Tambaqui (Colossoma macropomum) is one of the most widely produced freshwater species in South America (Araújo-Lima and Gomes, 2005) and the most produced native species in continental aquaculture, with great growth in 2016 with 136.99 thousand tons (IBGE, 2016). Interest in the species has risen due to its adaptability to intensive production systems and artificial feeding, fast growth, omnivorous feeding habit, high feed efficiency, and excellent taste and desirable texture (Araújo-Lima and Gomes, 2005).

The use of balanced and highly digestible diets is crucial for sustainable fish production. Digestibility coefficients provide an indication of the amount of the nutrient that is absorbed; the higher the digestibility of the nutrient, the better it will be utilized by fish, resulting in higher production performance and reducing excretion of nutrients in the production environment (Oliveira Filho and Fracalossi, 2006).

A previous study (Buzollo et al., 2018) highlighted the importance of studies on nutrient digestibility of conventional feed ingredients used by the aquafeed industry to maximize production and reduce operating costs and levels of nitrogen, phosphorus, and organic matter released into effluents of fisheries. However, no study has evaluated the digestibility of amino acids in ingredients used in commercial tambaqui diets. Thus, limited information is available for the formulation of balanced diets for commercial tambaqui production.

We aimed to determine the apparent digestibility coefficients of amino acids of 13 ingredients, which were divided into energetic (corn, wheat bran, broken rice, and sorghum) and proteic (corn gluten meal, soybean meal, poultry byproduct meal, salmon meal, fish meal [tilapia processing residue], wheat gluten meal, feather meal, cottonseed meal, and alcohol yeast [spray dried]) ingredients. The chemical score of each amino acid was calculated, and the amino acid profile of each ingredient compared to that of tambaqui white muscle to determine the limiting digestible amino acids of each ingredient.

The experimental trial was conducted in Jaboticabal, SP, Brazil (21°15'07.5" S, 48°19'46.0" W), in accordance with the ethical principles for animal experimentation adopted by the Brazilian College of Animal Experimentation (COBEA) and was approved by the Ethics Committee on Animal Use (case no. 016114/11).

In total, 420 tambaqui juveniles (mean initial weight: 70±8.58 g) were used in the study. The animals were kept in 28 tanks (430 L) provided with continuous aeration and water from a flowing artesian well (renewal rate: ∼10 times per day). The physicochemical parameters of the water were within the acceptable range for the species (Aride et al., 2004; Araújo Lima and Gomes, 2005): mean±SD, pH: 7.85±0.17; temperature: 29.72±0.34 °C; dissolved oxygen: 5.71±0.34 mg/L; electrical conductivity: 150.75±17.62 µS/cm; alkalinity: 88.67±0.82 µg/L; ammonia: 189.17±59.29 µg/L; nitrate: 419.96±100.28 µg/L; nitrite: 28.68±39.09 µg/L; and total phosphorus: 200.89±61.00 µg/L.

To determine the ADC of each ingredient, a reference diet was prepared to contain 237 g/kg of crude protein and 16.32 MJ/kg of gross energy (Table 1). The 13 test ingredients used in the experimental diets were obtained from four Brazilian industries: Guabi^® (sorghum, corn gluten meal, poultry byproduct meal, wheat gluten meal, feather meal, cottonseed meal, and alcohol yeast [spray dried]), Coplana^® (corn, wheat bran and soybean meal), Agromix^® (broken rice), and Grupo Ambar Amaral^® (fish meal [tilapia processing residue]), with exception of salmon meal that was imported from Chile, and were divided into two groups: energetic = corn, wheat bran, broken rice, and sorghum; and proteic = corn gluten meal, soybean meal, poultry byproduct meal, salmon meal, fish meal (tilapia processing residue), wheat gluten meal, feather meal, cottonseed meal, and alcohol yeast (spray dried). With these ingredients (Table 2), 13 test diets were formulated to contain 695 g/kg of the reference diet, 300 g/kg of the test ingredient (100 g/kg for wheat gluten meal due to the cohesive and viscoelastic properties of gluten that may provide result in a rubbery, dry pellet (Day at al., 2006), and 5 g/kg of chromium-III oxide (Cr₂O₃) used as the inert digestibility marker. For the preparation of diets, the ingredients were ground, manually mixed, moistened, and extruded using an Exteec extruder (Ex Micro model). Pellets were dried in an oven with forced-air ventilation at 55 °C for 24 h.

The digestibility coefficients of amino acids from the test ingredients were determined with the use of an inert marker (chromium-III oxide), according to Nose (1966). For fecal collection, 14 glass fiber collectors (80-L each) provided with continuous aeration and water circulation were constructed according to the modified Guelph system described by Abimorad and Carneiro (2004). Fecal collection from the four replicates of the 14 treatments (13 test diets and a reference diet) was divided into two periods. First period – distribution of replicates 1 and 2 in 28 feeding tanks. The adaptation to the diets was carried out for seven days. On day 8, feces were collected from replicate 1 (first 14 feeding tanks), and on day 9, from replicate 2 (another 14 feeding tanks). Second period – redistribution of the diets of replicates 3 and 4 in 28 feed tanks. The adaptation to the diets was carried out for seven days. On day 8, feces were collected from replicate 3 (first 14 feeding tanks), and on day 9, from replicate 4 (another 14 feeding tanks). In both periods, the fish from each replicate were fed to apparent satiation and transferred after the last feeding of the day (18.00 h) to the conical tanks and, therefore, the collections were performed during the night. Feces were collected into Falcon conical tubes (kept on ice to reduce feces degradation), every 3 h, for ease of animal management, according to previous project, until 6.00 h of the next day. All feces collected from each replicate were lyophilized using a Thermo Electron Corporation Fisher^®, freeze-dried, and analyzed.

Chromium-III oxide concentrations in diets and feces were determined by nitric-perchloric digestion according to the method described by Furukawa and Tsukahara (1966). The amino acids were analyzed using liquid chromatography in cationic exchange resin columns and post-column derivation with ninhydrin and an autoanalyzer. For the amino acid count, the samples were hydrolyzed with HCl 6 N for 22 h at 110 °C according to the method described by Moore and Stein (1963). Tryptophan was determined after the enzymatic hydrolysis with Pronase at 40 °C for 24 h, followed by a colorimetric reaction with 4-(dimethylamino)benzaldehyde in sulfuric acid 21.2 N and read at 590 nm. The tryptophan content was calculated according to Spies (1967).

The apparent digestibility coefficient (ADC) for a test ingredients was calculated from the amount of marker and amino acid in the reference diet, test diet, and feces according to the equation of Nose (1966):

The ADC for an amino acid in a test ingredient was calculated according to the following equation of Forster (1999):

in which a = AA contribution of the reference diet to the AA content of the test diet (% AA in reference diet × 0.695), and b = AA contribution of test ingredient to AA content of test diet (% AA in test ingredient).

Amino acid limitations in test ingredients were estimated by calculating the chemical score index (CSI) for each amino acid according to the following equation of Sgarbieri (1987):

The essential amino acid index (EAAI) of the test ingredients were calculated according to the following equation of Oser (1959):

in which a, b, c…j are the % digestible EAA of test ingredient protein; ap, bp, cp…jp are the % EAA in tambaqui muscle; and n = number of amino acids considered.

The two methods compare the amount of digestible AA in the ingredients with the amino acid profile of fish white muscle (Hepher, 1988). For these calculations, nine fish (mean weight: 42.0±5.76 g) from the same population were killed by ice-slurry immersion and, white muscle samples were taken for amino acid analysis.

The essential amino acid (EAA) with the lowest chemical score index was considered the first limiting amino acid of the ingredient. The EAAI was calculated from the geometric mean of all EAA scores. Protein quality is high in ingredients with higher EAAI values.

Proteic and energetic ingredients were analyzed separately. The trial with energetic and proteic ingredients were arranged in a randomized block design, with four replicates; energetic ingredients with four treatments (ingredients) and four replicates, whereas proteic ingredients with nine treatments (ingredients) and four replicates. Each block was considered as one round of fecal collection. The ADC values were subjected to ANOVA using the PROC GLM procedure of SAS (Statistical Analysis System, version 9.2). When significant differences were detected, treatment means were compared using Tukey's test at 5% significance level.

No fish mortality was observed during the experimental period. No effect of the fecal collection period (round) was observed; therefore, the ADC data of all test diets were analyzed together. Of the 13 proteic and energetic ingredients tested, only six had digestibility coefficients <70% for some amino acids, and significant differences in the ADC of amino acids were observed across ingredients (P<0.05).

High ADC values (>70%) were observed for most amino acids (Table 3), and only the ADC of arginine, phenylalanine, threonine, serine, and tyrosine for sorghum and threonine, serine, and tyrosine for broken rice were <70%. Additionally, corn (95%) and wheat bran (92%) had the highest ADC values, whereas broken rice (78.7%) and sorghum (74.6%) had the lowest ADC values.

Mean ADC_EAA and ADC_NEAA values of most proteic ingredients were high (>70%; Table 4). Only wheat gluten meal, feather meal, cottonseed meal, and alcohol yeast had ADC <70% for a few amino acids. Amino acid digestibility varied across proteic ingredients, but some EAA had low ADC common to a few ingredients: arginine (wheat gluten meal, feather meal, and alcohol yeast), isoleucine (alcohol yeast), lysine (wheat gluten meal and cottonseed meal), threonine (wheat gluten meal, cottonseed meal, and alcohol yeast), and valine (alcohol yeast).

The non-essential amino acids (NEAA) glycine and serine also had low ADC values for wheat gluten meal and alcohol yeast. Conversely, alanine had the highest ADC_NEAA values (>90%) in all proteic ingredients, whereas the other NEAA had a wide variation in ADC values. Additionally, corn gluten meal and soybean meal had the highest ADC_EAA (96.9 and 96.6%, respectively) and overall ADC_AA (corn gluten meal: 97.6%, soybean meal: 96.6%) values.

The crude and digestible amino acid compositions of proteic and energetic test ingredients were used to calculate the chemical scores and EAAI of ingredients relative to tambaqui white muscle protein (Table 5).

Tryptophan was the first limiting amino acid in ten ingredients (CSI: 0.06-0.51) and lysine was the second limiting amino acid in eight ingredients (CSI: 0.25-0.75).

Soybean meal had the highest EAAI (1.02) and was the most complete ingredient relative to the amino acid profile of juvenile tambaqui white muscle. Conversely, wheat gluten meal had the lowest EAAI (0.48).

Corn and wheat bran had the highest ADC_EAA and ADC_NEAA among energetic ingredients, in addition to the highest mean EAAI. In fish, corn digestibility depends on the digestibility capacity of each species (Halver and Hardy, 2002). Buzollo et al. (2018) evaluated the digestibility of crude protein, ether extract, and energy of some ingredients used in tambaqui diets and reported that corn had the highest ADC_protein (94.5%) of all ingredients tested. Guimarães et al. (2014), for tambaqui, and Abimorad et al. (2008), for pacu (Piaractus mesopotamicus) juveniles, also found high ADC_protein of corn (87.5 and 85.8%, respectively). Even though protein digestibility of corn by tambaqui and pacu is high, the comparison of amino acid profiles of ingredients and white muscle shows that corn protein quality was lower for tambaqui (EAAI: 0.84) than for pacu (EAAI: 1.03; Abimorad et al., 2008).

Other studies also reported lower protein and amino acid digestibility in wheat bran than in corn: Furuya et al. (2001) for Nile tilapia (Oreochromis niloticus), Abimorad et al. (2008) for pacu, and Wilson et al. (1981) for channel catfish (Ictalurus punctatus). According to Furuya et al. (2001), this reduced digestibility may be due to the shorter transit time of wheat bran in the gastrointestinal tract and its high content of crude fiber and non-starch polysaccharides. In fact, some of these polysaccharides, including pentosans and beta-glucans in triticale, may act as digestibility reducers, increasing intestinal viscosity and impairing enzymatic action (Furlan et al., 1997). Nevertheless, based on the high ADC values observed for tambaqui in this study, the digestibility of wheat bran and corn was not affected by crude fiber or polysaccharide content.

In this study, ADC of amino acids of broken rice were on average 16 and 13% lower than those of corn and wheat bran, respectively. These results are in agreement with our previous study (Buzollo et al., 2018), in which we observed low values of ADC_protein (71.21%) of broken rice for tambaqui. These low values may be related to the high levels of trypsin inhibitors in broken rice (Butolo, 2002). A similar ADC_protein of broken rice (81%) was reported by Abimorad and Carneiro (2004) for pacu. However, even lower ADC_protein values were reported for other carnivorous species: 43% for Pseudoplatystoma corruscans (Gonçalves and Carneiro, 2003) and 71% for hybrid striped bass (Morone saxatilis × M. chrysops) (Sullivan and Reigh, 1995). In fact, the enzymatic profile of carnivorous species does not support the use of starchy foods such as broken rice (Lundstedt et al., 2004). Nevertheless, higher ADC_protein values of broken rice than the ADC_protein of corn found in this study for tambaqui were reported for Rhamdia quelen (86%; Oliveira Filho and Fracalossi, 2006) and Nile tilapia (96%; Gonçalves et al., 2007), but these values may reflect methodological differences in fecal collection across studies.

Sorghum is the preferred substitute for corn due to its higher crude protein content and lower concentration of ether extract, lysine, and methionine in its composition (Antunes et al., 2007). In the current study, sorghum had larger quantities of phenylalanine, isoleucine, leucine, lysine, threonine, tryptophan, and valine than corn. Nevertheless, the ADC and EAAI of sorghum were low, indicating that for tambaqui, protein quality was significantly lower in sorghum than in the other energetic ingredients tested. Similar results were reported by Buzollo et al. (2018) for tambaqui and Pezzato et al. (2002), who found lower ADC_protein in sorghum than in corn for Nile tilapia. The low nutrient digestibility of sorghum may be due to tannins, which are an antinutritional factor found in many sorghum varieties (Rostagno, 1986).

In general, protein digestibility of energetic ingredients by tambaqui was high. Considering the large contribution of energetic ingredients in commercial diets, we conclude that they contribute significantly to meet the amino acid requirements of the species.

Corn gluten meal and soybean meal had the highest ADC of all proteic ingredients tested. High ADC_protein values of corn gluten meal have also been reported for tambaqui 98.09% (Buzollo et al., 2018) and other omnivorous and carnivorous species: 96% for Nile tilapia (Pezzato et al., 2002), 93.6% for largemouth bass (Micropterus salmoides; Portz and Cyrino, 2004), 92.3% for haddock (Melanogrammus aeglefinus; Tibbetts et al., 2004), 94.4% for cobia (Rachycentron canadum; Zhou et al., 2004), 95% for catfish (Oliveira Filho and Fracalossi, 2006), and 95.6% for pacu (Abimorad et al., 2008). In our study, corn gluten meal showed an imbalance in essential amino acid composition (EAAI: 0.82). The CSI for some amino acids such as leucine (2.21) and lysine (0.20) differed from this profile, and a similar imbalance in the same amino acids detected by CSI was also reported for pacu in a study by our research group (Abimorad et al., 2008). Thus, chemical scoring of AA in our study showed that, despite its high digestibility, availability of some amino acids in corn gluten meal is limited, which may hinder its use as the primary protein source in animal diets, further increasing its inclusion cost.

Soybean meal was the best protein source for tambaqui. This ingredient had the highest EAAI (1.02) and a balanced amino acid profile with chemical scores ranging from 0.60 for lysine to 1.46 for arginine. Similar to corn gluten meal, lysine had the lowest CSI (0.60) in soybean meal. Similar results have also been reported for other species, including channel catfish (Lim et al., 1998), Nile tilapia (Furuya et al., 2001; Köprücü and Özdemir, 2005), rainbow trout (Oncorhynchus mykiss; Cheng et al., 2003), largemouth bass (Portz and Cyrino, 2004), Murray cod (Maccullochella peelii peelii), Australian shortfin eel (Anguilla australis; De Silva et al., 2000), and pacu (Abimorad et al., 2008). Other studies observed reduced growth when using soybean meal as the primary protein source in carnivorous fish diets, which was mainly attributed to antinutritional factors and methionine deficiency (Anderson et al., 1993; Baeverfjord and Krogdahl, 1996; Degani, 1987; García-Gallego et al., 1998). Nevertheless, soybean meal is a potential substitute for protein sources such as fish meal and poultry byproduct meal in tambaqui diets.

Mean ADC values of poultry byproduct meal were significantly higher than those of fish meal. Conversely, Abimorad and Carneiro (2004) found no significant difference in ADC between the two ingredients for pacu. Chemical scores of EAA of poultry byproduct meal were high and showed little variation (0.75-1.58), except for tryptophan (0.31), which was limiting for tambaqui. Moreover, poultry byproduct meal had the third highest EAAI (0.93) of all ingredients tested. However, ADC of byproduct meals such as poultry byproduct meal may vary according to the composition and percentage of ingredients used in their production (Thompson et al., 2008).

In this study, the two fish meal sources tested, one made from tilapia filleting byproducts and produced in Brazil and one of Chilean origin made from salmon byproducts, had high ADC_AA for all amino acids. However, tilapia processing residue and salmon meal had only satisfactory amino acid profiles, with EAAI values of 0.96 and 0.80, respectively. In addition, the mean ADC for total amino acids was 5.0% higher in the poultry byproduct meal than in the processed tilapia residue. Crude protein and ash content indicate that poultry byproduct meal was a superior protein source for tambaqui over the processed tilapia residue: even though crude protein content was similar (poultry byproduct meal: 65.8%, processed tilapia residue: 60.2%), mineral matter content was higher in the processed tilapia residue (25.2%) than in the poultry byproduct meal (16.3%), indicating that a larger amount of bone was used in the processed tilapia residue production, resulting in an inferior ingredient.

Wheat gluten meal is an excellent protein source, but despite its high ADC values (mean: 84.5%), it had the lowest EAAI (0.48) as a result of the large variation in CSI and the low CSI of lysine (0.12) and tryptophan (0.06). Few studies have evaluated the digestibility of wheat gluten meal in fish (Buzollo et al., 2018; Allan et al., 2000; Robaina et al., 1999; Storebakken et al., 2000; Sugiura et al., 1998). Allan et al. (2000) also reported high ADC_AA of wheat gluten meal (100%) for Australian silver perch (Bydianus bydianus). However, none of the studies evaluated the CSI and EAAI of wheat gluten meal, and thus failed to determine the actual protein quality of wheat gluten meal for the species evaluated.

Similar to wheat gluten meal, feather meal and cottonseed meal also had high ADC, but an unbalanced amino acid profile, resulting in low chemical scores for tryptophan (feather meal: 0.06, cottonseed meal: 0.18) and low EAAI values (feather meal: 0.66, cottonseed meal: 0.61). Pezzato et al. (2002) compared the mean ADC_protein of proteic ingredients for Nile tilapia and reported that feather meal had the lowest ADC (29.1%), which was lower than the value observed for tambaqui. Feather meal hydrolysis was not as efficient at the time of that study, which may explain the low ADC found by Pezzato et al. (2002). Cottonseed meal had no harmful effects on tambaqui juveniles, despite the presence of gossypol, an antinutritional factor in cottonseed meal that can reduce its digestibility and affect biochemical processes by inhibiting enzyme activity (Beaudoin, 1985).

Mean ADC_AA values of alcohol yeast for tambaqui were low in this study. In our previous study (Buzollo et al., 2018), we also observed that alcohol yeast was not classified as a good source of protein and energy for juvenile tambaqui. Similar results were reported by Storebakken et al. (1998) for Atlantic salmon (Salmo solar), and the authors attributed the low digestibility of the alcohol yeast to the low digestibility of certain amino acids in its composition. In fact, essential amino acids of alcohol yeast such as arginine, threonine, and valine had low digestibility (<54%) by tambaqui juveniles. This low digestibility may be explained by the high inclusion level of alcohol yeast in test diets (300 g/kg), which has been generally lower in fish diets (Koch et al., 2015; Meurer et al., 2000; Sheikhzadeh et al., 2012). Moreover, the amino acid balance of alcohol yeast was suitable for the species, with little variation in chemical score and EAAI values (0.70).

This is the first study to combine apparent digestibility coefficients and chemical scores to evaluate a large number of ingredients used in fish diets for tambaqui. Our findings may improve least-cost diet formulations and enable effective substitution of ingredients that meet the limiting amino acid requirements of the species. Moreover, our findings may provide the basis for future studies on the digestible amino acid requirements for tambaqui.

Amino acids of proteic and energetic ingredients are well utilized by juvenile tambaqui. Corn and wheat bran have the highest mean ADC for total amino acids among energetic ingredients (95 and 92%, respectively), whereas corn gluten meal and soybean meal have the highest ADC for total amino acids among proteic ingredients (97.6 and 96.6, respectively).