Barramundi is an obligate carnivorous fish species that forms the basis of a significant aquaculture industry in south-east Asia and Australia( Reference Glencross 1 ). The development of high-nutrient density, formulated extruded feeds has been underpinned by the development of both a series of factorial bioenergetic nutritional models and foundation empirical studies( Reference Glencross 1 – Reference Glencross and Bermudes 5 ). These nutritional models have so far relied on the assumption that the dietary digestible energy (DE) source is irrelevant – that is, the dietary DE derived from protein, lipid and starch is utilised with equal efficiency, subject to key nutrients (e.g. protein) being provided at/or above minimum critical ratios to energy supply( Reference Glencross 4 – Reference Dumas, de Lange and France 10 ).
Each of the different macronutrients (starch, protein and lipid) supplies energy by distinct metabolic pathways. In aquatic animals, it is recognised that there are different levels of efficiency in the utilisation of each of these macronutrients for energy( Reference Saravanan, Schrama and Figueirdo-Silva 11 , Reference Glencross, Blyth and Bourne 12 ). It is now recognised that this difference requires an amendment of the digestible nutritional values of each macronutrient to those of metabolisable nutritional values and/or net energy (NE) nutritional values( Reference Hua, Birkett and De Lange 9 , Reference Glencross, Blyth and Bourne 12 – Reference Schrama, Saravanan and Geurden 14 ). A recent study by Schrama et al.( Reference Schrama, Saravanan and Geurden 14 ) examined the utilisation of both starch and lipid for growth by the omnivorous fish Nile tilapia (Oreochromis niloticus). These authors observed that each macronutrient had a different effect on the partial efficiencies of utilisation of DE (k DE ) by the fish, with dietary utilisation coefficients of 0·561 and 0·663 being observed for starch- and lipid-based diets, respectively. These observations clearly indicated that this fish species used lipid as an energy source for growth more efficiently. However, the third key macronutrient, protein, was not considered in this study. In the same study, Schrama et al.( Reference Schrama, Saravanan and Geurden 14 ) by reviewing the literature identified that there was a wide variability (0·31–0·82) in the k DE of different studies. It was suggested that the three primary reasons for this variability were as follows: different dietary macronutrient compositions, trophic level of the fish species and the composition of growth. In addition, there is increasing evidence that the roles of gluconeogenesis, glycolysis and β-oxidation play substantially different relative roles in energy provision in fish compared with other vertebrates( Reference Saravanan, Schrama and Figueirdo-Silva 11 , Reference Schrama, Saravanan and Geurden 14 – Reference Lansard, Panserat and Plagnes-Juan 17 ).
The objective of this study was to determine the partial efficiencies of utilisation of each of the different diets based on equivalent DE densities, but differing in ratio of each of the macronutrient energy substrates. By using a diet-by-ration factorial study, it was proposed that it would be possible to not only derive the partial efficiencies for each diet, but also by overlaying a multiple regression analysis of the responses to derive the discrete partial energetic efficiencies for each of the macronutrients. By determining these responses, it will help provide evidence for the true energetic role that each of the three macronutrients (protein, lipid and starch) plays as an energy source in diets when fed to barramundi.
Methods
Diet preparation
The diets used in this study were based on equivalent DE densities, but differed in the ratio of each of the macronutrient energy substrates. From this design, it will be possible to not only derive the partial efficiencies for each diet, but also by overlaying a multiple regression analysis of the responses to derive the discrete partial energetic efficiencies for each of the macronutrients used within each diet. The diets used in this study are based on diets used in the earlier study by Glencross et al.( Reference Glencross, Blyth and Bourne 12 ). In this experiment, each of the diets was formulated to be isoenergetic (15·3 MJ DE/kg) on a digestible nutrient basis based on the ingredient digestibility values determined by Glencross et al.( Reference Glencross, Blyth and Bourne 12 ). Most diets were also isoproteic (475 g/kg) on a digestible basis, with the exception of the ‘P’ diet, in which the digestible protein was 562 g/kg. An additional diet (C) was used as reference for diet specifications typically used in commercial diets.
Diets were prepared by mixing all the dry ingredients, and then processed by addition of oil and water (about 30 % of mash dry weight) while mixing to form a dough. The dough was then screw-pressed through a 4-mm diameter die using a pasta maker (Dolly; La Monferrina). The resultant moist pellets were oven-dried at 65°C for 12 h before being air-cooled, bagged and stored at −20°C. Formulations and composition of the diets are presented in Table 1.
Table 1 Formulation, composition and relative digestible contributions of the energy of each macronutrient in each of the experimental diets
C, control; P, protein; L, lipid; S, starch.
Fish handling
All animal procedures were approved by the CSIRO Animal Ethics Committee (approval no. A9/2011). Juvenile barramundi (Lates calcarifer) were obtained from a commercial hatchery (Betta Barra), and were on-grown to 69·6 (sd 0·75) g (n 480) in preparation for the experiment. During the on-growing period, all fish were fed the same diet (Marine Float; Ridley Aquafeeds) and maintained in 2×5000-litre seawater tanks. At trial initiation, forty fish were weighed on an electronic top-loading balance to 0·1 g accuracy to determine the mean and standard deviation of the population. Following this, twenty fish were allocated to 24×300-litre tanks on the basis of having to be within the mean (sd 1). The experiment was conducted at the Bribie Island Research Centre at Woorim, in a flow-through (3 litre/min), aerated, heated seawater tank array. Water temperature was maintained at mean 29·9 (sd 0·12)°C, and the dissolved oxygen level was 5·5 (sd 0·56) mg/l for the 42-d experiment.
Diets were manually supplied to the tanks. Three ration levels were used – a satiety level, 80 % of the initial satiety level and 60 % of the initial satiety level. The satiety rations were fed twice daily at 09.00–09.30 hours and 16.30–17.00 hours. Satiety levels were determined by feeding to slight excess all diets, and uneaten feed was accounted for. Correction factors were applied to determine solubilisation losses and pellet DM. The actual feed consumption within each tank was thus calculated on the basis of the methods reported by Helland et al.( Reference Helland, Grisdale-Helland and Nerland 18 ). The two restricted rations used in this study were based on 80 and 60 % of the measured initial demand, which was also consistent with the model of Glencross( Reference Glencross 4 ). These rations were not adjusted over time. Each treatment was duplicated within the 24-tank array. On the basis of the plan for using regression analysis in this experiment, it was proposed that a 3 rations×2 replicates design is stronger than a 2 rations×3 replicates approach.
Sample preparation and chemical analysis
Five fish were euthanised from the population at the beginning of the experiment as a representative initial sample. At the end of experiment, five fish from each tank were euthanised by immersion in an overdose of AQUI-S™ (AQUI-S) before being placed in an iced-seawater slurry. Following sample collection, each whole-fish sample was frozen before being minced using an industrial food processor to ensure sample homogeneity. Samples were then collected, and their moisture content was determined by oven drying at 105°C for 24 h, and a second sample was freeze-dried for chemical analysis. Freeze-dried fish samples were milled before analysis for DM, ash, fat, N and gross energy contents. Diet and faecal samples were analysed for DM, yttrium, N, lipid, starch and gross energy contents.
DM was calculated by gravimetric analysis following oven drying at 105°C for 24 h. Total yttrium concentrations were determined after mixed acid digestion using inductively coupled plasma-MS. Protein levels were calculated by determining total N using a CHNOS auto-analyser (Thermo Fisher Scientific), based on N×6·25. Total starch content of the diets was measured using an enzymatic method with the Total Starch Kit K-TSTA (Megazyme), following a modified Association of Official Analytical Chemists Method 996.11. Total lipid content of the diets was determined gravimetrically following extraction of the lipids using chloroform–methanol (2:1). Gross ash content was determined gravimetrically following loss of mass by combustion of a sample in a muffle furnace at 550°C for 12 h. Gross energy was determined by ballistic bomb calorimetry. All analyses were performed in accordance with the specifications of Association of Official Analytical Chemists( 19 ).
Diet digestibility analysis
At the end of the growth experiment and following sample collection, the remaining fish in the eight tanks fed to satiety were used for faeces collection. Fish were subjected to faecal stripping once daily about 6 h after feeding. Faecal stripping was carried out on the basis of the methods reported by Blyth et al.( Reference Blyth, Tabrett and Glencross 20 ). This involved netting of fish into a separate tank and rapid sedation to induce muscle relaxation. Once muscle relaxation occurred, fish were removed from the anaesthetic-containing water, stripped with gentle, manual abdominal pressure and faecal matter was expelled into a collection jar. Fish were then returned to their original tanks for recovery. Faeces samples were collected over a minimum of three stripping events, pooled within each tank and kept frozen for future analysis.
Differences in the ratios of DM, protein, lipid (insufficient faecal sample was available for starch analysis) or gross energy:yttrium in the feed and faeces for each treatment were calculated to determine the apparent digestibility (AD diet ) for each of the nutritional parameters examined in each diet based on the following formula:
$$AD_{{diet}} \,{\equals}\,\left( {1{\minus}\left( {{{Y_{{diet}} {\times}Parameter_{{faeces}} } \over {Y_{{faeces}} {\times}Parameter_{{diet}} }}} \right)} \right){\times}100,$$
where Y diet and Y faeces represent the yttrium content of the diet and faeces, respectively, and Parameter diet and Parameter faeces represent the nutritional parameters of concern (DM, protein or energy) in diet and faeces, respectively.
Protein and energy utilisation analysis
Protein (N×6·25) and energy (E) utilisation were determined on the basis of gain in both N and E over the experimental period, against the respective consumption of digestible N and E over the experimental period. Both gain and intake values were calculated on the basis of daily gain amount per unit body weight. To provide independence of size effects, modelling of protein, lipid and energy utilisation data was carried out with respect to known protein, lipid and energy body weight exponents for barramundi of x 0·7, x 0·9 and x 0·8, respectively( Reference Glencross and Bermudes 21 , Reference Salini, Poppi, Turchini and Glencross 22 ). Both protein energy and lipid energy utilisation were transformed to the energy body weight exponent value of x 0·8.
Nutrient and energy balance and deposition assessment
The net balances for protein (P), lipid (L) and energy (E) were calculated on the basis of the data derived from this study. The methods used for these calculations were based on those reported by Saravanan et al.( Reference Saravanan, Schrama and Figueirdo-Silva 11 ). Gross intake levels of each nutrient were determined on the basis of total feed intake for each tank multiplied by the percent composition of the feed being fed. Digestible intake levels were measured similarly on the basis of the digestibility of P, L and E from each diet. Faecal losses were determined as the reciprocal of the digestible levels. Retained nutrients and energy were determined on the basis of the net gain in nutrients and energy between fish at the end of the trial and those from the initial sample. Branchial and urinary N (BUN) were determined on the basis of the difference between digestible N intake and retained N with energy values based on 24·85 kJ×BUN using values reported by Saravanan et al.( Reference Saravanan, Schrama and Figueirdo-Silva 11 ). The metabolisable energy intake (MEI) was determined on the basis of the digestible energy intake (DEI) minus the branchial and urinary energy losses. Heat production (HP) was determined on the basis of the difference between metabolisable energy and retained energy (RE). Basal metabolism (HeE) was calculated on the basis of the reported fasting energy losses of 34·4 kJ/kg0·8 per d( Reference Glencross 4 ). The heat increment energy (HiE) was determined as follows – MEI minus RE and HeE. NE was determined as MEI minus HiE( Reference Bureau, Kaushik and Cho 23 ).
Statistical analysis
All values are means with their standard errors unless otherwise specified. Effects of diet treatment and ration levels were examined by multivariate ANOVA using the software package Statistica (StatSoft®). Levels of significance were determined using Fisher’s least significant difference test for planned comparisons, with critical limits being set at P<0·05. Regression figures presented were constructed using Microsoft Excel. Error terms for linear functions were determined using the regression feature of the Data Analysis package within Microsoft Excel. Multiple regression analysis was used to determine the component energy utilisation parameters based on having definitive assessments of protein energy utilisation efficiencies for each diet, which then enabled derivation, by multiple regression, of the contribution of both lipid energy and starch energy to the partial efficiency of energy utilisation in each diet( Reference Steel and Torrie 24 ).
Results
Effect of macronutrient energy bias on growth and body composition
There were significant differences between the effect of diets and feed ration levels on final weight, weight gain, feed intake and feed conversion ratio (FCR) (Table 2). Significant interaction effects on feed intake were observed between diet and ration levels, but none of the other performance parameters. There were no significant effects on survival attributable to diet, ration or the interaction term. Among fish fed to satiety, weight gain was the highest in fish fed Diet L and the lowest in fish fed Diet S. However, among fish fed to satiety, feed conversion was the highest in fish fed Diet P and the lowest in fish fed Diet S. Among fish fed to satiety, there were some significant differences in feed intake, with the highest intake in fish fed Diet S and the lowest in fish fed Diet P (Table 2).
Table 2 Growth and feed utilisation responses for each treatment
C, control; P, protein; L, lipid; S, starch; H, M and L, high, medium and low ration levels; D, R and D×R, P values for effects of diet, ration or the interaction, respectively; FCR, feed conversion ratio.
a,b,c,d,e,f Values within a row with unlike superscript letters were significantly different (P<0·05).
There was a significant effect of both feed ration level and diet on final live weight protein concentration, lipid concentration and energy content (Table 2). No significant differences were observed of diet on final live weight DM composition (Table 2). There were also significant interaction terms between diet and ration level on each of the parameters of final live weight DM, protein, lipid and energy concentrations. Key compositional differences were noted in fish fed Diet P, which had less lipid concentrations compared with fish fed Diet L. This effect was most notable at lower fixed ration levels (Table 2).
Effect of macronutrient energy bias on energy utilisation
The pairwise comparison within feed ration levels between each dietary treatment showed significantly different levels of energy retention between the starch diet and every other treatment (Table 3). The energy utilisation efficiencies (kJ/kg0·8 per d) for each diet were described by the following linear equations (Fig. 1):
$$y_{S} \,{\equals}\,0\! \cdot\! 508\left( \!{\pm0\! \cdot\! 010} \right)x-8\! \cdot\! 859\left( \!{\pm2\! \cdot\! 496} \right),\,R^{2} \,{\equals}\,0\! \cdot\! 998,$$
$$y_{L} \,{\equals}\,0\! \cdot\! 730\left( \!{\pm0\! \cdot\! 023} \right)x-29\! \cdot\! 821\left( \!{\pm5\! \cdot\! 461} \right),\,R^{2} \,{\equals}\,0\! \cdot\! 996,$$
$$y_{P} \,{\equals}\,0\! \cdot\! 715\left( \!{\pm0\! \cdot\! 012} \right)x-26\! \cdot\! 324\left( \!{\pm2\! \cdot\! 774} \right),\,R^{2} \,{\equals}\,0\! \cdot\! 999,$$
$$y_{C} \,{\equals}\,0\! \cdot\! 607\left( \!{\pm0\! \cdot\! 015} \right)x-8\! \cdot\! 686\left( \!{\pm3\! \cdot\! 717} \right),\,R^{2} \,{\equals}\,0\! \cdot\! 997.$$
Fig. 1 Energy gain (kJ/kg0·8 per d) by barramundi when fed different rations of each experimental diet. The regression equation of each diet is also shown. There was no significant difference in the linear regressions between the control, protein and lipid diet treatments. The regression equation of fish fed the starch diet was significantly different from each of the other treatments. 
, Control (y=0·6073x−8·6855, R
2=0·9974); 
, protein (y=0·7148x−26·324, R
2=0·9988); 
, lipid (y=0·7302x−29·821, R
2=0·9961); 
, starch (y=0·5078x−8·8594, R
2=0·9985).
Table 3 Protein (g/fish), lipid (g/fish) and energy (kJ/fish) balance analysis
C, control; P, protein; L, lipid; S, starch; H, M and L, high, medium and low ration levels; D, R and D×R, P values for effects of diet, ration or the interaction, respectively; FCR, feed conversion ratio; GPI, gross protein intake; FP, faecal protein; DPI, digestible protein intake; BUN(Peq), brachial and urinary N (protein equivalent); RP, retained protein; GLI, gross lipid intake; FL, faecal lipid; DLI, digestible lipid intake; RL, retained lipid; GEI, gross energy intake; FE, faecal energy; DEI, digestible energy intake; BUE, brachial and urinary energy; MEI, metabolisable energy intake; RE, retained energy; HP, heat production; HeE, basal metabolism; HiE, heat increment energy; NEI, net energy intake.
a,b,c,d,e,f Values within a row with unlike superscript letters were significantly different (P<0·05).
The coefficient of utilisation (k E ) was significantly lower for Diet S relative to each of the other diets. Similarly, the utilisation coefficient for Diet C was also significantly lower than that of Diets P and L. There was no difference in energy utilisation coefficients between Diets P and L. Maintenance energy demands (HEm) were calculated by extrapolating the linear regression to the intercept of the x-axis. From this, the following HEm values were derived – Diet S: 17·4 (sem 0·81) kJ/kg0·8 per d, Diet L: 40·8 (sem 0·98) kJ/kg0·8 per d, Diet P: 36·8 (sem 0·59) kJ/kg0·8 per d and Diet C: 14·3 (sem 1·14) kJ/kg0·8 per d. There were significant differences in the HEm values between Diets L and P relative to Diets S and C, but not within those pairings.
Effect of macronutrient energy bias on protein and lipid energy utilisation
The pairwise comparison within feed ration levels between each dietary treatment also showed significantly different levels of protein energy retention between starch diet and every other treatment (Table 3). The protein energy utilisation efficiencies (g/kg0·8 per d) for each diet were described by the following linear equations (Fig. 2):
$$y_{S} \,{\equals}\,0\! \cdot\! 412\left( \!{\pm0 \!\cdot\! 003} \right)x-1\! \cdot\! 302(\!\pm0\! \cdot\! 417\,R^{2} \,{\equals}\,0\! \cdot\! 994,$$
$$y_{L} \,{\equals}\,0\! \cdot\! 582\left( \!{\pm0\! \cdot\! 006} \right)x-8\! \cdot\! 094\left( {\!\pm0\! \cdot\! 572} \right),\,R^{2} \,{\equals}\,0\! \cdot\! 995,$$
$$y_{P} \,{\equals}\,0\! \cdot\! 556\left( \!{\pm0\! \cdot\! 005} \right)x-7\! \cdot\! 637\left( \!{\pm0\! \cdot\! 527} \right),\,R^{2} \,{\equals}\,0\! \cdot\! 996,$$
$$y_{C} \,{\equals}\,0\! \cdot\! 534\left( \!{\pm0\! \cdot\! 004} \right)x-0\! \cdot\! 088\left( \!{\pm0\! \cdot\! 588} \right),\,R^{2} \,{\equals}\,0\! \cdot\! 986.$$
Fig. 2 Protein energy gain (kJ/kg 0·8 per d) by barramundi when fed different rations of each experimental diet. The regression equation of each diet is also shown. There was no significant difference in the linear regressions between the control, protein and lipid diet treatments. The regression equation of fish fed the starch diet was significantly different from each of the other treatments. , Control (y=0·5335x−0·0885, R
2=0·9865); , protein (y=0·5565x−7·6372, R
2=0·996); , lipid (y=0·5818x−8·0943, R
2=0·9951); , starch (y=0·4122x−1·3025, R
2=0·9945).
The coefficient of utilisation was significantly lower for Diet S relative to each of the other diets. There was no difference in the protein energy utilisation coefficient (k PE ) between Diets P, L and C.
There were also different levels of lipid energy retention between the starch diet and every other treatment (Table 3). This resulted in the coefficient of utilisation being significantly higher for Diet S relative to each of the other diets. However, Diet P also had a significantly higher level of lipid energy utilisation relative to the lipid and control diets. The lipid energy utilisation efficiencies (kJ/kg0·8 per d) for each diet were described by the following linear equations (Fig. 3):
$$\!\!\!\!\!\!\kern-3pty_{S} \,{\equals}\,1\! \cdot\! 5478\left( \!{\pm0\! \cdot\! 015} \right)x-7\! \cdot\! 332\left( {\pm0\! \cdot\! 500} \right),\,R^{2} \,{\equals}\,0\! \cdot\! 991,$$
$$\kern-3pty_{L} \,{\equals}\,1\! \cdot\! 070\left( \!{\pm0\! \cdot\! 002} \right)x-19\! \cdot\! 619\left( \!{\pm1\! \cdot\! 469} \right),\,R^{2} \,{\equals}\,0\! \cdot\! 998,$$
$$\kern-3pty_{P} \,{\equals}\,1\! \cdot\! 387\left( \!{\pm0\! \cdot\! 006} \right)x-17\! \cdot\! 558\left( \!{\pm0\! \cdot\! 456} \right),\,R^{2} \,{\equals}\,0\! \cdot\! 994,$$
$$\kern-3pty_{C} \,{\equals}\,1\! \cdot\! 081\left( \!{\pm0\! \cdot\! 002} \right)x-8\! \cdot\! 375\left( \!{\pm0\! \cdot\! 183} \right),\,R^{2} \,{\equals}\,0\! \cdot\! 999.$$
Fig. 3 Lipid energy gain (kJ/kg0·8 per d) by barramundi when fed different rations of each experimental diet. The regression equation of each diet is also shown. There were no significant differences in the linear regressions among each of the control, protein, lipid and starch diet treatments. , Control (y=1·0812x−8·375, R
2=0·9987); , protein (y=1·3873x−17·558, R
2=0·9942); , lipid (y=1·0699x−19·619, R
2=0·9983); , starch (y=1·5478x−7·3317, R
2=0·9915).
When the lipid energy utilisation coefficients (k LE ) were examined relative to the dietary concentration of lipid, a strong but non-significant (P=0·127) linear relationship was observed (Fig. 4).
Fig. 4 Lipid energy utilisation coefficients relative to the dietary concentration of lipid (y=−0·0061x+1·9036, R 2=0·7618). Values are means with their standard errors.
Determination of macronutrient component contributions to energy utilisation
The different combinations of protein, lipid and starch among the diets in the present study allow for the analysis of the component contributions of each macronutrient to energy retention (Table 4). This assumes that each macronutrient is contributing a part of the dietary energy proportional to its content in the diet, its energetic value and a component utilisation value.
Table 4 Component energetic contributions from each macronutrient in each diet and the calculated and measured energetic parametersFootnote *
* Digestible energy value is derived from assumed energetic value of the digestible nutrient concentration in each diet. The calculated energy value of each diet is the sum of the component macronutrient digestible energy values. The measured energy value is the digestible energy measured from in vivo studies. Protein utilisation coefficients are derived from Equations (5)–(8). Lipid utilisation for diets protein and lipid, where starch was absent, are derived from Equations (10) and (11). Component lipid utilisation coefficients for each of the diets were derived from multiple regression of energy utilisation Equations (1) and (4). Similarly, component starch utilisation coefficients were derived by multiple regression of energy utilisation Equations (1) and (4).
On the basis of previously mentioned assumptions, each of the component energy utilisation values was derived using multiple regression analysis. For each of the diets, protein contribution can be defined by converting protein utilisation to protein energy utilisation, and thus defining the component protein energy utilisation (Fig. 2). Therefore, because we have a definitive assessment of protein energy utilisation efficiencies (see Equations (5)–(8)), we can also derive by multiple regression the remaining unknown variables, which constitute the contribution of both lipid energy and starch energy to the partial efficiency of energy utilisation in each diet (Tables 1 and 3). Although we have an assessment of the partial efficiency of lipid energy utilisation (Fig. 3), the fact that lipid energy gain in this representation also includes lipid deposited from non-lipid origins (i.e. starch and/or protein energy), it was necessary to derive the component lipid energy utilisation using multiple regression methods.
Effect of macronutrient energy bias on protein, lipid and energy budgets
There was a range of significant effects attributable to diet, feed ration level and the interaction term on the protein, lipid and energy budget parameters (Table 3). Exceptions to this were for the digestible protein intake (DPI), for which there were no significant interactions between diet and ration level. Gross protein intake (GPI) was highest in fish fed Diet P at ration level H with the corresponding lowest GPI at the same ration level for Diet L (Table 3). Faecal protein (FP) was also highest in fish fed Diet P, and this was consistent across each of the ration levels. The lowest FP was observed in fish fed diet L, again across each of the ration levels. DPI was highest in fish fed Diet P at ration level H, and although these differences were significant they were much smaller than those seen on GPI. Protein losses through BUN protein equivalent were highest among fish fed Diet S at ration level H, although the differences at lower ration levels were less obvious. Retained protein (RP) at the highest ration levels was similar in C, P and L diet groups, but significantly poorer in Diet S. The ratio of RP:DPI was highest among fish fed Diet C at ration level M. At ration level H, there was no significant difference among the RP:DPI for Diets C, P and L but for Diet S it was significantly lower (Table 3).
Gross lipid intake (GLI) was highest for fish fed Diet L at ration level H with the corresponding lowest GLI at the same ration level for Diet S (Table 3). Faecal lipid (FL) content was the highest among fish fed Diet P, and this was consistent across the ration levels. The lowest FL, across the ration levels, was found in fish fed both Diets C and S. Digestible lipid intake (DLI) was the highest among fish fed Diet L at ration level H, and for the other ration levels DLI was significantly higher compared with Diet L. Retained lipid (RL) at the highest ration levels was similar among fish fed diets C, P and S, but significantly higher among fish fed Diet L. The ratio of RL:DLI was the highest among fish fed Diet S, and this was consistent across ration levels. The ratio of RL:DLI was lowest among fish fed Diet L, and this too was consistent across ration levels. The ratio between RL:RP for Diets L and S were similar and significantly higher compared with fish fed Diets C and P. In most cases, this declined with declining ration, although no such effect was observed with Diet C (Table 3).
Gross energy intake (GEI) was highest among fish fed Diet S at ration level H with the corresponding lowest GEI at the same ration level being from Diet P (Table 3). Among the lower ration levels, there were no significant differences in GEI. These differences were also reflected in the DEI across treatments. Faecal energy was highest among fish fed both Diet C and S and lowest among fish fed Diet P. BUE losses were highest among fish fed Diet S at ration level H and M, although at the lowest ration level BUE was highest among fish fed Diet P. The highest MEI at ration level H was seen in fish fed Diet S, but at the two lower ration levels it was higher from Diet C. Lowest MEI values were from Diet P and the highest ration level (H), but at the two lower ration levels MEI intake was lower from Diet S. RE was highest among fish fed Diet L at ration level H, and lowest among fish fed Diet S at the lowest ration level, although the RE of fish fed Diet S was the lowest within each of the ration levels. HP was highest, and substantially so, in fish fed Diet S at ration level H, although differences at lower ration levels were less obvious. HeE had significant effects attributable to both diet and ration, but not interaction. HiE was highest among fish fed Diet S at ration level H, which was more than twice that of fish fed the same ration from Diet P. This effect was reversed at the lower ration levels with higher HiE values observed from Diet S at the two lowest ration levels. Net energy intake (NEI) was highest among fish fed Diet L and lowest among fish fed Diet S. Ration also had a clear effect on NEI, although differences between fish fed Diets C, P and L at each of the ration levels were nominal. The NEI values of fish fed Diet S were significantly lower at each ration level. The ratio of RE:DEI typically declined with declining ration. The RE:DEI values were similar between Diets P and L at similar ration levels, but significantly poorer when fed Diet S at each ration level except the lowest one. Diet C was a little different to the other diets and showed a largely consistent RE:DEI across the ration levels and at a high level (>50 %) (Table 3).
Discussion
The present study sought to define the relative contributions of each of the three macronutrients (protein, lipid and starch) in supplying DE in diets fed to juvenile barramundi. This has enabled an insight into the roles that these macronutrients play in contributing to energy provision in this species. Understanding this relationship is critical to fish nutrition because of the strong intrinsic link between fish growth, energy demand and diet energy density.
Effect of macronutrient energy bias on growth, feed utilisation and body composition
By using diets with equivalent levels of DE but with differences in proportions of protein, lipid or starch providing that energy, clear effects were seen in this experiment. For each of these treatments, the strategy of feeding each diet at specific ration levels has allowed us to build substantially on earlier findings from using these same diets, which were previously fed over a longer term( Reference Glencross, Blyth and Bourne 12 ). Therefore, in the present study, we focus our discussion on the effects within ration levels to allow us to examine the diet-specific effects. At the highest ration level, the responses of growth were generally consistent with the earlier study( Reference Glencross, Blyth and Bourne 12 ). In the earlier study, the best growth was seen with Diet P, whereas in the present study the best growth was seen with Diet L. However, in both studies, the poorest growth was seen with Diet S. At lower ration levels (M and L), growth was not consistent with the pattern seen at the H ration level. At lower ration levels, the best growth was seen among fish fed Diet P, followed by Diet L, and fish fed Diet S still performed the poorest. These results are directly comparable with those from our earlier study, and suggest that at the highest ration level, fed to apparent satiety, feed intake variability may have altered the responses. In another similar study by Saravanan et al.( Reference Saravanan, Schrama and Figueirdo-Silva 11 ) with rainbow trout fed either high- or low-protein diets with energy biased towards either starch or lipid, fish down-regulated their feed intake when fed starch-based diets. This observation was a direct contrast to the present study where barramundi increased their satietal intakes with starch-based diets. Differing again were the observations of Schrama et al.( Reference Schrama, Saravanan and Geurden 14 ), who observed in the omnivorous species tilapia that growth was not compromised with the use of starch as an energy source relative to that seen when lipid was used instead. We suggest that these differences are directly linked to the ability of tilapia to digest and utilise glucose from starch, whereas starch digestion by barramundi is comparatively poorer, and its ability to regulate blood glucose is questionable( Reference Stone 25 – Reference Wade, Skiba-Cassy and Dias 27 ). Clearly, there appears to be different nutritional capacity among different fish species to utilise starch as an energy source.
The responses of feed efficiency (FCR) to ration within each diet are consistent with observations of most studies on restricting nutrient/energy supply to fish, and the present findings are consistent in this regard with other findings from this species( Reference Glencross 4 , Reference Bermudes, Glencross and Austen 28 ). An advantage of using this pair-feeding regimen is that it allows for a very clear examination of the effect of diet composition on performance criteria independent of feed intake variability. However, we do acknowledge that this does potentially cause complications in the application of digestibility values across variable feed intake levels. Some of the clearest implications from the variation in energy supply by different macronutrients can be seen by the cross-diet comparison of FCR at each of the two lower ration levels in the present study.
Effects of each of the diets on fish body composition were noted primarily in terms of the whole-body lipid, DM and protein concentrations. One of the most notable compositional effects at the highest ration level (H) was the difference in lipid concentrations of fish fed Diet L relative to the other treatments, and that Diet P had the lowest lipid concentrations. These observations from the present study contrast those from an earlier study using these same diets, in that the lipid concentrations in fish fed Diet S are considerably lower and those of Diet L are higher( Reference Glencross, Blyth and Bourne 12 ). At lower ration levels in the present study this effect of the diets with considerable starch content (Diet C and S) on lipid concentration in the body is more consistent with our earlier study. Reasons for this discrepancy at the satiety (H) ration level are unclear. These present results (from the H ration) are, however, consistent with those of Schrama et al.( Reference Schrama, Saravanan and Geurden 14 ), who also noted higher levels of lipid in the whole body of fish (Tilapia) fed diets high in lipid, but less so in fish fed diets high in starch.
Effects of macronutrient bias on energy utilisation
The efficiency of energy utilisation (i.e. the ratio of gross energy gain as a function of DEI over a range of intake levels, expressed as k E ) differed among each of the treatments. In this study, the relationship between energy intake and gain was observed to be linear, with a calculated energy utilisation constant value that varied between k E =0·507 and 0·730, subject to diet. For Diet C (the most analogous to a commercial diet), k E =0·607, which is generally consistent with other k E values that have been determined for this species( Reference Glencross 4 , Reference Glencross and Bermudes 21 ). In an earlier study( Reference Glencross 4 ), a range in the values of k E of 0·61–0·76, with an average of 0·68, has been determined and shown to be marginally affected by fish size. In a subsequent study, the k E values have also shown to be influenced by temperature, with k E values ranging from 0·42 to 0·59 and being lower outside optimal thermal regimens( Reference Glencross and Bermudes 29 ).
In the present study, a range of k E values was observed and clearly related to the variation in macronutrients used to supply equivalent levels of DE in each of the diets. Those diets higher in starch had poorer k E values, with Diet C (135 g/kg starch) k E =0·607 and Diet S (225 g/kg starch) k E =0·507, compared with Diet P (17 g/kg starch) k E =0·715 and Diet L (29 g/kg starch) k E =0·730. A clear negative relationship between the k E values and diet starch concentration is seen (Fig. 5). Our findings in the present study are similar to those reported by Schrama et al.( Reference Schrama, Saravanan and Geurden 14 ), who also reported a range in k E values when diets were biased to either starch (k E =0·561) or lipid (k E =0·663). A key difference between these studies was that in the present one we can isolate this effect from differences in DE concentration of the diets, and clearly ascribe the effects solely to macronutrient supply differences. Some significant differences in HEm were observed among the different diets. For those diets largely devoid of starch the HEm was estimated to be 36·8–40·8 kJ/kg0·8 per d, whereas those diets with starch had HEm values estimated at 14·3–17·4 kJ/kg0·8 per d. However, an important constraint is that these are estimated values derived from extension of the linear regression functions to their intercept of the x-axis, and given that there were no ration levels below the HEm values these estimations are beyond the bounds of the data. As such we suggest that these differences may be an artefact of the extrapolation of the data set.
Fig. 5 Relationship between diet starch concentration and energy utilisation coefficient (k E ) values (y=−0·001x+0·747, R 2=0·987).
Effects of macronutrient bias on protein and lipid utilisation
The protein utilisation efficiency was determined as the amount of dietary digestible protein (g/kg0·7 per d) required to deposit a gram of protein in the body of the animal. In the present study values (k P ) determined in the present study ranged from k P =0·412 to 0·580 (data not shown). This compares well with values (k P =0·49–0·54) determined by Glencross( Reference Glencross 4 ) and Glencross & Bermudes( Reference Glencross and Bermudes 29 ) for barramundi of different sizes and at different temperatures. The values also compare well with other carnivorous marine species like the European seabass (Dicentrarchus labrax) for which a value of k P =0·52 was reported( Reference Lupatsch, Kissil and Sklan 30 ).
In the present study, a focus was made on the energy retention as protein energy retention. This was estimated based on its energy equivalent, in this case 23·6 kJ/g protein, and expressed relative to the metabolic body weight (W0·8) of the animal rather than its protein body weight (W0·7)( Reference Lupatsch, Kissil and Sklan 8 ). The calculated energy cost as DE (kJ) for deposition of protein from each diet varied and was shown to be significantly higher with the inclusion of starch in the diet. The energy cost values (1/k PE ) determined in the present study for protein deposition ranged from 1·72 to 2·43 kJ per kJ of protein energy deposited, with higher cost values of 1·87–2·43 being from diets higher in starch. This further supports that protein synthesis in the presence of higher dietary starch levels is more energetically expensive. This compares to other marine fish species (e.g. Sparus aurata, Dicentrarchus labrax and Epinephelus aeneus), and the freshwater carp (Cyprius carpio) which had 1/k PE values ranging from 1.78 to 1.90( Reference Lupatsch, Kissil and Sklan 8 , Reference Schwarz and Kirchgessner 31 ).
Lipid utilisation efficiency (data not shown) was determined as the amount of digestible dietary lipid (g/kg0·9 per d) required to deposit a gram of lipid in the body of the animal( Reference Glencross and Bermudes 21 ). In the present study, lipid utilisation efficiency values (k L ) ranged from k L =1·07 to 1·55. The utilisation of dietary lipid energy for lipid energy deposition to determine the partial efficiencies of k LE was also examined. What appeared unusual about these values is that they were all greater than 1. This implied that there was greater lipid energy deposition than lipid energy intake, resulting in a NE gain from this macronutrient and clearly indicating synthetic activity. Although a similar scenario for protein would be impossible, for lipid it demonstrates that there is lipid being synthesised from other macronutrient substrates (e.g. starch or protein). From diets low in lipid, it can be noted that the relative contribution to lipid synthesis from these other macronutrients is enhanced.
The energy cost (1/k LE ) for lipid gain in the present study ranged from 0·65 to 0·93 kJ per kJ of lipid deposited. This was similar to the range of values (0·83–0·86) reported by Glencross et al.( Reference Glencross, Hawkins and Evans 32 ) with rainbow trout (Oncorhynchus mykiss), but was substantially lower than that the 1·10, 1·11 and 1·31 reported by Lupatsch et al.( Reference Lupatsch, Kissil and Sklan 8 ) for three marine species (Sparus aurata, Dicentrarchus labrax and Epinephelus aeneus). In carp, the efficiency was estimated at 1·39( Reference Schwarz and Kirchgessner 31 ), demonstrating that lipid accumulation from lipid energy intake was a highly efficient process in barramundi, similar to other carnivorous species( Reference Glencross, Hawkins and Evans 32 ). That the energy cost of lipid gain is below one also demonstrates that this is an energetically efficient process in terms of energy storage. In contrast with the values of energy cost for protein deposition, which showed that the energetic cost for protein deposition was almost twice that of the energetic value for synthesis, lipid is so much more useful in terms of its storage mechanisms, because it uses less energy for storage than its own energetic value. One observation of note was the differences in the 1/k LE values, with Diet S having the lowest value of 1/k LE =0·65, showing that lipid storage from starch is very efficient.
Effects of macronutrient bias on component energy utilisation
Energy retention in fish consists almost exclusively of protein or lipid deposition; therefore, the efficiency of energy gain in terms of protein and lipid gain can be considered separately using multiple regression analysis as described first by Kielanowski( Reference Kielanowski 33 ). The comparison of the four diets in this study showed that the inclusion of starch in the diet had a significant effect on the gain of either protein or lipid relative to DEI, and a clear reduction in protein synthesis with the inclusion of this macronutrient in the diets.
When examining the components of energy utilisation, we have worked on the premise that it is the sum of the digestible value of protein, lipid and starch, their relative energetic proportions (%) in the diet and a discrete component utilisation (θk PE , θk LE or θk SE ) of each macronutrient that combines to provide the overall k E value for any particular diet (Table 4). Using this premise, we observed that the component protein energy utilisation value (θk PE ) was significantly impaired with the higher inclusion levels of dietary starch (Diet S θk PE =0·412 cf. Diet L θk PE =0·582). In diets with lower levels of digestible starch (e.g. Diet C θk PE =0·534; 111 g/kg), although a numerically lower θk PE was observed, it was not significantly reduced relative to diets with nominal levels of starch (e.g. Diet P θk PE =0·557).
The component lipid energy utilisation value (θk LE ) was highly variable compared with the other component energy utilisation values (θk PE or θk SE ) for the other macronutrients, with θk LE values ranging from 0·821 to 1·345 (Table 4). These determined values appear to reflect both the inclusion of dietary starch (e.g. Diet S θk LE =0·821 cf. Diet P θk LE =1·345), and influences of dietary lipid level on the component lipid energy utilisation (e.g. Diet P θk LE =1·345 cf. Diet L θk LE =1·036). We suspect that the variability in this component utilisation value reflects the responsive nature of the metabolism of lipids by this species in response to variable nutrient supply. In effect, what we are observing is an enhanced capacity of the animal to produce lipid from protein energy sources. Although it is less efficient than that from lipid or protein, there is still substantial lipid synthesis from starch energy occurring.
The component starch energy utilisation values (θk SE ) determined by using the multiple regression approach were determined to be the same across all diets (θk SE =0·438). Energy deposition from starch was clearly the least efficient of all the macronutrients (although a poorer θk PE was noted for Diet S). We suggest that barramundi has limited metabolic capacity to utilise starch-derived energy. Although it can produce lipids from glucose precursors, it clearly does so at a less-efficient rate than that seen from either protein or lipid directly.
Conclusions
The results of this study show that barramundi have clear metabolic inefficiencies associated with the inclusion of starch in their diet. With the increasing inclusion of starch in the diet of this species there was a reduction in the efficiency of protein (protein energy) utilisation and this contributed to an overall decline in the efficiency of energy utilisation. In the absence of starch, protein utilisation was constant and it was unaffected by its concentration in the diet. Collectively, the findings of this study support the notion that the concentration and type of macronutrient mix in a diet for barramundi has a significant effect on the ability of the fish to use those nutrients for energy. This finding suggests the existence of a metabolic mechanism that influences the ability of fish to utilise discrete nutrients for energy, independent of total energy intake.
Acknowledgements
The authors acknowledge the technical support of Nick Polymeris and Mike Anderson.
This study was supported by the Australian Centre for International Agricultural Research, Project FIS-2006-141.
B. D. G., D. B., S. I. and N. M. W. all contributed to the experimental design. B. D. G., D. B., S. C., N. B. and S. I. all contributed to the conduct of the experiment. D. B. manufactured the diets, N. B. and S. C. undertook most of the sample analysis. B. D. G. undertook the analysis of the data. B. D. G., S. I. and N. M. W. all contributed to the interpretation of the data and writing of the manuscript.
None of the authors has any conflicts of interest to declare.
