The convict surgeonfish, Acanthurus triostegus (Linnaeus, 1758), was chosen as a model to carry out such a study given its widespread distribution in the Indo-Pacific region (Randall 1961). At Moorea (French Polynesia), the larval stage is about 44 to 60 days, after which larvae colonize the reef at night and settle directly on shallow sandy areas with coral slabs (McCormick 1999, Lecchini and Galzin 2005). When sexual maturity is reached (standard length, SL=95 mm), adults move towards the barrier reef community, where they usually form large feeding schools (Randall 1961). Larvae feed on zooplankton while juveniles and adults graze on algae (Frédérich et al. 2012).

Sample collection was carried out in September-October 2009 in the lagoon of Moorea (17°30’S, 149°50’W; French Polynesia; Fig. 1). Settling larvae of Acanthurus triostegus (n=81), also called “naïve” larvae, were obtained from nets fixed to the reef crest at dawn (Lecchini et al. 2006). The first surgeonfish larvae, which were captured all at once (n=17) and formed one cohort, were sacrificed and represent the t₀ in the following framework. The other larvae (n=64), caught during another retrieval of nets, form a second cohort which was divided into two equal parts. Each half was allocated to a feeding treatment (n=32 for treatment 1 and n=32 for treatment 2). For a comparison between treatment 2 and juveniles collected in the field, we used the isotopic data of the epaxial musculature from 91 individuals (SL, 22.45-55.25 mm) captured at the same location and same date, and published in Frédérich et al. (2012). Juveniles were caught with a seine net on the beach zone (Fig. 1). Throughout the paper, we use the term “larvae” for individuals collected with crest nets and the term “juveniles” for fish maintained in aquaria. Treatment 1 consisted of providing commercially available granules, especially adapted for coral reef fish (SERA marin granules). These granules are composed of 52% proteins, 8.1% fat content, 3.5% crude fibres, 6.5% inorganic matter and 5.1% moisture. In treatment 2, the food sources were algae growing on stones/rubbles and found near beach shores. They were a mix of filamentous algae, red algae (Polysiphonia spp.) and green algae (Enteromorpha spp.) and represent the putative natural food of this species during the juvenile phase (Ogden and Lobel 1978, Frédérich et al. 2012).

The cohort of 64 larvae was equally separated into two aquaria of 200 L (100×50×40) filled with running sea water at a constant temperature of 26°C. Daylight simulation provided 12 h of light/darkness. In treatment 1, a total of 32 larvae were fed with granules three times per day. In treatment 2, the bottom of each aquarium was fully filled with stones or coral rubbles covered by turf algae, which were replaced daily. The experiments lasted 24 days, during which three to five individuals were randomly sampled every three days and then killed with an overdose of tricaine methanesulfonate (MS-222) according to ethical recommendations. All fish were measured and weighed. The SL was taken to the nearest millimeter with a Vernier caliper. Muscle tissue (0.5-2 cm³) from the epaxial section of the fish body and of the adductor mandibulae complex were sampled.

Samples of both types of muscle tissues and food sources (i.e. granules and algae) were dehydrated for 48 h at 60°C before being ground into a homogenous powder. After grinding, samples containing carbonates (brushed algae) were placed for 24 h under a glass bell with fuming HCl (37%) (Merck KGaA^©, for analysis quality) in order to eliminate calcareous material.

Samples were analysed on a V.G. Optima (Micromass) IR-MS coupled to an N-C-S elemental analyser (Carbo Erba). Routine measurements were precise to within 0.3% for both δ¹³C and δ¹⁵N values. Carbon and nitrogen isotope ratios were reported as per mil (‰) and expressed in the conventional delta notation (δ¹³C, δ¹⁵N) relative to the vPDB (Vienna Peedee Belemnite) standard and to atmospheric nitrogen standard, respectively. Certified materials were IAEA-N1 (δ¹⁵N=0.4±0.2‰) and IAEA CH-6 (sucrose) (δ¹³C=–10.4±0.2‰):

δX = [(Rsample/Rstandard) – 1] × 1000

where X is either ¹³C or ¹⁵N and R is the ratio of the heavy to the light isotope.

At settlement (t₀), significant differences in composition of carbon and nitrogen between the two types of muscles were checked using a Student t-test when variances were normally distributed and homogeneous (Shapiro-Wilk test and Levene test, respectively) and a Welch test when unequal variances were found even after transformation. For each food treatment lasting from t₃ to t₂₄, the relationship between δ¹³C, δ¹⁵N or elemental ratio of carbon to nitrogen (C/N) and time was investigated using two types of model: (1) an exponential model plateau followed by one phase decay and (2) a linear regression model. The equation of the exponential model plateau followed by one phase decay is:

δ_t = δ_f + (δ_i – δ_f)e^(–vt)

where δ_t is the stable isotopic composition of either adductor mandibulae or the epaxial musculature at the time when fish were collected from the aquaria, δ_i is the initial value before the diet switch, δ_f is the final isotopic composition equilibrated to the new diet, v is a measure of the turnover rate and t is the time that the fish were in the experiment (day) (Tieszen et al. 1983). This model allows for the calculation of the “Trophic Enrichment Factor” (TEF) which is the arithmetic difference in the nitrogen isotope ratio of consumer and diet (δ¹⁵N=δ¹⁵N_consumer−δ¹⁵N_diet).

The equation for the linear regression model is:

Y=β+αX

where β is the y-intercept and α the slope of the linear regression.

In order to perform the second model, a coefficient of correlation (r) was first calculated with Pearson’s product-moment method for parametric data and Spearman’s rank correlation coefficient was used for non-parametric data. If the results were significant (p<0.05), a linear regression model was performed and the coefficient of determination was calculated (R²). Normality of residuals of each model was tested using a Shapiro-Wilk test. If linear models were validated, analyses of covariance tests (ANCOVA) were performed to compare the dynamics of isotopic compositions between the two muscles. All statistical analyses were done with GraphPad Prism, version 5.0 (Motulsky 2007).

At settlement (t₀), δ¹³C values for the adductor mandibulae ranged from –20.73‰ to –19.55‰, δ¹⁵N values from 15.29‰ to 18.62‰ and carbon (C) to nitrogen (N) ratios were about 3.66 to 4.21 (Table 1). Compositions for epaxial musculature varied from –18.14‰ to –17.23‰ for carbon and from 17.14‰ to 18.96‰ for nitrogen, and the C/N ratio ranged from 3.22 to 3.28. Isotopic compositions in the adductor mandibulae differed significantly from values in the epaxial musculature (Welch test, δ¹³C, dF=23, p<0.05; Student t-test, δ¹⁵N, dF=22; p<0.05; C/N, dF=15, p<0.05). For the food sources, granules showed more negative values of δ¹³C and lower values of δ¹⁵N than algae (Table 2).

Acanthurus triostegus individuals fed with granules increased significantly in weight and size over time (linear regressions R²=0.64, p<0.05; R²=0.53, p<0.05, respectively; Fig. 2). Conversely, there was a significant negative correlation between weight and time for fish grazing on algae (linear regression R²=0.53, p<0.05; Fig. 2A). These fish did not increase in size during the experiment (t₆, 23.85 mm and 0.53 g; t₂₄, 22.50 mm and 0.31 g; Fig. 2B).

Results of linear regression models of both experiments are presented in Table 3. The turnover of δ¹³C in the adductor mandibulae and in the epaxial musculature showed the same increasing linear dynamics (ANCOVA, test for common slopes, F_1,47=0.0016, p>0.05; Fig. 3A). On the other hand, the δ¹⁵N values were decreasing for both types of muscles but their dynamics were different. Nitrogen values in the adductor mandibulae complex decreased linearly over time (mean value of 15.56‰ at the end of the experiment). The isotopic composition in epaxial musculature of each individual fitted an exponential model, whose values reached a plateau at 14.76‰ (Table 3, Fig. 3B). This exponential model allowed the calculation of the TEF, which is 6.56‰. The C/N ratio remained quite stable over time for each type of muscle. For both muscles, no significant correlation was found between the C/N ratio and time (Table 3, Fig. 3C).

When larvae were fed with turf algae, δ¹³C values of both muscles showed opposite dynamics during the experiment (ANCOVA, test for common slopes: F_1,48=31.45, p<0.05; Table 3). Values of δ¹³C increased linearly towards food source for the adductor mandibulae (i.e. reaching less negative values) while they decreased for the epaxial musculature (Fig. 3D) but values for each muscle tended towards –18.64‰ and –18.77‰, respectively. For both types of muscles, δ¹⁵N values did not change significantly over time (Table 3, Fig. 3E). The data did not fit exponential models (Table 3). The C/N ratio of the adductor mandibulae decreased linearly over time (r=0.41, p<0.05; Table 3, Fig. 3F). Conversely, no relationship was found between C/N ratio in epaxial musculature and time (Table 3).

No correlation was found between δ¹³C or δ¹⁵N and the SL of juvenile specimens from the field (p>0.05). Therefore, no model was estimated. However, the C/N ratio decreased linearly for both muscles (linear regression, adductor mandibulae Y=4.30–0.019X [R²=0.30; p<0.05] and epaxial musculature Y=3.28–0.004X [R²=0.39; p<0.05; Fig. 4]).

At settlement (t₀), the isotopic compositions differ significantly between muscles and this tissue-specific variation can be explained by physiological factors. Indeed, white muscles are less variable in δ¹³C and δ¹⁵N than red muscles (Pinnegar and Polunin 1999). The adductor mandibulae of fish have a mix of fast (i.e. white muscle) and slow fibres (i.e. red muscle) (Hernandez-Lagunas et al. 2005), while a thin superficial layer of red fibres covers the main white fibres making epaxial musculature (Rowlerson et al. 1985). Therefore, the amount of red fibres between the adductor mandibulae and the epaxial musculature in acanthurid larvae may be different and might significantly influence their isotopic and elemental compositions. A variation of the discrimination factors between muscle tissues can also explain this difference of isotopic compositions (Caut et al. 2009).

The C/N ratios are relatively high in A. triostegus larvae and juveniles (i.e. >3.2) (Table 1, Figs 3C, F and 4), suggesting the presence of carbon-rich compounds in muscle tissues (Sweeting et al. 2006). This probably comes from the lipid-rich planktonic diet of larvae, which allows them to store enough fat during the pelagic larval phase. In a natural population, this ratio decreased significantly when fish grew, meaning that lipid stores are used during the settlement phase (Fig. 4, McCormick and Molony 1992).

Many studies have revealed that the isotopic equilibrium can be reached after highly variable periods of time among fish species. Results of a feeding experiment on larvae of the Pacific blue fin tuna (Thunnus orientalis) showed that nitrogen values of fish approximated the isotopic composition of the food source after 20 days (Tanaka et al. 2014). In other experiments on larvae of the catfish Clarias gariepinus and Ictalurus punctatus, results of isotopic compositions reached the “plateau” phase between 4 to 10 weeks, respectively (Enyidi et al. 2013, Filbrun and Culver 2014). These studies were conducted on similar developmental stages (i.e. the larval stage) but tuna and catfishes did not undergo any drastic diet shift during ontogeny. Conversely, flatfishes observe a heavy metamorphosis and a diet shift from a pelagic food web towards a demersal habitat during growth. In the winter flounder (Pseudopleuronectes americanus) stable isotopes revealed that this diet shift occurred over a short period of 16 days (Bosley et al. 2002). In the present study, the isotopic equilibrium for A. triostegus was not reached after 24 days post-settlement although most marine ecologists agree that coral reef fishes observe a metamorphosis around this phase. Consequently, this study should probably be re-done over two or three months to define the turnover time.

According to varied functional demands between treatments, we expected to observe differences in isotopic dynamics between muscles. Indeed, we hypothesized a higher isotopic turnover for the adductor mandibulae complex than for the epaxial musculature, especially when fish are grazing on algae. Conversely, we expected the same turnover for the epaxial musculature in both treatments because the swimming activity did not vary.

When A. triostegus were fed on granules, the adductor mandibulae complex and the epaxial musculature showed the same linear pattern of variation for the isotopic compositions of carbon (Fig. 3A and Table 3). Both tissues displayed a relatively slow turnover, as δ¹³C data followed linear models. Surprisingly, linear regressions moved in an opposite direction to the value of the food source (Fig. 3A). As granules are composed of different elements (see Materials and Methods section), we suggest that the studied surgeonfish assimilated only some components of this food. The value of the “real” food (i.e. the components really incorporated) might have a different isotopic composition from the value for the whole granule and that would explain this unexpected isotopic dynamics.

For δ¹⁵N, the dynamics differs between the adductor mandibulae and the epaxial musculature (Fig. 3B) although the nitrogen isotopic compositions tend to decrease over time. Accordingly with growth and weight gain (Fig. 2), these isotopic dynamics are representative of a situation in which food contains high-energy ingredients, allowing synthesis of new muscle fibres. The equilibrium for the adductor mandibulae was not reached at the end of the experiment (24 days), contrary to epaxial musculature values, which fitted an exponential model, allowing a TEF of 6.56‰ to be calculated for nitrogen. This value is higher than the one found in another algivorous surgeonfish, Acanthurus sohal (TEF=4.69‰; Mill et al. 2007) but close to other TEFs recorded for tropical herbivorous fishes (Mill et al. 2007, Hata and Umazawa 2011). Currently, such a high TEF for herbivorous species is the subject of controversy among ecologists using stable isotopes. Various studies have found that discrimination factors depend on the species, the trophic position of the species, diet isotopic ratio and tissue, and a better quality food is apparently correlated with lower δ¹⁵N enrichment (Vanderklift and Ponsard 2003, Caut et al. 2009, Wyatt et al. 2010). However, Mill et al. (2007) demonstrated that high TEF was expected for herbivores, partly as a consequence of their very high ingestion rate. The observed TEF value of 6.56‰ might be partially determined by the selective assimilation of granule components (see above) or by a failure to properly model the isotope dynamics with the time-limited data.

Contrary to our hypothesis, muscle isotopic composition changed faster in the epaxial musculature than in the adductor mandibulae complex. This might be due to the fact that feeding on a more pelagic type of food (i.e. granules) does not require better-developed adductor mandibulae muscle than feeding on benthic algae (Frédérich et al. 2008). Perhaps granules or the “absence of biting on the benthos” do not induce the expected morphological and physiological changes (e.g. the development of large adductor mandibulae and robust oral jaws) required to efficiently graze on turf algae (Frédérich et al. 2008, 2012).

When algae were provided to A. triostegus juveniles, the δ¹³C turnovers were different between the two types of muscles. Values of δ¹³C in the adductor mandibulae increased towards the mean value of isotopic composition in algae (i.e. –15.54‰) (Table 2). For the epaxial musculature, δ¹³C values showed an opposite trend and shifted away from the food source value. This pattern suggests that epaxial musculature is probably used as a carbon source to favour the construction of a stronger jaw-closing system, increasing the efficiency to scrape algae. Moreover, the diet shift from “feeding on pelagic prey” to grazing could act as an environmental factor favouring or inducing morphological and physiological adaptations for algivory during ontogeny (Frédérich et al. 2012). The decreasing C/N ratio in the adductor mandibulae (Fig. 3F) sustains the hypothesis of a high level of activity and lipid loss in this muscle complex compared with the unchanged C/N ratio during treatment 1 (Fig. 3C; Schmidt et al. 2003, Logan et al. 2008).

During treatment 2, fish suffered from starvation, as suggested by their significant loss of weight (Fig. 2). Although we used the natural food source of A. triostegus (i.e. turf algae covering beach rocks or rubbles), the quantity of algae was dependent on the surface of the aquaria, inducing a limited amount of food source. In fact, starvation is a critical phase during which there is a continuous flow of molecules through different metabolic pathways, which are difficult to disentangle (Ferron and Leggett 1994). The breakdown of proteins and other molecules (i.e. catabolism) also occurs, leading to isotopic changes that are not related to diet change (Ferron and Leggett 1994). Therefore, fasting animals may differ in their stable isotope ratios from organisms fed ad libitum (Guelinckx et al. 2007). If diet did not contain enough carbohydrates, proteins can be catabolized in order to provide metabolic intermediates for the synthesis of other biologically important compounds (Wilson 1994). A study on the Atlantic salmon Salmo salar revealed that fasting individuals re-used their own body fat reserves by catabolism of stored lipid and protein reserves (Doucett et al. 1999). Here, the catabolism of endogenous fat stored from epaxial musculature is plausible but not so obvious in view of the constant C/N ratio during the treatment 2 (Fig. 3F). On the other hand, the decreasing δ¹³C values in the epaxial musculature might be due to the de novo synthesis (i.e. anabolism) of non-protein precursors such as carbohydrates for the development of other tissues such as the adductor mandibulae complex.

The C/N ratio dynamics were highly similar in “natural populations” and during algal treatment (Figs 3F and 4). This reinforces the assumption that A. triostegus may be subjected to catabolism or anabolism of non-protein precursors, inducing variation in isotopic compositions in different muscle types. This hypothesis is not strictly diet-specific but also physiologically due to internal metamorphosis during growth (Maruyama et al. 2001, Ankjærø et al. 2012). In this study, it is rather difficult to separate the effects of starvation and metamorphosis on the isotopic turnover dynamics in each muscle. Through our experimentation, we can argue that ecologists have to be careful about the significance of isotopic compositions in settling coral reef fish. A tissue-specific trade-off between sensitivity to changes in resource use (i.e. diet) and resistance to internal changes linked to metamorphosis probably occurs (Colborne and Robinson 2013).

This study is the first one exploring the dynamics of carbon and nitrogen turnover from two different muscles during the first days of post-settlement in a coral reef fish. As expected, we found some different patterns of variation in isotopic compositions over time between muscles. When artificial granules were provided, the turnover of both isotopes was faster in the epaxial musculature than in the adductor mandibulae. For algal treatment, the carbon turnovers showed opposite trends for each muscle but this difference was probably mitigated by starvation. Finally, a steady-state system in each muscle tissue of A. triostegus was not reached after a diet shift, at the end of the 24-day experiment. Overall, our results reveal that the metabolism of settling reef fish may be subjected to catabolism or anabolism of non-protein precursors, inducing variation in isotopic compositions.