Fatty acid signatures in seabird plasma are a complex function of diet composition: a captive feeding trial with herring gulls


  • R. Käkelä,

    Corresponding author
    1. Faculty of Biosciences, University of Joensuu, PO Box 111, FI-80101 Joensuu, Finland;
    2. Department of Biological and Environmental Sciences, University of Helsinki, PO Box 65, FI-00014 Helsinki, Finland;
      *Correspondence author. E-mail: reijo.kakela@helsinki.fi
    Search for more papers by this author
    • Present address. Department of Medical Biochemistry and Developmental Biology, Institute of Biomedicine, Biomedicum Helsinki, University of Helsinki, Haartmaninkatu 8, PO Box 63, FI-00014, Finland

  • R. W. Furness,

    1. Institute of Biomedical and Life Sciences, Graham Kerr Building, University of Glasgow, Glasgow G12 8QQ, UK; and
    Search for more papers by this author
  • S. Kahle,

    1. Institute of Avian Research ‘Vogelwarte Helgoland’, An der Vogelwarte 21, D-26386 Wilhelmshaven, Germany
    Search for more papers by this author
  • P. H. Becker,

    1. Institute of Avian Research ‘Vogelwarte Helgoland’, An der Vogelwarte 21, D-26386 Wilhelmshaven, Germany
    Search for more papers by this author
  • A. Käkelä

    1. Faculty of Biosciences, University of Joensuu, PO Box 111, FI-80101 Joensuu, Finland;
    Search for more papers by this author

*Correspondence author. E-mail: reijo.kakela@helsinki.fi


  • 1Tissue fatty acid signatures (FAS) can complement traditional methods of studying seabird diets. Although plasma lipid FAS are known to indicate dietary changes qualitatively, here we test whether they can be used to determine the proportions of different dietary items in a quantitative manner.
  • 2Captive herring gulls (Larus argentatus) were fed North Atlantic plaice Pleuronectes platessa (demersal species made available to wild seabirds by fisheries) and herring Clupea harengus (pelagic fish often found naturally in their diet) with different mixing ratios (0%, 10%, 20% and 50% herring).
  • 3Major fatty acids did not indicate diet, but several minor components in plasma, for example, 14 : 0, 18 : 3n-3, 18 : 4n-3 and C20–22 monounsaturated fatty acids (MUFA), showed good correlations with diet composition. Different fatty acids were incorporated from diet into plasma lipids with different calibration coefficients.
  • 4Together with dose-dependent but inefficient (low calibration coefficient) transfer of 22 : 1n-11 (a major fatty acid of herring) to the plasma FAS of the gulls, the percentages of potential chain shortening products of 22 : 1n-11, that is, 20 : 1n-11, 18 : 1n-11 and 16 : 1n-11 increased with increasing proportion of herring in the diet. Notably, the dietary supply of these fatty acids itself did not change. Thus the metabolic products of certain dietary fatty acids can reflect the amount of their dietary precursors in a quantitative way.
  • 5Despite the fact that many major fatty acids in FAS of seabird plasma are greatly modified by endogenous metabolism, several minor components of FAS (in this experiment 14 : 0, branched chain 17 : 0, 18 : 1n-7, 18 : 3n-3, 18 : 4n-3, C20–22 MUFA with their chain shortening products, and 22 : 4n-6) that can be accurately and reliably quantified by gas chromatography, vary proportionally to diet composition, allowing their use for monitoring temporal and spatial differences in seabird diet.


Quantitative information on seabird diet composition may give early warnings of ecological change in marine systems (Montevecchi 1993; Frederiksen et al. 2004; Wanless et al. 2004), and may contribute to multi-species assessment of commercially important fish stocks (e.g. North Sea sandeel Ammodytes marinus, ICES 2002). To use seabird diet data in this way, it is important to be able to obtain quantitative data on seabird diet composition. A large number of methods can be used to assess seabird diet directly, including shooting birds at sea to obtain stomach contents, observation or trapping of birds carrying food to chicks, collection of voluntary or forced regurgitations, collection of prey remains and regurgitated pellets (Barrett et al. 2007). However, all of these methods have biases of one sort or another which can make quantitative interpretation extremely difficult, and several are constrained by ethical, welfare and conservation issues, while most are restricted to a very short period of the year, while the seabirds are breeding and often just to when they are rearing chicks (Barrett et al. 2007).

In recent years there has been increasing interest in the development of indirect methods of diet assessment, such as serological methods, or the analysis of fatty acids and stable isotopes, in tissues or excrement, as these biochemical approaches can avoid some of the problems with direct assessments (Barrett et al. 2007). Probably the most promising of these biochemical methods for quantitative analysis of diet is the use of fatty acid signatures (FAS). Although use of FAS is quite well-established in studies of marine mammal diet (Iverson et al. 2004; Beck, Iverson & Bowen 2005; Cooper, Iverson & Heras 2005), analysis of FAS is just beginning to be applied to seabirds (Raclot, Groscolas & Cherel 1998; Dahl et al. 2003; Käkeläet al. 2005, 2006, 2007, Wang et al. 2007), and to investigate diet of invertebrates, fish and reptiles (Blair, Cree & Skeaff 2000; Kaushik et al. 2006; Stowasser et al. 2006). It seems likely that application of biochemical methods such as analysis of FAS alongside conventional diet sampling can provide a better understanding of seabird diet than the use of either method on its own (Käkeläet al. 2006), suggesting that a combined approach will be most productive.

Iverson et al. (2004) presented a statistical modelling technique that provides quantitative estimates of diet composition of piscivorous mammals from FAS obtained from adipose tissue samples (‘quantitative fatty acid signature analysis’, QFASA). They suggest that this technique has wide application for other marine predators, such as seabirds. However, obtaining adipose tissue samples by biopsies of live seabirds is much more difficult than sampling the large blubber layers of marine mammals, and so few equivalent studies using live biopsies have been published on seabirds (Iverson, Springer & Kitaysky 2007, Wang et al. 2007). In contrast, sampling blood from seabirds is relatively easy to do and can be done with safe protocols even when dealing with rare and endangered species, and so there is interest in whether FAS of blood can be used in quantitative or semi-quantitative assessment of seabird diet composition. Previously we showed that FAS of plasma samples from seabirds can indicate very clearly, at least in a qualitative way, switches in feeding between demersal and pelagic fish, due to changes in amounts of characteristic fatty acids of zooplankton and of the benthic bacterial foodweb in seabirds when feeding on pelagic or on demersal fish respectively (Käkeläet al. 2005). However, although experiments where we fed herring gulls (Larus argentatus) on pelagic or demersal fish caused a rapid and large shift in FAS in their plasma, it is not clear whether these FAS changes, or at least some components of these changes, are quantitative, and so indicate the proportions of different items in the diet.

Plasma fatty acids are derived from a mixture of polar and neutral lipids in varying ratio. Normally the polar lipids dominate and sometimes the plasma is poor in the neutral lipids that are expected to reflect the diet more than the polar ones. Given that there is also extensive metabolism of fatty acids by animals, which alters their FAS from that of their food (Budge, Cooper & Iverson 2004; Cooper, Iverson & Heras 2005; Käkeläet al. 2005; Cooper, Iverson & Rouvinen-Watt 2006; Stowasser et al. 2006), we did not expect to be able to show that the plasma FAS of seabirds is a direct quantitative reflection of the proportions of different foods in their diet. Instead we wanted to test whether some specific components of the plasma total lipid FAS could be used to indicate diet quantitatively, so we carried out feeding experiments with groups of seabirds in captivity given several different proportions of two kinds of fish known to differ markedly in fatty acid composition. Our aim was to investigate which, if any, of the components of the FAS show an approximately linear response to proportional composition of the diet.

Materials and methods

feeding trial and sampling

Twelve captive-bred adult herring gulls were housed in large outdoor aviaries (length 6·7 m, width 3 m, height 2·1 m) at the Institute of Avian Research ‘Vogelwarte Helgoland’, Wilhelmshaven, Germany (governmental institution with licence to keep birds) and fed 100% demersal fish, plaice Pleuronectes platessa (3·1 ± 1·1% lipid, wt/wt, on wet basis) for 24 days and after that sampled for blood (18 March 2005), which represented starting samples for the feeding experiment. Then the gulls were divided into three dietary groups (each group of four birds in their own cage) in which either (A) 10%, (B) 20% or (C) 50% of the plaice was replaced by pelagic herring Clupea harengus (undersized young individuals, 2·8 ± 0·6% lipid, wt/wt; in both species, the proportion of neutral lipids varied considerably, 15–50%, and averaged about one third of the total lipids). The dietary fish were bought fresh and stored frozen (at –18 °C) portioned in polyethylene bags for up to maximum of 3 weeks before use. The groups of birds were made as similar as possible. At the first/last sampling, the birds of the groups weighed: (A) 920 ± 90 g/850 ± 100 g, (B) 1020 ± 160 g/1030 ± 170 g, and (C) 1050 ± 190 g/1020 ± 190 g. In each group, three of the gulls were born 1994–1998, and one during 2004. In groups B and C there were two males and two females, but group A had only one male. Blood samples were collected again after the gulls had been feeding on these experimental diets for 11 and 21 days. The sampling times were chosen based on our earlier study (Käkeläet al. 2005) showing that after changing diet a clear shift in plasma FAS could be found in five days, which promised equilibrium in the percentages of fatty acid indicators by 11–21 days. The gulls were also weighed at every time of sampling. Each group was offered 1·0 kg fish per day, which was all consumed. The steady body weights of the birds showed that the differences in the amount of food taken by different individuals (causing potential inter-individual variability for the results of a group) were not large. The gulls were fed in the morning after sampling of blood. Thus the time interval from the last feeding to blood sampling was c. 24 h.

The blood samples were taken under licence of the Bezirksregierung Weser-Ems, Oldenburg (animal protection legislative body of Lower Saxony, Germany). Blood (1–2 mL per gull) was sampled by venipuncture from the vena ulnaris cutanea via a 24-gauge needle into a syringe lined with 100 µL of EDTA to prevent clotting. The blood was centrifuged (for 15 min at 8900 g at 4 °C) to separate plasma, which was stored at –20 °C under nitrogen for a maximum of 2 months before analysis.

fatty acid analysis

Fatty acid composition was determined in the total lipids of herring gull plasma at every sampling and in the diets, that is, homogenates of the whole plaice or herring (n = 5). Subsamples were directly transmethylated according to the recommendations of Christie (1993) by heating with 1% H2SO4 in methanol under nitrogen atmosphere and the formed fatty acid methyl esters (FAME) extracted with hexane in two steps. The dried and concentrated FAME were analysed by gas-liquid chromatography (GLC) using flame ionisation detection (FID) for quantification and mass detection for identification of the individual FAME (6890N network GC system with FID and 5973 mass selective detector, Agilent, USA; both operating lines equipped with DB-wax capillary columns, 30 m, ID 0·25 mm, film thickness 0·25 µm, J&W Scientific) as detailed previously (Käkeläet al. 2005). Quantifications were based on FID responses corrected according to the theoretical response factors (Ackman 1992) and calibrations with quantitative authentic standards. The fatty acid proportions were calculated as mol %, and the fatty acids were marked by using the abbreviations: [carbon number] : [number of double bonds]n–[position of the first double bond calculated from the methyl end] (e.g. 22 : 6n-3). All the quantitatively important polyunsaturated fatty acids (PUFA) were methylene-interrupted.

To study the biases in transfer of different fatty acids from diet to bird plasma, the plasma mol % of each fatty acid was divided by the corresponding % in the diet. These calibration coefficients were determined separately for all the dietary groups, that is, different dietary mixtures. In addition the plasma lipid class profiles and lipid contents of the whole fish homogenates were studied by thin-layer chromatography-FID analyser as detailed in Käkeläet al. (2005).

data analysis

The differences in the proportions of different fatty acids in the two dietary fish were tested by student's t-test. The differences in the values of the plasma samples of the gulls from different dietary groups collected at different times were studied by within subject analysis of variance (anova). If significant, subsequent pairwise comparisons were performed by repeated measures t-test. The interdependence of the dietary and plasma levels of indicator fatty acids were studied by mixed model regression accepting repeated measures (in spss 15).

As a multivariate approach to find out which fatty acids were mainly responsible for the differences between the groups, the FAS at day 0 and day 21 were subjected to principal component analysis (PCA) (Kvalheim & Karstang 1987). In PCA the dominance of the starting group (0% herring) with larger number of samples was removed by block normalizing the group size, and the dominance of the major components with large variations that did not carry dietary information was removed by dividing the log normalized fatty acid percentages by their means. The relative positions of the samples and variables were plotted using two new coordinates, the principal components PC 1 and PC 2 describing the largest and second largest variance among the samples. To quantify the observed differences between the samples from different experimental groups, space-filling models were created for two of the groups at a time and distances of the samples to these models were computed by using soft independent modelling of class analogies (SIMCA) at P < 0·05 level. The computations were performed by using SIRIUS program package (Pattern Recognition Systems, Bergen, Norway).


differences in fas of dietary plaice and herring

The plaice and young herring were selected for the dietary fish of the present feeding experiment based on their similar lipid content, 3 wt% ensuring as equal mass transfer of fatty acids as possible, and their very different fatty acid compositions characteristic of North Atlantic demersal and pelagic fish respectively (Fig. 1). The herring added to the plaice diet contained more 14 : 0, and many times more long chain monounsaturated fatty acids (MUFA) 20 : 1n-9 (5×) and 22 : 1n-11 (20×) than the plaice. In contrast, the plaice was rich in 18 : 0, and shorter MUFA 16 : 1n-7, 18 : 1n-9 and 18 : 1n-7. Among PUFA the herring was relatively rich in 18-carbon PUFA, and plaice had more of the 20- and 22-carbon PUFA, except that the largest PUFA component, 22 : 6n-3 and a minor 20 : 4n-3 were present in equal amounts in the two fish.

Figure 1.

Main fatty acids in the total lipids from whole fish homogenates of herring and plaice (mol%, mean ± SD, n = 5). The fish differed for all shown fatty acids, except 20 : 4n-3 and 22 : 6n-3, at least at P < 0·05 level (student's t-test).

changes in fas of herring gull plasma

In each dietary group with different mixing ratios, the change from the starting to the final FAS happened during the first 11 days, and the fatty acid levels showed little change from day 11 to day 21 (Fig. 2). The lipid class profiles of plasma did not differ between the samplings or dietary groups (62–75% polar lipids, 17–21% cholesteryl esters, 1% triaclyglycerols, and traces of other lipids) (P > 0·3 for all the different lipid classes in the 10%-, 20%- and 50% herring groups at the end of the experiment, anova). Despite the large differences in the dietary supply of different fatty acids, the plasma FAS of the herring gulls contained three major fatty acids with 15–20 mol%, that is, 18 : 0, 18 : 1n-9 and 20 : 4n-6, the mean levels of which did not differ significantly between the groups (see Fig. S1 in Supporting Information). Among the five 4–5% components (16 : 1n-7, 18 : 1n-7, 18 : 2n-6, 20 : 5n-3 and 22 : 6n-3) only 18 : 1n-7 showed a statistically significant (P < 0·05, anova followed by pairwise t-tests) trend, levels decreasing with the decrease of the dietary percentage of plaice. Most markers indicative of the dietary change were found in the minor fatty acid components (P < 0·05).

Figure 2.

Temporal changes in the herring gull plasma mol% of three fatty acids (a) 14 : 0, (b) 20 : 1n-9 and (c) 22 : 1n-11, which showed clear relative changes due to replacing 10%, 20% and 50% of the dietary plaice with herring (mean ± SD, n = 4). The values at different time points were analysed by within subject anova (P < 0·02 for all fatty acids and diets shown) followed by pairwise comparisons performed by repeated measures t-test, the probability levels of which are indicated on the plots. In general, the values of the 0-day and 11-day samples differed significantly and the values of the 11-day and 21-day samples were similar. The results are shown for the three quantitatively most important indicator fatty acids but the same patterns were evident and statistically significant for all the 12 fatty acids indicating the dietary change. NS = not significant.

PCA demonstrated that the best indicators of herring in the diet were the minor components 22 : 1n-11, 16 : 1n-11, 20 : 1n-11, 20 : 1n-9, 18 : 4n-3, 18 : 3n-3 and 14 : 0 which were clearly enriched along with the increase of herring in the diet. The first axis explained (after removing the dominance of the major components by standard normalization and weighting methods) 68% of the total variation (Fig. 3a). Farthest on left on this axis, although with small eigen values, were 22 : 4n-6, 18 : 1n-7, 20 : 2n-6, and the branched chain 17 : 0 more abundant in the plaice than in herring. The quantitativity of these PCA separations was tested by SIMCA, in which the samples of dietary groups fed 0 or 10%- vs. 50% herring fell inside their own statistical models separate from each other at the P < 0·05 level of confidence (as an example the class distance plot 10% vs. 50% herring is shown in Fig. 3b). The second axis PC2 explained 14% of the remaining variation. The dietary group that showed the largest spread in this direction was the starting group (0% herring), and the fatty acid that had the largest impact was 18 : 2n-6, plentiful in terrestrial sources and in commercial avian feeds. Therefore, this remaining variation has probably less to do with the experimental feeding but may represent individual differences in utilizing long-term lipid stores of the body or other metabolic factors.

Figure 3.

(a) Biplot of principal component analyses of plasma samples from herring gull fed the starting diet of plaice (0% = white squares), and diet in which 10% (black triangles), 20% (black diamonds) or 50% (black squares) of the plaice was replaced by herring (group size was block normalized, and the values log transformed and weighted by diving by the mean). The origin is marked with a cross in the middle of the plot and emphasized by the dashed lines starting from it. Fatty acids exceeding 0·1 mol% at least in one sample were used as loadings, and the grey patterned area near the origin shows the location of the fatty acids that had small influence on the separation of the samples (and for figure clarity were not marked individually). For clarity the samples belonging to each group were delineated. Abbreviations i and ai stand for the iso- and anteiso-branches in the carbon chain of a fatty acid. (b) Class distance plot (SIMCA, P < 0·05) of the samples from the groups fed 10% (triangles) and 50% herring (squares). Vertical dash line is the outer limit for the model of the 10%-group and horizontal dash line is the outer limit for the model of the 50%-group. The samples fall inside their own statistical models without clear outliers, and there are no samples in the area of model overlap. The samples of the gulls fed 0% and 50% herring were also separated at P < 0·05 level by SIMCA (not shown).

Mixing the plaice and herring (with the same content of total lipid) with known ratios enabled us to compare whether certain fatty acid markers are detected in the plasma in the proportions predicted from their abundance in the diet by assuming their additive/quantitative transfer from diet into plasma lipids. The dietary levels of the fatty acids in the mixed diets were calculated by weighting their percentages in the two fish according to the mixing ratio. The calibration coefficients between the plasma and dietary levels were very different for different fatty acids, and in the case of most fatty acids the coefficients differed also slightly for different mixing ratios (Table S1). A mixed model regression with repeated measures treating each bird as an individual, within subject factor in two classes (the starting, that is, baseline and experimental diets), and dietary fatty acid concentration as a fixed factor, showed that there were statistically significant effects of diet on plasma fatty acid for 12 fatty acids (Fig. 4, Table S2). Fatty acids 22 : 1n-11, 20 : 1n-9, 14 : 0 (Fig. 4a) and the minor components 18 : 3n-3 and 18 : 4n-3 (Fig. 4b) indicated the gradual increase of herring, rich in these fatty acids, in the diet. The correlations were best described by logarithmic fittings, but were also highly significant when interpreted by linear function (equations and r2-values by linear function on the plots; F-values and probability levels in Table S2). Interestingly, also two minor fatty acids 20 : 1n-11 and 16 : 1n-11 (Fig. 4c), which had the same small amounts in all diets, correlated significantly with the amount of herring, rich in 22 : 1n-11, in the diet of the birds. Also the biochemically related minor component 18 : 1n-11, seen as a shoulder of the main 18 : 1 isomer 18 : 1n-9, was increased several-fold, but the chromatographic separation was not sufficient to allow accurate quantification. The 18 : 1n-7 characteristic for plaice decreased with the herring supplementation (Fig. 4d), as did the minor fatty acids 17 : 0iso, 17 : 0anteiso, 20 : 2n-6 and 22 : 4n-6 (Fig. 4e).

Figure 4.

Correlations between the percentages of some indicative fatty acids in the diet and herring gull plasma: (a) 14 : 0, 20 : 1n-9 and 22 : 1n-11 rich in herring, (b) 18 : 3n-3 and 18 : 4n-3 rich in herring, (c) 20 : 1n-11 and 16 : 1n-11 likely produced by the gull from 22 : 1n-11 (the curve of 22 : 1n-11 included for comparison), (d) 18 : 1n-7 rich in plaice, and (e) 17 : 0iso (abbreviated i), 17 : 0anteiso (ai), 20 : 2n-6 and 22 : 4n-6 rich in plaice. Mixed model regression analyses were performed allowing repeated measures and showed statistically significant effects of diet on plasma fatty acid for all these 12 fatty acids (see Supplementary Table S2 for details). Means ± SD (n = 4) for each dietary mixture and logarithmic curves are shown. Equations, r2-values, and statistics (Table S2) are for the original data points by linear function.


Tissue FAS are likely to prove useful in future studies of seabird diets, and the potential of using the FAS, analysis of stable isotopes of nitrogen and carbon in tissues, and conventional dietary sampling methods (stomach contents, excrement, pellets etc.) in combination offers a possibility for even more powerful analyses of seabird diets (Barrett et al. 2007; Käkeläet al. 2007). The FAS provide a potential to investigate the winter diet of individual birds caught during the breeding season if adipose tissue can be sampled (Furness et al. 2006), without most of the associated biases affecting conventional analysis of diet from stomach contents analysis (Barrett et al. 2007). They provide an opportunity to investigate recent diet of individuals from blood plasma samples taken at capture, either during the breeding season or indeed away from colonies at other times of year when conventional sampling of diet by non-destructive means is extremely difficult (Barrett et al. 2007). In contrast to most conventional methods of diet study, FAS therefore has the potential to make major contributions to our understanding of the trophic role of seabirds at times of year when they are not rearing chicks. Although sampling adipose tissue from live seabirds is not a method that can be routinely used for all seabirds, taking blood samples is extremely easy, and this extra procedure causes minimal disturbance to birds beyond that of capture and handling itself. We have already shown that seabird plasma fatty acids in captive seabirds alter in relative abundance as a consequence of switching diet between contrasting prey types (Käkeläet al. 2005), and that these qualitative indicators of diet type also show variation in wild seabirds that corresponds to expected differences in diets of individuals and of different species (Käkeläet al. 2006, 2007). So the ability to use FAS to calculate from a two-source mixing model, the proportional composition of the diet, for example as the proportion of pelagic or the proportion of demersal fish consumed, would be a very useful development. However, when using mixing models the lipid poor and lipid rich items must be weighted differently, according to the actual mass of fatty acids they are providing.

In the present study we have found that the most abundant fatty acids in plasma tend to provide very little or no indication of diet, and that many fatty acids show changes in abundance that relate to physiology rather than to diet. Nevertheless, our data do indicate that some of the less frequent fatty acids show quantitative changes in abundance that correlate closely with the proportions of pelagic rather than demersal fish in the diet. The reproducibility and accuracy of modern GLC allows reliable detection of fatty acids 1–2 orders of magnitude less abundant than the ones used here as indicators. These results imply that plasma FAS can be used as a quantitative measure of diet composition in birds, but also indicate the need to improve understanding of the physiological processes affecting the FAS.

Fatty acids of avian plasma originate from phospholipids, cholesteryl esters, triacylglycerols, unesterified so called free fatty acids, and other minor lipids, distributed differently between different lipoprotein particles and carrier proteins. The fatty acids from these different lipids reflect the dietary fatty acid supply to different extents according to different time-windows (Castillo et al. 2002; Skeaff, Hodson & McKenzie 2006; de Beer et al. 2008). In birds, the ingested lipids are hydrolyzed and absorbed as fatty acids in the intestine, and then incorporated mainly into the lipids of large lipoprotein particles called portomicrons that are drained from intestinal walls by portal blood directly to the liver before entering general circulation (Bensadoun & Rothfield 1972; Hermier 1997). Thus, a couple of hours after absorption, the portomicron neutral lipids with fatty acids from the recent meal are elevated (Sanz et al. 2000; Lien, Jan & Chen 2005). However, the fatty acids in the neutral lipids of all lipoproteins are not exclusively from recent diet, and when more time has elapsed from the last meal the phospholipid fatty acids comprise most of the plasma FAS. Despite the fact that free-ranging seabirds from the wild have very variable lipid profiles of plasma, most samples from the wild have had phospholipid contents clearly exceeding those of the cholesteryl esters and triacylglycerols (Guglielmo et al. 2002; Guglielmo, Cerasale & Eldermire 2005; Käkeläet al. 2006; Vaillancourt & Weber 2007). Although the incorporation of dietary fatty acids into phospholipids is largely controlled by endogenous biochemical preferences, the conservative FAS of plasma phospholipids are also affected by large changes in dietary fatty acid composition (Gillotte et al. 1998; de la Presa-Owens, Innis & Rioux 1998). Thus every individual from the wild has its momentary metabolic status, and would be worthy of a detailed study of the lipoprotein and lipid class profiles, and subsequent analyses of the FAS in the different lipoprotein fractions and lipid classes separately. However, in ecological studies, often involving samples of hundreds of individuals, it is cost-effective to analyse the FAS of plasma total lipids, already shown to indicate switches in seabird diet remarkably well (Käkeläet al. 2005). The present study has developed the method by showing that several indicator fatty acids can quantitatively reflect dietary changes.

Despite marine food webs being complex and potential dietary items numerous, with seasonally changing abundance and size-related interspecies variability of fatty acid composition, specific characteristics of the FAS of demersal vs. pelagic fish enable the estimation of the relative contributions of these types of fish in seabird diets (Iverson, Frost & Lang 2002; Käkeläet al. 2005, 2006, 2007). In the present feeding experiment we used different mixtures of one demersal and one pelagic species that have the same total lipid content and as different FAS as possible, which is thus the simplest possible experiment to test whether certain fatty acid components in the total lipid samples of bird plasma indicate diet in a quantitative manner. Compared to ecological studies of free-ranging seabirds, this experiment is a simplification also in another way; all the birds were sampled 24 h after the last meal, which removes most of the variation in FAS arising from temporal changes of the ratio of polar and neutral lipids in the plasma. However, it should be recalled that also in the wild seabirds, as shown for the great skua Stercorarius skua (Käkeläet al. 2006) the main part of the plasma fatty acids come from phospholipids (in the skua, more than 70% originated from phospholipids, 20–25% from cholesteryl esters and only 3% from triacylglycerols; the relative variation was the highest for the triacylglycerols). Thus, the use of total lipid FAS can reveal temporal or spatial shifts in seabird diets.

The fast equilibration of the plasma FAS was entirely consistent with conclusions from our previous study (Käkeläet al. 2005) showing that a large shift in the diet can be detected in the FAS of plasma total lipids in a few days. The present data indicated equlibration of the plasma FAS after 11 days on the same diet. The fast equilibration may be due to the high metabolic rate of birds, the pronounced role of the liver in avian lipid metabolism, and the small volume of slowly-turning over adipose tissues of the birds (Griffin et al. 1992; Bryant & Furness 1995). The fatty acid turnover of avian adipose tissues is relatively fast; for example a 20-day half-life in chickens (Foglia et al. 1994).

The very different calibration coefficients between plasma and diet for different fatty acids showed that the whole FAS is not very useful in indicating the seabird diet. First, 16 : 1n-7, 18 : 0 and 18 : 1n-9, which were major components in plasma, differed considerably in the two diets, but showed no response to the different mixing ratios of the diets. These fatty acids are likely distributed largely to adipocytes in different parts of the body but can also be synthesized de novo, and thus their levels in different tissues can be readily adjusted according to tissue and lipid-specific preferences of the species (Miyazaki & Ntambi 2003). Second, the phospholipids of lipoprotein shells, mainly phosphatidylcholine, conserve and incorporate preferentially 20 : 4n-6, which is then greatly enriched compared to the other major PUFA available from the diet (Mosconi et al. 1990). In this study the decrease in the dietary ratio of n-6PUFA : n-3PUFA was modest and thus faded out in the plasma FAS of the herring gulls through endogenous metabolism. However, the ratio of 20 : 4n-6 to 20 : 5n-3 in the plasma of captive herring gulls and wild seabirds has previously been found to change when even more pronounced changes occurred in the dietary ratio of n-6 PUFA : n-3 PUFA than in the present experiment (Käkeläet al. 2005, 2006, 2007).

Although the endogenous metabolism tends to mask the influence of diet on the major fatty acids, a 3%-component 18 : 1n-7 was affected by the dietary supply. This isomer of 18 : 1 is plentiful in plaice, as in the other North Atlantic demersal fish species, and originates largely from benthic microbial sources (Käkeläet al. 2005; Pistocchi et al. 2005). Birds fed plaice may have got one part of their 18 : 1n-7 from this food, but additionally they may have elongated some of the 16 : 1n-7, also plentiful in plaice, to 18 : 1n-7 by themselves. Both ways the consequence would be the observed linear response of 18 : 1n-7 to the dietary change.

The best indicators of diet were found in the minor fatty acid components of plasma that are not the targets of large scale endogenous metabolism in the bird. These included 14 : 0, the long chain MUFA and C18-PUFA. The 14 : 0 reflects diet successfully because the endogenous fatty acid synthase of vertebrates, when using acetyl-CoA as a primer, releases 16 : 0 as the exclusive product, and by doing so passes the 14-carbon stage without releasing 14 : 0 (Smith, Witkowski & Joshi 2003). In addition, tissues do not synthesize large amounts of MUFA longer than 18 carbons, necessary mainly for the synthesis of sphingolipids (Sastry 1985), and thus the C20–22 MUFA are mainly of dietary origin (Saito & Kotani 2000). The metabolism of PUFA is largely heading to the formation of C20–22 PUFA, having the highest biological value, for example in structuring of cellular membranes. However, since the birds obtain a lot of C20–22 PUFA directly from the diet, further elongation and desaturation of C18 PUFA precursors to their C20–22 derivatives is probably slow, and thus a part of the dietary C18 PUFA remains to be incorporated into body lipids unaltered.

Even among the fatty acids indicating diet, the transfer of the signal from diet into plasma lipids differed: the calibration coefficients of 14 : 0 and 20 : 1n-9 were about 1/5, but for 22 : 1n-11 < 1/10. Moreover, the calibration coefficients for the long MUFAs, especially for 22 : 1n-11, decreased when their dietary supply increased. Clearly, mixing of the dietary fatty acids with those from endogenous pools (i.e. passive dilution) cannot explain this selectivity. More likely, tissues do not accumulate much 22 : 1n-11 since excess retention of this fatty acid would cause tissue lipidosis (due to poor mitochondrial oxidation) and lead to relatively viscous lipid depots (due to the high melting point of the long chain acid) (Bremer & Norum 1982). This is prevented by activating peroxisomal chain shortening (partial β-oxidation) of the long chain fatty acids (Reddy & Hashimoto 2001). Consequently, the high supply of 22 : 1n-11 from the diet should then lead to increases of its chain shortening products, that is, 20:1n-11, 18 : 1n-11 and 16 : 1n-11. Indeed, and without any difference in the dietary supply of them, these fatty acids were clearly elevated along with the increasing proportions of herring in the diet of the gulls of this study. Thus, in addition to fatty acids themselves, their metabolically modified products can indicate dietary changes. Similar accumulation of the chain shortening products instead of the originally consumed 22 : 1n-11, abundant in diet, has been found for piscivorous mammals (Cooper, Iverson & Heras 2005; Cooper, Iverson & Rouvinen-Watt 2006). The 14 : 0, 18 : 3n-3 and 18 : 4n-3, derived from phytoplankton via several trophic steps (Ackman, Tocher & McLachlan 1968; Rossi et al. 2006), have less such biochemical restrictions, and they responded in a linear way to the increased proportion of herring in the diet, and were excellent markers.

Minor branched- and odd-chain saturated fatty acids (SFA), especially the iso- and anteiso- forms of 17 : 0, indicated, although not very strongly, the presence of demersal fish in the diet of the gulls. The branched-chain SFA are produced not only by demersal microbial communities, that provide bottom feeding fish with these specific fatty acids, but also by intestinal (Gram-positive) microbes of birds and other vertebrates (Rütters et al. 2002; Stoeck et al. 2002; Or-Rashid, Odongo & McBride 2007; Rehman et al. 2007). Despite this, the larger supply of the branched- and odd-chain SFA from demersal fish diet was not masked by microbial lipids originating from the gulls’ own digestive tract. In addition, the preference of plasma lipids to incorporate n-6 PUFA probably helped to transfer the slightly higher status of minor fatty acids 20 : 2n-6 and 22 : 4n-6 from demersal fish to the plasma of birds fed these fish.

In summary, it is important to note that the metabolism of different fatty acids forming plasma FAS is different. Endogenous metabolism can mask dietary differences or changes for many major fatty acids but this study shows that at the same time several minor ones (e.g. in this experiment, 14 : 0, 18 : 3n-3, 18 : 4n-3, and C20–22 MUFA with their chain shortening products), quantified accurately by GLC, carry dietary information in a quantitative manner. The metabolic modifications can be sufficiently elucidated in representative calibration experiments, and the FAS of plasma total lipids can be used to detect recent diets of seabirds when adipose tissue samples, representing a larger, and probably much earlier, time window, cannot be collected. In simple cases, good quantitative estimations can be achieved, but in complex marine food webs the task is challenging and requires large data-bases of dietary items covering individuals of different age and size, collected at different times. However, the method is an excellent tool for observing temporal or spatial differences in the diet of seabird populations and may provide information on changes in marine food webs and oceanographic variations.


Authors thank Ulrike Strauß, Adolf Völk and Johannes Wieland for preparing the fish diets. This work was supported by EU contract Q5RS-2001-00839 ‘DISCBIRD’.