Trophic niche differences between two congeneric goby species : evidence for ontogenetic diet shift and habitat use

Food partitioning is one of the primary mechanisms facilitating the stable cooccurrence of competing species, but very few studies have investigated how food resource use of competing and closely related species varies with life-history stages and habitats. In Lake Erhai (China), the trophic niche of 2 congeneric and co-occurring invasive goby species (Rhinogobius cliffordpopei and R. giurinus) was examined to test the existence of an ontogenetic diet shift and to determine potential differences in trophic niche between species within a habitat and between habitats (i.e. littoral and profundal) within species. Results obtained from gut content and stable isotope analyses revealed a strong trophic niche shift during ontogeny, whereby juveniles mainly consumed macrozooplankton while sub-adults and adults increasingly consumed aquatic insects and Tubificidae. The trophic niche of juveniles was similar between species in each of the 2 habitats but significantly different for sub-adults and adults, notably in littoral habitat. Moreover, the trophic niche was similar between habitats for juveniles of each species, whereas it differed significantly between habitats for sub-adults and adults. This study demonstrates the importance of ontogeny and shows that habitat use can significantly affect food resource use and trophic relationships between 2 co-occurring fish species.


INTRODUCTION
Niche theory predicts that species-specific specialization in resource use is a primary mechanism allowing the stable coexistence among competing species within a local community (Chesson 2000, Kyla fis & Loreau 2011).Specialization in food re sources is crucial since it may substantially enhance coexistence by reducing interspecific competition (Gabler & Amundsen 2010, Kleynhans et al. 2011).In fish, for instance, some closely related species co-existing in the same ecosystem display strong food resource differentiation and trophic niche specialization (Colloca et al. 2010, Davis et al. 2012).Since the diet and foraging behavior of fishes are often strongly driven by environmental conditions such as prey availability (Shimose et al. 2010), competition (Kaspersson et al. 2010), and predation (Hammerschlag et al. 2010), the patterns of food use among potentially competing fishes might therefore also vary between habitats with different environmental characteristics (Davis et al. 2012).However, few studies have focused on the potential effects of habitat characteristics on trophic resource use among competing and closely related fish species.
Most animals show complex life cycles with significant shifts in their ecological niches during onto geny, and conspecific individuals can display varying trophic resource exploitation due to morphological, physiological, and/or behavioral differences be tween life-history stages (Specziár & Rezsu 2009).The magni tude of resource partitioning among competing species may therefore differ during ontogeny.Consequently, investigations based on a single life-history stage could potentially provide inaccurate information to elucidate the trophic relationships be tween co-occurring species (Specziár & Rezsu 2009).Fishes have complex age/size-structured populations, and the range of body sizes between conspecific individuals can sometimes span several orders of magnitude (Specziár & Rezsu 2009, Davis et al. 2012).During onto geny, strong niche shifts can oc cur, with individuals belonging to different trophic guilds during their life cycle.As a result, the patterns of food resource use among co-occurring fish species are often made more complex by ontogeny-dependent interactions (Naka yama & Fuiman 2010, Davis et al. 2012).Consequently, investigating food re source use between competing fish species at different life-history stages is crucial for a better understanding of their trophic relationship.
Rhinogobius cliffordpopei (Rutter 1897) and R. giuri nus (Nichols 1925) are highly invasive fish species in Yunnan-Guizhou Plateau of China (Xie et al. 2001, Yuan et al. 2010).They were inadvertently introduced into most lakes of the Yunnan-Guizhou Plateau in the 1950s and 1960s and introduced simultaneously to Lake Erhai in 1961 (Du & Li 2001, Xie et al. 2001, Yuan et al. 2010).They represent a rare case where 2 introduced species became simultaneously abundant and are considered to be one of the major causes of the decline and/or extirpation of native fishes in Lake Erhai (Du & Li 2001).R. cliffordpopei spawns from February to June, with a spawning peak occurring in March and April.Spawning activity of R. giurinus occurs from April to August with a peak activity during May and June (Guo et al. 2013).In lakes along the middle and lower reaches of the Yangtze River in China, where the 2 gobies are native, species diets (sub-adults and adults) are mainly large zooplankton (Cladocera and Copepoda) and aquatic insects (chironomid larvae) (Xie et al. 2000, 2005, Zhang 2005).However, no study has focused on ontogenetic trophic niche shifts and trophic relationships between the 2 species.Moreover, a strong spatial segregation between the 2 species occurs in Lake Erhai (Guo et al. 2012), and we aimed to test whether this is associated with food resource partitioning within and between habitats, a process that may facilitate the co-occurrence of the 2 species at high abundance.Specifically, the trophic niches of the 2 species were quantified using gut content analysis (GCA) and stable isotope analysis (SIA), and we aimed to determine (1) how the trophic niche shifts during ontogeny for each species; (2) whether the trophic niche in each habitat differs between species; and (3) whether the trophic niche of each species differs between habitats.

Study area
Lake Erhai is located in the Yunnan-Guizhou Plateau, southwestern China.It is a freshwater lake with a water surface area of 250 km 2 and maximum water depth of 21 m in 2010 (see details in Guo et al. 2012).The fish assemblage of the lake is currently composed of 28 fish species (7 native and 21 nonnative species).Rhinogobius cliffordpopei and R. giurinus are the most abundant benthic fish species.The littoral habitat (water depth < 6 m) of the lake was characterized by relatively high abundances of submerged macrophytes, R. cliffordpopei, and certain other fish species but low abundance of R. giurinus (including juveniles and adults, Guo et al. 2012).Conversely, the profundal habitat (water depth ranging from 12 to 20 m) was characterized by the absence of submerged macrophytes, high abundance of R. giurinus, but low abundance of R. cliffordpopei and other fish species (Guo et al. 2012).The main food resources available for the 2 gobies differed between the 2 habitats, i.e. the abundances of macrozooplankton and aquatic insects were higher in the littoral habitat compared to the profundal habitat, while the abundance of Tubificidae and shrimp larvae were higher in profundal habitat compared to littoral habitat (Guo et al. 2012).

Fish sampling
Rhinogobius cliffordpopei and R. giurinus were collected using benthic fyke nets from May to August 2010 (total 4 mo).The nets comprised 20 traps (each 0.60 × 0.62 × 0.35 m), 2 end-traps (1 m each), and 2 end-pockets (0.5 m each; Guo et al. 2012).The stretched mesh size was 0.4 cm.Five and 3 sites were sampled in the littoral and the profundal habitat, respectively (details in Guo et al. 2012).At each sampling site, 8 nets were deployed separately with a stone in each end of the nets.After 24 h, the catches in the end pockets were collected and identified to the species level, counted, and batch-weighed.In each month, all R. cliffordpopei and R. giurinus in the same habitat were pooled, and then sub-samples were taken for GCA and SIA.The samples for GCA were preserved in 8% formalin for 2 wk and then transferred to 75% ethanol for storage while the samples for SIA were stored at −20°C.

Gut content analyses
Total body length (L T ) and body mass (M T ) of each specimen analyzed for GCA were measured to the nearest mm and 0.01 g.Since there was no clear stomach structure for the 2 species, the foreguts (i.e.section of intestine from the esophagus to the posterior end of the first loop) were removed and preserved in 75% ethanol (Xie et al. 2000(Xie et al. , 2005)).Food items in non-empty foreguts (77.8% of the foreguts) were counted individually in a counting chamber and subsequently identified to the lowest possible taxonomic level under a dissecting microscope.To determine the percentage of weight of diet items, the weight of each food category was estimated.Following Chen (1981) and Zhang (2005), the estimated mean weight of Cladocera and Ostracoda was set at 0.023 mg ind.−1 , and the estimated mean weight of Copepoda and copepod larvae was set at 0.014 and 0.003 mg ind.−1 , respectively.The indirect volumetric analysis was used to estimate the volume of fish eggs, plant materials, and unidentified items (Hyslop 1980).The weight of these items was estimated as equal to their volume (specific gravity was assumed to be 1).After removing surface ethanol by blotting on tissue paper, shrimp larvae, aquatic insects, Gastropoda, and fish larvae were weighed to the nearest 0.1 mg (Hyslop 1980).For each food category, frequency of occurrence (FO), percentage of number (PN), and percentage of weight (PW) were calculated using the following formulas: where O i is the number of the guts that contain food category i, and n is the number of individuals used for gut content analysis, where N i is the number of an individual food category i, n is the total number of food categories identified in all guts and where W i is the weight of food category i, n is the total number of food categories identified in all guts (Hyslop 1980).Finally, diet compositions of the 2 goby species were estimated using an index of relative importance (%IRI) that was calculated using the formula: (4) where IRI i = (PN i + PW i ) / FO i , n is the total number of food categories identified in all guts (Assis 1996).FO, PN, PW, and %IRI i were calculated at the individual level for each life-history stage of each species.

Stable isotope analyses
To determine the stable isotope values of potential food resources in each habitat, macrozooplankton, Tubificidae, aquatic insects, and shrimp larvae were collected in May 2010.Macrozooplankton were collected with hand nets (standardized mesh size 64 μm) and counted under a dissecting microscope.Tubificidae and aquatic insects were sampled using a Peterson dredge (1/16 m 2 ; details in Guo et al. 2012).Macrozooplankton included Cladocera (dominated by Daphniidae, Chydoridae, and Bosminidae) and Copepoda (dominated by Cyclopoida).Tubificidae were dominated by Limnodrilus, and aquatic insects were dominated by Chironomidae.Shrimp larvae were collected in benthic fyke nets and were dominated by Atyidae (L T was 8.4 ± 5.1 mm, mean ± SD).Macrozooplankton, Tubificidae, and aquatic insects were kept alive in distilled water for 24 h for gut clearance.

Statistical analyses
Chi-squared (χ 2 ) tests were used to test for differences in diet compositions (%IRI) between lifehistory stages (juveniles, sub-adults, and adults) within species, between species (Rhinogobius cliffordpopei and R. giurinus), and habitats (littoral and profundal) within the same life-history stage (Gallagher & Dick 2011).The dietary niche breadth of each life stage of each species was estimated using the niche breadth (B) of Levins (1968): (5) where P i is the frequency of occurrence of prey item i in the diet of a consumer, and n is the number of prey groups.B values increase with dietary niche breadth.Moreover, Pianka's (1973) index (O) was used to determine trophic niche overlap between life-history stages in each species, between species at the same life stage within a habitat, and between habitats at the same life stage in each species.O was calculated as: (6) where P i is the frequency of occurrence of prey items i in the diet of species j and k, and n is the number of prey groups.O varies between 0 (total separation) and 1 (total overlap).A modification of the graphical Costello method was used to illustrate feeding strategy of the 2 species (Amundsen et al. 1996).The analysis is based on a 2-dimensional representation of prey-specific abundance (P i ) and frequency of occurrence (%O) of the different prey types in the diet.(7) where P i is the prey-specific abundance of prey i, S i is the stomach content (number) comprised of prey i, and St i is the total stomach content in only those predators with prey i in their stomach (Amundsen et al. 1996).
After examination of normality and variance homogeneity of the data using Kolmogorov-Smirnov and Levene's tests, 3-way AN OVAs were used to test for differences in stable isotope values (δ 13 C and δ 15 N were inverse-transformed and tested separately) between life-history stages, species, and habitats.If species and/or habitats had a significant effect on the isotope values of fish, Mann-Whitney tests were then used to test for differences in stable isotope values between species within habitats and/or between habitats within species at the same life-history stage.
The relative contribution of each potential prey to the diet of the 2 species and different life stages was estimated using stable isotope mixing models, i.e.Stable Isotope Analysis in R (SIAR, http://cran.r-project.org/ web/ packages/ siar/; Inger et al. 2010).Since there were no available trophic fractionation factor (TFF) values for the 2 goby species, we used an average TFF with a large SD as suggested by Inger et al. (2010).Specifically, we used TFFs of 1.0 ± 1.0 ‰ (mean ± SD) for δ 13 C and 3.3 ± 1.0 ‰ for δ 15 N (e.g.Inger et al. 2010).Mean δ 13 C and δ 15 N values (± SD, n = 3) of potential prey in each habitat were used in the models.As life-history stages and habitats significantly affected the values of δ 13 C and δ 15 N (see 'Results'), the models were run separately for the differ- ent life-history stages in each habitat.All statistical analyses were conducted in R 2.14.2 (R Development Core Team 2012).
Life-history stages had significant effects on δ 13 C and δ 15 N values (3-way ANOVA, p < 0.001, Table 2).Both species generally showed more enriched δ 13 C and δ 15 N values for juveniles compared to adults in the 2 habitats (Fig. 5).The mixing models estimated that the contribution of macrozooplankton was one order of magnitude higher for juveniles (43.4−74.1%)than for adults (4.2− 7.4%) of the 2 species in both habitats (Fig. 5).The contribution of aquatic insects and Tubificidae to the diet of juveniles of the 2 species (3.4−9.2%) was relatively low while they represented the main food resources of sub-adults and adults in both habitats (27.0−53.2%,Fig. 6).
δ 13 C and δ 15 N values differed significantly between the 2 species in both habitats (3-way ANOVA, p < 0.05, Table 2).In the littoral habitat, stable isotope values of juveniles did not differ significantly between species (Mann-Whitney tests, p = 0.289 in δ 13 C and p = 0.239 in δ 15 N), while sub-adults of Rhino go bius cliffordpopei showed significantly enriched δ 13 C (Mann-Whitney tests, p < 0.001) and δ 15 N (Mann-Whitney tests, p = 0.003) compared to R. giurinus (Fig. 5).Adults of R. cliffordpopei in the littoral habitat were significantly enriched in δ 13 C (Mann-Whitney tests, p = 0.010) but significantly more depleted in δ 15 N (Mann-Whitney tests, p = 0.015) than R. giu rin us (Fig. 5).In the littoral habitat, mixing models predicted that macrozooplankton and shrimp larvae contributed more to the diets of sub-adult R. cliffordpopei than R. giurinus, whereas aquatic insects and Tubificidae contributed more to the diets of sub-adult R. giurinus than R. cliffordpopei (Fig. 6).For adults in this habitat, R. cliffordpopei consumed fewer aquatic insects but more Tubificidae than R.
Stable isotope values of the 2 species differed significantly between habitats (3-way ANOVA, p < 0.001, Table 2).Specifically, sub-adults and adults of Rhinogobius giurinus displayed significantly enriched δ 13 C (Mann-Whitney tests, p < 0.001) and depleted δ 15 N (Mann-Whitney tests, p < 0.001) values in littoral habitat compared to profundal habitat, whereas δ 13 C and δ 15 N of R. cliffordpopei were similar for the 3 life-history stages between habitats (Mann-Whitney tests, p = 0.154, 0.245, and 0.378 in δ 13 C and p = 0.093, 0.356, and 0.289 in δ 15 N for juveniles, subadults, and adults, respectively) (Fig. 5).The mixing models predicted that subadults and adults of R. giurinus consumed more aquatic insects and fewer Tubificidae in littoral habitat than in profundal habitat.The contribution of macrozooplankton and shrimp larvae to the diet of each species was similar between habitats (Fig. 6).For R. cliffordpopei, the dietary contribution of the different prey did not differ between the 2 habitats (Fig. 6).

DISCUSSION
Macrozooplankton (Cladocera and Copepoda) were the main prey consumed by juveniles and were also an important food resource for sub-adults of the 2 goby species in the 2 habitats of Lake Erhai (Figs. 1 & 5).Our results are consistent with findings in lakes along the middle and lower reaches of the Yangtze River, including Liangzi Lake (Xie et al. 2005) and Biandantang Lake (Xie et al. 2000, Zhang 2005), where the 2 species are native.Both gobies shifted their diet by feeding increasingly on macrozoobenthos as they grew.Most goby species mainly consume macrozoobenthos when they settle on lake bottoms after their initial pelagic lifehistory stages (Kakareko et al. 2005, Borza et al. 2009, Grabowska et al. 2009).Of the macrozoobenthos, chironomid larvae are among the main prey consumed by species belonging to the genus Rhinogobius (Zhang 2005, Rusuwa et al. 2009).The estimated contribution of Tubificidae to the diet of sub-adults and adults was higher when using SIA than with GCA.These differences could be caused by the high digestibility and low detectability of Tubificidae in GCA, which do not affect estimates based on SIA (Inger et al. 2010, Polito et al. 2011, Cucherousset et al. 2012).In the present study, the trophic niche of the 2 goby species showed a strong ontogenetic diet shift from macrozooplankton to aquatic insects and Tubificidae.A similar shift has been observed for R. brunneus in the Ado River of Japan, where the 3 most important food items were Fig. 5. Rhinogobius cliffordpopei and R. giurinus.Stable isotope values (δ 13 C and δ 15 N, ‰) of R. cliffordpopei (circles), R. giurinus (triangles), and their potential food resources (mean ± SD, n = 3) in (A) littoral habitat and (B) profundal habitat of Lake Erhai (China) in 2010.Juveniles, sub-adults, and adults are displayed with white, grey, and black symbols, respectively.Potential food resources were macrozooplankton (MZ), shrimp larvae (SL), aquatic insects (AI), and Tubificidae (TU) Ephemeroptera, Diptera, and Cyano phyta for juveniles and Ephemeroptera, Trichoptera, and detritus for adults (Rusuwa et al. 2009).Moreover, we found that δ 15 N decreased from juveniles to adults.This pattern was driven by the fact that Cladocera and Copepoda (i.e.main diets of juveniles) had higher δ 15 N than aquatic insects and Tubificidae (i.e.main diets of adults).Generally, phytoplankton is the primary resource of zooplankton (Pietrzak et al. 2010, Yin et al. 2010).Aquatic insects and Tubificidae consume mostly detritus (e.g.leaf litter) and/or periphyton (e.g.particulate organic matter, POM; Füreder et al. 2003, Compson et al. 2013).Several studies have observed that algae could have more enriched δ 15 N values than detritus or POM (France & Schlaepfer 2000, Vizzini et al. 2002, Bunn et al. 2003), as these primary producers represent 2 different food chains.Potential higher δ 15 N values of primary producers in Lake Erhai were therefore likely to lead to higher δ 15 N values in zooplankton compared to aquatic insects and Tubi fi cidae, and thus to the observed de crease in δ 15 N during ontogeny of the 2 goby species.
In the littoral habitat, juveniles of the 2 gobies showed a similar trophic niche while sub-adults and adults displayed different trophic niches.Body size has a strong influence on trophic niche since it affects foraging efficiency (e.g.gape size) and/or competitive ability (Colloca et al. 2010, Nakayama & Fuiman 2010, Wasserman 2012).Since small individuals are gape-limited and are almost exclusively zooplanktivorous in the pelagic habitat, the competing fishes at this life-history stage rarely show clear trophic niche specialization (Colloca et al. 2010, Naka ya ma & Fuiman 2010, Rodrigues & Vieira 2010).Sub-adults or adults (larger-sized individuals) usually have better locomotion capacity and higher feeding efficiency (Borcherding et al. 2010), which are associated with a broader spectrum of prey size.Therefore, these individuals can specialize and partition different food resources, which may allow species co-occurrence among closely related or ecologically similar species within a community (Colloca et al. 2010, Davis et al. 2012).Several studies have demonstrated that sub-adult or adult gobies can exhibit food partitioning when they become abundant outside of their native ranges (Kaka reko et al. 2005, Borcherding et al. 2013).For instance, in the Danube River (Hungary), Ponticola kessleri mainly consumed Dikerogammarus spp.while Neogobius melanostomus principally foraged on chironomid larvae in spring (Borza et al. 2009).
In the profundal habitat, SIA indicated that the trophic niche of sub-adults and adults was similar between species, whereas their diets appeared different when analyzed using GCA.GCA provides a short-term estimate of individual diets, reflecting an individual's recent feeding history, while SIA is a more integrative tool that provides longer-term dietary records (Post 2002, Polito et al. 2011, Cucherousset et al. 2012).Therefore, it is very likely that over a season of growth, sub-adults and adults of the 2 species displayed different patterns of food partitioning in the 2 habitats, with different trophic niches in littoral habitat and similar trophic niches in profundal habitat.Trophic resource use among competing fishes may vary with environmental factors in - cluding habitat conditions (e.g.stable versus highly variable ecosystem, Davis et al. 2012), prey availability (Shimose et al. 2010), or density of competitors (Kim & Grant 2007, Kaspersson et al. 2010).In Lake Erhai, the 2 studied habitats are characterized by strong differences in depth, submerged macrophytes, food abundances, and density of predators (Guo et al. 2012).In addition, the abundance of Rhino gobius cliffordpopei was highest in the littoral habitat (water depth < 6 m), followed by sub-littoral habitat (water depth ranging from 6 to 12 m) and lowest in profundal habitat (water depth ranging from 12 to 20 m), whereas the abundance of R. giurinus were highest in the profundal habitat, followed by sub-littoral habitat and lowest in littoral habitat (Guo et al. 2012).Density of competitors (the 2 goby species) is also an important factor that influences foraging behavior and competitive processes (Kim & Grant 2007, Kaspersson et al. 2010).Therefore, further investigations are needed to elucidate the specific effects of those factors on the patterns of competition and food partitioning be tween the 2 invasive species.
Invasive gobies, in many cases, consume different diets in the face of spatial (Kakareko et al. 2005, Grabows ka et al. 2009, Rusuwa et al. 2009) and temporal (e.g.seasons, Grabowska et al. 2009, Rusuwa et al. 2009;months, Borcherding et al. 2013) variation in food availability.In our study, juveniles of the 2 goby species did not exhibit clear differences in diet compositions between littoral and profundal habitats due to the high dietary contribution of zooplankton (Rodrigues & Vieira 2010, Wasserman 2012).However, sub-adults and adults generally consumed more macrozooplankton and chironomid larvae in littoral habitat but more Tubificidae in profundal habitat.In Lake Erhai, the density of macrozooplankton and aquatic insects in littoral habitat is almost twice as high as in profundal habitat, whereas the Tubificidae are much more abundant in profundal habitat than in littoral habitat.The diets of sub-adults and adults seem to closely match the abundances of available food resources in the 2 habitats.Similar patterns were observed in Neogobius fluviatilis and N. gymnotrachelus in River Vistula of Poland (Kaka reko et al. 2005).
In conclusion, the present study demonstrated that these 2 closely related species displayed a strong trophic shift during ontogeny, with juveniles of the 2 goby species showing similar trophic niches while sub-adults and adults partitioned their food resources within the same habitat.Moreover, these 2 species showed different diet composition when food avail-ability differed between habitats.Understanding the competing interactions triggering food partitioning is crucial, and further experimental investigations should examine how exploitive competition (e.g.food intake, feeding rate) or interference competition (e.g.chase rate, attack rate, food holding capacity) varies with habitat conditions (food resource levels, population densities, predation risks) and mechanistically drives trophic niche segregation.

LITERATURE CITED
Amundsen PA, Gabler HM, Staldvik FJ (1996) A new approach to graphical analysis of feeding strategy from stomach contents data -modification of the Costello (1990) method.J Fish Biol 48: 607−614 Assis CA (1996) A generalized index for stomach contents analysis in fish.Sci

Fig. 4 .
Fig. 4. Rhinogobius cliffordpopei and R. giurinus.Feeding strategy diagrams based on the Costello method: prey-specific abundance plotted against frequency of occurrence of prey in the diet of 2 goby species in Lake Erhai (China).(A) R. cliffordpopei in littoral habitat, (B) R. giurinus in littoral habitat, (C) R. cliffordpopei in profundal habitat, (D) R. giurinus in profundal habitat