitors, Foxg1 suppresses the adoption at later stages of an early-born cell fate, namely, Cajal-Retzius cells (Hanashima et al., 2004).In the developing chicken otic placode, Foxg1 represses markers of other lineages, such as the olfactory and lens placodes, and epidermis (Anwar et al., 2017).Hence, it is possible that Foxg1 acts in the central zone of lateral line sensory ridges in electroreceptive fishes as a pioneer factor for neuromast/hair cell formation, and/or that it represses ampullary organ/electroreceptor formation.Such a role would be consistent with lateral line expression of Foxg1 only in electroreceptive fishes: Foxg1 is not expressed in the developing lateral line of zebrafish or Xenopus (e.g., Dirksen and Jamrich, 1995;Papalopulu and Kintner, 1996;Toresson et al., 1998;Eagleson and Dempewolf, 2002;Duggan et al., 2008;Zhao et al., 2009).
Overall, these data lead us to propose the speculative hypothesis that electrosensory organs may be the 'default' fate within lateral line sensory ridges in electroreceptive vertebrates, and that Foxg1 represses this fate to enable mechanosensory neuromasts and hair cells to form.To test these hypotheses directly, it will be important in the future to identify global changes in gene expression and chromatin accessibility in the absence of Foxg1.

Summary and Perspective
Overall, we have found that the 'hair cell' transcription factor Atoh1 is required for the formation of lateral line electroreceptors as well as hair cells, consistent with a close developmental relationship between these putative 'sister cell' types.Electrosensory-restricted Neurod4 may act redundantly with other Neurod family members expressed in developing ampullary organs.
Mechanosensory-restricted Foxg1 represses the formation of electroreceptors within neuromast lines, suggesting the surprising possibility that electroreceptors are the 'default' fate within lateral line sensory ridges and raising interesting developmental and evolutionary questions for future investigation.

Animals
Fertilized sterlet (Acipenser ruthenus) eggs were obtained from the breeding facility at the Research Institute of Fish Culture and Hydrobiology, Faculty of Fisheries and Protection of Waters, University of South Bohemia in České Budějovice, Vodňany, Czech Republic, and staged according to Dettlaff et al. (1993).For detailed information about sterlet husbandry, in vitro fertilization and the rearing of embryos and yolk-sac larvae, see Stundl et al. (2022).Each fertilization used a mix of sperm from three different males, so each batch was a mix of siblings and half-siblings.

Embryo injection
On the day of injection, 1200 ng gRNA were mixed with 2400 ng Cas9 protein with NLS (PNA Bio) in 5 μl nuclease-free water and incubated for 10 minutes at room temperature to form ribonucleoprotein (RNP) complexes.For gRNA multiplexing, two RNP mixes were combined 1:1 to a final volume of 5 μl, and 0.5 μl of 10% 10,000 MW rhodamine dextran (Invitrogen) added to better visualize the injection mixture and allow selection of properly injected embryos using rhodamine fluorescence.Injection mixtures were kept on ice throughout the injection session.10 μl glass microcapillaries (Drummond Microcaps) were pulled in a capillary needle puller (PC-10, Narishige) set to 58 °C with two light and one heavy weights, in single-stage pulling mode.Fertilized sterlet eggs were manually dechorionated using Dumont #5 forceps.
A 1000 μl pipette tip cut to the same diameter as a dechorionated sterlet egg was used to prepare a series of wells in an agar plate to allow ideal egg positioning during injection using an automatic microinjector (FemtoJet 4x, Eppendorf), set to 100 hPa.Approximately 20 nl of the injection mixture (corresponding to approximately 4.8 ng gRNAs and 9.6 ng Cas9) were injected into fertilized eggs or two-cell stage embryos, targeting the animal pole at a 45° angle.
Injected embryos were moved to a clean Petri dish and, for optimum Cas9 efficiency, kept at room temperature until the end of the injecting session or until at least the 32-cell stage, then moved to a 16°C incubator.No more than 30 eggs were kept per 90 mm Petri dish.Unfertilized eggs and dead embryos were removed at the end of the injection day.Petri dishes were checked regularly for dead embryos and the water was changed at least twice a day before gastrulation was completed, and once daily post-gastrulation.Hatched larvae were kept for approximately 16 days post fertilisation until stage 45, then euthanized by MS-222 overdose and fixed with modified Carnoy's fixative (see above).Fixed larvae were then dehydrated stepwise into 100% ethanol and stored at -20°C.

Gene cloning, in situ hybridization and immunohistochemistry
Total RNA was isolated from sterlet embryos using Trizol (Invitrogen, Thermo Fisher Scientific), following the manufacturer's protocol, and cDNA synthesized using the Superscript III First Strand Synthesis kit (Invitrogen, Thermo Fisher Scientific).We used our sterlet transcriptome assemblies (from pooled yolk-sac larvae at stages 40-45; preprint: Minařík et al., 2023), which are available at DDBJ/EMBL/GenBank under the accessions GKLU00000000 and GKEF01000000, to design primers for Neurod4 (forward: GAGAGAGCCCCAAAGAGACGAG; reverse: CTGCTTGAGCGAGAAGTTGACG). cDNA fragments amplified under standard PCR conditions were cloned into the pDrive cloning vector submitted for sequencing (Genewiz by Azenta Life Sciences).To analyze CRISPR editing efficiency, Sanger trace files were uploaded to the Synthego Inference of CRISPR Edits (ICE) tool (https://ice.synthego.com;Conant et al., 2022).

Imaging and image processing
Larvae were placed in a slit in an agar-coated Petri dish with PBS and imaged using a Leica MZFLIII dissecting microscope equipped with either a MicroPublisher 5.0 RTV camera (QImaging) controlled by QCapture Pro 6.0 or 7.0 software (QImaging), or a MicroPublisher 6 color CCD camera (Teledyne Photometrics) controlled by Ocular software (Teledyne Photometrics).For most larvae, a stack of images was acquired by manually focusing through the sample, then Helicon Focus software (Helicon Soft Limited) was used for focus stacking.
Adobe Photoshop (Adobe Systems Inc.) was used to process images.

Figure 1 .
Figure 1.Atoh1 is required for the differentiation of lateral line hair cells and electroreceptors.Sterlet crispants at stage 45 after in situ hybridization (ISH) for the hair cell and electroreceptor marker Cacna1d, or the electroreceptor-specific marker Kcnab3.Black arrowheads indicate examples of neuromasts; white arrowheads indicate examples of ampullary organs.(A-D) In a control Tyr crispant, Cacna1d expression shows the normal distribution of hair cells in lines of neuromasts, and electroreceptors in fields of ampullary organs flanking the neuromast lines (lateral view: A,B; ventral view: C,D).(E-J) In an Atoh1 crispant (from a different batch to the Tyr crispant shown in A-D), Cacna1d expression is absent on the left side of the head (E,F), except for a few isolated organs in the otic region and on the operculum.Post-ISH Sox2 immunostaining (E 1 ,F 1 ) shows that neuromast supporting cells are still present.A ventral view (G,H) and a lateral view of the right side of the head (I,J; image flipped horizontally for ease of comparison) reveal a unilateral phenotype, with Cacna1d-expressing hair cells and electroreceptors mostly absent from the left side only of the ventral rostrum (asterisk in G,H) and present on the right side of the head (I,J).(K,L) Lateral view of a control Tyr crispant after ISH for Kcnab3, showing the position of electroreceptors in ampullary organs.(M,N) Ventral view of another Tyr crispant showing Kcnab3 expression in ampullary organs.(O-P 1 ) Lateral view of an Atoh1 crispant

Figure 2 .
Figure 2. Atoh1 is required for Pou4f3 and Gfi1 expression in ampullary organs and neuromasts.Sterlet crispants at stage 45 after in situ hybridization for transcription factor genes expressed by developing hair cells.Black arrowheads indicate examples of neuromasts; white arrowheads indicate examples of ampullary organs.(A,B) In a control Tyr crispant, Pou4f3 expression is detected in both neuromasts and ampullary organs.(C-D 1 ) In an Atoh1 crispant, Pou4f3 expression is absent from both neuromasts and ampullary organs, except for a few isolated organs in the postorbital region.Post-ISH Sox2 immunostaining (C 1 ,D 1 ) shows that neuromast supporting cells are still present.Most ampullary organs are not visible as Sox2 expression in ampullary organs is significantly weaker than in neuromasts (Supplementary Figure S1A,B) and often not detectable after post-ISH immunostaining.(E,F) In a control Tyr crispant, Gfi1 expression is detected in both neuromasts and ampullary organs.(G-H 1 ) In an Atoh1 crispant, Gfi1 expression is absent from both neuromasts and ampullary organs, except for a few isolated organs in the supra-and infraorbital region.Post-ISH Sox2 immunostaining (G 1 ,H 1 ) shows that neuromast supporting cells are still present.(Most ampullary organs are not visible due to weaker Sox2 immunostaining.)Abbreviations: b, barbel; di, dorsal infraorbital ampullary organ field; e, eye; n, naris; S, stage; vi, ventral infraorbital ampullary organ field.Scale bars: 200 μm.

Figure 3 .
Figure 3. Foxg1 represses ampullary organ and electroreceptor formation within neuromast lines.Sterlet crispants at stage 45 after in situ hybridization (ISH) for the hair cell and electroreceptor marker Cacna1d, or the electroreceptor-specific marker Kcnab3.Black arrowheads indicate examples of neuromasts; white arrowheads indicate examples of ampullary organs.(A,B) Lateral view of a control Tyr crispant.Cacna1d expression shows the normal distribution of hair cells and electroreceptors.Note that ampullary organs have significantly more Cacna1d-expressing receptor cells than neuromasts.(C,D) Ventral view of a second control Tyr crispant.Cacna1d expression reveals the infraorbital neuromast line on both sides of the ventral rostrum, flanked by the dorsal infraorbital (di) and ventral infraorbital (vi) ampullary organ fields.(E,F) Lateral view of a Foxg1 crispant.Cacna1d expression reveals that distinct neuromast lines are missing and the corresponding space is filled by putative ectopic ampullary organs, based on the large, ampullary organ-like clusters of Cacna1d-expressing cells.The dorsal and ventral infraorbital ampullary organ fields seem to have fused across the missing neuromast line (compare with A,B).(G,H) Ventral view of a second Foxg1 crispant.Cacna1d expression reveals an apparent fusion of the dorsal infraorbital (di) and ventral infraorbital (vi) ampullary organ fields across the missing infraorbital neuromast lines on both sides (compare with C,D).(I-K) In a third

Figure 4 .
Figure 4. Ectopic ampullary organs in Foxg1 crispants express ampullary organ-specific transcription factor genes Mafa and Neurod4.Sterlet crispants at stage 45 after in situ hybridization (ISH) for ampullary organ-restricted transcription factor genes.Black arrowheads indicate examples of neuromasts; white arrowheads indicate examples of ampullary organs.(A-D) In an uninjected sibling/half-sibling (eggs were fertilized in vitro with a mix of sperm from three different males), Mafa expression is restricted to ampullary organs (lateral view: A,B; ventral view: C,D).(E-J) A Foxg1 crispant.On the left side of the head (E,F), several Mafa-expressing ectopic ampullary organs are present in the space where the infraorbital neuromast line would normally run, such that the dorsal and ventral infraorbital ampullary organ fields seem to have fused (compare with A,B).Post-ISH Sox2immunostaining (E 1 ,F 1 ) shows that neuromasts are still present both proximally and distally to the sites of ampullary organ field fusion.In ventral view (G,H), ectopic ampullary organs (green arrowheads) are seen bilaterally, within the spaces where the infraorbital neuromast lines run on either side of the ventral rostrum (compare with C,D).On the right side in lateral view (I,J; image flipped horizontally for ease of comparison), a single Mafa-expressing ectopic ampullary organ (green arrowhead) is also present in the space where the infraorbital neuromast line runs (compare with A,B).(K-N) In an uninjected

Table 1
(Conant et al., 2022)oth ohnologs have been retained: one on chromosome 45 and the other annotated on an unplaced genomic scaffold, with 99.45% nucleotide identity (and 99.46% amino acid identity) in the coding sequence.Our sgRNAs target both Neurod4 ohnologues without mismatches.Our genotyping primers were designed before chromosome-level sterlet genomes were available; comparison with the reference genome showed five mismatches in the forward primer used for genotyping larvae targeted with sgRNAs 1 and 2 against the chromosome 45 ohnolog (Supplementary Table1).It was not possible to tell from our genotyping data whether the primers amplified both ohnologs or only one, however, as the remaining sequence targeted by the primers is identical between the two ohnologs.Genotyping and ICE analysis(Conant et al., 2022)showed successful ).This had no detectable effect on expression of electroreceptor-specific Kcnab3 (n=0/44 across nine batches, Supplementary Table 1) or the hair cell/electroreceptor marker Cacna1d (n=0/4 across two batches, Supplementary Table 1).Examples of Neurod4 crispants after ISH for Kcnab3, plus a Tyr control crispant for comparison, are shown in Supplementary Figure S4B-D.Our sgRNAs were designed before chromosome-level sterlet genomes were disruption of the Neurod4 gene arising from six different combinations of the sgRNAs (Supplementary Table 1; examples are shown in Supplementary FigureS4E-I).Not all sgRNA combinations were genotyped, but all Neurod4 crispants counted (n=44) included at least one sgRNA confirmed to disrupt the Neurod4 gene via genotyping of other embryos (Supplementary Table1).
Upon reaching the desired developmental stages, embryos and yolk-sac NEB) and stored at -80 °C before use.Alternatively, chemically modified synthetic gRNAs were ordered directly from Synthego (CRISPRevolution sgRNA EZ Kit).
České Budějovice, Czech Republic, and by the Ministry of Agriculture of the Czech Republic (reference number: MSMT-12550/2016-3).Experimental fish were maintained according to the principles of the European Union (EU) Harmonized Animal Welfare Act of the Czech (