SEARCH

SEARCH BY CITATION

Keywords:

  • amino acids;
  • isotopic routing;
  • Nile tilapia;
  • protein synthesis;
  • stable isotopes

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

1. Further advancement in the use of stable isotope analysis in animal ecology and physiology requires a better understanding of how organisms incorporate the macromolecular elements they consume into the tissues they biosynthesize.

2. Mixing models used to infer diets from isotopic data assume that assimilated macromolecules are dissembled into elements and then reassembled in animal tissues.

3. To test this assumption, we fed Nile tilapia diets with contrasting levels of protein and in which the carbon isotopic composition of protein differed from that of other macronutrients (carbohydrates and lipids). We then analysed the δ13C values of individual tilapia amino acids using compound-specific stable isotope analysis (CSIA).

4. The incorporation of dietary protein carbon was dependent on dietary protein content and on each amino acid’s biosynthesis pathway. The δ13C values of glycolytic amino acids, such as glycine, serine and alanine, had roughly constant values that reflected a large contribution of dietary carbohydrates and a small contribution of dietary protein. The δ13C values of aspartate and glutamate that are biosynthesized from Krebs cycle intermediaries paralleled bulk diet.

5. The δ13C values of indispensable amino acids resembled that of carbohydrates at low protein intakes but tended asymptotically to that of the δ13C values of their corresponding dietary amino acid as protein intake increased. This pattern is consistent with assimilation of indispensable amino acids of microbial origin by tilapias fed low protein food.

6. Our results suggest that the assumptions of mixing models are sound in situations where omnivores consume protein-deficient diets, as the elemental constituents used to biosynthesize amino acids in tissues may be derived from various ingested macromolecules (e.g. protein, carbohydrates, or lipids) and their elemental components. In contrast, for omnivores that consume sufficient amounts of protein, macromolecular (e.g. protein) routing likely occurs and mixing model assumptions are violated.

7. Our study shows that CSIA is a novel method to quantify the contribution of symbiotic microbes to the amino acid homeostasis of animals.


Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Stable isotope mixing models are often used to determine the relative contribution of various dietary sources to an animal’s diet (reviewed by Martínez del Rio et al. 2009). These models assume that ingested macromolecules (e.g. protein, lipids and/or carbohydrates) are dissembled into their elemental components (e.g. carbon and nitrogen) and then reassembled into the molecules used in tissue synthesis (Van der Merwe 1982; Martínez del Rio et al. 2009). This ‘scrambled egg’ assumption of mixing models is particularly unrealistic for amino acids because they are biosynthesized not from single carbon atoms but from various carbon skeletons. For example, the carbon skeletons of indispensable (i.e. essential) amino acids must be derived intact from assimilated nutrients as they cannot be biosynthesized de novo by many eukaryotes. Even for dispensable (i.e. non-essential) amino acids, the preferential inclusion of assimilated dietary amino acids in the consumer’s proteins is expected, as the use of premanufactured carbon skeletons reduces the costs of synthesis (Horton et al. 2002). An implication of these observations is that contrary to the predictions of mixing models, the carbon isotope (δ13C) value of animal protein in tissues should resemble not that of bulk diet but to be closer to that of dietary protein (Ambrose & Norr 1993). Direct incorporation of an amino acid from dietary protein into animal tissue is often referred to as isotopic routing (Schwarcz & Schoeninger 1991). The routing hypothesis has received support in bulk tissue experiments on birds (Podlesak & McWilliams 2006), mammals (Ambrose & Norr 1993; Voight et al. 2008), and fish (Kelly & Martinez del Rio 2010). While these experiments demonstrate that routing produces results that deviate from expectations based on mixing models, they were not able to characterize the isotopic patterns among specific amino acids that contribute to routing because they were conducted at the bulk tissue level.

Compound-specific stable isotope analysis (CSIA) offers an informative method for tracing the elemental origins of biomolecules used to biosynthesize tissues, but has been used sparingly to investigate routing in controlled experiments in comparison with bulk tissue analysis. The δ13C values of individual amino acids in a single protein can vary by >20‰, but similarity in the overall pattern of amino acid δ13C values among animals with different diets suggests that biosynthetic pathway is an important factor in driving the observed differences in δ13C values among amino acids in a single tissue (Hare et al. 1991). Recent controlled feeding experiments on vertebrates show that indispensable amino acids were routed directly from diet with little isotopic fractionation, whereas the δ13C values of dispensable amino acids can differ from those in diet by >20‰ (Howland et al. 2003; Jim et al. 2006; McMahon et al. 2010). Jim et al. 2006 estimated a minimum estimate of c. 50% routing of dietary amino acid carbon into rat bone collagen, with indispensable amino acids accounting for nearly half (c. 22%) of this total. Determining which dietary macromolecules provide the source of the other c. 50% of carbon in an animal’s proteinaceous tissues is an open question. The δ13C values of dispensable amino acids biosynthesized from glycolytic (e.g. glycine), or Krebs cycle (e.g. glutamate) precursors, however, are not closely correlated with their respective dietary amino acid but rather with bulk diet δ13C value (Howland et al. 2003; Jim et al. 2006). Moreover, in an experiment designed to assess resource allocation to reproduction in Lepidopterans, O’Brien, Fogel & Boggs (2002) found that dispensable amino acids in eggs were biosynthesized from ingested nectar sugars, which resulted in a large fraction (c. 50%) of total amino acids in butterfly protein being derived from non-protein sources.

To test the routing hypothesis in an omnivorous fish, we fed Nile tilapia (Oreochromis niloticus) a combination of diets in which the δ13C value of dietary protein (C3 casein, −27‰) was different from that of dietary carbohydrates and lipids (C4 corn, −11‰). To examine how routing responds to dietary protein quantity, we varied casein content from 3·75% to 30·0% (by mass). As predicted by the routing hypothesis, the isotopic value of bulk fish muscle was significantly different from that of bulk diet (and hence different from that predicted by a mixing model) when the protein content of diet was higher than 3·75% (Kelly & Martinez del Rio 2010). At the lowest protein content (3·75%), however, the δ13C value of bulk muscle was statistically indistinguishable from that of bulk diet (−13‰). This result was surprising because it suggested that the carbon in indispensable amino acids must be derived from non-protein dietary sources. Approximately 25% of the carbon in fish muscle protein is comprised of indispensable amino acids (Ozyurt & Polat 2006). Thus, if we assume that casein is the only source for indispensable amino acids and that the remaining carbon in dispensable amino acids was derived from the non-protein dietary sources (e.g. carbohydrates and lipids), tilapia muscle should have a δ13C value of c.−18‰, which is c. 5‰ lower than the observed δ13C value of tilapia muscle (−13‰) fed a low protein (3·75%) diet. It is possible that the source of this 13C-enriched carbon in tilapia muscle was protein biosynthesized by the fishes’ gastrointestinal microbiota from carbon derived from non-protein dietary sources. To test this hypothesis, we used CSIA to determine the isotopic composition of individual amino acids in the same tilapia muscle analysed by Kelly & Martinez del Rio (2010).

Previous studies and theoretical considerations suggest that the isotopic composition of amino acids of tilapia muscle in our experiment would be driven by two main factors: (i) the amino acid biosynthetic pathway (Hare et al. 1991; O’Brien, Fogel & Boggs 2002; Howland et al. 2003; Jim et al. 2006) and (ii) the protein content of diet. Other factors such as growth rate and metabolic state are also important considerations, because several amino acids (e.g. proline) can be conditionally indispensable when de novo synthesis does not meet requirements for growth (Ball, Atkinson & Bayley 1986; Chung & Baker 1993; Kirchgessner, Fickler & Roth 1995; Dabrowski et al. 2005); such factors are addressed in the Discussion. We established three amino acid categories a priori (Fig. 1). The first category includes dispensable amino acids biosynthesized from precursors derived during the first steps of glycolysis (Horton et al. 2002), including glycine (Gly), serine (Ser) and alanine (Ala). Alanine can also be biosynthesized in a second pathway from aspartate via oxaloacetate, an intermediate in the Krebs cycle. Because our experimental diets had a relative excess of carbohydrates relative to proteins or lipids, the isotopic composition of ‘glycolytic’ amino acids should reflect that of carbohydrates and be relatively independent of the amino acid composition of the diet given the small range (0–30%) in diet protein content (Fig. 1). The second category includes the amino acids such as aspartate (Asp), glutamate (Glu), arginine (Arg), proline (Pro) and possibly alanine (see earlier) that are biosynthesized from precursors derived from Krebs cycle reactions (Horton et al. 2002). Because these inputs into the Krebs cycle originate from dietary carbohydrates, lipids and protein, the carbon isotope composition of this second class of dispensable amino acids should reflect a mixture of dietary sources and be similar to the δ13C value of bulk diet (Fig. 1). Finally, the third category includes the indispensable amino acids – isoleucine (Ileu), leucine (Leu), phenylalanine (Phe), threonine (Thr) and valine (Val). If indispensable amino acids biosynthesized by microbes in the gut provided a significant source to the consumer, then indispensable amino acids in consumer tissues would have intermediate δ13C values between those of the corresponding dietary amino acid and the isotopic composition of bulk diet. Conversely, if microbial contribution was minimal, the δ13C value of indispensable amino acids should tend towards values similar to that of the amino acid in dietary protein (casein) with increasing dietary protein content. Figure 1 illustrates these expectations graphically. In the Models and Statistical Analyses section in Materials and Methods, we transform the expectations described previously into mathematical models.

image

Figure 1.  Conceptual diagram of how the δ13C value of three categories of amino acids were expected to change with increasing dietary protein content. Asterisk for alanine (Ala) denotes that this amino acid can be biosynthesized from glycolytic and/or Krebs cycle precursors. Glycolytic amino acids (Ser, Gly, Ala) were expected to mirror the isotopic composition of dietary carbohydrates irrespective of dietary protein content. Amino acids derived from Krebs cycle intermediaries (Glu, Asp, Ala) were expected to track the isotopic composition of bulk diet. The δ13C value of indispensable amino acids (Phe, Thr, Leu, Ileu, Val) were expected to show a nonlinear response with increasing dietary protein because they could be sourced directly from dietary amino acids and/or indispensable amino acids biosynthesized by gut microbes that utilize bulk diet as a carbon source.

Download figure to PowerPoint

Materials and methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Tilapia experimental design and sample collection

Fingerling tilapias were purchased in June 2007 (AmeriCulture, Inc., Animas, NM, USA) and were housed in the University of Wyoming’s Red Buttes Animal Care Facility. Fish with an average initial mass (±SD) of 0·51 g (±0·07) were randomly assigned to one of four different experimental groups and placed individually in tanks (25·5 × 25·5 × 15 cm) at 24 °C (±1 °C) and a 12L : 12D hour photoperiod. The throughput rate of the holding tanks was 80–90 mL min−1. The four experimental groups were fed diets that varied in (i) relative protein content (from 3·75% to 30% dry weight) and (ii) isotopic composition (Table 1); refer to Kelly & Martinez del Rio (2010) for details regarding the composition of each diet treatment. We varied the protein content from 3·75% to 30% in an attempt to capture the natural variation in the diet of wild omnivorous tilapia, which are known to vary seasonally (Lim & Webster 2006). The mean δ13C value (±SD) of the protein source (casein) was different (−27 ± 0·1‰) from that of all other dietary components – corn starch, corn syrup, corn oil, cellulose, methyl cellulose – that had mean δ13C values of −11‰ (±0·1). Fish were fed once daily at a ration that exceeded their maximal intake. CO2 asphyxiation was used to euthanize five fish per treatment (AVMA 2007), chosen at random when the average mass of all fish in the treatment was equal to c. 300% of starting mass. Because growth rate differed among fish, the time at which each fish was euthanized varied between 28 and 133 days. This protocol was approved by the University of Wyoming’s IACUC committee (#A-3216–01). Euthanized fish were immediately stored frozen at −20 °C. Dissected lateral myotome muscles from each fish were dried at 40 °C and ground into a homogenous mixture. Lipid was then extracted from the mixture with three sequential petroleum ether extractions.

Table 1.    Mean δ13C values of individual amino acids measured in each of the four protein treatments (3·75%, 7·5%, 15·0%, 30·0%) and dietary protein source (casein)
Amino acidPathwayCaseinProtein content
3·75% (5)7·5% (4)15·0% (4)30·0% (4)
  1. Numbers in parentheses adjacent to δ13C values represent standard deviation; numbers in parentheses adjacent to protein contents represent the number of tilapia analysed in each treatment. Amino acid pathways as described in the text are labelled; asterisk denotes that arginine and proline can be conditionally indispensable in rapidly growing animals such as the tilapia examined in our experiments.

Glycine (Gly)Glycolytic−7·8 (6·4)−2·3 (2·2)−7·5 (1·8)−7·3 (3·0)−8·2 (4·2)
Serine (Ser)Glycolytic−26·3 (4·0)−5·6 (7·1)−10·4 (3·2)−11·4 (3·3)−13·5 (2·7)
Alanine (Ala)Glycolytic/Krebs−19·7 (0·8)−9·1 (2·2)−9·9 (2·3)−12·0 (0·7)−12·0 (2·0)
Aspartate (Asp)Krebs−23·4 (0·9)−10·2 (1·9)−12·2 (1·2)−13·6 (1·4)−15·3 (1·9)
Glutamate (Glu)Krebs−25·4 (0·3)−12·1 (2·0)−13·6 (1·6)−14·6 (1·4)−16·5 (1·5)
Arginine (Arg)Krebs*−18·3 (1·4)−10·3 (1·3)−13·5 (2·1)−13·1 (1·7)−14·4 (1·7)
Proline (Pro)Krebs*−25·2 (0·6)−17·0 (3·1)−19·7 (1·5)−20·7 (2·4)−22·7 (0·8)
Isoleucine (Ileu)Indispensable−27·6 (2·1)−33·9 (3·6)−36·1 (1·6)−36·1 (1·6)−40·2 (1·5)
Leucine (Leu)Indispensable−34·8 (0·7)−22·5 (3·7)−23·6 (2·9)−25·8 (1·7)−29·2 (1·6)
Phenylalanine (Phe)Indispensable−32·0 (0·3)−24·0 (3·1)−25·6 (2·0)−27·2 (1·4)−29·6 (1·1)
Threonine (Thr)Indispensable−18·4 (2·2)−10·3 (2·6)−12·8 (1·8)−14·3 (1·5)−16·6 (2·3)
Valine (Val)Indispensable−31·8 (1·9)−19·6 (3·3)−21·7 (1·8)−23·4 (2·3)−26·7 (2·1)

Compound-specific (amino acid) stable isotope analysis

We measured the δ13C value of individual amino acids in various tissues using gas chromatograph combustion isotope ratio mass spectrometry. Approximately 1·5 mg of dried tilapia muscle or casein was hydrolysed to amino acids in 1 mL of 6N hydrochloric acid at 110 °C for c. 20 h. Hydrolysed amino acids were then derivatized to N-trifluoroacetic acid isopropyl esters (Fantle et al. 1999; O’Brien, Fogel & Boggs 2002) in several batches and run in triplicate. Derivatized samples were injected into a Varian 3400 gas chromatograph for separation, converted to CO2 gas via a combustion interface and analysed with a Finnigan Delta Plus XL (Thermo Scientific's mass spectrometry division is in Bremen, Germany) isotope ratio mass spectrometer. Isotopic results are expressed as δ values, δ13C = 1000* [(Rsample − Rstandard)/Rstandard], where Rsample and Rstandard are the 13C/12C ratios of the sample and standard, respectively. The internationally excepted standard is Vienna-Pee Dee Belemnite limestone for carbon, and the units are expressed as parts per thousand or per mil (‰).

Using this derivatization method, measurements of δ13C value for a given amino acid include carbon from both the amino acid and the derivatization reagents (e.g. isopropanol), which can be heavily fractionated depending on the synthesis pathway and on the number of exchangeable vs. non-exchangeable carbon atoms in each amino acid. Amino acid standards of known δ13C composition were derivatized and analysed along with each batch of samples. The results of the standards reflect the effect of carbon added by the derivatization reagents and provided the information required to calculate the δ13C value of amino acid carbon in the samples. Individual amino acid δ13C values (δ13CAAsample) were calculated using the equation:

  • image((eqn 1))

The terms δ13Cdsa and δ13Cdst refer to the derivatized sample and standard, respectively, and std refers to the standard. pstd is equal to the proportion of carbon in the derivative that was sourced from the amino acid (Fantle et al. 1999; O’Brien, Fogel & Boggs 2002). We resolved 12 amino acids using these methods, and seven of which are typically considered dispensable for most eukaryotes – glycine, serine, alanine, aspartate, glutamate, arginine and proline – and five of which are considered indispensable, including threonine, valine, leucine, isoleucine and phenylalanine. A set of amino acid standards were analysed every three unknown samples. The standard deviation of δ13C values among standards ranged from 0·8‰ for phenylalanine to 2·6‰ for glycine (n = 15). The average δ13C standard deviation for all amino acids measured was 1·3‰.

Models and statistical analyses

The relationship between amino acid δ13C value and per cent protein (%) in diet was fitted to models that represent three possible origins (e.g. glycolytic, Krebs or microbial dilution):

Model 1

For dispensable amino acids biosynthesized from glycolytic precursors, we used the constant function:

  • image((eqn 2))

where β0 represents the δ13C value of carbohydrates plus a discrimination factor because of potential fractionation during biosynthesis (i.e. β0 = δ13Ccarbohydrates + Δ).

Model 2

For dispensable amino acids derived from Krebs cycle intermediaries, a linear function was used:

  • image((eqn 3))

which represents the assumption of mixing in cataplerotic reactions. If we interpret eqn 3 as a linear mixing model, β0 is as defined earlier, whereas β1 is the change in δ13C value per unit increase in the percentage of protein in the diet (with units ‰ per % protein).

Model 3

The δ13C value of indispensable amino acids was estimated from a mixing model in which the contributions of amino acids from direct incorporation from the diet are related to potential microbial inputs:

  • image((eqn 4))

where δ13CAA is the isotopic composition of the dietary amino acid, and p(%) is the fraction of the amino acid derived directly from diet. The δ13CM is the isotopic composition of the amino acid biosynthesized by gut microbes, which we assume is more similar to the isotopic composition of the bulk diet than to the corresponding dietary amino acid. We use the symbol (%) to signify that the value of p is a function of the protein percentage in diet (see eqn 2). In this experiment, we have hypothesized that the fraction p depends on the protein content in the diet (%). The simplest form for the dependence of p on % assumes that the relative amount of the amino acid derived directly from diet is proportional to the % of protein in diet (i.e is equal to α%, where α is a proportionality constant), whereas the contribution of microbially derived carbon in these amino acids is constant (λ). Thus,

  • image((eqn 5))

The parameter λ/α in eqn 5 represents the ratio of the contribution of microbes relative to that of the corresponding dietary amino acid when the protein content of the diet is very small. Therefore,

  • image((eqn 6))

Equation 6 describes a process in which the isotopic signal of microbially biosynthesized amino acids become diluted as the contribution of dietary indispensable amino acids increases. Thus, for indispensable amino acids derived from a mixture of microbial synthesis and diet, we used a re-parameterization of eqn 6 that leads to Model 3:

  • image((eqn 7))

where β0 = δ13CM, β1 = δ13CAA − δ13CM, and β2 = λ/α. To fit the three models (eqns 2, 3, and 7), we used standard linear squares methods, including a Levenberg-Marquardt method in the case of eqn 7 (Gill & Murray 1978). We used small sample Akaike Information Theoretical Criteria (AICc) to assess the relative support for each model given our data (Burnham & Anderson 2002). If two models (i and j) differed in AICc by more than 3 (i.e. if Δij = AICci − AICcj), we deemed the model with the lowest AICc as better supported (Burnham & Anderson 2002). If the difference in AICc between two models was ≤3 (i.e. if the evidence ratio between two models was <2·7), we considered that both models are similarly supported and used other criteria to evaluate the merit of each competing model.

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

As reported in Kelly & Martinez del Rio (2010), dietary protein content had a significant effect on fish growth rates, but only for tilapia in the 30% protein treatment. Growth rates for tilapia in the 3·75%, 7·5% and 15% dietary protein treatments were not different from one another.

The δ13C value of individual amino acids varied with the content of protein in diet (Table 1). The relationship between an amino acid’s δ13C value and the protein content in diet revealed three distinct patterns. At low dietary protein concentrations, glycine, serine and alanine had δ13C values that were more positive than the δ13C value of dietary carbohydrates, but rapidly decreased to relatively constant values (Fig. 2). The difference between the δ13C value of these amino acids and of carbohydrates at low dietary protein content (3·75%) is approximately equal to the discrimination factor (Δ) in the equation (β= δ13Ccarbohydrates + Δ) that results from isotopic fractionation during biosynthesis. Our model selection procedure (Table 2) ranked the linear model (Model 2) as superior to the constant model (Model 1). Because it is evident that the relationship between δ13C value and dietary protein in this situation was nonlinear (Fig. 2), however, we used an ad hoc exponential model that provides a more flexible fit than the linear model:

  • image((eqn 8))

This equation (Model 4) was selected as the model with the strongest support in all the amino acids assumed a priori as having glycolytic precursors (Table 2). For alanine and serine, the asymptote (β1) of this model did not differ significantly (t = 1·4 and 0·69, P > 0·1, respectively) from the δ13C value of carbohydrates (−11‰). The β1 value for glycine was significantly more positive than that of carbohydrates in the diet (β1 ± SE = −7·7 ± 1·0, t = 3·3, P < 0·05). For consistency, we applied this ad hoc model to all amino acids (Table 2).

image

Figure 2.  Change in the δ13C value of glycolytic amino acids (Gly, Ser, Ala) in tilapia muscle as a function of dietary protein content. As in Fig. 1, the coarse dashed line is the δ13C value of dietary carbohydrates (−11‰); the fine dashed line is the δ13C value of the amino acid in dietary protein; the solid line represents the δ13C value of bulk diet carbon. In all cases, the nonlinear model (eqn 8) provided the best fit to the change in δ13C values of glycolytic amino acids with increasing dietary protein content (Table 2).

Download figure to PowerPoint

Table 2.    We used Akaike Information Theoretic Criteria corrected for small samples (AICc = nlog(inline image) + 2K(n/(n − − 1) to compare the support of four alternative models given the data (Burnham & Anderson 2002)
Amino acidModel 1Model 2Model 3Model 4Best model r2
β0β0 + β1%inline imageinline image
  1. Numbers in bold represent the AICc value of the model with strongest support. If two models differed in AICc by more than 3, we deemed the model with the lowest AICc as better supported (Burnham & Anderson 2002). If the difference in AICc between two models was ≤3, we considered both models are similarly supported and used other criteria to evaluate the merit of each competing model. Note that β0 and β1 are defined differently for each model; see text (Materials and Methods and Results) for details. Note that whereas Models 1, 2 and 3 have clear mechanistic interpretations, Model 4 is used only for descriptive purposes. We used inline image to describe the goodness-of-fit of the model best supported by data.

Glycine (Gly)44·043·2No Fit42·90·40
Serine (Ser)56·451·4No Fit48·00·61
Alanine (Ala)26·124·6No Fit23·20·54
Arginine (Arg)31·626·6No Fit23·10·57
Aspartate (Asp)34·620·517·519·20·76
Glutamate (Glu)30·716·316·616·70·63
Proline (Pro)40·129·126·527·00·70
Isoleucine (Ileu)41·926·730·230·40·64
Leucine (Leu)41·928·227·127·70·76
Phenylalanine (Phe)39·036·525·026·20·79
Threonine (Thr)41·827·123·624·20·72
Valine (Val)45·431·330·431·00·71

Aspartate and glutamate, which have Krebs cycle intermediary precursors, had similar, albeit not identical patterns (Fig. 3). In both cases, δ13C values decreased monotonically as dietary protein content increased, and the δ13C value of each amino acid was intermediate between the δ13C value of dietary carbohydrates and the carbon isotope value of bulk diet (Fig. 3). Model 2 (linear function, eqn 3), however, was better supported by data for glutamate, whereas Model 3 (microbial dilution, eqn 7) was better supported for aspartate (Table 2). It is notable that for both aspartate and glutamate, the slope of the linear model was similar to that of the relationship between the δ13C value of bulk diet and dietary protein content denoted by the bold solid line in Fig. 3. The slopes (±SE) of the linear model for aspartate and glutamate were 0·19 (±0·04) and 0·17 (±0·03) per mil, respectively, for every 1% increase in the amount of dietary protein. Although Model 4 (exponential function, eqn 8) and Model 3 (microbial dilution, eqn 7) described the data for aspartate and glutamate well (Table 2), the asymptotic values that these models predict were higher than the δ13C value of these amino acids in diet. Asymptotic values (±SE) were −15·7 ± 8‰ and −18·1 ± 9·0‰ for aspartate and glutamate, respectively, and both were significantly higher (> 8·0, P < 0·001) than the respective δ13C value of the amino acid in diet (Fig. 3; Table 1).

image

Figure 3.  Change in the δ13C value of Krebs cycle amino acids (Glu and Asp) in tilapia muscle as a function of dietary protein content; see caption for Fig. 1 for definitions of the solid and dashed lines. A linear function (eqn 3) provided the best fit for glutamate with increasing dietary protein content, whereas the microbial dilution model (eqn 7) was better supported by the data for aspartate (Table 2).

Download figure to PowerPoint

With the exception of isoleucine, the data supported Model 3 (microbial dilution, eqn 7) over the constant and linear models for all indispensable amino acids (Table 2). For isoleucine, Model 2 (linear function, eqn 3) received the strongest support (Table 2). All indispensable amino acids had δ13C values that were intermediate between that of the value of the corresponding dietary amino acid and the isotopic composition of bulk diet (Fig. 4). To further confirm the adequacy of the microbial dilution model, we estimated the δ13C value of the amino acid found in diet from the asymptotic value reached by eqn 7 when the protein content in diet is high (i.e. the limit of eqn 7 when protein % tends to infinity, which equals β0 + β1). The model very accurately predicted that the δ13C value of the amino acid in tissues would become equal to that in diet when protein content is >30%.

image

Figure 4.  Change in the δ13C value of indispensable amino acids (Phe, Thr, Leu, Ileu and Val) in tilapia muscle as a function of dietary protein content; see caption for Fig. 1 for definitions of the solid and dashed lines. In all cases, the nonlinear microbial dilution model (eqn 7) provided the best fit to the change in δ13C values of indispensable amino acids with increasing dietary protein content (Table 2).

Download figure to PowerPoint

Arginine and proline had patterns that linked them with different categories (Fig. 5). Arginine had a pattern that resembled that of glycolytic amino acids (Gly, Ser, Ala), and Model 4 (exponential function, eqn 8) provided the best fit to the data (Table 2). Unlike glycine, serine and alanine, the asymptotic value (β0) for arginine was significantly lower (−13·8 ± 0·6‰) than the δ13C value of dietary carbohydrates. In contrast, the dependence of proline’s δ13C value on dietary protein content (Fig. 5) strongly resembled that of indispensable amino acids. For proline, Model 3 (microbial dilution, eqn 7) was marginally better supported by the data (Table 2) than Model 4 (exponential function, eqn 8) and the asymptotic value (±SE) of the model (−24·2 ± 5·0) was statistically indistinguishable (t = 0·2, P > 0·5) from that of the δ13C value (−25·2 ± 1·0) of dietary proline (Table 1).

image

Figure 5.  Change in the δ13C value of amino acids arginine and proline in tilapia muscle as a function of dietary protein content; see caption for Fig. 1 for definitions of the solid and dashed lines. For arginine, the nonlinear model (eqn 8) provided the best fit to the change in δ13C value with increasing dietary protein content (Table 2). For proline, the nonlinear microbial dilution model (eqn 7) provided the best fit to the change in δ13C value with increasing dietary protein content (Table 2).

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

With a few exceptions, the relationship between an amino acid’s δ13C value and protein was that expected a priori from the amino acid’s biosynthetic pathway. In some cases, however, our simplistic models predicted the form of the relationship poorly. For example, it is clear that a constant function is a poor descriptor for the patterns exhibited by the amino acids derived from glycolytic precursors. In other cases, however, our models provided a good fit, especially for indispensable amino acids.

Glycine, serine and alanine

Because glycine, serine and for the most part alanine are biosynthesized from glycolytic precursors and our experimental diets were rich in carbohydrates, we (incorrectly) hypothesized that their isotopic composition would resemble that of diet and that it would be relatively independent of dietary protein content. In contrast, we found that the δ13C values of all these amino acids were more positive at the lowest protein content (3·75%), and then became relatively constant with increasing dietary protein (Fig. 2). Model 4 (exponential function, eqn 8) with a relatively high rate constant best described this pattern (Table 2). Two factors are likely driving the observed pattern. First, the δ13C value of these amino acids represents the mixing of a large fraction of carbon derived from carbohydrates and a small contribution from dietary amino acids. The amino acid contribution from diet is negligible at very low protein intakes but increases to a small constant fraction that makes a measurable difference if the dietary protein content exceeds c. 7·5%. Second, the biosynthesis of glycolytic amino acids from carbohydrates likely has a positive discrimination factor as described in the following paragraphs (Fig. 2).

Alanine can be biosynthesized via two distinct pathways – one uses pyruvate in glycolysis as the precursor and the other utilizes aspartate derived from oxaloacetate, an intermediate in the Krebs cycle. Our data suggests that the former pathway was most prevalent in our experiment because the asymptotic δ13C value of tilapia muscle alanine more closely resembled the δ13C value of dietary carbohydrates than bulk diet (Fig. 2). In the glycolytic pathway, a positive discrimination factor between carbohydrates and alanine is likely as it is biosynthesized in one step by the transamination of pyruvate. DeNiro & Epstein (1977) have shown that de-carboxylation of pyruvate to coenzyme-A discriminates against 12C (see also Melzer & Schmidt 1987), and as a consequence, the remaining pyruvate pool must be enriched in 13C relative to assimilated carbohydrates.

Further enrichment of glycine could result from the transfer of 12C from serine to tetrahydrofolate (Horton et al. 2002). The 13C-enrichment of glycine relative to diet is a well-documented observation (Tuross, Fogel & Hare 1988; Howland et al. 2003; Jim et al. 2006) that seems to explain why the δ13C value of bone collagen, composed of c. 33% glycine (Eastoe 1957), is c. 4‰ more positive than that of other tissues commonly analysed in ecology (Koch 2007). The high δ13C values of serine and glycine at low protein intakes could result from enrichment in the isotopic composition of the carbohydrate pool, which at low protein concentrations would experience great metabolic demand. These amino acids are derived primarily from 3-phosphoglycerate, which can be biosynthesized directly from glucose or from the, presumably 13C-enriched, pyruvate pool (Walsh & Sallach 1966).

Aspartate and glutamate

The carbon isotope composition of aspartate and glutamate exhibited similar patterns with increasing protein intake. These patterns qualitatively resembled those predicted by our a priori linear model (Model 2, eqn 3) and are close to those expected from amino acids biosynthesized from Krebs cycle precursors (Fig. 3). A linear model (Model 2) was best supported by the glutamate data, but a nonlinear microbial dilution model (eqn 7) was better supported in the case of aspartate (Table 2). There are two reasons to hesitate to accept the nonlinear model as the best fit for aspartate. First, the difference in the AICc values between the linear and nonlinear model was small (Table 2), indicating at least partial support for the linear model. Second, the slope of the linear model relating δ13C value with % protein in diet was not significantly different from that of bulk diet with % protein in diet (Fig. 3).

Aspartate and glutamate are biosynthesized from oxaloacetate and α-ketoglutarate, respectively. These two compounds are generated by a variety of precursors in anaplerotic reactions within the Krebs cycle. Multiple cycling of metabolites through the Krebs cycle could result in carbon skeletons are that have been ‘shuffled’, because several of the intermediates in the Krebs cycle (e.g. succinate and fumarate) are symmetrical molecules. This observation suggests that the isotopic analysis of aspartate and glutamate in various tissues may be particularly useful in studies attempting to reconstruct bulk diet (Howland et al. 2003).

Indispensables: isoleucine, leucine, phenylalanine, threonine and valine

With the exception of isoleucine, our results for the indispensable amino acids were best characterized by Model 3, the nonlinear microbial dilution model (Table 2 and Fig. 4). Isoleucine was best characterized by a linear function (Table 2), but its overall carbon isotope composition was more similar to that of other indispensable amino acids than those associated with glycolytic or Krebs precursors (Fig. 4). At high dietary protein contents, the asymptotic value of all indispensable amino acids closely resembled that of the corresponding amino acid in dietary protein (Fig. 6), suggesting that indispensables are being directly routed from diet (Fig. 4).

image

Figure 6.  Relationship between the asymptotic δ13C value predicted by the microbial dilution model (eqn 7) and δ13C value of the dietary amino acid for the indispensable amino acids such as Phe, Thr, Leu, Ileu, Val, and conditionally indispensable Pro.

Download figure to PowerPoint

The isotopic composition of the indispensable amino acids at low dietary protein contents, however, was significantly higher than the corresponding amino acid δ13C value in dietary protein. For example, the mean δ13C value (±SD) of tilapia threonine (−10·3 ± 2·6‰) was indistinguishable from that of dietary carbohydrate at low dietary protein content (3·75%). Likewise, the mean δ13C value of tilapia valine (−19·6 ± 3·3‰) at low dietary protein content was more similar to that of bulk diet than the δ13C value of valine (−31·8 ± 1·9‰) in dietary protein (Table 1). The other three indispensable amino acids (Ileu, Leu and Phe) had δ13C values intermediate between bulk diet and dietary protein. There are several potential reasons for this pattern, because all of these amino acids (Thr, Val, Ileu, Leu and Phe) are known to be indispensable to rapidly growing tilapia (Santiago & Lovell 1988). First, isotopic fractionation can occur during the catabolism of amino acids from available bodily tissues (e.g. muscle, bone) leading to higher δ13C values in tilapia muscle vs. dietary amino acids (Macko & Estep 1985). Tilapia were fed a daily ration that exceeded their maximal intake to maintain rapid growth and tripled their mass during the course of the experiment (Kelly & Martinez del Rio 2010); thus, it is unlikely that tissue amino acids were catabolized for energy. Second, it is possible that tilapia were acquiring some of their indispensable requirements by grazing algae, microbial biofilms, phytoplankton, or other detritus in the holding tanks (Bowen 1980); however, this source was likely minimal because we frequently cleaned the tanks and the throughput rate was high (80–90 mL min−1) in comparison with the size of the tanks (25·5 × 25·5 × 15 cm). Lastly, this pattern could be caused by the in vivo synthesis of amino acids from the bulk diet (i.e. carbohydrates) by gut microflora to provide indispensable amino acids to tilapia when they are fed protein-deficient diets (Kelly & Martinez del Rio 2010).

One of the many functions of the gastrointestinal microbiota is to provision hosts with essential molecules (e.g. vitamins). The microbial contribution to the host’s nutrition in terms of amino acids has been relatively well studied in ruminants (Kung & Rode 1996; Karasov & Carey 2008); however, whether a majority of amino acids can originate from microbiota for non-ruminants like tilapia remains much less clear (Fuller & Reed 1998). A number of studies have relied on the oral administration of non-specific 15N as urea or ammonium chloride to label lysine biosynthesized by microbes and trace its flux into the host because animals do not transaminate lysine (D’Mello 2003). The method has the serious shortcoming of being limited to lysine. Nonetheless, it has revealed a significant contribution of microbial lysine to the lysine in the host’s body protein (reviewed by Metges 2000).

How much does microbiota contribute to the indispensable amino acid budget of tilapia?

Our results point to a novel method to quantify the contribution of the microbiota to the indispensable amino acid intracellular pool. The nonlinear microbial dilution model (Model 3) seems to describe adequately well the variation of δ13C values of most of the indispensable amino acids with changes in dietary protein intake (Fig. 4). The terms in the equation that represent Model 3 are biologically relevant and simple to interpret: β= δM, β+ β1 = δAA and inline image equal the fraction of carbon in an amino acid derived directly from diet (compare eqns 6 and 7). Thus, the fraction of carbon in an amino acid derived from microbes, which we will designate by pM(%), is equal to inline image.

Inspection of the estimates that result from fitting isotopic values for each amino acid to eqn 7 revealed that this equation accurately estimates δAA (Fig. 6). The estimated isotopic values of the amino acids biosynthesized by microbes from carbohydrate precursors (δM), however, are unrealistically enriched in 13C (Fig. 4). These estimated δM values ranged from 13·5‰ to 22·0‰. Considering that the non-protein carbon sources had a δ13C value of −11‰, the observed δ13C values would require an unreasonably large discrimination against 12C during microbial biosynthesis. Previous work suggests that microbial synthesis of amino acids by heterotrophic microbes has modest discrimination values and often discriminates against 13C (Blair et al. 1985; Macko et al. 1987). Because an incorrect estimate of δM will lead to biased estimates of pM(%), the model needs to be constrained with realistic values for δM.

To estimate pM(%), we constrained the value of β= δM in the microbial dilution models using Δsource amino acid values estimated by Larsen et al. (2009) for an Actinobacterium and Rhodococcus spp. grown on a medium with sucrose as the sole carbon source (Table 3). Table 3 lists the δM values that we used to constrain the microbial dilution model and the AICc values for this model. Despite the assumption that these microbes are similar to gut microbiota, the AICc values revealed that this new constrained model performed well relative to the unconstrained model. Indeed, the estimated δAA value was very similar to that estimated by the unconstrained model and to the value of the amino acid in diet. Furthermore, the curves generated by both models were indistinguishable over the range of dietary protein content (3·75–30%) in our experiments. More strikingly, the pattern of change in amino acids derived from microbes, pM(%), was remarkably similar among all indispensable amino acids (Fig. 7). The model estimated substantial microbial contributions ranging from c. 55% to 70% of indispensable amino acids when fish were fed diets with low (3·75%) protein content, but declined rapidly as dietary protein content increased. The model suggests that when growing tilapia were fed diets with optimal protein content (c. 45%, Lim & Webster 2006), the contribution of microbes to amino acids was small (10–20%) but significant (Fig. 7). Again, we emphasize the tentative and speculative nature of this analysis. Only direct measurement of Δsource amino acid for gastrointestinal microbiota will permit quantitative estimates of the contribution of the gut microbiota to the assimilation of indispensable amino acids.

Table 3.    We used Δ13Csource amino acid values for two heterotrophic bacteria (Larsen et al. 2009) to constrain the δM = β0 value included in the microbial dilution model. Judging from its lower AICc values, this new model performed well relative to the ‘unconstrained’ model. Constraining the value of δM permitted fitting the microbial dilution model for isoleucine
Amino acidΔ13Camino acid sourceδMAICc
Isoleucine (Ileu)−2·4−13·430·4
Leucine (Leu)−1·1−12·126·7
Phenylalanine (Phe)−4·0−15·116·0
Threonine (Thr)7·0−4·020·7
Valine (Val)−2·0−13·027·5
image

Figure 7.  Relationship between the fractional contribution (%) of indispensable amino acids derived from gastrointestinal microbiota with increase in casein (protein) content of diet.

Download figure to PowerPoint

Outliers: arginine and proline

Our model approach linked arginine and proline to different precursor categories. Arginine is considered to be a conditionally indispensable amino acid because it can be biosynthesized de novo via conversion of citrulline, but this pathway is energetically expensive and may not produce sufficient arginine, especially in rapidly growing (i.e. young) animals. Citrulline can be derived from multiple sources, but one of the most common precursors is orthinine, which is produced via the catabolism of glutamate, a Krebs cycle intermediate. For arginine, however, our modelling approach identified its isotopic pattern as being more similar to that of the dispensable glycolytic amino acids (Gly, Ser, Ala) than indispensable amino acids. Arginine δ13C values were slightly higher than dietary carbohydrates at low dietary protein content but then decreased to values intermediate between bulk diet and dietary carbohydrates as dietary protein content was increased (Fig. 5). This suggests that sufficient amounts of arginine can be biosynthesized de novo from glutamate (Caldovic & Tuchman 2003) derived from bulk diet when diets are protein deficient. As dietary protein was increased to moderate levels (c. 7·5%), tilapia arginine δ13C values dropped below that of dietary carbohydrates and remained relatively constant with increasing dietary protein content. This pattern is similar to that observed for the glycolytic amino acids (Gly, Ser and Ala) and suggests that some dietary arginine was being routed into tilapia muscle when moderately available in diet (Fig. 5).

Like arginine, proline can be biosynthesized de novo from glutamate, a dispensable amino acid biosynthesized from precursors derived from Krebs cycle reactions that closely tracked the δ13C value of bulk diet in our experiments (Fig. 3). Our modelling approach (Table 2), however, identified Model 3 (microbial dilution, eqn 7) as having a similar amount of support as Model 4 (exponential function, eqn 8). In addition, proline’s overall carbon isotope composition with increasing dietary protein content was more similar to that of indispensable amino acids than amino acids biosynthesized from Krebs cycle precursors. If proline was being biosynthesized from glutamate at low dietary protein contents (3·75%), then tilapia proline δ13C values would be higher than the bulk diet δ13C value because of a positive fractionation during de novo biosynthesis, which is similar to the pattern observed for aspartate and glutamate (Fig. 3). At low dietary protein contents, however, tilapia proline δ13C values were intermediate between that of bulk diet and dietary protein sources (Fig. 5 and Table 1). This pattern is more similar to that of indispensable amino acids (Fig. 4) than dispensable amino acids biosynthesized from Krebs cycle precursors (Fig. 3). Furthermore, proline δ13C values decreased significantly with increasing protein content such that its asymptotic δ13C value at high dietary protein content (30%) was indistinguishable from that of proline in dietary protein (Fig. 5). Thus, there was little carbon isotope fractionation of proline from diet to tissue at high dietary protein contents, which is a pattern expected of indispensable amino acids when readily available in diet. In agreement with data from recent experiments (Ball, Atkinson & Bayley 1986; Chung & Baker 1993; Kirchgessner, Fickler & Roth 1995; Dabrowski et al. 2005), our results suggest that proline is conditionally indispensable for tilapia fed protein-deficient diets (3·75%) and is directly routed with minimal isotopic fractionation at high dietary protein contents (30%).

Implications for isotopic ecology: protein routing and microbial contributions

Overall, our results have implications for protein routing and the interpretation of isotopic data in ecology and more broadly for the contribution of gut microbiota to the budget of indispensable amino acids in animals. Our results show that routing of dietary protein into consumer proteinaceous tissues occurs in a rapidly growing omnivorous ectotherm. When coupled with recently published data for bulk muscle tissue from the same fish (Kelly & Martinez del Rio 2010), the amino acid-specific patterns reported here show that ‘scrambled egg’ assumption made by stable isotope mixing models is not valid for omnivores, such as tilapia, that consume moderate amounts of protein (≥10%). At lower protein concentrations, the isotopic composition of tilapia resembled that of diet. This resemblance at the level of bulk tissue, however, was the result of complex underlying patterns at the level of individual amino acids. The finding of isotopic routing has significant implications for any applied study that uses isotopic data to characterize omnivore diets, especially for animals that often experience seasonal or spatial changes in protein availability. In agreement with available evidence from experiments focused on bulk tissue analysis (Podlesak & McWilliams 2006; Voight et al. 2008; Kelly & Martinez del Rio 2010), our results suggest that for animals fed diets with moderate to high amounts of protein, mixing models can overestimate the contribution of dietary protein sources to diet. At moderate to high protein diet levels, the proteinaceous tissues (e.g. muscle, feather, blood) routinely analysed in ecology tend to preferentially reflect the isotopic composition of dietary protein. Martínez del Rio & Wolf (2005) constructed models that assume perfect routing in which for animals that consume moderate to high amounts of protein, the isotopic composition of proteinaceous tissues exclusively reflects that of dietary protein. Our analysis suggests that these models may be incorrect. Even at high protein levels, the biosynthesis of dispensable amino acids might add carbon from non-protein dietary sources (carbohydrates and lipids) to the proteinaceous tissues biosynthesized by the animal. Ultimately, future experiments that utilize compound-specific stable isotope analysis (CSIA) could lead to a new generation of mixing models for omnivores that include more mechanistic detail and hence lead to more accurate diet reconstructions.

Tilapia appeared to be able to acquire their indispensable amino acid requirements directly from gastrointestinal microbiota when dietary sources were insufficient. This observation highlights yet another critically important process mediated by the symbiosis between monogastric omnivorous animals and their gut microbiota. Our approach also shows that CSIA of individual amino and/or fatty acids is a powerful proxy (Evershed et al. 2007) for exploring questions that lie at the intersection between microbiology and animal nutrition.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Tilapia feeding experiments were supported by the National Science Foundation (IBN-0114016). Robert Carroll assisted in fish care, while J. Bobbit and S. Devries helped us build tanks and maintained constant water temperatures in the face of a Wyoming winter. This manuscript benefited from the comments of A.C. Jakle. S.D.N. was partially funded by the National Science Foundation (ATM-0502491), Carnegie Institution of Washington and the W.M. Keck Foundation (072000).

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  • Ambrose, S.H. & Norr, L. (1993) Experimental evidence for the relationship of the carbon isotope ratios of whole diet and dietary protein to those of bone collagen and carbonate. Prehistoric Human Bone: Archaeology at the Molecular Leve (eds J.B. Lambert & G. Grupe), pp. 137. Springer-Verlag, Berlin.
  • Ball, R., Atkinson, J. & Bayley, H. (1986) Proline as an essential amino acid for the young pig. British Journal of Nutrition, 55, 659668.
  • Blair, N., Leu, A., Muñoz, E., Olsen, J., Kwong, E. & Des Marais, D. (1985) Carbon isotopic fractionation in heterotrophic microbial metabolism. Applied and Environmental Microbiology, 50, 9961001.
  • Bowen, S.H. (1980) Detrital nonprotein amino acids are the key to rapid growth of Tilapia in Lake Valencia, Venezuela. Science, 207, 12161218.
  • Burnham, K.P. & Anderson, D.R. (2002) Model Selection and Multimodel Inference: A Practical Information-Theoretic Approach, 2nd edn. Springer-Verlag, New York.
  • Caldovic, L. & Tuchman, M. (2003) N-acetylglutamate and its changing role through evolution. Biochemical Journal, 372, 279290.
  • Chung, T.K. & Baker, D.H. (1993) A note on the dispensability of proline for weanling pigs. Animal Production, 56, 407408.
  • Dabrowski, K., Terjsen, B.F., Zhang, Y., Phang, J.M. & Lee, K.J. (2005) A concept of dietary dipeptides: a step to resolve the problem of amino acid availability in the early life of vertebrates. Journal of Experimental Biology, 208, 28852894.
  • DeNiro, M.J. & Epstein, S. (1977) Mechanism of carbon isotope fractionation associated with lipid synthesis. Science, 197, 261263.
  • D’Mello, J.P.F. (2003) Amino Acids in Animal Nutrition. CABI Publishing, London.
  • Eastoe, J.E. (1957) The amino acid composition of fish collagen and gelatin. Biochemical Journal, 65, 363368.
  • Evershed, R.P., Bull, I.D., Corr, L.T., Crossman, Z.M., van Dongen, B.E., Evans, C.J., Jim, S., Mottram, H.R., Mukherjee, A.J. & Pancost, R.D. (2007) Compound specific stable isotope analysis in ecology and paleoecology. Stable Isotopes in Ecology and Environmental Science (eds R. Michener & K. Lajtha), pp. 480539. Blackwell Publishing, Boston.
  • Fantle, M.S., Dittel, A.I., Schwam, S.L., Epifanio, C.E. & Fogel, M.L. (1999) A food web analysis of the juvenile blue crab, Callinectes sapidus, using stable isotopes in whole animals and individual amino acids. Oecologia, 120, 416426.
  • Fuller, M.F. & Reed, P.J. (1998) Nitrogen cycling in the gut. Annual Review of Nutrition, 18, 385411.
  • Gill, P.E. & Murray, W. (1978) Algorithms for the solution of the nonlinear least-squares problem. SIAM Journal on Numerical Analysis, 15, 977992.
  • Hare, P.E., Fogel, M.L., Stafford, T.W., Mitchell, A.D. & Hoering, T.C. (1991) The isotopic composition of carbon and nitrogen in individual amino acids isolated from modern and fossil proteins. Journal of Archaeological Science, 18, 277292.
  • Horton, R., Moran, L.A., Scrimgeour, G., Perry, M. & Rawn, D. (2002) Principles of Biochemistry. Prentice Hall, New York.
  • Howland, M.R., Corr, L.T., Young, S.M.M., Jones, V., Jim, S., Van Der Merwe, N.J., Mitchell, A.D. & Evershed, R.P. (2003) Expression of the dietary isotope signal in the compound-specific δ13C values of pig bone lipids and amino acids.
  • Jim, S., Jones, V., Ambrose, S.H. & Evershed, R.P. (2006) Quantifying dietary macronutrient sources of carbon for bone collagen biosynthesis using natural abundance stable carbon isotope analysis. British Journal of Nutrition, 95, 10551062.
  • Karasov, W.H. & Carey, H.V. (2008) Metabolic teamwork between gut microbes and hosts. Microbe, 4, 323328.
  • Kelly, L.J. & Martinez del Rio, C. (2010) The fate of carbon in growing fish: an experimental study of isotopic routing. Physiological and Biochemical Zoology, 83, 473480.
  • Kirchgessner, v.M., Fickler, J. & Roth, F. (1995) Effect of dietary proline supply on N-balance of piglets. Journal of Animal Physiology and Animal Nutrition, 73, 5765.
  • Koch, P.L. (2007) Isotopic study of the biology of modern and fossil vertebrates. Stable Isotopes in Ecology and Environmental Science (eds R. Michener & K. Lajtha), pp. 99154. Blackwell Publishing, Boston.
  • Kung, L. & Rode, L.M. (1996) Amino acid metabolism in ruminants. Animal Feed Science and Technology, 59, 167172.
  • Larsen, T., Taylor, D.L., Leigh, M.B. & O’Brien, M. (2009) Stable isotope fingerprinting: a novel method for identifying plant, fungal, or bacterial origins of amino acids. Ecology, 90, 35263535.
  • Lim, C. & Webster, C.D. (2006) Tilapia: Biology, Culture, and Nutrition. Haworth Press, New York.
  • Macko, S.A. & Estep, M.L. (1985) Microbial alteration of stable nitrogen and carbon isotope compositions of organic matter. Organic Geochemistry, 6, 787790.
  • Macko, S., Fogel, M.L., Hare, P. & Hoering, T. (1987) Isotopic fractionation of nitrogen and carbon in the synthesis of amino acids by microorganisms. Chemical Geology, 65, 7992.
  • Martínez del Rio, C. & Wolf, B. (2005) Mass-balance models for animal isotopic ecology. Physiological and Ecological Adaptations to Feeding in Vertebrates (eds J.M. Starck & T. Wang), pp. 141174. Science Publishers, Enfield, NH.
  • Martínez del Rio, C., Wolf, N., Carleton, S. & Gannes, L.Z. (2009) Isotopic ecology ten years after a call for more laboratory experiments. Biological Review, 84, 91111.
  • McMahon, K.W., Fogel, M.L., Elsdon, T. & Thorrold, S.R. (2010) Carbon isotope fractionation of amino acids in a marine fish. Journal of Animal Ecology, 79, 11321141.
  • Melzer, E. & Schmidt, H.L. (1987) Carbon isotope effects on the pyruvate dehydrogenase reaction and their importance for relative carbon-13 depletion in lipids. Journal of Biological Chemistry, 262, 81598164.
  • Metges, C.C. (2000) Contribution of microbial amino acids to amino acid homeostasis of the host. Journal of Nutrition, 130, 1857S1864S.
  • O’Brien, D.M., Fogel, M.L. & Boggs, C.L. (2002) Renewable and nonrenewable resources: amino acid turnover and allocation to reproduction in Lepidoptera. Proceedings of the National Academy of Sciences of the United States of America, 99, 44134418.
  • Ozyurt, G. & Polat, A. (2006) Amino acid and fatty acid composition of wild sea bass (Dicentrarchus labrax): a seasonal differentiation. European Food Research and Technology, 222, 316320.
  • Podlesak, D.W. & McWilliams, S.R. (2006) Metabolic routing of dietary nutrients in birds: effect of diet quality and macronutrient composition revealed using stable isotopes. Physiological and Biochemical Zoology, 79, 534549.
  • Santiago, C.B. & Lovell, R.T. (1988) Amino acid requirements for growth of Nile tilapia. Journal of Nutrition, 118, 15401546.
  • Schwarcz, H.P. & Schoeninger, M.J. (1991) Stable isotope analysis in human nutritional ecology. Yearbook of Physical Anthropology, 34, 283321.
  • Tuross, N., Fogel, M.L. & Hare, P.E. (1988) Variability in the preservation of the stable carbon and nitrogen isotope ratios of collagen in fossil bone. Geochimica et Cosmochimica Acta, 52, 929935.
  • Van der Merwe, N.J. (1982) Carbon isotopes, photosynthesis, and archaeology. American Scientist, 70, 596606.
  • Voight, C.C., Rex, K., Michener, R.H. & Speakman, J.R. (2008) Nutrient routing in omnivorous animals tracked by stable carbon isotopes in tissue and exhaled breath. Oecologia, 157, 3140.
  • Walsh, D.A. & Sallach, H.J. (1966) Comparative studies on pathways for serine biosynthesis in animal tissues. The Journal of Biological Chemistry, 241, 4068.