Oecologia

, Volume 162, Issue 3, pp 571–579

The impact of protein quality on stable nitrogen isotope ratio discrimination and assimilated diet estimation

Authors

    • Department of Natural Resource Sciences and School of Biological SciencesWashington State University
  • Laura A. Felicetti
    • Department of Natural Resource Sciences and School of Biological SciencesWashington State University
  • Scott T. Florin
    • School of Biological SciencesWashington State University
Physiological ecology - Original Paper

DOI: 10.1007/s00442-009-1485-8

Cite this article as:
Robbins, C.T., Felicetti, L.A. & Florin, S.T. Oecologia (2010) 162: 571. doi:10.1007/s00442-009-1485-8

Abstract

Accurately predicting isotopic discrimination is central to estimating assimilated diets of wild animals when using stable isotopes. Current mixing models assume that the stable N isotope ratio (δ15N) discrimination (∆15N) for each food in a mixed diet is constant and independent of other foods being consumed. Thus, the discrimination value for the mixed diet is the combined, weighted average for each food when consumed as the sole diet. However, if protein quality is a major determinant of ∆15N, discrimination values for mixed diets may be higher or lower than the weighted average and will reflect the protein quality of the entire diet and not that of the individual foods. This potential difference occurs because the protein quality of a mixed diet depends on whether, and to what extent, the profiles and amounts of essential amino acids in the individual foods are complementary or non-complementary to each other in meeting the animal’s requirement. We tested these ideas by determining the ∆15N of several common foods (corn, wheat, alfalfa, soybean, and fish meal) with known amino acid profiles when fed singly and in combination to laboratory rats. Discrimination values for the mixed diets often differed from the weighted averages for the individual foods and depended on the degree of complementation. ∆15N for mixed diets ranged from 1.1‰ lower than the weighted average for foods with complementary amino acid profiles to 0.4‰ higher for foods with non-complementary amino acid profiles. These differences led to underestimates as high as 44% and overestimates as high as 36% of the relative proportions of fish meal and soybean meal N, respectively, in the assimilated mixed diets. We conclude that using isotopes to estimate assimilated diets is more complex than often appreciated and will require developing more biologically based, time-sensitive models.

Keywords

Assimilated dietIsotope discriminationNitrogenProtein qualityStable isotopes

Introduction

Stable isotopes have become an important tool in studies of diet and trophic interactions (Thompson et al. 2005; Crawford et al. 2008). The opportunity to understand the sources and relative proportions of assimilated C, N, or S from small tissue samples without quantifying the myriad of foraging, digestive, metabolic, and productive processes is extraordinarily appealing to both basic and applied ecologists. However, in moving rapidly to field application, many assumptions critical to using the technique have not been tested or understood (Martinez del Rio et al. 2009).

For example, an essential step in many of these studies is assigning a discrimination value to each major dietary component to create mixing model polygons (Ben-David and Schell 2001; Phillips 2001; Phillips and Gregg 2001; Caut et al. 2008a, b). Thus far, current mixing models treat the assigned discrimination value for each food as a constant that is frequently determined with captive animals (e.g., Hobson and Clark 1992a; Hilderbrand et al. 1996; Felicetti et al. 2003; Robbins et al. 2005; Caut et al. 2008a). Such discrimination values vary with some characteristic of the food (e.g., protein quality or quantity) or animal (e.g., specific tissue, level of intake, growth rate, or metabolic rate; Fantle et al. 1999; Roth and Hobson 2000; Gaye-Siessegger et al. 2004; Robbins et al. 2005; Miron et al. 2006; Gaye-Siessegger et al. 2007; Caut et al. 2008a; Martinez del Rio et al. 2009).

Regressions between diet and consumer isotope values produced using captive animals (e.g., Fig. 1) are frequently used to estimate isotope values for foods that can not be measured directly (DeNiro and Epstein 1981; Hilderbrand et al. 1996; Felicetti et al. 2003; Caut et al. 2008b). However, the regression coefficients, which are frequently very high, are misleading as these are autocorrelations in that the x-axis is diet and the y-axis is diet plus discrimination. When diet and discrimination are plotted directly, the regression coefficient is lower (Fig. 1) and suggests the potential for a larger error in estimating unknown discrimination values.
https://static-content.springer.com/image/art%3A10.1007%2Fs00442-009-1485-8/MediaObjects/442_2009_1485_Fig1_HTML.gif
Fig. 1

The relationships between the N isotope value [stable N isotope ratio (δ15N)] of the diet and a δ15N of the animal or b δ15N discrimination (∆15N) between the diet and animal for whole blood, plasma or serum of mammals and birds. The values are for single item diets or mixed diets in which protein was supplied by a single ingredient (Steele and Daniel 1978; Hobson and Clark 1992a; Hobson and Welch 1992; Hobson et al. 1996; Hilderbrand et al. 1996; Jenkins et al. 2001; Ben-David and Schell 2001; Haramis et al. 2001; Bearhop et al. 2002; Lesage et al. 2002; Pearson et al. 2003; Felicetti et al. 2003; Sponheimer et al. 2003; Cherel et al. 2005; Arneson and MacAvoy 2005; Darr and Hewitt 2008; this study)

Because none of the mixing models incorporate the consequences of dietary interactions between foods, the discrimination value for mixed diets becomes the weighted average for each dietary component (Ben-David and Schell 2001; Phillips 2001; Phillips and Gregg 2001; Phillips and Koch 2002; Caut et al. 2008b). If this assumption regarding metabolic independence of the foods in mixed diets is correct, variation along the regression between dietary stable N isotope ratio (δ15N) and discrimination (∆15N) for mixed diets (Fig. 2) should be similar or less than that for single item diets (Fig. 1). Unfortunately, although the regressions for single item and mixed diet discriminations are significant and virtually identical (analysis of covariance, no difference between slopes and intercepts, F = 1.99, P = 0.29; SAS 1998), variation between the dietary isotope value and discrimination is more than twice as large for mixed diets (Fig. 2) than for single item diets (Fig. 1; test for equality of two variances, F = 1.85, P = 0.05; Zar 1996: p 139). This may indicate that current mixing models used to predict assimilated diets and the assumptions underlying them are not valid and that the error in diet estimates is larger than may be acceptable.
https://static-content.springer.com/image/art%3A10.1007%2Fs00442-009-1485-8/MediaObjects/442_2009_1485_Fig2_HTML.gif
Fig. 2

The relationships between the δ15N of the diet and a δ15N of the animal or b15N between the diet and animal for whole blood, plasma or serum of mammals and birds. The values are for when mixed diets with multiple protein sources were fed (Hobson and Clark 1992b; Hilderbrand et al. 1996; Roth and Hobson 2000; Ben-David and Schell 2001; Jenkins et al. 2001; Haramis et al. 2001; Felicetti et al. 2003; Pearson et al. 2003; Ogden et al. 2004; Cherel et al. 2005; Arneson and MacAvoy 2005; Miron et al. 2006; Darr and Hewitt 2008; Tsahar et al. 2008; this study). Dashed lines are the regressions from Fig. 1. For abbreviations, see Fig. 1

Particularly relevant to N isotope discrimination is the recognition that amino acid profiles in specific foods can be inadequate relative to animal metabolism and that animals can consume foods with complementary amino acid profiles (Murphy and Pearcy 1993). Complementation would produce higher quality dietary protein and therefore potentially lower N isotope discrimination for mixed diets than would be predicted from the weighted average of each dietary component (Robbins et al. 2005). While the endpoints or single item discrimination values for each major dietary component could be correct in defining the mixing model polygon, the mixed diet estimates would be incorrect if discrimination values are not additive and linear.

The increased variation in discrimination when animals consume mixed diets relative to single item diets suggests at least three potential problems: analytical errors, such as obtaining representative samples and isotope values for mixed diets due to the extremely small sample sizes (<1 mg) used by modern isotope ratio mass spectrometers; lack of cause-effect between the diet’s isotope value and discrimination; and non-independence in the metabolism of the individual foods in mixed diets. In this study, we assess the assumption about the independence of metabolism and therefore the additive nature of discrimination values in going from single item to mixed diets. Certainly, a vast literature has accumulated over decades about the metabolic benefits of mixing foods with differing nutrient profiles that might relate to isotope discrimination (e.g., Murphy and Pearcy 1993; Klasing 1998; Simpson et al. 2004; Robbins et al. 2007).

The specific hypotheses that we tested were:

  1. 1.

    Protein quality is the primary dietary determinant of N discrimination as compared to C:N ratios or N content.

     
  2. 2.

    Foods with complementary amino acid profiles that are consumed in mixed diets will produce diets of higher protein quality and, therefore, lower discrimination values than the weighted average of the individual foods.

     
  3. 3.

    Foods with non-complementary amino acid profiles that are consumed in mixed diets will produce diets of either equal or lower protein quality and, therefore, equal or higher discrimination values than the weighted average of the individual foods.

     
  4. 4.

    Discrimination values assigned to foods in mixed diets are not constants and will require some understanding of amino acid profiles, animal metabolism, and temporal foraging patterns to determine appropriate N discrimination values (Murphy and Pearcy 1993; Robbins et al. 2005; Caut et al. 2008a).

     

Hypotheses 1–3, if true, would produce the increased variation in N discrimination when mixed diets are fed relative to single item diets.

Materials and methods

Feeding trials: animals

Fourteen male, Sprague–Dawley laboratory rats were used in all feeding trials. Each feeding trial lasted 21 days to ensure that plasma had equilibrated with the diet (Hobson and Clark 1992b; Hilderbrand et al. 1996). Blood samples were collected in heparinized tubes at the end of each feeding trial. Plasma was separated, frozen, and freeze-dried. All rats were fed ad libitum to promote positive energy and protein balance, weight gain, and therefore minimal tissue mobilization. Rats were weighed weekly.

Feeding trials: diets

To test the various hypotheses, several foods with known amino acid composition and large differences in protein quantity and quality were sought (Table 1; Fig. 3). These foods were to be fed both singly and in combination to determine how discrimination varied. All feeds were purchased as single batches and the diets were mixed prior to the study to ensure that composition and isotopic values were constant.
Table 1

Dietary composition (100% dry matter basis), isotope values (‰), and discrimination (‰) by laboratory rats consuming basal diets used to determine the effect of protein quality on stable N isotope ratio (δ15N) discrimination

Item

Diets

Corn

Wheat

Alfalfa

Wheat + alfalfa

Soybean

Fish

Corn meal

96.8

     

Wheat meal

 

96.8

 

58.8

  

Alfalfa

  

99.0

38.9

  

Soybean meal

    

96.8

 

Fish meal

     

96.8

Limestone

2.2

2.2

1.3

2.2

2.2

Trace mineral salt with selenium

0.7

0.7

0.7

0.7

0.7

0.7

Vitamin supplement

0.3

0.3

0.3

0.3

0.3

0.3

N

1.4

2.0

3.0

2.4

6.9

10.6

C

47.4

43.4

44.1

43.7

45.8

41.4

Dietary isotope values

 N

0.3

1.7

0.3

1.0

1.2

13.6

 C

−11.5

−25.2

−25.8

−25.4

−24.6

−18.8

Animal isotope values

 N

4.7 ± 0.1

6.8 ± 0.2

6.4 ± 0.2

6.9 ± 0.1

16.9 ± 0.1

 C

−12.8 ± 0.2

−22.0 ± 0.1

−23.0 ± 0.1

−22.6 ± 0.1

−16.6 ± 0.2

Discrimination

 N

4.4 ± 0.1

5.1 ± 0.2

5.4 ± 0.2

5.7 ± 0.1

3.3 ± 0.1

 C

−1.3 ± 0.2

3.2 ± 0.1

2.4 ± 0.1

2.0 ± 0.1

2.2 ± 0.2

https://static-content.springer.com/image/art%3A10.1007%2Fs00442-009-1485-8/MediaObjects/442_2009_1485_Fig3_HTML.gif
Fig. 3

The relative amino acid requirements of laboratory rats in comparison to the proportions provided by corn, wheat, alfalfa, soybean meal, and fish meal [National Research Council (NRC) 1994, 1995]. Although all amino acids vary, lysine, methionine and cystine are the most limiting amino acids in these foods. Note that although none of the foods have the optimum amounts of methionine and cystine for maintenance and growth when expressed as a percent of total dietary protein, all of the foods have enough of these two amino acids for maintenance and growth when expressed as a percent of the diet

The proteins of cereal grains and forages are known to be of relatively modest quality because of amino acid deficiencies or imbalances. For example, corn and wheat, while containing dietary protein above the recommended level for maintenance (5%), have levels which are below that recommended for maximum growth (15%) and are low in their relative amounts of the amino acids lysine, methionine and cystine (Fig. 3). Alfalfa, while containing dietary protein above the level recommended for maximum growth, has low relative amounts of lysine, methionine and cystine (Mitchell 1924; NRC 1994, NRC 1995; Klasing 1998).

Fish meal and soybean meal are very high in total protein content and commonly used as protein supplements (Elangovan and Shim 2000). Fish meal is a high quality protein that is rich in essential amino acids, including lysine, and higher in methionine and cystine than many plant proteins (Fig. 3; Robbins et al. 2005). Thus, the amino acids of fish meal are complementary to those in corn, wheat, and alfalfa when consumed in mixed diets. If protein quality is the major determinant of N discrimination, mixed diets of fish meal and corn, wheat, or alfalfa should have N discrimination values that are lower than what would be predicted from the weighted average of each dietary component.

Soybean meal, while relatively high in lysine, is relatively low in methionine and cystine and therefore of lower protein quality than fish meal (Fig. 3). Although soybean meal and corn or wheat are often referred to as having complementary amino acid profiles because the grains are higher in methionine and cystine than soybean meal while soybean meal is higher in lysine than the grains (Klasing 1998), neither corn, wheat nor soybean meal have the minimum recommended relative levels of methionine and cystine (Fig. 3). While both soybean meal and fish meal have ample methionine and cystine when expressed as a percent of the diet because of their very high protein content, only fish meal has sufficient amounts when combined with corn, wheat, and alfalfa to produce a diet with relatively high protein quality. Thus, mixtures of soybean meal and corn, wheat or alfalfa should not be as complementary as when the latter are fed with fish meal. Therefore, if protein quality is the major determinant of N discrimination, mixed diets of soybean meal and corn, wheat or alfalfa should produce N discrimination values that are either equal to or greater than the weighted average of each dietary component (Mitchell 1924; Lewis and Morris 1983; NRC 1994, 1995; Klasing 1998; Robbins et al. 2005).

Because grains and fish meal can be deficient in several minerals or vitamins (e.g., Ca, thiamin, vitamin E, or biotin), mineral and vitamin supplements were added to all diets to minimize potential deficiencies. Limestone was not added to alfalfa because of its high Ca content (Robbins 1993). Similarly, to reduce the chance of any deficiency developing long term and to reduce variation in the N isotope signatures of rats at the start of each 21-day trial, all rats were fed laboratory rat chow (Harlan 8640; Madison, Wis.) for 1 week prior to each feeding trial. The top four ingredients in the rat chow were soybean meal, corn, wheat and fish meal, four of the five foods used in the current study. All diets were finely ground to minimize selection, which was never observed.

Feeding trials: protocol

Corn, wheat, soybean meal, and Menhaden (Brevoortia tyrannus) fish meal were fed both singly (Table 1) and in combination (Tables 2, 3). Because the rats could not maintain their weight when fed only alfalfa, alfalfa was combined with wheat with the expectation of producing a diet of relatively modest protein quality and therefore high discrimination value. Corn, wheat, and the alfalfa + wheat combination were fed with two levels of fish meal (Table 2) and one level of soybean meal (Table 3). To determine the appropriate ratio of each food in the mixed diets to provide either 25 or 50% of the N from each ingredient, 5-day, total collection digestion trials were run with five rats consuming each of the five basal diets (Table 1). Rats were housed during the digestion trails in suspended, metal-floored, metabolism cages. True digestibilities of N were estimated as the slope of the regression between dietary protein content and digestible amount (g/100 g diet) and averaged 92.8 ± 4.4% for the five diets. Dietary proteins that are not heat damaged, bound to tannins, or keratinized are highly digestible (Robbins 1993; Robbins et al. 1987).
Table 2

Dietary composition (100% dry matter basis), isotope values (‰), and discrimination (‰) by laboratory rats consuming mixed fish meal diets used to evaluate the effect of protein quality on δ15N discrimination

Item

Diets

Corn:fish

Wheat:fish

Wheat + alfalfa:fish

Dry matter source

88.3:11.7

95.8:4.2

84.0:16.0

94.0:6.0

81.4:18.6

92.9:7.1

Protein source

50:50

75:25

50:50

75:25

50:50

75:25

N

2.5

1.8

3.4

2.5

3.9

3.0

C

46.7

47.1

43.1

43.3

43.3

43.5

Dietary isotope values

 N

6.9

3.6

7.6

4.7

7.3

4.1

 C

−12.3

−11.8

−24.2

−24.8

−24.2

−25.0

Animal isotope values

 N

9.7 ± 0.2

7.4 ± 0.2

10.7 ± 0.2

8.3 ± 0.2

10.9 ± 0.2

8.4 ± 0.1

 C

−13.7 ± 0.1

−12.9 ± 0.3

−20.6 ± 0.1

−21.5 ± 0.1

−20.9 ± 0.2

−22.1 ± 0.1

Discrimination

 N

2.8 ± 0.2

3.8 ± 0.2

3.1 ± 0.2

3.6 ± 0.2

3.6 ± 0.2

4.3 ± 0.1

 C

−1.4 ± 0.1

−1.1 ± 0.3

3.6 ± 0.1

3.3 ± 0.1

3.3 ± 0.2

2.9 ± 0.1

The diets were produced by blending the corn, wheat, and wheat + alfalfa diets of Table 1 with fish meal in the proportions listed under Dry matter source. This produced diets in which fish meal provided either 25 or 50% of the protein

Table 3

Dietary composition (100% dry matter basis), isotope values (‰), and discrimination (‰) by laboratory rats consuming mixed soybean meal diets used to evaluate the effect of protein quality on δ15N discrimination

Item

Diets

Corn:soybean

Wheat:soybean

Wheat + alfalfa:soybean

Dry matter source

83.1:16.9

77.6:22.4

74.1:25.9

N

2.3

3.1

3.6

C

47.1

43.9

44.2

Dietary isotope values

 N

0.8

1.5

1.1

 C

−13.6

−25.1

−25.2

Animal isotope values

 N

6.2 ± 0.2

6.8 ± 0.2

6.6 ± 0.1

 C

−16.2 ± 0.4

−22.7 ± 0.2

−23.0 ± 0.2

Discrimination

 N

5.4 ± 0.2

5.3 ± 0.2

5.5 ± 0.1

 C

−2.6 ± 0.4

2.4 ± 0.2

2.2 ± 0.2

The diets were produced by blending the corn, wheat, and wheat + alfalfa diets in Table 1 with soybean meal in the proportions listed under Dry matter source to produce diets in which soybean meal provided 50% of the protein

Nutritional, isotopic, and statistical analyses

N content of all diets was determined in duplicate using a C–N TruSpec analyzer (LECO; St. Joseph, Mich.). δ13C and δ15N values of diets and freeze-dried plasma were determined with a continuous-flow isotope ratio mass spectrometer (Delta PlusXP; Thermo Finnigan). Mean δ13C and δ15N dietary values are based on the analyses of five samples per diet. Results are reported as per mil (‰) relative to National Bureau of Standards (NBS) 19 (TS-limestone; δ13C) and atmospheric N (δ15N). Laboratory reference standards were interspersed throughout each analysis to ensure maintenance of calibration. Analytical error was 0.1‰ for both isotopes.

A two-source linear mixing model (Eqs. 6 and 7; Martinez del Rio et al. 2009) was used to estimate the assimilated dietary proportions of N for each mixed diet from the tissue isotope values for rats consuming each of the two sources and their mixture. Those proportions estimated from the N isotope values were then compared to the known assimilated dietary proportions (i.e., 50:50 or 75:25) to estimate error. The C values were not used because: (1) this study focused purely on understanding N discrimination; and (2) soybean meal, wheat, alfalfa, and the mixed diets created from them had virtually identical C values characteristic of C3 plants (Tables 1, 3) and, therefore, could not have been used.

Linear least squares regressions were used to model all relationships (SAS 1998). Differences in slopes of regressions were tested using small sample t-tests (Kleinbaum and Kupper 1978). Paired t-tests were used to test for differences in discrimination between diets. A P-value of ≤ 0.05 was considered significant. Means are reported ±1 SD.

Results

Laboratory rats either maintained or gained mass during each 21-day feeding study. Rates of gain (g/day) for the single source protein diets (Table 1) were 0.8 ± 0.7 for corn, 1.0 ± 0.5 for wheat, 3.3 ± 0.6 for soybean meal, and 3.5 ± 0.9 for fish meal. Rates of gain (g/day) were 1.8 ± 0.9 for the three (two fish and one soybean meal) mixed corn diets, 1.4 ± 0.8 for the three mixed wheat diets, 0.1 ± 0.7 for the four mixed wheat + alfalfa diets (Tables 1, 2, 3).

For the five basal diets fed to laboratory rats (Table 1), ∆15N was lowest when consuming high protein fish meal (3.3 ± 0.1‰) and highest when consuming the various plant-based diets, including high protein soybean meal (5.7 ± 0.1‰). Thus, neither dietary N content nor the C:N ratio were correlated with ∆15N. This occurred for both the entire data set (Fig. 4) and when each of the lower N basal diets was supplemented with either fish meal, which decreased ∆15N, or soybean meal, which either maintained or increased ∆15N (Fig. 5).
https://static-content.springer.com/image/art%3A10.1007%2Fs00442-009-1485-8/MediaObjects/442_2009_1485_Fig4_HTML.gif
Fig. 4

The relationships between ∆15N and the C:N ratio and N content of diets consumed by laboratory rats in the current study

https://static-content.springer.com/image/art%3A10.1007%2Fs00442-009-1485-8/MediaObjects/442_2009_1485_Fig5_HTML.gif
Fig. 5

Observed and expected ∆15N when laboratory rats consumed foods with either complementary (fish meal and corn, wheat, or alfalfa) or non-complementary (soybean meal and corn, wheat, or alfalfa) amino acid profiles. The dashed lines that connect the 0 and 100% experimental values are where the expected discrimination values should be for the mixed diets if they were the weighted average of the two individual foods. Significance was determined by creating 95% confidence intervals (CI) for the endpoints using the largest observed SD amongst all diets for each graph. In addition, the interpolated values at 25 and 50% fish meal and 50% soybean meal were determined. Ninety-five percent CIs for those points were assigned using the same maximum SD for each graph. Comparisons between the observed means and interpolated value on the line were made based on overlap-nonoverlap of the 95% CIs. If the CIs did not overlap, the two values were statistically different (α < 0.05). All discriminations for 25 and 50% fish meal were significantly different from expected, whereas only the 50% corn:soybean meal was significant

As hypothesized, the addition of fish meal to the corn, wheat, and wheat + alfalfa diets at either 25 or 50% of total dietary N produced lower ∆15N than predicted as weighted averages (Table 2; Fig. 5). Across all mixed fish meal diets, the reduction in the expected ∆15N averaged −0.8 ± 0.3‰ (range, −0.3 to −1.1‰), which produced a mean ∆15N of 3.5 ± 0.5‰. However, discrimination decreased in all three cases when each basal diet was supplemented with fish meal at the 50% N level (3.2 ± 0.5‰) relative to the 25% level (4.0 ± 0.4‰; Table 2; all ts ≥ 5.4, Ps < 0.001). The lower than expected ∆15N for all mixed fish meal diets produced a 22 ± 12% (range, 12–44%) underestimate of the relative proportion of fish meal N in the assimilated diets.

In contrast to the lowered ∆15N when fish meal was added to the plant-based diets, addition of soybean meal to the corn, wheat, and wheat + alfalfa diets produced either equal or higher ∆15N values than predicted (Table 3; Fig. 5). Across all mixed soybean meal diets, the difference between the observed and expected discrimination averaged 0.1 ± 0.3‰ (range, −0.1 to 0.4‰), which produced a mean ∆15N of 5.3 ± 0.2‰. The difference between the observed and expected ∆15N for the mixed soybean meal diets decreased as the ∆15N and, presumably, protein quality of the individual components became more similar (Fig. 5). These similarities or differences produced overestimates of the relative proportion of soybean meal N in the mixed diets that ranged from zero for the wheat:soybean meal and wheat + alfalfa:soybean meal diets to 36% for the corn:soybean meal diet.

Discussion

15N for the five basal diets fed to laboratory rats during the current study are similar to those measured by other investigators. For example, the ∆15N for fish meal in the current study (3.3 ± 0.1‰) is similar to the 3.2‰ measured in dunlins (Calidris alpine pacifica; Ogden et al. 2004) and 2.9‰ in laboratory mice (Arneson and MacAvoy 2005). ∆15N for grains, soybeans, and alfalfa fed in the current study (5.3 ± 0.4‰, n = 7) are similar to those measured for forages and grains previously fed to ruminants (5.1 ± 0.8‰, n = 6; Steele and Daniel 1978; Sponheimer et al. 2003; Darr and Hewitt 2008).

15N for the mixed fish meal diets with complementary amino acid profiles were lower and for the mixed soybean meal diets with non-complementary amino acid profiles were either the same or higher than the weighted averages of the individual foods as hypothesized based on overall dietary protein quality. While it is not possible to control both protein quality and quantity when feeding individual foods and their mixtures as in the current study, the observed differences indicate that understanding ∆15N for mixed diets and ultimately predicting assimilated diets of wild animals are not elemental or dilution processes based solely on either dietary C:N ratios or N content, but are far more complex biological processes (Robbins et al. 2005). Thus, the addition of protein to basal, lower protein diets and either the increase (Pearson et al. 2003) or decrease (Tsahar et al. 2008) in ∆15N can not be used to support the view that protein quantity is a primary determinant of ∆15N because protein quality also changes (Martinez del Rio et al. 2009).

Unfortunately, the relationships between the dietary δ15N value for single-source protein diets and either the animal’s δ15N or the diet-animal ∆15N (Fig. 1) are unlikely to provide accurate estimates of the ∆15N for estimating the assimilated proportions in mixed diets. The good relationships of Fig. 1 are not cause-effect, but reflect the generally increasing protein quality and decreasing ∆15N with trophic level (Robbins et al. 2005; this study). The increased variation in these relationships for mixed diets (Fig. 2) occurs because the ∆15N of mixed diets reflects the amino acid profile and protein quality of the entire diet and not the weighted ∆15N of the individual foods (Fig. 5). Thus, the use of ∆15N determined by feeding captive animals commercial diets of either unknown protein quality or very high protein quality (e.g., laboratory rodent and poultry diets) will not be appropriate in most cases for estimating assimilated diets of wild animals.

For example, the five lowest discrimination values below the mixed diet regression of Fig. 2 are for animals fed laboratory rodent chows and commercial poultry diets that are generally formulated with fish meal or purified amino acids, particularly methionine and lysine, to maximize protein quality (Hobson and Clark 1992a; Haramis et al. 2001; Jenkins et al. 2001; Ogden et al. 2004; Arneson and MacAvoy 2005). At the other extreme, four of the six highest values above the regression of Fig. 2 are pelleted, non-commercial diets, whose discrimination values may partially reflect protein damage due to the heat and pressure of pelleting (Van Soest 1994; Hilderbrand et al. 1996; Felicetti et al. 2003)). The removal of such diets as well as those containing extracted or purified proteins as not representative of what free-ranging wild animals might consume produces a very different discrimination regression (Fig. 6). Although the regression coefficient is much higher in comparison to that for all mixed diets (Fig. 2), the slope of the regression is 50–100% greater than that for all mixed diets (Fig. 2) and diets with single source proteins (Fig. 1). Thus, although all of the discrimination regressions are statistically significant, the lack of a cause-effect biological link between the variables minimizes their predictive power.
https://static-content.springer.com/image/art%3A10.1007%2Fs00442-009-1485-8/MediaObjects/442_2009_1485_Fig6_HTML.gif
Fig. 6

The relationship between the δ15N of the diet and ∆15N between the diet and animal for whole blood, plasma or serum of mammals and birds. The values are for mixed diets with multiple protein sources that do not contain extracted or purified proteins (e.g., casein or soya) and exclude commercial diets and pelleted diets (Hobson and Clark 1992b; Hilderbrand et al. 1996; Roth and Hobson 2000; Ben-David and Schell 2001; Jenkins et al. 2001; Haramis et al. 2001; Felicetti et al. 2003; Pearson et al. 2003; Ogden et al. 2004; Cherel et al. 2005; Arneson and MacAvoy 2005; Miron et al. 2006; Darr and Hewitt 2008; this study). Dashed lines are the discrimination regressions from Figs. 1 and 2

In summary, several of the assumptions underlying the current use of isotopes to estimate assimilated diets of wild animals may not be correct. Future studies and models may need to incorporate knowledge of animal metabolism, amino acid profiles of foods relative to animal requirements, and the timing of food consumption. Discrimination values for foods in mixed diets are not constants, need not be additive and linear, and will vary with temporal aspects of food consumption and the degree of amino acid complementation determined by the entire diet. For example, discrimination values and mixing models used to estimate assimilated diets for grizzly bears consuming salmon and plant matter will differ depending on when those foods are consumed. If these are seasonal diets (e.g., plant matter in the spring and salmon in the summer and fall), discrimination values will be those for the individual foods without the effects of complementation; and current linear mixing models may be adequate to predict assimilated diet. However, if these foods are being consumed in a mixed fall diet, discrimination values will likely reflect dietary protein complementation; and a yet to be developed mixing model may be needed to predict assimilated diet.

Acknowledgments

The project was approved by the Washington State University Institutional Animal Care and Use Committee (no. 3762) and funded by the Nutritional Ecology Research Endowment. We appreciate the assistance of M. Van Daele.

Copyright information

© Springer-Verlag 2009