Mudskippers brood their eggs in air but submerge them for hatching

Mudskippers (family Gobiidae, subfamily Oxudercinae) are specialized teleosts adapted for amphibious life on mudflats. During low tide, mudskippers volitionally emerge from water and feed, defend territories, and engage in courtship on the mud surface (Clayton,1993). Mudskipper terrestrial activity is supported by a capacity for aerial respiration, to which epithelia in the gills, the buccal-branchial cavity, and the skin all contribute in varying proportions, depending on species (Graham, 1997). The large buccal-branchial cavity of mudskippers is important for holding a substantial air volume for respiration, as well as for territorial displays and burrowing (Graham et al.,2007). For example, the Malaysian mudskipper Periophthalmodon schlosseri (Pallas) has a buccal-branchial cavity volume that is 16% of its body volume (Aguilar et al.,2000), which is much larger than the cavity volume of 2–6%reported in gobies not specialized for amphibious air breathing(Gee and Gee, 1991; Gonzales et al., 2006). The anatomical and physiological specializations of mudskippers enable them to maintain metabolic rates in air that are generally equal to or higher than in water (Graham, 1997; Kok et al., 1998; Takeda et al., 1999).

In contrast to activities on the mudflat surface, little is known about mudskipper behaviour in burrows during high tide. Most mudskippers excavate burrows and use them as a refuge from predators, for protection from desiccation and temperature extremes, and for nesting. Although spawning in burrows affords egg protection, the severe hypoxia of burrow water(Ishimatsu et al., 2000)raises the question of what mechanisms are used to ensure adequate O₂ for developing mudskipper eggs(Gordon, 1995). Mudskipper burrows often have one or more horizontal or upturned chambers near their terminus where monolayers of fertilized eggs develop(Brillet, 1976; Clayton and Vaughan, 1986; Kobayashi et al., 1971). Recent studies document an air-deposition behaviour by burrow-guarding mudskippers in the field (Ishimatsu et al., 1998; Ishimatsu et al.,2000) and in artificial laboratory burrows(Lee et al., 2005), prompting the hypothesis that eggs are spawned and develop within a burrow's air-containing space. We tested this and related aspects of burrow incubation of eggs by conducting field and laboratory studies of the Japanese mudskipper Periophthalmus modestus Cantor. In the spawning season (mid-May to late August in our study area) male P. modestus excavate `J' shaped burrows (Fig. 1) having two or three openings on the mudflat surface(Kobayashi et al., 1971). Using nuptial colouration and courtship displays the male attracts a female into its burrow, where spawning occurs, and the fertilized eggs are deposited in the terminus (Matoba and Dotsu,1977). Thereafter, the male tends its burrow for about 1 week until the eggs hatch.

Field and laboratory studies were carried out to determine the role of the male in securing egg development and hatching in burrows. The questions we studied were as follows. (1) Does burrow air occur in the egg chamber? (2)Does a burrow-guarding male deposit air before or after egg spawning? (3) How are the oxygen levels of the egg chamber maintained during egg-incubation periods? (4) Does egg chamber hypoxia stimulate male's behaviour to restore egg-chamber P_O2? (5) Do mudskipper eggs really develop in air, and what is the trigger for hatching? Integration of the data obtained provides strong evidence that during each low tide the burrow-guarding male adds air into an air-filled egg chamber, thereby refreshing egg-chamber O₂. The air-adding behaviour is at least partially governed by egg-chamber P_O2. Once egg development is complete, the male removes egg-chamber air to induce hatching.

Field studies on Periophthalmus modestus Cantor burrows occurred from 1998 through 2006 on mudflats in Ashikari, Saga Prefecture, Japan. The burrows of P. modestus could be easily differentiated from those of two other sympatric intertidal gobies [Boleophthalmus pectinirostris(Linnaeus), also an oxudercine, and Odontamblyopus lacepedii(Temminck & Schlegel), subfamily Amblyopinae] by the presence of mud pellets (i.e. mouthfuls of mud ejected during burrow excavation and maintenance) around the openings.

Samples for burrow-water P_O2 determinations were taken in glass tuberculin syringes (1 ml volume), the tips of which were inserted to a depth of at least 5 cm into the water standing in burrow shafts at low tide. The filled syringes were immediately stored in ice water and transported to the Saga Prefectural Ariake Fisheries and Development Center where, following anaerobic centrifugation to separate the suspended mud particles, the P_O2 of the supernatant in each was measured using an O₂ electrode (E101, Cameron Instrument Co.,TX, USA) and a meter (OM200, Cameron Instrument Co.). The total gas volume and its P_O2 and P_CO2levels were compared for 20 guarded egg chambers (i.e. eggs present) and 12 pre-spawning egg chambers (i.e. male courting, no eggs in the chamber). Egg-chamber gas was obtained by forming a shallow depression over the burrow,lining it with plastic sheeting and placing a thin layer of water therein, and then compressing the underlying burrow to force air up under the plastic where it could be withdrawn into a glass syringe for determination of volume and gas content. Gas P_O2 and P_CO2 were measured by chromatography (Shimadzu GC-14B, Kyoto, Japan). Determinations were also made for both the average particle size of the mud comprising the layer between the mudflat surface and the egg chamber (Shimadzu SALD-3100 Laser Diffraction Particle Size Analyzer)and the redox potentials (ORP electrode, 9678BN and 230Aplus meter,ThermoOrion, MA, USA) of mud from the egg-chamber surface, the shaft walls,and at sites remote from the burrow. A combination pH electrode (ThermoOrion 8163BN) was used as a reference electrode to improve stability of the redox measurements.

Burrow structure and the instrumentation procedures are detailed in Fig. 1. For each burrow selected for instrumentation the first step was to remove standing water in the shaft to prevent it from accidentally inundating the eggs when the egg chamber was opened. The mud over the chamber and surrounding the burrow shaft was then excavated down to the level of the chamber's roof. When the chamber was opened, the presence of eggs was verified and their state of development assessed by immersing a small number of eggs in water. If immersed eggs did not hatch within 5 min, the chamber was considered to contain relatively young eggs and instrumentation proceeded. The roof of the egg-chamber was fitted with the upper-half of a plastic bottle into which both an endoscope(HSCI-S10M, HOGA, Kyoto, Japan) and an O₂ electrode (ThermoOrion 083025A) had been mounted. For two burrows a small Teflon™ tube (1 mm i.d.) was also inserted into the chamber roof so that egg-chamber O₂ levels could be experimentally modulated (see below). Stainless steel impedance electrodes were inserted in the burrow section below the egg chamber (Fig. 1). Mud was then replaced to cover and secure the instruments and the section of burrow shaft destroyed by digging was reconstructed. Unless egg-chamber excavation and instrumentation caused the burrow-guarding male to abandon the nest completely(in such cases the eggs did not hatch), it would re-enter the burrow and resume its activities. However, the male usually did not use the hand-constructed opening, but built new connections to the vertical shaft. To monitor the tidal cycle, a Tygon™ tube connected to a pressure transducer (Keyence AP-81A/AP-12, Osaka, Japan) was deployed on the mud surface at nearly the same height as the reconstructed burrow opening.

Cables connecting the measuring devices to their monitors (ca. 10 m long)extended to a tent pitched on a nearby pier (ca. 3 m above the mudflat surface). Signals from the O₂ electrode were fed to and saved every 5 min in an O₂ meter (ThermoOrion 835A). The O₂electrode (a nominal drift of 0.1% per day) was calibrated with humidified N₂ gas and air immediately before egg-chamber instrumentation and after termination of the observations. Egg-chamber P_O2 data were corrected for hydrostatic pressure using measured water depth and specific gravity of 1.02 for seawater,and drift corrections assumed a linear drift over the observation period. Endoscope video recording was kept minimal to lessen the effects of light on the guarding male's behaviour. Signals from the tide-monitoring pressure transducer were digitized and stored every 5 min in a data acquisition system(Keyence NR-1000).

Preliminary laboratory experiments indicated that the passage of dead P. modestus through a pipe generated impedance signals of more than±500 mV, and it was expected that this criterion could be used to indicate the number of trips made by the male into the burrow chamber. Burrow impedance signals were amplified with an impedance converter (model 2991, UFI,CA, USA), digitized by another data acquisition system (Keyence NR-2000), and stored in memory cards (32 MB) at a sampling rate of 2 Hz. The digitized data were analysed with software (Wave Shot v.1.05 and Flex Logger v.1.07,Keyence).

The two burrows in which the egg chamber had been fitted with a small Teflon™ tube (Fig. 1)were used to test the effects of a reduction in egg-chamber O₂level on fish passage through the burrow section connecting to the chamber(indicated by impedance spikes) as a possible indicator of air-adding behaviour. For these tests humidified hypoxic-gas mix, in which the P_CO2 was adjusted to minimize changes in egg-chamber P_CO2, was introduced into the egg chamber to reduce its P_O2 to within 0.25 kPa of the levels observed at the beginning of the low tide period. Hypoxic gas injections were made approximately 1 h before the burrow was covered by water of the rising tide and while the burrow-guarding male was on the mudflat surface. After egg incubation and hatching were completed, egg-chamber air was withdrawn through the Teflon™ tube to determine the volume for the computation of air-adding frequency (see Appendix).

Five intact egg chambers were transported to the laboratory for studies of egg-hatching success in relation to time and for comparisons of the hatching success of eggs in air and in water. After clearing mud from around the chamber, the column of mud surrounding the intact egg chamber (i.e. a complete, mud-walled, air-filled chamber having a ventral opening into the burrow shaft) was extracted by sliding a bottomless plastic bottle over it,placing a plastic plate under the open end of the bottle and lifting it out. Only egg chambers with relatively young eggs (as defined above) were extracted and, once in the laboratory, the chambers were incubated in high humidity at 23°C. About 50 eggs were removed from each chamber and used in daily 4 h tests of hatching competence (done by immersing eggs in freshwater, 50%seawater and full-strength seawater). Also tested was the viability of eggs kept submerged in hypoxic water. The total number of eggs in each egg chamber was estimated from the density of eggs and total surface area of the chamber.

The standing water in the burrow shafts was extremely hypoxic[P_O2=1.47±1.14 kPa (mean ± s.d., N=13); the P_O2 of air-equilibrated water is 20.6 kPa]. The average gas volume of the egg chambers was 47±19 ml (N=17) with a P_O2 of 13.8±3.0 kPa and a P_CO2 of 1.11±0.57 kPa (N=20) and, as in other mudskipper burrows(Ishimatsu et al., 1998; Ishimatsu et al., 2000), there was a significant inverse relationship between P_O2 and P_CO2(P_CO2=2.65–0.11P_O2, r=–0.577, N=20, P=0.008, Fig. 2). The egg-chamber volumes of `still courting' (i.e. no eggs present) and `egg-guarding' males did not differ (47±19 vs 43±15 ml, N=12) and,while the P_O2–P_CO2regression slopes of the two were also not different (ANCOVA, P=0.141), the mean P_O2 of the `no-egg'chambers was significantly less (6.15±4.66 kPa vs13.8±3.0 kPa, P=0.04, t-test, d.f.=30).

The method of egg-chamber excavation did not allow visual verification that the egg chambers contained air (i.e. we could not rule out the possibility that water had drained out of the chamber during excavation). However, our supposition that the egg chambers normally contain air was supported on the basis of the colour and the redox potentials measured for the mud lining the chamber wall. Mud on the wall of the egg chamber had a uniform brownish-gold colour rather than the dark grey colour of mud on the surface of other burrow regions. Also, the positive redox potential measured for egg-chamber mud(82.9±23.4 mV, reflecting a greater oxidation state) was significantly higher than those of mud lining the burrow vertical shaft below the water line(–29.3±60.2) and mud positioned at the same depth (20 cm under the surface) but not near the burrow (–117.1±22.4, N=8; P<0.01 Sheffé test for pairwise comparisons). The average particle size of mud comprising the layer between the mudflat surface and the egg chamber was 7.3±0.5 μm (N=5). The estimated average number of eggs in a P. modestus burrow was 5200±765(N=11).

Of 44 attempts to instrument burrows, only 6 records were obtained in which egg-chamber P_O2, endoscope video recordings,and impedance data spanned 4 days or longer.

The P_O2 records documented tidal-cycle related changes; O₂ increased during low tide but dropped during high tide (Fig. 3A). The corresponding endoscope records verified that the eggs remained in air, that the male occasionally entered the egg chamber and that it was not adversely affected by the presence of air therein. The quality of the simultaneous burrow impedance records did not enable full quantification of male movement pattern. However, a semi-quantitative pattern of fish movement was discernible and, in the majority of cases, this showed that the male's passage through the burrow section under the egg chamber (and entry into the egg chamber) occurred mainly during low tide (Fig. 3A). The increase in egg-chamber P_O2 during low tide was a common finding for all of the instrumented burrows. The average egg-chamber P_O2 was 12.8±0.99 kPa (N=4) at the beginning of low tide exposure (i.e. burrow opening exposed to air) and was 16.8±0.32 kPa at the end of the low tide. The correspondence between the increase in P_O2 and the impedance records suggests that during low tide the male adds gulps of air to the egg chamber and this raises P_O2.

More irregularity in the impedance records was found for burrows that were sufficiently high on the mudflat to remain uncovered by the highest neap tides; these showed impedance signals during both low and high tide. While also reflecting the tidal cycle, the P_O2 in these higher-positioned egg chambers showed small, intermittent step increases during `high tide' that also correlated with impedance-indicated movement. In one extreme case, the monitored burrow remained uncovered for 4 days after instrumentation. The subsequent increase in tidal amplitude and the covering of the burrow by rising water coincided with hatching(Fig. 3B).

To examine possible role of egg-chamber hypoxia as a modulator of air-adding behaviour, a volume of hypoxic gas was injected into the egg chamber to reduce P_O2 to within 0.25 kPa of the level observed at the beginning of the low tide period. Fig. 4A,C show the gradual rise in egg-chamber P_O2 that occurred in two burrows from the beginning of the low-tide period through about 7 h and up to the time of hypoxic-gas introduction (arrowheads). Although hypoxic gas injection reduced egg-chamber P_O2 to near the level observed at the beginning of each low tide, the male fish was able to rapidly restore the P_O2 within 1 h. For both of these burrows ANCOVA comparisons show a significantly greater rate of P_O2 increase for the 60 min after hypoxia injection than for the first 60 min of low-tide exposure, which implies that air-adding behaviour was not solely modulated by egg-chamber air P_O2. This possibility is further supported by the fact that the air-adding frequency (f_a) following hypoxic gas injection lies far above the calculated f_a–P_O2 relationships(Fig. 4B,D).

Eggs from the incubated chambers developed normally in air and had embryos ready for hatching (indicated by frequent turning) within 6–7 days of their estimated fertilization time(Kobayashi et al., 1972). This contrasts to eggs from these same burrows which, when placed in hypoxic water(P_O2≈2 kPa, simulating typical burrow-water conditions), succumbed within 2 days.

Although the laboratory-incubated eggs developed in air, they did not hatch and eventually died. It was found that water immersion (either in freshwater or seawater) was necessary to trigger egg hatching. Hatching competence tests on developing eggs over a 12-day period showed a 5–6 day window within which immersion must take place to ensure a mean successful hatch rate of 80%(Fig. 5). Eggs remaining in air beyond this time had a reduced hatching competence. Although hatching occurred in water of different salinities (i.e. freshwater, 50% seawater and full-strength seawater), larvae induced to hatch in freshwater did not survive unless transferred to higher salinities.

Endoscope video records (see Movie 1 in supplementary material) documented the male's role in hatch induction (Fig. 6, Table 1). Approximately 80 min after the burrow openings were inundated by the rising tide, the egg-guarding male triggered egg hatching by expelling air from the egg chamber, which immersed the eggs. Air removal was done by transporting air gulps into the burrow shaft leading to the openings(Fig. 1) and releasing them. As this was occurring bubbles were observed exiting from the burrow openings. Video records showed that flooding the chamber required the removal of an average of 103±42 air gulps, which gave an almost exact volume of total egg-chamber air (47 ml) based on the buccal-branchial capacity of P. modestus (0.46 ml in a 2.6 g fish, Table 1). Hatching occurred during evening or nocturnal rising tides(Fig. 7) but was not correlated with the lunar cycle.

Several littoral and sublittoral fish species deposit demersal eggs on surfaces in the intertidal zone that will emerge from the water as the tide recedes after spawning. This affords the eggs protection from aquatic predators and development in a habitat that is warmer and has greater O₂ availability. After completing development, the embryos hatch typically by inundation in high tide to be scattered as pelagic larvae(DeMartini, 1999; Helfman et al., 1997; Martin et al., 2004). Mudskippers differ from these fishes that spawn at the water's edge since a lack of cover and the extreme conditions on the mudflat preclude placement of eggs on the open surface and make it impossible for an adult mudskipper to remain with the eggs and care for them. Sequestration in burrows protects mudskipper eggs from predation and tidal extremes in current flow, temperature and salinity. However, the burrow water is extremely hypoxic(Gordon, 1995; Gonzales et al., 2006) (data in the present study), so this requires a solution for the O₂supply problem which, as demonstrated here for P. modestus, is done by egg development in air and refinements in parental care.

This study documents the critical importance of the male mudskipper in excavating a burrow, complete with an air chamber, at a position in the intertidal zone where it will have access to atmospheric air during low tide. Even though direct evidence for the presence of air in egg chambers could not be obtained, the differences in mud colour and the positive redox potential of chamber mud, the successful aerial egg development in the laboratory, and the burrow-guarding male's air-adding behaviour all attest to air deposition in egg chambers of intact P. modestus burrows. It should be noted that for another mudskipper, Periopthalmodon schlosseri, which is much larger than P. modestus, it has been possible to confirm the presence of air in egg chambers of intact burrows by directly inserting an endoscope through a much larger vertical shaft (A. Ishimatsu, T. Takeda, Y. Tsuhako and K. H. Khoo, manuscript in preparation).

Just as air access at low tide is required for egg-chamber O₂replenishment during incubation (Fig. 3), submersion of the burrow opening by a rising tide is necessary for the male to induce hatching and larval escape by flooding the egg chamber(Fig. 6). In addition to being in synchrony with the tidal cycle, the male's egg-chamber management behaviour is also modulated by its P_O2(Fig. 4). The aerial-egg development and water-induced hatching demonstrated for P. modestusin this study agrees with observations of Brillet(Brillet, 1976) who, while not detailing methodology, reported that the eggs of P. sobrinus [=P. argentilineatus Valenciennes (Murdy,1989)] required both air incubation and subsequent submersion for hatching and did not survive if incubated in water or kept in air but not immersed after development.

On the basis of the presence of air in the egg chambers of males engaged in courtship (i.e. no eggs present), we speculate that both spawning and fertilization occur in air, although this has not been observed. Details about fish aerial spawning behaviour were recently described for the rockhopper blenny Andamia tetradactyla (Bleeker)(Shimizu et al., 2006). The female deposits a monolayer of eggs in rock crevices along the splash zone,which the male fertilizes by rubbing its genital papillae over them. Periophthalmus modestus and other mudskippers may have a similar spawning behaviour and, while this is not documented, egg-chamber videos(Movie 1 in supplementary material) show that males can emerge into the air phase and both turn and crawl on the mud and egg surfaces.

The comparable volumes of the `no-egg' and `egg-containing' chambers imply that the egg chamber is maximally filled with air at construction and that air additions during egg incubation displace excess air into the burrow shaft from where it ultimately exits the burrow. The pre-spawning deposition of air inside an egg chamber may be beneficial for maintaining egg-chamber P_O2 during the incubation period because it would oxidize the highly reducing surrounding mud and thus lessen the total O₂ need once eggs are added.

The correlation between male activity (impedance data) near the egg chamber and its increase in P_O2(Fig. 3) most likely reflects transport of mouthfuls of fresh air into the chamber. Because burrow shafts are water-filled during low tide, it is not possible that air in the burrow was trapped during low tide as suggested for P. sobrinus(Brillet, 1976). Also, the small particle size of the overlying dense mud, its low redox potential, and the egg-chamber's position 20 cm under the mud surface(Fig. 1), exclude the possibility that the high egg-chamber P_O2results from the diffusion of atmospheric air. That the decline in egg-chamber P_O2 during high tide reflects egg-mass respiration is suggested by the product of the measured rate of O₂consumption of a developing mudskipper egg [76.5± 12.0 nl h^–1/hatch-competent egg(Etou et al., 2007)] and our estimate of the number of eggs in the chamber (5200). Assuming an egg-chamber air volume of 47 ml and a temperature of 24°C, the respiring egg mass would consume 1.1–2.0 ml of O₂ during a daytime high-tide period on the day of hatching (duration ranging 170–310 min, N=4; note that hatching occurred at the nocturnal high tide of the same day), and this accounts for 75–100% of the observed egg chamber aerial O₂ depletion of the high-tide periods (1.2–2.3 ml O₂) calculated from the slope of egg-chamber air P_O2 over time (–0.013 to –0.018 kPa min^–1).

The lowest egg-chamber P_O2 levels recorded at the end of high tide ranged from 8.2 to 14.5 kPa (N=8). These are above the critical P_O2 (i.e. the O₂partial pressure below which aerobic metabolism declines) of a P. modestus egg [4.3 kPa (Etou et al.,2007)]. Although the high-tide decline in P_O2 did not reach the eggs' critical P_O2, this level would have eventually been reached if the male had not added air during the subsequent low tide. Also,variables such as chamber volume, egg mass, the burrow's position in the intertidal zone, tidal amplitude, and whether or not the male also requires egg-chamber O₂ for respiration could all combine to cause greater reductions in egg-chamber O₂ during high tide.

Most nest-guarding fishes respond to hypoxia with fanning or other ventilation behaviours to augment O₂ delivery to the eggs(Jones and Reynolds, 1999; Takegaki and Nakazono, 1999). For species guarding eggs in the intertidal zone, fanning may be needed during low tide when pools are isolated and convective O₂ delivery is reduced. In this respect the air-adding behaviour of P. modestusparallels fanning in being hypoxia-induced and linked to the tidal cycle. A principal difference between the two is that, while requirements for fanning can vary depending on variations in the physical and biotic factors around an underwater nest, the small, mud-enclosed burrow egg chamber will always require the addition of air, which can only be done during low tide. The male must therefore ensure that there is sufficient O₂ to meet the requirement of the eggs during the subsequent high tide.

The mechanism enabling P. modestus to monitor O₂ levels in its egg chamber may not differ from the O₂ sensing mechanisms of other fish species. Most fishes have O₂-sensitive chemoreceptors that monitor the external environment and these are often positioned in the buccal-branchial cavity, associated with the gills or other structures(Milsom and Burleson, 2007). These external receptors have an important role in rapid responses to hypoxia,which appears important for air-breathing fishes(Graham, 1997). For Pn. schlosseri aerial hypoxia is a more important driver of air ventilation than aquatic hypoxia (Aguilar et al.,2000). While the air-adding behaviour of P. modestus is likely a ventilatory response driven by external O₂ receptors, it requires a more sophisticated behavioural combination of air gulping at the water surface, transporting mouthfuls of air down through the burrow shaft,releasing these into the egg chamber and repeating this action until egg-chamber P_O2 reaches a level sufficient to sustain the O₂ consumption of the eggs during the following high-tide period. The injection of hypoxic gas into the egg chamber late in the low-tide period increased the male's air-adding frequency and egg-chamber O₂ was restored to its pre-hypoxia injection level within 1 h(Fig. 4). The far higher rate of air addition triggered by this treatment implies that the male somehow modulates this behaviour by integrating information on egg-chamber P_O2 that must be attained by the end of low tide and the time available before the next submersion. This could be tested through hypoxic gas injections at different times during low tide.

A male P. modestus can readily add air during low tide, but numerous obstacles prevent this at high tide. When the burrow is submerged (by as much as 2 m at high tide in our study site), the male would need to swim to the surface, gulp air, and return it to the burrow. Direct observations of male activity during high tide are difficult because the water covering the mudflat at high tide is extremely turbid. Nevertheless, mudskippers were never observed surfacing to gulp air during high tide. If the male's primary duty is to guard the nest, it may spend high tide near the burrow openings where,while intercepting potential predators, it could respire aquatically(Tamura et al., 1976). Both endoscope and impedance data suggest that the male does not spend much time in the egg chamber during high tide. However, this conclusion is tentative because endoscope recording times were kept minimal to lessen disturbance from its light and the lower number of impedance spikes at high tide does not indicate that the fish is not in the egg-chamber.

Inundation is the most common hatch mechanism for the aerially spawned eggs of fishes (Yamagami, 1988) and amphibians (Warkentin, 2002). Egg submergence may be by rain-caused flooding or by cyclic change in water height (spring tides). In cases where immersion occurrence is less predictable, extra yolk supply within the egg extends hatch competency(Moffatt and Thomson, 1978). Egg hatching for P. modestus differs from these general examples in that, once development is complete, the guarding male triggers hatching by inundating the egg chamber. This behaviour, which usually occurs during a nocturnal rising tide (Fig. 7),may be initiated through the male's capacity to sense that egg development is complete (e.g. increased vibrations from turning larva) or through an endogenous factor (e.g. a tidal rhythm) regulating its behaviour. However,because P. modestus eggs only have a 5–6 day window for hatching competence, the male's initial selection of burrow position within the intertidal zone and the timing of spawning in its burrow relative to the tidal cycle are important factors in hatching success.

It is not known how hatched larvae make their way from the egg chamber to the open water where they will live in the plankton for about 50 days(Kobayashi et al., 1972). Preliminary field experiments suggest that relatively few of the hatched larvae are able to find their way out of the burrow. However, rapid escape is important as survival of newly hatched larvae kept in hypoxic water(P_O2=2.0 kPa) quickly decreased with time (35%in 3 h and <10% in 8 h) (Etou et al.,2007). Whether or not the guarding male is involved in transporting larvae out of the burrow remains to be determined.

Fishes in the family Gobiidae are typically substrate brooders and attach adhesive eggs to rocks, in depressions, or to the inside of enclosed spaces including burrows (Breder and Rosen,1966; Thresher,1984). In addition to egg guarding, male gobies carry out other parental-care behaviours such as preening and fanning that have been described for other species (Blumer,1982; Breder and Rosen,1966). As common among gobies, these traits appear to have developed before mudskipper's invasion of the land and have set a stage for the development of their novel reproductive strategy. Acquisition of air-breathing capability is a vital prerequisite for brooding eggs in a burrow filled with hypoxic water, which demands the storage of air within the burrow and aerial egg development. Several of the other species classified with mudskippers in the subfamily Oxudercinae also live in burrows and have the capacity to gulp air at low tide, suggesting that they are also air breathers(Graham et al., 2007). Another mudflat dweller, the eel goby Odontamblyopus lacepedii, which is sympatric with P. modestus, has recently been shown to be a facultative air breather (Gonzales et al.,2006). Odontamblyopus, however, is not amphibious and is restricted to its extensive mudflat burrow where, during low tide, it breathes air to sustain metabolism. The eel goby's reproductive behaviour is not known,however, fish that spawned in captivity attached their eggs along the upper,inner side of the submerged clay pipes in their holding tank(Dotsu and Takita, 1967),implying that burrow spawning may occur. If O. lacepedii and some of the other oxudercines are burrow-brooders then, because of severely hypoxic burrow water (Gonzales et al.,2006; Ishimatsu et al.,2000), both air storage and air-adding behaviour during low tide are likely requirements for reproductive success. Investigations into the reproductive strategies of other gobies of Oxudercinae and related subfamilies would shed light onto the evolution of subterranean aerial brooding of mudskippers.

In summary, mudskippers maximize their potential for reproductive success in the mudflat habitat with eggs that are capable of developing within the air placed in the burrow egg chambers by the male. In addition to guarding the nest, the male carries out elaborate egg-care behaviours that include gulping air and transporting it to the egg chamber at low tide to ensure adequate oxygen supply for the subsequent high tide (when air cannot be obtained), the orchestration of submergence hatch induction by removing air and thus flooding the egg chamber during a rising tide, and possibly the facilitation of burrow escape by the newly hatched larvae.

This study was partially supported by a Grant-in-Aid for Scientific Research from the Japan Society for Promotion of Science (JSPS, 13854006). Fellowships for J.B.G. and H.J.L. were from JSPS (Short-term invitation fellowship for research in Japan, RC29937012) and the Japanese Ministry of Education, Culture, Sports, Science and Technology (Summer Program). The town of Ashikari, Saga, Japan is gratefully acknowledged for its hospitality and support of the field studies. The Saga Prefectural Ariake Fisheries Research and Development Center generously offered a workplace for our gas analysis. Determinations of mud particle size were done in Seikai National Fisheries Institute, Nagasaki, Japan.