Sensory thresholds enable animals to regulate their behavioral responses to environmental threats. Despite the importance of sensory thresholds for animal behavior and human health, we do not yet have a full appreciation of the underlying molecular-genetic and circuit mechanisms. The larval zebrafish acoustic startle response provides a powerful system to identify molecular mechanisms underlying establishment of sensory thresholds and plasticity of thresholds through mechanisms like habituation. Using this system, we identify Cadherin-16 as a previously undescribed regulator of sensory gating. We demonstrate that Cadherin-16 regulates sensory thresholds via an endocrine organ, the corpuscle of Stannius (CS), which is essential in zebrafish for regulating Ca²⁺ homeostasis. We further show that Cadherin-16 regulates whole-body calcium and ultimately behavior through the hormone Stanniocalcin 1l (Stc1l), and the IGF-regulatory metalloprotease, Papp-aa. Finally, we demonstrate the importance of the CS through ablation experiments that reveal its role in promoting normal acoustic sensory gating. Together, our results uncover a previously undescribed brain non-autonomous pathway for the regulation of behavior and underscore Ca²⁺ homeostasis as a critical process underlying sensory gating in vivo.

Cadherin-16 regulates acoustic startle response thresholds and habituation learning

At 5 days post-fertilization (dpf), larval zebrafish respond to high-intensity acoustic stimuli via a short-latency acoustic startle response, or short-latency C-bend (SLC) [2,29]. At this stage, zebrafish are also capable of distinguishing between high-intensity stimuli and low-intensity stimuli. They respond to high-intensity stimuli with fast responses and are able to ignore lower intensity stimuli or respond with long-latency C-bends [3,11,29]. Sensorimotor gating mechanisms, including the developmental establishment of acoustic startle response thresholds, enable animals to make distinctions between threatening and non-threatening acoustic stimuli [3]. Moreover, thresholds established during development can be transiently modified in 5 dpf larvae through plasticity mechanisms like habituation [7,8,11].

Through a forward genetic screen, a large collection of molecular regulators of sensorimotor gating were identified, including genes regulating (1) initial establishment of acoustic startle response thresholds, (2) plasticity of thresholds through habituation, and (3) the decision to perform a short-latency versus a long-latency C-bend [3,11–15]. irresistible^p173 mutants were identified based on their inability to modulate response frequency through habituation and their hypersensitivity to low-level acoustic stimuli [11]. Specifically, a “sub-threshold” stimulus designed to elicit SLC responses only 5%–20% of the time elicited SLC responses more than 30% of the time in irresistible mutants. To understand further how acoustic startle thresholds are affected, we exposed irresistible mutants to a series of acoustic stimuli, ranging in intensity from 0.54g to 51.1g as previously described (see “Materials and methods”) [3,4]. irresistible mutants exhibit an increased sensitivity to acoustic stimuli, responding at higher rates than their siblings across multiple stimulus intensities, indicating deficits in sensory gating (Fig 1A). Next, we measured habituation by presenting animals with 40 high-intensity (51.1g) stimuli, each separated by a 3-s inter-stimulus interval (ISI). We found that irresistible mutants continue to respond at a high rate throughout the habituation assay, indicating deficits in the ability to dynamically tune acoustic startle response thresholds (Fig 1B and 1C).

Despite the dramatic impacts on their ability to threshold acoustic stimuli, irresistible mutants are adult-viable and fertile. To examine whether irresistible specifically regulates acoustic sensory gating, or has broader effects, we tested visual startle response rates (O-bend responses to whole-field loss of illumination or dark flash) [30] (Fig 1D), habituation to dark flash stimuli [7,31,32] (Fig 1E), light flash responses [30] (Fig 1F), visuomotor responses [33] (Fig 1G–1I), and ability to respond to thermal stimuli [34] (Fig 1J). We found no significant differences between irresistible mutants and their siblings, consistent with a specific deficit in developmental and acute regulation of acoustic thresholds in mutant larvae.

To map the genetic locus responsible for the irresistible phenotype, we conducted whole-genome sequencing followed by homozygosity mapping as previously described [11]. This uncovered a premature stop codon (Y657*) in the cdh16 gene, encoding the calcium-dependent cell-adhesion protein, Cadherin-16. Like other members of the 7D family of cadherins, Cadherin-16 has 7 extracellular cadherin domains, a transmembrane domain (TM), and a short intracellular domain [35,36]. Y657* results in a termination codon after the sixth Cadherin domain, prior to the TM domain (Fig 2A). To determine whether the premature stop codon in the cdh16 locus is causal for the acoustic hypersensitivity and habituation phenotypes, we used CRISPR-Cas9 genome editing to generate an independent loss-of-function allele, co79, in cdh16 in wild type animals. co79 results in a 10 bp deletion in exon 2, resulting in a frameshift and premature stop codon (F38del[SPSCQISL*]FSX8) (Fig 2A). Like p173, animals homozygous for the co79 mutant allele are hypersensitive to acoustic stimuli and fail to habituate (Fig 2B–2D). Conversely, habituation and startle sensitivity assays demonstrated that animals heterozygous for either co79 or p173 are indistinguishable from their wild type siblings on these measures (Fig 2E–2G). Finally, through complementation testing, we determined that larvae carrying a combination of both mutant alleles (cdh16^p173/co79) fail to habituate and exhibit hypersensitivity to low-intensity acoustic stimuli (Fig 2E–2G). Together, these data demonstrate that Cadherin-16 regulates the establishment and dynamic tuning of acoustic startle thresholds through habituation.

Cadherin-16 expression is sufficient after the development of the acoustic startle circuit to restore acoustic startle thresholds and habituation

The neuronal circuits required for the performance of the acoustic startle response are in place by 4 dpf. By this stage, animals reliably perform acoustic startle responses to high-intensity stimuli and exhibit robust habituation learning [2,37]. Cadherin proteins regulate many developmental processes throughout the body, including the assembly of neuronal circuits [38]. Therefore, we wondered whether cdh16 is required for the assembly of the acoustic startle circuit, or whether it might be required for the maintenance, function, or maturation of the acoustic startle circuit. To test this, we generated a transgene expressing cdh16 under the control of the hsp70 heat-shock activated promoter. We found that ubiquitous, heat-shock induced expression of cdh16 at 3 and 4 dpf rescued acoustic startle thresholds and habituation at 5 dpf, consistent with a role for Cdh16 during development. The same manipulation had no significant effect in sibling animals overexpressing cdh16 (Fig 3A and 3B). However, expression of cdh16 at 2 and 3 dpf did not restore normal behavior measured at 5 dpf, suggesting that maintenance of Cadherin-16 expression at the time of behavior testing is required for the regulation of sensory-evoked behaviors (Fig 3C and 3D). Moreover, we determined that heat-shock induced expression of cdh16 at 5 and 6 dpf rescued acoustic startle thresholds and habituation measured six hours after heat shock on day 6, consistent with a role for Cdh16 in the regulation of sensory processing after the establishment of the acoustic startle circuit (Fig 3E and 3F). To broadly examine how cdh16 might impact neuronal development, we performed whole-brain morphometric analyses in cdh16 mutants versus siblings across 294 molecularly-defined brain regions [39,40]. This unbiased approach for assessing brain development revealed no substantial changes in size across these regions (Figs 3G, S1). Although this approach is designed to detect substantial changes in region-by-region volume, and likely would not resolve finer changes, these findings are consistent with our heat-shock rescue data and underscore a role for Cdh16 in regulating nervous system function rather than early nervous system development.

cdh16 is expressed in the corpuscles of Stannius

In mammals, cdh16 is primarily expressed in the kidney and the thyroid [41,42]. In zebrafish, while cdh16 expression in the brain has been documented at 10 days post-fertilization [43], others have found that at earlier stages cdh16 is primarily expressed in the developing pronephros or embryonic kidney [44]. Around 2 days post-fertilization, cdh16 expression becomes largely restricted to an endocrine organ called the corpuscle of Stannius (CS), which is extruded from the pronephros and secretes Stc1l, a Ca²⁺-regulatory hormone [44]. Given our finding that cdh16 is required for sensory gating after 2 dpf, we wondered whether cdh16 expression might persist in the CS beyond this early developmental time point. To address this question, we used in situ hybridization chain reaction (HCR), examining cdh16 expression in whole-mount embryos and larvae from 24 h post-fertilization (hpf) through 144 hpf (Fig 4A–4G). At 5 dpf, we observed strong cdh16 expression in the CS. Consistent with earlier reports of cdh16 expression [44], we also observed weak signal in other tissues, including in the head (Fig 4A). To distinguish between true cdh16 expression and non-specific probe binding, we generated a large deletion in the cdh16 locus, cdh16^co120. We then designed an HCR probe set, in which all probes bind to target sequences contained within the large deletion (S2A Fig). Finally, we repeated our HCR experiment in cdh16^co120 mutant animals. In mutant larvae carrying the large deletion, we found that signal in the CS was lost, indicating that this signal reports true cdh16 expression. Conversely, weak signal observed in other tissues, including the head, was maintained, consistent with this signal arising due to non-specific probe binding (S2B Fig). These findings are consistent with Daniocell data that indicate larval expression of cdh16 in the CS, but not in neural or glial cell types [45,46]. Indeed, at all time points, we found that cdh16 was strongly expressed either in the pronephros (24 hpf, Fig 4B) or the CS (48–144 hpf) (Fig 4C–4G), and we have no evidence to support cdh16 expression in the brain at 5 dpf. Based on these data, we predicted that Cdh16 is required outside the brain to regulate acoustic startle thresholds.

Cadherin-16 promotes the function of PAPP-AA through the regulation of the hormone Stanniocalcin 1l

Morpholino knockdown of cdh16 in embryonic zebrafish leads to a dramatic increase in the expression of stc1l [44]. In zebrafish and mammals, Stc1 inhibits the metalloprotease PAPP-AA [23,25,47,48], which is a known regulator of acoustic startle sensitivity and habituation in larval zebrafish [11]. pappaa mutants largely phenocopy cdh16 with one exception: pappaa mutants are not responsive to dark-flash, or whole-field loss of illumination [11,49]. We hypothesized that excessive stc1l expression in cdh16 mutants inhibits pappaa, precluding appropriate acoustic startle thresholding and plasticity of thresholds through habituation. To test our hypothesis, we set out to confirm that cdh16 mutants, like cdh16 morphants, show increased expression of stc1l. Using RT-qPCR, we found that as in cdh16 morphants, stc1l expression was strongly increased in cdh16 loss-of-function mutants (Fig 5A). Next, we wondered whether loss of cdh16 might lead to a change in stc1l expression in the brain. To test this, we dissected 5 dpf larval zebrafish, separating the trunk and the head, and performed RT-qPCR in each tissue independently in mutants and siblings. We found that while stc1l was strongly upregulated in the trunk (which contains the CS) (Fig 5B), there was no change in the head (Fig 5C), consistent with a CS-specific role of Cdh16 in regulating stc1l expression. Finally, we wondered how Cdh16 might suppress stc1l expression. Recent work shows that mutations in sox10 increase the number of stc1l-positive cells in the CS, consistent with a possible role in regulating the proliferation or survival of CS cells [50]. To test whether a similar mechanism might underly the role of Cdh16 in suppressing stc1l expression, we performed HCR for stc1l in cdh16^p173 mutant and sibling animals (Fig 5D and 5E), quantifying the stc1l signal per CS as well as the number of cells comprising each CS. Consistent with our RT-qPCR data, we found that stc1l signal per CS was increased in cdh16 mutants relative to their siblings (Fig 5F). Moreover, consistent with a role for Cdh16 in regulating the proliferation or survival of CS cells, we found that the total number of cells comprising each CS was increased in cdh16 mutant animals relative to their siblings (Fig 5G). Next, we quantified stc1l signal per cell and found that this was also increased in cdh16 mutants (Fig 5H), consistent with a role for Cdh16 in regulating both stc1l expression and proliferation or survival of stc1l-expressing cells. Finally, we wondered whether loss of cdh16 might affect the development of the Mauthner cell, a reticulospinal neuron that receives acoustic inputs from the eighth nerve, and sends outputs to motor neurons along the spinal cord, driving the SLC behaviors that are hypersensitive in cdh16 mutant animals. We measured Mauthner soma length (S3A Fig), lateral dendrite length (S3B Fig), and total Mauthner length from the tip of the lateral dendrite to the axon initial segment (S3C Fig). None of these three measures were significantly different between mutant and WT animals (S3D and S3E Fig). Together, these data indicate that Cdh16 functions in the CS to suppress stc1l expression in part by suppressing proliferation or maintenance of cells comprising this endocrine organ.

Based on these findings, we predicted that since hypersensitive cdh16 mutants overexpress stc1l, stc1l loss-of-function would lead to hyposensitivity to acoustic stimuli. To test this, we performed F₀ CRISPR mutagenesis experiments, injecting wild type embryos at the 1-cell stage with Cas9 together with either 3 control guides [51] or together with 3 guides that we designed against stc1l. We found that loss of function in stc1l leads to severe pericardial edema, which becomes apparent by 5 dpf as previously described [23]. Therefore, we tested behavior in larvae injected with stc1l guides (stc1l crispants) at 4 dpf, before severe pericardial edema develops. At 4 dpf, wild type zebrafish larvae are less responsive to acoustic stimuli, but as predicted, we found that stc1l crispants were even less responsive to acoustic stimuli than their control guide injected siblings (Fig 6A). Next, to test our hypothesis that the cdh16 mutant phenotype arises due to overexpression of Stc1l, we injected a construct expressing V5-stc1l under the control of a heat-shock promoter to transiently overexpress Stc1l in wild type embryos. We then heat-shocked the injected embryos (as well as embryos injected with a construct containing the heat-shock promoter alone) at 4 dpf and tested their behavior the next day at 5 dpf. Consistent with our hypothesis that hypersensitivity is a result of overexpressed Stc1l, animals transiently overexpressing Stc1l were hypersensitive to acoustic stimuli relative to control-injected siblings (Fig 6B).

Next, we set out to test whether stc1l overexpression in cdh16 mutants is the cause of the hypersensitivity phenotype. For this, we performed the same CRISPR-Cas9 F₀ mutagenesis in cdh16 mutants and siblings. Consistent with our model, stc1l loss-of-function in cdh16 mutants resulted in hypo-responsiveness to acoustic stimuli (Fig 6C).

Previous work shows that Stc1l limits Ca²⁺ uptake by inhibiting Papp-aa [23]. Therefore, we predicted that pappaa loss-of-function would suppress the hyposensitive phenotype observed in stc1l crispants, and that loss-of-function of both genes would resemble single mutants for pappaa. Indeed, we found that pappaa mutants injected with stc1l guides were hypersensitive, showing no difference relative to control-guide injected pappaa mutants (Fig 6D). If the function of cdh16 is to release pappaa from inhibition by inhibiting stc1l, then animals carrying loss-of-function mutations in both cdh16 and pappaa should be no more hypersensitive to acoustic stimuli than single mutants for either gene. Indeed, our crispant experiments are consistent with this model, as pappaa mutants injected with guides against cdh16 were no more hypersensitive than pappaa mutants injected with control guides (Fig 6E). Finally, we set out to understand whether loss of cdh16 affects expression of pappaa. Stc can both down-regulate the expression of pappaa [23] and separately can inhibit its enzymatic function [47,48]. To address this question, we conducted RT-qPCR experiments to assess whole-body pappaa expression in cdh16^p173 mutants relative to WT. We found that pappaa RNA expression levels were not changed in our cdh16 mutants relative to WT (Fig 6F). Together, our data support a model in which Cdh16 suppresses stc1l and promotes pappaa function rather than directly regulating its expression, though future experiments are needed to carefully assess the enzymatic function of pappaa in cdh16 mutant animals.

The corpuscles of Stannius and Ca²⁺ homeostasis are crucial regulators of acoustic sensory thresholds

Thus far, our data are consistent with a model in which Cdh16 and Papp-aa regulate Ca²⁺ homeostasis to promote acoustic startle thresholds and habituation. Importantly, in addition to its expression in Ca²⁺-regulatory ionocytes, pappaa is expressed in cells surrounding neuromasts, as well as in the retina and brain, including in the acoustic startle circuit [11,25,49,52,53]. However, it is not yet known whether pappaa expression in the brain or potentially in the ionocytes regulates sensory thresholds. First, to test whether cdh16 mutants are hypocalcemic, we performed a colorimetric assay for whole-body Ca²⁺ content. Consistent with a model in which loss of cdh16 leads to excessive stc1l, which downregulates pappaa and ionocyte proliferation and function to ultimately impair Ca²⁺ uptake, we found that cdh16 mutants are hypocalcemic relative to their siblings (Fig 7A). Next, we wondered whether cdh16 mutants might be hypocalcemic as a result of impaired ionocyte proliferation. To assess this, we used HCR probes against trpv6 to label NaR ionocytes in the ventral epithelium. Using the trpv6 signal to count individual ionocytes in mutants versus siblings, we found that mutant animals had a subtle, but significant reduction in NaR ionocyte number (Fig 7B–7D). These data are consistent with loss of cdh16 resulting in overexpression of stc1l, hyper-suppression of Papp-aa function, and ultimately resulting in reduced ionocyte proliferation, reduced Ca²⁺ uptake, and impaired gating. Supporting the idea that reduced Ca²⁺ might cause hypersensitivity to acoustic stimuli, prior work has demonstrated that zebrafish raised in high-Ca²⁺ media are hyposensitive to acoustic stimuli [54]. To test whether low Ca²⁺ media would result in acoustic hypersensitivity, we exposed larvae to low-Ca²⁺ media (0.001 mM) for 4 h. We found that this short-term treatment resulted in acoustic hypersensitivity (Fig 7E) and animals exposed to this treatment trended towards a failure to habituate (Fig 7F). We note that short-term exposure to 0.02 mM Ca²⁺ surprisingly caused the opposite phenotype: animals were hyposensitive and trended toward improved habituation. Finally, to test how short-term exposure to media with altered Ca²⁺ concentration affects whole-body Ca²⁺, we again used our colorimetric assay (S4A Fig). We found that while high Ca²⁺ (10 mM) significantly elevated whole-body Ca²⁺, exposure to low Ca²⁺ trended toward a lower level of whole-body Ca²⁺, but was not significant (p = 0.093).

We additionally examined visually evoked behaviors (S4B–S4D Fig). Like pappaa mutants, animals exposed to low Ca²⁺ media (0.001 mM) show reduced responsiveness to dark-flash stimuli, consistent with low Ca²⁺ in pappaa mutants as an important driver of both phenotypes. These data highlight that low Ca²⁺ and loss of pappaa both cause reduced escape responses to whole-field loss of illumination and increased responsiveness to acoustic stimuli.

Finally, our data suggest that cdh16 regulates sensory thresholds through its function in the CS. To test this, we used a 532 nm pulse laser to ablate the CS in wild type animals expressing her6:mCherry [55], a transgene that labels the CS at 3–4 dpf. Based on the overexpression of stc1l in the CS of cdh16 mutants, and the suppression of hypersensitivity in cdh16^p173; stc1l crispants, we predicted that ablation of the CS would result in hyposensitivity similar to that observed in stc1l crispant animals. Importantly, CS-ablated animals largely did not display pericardial edema at 5 dpf (S4E and S4F Fig). Those with pericardial edema were excluded from analysis. Consistent with a function for cdh16 in the CS, we found that compared to their sham-ablated counterparts, CS-abated wild type animals were hyposensitive to acoustic stimuli (Figs 7G and S4G–S4J).

Fig 8. Proposed model.

(A) In wild type animals, Cdh16 suppresses stc1l expression in the corpuscles of Stannius. This limits the ability of Stc1l to suppress the function of Papp-aa, allowing for some proliferation and function of ionocytes. As a result, Ca²⁺ is taken up from the environment and normal acoustic startle thresholds are maintained. (B) In cdh16 mutant animals, suppression of stc1l expression is relieved and stc1l is overexpressed. This results in hyper-inhibition of Papp-aa. As a result, Ca²⁺ uptake is severely limited, animals are hypocalcemic, and acoustic response thresholds are lowered.

https://doi.org/10.1371/journal.pbio.3003164.g008

Limitations

Although we demonstrate a genetic relationship between cdh16, stc1l, and pappaa, we do not determine a mechanism through which Cdh16 and Stc1l affect Papp-aa. Prior work has shown that loss of stc1l results in increased pappaa expression [23]. Although we would have predicted that cdh16 mutants, which overexpress stc1l, would therefore have decreased pappaa expression, we find that pappaa expression at the whole-animal level is not affected via RT-qPCR. This leaves open the possibility that Cdh16 and Stc1l regulate Papp-aa by regulating its enzymatic activity. Indeed, in vitro experiments demonstrate that Stc1l suppresses Papp-aa enzymatic activity [47,48]. A second possibility is that Cdh16 and Stc1l influence pappaa expression in a regionally-restricted manner not detected by our RT-qPCR experiments. Future work could begin to disentangle these mechanisms by assessing pappaa expression using in situ hybridization to visualize region-specific changes in cdh16 mutants, for example in the retina versus in the ionocytes.

Although we find no evidence that cdh16 is expressed in the brain, we did not test cell-type specific rescue for cdh16 within the CS in this study. It therefore remains possible that cdh16 could function outside of the CS. Nonetheless, our findings that cdh16 is expressed in the CS throughout embryonic and larval development, that it suppresses stc1l expression, including by regulating CS size, that overexpression of stc1l alone recapitulates the cdh16 mutant phenotype, and that CS ablations cause hyposensitivity, strongly support a model in which cdh16 functions in the CS to regulate acoustic startle thresholds in larval zebrafish.

We observe some complexity in the relationship between environmental Ca²⁺ and sensitivity to acoustic stimuli and in the relationship between environmental Ca²⁺ and whole-body Ca²⁺. First, while our lowest Ca²⁺ concentration causes hypersensitivity, 0.02 mM surprisingly causes hyposensitivity. We speculate that this unexpected result may reflect engagement of compensatory mechanisms that drive animals toward hyposensitivity. However, why such compensatory mechanisms are not engaged in 0.001 mM Ca²⁺ is not clear. We note that wild type zebrafish larvae show a remarkable ability to cope with low environmental Ca²⁺ in terms of maintaining bone-mineralization [50]. However, the specific mechanism underlying the complexity in the behavioral response to lowered Ca²⁺ remains unexplained. Secondly, while exposure to high environmental Ca²⁺ caused a significant increase in whole-body Ca²⁺ as measured by our colorimetric assay, low Ca²⁺ trended toward but did not cause a significant decrease. We speculate that the subtle effect of reducing environmental Ca²⁺ may reflect a limitation of our assay: perhaps it largely measures Ca²⁺ reserves in bone, which is present by 5 dpf, and which might outweigh fluctuations in serum Ca²⁺.

Taken together, our studies support a model in which Cdh16 suppresses Stc1l both by suppressing the expression of stc1l and proliferation or survival of Stc1l-expressing cells in the CS. Cdh16 continues to play this role throughout larval development rather than during the specification or assembly of the CS. In cdh16 mutants, overexpressed Stc1l functions to suppress Papp-aa and ionocyte proliferation, and ultimately promotes hypocalcemia and hyper-responsiveness to acoustic stimuli. This work highlights a previously unappreciated role for Ca²⁺ homeostasis in the regulation of acoustic response thresholding and identifies a new brain non-autonomous pathway for the regulation of behavior.

Ethics statement

All procedures were approved by the University of Colorado Anschutz Medical Campus School of Medicine Institutional Animal Care and Use Committee (IACUC). Protocol #1,127. The CU Anschutz IACUC follows the US Public Health Service’s Policy on Humane Care and Use of Laboratory Animals and Guide for the Care and Use of Laboratory Animals.

Experimental model and subject details

Zebrafish larvae were obtained from pairwise or group crosses of adult zebrafish carrying mutations or transgenes of interest on the TLF (WT) background. Larvae were raised at 28.5 °C in E3 media and sorted for normal development.

The p173 allele of cdh16 and the p170 allele of pappaa were recovered from a forward genetic screen [11]. Mutants were genotyped using proprietary allele specific primer sequences (LGC genomics) and the KASP assay method, which utilizes FRET to distinguish between alleles. For genotyping of p173 in the context of Tg[hsp70:cdh16-p2a-mkate], CAPS primers 107 and 108 were used in combination with MseI (see Table 1).

co79 and co120 mutant alleles were generated using CRISPR-Cas9 mutagenesis. To create these alleles, wild type embryos were injected with either sgRNA 622 (co79) or sgRNAs 622 and 867 (co120) (Tables 1 and 2). sgRNAs were designed using ChopChop [72], purchased from IDT, and reconstituted to 200 μM using the IDT-provided duplex buffer. sgRNAs were combined with tracrRNA, also purchased from IDT, to form a 50 μM duplex by heating at 95 °C in a thermocycler for 5 min, followed by cooling to RT for 10 min. Injection mixes were prepared by mixing 1 μL of 50 μM duplex together with 1 μL Cas9 protein (5 mg/mL) obtained from PNA Bio and 1uL phenol red. cdh16^co79 mutations were genotyped by PCR with primers 657 and 658 (Table 1). cdh16^co120 mutations were genotyped via PCR with primers 931, 932, and 658 (Table 1).

Transgenic animals carrying Tg[hsp70:cdh16-p2a-mkate] (co113) were generated by cloning the cdh16 cDNA from total zebrafish RNA at 5 dpf into pME-cdh16-p2a-mKate. Gateway cloning was used to recombine pME-cdh16-p2a-mKate into a pDest vector containing the hsp70 promoter and I-SceI restriction sites, generating hsp70-cdh16-p2a-mKate. I-SceI transgenesis was performed as previously described [73] by injecting I-SceI and the hsp70-cdh16-p2a-mKate plasmid into 1-cell stage TLF embryos. G₀ injected larvae were raised, outcrossed, and heat-shocked at 37 °C in a thermocycler for 45 min to identify carriers. Larvae expressing the transgene were identified by screening for mKate using a fluorescent stereomicroscope (Leica M205FCA). For behavior experiments, animals were pre-screened for fluorescence and genotyped post-hoc using primers 107 and 108 (Table 1).

Transgenic animals carrying Tg[hsp70-V5-stc1l] were generated by cloning V5-stc1l (obtained as a custom gene block from IDT) (Table 3) to generate pME-V5-stc1l through HiFi DNA Assembly. Gateway cloning was then used, as described above, to recombine pME-V5-stc1l into the pDest vector containing the hsp70 promoter and I-SceI restriction sites, to generate hsp70-V5-stc1l. I-SceI transgenesis was performed as described above; we injected either hsp70 vector alone or hsp70-V5-stc1l. G₀ embryos were raised to 4 dpf and heat-shocked at 37 °C in a thermocycler for 45 min. Behavior was then tested at 5 dpf. Presence of hsp70-V5-stc1l was confirmed after behavior testing using cloning primer 1,033 (5′-CTTGTTCTTTTTGCAGgccaccATGCTCCTGAAAAGCGGATTTC-3′) and genotyping primer 706 (5′-CAGCAGAGGGTTTGGGATAG-3′), (annealing temperature: 56 °C, extension time: 30 s, product size: 118 bp). All G₀ larvae injected with the hsp70-v5-stc1l construct genotyped positive; all that were injected with the control plasmid genotyped negative.

Transgenic animals carrying the gal4 driver Tg[gffDMC130a] were provided by the lab of Dr. Koichi Kawakami [74]. Transgenic animals carrying Tg[UAS:Gap43-citrine] were provided by the lab of Dr. Jonathan Raper [75]. Animals carrying Tg[her6:mCherry] [55] were provided by the lab of Dr. James Nichols and outcrossed to TLF for ablation experiments.

To generate conceptual translations of each allele, SMART domain-prediction software was used [76]. SMART identified Cadherin repeats 1–6 based on the full-length protein sequence. Cadherin repeats 7 was not originally identified, however SMART identified a seventh cadherin repeat when the final portion of the extracellular domain was searched alone.

Behavior testing

Before testing their response to acoustic and visual stimuli, larvae were acclimated to the behavior room inside an incubator kept at 28 °C for 30 min. To measure acoustic startle thresholds, six increasingly intense acoustic stimuli were administered 5 times each, 40 s apart, after which acoustic startle response habituation was measured by providing 40 stimuli with a 3-s interstimulus interval (ISI). Visual motor responses (VMR) were measured by first dark-acclimating larval zebrafish inside the behavior arena. Next, the lights were turned on for a 7-min period to assess initial visual motor reactivity in response to light. Then, the lights were turned off for 7 min to assess the initial visual motor response to darkness. Light flash reactivity was examined by first dark-acclimating larval zebrafish inside the behavior arena. Next, larvae were exposed to 10 pulses of light with a one second duration, 30 s apart. To assess dark flash reactivity, 6 dpf larvae were acclimated to the light inside the behavior arena. Following this, the lights were extinguished 5 times in pulses lasting 1 s with a 1-min ISI. To assess dark-flash habituation 60 additional dark flash stimuli were administered with a 10-s ISI. During these final stimuli, the camera recorded behavior during every other stimulus. For the above-described behavior assays, larvae were loaded onto a custom-made acrylic 6 × 6 well-plate attached to a mini-shaker (Brüler and Kjær, Model 4,810), which was used to deliver the acoustic stimuli. A cover was placed over the rig for assays of visually evoked behaviors.

Behavior was recorded with a high-speed camera (FASTCAM Mini UX50 Type 160K-M-32G) placed above the plate and an LED light pointed at the behavior arena was used for light stimuli. Acoustic stimuli were calibrated using an accelerometer (PCB Piezotronics, Y355B03) and stimulus intensities are reported in g or acceleration due to gravity. To analyze behavior, video files were background-subtracted and then analyzed using FLOTE, Batchan [29], and Microsoft Excel. Statistical analyses and graphing were performed using Graphpad Prism.

Larvae were tested for thermal behavior using a 96-well (square wells) plate loaded into a DanioVision observation chamber running EthoVision XT 11.5 software (observation chamber and software, Noldus, Leesburg, VA). The temperature in the observation chamber was set using a temperature control unit. Larvae were acclimated to the baseline temperature of 28.5 °C for 30 min, after which their total distance moved was recorded for 2 min. The temperature was then raised to 33.5 °C, and fish were recorded again for 2 min. All behavioral assays were performed at 5 dpf, except for our dark flash assay, which was performed at 6 dpf.

Heat-shock induced cdh16 rescue and stc1l overexpression

To induce expression of hsp70-cdh16-p2a-mKate, zebrafish embryos or larvae were placed in a 96-well plate at a density of no more than 5 larvae per well. The plate was heated to 37 °C for 45 min using a thermocycler. Larvae were then recovered to petri dishes for at least 5 h before behavior testing. Similarly, to overexpress stc1l via hsp70-V5-stc1l, 4 dpf larvae were placed into a 96-well plate at a density of no more than 5 larvae per well. The plate was heated to 37 °C for 45 min using a thermocycler. The larvae were then recovered to petri dishes for 24 h before testing at 5 dpf.

Crispant (F₀) mutagenesis and behavior analysis

sgRNAs targeting cdh16 (622, 623, 867) and stc1l (942, 943, 944) were designed using ChopChop [72]. Scrambled sgRNAs (759, 760, and 761) were used as controls and were designed by IDT as previously described [51]. The sgRNAs were purchased from IDT and reconstituted to 200 μM stocks using the IDT-provided duplex buffer. sgRNAs were then combined individually with tracrRNA, also purchased from IDT, to form a 61 μM duplex by heating at 95 °C in a thermocycler for 5 min, followed by cooling to RT for 10 min. Injection mixes were prepared by mixing 1 μL of duplex together with 1 μL Cas9 nuclease V3 (10 μg/μL; IDT Cat #1,081,059). 1nl of injection mix was injected in the yolk at the single cell stage, before the cell inflates.

The mutation rate in crispants was assessed by PCR using primers flanking the sgRNA target sequences to detect indels and large deletions. Following behavioral analysis, we genotyped larvae injected with gene-specific sgRNAs and larvae injected with control sgRNAs to confirm guide efficiency (see Tables 2 and 4).

Hybridization chain reaction (HCR) FISH staining

HCR probes, hairpins, and buffers were purchased from Molecular Instruments. Staining for experiments utilizing cdh16, cdh16 (deletion set), stc1l, and trpv6 HCR probes was performed using the manufacturer’s protocol: “HCR RNA-FISH protocol for whole-mount zebrafish embryos and larvae (Danio rerio)” with the following modifications: we did not apply PTU to inhibit melanogenesis, except in the case of trpv6 HCR to quantify NaR cells, we used 30 larvae per Eppendorf tube, and we used 8 μl of 1 μM cdh16 and cdh16 (deletion set) probe solutions, all other probes were used at 4 μl of 1 μM probe set solution in 500 μL of Probe Hybridization Buffer. For CS imaging, animals were mounted laterally in 1.5% low-melt agarose in PBS and imaged using a 63× objective on a 3i Marianas Spinning Disk Confocal Microscope. For ionocyte imaging, animals were mounted with their ventral side facing the objective. For whole-fish images, 8 individual 20× images were acquired to capture the entire larva using the 8 × 1 montage function in SlideBook and then stitched using the legacy montage function.

Calcium manipulations

To create Ca²⁺-supplemented media, we first created a stock solution of 60× E3 embryo media without Ca²⁺: 300 mM NaCl, 10.2 mM KCl, and 19.8 mM MgSO₄·7H₂O. A separate stock solution of 60× CaCl₂·2H₂O (Sigma CAS#:10035-04-8) was also made. Ca²⁺ concentrations of 10 mM Ca²⁺, 0.33 mM (Normal), 0.02 mM, and 0.001 mM were generated by mixing 60× E3 and 60× CaCl₂ in the appropriate ratios. At 5 dpf, larvae were rinsed three times out of E3 media containing normal Ca²⁺ (0.33 mM Ca²⁺), and into one of the four different Ca²⁺-supplemented media concentrations four hours before performing behavior and then tested in those same Ca²⁺ concentrations.

Corpuscle ablations

Tg[her6:mCherry] embryos were screened for mCherry expression at 3 dpf using a fluorescent stereomicroscope. Four dpf mCherry-positive larvae were live-mounted laterally in 1.5% low-melt agarose (Lonza Cat# 50,101) in E3 embryo media on a 3.5 cm glass-bottom dish. The CS were identified and then ablated using 532 nm pulse laser attached to a 3i Marianas spinning disk confocal microscope with a 63× objective. To ensure complete ablation, an average of 3 laser pulses were administered per corpuscle (laser pulses were delivered until the CS was eliminated). For sham ablations, a target region posterior to the kidney and yolk extension was located and ablated, after which the CS were re-located and confirmed to be undamaged. Ablated and sham-ablated larvae were then unmounted and placed in a 6 cm petri dish with fresh E3 to recover for approximately 21 h, after which they were behavior tested at 5 dpf for acoustic startle thresholds and habituation.

WGS and molecular cloning of cdh16

Molecular cloning of the cdh16 allele was performed as previously described [11,14]. Pools of 50 behaviorally identified p173 mutant larvae were collected and used to prepare genomic DNA (gDNA) libraries. gDNA was sequenced with 100-bp paired-end reads on the Illumina HiSeq 2000 platform, and homozygosity analysis was done using 463,379 SNP markers identified by sequencing gDNA from ENU-mutagenized TLF and WIK males as described previously [11].

Calcium content assays

Whole-body Ca²⁺ was quantified using a colorimetric assay kit (Abcam ab102505). For the first experiment measuring Ca²⁺ in cdh16^p173, 2 dpf larvae were live tail-clipped and genotyped for the cdh16^p173 allele. At 4 dpf, 10–15 larvae were pooled in each of 6 Eppendorf tubes: three WT biological replicates and three mutant biological replicates. For the second experiment, Ca²⁺ was measured following acute (four hour) exposure of wild type larvae to four Ca²⁺ concentrations at 5 dpf (10 mM, 0.33 mM, 0.02 mM, 0.001 mM). Then 10–15 larvae were pooled in 12 Eppendorf tubes: three WT biological replicates for each of the four Ca²⁺ concentrations. The assays were then performed as previously described [50].

RT-qPCR

Larvae were dissected to remove the distal tip of the tail for genotyping. To generate cDNA from heads and trunks, larvae were dissected to isolate the head from the trunk at the base of the hindbrain. Tissue to be used for RT-qPCR was placed into RNAlater (Sigma Cat# R0901-100ML) and stored at 4 °C. Following genotyping, whole larvae (Figs 5A and 6F), trunks (Fig 5B), or heads (Fig 5C) were pooled by genotype and total RNA was extracted using Trizol/Chloroform followed by the RNeasy Plus Mini Kit (Qiagen Cat# 74,143). cDNA pools were generated using SuperScript II Reverse Transcriptase (Invitrogen Cat# 11,904−018). qPCR was performed with LUNA qPCR MasterMix (NEB Cat# M3003) on a QuantStudio 3 Real-Time PCR System (Fisher Cat# A28566) using qPCR primers (Table 5) designed for each target gene. Expression levels of target genes were normalized to gapdh.

Analysis of stc1l in the corpuscles of Stannius

To measure stc1l fluorescence intensity, images acquired at 63× magnification of the stc1l-labeled CS were converted to OME-TIFFs. TIFFs were sum projected over their entire z-stack using FIJI. Regions of interest were then drawn freehand around the borders of each corpuscle, or around the border of the entire structure when the two corpuscles overlapped in Z. Fluorescence intensity was then measured and divided by the number of corpuscles (2 CS were observed in every animal except one mutant that had 3 CS). To count the total number of stc1l+ cells in the CS, z-stacks of the CS were visually inspected using Slidebook imaging software and marking each cell using the freehand ROI tool to assign it a number. This number was divided by the number of corpuscles, to quantify the number of stc1l+ cells per CS. Finally, the intensity per CS was divided by the number of stc1l+ cells per CS, to quantify the relative intensity of stc1l signal per CS cell.

Epithelial ionocyte (NaR cell) quantification

Two z-stacks encompassing trpv6 expression in NAR cells along the jaw and the skin overlying the swim bladder were acquired at 20× magnification for each larva. To quantify the number of NaR cells, we stitched swim bladder and jaw images using FIJI. A maximum projection was then made from the combined z-stack and the despeckle function was used to reduce noise. An image threshold was applied with 6,725 as the minimum value and 65,535 as the maximum value. The watershed function was then used to further delineate the cell borders from one another. When necessary, an ROI was drawn to exclude any autofluorescence in the eyes, and the included region was analyzed for the total number of particles, taken to be the total number of NaR cells in this region.

Mauthner morphology quantification

To assess Mauthner morphology, WT and cdh16^p173 mutant larvae heterozygous for each of Tg[Gap43:Citrine] and Tg[Gffdmc130a] were fixed overnight in sweet fix (4% paraformaldehyde, 4% sucrose, 1× PBS). After 3 washes in PBT, tails were clipped, lysed, and KASP genotyped for the cdh16^p173 allele. Brains were then dissected from wild type and cdh16 mutant larvae, mounted dorsal side down in Vectashield, and assigned random numbers by an independent investigator to ensure blinding during imaging and analysis. Z-stacks of Mauthner neurons were acquired using a 3i Marianas spinning disk confocal microscope with a 63× objective. Maximum intensity projections of the soma and lateral dendrites were generated in FIJI and measurements were performed using the straight-line tool. Mauthner neuron morphology was quantified as follows: Whole Mauthner length was measured from the axon initial segment to the distal-most tip of the lateral dendrite. Soma length was measured from the axon initial segment to the first narrowing at the soma-lateral dendrite boundary. Lateral dendrite length was measured from this point to the distal-most tip of the distal-most lateral dendrite branch. Two Mauthner neurons were measured and displayed per fish, except in cases where only one Mauthner neuron was labeled (for WT, n = 13 showed labeling in both Mauthners, n = 3 had only one labeled; for cdh16 mutant, n = 9 showed labeling in both Mauthners, n = 1 had only one labeled).

Quantification and statistical analysis

Statistical tests were performed in Graphpad PRISM 9 and 10. To determine normality for each data set, the D’Agostino and Pearson test was performed. In normally distributed data, an unpaired T test, one-way ANOVA, or two-way ANOVA was performed as needed. To account for multiple comparisons in two-way ANOVAs, the Šidák’s multiple comparisons test was performed when comparing means across one variable while the Tukey’s multiple comparisons test was used to compare means between all experimental groups. In datasets that are not normally distributed, a Mann–Whitney test was executed to compare two groups and a Kruskal–Wallis test was used to compare between greater than two groups.

Whole brain morphometric imaging and analysis

Six dpf larvae (n = 60) from a cdh16^p173 heterozygote incross were acclimated to the behavior testing room for 30 min. Following acclimation, the larvae were placed in a cell strainer within a 6 cm petri dish containing E3 for 30 min. Finally, spontaneous behavior was recorded for 16 min before the cell strainer was removed and placed into a 6 well dish containing 4% paraformaldehyde in PBT (PBS-Triton 0.25%) for 45 s to flash-fix the larvae. The cell strainer was then transferred to a solution of 4% paraformaldehyde in PBS, incubating at 4 °C overnight. Larvae were moved from the cell strainer to a 1.5 mL tube and washed with PBT for three, 5-min washes. To increase the ratio of mutants to WT larvae included in the imaging experiment, tail clips were collected from each sample, lysed, and KASP genotyped for the cdh16^p173 mutation. Wild type and mutant larvae were pooled at a 1:1 ratio into a 1.5 mL tube containing PBT and stained according to a previously developed immunohistochemistry protocol for MAP-mapping [40] with procedural alterations [17]. Finally, samples were mounted onto a glass-bottom dish using 1.5% low-melt agarose made with PBS. Each larva was positioned with the dorsal portion of its brain facing the glass bottom of the dish. Whole-brain z-stacks were collected for each sample using an LSM780 microscope with a 20× objective and 2 × 1 tile scanning. Larvae were unmounted from the agarose and gDNA was prepared for KASP genotyping. Morphometric analysis of cdh16^p173 mutants was then performed as previously described [39,40]. Differences in whole brain morphology were examined by assessing the significant delta medians of mutants over WT.

S1 Fig. Whole-brain morphometric analysis reveals minimal changes to region-by-region brain volume.

(A) Summary of whole-brain morphometric data for 6 dpf cdh16^p173 mutants (n = 13) as compared to siblings (n = 16). Region-by-region differences in volume are indicated in yellow (regions that are larger in mutants) or cyan (regions that are smaller in mutants). Image is a summed stack of the significant delta medians of mutants over wild types. Note there are no colored pixels within the brain, indicating no significant differences between mutants and siblings across the annotated brain regions.

https://doi.org/10.1371/journal.pbio.3003164.s001

(TIF)

S2 Fig. Cadherin-16 is expressed in the CS.

(A) Schematic of the large (approximately 49 kB) deletion allele that we generated in the cdh16 locus (cdh16^co120). (B) in situ HCR image of cdh16^co120 mutant larva using only probes that bind within the large deletion. The corpuscles of Stannius are not labeled. All other expression, including weak labeling in the head, can be considered background (representative image in Fig 4A). Shown is a representative image of cdh16^co120; n = 4 larvae imaged.

https://doi.org/10.1371/journal.pbio.3003164.s002

(TIF)

S3 Fig. Cadherin-16 does not regulate Mauthner cell morphology.

(A–E) Analysis of the Mauthner neurons responsible for the acoustic startle (SLC) response. (A–C) cdh16^p173 mutants (n = 19) and wild type siblings (n = 29) have no differences between Mauthner soma length (A) p = 0.3169, lateral dendrite length (B) p = 0.4054, or total length (C) p = 0.9248, unpaired t test. Error bars represent SD. (D–E) No morphological differences are detected between the Mauthner cells of cdh16^p173 mutant (bottom image) and WT larvae (top image). The data underlying this figure can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3003164.s003

(TIF)

S4 Fig. Ca²⁺ homeostasis and the corpuscles of Stannius regulate sensory gating.

(A–D) E3 media with differing Ca²⁺ concentrations were applied to WT larvae at 5 dpf. Whole-body Ca²⁺ was measured 4 h later (A) or behavioral assays were performed 4 h later (B–D). (A) Exposure to high (10 mM) Ca²⁺ caused a slight but significant increase in whole-body Ca²⁺ levels, compared to normal 0.33 mM Ca²⁺ (*p = 0.024). Exposure to media containing low Ca²⁺ (0.02 and 0.001 mM) did not significantly affect whole-body Ca²⁺ levels (n = 3 biological replicates per condition, p = 0.8297, p = 0.093; one-way ANOVA with Dunnett’s multiple comparisons test). Error bars represent SD. (B) As is observed in pappaa mutant larvae, animals in 0.001 mM Ca²⁺ (n = 17) show decreased responding to dark flashes relative to siblings in 0.33 mM Ca²⁺ (n = 18) ****p < 0.0001, Kruskal–Wallis test with Dunn’s multiple comparisons test. Error bars represent SD. (C) Animals in the lowest (0.001 mM) Ca²⁺ concentration (n = 18) were more responsive to the lights-on stimulus in the visual motor assay as compared to their siblings in a normal 0.33 mM Ca²⁺ concentration (n = 18), p = 0.0033, Kruskal–Wallis test with Dunn’s test for multiple comparisons. Error bars represent SD. (D) Animals in 0.001 mM Ca²⁺ (n = 18) displayed more robust responses to a light flash than their siblings in 0.33 mM Ca²⁺ (n = 18) ***p = 0.0009, Kruskal–Wallis test with Dunn’s test for multiple comparisons. Error bars represent SD. (E–F) Images of 5 dpf WT larvae 24 h after either CS ablation (left) or sham ablation (right). Larvae with ablated corpuscles do not have visible pericardial edema. (G–J) stc1l HCR to visualize the CS after sham ablation (G, I) or CS ablation (H, J). Only a few stc1l-positive cells are present in the CS region 4 h after CS ablation (H), and stc1l expression is strongly reduced. By 24 h post-CS ablation, the structure has partially regenerated (J). Imaged n = 10 CS-ablated 4 h post-ablation, n = 5 sham-ablated 4 h post-ablation, n = 10 CS-ablated 24 h post-ablation, n = 5 sham-ablated 24 h post-ablation. The data underlying this figure can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3003164.s004

(TIF)