Oecologia

, Volume 146, Issue 1, pp 89–97

Isotopic enrichment in herbivorous insects: a comparative field-based study of variation

Authors

    • Department of EntomologyUniversity of California
  • Jay A. Rosenheim
    • Department of EntomologyUniversity of California
Community Ecology

DOI: 10.1007/s00442-005-0170-9

Cite this article as:
Spence, K.O. & Rosenheim, J.A. Oecologia (2005) 146: 89. doi:10.1007/s00442-005-0170-9
  • 247 Views

Abstract

Researchers will be able to use stable isotope analysis to study community structure in an efficient way, without a need for extensive calibrations, if isotopic enrichment values are consistent, or if variation in enrichment values can be predicted. In this study, we generated an experimental data set of δ15N and δ13C enrichment means for 22 terrestrial herbivorous arthropods feeding on 18 different host plants. Mean enrichments observed across a single trophic transfer (plants to herbivores) were −0.53±0.26‰ for δ13C (range: −3.47‰ to 1.89‰) and 1.88±0.37‰ for δ15N (range: −0.20‰ to 6.59‰). The mean δ13C enrichment was significantly lower than that reported in recent literature surveys, whereas the mean δ15N enrichment was not significantly different. The experimental data set provided no support for recent hypotheses advanced to explain variation in enrichment values, including the proposed roles for consumer feeding mode, development type, and diet C:N ratio. A larger data set, formed by combining our experimental data with data from the literature, did suggest possible roles for feeding mode, nitrogen recycling, herbivore life stage, and host plant type. Our results indicate that species enrichment values are variable even in this relatively narrow defined group of organisms and that our ability to predict enrichment values of terrestrial herbivorous arthropods based on physiological, ecological, or taxonomic traits is low. The primary implications are that (1) mean enrichment may have to be measured empirically for each trophic link of interest, rather than relying on estimates from a broad survey of animal taxa and (2) the advantage of using stable isotope analysis to probe animal communities that are recalcitrant to other modes of study will be somewhat diminished as a consequence.

Keywords

ArthropodCommunity ecologyFood webTerrestrial systemTrophic position

Introduction

Stable isotope analysis has been used extensively in the study of aquatic ecosystems (see Post 2002) and is being employed increasingly in the study of terrestrial communities (Bluthgen et al. 2003; Davidson et al. 2003; Hobson and Clark 1992; Neilson et al. 1998; Ostrom et al. 1997; Patt et al. 2003; Scrimgeour et al. 1995). In particular, stable isotope analysis has been demonstrated to be especially useful in terrestrial systems where other means of study are logistically difficult or impossible, such as in cryptic soil macro- and mesofaunal food webs (Eggers and Jones 2000; Ponsard and Arditi 2000; Ruess et al. 2004; Scheu and Falca 2000). Many component species of terrestrial communities have variable diets, including omnivores that feed on both plants and prey (Coll and Guershon 2002) and predators that feed on both herbivores and on other predators (Rosenheim 1998). As a rapid tool to assess diet and trophic position, stable isotope analysis can be an important probe into the ecological roles of these omnivores (Davidson et al. 2003; Tillberg and Breed 2004). Naturally occurring stable isotopes of carbon (C) and nitrogen (N) can be used to study community structure, because heavier isotopes are preferentially retained and lighter isotopes are preferentially lost through excretion or respiration when one organism eats another (Deniro and Epstein 1978; Deniro and Epstein 1981). If trophic enrichment values are consistent, we will be able to use stable isotope analysis efficiently without a need for extensive calibrations. However, if there is a substantial amount of unexplained variation, such calibrations will be necessary.

Sensitivity analysis has shown that the primary producer-herbivore link contributes the greatest amount of error to trophic position estimates (Vander Zanden and Rasmussen 2001). It has been suggested that this variation may not pose a significant problem for some applications of stable isotope analysis, such as for studying top predators in freshwater ecosystems (Post 2002). This assertion arises from the knowledge that the stable isotope content of these top predators reflects the averaging of many pathways within the food web, producing a general enrichment mean that is fairly robust and similar across food webs and ecosystems. We hypothesize, however, that this assumption will frequently be violated in terrestrial arthropod communities, because it is often the case that one or a few abundant prey heavily dominate the diets of generalist predators (e.g., Evans 1982; Rosenheim et al. 1999). Thus, it is important to reconsider the amount and importance of variation in isotopic enrichment across a single trophic transfer when using stable isotope analysis to study arthropod communities.

Several hypotheses seeking to explain variation in enrichment of δ15N and δ13C exist in the literature. One potentially important factor is diet quality, which has been shown to influence consumer δ15N and δ13C enrichment (Adams and Sterner 2000; Vanderklift and Ponsard 2003; Webb et al. 1998). Animals feeding on poor diets with high C:N ratios may recycle internal nitrogen stores, leading to increased enrichment values (Hobson and Clark 1992). A consumer’s feeding mode (e.g., vascular vs. non-vascular; (McCutchan et al. 2003; Pinnegar et al. 2001) and the presence of nitrogen upgrading or recycling endosymbionts (Davidson et al. 2003) may also influence observed isotopic ratios. Additionally, it has been hypothesized that fractionation due to metamorphosis may contribute to differences in isotopic ratios observed across different arthropod life stages (Patt et al. 2003). As a corollary of this hypothesis, it is also possible that enrichment values may differ for arthropods that undergo hemimetabolous versus holometabolous development. Three meta-analyses of the expanding stable isotope analysis literature have been conducted to examine the mean and variance of isotopic enrichments and to correlate variation in enrichment values with various biological and ecological factors, including several of the hypotheses mentioned above (McCutchan et al. 2003; Post 2002; Vanderklift and Ponsard 2003). These meta-analyses have been valuable in identifying patterns of variation in enrichment values and testing a variety of putative mechanisms. In evaluating whether trophic enrichment calibrations will be necessary for the accurate use of stable isotope analysis in ecological studies, the magnitude of inter-specific variation is the key metric. A single, methodologically uniform study surveying an array of taxa will generate the most accurate estimate of this variation in enrichment values across a single trophic transfer.

Here, therefore, we attempt to complement published meta-analytical studies of enrichment variation by building a comparative data set experimentally, using a uniform methodology and a subject group that is relatively uniform both taxonomically and ecologically (herbivorous terrestrial arthropods). The goals of this study are: (1) to evaluate the applicability of enrichment estimates generated from broad taxonomic surveys to the study of terrestrial arthropods, (2) to quantify variation in isotopic enrichment that exists across different combinations of host plants and their herbivores and search for correlates of observed enrichment variation, and (3) to assess the repeatability over time of an enrichment value observed for a given plant-herbivore combination.

Materials and methods

Host plant tissue samples

Perhaps the most significant problem associated with the use of stable isotope analysis in studies of community structure has been the lack of appropriate experimental controls, in which consumers are fed known diets and their resulting isotopic enrichment measured (Eggers and Jones 2000; Gannes et al. 1997; Handley and Scrimgeour 1997). The ideal control for a laboratory study would be a laboratory-based control, however, for researchers applying stable isotope analysis to problems in field ecology (as opposed to lab micro- or meso-cosms), field-based controls will be the most appropriate.

Our study circumvented this issue via field collection of herbivore–host plant pairs (an herbivore and the plant upon which it fed) in which the herbivore species was relatively sessile (e.g., gall formers, leaf miners, and the immature stages of externally feeding herbivores). The benefit of a field-based study is that field collected individuals should reflect isotopic enrichment under natural conditions. The feeding habits of our chosen herbivores make it possible to calculate stable isotope enrichment for the herbivore relative to the same plant individual, and often the same plant organ, that had previously been eaten. Plant tissue was collected using a hole-punch (0.25 cm2) or by taking an equivalent cutting of material (e.g., stem tissue) directly adjacent to the most recent feeding site of the insect. For chewing insects (e.g., caterpillars) and, to a lesser degree, for cell content feeders (e.g., leaf hoppers and mites), the plant tissues thus collected should closely reflect those the herbivore consumed. We emphasize, however, that for those arthropods that feed on host plant phloem or xylem, we were unable to obtain the most relevant measure of plant baseline isotopic values (i.e., phloem or xylem fluid values). We discuss the possible significance of this difficulty of obtaining ideal host plant baseline values below.

Processing of samples

A total of 22 insect herbivore–plant pairs from natural and agricultural terrestrial ecosystems were analyzed in this study (S1a). All samples for each herbivore–plant pair were collected at the same location and time. Internally feeding herbivores were dissected from the host plant in the field. Care was taken to ensure that plant material was free from insect frass or decay. To avoid microbial decomposition and possible subsequent alteration of isotopic signals, field collected samples were immediately placed on ice and returned to the laboratory where they were either frozen until prepared for analysis or were immediately placed in a drying oven at 60°C. Many herbivores are subject to attack by parasitoids; it is possible that the presence of internally-feeding parasitoid larvae could affect the isotopic values of the host. However, this study, and all similar studies previously published, did not explicitly exclude parasitized individuals (although moribund and dead individuals were excluded).

Unassimilated plant material in the insect gut could contribute to a distorted isotopic enrichment measurement. Therefore, prior to drying, insects were dissected in the laboratory to remove the gut and gut contents. This was done with all arthropods except for the insect pupae, because insects generally void their guts at pupation, and for scale insects, which have guts that are too delicate to dissect readily. Insect and plant materials were held in a drying oven at 60°C until submitted for stable isotope analysis (a minimum of 2 days).

Stable isotope analysis

In general, five individual insects were aggregated to form a single sample for isotope analysis (one replicate of the herbivore measurement), as were their five corresponding plant specimens (one replicate for the plant measurement). However, in some cases, low insect abundance mandated that samples be composed of fewer individuals, while in other cases the sensitivity of the analytical equipment required that larger numbers of arthropods be combined in a single sample (e.g., for scale insects). We always used whole insect bodies for analysis because initial explorations showed that different portions of the insect’s body (e.g., head vs. thorax + abdomen) exhibited significantly different δ15N isotope signatures. Generally, the total dry masses of the insect samples were 0.06–1.0 mg and plant samples were 2.0–3.0 mg. Dried samples exceeding these limits were ground with a mortar and pestle and 2.0 mg (insects) and 3.0 mg (plants) of homogenized material were submitted for analysis. Samples were analyzed at the UC Davis Stable Isotope Facility using a PDZ Europa ANCA-GSL elemental analyzer in line with a PDZ Europa Hydra 20/20 continuous-flow isotope ratio mass spectrometer. The data for each sample include total N and C and 15N and 13C delta values. Delta values are given by the equation δX=[(Rsample/Rstandard) − 1]×1,000, where R=the isotopic ratio (heavy/light). Sample precision (SD) was 0.18‰ for δ15N and 0.05‰ for δ13C. The mean isotopic enrichment (Δ) was calculated for each herbivore–plant pair using the formula
$$\Delta = \sum\limits_{i = 1}^{i=N_{\rm tot}} \frac{{\delta \hbox{insect}\;\hbox{sample}\,X_i - \delta \hbox{plant}\;\hbox{sample}\,Y_i}}{N_i} $$
(1)
where ‘i’ indicates the matched samples for herbivore–plant pair X–Y, and ‘Ntot’ is the total number of matched samples analyzed for pair X–Y.

Temporal variation

Recent work in an aquatic system has suggested that temporal variation in isotopic values at the primary consumer level can be significant and could influence relative trophic position estimates (O’Reilly et al. 2002). To quantify the temporal variation that exists within the herbivore–plant combinations in this study, six herbivore–plant pairs were re-collected and analyzed approximately 1 year after their initial collection dates. Herbivore–plant pairs were chosen that encompassed a wide range of enrichment values and that were available during the re-collection period (March–April 2004) (S1a). Samples were re-collected from the same location, but not necessarily from the same individual plants that had been sampled previously. Additionally, to determine if changes in herbivore Δ δ15N are correlated with temporal changes in host plant C:N ratio, we calculated the shift in consumer Δ δ15N (Δ δ15Nshift=Δ δ15Nyear1 −Δ δ15Nyear2) and host plant C:N ratio (C:Nshift = meanC:Nyear1-meanC:Nyear2).

Literature review

We supplemented our experimental data with previously published estimates of mean trophic enrichment of δ13C and δ15N (S1b) for herbivorous terrestrial insects and mites. Our goal was to generate the largest possible data set for comparison. Thus many studies were included, which differed from our own internal methodology (laboratory feeding trials; studies of generalist herbivores and/or mobile life stages; and studies which did not dissect the herbivores’ guts). However, studies in which the herbivores were reared on non-plant (e.g., yeast), artificial, or unknown diets were excluded. Henceforth, the terms “experimental data set” and “full data set” will refer to the internally generated and literature supplemented data sets, respectively.

Data analysis

Our analyses treated the mean isotopic enrichment for a particular herbivore–plant pair as the basic unit of replication. However for two genera of our sampled herbivores, Liriomyza spp. (three herbivore–host plant combinations) and Empoasca spp. (three herbivore–host plant combinations) there was evidence of significantly reduced variability for δ13C enrichment values across the replicate measurements. To be conservative, we therefore took the averages of the δ13C enrichment values for these two herbivore genera, treating them as a single replicate. The δ13C enrichment data from the literature survey were treated similarly. For both Andricus kingii feeding on Quercus sp. and Junonea coenia feeding on Plantago lanceolata and Kickxia elatine, enrichment means were available for two life stages (larva and pupa (A. kingii) and pupa and adult (J. coenia)). To ensure the independence of our observations, the pupal stage means for these species were excluded from all analyses except those specifically comparing life stages. Data are presented as means ±1SE, unless otherwise noted.

Differences in mean δ15N and δ13C enrichment among herbivore–plant pairs and groupings of pairs were analyzed using the Wilcoxon or Kruskal-Wallis tests. Because the δ13C enrichments observed in this study were significantly lower than those observed previously (see Results), differences in δ13C enrichments in the full data set were analyzed using a one-way non-parametric ANOVA blocked by data source (this study vs. previously published studies), an extension of the Friedman test (N. Willits, UCD Statistics Department, personal communication). Levene’s test was used for variance comparisons. Regression analysis was used to examine the relationship between the first and second measurements of enrichment values for the herbivore–host plant pairs that were re-collected to quantify temporal variation in isotopic enrichment. Spearman’s rank correlation coefficient test was utilized to examine the relationship between host plant C:N ratio and herbivore isotope ratio enrichment.

We used stepwise regression analysis to develop two enrichment models, whose terms were selected from the enrichment correlates determined to be significant in the one-way and two-way analyses described above. The distribution of δ13C enrichments in the full data set was non-normal (W=0.93, P=0.01), therefore to satisfy the assumptions of this analysis these data were square-root transformed (after adding a constant to make all values positive). The criterion for entering the model was P≤0.25, and the criterion for retaining a term in the model was P≤0.10. The resulting models were analyzed using one-way (for δ15N) and two-way (for δ13C) ANOVA. Although the full data set was used for these multi-factor analyses, we recognize that a still larger data set would have been better. However, because many of the correlates of enrichment values identified by our analyses and in previous meta-analyses are themselves highly intercorrelated (e.g., taxon and feeding mode), we believe a multivariate approach is needed, and we submit these models as a first step toward a clearer dissection of the mechanisms underlying variation in isotopic enrichment values.

The sampling design used in this study was sub-optimal in that not all samples were collected and analyzed simultaneously. Although we attempted to maintain a consistent protocol for sample processing and analysis, it is still possible that our procedures varied over time in subtle ways, and that this temporal variation contributed to the variance we report in enrichment values. To address this possibility, we also report analyses of a subset of the experimental data that was collected and analyzed simultaneously. All data analyses were conducted on the SAS JMP 3.2.6 platform.

Results

Overall enrichment means

The herbivore–plant pairs had a combined mean δ15N enrichment of 1.88±0.37‰ (22 pairs; the pupal stage of A. kingii was not included) and a δ13C enrichment of −0.53±0.26‰ (18 pairs) (i.e., herbivores were depleted in 13C relative to their host plants). ANOVA revealed significant variation in enrichment values among the herbivore–plant pairs (δ15N: χ2=73.19, P<0.0001; δ13C: χ2=67.01, P<0.0001) (Fig. 1). The literature review yielded 27 estimates of δ15N enrichment, with a mean of 1.75±0.39‰, and 27 estimates of δ13C enrichment, with a mean of 0.74±0.26‰. Our overall estimate of δ15N enrichment did not differ significantly from that obtained from the literature review, but the δ13C enrichment estimate was significantly lower (−0.53±0.26‰ vs. 0.74±0.26‰, Z=−3.41, P=0.0007; Fig. 2a).
https://static-content.springer.com/image/art%3A10.1007%2Fs00442-005-0170-9/MediaObjects/442_2005_170_Fig1_HTML.gif
Fig. 1

δ15N and δ13C enrichment (mean ± SE) of 22 herbivore–plant pairs in the experimental data set. Numbers correspond to herbivore species listed in ESM. To ensure independence of observations, data for A. kingii pupae (pair no. 2) are not presented. The expected mean for “terrestrial organisms” (McCutchan et al. 2003) is represented by filled diamond, and the mean for the experimental data set by filled circle. Standard error for some pairs is too small to show

https://static-content.springer.com/image/art%3A10.1007%2Fs00442-005-0170-9/MediaObjects/442_2005_170_Fig2_HTML.gif
Fig. 2

a δ15N and b δ13C enrichment (mean  ± SE) among herbivores classified according to factors that may be associated with the magnitude of isotope ratio enrichment. None of the observed effects are statistically significant. Standard deviation in c δ15N and d δ13C enrichment for herbivore classifications. Numbers within each bar indicate the sample size n. *P≤0.05; **P≤0.005

Variation in enrichment values

In this study, δ15N and δ13C enrichment ranged from −0.20‰ to 6.59‰ and −3.47‰ to 1.89‰, respectively. This range is equivalent to the cumulative enrichment expected of ca. 3.5 trophic transfers assuming an average enrichment of 1.88‰. Standard deviations of δ15N and δ13C enrichment were 1.73‰ and 1.12‰, respectively. These standard deviations did not differ significantly from those observed in the literature survey (SD Δ δ13C: 1.12‰ vs.1.46‰, df=1,43, F=0.86, P=0.36; SD Δ δ15N: 2.01‰ vs. 1.73‰, F=1.18, P=0.28; Fig. 2b).

Baselines and collection method

Enrichment means and variances for samples collected at the same site (4) and time did not differ significantly from those that were not (18) (Δ δ13C: −0.06±0.49‰ vs. −0.63±0.24‰, Z=0.72, P=0.47; Δ δ‰15N: 2.43±0.52‰ vs.1.76‰±0.44, Z=1.32, P=0.19). There were also no significant differences in either mean enrichment values or variance in enrichment values when we compared the vascular feeders, for which we had sub-optimal measures of the plant baseline, with the non-vascular feeders, for which we had more nearly optimal measures of the plant baseline (Fig. 2a, b).

Temporal variation in enrichment values

Regression analysis suggested that δ13C enrichment values measured for particular herbivore–plant pairs in years 1 and 2 were not correlated (r= −0.28, F1,4=0.33, P=0.59). δ15N enrichment value measurements were more repeatable (r=0.79, F1,4=6.73, P=0.06), although small sample size (n=6) rendered this result only marginally significant (Fig. 3). Regression of herbivore Δ δ15Nshift and host plant C:Nshift revealed a positive but not a significant trend (r=0.71, F1,4=4.02, P=0.116; Fig. 4).
https://static-content.springer.com/image/art%3A10.1007%2Fs00442-005-0170-9/MediaObjects/442_2005_170_Fig3_HTML.gif
Fig. 3

Relationship between the mean δ15N enrichment of six herbivore samples collected approximately 1 year apart (r=0.79, P=0.06). Δ represents the difference in enrichment between herbivores and plants. Numbered data points correspond to the herbivore species listed in ESM

https://static-content.springer.com/image/art%3A10.1007%2Fs00442-005-0170-9/MediaObjects/442_2005_170_Fig4_HTML.gif
Fig. 4

Relationship between host plant C:N ratio shift (C:Nyear 1–C:Nyear2) and herbivore δ15N enrichment shift (Δ δ15Nyear1 –Δ δ15Nyear2) for six herbivore–plant pair samples collected approximately 1 year apart (r=0.71, P=0.116). Δ represents the difference in enrichment between herbivores and plants. Numbered data points correspond to the herbivore species listed in ESM

Enrichment correlates

Spearman’s rank correlation coefficient test was used to examine the relationship between host plant quality (measured as its C:N ratio) and herbivore isotopic enrichment in the experimental data set. There was no relationship between the C:N ratio of the host plant and herbivore isotopic enrichment (δ15N enrichment: rs= 0.16, P=0.48; δ13C enrichment: rs=−0.013, P=0.96). Analysis of the full data set also revealed no correlation between isotope ratio enrichment and diet C:N ratio (for Δ δ15N: rs=0.08, P=0.71; for Δ δ13C: rs=−0.20, P=0.38).

Several categorical comparisons of enrichment means and variances for both stable isotopes were conducted using the experimental data set, as summarized in Fig 2a and b. Tests for the influence of feeding mode, the presence of N-recycling/upgrading endosymbionts, development mode (metamorphosis type and generations per year), and host plant type (woody or herbaceous) revealed no significant differences in mean Δ δ15N or Δ δ13C for any of the comparisons. The δ13C enrichments of herbivores feeding on herbaceous plants were less variable than those consuming woody hosts (SD: 0.35‰ vs. 1.82‰, F1,16=11.36, P=0.004), and herbivores that have multiple generations per year had less variation in mean Δ δ13C than did those with a univoltine life history (SD: 0.82‰ vs. 1.82‰, F1,15=4.62, P=0.048). Though similar trends were suggested for the variance in Δ δ15N, differences were non-significant. Sample sizes for many of these comparisons were small, and thus standard errors and associated confidence intervals were relatively large in some instances. Though recognizing that the results must therefore be interpreted conservatively, we find no evidence that trophic enrichment is affected consistently by diet quality, consumer feeding mode, the presence of N-recycling endosymbionts, development type, or host plant type.

When the same analyses were conducted with the full data set, significant or marginally significant effects of feeding mode and life stage on mean Δ δ15N, and of life stage, N recycling, and host plant type on mean Δ δ13C were found (S2). These analyses benefit from having a larger sample size and increased power, but include data from a myriad of sources with different methodologies. We feel most confident in the results from the internally consistent data set generated by this study, yet all of these analyses are meant to suggest possible trends and areas for further research rather than representing definitive tests of their respective hypotheses.

Multi-factor model

Of the factors showing some correlation with Δ δ15N (life stage, development mode, feeding mode, and generations per year) and Δ δ13C (life stage, host plant woodiness, and source) in the univariate analyses, only data source (for Δ δ13C) and life stage (for Δ δ15N and Δ δ13C) were retained in the final models identified by stepwise regression analysis (Table 1). Though nitrogen recycling also showed some correlation with Δ δ13C, insufficient degrees of freedom prevented its inclusion in the multi-factor model.
Table 1

Analysis of variance results for enrichment models developed through stepwise regression

Variable

df

SS

F

P

δΔ 13C

Life style {P vs. J&A}

1

0.064

4.66

0.039

Source

1

0.050

3.62

0.066

δΔ 15N

Life style {J vs. P&A}

1

14.53

5.04

0.031

Criteria for entering model: P≤0.25. Criteria for remaining in model: P≤0.1

Abbreviations: J juvenile, P pupae, A adult

Discussion

The overall δ15N enrichment mean in this study was 1.88±0.37‰, and the δ13C enrichment mean was −0.53±0.26‰. Although overall Δ δ15N was not significantly different from the mean compiled from the literature, our Δ δ13C was significantly lower, reflecting a depletion rather than an enrichment of δ13C. This was an unexpected result, but may reflect the large number of juvenile herbivores sampled and the influence of effects related to metamorphosis (see discussion below). Another possibility is that because we intentionally focused on sessile species, the herbivores in this study may have had a larger ratio of lipids to muscle and chitin. Lipids are known to be generally depleted in 13C, while muscle and chitin are usually enriched (Deniro and Epstein 1978; Webb et al. 1998).

Perhaps the most important result of our study was the substantial level of variability in δ13C and δ15N enrichment values observed for different herbivore–host plant pairs (for Δ δ13C, SD=1.12‰, range=−3.47‰ to 1.89‰; for Δ δ15N, SD=1.73‰, range=−0.20‰ to 6.59‰). The observed range in variation of δ15N enrichment among species is similar to the total enrichment expected across three trophic levels, assuming an average enrichment of 2.3‰ (see ‘terrestrial organisms’, McCutchan et al. 2003) and ca. 3.5 trophic levels, assuming an average enrichment of 1.88‰, as measured here. The variation in enrichment values that we observed was similar in magnitude to that measured in previously published meta-analyses. Thus, this variation appears not to be an artifact of variable procedures used across different studies, but rather to represent true variation in the biological processes underlying stable isotope fractionation. We were unable to identify any significant correlates of the magnitude of trophic enrichment, and found little support for previously advanced hypotheses linking trophic enrichment to the C:N ratio of the food resource, feeding mode, development mode, or host plant type. Instead, our results suggest that trophic enrichment values are repeatable (at least for the Δ δ15N values) but idiosyncratic characteristics of a particular herbivore–host plant combination. The primary implication is that those using stable isotope analysis to probe the structure and function of arthropod communities may not be able to rely on a mean enrichment value derived from a broad survey of animal taxa.

Various hypotheses have been advanced seeking to explain variation in isotopic enrichment values. Our discussion of these hypotheses will focus primarily on enrichment of δ15N, as it often plays a primary role in estimates of trophic position.

The meta-analysis performed by McCutchan et al. (2003) suggested that consumers that feed on fluids (blood and plant juices) have significantly lower δ15N enrichments than non-fluid feeders. This may result from fluid feeders ingesting labile amino acids directly, rather than having to break down proteins (Pinnegar et al. 2001). We conducted similar comparisons with vascular and non-vascular feeders. The experimental data set did not support this hypothesis, though the full data set did. The conflicting results could be due to the greater proportion of aphids among vascular feeders in the full data set, if recycling of light (15N depleted) ammonia waste by endosymbionts leads aphids to have lower δ15N enrichment. Our inability to detect a difference either between N-recyclers (aphids) and non-recyclers, or aphids and other vascular feeders does not support this alternative hypothesis however. Indeed, it has been suggested that all phloem feeding Homoptera are likely to be associated with nutrient enhancing endosymbionts (Wilkinson and Ishikawa 2001), and N-recycling bacteria are known to occur in ants, termites, and cockroaches, taxa that also feed on nitrogen poor diets (Cochran 1985; Hongoh and Ishikawa 1997; Potrikus and Breznak 1977; Sasaki et al. 1996). As mentioned previously, vascular feeding arthropods are also problematic with regard to obtaining an appropriate isotopic baseline sample (e.g., phloem). In lieu of vascular fluids, other tissues more easily sampled are often collected from the host plant/animal. The problem is that vascular fluids may have significantly different isotopic values from other tissues or the organism as a whole (Pinnegar et al. 2001; Yoneyama et al. 1997). A more definitive test of these hypotheses therefore will await a good sample (many species) of phloem and xylem feeders for which the plant vascular tissue baseline isotopic values are measured directly.

Nitrogen availability has also been investigated for its effects on consumer isotopic enrichment. Nutritional stress and/or starvation have been hypothesized to increase metabolic recycling of nitrogen within an organism, leading to increased δ15N fractionation and subsequently higher enrichment values (Oelbermann and Scheu 2002). It follows that food quality should be inversely related to consumer enrichment, with increasing enrichment seen with decreasing food quality (i.e., protein or nitrogen content). The empirical results, however, are mixed. For Daphnia magna (Adams and Sterner 2000) and Locusta migratoria (Webb et al. 1998), δ15N enrichment decreased with increasing food quality, but the spider Pardosa lugubris showed the reverse effect: increasing δ15N enrichment with increasing food quality (Oelbermann and Scheu 2002). Results from meta-analyses have also differed, with Vanderklift and Ponsard (2003) finding a negative relationship and McCutchan et al. (2003) finding no relationship between diet quality and δ15N enrichment. Trophic, ecological, and taxonomic differences among the organisms likely contribute to the mixed results. In the current study, we found no relationship between herbivore δ15N enrichment and the C:N ratio of the paired host plant. We did find a marginally non-significant trend for the shift in host plant C:N with the corresponding shift in herbivore δ15N enrichment, suggesting that further study might be warranted to see if herbivore enrichment values track shifts in host plant C:N ratios.

It has been shown that starvation and nutritional stress can cause increased δ15N enrichment (Adams and Sterner 2000; Oelbermann and Scheu 2002). By extension, we can hypothesize that holometabolous insects might exhibit an increase in δ15N during the pupal stage, when they undergo metamorphosis and are unable to forage, and must therefore rely on internal nutrient stores and the histolysis of larval tissues for maintenance and development (Chapman 1998). Patt et al. (2003) developed a model that successfully predicted adult Chysoperla carnea isotopic compositions only when a novel factor accounting for the effect of metamorphosis was included. If the effect of metamorphosis is important and general among arthropods, then hemimetabolous herbivores (which are generally capable of foraging throughout development) may have lower δ15N enrichment than holometabolous herbivores. Likewise, juveniles that have not undergone metamorphosis should have lower enrichments than adult insects. Our experimental data did not reveal a difference in δ15N value between hemimetabolous and holometabolous species, though such a difference may only be detectable among adult insects after metamorphosis is complete (while this study focused on juvenile stages). When supplemental data from the literature were included, however, the average δ15N enrichment of the juvenile stages was found to be significantly lower than that of adults, supporting a role for metamorphosis effects. Comparative studies of conspecific larvae and adults reared on the same diet will be useful in determining the magnitude and generality of the effect of metamorphosis on arthropod enrichment values.

Recent meta-analyses examining enrichment correlates have not utilized multivariate analyses (McCutchan et al. 2003; Vanderklift and Ponsard 2003). An obvious constraint to the use of multivariate analyses is sample size, a limitation also faced in this study. However we feel the full data set is of sufficient size to enable us to begin exploring multivariate analyses of arthropod enrichment correlates. Our multivariate statistical models suggest that life stage is an important influence on both arthropod Δ δ15N and Δ δ13C values. As more data become available in the literature, we expect that future meta-analyses seeking to explain enrichment variation will increasingly utilize multivariate models, generating more robust conclusions.

Acknowledgments

We wish to thank UCD Student Farm and Sagehen Creek Research Station for granting permission for insect collection; P. Ward, T. Kondo, J. DeBenedictis, and the UCD Herbarium for their help with specimen identification; T. Mittler and N. Willits for technical advice; G. Langellotto for sharing unpublished data; C. Armer, J. Harmon, R. Karban, G. Langellotto, S. Scheu, and L. Yang, who provided helpful comments on the manuscript; and the three anonymous reviewers whose comments improved the revised manuscript. This work was supported by funds from USDA NRICGP grant 2001–35302–10955 to JAR. The experiment in this study was conducted in accordance with the laws of the United States of America.

Supplementary material

442_2005_170_MOESM1_ESM.pdf (96 kb)
Supplementary material

Copyright information

© Springer-Verlag 2005