Important impacts of tissue selection and lipid extraction on ecological parameters derived from stable isotope ratios


Correspondence author. E-mail:


  1. The nitrogen (δ15N) and carbon (δ13C) isotope ratios of animal tissues can help identify the composition of diets and open up a myriad of ecological applications. However, consumers do not ingest or assimilate all components of food items, and it is not well understood how sampling different tissues of sources and consumers may affect isotopic values ascribed, and thereby how such variation affects derived ecological measures.
  2. Utilizing a simple prey–predator feeding relationship in insects, we examined isotopic differences in soft, exoskeleton and whole tissues using samples with and without lipid extraction. As a derived ecological measure, we calculated trophic discrimination factors, changes in δ15N or δ13C between source and consumer, for the different prey–predator tissue combinations.
  3. Lipid extraction did not affect δ15N values, and we found significant tissue differences in δ15N that varied between prey and predator. Lipid extraction enriched δ13C values in most instances, and it was only after extraction of lipids that we observed consistent depletion of δ13C in exoskeleton relative to soft tissues in prey and predator.
  4. Isotopic differences between tissue types propagated marked variation in derived ecological parameters. Common sampling practice using whole tissue for prey and predator (whole/whole) resulted in a trophic discrimination factor of 0·48‰ for δ15N, compared with correct factors of 0·97‰ (soft/whole) and 2·18‰ (soft/soft) using prey soft tissue actually ingested by the predator. For δ13C, variation across discrimination factors was less, with whole/whole tissue of −0·14‰, whilst correct factors were −0·55‰ (soft/whole) and −0·04‰ (soft/soft).
  5. Our results indicate that tissue selection and preparation are important considerations for isotopic studies using arthropods. Lipid extraction is necessary to derive accurate δ13C values based on proteins, whilst consequences of tissue selection are likely context-dependent: In poorly defined systems where sources are isotopically similar or have larger variance, our results indicate that tissue selection within sources is important to avoid significant error, whether estimating trophic positions or dietary proportions using mixing models. In such cases, we strongly recommend exclusion of source materials not assimilated in consumers.


Stable isotope ratios of nitrogen (N15/N14, termed δ15N) and carbon (C13/C12, termed δ13C) are increasingly used for testing a broad suite of ecological theories. Isotopic data are frequently employed as a characterization of dietary niche (Newsome et al. 2009), with δ15N and δ13C signatures in consumer protein allowing inference of prey sources (DeNiro & Epstein 1978, 1981). Typically, δ15N enriches from food source to consumer, and although this enrichment is variable (e.g. 0·6–5·4‰; Post 2002), average δ15N enrichment of between 2·5‰ and 3·4‰ (Post 2002; Vanderklift & Ponsard 2003; Caut, Angulo & Courchamp 2009) is commonly used for the estimation of species' trophic positions (Post 2002) and food chain length (Vander Zanden & Fetzer 2007). Conversely, δ13C typically enriches by <1‰ from diet to consumer (Post 2002; Caut, Angulo & Courchamp 2009), and because δ13C often varies between basal resources, δ13C is generally used to trace prey–consumer interactions or food chains (Post 2002). Thus, changes in δ15N and δ13C between the food source and its consumer, termed discrimination factors (Martinez del Rio et al. 2009), can allow for the elucidation of trophic relationships, or for inference of diet-related mechanisms driving ecological or evolutionary processes (Post 2002; Bearhop et al. 2005; Vander Zanden & Fetzer 2007). Combining δ15N and δ13C as an investigative tool, isotopic dietary information has been instrumental in developing empirical understanding in a number of research areas, including trophic relationships (Syvaranta & Jones 2008; Newsome et al. 2009), dietary reconstruction using mixing models (Moore & Semmens 2008; Parnell et al. 2010), population niches (Bearhop et al. 2004; Newsome et al. 2007) and community food web structure (Layman et al. 2007, 2011; Jackson et al. 2011). However, to improve research in these areas, hypothesis testing requires isotopic data that accurately reflect dietary pathways, and so, it is important that sampling protocols for providing δ15N and δ13C values are appropriate to the questions being tested (Martinez del Rio et al. 2009; Boecklen et al. 2011).

Many factors are known to affect δ15N and δ13C values, some of which are well studied, including age, size, diet quality, habitat, season, trophic position, consumer's nutritional state and mode of excretion (Caut, Angulo & Courchamp 2009; Martinez del Rio et al. 2009; Boecklen et al. 2011 and references therein) and where appropriate, researchers can control for these factors. One factor that has not been thoroughly considered is the extent to which different tissue types within an individual differ in their δ15N and δ13C values. Such tissue differences can occur because of the following: differing amino acid structure of tissues (Martinez del Rio et al. 2009), differing protein turnover rates amongst tissues (Tieszen et al. 1983; Arneson, MacAvoy & Basset 2006), differential metabolic routing of nutrients to tissues (Voigt et al. 2008) and ontogenetic tissue synthesis coupled with temporal or ontogenetic diet shifts (O'Brien, Boggs & Fogel 2005).

Tissue-specific sampling is routinely practised for vertebrate studies, including fish (Perga & Gerdeaux 2005), birds (Bearhop et al. 2002) and mammals (DeNiro & Epstein 1981; Tieszen et al. 1983). In contrast, research has rarely explored tissue sampling of invertebrates (Vanderklift & Ponsard 2003), in particular arthropods (Caut, Angulo & Courchamp 2009; Boecklen et al. 2011), which are a major component of most food webs. Tissue-specific sampling for arthropods can be problematic and laborious, and current common practice utilizes whole tissue (Vanderklift & Ponsard 2003; Mateo et al. 2008; Caut, Angulo & Courchamp 2009 and references therein for all). Consequently, most literature estimates of arthropod discrimination factors, utilized to reconstruct diet and calculate subsequent ecological parameters, are based on whole tissue. In reviews of laboratory arthropod studies, Vanderklift & Ponsard (2003) and Caut, Angulo & Courchamp (2009) report 82 estimates of δ15N discrimination factors from 18 studies and 73 estimates of δ13C discrimination factors from 14 studies, in which all but one study (five estimates of δ15N only) utilized whole tissue.

Arthropod anatomy is crudely characterized as soft internal tissues (broadly composed of proteins, sugars and fats) contrasting with a hardened exoskeleton largely constructed of chitin embedded with protein (scleratin), which in aquatic species also often contains inorganic carbon in the form of CaCO3. Given such differences in tissue composition, in isotopic studies, tissue selection is likely to be important because many consumers feed selectively on the prey they capture; >70% predatory terrestrial arthropods ingest soft internal components but not cuticles of arthropod prey (Cohen 1995). Thus, the inclusion of exoskeleton, inherent in measures of whole tissue, has the potential to introduce error into estimates of δ15N and δ13C in instances of selective assimilation by consumers.

Few studies have explicitly compared δ15N and δ13C in arthropod tissues, especially between soft and exoskeleton tissues (Macko et al. 1989; Tibbets, Wheeless & Martinez del Rio 2008), and it remains untested how tissue selection may lead to erroneous estimates of subsequent isotopic measures such as discrimination factors, which are key to dietary reconstruction. It is therefore of importance to establish when and whether tissue differences should be considered sufficient to change the way we sample and process arthropods prior to generating isotopic data.

Additionally, to accurately estimate δ13C of proteins within tissues, it is accepted practice to first remove free lipid contained within. Lipid is naturally depleted in δ13C (DeNiro & Epstein 1977), and its concentration varies between tissues. The extraction process is well studied but has produced mixed results (Pinnegar & Polunin 1999; Sweeting, Polunin & Jennings 2006; Bodin, Le Loc'h & Hily 2007). Failure to extract lipids can lead to erroneous conclusions (Tarroux et al. 2010), but it is currently underutilized (Mateo et al. 2008) likely because of the lack of clarity about when it is needed (Post et al. 2007). It has been shown that some invertebrates contain significant concentrations of lipid (Meier, Meyer & Meyns 2000), but to date, few studies have considered tissue-specific effects of extraction across multiple tissues (Pinnegar & Polunin 1999; Sweeting, Polunin & Jennings 2006; Logan & Lutcavage 2008), particularly in arthropods (Bodin, Le Loc'h & Hily 2007; Mateo et al. 2008 and references therein).

In this study, we addressed three questions: (i) How consistent are differences in δ15N or δ13C signatures of whole, exoskeleton and soft tissue? (ii) How important are tissue-specific differences in δ15N or δ13C in explaining differential estimations of discrimination factors for δ15N and δ13C? (iii) How does lipid extraction differentially alter δ15N or δ13C signatures of whole, exoskeleton or soft tissue?

Materials and methods

Herbivorous grain aphids (Sitobion avenae) of a common stock and predatory 1st generation larvae of wild-caught hoverfly (Syrphus vitripennis) yielded soft and exoskeleton tissue to test for differences in δ15N and δ13C between tissue types within each species. We reared aphids on two independent food plants; one based on a C3 photosynthetic pathway (wheat Triticum aestivum) and the other on a C4 pathway (maize Zea mays), enabling separation of the plants on a δ13C axis and thus broadening generality of any observed patterns. Aphids were raised under a 16 : 8 light/dark cycle in 70% humidity. Plants were raised on a common source of homogenized compost and distilled water, and introduced to aphids at 20 days (wheat) or 30 days (maize). Randomly collected aphids of all ages were frozen (−20 °C) for later dissection. 15 gravid wild-caught hoverfly females were induced to lay eggs in the laboratory and emergent larvae randomly assigned to feed on either wheat or maize aphids, under a 16 : 8 light/dark cycle at 70% humidity. Hoverfly larvae entered pupation 8–10 days after hatching and after 72 h pupation were frozen (−20 °C) for later dissection. Prior experimentation identified 72 h pupation as suitable to provide exoskeleton tissue in the form of exuviae, whilst terminating pupation before larval metamorphosis was judged to have sufficiently altered body form and thus potentially caused significant shifts in δ15N and δ13C of tissues. Notably, Tibbets, Wheeless & Martinez del Rio (2008) showed δ15N of whole tissue for half pupated Diptera to not differ from larvae. 72 h pupation for our hoverflies represented <one-third total pupation time.

Tissue Preparation and Lipid Extraction

For each of 15 replicates; 80 pooled aphids were dissected into soft and exoskeleton components to provide enough material for two lipid treatments, one with lipids present (+L) and one with lipids extracted (−L), whilst a separate sample of 80 pooled aphids provided whole tissue for each lipid treatment. For hoverflies, we collected 15 replicates of individual larvae, each of which provided four tissue treatments; having split pupae and separated soft and exoskeleton (exuviae) tissues, for each larva, half of the soft and exoskeleton tissue underwent lipid extraction. Thus, there were six tissue treatments for aphids: (i) soft+L, (ii) soft−L, (iii) exoskeleton+L, (iv) exoskeleton−L, (v) whole+L, (vi) whole−L; and four tissue treatments for hoverfly (i–iv). All samples were dried at 45 °C for >48 h and then homogenized. For samples undergoing lipid extraction, tissue was subsequently immersed in 2 : 1 chloroform/methanol solution for 50 min to remove free lipid, and then left to air dry.

Stable Isotope Analysis

For all samples, 0·5 mg ± 0·05 dried material was enclosed in tin capsules. Stable isotope analysis (SIA) was conducted at the Food and Environment Research Agency, York, UK. Samples were analysed for δ15N and δ13C in a Fisons EA1108 elemental analyser (Carlo Erba Instruments, Milan, Italy), coupled with an Isoprime isotope ratio mass spectrometer (GV Instruments, Manchester, UK). Stable isotope ratios are reported in delta (δ) notation where δ15N and δ13C = [(Rsample/Rstandard)−1] × 1000, where R is 15N/14N or 13C/12C. Isotope ratios are expressed in per mil (‰) relative to the ratio of international reference standards (Rstandard) which are atmospheric nitrogen and Vienna PeeDee Belemnite (VPDB) for nitrogen and carbon, respectively. Measures of standards placed throughout samples exhibited acceptable instrument reproducibility of <0·09‰ (SD) for δ15N and <0·18‰ (SD) for δ13C using collagen standard, insect whole tissue standard (cockroach; Nauphoeta cinerea) and sucrose C4 plant standard.

Data Analysis

Analyses were employed separately for aphids and hoverflies using general linear mixed models (GLMM) to test how response variables δ15N or δ13C were affected by explanatory variables food chain (levels = wheat or maize), tissue (levels = soft, exoskeleton or whole) and lipid extraction [levels = lipid present (+L) or lipid extracted (−L)]. The random effect replicate was incorporated to account for non-independence of paired tissue samples (soft+L, exoskeleton+L, soft−L, exoskeleton−L,) taken from each replicate. Analyses were conducted using R version 2.14.1 (R Development Core Team 2011) using ‘lme’ from the nlme package (Pinheiro et al. 2010). Model simplification used backwards stepwise regression from a maximal model and anova model comparisons to identify non-significant model terms for elimination. Homogeneity of variances and normality of model residuals were checked in all instances.

Differences between tissue types were calculated within each paired sample. For aphids only, whole tissue replicates were paired at random with soft and exoskeleton treatment replicates from which they were independent. For all pairings, = 15. Mean ± SD for each tissue comparison was calculated on these 15 pairings.

Mean ± SD of δ15N and δ13C discrimination factors was calculated from differences between species (i.e. hoverfly whole–aphid soft), based on bootstrap resampling using all replicate pairings (i.e. 15 × 15, = 225), for each tissue combination. Hoverfly whole tissue δ15N and δ13C values were estimated as proportional distances between soft and exoskeleton for each isotope, based on the mass/balance ratio of dry hoverfly exoskeleton to soft tissue (average = 22%, 78%, respectively) for each of the 15 replicates, producing average values of −2·3‰ ± 0·5‰ for δ15N and −31·0‰ ± 0·2‰ for δ13C. This method was validated as suitable through comparing mass/balance calculated aphid whole tissue values (using aphid exoskeleton–soft tissue; average = 49%, 51%, respectively) against known aphid whole values and finding no significant difference. Discrimination factors were calculated using wheat food chain, lipid-extracted aphid and hoverfly data only.

Literature Review

In addition to our empirical study, to assess consistency of any differences in δ15N or δ13C signatures between whole, exoskeleton and soft tissues, we also conducted a literature review to collate all estimates of isotopic values within these tissues in arthropods. We identified 11 studies reporting 26 estimates of δ15N and 18 estimates of δ13C for combinations of soft, whole and exoskeleton (or chitin extract) tissues, across 22 species of arthropods.


Lipid Extraction & Tissue Selection

Whilst δ13C models had a greater number of significant terms than δ15N, patterns for either δ15N or δ13C were consistent across species (Table 1).

Table 1. Effects of tissue type, food chain and lipid extraction on aphid and hoverfly δ15N and δ13C (‰). Four separate GLMMs were used
Full model termsδ15N Aphidδ15N Hoverflyδ13C Aphidδ13C Hoverfly
  1. Test statistic is chi-square in all instances, with degrees of freedom in brackets. Significant effects are indicated by ***, with < 0·001 in all instances.

Tissue13·62 (2)***388·19 (1)***68·05 (4)***277·06 (2)***
Food chain100·52 (1)***116·93 (1)***297·21 (2)***244·73 (2)***
Lipid extraction0·02 (1)2·79 (1)482·82 (4)***245·65 (3)***
Tissue × lipid extraction0·64 (2)1·60 (1)58·72 (2)***173·68 (1)***
Food chain × lipid extraction0·14 (1)3·70 (1)62·19 (1)***16·07 (1)***
Food chain × tissue0·41 (2)0·01 (1)1·47 (2)2·31 (1)
Food chain × lipid extraction × tissue2·51 (2)0·05 (1)3·19 (2)0·01 (1)

Lipid Extraction


Lipid extraction did not effect δ15N of either aphid or hoverfly tissues (Table 1).


For both species, significant interactions between lipid extraction and tissue type (Table 1) indicated tissue-specific effects on δ13C were dependent upon the application of lipid extraction. Broadly across both species, δ13C of tissues was significantly enriched by lipid extraction, and these differences were large (range 1·3–2·9‰, Fig. 1, Table 2).

Table 2. Mean ± SD (‰) of difference between tissue treatments for δ15N and δ13C, based on paired samples = 15. Values are given in reference to CAPITALIZED tissue. +L = with lipid, −L = lipid removed
δ15N (‰)AphidHoverfly
Lipid extraction
SOFT (−L) – soft (+L)−0·16 ± 0·670·20 ± 0·80−0·23 ± 0·570·14 ± 0·56
EXOSKELETON (−L) – exoskeleton (+L)−0·03 ± 0·49−0·02 ± 0·710·04 ± 1·97−0·11 ± 0·55
WHOLE (−L) – whole (+L)0·33 ± 0·52−0·11 ± 0·67  
Tissue differences (post-extraction)
SOFT – exoskeleton−0·21 ± 0·39−0·54 ± 0·695·22 ± 0·705·19 ± 0·80
SOFT – whole−0·08 ± 0·74−0·47 ± 1·33  
WHOLE – exoskeleton−0·14 ± 0·62−0·07 ± 1·12  
δ13C (‰)
 Lipid extraction
SOFT (−L) – soft (+L)2·90 ± 0·362·02 ± 0·392·07 ± 0·181·73 ± 0·33
EXOSKELETON (−L) – exoskeleton (+L)1·94 ± 0·311·35 ± 0·280·18 ± 0·22−0·15 ± 0·54
WHOLE (−L) – whole (+L)2·42 ± 0·331·83 ± 0·47  
 Tissue differences (post-extraction)
SOFT – exoskeleton0·60 ± 0·240·54 ± 0·302·28 ± 0·252·16 ± 0·40
SOFT – whole0·40 ± 0·380·24 ± 0·44  
WHOLE – exoskeleton0·20 ± 0·270·30 ± 0·33  
Figure 1.

Mean ± SD (‰) of δ15N and δ13C of different tissue types and lipid treatments, shown by species and food chain. +L = with lipid, −L = lipid removed. = 15 in all instances. Exo = exoskeleton.

For aphids, all tissues were significantly enriched in δ13C after the extraction of lipids. A significant interaction between lipid extraction and tissue type indicated that the effects of lipid extraction differed in strength between tissue types: soft > whole > exoskeleton. This pattern was consistent across food chains (Fig. 1, Table 2). Similarly for hoverflies, a significant interaction between tissue type and lipid extraction showed extraction effects on δ13C were dependent on tissue type (Table 1). Figure 1 and Table 2 show that hoverfly soft tissue was strongly enriched in δ13C as a result of the extraction process (average across food chains = 1·9‰) but that hoverfly exoskeleton was not.

For both aphid and hoverfly, whilst overall patterns of tissue response to lipid extraction were consistent for both food chains, the significant interaction between lipid extraction and food chain indicated the magnitude of tissue δ13C enrichment following lipid extraction was greater for tissues on the wheat than maize food chain (average enrichment across tissues: wheat = 2·3‰ ± 0·4‰, maize = 1·7‰ ± 0·3‰; Table 2).

Importantly, differences in δ13C between tissue types were only detectable after lipid extraction (Fig. 1, Table 2).

Tissue Selection


In both the aphid and hoverfly models, tissue type and food chain significantly influenced δ15N (Table 1). There were no significant interaction terms. Tissue effects on δ15N differed between aphid and hoverflies. For the aphid model, significant but very small tissue effects showed that soft tissue δ15N was 0·1–0·5‰ less than either exoskeleton or whole tissue (Fig. 1 and Table 2). Conversely, hoverfly soft tissue was significantly enriched in δ15N relative to exoskeleton, and this difference was large, with a mean difference consistent across food chains of 5·2‰ (Fig. 1 and Table 2).

Literature-reviewed estimates of δ15N consistently showed that soft > whole > exoskeleton or chitin (Table 3). δ15N of soft tissue was enriched relative to exoskeleton on average by 5·9‰ ± 0·6‰ (one study, three estimates), whilst estimates of whole tissue were enriched relative to exoskeleton on average by 4·1‰ ± 2·5‰ (five studies, 13 estimates). These findings concur with our hoverfly observations but contrast our aphid observations.

Table 3. Survey of literature that explicitly reports δ15N and δ13C (‰) values across different soft and exoskeleton tissue components in Arthropods
ReferencesArthropoda PhylumClassCommon nameHabitatTissue 1Tissue 2δ15N Difference (=Tissue 1–2)δ13C Difference (=Tissue 1–2)Notes
  1. *Non-acid washed estimates used; **Exoskeletons obtained by rinsing with KOH.

  2. To the best of our knowledge, prior to stable isotope analysis of δ13C; (a) lipid extraction stated as having been conducted; (b) lipid extraction not stated and (c) correction factor applied.

Macko et al. (1989)CrustaceaMalacostracaLobsterAquaticSoft (muscle)Exoskeleton (carapace)6·670·61b
Macko et al. (1989)CrustaceaMalacostracaBrown ShrimpAquaticSoft (muscle)Exoskeleton (carapace)5·460·36b
Macko et al. (1989)CrustaceaMalacostracaTiger ShrimpAquaticSoft (muscle)Exoskeleton (carapace)5·680·79b
Macko et al. (1989)CrustaceaMalacostracaLobsterAquaticSoft (muscle)Chitin Extract7·501·23b
Macko et al. (1989)CrustaceaMalacostracaBrown ShrimpAquaticSoft (muscle)Chitin Extract5·641·57b
Macko et al. (1989)CrustaceaMalacostracaTiger ShrimpAquaticSoft (muscle)Chitin Extract7·181·53b
Macko et al. (1989)CrustaceaMalacostracaMantan ShrimpAquaticSoft (muscle)Chitin Extract10·681·51b
Montoya, Wiebe & McCarthy (1992)CrustaceaMalacostracaAmphipodAquaticWholeExoskeleton (carapace)3·80 
Montoya, Wiebe & McCarthy (1992)CrustaceaMalacostracaAmphipodAquaticWholeExoskeleton (carapace)4·60 
Currin, Newell & Paerl (1995)CrustaceaMalacostracaCrabAquaticSoft (muscle/gills)Whole1·0/0·5−0·5/−1·2b
Yokoyama et al. (2005)CrustaceaMalacostracaGhost Shrimp AAquaticSoft (muscle)Whole1·20−1·00b*
Yokoyama et al. (2005)CrustaceaMalacostracaGhost Shrimp AAquaticWholeExoskeleton3·30−3·80b*
Yokoyama et al. (2005)CrustaceaMalacostracaGhost Shrimp BAquaticSoft (muscle)Whole0·90−0·70b*
Yokoyama et al. (2005)CrustaceaMalacostracaGhost Shrimp BAquaticWholeExoskeleton3·50−1·60b*
Perga (2010)CrustaceaBranchiopoda Bosmina AquaticWholeExoskeleton−0·80c**
Perga (2010)CrustaceaBranchiopoda Daphnia AquaticWholeExoskeleton7·901·40c**
Perga (2011)CrustaceaBranchiopoda Daphnia AquaticWholeExoskeleton9·00 **
DeNiro & Epstein (1978)UniramiaInsectaG'hopper/Beetle/BugTerrestrial Diet Chitin Extract−0·1 to −0·7b
DeNiro & Epstein (1981)UniramiaInsectaGrasshopperTerrestrial Diet Chitin Extract6·60 
DeNiro & Epstein (1981)UniramiaInsectaMilkweed BugTerrestrial Diet Chitin Extract8·60 
Webb, Hedges & Simpson (1998)UniramiaInsectaLocust (on diet A)TerrestrialSoft (muscle)Chitin Extract6·801·50a
Webb, Hedges & Simpson (1998)UniramiaInsectaLocust (on diet B)TerrestrialSoft (muscle)Chitin Extract12·000·90a
Gratton & Forbes (2006)UniramiaInsectaBeetleTerrestrialSoft (various)Exoskeleton (various)+0·3 to −0·8b
Tibbets, Wheeless & Martinez del Rio (2008)UniramiaInsectaSilkworm MothTerrestrialWhole (larvae)Exoskeleton (exuviae)1·30 
Tibbets, Wheeless & Martinez del Rio (2008)UniramiaInsectaWax MothTerrestrialWhole (larvae)Exoskeleton (exuviae)1·70 
Tibbets, Wheeless & Martinez del Rio (2008)UniramiaInsectaTobacco MothTerrestrialWhole (larvae)Exoskeleton (exuviae)−1·00 
Tibbets, Wheeless & Martinez del Rio (2008)UniramiaInsectaButterflyTerrestrialWhole (larvae)Exoskeleton (exuviae)0·50 
Tibbets, Wheeless & Martinez del Rio (2008)UniramiaInsectaFlesh FlyTerrestrialWhole (larvae)Exoskeleton (exuviae)3·90 
Tibbets, Wheeless & Martinez del Rio (2008)UniramiaInsectaBeetleTerrestrialWhole (larvae)Exoskeleton (exuviae)−1·20 


Following lipid extraction, across species and food chains, consistent significant differences in δ13C between tissue types were found (Table 1). Soft tissue was significantly enriched in δ13C relative to exoskeleton, with the magnitude greater in hoverflies (0·6‰ and 2·2‰ for aphids and hoverflies, respectively, averaged across food chains). Additionally for aphids, the pattern of tissue differences in δ13C was as follows: soft > whole > exoskeleton (Fig. 1, Table 2).

For reviewed studies, patterns of δ13C between tissue types were inconsistent (Table 3), contrasting with the consistency we observed across aphids and hoverflies.

Tissue-Specific Trophic Discrimination Factors

In our study, differences between source and consumer tissues propagated notable variation in the derived estimates of discrimination factors (Table 4).

Table 4. Mean ± SD (‰) discrimination factors for δ15N and δ13C (calculated as hoverfly predator–aphid prey) for all prey–predator tissue combinations. All values based on bootstrap resampling (= 225). Common sampling practice whole tissue (whole/whole) discrimination factor is denoted *. Correct factors based on prey soft tissue actually assimilated by predators in our experiment are marked with **(soft/whole) and ***(soft/soft). Values based on lipid-extracted wheat food chain data
Tissue combinationDiscrimination factor
Aphid tissueHoverfly tissueMean δ15NMean δ13C
SoftSoft***2·18 ± 0·97−0·04 ± 0·27
WholeSoft1·69 ± 0·830·36 ± 0·29
ExoskeletonSoft1·84 ± 0·760·56 ± 0·25
SoftWhole**0·97 ± 0·92−0·55 ± 0·26
WholeWhole*0·48 ± 0·78−0·14 ± 0·28
ExoskeletonWhole0·63 ± 0·700·05 ± 0·24
SoftExoskeleton−3·2 ± 1·06−2·32 ± 0·36
WholeExoskeleton−3·69 ± 0·94−1·91 ± 0·37
ExoskeletonExoskeleton−3·55 ± 0·87−1·72 ± 0·34


For common sampling practice, where whole tissue is used for prey and predator (whole/whole), a discrimination factor of 0·48‰ for δ15N was obtained, compared with correct factors of 0·97‰ (soft/whole) and 2·18‰ (soft/soft) based on prey soft tissue actually ingested by the predator.

For δ15N, overall range in observed discrimination factors for all source–consumer tissue combinations extended from enrichment to depletion (2·18‰ to −3·69‰). The upper and lower boundaries of these discrimination factors were determined by the differences in hoverfly soft and exoskeleton tissues (Table 4).


For δ13C, the differences between discrimination factors were less, with common sampling practice tissues (whole/whole) of −0·14‰, whilst correct factors were −0·55‰ (soft/whole) and −0·04‰ (soft/soft).

For δ13C, overall range in observed discrimination factors for all source–consumer tissue combinations also extended from enrichment to depletion (0·56‰ to −2·32‰). The upper and lower boundaries of these discrimination factors were also determined by the differences in hoverfly soft and exoskeleton tissues (Table 4).


We measured for the first time tissue-specific trophic discrimination factors in an arthropod predator–prey system and, perhaps not surprisingly, our results show that estimates of trophic discrimination factors can be markedly affected by tissue selection. In our literature review, we additionally identify strong evidence that large tissue differences are frequent across arthropod taxa, suggesting that tissue effects on trophic discrimination factors, such as we have demonstrated, may also be frequent. Tissue differences as shown in our findings may explain some of the variation around commonly used average trophic discrimination factors taken from the literature (Post 2002; Vanderklift & Ponsard 2003; Caut, Angulo & Courchamp 2009), and we speculate that such variation is actually error variation due to inappropriate source tissue selection when consumers feed selectively. Given some of the observed error variation in trophic discrimination factors is likely large enough to affect subsequent ecological conclusions, consideration of source tissue selection to best represent assimilation in consumers is therefore of importance in isotopic ecology more generally. This is particularly the case for arthropods given that many consumers only feed on arthropod soft tissue components (Cohen 1995) although a majority of studies presently use whole tissues (Vander Zanden & Rasmussen 2001; Vanderklift & Ponsard 2003; Caut, Angulo & Courchamp 2009). We acknowledge arthropod tissue sampling may be laborious or difficult, but to improve future ecological conclusions derived from isotopic data, we therefore recommend the use of arthropod soft tissues to best represent dietary sources in consumers that do not assimilate exoskeleton.

How Consistent are Differences in math formula15N or math formula13C Signatures of Whole, Exoskeleton and Soft Tissue?


We found marked differences in δ15N between exoskeleton and soft tissues, which varied between species. Significant enrichment of aphid exoskeleton (and whole tissue) relative to soft tissue was very small and variable between food chains (mean δ15N = 0·3‰), whilst conversely, significant depletion of hoverfly exoskeleton relative to soft tissue was large and consistent between food chains (mean δ15N = 5·2‰). Such contrast in the direction, variability and magnitude of this exoskeleton – soft tissue δ15N relationship suggests species-specific tissue compositions. Although significant, small enrichment of exoskeleton over soft tissues for aphids in our study is of doubtful ecological importance, although more generally this lack of depletion in exoskeleton relative to soft tissue is uncommon in the literature (Tibbets, Wheeless & Martinez del Rio 2008; and see Table 3) and the mechanisms are poorly defined. Conversely, depletion of δ15N in hoverfly exoskeleton (exuviae) concurs with limited available results showing depletion of 0·5–3·9‰ in insect exuviae relative to whole larvae tissue, as a consequence of high chitin content in the exoskeleton (Tibbets, Wheeless & Martinez del Rio 2008). Notably, the magnitude of depletion shown for hoverfly exoskeleton relative to soft tissues also concurs closely with that shown for aquatic crustaceans (Macko et al. 1989; Yokoyama et al. 2005) and is less depleted than some others ≈8–9‰ (Perga 2010, 2011; Table 3). Therefore, whilst insect exuviae have rarely been utilized in isotopic studies elsewhere, given tissue differences identified by our review in Table 3, exuviae is hence not unrepresentative of arthropod exoskeleton more generally. Our results and those identified by our review show that large differences exist in δ15N between component arthropod tissues known to be differentially assimilated or avoided by consumers. We therefore recommend researchers use only source tissues assimilated by consumers, and when consumer feeding habits are unknown, urge caution on the inclusion of exoskeleton in arthropod prey.


Across species, we found significant δ13C depletion in exoskeleton relative to soft tissue, with the magnitude of tissue differences being greater in hoverflies than aphids (mean = 2·2‰ and 0·6‰, respectively). Our results fall within a broader and less consistent arthropod literature (reviewed in Table 3). Such divergent literature results likely represent some species-level tissue differences in δ13C, although it is noteworthy that only Webb, Hedges & Simpson (1998) records utilizing lipid extraction on samples prior to stable isotope analysis. In our study, enrichment of soft tissue δ13C relative to exoskeleton was only apparent after extraction of lipids.

How Important are Tissue-Specific Differences in δ15N or δ13C in Explaining Differential Estimations of Discrimination Factors for δ15N or δ13C?


We found marked variation in δ15N estimates of discrimination factors across different consumer–prey tissue combinations. Such variation was more influenced by changing hoverfly tissues than aphid. Observed variation in δ15N discrimination factors between common sampling practice whole tissues of 0·48‰ contrasted with correct factors based on soft tissues of 0·97‰ (soft/whole) and 2·18‰ (soft/soft). Importantly, this result demonstrates that the error inherent in using whole tissue δ15N estimates, when consumers do not assimilate exoskeleton, is of notable magnitude. This strongly suggests that tissue selection requires consideration to avoid propagating such error when utilizing discrimination factors to quantify food chain length, trophic positions of consumers or source contributions to consumer diets using mixing models, for instance. By providing empirical estimates of tissue effects on discrimination factors in this study, our findings develop the literature reviewed in Table 3 which show large δ15N differences between soft and exoskeleton tissue components, but crucially do not directly compare these with whole tissues. Such a comparison between soft, exoskeleton and whole tissue, as in our study, is necessary to understand whether whole tissue can comprise enough exoskeleton material by mass, that exoskeleton differences from soft tissue can notably affect whole tissue δ15N. Tissue-specific variation in discrimination factors as shown in this study demonstrates that this can be the case.

Widely cited review studies of invertebrate δ15N discrimination factors (collectively >1400 citations on Google Scholar) are composed of >90% arthropods, of which >95% utilize whole tissue (Vander Zanden & Rasmussen 2001; Vanderklift & Ponsard 2003; Caut, Angulo & Courchamp 2009). We therefore speculate that reviewed δ15N whole tissue values comprise exoskeleton tissues that, if used as discrimination factors to parameterize sources in isotopic models in subsequent studies, will constitute error when consumers do not assimilate exoskeleton. In instances where such error is significant, this will propagate and affect subsequent trophic estimates and ecological conclusions.


Effects of tissue type on δ13C discrimination factors were less than those of δ15N, with a common sampling practice whole tissue discrimination factor of −0·14‰, compared with correct soft tissue discrimination factors of −0·55‰ and −0·04‰. More generally, smaller variation in estimates of δ13C than δ15N discrimination factors is a consequence of lesser fractionation in δ13C than δ15N from prey to predator, as well-established in the literature (DeNiro & Epstein 1978, 1981). This causes smaller differences in δ13C than δ15N both within and between aphid and hoverfly tissues and hence smaller differences in δ13C than δ15N discrimination factors. Smaller observed differences in δ13C discrimination factors may seem of lesser consequence, but such differences must be considered relatively; in trophic systems where δ13C ranges are narrow (i.e. many food chains), small tissue differences in δ13C may still be important in affecting overall conclusions, for instance in discerning between sources in food chains or mixing models (Post et al. 2007; Tarroux et al. 2010). Thus, we would urge caution in the use of source tissues that are known to not be consumed and advise that researchers consider their context of use.

How does Lipid Extraction Differentially Alter δ15N or δ13C Signatures of Whole, Exoskeleton or Soft Tissue?

Extraction of lipids did not significantly detrimentally affect δ15N of aphid or hoverfly tissues. In contrast, significant enrichments in δ13C were recorded after extraction for soft, whole and exoskeleton tissues of aphids and soft tissue of hoverflies (means range 1·7–2·5‰). The direction and magnitude of our results concur with other limited empirical evidence from arthropods (Bodin, Le Loc'h & Hily 2007; Logan et al. 2008). The exception in our study was hoverfly exoskeleton, which showed no δ13C enrichment, in contrast to aphid exoskeleton. This apparent discrepancy is likely explained by differing exoskeleton composition; for example, insect cuticles vary in their configurations of proteins (notably sclerotin) and lipid as a function of cuticle rigidity and waterproofing (Wigglesworth 1970, 1985). Many studies likely assume insect exoskeleton to be composed largely of chitin without significant lipid, and thus without a need to be subjected to lipid treatment. This is clearly not the case with aphid exoskeleton, and thus, important species-specific differences in exoskeleton structure may necessitate treatment to remove significant lipid components. Notably, a significant interaction between food chain and lipid extraction showed extraction effects to be greater for aphid and hoverfly tissues on the wheat than maize food chain. We speculate such an effect is a consequence of differences in nutritional content between primary producers being propagated to consumers, as shown elsewhere (Wilson, Sternberg & Hurley 2011), with proportionally more lipid derived by aphids feeding on wheat plants than maize. This result infers that lipid extraction effects were mediated by diet. Importantly, more generally, the magnitude of effects of lipid extraction on δ13C observed in our results fall within the range (>2‰) of those shown to have potentially significant consequences for deriving subsequently spurious ecological conclusions (Tarroux et al. 2010) suggesting that lipid in insect samples must be accounted for.

In conclusion, our results show that significant differences exist between component tissues of arthropods that are known to be selectively assimilated by consumers and that such differences propagate notable error variation between discrimination factors. This will affect subsequent trophic measures and potentially ecological conclusions. Currently, tissue selection–based error is not accounted for in a majority of isotopic studies that use arthropods. Implications of this study for practitioners of isotopic studies are best interpreted in the context that researchers intend to use them; in poorly defined systems where sources are isotopically similar or have larger variance, our results indicate that tissue selection within sources is important to avoid significant error, whether estimating trophic positions or dietary estimates using mixing models. When researchers are without prior knowledge, we recommend only using arthropod soft tissue components and excluding exoskeleton material. We also conclude that lipid extraction is necessary to derive accurate δ13C values based on proteins for arthropod tissues. We additionally call for further research to test tissue selection effects upon derived isotopic measures and ultimate ecological conclusions, and given the laborious and difficult nature of arthropod dissection, suggest such research will be instrumental in testing potential mathematical or mass/balance corrections as a potential alternative to dissection when necessary.


We thank the Food and Environment Research Agency (FERA) for funding this research through their seedcorn programme, Anna Cucknell for laboratory expertise and an anonymous reviewer for useful comments.

Data accessibility

Data available from the Dryad digital repository. doi: 10.5061/dryad.7tm07 (Perkins et al. 2013).