Paternal care regulates the timing, synchrony and success of hatching in a coral reef fish

Abstract

In oviparous species, the timing of hatching is a crucial decision, but for developing embryos, assessing cues that indicate the optimal time to hatch is challenging. In species with pre-hatching parental care, parents can assess environmental conditions and induce their offspring to hatch. We provide the first documentation of parental hatching regulation in a coral reef fish, demonstrating that male neon gobies (Elacatinus colini) directly regulate hatching by removing embryos from the clutch and spitting hatchlings into the water column. All male gobies synchronized hatching within 2 h of sunrise, regardless of when eggs were laid. Paternally incubated embryos hatched later in development, more synchronously, and had higher hatching success than artificially incubated embryos that were shaken to provide a vibrational stimulus or not stimulated. Artificially incubated embryos displayed substantial plasticity in hatching times (range: 80–224 h post-fertilization), suggesting that males could respond to environmental heterogeneity by modifying the hatching time of their offspring. Finally, paternally incubated embryos hatched with smaller yolk sacs and larger propulsive areas than artificially incubated embryos, suggesting that paternal effects on hatchling phenotypes may influence larval dispersal and fitness. These findings highlight the complexity of fish parental care behaviour and may have important, and currently unstudied, consequences for fish population dynamics.

1. Introduction

Hatching represents perhaps the most vulnerable life-history transition in the development of oviparous animals, which makes choosing the optimal time to hatch a crucial determinant of survival. There is broad evidence that embryos can use environmental cues that indicate the likelihood of mortality within or outside the egg (i.e. visual, chemical, or vibrational cues and changes in their physical environment) to make adaptive behavioural decisions regarding when to hatch (reviewed in [1]). In a variety of taxa, embryos will either accelerate or delay hatching in response to environmental cues that indicate the risks (e.g. egg or larval stage predators [2–5], pathogens [6–8] or adverse conditions for development either within or outside the egg [9–11]) and opportunities (e.g. synchronized hatching with siblings [12–14], or favourable conditions for development within or outside the egg [15–17]) associated with hatching. Nevertheless, for immobile embryos in which sensory systems have not yet fully developed and where only limited information is available regarding the external environment, reliably detecting environmental cues and weighing the risks and opportunities of transitioning between life stages is inherently challenging [18,19].

Parents often go to great lengths to provide care that increases the survival and growth of their offspring [20]. In oviparous animals, a variety of pre-hatching parental care behaviours have evolved (e.g. nest guarding, cleaning, fanning and incubation) that improve the survival of embryos and, in turn, the fitness of parents [21]. More recently, however, it has been recognized that parents may also assist and/or regulate the hatching process (termed ‘hatching care’ by Mukai et al. [22]). Parental hatching care is divided into two categories: hatching assistance, where parents decrease the physical burden of hatching on embryos, and hatching regulation, where parents induce hatching either with or without reducing the physical burden on embryos [22]. Hatching assistance, such as when parents help their embryos breakdown the outer egg capsule or excavate buried nests to help free hatchlings, is relatively common across animal taxa ranging from spiders and crabs to reptiles and birds (e.g. [16,23–26]). These parental care behaviours assist embryos that have independently begun the hatching process, without directly influencing the initiation of hatching. By contrast, hatching regulation requires parents to assess environmental cues and make decisions regarding when to induce their embryos to hatch. In some species, parents provide vibrational or chemical cues or change environmental conditions within the nest that induce their embryos to hatch independently (e.g. [16,22,27–29]), while in others, parents physically remove embryos from the egg capsule (e.g. [30,31]). Parents may induce hatching to help their offspring escape embryo stage risks (e.g. [31–33]), or to time hatching in relation to the risks and/or opportunities that their offspring will experience post-hatching (e.g. [15,28,34]). However, despite the phylogenetically diverse examples of the role that parental hatching care can play in the regulation of hatching time, we know little about the influence of parental hatching care on the timing of hatching in fishes.

Ray-finned fishes are the most speciose group of vertebrates, and approximately 25% of fish species provide parental care to their offspring [35]. Pre-hatching parental care behaviours take a wide variety of forms in fishes, ranging from nest construction and guarding, to tending, fanning, and mouthing the clutch as embryos develop, and, in rare cases, parents also continue to defend and provision their offspring post-hatch [36,37]. However, there are remarkably few examples of hatching care in fishes. In some demersal spawning cryptobenthic fishes (defined as cryptic small-bodied, bottom-dwelling fishes that typically have high mortality rates and whose populations require continuous replenishment from local larval recruitment [38,39]) parents appear to assist in the hatching process by increasing the time spent fanning or mouthing their embryos during hatching [26,40]. This increase in fanning behaviour helps embryos break down their chorion in hairy blennies Rhabdoblennius nitidus [26]. In the same species, males may remove the last remaining embryos of the clutch and ‘spit’ out hatched larvae, thus potentially increasing hatching success [26]. Moreover, mudskippers in the genus Periophthalmus have been reported to induce hatching by flooding the air-filled chamber where their clutch is located within their burrow. Males flood the burrow during high tide, which allows hatchlings to exit the burrow and be transported to deeper water where they complete larval development [28,41]. These examples suggest that parental hatching care could play an important role in regulating hatching time in demersal spawning fishes, with unknown consequences for the morphology and dispersal potential of fish larvae.

We investigated parental hatching care in the sponge-dwelling neon goby Elacatinus colini, a cryptobenthic coral reef fish endemic to the Belizean barrier reef. Elacatinus colini typically live within tube sponges (figure 1a,b) located on lagoonal patch reefs where they occur in male/female breeding pairs (figure 1c) or large groups that are likely composed of a male and multiple females (figure 1d). Breeding pairs lay their eggs in a monolayer on the inner wall at the base of a sponge tube (figure 1c). Males provide pre-hatching parental care to the embryos by guarding against predators, fanning the clutch to oxygenate the embryos, and mouthing the clutch to remove detritus and dead or diseased embryos (J.E.M. 2014, personal observation). Upon hatching (7–8 days post-fertilization (dpf) [42]), E. colini larvae immediately leave the sponge to complete a 28–58 day pelagic larval stage [42,43]. During an initial study [42], we observed that E. colini embryos hatched immediately and up to several days early when they were moved from their parent's aquarium to a separate larval rearing system. Further, when the embryos were left to hatch in the parent's aquarium, male E. colini were observed ‘spitting’ hatched larvae from the entrance of their spawning shelters. Taken together, these observations suggest that E. colini may provide an interesting system for studying hatching plasticity and the role that male parents may play in assisting or regulating the hatching process in a cryptobenthic coral reef fish.

Figure 1.

Natural context for parental hatching care and hatching plasticity in the sponge-dwelling neon goby, Elacatinus colini. (a) In situ photograph of a pink vase sponge (Niphates digitalis). (b) Elacatinus colini live within the sponge tube and occur in social groups ranging in size from (c) an individual breeding pair to (d) a large colony with a single male tending embryos at the base of the sponge and many smaller females or subadults clustered around the osculum. Brittlestars (Ophiothrix sp.), crabs, shrimp and other conspecifics are common co-inhabitants of sponges (Majoris, personal observation). White arrow indicates an Ophiothrix sp. brittlestar. Symbols indicate a male (♂) and female (♀) E. colini. (Online version in colour.)

Here, we use a laboratory experiment to demonstrate that male E. colini actively regulate when their embryos hatch. In doing so, males influence the age, synchrony, hatching success and morphological traits of hatchlings, thus profoundly influencing the potential fitness of their offspring. Specifically, we tested whether male E. colini actively regulate the timing of hatching by comparing the hatching process of embryos that were paternally incubated with artificially incubated embryos that were non-stimulated, and artificially incubated embryos that were experimentally stimulated (i.e. shaken) to assess their capacity to accelerate hatching. Furthermore, we examined the morphological consequences of paternal hatching regulation by comparing hatchling phenotypes among treatments.

2. Methods

(a) . Study population

We investigated the potential for paternal hatching care and embryo hatching plasticity to occur in a captive breeding population of the sponge-dwelling neon goby Elacatinus colini. All breeding pairs of E. colini (n_pairs = 15; e.g. figure 1c) were collected from sponges on lagoonal patch reefs located off the coast of central Belize, then transported to Boston University where each wild-caught pair was housed in a 76 l aquaria connected to a recirculating seawater filtration system. The aquarium room was maintained on a 12 L : 12 D light cycle. To simulate sunrise and sunset, overhead lights in the aquarium room increased gradually in brightness from 07.45 to 08.00 and decreased in brightness from 19.45 to 20.00. Weekly 50% water changes were completed by removing water from the reservoir and refilling the system with artificial seawater. The water quality parameters were monitored daily and maintained at a temperature of 27–28°C, salinity 35–37 ppt, pH 8.1–8.4, NH₃ < 0.25 ppm, NO₂ < 0.25 ppm, NO₃ < 5 ppm.

(b) . Observation of hatching time

(i) . Paternally incubated embryos

To test the hypothesis that parents provide hatching care (assistance and/or regulation), E. colini breeding pairs were conditioned to spawn in acrylic shelters that were open on one side and had a removable polyvinyl chloride (PVC) insert (figure 2a). The shelters were attached by suction cups to the front wall of the aquaria so that parental care behaviour and embryo development could be observed through the glass. An opaque black plastic cover was placed on the outside glass to provide a dark, enclosed space within each shelter. To determine the approximate time of day when spawning occurred (henceforth ‘spawning time’), the shelters were checked for new clutches daily at 06.00, 14.00 and 22.00. On average, E. colini pairs began spawning within 12.2 ± 10.3 days, they laid clutches of 168 ± 83 eggs every one to two weeks, and the embryos developed for 6–7 days before hatching [42].

Figure 2.

Male sponge-dwelling neon gobies regulate hatching. (a) A male Elacatinus colini tending a clutch of embryos within a spawning shelter. White dashed boxes indicate the field-of-view of the infrared cameras. (b–g) Images of paternal hatching care extracted from videos. One camera was focused on the clutch of embryos within the spawning shelter (b–d), and a second camera was focused on the entrance to the spawning shelter (e–g). To induce hatching, (b) a male goby selected a patch of embryos and (c,d) physically removed the embryos from the clutch using its mouth. Black dashed ovals indicate a patch of embryos removed by the male goby. (e) After removing a patch of embryos, the male swam to the entrance of the shelter holding hatchlings in its mouth, and (f,g) spat free-swimming larvae into the water column. White dashed circles indicate a group of larvae spat out by the male goby. Males repeated this behaviour until they hatched all of the embryos in the clutch. (h) Age and time of day when male neon gobies induced hatching, stratified by the time of day when the clutch of embryos was initially observed in the spawning shelter (06.00, 14.00 or 22.00). White endpoints indicate the age of the first and last larva that hatched from a clutch. White shading indicates daylight, grey shading indicates darkness in the laboratory. See online supplementary information for representative videos (electronic supplementary material, videos S1–S2). (Online version in colour.)

To observe parental care behaviours and the timing of hatching, the plastic cover on the outside of the aquarium was replaced with a pair of custom-made infrared video cameras 1–2 days before the embryos were expected to hatch. Each camera was equipped with infrared LED lights (940 nm) that provided illumination within the spawning shelter, invisible to the fish. One camera was focused on the developing embryos inside the shelter, and a second camera was focused on the entrance to the spawning shelter. The video recordings were checked every 8 h at 06.00, 14.00 and 22.00 to determine whether the embryos had hatched, and the aquarium was checked using a flashlight to determine if any larvae were swimming in the water column. If hatching had occurred, larvae were sampled from the water column and photographed for morphological analyses. Each video recording was watched to observe parental care behaviour during hatching and to determine the beginning and end of the hatching process. These videos were also used in the paternally stimulated (PS) hatching treatment described in the ‘Manipulation of hatching time’ section below.

(c) . Manipulation of hatching time

(i) . Artificially incubated embryos

To test the hypothesis that embryos accelerate hatching time in response to vibrational cues, acrylic flow chambers (henceforth ‘flumes’; dimensions: L 30 cm, W 5 cm, H 5 cm) were used to artificially incubate E. colini embryos and collect hatchlings. The flumes were designed to fit the removable PVC inserts on which E. colini breeding pairs attached their eggs within the spawning shelters and to provide a gentle water current over the clutch to artificially incubate the embryos. When embryos hatched, the water current pushed the free-swimming larvae over a spillway and into a collection cup at the downstream end of the flume. Collection cups were fitted with 75 µm mesh screens that retained larvae within the cup while allowing water to exit the cup and return to a temperature-controlled reservoir that was equipped with mechanical filtration and a UV sterilizer (henceforth ‘hatching system’). Daily 50% water changes were completed by removing water from the reservoir and refilling the system with artificial seawater. The water quality parameters within the hatching system were monitored daily and adjusted to match the parameters of the breeding system: temperature 27–28°C, salinity 35–37 ppt, pH 8.1–8.4, NH₃ < 0.25 ppm, NO₂ < 0.25 ppm, NO₃ < 5 ppm.

Male E. colini provides pre-hatching parental care by defending, fanning and mouthing the embryos during the entire 7–8 day incubation period. During pilot trials, we observed that embryos were more likely to survive until hatching if the clutch was cared for by the male for the first 48 h before being transferred to a flume for artificial incubation, than if the clutch was moved to a flume immediately after it was first observed in the spawning shelter. Therefore, for all treatments we allowed the male to incubate the clutch for 48 h before the PVC insert was removed and transferred to a flume that was submerged within the parent's aquarium. The flume and clutch were then immediately removed from the parent's aquarium, photographed to determine the total number of embryos in the clutch, and connected to the hatching system to begin artificial incubation. The flow rate within the flume was calibrated to 5 cm s⁻¹ (300 ml min⁻¹), which was approximately the mean flow rate measured from tube sponges at our field site in Belize (J.E.M. 2013, unpublished data) and resulted in continuous but gentle agitation of the embryos in the clutch. Once the flow rate was calibrated, an opaque black cover was placed on the flume to simulate the enclosed conditions of the parents spawning shelter and overhead lighting was provided on a 12 L : 12 D light cycle from 08.00 to 20.00.

To experimentally test the hypothesis that embryos accelerate hatching in response to vibrational cues, we compared the hatching time of artificially incubated embryos that were non-stimulated (NS), and artificially incubated embryos that were experimentally stimulated (ES) to hatch, with paternally incubated embryos that were paternally stimulated (PS – observation protocol described in the ‘Paternally incubated embryos' section above). When a clutch was first observed in the spawning shelter (06.00, 14.00 or 22.00), it was randomized into either NS, ES or PS treatments. A single clutch from each breeding pair was used in each treatment (NS: n_clutches = 12, ES n_clutches = 11, PS: n_clutches = 14). Of the 15 breeding pairs in our study population, 10 pairs contributed a clutch to all three treatments while 5 pairs contributed a clutch to only one or two treatments. Data from all clutches were used in our statistical analyses.

(ii) . Non-stimulated hatching

Clutches assigned to artificially incubated non-stimulated (NS) treatments were transferred to a randomly assigned flume and position within the hatching system to control for potential tank or spatial effects on hatching time. Then, a cover was placed on the flume and the clutch was left undisturbed until the trial was complete. Once the first embryo hatched from a clutch, an NS treatment was considered complete after three consecutive sampling intervals (i.e. 24 h) without additional hatching. After this period, the lid was removed from the flume and any unhatched embryos were photographed and counted.

(iii) . Experimentally stimulated hatching

After a clutch assigned to the artificially incubated experimentally stimulated (ES) treatments was transferred to the hatching system, forceps were used to vigorously shake the PVC insert on which the clutch was attached for 1 min—alternating between two cycles of shaking the clutch from side-to-side for 15 s, front-to-back for 15 s—then a cover was placed on the flume. This approach provided a standardized, repeatable vibrational cue that induced a strong hatching response without specifically emulating the frequency or intensity of a vibrational cue that embryos may experience in their natural environment. The clutch was shaken every 8 h at 06.00, 14.00 or 22.00. Once embryos began to hatch, an ES treatment was discontinued after three consecutive sampling intervals (i.e. 24 h) without additional hatching and any unhatched embryos were photographed and counted.

(iv) . Sampling intervals

To quantify the effect of the experimental treatments on hatching time and larval phenotypes, collection cups from both the NS and ES treatments were removed and replaced with an empty cup at the start of each 8 h sampling interval, and this procedure was repeated 30 min later. This approach resulted in an initial 30 min sample that collected larvae of known age that had hatched spontaneously in the NS treatment, or rapidly in response to the experimental stimulus in the ES treatment, and a subsequent 7.5 h sample that collected larvae of less precise age that took longer to hatch following the experimental-stimulus or hatched spontaneously in the interval between experimental stimuli. Larvae that hatched in each interval were counted and a subset of individuals that were collected during the 30 min sampling interval was photographed for morphological analyses.

(d) . Quantifying the timing, synchrony and success of hatching

To compare the timing of hatching among treatments, we calculated the proportion of a clutch that hatched during each 8 h sampling interval (i.e. number of larvae collected/total number of embryos in a clutch), then determined the mean proportion of embryos that hatched during each interval among clutches within each treatment. For the PS treatment, we watched each video recording to determine the onset of hatching for each clutch. To compare hatching age among treatments, we calculated the mean hatching age of embryos within a clutch, then determined the mean hatching age of clutches in each experimental treatment (NS, ES, PS). To compare hatching synchrony among treatments, we calculated the hatching duration for the NS and ES treatments as the time between the first and last sampling interval in which larvae were collected. For PS treatments, each video recording was watched to determine the age of the first and last embryo that hatched from each clutch. These time points were used to calculate the hatching duration of each clutch. To compare hatching success among treatments, we determined the proportion of the clutch that hatched by dividing the total number of larvae collected during the experiment by the total number of embryos in the clutch. All embryos that were visible in the PS video recordings were removed by the male during the hatching process; therefore, we recorded this as complete hatching success.

(e) . Quantifying the effects of variation in hatching time on larval phenotypes

Live larvae from each treatment (NS, ES and PS) were anaesthetized with MS-222 (250 mg l⁻¹) and photographed in lateral view using a digital camera (Canon PowerShot S5 IS) attached to a dissection microscope (Zeiss Stemi-2000). We measured the total length (TL), eye diameter (ED), yolk sac area (YA) and propulsive area (PA—area of the body and fins excluding the head and gut region) from the images of larvae that were sampled at the mean hatching age of each clutch using ImageJ v. 1.47v [44].

(f) . Statistical analyses

All statistical analyses were performed in R v. 4.1.0 [45]. We used generalized linear Bayesian mixed models to determine whether ‘hatching age’, ‘synchrony’ and ‘success’ were related to ‘treatment’ and/or ‘spawning time’ (and their interaction) while accounting for the lack of independence among clutches from the same parents by including ‘parent ID’ as a random effect [46]. Specifically, to analyse hatching age, we used the age of each hatchling as a replicate (n = 5938) and ran the model using a negative binomial error structure, with an interaction effect between treatment and spawning time. The leave-one-out-cross-validation information criterion (LOOIC) [47] indicated that a model that included the interaction effect outperformed the model without an interaction effect. For hatching synchrony, we calculated the total duration of hatching for each clutch (n = 37) and again ran a model with a negative binomial error structure. LOOIC indicated no clear improvement of the model with the inclusion of an interaction effect between treatment and spawning time. Therefore, we omitted the interaction effect in the final model. Finally, for hatching success, we ran a model testing the effect of treatment and spawning time on the proportion of hatched larvae per clutch (n = 37) using a beta distribution. Since the beta distribution is bounded between 0 and 1, we applied a small adjustment to convert 1 s and 0 s to 0.9999 and 0.0001, respectively. We ran the beta regression model without the interaction effect since the LOOIC suggested no improvement. For all Bayesian regression models, we used flat, non-informative priors and ran four chains of 5000 iterations, with a warm-up of 2500 iterations. We checked chain convergence visually and used fitted values to visualize model outcomes.

To compare hatchling phenotypes among treatments, a principal components analysis (PCA) was conducted with the scaled and centered larval morphometrics (total length, eye diameter, yolk sac area, propulsive area) using the ‘FactoMineR’ package [48]. Morphological differences among the treatment groups (NS, ES and PS) were assessed using a permutational multivariate analysis of variance (PERMANOVA) completed using the ‘adonis’ function from the ‘vegan’ package [49]. Pairwise PERMANOVAs with Bonferroni p-value corrections for multiple comparisons were used to identify differences in morphological traits among treatment groups.

3. Results

(a) . Male gobies assist and regulate hatching

Infrared video recordings revealed that all of the male Elacatinus colini in our study population provided hatching care by physically removing embryos from the clutch with their mouth (figure 2b–d), transporting hatchlings to the entrance of the spawning shelter (figure 2e), and spitting free-swimming larvae into the water column (figure 2f,g; see electronic supplementary material, videos S1–S2). This sequence was repeated until all of the embryos in the clutch had hatched. Males invariably induced hatching 7 days post-fertilization within approximately 2 h of artificial sunrise in the laboratory (artificial sunrise: 07.45–08.00; initiation of hatching: mean = 07.02, range = 06.05–08.08). While the timing of paternally regulated hatching was highly consistent, the age of embryos at hatch varied by up to 15.8 h (age at hatch: range = 153.4 h post-fertilization [hpf] – 169.2 hpf) depending on the time-of-day when a breeding pair had spawned.

(b) . Hatching care affects the age, synchrony and hatching success of embryos

The pattern of hatching differed among treatment groups and spawning times (figure 3a). Relative to the mean hatching age of embryos in the PS treatment, embryos in the ES and NS treatments hatched 14.9% and 4.7% earlier, respectively. The youngest embryo hatched at 80hpf from a clutch in the ES treatment, representing a 50.7% acceleration in hatching age relative to the mean hatching age of embryos in the PS treatment. The oldest embryo hatched at 224 hpf from a clutch in the NS treatment, representing a 38% delay relative to embryos in the PS treatment.

Figure 3.

Effect of experimental treatments on hatching patterns, age, synchrony and success in sponge-dwelling neon gobies. (a) Mean proportion of a clutch that hatched in each 8-h sampling interval compared among treatment groups and stratified by the time of day when clutches were first observed. (b) Mean age of embryos at hatch, (c) duration between the first and last embryo to hatch from a clutch, and (d) proportion of the embryos in a clutch that hatched successfully compared among treatments. Violin plots indicate the distribution of the posterior fitted values and their 95% intervals, while points represent the raw data. Treatments: ES—experimentally stimulated; NS—non-stimulated; PS—paternally stimulated. (Online version in colour.)

(i) . Hatching age

The Bayesian mixed model indicated that there was an effect of treatment, spawning time and their interaction on hatching age (figure 3b). Both NS and PS treatments had a 100% probability of producing older hatchlings than the ES treatment, with PS, on average, producing the oldest hatchlings (median posterior density estimate = 169.3 hpf, lower 95% credible interval = 166.3 hpf, upper 95% credible interval = 172.4 hpf) and NS producing intermediate-aged hatchlings (156.9 hpf [154.1, 159.7]). Similarly, spawning times at 14.00 and 22.00 both had a 100% probability of decreasing hatchling age relative to embryos spawned at 6.00, with embryos spawned at 22.00 predictably being the youngest hatchlings (127.8 hpf [125.2, 130.6]). The interaction effect indicated that the PS treatment resulted in the strongest effect of spawning time. While the difference in age between embryos spawned at 6.00 and 22.00 was 11.1% in ES clutches and only 2.4% in NS clutches, PS clutches decreased in age by 22.2% (figure 3b).

(ii) . Hatching synchrony

Hatching in the ES and NS treatments occurred gradually over several days, with peak hatching typically occurring in the evening sampling intervals (figure 3a). By contrast, embryos in the PS treatment hatched exclusively in the morning on the 7th day of development (figure 2h and figure 3a). The Bayesian mixed model confirmed that the PS treatment (16.6 min [9.6, 29.0]) had a 100% probability of decreasing hatching duration compared to the ES treatment (3137 min [1955, 5364]). By contrast, there was no clear difference between the other treatments as the NS treatment (3536 min [2165, 6298]) had only a 70.6% probability of increasing the duration of hatching compared to the ES treatment (with their parameter estimates broadly overlapping; figure 3c). Clutches spawned at 22:00 h had a 96.5% probability of having longer hatching durations compared to clutches spawned at 6.00, while the 14.00 spawning time had no clear effect (55.1% probability of increased duration).

(iii) . Hatching success

Artificially incubated embryos in both the ES and NS treatments developed normally and hatched without parental assistance. However, there was a 99.8% probability that the PS treatment (proportion hatched: 98.3 [96.0, 99.4]) increased hatching success compared to the ES treatment (94.4 [88.6, 97.5]), while the NS treatment (84.5 [73.2, 91.9]) had a 99.5% probability of decreasing hatching success (figure 3d). The spawning time at 22:00 h (97.9 [93.5, 99.5]) had a 97.6% probability of positively affecting hatching success compared to the 6:00 h spawning time, while the 14:00 h spawning time (95.0 [89.3, 97.9]) showed no clear effect (65.1% probability of a positive effect; figure 3d). Embryos that remained unhatched at the end of ES and NS treatments had typically developed an opaque, milky-white coloration which suggests that the embryos died from a fungal or bacterial infection.

(c) . Larval phenotypes differ among treatment groups

Differences in larval morphology were observed among the three treatment groups (figure 4). A PCA of the morphological traits of hatchlings indicated that PC1 and PC2 explained 84.4% of the variation in morphology among larvae (figure 4a). Scores on PC1 were primarily driven by total length, propulsive area and eye diameter, whereas scores on PC2 were primarily driven by yolk sac area (electronic supplementary material, table S1). Convex hulls overlaid on the PCA biplot identified clustering of larvae by treatment groups. A PERMANOVA indicated that there was strong evidence for an effect of ‘treatment’ (PERMANOVA: F = 9.45, d.f. = 2, R² = 0.11, p = 0.0001; figure 4a), and the interaction between ‘treatment × spawning time’ on larval morphology (PERMANOVA: F = 9.84, d.f. = 3, R² = 0.18, p = 0.0005), but no independent effect of ‘spawning time’ (PERMANOVA: F = 1.87, d.f. = 2, R² = 0.02, p = 0.4242). Hatchlings from ES and PS treatments occupied non-overlapping morphospaces (pairwise PERMANOVA: p = 0.0006), and larvae from NS and PS treatments also differed in morphology (pairwise PERMANOVA: p = 0.0003), primarily driven by differences in yolk sac area, total length and propulsive area. However, there was only weak evidence that larvae from ES and NS treatments differ in morphology (pairwise PERMANOVA: p < 0.0477; figure 4).

Figure 4.

Effect of experimental treatments on the phenotype of newly hatched larvae. (a) Biplot of the first and second dimensions of a principal components analysis of standardized larval morphometrics. Convex hulls illustrate the morphospace of larvae in each treatment group; vectors indicate the loadings of the morphological traits to the principal components, with the length of the arrows indicating strength of the loading. (b) Photomicrographs of larvae collected at the mean hatching age of a representative clutch of embryos from each treatment group. White arrows point to the location of the remaining yolk sac. Morphometrics: TL—total length, ED—eye diameter, PA—propulsive area, YA—yolk sac area. Treatments: ES—experimentally stimulated; NS—non-stimulated; PS—paternally stimulated. (Online version in colour.)

4. Discussion

In oviparous species with parental care, hatching plasticity provides parents with an opportunity to determine the optimal time for their embryos to hatch. Here, we demonstrate that male neon gobies both assist and regulate hatching by removing embryos from the clutch with their mouths and spitting free-swimming larvae into the water column. Independent of the time of day that embryos were laid, all males in our breeding population synchronized hatching within 2 h of sunrise on the 7th day post-fertilization (dpf). Paternally incubated embryos hatched later in development, more synchronously, and had a higher hatching success rate than embryos in artificially incubated treatments that were either non-stimulated or experimentally stimulated. While the spawning time strongly affected the age of hatchlings in paternally regulated clutches, there was no evidence that spawning time affected the hatching duration, success or morphology of larvae. Artificially incubated embryos hatched without parental assistance, and, on average, accelerated hatching relative to the mean hatching time of embryos in the paternally incubated treatment. However, their hatching was spread out over longer periods, generally less successful, and was associated with morphological tradeoffs across life stages. Thus, paternal regulation of hatching may substantially influence the survival and dispersal potential of neon goby larvae and could play a crucial role in the early life history of other demersal spawning fishes.

Examples of parental hatching care are common across taxa and have been characterized as behaviours that either assist or regulate the hatching process [29]; however, there are few examples of these parental care behaviours in fishes. Sano et al. [26] demonstrated that male hairy blennies assist hatching by vigorously fanning their embryos to help break down the egg capsule, a behaviour that could also induce their embryos to initiate hatching. These fishes occasionally removed embryos from the clutch to spit out hatched larvae. However, this behaviour appears to play a secondary role in the hatching process of hairy blennies as it was only performed to remove a few embryos that remained unhatched at the end of the hatching period. By contrast, we found that male neon gobies provided both hatching assistance and regulation by physically hatching all of the embryos in their clutch at sunrise on the 7th day of development, which, to our knowledge represents the first clear evidence of active hatching regulation in coral reef fishes. In addition to regulating the time of day that embryos hatched, paternally incubated clutches hatched later in development and had higher hatching success rates than those that were artificially incubated. This suggests that, despite having the ability to hatch independently, embryos that received paternal care waited until male gobies initiated the hatching process. The wide window of hatching competence in this species (range: 80–224 hpf) suggests that male gobies may have the ability to either accelerate or delay hatching in response to environmental cues that indicate the optimal time for their embryos to hatch. Many reef fishes engage in vigorous fanning and mouthing activity on the night of hatching [50]. If these behaviours are a form of parent–offspring communication that induces hatching [51], then parental hatching regulation could be common in demersal spawning fishes.

Despite goby embryos having a wide window of hatching competence, there was little variation in the time of day when males hatched their embryos. This suggests that, under natural conditions, there is likely a fitness advantage associated with inducing hatching shortly before sunrise. Newly hatched reef fishes are particularly vulnerable to predation as they transition from the benthos to the pelagic environment where they complete larval development. Hatching after dark is thought to be an adaptive behaviour that helps larvae avoid reef-based planktivorous predators [52], and may help to explain why many reef fishes hatch shortly after sunset [40,50]. For E. colini, in contrast, hatching during sunrise could induce the phototactic hatchlings to immediately swim up and away from the reef before planktivores begin feeding. However, artificially incubated embryos that hatched independently typically chose to hatch in the evening. This may indicate that the optimal hatching time shifts in the absence of paternal care, or that embryos may have difficulty independently assessing cues that indicate the optimal time to hatch. The adaptive value and natural context of hatching plasticity in this species will require further experimental evaluation.

While embryos that received paternal care waited until males initiated the hatching process, our artificial-incubation and stimulation experiment revealed that neon goby embryos can hatch independently in response to vibrational stimuli and have a wide window of hatching competence. Yet, these embryos hatched in smaller numbers, at earlier ages, and less synchronously than paternally incubated embryos. While reduced hatching success naturally compromises fitness because unhatched embryos eventually perish, hatching early, less synchronously and underdeveloped could benefit neon gobies, and other demersal spawning species if it allows embryos to escape immediate mortality from egg-stage risks. In other taxa, early hatching has been documented in response to risks associated with embryo stage predators [3–5], variation in the quality of parental care [10], hypoxia [53], dehydration [9,54] and fungal infections [6–8]. Embryos that were artificially incubated and experimentally stimulated hatched early and rapidly in response to vibration cues that may mimic, for instance, parental induction or a predator attack, while embryos in artificially incubated non-stimulated clutches may have hatched early in response to cues that indicated parental absence or suboptimal developmental conditions. Further, the unhatched embryos in several artificially incubated clutches developed an opaque milky white coloration indicative of a fungal or bacterial infection. Neighbouring embryos may have hatched early and asynchronously to escape this dangerous but relatively slow-acting embryo-stage threat [7]. Plasticity in hatching age suggests that E. colini embryos can independently accelerate or delay hatching in response to a variety of environmental cues that indicate the optimal time to hatch.

Early hatching is often associated with tradeoffs across life-history stages. In both red-eyed treefrogs and glass frogs, in which embryos hatch early to escape embryo-stage risks, newly hatched tadpoles are less well-developed, compromising their diving ability and resulting in increased predation rates [55,56]. In our study, goby embryos that hatched younger without paternal regulation had large yolk sacs which provide a nutritional reserve that extends the period before they need to feed exogenously (approx. 8–12 h). By contrast, older hatchlings with minimal yolk reserves must begin feeding immediately [42]. However, younger hatchlings were also smaller in size and had smaller propulsive areas than embryos that hatched older following paternal care, which suggests that they may have weaker swimming abilities that could reduce their ability to capture prey, escape predators and influence their dispersal [57]. For neon gobies (and probably many other cryptobenthic species), dispersal appears to be spatially limited [58] and results in highly localized replenishment of adult populations [39]. Increased larval competency at hatching, as conferred by paternal regulation of hatching age, may underpin these dynamics by ensuring that hatchlings are able to influence their dispersal trajectories [59].

Beyond hatchling age and resulting morphological trade-offs, synchronous hatching in other taxa that do not provide post-hatching parental care allows hatchlings to avoid predation by cannibalistic conspecifics [33], overwhelm predators as they emerge from the nest [60], and form cooperative groups that improve foraging [61]. In neon gobies, males drastically increased hatching synchrony by rapidly hatching all of the embryos in their clutch, transporting hatchlings to the entrance of the spawning shelter and spitting free-swimming larvae into the water column. Synchronizing when their embryos hatch may help to increase their offspring's chance of overwhelming predators and cannibalistic conspecifics (e.g. figure 1d; J.E.M. 2014, personal observation) as they leave the sponge and swim towards surface waters, thus increasing the number of offspring that survive the first moments of their free-swimming larval stage. Furthermore, synchronous hatching could provide a mechanism for offspring to disperse in cohesive sibling groups [62]. However, it should be noted that genetic evidence suggests that spatial patterns of relatedness in reef fishes can be explained by limited dispersal distances alone, without invoking active sibling cohesion during dispersal [63,64].

5. Conclusion

The embryos of demersal spawning cryptobenthic fishes, many of which are laid deep within crevices or burrows on the reef, may have limited access to external cues that would indicate the optimal time to hatch. Thus, parental induction of hatching could offer an alternative mechanism for assessing environmental cues to optimize the timing and synchrony of hatching. We demonstrate that, in sponge-dwelling neon gobies, paternal regulation of hatching results in distinct larval phenotypes that may improve their offspring's ability to survive and influence their dispersal. For cryptobenthic species with specialized benthic microhabitat preferences [43] and high population turnover [39], paternal hatching regulation could represent an adaptation that improves larval retention and local population replenishment. Furthermore, the potential for paternal regulation to result in distinct larval phenotypes and parent–offspring conflicts over the timing of hatching make E. colini a promising study system for understanding the early life-history dynamics and behavioural ecology of cryptobenthic fishes.

Ethics

All applicable international, national and/or institutional guidelines for the care and use of animals were followed. All procedures with live animals were reviewed and approved by the Boson University IACUC (protocol nos 13-021 and 10-036). Collection and export of live fish were approved by Belize Fisheries (research permit no. 000018-13, export permit no. GEN/FIS/15/04/2013 (39) VOL.IX).

Data accessibility

Data and code for statistical analyses are available from the Dryad Digital Repository (https://doi.org/10.5061/dryad.ngf1vhhx8) [65].

The PCA results table and videos illustrating paternal hatching care are provided in the electronic supplementary material [66].

Authors' contributions

J.E.M.: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, visualization, writing—original draft, writing—review and editing; F.A.F.: data curation, investigation, methodology, writing—review and editing; C.M.B.: data curation, investigation, methodology, writing—review and editing; S.J.B.: formal analysis, visualization, writing—review and editing; K.M.W.: conceptualization, methodology, supervision, writing—review and editing; P.M.B.: conceptualization, formal analysis, funding acquisition, investigation, methodology, project administration, supervision, writing—review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Conflict of interest declaration

We declare we have no competing interests.

Funding

Funding support was provided by an NSF grant to P.M.B. (grant no. OCE-1260424), and a Dana Wright Summer Research Fellowship awarded by the Boston University Marine Program to J.E.M.