Comparison of methods for stable isotope ratio (δ13C, δ15N, δ2H, δ18O) measurements of feathers

Authors


Summary

  1. Natural variations in the stable isotope ratios of bioelements in bird feathers are being increasingly used by animal ecologists to investigate different aspects of bird life. However, to ensure reliability of the data, a critical and very delicate aspect is the preparatory phase (cleaning, drying and subsampling) and the proper analysis, mainly in relation to δ2H and δ18O, respectively, for the presence of exchangeable Hs and of nitrogen and sulphur in keratin.
  2. With respect to determination of the isotope ratios of C, N, O and H, in this work, we compare the cleaning mixture most commonly used in the literature (chloroform : methanol 2 : 1) with diethylether : methanol 2 : 1, which avoids the use of the carcinogenic solvent chloroform. We also compared oven-drying with air-drying of samples, as well as subsampling of feathers by cutting with surgical scissors or cryogenic pulverization. Finally, we investigated whether stable isotope ratios varied along the vane and between the rachis and vane.
  3. The different methods compared in the three preparatory stages showed no differences performance-wise and can therefore be used interchangeably. Variability in stable isotope ratios can be considerable, both along the vane and between rachis and vane, which is because their compositions register changes in diet, area and climate. However, in this specific study, when the parts of the feather closest to the calamus were removed, the delta values were clearly more homogeneous. Finally, we demonstrate that a casein with a known δ2H value, although probably differing from keratin in the number of exchangeable Hs, can be used to normalize the δ2H values of feathers, although only in the range of values close to that of the reference material. In determining δ18O, the use of a longer gas chromatography-GC column, its frequent change and the use of a linear equation built with matrix match equivalent reference materials seems to reduce the drift of GC column performance due to the presence of nitrogen and the accumulation of sulphur.

Introduction

Natural variations in the stable isotope ratios of key bioelements (C, N, H, O) have been increasingly used by avian ecologists over the past decades (Hobson & Wassenaar 1997; Marra, Hobson & Holmes 1998; Inger & Bearhop 2008). The isotope approach has allowed researchers to investigate trophic relationships in food webs (Kelly 2000; Boecklen et al. 2011), discriminate between animals living in different biomes (Bolton et al. 2008), undertake palaeodietary studies (Burnham, Burnham & Newton 2009), and trace the breeding and wintering origins of migratory animals (Cherel et al. 2006; Chang et al. 2008). The use of stable isotopes in these studies relies on two main principles. First, the isotopic composition of animal tissues can reflect diet and/or environmental water in a predictable manner, and secondly, because tissues turn over at different rates, they assimilate this information over different time-scales and, if the animal is mobile, over different spatial scales. The increasing popularity of the stable isotope method to investigate the ecology of individual species is due, in part, to the fact that conventional tools for investigating animal diets and movements are usually hampered by, for example, the low percentage of recapture of ringed birds and/or geographical bias in surveys. It is therefore of paramount importance to develop a suitable analytical method for obtaining rational isotopic values to avoid mistakes when assigning the geographical origin rather than the diet, when evaluating the data. Indeed, despite the enormous potential of stable isotope ratios, a number of issues have still to be addressed, in particular, with regard to pre-treatment of the feather and determination of δ2H and δ18O (Wassenaar & Hobson 2000a,b, 2003; Paritte & Kelly 2009; Qi, Coplen & Wassenaar 2011; Meier-Augenstein, Hobson & Wassenaar 2013).

Different methods have been used for pre-treating feather samples for isotopic analyses, particularly with respect to the key stages of cleaning, drying and subsampling. With the exception of Bensch, Bengtsson and Akesson (2006), most of the preparation methods involve cleaning the feathers to eliminate contamination of the feather keratinous material with surface oils (Wassenaar 2008), which may lead to deviations in the δ13C, δ2H and δ18O values. Various reagents have been used to clean feathers: water (Wolf & Hatch 2011), detergents (Chamberlain et al. 1997; Lott, Meehan & Heath 2003; Meehan et al. 2003), sodium hydroxide (Bearhop et al. 2002; Pain et al. 2004; Atkinson et al. 2005; Bolton et al. 2008; Ramos et al. 2009; Evans et al. 2012) and various mixtures of organic solvents, including acetone (Hilton et al. 2006; Horacek 2011), petroleum ether (Chang et al. 2008), chloroform-ether 2 : 1 (Hobson & Wassenaar 1997; Cherel, Hobson & Weimerskirch 2000; Cherel et al. 2006) and hexane-acetone (Hill et al. 2012). A review of the literature revealed that the washing mixture most frequently used (more than 30 papers from 2000 to the present) is chloroform-methanol 2 : 1, as proposed by Hobson, Atwell and Wassenaar (1999) and Wassenaar & Hobson (2000b). Recent examples include Bortolotti, Clark and Wassenaar 2013; and Reichlin et al. 2013; and Storm-Suke et al. 2012. This mixture has been used to clean other keratinous materials such as claws (Rolshausen, Hobson & Schaefer 2010), and the hair of other animal species such as bats (Popa-Lisseanu et al. 2012) and felines (Pietsch et al. 2011), while Chesson et al. (2009) used it for cleaning human, horse and sheep hair. The critical aspect of this mixture of solvents is that chloroform has been classified as a carcinogen and mutagen (e.g. by the National Institute for Occupational Safety and Health – NIOSH), so various precautions need to be taken when handling it.

After cleaning, the next stage in the pre-treatment of feathers is to remove the remaining solvent by air-drying the samples in a fume hood (Hardesty & Fraser 2010) or drying them in an oven at a temperature varying between 40 °C and 70 °C (Bearhop et al. 2002; Browne et al. 2011).

The final pre-treatment phase is the subsampling of feathers. The small sample size required for isotopic assays (typically 0·2–2 mg) could lead to problems arising from possible biological or isotopic heterogeneities, which have already been detected for δ2H within the vane and between the vane and rachis (Wassenaar & Hobson 2006). Therefore, how finely the sample is ground and its degree of homogeneity could influence the values of all four isotope ratios. The two grinding methods described in the literature are cryogenic pulverization (Bearhop et al. 2002; Wassenaar & Hobson 2003) and sectional cutting with surgical scissors (Cherel, Hobson & Weimerskirch 2000; Hilton et al. 2006; Pietsch et al. 2011; Wassenaar & Hobson 2006; Storm-Suke et al. 2012) . Cryogenic pulverization of feathers is very laborious and notoriously difficult, and most researchers prefer sectional cutting of subsamples for ease of handling, particularly if the feathers are relatively isotopically homogenous. In addition, opinions vary as to whether to include the rachis and/or the entire length of the feather in the sample. Evans et al. (2012) suggested including the rachis to obtain a more complete measure of the resources used in feather growth; on the other hand, Wassenaar and Hobson (2006) suggested the rachis should be avoided because of the observed greater isotope enrichment (2H) in the rachis than in the vane and potentially greater isotopic variation in rachis measurements.

In hydrogen stable isotope analysis of proteins in general, and keratin in this specific case, account has to be taken of the fact that a proportion of the total hydrogen present is exchangeable. This requires specific methods to account for the variability of hydrogen isotopic ratios from labile positions depending on the hydrogen isotopic ratio of ambient air humidity. Because of this, since 2000, there has been agreement among some authors to use the comparative equilibration method (Wassenaar & Hobson 2000a,b, 2003; Meier-Augenstein, Hobson & Wassenaar 2013), whereby the samples and at least two keratin references are equilibrated with air humidity under the same conditions. The reference materials must be matrix match equivalent and have a known 2H isotopic composition of the non-exchangeable H component (Meier-Augenstein, Hobson & Wassenaar 2013). On the other hand, δ18O analysis in nitrogen- and sulphur-bearing organic materials, such as keratin, is complicated by the presence of a significant N2 peak before the CO peak, which can tail in the latter, and by worsening of the gas chromatographic (GC) column performance due to sulphur binding to it. To solve this, Qi, Coplen and Wassenaar (2011) suggest using a longer GC column and a valve to divert nitrogen, in addition to a frequent change of the column.

In this work, we evaluate preparatory methods in the isotopic analysis of C, N, O and H in feathers, which would yield meaningful results comparable with the data reported in the literature. In particular, a solvent mixture (diethyl ether-methanol 2 : 1), which did not contain the problematic chloroform, was compared with the commonly used chloroform-methanol mixture for cleaning feathers. We also compared methods for drying (oven- or air-dried) and grinding (cryogenically powdered or cut with surgical scissors) the sample, and assessed within-feather isotopic variance (along the vane, and between the vane and rachis). Finally, results from the more problematic δ2H and δ18O analyses in the reference materials were critically assessed.

This is the first enquiry into all four stable isotope ratios of C, N, O and H on single samples; they have normally been examined separately in previous papers, with the exception of an article by Chang et al. (2008), who determined δ13C, δ15N, δ2H and δ18O in bird feathers, but only in the vane and without any methodological considerations.

Materials and methods

Sampling and Preparation

All the feather samples were collected during ringing activities on the Alps by the Museum of Natural Sciences - MUSE of Trento. Thirty-one feather samples (Psilopsiagon aurifrons n = 6, Apus apus n = 11, Turdus merula n = 8 and Carduelis spinus n = 6) were used to compare the cleaning procedures, necessary step to eliminate contamination of the feather keratinous material with surface oils (Wassenaar 2008), which may lead to deviations in the δ13C, δ2H and δ18O values. Each feather sample was split into two halves along the rachis, and one half cleaned with diethyl ether-methanol 2 : 1 (Mix 1), the other half with chloroform-methanol 2 : 1 (Mix 2), as described by Wassenaar and Hobson (2000b). The feathers were placed in a vial using tweezers and cleaned in solvent mixture (by sonicating for 1 min); this process was repeated for a total of three washes.

Another 15 feathers of Psilopsiagon aurifrons (n = 4), Apus apus (n = 7) and Turdus merula (n = 4) were each longitudinally split into two halves. After cleaning, one half was dried in an oven at 60 °C to constant weight while the other half was air-dried in a fume hood.

To check for potential differences between vane and rachis, the vane and the rachis of 21 feather samples (Psilopsiagon aurifrons n = 4, Apus apus n = 9, Turdus merula n = 4 and Carduelis spinus n = 4) were separated along their length (Fig. 1). The values of the vane were compared with the values of the rachis within each sample.

Figure 1.

Sampling of feather vane and rachis to investigate variability in SIRs between the vane and the rachis.

Another three feathers (Apus apus, Turdus philomelos and Carduelis spinus) were used to assess potential natural isotope variation in the vane. The vanes were sequentially sampled along the length of the feather to obtain ten subsections. The values of the ten subsections within each vane were compared.

Finally, three feathers (Apus apus, Turdus philomelos and Carduelis spinus) were longitudinally split into two, and one half was very thinly cut using surgical scissors while the other half was cryogenically powdered.

The feathers were randomly selected for each experiment. The samples came from wild birds without controlling diets, as the aim of this study was simply to investigate the natural variation of the ratios in the feather from a ‘chemical’ point of view.

In the experiment to compare the solvent mixtures, the samples were air-dried and cut with scissors. Having obtained the results of the solvent comparison, the samples in all the other tests were cleaned with the diethyl-ether:methanol mixture.

Stable Isotope Ratio Analysis

After cleaning, drying and homogenizing, the feather samples were weighed using a microbalance (CP2P, Sartorius AG, Goettingen, Germany); about 0·250 mg of each sample were placed in tin capsules for δ13C and δ15N analyses, and about 0·250 mg placed in silver capsules for δ18O and δ2H analyses (Säntis Analytical AG, Teufen, Switzerland). Each sample was weighed in two capsules, and the analysis was carried out; if the difference of the two values obtained was lower than the repeatability, then the result was considered valid and the final result was calculated as the average of the two values.

The comparative equilibration method was used for δ2H analysis, and the samples and keratin standards (CBS – Caribou Hoof Standard and KHS – Kudu Horn Standard) were left at laboratory air moisture for at least 96 h, then placed in a desiccator with P2O5 insufflated with nitrogen. Samples were then loaded onto the autosampler tray in about 3 min, then returned to the desiccator for at least 2 h, and finally put on the carousel in <1 min, sealed with a cover and purged with nitrogen. In each run, an empty capsule was included as a blank to make sure there was no infiltration of N2 through the autosampler.

A Delta Plus V Isotope Ratio Mass Spectrometer (ThermoFinnigan, Bremen, Germany) equipped with a Flash EA 1112 Elemental Analyzer (ThermoFinnigan) was used to measure δ13C and δ15N. The EA was connected to a Porapack QS 80/100 mesh GC column 3 m long (ThermoFinnigan) and the flow rate set to 120 mL He min−1.

Measurements of δ18O and δ2H were carried out with a Delta Plus XP IRMS connected with a TC/EA pyrolyzer (ThermoFinnigan). The pyrolysis reactor was maintained at 1450 °C and the gas chromatography (GC) column at 80 °C. The GC column was a molecular sieve 5Å 80 cm long (Restek Corporation, Bellefonte, PA, USA) with the flow rate set to 80 mL He min−1.

The isotope ratios were expressed in δ‰ relative to V-PDB (Vienna – Pee Dee Belemnite) for δ13C, V-SMOW (Vienna – Standard Mean Ocean Water) for δ18O and δ2H, and AIR for δ15N, according to the following formula:

display math

where Rs is the isotope ratio measured for the sample and Rstd is the isotope ratio of the international standard.

The δ13C and δ15N isotopic values were calculated using a homogenized in-house protein standard which was itself calibrated against international reference materials: l-glutamic acid USGS 40 (IAEA-International Atomic Energy Agency, Vienna, Austria), fuel oil NBS-22 (IAEA) and sugar IAEA-CH-6 for 13C/12C; l-glutamic acid USGS 40 and potassium nitrate IAEA-NO3 for 15N/14N.

The δ2H and δ18O values were calculated against CBS (Caribou Hoof Standard δ2H = −197 ± 2 ‰ and δ18O = +2·4 ± 0·1 ‰) and KHS (Kudu Horn Standard, δ2H = −54 ± 1 ‰ and δ18O =+21·2 ± 0·2 ‰) through the creation of a linear equation (Wassenaar & Hobson 2003). A commercial keratin and casein were analysed as independent controls for each batch for δ2H and δ18O determinations. All results reported for non-exchangeable H are expressed in the typical delta notation in units per mil (‰) and normalized relative to the Vienna Standard Mean Ocean Water–Standard Light Antarctic Precipitation (VSMOW–SLAP) standard scale. A sample of NBS-22, a fuel oil that doesn't exchange, was routinely included as a check of system performances.

Measurements of the reference and control materials were precise for the two analysis periods; the typical means and relative standard deviations are described in Results and Discussion section.

Measurement uncertainty, expressed as 1 standard deviation when measuring the same sample 10 times, was 0·2 ‰ for 13C/12C and 15N/14N, 0·3 ‰ for 18O/16O and 2 ‰ for 2H/1H.

Statistical Analysis

Normality of distribution was checked using the Kolmogorov–Smirnov and Shapiro–Wilk tests. Homoscedasticity was monitored with Levene test; skewness and kurtosis were checked using Pearson's indices. A repeated measures anova and an honestly significant difference (HSD) Tukey test were carried out on the data set to identify differences between the various groups. The data were analysed using Statistica v. 9 (StatSoft Italia srl, Padua, Italy).

Results and discussion

Diethyl Ether-Methanol 2 : 1 vs. Chloroform-Methanol 2 : 1

Figure 2 shows a comparison of the data obtained from washing the feathers with diethyl ether-methanol 2 : 1 (Mix1, y-axis) or with chloroform-methanol 2 : 1 (Mix2, x-axis). The ranges of variation for the four stable isotope ratios of the samples covered most of the values usually seen in bird feathers, δ13C from −24 ‰ to −19 ‰, δ15N from +2 ‰ to +14 ‰, δ2H from −120 ‰ to +20 ‰ and δ18O from +9 ‰ to +15 ‰. Some of the δ2H values were far from those of the standards, but absolute values are not so important here as we only used these data to compare two solvent mixtures. The repeated measures anova and the Tukey test showed that there were no significant differences (P > 0·05) between the two groups of feathers cleaned with the different solvents for any of the isotopic ratios. The data were then analysed using a Pearson correlation test, and highly statistical correlations (P < 0·001) between the feathers cleaned with the two different mixtures were found for all four parameters. The relationships were as follows: δ13Cmix1 = −2·3022 + 0·8959* δ13Cmix2 (R2 = 0·97), δ15Nmix1 = 0·1914 + 0·9777* δ15Nmix2 (R2 = 0·99), δ2Hmix1=−2·5776 + 0·984* δ2Hmix2 (R2 = 0·99), and δ18Omix1=−0·6102 + 1·0421* δ18Omix2 (R2 = 0·97). The residuals were randomly dispersed, so the linear regression model was deemed appropriate for the data. Furthermore, residuals were normally distributed, their variances were homogeneous, and kurtosis and asymmetry of the distributions were successfully checked. The mean differences between the two cleaning methods were 0·2 ± 0·1 ‰ for δ13C, δ15N and δ18O, and 2 ± 2 ‰ for δ2H, taking into account analytical uncertainty. Given these results, it can be concluded that the two solvent mixtures do not differ performance-wise and can therefore be considered alternatives. The option of using the diethyl ether-methanol mixture means carcinogenic chloroform can be avoided and replaced by a solvent with highly similar characteristics but without the toxicity problems.

Figure 2.

Comparison of solvent mixtures for cleaning feathers. Mix1, Diethyl ether-methanol 2 : 1; Mix2, chloroform-methanol 2 : 1.

Drying the Samples: Oven versus Fume Hood

Figure 3 shows comparison of the data obtained air-drying feathers in a fume hood (x-axis) and oven-drying at 60 °C (y-axis). In this comparison, the data again show a rather large range of values for the four isotopes (δ13C from −18 ‰ to −24 ‰, δ15N from 8 ‰ to 15 ‰, δ2H from −120 ‰ to 0 ‰ and δ18O from +9 ‰ to +15 ‰). A repeated measures anova showed there to be no significant differences between the two groups across all the parameters (P > 0·05). A Pearson test showed the two groups of data to be highly correlated for all the ratios (P < 0·001) (Fig. 3), the relationships being δ13CE = 2·1665 + 1·0945* δ13CNE (R2 = 0·99), δ15NE =−0·0625 + 1·0082* δ15NNE (R2 = 0·99), δ18OE = 1·8627 +0·8808* δ18ONE (R2 = 0·91) and δ2HE = −0·467 + 0·9989* δNE (R2 = 0·99). The residuals were randomly dispersed, confirming the appropriateness of the linear regression model for the data. Furthermore, residuals were normally distributed, their variances were homogeneous, and kurtosis and asymmetry of the distribution were successfully checked. The mean differences between the two drying methods were 0·2 ± 0·1 ‰ for δ13C and δ15N, 0·4 ± 0·2 ‰ for δ18O and 3 ± 1 ‰ for δ2H, in accordance with analytical uncertainty. In view of these results, it can be concluded that the two drying methods do not differ substantially and can be considered alternatives.

Figure 3.

Comparison of two drying methods. E, oven-dried at 60 °C; NE, air-dried in a fume hood.

Cryogenically Powdered vs. Scissor Cutting

Table 1 shows the results of determination of the four stable isotope ratios in the cryogenically powdered and homogenized feathers and those cut with surgical scissors. Differences in stable isotope ratios between the two halves of the feather fall within the ranges of repeatability. A repeated measures anova showed no significant differences between the two groups (P > 0·05) across any of the parameters. The results show that the two methods do not differ substantially and can be considered alternatives.

Table 1. Comparison of δ13C, δ15N, δ2H and δ18O values obtained from two halves of the same feathers, half cryogenically powdered (N2), half cut with surgical scissors (C). A, Apus apus; C, Carduelis spinus; T, Turdus philomelos
 δ13C (‰)δ15N (‰)δ2H (‰)δ18O (‰)
A_N2−20·411·8−1412·8
A_N2−20·312·0−1212·6
Mean−20·311·9−1312·7
SD0·10·110·1
A_C−20·712·1−1412·7
A_C−21·012·1−1313·4
Mean−20·912·1−1313·0
SD0·20·010·5
C_N2−24·78·5−859·6
C_N2−24·88·5−909·7
Mean−24·88·5−889·7
SD0·10·040·1
C_C−24·98·7−9197
C_C−24·98·8−909·6
Mean−24·98·8−919·7
SD0·00·110·1
T_N2−25·14·8−8710·4
T_N2−25·14·6−9110·6
Mean−25·14·7−8910·5
SD0·00·130·2
T_C−25·24·2−8510·5
T_C−25·24·4−909·7
Mean−25·24·3−8810·1
SD0·00·140·5

Vane–Rachis and Intra–Vane Variability

Table 2 reports the values of the rachis and the vane of each feather and the differences between them for all the four isotopes (mean differences: δ13C −0·4 ‰, δ15N 0·2 ‰, δ18O 1·1 ‰ and δ2H 6·4 ‰). A paired t-test was performed on these data to ascertain whether there were any statistically significant (P < 0·001) differences between the values from rachis and vane. The results of the paired t-test are given in Table 2, and the number of asterisks indicating the level of significance (* = P < 0·05, ** = P < 0·01 *** = P < 0·001). The rachis–vane difference was significant in at least one sample for each stable isotope, in particular, δ18O. Unlike Wassenaar and Hobson (2006), who found systematically lower δ2H values in the rachis than in the vane of feathers of different origins, we found Carduelis spinus to have higher δ2H values in the rachis than in the vane. No data have been published on the differences between the rachis and the vane for the other stable isotope values. As with δ2H, statistically significant differences between rachis and vane were sometimes found for δ13C, δ15N and δ18O, although not consistently nor in the same direction. We therefore strongly agree with Wassenaar and Hobson (2006), who suggest removing the rachis before undertaking stable isotope analyses.

Table 2. Variability in stable isotope ratios of C, N, H and O between vane and rachis
SpacesNumber of featherδ13C (‰)δ15N (‰)δ2H (‰)δ18O (‰)
VaneRachisDifference vane–rachisSign. t-testVaneRachisDifference vane–rachisSign. t-testVaneRachisDifference vane–rachisSign. t-testVaneRachisDifference vane–rachisSign. t-test
  1. Sign. t-test = significance resulting submitting the two analytical replicates of vane and rachis to a paired t-test.*P < 0.05; *P < 0.01; ***P < 0.001.

Psilopsiagon aurifrons 1−23·4−23·50·18·28·20·0−119−12129·99·10·8***
2−23·0−23·20·28·58·50·0−120−1244*10·59·80·7***
3−22·8−22·4−0·47·77·60·1−98−99111·911·50·4***
4−23·3−23·50·28·28·10·1−121−123210·19·80·3
Apus apus 1−19·7−19·3−0·4*9·99·60·3−12−19713·312·11·2**
2−20·6−19·3−1·3*11·410·01·4**−5−1611*14·112·61·5**
3−21·1−18·5−2·6*12·812·70·13−1013**14·012·71·3**
4−23·4−23·40·013·913·20·7*5−1520**13·912·61·3**
5−20·7−21·40·7*12·011·80·2−8−10212·511·60·9**
6−14·8−14·7−0·112·412·40·0−7−169*13·712·51·2**
7−23·2−23·20·013·312·80·55−16**13·412·60·8**
8−19·9−18·5−1·4*12·712·70·07−310**13·612·90·7**
9−23·1−23·10·012·811·71·1*044−4413·511·91·6**
Turdus merula 1−22·7−22·70·08·38·20·1−8799−18611·511·50·0
2−23·0−22·8−0·28·78·8−0·1−81−86512·111·50·6
3−23·0−22·8−0·28·78·40·3−80−85512·613·2−0·6
4−23·0−23·10·18·78·50·2−83−87412·111·50·6
Carduelis spinus 1−22·1−21·6−0·5*3·53·9−0·4*−102−94−8*10·811·4−0·6
2−21·8−21·80·04·14·10·0−93−92−113·312·11·2***
3−21·4−21·40·04·55·0−0·5*−83−81−213·212·21·0***
4−21·6−21·60·04·14·2−0·1−86−86013·512·21·3***

We also examined the extent to which δ13C, δ15N, δ18O and δ2H vary along the vane (Table 3). To do this, we compared the values of the 10 subsections of the vane of each of the three sample feathers and computed mean and standard deviation. Variability along the vane was either very low, the results at the level of repeatability limit, or very high (up to 0·6 ‰ for δ13C, 1·5 ‰ for δ15N, 22 ‰ for δ2H, 1·4 ‰ for δ18O), depending on the sample and the isotope. This means that, as expected, the isotopic values can vary greatly along the vane, possibly due to shifts in diet or habitat during the growth of the feather, as already observed for δ2H (Wassenaar & Hobson 2006). However, the section of each feather with the least homogeneous values was the ‘younger’ one, that is, the part closest to the quill/calamus (subsamples 1, 2 and 3). Removing the values of the three segments nearest the calamus from the calculation of the means and standard deviations of the four isotopes yielded delta values that were clearly more isotopically homogeneous. For this reason, in this specific situation, we suggest eliminating this part of the feather before isotopic analysis, in agreement with Chang et al. (2008). However, the decision whether or not to do this has to be made on a case by case basis (Wassenaar 2008).

Table 3. Variability in stable isotope ratios of C, N, H and O along the feather vane
 δ13C (‰)δ15N (‰)δ2H (‰)δ18O (‰) δ13C (‰)δ15N (‰)δ2H (‰)δ18O (‰) δ13C (‰)δ15N (‰)δ2H (‰)δ18O (‰)
  1. T, Turdus philomelos; C, Carduelis spinus; and A, Apus apus. 1 = feather base subsection, 10 = feather tip subsection.

T-1−24·30·9−11010·4C-1−22·94·0−14614·2A-1−23·513·9−4715·4
T-2−24·81·49712·3C-2−22·94·0−12114·2A-2−22·714·0−3813·2
T-3−24·52·4−10011·0C-3−21·94·1−11613·1A-3−23·113·1−2912·9
T-4−23·83·1−9610·5C-4−21·83·9−8313·0A-4−22·412·7−2413·1
T-5−24·33·4−968·4C-5−21·44·2−8212·3A-5−22·612·2−1212·6
T-6−24·14·1−978·3C-6−21·54·5−8412·6A-6−22·012·3−1012·0
T-7−24·04·8−968·2C-7−21·74·6−8612·5A-7−22·212·5−811·7
T-8−23·84·6−908·8C-8−21·64·7−8812·6A-8−22·312·3−1111·9
T-9−23·95·0−949·4C-9−21·24·5−8812·4A-9−22·312·2−511·8
T-10−23·85·0−989·0C-10−21·24·9−8813·3A-10−21·612·3−1811·9
Mean−24·13·5−979·6Mean−21·84·3−9813·0Mean−22·512·7−2012·6
SD0·41·551·4SD0·60·3220·7SD0·50·7141·1
Mean (samples 4–10)−23·94·3−958·9Mean (samples 4–10)−21·54·5−8612·6Mean (samples 4–10)−22·212·4−1212·1
SD (samples 4–10)0·20·830·8SD (samples 4–10)0·20·330·4SD (samples 4–10)0·30·270·5

Results of the Reference Materials

Results of the laboratory reference and control materials are shown in Table 4 for δ2H and δ18O, the most problematic measurements (Qi, Coplen & Wassenaar 2011; Meier-Augenstein, Hobson & Wassenaar 2013). Measurements of the two keratin laboratory reference materials (CBS and KHS) carried out over two 3-month periods in 2011 and 2012 (July–September 2011 and June–August 2012) and corrected for linear instrumental drift were found to be precise. The typical mean ± standard deviation values of CBS and KHS were, respectively, −196 ± 2 ‰ (n = 102) and −54 ± 2 ‰ (n = 103) for δ2H, and 2·3 ± 0·0 ‰ and 21·3 ± 0·3 ‰ for δ18O. The 6-month running means and standard deviations of keratin were −79 ± 3 ‰ for δ2H and 8·7 ± 0·3 ‰ for δ18O (n = 27). Overall, we consider these results to be satisfactory as they are comparable to those obtained for repeatability in much more ‘protected’ conditions (the same sample run consecutively 10 times by the same operator). Furthermore, for δ18O, it seems that the use of a longer GC column (80 cm) than the model normally used (60 cm) allowed better separation of CO from the interfering N2 during δ18O analysis, allowing complete separation of the two peaks. Furthermore, its frequent and immediate change in case of worsening performance (closer retention times of H2 and CO peaks, flattened shape of CO peak) due to the accumulation of sulphur as well as the use of two matrix match equivalent keratin reference materials to correct the values of feather samples mitigated the problems inherent in δ18O determination (Qi, Coplen & Wassenaar 2011).

Table 4. Trends in the δ2H and δ18O values of reference (CBS and KHS keratins) and control (commercial keratin and casein) materials
 CBS (n = 102)KHS (n = 103)Commercial keratin (n = 27)Commercial casein (n = 28)
δ2H δ18Oδ2Hδ18Oδ2Hδ18Oδ2Hδ18O
Mean−1962·3−5421·3−798·7−1519·7
SD20·020·330·320·3

Despite what has been reported in the literature (Meier-Augenstein, Hobson & Wassenaar 2013), and despite being a different proteic material from keratin, and therefore probably having a different number of exchangeable hydrogen atoms, the commercial casein control material run in each δ2H and δ18O analytical batch yielded surprisingly very repeatable results over the 6-month running period (−151 ± 2 ‰ and 9·7 ± 0·3 ‰, respectively). Therefore, we decided to investigate whether casein could be a useful standard for keratinous materials. To this end, all the data were recalculated using casein as a normalization standard and taking its ‘true’ value as that found with the keratin correction mode. The recalculated keratin values for δ2H were −185 ‰ instead of −197 ‰ (∆ = +12 ‰) for CBS and −78 ‰ instead of −54 ‰ (∆ = −24 ‰) for KHS, while for δ18O, they were 2·8 ‰ instead of 3·8 ‰ (∆ = −1 ‰) for CBS and 21·3 ‰ instead of 20·3 ‰ (∆ = 1 ‰) for KHS. If we look at the samples, we see that the mean ± standard deviation of the difference between the values calculated with keratin correction and those calculated with casein normalization were −14 ‰ ± 5 ‰ for δ2H and 0·4 ‰ ± 0·5 ‰ for δ18O. Data agreement was much better in the samples with values similar to those of casein; with respect to δ2H, the mean difference changed from −2 ‰ ± 0·6 ‰ for sample values up to −138 ‰, increasing to −5 ‰ ± 2 ‰ for sample values up to −120 ‰ and to −20 ‰ ± 3 ‰ for samples with values ranging from −80 ‰ to −57 ‰. Similarly, the mean difference for δ18O changed from 0·3 ‰ ± 0·2 ‰ for sample values up to 11·9 ‰, increasing to 0·6 ‰ ±0·9 ‰ for samples with values ranging from 15 ‰ to 23 ‰.

Finally, we tried to correct the isotope values using KHS and casein data, using a calibration curve (not CBS because the range between the two standards would be too narrow). The 6-month running mean ± standard deviation values of CBS, KHS and casein resulting from this calibration were, respectively, −196 ± 1 ‰, −54 ± 2 ‰ and −151 ± 1 ‰ for δ2H, and 2·6 ± 0·3 ‰, 21·3 ± 0·2 ‰ and 9·7 ± 0·2 ‰ for δ18O. Regarding the samples, the mean ± standard deviation values of the difference between the values calculated with keratin correction and those calculated with KHS–casein calibration were −1 ‰ ± 1 ‰ for δ2H and 0·1 ‰ ± 0·1 ‰ for δ18O.

It seems, therefore, that even though it probably has a different number of exchangeable Hs, casein can be used to normalize the δ2H values of keratin if the range of values is close to that of the reference material (±30 ‰). If there is a wider range of variability, a linear equation built on reference standards covering a wide range of values should be used, as suggested by Meier-Augenstein, Hobson and Wassenaar (2013).

Conclusions

This paper reports an investigation and critical assessment of the effects of different sample preparation treatments on determining the stable isotopes of δ13C, δ15N, δ18O and δ2H in bird feathers. In particular, the results demonstrated that the solvent mixture diethyl ether-methanol 2 : 1 can be used instead of chloroform-methanol 2 : 1 as a cleaning mixture, thus avoiding a carcinogenic solvent. Furthermore, it was shown that there were no significant differences between air-drying and oven-drying feathers, and between cryogenically pulverizing feathers and thinly cutting them with surgical scissors. Moreover, variability along the vane and between the vane and rachis was shown for the four isotopes, but it was found that removing the part of the feather closest to the calamus resulted in more homogeneous delta values. This last finding shows that the choice feather part to be analysed has a bearing on the information the researcher wishes to derive from the isotopic composition. Finally, we demonstrated that a casein with a known δ2H value, but probably with a different number of exchangeable Hs with respect to keratin, can be used to normalize the δ2H values of feathers, but only if the range of values is close to that of the reference material. Where there is a greater degree of isotopic variability, a linear equation built on reference standards covering a wide range of values should be used. Similarly, in δ18O determination, using a longer GC column, and immediately changing it in case of worsening performance, and using two keratin reference materials to correct the values of feather samples, fewer problems arose due to the accumulation of sulphur and nitrogen interference.

Acknowledgements

The authors would like to thank their collaborators at the MUSE ringing station at Bocca di Caset (Progetto Alpi) for collecting the samples.

Ancillary