isl1a coordinates onset and termination of regeneration of the zebrafish lateral line

Abstract

Regenerative capacity varies widely across tissues and species. Most regeneration studies have focused on identifying genes and pathways that promote regeneration, whereas mechanisms that inhibit regeneration to enforce organ size control and prevent excessive proliferation remain less well defined. Here, we identify isl1a as a dynamically expressed negative regulator of lateral line hair cell regeneration. isl1a is highly expressed during homeostasis but becomes transiently downregulated following hair cell loss. Loss of isl1a leads to proliferative expansion of support cells, producing enlarged sensory organs with increased hair cell numbers after injury. Mechanistically, isl1a downregulation increases chromatin accessibility at pro-proliferative loci, including AP-1 transcription factors, accompanied by sustained expression of injury- and regeneration-responsive genes. Together, these findings demonstrate that injury-induced downregulation of isl1a promotes the initiation of regeneration, while its subsequent re-expression restricts the duration of the response, thereby fine-tuning hair cell regeneration and maintaining appropriate support cell numbers.

Introduction

The capacity to regenerate upon injury varies widely across tissues and organisms, and understanding the mechanisms that govern this regenerative potential remains a central question in developmental biology. One example of this variability is found in sensory systems. Mechanosensory hair cells in the vertebrate inner ear are essential for hearing and balance, but their loss is irreversible in mammals. In contrast, the zebrafish sensory lateral line system robustly regenerates hair cells after injury^1–3, making it a powerful model to study the cellular and molecular basis of regeneration.

The lateral line is composed of mechanosensory organs, called neuromasts, that are organized in linear patterns along the head and trunk (Fig. 1a). During development, trunk neuromasts are periodically deposited by a lateral line primordium that originates posterior to the otic placode and migrates to the tip of the tail⁴. At 5 days post fertilization (dpf), mature neuromasts consist of 50–70 cells and contain mechanosensory hair cells that resemble the mammalian inner ear hair cells in morphology and function⁵. These hair cells are surrounded by distinct populations of support cells and an outer ring of mantle cells that form the boundary of the organ (Fig. 1b, c)^6,7. Central support cells, located in the center of the organ below the hair cells, serve as the primary hair cell progenitor population and divide symmetrically to give rise to new hair cells during both homeostasis and regeneration^7,8. Support cells adjacent to mantle cells also proliferate symmetrically but give rise to support cells^6,7. Although several signaling pathways such as Notch, Fgf, and Wnt have been implicated in the hair cell regeneration process^7,9–11, the transcriptional regulators that control progenitor activation and regenerative competence are not fully understood.

Fig. 1. Temporal dynamics of isl1a expression in development and regeneration.

a Confocal maximum intensity projection of a 5 dpf zebrafish larva. Lateral line is labeled in cyan by Tg(she:H2B-EGFP). Scale bar = 500 μm. b, c Schematic of a neuromast showing a top (b) and lateral view (c) in homeostasis⁹. Hair cells are labeled in green, central support cells in blue and amplifying support cells in red. d, e t-SNE plots of the different neuromast cell populations at 5 dpf in homeostasis⁹ (d), and the central support cell-specific expression of isl1a (e). f–h Confocal maximum intensity projections of HCR-FISH for isl1a in homeostatic wild-type embryos at 32 hpf (f), 2 dpf (g) and 5 dpf (h). Scale bar = 20 μm. i Heatmap of isl1a average gene expression in lateral line cell types in a regeneration timecourse generated from Baek et al. regeneration scRNA-seq shinyapps.io²⁰. Hair cell progenitors (HC-prog) and hair cells (HCs). j–p Confocal maximum intensity projections of HCR-FISH for isl1a (magenta dotted signal) at 5 dpf in homeostasis and during regeneration. isl1a is highly expressed in the central cells in homeostasis (j), but significantly decreased at 0 min (k), 30 min (l) and up to 1 h (m) post-neomycin treatment. By 3 h (n) isl1a expression increases in the central cells, and by 5 h (o) and 10 h (p), it returns to levels similar to homeostasis. Scale bar = 20 μm. b–e are adapted from Lush et al. eLife 8:e44431 (2019)⁹, under the terms of the Creative Commons Attribution License.

LIM-homeodomain (LIM-HD) transcription factors represent an interesting starting point to understand the transcriptional changes associated with successful regeneration, as they have been implicated in the regulation of cell proliferation, differentiation, and survival across tissues during development^12–14. Within this family, Islet-1 (Isl1) is of particular interest due to its critical role in auditory neuron development^15–17 and the regulation of hair cell differentiation and survival in the mouse auditory system. Ectopically induced co-expression of Isl1 and Atoh1 in supporting cells enhances their ability to transdifferentiate into hair cells¹⁸. Moreover, overexpression of Isl1 in the postnatal hair cells also promotes hair cell survival and prevents hearing loss caused by aging or acoustic overstimulation¹⁹. However, the molecular mechanism by which Isl1 regulates these functions is not clear.

Recent single-cell RNA sequencing (scRNA-seq) of zebrafish neuromasts revealed that isl1a is selectively expressed in central support cells and dynamically responds to hair cell death^9,20. Because of its dynamic expression pattern and roles in progenitor regulation, isl1a is an excellent candidate as a regulator of the regeneration process. We explored if isl1a acts as a central modulator of the transcriptional programs leading to successful regeneration, and more specifically, whether isl1a integrates injury signals with downstream transcriptional responses involved in cell proliferation and differentiation.

Using CRISPR/Cas12a-mediated genome editing, we generated a whole-gene deletion of isl1a and investigated its role in regulating neuromast cell number, cell type composition, regenerative dynamics, and transcriptional and epigenomic responses in the zebrafish lateral line. We show that loss of isl1a leads to an expansion of both central support cell and hair cell populations, with little to no effect in other cell populations, demonstrating that isl1a acts as a negative regulator of the progenitor pool size in neuromasts. We also show that loss of isl1a results in enhanced chromatin accessibility that leads to the prolonged transcriptional activation of pro-proliferation- and regeneration-related genes. The cumulative impact of these broad transcriptional changes results in a striking increase in support cell numbers. Our data demonstrate that isl1a is a critical negative regulator of regeneration, providing new insights into the transcriptional and epigenomic constraints that orchestrate the regenerative response.

Results

isl1a is expressed in central support cells and hair cell progenitors during development and regeneration

Our scRNA-seq analysis showed that mature 5 dpf neuromasts highly express isl1a in central support cells, which give rise to hair cell progenitors⁹ (Fig. 1d, e). Using in situ hybridization chain reaction (HCR-FISH) in developing wild-type embryos, we found that isl1a was not expressed in the migrating primordium or the first deposited neuromast at 32 h post fertilization (hpf) (Figs. 1f and S1a). isl1a starts to be expressed in the central region of deposited neuromasts at 2 dpf (Fig. 1g). As neuromasts mature, isl1a expression gradually becomes more restricted to the central support cell population at 5 dpf (Fig. 1h).

Given the robust expression of isl1a in central support cells during development, we wondered how isl1a expression changed throughout the regeneration process. scRNA-seq regeneration data throughout hair cell regeneration showed that isl1a is dynamically regulated during the first 10 h of regeneration after neomycin-induced hair cell death²⁰. Its expression exhibited rapid changes in response to injury, with a marked upregulation observed specifically in central support cells and hair cell progenitors within 3 h following neomycin treatment (Fig. 1i). As hair cell progenitors differentiate into young and then mature hair cells, the expression of isl1a becomes progressively downregulated (Figs. 1i and S1b). No significant changes in expression were observed in other cell populations after injury (Fig. S1c). To validate the changes in mRNA expression levels in the neuromast during regeneration, we performed HCR-FISH after neomycin-induced hair cell death (Fig. 1j–p). Consistent with the transcriptomic data, isl1a expression was initially downregulated after hair cell death and was subsequently upregulated between 1–3 h after neomycin treatment (Fig. 1m, n and S1d). This corresponds to the key time window of regeneration immediately before hair cell specification and progenitor proliferation commences²⁰. Taken together, these data show that isl1a is expressed in homeostatic central support cells and hair cell progenitors after neuromasts have been deposited and is dynamically regulated during regeneration.

Loss of isl1a expression leads to an increase in central support cell and hair cell numbers in homeostasis

To assess the function of isl1a in neuromast development and regeneration, we analyzed an isl1a stop codon mutation (isl1a^sa29/sa29)^21,22 and generated a whole-gene deletion mutant using CRISPR/Cas12a to avoid possible effects on the phenotype by transcriptional compensation²³ (isl1a^psi97/psi97, Fig. S2a). The isl1a^sa29/sa29 mutation removes part of the LIM-domain and the entire homeobox, but truncated isl1a mRNA is still transcribed (Fig. S2f, g). In contrast, no isl1a mRNA is transcribed in the whole gene deletion larvae (Fig. S2h, i). Because we did not detect differences in the severity of the phenotypes, we subsequently performed all experiments with the isl1a^psi97/psi97 null allele, hereafter referred to as isl1a^−/−.

At 5 dpf, the isl1a^−/− larvae were characterized by a severe heart edema, bradycardia and irregular heartbeat. Mutants also had a small jaw, failed to extend axonal projections (Fig. S2b–e), and showed motility defects and a failure to inflate their swim bladder. Nonetheless, the sensory lateral line formed normally in isl1a^−/− larvae, with a similar number and position of neuromasts as their sibling counterpart (Fig. 2a, b).

Fig. 2. Developmental and homeostatic phenotypes of isl1a mutants.

a, b Confocal maximum intensity projection of 5 dpf zebrafish sibling (a) and isl1a^−/− (b) larva. Lateral line is labeled in cyan by Tg(she:H2B-EGFP) and all other cells by DAPI. Scale bar = 500 μm. c, d Confocal maximum intensity projections of zebrafish sibling (c) and isl1a^−/− (d) neuromasts throughout development (2–7 dpf). Hair cells are labeled in magenta by Tg(myo6b:H2B-mScarlet-I) and all other cells by DAPI. Scale bar = 20 μm. e–g Quantification of the number of neuromast cells at different developmental stages. The total number of neuromast cells was quantified by DAPI (e), myo6b:H2B-mScarlet-I⁺ cells were quantified as hair cells (f), and the hair cell ratio was calculated by dividing the number of hair cells by the total cell number in the neuromast at each stage (g). n = 28 neuromasts from 7 larvae per condition. Data are represented as mean ± standard deviation (SD). **p < 0.01, ***p < 0.001, ****p < 0.0001 by two-way ANOVA (genotype, timepoint) with Šidák’s multiple comparison test. h, i Confocal maximum intensity projections of 5 dpf sibling (h) and isl1a^−/− (i) neuromasts, where hair cells are labeled in magenta by Tg(myo6b:H2B-mScarlet-I), central support cells in cyan by Tg(fgf3:H2B-mturquoise2), and mantle cells in yellow by ET20:EGFP. j–m Cell quantification of myo6b:H2B-mScarlet-I⁺ hair cells (j), fgf3:H2B-mturquoise2⁺ central support cells (k) and ET20:EGFP⁺ mantle cells (l). DV and AP cells were calculated by subtracting all fluorescently labeled cells from the total neuromast cell number (m). n = 20 neuromasts from 5 larvae per condition. Data are represented as mean ± SD. **p < 0.01 by unpaired two-tailed t-test in (j, k) and two-tailed Mann–Whitney U test in (l, m).

To test the role of isl1a in lateral line primordium migration and neuromast formation, we analyzed isl1a^−/− larvae with the reporter line Tg(cxcr4b:H2A-GFP)²⁴. This transgenic line labels all nuclei in the lateral line primordium and its deposited neuromasts. We quantified the cell number in the migrating primordium and the first deposited neuromast at 32 hpf in isl1a^−/− larvae and sibling controls (Fig. S2j, k) and found no significant differences between sibling and mutant larvae at this stage (Fig. S2l, m). These results demonstrate that neither primordium migration nor newly deposited neuromasts were affected by the loss of isl1a. To determine if isl1a influenced neuromast development as the organs mature, we analyzed a lateral line-specific reporter line (Tg(she:H2B-EGFP))²⁵ and a hair cell-specific reporter line (Tg(myo6b:H2B-mScarlet-I))²⁵. Quantification of the total number of neuromast cells and hair cells in isl1a^−/− larvae revealed a gradual increase in neuromast cell number and hair cell number in the isl1a^−/− larvae that became significantly different from siblings at 5 dpf (Fig. 2c–f). However, when we calculated the ratio of hair cells to total neuromast cell number (hair cell ratio), there was no significant difference (Fig. 2g). This shows that, although isl1a^−/− neuromasts possess significantly more hair cells by 5 dpf, this increase is proportional to the increase in total neuromast cell numbers.

Since neuromasts are not only comprised of hair cells, we examined whether the increase in neuromast cell number in isl1a^−/− larvae also affected other cell populations. Using a combination of transgenic lines, we quantified hair cells (Tg(myo6b:H2B-mScarlet-I)), central support cells (Tg(fgf3:H2B-mturquoise2))⁹, and mantle cells (ET20:EGFP)²⁶ in 5 dpf homeostatic sibling and isl1a^−/− neuromasts (Fig. 2h, i). isl1a^−/− neuromasts showed an increase in hair cells and central support cells (Fig. 2j, k), but not in mantle cells or other support cell types (Fig. 2l, m). Altogether, and consistent with its spatial and temporal expression, these data indicate that isl1a is not required for posterior lateral line primordium migration or neuromast deposition but is essential for regulating the number of central support cells as neuromasts mature.

isl1a^−/− neuromasts show enhanced hair cell regeneration driven by overproliferation

Since isl1a^−/− neuromasts are characterized by an increased number of central support cells, we wondered if loss of isl1a would also lead to an enhanced regenerative response following injury. We therefore ablated hair cells at 5 dpf with neomycin and quantified the number of regenerated hair cells (Figs. 3a, b and S3a, b). At 48 h after ablation, isl1a^−/− neuromasts showed a significant increase in total cell and hair cell number (Fig. 3c, d).

Fig. 3. isl1a modulates hair cell regeneration and support cell proliferation.

a–b Confocal maximum intensity projections of sibling (a) and isl1a^−/− (b) neuromasts at 5 dpf in homeostasis and 48 h post-neomycin treatment. Hair cells are labeled in magenta by Tg(myo6b:H2B-mScarlet-I) and all other cells by DAPI. Scale bar = 20 μm. c, d Quantification of the number of neuromast cells in regeneration. At 48 h post-neomycin treatment, the total number of neuromast cells was quantified by DAPI (c), myo6b:H2B-mScarlet-I⁺ cells were quantified as hair cells (d). n = 24 neuromasts from 6 larvae per condition. Data are represented as mean ± SD. ***p < 0.001, ****p < 0.0001 by unpaired two-tailed Welch’s t-test in (c) and unpaired two-tailed t-test in (d). e–f Maximum intensity projection of confocal z-stacks showing EdU staining 24 h post hair cell death in siblings (e) and isl1a^−/− (f) larvae. Hair cells are labeled in magenta by Tg(myo6b:H2B-mScarlet-I), neuromast cells are labeled in cyan by Tg(she:H2B-EGFP) and EdU cell are labeled in yellow. Scale bar = 20 μm. g EdU differentiation index (EdU⁺ hair cells/total neuromast cell number), amplification index (EdU⁺ support cells in contact with mantle cells/total neuromast cell number) and total EdU index (all EdU⁺ cells/total neuromast cell number). n = 24 neuromasts from 6 larvae per condition. Data are represented as mean ± SD. ***p < 0.001, ****p < 0.0001 by multiple unpaired two-tailed Welch’s t-tests or two-tailed Mann-Whitney U tests when normality was not met, with Holm-Šidák correction for multiple comparisons. h–i Maximum intensity projections of confocal z-stacks showing EdU staining 24 h post hair cell death in siblings (h) and hs:isl1a OE (i) larvae. Hair cells are labeled in magenta by Tg(myo6b:H2B-mScarlet-I), neuromast cells are labeled in cyan by Tg(she:H2B-EGFP) and EdU cells are labeled in yellow. Scale bar = 20 μm. j EdU differentiation index (EdU⁺ hair cells/total neuromast cell number), amplification index (EdU⁺ support cells in contact with mantle cells/total neuromast cell number) and total EdU index (all EdU⁺ cells/total neuromast cell number). n = 24 neuromasts from 6 larvae per condition. Data are represented as mean ± SD. *p < 0.05, ****p < 0.0001 multiple unpaired two-tailed Welch’s t-tests or two-tailed Mann–Whitney U tests when normality was not met, with Holm-Šidák correction for multiple comparisons.

We then asked whether the increase of hair cells in isl1a^−/− mutants was the result of overproliferation of central support cells specifically. Previous work has demonstrated that during regeneration in the zebrafish lateral line, two distinct support cell populations are proliferating. Peripheral support cells divide and give rise to two new support cells (amplifying divisions), and hair cell progenitor cells divide and give rise to two hair cells (differentiating divisions)⁷. We defined amplifying divisions as EdU+ support cells in direct contact with mantle cells, which have been proposed to be the niche for amplifying cells⁷. We performed EdU incorporation assays for the first 24 h after hair cell death and calculated the EdU indexes (Fig. 3e, f). We observed a significant increase in both differentiating and amplifying cell divisions in isl1a^−/− mutants (Fig. 3g). These differences in the EdU indexes were preserved when incorporating EdU from 24 h to 48 h after hair cell death (Fig. S3c–f). Together, these data indicate that the increased hair cell production in isl1a mutants is driven by increased proliferation of central support cells, the direct progenitors of hair cells, as well as increased amplifying divisions that can give rise to hair cells if medially displaced. The increase in amplifying cell proliferation is likely a non-cell-autonomous effect as isl1a expression in amplifying support cells is negligible (Fig. S1e).

To further test if isl1a normally inhibits cell proliferation in neuromasts, we generated a heat shock isl1a overexpression line Tg(hsp70l:isl1a;myl7:EGFP), hereafter (hs:isl1a OE) (Fig. S3g–j). We heat-shocked 5 dpf larvae and performed a regeneration assay with EdU incorporation for 24 h (Fig. 3h, i). Calculation of the EdU indexes revealed a significant decrease in both differentiating and amplifying cell divisions in hs:isl1a OE neuromasts when compared to siblings, with an overall decrease in cell proliferation (Fig. 3j). In summary, we demonstrated by both loss- and gain-of-function experiments that during regeneration, isl1a negatively regulates support cell proliferation.

The role of isl1a in regeneration is mediated by its neuromast-autonomous activity in the neuromast

The EdU incorporation assay in regenerating isl1a^−/− neuromasts showed not only an increase in cell proliferation (EdU⁺ cells) in the neuromast but also in the surrounding skin (Figs. 3f and S3e). Although isl1a does not appear to be expressed in the skin cells in zebrafish larvae at 5 dpf, it is expressed in sensory neurons^27–29 that are mostly missing in isl1a^−/− larvae. Work by others has shown that axonal terminations can regulate epidermal adhesion³⁰, which in turn can regulate proliferation in the skin³¹. We therefore wondered if the proliferation defect in neuromasts was secondary to the requirement of isl1a in the surrounding neurons/skin or if the neuromast proliferation defect was due to a neuromast-autonomous role of isl1a. To generate a neuromast-specific isl1a deletion, we transplanted cells from isl1a^−/− and sibling embryos, expressing two fluorescent reporters that label lateral line cell membranes (Tg(cldnb:lynEGFP))³² and nuclei (Tg(cldnb:H2A-mCherry))⁷, into TU wild-type embryos (Fig. 4a). We then ablated hair cells in transplanted larvae at 5 dpf and quantified hair cell and neuromast cell numbers 48 h after injury (Fig. 4b–e). Transplanted isl1a^−/− cells led to a significant increase in neuromast cell number and hair cell number when compared to transplanted sibling cells (Fig. 4f, g). This recapitulates the observed phenotype in non-manipulated larvae, demonstrating that the neuromast phenotype is due to the requirement of isl1a in neuromasts and not in the surrounding tissues. The increase in size and hair cell number persisted in transplanted fish raised to 30 dpf (Fig. S4a–d). Given that loss of isl1a leads to a specific increase in the central support cell and hair cell pools in homeostasis, we next asked whether isl1a affects only these cell populations in regeneration. We again used a combination of transgenic lines and HCR-FISH to label different neuromast cell populations and quantified these populations in isl1a^−/− and sibling neuromasts 48 h after hair cell death (Fig. 4h, i). Consistent with the homeostatic analysis, only the central support cells and hair cells displayed a significant increase in isl1a^−/− neuromasts in regeneration (Fig. 4j, k), whereas dorso-ventral (DV), anterior-posterior (AP) and mantle cell populations remained unaffected in mutants (Fig. 4l, m), indicating an expansion of the isl1a-positive population. Importantly, although we detected a small statistical difference, the overall ratio of hair cells to central support cells remained comparable between genotypes (Fig. 4n), suggesting that the increased hair cell number reflects proportional expansion of the central support cell pool rather than enhanced differentiation of support cells into hair cells.

Fig. 4. isl1a plays a neuromast-autonomous role in support cells during regeneration.

a Schematic of cell transplantation experiment. b–e Confocal maximum intensity projections of sibling (b) and isl1a^−/− (d) donor larvae 48 h post-neomycin treatment showing cldnb:H2A-mCherry+ cells in black. Gamma was modified in (b) and (d) to allow for visualization of dim structures. Scale bar = 500 μm. cldnb:H2A-mCherry+ cells in magenta represent the transplanted cells in neuromasts from sibling (c) and isl1a^−/− (e) donor larvae. DAPI was used to label all other cells in the host larvae. Scale bar = 20 μm. f, g Quantification of the number of neuromast cells in regeneration in transplanted larvae and donor controls (not manipulated larvae from the same clutch). At 48 h post-neomycin treatment, the total number of neuromast cells was quantified by DAPI (f), hair cells were quantified by their distinct DAPI incorporation pattern and position in the neuromast (g). n = 18 neuromast from 5 larvae in sibling group and n = 20 neuromast, 4 larvae in isl1a^−/− group. Data are represented as mean ± SD. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 by multiple unpaired two-tailed Welch’s t-tests with Holm-Šidák correction for multiple comparisons. h–i Confocal maximum intensity projections of sibling (h) and isl1a^−/− (i) neuromasts 48 h post-neomycin treatment. Hair cells are labeled in magenta by Tg(myo6b:H2B-mScarlet-I) and central support cells in cyan by Tg(fgf3:H2B-mturquoise2). sost-B1 HCR-FISH (magenta dotted signal) labels DV cells and si:ch73-261i21.5-B4 HCR-FISH (yellow dotted signal) labels AP cells. Scale bar = 20 μm. j-n Cell quantification of myo6b:H2B-mScarlet-I⁺ hair cells (j), fgf3:H2B-mturquoise2⁺ central support cells (k), sost⁺ DV cells (l), si:ch73-261i21.5⁺ AP cells (m), and the ratio of hair cells to central support cells (HC to CSC) (n). n = 25 neuromast from 5 larvae per condition. Data are represented as mean ± SD. *p < 0.05, ****p < 0.0001 by two-tailed Mann–Whitney U test in (j, l, m), and two-tailed unpaired Welch’s t-test in (k, n).

Together, the transplantation experiments and mutant analyses of neuromast cell populations demonstrate that during regeneration, isl1a limits central support cell proliferation cell-autonomously. Amplifying cell proliferation is upregulated secondarily by loss of isl1a, as isl1a is not expressed in this cell population. How central support cells signal to amplifying cells remains to be determined.

isl1a represses pro-proliferative signaling and modulates regeneration-associated gene expression

To investigate how isl1a regulates gene expression in the lateral line, we performed bulk RNA sequencing (bulk RNA-seq) on FAC-Sorted neuromast cells isolated from isl1a^−/− and sibling 5 dpf larvae in homeostasis and 3 h post-neomycin treatment, when isl1a is most highly re-expressed after hair cell death (Figs. 5a and 1n). Differential gene expression analysis revealed that during homeostasis, approximately 400 genes are differentially expressed between isl1a mutants and siblings, with the number of upregulated genes strongly increasing 3 h after hair cell death (Fig. 5b). This supports the idea that isl1a regulates regeneration of the lateral line by mostly repressing gene expression (Fig. S5a–c and Supplementary data 1). Heatmaps of the top up- and downregulated genes illustrate the broad transcriptional shift in isl1a mutants in regeneration (Fig. 5c, d).

Fig. 5. isl1a restricts pro-proliferative signaling and fine-tunes early transcriptional programs in hair cell regeneration.

a Schematic overview of bulk RNA-seq experiment. Samples were collected at two different timepoints. Tg(she:H2A-mCherry) reporter was used to FAC-Sort lateral line cells for each condition. b Bar plot showing total number of differentially expressed genes (DE) between sibling and isl1a^−/− neuromasts in homeostasis and regeneration (3 h post neo). Triplicates for each condition were merged prior to DE analysis. c–d Heatmaps representing the top 30 upregulated genes in regeneration at 3 h post-neomycin in isl1a^−/− (c) and the top 30 downregulated genes at 3 h post-neomycin in isl1a^−/− (d) by bulk RNA-seq. e Transcriptional modules after neomycin-induced hair cell death, as described in Baek et al.²⁰. f Heatmap of Module 1 upregulated genes that are upregulated at 3 h post-neomycin in isl1a^−/− by bulk RNA-seq. g–j HCR-FISH for junba (g, h) and dusp1 (i, j) in a regeneration time course to compare their mRNA expression levels between sibling and isl1a^−/− at each regeneration time point. Scale bar = 20 μm. k Heatmap of Module 2 upregulated genes that are upregulated at 3 h post-neomycin in isl1a^−/− by bulk RNA-seq. l–o HCR-FISH for lima1a (l, m) and cbx7a (n, o) in a regeneration time course to compare their mRNA expression levels between sibling and isl1a^−/− at each regeneration timepoint. Scale bar = 20 μm. p Heatmap of Module 3 upregulated genes that are upregulated at 3 h post-neomycin in isl1a^−/− by bulk RNA-seq. q–r HCR-FISH for ctnnb1 in a regeneration time course to compare its mRNA expression level between sibling (q) and isl1a^−/− (r) at each regeneration timepoint. Scale bar = 20 μm.

To characterize the role of isl1a during regeneration, we performed pathway enrichment analysis³³ (Fig. S5d) using the differentially expressed gene list at 3 h after hair cell death. Most differentially upregulated genes in the isl1a^−/− regeneration dataset were associated with pathways involved in the regulation of cell proliferation. Among the most highly enriched terms were tight junctions and focal adhesion junctions, MAPK, Wnt, and ErbB signaling, supporting the hypothesis that isl1a negatively regulates cell proliferation. A similar pathway analysis with the downregulated genes did not reveal any enriched pathways.

We next asked when, during regeneration, isl1a plays its most critical role. Our previous scRNA-seq regeneration analysis identified three sequentially activated modules in regeneration: an immediate inflammatory response along with the downregulation of progenitor cell maintenance genes (Module 1), followed by transient regeneration-specific gene activation (Module 2), and, finally, a robust reactivation of developmental programs (Module 3) (Fig. 5e). To determine whether isl1a plays a role during a specific transcriptional phase, we systematically compared our bulk RNA-seq differential expression data at 3 h after neomycin to each of the module-specific gene sets defined in the scRNA-seq regeneration dataset. Comparison with Module 1 revealed that several injury and inflammatory response genes, that are normally only expressed for 30 min, failed to be downregulated in isl1a^−/− neuromasts at 3 h post-neomycin (Fig. 5f). We validated this prolonged mRNA expression in isl1a^−/− neuromasts with HCR-FISH for junba and dusp1 in a regeneration time course that matches the scRNA-seq data (Figs. 5g–j and S5e, f). This suggests that isl1a normally limits the duration of expression of Module 1 genes, although these genes are eventually downregulated even in the mutant. Similarly, we found that several Module 2 regeneration-specific genes were also upregulated in isl1a^−/− neuromasts at 3 h after neomycin treatment (Fig. 5k), a result that we confirmed by HCR-FISH for lima1a and cbx7a (Figs. 5l–o and S5g, h).

Fewer Module 3 genes were differentially expressed in the isl1a^−/− neuromasts (Fig. 5p), and among these, several were associated with the Wnt signaling pathway, such as ctnnb1 (Figs. 5q, r and S5i, j). We and others previously demonstrated that Wnt, Notch, and Fgf signaling are crucial for hair cell regeneration in the lateral line^7,9–11. We therefore examined if members of these signaling pathways were differentially expressed in the 3 h post-neomycin transcriptomic data. Our data showed that there were only subtle differences in mRNA expression of Notch and Fgf signaling candidate genes between sibling and mutant neuromasts (Fig. S5k, l), suggesting that isl1a is not regulating these pathways.

In summary, comparison with genes belonging to pre-established regeneration modules demonstrates that isl1a plays a role in regulating the temporal dynamics of gene activation during the early phase of hair cell regeneration. Also, isl1a is regulating proliferation partially through Wnt signaling, and isl1a broadly restricts regenerative gene expression dynamics in parallel to Notch and Fgf signaling.

isl1a likely restricts chromatin accessibility of proliferation-associated regulatory regions early in regeneration

Given that in mouse cardiac cell progenitors, Isl1 acts as a pioneer transcription factor with the ability to drive chromatin opening^34,35, we asked whether isl1a could similarly modulate chromatin accessibility during neuromast regeneration. To address this, we conducted ATAC-seq on FAC-Sorted neuromast cells from sibling and mutant larvae at 5 dpf in homeostasis, 1 h, and 3 h post-neomycin treatment (Fig. 6a). Differential chromatin accessibility analysis revealed that isl1a^−/− neuromasts exhibit a greater number of differentially accessible regions (DARs) at 1 h post injury when compared to homeostasis and 3 h post-neomycin timepoints, with a particularly strong increase in accessibility at this timepoint (Fig. 6b and Supplementary data 2). This finding suggests that isl1a regulates neuromast regeneration by reducing chromatin accessibility and limiting access to the underlying regulatory DNA. We therefore asked if the observed opening of the chromatin in isl1a^−/− larvae impacted the transcriptional regulation captured at 3 h in regeneration. First, we assigned DARs to genes across timepoints by associating each DAR to its nearest promoter. Cross-referencing the DARs from 1 h post-neomycin with differentially expressed genes at 3 h post-neomycin revealed a strong correspondence between early chromatin opening in isl1a^-/- and later upregulation of regeneration-related genes (Fig. 6c). This suggests that epigenomic regulation at 1 h post-neomycin indeed drives gene expression changes captured at 3 h post-neomycin.

Fig. 6. isl1a deletion alters chromatin accessibility dynamics during early regeneration.

a Schematic overview of bulk ATAC-seq experiment. Samples were collected at three different timepoints. Tg(she:H2A-mCherry) reporter was used to FAC-Sort lateral line cells for each condition. b Bar plot showing total number of differentially accessible regions (DARs) in homeostasis and regeneration (1 h and 3 h post neo) in isl1a^−/− samples. Differential accessibility was calculated comparing isl1a^−/− against sibling data using the DESeq2 framework⁸¹. Triplicates for each condition were merged prior to DE analysis. c Average accessibility plots showing the normalized ATAC-seq signal values at ATAC-seq peaks that were closest to a differentially expressed gene identified at 3 h post neo by bulk RNA-seq. ATAC-seq peak summits were used as the anchor point for plotting chromatin accessibility data. d Bar plot of the top enriched transcription factors (TFs) in isl1a^−/− most accessible regions at 1 h post neo compared to non-differential accessible regions as background. The TF list was obtained by running the MEME suite tool AME⁸² and using the JASPAR 2022 vertebrate core TF motif database⁸³ as reference. e Normalized ATAC-seq genome tracks representing the topmost accessible regions at 1 h post neo that are associated with TFs and are also upregulated at 3 h post-neomycin in the transcriptomic data. f Classification of differentially accessible ATAC-seq peaks (associated with differential transcriptomic data 3 h after neomycin and with at least one Isl1 motif) into four categories based on whether the DAR is more open or closed in isl1a mutants, and whether the associated gene is transcriptionally up- or downregulated. g Heatmap representing the log2FC of TFs, from (f) in the ATAC-seq 1 h post neo and the RNA-seq 3 h post neo data. h Normalized ATAC-seq track of the TF prdm1a from (g), where the DAR containing Isl1 motifs is highlighted. There are other DARs represented that do not contain Isl1 motifs. i Confocal maximum intensity projection of HCR-FISH for prdm1a in sibling and isl1a^−/− neuromasts at 3 h post-neomycin. Scale bar = 20 μm.

To uncover broader transcription factor activity in our dataset independent of direct isl1a regulation, we analyzed the most accessible regions in isl1a^−/− neuromasts to identify enriched transcription factor motifs. We found a strong enrichment of AP-1 family motifs (including jun, jund, junb, atf3, fos, fosl1 and fosl2 (Fig. S6a)), as well as motifs that belong to the ETS and homeodomain superfamilies (Fig. 6d). Many of the transcription factors enriched in the isl1a^−/− ATAC-seq data, such as atf3, jun, bach2b, and junba, were themselves significantly more accessible and upregulated in the isl1a^−/− transcriptomic data in regeneration (Fig. 6e). These results suggest that the loss of isl1a facilitates the expression of AP-1 family members to amplify the pro-proliferative and injury-responsive gene expression programs during regeneration.

To test whether isl1a potentially represses chromatin accessibility directly, we curated a list of only differentially accessible ATAC-seq peaks that were associated with a differentially expressed gene in the 3 h post-neomycin transcriptomic data. Of the 805 DARs from this curated list, we retrieved 112 DARs that contained at least one Isl1 motif and were therefore presumed to be direct targets of isl1a. We classified these regions into four categories based on whether the DAR is more open or more closed in isl1a^−/− mutants compared to siblings, and on whether the gene associated with it is transcriptionally up- or downregulated (Fig. 6f). Among the 112 DARs with an Isl1 motif, 86 were more open and associated with an upregulated gene in isl1a mutants. This result suggests that isl1a acts mostly as a repressor by restricting chromatin accessibility and thereby downregulating the target genes during regeneration.

To place isl1a in the regeneration gene regulatory network, we identified the transcription factors associated with the 112 DARs in regeneration that contained at least one Isl1 motif (Fig. 6g). prdm1a, bach2b, nr3c1, jund, tfe3a and smad3a stood out for showing both increased chromatin accessibility at 1 h post-neomycin and transcriptional upregulation at 3 h post-neomycin in isl1a^−/− larvae (Fig. 6h, i and S6b, c). These transcription factors have been linked to cell proliferation, and they promote or restrict proliferation in a context-dependent manner^36–42. Out of the DARs associated with prdm1a in our ATAC-seq data, only one intergenic region contains at least one Isl1 motif (Fig. 6h). This regulatory region could be responsible for the transient central support cell expression driving central support cell proliferation⁴³, which remains to be validated experimentally. The presence of Isl1 motifs in proliferation-associated transcription factors that become more accessible and upregulated in the isl1a^−/− larvae suggests that isl1a may directly repress their activation during homeostasis by limiting their chromatin accessibility.

Collectively, these results indicate that isl1a acts as an activator and repressor during homeostasis, as its loss leads to up- and downregulation of genes. However, during regeneration, it mainly acts as a repressor that needs to be rapidly downregulated after hair cell loss to allow upregulation of AP-1 factors and chromatin opening of pro-proliferative loci, leading to support cell proliferation and hair cell regeneration. The dynamic regulation of isl1a modulates the initiation of regeneration but also limits the magnitude and duration of the regenerative response.

Discussion

Although the transcriptional programs underlying hair cell regeneration have been extensively mapped in zebrafish^11,20, the upstream regulators that coordinate the timing and magnitude of this response are not fully understood. Here, we identify isl1a as a component of the gene regulatory network that modulates regeneration in the zebrafish lateral line. We demonstrate that isl1a constrains proliferation during homeostasis and tempers the regenerative response by limiting progenitor expansion and chromatin accessibility at regeneration-associated loci. Regeneration-associated genes (e.g., lima1a, cbx7a, sall1b, dusp5), which are normally expressed in a transient, time-restricted manner, remain upregulated for an extended period upon hair cell death in isl1a^−/− neuromasts. These cellular and molecular alterations culminate in overproliferation of support cells, leading to increased hair cell numbers. Despite broad upregulation of proliferation-associated genes in isl1a mutants, only populations known to proliferate in neuromasts show increased proliferation, namely central support cells and amplifying cells^7,44. Notably, strong pro-proliferative perturbations such as Wnt activation similarly drive expansion within specific cell populations rather than uniformly across the neuromast⁷. What makes only certain cell populations competent to respond to pro-proliferative factors in neuromasts remains to be determined.

In wild-type neuromasts, isl1a mRNA decreases rapidly after injury, coinciding with the onset of regeneration-induced gene expression. It also coincides precisely with the downregulation of Notch and Fgf signaling pathway components^7,9, both of which inhibit proliferation during homeostasis and are reactivated 5 h after hair cell death. This temporal correlation, along with the fact that these pathways inhibit regeneration if overexpressed, raises the question of whether isl1a, Notch, and Fgf interact. Our data suggest that isl1a acts in parallel with Notch and Fgf signaling. Notch pathway members were not differentially expressed in the isl1a^−/− regeneration dataset, and pharmacological Notch inhibition did not lead to changes in isl1a mRNA expression. Likewise, the Fgf pathway was not significantly affected in our regeneration data.

This begs the question of how isl1a inhibits proliferation. isl1a appears to limit proliferation in part through the coordinated modulation of regenerative signaling cascades as revealed by the upregulation of KEGG pathways (e.g., MAPK, Wnt, ErbB) that have been implicated in the control of proliferation and regeneration^7,10,45–48. However, whether isl1a limits regeneration through the induction of a specific signaling pathway or through the coordinated control of MAPK, Wnt, ErbB, tight junctions and focal adhesion in parallel remains unclear. Our data support the latter, as many genes across these pathways exhibit modest but significant transcriptional changes in isl1a^−/− neuromasts. While the effect on individual genes may be subtle, the cumulative impact of these broad transcriptional changes results in a striking increase in support cell numbers. This highlights the ability of transcription factors to mediate widespread, coordinated regeneration responses through the regulation of many targets.

In addition to changes captured in signaling pathways, isl1a^−/− neuromasts also exhibit altered temporal dynamics of gene expression. Several regeneration-associated genes previously identified as transiently upregulated for only about 1 h during early stages of regeneration²⁰ displayed prolonged expression in isl1a^−/− neuromasts over the regeneration time course. This prolonged activation mirrors the temporal misregulation seen in other models of enhanced regeneration, where delayed silencing of regeneration-responsive enhancers leads to sustained gene activation that can impact tissue repair^49,50.

The ATAC-seq analysis provides mechanistic insight into how isl1a regulates the regenerative capacity. isl1a mutants exhibit an increase in chromatin accessibility at 1 h after neomycin treatment (Fig. 6c), indicating that isl1a normally restricts the access to regulatory regions that drive the earliest stages of the regenerative gene expression program (Module 1 genes). This model agrees with findings in other contexts in which chromatin accessibility dynamics shape the regenerative potential^51–54. In addition, a targeted approach focusing on DARs that contain predicted Isl1 motifs highlighted potential direct regulatory targets of isl1a. These genes included several transcription factors such as prdm1a, bach2b, and nr3c1, placing isl1a at the node of a regeneration-responsive gene regulatory network. However, most early opening DARs lack Isl1 motifs, suggesting that many of these regulatory regions are unlikely to be direct targets and instead reflect an indirect mechanism in which loss of isl1a permits early engagement of injury-response factors that can recruit chromatin remodelers to open regenerative regulatory elements⁵⁵.

Consistent with this model, many of the identified DARs are enriched for AP-1 transcription factor motifs, including jun, junba, atf3, fosl1, and fosl2 (Fig. S6a), which are among the earliest transcription factors induced after injury²⁰, and even more highly upregulated in the mutants. These AP-1 transcription factors drive proliferation and cellular reprogramming and promote chromatin accessibility during heart regeneration^55–58. Candidate chromatin remodelers and transcription cofactors that may interact with isl1a to promote the activation of AP-1 transcription factors include Smarca4 and Brn3b/Pou4f2. The chromatin remodeler SWI/SNF ATPase Smarca4 regulates progenitor plasticity and differentiation^59,60 and directly interacts with Isl1 during cardiogenesis, where it acts as a pioneer factor³⁴. However, Smarca4 loss leads to failure of cardiomyocyte differentiation and isl1a mutants do not possess defects in hair cell differentiation. Also, isl1a does not act as a pioneer factor in the lateral line. In the developing retina, Isl1 cooperates with Brn3b/Pou4f2 to regulate genes controlling neuronal differentiation and cell cycle exit^61,62. In the zebrafish lateral line, we did not detect enrichment of Brn3/Pou4 motifs at isl1a-dependent regulatory regions. Also, isl1a mutants do not possess hair cell differentiation defects that would be expected from a decrease in Brn3/Pou4 levels. Nevertheless, isl1a possibly acts with other cofactors in the lateral line. Identifying these cofactors of isl1a during regeneration will require future biochemical and genetic studies.

Previous studies have established the importance of central support cells as a progenitor pool for hair cell regeneration in zebrafish neuromasts^2,6. Here, we show that isl1a limits the regeneration response by constraining progenitor proliferation and hair cell production, largely through modulation of chromatin accessibility.

Although isl1a is commonly described as an activator of proliferation, differentiation, and survival across tissues during development^34,63, our findings reveal that during lateral line regeneration, it acts mostly as a repressor, similar to its function during chick dorsal root ganglion formation⁶⁴. In the mouse heart, Isl1 promotes proliferation and cardiomyocyte fate specification by acting as a pioneer transcription factor in cardiac cell progenitors³⁴.

Isl1 also controls neuronal differentiation during retinal development, where it promotes cell cycle exit and specification of distinct neuronal subtypes⁶⁵. Therefore, the question arose whether isl1a could similarly promote hair cell differentiation while suppressing proliferation in neuromasts. However, our data show that isl1a is not a differentiation factor during lateral line regeneration. In isl1a mutants, support cells retain the ability to differentiate into hair cells and even generate more hair cells during regeneration, arguing against a model in which loss of isl1a maintains cells in an undifferentiated proliferative state. Consistent with previous work showing that proliferation and differentiation are uncoupled in the zebrafish lateral line⁴⁴, our findings demonstrate a previously underappreciated role for isl1a in restricting the magnitude and duration of regenerative proliferation, rather than cell differentiation. In the mammalian auditory system, postnatal overexpression of Isl1 in hair cells protects them against age and noise-induced damage and improves survival¹⁹, while co-overexpression of Isl1 and Atoh1 enhances supporting-to-hair-cell conversion¹⁸. The latter finding suggests a role for Isl1 in hair cell differentiation in the mouse cochlea that is different from its role in the lateral line. To date, the consequences of Isl1 loss of function on chromatin accessibility or proliferation in mouse supporting cells following injury have not been described.

Together, these observations show that Isl1 activity is context-dependent and that it can function as an activator or repressor depending on the cellular environment and developmental stage. In zebrafish lateral line regeneration, isl1a constrains progenitor expansion via chromatin regulation. Whether the effects of Isl1 during development and regeneration in other organisms and tissues are also caused by chromatin accessibility changes remains to be determined.

Our findings demonstrate that isl1a plays an unexpected, pivotal role in limiting the regenerative response of the zebrafish lateral line by constraining chromatin accessibility and transcription of proliferation- and regeneration-specific genes. Injury-induced downregulation of isl1a promotes the initiation of regeneration, while its subsequent re-expression restricts the duration of the response, thereby fine-tuning hair cell regeneration and maintaining appropriate support cell numbers.

Methods

Zebrafish strains and husbandry

Zebrafish experiments were performed under the guidelines of the Stowers Institute for Medical Research IACUC-approved protocol 2023-159. Embryos and larvae were maintained in 0.5× E2 medium (7.5 mM NaCl, 0.25 mM KCl, 0.5 mM MgSO4, 75 mM KH2PO4, 25 mM Na2HPO4, 0.5 mM CaCl2, 0.5 mg/l NaHCO3, pH 7.4) and incubated at 28.5 °C. Embryos and larvae were euthanized by overdose of tricaine methanesulfonate (MS-222) prior to fixation or dissociation for downstream analyses. The following published transgenic lines were used in this study: Tg(cxcr4b:H2A-GFP)²⁴, Tg(she:H2B-EGFP)^psi59Tg25 referred in the paper as Tg(she:H2B-EGFP), Tg(myo6b:hist2h2l-mScarletI)^psi66Tg25 referred in the paper as Tg(myo6b:H2B-mScarlet-I), Tg(fgf3:H2B-mturquoise2)^psi60Tg9 referred in the paper as Tg(fgf3:H2B-mturquoise2), Et(krt:EGFP)^sqet20ET26 referred in the paper as ET20:EGFP, Tg(she:H2A-mCherry)^psi57Tg25 referred in the paper as Tg(she:H2A-mCherry), Tg(-8.0cldnb:lynEGFP)^zf106Tg32 referred in the paper as Tg(cldnb:lynEGFP), and Tg(-8.0cldnb:H2A-mCherry)^psi4Tg7 referred in the paper as Tg(cldnb:H2A-mCherry).

Transgenic line generation

For the generation of Tg(hsp70l:isl1a;myl7:EGFP)^psi98Tg, isl1a cDNA was amplified from total cDNA using Phusion High Fidelity Polymerase (New England Biolabs, E0553). The PCR product was gel extracted according to the manufacturer’s protocol (Monarch® DNA Gel Extraction Kit) and cloned into a pME vector digested with EcoRI using the following homology arms: 5’-GTATCGATAAGCTTGATATCGAATTCACC-3’ and 5’- GTGGATCCCCCGGGCTGCAGGAATT-3’ through Gibson assembly⁶⁶. This plasmid was then used in a standard LR Gateway reaction (Thermo Fisher Scientific, 11791020) with p5E-hsp70, p3E-polyA and pDestTol2CG2 plasmids from the tol2 transgenesis kit⁶⁷. The injection mix consisted of 70 ng/µL DNA, 200 ng/µL tol2 mRNA, 0.3 M KCl and 0.05% phenol red in dH₂O. 1–3 nL was injected into the cell of one-cell stage zebrafish embryos. Since pDestTol2CG2 contains the cmlc2:EGFP marker, GFP⁺ larvae were screened at 24 hpf and raised to adulthood. Founders were screened for germline transmission of GFP.

Generation of isl1a mutant

CRISPR/Cas12a was used to delete the entire isl1a gene. sgRNAs (5’-GCTAGCCCGTAGGCTACTTT-3’) and (5’-GCTAAACAGATGCTGCAAAT-3’) targeting approximately 300 base pairs upstream and 500 base pairs downstream of the gene respectively were designed using DeepCpf1⁶⁸. Lba-Cas12a sgRNA guides were ordered from IDT. The injection mix consisted of 12 µM sgRNA guide mix, 20 µM Lba-Cas12a protein (New England Biolabs), 0.3 M KCl and 0.05% phenol red in dH₂O. This mix was incubated at 37 °C for 10 min then kept on ice. 1–3 nL was injected into the cells of one-cell stage zebrafish embryos. Correct deletions were verified in several embryos at 48 hpf while the remaining embryos were raised to adulthood then screened for germline mutations. Forward (5’-GCGCAGTTTAAAACGGTCAT-3’) and reverse (5’-ATTTGCGCTTTTGGCATTAC-3’) primers were used to genotype the deletion allele. To genotype the wild-type allele a reverse (5’-GTTCGAGGTGTTGCAGGACT-3’) primer with the above forward primer were used.

Hair cell ablation and cell proliferation analysis

To induce hair cell death, 5 dpf zebrafish larvae were incubated in 300 µM neomycin (Sigma-Aldrich) in 0.5× E2 medium for 30 min. Larvae were washed with 0.5× E2 medium five times and allowed to recover in 0.5× E2 medium at 28.5 °C for specific timepoints depending on the experiment.

To label dividing cells, 5 dpf larvae were incubated in 3.3 mM EdU (Carbosynth) with 1% DMSO in 0.5× E2 medium. Larvae were incubated in EdU medium immediately after neomycin treatment for 24 h or from 24 h to 48 h, depending on the experiment, and fixed in 4% paraformaldehyde (PFA) overnight at 4 °C. EdU staining and GFP immunostaining were performed as previously described⁴⁴. Cell counts were performed manually in Imaris as previously described^7,69 and EdU indexes were calculated by dividing EdU+ cells by total neuromast cell number.

Heat shock paradigm

For heat shock induction of isl1a at 5 dpf, transgenic zebrafish (Tg(hsp70l:isl1a;myl7:EGFP)) were incubated at 37 °C for 1 h, followed by 1 h recovery at 28.5 °C, 1 h at 39 °C, 1 h at 28.5 °C, 1 h at 39 °C, and 1 h at 28.5 °C. Larvae were either fixed at this point or treated with neomycin as described above and incubated in EdU medium for 24 h. Larvae were continuously heat shocked at 37 °C for 1 h, followed by 2 h recovery at 28.5 °C, until the 24 h timepoint, then fixed in 4% PFA overnight at 4 °C. EdU staining and GFP immunostaining were performed as previously described⁴⁴.

In situ hybridization chain reaction (HCR-FISH)

In situ HCR v3.0 was performed according to the manufacturer’s instructions (Molecular Instruments)⁷⁰ with modifications as described previously⁷¹. Samples were incubated for 2 days in the amplifier mix before washing and staining with DAPI (1 µg/mL). The following probes were used: isl1a-B5, sost-B1, si:ch73-261i21.5-B4, junba-B1, dusp1-B2, lima1a-B2, cbx7a-B4, ctnnb1-B1, prdm1a-B1, nr3c1-B4, and bach2b-B5. Amplifiers were conjugated with either Alexa488, Alexa546, or Alexa647.

Immunohistochemistry

Zebrafish were fixed in 4% PFA overnight at 4 °C, then washed in PBS/0.8% Triton X-100 (PBSTx) 3 times for 10 min each at room temperature (RT). Samples were placed in acetone for 7 min at −20 °C, followed by a 15-min wash in dH₂O. Larvae were then blocked for 1 h in 10% normal goat serum (GEMINI BIO-PRODUCTS, #100-109) in PBSTx. Anti-acetylated tubulin (Sigma-Aldrich, T6793) was diluted to 1:1000 in blocking buffer and incubated overnight at 4 °C. Larvae were then washed in PBSTx at RT and incubated in secondary antibody, goat anti-mouse Alexa-488 (Thermo Fisher Scientific, A11001) 1:1000 in blocking buffer, overnight at 4 °C. The samples were washed in PBSTx and stained with DAPI (1 µg/mL).

Cell counts

Cell counts were performed manually in Fiji or Imaris as previously described^7,69. For 32 hpf cell counts, GFP+ cells were counted in the primordia and first deposited neuromasts of Tg(cxcr4b:H2A-GFP) embryos.

For cell counts from 2 to 7 dpf, Tg(she:H2B-EGFP) was used to count total neuromast cells and Tg(myo6b:H2B-mScarlet-I) for hair cells. For cell population analysis at 5 dpf the following lines were used for each cell population: mantle cells (ET20:EGFP), central support cells (Tg(fgf3:H2B-mturquoise2)), and hair cells (Tg(myo6b:H2B-mScarlet-I)). All neuromast cells negative for any of the markers above were classified as AP-DV cells. For cell population analysis at 48 h post-neomycin treatment a combination of transgenic lines and HCR-FISH labeling were used to label the cell types as follows: neuromast cells (Tg(she:H2B-EGFP)), central support cells (Tg(fgf3:H2B-mturquoise2)), hair cells (Tg(myo6b:H2B-mScarlet-I)), AP cells (si:ch73-261i21.5-B4 HCR-FISH), DV cells (sost-B1 HCR-FISH).

Transplantation experiment

Adult isl1a^psi97/+/Tg(cldnb:lynEGFP)/Tg(cldnb:H2A-mCherry) zebrafish were incrossed to generate embryos for cell transplants. Cells were transplanted into the presumptive placode region of TU wild-type embryos⁷² between 3 and 5 hpf. Host embryos were screened for lateral line labeling at 2 dpf, then raised to 5 dpf and treated with neomycin as mentioned above. After 48 h post-treatment, larvae were fixed in 4% PFA overnight at 4 °C, washed with PBS/0.05% Tween-20 and stained with DAPI (1 µg/mL) for imaging. Donor embryos were raised to 3 dpf for genotyping (based on morphology and by PCR). Cell counts were performed manually in Imaris as previously described^7,69. Neuromasts with >50% positive transplanted cells were considered for statistical analysis.

Confocal imaging

Images were acquired on a Zeiss LSM780 with Zeiss Plan-Apochromat 10×/0.45 M27 (air) and LD C-Apochromat 40×/1.1 W Korr M27 (water) objective lenses and a Nikon Ti2 Yokogawa CSU-W1 spinning disk head equipped with a Hamamatsu Orca Fusion sCMOS with CFI Apo LWD 40× WI 1.15 NA Lambda S and CFI Plan Apochromat Lambda D 10x objective lenses.

RNA-seq embryo dissociation and FACS

Adult isl1a^psi97/+/Tg(she:H2A-mCherry) zebrafish were incrossed to generate embryos for RNA-seq experiments. Experiments were performed in triplicate using 5 dpf isl1a sibling and isl1a^−/− larvae. mCherry+ fish were sorted at 3 dpf for the isl1a mutation (based on heart edema and lack of motility). Five hundred larvae were collected for each replicate and processed as follows, either in homeostasis or 3 h post-neomycin treatment. Larvae were anesthetized using MS-222 and cells dissociated on ice using 0.25% Trypsin-EDTA (Thermo Fisher Scientific, 25200056) as previously described⁴⁴ with a modified dissociation time of 3 min. To stain nucleated cells, the suspension was incubated with DRAQ5 (1:200 dilution, Thermo Fisher Scientific, 65-0880-96) and DAPI (2 µg/mL) for 5 min on ice. Cells were sorted at the Stowers Institute Cytometry Core Facility using the BD Influx Cell Sorter (BD Biosciences) instrument. Samples were first gated by forward and side scatter to remove debris, followed by singlet gating based on forward scatter area versus height. Nucleated cells were identified as DRAQ5+/DAPI− events. Finally, single nucleated DRAQ5+/DAPI− cells were gated for mCherry^high expression and sorted into 1 mL of TRIzol (Invitrogen, 15596026).

RNA-seq library preparation and sequencing

40,000 mCherry+ cells for each condition were used for total RNA extraction. Total RNA was isolated as previously described⁷³ with the following modifications: samples sorted into TRIzol were vortexed for 10 s and incubated for 5 min at RT. To precipitate the RNA 0.5 mL of isopropanol and 1 µL of GlycoBlue (Invitrogen, AM9515) were added to each sample. Samples were then mixed by quick vortex and incubated overnight at −20 °C. Pellet was resuspended in 30 µL of RNase-free H₂O, and the quality of RNA samples was measured with the Bioanalyzer Total RNA Pico assay (Agilent Technologies). SMART-Seq v4 Ultra Low Input RNA Kit for sequencing (Takara ClonTech, 634891) was used for cDNA synthesis and Nextera XT DNA Sample Preparation Kit (Illumina, FC-131-1096) for library construction. Libraries were purified using AMPure XP Beads (Beckman Coulter, A63882) and quantified using a Bioanalyzer High Sensitivity DNA assay (Agilent Technologies) and Qubit Fluorometer 3.0 (Life Technologies). Libraries were pooled and sequenced on an Illumina NextSeq 500 instrument in 2 × 75 bp paired-end mode.

RNA-seq data processing and analysis

Raw reads were demultiplexed into Fastq format, allowing up to one mismatch using Illumina bcl2fastq2 (v2.18). Reads were aligned to UCSC genome danRer11 with the STAR aligner (v2.7.3a), using Ensembl 102 gene models. TPM values were generated using RSEM (v1.3.0). Genes with fewer than 10 total counts across all samples were excluded prior to normalization. Samples were processed in R (v4.3.0) using the DeSeq2 package (v1.42.0) in both timepoints collected. Read counts were normalized, and differential gene expression was performed using DESeq2 default parameters. The analysis was performed comparing isl1a^−/− mutant samples to sibling samples at each time point. Genes with an adjusted p-value < 0.05 and an absolute log2 fold change (log2FC) > 0.5 were considered significant. Variance-stabilizing transformation (VST) was applied to the normalized counts for visualization. Heatmaps were generated in R using the pheatmap package, and Euler plots were generated in R using the eulerr package.

ATAC-seq embryo dissociation and FACS

Adult isl1a^psi97/+/Tg(she:H2A-mCherry) zebrafish were incrossed to generate embryos for ATAC-seq experiments. Experiments were performed in triplicate using 5 dpf isl1a sibling and isl1a^−/− larvae. mCherry+ fish were sorted at 3 dpf for the isl1a mutation (based on heart edema and lack of motility). Five hundred larvae were collected for each replicate and processed as follows either in homeostasis, 1 h or 3 h post-neomycin treatment. Larvae were anesthetized using MS-222 and cells dissociated on ice using 0.25% Trypsin-EDTA (Thermo Fisher Scientific, 25200056) as previously described⁴⁴ with a modified dissociation time of 3 min. To stain nucleated cells, the suspension was incubated with DRAQ5 (1:200 dilution, Thermo Fisher Scientific, 65-0880-96) and DAPI (2 µg/mL) for 5 min on ice. Cells were sorted at the Stowers Institute Cytometry Core Facility using the BD FACSMelody Cell sorter (BD Biosciences) instrument. Samples were first gated by forward and side scatter to remove debris, followed by singlet gating based on forward scatter area versus height. Nucleated cells were identified as DRAQ5+/DAPI− events. Finally, single-nucleated DRAQ5+/DAPI− cells were gated for mCherry^high expression and sorted into 1X DPBS.

ATAC-seq library preparation and sequencing

For all ATAC-seq experiments, 60,000 mCherry+ sorted cells were used. Following sorting, cells were incubated in 50 µL of lysis buffer (10% Tween-20, 10% IGEPAL, 1% Digitonin (Promega, G9441) and DPBS) and chilled on ice for 10 min. Next, 1 mL of DPBS was added, and nuclei were spun down for 10 min at 500 × g at 4 °C. After removing the supernatant, the nuclear pellets were resuspended in 50 µL of ice-cold ATAC RSB buffer (10 mM Tris-HCl, pH 7.4, 10 mM NaCl, 3 mM MgCl₂) containing 0.1% IGEPAL CA-630, 0.1% Tween-20, and 0.01% Digitonin. Samples were incubated on ice for exactly 3 min before washing with 1 mL of ATAC RSB buffer containing 0.1% Tween-20⁷⁴. The nuclei were then pelleted again by centrifuging for 10 min at 500 × g at 4 °C. The Tn5 transposition reaction was performed as described previously^75,76. Tagmentation was performed at 37 °C for 30 min at 1000 rpm in a 50 µL total reaction volume consisting of 10 µL of 5x Tagment DNA Buffer (50 mM Tris-HCl pH 7.4, 25 mM MgCl_2, DMF 50%), 0.5 µL 10% Tween-20, 0.5 µL 1% Digitonin, 16.5 µL 1x PBS, and 0.4 μL of TD Enzyme (Illumina, 15027916). The DNA fragments were purified using the Monarch PCR & DNA Cleanup Kit (NEB, T1030). Libraries were constructed using Illumina Nextera Dual Indexing, and qPCR was used to prevent over-amplification as described⁷⁶. Libraries were pooled and sequenced on an Illumina NextSeq 500 instrument in 2 × 75 bp paired-end mode.

ATAC-seq data processing

ATAC-seq data were processed using a previously described pipeline⁷⁴ with minor modifications, as detailed below. Paired-end reads were aligned to the Danio rerio genome assembly danRer11 using Bowtie2⁷⁷ (v2.3.5.1). Picard⁷⁸ (v2.23.8) was then used to mark aligned BAM files and remove PCR duplicates based on unique fragment coordinates. BAM files were reoriented to correct for the Tn5 enzymatic offset introduced during transposition by shifting the reads +4/−4 and were then filtered to remove any fragments greater than 600 bp. Coverage bigwig files were generated in R, and normalized coverage was concurrently generated using reads-per-million (RPM) normalization by scaling the coverage to the total number of reads in the sample divided by 1 million. ATAC-seq peaks were called using MACS2⁷⁹ (2.2.7.1) with default parameters. For all downstream analyses, only called peaks that were reproducible across replicates after applying the Irreproducible Discovery Rate filter⁸⁰ were used. ATAC-seq coverage tracks were displayed using pyGenomeTracks (v3.8).

Differential chromatin accessibility analysis

To determine the differentially accessible regions (DARs) between siblings and isl1a^−/− mutants at all time points, we performed DESeq2 analysis with default parameters and FDR = 0.05⁸¹. First, we calculated the total ATAC-seq signal in a 2000 bp window centered on each ATAC-seq peak for all replicates of each experimental condition. In this manner, we produced a DESeq2 model encompassing ATAC-seq counts from all experiments. We then used DESeq to conduct pairwise comparisons between sibling and isl1a^−/− mutant conditions at each time point and calculated the log2FC (isl1a^−/−/sibling) values. Peaks with a log2FC value > 0.5 and an adjusted p-value of <0.05 were classified as a gain in chromatin accessibility, while peaks with a log2FC value < −0.5 and an adjusted p-value of <0.05 were classified as a loss in chromatin accessibility. Aggregate ATAC-seq signal plots were generated in R using ggplot2 and cowplot packages, and heatmaps were generated using the pheatmap package.

ATAC-seq motif enrichment analysis

Using the list of DARs at 1 h post-neomycin, a FASTA file was generated for the top 200 peaks with increased accessibility in isl1a^−/−. The MEME suite tool AME⁸² (v5.3.0) was then used to search for enriched TF motifs, with a background sequence set consisting of a sample of 200 non-differential accessible peaks for the same timepoint. The JASPAR 2022 vertebrate core TF motif database⁸³ was used as a reference for AME. Motif enrichment bar plot was generated in R using the ggplot2 package.

To predict potential isl1a direct regulation, the lists of DARs for homeostasis and 1 h post-neomycin were cross-referenced to the differentially expressed gene list at 3 h post-neomycin. Only the regions that matched transcriptomic differentially expressed genes were kept. The resulting lists of DARs were subsequently scanned for Isl1 binding sites (JASPAR2024 Motif ID: MA1608.2, derived from mouse ChIP-seq data). DARs containing at least one Isl1 motif were considered putative Isl1 targets. Each motif-containing DAR was annotated for chromatin accessibility change, direction of associated gene expression changes and whether the associated gene encoded a transcription factor or cofactor⁸⁴. This classification allowed for systematic assessment of potential Isl1-dependent regulation.

Image analysis

Image processing and analysis were performed in ImageJ⁸⁵. For quantification of HCR-FISH fluorescence intensity in lateral line neuromasts, images were Z-projected (maximum intensity projection), and background was subtracted using the “Subtract Background” function with a rolling ball radius of 50 pixels. An ROI was drawn around the neuromast, and the fluorescence intensity was measured using the “Measure” function. The mean fluorescence intensity values were used for statistical analysis and displayed in GraphPad Prism 10 (v10.4.1).

Statistical analysis

All statistical tests and plots, unless specified, were done in GraphPad Prism 10 (v10.4.1). Normal distribution of data was assessed using the Shapiro–Wilk test. For pairwise comparisons between sibling and mutant groups, unpaired two-tailed t-tests were used, Welch’s correction was applied when variances were unequal, and two-tailed Mann–Whitney U tests when normality was not met. p-values were adjusted for multiple comparisons using the Holm-Šidák method. For multiple timepoint comparisons, data from sibling and mutant groups were analyzed using two-way ANOVA (genotype, timepoint) with Šidák’s multiple comparison test. For comparison across multiple groups with unequal variances, Brown-Forsythe and Welch ANOVA followed by Dunnett’s T3 multiple comparisons test was applied. A p-value < 0.05 was considered significant.

Author contributions

P.M.G. and T.P. conceived the project and designed experiments. M.E.L. designed and generated the isl1a mutant line. P.M.G. and K.J.B. performed experiments. P.M.G. analyzed experimental data and prepared figures. P.M.G., N.T.T.T., K.J.B., and A.H. performed computational and biostatistical analysis. P.M.G. and T.P. wrote the manuscript with input from all authors.

Data availability

Further information, resource and reagent requests should be directed to the Lead Contact, Tatjana Piotrowski (pio@stowers.org). Fish lines generated in this study are available from the lead contact. Sequencing data generated in this study have been deposited in the Gene Expression Omnibus (GEO) database under accession number: GSE314610 and are available at: https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE314610. Additional source data underlying this study can be accessed from the Stowers Institute for Medical Research Original Data Repository at: https://www.stowers.org/research/publications/libpb-2593.

Code availability

All analyses were performed using publicly available software and previously described pipelines. Software versions and appropriate citations are provided in the Methods section. No custom code was generated in this study.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41536-026-00489-4.