Individual dietary specialization reduces intraspecific competition, rather than feeding activity, in black amur bream (Megalobrama terminalis)

Individual specialization and high plasticity in feeding activity are common in natural populations. However, the role of these two in intraspecific competition is unclear. In this study, the rhythm of feeding activity, dietary composition, niche width, niche overlap, and individual specialization was explored in four different size groups of black amur bream (Megalobrama terminalis), using microscopic identification of foregut contents and stable isotope analysis (δ13C and δ15N) of dorsal muscle. Both methods observed ontogenetic shifts in dietary preference and individual specializations, and revealed that the total niche width of large individuals was greater than small individuals. Mixed linear models indicated that feeding activity was significantly influenced by time (p < 0.0001), and no significant changes among size groups was evident (p = 0.244). Niche overlaps revealed that there was intensive diet competition between different size groups of black amur bream. Individual specialization in small juveniles was likely to be stronger than sub-adult and adult groups. Pearson’s correlation analysis revealed that the individual specialization was positively correlated with mean diet similarity within a group. The results indicated that intraspecific competition is reduced mainly by individual dietary specialization, rather than shift in feeding activity.

www.nature.com/scientificreports/ variations among populations for evaluation of population dynamics. Individual specialization in diet plays a vital role in intraspecific variation, such as intestine length, and jaw structure 17 . Intra-and interspecific competition, diversity of available resources, and predation may modify the strength of individual specialization 11 . Optimal foraging theory (OFT) hypothesizes that the animal's diet will be such that the net rate of energy intake is maximized, depending on individuals' capacity to capture and digest those resources 18 . OFT explains why dietary differences between individuals of a population occur. Changes in consumer choice occur as the number of competing individuals increases 19 . The range of food items of a species increases as the number of individuals increases within a population, and the extent of individual specialization will also increase 20 . The most competitive or dominant individuals in a population negatively affect the foraging efficiency of weaker conspecifics, that then lose the best foraging areas, and have reduced access to preferred resources [21][22][23] . This demonstrates that intraspecific competition strengthens individual specialization.
In the present study, the role of feeding rhythm divergence, and individual specialization in reducing intraspecific dietary competition in a freshwater ecosystem was explored. Black amur bream (Megalobrama terminalis) is an economically important, omnivorous species which is abundant in the lower reaches of the Pearl River in China. It has been reported that black amur bream mainly consume golden mussel (Limnoperna fortunei), Asian clam (Corbicula fuminea), organic debris, and aquatic plants 24 . Small subunit ribosomal DNA sequencing analysis revealed that the black amur bream presented dietary shifts during gonad maturation 25 . During the past decade, biomass percentage of black amur bream in catches has decreased by 18.3% 26,27 , and the stock status of this species is overexploited 28 . Therefore, it is important to assess the implications of population niche, niche overlap, and variation in individual specialization of black amur bream in fish stock decline.

Results
Gut content composition. A total of 294 black amur bream were collected; the size of specimen ranged from 90 to 345 mm TL with a mean size (± SD) of 212 ± 50 mm. All samples were divided into four size groups based on standard length and gonadosomatic index. Among them 65 (22%), 91 (31%), 96 (33%), and 42 (14%) individuals were small juvenile, large juvenile, sub-adult and adult, respectively (Table 1). A total of 28 prey items from 37 intestinal tracts were assigned to taxa. The prey items that were identified to the order level represent over 98% (%W and %N) of all prey examined in each group ( Table 2). The percent by weight of gut contents was dominated by detritus across all groups with an average biomass percentage of > 84%. Small juveniles consumed a wider variety of prey that consisted of 20 prey taxa. Besides detritus, the most common prey taxa were Chaetophorales (5.5%W), followed by Coscinodiscales (0.2%W), and Araphidiales (0.2%W). Large juveniles fed on three major prey taxa besides detritus, Mytiloida, Chaetophorales and Ulvales, with biomass percentages of 13.8%, 1.4% and 0.1%, respectively. Sub-adult individuals mainly fed on detritus and Mytiloida, comprising 96.8% and 3.1% by weight respectively. Detritus and Mytiloida also dominated the adult diet, comprising 92.7% and 7.2% by weight, respectively.
The percent by number of prey taxa in a given individual gut was different to the percent by weight due to the different volumes. The amount of detritus was hard to determine, and we excluded while calculating the percent by number. Four prey taxa, Coscinodiscales, Chaetophorales, Chlorococcales and Aulonoraphidinales dominated the small juvenile diet, cumulatively comprising 87.1% by number ( Table 2). The prey taxa, Chaetophorales, dominated the large juvenile diet, comprising 43.4% by number. Sub-adults mainly fed on Coscinodiscales comprising 45.0% by number, and Aulonoraphidinales and Mytiloida were also recorded in rare instances ( Table 2). The most abundant prey taxon was Coscinodiscales (50.2%N) in the adult group, and Araphidiales, Biraphidinales, Aulonoraphidinales, and Mytiloida were also found in some adult individuals.
Overall, the gut content compositions of the different size groups did not vary greatly, as indicated by small variations for the major prey taxon detritus in percent by weight (Table 2). However, the prey taxa varied greatly between size groups in percent number. Black amur bream demonstrated different diet preferences in relation to body length.
Feeding activity. During the sampling period of 24 h, the black amur bream was considered to be actively feeding at all times (Fig. 1). The highest fullness index (FI) of small juvenile was observed around 16:00, and the lowest around 13:00. After sunset, before mid-night (19:00-22:00), the gut fullness stayed at an intermediate level. From 1:00 to 10:00, the gut fullness was at a low level. The large juveniles generally had a greater fullness than small juveniles, and feeding activity did not vary across all times. The highest gut fullness in sub-adults was observed around 1:00, and sharply decreased by 4:00, and continued to decline until 7:00. During the late morn- www.nature.com/scientificreports/ ing, (7:00-13:00), gut fullness was observed to be low. From 16:00 to 22:00, the gut fullness slightly increased and stayed at a medium level. Mean gut fullness of adult individuals was low during the early morning (1:00-7:00), and declined to a minimum at 10:00. During the afternoon, the gut fullness greatly increased and peaked at 16:00, followed by slight decline after sunset (19:00-22:00). The FF was the response of the mixed linear model; size group, sampling time and the interaction of these two factors were fixed effects; sampling season was the random effect. The results of the ANOVA tested models indicated that sampling time (χ 2 = 21.432, p < 0.0001) was a significant influence on the feeding activity. The size group did not significantly affect the fish feeding activity (χ 2 = 1.358, p = 0.244). This demonstrated that there was no significant change in feeding rhythm among individuals.
Trophic niche widths. A total of 46 samples of four size groups of black amur bream and 40 samples of potential prey sources were analyzed for δ 13 C and δ 15 N. There was a high degree of isotope overlap in δ 13 C and δ 15 N between the size groups ( Table 3). The values for black amur bream ranged from − 28.44 to − 22.08‰ for δ 13 C and from 7.23 to 21.76 ‰ for δ 15 N. Curve estimation indicated that the δ 15 N value of the black amur bream is a linear function of standard length. All coefficients were statistically significant, and the R-squared value was approximately 0.133 (Fig. 2B, δ 15 N = 8.90 + 0.02 × SL, p = 0.013). The δ 13 C value was negatively correlated (linearly) with fish standard length, but the relationship was not significant (p = 0.134, Fig. 2A). The δ 13 C and     Table 3. The C/N ratio (% weight) in potential prey taxa ranged from 3.17 to 77.32, where M. nipponense had the minimum value, and riparian C 4 plants had the maximum value. The mean proportional contribution of potential prey sources to black amur bream varied among taxa, observing a range from 0.02 to 0.15 ( Table 4). The contribution of 12 potential taxa to small juveniles ranged from 0.05 to 0.10, with C. fluminea in the highest proportion, and C 4 _P in the lowest proportion. The mean contributions of other prey to small juveniles were almost similar. The riparian C4 plant contributions were always low across all size groups. The mean proportional contributions of zooplankton, M. nipponense and C. fluminea were high, demonstrated to be very important for black amur bream. The contributions of M. nipponense and C. fluminea increased from small to adult individuals, and contributions of prey sources within groups also increased with body length. This appeared to be a dietary shift during fish growth.
The total trophic niche width was assessed using both gut contents and stable isotopes through TNW and SEA c , respectively. The higher values indicated a greater niche width in a given group. Both methods showed consistent results, with the trophic niche width increasing with body length. The values of the TNW and SEA c varied among size groups: adult > sub-adult > large juvenile > small juvenile (Table 5).
Niche overlap. The pairwise Morisita's dietary overlap index was generally > 0.6 between any two size groups, and very close to 1. Both carbon and nitrogen isotopic values exhibited no significant difference between the four size groups (ANOVA, p = 0.156 and 0.104 for δ 13 C and δ 15 N respectively). These similarities suggested niche partitioning among the four size groups, with large SEA c overlap (Fig. 3). The SEA c of four size groups overlapped with each other on the isotope biplot, with the largest overlap area occurring between sub-adult and adult, whereas the smallest overlap occurred between small juveniles and adult (Table 6). Both methods (Morisita's index and SEA c overlapping) indicated a high potential for resource competition within the population of black amur bream.
Individual specialization. Individual prey specialization was assessed using both gut contents and stable isotopes through the WIC/TNW g and CD, respectively. Low WIC/TNW g and high CD both indicated less similarity between individuals. Both methods provided higher individual specialization in small juveniles than in sub-and adult groups. Pearson's correlation analysis revealed that mean pairwise diet similarity was negatively correlated with WIC/TNW g (Pearson's correlation = − 0.642, p = 0.358).
The NR was found to be larger for sub-adults and adults than for juvenile groups, whereas CR was high in the sub-adult group, and low in the adult group (Table 5). A slightly higher mean trophic diversity (assessed by CD) was found in small juveniles compared with other size groups. The RPP test indicated that the CD in the small juveniles did differ significantly from the sub-adult (p = 0.026) and adult (p = 0.019) groups. The MNND did not differ significantly between size groups (p > 0.05). The Hotelling's T 2 tests revealed that centroid location did not differed significantly (Euclidean distance between centroids was not significantly different from zero) for the contrast of the small juveniles and adults (Hotelling's T 2 = 5.81, p = 0.089). Table 3. Summary statistics (mean ± SE) of δ 13 C, δ 15 N, and C/N in the different size groups of black amur bream and potential prey sources in the sampling site. Values are mean ± SD.

Discussion
Although black amur bream have been an economically important species in the Pearl River basin for many decades, only a handful of studies have explored their feeding ecology 24,25 . The goal of the present study was to improve the understanding of intraspecific dietary competition, and its influencing factors, which is important to the protection and sustainable use of black amur bream. Gut contents and stable isotope analyses were both used to compare feeding rhythm, niche overlap, and individual specialization among four size groups.
Prey composition different revealed by gut content and stable isotope. Remains of prey in the foregut samples were difficult to identify by microscopic observation due to digestion. The proportions of prey items in black amur bream were very different by weight or by number. Dietary preference shifts were observed during fish growth, where juveniles preferred to consume phytoplankton and adult consumed more animal material. Dietary shift is common in fish and are closely related to the ontogenetic changes in energy demands, and foraging performance 29,30 . Dietary composition of the black amur bream is supported by a previous study in the Pearl River with a high consumption of detritus 24 . Small subunit ribosomal DNA (18S rDNA) sequenc- www.nature.com/scientificreports/ ing also detected juvenile black amur bream with highly abundant Streptophyta, and adults with more benthic animals 25 . The 18S rDNA sequencing method identified prey items along with intestinal microbes, while gut content identification can more accurately determine the contribution of each item. However, microscopic evaluation of foregut contents overestimated detritus, but this did not great affect the calculation of dietary overlap due to the high percentage of detritus in each size group.
Stable isotope analysis provides a powerful tool to characterize carbon sources and trophic position 31 , but is biased to detecting specific trophic interactions as isotope values in potential prey often overlap 32 . The stable isotope analysis showed that almost all prey taxa contribute to black amur bream nutrition (Table 4), which may be attributed to the trophic interactions among prey items. However, isotope analysis revealed the variation in the contribution of given prey items to the four size groups with results concurring with gut content analysis. The number of prey items determined by isotope analysis was much less than visual, and molecular identification of gut contents 25 . Two isotope ratios powerful evaluated the flow of organic material from three prey items to consumers 33 , but it is hard to determine more than three preys due to the variation in trophic fractionations between preys and consumers [34][35][36] .

High intraspecific dietary competition of black amur bream. In addition to the shift in dietary
preference, there is an increase in the niche width of black amur bream during fish growth. Competing theories assumed that ecosystem productivity increases the niche width of omnivores, which decreases as consumers utilize distinctive prey items 37,38 . Trophic niche width is also closely related to the intensity of competition, which tended to increase under some environmental conditions 39 . Large individuals are generally superior to small sized individuals when competing for food, with a higher growth rate and wider dietary sources 40 . The niche width of adult black amur bream was larger than juvenile individuals may mainly due to ontogenetic changes in feeding ability.
Both methods visual identification and isotope analysis consistently revealed high intraspecific dietary competition in black amur bream. It has been reported that the primary production of phytoplankton in the mainstream of the Pearl River is less than 400 mg (m 2 d) −1 , and biomass of riparian and submersed plants is very low 41,42 . Low Table 4. Contributions of different potential prey taxa to different size groups of black amur bream, according to stable isotope Bayesian mixed models. CI95%, lower-higher confidence intervals.  www.nature.com/scientificreports/ abundance of available prey resources could be the reason for high intraspecific competition, and busy feeding activities occurring during the 24 h sampling period seem to prove this hypothesis (Fig. 1). The proportion of starving black amur bream (FF = 0 or 1) was about 11% by number.

Individual specialization reduces intraspecific dietary competition. Variation among individuals
within populations is an evidence of competition, and has important consequences for ecological interactions 13 . Natural selection holds that competition intensity should decline over time among sympatric populations due to adaptive change, facilitating natural population persistence 43,44 . Generally, high individual specialization represents populations that possess considerable diversity and great variation, promoting resource portioning and coexistence. The highest individual specialization was found in small juvenile black amur bream (lowest WIC/ TNW), which may be explained by the high individual variation cause by limited foraging abilities within this group. Diet similarity between individuals in a group was negatively correlated with WIC/TNW, which demonstrated that intraspecific diet competition was positively correlated with individual specialization. On average, the WIC/TNW value was 0.66 ± 0.21 (n = 78 populations), and individuals' niches were narrower than population niches 11 . The value of WIC/TNW g in small juvenile black amur bream was much less than the average level, indicating high intraspecific competition. Low foraging efficiency and high conspecific densities may contribute  www.nature.com/scientificreports/ to high competition in the small juvenile group 45,46 . Low variation among individuals in sub-adults and adult may be due to exploitation competition, where one individual depletes a food patch before another one arrives 4 . An extremely high value of WIC/TNW g was found in the large juvenile group (Table 5), probably because of the low prey taxonomic richness, which may underestimate the dietary variations among individuals within this group. The CD demonstrated same variation patterns with WIC/TNW g . However, CD in the small juveniles did not significantly differ from the large juveniles (p = 0.330), which proves that WIC/TNW underestimates the dietary variations in the large juvenile group.

Conclusions
Ontogenetic changes in dietary preference and plasticity in feeding activity are common in fish; however, the role of these two in intraspecific competition is unclear. The results showed that juvenile black amur bream preferred to consume phytoplankton and adult consumed more animal material. However, there were no ontogenetic changes in rhythms of feeding activity of black amur bream. Although dietary shift was found in black amur bream, there was great dietary overlap and intense competition between different sized individuals. Moreover, diet similarity within a size group was positively correlated with individual specialization. The results indicated that intraspecific diet competition was reduced by individual specialization, rather than divergence in feeding activity. The challenge for the future is to understand the contribution of genetic variance and morphological feeding differences to individual specialization, and how the strength of individual specialization changes with ecological interactions.

Materials and methods
Sample collection and preparation. The Pearl River is the longest river in southern China, with a total length and annual discharge of 2214 km and 10,000 m 3 s −1 , respectively. The present study was conducted in the Zhaoqing section (23°02ʹ-04ʹ N, 112°25ʹ-31ʹ E) in the lower reaches of the Pearl River. Fish samples were collected in the dry season of 2015-2017, and the flood season of 2017 and 2019. Sampling was carried out with circular cast nets (15 m diameter, mesh size 4 cm). Eight daily hauls of 1 h each were performed every 3 h throughout a 24 h period to cover the entire diel cycle. The samples, kept on ice immediately and subsequently refrigerated, were transported to the laboratory. Each fish was measured for standard length (SL) to the nearest 1 mm, weighed for body weight (W t ), gonad weight (W g ) and gut weight (W gut ) to the nearest 0.1 g, and sex and maturity stage was recorded. The foregut was removed, and the contents were weighed to the nearest 0.0001 g. A sample of dorsal muscle was dissected and stored at − 20 °C for stable isotope analysis. Gonadosomatic index (GSI) and fullness index (FI) were calculated by the biomass percentage of gonad and gut to body mass, respectively. The GSI and FI were defined as follows: Potential prey sources including phytoplankton, zooplankton, invertebrate, riparian C 4 plants, submersed plants, benthic detritus and sediment were collected in the dry season of 2017 and 2018. Phytoplankton was collected by a horizontally hauled phytoplankton net (25#, 64 μm) from 0.5 m below the surface, and repeated until enough sample was collected. The phytoplankton samples were suspended in distilled water for at least 2 h, and then passed through plankton net (13#, 112 μm) to remove zooplankton and large particles. The phytoplankton was filtered onto pre-combusted Whatman GF/F glass fiber filters (0.7 μm, 450 °C for 8 h). Zooplankton samples were collected using the plankton net (13#), with the same preparation method as phytoplankton.
Benthic macroinvertebrates were collected with a weighted Petersen grab (1/16 m 2 ), and the samples were sieved through a 420 μm sieve 47 . In the laboratory, benthic animals were picked and kept in distilled water for 12 h to empty their gut contents. The benthic detritus was collected from the residue after the benthic animals were removed. Sediment samples were collected using Petersen grab, benthic animals were removed, and the sample was divided into two parts at each sampling site. One part was acidified with hydrochloric acid vapor in an enclosed space for 24 h to remove any carbonate for carbon isotope analysis, and the other part used for nitrogen isotope analysis. Plants were collected by hand, and then rinsed in distilled water. Riparian C 4 plants were dominated by Hemarthria altissima and Cynodon dactylon, and submersed plants included Potamogeton crispus and P. pectinatus. Several individuals (5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18)(19)(20) of small species like Corbicula fluminea and Limnoperna fortune were pooled to make a single sample of adequate sample mass. The preparation of macroinvertebrates was as previously described 48 .
All samples for stable isotope analyses were freeze-dried at < 40 Pa and < − 40 °C for 48 h (FD-1-50 Plus, BIOCOOL) and then ground into a fine powder using an automated vibration ball grinder (except for the filtered samples). The stable isotopes δ 13 C and δ 15 N were measured using a Thermo DELTA V Advantage Isotope Ratio Mass Spectrometer with external Flash EA1112 HT Elemental Analyzer equipment. Stable isotope ratios were expressed as parts per thousand (‰) of the international standards. The Vienna Pee Dee Belemnite and atmospheric nitrogen were used as the standards for carbon and nitrogen, respectively. Analytic precision was ± 0.1‰ for δ 13 C and ± 0.2‰ for δ 15 N, respectively. The C% or N% of each sample was calculated using the relative peak area of the sample to a working standard.
Feeding activity. Fish fullness (FF) was divided into six levels by visual observation of morphology 49 where 0 is an empty intestine, 1 is 1/4 full intestines, 2 is 1/2 full intestines, 3 is 3/4 full intestines, 4 is full intestines, and 5 is intestinal distension. www.nature.com/scientificreports/ The FF was a response variable, and size group and sampling times were fixed effects. A mixed linear model was fitted to analyze the effects of body size and sampling time on feeding activity. Time-and group-level predictors were included in the model (full model: FF ~ Group + Time + Group × Time + (1|Season)), and new models were produced by reducing fixed effects step by step. To test the significance of fixed effects, the full model and other models were compared using ANOVA. Mixed linear models were fitted by the "lme4" package 50 . All statistical analyses were performed using R 3.6.1 51 .
Trophic niche widths. Foregut contents were identified, and sorted into taxonomic groups down to the order level if possible. Prey importance was expressed by relative numerical abundance (%N), and relative mass abundance (%W). Mean proportional contribution of potential prey sources to black amur bream based on stable isotope ratios was calculated using the Stable Isotope Analysis in R (siar) package 52 . The black amur bream consumed mixtures of plant and animal material, thus the trophic enrichment factor (TEF) was assigned as 1.3 ± 0.3 ‰, and 2.3 ± 0.18 ‰ for δ 13 C and δ 15 N, respectively 53 .
The mean distance to centroid (CD) measured the Euclidean distance of each sample to the mean δ 13 C and δ 15 N values, to indicate average degree of trophic diversity. The CD can be partitioned into trophic level diversity (δ 15 N range: NR), and carbon matter source diversity (δ 13 C range: CR). The isotopic niche width was based on the convex hull area (TA) that was calculated by the smallest convex polygon bounding the individuals in δ 13 C and δ 15 N niche space 54 . The TA directly measured population niche width, reflecting variation in niche dimensions along two isotopes, is widely used to describe the isotopic niche width of fish 55 . The standard elliptical area (SEA) also represents dietary niche partitioning for each group, and contains 40% of studied individuals of a population without being affected by outliers 56 . The SEA c was corrected by sample size in a bivariate distribution, being unbiased with respect to sample size 54 . The mean nearest neighbor distance (MNND) measured the trophic similarity between individuals through Euclidean distances in the biplot space 54 .
TA, SEA, SEA c were calculated using the Stable Isotopes Bayesian Ellipses (SIBER) package 54 . The differences between size groups for the centroid location, CD, and MNND were tested using a residual permutation procedure (RPP) and the parametric Hotelling's T 2 test 57 .
Niche overlap. Dietary overlap was calculated using the simplified Morisita's index 58,59 : where C ij , dietary overlap index for predators i and j; p ik and p jk , biomass proportions of predators i and j with prey k in their foregut. A C ij value over 0.6 is considered a significant overlap.
Isotopic niche overlap was assessed by the extent of the overlapping area between the SEA c of two groups, which was measured by the percentage of overlapping SEA c and compared between size groups 60 .
Individual specialization. The total niche width (TNW) of a population is the variance of total resource use of all individuals, can be divided into two components: the variation within individuals (WIC), and the variance between individuals (BIC) 61 . The relative degree of individual specialization can be assessed by the proportion of TNW explained by WIC, WIC/TNW, and ranges from 0 to 1. Smaller values of WIC/TNW indicate a lower individual overlap, and higher individual specialization 62 . The TNW and WIC of each group was expressed as follows 62,63 : where p i is the proportional numerical abundance of all resources used by individual i; q k is the proportion of the kth resource category in a group's niche, and p ik is the proportion of resource k used by individual i. Diet similarity calculates the mean pairwise diet similarity between all individuals in a group. The TNW, WIC/TNW, and diet similarity were calculated for each group by RInSp 64 .