A sensation for inflation: initial swim bladder inflation in larval zebrafish is mediated by the mechanosensory lateral line

The swim bladder of teleost fish provides for neutral buoyancy in the water column, thereby minimizing the energy expenditure associated with locomotion (Alexander, 1966). Larval teleosts initiate swim bladder inflation by swimming up to the surface to ingest air using a set of behaviors we refer to as ‘surfacing’ (Evans and Damant, 1928; Lindsey et al., 2010; von Ledebur, 1928). Although surfacing poses a significant risk of predation, larval fish are highly motivated to reach the surface to intake air. For example, pre-inflated lake trout larvae will swim tens to hundreds of meters at a constant rate when attempting to access the surface (Tait, 1960). Surfacing is a complex task that requires a combination of newly developed physical features, sensory systems and motor skills (Bailey and Doroshov, 1995; Chatain, 1989; Doroshev et al., 1981; Kimmel et al., 1995; Kitajima et al., 1994): (i) Larvae must discriminate up from down; (ii) they must be able to move directionally towards the surface; (iii) they must appropriately sense the air–water interface; and (iv) be able to intake air through the mouth upon arrival at the surface. Ingested air is then moved by peristalsis into the swim bladder through the pneumatic duct that connects the gut and swim bladder (Doroshev et al., 1981; Rieger and Summerfelt, 1998; Tait, 1960). Together, surfacing represents perhaps the most complex behavior that larvae perform prior to achieving neutral buoyancy (Fero et al., 2011).

Little is known about the sensory cues that instruct the surfacing behavior. Perturbations to the vestibular system typically result in hypoinflation (Nicolson et al., 1998; Schoppik et al., 2017). The inability to detect gravity by the utricular otolith is specifically responsible for the phenotype, presumably because larvae are unable to orient their movements upwards towards the surface (Riley and Moorman, 2000). Photosensory cues may also play a role in directional orientation and surface detection. However, which photosensory systems are involved has not been defined and the effects of altering lighting conditions varies between species (Stuart and Drawbridge, 2012; Trotter et al., 2003; Villamizar et al., 2009). Even less is known about how larval fish detect the air–water interface itself. Fish and some amphibians possess an additional hair cell-based sensory system – the mechanosensory lateral line – that contributes to detection of near-field hydrodynamic information originating from water currents, predators and prey, conspecifics, abiotic objects and surface waves (Anneser et al., 2020; Bleckmann, 2006; Butler and Maruska, 2016; Carrillo and McHenry, 2016; Mogdans, 2019; Montgomery et al., 1997; Stewart et al., 2013; Venuto et al., 2022). Specific to surface detection, surface-feeding fish use their lateral line to detect surface waves created by their prey (Bleckmann and Schwartz, 1982). Furthermore, Japanese flying fish are predicted to use their ventrally located lateral line to sense transitions through the air–water interface (Gibbs, 2004; Tsukamoto and Yoshino, 1957). While the lateral line is a strong candidate to aid in surfacing behaviors that lead to initial swim bladder inflation in larval fish, its specific role in this important behavior has not been explored.

A CRISPR-Cas9 knockout of LHFPL tetraspan subfamily member 5b (lhfpl5b) is the first genetic zebrafish mutant where the lateral line is non-functional from birth but hearing and balance are normal (Erickson et al., 2020). These mutants provide an opportunity to uncover novel roles for the lateral line. Unexpectedly, homozygous mutant larvae present with a hyperinflation phenotype characterized by over-filled swim bladders and positive buoyancy, which is a very rare finding among swim bladder defects (Abbas and Whitfield, 2009; McCune and Carlson, 2004). Using several genetic and chemical techniques to manipulate lateral line function, we show that lateral line sensory information is used by larval zebrafish to appropriately fill their swim bladders, probably by detecting surface tension at the air–water interface using mechanosensory neuromasts located on their heads. Overall, our study uncovers a novel role for the mechanosensory lateral line mediating initial inflation of the swim bladder.

Adult zebrafish (Danio rerio) were maintained and bred using standard procedures (Westerfield, 2000). All experiments used larvae at 2–6 days post-fertilization (dpf), which are of indeterminate sex at this stage. Except where otherwise stated, animals were placed in clear, plastic tubs (dimensions: 15 cm×10 cm×4 cm, E3 volume: 460 ml) at 2 dpf and maintained at 28.5°C on a 14 h:10 h light:dark cycle at 500–700 lux in E3 embryo medium (5 mmol l⁻¹ NaCl, 0.17 mmol l⁻¹ KCl, 0.33 mmol l⁻¹ CaCl₂, 0.33 mmol l⁻¹ MgCl₂, buffered with NaHCO₃). Unhatched embryos were manually dechorionated before placing larvae in experimental conditions at 2 dpf. It is worth noting that there was no obvious difference in wild-type versus lhfpl5b mutant hatching rate, though this was not explicitly studied. Animal research complied with guidelines stipulated by the Institutional Animal Care and Use Committees at East Carolina University (protocol D372) and Amherst College (assurance number 3925-1 with the Office of Laboratory Animal Welfare).

The zebrafish mutant allele lhfpl5b^vo35 and transgenic lines Tg(myo6b:EGFP-lhfpl5a)vo23 and Tg(myo6b:EGFP-PA)vo68 were used in this study (Erickson et al., 2017, 2020). For all experiments, lhfpl5b^vo35 homozygotes were identified from their hetero- and homozygous wild-type siblings by lack of FM 1-43 (Thermo Fisher) labeling of lateral line hair cells.

To create a fish line that expresses the Channelrhodopsin-2 (ChR2) optogenetic protein solely in the hair cells of the lateral line, approximately 2 kb of DNA upstream of the start codon of the lhfpl5b gene was cloned into the p5E plasmid of the recombination-based multisite Gateway cloning system (Thermo Fisher). A Tol2 transposon backbone was used to create the injectable expression plasmid lhfpl5b2028:ChR2-EYFP-PA. This plasmid was then co-injected with transposase mRNA into single-cell embryos (Kwan et al., 2007) to create the Tg(lhfpl5b2028:ChR2-EYFP-PA) zebrafish line.

To ascertain the buoyancy status of larval fish, larvae were anesthetized with MS-222 (Western Chemical Inc., Ferndale, WA) and placed at the vertical midway point of a 50 ml conical tube containing E3 medium. After 30 s, the end position of the fish was recorded and the fish was removed from the tube for imaging under a microscope. If the fish reached the top or bottom of the tube before 30 s, then the time it took to reach that location was recorded. The total height of the water column was 8.9 cm, and the fish was placed half-way at 4.45 cm. The rate of buoyancy was determined by dividing the centimeters travelled by the seconds it took to travel that distance (cm s⁻¹). Over-inflated fish were categorized by larvae floating to the top of the tube in 30 s, while under-inflated fish sunk to the bottom of the tube. Neutrally buoyant or regular-inflated larvae remained near the middle of the water column. This test was conducted during the imaging portion at the end of each experiment, with the exception of the oil–water interface experiment where buoyancy could not be accurately determined because of the greater density of oil compared with air.

Swim bladders were imaged using a SteREO Discovery.V8 microscope (Zeiss) equipped with a Gryphax Arktur camera (Jenoptik). Lateral and dorsal images of the swim bladder were taken for each fish. Swim bladder measurements were conducted using the measurement tool in Adobe Photoshop where measurements were calibrated according to the image of the stage micrometer taken when imaging experiments. The major axis and minor axis of the lateral view of the swim bladder was measured as well as the minor axis of the dorsal view. All measurements were taken from numbered images so that the measurer was blinded to the genotype/condition. These meaurements were halved and used in the equation for the volume of an ellipsoid: V=4/3πabc to find the volume of the swim bladder (Lindsey et al., 2010).

At 0 dpf, zebrafish embryos were collected and immediately placed in plastic tubs filled with E3 (detailed above) in an unlit incubator. Larvae were not brought into light until 6 dpf, at which point lateral line mutants and wild-type siblings were identified by FM 1-43 labeling. After sorting, larvae were analyzed for buoyancy. The experiment was repeated three times using 20–21 larvae (combined mutant and wild-type) per trial.

Mutant larvae and wild-type siblings were sorted at 2 dpf into glass cylinders (diameter: 9.5 cm; E3 volume: 425 ml) equipped with fitted, wire mesh filter plungers (Bodum Inc., USA). Conditions were as follows: (i) wild-type with open access to the surface, (ii) wild-type with blocked access to the surface, (iii) mutant with open access to the surface and (iv) mutant with blocked access to the surface. All containers were filled half-way with E3 and in the experimental condition (‘blocked access’), the filter was pressed down below the surface for both wild-type and mutant larvae. The control condition used the same volume of E3 but the filter remained above the surface so that larvae could access the exogenous air. Conditions remained constant until 6 dpf when larvae were imaged and analyzed for swim bladder inflation. The experiment was repeated three times using 15–25 larvae per condition.

Wild-type and lhfpl5b^vo35 larvae were sorted at 2 dpf and blocked from having access to the surface until 09:00 h on 4 dpf when individuals were immediately placed into clear rectangular containers (dimensions: 3 cm×3 cm×3 cm; E3 volume: 20 ml) for video recording. Recordings lasted 12 h, from 09:00 h to 21:00 h on 4 dpf. At the end of filming, larvae were imaged and analyzed for buoyancy and swim bladder volume, as described. All surfacing videos were numbered for blind analysis. Surfacing analysis was conducted by manual observation of each individual larvae from the videos (30 frames per second) and manually counting visits made to the surface (counted as reaching the meniscus), the time frame of the video at which the surfacing event occurred and the duration of time spent at the surface per surfacing event. Recording was repeated six times with both mutant wild-type larvae present in each recording. Larvae that did not inflate their swim bladders by the end of filming (5 larvae total: 1 wild-type and 4 mutants) were excluded from this analysis owing to lack of surfacing events.

For all experiments, wild-type and lhfpl5b mutant larvae were treated with ototoxin with untreated siblings acting as controls (n=7–15 larvae per condition per trial; 4 trials) For single treatments, surface access was blocked as described above for untreated and treated larvae until 09:00 h on 4 dpf when treatment occurred. Single 30 min treatments with 100 µmol l⁻¹ neomycin sulfate (EMD Millipore) in E3 were carried out and repeated 50 µmol l⁻¹ neomycin treatment was as previously described (Venuto and Erickson, 2021). Full ablation was carried out with 10 µmol l⁻¹ copper sulfate (Ward's Science, Rochester, NY) with a 30 min exposure. For experiments where head and trunk neuromasts were selectively ablated with CuSO₄, anesthetized larvae were moved to a rectangular plate (4.5 cm×3 cm×1 cm) coated with Sylgard (Electron Microscopy Sciences, Hatfield, PA, USA). After removing excess water, a thin layer of petroleum jelly (Vaseline, Unilever, USA) was placed horizontally at the back of the head. Next, 30 µmol l⁻¹ CuSO₄ or E3 was pipetted onto the appropriate region depending on the experimental condition. Immediately following the 13 min treatment, larvae were placed in fresh E3 in clear tubs detailed above for 24 h until 09:00 h on 5 dpf when larvae were imaged and analyzed for swim bladder inflation. For all experiments, larvae were imaged and analyzed for swim bladder inflation at 09:00 h on 5 dpf. Each experiment was repeated three times using 10–24 larvae per condition.

Larval zebrafish from crosses of lhfpl5b^+/− and Tg(myo6b:EGFP-lhfpl5a)vo23Tg; lhfpl5b^+/− were sorted at 2 dpf for transgene expression. At 6 dpf, larvae were numbered, imaged for swim bladder inflation and genotyped for the lhfpl5b^vo35 allele as previously described (Erickson et al., 2020). The experiment was repeated twice using 26–48 larvae per condition.

After blocking access to the surface until 4 dpf, we exposed wild-type and lateral line mutants, with or without the lhfpl5b2028:ChR2-EYFP-PA transgene, to LED blue light (470 nm) with an intensity of 730 lx over their holding tanks against a dark background. Light flashed for 25 ms at 1 s intervals. Control larvae were kept in the same background, without blue light exposure. Conditions were held constant until 6 dpf when larvae were sorted for genotype, imaged and analyzed for swim bladder inflation. The experiment was repeated three times using 15–25 larvae per condition.

Mutant larvae and wild-type siblings were sorted at 2 dpf into clear cylindrical containers (diameter: 5 cm, volume: 60 ml) with embryo medium only or embryo medium with a 2 mm thick surface layer of lab grade mineral oil (Ward's Science, Rochester, NY, USA). Oil on the surface decreases surface tension (Johansen, 1924) and has been previously used in experiments involving the zebrafish swim bladder (Ehrlich and Schoppik, 2017). Conditions remained constant until 6 dpf when larvae were imaged and analyzed for swim bladder volume. Because oil-filled larvae are negatively buoyant, we did not perform buoyancy tests and classified larvae as over-filled based on swim bladder volume alone. The experiment was repeated three times using 15–25 larvae per condition.

All graphs and statistical tests were done using R (https://www.r-project.org/). One-way ANOVAs were used to compare the conditions as a whole (genotype independent variable plus environment independent variable) with Tukey post hoc tests for pairwise comparisions (Table S1). Chi-squared tests were used to analyze proportion data for over and under-inflation. For the over-inflation comparisons, the two categories used in the comparison between conditions/genotypes were (i) over-inflation and (ii) all other inflation (regular and under), meaning that totals per category/genotype accounted for 100% of the population. Similarly, for the under-inflation comparisons, the two categories used in the comparison between conditions/genotypes were (i) under-inflation and (ii) all other inflation (regular and over). P-values less than 0.05 were considered significant.

The lateral line has not been previously implicated in initial inflation of the larval swim bladder. As such, it was unexpected to observe that lateral line mutants (lhfpl5b^−/−) exhibited hyperinflated swim bladders by 5–6 dpf (Fig. 1A). Based on our swim bladder volume and buoyancy measurements, we categorized each larva in the mutant (n=31) and wild-type (n=33) populations as over-, regular- or under-inflated (Fig. 1B). On average, 53.7±6.4% (mean±s.d.; n=41) of lateral line mutants over-inflated their swim bladders and 7.3±7.1% (n=41) under-inflated them. For wild-type larvae (including lhfpl5b^vo35 heterozygotes), 0±0% (n=41) over-inflated and 4.9±3.7% (n=41) under-inflated their swim bladders during initial inflation on average (Fig. 1C). The proportion of over-inflation was significantly different between mutant and wild-type larvae (χ²: 70.45, P<0.00001), while the proportion of under-inflation was not (χ²: 0.3546, P=0.552). Average swim bladder volume of mutant fish was larger (0.012±0.002 mm³) than that of wild-type siblings (0.0058±0.00043 mm³; t₈₀=17.4, P<0.0001) (Fig. 1D). The proportions of neutrally and positively buoyant larvae did not change when raised in total darkness (Fig. S1). This unexpected hyperinflation phenotype in lhfpl5b^vo35 homozygotes led us to investigate if the lateral line contributes important sensory information during the surfacing behavior.

For most teleosts, initial swim bladder inflation requires an intake of exogenous air, typically from the water's surface. To test if the hyperinflation phenotype of lateral line mutants requires access to surface air, we physically blocked access to the surface from 2 to 6 dpf (Fig. 2A). As expected, for wild-type larvae, blocking access to the surface led to significantly more underinflated larvae (Fig. 2B) and a significant decrease in the average swim bladder volume (blocked: 0.0037±0.001 mm³, n=49; open: 0.0066±0.0009 mm³, n=63) (one-way ANOVA with Tukey post-test, P=0.002) (Fig. 2C). For lhfpl5b mutant siblings, blocking access to the surface completely prevented the hyperinflation phenotype, with 0±0.0% (n=43) of blocked lhfpl5b mutants exhibiting positive buoyancy, which is significantly less than mutants with access to the surface (χ²: 70.45, P<0.00001; Fig. 2B). Likewise, the average swim bladder volume of blocked mutants (0.0033±0.0015 mm³) was significantly smaller than mutants with open access (0.014±0.0027 mm³, n=58) (one-way ANOVA with Tukey post-test, P<0.00001). Blocked mutant and wild-type siblings had statistically similar swim bladder volumes (one-way ANOVA with Tukey post-test, P=0.972). From these data, we concluded that the lhfpl5b mutant hyperinflation phenotype requires intake of exogenous air and is not caused by abnormal gas exchange from the surrounding medium or excess internal gas production.

We next examined if lateral line defects are responsible for the hyper-inflation phenotype of lhfpl5b mutants by restoring lateral line function in mutant larvae. We used the Tg(myo6b:EGFP-lhfpl5a)vo23 transgene, which uses a hair cell-specific promoter (myo6b) to drive a GFP-tagged lhfpl5a gene (Erickson et al., 2017). It has been shown that this lhfpl5a transgene rescues lateral line hair cell function in lhfpl5b^−/− mutant fish (Erickson et al., 2020). Using this rescue method, we observed no cases of hyperinflation in transgenic mutants (0±0.0%, n=17), which was significantly different from non-transgenic mutants (χ²: 82.12, P<0.00001; Fig. 3A). The average volume of the transgenic mutant swim bladder (0.0057±0.00001 mm³) was significantly smaller than non-transgenic, lhfpl5b mutant siblings (0.0109±0.0031 mm³, n=18) (one-way ANOVA with Tukey post-test, P=0.000003), but statistically similar to that of wild-type transgenics (0.0057±0.0005 mm³, n=57) (one-way ANOVA with Tukey post-test, P=0.979; Fig. 3B). Because the restoration of lhfpl5 function specifically in hair cells mitigates hyperinflation in lhfpl5b mutants, we concluded that the phenotype is due to defects in sensory hair cell function and not caused by an unrecognized role for lhfpl5b in another sensory organ or the swim bladder itself.

To further test if lateral line defects are responsible for the hyper-inflation phenotype seen in lhfpl5b mutants, we ablated lateral line hair cells in wild-type larvae using either neomycin sulfate (neo) or CuSO₄ (Fig. S2 and Fig. 4). Single or repeated neomycin treatments during the initial swim bladder inflation period of 3–4 dpf resulted in a significant increase in the proportion of positively buoyant larvae (single neo, χ²: 9.95, P=0.0016, n=58; repeated neo, χ²: 34.88, P<0.0001, n=47; Fig. S2A), but a non-significant increase in the average swim bladder volume (one-way ANOVA with Tukey post-test: single neo, P=1.0; repeated neo, P=0.578; Fig. S2B). By comparison, a single CuSO₄ treatment resulted in a hyperinflation phenotype that was statistically indistinguishable from lateral line mutants. An average of 59.6±7.7% (n=72) of CuSO₄-treated wild-type larvae over-inflate their swim bladders, which was significantly different from wild-type controls (χ²: 39.69, P<0.0001, n=44; Fig. S2A). The average swim bladder volume of the CuSO4-treated wild-type larvae (0.0126±0.0027 mm³) was statistically similar to the average swim bladder volume of untreated lateral line mutants (0.0139±0.0007 mm³, n=45) (one-way ANOVA with Tukey post-test, P=0.813) (Fig. S2B).

Since the anterior lateral line is likely to be the only part of the sensory organ to interact with the air–water interface during surfacing, we predicted that disruption of the head neuromasts would result in hyperinflation and that trunk neuromasts would not play a role (Fig. 4; Fig. S3). Selective ablation of the anterior region with CuSO₄ resulted in an average of 35.9±10.1% (n=38) hyperinflated larvae, which was significantly different from unablated controls (χ²: 39.69, P<0.0001; Fig. 4B). Wild-type larvae with head-specific ablations exhibited a significant increase in average swim bladder volume (0.016±0.0013 mm³) compared with untreated wild-type siblings (0.0012±0.0013 mm³, n=36) (one-way ANOVA with Tukey post-test, P=0.0365) and statistically comparable swim bladder volumes to that of full CuSO₄-treated larvae (0.016±0.0029 mm³, n=38) (one-way ANOVA with Tukey post-test, P=0.987) (Fig. 4C). Conversely, tail-specific CuSO₄ treatments did not cause hyperinflation in wild-type fish and resulted in statistically similar swim bladder volumes to untreated siblings (0.012±0.0031 mm³, n=34) (one-way ANOVA with Tukey post-test, P=0.835) (Fig. 4C). The lhfpl5b mutant hyperinflation phenotype was not affected by either neomycin or CuSO₄ treatments (Fig. S2). Taken together, we concluded that the anterior region of the lateral line system is critical to surface detection during initial swim bladder inflation.

Surface tension is a characteristic that makes the air–water interface distinct from the rest of the water column. We predicted that larval fish sense this stimulus with their anterior lateral line neuromasts and that by decreasing the interfacial tension we would observe abnormal swim bladder inflation in wild-type larvae while lateral line mutants would be unaffected. The interfacial tension between oil and water is approximately half that of air and water (Johansen, 1924). We layered oil onto the surface of the embryo media and allowed larvae to perform surfacing behaviors between 2 and 6 dpf (Fig. 5). Using this experimental design, we found that the average swim bladder volume of wild-type larvae exposed to an oil–water interface (0.016±0.0029 mm³, n=32) was significantly greater than the control group with access to an air–water interface (0.007±0.001 mm³, n=28) (one-way ANOVA with Tukey post-test, P<0.0001). Furthermore, oil-exposed wild-type larvae had statistically similar swim bladder volumes to that of lhfpl5b mutants (0.014±0.0038 mm³, n=24) (one-way ANOVA with Tukey post-test, P=0.886) (Fig. 5C). Lateral line mutants overfilled their swim bladders to statistically similar proportions when presented with either an air–water or oil–water interface (0.01±0.0026 mm³, n=28) (one-way ANOVA with Tukey post-test, P=0.133). We concluded that surface tension is one of the surface characteristics the lateral line can detect during initial swim bladder inflation.

If hydrodynamic feedback informs the surfacing behaviors of larval fish, then stimulus-independent manipulations of lateral line activity should alter their ability to correctly inflate their swim bladders. To test this, we created the Tg(lhfpl5b2028:ChR2-EYFP-PA) transgenic line that expresses the blue light-activated channel Channelrhodopsin-2 specifically in lateral line hair cells (see Fig. S4). We predicted that optogenetic stimulation of lhfpl5b mutants, whose lateral lines are insensitive to physical stimuli, would reduce the frequency of the hyperinflation phenotype. Consistent with these predictions, under blue-light stimulation, lateral line mutants with the lhfpl5b:ChR2 transgene did not exhibit positive buoyancy and had a significant decrease in average swim bladder volume (0.0046±0.0034 mm³) compared with unstimulated transgenic mutants (0.012±0.0031 mm³) (one-way ANOVA with Tukey post-test, P=0.00127) (Fig. 6B). Transgenic wild-type larvae under blue light exhibited a non-significant decrease in swim bladder volume (0.0052±0.004 mm³) compared with unstimulated transgenic siblings (0.0079±0.002 mm³) (one-way ANOVA with Tukey post-test, P=0.458). To demonstrate that blue light stimulation was reducing swim bladder volumes via activation of lateral line hair cells, we treated wild-type lhfpl5b:ChR2 larvae with CuSO₄. As expected, we observed hyperinflation following CuSO₄ treatment in stimulated wild-type lhfpl5b:ChR2 larvae, with an average swim bladder volume (0.012±0.0037 mm³) that was statistically similar to experimental non-transgenic lhfpl5b mutants (0.011±0.0024 mm³) (one-way ANOVA with Tukey post-test, P=0.998) (Fig. S4B). CuSO₄ treatments in lhfpl5b mutants did not kill hair cells because of the non-functional mechanotransduction channels, but rather acted as a control for potential behavioral side effects of the CuSO₄ treatments. Overall, optogenetic activation of mechanically insensitive lateral line organs in lhfpl5b mutants prevented the hyperinflation phenotype. Together, these data support the hypothesis that the lateral line provides crucial sensory information during initial swim bladder inflation in larval zebrafish.

Our results suggested that larval fish are using their head neuromasts to sense the interfacial tension between air and water at the surface. Even in wild-type zebrafish, the interaction with the surface during initial swim bladder inflation is not well characterized. To study this behavior in more detail, we quantified the number of unique visits to the surface and the amount of time per visit spent within the meniscus region from videos of wild-type (n=7) and mutant (n=14) larvae at 4 dpf. These counts excluded larvae (n=5 total: 1 wild-type and 4 mutant) that under-inflated owing to lack of surfacing events. lhfpl5b mutants that over-inflated their swim bladders made more visits (87.3±17.4) to the surface compared with both wild-type (42.0±16.3) and mutant siblings that inflated to normal proportions (31.9±19.2) (one-way ANOVA with Tukey post-test, P=0.0004) (Fig. 7A). Mutants destined for hyperinflation also spent more time (139.0±31.6 s) at the surface than normally inflated wild-type (51.9±24.7 s) and mutant siblings (42.3±29.5 s) (one-way ANOVA with Tukey post-test, P=0.00006) (Fig. 7B). Since excess surfacing could be due to some mutants initiating the behavior earlier, we also recorded the time of the first surfacing event and found no significant difference between neutrally buoyant and over-inflated larvae (Fig. 7C). We concluded there is a correlation between how larvae interact with the surface during the initial swim bladder inflation period and their resulting swim bladder volume and buoyancy status.

The intake of exogenous air from the surface is a common feature for intitial swim bladder inflation in many species of fish (Furukawa et al., 2022; Goolish and Okutake, 1999; Jones and Marshall, 1953; Korsøen et al., 2012; von Ledebur and Wunder, 1937). However, few studies have considered what sensory information is required to achieve neutral buoyancy. In this study, we describe a new role for the mechanosensory lateral line in mediating initial inflation of the swim bladder in zebrafish (Fig. 8). Our data support a model in which larval zebrafish use their head neuromasts to detect the air–water interface in support of their transition to neutral buoyancy.

Zebrafish lhfpl5b mutants are the first genetic model for the specific loss of lateral line function in any organism (Erickson et al., 2020). These ‘lateral line mutants’ exhibit an uncommon swim bladder hyperinflation phenotype that prompted us to investigate this novel connection between lateral line sensation and initial inflation of the swim bladder (Fig. 1). Previous studies have demonstrated that restricting access to surface air during the initial inflation period prevents swim bladder inflation in larval zebrafish (Goolish and Okutake, 1999; Lindsey et al., 2011). In addition, one study showed that a proportion of solute carrier family 12 member 2 (slc12a2/nkcc1) mutant zebrafish over-fill their swim bladders and that blocking access to the surface prevents this phenotype (Abbas and Whitfield, 2009). This gene is implicated in endolymph production in the inner ear, but further studies are required to determine the cause of over-inflation in slc12a2 mutants. Likewise, we find that blocking access to the air–water interface eliminates over-inflation in lateral line mutants and greatly decreases average swim bladder volumes of both mutant and wild-type larvae (Fig. 2). Our method of blocking access to the air–water interface resulted in a small proportion of larvae (both wild-type and mutant) inflating their swim bladder to normal proportions, although hyperinflation was never observed. Inflation was likely achieved by larvae accessing air bubbles that remain in the water after submerging the filter (Goolish and Okutake, 1999). Overall, the hyperinflation phenotype in lateral line mutants requires an intake of exogenous air, pointing to a sensory defect in lhfpl5b mutants rather than a physiological one.

In support of this sensory-deficit hypothesis, we show the hyperinflation phenotype is prevented by either transgenic rescue of lhfpl5 function specifically in hair cells or optogenetic stimulation of mechanically insensitive lateral line hair cells in lhfpl5b mutants (Figs 3 and 6). Furthermore, chemical ablation of lateral line hair cells in wild-type larvae prior to initial inflation can phenocopy the lhfpl5b mutant. A single treatment with CuSO₄ caused similar swim bladder over-inflation proportions to lateral line mutants (Fig. S2). Repeated neomycin treatments over 3–4 dpf caused the next highest percentage of swim bladder over-inflation and a single neomycin treatment resulted in only a few cases of over-inflation in wild-type larvae. These results agree with previous work showing rapid hair cell regeneration following neomycin treatment and comparatively slower recovery (approximately 24 h for full recovery) after CuSO₄ exposure (Hernández et al., 2006; Mackenzie and Raible, 2012; Murakami et al., 2003; Santos et al., 2006; Venuto and Erickson, 2021). Thus, via multiple genetic and chemical approaches, our data show that lateral line hair cell activity can modulate swim bladder inflation in larval zebrafish.

Ototoxins have been extensively used to study lateral line-mediated behaviors in free-swimming larvae and adult fish (Baker and Montgomery, 1999; Bleckmann, 2006; Coffin and Ramcharitar, 2016; Montgomery et al., 1997; Pavlov and Tyuryukov, 1993). However, we are not aware of any behavioral studies where the lateral line was disrupted prior to larvae achieving neutral buoyancy, likely explaining why this connection between the lateral line and initial swim bladder inflation has not been noted in current literature. Interestingly, an 1896 study predicted a relationship between the lateral line and swim bladder inflation, since disruption of the lateral line in adult goldfish resulted in ‘swelling of the swim bladder’ and positive buoyancy (Richard, 1896). However, it is not clear if the phenotypic similarities between these adult goldfish and our experimental larval zebrafish share the same underlying causes. Regarding roles for the lateral line in early life behaviors, a close parallel to our findings is the report that unhatched red-eyed treefrogs use their lateral line organ to mediate mechanosensory-cued hatching (Jung et al., 2020). In this case, the frog lateral line can sense vibrational stimuli, such as that produced by predators, promoting escape hatching behaviors even before the onset of vestibular function. Similarly, our data suggest that the lateral line is modulating early locomotor activities in zebrafish prior to the development of fully formed vestibular or rheotactic responses (Bagnall and Schoppik, 2018; Mo et al., 2010; Olive et al., 2016). As such, the involvement of lateral line in the surfacing behavior may represent the earliest behavioral role for the lateral line yet described.

We expanded on the full-body lateral line ablation experiments with specific ablations of either the head or truck neuromasts with CuSO₄. Previous studies have shown that head neuromasts are involved with detection of surface features and prey capture (Bleckmann, 1988; Carrillo and McHenry, 2016; Müller and Schwartz, 1982; Schwartz and Hasler, 1966). Indeed, ablation of the head neuromasts in wild-type larvae yielded similar swim bladder hyperinflation levels to larvae that experienced full body treatments. While this specific experiment did not include lhfpl5b mutants, we found that head-treated larvae generally recapitulated the hyperinflation phenotype of lhfpl5b mutants. Ablation of the tail neuromasts had no effect on either swim bladder volume or buoyancy (Fig. 4). These findings suggest that the head neuromasts interact with the air–water interface during surfacing and that the posterior lateral line is dispensable for achieving neutral buoyancy. As such, our findings also provide additional support for the anterior lateral line as a central mediator of surface-related behaviors in fish.

The interfacial tension between air and water represents a unique hydrodynamic stimulus that may be detected by the lateral line. The tension between oil and water is half that of air and water (Johansen, 1924). Thus, by adding a layer of oil to the surface, we predicted that surface detection would become increasingly difficult and result in over-filling of the swim bladder. We found that, when given access to an oil–water interface, wild-type larvae over-fill their swim bladders with oil to a statistically similar volume as lateral line mutants that overinflate with air (Fig. 6). We interpret this result as support for the hypothesis that the lateral line senses interfacial tension between air and water as larvae interact with the surface during initial swim bladder inflation. This interpretation is further supported by the observation that lhfpl5b mutants (which are already insensitive to interfacial tension) exhibit similar swim bladder volumes when exposed to either oil or air. However, a possible complementary interpretation for these data is that larvae that fill their swim bladders with oil may be deprived of the normal feedback on buoyancy levels as they intake from the surface. Since oil is denser than air, oil-filled larvae are negatively buoyant. In this scenario, larvae may continue to visit the surface for air intake in an attempt to achieve neutral buoyancy. However, future studies are required to examine if such a feedback mechanism exists in larval zebrafish.

Since our data suggest that the air–water interface provides a distinct stimulus to the lateral line, we investigated whether surface interactions differed between hyperinflated lhfpl5b mutants and their wild-type counterparts. As predicted, over-inflated mutants took more visits and spent more time at the surface than neutrally buoyant mutant and wild-type siblings (Fig. 7). Excess breaching of the surface occurred despite over-inflated and neutrally buoyant larvae exhibiting similar timing for their first surfacing attempts. While it is unclear whether larvae ingest air during the entire time spent above the air–water interface, it is known that surface breaches are observed during initial inflation (Goolish and Okutake, 1999; Lindsey et al., 2010; Rieger and Summerfelt, 1998). Therefore, the increase in surface visits probably leads to an increase in air deposition to the swim bladder in lateral line mutants.

Interestingly, approximately half of all lateral line mutants inflate their swim bladders to normal proportions and continue to develop into viable adults. To account for this fact, we hypothesize that additional somatosensory, photosensory or mechanosensory systems may compensate for the loss of lateral line sensation and help mutant larvae achieve neutral buoyancy (Riley and Moorman, 2000; Suchocki and Sepulveda-Villet, 2019; Trotter et al., 2003). The lateral line is involved several multi-sensory behaviors, including flow orientation (Bak-Coleman et al., 2013; Coombs et al., 2020; Liao, 2006), prey capture (Carrillo and McHenry, 2016; Nelson et al., 2002; New, 2002) and predator avoidance (Free et al., 2019; Jung et al., 2020; Stewart et al., 2013), and we predict that surfacing can be added to this list.

The photosensory and mechanosensory systems exhibit distinct organizational patterns for primary input to the central nervous system: even the anterior and posterior lateral line ganglia primary projections are somatotopically organized in zebrafish larvae (Alexandre et al., 1999; Fame et al., 2006; Hale et al., 2016; Liao and Haehnel, 2012; McCormick, 1989; Mirjany and Faber, 2011). The Mauthner neurons integrate multisensory input at the level of the hindbrain, but the surfacing behavior is unlikely to use this circuit. For higher-order processing, however, information from the lateral line, vestibular and photosensory systems are integrated in various brain regions. The thalamus is one such sensory integration center in zebrafish (Mueller, 2012). Socially relevant mechanosensory and photosensory information is represented in a thalamic region marked by the expression of parathyroid hormone 2 (pth2) and somatostatin 7 (sst7) (Anneser et al., 2020; Kappel et al., 2022; Venuto et al., 2022). Vestibular information is also represented in the thalamus (Favre-Bulle et al., 2018; Migault et al., 2018), although it is not clear if the exact same region is involved. Photosensory, lateral line and vestibular/auditory information is also integrated in the tectum (Favre-Bulle et al., 2018; Filosa et al., 2016; Isa et al., 2021; Migault et al., 2018). In fact, neurons in the periventricular layer (PVL) of the tectum, which is similar to the superior colliculus in mammals, respond to visual, auditory and lateral line inputs (Ingle, 1973; King et al., 1996; Perrault et al., 2011; Thompson et al., 2016). However, it is hypothesized that most of these modalities are processed in parallel at earlier stages of development (Thompson et al., 2016). It is worth noting that these sensory systems are in developing stages during surfacing behaviors.

Photosensory cues did not alter the buoyancy status of either wild-type or lhfpl5b mutant larvae (Fig. S1), suggesting that photosensory systems play, at most, a subtle modulatory role in the surfacing behavior of larval zebrafish under laboratory conditions. The vestibular system is probably required to coordinate the vertical orientation and climbing portions of surfacing (Bagnall and Schoppik, 2018; Ehrlich and Schoppik, 2017, 2019). Meanwhile, our data support the idea that the lateral line is responsible for the surface detection and breaching aspects of surfacing. However, given that both the vestibular and lateral line systems are providing spatial information during the surfacing behavior, there may be crosstalk between them. For example, the vestibular system must quickly adapt to a dynamic increase in buoyancy over the course of multiple breaching events, so feedback from the lateral line during surface interactions may help fine tune the vestibular sense. However, the link between these systems has not been well investigated during deliberate behaviors such as surfacing (Chagnaud et al., 2017; Liu et al., 2022). So, although sensory integration exists in zebrafish and seems necessary for larvae to achieve neutral buoyancy, the integration dynamics are unclear during the time frame of initial swim bladder inflation and require future investigation.

The authors would like to thank Dr Fadi Issa (ECU) for his support and technical advice. The authors also thank the Issa, Trapani and Erickson lab members for fish care and maintenance. Microscopy was performed with support of the ECU Biology Imaging Core Facilities.