Rapid embryonic development supports the early onset of gill functions in two coral reef damselfishes

Gill formation is an essential process in the early developmental stages of many aquatic organisms including fish, for which the gills are the primary site for vital physiological processes, including oxygen (O₂) uptake, iono/osmoregulation, acid–base regulation and ammonia excretion (Evans et al., 2005). Gills also provide a secondary site for hormonal control, immune defence and food capture (Evans, 1998). It is therefore surprising that, for teleost fishes, the structure and function of gills during early life stages are relatively understudied when compared with adult life stages. As far back as the 1980s, studies started exploring the onset of gill function at early life stages (Burggren and Pinder, 1991; Rombough, 1988; Kimmel et al., 1995) and the external and internal drivers of gill formation during development (Brauner and Rombough, 2012; Rombough, 2007; Baker et al., 2015; Rombough, 2002; Fu et al., 2010; Rombough and Drader, 2009; Rombough, 1999, 1998). These efforts are already showing applied importance in conservation biology, where exposure to environmental stressors – during any life stage – can lead to impairment of gill structure and function (Hess et al., 2017, 2015; Berry et al., 2016). However, our understanding of gill formation and its developmental significance across teleost fishes is still emerging.

Currently, there are two hypotheses explaining the first function of the developing teleost gill. The first, proposed by Krogh (1941), suggests that gills develop to meet the O₂ demands of a developing fish (Krogh, 1941). During early development, fish embryos and larvae absorb O₂ across the skin and yolk-sac epithelium directly into the tissues or the plasma (Feder and Burggren, 1985). However, this mode of O₂ uptake becomes less sustainable as the surface area to volume ratio (SA/V) of embryos declines with growth (Rombough and Moroz, 1997). Thus, the oxygen hypothesis argues that limitations to O₂ uptake through cutaneous pathways necessitates gill development (reviewed in Rombough, 2007). The second hypothesis is the ionoregulatory hypothesis, which suggests that the first function of fish gills is for ion transport to achieve acid–base, ion, water and ammonia regulation (reviewed by Rombough, 2007; Brauner and Rombough, 2012; Burggren and Dubansky, 2018). Cutaneous iono/osmoregulatory processes become insufficient before processes associated with respiration, and this is because cutaneous iono/osmoregulatory processes face the same SA/V constraints as respiration, but are further restricted by ion transporters localized within specialized cells known as ionocytes (Li et al., 1995). Functioning ionocytes require an apical membrane that is exposed to the external environment and a basal membrane directly contacting the blood (Alderdice, 1988). As the skin thickens, the likelihood of ionocytes reaching both the external environment and the blood significantly decreases, thus limiting the transport capacity of ionocytes before limitations to respiration occur (Alderdice, 1988). Ionoregulatory mechanisms functioning at the gills prior to respiratory functions have been found in all investigated fishes to date (Rombough, 2002; Fu et al., 2010; Rombough, 1999; Hiroi et al., 1998; Zimmer and Wood, 2015; Zimmer et al., 2014). Iono/osmoregulation, therefore, is increasingly thought to be the primary process underpinning gill development.

The majority of studies investigating the drivers of gill development have focused on model species such as tilapia Oreochromis mossambicus (Li et al., 1995), zebrafish, Danio rerio (Rombough, 2002), and, most recently, the commercially important rainbow trout, Oncorhynchus mykiss (Fu et al., 2010; Zimmer and Wood, 2015; Zimmer et al., 2014). Nevertheless, these species do not represent all fishes, especially fishes living in contrasting and challenging ecosystems. Coral reef fishes, during early life stages, often exhibit extreme physiological and demographic characteristics that remain poorly understood. For example, clutch mortality rates during embryogenesis can approach 100% (Emslie and Jones, 2001). During pelagic larval stages, mortality remains extremely high (Goatley and Bellwood, 2016); yet, coral reef fishes have the fastest sustained swimming speeds and highest O₂ uptake rates (Ṁ_O2) relative to their body size of all ectothermic vertebrates studied to date (Nilsson et al., 2007; Stobutzki and Bellwood, 1994, 1997; Fisher et al., 2005; Downie et al., 2020). Upon settlement, coral reef fishes cope with fluctuating O₂ levels that can fall below 20% air saturation at night (Nilsson and Östlund-Nilsson, 2004), making them some of the most hypoxia-tolerant species studied so far (Nilsson et al., 2007). Additionally, developing and residing in tropical, warm waters of coral reefs means these fishes may already live close to their upper thermal limits (Rummer et al., 2014; Gardiner et al., 2010; Johansen and Jones, 2011; Pörtner et al., 2017), accelerating growth and metabolic turnover, and thus creating greater O₂ and carbon dioxide (CO₂) exchange and waste product removal. These aforementioned physiological and demographic characteristics are not yet fully understood, but are almost certainly supported by highly functioning respiratory and ionoregulatory systems. However, the link between gill development and function to support these life history characteristics in coral reef fishes remains unknown. Therefore, expanding our knowledge of gill development in coral reef fishes presents an important opportunity to understand how key developmental events occur in species with unique physiological traits and a challenging environment shaped by high mortality rates.

Here, we examined embryonic development in two coral reef damselfish species to generate a more comprehensive understanding of oxygen uptake and ionoregulatory processes early in the ontogeny of fishes. The aims of this study were to quantify the rate of gill development in reef fish and define when and in what order respiration and ion regulation transitions to the gills. The spiny chromis (Acanthochromis polyacanthus) and the cinnamon anemonefish (Amphiprion melanopus) were chosen as the study species because they come from one of the most speciose and ecologically dominant coral reef fish families, Pomacentridae (Cowman, 2014; Ceccarelli et al., 2001). The two species also exhibit opposing life histories; A. polyacanthus resides on the reef after hatching close to its parents, receiving additional parental care, whereas A. melanopus spends its larval phase in the pelagic environment before finding a suitable coral reef habitat on which to settle as a juvenile. For these reasons and others mentioned above, we hypothesize that coral reef fishes will exhibit a rapid rate of development in terms of oxygen uptake and ionoregulatory processes when compared with previously examined species. We might expect to see differences between the two species as a reflection upon their larval life histories. In both species, we examined the gross morphology and histology of gills during early ontogeny. We used respirometry to determine changes in Ṁ_O2 across development, and we used phenylhydrazine (PHZ) to inhibit haemoglobin (Hb) function and provide insights into the role of the gills in respiration. We expect Hb to begin functioning for oxygen transport soon after it first appears to support the aerobically demanding early life history traits of coral reef fishes (i.e. strong swimming abilities and very high mass-specific Ṁ_O2). PHZ destroys Hb indirectly by oxidizing the haeme group (Fe³⁺; methaemoglobin), which causes Heinz body formation, and ultimately decreases red blood cell (RBC) concentrations and the blood's capacity for oxygen binding and transport (Shetlar and Hill, 1985). Key components of ionoregulatory functions were explored by documenting the appearance of ionocytes within cutaneous and branchial locations using immunocytochemistry. These techniques allowed us to estimate when gills begin functioning for oxygen uptake and ion regulation, and to produce a time line of these traits to compare with other, previously described species.

James Cook University Animal Ethics Committee approved the following experiments under ethics protocol A2407. Import of antibodies into Australia was approved by the Australian Government, Department of Agriculture and Water Resources under permit #0002256071.

Embryos from Acanthochromis polyacanthus (Bleeker 1855) and Amphiprion melanopus Bleeker 1852 were sourced from established breeding pairs held in 60 l holding tanks (∼4000 l recirculating system; UV-filtered seawater) in aquaria facilities at the Marine and Aquaculture Research Facility at James Cook University. Breeding pairs were exposed to ambient summer photoperiod (approximately 14 h:10 h light:dark), salinity was maintained at 36 ppt, water temperature at 28.5±1°C, and fish were provided with half a terracotta pot for shelter and embryo deposition. Fish were fed twice daily (morning and evening) to satiation with INVE Aquaculture NRD 1.2 mm food pellets (ProAqua Pty Ltd, Coorparoo, QLD, Australia). Pots were checked for egg deposition during morning feeding. For experiments, embryos were collected the morning of the trial, which ensured embryos received maximum parental care by remaining with parents until needed. This ranged from 1 day post-deposition until 1 day prior to hatching. For A. melanopus, hatching generally occurred at 7 days post-fertilization (dpf), and for A. polyacanthus, hatching occurred at 10 dpf. Six embryos per age across a minimum of three breeding pairs were used for histological and immunocytochemistry protocols (Li et al., 1995). For respirometry trials, a minimum of 10 embryos (Pasparakis et al., 2016) obtained from at least three breeding pairs were used per age and treatment.

Embryos were removed from pots and examined under a dissecting microscope (Olympus SZX7 Stereomicroscope System) to observe the onset of the heartbeat, red blood and the presence of gills. Micrographs of the embryos were obtained using an Olympus SC50 digital camera. Embryos were then euthanized by being placed into ice cold water and fixed in Bouin's solution, where they remained for 12–24 h before being transferred into 70% ethanol, where they remained until histological preparation.

To prepare samples for histological assays, all embryos were dechorionated under a dissecting microscope (Olympus SZX7 Stereomicroscope System), micrographs were obtained for  measurement of total length (measured in ImageJ v.1.51s, National Institutes of Health, Rockville, MD, USA), and embryos were placed into cassettes. Samples underwent a series of ethanol and wax infiltrations (Shandon Southern Duplex Processor BS5), embedded in paraffin blocks (Shandon Histocentre 3, Thermo Electron Corporation), and then sectioned at 5 µm thickness using a microtome. Samples were embedded laterally and sectioned from when gill tissues were first observed (preliminary inspection using a compound microscope, Olympus BX41 Microscope) and then subsequently sampled every 10–20 µm. In early ages when gill tissue was not yet formed, sections were sampled when adjacent tissues (e.g. eye, trunk and yolk sac) were in view. Slides were stained using Mayer's Hematoxylin and Eosin, and examined under a compound microscope; images were then taken using an Olympus SC50 digital camera. The presence of morphological features including gill arches, rows of filaments and lamellae was recorded from micrographs.

We measured oxygen uptake rates (reported as Ṁ_O2) in developing embryos (A. polyacanthus: 4.9 mg; A. melanopus: 1.6 mg) to determine when blood oxygen transport becomes critical across embryogenesis. However, prior to measuring Ṁ_O2, individuals were either control or PHZ exposed (2 mg l⁻¹) for 3 h (Jacob et al., 2002; Pelster and Burggren, 1996). Exposure time and the movement of PHZ across the chorion were determined by examining control and PHZ-exposed embryos under a dissecting microscope (Olympus SZX7 Stereomicroscope System) and characterizing the pigmentation of the blood flow (i.e. after 3 h exposure to PHZ, blood flow appeared clear to yellow in coloration instead of red, indicating that Hb was destroyed; see Fig. S1; Pelster and Burggren, 1996). Ṁ_O2 was subsequently measured in embryos daily, from 1 dpf until hatching. Measurements ceased in all animals upon hatching because mortality was 100% if newly hatched larvae were exposed to PHZ.

Embryos were euthanized as described above and fixed in 4% paraformaldehyde for 24 h at room temperature before being transferred into 70% ethanol and stored at 2°C until further analysis. To prepare embryos for immunocytochemistry, embryos were dechorionated using super fine tip forceps (Dumont) under a dissecting microscope (Olympus SZX7 Stereomicroscope System) and photographed using an Olympus SC50 digital camera for total length. All immunocytochemistry protocols were obtained and modified from McCormick et al. (2003) and Hiroi et al. (2008). Wholemount preparation was used to observe the skin and yolk sac, whereas, sectional preparation was used to observe gill tissue.

Dechorionated embryos were rinsed in 0.01 mol l⁻¹ phosphate-buffered saline (PBS; 0.138 mol l⁻¹ NaCl, 1.9 mmol l⁻¹ NaH₂PO₄, 8.1 mmol l⁻¹ Na₂HPO₄ in deionized water, pH 7.4) for 15 min at room temperature, rinsed again in 30% sucrose/PBS solution, and rinsed a third time in PBS before a 30 min incubation in 2% normal goat serum.

Similar to wholemount preparation, dechorionated embryos were rinsed in 0.01 mol l⁻¹ PBS for 15 min at room temperature and then rinsed again in 30% sucrose/PBS solution before being placed into Cryomolds, covered with embedding medium and frozen at −20°C.

Sections were cut to 5 µm (−20°C; samples were positioned and sectioned as described above) using a Leica CM1860 UV Cryostat and placed on slides to dry. Slides were rinsed with PBS following a 30 min incubation in 2% normal goat serum. Slides and wholemounts were incubated in α5 anti-Na⁺,K⁺-ATPase antibody (monoclonal; α5 was designed to a highly conserved region of Na⁺,K⁺-ATPase and deposited in the Developmental Studies Hybridoma Bank by D. M. Fambrough) at 1:1500 dilution in antibody dilution buffer (0.01% NaN₃, 0.1% bovine serum albumin, 2% normal goat serum and 0.02% keyhole limpet hemocyanin in PBS) overnight at 4°C to label ionocytes. The following morning, the samples were rinsed with PBS and incubated a final time with 1:1000 Alexa 488 goat anti-mouse (Life Technologies Australia Pty Ltd, Mulgrave, VIC, Australia) for 2 h at room temperature and rinsed with PBS.

Wholemounts and sectioned samples were examined on an inverted confocal fluorescence microscope (Zeiss LSM 710). Images were taken using Zen 2009 Light Edition (v.5.5.285) confocal software, using a combination of z-stacks and continuous imaging. To analyse images obtained from wholemount samples, ionocyte density and size were measured in ImageJ (v.1.51s, National Institutes of Health) and recorded in regard to their location. A grading system (as described by Hiroi et al., 1998) was used for sectional samples of gills obtained because of the complex structure and size of the developing gill tissue. Grades were given to each sample as follows after visually inspecting four individual sections (see Fig. 4 for visible progression of ionocyte density; Hiroi et al., 1998): 0, absent (no ionocytes visible); 1, sparse (some ionocytes visible); 2, moderately dense (several clusters of ionocytes but some areas on filament without ionocytes); or 3, highly dense (filament tissue completely covered with ionocytes or majority covered with minimal areas lacking ionocytes).

All statistical analyses were performed in R version 1.1.453. Linear mixed-effects models (LME; nlme package) were used to observe changes in Ṁ_O2 as a function of embryonic age (dpf). Model selection was conducted by removing non-significant main effects or interactions, except for mass, which was always included in the model. Once visually assessed through Q–Q plots and residual versus fitted plots and compared against un-transformed Ṁ_O2 data using the Akaike information criterion (AIC), Ṁ_O2 was log transformed for both species to meet assumptions of normality. In LMEs for both species, logged Ṁ_O2 was used as a response variable, age and treatment were fixed effects, mass was used as a covariate and clutch was used as a random effect.

LME models were also used to determine changes in ionocyte density and size as a function of embryonic age (dpf). Model selection was conducted by removing non-significant main effects or interactions. Models were visually assessed through Q–Q plots and residuals versus fitted plots to assess normality and equal variances. As a result, and upon comparison with un-transformed Ṁ_O2 data using AIC, ionocyte size was log transformed for both species to meet normality. In the LMEs for both species, ionocyte density or logged size was used as the response variable with age and tissue as fixed effects, and clutch was included as a random effect.

LME models with significant interactions and main effects (i.e. if no significant interaction was detected) were compared using estimated marginal means (EMMs; emmeans package) by comparing the specified variable(s) in the model.

In A. polyacanthus, at 1 dpf, the yolk sac and rudimentary body and head regions were evident. At 2 and 3 dpf, the heart began to beat, and red blood cells started circulating (Fig. 1). Gill arches appeared as early as 3 dpf but were evident by 4 dpf in all investigated embryos (Fig. 1). After gill arches formed, gill filaments were evident, appearing between 4 and 5 dpf, but most commonly at 5 dpf (Figs 1 and 2), after which, lamellar formation occurred. Lamellae were visualized as early as 5 dpf, most commonly appearing at 6–7 dpf (Figs 1 and 2), but as late as 8 dpf. At 8 and 9 dpf, lamellae proliferated across all filaments. Hatching generally occurred at 10 dpf (±1 dpf).

In Amphiprion melanopus, the yolk sac and rudimentary body and head regions were evident at 1 dpf. The heart began to beat and red blood cells started circulating at 2 and 3 dpf, respectively (Fig. 1). In the majority of A. melanopus embryos examined, gill arches formed at 3 dpf (Fig. 2); however, in one embryo, gill arches formed at 4 dpf (Fig. 1). Gill filaments appeared predominantly at 4 dpf, but in all embryos by 5 dpf. Lamellae formed soon after at 5 dpf (Fig. 1), although one A. melanopus embryo was found to develop lamellae at 6 dpf (Fig. 1). Hatching occurred at 7 dpf (±1 dpf).

Significant differences in Ṁ_O2 were detected in A. polyacanthus between treatments (LME, F_1,370=4.7331, P=0.0302) and among ages (LME, F_8,370=3.2493, P=0.0014; Table S2), whereas treatment (LME, F_1,262=0.002, P=0.9671) and age (LME, F_5,262=0.161, P=0.9766) were not significant in A. melanopus. In A. polyacanthus, PHZ-exposed embryos exhibited an Ṁ_O2 that was 4% (0.174±0.0705 mg O₂ kg⁻¹ h⁻¹ log transformed; t₃₇₀=2.468, P=0.0140) lower than in control embryos. In both species, Ṁ_O2 increased with age, as expected (Fig. 3), although this trend was not significant in A. melanopus. In A. polyacanthus, age was only significant between 1 and 3 dpf, as well as between 3 and 7 dpf (P<0.05; Table S2). Significant interactions between Ṁ_O2 and age and treatment were not detected in either species (P>0.05; removed from model); however, 100% mortality occurred in all fish exposed to PHZ at hatching.

Extra-branchial ionocytes first appeared in the yolk sac and trunk (i.e. the skin covering the main body compartment) at 2 dpf in A. polyacanthus but at 1 dpf in A. melanopus (Fig. 4). In A. polyacanthus, a significant interaction was detected between ionocyte density and age and location (LME, F_7,89=3.5765, P=0.002; Table S3). Ionocyte density increased significantly in the trunk of A. polyacanthus embryos at 4–5 dpf (an increase of 500%; Fig. 5A; Table S4), after which (i.e. 6–9 dpf), ionocyte density decreased back to similar levels to those estimated at 2–3 dpf (Fig. 5A; Table S4). Ionocyte density in the yolk sac remained similar across development (Figs 4A and 5A; Table S3). In A. melanopus, no significant changes in ionocyte density in either tissue were found over increasing embryonic age (Fig. 5B; Table S3). However, ionocyte size changed significantly in both locations (LME, F_5,810=2.525, P=0.028; Table S3), decreasing by 60% and 67% in the trunk and yolk sac of A. melanopus with increasing age (LME, P≤0.05; Fig. 6B; Table S5). Similarly, in A. polyacanthus, ionocyte size in the yolk sac decreased with increasing age (∼60% decrease; LME, P≤0.05), although ionocytes at 9 dpf were not significantly different in size from those at 5 and 6 dpf (Fig. 6A; Table S5). Along the trunk of A. polyacanthus, ionocyte size changed with age (LME, P≤0.05); however, the trend was not the same as it was in other tissues (Fig. 6A; Table S5).

In both species, branchial ionocytes were only found in the gill filaments, at the base of the lamellae (interlamellar region), although not in the lamellae themselves (Fig. 4). Ionocytes appeared as soon as filaments started to form, which occurred at 4–5 dpf in A. polyacanthus and at 4 dpf in A. melanopus (Fig. 7). Moderately to highly dense clusters of ionocytes were found in the filaments at 7 dpf in A. polyacanthus, and moderately dense clusters were found at 5 dpf in A. melanopus, simultaneous with the formation of lamellae (Fig. 7).

Our results show that, when compared with most of the previously investigated fishes, A. polyacanthus and A. melanopus exhibit rapid gill development during embryogenesis. It is likely that, during embryonic development, the gills have limited contributions to the overall respiratory and ionoregulatory functions, but rather are preparing to become the primary site for these functions upon hatching. This pattern of early gill formation may occur to sustain the high demands of the challenging lifestyle that coral reef fishes endure post-hatching and when the larvae transition from the pelagic environment to metamorphose into juveniles that reside on tropical coral reefs. Coral reef fishes have inherently high mortality rates (Emslie and Jones, 2001; Goatley and Bellwood, 2016), and this may be related to mismatches between the fast pace of development and the efficiency of key physiological functions. Disruptions to gill development during embryogenesis and to gill functions during the larval phases may become more frequent and intense with increasing anthropogenic stressors, perhaps exacerbating the already high mortality rates of this group of coral reef fishes.

In this study, filament and lamellar formation were evident by 5 and 7 dpf, respectively, in A. melanopus and A. polyacanthus. At 1 day prior to hatching, gill structure in both species started resembling those of juvenile and adult fishes, although they were smaller in size (i.e. transitioning from bud-like structures to a more defined elongated shape; Rombough, 2011). In the majority of fish species that have been investigated to date, gill development does not commence until late in embryonic development, with respiratory tissues, including lamellae, often forming after hatching (Li et al., 1995; Gao et al., 2016; Padrós et al., 2011; Hachero-Cruzado et al., 2009; Rombough, 2002; Rombough, 1999; Hwang, 1990). This is probably due to other environmental parameters influencing their developmental pace (Burggren and Dubansky, 2018). The freshwater cichlids (e.g. tilapia) were previously thought to exhibit the fastest rate of gill development among fishes, with the first signs of gills noted at 7 dpf or 2 days post-hatching (dph; Li et al., 1995); however, more recently studies have shown that zebrafish (Danio rerio) begin gill development at 3 dpf with lamellae forming later at 12 dpf (Rombough, 2002), and the tropical gar (Atractosteus tropicus) form gill filaments at 5.5–6.5 dpf or 2.5 dph, with lamellae formation beginning at 11–12 dpf or 9 dph (Burggren et al., 2016). The species examined in the present study, therefore, fall within the high extremes of known developmental rates among teleost fishes.

Alongside rapid structural development, oxygen demand (measured as Ṁ_O2 in this study) also increases during embryogenesis, peaking between 5 and 7 dpf. During embryogenesis, oxygen uptake rates in these two damselfish species were unaffected by PHZ exposure regardless of age; however, PHZ-treated embryos did exhibit a trend of having a lower Ṁ_O2 than control embryos, but this trend was not significant until hatching. This suggests that Hb – during embryogenesis – does not playing a key role in oxygen uptake, unlike what has been documented in the majority of adult teleost fishes (Cameron and Wohlschlag, 1969). Indeed, during embryogenesis, oxygen uptake in these two damselfish species is most likely occurring across the skin epithelium (Feder and Burggren, 1985; Pelster and Burggren, 1996), but they become reliant on Hb for O₂ transport during their larval stage, possibly related to the onset of their profound swimming abilities and increased mass-specific Ṁ_O2. This finding shows that these two study species develop Hb O₂ transport rapidly and early in ontogeny. This change in Hb O₂ transport before and after hatching is surprising, given that Ṁ_O2 in zebrafish at hatching (<4 dpf) was unaffected by PHZ treatment at similar exposure levels (Pelster and Burggren, 1996), but showed similar trends (i.e. lower Ṁ_O2 with PHZ exposure). Although Ṁ_O2 in PHZ-exposed embryos was not significantly lower than that in control embryos, Hb may still be playing some role in O₂ transport that was not captured in this experiment; for example, because of a lack of sufficient water flow across the gills due to irregular buccal pumping during early development (Rombough, 2011). Moreover, fine scale measurements (e.g. hourly) may provide greater insight. Future investigations should focus on early development of oxygen uptake machinery, the transition from cutaneous to branchial pathways, and Hb O₂ transport activation in coral reef fish embryos and early life stages of other fish species.

We also found early development of key components of the ionoregulatory system in both A. polyacanthus and A. melanopus when compared with previously studied teleosts. Ionocytes first appeared on the gills at 3 dpf, simultaneously with gill arch and filament formation, thus outpacing zebrafish (5.5 dpf; Rombough, 2002). Moreover, ionocyte density increased substantially during this short developmental window between 4 and 6 dpf, a period coinciding with a decreased density of cutaneous ionocytes in A. polyacanthus and a decrease in cutaneous ionocyte size in both species. Morphometric changes to cutaneous ionocytes that coincide with rapid expansion of branchial ionocytes may indicate the onset of ionoregulatory processes transitioning to the gills.

This accelerated pace of gill development observed in these two coral reef fish species may occur for a number of reasons. Marine fishes are faced with greater ionoregulatory and oxygen uptake limitations than freshwater fishes because oxygen solubility is lower in seawater than in freshwater. The energetic requirements of osmoregulation are thought to be higher in seawater than in freshwater, though the precise requirements for each are still unknown (Kirschner, 1995; Ern et al., 2014). Additionally, ectothermic, tropical species exhibit higher metabolic rates than their temperate water counterparts (Angilletta et al., 2002; Pörtner et al., 2017), therefore creating immense pressure on physiological systems to ensure efficient gas exchange and ion balance. Rapid onset of gill development may also be a necessary response to high mortality rates endured by these coral reef fish species and the overall challenging lifestyle of coral reefs (Goatley and Bellwood, 2016; Emslie and Jones, 2001). Therefore, the rapid gill development documented for these two species of coral reef fishes may be related to the fact that they are marine, tropical ectotherms, and coral reef fish species, or a combination thereof.

Larval coral reef fishes exhibit an array of early life history traits that are likely to rely on the rapid pace of gill development observed in this study. Firstly, coral reef fishes exhibit high growth rates (e.g. organogenesis and morphogenesis) and rapid yolk sac consumption (Green and McCormick, 2001; Donelson et al., 2012), generating high loads of metabolic waste (Wright et al., 1995; Dhiyebi et al., 2013; Zimmer et al., 2014). During early life stages, nitrogenous waste must be removed through Na⁺/NH₄⁺ exchangers that are located within cutaneous ionocytes; such cutaneous processes are curtailed with declining SA/V ratios of embryos, thus necessitating increasing reliance on branchial pathways. Secondly, larval coral reef fishes undergo major changes in aerobic capacity upon hatching, which may be related to early gill development, both structurally and physiologically. Of all coral reef fishes, the pomacentrids have been investigated most thoroughly with respect to metabolic performance and exhibit among the highest recorded mass-specific Ṁ_O2 (e.g. 5250 mg kg⁻¹ h⁻¹) for ectothermic vertebrates (Nilsson et al., 2007) and the fastest sustained swimming speeds (i.e. approximately 50 body lengths per second) examined to date (Stobutzki and Bellwood, 1994, 1997; Downie et al., 2020). Our results suggest that these high demands may be aided by structural and functional changes to the gills during early development, including enhanced structure for ion transport and increased aerobic capacity.

Understanding whether gills form primarily for respiratory or ionoregulatory purposes is crucial to identifying the processes driving the evolution and development of fishes. For example, Fu et al. (2010) and Zimmer et al. (2014) found that Na⁺ and NH₄⁺ movements in rainbow trout larvae transitioned from cutaneous pathways to the gills simultaneously at 15 dph, but before O₂ uptake, which occurred at 27 dph (Zimmer et al., 2014). These authors also provide evidence supporting sodium uptake and ammonia excretion as one of the first functions performed by the developing gill, thus supporting the ionoregulatory hypothesis. Our study did not directly distinguish between cutaneous and branchial pathways for respiration and ionoregulation. However, our study shows that ionoregulatory tissues (i.e. gill filaments) and ionocytes appear prior to respiratory tissue (i.e. gill lamellae) formation and before respiration becomes dependent on the gills, suggesting that gills may initially function for ionoregulatory processes in these coral reef fish species as well. The presence of ionocytes in gill filaments prior to lamellae formation is a phenomenon that has been observed in numerous fish species (Li et al., 1995; Gao et al., 2016; Padrós et al., 2011; Hachero-Cruzado et al., 2009; Rombough, 2002, 1999; Hwang, 1990), suggesting that gills perform ionoregulatory processes first. Therefore, our findings also lend support to the ionoregulatory hypothesis. Although morphological trends of ionocytes can be used as a tool to determine the onset of ionoregulatory processes at the gills (Brauner and Rombough, 2012), this only approximates the relative roles of cutaneous and branchial mechanisms at this stage of development.

Further investigation is needed to determine when and in what order the respiratory and ionoregulatory processes transition from the cutaneous surfaces to the gills in coral reef fishes (i.e. oxygen versus ionoregulatory hypothesis). Coral reef fishes present a unique group that contrast with previously investigated teleost fishes, as they exhibit physiologically demanding life history traits that are most likely the definitive drivers underpinning early gill development. With this information, we can then identify potential mismatches between gill development and function and whether this is related to high mortality rates exhibited during their larval phase. This information should be extremely powerful in determining how anthropogenic stressors interfere with growth and survival, especially during key developmental windows, which could collectively improve predictions for future coral reef fish populations.

In summary, gill formation during embryonic development in coral reef fishes occurs at a rapid rate, perhaps to support high-performance traits exhibited during their larval life stage and in preparation for the challenging lifestyle on tropical coral reefs. Many coral reef fish species have now been documented to possess high Ṁ_O2, fast swimming speeds and remarkable hypoxia tolerance (Leis and McCormick, 2002; Stobutzki and Bellwood, 1994, 1997; Nilsson et al., 2007); therefore, it is likely that the fast-paced developmental patterns observed here may extend beyond these two study species and beyond species within the Family Pomacentridae. Furthermore, this study has outlined a critical developmental window in two coral reef fish species, perhaps one that may be common for other coral reef fish species. This developmental window may therefore be an appropriate target for future investigations, in particular when identifying potential mismatches between environmental pressures and the pace at which these key respiratory and ionoregulatory systems develop.

The authors thank the J.L.R. and S.D.M. labs for their assistance with animal husbandry, Adam Downie and Bjorn Illing for all their insight and helpful discussions, James Cook University Marine and Aquaculture Facilities for resources, M. J. McWilliam for editorial work, R. Jones for statistical advice, S. and L. Reilly, K. Cuthbertson and S. Askew for immunocytochemical and histological support, and E. Walsh for illustrations. Any use of trade, product, or firm names in this paper is for descriptive purposes only and does not imply endorsement by the US Government.