Embryo oxygenation in pipefish brood pouches: novel insights

Parental care is a phylogenetically widespread phenomenon (Clutton-Brock, 1991; Reynolds et al., 2002). In fish, lack of parental care is the norm, but in species that do provide parental care, paternal care is more common than maternal or biparental care (Gross and Sargent, 1985; Reynolds et al., 2002). Parental care, which in fish is largely expressed as fanning behaviour and brood protection (Blumer, 1979, 1982; Clutton-Brock, 1991; Lindström and St Mary, 2008), has generally been linked to greater offspring survival and to the production of larger eggs (Shine, 1978; Sargent et al., 1987; Kolm and Ahnesjö, 2005).

Aquatic environments have much lower O₂ availability compared with air (Keister et al., 2000). This fact has led to the assumption that egg size evolution is constrained in aquatic environments because of the limited O₂ availability (Rombough, 1988). Because O₂ travels by diffusion through the pores in the egg membrane, it is thought that embryos from larger eggs have greater difficulties acquiring sufficient O₂ for successful development, as a result of the unfavourable surface to volume ratio characteristic of sphere-shaped eggs. This assumption has lingered in the literature for decades despite remaining largely untested (van den Berghe and Gross, 1989; Quinn et al., 1995; Einum et al., 2002). At the same time, the evolution of parental care in fishes has been suggested to have facilitated the evolution of larger egg sizes (reviewed in Kolm and Ahnesjö, 2005), because an important parental task is the provision of O₂ to the developing embryos, through fanning behaviour or choice of well-oxygenated nest sites (Lukas and Orth, 1995; Takegaki and Nakazono, 1999; Takegaki, 2001; Wisenden et al., 2009). In this way, the parental care period can be viewed as a ‘safe harbour’ to the embryos and, with larger egg sizes, this phase can be prolonged (Shine, 1978; Kolm and Ahnesjö, 2005). Can the syngnathid male pregnancy have a similar function?

Because of their unique modes of parental care and their diverse mating systems, pipefishes and seahorses (family Syngnathidae) have been extensively studied over the past decades. All syngnathids show paternal care by carrying the offspring on their body throughout embryo development. Depending on genus, the eggs may be attached to the skin of the male or nurtured inside a brood pouch, consisting of two skin flaps common among pipefishes, or in a sack-like structure typical of seahorses, placed ventrally on the male's tail or upper body (Herald, 1959; Dawson, 1985; Wilson et al., 2001,, 2003). During mating, the female transfers her eggs to the male, by inserting her ovipositor through an opening at the top of the brood pouch or sliding it along the male's body. More complex brood pouches enable males to physically protect the embryos from external factors such as predators, and to provision them with nutrients, osmoregulation and oxygenation during brooding (Wilson et al., 2001). Male provisioning of nutrients to the developing young has been shown in a number of pipefish species (Haresign and Shumway, 1981; Ripley and Foran, 2006, 2009; Kvarnemo et al., 2011). Also, the function of the brood pouch in osmoregulation is well established (Quast and Howe, 1980; Carcupino et al., 1997; Partridge et al., 2007; Ripley, 2009). Oxygenation during brooding, however, has remained surprisingly less explored (Carcupino et al., 2002; Monteiro et al., 2005; Stölting and Wilson, 2007), although the presence of a vascularised dermis in all pouch types studied so far suggests a common respiratory function to all species, independent of other types of care provided during brooding (Carcupino et al., 2002).

Here, we used the broad-nosed pipefish, Syngnathus typhle Linnaeus, to explore paternal ability to actively provide O₂ for successful embryo development. Females mature large, nutritious eggs (egg diameter range 1.5–2.2 mm; Braga Goncalves et al., 2010,, 2011) continuously, allowing them to mate and transfer multiple batches of eggs within a short period of time (Sogabe and Ahnesjö, 2011). Males typically mate with one or several females, until their pouches are filled, at which point they commence brooding (Berglund et al., 1988, 1989; Jones et al., 1999). The number of embryos a fully mated male can brood at a time can vary substantially (between 30 and 180 embryos) because, as males grow, the brood pouch becomes longer and wider, allowing males to accommodate more and longer rows of embryos (I.B.G., I.A. and C.K., personal observation). Similar to Syngnathus floridae (Ripley and Foran, 2006), S. typhle males have two ventral pouch flaps along the tail that seal in the middle at the onset of brooding, isolating the embryos from the external environment (Fig. 1). At parturition, the seal is broken, allowing the pouch flaps to separate, and the free-living juveniles are expelled from the pouch by paternal shakes and body contractions. Because S. typhle males provide extensive care to the developing offspring during a prolonged brooding period (Berglund et al., 1986b), this is an ideal species in which to investigate paternal ability to oxygenate the embryos and to maintain O₂ levels during a significant proportion of the brooding period, in both high and low ambient O₂ conditions. By exploring brooding males' ability to provide O₂ to the developing embryos, we provide insight into the hypothesis that paternal embryo oxygenation may have promoted the evolution of larger eggs in the syngnathid species that have enclosed brood pouches (reviewed in Braga Goncalves et al., 2010; Braga Goncalves et al., 2011).

The aims of this study were to measure pouch fluid O₂ saturation levels at regular intervals (every 6 days) and in different sections of the pouch during 24 days of brooding, and assess whether: (1) males are able to maintain O₂ saturation levels in the pouch fluid that are high enough to support embryo development, even in low ambient O₂, (2) pouch fluid O₂ saturation levels are uniform throughout the brood pouch, (3) embryo size and survival are affected by long-lasting low ambient O₂, and (4) male body size affects paternal ability to oxygenate the embryos. Under normoxia, we expected brooding males to be able to provide O₂ to sustain normal embryo development, and to be able to keep pouch fluid O₂ saturation levels similar to the surrounding water. Under hypoxia, however, male condition should be negatively affected, as O₂ demands of both the male and the offspring are expected to increase while metabolic efficiency is decreased (Kramer, 1987). For that reason, we expected a reduced ability of males brooding in hypoxia to oxygenate the developing embryos and this impact was expected to be more noticeable at more advanced stages of development, as embryos then are larger and require more O₂ (Berglund et al., 1986b). Like other fishes, S. typhle individuals grow throughout their lives and it is well documented that larger males are preferred partners and have greater reproductive success than smaller males (Berglund and Rosenqvist, 1990,, 2003; Ahnesjö, 1992b). Thus, as small males brood fewer embryos (lower reproductive value) because of their own smaller size, they may gain more from directing their resources into self-maintenance and growth to improve their future reproductive success than from providing excellent care. In other words, we might expect small males to provide lower quality of care to the developing embryos, here measured as lower pouch fluid O₂ saturation levels.

Analysis of pouch fluid O₂ saturation levels revealed several significant effects (Table 1). Male standard body length (SL) had a significant effect on O₂ saturation levels (Table 1), with larger males having, on average, higher pouch fluid O₂ levels (Table 1, Fig. 2). Despite high individual variation and significant male × day variation, males brooding in hypoxia showed significantly lower pouch fluid O₂ saturation levels (mean±s.e. of all days and pouch sections combined for each male, normoxia: 51.3±2.4%, hypoxia: 32.5±1.7%; Table 1). In both treatments combined, O₂ saturation levels in the bottom section of the pouch (45.1±2.2% O₂) were significantly higher than in the middle section (39.1±2.0% O₂, t=3.75, d.f.=135, permutational probability P_perm<0.001) and tended to be higher than in the top section of the pouch (41.2±2.6% O₂, t=1.85, d.f.=135, P_perm=0.071), whereas top and middle sections did not differ significantly (t=1.34, d.f.=135, P_perm=0.183). Furthermore, average O₂ saturation levels decreased over the course of the brooding period (Table 1, Fig. 3). Pair-wise post hoc comparisons showed that, irrespective of treatment, O₂ levels were significantly higher on day 6 than on any other day (day 6–day 12: t=3.51, d.f.=33, P_perm=0.002; day 6–day 18: t=4.26, d.f.=33, P_perm<0.001; day 6–day 24: t=5.36, d.f.=33, P_perm<0.001). O₂ levels were also significantly higher on day 12 than on day 24 (t=2.21, d.f.=33, P_perm=0.031) but not on day 18 compared with day 24 (t=1.68, d.f.=33, P_perm=0.096; Fig. 3).

Despite significant variation between brooding aquaria, reduced O₂ treatment affected overall embryo survival and dry mass negatively (Table 2). Separately, we detected no effects of treatment or tank on relative embryo survival, but embryo mass was significantly lower in the hypoxia treatment (Fig. 4A) and varied significantly across brooding aquaria (Table 3).

Embryos brooded in hypoxia were significantly shorter in length than those brooded in normoxia (Table 4). In addition, embryo length was significantly affected by the section of the pouch in which the embryos had developed (Table 4), such that embryos brooded in the bottom section were significantly longer than those brooded at the top (t=2.42, d.f.=30, P_perm=0.018), with the embryos from the middle section presenting intermediate lengths, not significantly different from either of the other sections (top–middle: t=1.08, d.f.=30, P_perm=0.284; middle–bottom: t=1.68, d.f.=30, P_perm=0.104; Fig. 4B).

Male condition was significantly affected by O₂ treament, with males that brooded in hypoxic conditions presenting lower hepatosomatic indices (HSI; mean±s.e: 1.98±0.09%) than males that had brooded in normoxia (mean±s.e.: 2.81±0.17%, F_1,5.27=15.26, P_perm<0.001).

Overall, mean embryo survival in the two pouch areas between the three insertion points was significantly correlated to the mean survival of embryos in the areas where the O₂ probe was introduced in the pouch (Pearson correlation: r=0.71, P<0.001, N=34). However, in a paired comparison, the insertion of the probe in the pouch had a negative effect on local embryo survival, as embryos from the target areas presented significantly lower survival (mean±s.e.: 71.5±2.9%) than embryos in the areas between the insertion points (mean±s.e.: 80.8±3.1%, paired t-test: t=−4.89, P<0.001, d.f.=33).

Using a very fine oxygen probe, this study is the first to measure O₂ saturation levels inside the brood pouches of pregnant pipefish males. We investigated the ability of brooding broad-nosed pipefish (S. typhle) males to oxygenate the developing embryos in high and low O₂ conditions during the first 24 days of brooding and we found that pouch fluid O₂ saturation levels were clearly affected by the ambient O₂ conditions experienced by the males. Pouch fluid levels were significantly lower in males brooding in hypoxic water than in males brooding in normoxic water and, importantly, we found that in both normoxia and hypoxia, O₂ saturation levels inside the brood pouches were much lower than those of the surrounding environment. Thus, in accordance with our predictions, hypoxia negatively affected the male's ability to oxygenate the developing brood, resulting in smaller, though not fewer, embryos. In addition, smaller males demonstrated a lower ability to oxygenate the embryos compared with larger males. This finding confirms our expectations and provides an additional explanation for the strong female preference for large partners reported in this population (Berglund and Rosenqvist, 1990, 2003).

Pregnancy, although ubiquitous in therian mammals and common in squamate reptiles, also occurs in fishes (Mank and Avise, 2006; Stölting and Wilson, 2007). The pipefish species investigated here has a completely closed pseudo-placental structure (the brood pouch) with blood vessels and capillaries through which O₂ is transported to the embryos during the prolonged brooding period (Ripley et al., 2010). Our findings, which are based on repeated measurements of O₂ saturation levels in the pouch fluid during the first half of embryo development, provide important new insights into the evolution of male pregnancy in syngnathids, by unveiling an important limitation in the brooding ability of this brood pouch-adorned species: at any ambient O₂ saturation, embryos of S. typhle have less O₂ available for their development than embryos that are in direct contact with the surrounding water. Thus, our results provide evidence against the notion that brood pouches should be viewed as well-oxygenated structures that promote the evolution of larger eggs in syngnathids (Shine, 1978; Kolm and Ahnesjö, 2005). They also put into question the generally accepted hypothesis that more complex types of brood pouches, like that of S. typhle, equal better paternal care (Wilson et al., 2001).

Interestingly, we found differences in O₂ saturation levels between the different sections of the pouch. O₂ levels were lowest in the middle section, which may result from a greater number of embryos in this section, where the pouch is widest and embryos in multiple layers are most commonly observed (I.B.G., I.A. and C.K., personal observation). The highest levels of O₂ were found in the bottom section of the pouch, coupled with the presence of significantly larger embryos. Because of the way eggs are transferred into a male's pouch and the possibility for males to mate with multiple females before the pouch is completely filled (Berglund et al., 1988; Jones et al., 1999), the eggs at the bottom of the pouch are the first to be received and fertilised by the male, and, arguably, are the first ones to begin to develop. This potential difference in the timing of initiation of development, in combination with a usually smaller number of embryos in this area, may explain why embryos at the bottom of the pouch were better oxygenated and larger than those throughout the rest of the pouch. Indirect negative effects of egg density on embryo development have been shown in other species, including the rainbow trout, Onchorhynchus mykiss (Dhiyebi et al., 2013), due to changes in water flow in the vicinity of the eggs.

Pouch fluid O₂ saturation levels decreased throughout the brooding period, presenting a sharp decrease between days 6 and 12, irrespective of O₂ treatment. This decrease in O₂ availability in the pouch is likely to be a result of increased respiration by the embryos as they grow (Berglund et al., 1986b). Greater O₂ consumption by embryos with the progression of development is well known and has been recorded, for instance, in rainbow trout (Rombough, 1988). Moreover, in a tropical clown fish, Amphiprion melanopus, caring males have been reported to increase fanning effort as the embryos develop and O₂ demands increase (Green, 2004; Green and McCormick, 2005). However, it is also possible that brooding S. typhle males struggle to maintain O₂ levels over the brooding period. In our study, it was not possible to determine to what extent the lower within-pouch O₂ saturation levels later in the pregnancy are the result of reduced paternal ability to supply O₂, increased O₂ consumption by the embryos, or a combination of the two.

Embryo survival did not differ between the O₂ treatments, but embryos were significantly smaller in the hypoxic treatment at the termination of the experiment. Body size is a key fitness correlate in S. typhle, both in adults and in juveniles. Because by parturition the yolk sacs have usually been fully resorbed (Sommer et al., 2012), juveniles switch to external feeding once released from the paternal brood pouch. A previous study has shown that larger juvenile size at birth is associated both with subsequent faster growth rates, presumably due to higher foraging success, and with higher survival, due to greater success at escaping predators (Ahnesjö, 1992a). Thus, the significant reduction in embryo size we found in just under half the length of a full brooding period, due to ambient hypoxia, is indicative of larger size differences by parturition and suggestive of important detrimental fitness consequences.

Low ambient O₂ conditions lead to costly increases in parental effort. For instance, in common gobies, Pomatoschistus microps, males caring in hypoxic conditions fan more and lose more weight, but maintain similar hatching success to males caring in normoxic conditions (Jones and Reynolds, 1999). In our study, brooding in hypoxic conditions significantly lowered male condition (measured as HSI), showing that brooding in poorly oxygenated waters is energetically costly to the males. However, male condition was not related to male length. This result that small and large males showed similar condition within O₂ treatment is interesting, because large males maintained significantly higher within-pouch O₂ saturation levels compared with smaller males. This difference in quality of care is likely to be important and may further explain the strong preference shown by females of all sizes for large mates (Berglund et al., 1986a; Berglund and Rosenqvist, 2003). As sexual selection acts on body size in both sexes, small males may be less willing to invest in current reproductive events if future reproductive opportunities are likely to yield higher fitness pay-offs (Magnhagen, 1990).

We used a technique that allowed us to directly assess O₂ saturation levels in the pouches of brooding pipefish males. Specifically, this technique enabled us to assess pouch fluid O₂ saturation levels throughout the pouch and saturation changes over a long period of time, within the same individuals, i.e. minimising the number of experimental subjects. The technique worked well and measurements were repeatable and reliable. However, the use of the O₂ probe, which is an invasive technique, affected embryo survival negatively, with a slightly higher number of unviable embryos occurring around the insertion points than in the sections between the insertion points. This effect suggests that, even though unique and valuable information is obtained, this technique should be used with caution in future studies, in particular for repeated measurements in the same points.

To conclude, we report two important findings. Firstly, we found that pouch fluid O₂ saturation levels are generally lower than those of the surrounding water, vary throughout the brood pouch and decrease over time. These results highlight a potential O₂ limitation to brooding within the closed environment of the brood pouch. The evolution of the brood pouch among syngnathids has been conceptually linked to more advanced types of care because of the potential to osmoregulate (Quast and Howe, 1980; Carcupino et al., 1997, 2002; Partridge et al., 2007; Ripley, 2009) and to provide nutrients (Haresign and Shumway, 1981; Ripley and Foran, 2006, 2009; Kvarnemo et al., 2011) to the developing young – abilities that are limited in species that do not have brood pouches (Berglund et al., 1986b; Carcupino et al., 2002). Therefore, it is important to note that, in terms of oxygenation, the presence of a brood pouch is here shown to limit, not ease, embryo access to O₂ and thus limit their development.

Secondly, within-pouch O₂ saturation levels were positively correlated with male length. We know that S. typhle females strongly prefer to mate with larger partners compared with smaller ones (Berglund et al., 1986a; Berglund and Rosenqvist, 2003) but direct evidence of differences in parenting quality of small and large males is still lacking (Braga Goncalves et al., 2010,, 2014). Here, we provide direct evidence that, in terms of oxygenation, large males provide better care to the developing offspring than small males do, which is particularly important when embryo development is likely to be restricted by O₂ availability, as shown by the negative effects on embryo length and mass. This study, thus, provides important contributions to the understanding of the function of the brood pouch during male pregnancy in syngnathids in general (Stölting and Wilson, 2007), and improves the knowledge of factors guiding mate choice in this model species in particular.

The study was carried out on the Swedish west coast at the Sven Lovén Centre for Marine Sciences, Kristineberg (58°15′N, 11°28′E) between May and July 2009. Syngnathus typhle pipefish were collected from eelgrass (Zostera marina) meadows with a beam trawl pulled behind a boat in the vicinity of the marine station.

Fish were brought into the lab, separated by sex and kept in storage barrels (225 l). Both barrels and brooding aquaria were provided with artificial eelgrass, a flow-through system of continuously renewed natural seawater providing natural salinity conditions, and light 16 h per day. Air and water temperatures were kept constant at 15°C. Fish were fed three times per day with Artemia salina and supplemented with wild-caught shrimps (Praunus flexuosus and Crangon crangon) and copepod species. Storage barrels were cleaned daily and brooding aquaria every third day.

Matings were performed in barrels by introducing groups of males and females together. Because a large S. typhle female produces enough eggs to fill up almost three males of her size during the course of one male pregnancy (Berglund and Rosenqvist, 1990; Ahnesjö, 1995), females and males were put together in a ratio of 1:2 (eight females and 16 males). This ratio ensured that males mated and had their brood pouches filled up quickly. Males were kept in the barrels until their pouches were filled (24–72 h). One day after being fully mated, males were measured and individually identified before being transferred into brooding aquaria (approximately 140 l), where they were kept in groups of six or seven individuals for a period of 24 days. Brooding aquaria had either fully oxygenated water (100% O₂ saturation) or hypoxic water (40% O₂ saturation). The concentration of dissolved O₂ was decreased with the use of a MiniModule 1.7×5.5 Membrane Contactor (Liqui-Cel, Celgard, Inc., Charlotte, NC, USA). This unit enables the flow of nitrogen in a counter-current system in relation to the direction of the water flow, continuously removing O₂ from the water. Hence, we were able to provide a flow-through seawater system in the hypoxic aquaria, similar to the conditions in the normoxic aquaria. O₂ levels in the treatment aquaria were checked daily using a portable O₂ meter (Handy Delta, OxyGuard, Tekno Trading AB, Säffle, Sweden) and we adjusted water and nitrogen flow when necessary. In the normoxic brooding aquaria, flowing seawater was supplemented with air flowing through air stones to maintain high ambient water O₂ saturation conditions. Males were allocated to treatments at random, only ensuring that treatments had similar sample sizes and body size ranges.

The brooding period was set to 24 days, which amounts to little over half of the total brooding period of S. typhle at our experimental temperature. At experiment termination, males were killed with an overdose of anaesthetic (10 min in 1 ml 2-phenoxyethanol/litre seawater) followed by severing the spinal column posterior to the operculum. Each male was preserved in 70% ethanol in individual vials for posterior assessment of embryo survival and size. Ethical approval for this study was given by the Swedish Animal Welfare Agency (dnr. 111-2007).

The SL and body depth of all females, to the nearest mm, were taken before they were introduced into the mating barrels. The SL of males was similarly measured before they were introduced into the mating barrels and again after mating for identification purposes. A range of male sizes (SL range: 150–203 mm, mean±s.e.: 177±2.5 mm, N=34) was used to assess differences in paternal care in relation to body size. Because egg size correlates positively with female SL (Berglund et al., 1988), only large females (SL range: 231–276 mm, mean±s.e.: 249±1.87 mm; body depth range: 9.6–13.7 mm, mean±s.e.: 11.3±0.15 mm) were chosen to attempt to standardise egg size among all males and, thus, minimise confounding effects of varying egg sizes on embryo length and mass. Within each brooding aquarium, males could be recognised by their SL and colour, as individuals' body coloration may range from silvery black to brown and pale green. After mating, all females were released close to the site of capture.

The barometric pressure at the study site ranged between 747.80 and 777.82 mmHg, with an average of 759.99 mmHg during the course of the experiment. Hence, we assumed an atmospheric pressure of 760 mmHg (or 1 atm, as expected at sea level). At a temperature of 15°C, O₂ solubility in seawater at 24 ppt salinity is 8.70 mg l⁻¹ O₂. Thus, at 100% O₂ saturation, seawater had 8.70 mg l⁻¹ O₂, corresponding to a partial pressure of 156 mmHg; at 40% O₂ saturation, O₂ concentration was 3.5 mg l⁻¹ O₂ and the respective partial pressure was 63 mmHg. O₂ measurements were taken with a Foxy-AL300-AF probe, connected to a laptop through a MFPF100-1 O₂ sensor (OOISENSORS, Ocean Optics, B.V., Duiven, The Netherlands). The O₂ sensor was calibrated weekly, in the same temperature-controlled room where the experiments were carried out and measurements taken. Two standards were used for calibration. The first standard, at 0% O₂ saturation, was obtained by bubbling nitrogen into the sample until O₂ saturation stabilised at 0%. The second standard, at 100% O₂ saturation, was obtained by continuously bubbling air into the sample. Both standards were made of seawater collected from the same taps used to feed water into the experimental tanks, and thus had similar salinity levels. The O₂ sensor was set to take measurements every 5 s.

On days 6, 12, 18 and 24 of the brooding period, O₂ saturation levels (%) were measured from three different sections inside the brood pouch (top, middle and bottom; Fig. 1A). Each male was briefly removed from the experimental aquarium and placed into a silicon tube that had a middle section cut out (Fig. 1C). Because the fish could turn around in the tube, an assistant helped to hold the males in place while measurements were recorded by gently squeezing the sides of the tube at head and tail points. The tube was filled with water from the experimental aquaria in order to keep the surrounding O₂ saturation levels similar to experimental conditions. For each section of the pouch, the needle of the probe was inserted into the brood pouch along the sealed line where the two pouch flaps meet and kept in place for 1–2 min, until three measurements were recorded within 1% of O₂ saturation. The average value of these three recordings was used in the analyses. In between consecutive measurements at the same site, the needle was kept in place. Once the measurements in one section were concluded, the probe was rinsed and put back in the 100% O₂ saturation standard to confirm the quality of the readings, before it was inserted in the next section. For each male, the whole procedure took less than 5–10 min, from the moment the male was removed from the experimental tank, until it was returned. If the male was kept in the tube for more than 5 min, the water was replaced with new water. Measurements were recorded from the same entrance points in the pouch throughout the experiment. A small scar that closed up between collection days (i.e. in between day 6 and day 12, day 12 and day 18, etc.) formed at each of the entrance points. The scar thus marked the exact point of the previous recording and ensured that we took our readings from the same points in the pouch each time.

The brood pouches of all males were dissected and overall embryo survival was assessed for each male by dividing the number of developing embryos by the total number of eggs transferred by the females (sum of undeveloped and developing embryos, and unviable eggs in the pouch). The eggs in the pouch were further separated into the three measurement sections and the two sections in between to assess relative embryo survival in each section and to compare average embryo survival between the two sections: with or without an entrance point for the O₂ probe.

Five embryos each from the top, middle and bottom of the pouch of each male were separated from the egg membrane and yolk sac. Pictures of the embryos were taken using a camera (Leica DFC420 C) attached to a stereo microscope (Leica MZ16 A). The total length (tip of rostrum to tip of tail) of each embryo (±0.01 mm) was measured from the photographs using Leica Application Suite, version 2.7.0.RI (Build: 1294, Heerbrugg, Switzerland). For each male, the average embryo length was calculated for each of the three pouch sections.

Ten embryos, selected at random from the brood pouch, were separated from the egg membrane and yolk sac, and dried in a heating cupboard (60°C) for 1 week then weighed twice on a Sartorius LE26P microbalance (±2 µg). To calculate the average dry mass per embryo for each male, we divided the average of these values by 10.

We dissected the males and removed their livers. Both males and livers were dried in a heating cupboard (60°C) for 1 week and weighed on a Sartorius LE26P microbalance (±2 µg) twice each. The average liver mass was divided by the total body mass (excluding the liver mass; Tomkins and Simmons, 2002) to calculate the HSI for each male, reported as percentages.

We used PERMANOVA+ for PRIMER v6 (PRIMER-E, Plymouth, UK) and SPSS 17 (SPSS Inc., Chicago, IL, USA) to perform all analyses. The Permdisp routine was used to assess the spread of the data and ensure that differences detected were due to differences in means rather than differences in dispersion of the data. Data were analysed with the Permanova routine in PERMANOVA+ because of the small and unequal sample sizes between brooding aquaria, which reduced the reliability of parametric tests.

Before conducting the analyses, O₂ readings and relative embryo survival were arcsine transformed. Variables were also normalised (each variable had its mean subtracted and was divided by its standard deviation) before multivariate analyses, in order to achieve a common scale. Resemblance matrices based on Euclidean distances (Clarke and Gorley, 2006; Anderson et al., 2008) were produced for each statistical test.

We used a nested-ANCOVA design to explore the effects of male SL (covariate), O₂ treatment [fixed factor (FF): 100% and 40% O₂], tank [random factor (RF), nested within O₂ treatment, three aquaria per treatment], male (RF, nested within tank, 34 individuals), pouch section (FF: top, middle and bottom) and day (FF: days 6, 12, 18 and 24) on pouch fluid O₂ saturation levels. The effects of male SL, O₂ treatment and brooding tank on overall embryo dry mass and relative survival were determined with a nested-MANCOVA design, followed by separate nested ANOVA on the individual response variables. The effects of male SL, O₂ treatment, brooding tank and pouch section on embryo length were analysed using a nested ANCOVA design. A nested ANCOVA design was also used to explore the effects of male SL, O₂ treatment and tank on male condition (HSI). Male SL was removed from the analyses of embryo length, dry mass and relative survival, and of male HSI, because it did not affect the response variables significantly (all P_perm>0.65). Models were sequentially reduced by pooling non-significant factors (interactions and main factors, when P_perm>0.15), one at a time, starting with highest degree interactions followed by within-level lowest significance factors, to improve the power of the remaining factors. ANCOVA models were tested using type I sums of squares; ANOVA-type models were tested using type III sums of squares. The following options were chosen for all analyses: fixed effects sum to zero, model: permutation of residuals under a reduced model, number of permutations: 9999. Significance level for all tests was set at P<0.05, except for removed terms as explained above.

We thank M. Hällbom, K. Olsson, I. Duranovic and G. Sagebakken for assistance in the field, A. Berglund and G. Rosenqvist for equipment, M. J. Anderson for statistical advice and G. Sagebakken, O. Jennersten and L. Hammar for photographs. We are grateful for the facilities and support provided at Sven Lovén Centre for Marine Sciences – Kristineberg, University of Gothenburg.