Evidence for energy reallocation, not oxygen limitation, driving the deceleration in growth of adult fish Free

For almost all fish species, ontogenetic changes in body size (growth) are biphasic – juvenile growth is rapid and adult growth declines. Despite traditional mathematical models accounting for these differences, including the renowned von Bertalanffy growth function (Von Bertalanffy, 1938), there is a long history of discussion over the underlying mechanisms responsible for growth deceleration in adult fish (Pauly, 1997; Atkinson and Sibly, 1997). Our relatively poor understanding of the drivers of growth (Marshall and White, 2019b) has been exposed by the effects of climate change, where passionate debates prevail as to why fish often grow to a smaller adult size in response to warming (Marshall and White, 2019a; Pauly and Cheung, 2018; Lefevre et al., 2017). The debates appear to have reached a stalemate, and it is clear that targeted empirical data are required if we are to reach a consensus on the underlying mechanisms.

Several key hypotheses attempt to explain the biphasic growth trends in fishes. For example, the deceleration of adult growth has been attributed to an optimal reallocation of resources (i.e. energy) (Kozłowski, 1996; Kozłowski et al., 2004). Kozłowski et al. (2004) suggest that, through ontogeny, an organism will allocate energy towards somatic growth until sexual maturation, which subsequently triggers a ‘switch’ to divert most energy towards reproduction thereafter. Growth continues to slow throughout adulthood because increasingly more energy is diverted towards reproduction (Fig. 1A). The size and age at which this switch (i.e. sexual maturation) occurs can vary under different environmental conditions, which is proposed to be an evolutionary result of selection pressures on factors such as mortality (Marshall and White, 2019a; Kozłowski, 1996). Based on this notion, if fish have access to surplus energy (i.e. high feed availability) at the time of maturation, we might expect the deceleration in growth to be less pronounced than in individuals without a food surplus (Fig. 1A).

Another possibility is that the slowing of growth is a consequence of size-specific geometric constraints. Pauly (2021) proposed the gill-oxygen limitation (GOL) hypothesis, which suggests that as gills grow in a two-dimensional manner, they cannot keep pace with the increased oxygen needs of the three-dimensional body as it grows. Thus, the GOL hypothesis posits that at a certain size when oxygen supply and demand intersect, the gills can only provide sufficient oxygen for bodily maintenance, driving growth to decelerate thereafter (Fig. 1B). Here, sexual maturation is said to be triggered by respiratory stress and hypercapnia, and reproduction as a whole does not necessarily cause the slowing of growth (Pauly, 1984, 2021). As supplemental oxygen (hyperoxia) can increase the oxygen uptake capacity of fish (Skeeles et al., 2022), it may be expected based on the GOL hypothesis that the deceleration in adult growth could be circumvented with surplus oxygen (Fig. 1B).

As far as we are aware, the ideas behind the ‘energy reallocation’ and ‘oxygen limitation’ explanations for adult growth deceleration have not been empirically tested, which is a key factor as to why debates prevail (Audzijonyte et al., 2019). To provide the missing empirical data, we tested whether adult growth is the product of resource reallocation or oxygen limitation in differently sized, female common galaxias (Galaxias maculatus) for the first 100 days post-maturation. Subsets of fish were given supplementary oxygen, supplementary energy, or a combination of the two to assess whether these environmental interventions could alleviate any constraints on the adult growth trajectory (see dashed lines in Fig. 1).

The common galaxias [Galaxias maculatus (Jenyns 1842)] is a salmoniform fish distributed across the Southern Hemisphere. In Victoria, Australia, these fish spend their first ∼5 months as larvae at sea before migrating up rivers, where they reach sexual maturation and develop into adults within the subsequent ∼6 months (Barbee et al., 2011). They can be iteroparous, living up to 3 years and attaining a maximum size of 190 mm total length (TL), but usually around 100 mm TL (Allen et al., 2002). Similar to most other fishes, their adult growth is known to decelerate as they approach maximum size (Mitchell, 1989).

Juvenile G. maculatus were caught with box traps in the Cumberland River mouth, Lorne, VIC, Australia, in early November 2020. Fish were transported by road to Deakin University's Queenscliff Marine Science Centre and left to habituate to captivity in two 125 l aerated tanks at 16.5°C and at a salinity of ∼7 ppt. In early January 2021, 440 fish were evenly distributed across two recirculating rack systems, each housing ten 25 l rectangular tanks (∼22 individuals per tank; means±s.d.: TL=54±4 mm, mass=1.17±0.41 g). The tanks in each rack drained into a 200 l sump housing a protein skimmer, and a biological and a mechanical filter, and water subsequently passed through a UV steriliser in transit back to the tanks. Both racks were raised to a summer temperature of 20°C over 16 h using a TECO chiller-heater (model TK-6000, Ravenna, Italy). One rack was kept normoxic (air saturation O_2,sat=100%) by bubbling air in the sump, whereas the other was kept hyperoxic (O_2,sat=150%) by bubbling pure oxygen near the supply pump within the sump. This ensured that the oxygen was dissolved by the supply pump in transit to the tanks. Oxygen levels within the tanks were monitored daily (YSI Pro2030 handheld meter, YSI Incorporated, Yellow Springs, OH, USA) and adjusted accordingly with a valve at the oxygen bottle. The fish were fed to satiation once a day (Otohime, granular size C2; BMAQUA, Frederickton, NSW, Australia), kept at a salinity of 7 ppt and reared until the end of their first reproductive window in September 2021 (Fig. S1). All experiments were conducted in accordance with the guidelines set by Deakin University Ethics Committee (no. B27-2018), which complies with the Australian Code for the Care and Use of Animals for Scientific Purposes set out by the Australian Federal Government.

The end of July marks the conclusion of this G. maculatus population's natural reproductive season (Barbee et al., 2011). These fish rely on spring tides to initiate spawning in the wild (Mitchell, 1989) and, therefore, natural spawning did not occur in captivity. Instead, individuals reabsorbed their gonads, visually evident in the contraction of the abdomen. On 27 August 2021, five to six females spanning the entire available size range (mass=2.55–15.78 g, TL=74–132 mm) were caught from each tank and rack. After the fish were checked for signs of gonad reabsorption (noticeable from the hardening of eggs when gently massaging the abdomen), they were anaesthetised (Aqui-S, 0.4 ml in 15 l water), injected subcutaneously with unique elastomer tag combinations in the mid-dorsal region (allowing for individual identification) and subsequently released back into their respective tanks with the untagged fish. Fish were re-caught on 1 September 2021, lightly anaesthetised, identified by their tag, and measured for length and mass. These measurements were taken as each individual's length and mass at maturation, and hence the start of their adulthood (Fig. S1). We subsequently tracked the individuals' growth for the first 100 days of their adult life, measuring length and mass at 70 and 100 days post-maturation. During this 100-day period, five tanks per rack were fed to satiation once every weekday before 10:00 h, whereas the other five tanks per rack were fed to satiation twice every weekday, with their first feed before 10:00 h and the second after 15:00 h. This resulted in an orthogonal treatment design of standard energy supply (fed once per day), excess energy supply (fed twice per day), standard oxygen availability (normoxia, O_2,sat=100%) and supplementary oxygen availability (hyperoxia, O_2,sat=150%).

To understand whether energy intake differed between treatments, we estimated each tank's feed consumption over 2 days roughly halfway through the experiment (Fig. S1; 26 and 28 October 2021). First, the inflow pumps were turned off 10 min before feeding to ensure any small food particles did not leave the outflow. The fish within each tank were subsequently fed 2 g of feed and, after 15 min, photographs were taken of any leftover food on each tank floor prior to the inflow pumps being turned back on. This procedure, excluding the taking of photographs, was initiated a month before feeding measurements to habituate fish to a lack of water movement during feeding. To quantify feed consumption, the percentage of leftover food in each photograph was compared with photographs of blank tanks, which represented what the tank floor would look like if no food was eaten (i.e. 100% food coverage). From this comparison, the amount of food consumed from the 2 g provided was calculated. As estimating the percentage coverage of leftover food in each photograph was somewhat subjective, we used the median value of estimates from five independent photograph reviewers.

For the majority of the growth analyses, we focused on the change in the TL of individuals as this is the parameter that has shown to immediately decelerate during adulthood in fishes, and because most fish lost mass from the reabsorption of their gonads. Data for five fish were excluded from analyses because of measurement/recording errors (deemed an error if the TL decreased by more than 2 mm). Using linear mixed-effects (LME) models (lmer function; Bates et al., 2015), we first assessed the change in the length of fish as a function of time (continuous, 0–100 days post-maturation), feed regime (fed once versus twice a day), oxygen treatment (normoxia versus hyperoxia) and their full interaction, whilst including each individual as a random intercept to account for repeated measures. We initially included replicate tank as an additional random effect, but this explained very little variance in growth and was therefore removed.

We subsequently conducted an optimum model selection protocol, based on the guidelines of Zuur et al. (2009), to identify the best explanatory variables of adult growth in G. maculatus. An LME model was first fitted using maximum-likelihood (ML) estimation with length (log-transformed) as a function of the full interaction between time (days post-maturation), feed regime, oxygen treatment and the length of each fish at sexual maturation (continuous, log-transformed and centred to the mean), with individual as a random intercept. Including each fish's length at sexual maturation allowed us to identify whether constraints on adult growth differed for smaller or larger fish. Fixed effects were then dropped systematically and each new model was fitted with ML estimation. The quality of each model was subsequently compared using the sample-size-adjusted Akaike information criterion (AICc). The best model was selected based on the lowest AICc and refitted with restricted ML estimation to assess the effect size and confidence intervals of each variable. If ΔAICc was <2 between models, the model with the most applicable explanatory variables to test the hypotheses was chosen.

For the feed intake analysis, linear relationships were established between grams of food consumed and tank biomass for each of the four treatment groups (five tanks per group and two measurements per tank), then residuals were calculated for each tank and added to the regression-derived estimate of food consumed per 100 g of biomass. Here, we used a LME model with feed intake as a factor of the full interaction between feed regime (once versus twice) and oxygen treatment (normoxia versus hyperoxia), and replicate tank as a random intercept to account for repeated measures.

All statistics were performed in R (version 4.12; https://www.R-project.org/) and LME models were fitted using the lme4 package (Bates et al., 2015). Wherever LME models were used, assumptions of normality and homogeneity of variance were thoroughly assessed with a variety of residual plots, and the significance of predictor variables was estimated using a type III ANOVA. For visual purposes, in the figures, only measurements from days 0 and 100 are shown, but all statistics were conducted on measurements from days 0, 70 and 100.

Adult TL significantly increased over the experimental period by an average of 5 mm [Table S1, row 3, 95% confidence interval (CI) 4–7 mm]. This growth was considerably improved by additional energy in the diet, as the average TL of fish fed twice a day increased by 7 mm (Fig. 2A; Table S1, row 5, 95% CI 4–9 mm). In contrast, supplementary oxygen had no effect on adult growth (Fig. 2B) nor did its interaction with additional energy (Table S1, rows 7 and 8, respectively).

The optimum model, which described 96% of variance in the relative adult growth of G. maculatus, comprised the full interaction between time post-maturation, size at sexual maturation and feed regime as explanatory variables of TL (Table 1). TL increased with time, but this increase was significantly less for fish that sexually matured at a larger size compared with smaller conspecifics (Fig. 3; Table S2, row 5, factor: length at maturation×time; slope effect −1.128×10⁻³; 95% CI −1.87×10⁻³ to −3.86×10⁻⁴). Interestingly, additional energy (fish fed twice a day) tended to alleviate this size-related effect (Fig. 3; Table S2, row 8, factor: length at maturation×time×feed level; slope effect 7.44×10⁻⁴; 95% CI −3.58×10⁻⁴ to 1.85×10⁻³). Although this three-way interaction effect between length at maturation, time and feed regime was not statistically significant, the presence of the factor in the optimum model, as well as the positive skew of the confidence intervals, reveals its importance to growth. Thus, the adult growth of larger fish was, to some degree, energy limited (Fig. 3). Again, the importance of oxygen was minor, hence the omission of this factor from the best model (Table 1).

Body mass generally decreased over time for fish fed once a day. Specific mass change (percentage of body mass per day) was significantly less than zero in normoxia but only tended in the same direction for hyperoxia (Fig. 4A). Normoxic and hyperoxic fish that were fed twice per day consumed additional food (3.15±0.26 g and 3.34±0.45 g per 100 g biomass per day, respectively) compared with the equivalent oxygen groups fed once per day (1.88±0.04 g and 1.91±0.09 g per 100 g biomass per day, respectively) (Fig. 4B; Table S3), which enabled the former groups to maintain body mass (Fig. 4A).

Enthusiastic debates remain as to whether the slowing of adult growth in fish is a product of energy reallocation, oxygen limitation induced by size-related geometric constraints, or some other factor(s). The findings of the present study were not as clear-cut as we had anticipated, but we found evidence that energy availability, not oxygen limitation, is a more important factor in regulating the adult growth trajectory of G. maculatus. Supplementary oxygen elicited no change to the growth of female fish over the initial 100 days of adulthood, yet fish that were offered additional energy exhibited improved growth. Interestingly, we found that these energy-related growth improvements were mostly prevalent in larger adults, suggesting a size dependence to the observed growth limitations.

Hyperoxia has been shown to elevate the maximum oxygen uptake capacity of fish (Brijs et al., 2015; McArley et al., 2021a, 2018), including G. maculatus (Skeeles et al., 2022), which may function by alleviating a potential gill-oxygen diffusion limitation (McArley et al., 2022). Thus, if size-related constraints on gill surface area and oxygen uptake were responsible for the deceleration of growth in adult fish, as implied by the GOL hypothesis, then growth improvements would be expected under hyperoxia. We did not find evidence for this idea in G. maculatus chronically held in hyperoxia, adding to a growing body of literature (reviewed by McArley et al., 2021b) questioning the suggestion that fish growth is oxygen limited under normoxic conditions. Furthermore, smaller female G. maculatus did not show signs of growing to match the size of their larger conspecifics, despite unarguably having the gill surface area to do so, illustrating that factors other than gill-oxygen uptake are at play in shaping the adult growth of fishes (Kozłowski, 1992).

In contrast to the negligible effects of oxygen, supplementary energy improved the adult growth of G. maculatus. As suggested by Kozłowski (1996), a proportion of energy is diverted towards reproduction after an animal with indeterminate growth reaches sexual maturation. Animals must generally fulfil maintenance requirements first (as shown in Jokela and Mutikainen, 1995) and can then allocate energy towards reproduction (usually in the form of energy reserves for the next reproductive season) and somatic growth, either simultaneously (Rijnsdorp, 1990) or interchangeably (Vahl, 1981). This may explain the improved adult growth of fish fed twice versus once a day in the present study, because after allocating energy to maintenance and reproductive preparedness, these individuals had surplus energy available for continued growth. Similarly, Beauvieux et al. (2022) found that the growth of adult sardines Sardina pilchardus during their reproductive winter months improved when supplied with additional energy. Analogous evidence comes from studies showing that surplus energy can bolster the growth of fishes during the energy-taxing reproductive vitellogenesis process (Yoneda and Wright, 2005; Kennedy et al., 2008). Thus, our results contribute to the idea that there is a trade-off between energy allocation towards growth and reproduction in adult fishes, suggesting that the deceleration of adult growth stems – at least partly – from the onset of energy reallocation towards reproduction.

As revealed by our optimal explanatory model (Table 1; Table S2), the adult growth of G. maculatus that matured at a larger size was less than that of their smaller counterparts, and their growth was also more responsive to supplementary energy (Fig. 3). These intraspecific, size-dependent variations in growth at summer temperatures are supported by Lindmark et al. (2022), who found that optimum temperatures for net energy gain (consumption minus resting metabolism) decreased with body size within fish species. Our findings, combined with those of others, may provide evidence for the benefits of reproducing small in warmer settings, reducing the likelihood of being energy limited during adulthood and enabling adequate allocation towards maintenance, reproduction and continued somatic growth (Van Noordwijk and De Jong, 1986). Further investigation into the allometry of resource acquisition and partitioning between growth and reproduction in fish is needed (Jutfelt et al., 2021; Audzijonyte and Richards, 2018), but there is evidence that reproductive requirements may be disproportionately greater for larger individuals and this may be exacerbated with warming (Moffett et al., 2022; Barneche et al., 2018). As optimal resource allocation forms the central tenet of life history theory, we encourage eco-physiologists and life history theorists to work together in understanding body-size responses of fish to climate warming (Roche et al., 2022).

Mounting observations show that many fish species are entering adulthood at an earlier age and smaller size in response to warming (Gardner et al., 2011; Sheridan and Bickford, 2011). Our results suggest that energy availability may be of increasing importance in shaping the size to which fish grow at warmer temperatures (Arendt and Reznick, 2005; Dhillon and Fox, 2004), implying that productive ecosystems will be crucial for adult fish to maximise growth and their associated reproductive potential. Thus, to help mitigate the effects of climate warming and sustain fish populations, management strategies that have been shown to enhance productivity, such as no-take marine-protected areas (Lester et al., 2009), may play an increasingly important role. Moreover, reduced levels of mortality within these protected areas may allow fish to invest more in juvenile growth, delay the onset of sexual maturation and enter adulthood at a larger size (Álvarez-Noriega et al., 2023), which is particularly important given the hyperallometry of reproduction in fish (Barneche et al., 2018).

We thank Hanna Scheuffele for experimental assistance and James Redmond, Sam Wines and Lisa Grubb for their technical support. We also thank Elizabeth Hoots, Luis Kuchenmüller and Beichen Yang for reviewing photographs for the food intake analyses.