Introduction

Strabismus, or misalignment of the eyes, is a heritable disorder of vision affecting ∼2% of the population, and is the most common childhood disorder of vision ¹. It is frequently associated with amblyopia (vision loss resulting from inadequate visual experience during development ²), loss of binocular vision, and decreased quality of life in both children and adults ³. Incomitant strabismus, where the degree of misalignment differs based on gaze angle, can arise from mutations in genes that regulate the development of the extraocular motor system ^4,5. While some forms of strabismus are caused by loss-of-function of known transcription factors (e.g. Duane retraction syndrome (DRS) following monoallelic MAFB loss of function ⁶ and congenital fibrosis of the extraocular muscles (CFEOM) resulting from biallelic mutation of PHOX2A ⁷), causative genes for most forms of congenital strabismus remain unknown. Recent work has highlighted several novel candidate genes relevant to ocular congenital cranial disinnervation disorders (OCCDs) ⁸, but our understanding of disease etiology is limited by a lack of insight into molecular mechanisms of extraocular motor neuron development.

The six muscles that move the vertebrate eye are controlled by six corresponding pools of motor neurons distributed across three cranial nuclei: nIII (oculomotor, 4 pools), nIV (trochlear, 1 pool), and nVI (abducens, 1 pool). Axons from each pool follow stereotyped trajectories to innervate a single muscle, and each muscle pulls the eye in a well-defined direction ^9–11. nIII and nIV motor neurons are responsible for effecting compensatory eyes-up/eyes-down movements following vertical/torsional destabilization of the body or head. This highly conserved vestibulo-ocular reflex enables visual perception of the world to remain stable despite perturbations from self-motion. The vestibulo-ocular reflex circuit ^12,13, including nIII and nIV motor neurons ¹⁴, is organized topographically by birthdate and muscle target: dorsal pools innervating the extraocular muscles that pull the eyes downward become post-mitotic earliest, while later born ventral neurons innervate the muscles responsible for upward movements. Birthdate could therefore serve as a handle to segregate functional subpopulations of extraocular motor neurons.

The larval zebrafish, a small transparent model vertebrate, offers reliable genetic access to extraocular motor neurons ^15–17. Like all vertebrates, the zebrafish has a reliable and well-characterized vestibulo-ocular reflex ^18–22, with comparatively simple and well-understood anatomy ²³. Recent work in zebrafish has also demonstrated important connections between transcriptomic, anatomical, and functional characteristics of motor neuron subpopulations in the spinal cord ^24–26. We hypothesized that a similar characterization of extraocular motor neuron transcriptional diversity might define genes that regulate development and reveal novel disease candidates.

To investigate the molecular logic that ensures proper extraocular motor neuron development, we began by performing bulk and single-cell transcriptional sequencing of extraocular motor neurons. In-situ hybridization experiments revealed a set of genes whose spatially- and temporally-restricted expression patterns in nIII/nIV extraocular motor neurons suggest a role in subpopulation development. Next, we evaluated nIII/nIV-dependent eye movements after constitutive loss of four candidate genes: sim1a, phox2a, nav2a, and onecut1. Loss of sim1a profoundly disrupted the torsional vestibulo-ocular reflex without loss of motor neurons; both nIII/nIV motor neurons and vestibuloocular reflex behavior were absent in phox2a mutants. Finally, in-situ hybridization across the vestibulo-ocular reflex circuit confirmed that sim1a expression was confined to nIII/nIV motor neurons. Our work reveals that sim1a regulates proper functional development of extraocular motor neurons. Moreover, our experiments establish a pipeline to illuminate the molecular logic that regulates extraocular motor development, a major step towards identifying and understanding the genetic determinants of incomitant strabismus.

Results

Bulk sequencing identifies genes that label early and late-born subsets of extraocular motor neurons

Pools of motor neurons in nIII/nIV are organized topographically in both space ²⁷ and across developmental time ¹⁴. Motor neurons that control the medial and inferior rectus muscles (MR, IR) are located in dorsal nIII, while those that control the superior rectus and inferior oblique muscles (SR, IO) are located ventrally. Dorsal nIII motor neurons are born first, followed by those in nIV (which control the superior oblique muscle, SO), and then ventral nIII Figure S1. We therefore hypothesized that comparing bulk gene expression profiles between dorsal and ventral motor neurons would yield candidate marker genes.

To isolate transcripts from early (IR/MR) and late-born (SO/SR) motor neurons, we used an optical tagging approach (Figure 1A). The transgenic line Tg(isl1:Kaede) ¹⁶ expresses the photolabile protein Kaede under control of the isl1 promoter that labels motor neurons in three nIII pools (SR,MR,IR) and nIV (SO) ¹⁵. We briefly exposed larvae at 33 hours post-fertilization (hpf) to ultraviolet light, irreversibly converting the Kaede from green to red. At 33 hpf, the bulk of dorsal nIII neurons, but no ventral nIII neurons, have been born (Figure S1). We then dissected nIII/nIV from photoconverted larvae at 50 hpf, dissociated the neurons, and used fluorescence-activated cell sorting (FACS) to separate early- (red/green) and late-born (green only) neurons before RNA extraction and bulk sequencing. Almost all motor neurons are born by 50 hpf (Figure S1). We sequenced an average of 3,326 early-born and 1,577 late-born cells per replicate across four technical repeats (N=137±13 fish per repeat). Based on prior counts of ∼300 isl1+ motor neurons per fish in nIII/nIV, we estimate a post-FACS yield of ∼10%.

Figure 1:

Analysis of differentially-expressed genes between early and late-born motor neurons identified 695 candidate genes (411 higher-expressed in early-born neurons, 284 higher in late-born neurons) (Figure 1B). The 695 candidates included genes in functional categories implicated in neuronal differentiation and development such as cell adhesion and transcription factors (Figure 1C). Next, we used fluorescent in situ hybridization to evaluate the expression patterns of 25 differentially expressed genes at 50 hpf. Of the 13 genes predicted to be highly- expressed in early-born neurons, 10 were expressed exclusively in early-born neurons in dorsal nIII and the rest were expressed in both dorsal nIII and late-born nIV neurons (Figures 1D and S2). Across the candidates we evaluated, we did not see expression of genes enriched in early-born neurons in ventral nIII at 50 hpf. Of 12 genes predicted to be highly expressed in late-born neurons, 8 were expressed exclusively in late-born neurons in ventral nIII and/or nIV, 2 were expressed in both late-born and early-born neurons, and only 2 were expressed in early-born neurons (Figures 1E, 1F and S2).

We propose that differential evaluation of gene expression in early and late-born extraocular motor neurons in nIII/nIV can identify selectively-expressed transcripts that reflect temporal differences among extraocular motor neurons in nIII/nIV.

Single-cell sequencing identifies genes expressed in sub-populations of nIII/nIV motor neurons

Temporally-defined categories broadly segregate developing neurons in nIII/nIV, but each category includes distinct pools of motor neurons. For example, late-born neurons includes motor neurons that comprise both SR and SO pools. To further validate our bulk sequencing dataset, and to refine our transcriptional classification, we performed single-cell RNA sequencing of neurons in nIII/nIV. We extracted neurons from nIII/nIV in the Tg(isl1:GFP) line (Figure 2A) in 48 hpf larvae. We then used FACS to isolate and place GFP⁺ neurons for plate-based CEL-Seq2²⁸ transcriptomic analysis.

Figure 2:

Of 192 sequenced cells, 84 passed stringent criteria for further analysis (Methods). The presence of phox2 marks them as extraocular motor neurons in nIII/nIV (Figure 2B). Notably, extraocular motor neurons are born over an ∼28 hour period window from 22-50 hpf (Figure S1). Consequentially, some neurons in our dataset were newly-born (in ventral nIII), while others were over a day old (in dorsal nIII). This age-related gradient can be seen in the different expression levels of markers of functionally mature motor neurons such as chata and vachta (Figure 2B).

We used Uniform Manifold Approximation and Projection (UMAP) to visualize the continuum of motor neuron profiles; subsequent clustering suggests three groups with distinct RNA expression profiles (Figure 2C). The majority of the top 20 genes marking clusters 0 (12/20) and 2 (15/20) were genes we previously identified as differentially-expressed between early- and late-born categories, respectively (Figure 2D). Cluster 1 contained a small number (5/20) of genes previously identified from both early- and late-born categories, as well as many novel candidates. Qualitatively, expression levels of marker genes were consistent with either selective enrichment in a particular cluster (e.g. Cluster 0, sim1a, Figure 2E) or gradients of differential expression across all neurons (e.g. Cluster 2 pik3r3b, Figure 2E).

We repeated our fluorescent in situ validation on 26 of the top marker genes across the three clusters. RNA expression patterns largely matched predictions from sequencing data (Figure S2). Cluster 0 gene markers were expressed in motor neurons in dorsal nIII (Figures 2F and 2G). Cluster 1 gene markers were expressed in nIV, with three genes expressed only in nIV (inab, mafba, and nefma) and the rest expressed in nIV and dorsal nIII (Figures S2, 2F and 2H). Genes that were highly expressed in Cluster 2 (Figure 2E) were predominantly expressed in ventral nIII (Figures 2F and 2I).

Taken together, our single-cell dataset confirms our previous temporal categorization (Clusters 0 & 2 as early and late-born, respectively) and differentiates motor neurons in nIV (Cluster 1). The genes identified here are therefore candidates that might contribute to the development of selective components of the extraocular motor system.

Select candidate genes remain subpopulation-specific across early oculomotor development

Selective gene expression patterns might define bona fide subpopulation-specific markers. However, the motor neurons we sequenced are different ages, and the selectivity might reflect differential sampling at specific times during development. To review: At 33 hpf, early-born dorsal nIII neurons are 0–10 hours old. By 50 hpf, all nIII and nIV neurons have been born. The oldest dorsal nIII cells are 30 hours old and have begun axogenesis, and the youngest population of ventral nIII neurons are 0–10 hours old. At 80 hpf, the oldest cells are nearing 60 hours old and the youngest motor neurons are 30 hours old; axogenesis is complete and synaptogenesis on to specific muscle targets has begun ²⁹.

To determine if the subpopulation selectivity we observed persisted across early development we performed in situ hybridization experiments on select candidate genes at 33, 50, and 80 hpf. Both sim1a and nrp1a had persistent expression in dorsal nIII motor neurons (Figure 3A). Global expression levels for nrp1a varied across developmental age with particularly strong expression at 50 hpf; at this stage, expression was visible in both dorsal nIII and nIV. Expression of nav2a and tacc2 persisted in ventral nIII (Figure 3B). Finally, in contrast to nIII markers, expression of inab and mafba were only strongly expressed in nIV at 50 hpf.

Figure 3:

Taken together, evaluating expression at multiple timepoints suggests that our sequencing dataset contains both genes whose activity persists in specific subpopulations (e.g. sim1a) and genes with selective but transient expression (e.g. inab/mafba). We propose that transcriptional heterogeneity among motor neurons reflects true differences among subpopulations, rather than just differences in age.

Subpopulation-specific gene mutations cause vertical eye rotation deficits

We next combined a genetic loss-of-function approach with an eye movement assay to evaluate what role — if any — three candidate genes play in ocular motor function. The transcription factor sim1a was upregulated in dorsal nIII neurons in both bulk (Figure 1D) and single-cell sequencing (Figures 2E and 2H) approaches, and remained so through early development (Figure 3A). nav2a was expressed in ventral nIII neurons (Figure 2F) and also remains subpopulation-specific (Figure 3B). The transcription factor onecut1 was expressed in nIV and dorsal nIII neurons (S2B) and has been implicated in spinal motor neuron diversity ³⁰. We generated loss-of-function alleles of sim1a, nav2a, and onecut1 using CRISPR/Cas9 mutagenesis (Figure S3). Both nav2a and onecut1 homozygous mutants were morphologically normal, but nav2a mutants did not survive past 7 dpf. sim1a homozygous mutants did not inflate their swim bladder and also did not survive past 7 dpf. Finally, we evaluated a previously validated phox2a loss-of-function allele that results in near-total loss of isl1+ neurons in both nIII and nIV ²⁰.

To quantify nIII/nIV-derived ocular motor function we measured the torsional vestibulo-ocular reflex following pitch (nose-up/nose-down) tilts at 6–7 dpf using a previously validated assay/apparatus ²¹. Briefly, fish were immobilized with one eye freed, placed on a rotating platform, and tilted in darkness (Figure 4A). The angle of the platform and the eye’s rotation were tracked throughout time (Figures 4B and 4C); behavioral performance is defined as the gain, or the ratio of the peak eye velocity to the peak platform velocity (35 °/sec). This vestibulo-ocular reflex can be reliably elicited after 4 dpf, but as it does not mature until 9 dpf, we expect the gain to be below 1^18,21.

Figure 4:

As expected, phox2a homozygotes did not move their eyes in response to either nose-up or nose-down pitch tilts (Figure 4D). Similarly, the vestibulo-ocular reflex gain was markedly reduced in sim1a homozygotes (Figure 4E). In contrast, gain was indistinguishable between wild-type / heterozygous siblings and larvae with mutations in either nav2a or onecut1 (Figures 4F and 4G). We propose that sim1a is indispensable for a proper nIII/nIV-dependent vestibulo-ocular reflex.

The vestibulo-ocular reflex circuit consists of sensory afferents in the stato-acoustic ganglion, central projection neurons in the tangential nucleus and motor neurons in the oculomotor and trochlear nuclei (Figure 5A). Defects in any of these areas could compromise the vestibulo-ocular reflex. phox2a expression was only observed in nIII/nIV (Figure 5B), and phox2a mutants have few or no isl1+ oculomotor/trochlear cells (Figure 5C). In contrast, we observed no differences in the number of isl1+ motor neurons in nIII and nIV in any of our novel mutant alleles (Figures 5D to 5F). Finally, sim1a was not expressed in either the stato-acoustic ganglion or the tangential nucleus at 33, 50, and 80 hpf (Figure 5G). Our data suggests that sim1a does not specify, but instead regulates proper functional development of nIII/nIV extraocular motor neurons.

Figure 5:

Discussion

We measured transcriptional diversity between subpopulations of extraocular motor neurons and discovered a novel role for sim1a in ocular motor development. Validated bulk sequencing identified markers of early- and late-born subsets of extraocular motor neurons in the oculomotor and trochlear nuclei. Next, single-cell sequencing and fluorescent in-situ hybridization experiments revealed differentially-expressed genes marking specific subpopulations that remain subpopulation-specific through early development. Finally, we used the vestibulo-ocular reflex to evaluate loss-of-function phenotypes for four candidate genes. Of these, sim1a, a novel candidate expressed in early-born dorsal nIII neurons, is necessary for proper tilt-evoked eye movements. We propose that our pipeline from transcriptomics to behavior can discover and define roles for candidate genes responsible for extraocular motor neuron development. This work is therefore a key step towards understanding the etiology of congenital incomitant strabismus.

Loss of sim1a disrupts the vestibulo-ocular reflex

SIM1, after Drosophila single minded (sim), is a basic helix-loop-helix transcription factor ³¹ that is important for neural development in flies ³², mice ³³, and humans ³⁴. In zebrafish, sim1a has also been implicated in the development of the kidney and associated cells ³⁵, acting downstream of retinoic acid ³⁶ and Notch/Tbx2a ³⁷ signaling pathways to specify cell fate. Similarly, sim1a specifies diencephalic dopaminergic neuron ³⁸, hypothalamic neuron ^39–41, and V3 spinal interneuron fates, the latter depending on exposure to Notch ⁴². In the hypothalamus, sim1a interactions with its binding partner arnt2 regulate guidance of hypothalamo-spinal axons by interfering with Robo2-mediated repulsion ⁴³. Prior work thus establishes sim1a as an excellent candidate to specify fate and/or regulate axonal projections of extraocular motor neurons in nIII.

What role might sim1a play in the development of nIII motor neurons? We did not observe any differences in the number of nIII isl1:GFP⁺ motor neurons in sim1a mutants, and we see partial preservation of both up/down eye movements. Together, these findings suggest aberrant axonal guidance or functional disruptions rather than a fate change after loss of sim1a. Intriguingly, in mice, loss of Sim1 disrupts the dorsal clustering and electrophysiological diversification of early-born, but not late-born, V3 spinal interneurons ⁴⁴. Future experiments in sim1a mutants characterizing nIII nerve branching/muscle innervation ²⁹ and functional responses of dorsal nIII motor neurons to vestibular stimuli ⁴⁵ will speak to the anatomical and functional contributions of sim1a.

sim1a expression is restricted to dorsal nIII, which contains inferior rectus motor neurons that move the eyes down after nose-up pitch tilts. However, sim1a mutants show markedly reduced vestibulo-ocular reflex gains in both directions on the pitch axis. While several cases of deletions on the long arm of chromosome 6, which contains the human SIM1, present with strabismus ^46–48, (though see ⁴⁹) comprehensive characterization of the strabismus from these patients is not available. The absence of sim1a expression elsewhere in the vestibulo-ocular reflex circuit argues against an upstream deficit in sensory/central processing of tilts. Notably, humans show similarly “paradoxical” vestibulo-ocular reflex impairments in cases of fourth nerve palsy in adulthood, thought to reflect adaptive plasticity ⁵⁰. Alternatively, impaired basal tension of extraocular muscles would shift the eyes’ resting orientation, disrupting both up and down rotations. Future experiments could measure visually-evoked vertical eye movements ²² and/or medial rectus-dependent horizontal eye movements ⁵¹ in sim1a mutants to better define the nature of the behavioral deficits.

Mutations of nav2a and onecut1 do not impair the vestibulo-ocular reflex

nav2a is orthologous to unc-53 in C. elegans, where it is essential for axonal and cell migration ⁵², and sickie in D. melanogaster, where loss-of-function is homozygous lethal ⁵³. Mouse models of Nav2 loss-of-function show cerebellar malformations, similar to a patient with a NAV2 deletion who also exhibits deficits in voluntary eye movements ⁵³. While the vestibulo-ocular reflex was normal in our nav2a mutants, their premature lethality is consistent with an important role for nav2a in development.

onecut1 frameshift mutants had a normal vestibulo-ocular reflex and no obvious morphological defects. The onecut family is a highly conserved family of transcription factors whose members are implicated in development of various nervous system cell types ^54,55, including upstream of isl1 as regulators of spinal motor neuron diversity ³⁰. Redundancy of these similar transcription factors ^56,57 might compensate for the loss of onecut1. Compensation for mutations to onecut1 may also occur through the process of transcriptional adaptation, in which degradation of mutant mRNA sequences triggers upregulation of similar genes ⁵⁸. Mutagenesis targeting combinations of onecut factors (such as the zebrafish onecut2, onecut3a, onecut3b, and/or onecutl), or that creates an allele that fails to transcribe the mutant gene, thus eliminating the potential for degradation of mutant transcripts, could clarify onecut factors’ contributions to extraocular motor neuron development. Finally, though we did not observe any cellular, anatomical, or behavioral phenotypes through 7 dpf in homozygous mutant onecut1 larvae, we cannot rule out the emergence of deficits later in development.

Genes directly or indirectly implicated in oculomotor disorder

In addition to the genes selected for mutagenesis and subsequent experiments, our sequencing highlights several extraocular motor neuron subpopulation-specific genes that have been directly, or indirectly, implicated in oculomotor disorder. Duane Retraction Syndrome (DRS), an oculomotor disorder that presents with maldevelopment of the abducens nerve (nVI) and restricted horizontal eye movements, is caused by loss of MAFB function ^6,8. This is consistent with recent analysis of gene expression in the zebrafish hindbrain showing transient mafba expression in hindbrain rhombomeres 5/6⁵⁹, where the extraocular motor neurons that innervate the lateral rectus muscle aggregate in the abducens nucleus (nVI). In Mafb knockout mice, not only is nVI absent, but nIII shows aberrant branching as well ⁶. Interestingly, we observed mafba expression in trochlear (nIV) cells. This mafba expression was restricted to nIV, but only present at 50 hpf, at which point both dorsal nIII and nIV cells are in the process of extending their axons towards target muscles ²⁹. Our transcriptomic and expression data expands the view of MAFB homologs in regulating extraocular muscle targeting.

Mutations in the gene SEMA3F, which encodes a member of the semaphorin signaling protein family, are associated with nIII hypoplasia ⁸. nIII axons defasciculate and fail to reach muscle partners in sema3f mutant zebrafish ⁸. Neuropilin 2 (Nrp2) is a receptor for Sema3f ⁶⁰. Consistent with a role for Sema3f/Nrp2a signaling in oculomotor axon guidance, our in situ hybridization shows nrp2a RNA in early-born dorsal nIII cells. However, in bulk sequencing data, nrp2a was more strongly expressed in the late-born population . One possibility is that the expression is due to off-target binding of nrp2a in situ hybridization probes: the in situ expression pattern is similar to that of nrp1a, which was expressed most strongly in dorsal nIII at 50 hpf. While Neuropilin 1 (Nrp1) is a receptor for Sema3a ⁶¹, our data implicates class 3 semaphorins broadly as mediators of oculomotor axon guidance. Regulation of Nrp1 in zebrafish caudal primary (CaP) motor neurons contributes to CaP axon guidance in the spinal cord ⁶². Future work that similarly modulates levels of nrp1a expression in nIII cells would speak to the role of these guidance cues.

Trisomy 10 and duplications of the long arm of chromosome 10, which contains the intermediate filament encoding gene INA, have been linked to strabismus ^63,64. Our in situ hybridization showed expression of the homolog inab specifically in nIV at 50, but not 33 or 80, hpf. By 50 hpf, nIV axons have crossed the midline and begun to grow ventrally towards the eye ²⁹. Notably, inab is required for proper axon branching of subsets of zebrafish spinal motor neurons ⁶⁵. Future investigations of extraocular motor neuron axonal morphology ²⁹ could determine whether inab plays a similar role in extraocular motor neurons.

Limitations

The major limitations of this study are common to all transcriptomic and loss-of-function work. While we used bulk sequencing, plate-based single-cell sequencing, and in situ validation, our datasets are unlikely to be comprehensive with respect to lowly-expressed genes. While it is beyond the scope of this paper, recent zebrafish datasets ^66–68 include nIII/nIV; together with our data, these might provide a more complete picture of extraocular motor neuron transcriptomes. We only observed impaired vestibulo-ocular reflex behavior after loss of sim1a; as discussed above, we can only speculate with respect to why nav2a and onecut1 did not show similar impairment.

More specifically, our reagents and choice of behavior limits what we can observe. The Tg(isl1:GFP) line we used to isolate extraocular motor neurons does not label inferior oblique or nVI (abducens) motor neurons ¹⁵. Future work with broader transgenic lines ⁶⁹ would allow a more comprehensive characterization. Finally, our vestibulo-ocular reflex assay does not probe horizontal eye movements produced by the medial (nIII) or lateral rectus (nVI) motor neurons, and is limited with respect to stimulus kinetics. As discussed above, visual stimulation ²² and characterization of horizontal eye movements ⁵¹ could more comprehensively probe potential behavioral deficits.

Conclusion

This study establishes a pipeline to discover the genes responsible for normal development of vertebrate extraocular motor neurons. Our assay reveals considerable diversity among developing motor neuron pools. Further, a loss-of-function approach coupled with a vestibuloocular reflex assay implicates sim1a in the development of extraocular motor neurons. Broadly, our work establishes a powerful approach to yield novel insights into the molecular mechanisms that govern development of the extraocular motor system in health and disease.

Materials and methods

Fish Care

All procedures involving zebrafish larvae (Danio rerio) were approved by the Institutional Animal Care and Use Committee of New York University Langone Health. Fertilized eggs were collected and maintained at 28.5° C on a standard 14/10 hour light/dark cycle. Before 5 dpf, larvae were maintained at densities of 20–50 larvae per petri dish of 10 cm diameter, filled with 25–40 mL E3 with 0.5 ppm methylene blue. After 5 dpf, larvae were maintained at densities under 20 larvae per petri dish and were fed cultured rotifers (Reed Mariculture) daily. Zebrafish larvae at the ages we studied have not yet differentiated their sex, and so sex was not considered as a behavioral variable.

Transgenic Lines

The Tg(isl1:Kaede) line ¹⁶ was used for bulk sequencing experiments and in situ hybridization. The Tg(isl1:GFP) line ¹⁵ was used for single-cell sequencing, in situ hybridization, and motor neuron counts. Triple transgenic lines were used to identify the statoacoustic ganglion and tangential nucleus that carried the following alleles: Tg(-17.6isl2b:GFP) ⁷⁰;Tg(6.7Tru.Hcrtr2:GAL4-VP16) ^19,71; Tg(UAS:E1b-Kaede) ⁷². Motor neuron imaging experiments and in situ hybridization experiments were done on the mitfa^-/- background to remove pigment.

Ocular Motor Neuron Birthdating

Kaede-expressing embryos were photoconverted as described in ¹³. Briefly, at 18, 22, 26, 30, 34, 42, 46, 50, or 54 hpf the entire fish was exposed to 405 nm LED light for 5 mins, and then raised in a dark incubator to prevent background photoconversion of Kaede fluorophore generated after the photoconversion timepoint. Photoconverted fish were then mounted dorsally and imaged at on a Zeiss LSM800 confocal microscope with a water-immersion objective (Zeiss W Plan-Apochromat 20x/1.0). To account for background conversion of Kaede due to incident light exposure during the experiment, non-converted controls were also imaged and were used to set a threshold level of red fluorescence. For each photoconversion timepoint, image stacks were manually evaluated in FIJI ⁷³ to determine cell location and whether each cell was Kaede-red positive. Cells that were Kaede-red positive were considered born prior to that time-point.

Ocular Motor Neuron Dissection, Dissociation, and Flow Cytometry

Tg(isl1:Kaede) larvae were photoconverted as above at 33 hpf by exposing the entire fish to a 405 nm LED light for 5 minutes. To prevent background photoconversion, larvae were raised in a dark incubator and were kept in the dark during motor neuron dissections. Ocular motor neurons were harvested from 50 hpf photoconverted larvae. Three experimenters (D.G., K.R.H., and P.L.) harvested neurons in parallel. Larvae were anesthetized in MESAB in Earle’s Balanced Salt Solution with calcium, magnesium, and phenol red (EBSS, Thermo Fisher Scientific 24010043) and larvae were positioned dorsal-up in an 3% agarose-molded petri dish with triangle wells. Fluorescence in motor neurons was visualized using a SugarCube LED Illuminator (Ushio America, Cypress CA) using 10x eyepieces on a stereomicroscope (Leica Microsystems, Wetzlar, Germany). Cells were harvested using a thin wall glass capillary tube (4 inch, OD 1.0MM, World Precision Instruments) into EBSS in a non-stick Eppendorf tube and kept on ice until dissociation. Cells were dissociated in 20 units/mL of papain prepared in EBSS (Worthington Biochemical), 2000 units/mL of deoxyribonuclease prepared in EBSS (Worthington Biochemical), and 100 mg/mL of Type 1A Collagenase (Sigma Aldrich) prepared in Hanks Buffered Salt Solution without calcium/magnesium (HBSS, Thermo Fisher Scientific). Cells were incubated for 45 minutes at 31.5° C with a gentle vortex every 10–15 min, then passed through a 20 µm filter and centrifuged for 10 mins at 300 x g. After removing supernatant, neurons were resuspended in L15 (Thermo Fisher Scientific) with 2% fetal bovine serum (Thermo Fisher Scientific). Cell health was evaluated using DAPI, applied at 0.5 µg/ml (Invitrogen) and incubated on ice for 30-45 mins prior to flow cytometry. Flow cytometry was performed using a Sony SH800z cell sorter (100 ţm nozzle, 20 psi) to isolate single neurons. Four controls were run for setting FACS gates: (1) non-fluorescent cells, (2) non-fluorescent cells + DAPI, (3) green fluorescent cells from unconverted Tg(isl1:Kaede) + DAPI and (4) red fluorescent cells from photoconverted Tg(isl1:Kaede) + DAPI. On average, less than 0.5% of cells were DAPI-positive and excluded. Experimental cells were gated for green fluorescence (late-born) or red + green fluorescence (early-born). Cells were sorted into an Eppendorf tube containing 700 µl of lysis buffer (RNAqueous Micro Total RNA Isolation Kit, Thermo Fisher Scientific) for downstream bulk RNA sequencing.

Bulk RNA Sequencing and Analysis

RNA isolation was performed using an RNAqueous Micro Total RNA Isolation Kit (Thermo Fisher Scientific). RNA concentration and quality was evaluated using an RNA 6000 Pico Kit and a 2100 Bioanalyzer system (Agilent Technologies, Santa Clara, California). The majority of our samples had low RNA concentrations and accordingly did not have RIN scores, but for larger concentration samples RIN scores were routinely above 8 indicating that the extraction process did not affect RNA integrity. RNA sequencing was performed by NYU Langones Genome Technology Center (RRID:SCR_017929). Libraries were prepared using the low-input Clontech SMART-Seq HT with Nxt HTkit (Takara Bio USA) and sequenced using an Illumina NovaSeq 6000 with an S1 100Cycle Flow Cell (v1.5). Reads were mapped to GRCz11, and differential gene expression analysis was performed using the DESeq2 pipeline. Candidate genes were considered significantly differentially expressed if they had a fold-change greater than 2 and an adjusted p-value less than p=0.01.

Gene ontology (GO) analysis was performed in PANTHER19.0 (https://pantherdb.org/), testing for over-represented GO terms for molecular function among the 695 significantly differentially expressed genes ^74,75. Statistical significance for GO terms was determined using a binomial test, with Bonferroni correction for multiple tests.

Single-cell RNA Sequencing and Analysis

We dissected and dissociated nIII/nIV as above from 48 hpf Tg(is1:GFP) fish. We used a Sony SH800Z to FACS green cells and dispense them into plates for library preparation using a CEL-Seq2²⁸ approach followed by sequencing performed by the NYU Genome Technology Center. After mapping reads to GRCz10, Seurat v4⁷⁶ was used to exclude cells (<1,100 genes, >15,000 genes, >4% mitochondrial counts), normalize, and identify variable features. The top 2000 variable genes were used for dimensionality reduction by PCA; the top 21 principal components were retained for downstream analyses. Two 96-well plates yielded 84 cells that met our criteria, with 17,440±6,559 unique molecular identifiers per cell across 4,412±1,051 unique genes/cell.

Fluorescent In Situ Hybridization and Imaging

Experiments were performed using Hybridization Chain Reaction (HCR) for whole-mount zebrafish larvae ⁷⁷. Probes were generated using the HCR 3.0 probe maker using the sense sequence of the canonical gene cDNA from NCBI ⁷⁸. Larvae were from the Tg(isl1:Kaede) background, photoconverted as above at 33 hpf. Pools of 6–8 larvae were fixed in 1.5 mL microcentrifuge tubes overnight with 4% PFA in 1x Phosphate-Buffered Saline (PBS) at 4°C and stored in 100% methanol at -20° C. Subsequently, HCR was performed as described in ⁷⁹, with adjustments to proteinase K incubation time based on age (33 hpf: 17 minute incubation, 50 hpf: 21 minutes; 80 hpf: 33 minutes). HCR experiments used buffers and amplifiers from Molecular Instruments (Los Angeles, CA). Samples were stored in 1x PBS at 4° C until imaging. Larvae were mounted dorsally in 2% low-melting temperature agarose (Invitrogen 16520-050) in 1x PBS and imaged on a Zeiss LSM800 confocal microscope with 20x water-dipping objective. Laser power was kept consistent for all fluorophores across all imaged fish. White levels in representative images have been been adjusted for optimal viewing of fluorescence signal; all images from the same channel within the same figure have been adjusted to have the same minimum and maximum pixel intensity to facilitate comparison within figures.

CRISPR-Cas9 Mutagenesis

Gene-specific CRISPR target sequences were selected using CRISPRscan ⁸⁰, and were synthesized as DNA oligos with a 5’ T7 promoter (TAATACGACTCACTATA) and a 3’ overlap region (GTTTTAGAGCTAGAA). The DNA target sequence for nav2a was AGCGCCGCCCCGGTGGCCAA, for sim1a was GGAGGGCAGAGGCAGCAGTT, and for onecut1 was GTGGCCCCGGGGTGCGCGTACGG. Single guide RNA (sgRNA) was generated as described in ⁸¹, with the following modifications: in vitro transcription was performed using the AmpliScribe T7 Flash Transcription Kit (LGC Biosearch Technologies) according to the manufacturer protocol and allowing the transcription reaction to incubate at 37° for 4 hours. sgRNA was precipitated using sodium acetate as described in ⁸². Crispants were generated by injecting zebrafish embryos at the 1-cell stage with 1 nL of injection mixture containing a final concentration of 4 uM EnGen Spy Cas9 NLS protein (NEB M0646), 100 ng/µl sgRNA, and 10% phenol red in a Cas9 protein buffer solution (10 mM MgCl₂, 200 mM KCl, 20 mM Tris Buffer pH 8.0, ⁸¹). Injection mixture was incubated at 37°C for 5 minutes prior to injections to promote formation of the Cas9/sgRNA complex.

Injected fish were then raised and founders (F0) identified by out-crossing and sequencing pools of 10–20 embryos using Sanger sequencing of an approximately 500 base pair region around the target sequence. One allele of each gene was identified and used for experiments: sim1a^d17 has a 17 bp deletion from base pairs 693 to 709. nav2a^d8 has an 8 bp deletion from base pairs 151590 to 151597. onecut1^d14 has a 14 bp deletion from base pairs 826 to 839. Each mutation was a frameshift mutation, causing a predicted premature stop codon. Studies using phox2a, nav2a, and onecut1 stable line mutants were performed on larvae from in-crosses of F2 or later generations. For sim1a, studies were performed on larvae from in-crosses of the F1 generation.

Eye Rotation Behavioral Assay

Body tilts were delivered and eye rotations measured and analyzed as per ²¹. Briefly, 6–7 dpf larvae were mounted in 2% low-melting temperature agarose on a piece of Sylgard 184 (Dow Corning), the left eye was freed, and the Sylgard was placed in a 10 mm optical glass cuvette filled with 1 mL of E3. The cuvette was then mounted to a rotating platform. The eye was imaged with a 5x objective (Olympus MPLN, 0.1 NA) onto a machine vision camera (Guppy Pro 2 F-031, Allied Vision Technologies) which acquired a 100x100 pixel image of the left eye of the fish (6 µm/pixel) at 200 Hz. Images were processed on-line to derive an estimate of torsional angle (LabView 2014, National Instruments), and data was analyzed using custom MATLAB scripts (Mathworks, Natick MA).

Each experiment consisted of 50 cycles of four steps. Steps were ± 15° towards and away from the horizon, with a trapezoidal velocity profile that peaked at 35°/sec, peak acceleration 150°/sec². Eye and platform data was processed as per ²¹; each step was evaluated manually to exclude trials with rapid deviations in eye position indicative of horizontal saccades or gross failure of the pattern-matching algorithm. The response to each step for a given fish was defined as the peak eye velocity occurring over the first second after the start of the step, averaged across all cycles.

Ocular motor neuron imaging and counts

Larvae between 5–7 dpf were mounted dorsally in 2% low-melting temperature agarose in E3 and imaged on a Zeiss LSM800 confocal microscope with a water-immersion objective (Zeiss W Plan-Apochromat 20x/1.0). Stacks spanned ∼90 µM, sampled every 1.5 µM. Analysis was performed in Fiji/ImageJ using the Cell Counter plugin. Stacks of nIII/nIV were subdivided in the dorsoventral axis as described in ¹⁴ to facilitate localization. A point ROI was dropped over each neuron in the plane in which the soma was brightest, and the number of neurons belonging to nIV, dorsal nIII, and ventral nIII in each plane was recorded.

Data & Code

All data, raw and analyzed, as well as code necessary to generate the figures will be available at 10.17605/OSF.IO/JR4GS

Data availability

All data, raw and analyzed, as well as code necessary to generate the figures will be available at 10.17605/OSF.IO/JR4G

Supplementary figures

Figure S1:

Figure S2:

Figure S3:

Acknowledgements

Research was supported by the National Eye Institute under award number R01EY035691, the National Institute on Deafness and Communication Disorders of the National Institutes of Health under award numbers R01DC017489, F31DC019554, and F31DC020910, and the National Institute of Neurological Disorders and Stroke under award numbers F99NS129179, T32NS086750, and the National Cancer Institute P30CA016087. The authors would like to thank Elizabeth Engle for confirming the absence of known SIM1 human variants, Itai Yanai and Felicia Kuperwaser for help with CEL-Seq2, the NYU Langone Health Genome Technology Center for sequencing, alignment, and analysis, and members of the Schoppik and Nagel Labs for their helpful feedback.

Additional information

Author contributions

Conceptualization: DS, KRH; Methodology: KRH, DG, BR; Investigation: KRH, BR, HG, GX, CQ, DG, PL, EG; Visualization: KRH, EG; Writing: EG, KRH; Editing: DS, EG; Funding Acquisition: KRH, DS; Supervision: KRH, DS.

Funding

HHS | NIH | National Eye Institute (NEI) (R01EY035691)

• David Schoppik

HHS | NIH | National Institute on Deafness and Other Communication Disorders (NIDCD) (R01DC017489)

• David Schoppik

HHS | NIH | National Institute on Deafness and Other Communication Disorders (NIDCD) (F31DC019554)

• Kyla Rose Hamling

HHS | NIH | National Institute on Deafness and Other Communication Disorders (NIDCD) (F31DC020910)

• Paige Leary

HHS | NIH | National Institute of Neurological Disorders and Stroke (NINDS) (F99NS129179)

• Dena Goldblatt

HHS | NIH | National Institute of Neurological Disorders and Stroke (NINDS) (T32NS086750)

• Kyla Rose Hamling