Wing morphology is linked to stable isotope composition of nitrogen and carbon in ground beetles ( coleoptera : carabidae )

As movement is expensive in terms of energy required, mobile animals might have to utilize more energy rich resources than less mobile animals. As mobile animals are possibly more effective foragers we hypothesize a positive correlation between mobility and trophic niche width. We address this link using information on the trophic level of 35 winged, wingless and wing dimorphic species of ground beetles studied on 18 lake islands and at two mainland sites in northern Poland. Trophic analyses using stable isotope values (δ15 N, δ13C) revealed that winged individuals of wing dimorphic species are characterized by broader isotopic niches than wingless individuals. Macropterous species are characterized by depleted δ13C values, which can be interpreted in terms of lipid­rich prey selec­ tion. Wing dimorphic species are characterized by higher trophic levels, as inferred from δ15 N values, than winged species. Numbers of islands colonized by particular species were not correlated with δ15 N values, δ13C values or isotopic niche width. our results point to a relationship between diet and wing morphology in ground beetles.


IntroductIon
The ability of animals to move is a key component of many important activities like foraging, reproduction or predator avoidance (Begon et al., 2006).In particular, it determines the ability to disperse and influences energy budgets that affect other life history traits (e.g.Dingle, 1978;Roff, 1990;Guerra, 2011).While movement is ex pensive in terms of energy (Zera & Denno, 1997;Combes & Dudley, 2009;Bonte et al., 2012), mobile animals might utilize different resources than less mobile animals.In ad dition, the ability to use distant and diverse food resources might correlate with the degree of polyphagy (Real & Ca raco, 1986), or alternatively less mobile animals might be dependent on local, often poor quality resources, whereas mobile animals can search for and select high quality food items; so mobile animals might be characterized by a nar rower trophic niche.In consequence, trophic niche breadth and possibly trophic level might be linked to their ability to move.Such differences should be observable both between individuals of the same species with different mobilities and between species, although this hypothesis has so far not been verified.Here, we address the link between wing morphology of ground beetles and resource use at both in tra-and interspecific levels.
Ability to move and trophic level are not easily measured traits and appropriate data are available for only a small number of taxa.In this respect probably the best studied are ground beetles (Kotze et al., 2011).While wing mor phology and dispersal ability in Carabidae are not linked in a simple way due to e.g.autolysis of flight muscles (Den Boer et al., 1980, Desender, 2000), beetles can be un equivocally classified according to wing morphology into winged, wingless and wing dimorphic species (Den Boer et al., 1980).
In contrast to wing morphology, trophic levels are diffi cult to estimate in carabids and major compilations rely on feeding experiments, gut content analysis and field obser vations (Hengeveld, 1980;Harwood et al., 2001;Juen & Traugott, 2006).Stable isotope analysis promises a better Gravel et al. (2011) might also apply to the relationship between trophic niche width and distributions of carabids.
Stable isotope analysis can be used to test these predic tions in detail.Below, we link trophic niche space, wing morphology and the spatial distributions of ground beetles to test for an unexplored relationship between key life his tory traits.We predict that (1) winged beetles are character ized by larger trophic niches than wingless beetles, and that (2) winged beetles are characterized by lower δ 13 C values than wingless beetles.We test these predictions within and between species.

study sites and sampling
We sampled carabid beetles on 18 islands and at two adja cent mainland sites (Fig. 1) in two lakes in NE Poland: lake Mamry (54°00´-54°10´N, 21°30´-21°52´E) and lake Wigry (54°00´-54°05´N, 22°01´-22°09´E).The islands varied in size (0.15-38.82 ha) and distance to the nearest mainland (30-375 m).In both archipelagos humid Alder (Alnion glutinosae) and lime oak forests (Tilio-Carpinetum betuli) dominate.In addition, there were abandoned pastures on three of the islands in lake Wigry (Arrhenatherion and Cynosurion alliances).All habitats on each island were sampled along a total of 27 transects.on each island and in each habitat we established transects composed of 10 pit fall traps (0.5 l plastic mug, mouth diameter 120 mm, plastic roof, filled with pure monoethylene glycol) set 10 m apart.The traps were emptied weekly and animals were preserved in 96% alco assessment of trophic level and diet type (layman et al., 2007;Martinez del rio et al., 2009;Boecklen et al., 2011).The carbon isotope value of body tissues (δ 13 c, reflecting the 13 c ⁄ 12 C value) is approximately stable across trophic levels but depends on resource type and habitat, while the nitrogen isotope value (δ 15 N, reflecting the 15 N ⁄ 14 N value) increases in insects by about 2.5‰ per trophic level (Mc cutchan et al., 2003;Vanderklift & Ponsard, 2003;Ikeda et al., 2010) and thus indicates the relative position of an in dividual within its food chain.Hence variation in δ 13 C and δ 15 N values delineates the "isotopic niche" that is distinct from, but in many circumstances should align closely with, the actual trophic niche (Newsome et al., 2007;layman et al., 2012).In ground beetles, stable isotope analysis has already proven to be a powerful tool for inferring trophic differences both between species and between individuals of the same species (e.g. okuzaki et al., 2009, 2010;Sa sakawa et al., 2010).In particular, this method has revealed that carabids feed on different diets ranging from living and decomposing plant material to epigeic predatory ar thropods across more than three trophic levels (Zalewski et al., 2014).Their broad spectrum of trophic preferences and the variation in their wing morphology, within and between species, make Carabidae a unique model for investigating the relationship between mobility and trophic position.
In particular wing dimorphic species (species is com posed of winged and wingless individuals) provide an opportunity to analyse the relationship between mobility and trophic level.Winged individuals of wing dimorphic insects are more mobile than wingless individuals, even if they have already lost the ability for effective flight (So cha & Zemek, 2003).As flight and development of wings are resource costly, winged beetles are expected to have higher energy requirements than their wingless counter parts (Bonte et al., 2012).In arthropods, predators can be lipid limited (Wilder et al., 2013) and might have to seek for lipid rich (i.e.energy rich) food.This should be mostly manifested in winged beetles.As lipids are 13 Cdepleted (Post et al., 2007, Tarroux et al., 2010), winged animals ought to be characterized by depleted δ 13 C values.In addi tion, the higher activity of winged individuals might lead to a more diversified diet due to their greater chances of encountering different types of prey.Consequently, winged individuals of dimorphic species might be expected to be more omnivorous.These patterns should be observable both within dimorphic species where both winged and wingless individuals exist and among species with differ ent wing morphologies.
There have been several attempts to integrate food web theory into models of species spatial distribution and is land biogeography (Holt, 2010;Gravel et al., 2011).These models predict a positive correlation between trophic niche breadth and individual home range in heterogeneous land scapes (Gravel et al., 2011).In ground beetles species spa tial distributions are indeed connected to wing morphol ogy (e.g.Den Boer, 1990;Gutierrez & Menéndez, 1997;Zalewski & ulrich, 2006).Therefore, the prediction of hol.This procedure does not influence the isotopic composition of carbon and nitrogen in beetles (Zalewski et al., 2012), however some caution is necessary since similar procedures affected car bon signatures in some other animals (Tillberg et al., 2006).emp tied traps were refilled with fresh monoethylene glycol.Sampling was conducted over periods of four weeks in June and August 2010, respectively.Wing morphology and species average body length was assessed using data in the literature (Den Boer et al., 1980, lindroth & Bangsholt, 1985, Hůrka, 1996) and in addition for the seven wing dimorphic species wing morphology was de termined by visual inspection.

stable isotope analysis
In total we determined the isotopic values of 1155 individual ground beetles belonging to 57 species (see Appendix 1) and the following analyses are based on 35 ground beetle species of which at least five individuals were analyzed.The stable isotope data used in this study was already published in Zalewski et al. (2014).In particular the present analyses include the results for 16 winged species (370 individuals), 11 wing dimorphic species (647 individuals) and 8 wingless species (90 individuals) (see Ap pendix 1).This sample contained 553 individuals belonging to 7 wingdimorphic species (Agonum fuliginosum, Carabus granulatus, Notiophilus palustris, Agonum (Oxypselaphus) obscurus, Pterostichus melanarius, P. minor, P. strenuus) for which wing morphs were recorded and were used in comparisons of stable isotope composition between morphs of particular species.
Isotopic values were determined using the standard procedures fully described in Zalewski et al. (2014).Isotope values are ex pressed in delta ("δ") units as a deviation from the international standards and recalculated in terms of parts per thousand (‰), according to the formula: δ 13 c or δ 15 N (‰) = (r sample / R standard -1) × 1000, where R is the value of heavy/light isotope content for the element studied.The international standards were Pee Dee Belemnite for δ 13 c and atmospheric nitrogen for δ 15 N. To account for differences in isotopic baseline values recorded for different sampling sites, we collected litter samples along each trap line (in total 5 samples per line) in the middle of June and August 2010.Baseline samples were dried at 60°c for 48-70 h and ana lyzed following the methodology used for beetles.The carbon isotopic signature of the baseline was different for meadows and forested sites, but not for the different islands (DudekGodeau et al., in prep.), therefore the mean value of δ 13 C litter was calculated for each of the three habitats sampled (alder and lime forests and meadows) and was used for baseline correction (δ 13 C beetle = δ 13 C raw beetle -δ 13 C average litter in habitat ).Due to the high variability of δ 15 N litter values (mean = -2.7,SD = 1.3, n = 135), the mean δ 15 N was calculated for each trap line and used for the baseline correction (δ 15 N beetle = δ 15 N raw beetle -δ 15 N average litter on trapline ).We note that there was only a single habitat on most islands and that the vast major ity of species were recorded only in one habitat.This limited vari ability makes the inference of habitat specific trophic differences challenging.Therefore, we use habitat only for appropriate base line correction but focus on trophic differences between islands.We further note that the use of single traps might introduce some degree of pseudoreplication as traps from a particular trap line might share common habitat specific isotope signals that would reduce intra-island variability in isotopic space.However, as all species will be affected similarly this possible bias should not affect the assessment of differences with respect to flight ability.

data analysis
The standard ellipse area is proposed as a metric of isotopic niche width that is unbiased with respect to sample size, partic ularly for the Bayesian method, which accounts for the greater uncertainty associated with smaller sample sizes (Jackson et al. 2011).estimates of the Bayesian standard ellipse area (SeA) were calculated using the package r, SIAr (Parnell et al., 2010).
To statistically compare the size of ellipses for different catego ries we used the mean of SeA calculated based on 10,000 itera tions.We calculated respective SeA's for all 35 species (at least five individuals analyzed) and related SEA's and trophic posi tions (δ 15 N and δ 13 c values) to flight ability using general linear modelling (covariance analysis with body size as the continuous predictor, based either on all individuals or on species averages) and Tukey post hoc comparisons with orthogonal sums of squares and asymptotic error estimations as implemented in Statistica 7.0.The same individualbased linear model structure with body length as metric covariate was also used to assess differences in isotopic ratios between islands, species and wing morphs.A critical point when using an individual based analysis might be the possible nonindependence of data due to species member ship.In a previous study (Zalewski et al., 2014) we demonstrated the high intraspecific variability in isotopic niche spaces that fre quently exceeded the interspecific differences.Nevertheless, in order to account for this possible source of error we followed the approach of ulrich et al. ( 2014) and reduced the error degrees of freedom (1099) to the total number of traps (270) in the individ ual based modelling.This approach should maximally limit the possible nonindependence induced by trapping many individuals in the same trap.
Pearson correlation analysis was used to infer relationship be tween δ 15 N, δ 13 C and SeA values of species of ground beetles and the number of islands colonized.

results
General linear modelling revealed highly significant dif ferences between populations from different islands (P < 0.0001) and between individuals of winged, wingless and dimorphic species (P < 0.0001) with respect to δ 15 N and δ 13 c values (Table 1).We also found highly significant island × flight ability interaction terms indicating that the table 1.General linear modelling (orthogonal sums of squares) for litter-corrected δ 15 N and δ 13 C values depending on island and wing morphology as categorical and body size as a covariate.To account for the possible statistical nonindependence of individu als from the same trap we reduced the error degrees of freedom to the total number of traps (270) instead of using the total number of individuals (1155).δ 13 C value: r 2 = 0.38, P < 0.001, δ 15 N value: r 2 = 0.65, P < 0.001.relationship between flight ability and stable isotope value is different across islands (Table 1).Tukey post hoc com parisons pointed to significant differences in δ 15 N and δ 13 C values between the flight ability groups.Wingless and di morphic species had significantly (P < 0.01) higher δ 13 C values than winged species (Fig. 2E).With respect to δ 15 N, dimorphic species had significantly (P < 0.01) higher val ues than winged species (Fig. 2D).

Variable
We did not find differences in isotopic niche breadth (SEAs) between the three flight ability groups (Table 2, all pair-wise comparisons: P > 0.05; Fig. 2F) and SEAs were not correlated with the number of islands colonized, to tal abundance or body length (Table 2, Fig. 2c).δ 15 N and δ 13 c values were not significantly linked to the number of islands colonized (Fig. 2A, B), although species that had colonized more than half of the islands studied occupied nearly exclusively higher trophic levels of δ 15 N > 5.0 (Fig. 2A).
Intraspecific comparisons of wing morphs of seven wing dimorphic species (winged vs. wingless individuals) did not point to significant intraspecific differences with re spect to δ 15 N and δ 13 C values (except for P. melanarius).However, in three out of the four species for which we could calculate SeAs the macropterous individuals had larger SeAs (Table 3) than brachypterous individuals (p < 0.01).

dIscussIon
Our study is apparently the first to compare trophic posi tion and wing morphology in ground beetles.We found, based on their δ 15 N values, that wing dimorphic species characteristically occupied higher trophic levels than winged species (Fig. 2D).The average δ 13 C values also differed between the three mobility groups (Table 1, Fig. 2E).Finally, we showed that isotopic niche space (SEA) of winged individuals is larger than that of wingless individu als in wing dimorphic species (Table 3).
Intraspecific variation in their ability to move might be common, but is usually difficult to measure (Bullock et al., 2002).Wing dimorphic insects, particularly ground beetles might serve as a suitable model (Roff, 1986).Contrary to our prediction, we found that in general macropterous in table 2. General linear modelling (orthogonal sums of squares) for standard ellipse areas (SeA) depending on wing morphology (categorical), trophic position (corrected δ 13 c and δ 15 N), number of sites colonized and total abundance (metrical): N = 33; r 2 = 0.53, P = 0.008.2009) record differ ences in the proboscis lengths of winged and wingless morphs of seed-feeding Hemiptera.In addition, in crickets the effectiveness of food assimilation might also differ be tween morphs (Mole & Zera, 1993).Finally, Karpestam & Forsman (2013) record higher δ 15 N values for winged morphs of a wing dimorphic grasshopper in one popula tion and lower values in another.These authors interpreted their results in terms of the ability of winged grasshoppers to find and capture high quality food items.Although δ 15 N and δ 13 C values for the different wing morphs of carabids do not differ, morphs significantly dif fered with respect to isotopic niche breadth expressed in terms of SeA (Table 3).Apparently, macropterous beetles with potentially higher mobility and energy requirements obtain resources from different trophic levels and from different sources of carbon.This finding supports our first hypothesis that winged individuals of generalist predators should be more omnivorous (Wilder et al., 2013).

Variable
Broader isotopic niche spaces (SeAs) of winged indi viduals of dimorphic species (Table 3) are in line with recent theory that predicts that ecological generalists are highly mobile (Bonte et al., 2003;Billiard & lenormand, 2005;ronce, 2007).Fig. 2F indicates a similar interpre tation, although the recorded differences in niche space between wingless and winged species are statistically not significant (t-test: P = 0.24 and Table 2).Winged species were however characterized by strongly depleted δ 13 C val ues (Fig. 2E), as predicted by hypothesis 2. In this case, macropterous beetles might search for lipidrich prey.As lipids are generally 13 c depleted (Post et al., 2007), mobile species (or possibly specimens of wing dimorphic species) should have comparably lower δ 13 C values.This interpre tation is supported by data for Pterostichus melanarius, the only dimorphic species that was characterized by differ ences in δ 13 C (Table 3).our overall results are in line with winged species preferring more lipid rich prey.
Trophic niche width estimated using SeAs was not re lated to island occupancy by ground beetles (Table 2, Fig. 2C) which does not corroborate recent theoretical studies aimed at linking trophic ecology with island biogeography (Holt, 2010, Gravel et al., 2011).Indeed, differences be tween species in terms of the widths of their trophic nich es might not be related to their being habitat generalists.
Similarly the other finding that widespread species occupy higher trophic levels (Fig. 2A), while species with a re stricted occurrence occupy all trophic levels, contradicts popular expectations.In particular Holt (2010) predicts a negative correlation between trophic level and home range due to the lower population densities of trophically high ranking species.Calcagno et al. (2011) states that if preda tors are constrained by the presence of their prey (see also Piechnik, 2013), which should restrict predator occur rences, then predators need to have high dispersal abili ties in order to sustain viable populations.Apparently these models do not reveal the important constraints determining ground beetle spatial distributions.
our results add to our knowledge of the relationship be tween two important ecological features of carabids: the trophic level of beetles and their wing morphology.How ever, there are some methodological pitfalls, which require further study.In particular, the δ 15 N value is still only a proxy of the real trophic level within a food web (Martinez del Rio et al., 2009) and the difference in SeAs of winged and wingless individuals of wing dimorphic species can be biased due to errors in the baseline correction for ani mals that move from one habitat to another.However, as we tested different baseline corrections (DudekGodeau et al., in prep.), which gave qualitatively identical results, our findings indicate that in the system studied, which is composed of species of beetles with the whole range of mobility and feeding strategies, trophic characteristics ex pressed in terms of stable isotope composition of nitrogen and carbon are indeed linked to wing morphology.While these links are not always in accordance with theoretical predictions, areas of uncertainty identified in the present study could direct further research.table 3. Average litter-corrected δ 13 c and δ 15 N values (± one standard error) and standard ellipse areas as approximations of the isotopic niche spaces (SeA) of wing dimorphic species.Due to sample size constraints it was only possible to estimate SeA for four species.Significant differences (t-test) between macropterous and brachypterous individuals (species): * P < 0.05; ** P < 0.01; *** P < 0.001.N -respective sample sizes.Note that δ values and SEA areas are measured in different units (value and eigenvalue scale) and thus are not directly comparable.

Fig. 2
Fig. 2. δ 15 N (A), δ 13 c (B) and SEA (c) values for species of ground beetles were not related to island colonization (Pearson correla tions: wingless species (blue dots), winged species (violet dots), wing dimorphic species (red dots): all |r| < 0.2, P > 0.2).respective averages are presented in D, E and F in the form of box-whiskers plots, which indicate the medians, 25-75 percent quartiles (boxes) and minimum and maximum values (whiskers).Stars indicate significant (P < 0.01) difference between sets of species (Tukey post hoc comparisons).