Coexistence of Two Closely Related Cyprinid Fishes ( Hemiculter bleekeri and Hemiculter leucisculus ) in the Upper Yangtze River, China

: Species coexistence is one of the most important concepts in ecology for understanding how biodiversity is shaped and changed. In this study, we investigated the mechanism by which two small cyprinid ﬁshes ( H. leucisculus and H. bleekeri ) coexist by analyzing their niche segregation and morphological di ﬀ erences in the upper Yangtze River. Morphological analysis indicated that H. leucisculus has posteriorly located dorsal ﬁns, whereas H. bleekeri has a more slender body, bigger eyes, longer anal ﬁn base, and a higher head. Niche segregation analysis showed spatial and trophic niche segregation between these two species: on the spatial scale, H. leucisculus was more widely distributed than H. bleekeri , indicating that H. leucisculus is more of a generalist in the spatial dimension; on the trophic scale, H. bleekeri had a wider niche than H. leucisculus . Therefore, these two species adopt di ﬀ erent adaptation mechanisms to coexist


Introduction
Species coexistence is one of the most important concepts in ecology for understanding how biodiversity is shaped and changed [1,2]. The mechanism that enable species-especially closely related, ecologically similar species-to coexist remains of interest in community ecology [3]. Although similar species commonly coexist in nature, classical niche theory suggests that complete niche overlap is evolutionally impossible [4,5]. Niche segregation, a process by which competing species evolve different patterns of resource-use under evolutionary pressures, is widely used to explain how similar species coexist [6][7][8]. Variation in resource-use can generally be separated into three aspects: space, time and food [9,10]. However, competing species tend to be segregated by at least one niche dimension [8,10,11].
Studies across diverse taxa, including birds [12], invertebrates [3,13], microbes [14,15], plants [16,17], fishes [1,8,18,19], amphibians and reptiles [20,21], have found that niche segregation is the main mechanism promoting coexistence. Rossier [22] confirmed the fish undergo spatial and temporal separation by identifying the spatial (distance from the shore) and seasonal (summer-winter) distributions of the ichthyofauna in the littoral zones of Lake Geneva. Knickle and Rose [5] studied the spatial and temporal movement patterns of sympatric juvenile Gadus morhua and Gadus ogac, using high-resolution radio-acoustic positioning in a coastal area of Newfoundland. Carniatto et al. [1] confirmed two sympatric, morphologically similar species of Moenkhausia with segregated food

Morphological Measurement
Samples of H. leucisculus and H. bleekeri for morphological measurements were collected from Mudong town in the upper Yangtze River during the period May-June 2016 ( Figure 3). A total of 58 specimens of these two species were measured (29 of each species), and 20 morphometric characteristics were measured for each specimen [43] (Figure 4). All measurements were taken by the

Morphological Measurement
Samples of H. leucisculus and H. bleekeri for morphological measurements were collected from Mudong town in the upper Yangtze River during the period May-June 2016 ( Figure 3). A total of 58 specimens of these two species were measured (29 of each species), and 20 morphometric characteristics were measured for each specimen [43] (Figure 4). All measurements were taken by the

Morphological Measurement
Samples of H. leucisculus and H. bleekeri for morphological measurements were collected from Mudong town in the upper Yangtze River during the period May-June 2016 ( Figure 3). A total of 58 specimens of these two species were measured (29 of each species), and 20 morphometric characteristics were measured for each specimen [43] (Figure 4). All measurements were taken by the same investigator to minimize artificial error, and morphometric characteristics were measured to the nearest 0.01 mm with a digital vernier caliper, except for body length, which was accurate to 1 mm. same investigator to minimize artificial error, and morphometric characteristics were measured to the nearest 0.01 mm with a digital vernier caliper, except for body length, which was accurate to 1 mm.
A previous study identified the different morphological traits between the two species by analyzing the morphometric data of the specimens collected from the different locations (Table S1) [44]. Those results coincided with the results of the present study, demonstrating that the two species evidently had different morphological traits no matter where the specimens were collected. Moreover, many studies have performed morphometric analysis on fish using 22-32 specimens of a species [45][46][47][48]. Therefore, the number of specimens analyzed in the present study was enough for a robust morphometric analysis.   same investigator to minimize artificial error, and morphometric characteristics were measured to the nearest 0.01 mm with a digital vernier caliper, except for body length, which was accurate to 1 mm.
A previous study identified the different morphological traits between the two species by analyzing the morphometric data of the specimens collected from the different locations (Table S1) [44]. Those results coincided with the results of the present study, demonstrating that the two species evidently had different morphological traits no matter where the specimens were collected. Moreover, many studies have performed morphometric analysis on fish using 22-32 specimens of a species [45][46][47][48]. Therefore, the number of specimens analyzed in the present study was enough for a robust morphometric analysis.   BD, Body depth; ED, Eye diameter; HL, Head length; HD, Head depth; DprD, Predorsal fin distance; DPrP 1 , Prepectoral fin distance-P 1 is an abbreviation for pectoral fin; LP 1 , Pectoral fin length; DPrP 2 , Prepelvic fin distance-P 2 is an abbreviation for pelvic fin; LP 2 , Pelvic fin length; DPrA, Pre-anal fin distance; LD, Dorsal fin length; LA, Anal fin length; LAb, Anal fin base length; CPL, Caudal peduncle length; CPD, Caudal peduncle depth [43].
A previous study identified the different morphological traits between the two species by analyzing the morphometric data of the specimens collected from the different locations (Table S1) [44]. Those results coincided with the results of the present study, demonstrating that the two species evidently had different morphological traits no matter where the specimens were collected. Moreover, many studies have performed morphometric analysis on fish using 22-32 specimens of a species [45][46][47][48]. Therefore, the number of specimens analyzed in the present study was enough for a robust morphometric analysis.

Data Sources
Our data consisted of a 1:250,000 drainage map and 90-m high-resolution Shuttle Radar Topography Mission (SRTM) Digital Elevation Models (DEM). The drainage map was provided by the National Geomatics Center of China and the SRTM DEM data were provided by the National Aeronautics and Space Administration (NASA) and the USGS National Imagery and Mapping Agency (NIMA) at 90-m resolution (http://srtm.csi.cgiar.org) [49]. These data provided information on the river, sub-catchment boundaries and altitude needed to calculate parameters. The information concerning the H. leucisculus and H. bleekeri distributions were obtained from surveys from our research group and other literature.

Technical Procedure for Classifying River Habitats
The habitat in this study was analyzed on the macro scale. The following habitat classification methods were used. First, appropriate classification indexes were chosen based on references. Second, the sub-catchment was generated on the ArcGIS platform based on the SRTM DEM data and the river drainage map. Then, the values of indexes within each sub-catchment were calculated with ArcGIS software, the indexes were analyzed by the cluster analysis method in R software, and the number of habitat types was determined according to the clustering results. Finally, each habitat type was named according to the dominant habitat factors.

Choice and Calculation of Habitat Classification Index
According to the scale characteristics of the habitats, the classification index should be able to reflect the medium-or large-scale characteristics of the river system, such as the physical form, water system structure, and scale. River slope, sinuosity, drainage density, stream order and altitude were selected as the classification indexes, based on a literature review [50] and the river characteristics.
Stream order [50] and altitude indexes were directly extracted from the 1:250,000 drainage map and SRTM DEM data in ArcGIS software.
The slope (m/km) of a reach was calculated using Equation (1)  The sinuosity of a reach was calculated using Equation (2) [52]: Drainage density (km/km 2 ) was calculated using Equation (3) [52]: The parameters, such as straightline downvalley distance, channel distance, channel length and drainage area, used in Equations (1)-(3), were directly extracted from the 1:250,000 drainage map and SRTM DEM data in ArcGIS software. we grouped all of the collected samples according to their standard body length, and randomly selected samples from each group. Thus, the average body length of the selected samples was approximately equal to the average body length of the population (Table S2). Our aim was to make the selected samples as representative of the population as possible. One dorsal muscle tissue (about 0.3 mg) was used for stable isotope analysis (SIA). δ 13 C and δ 15 N were measured with a Delta Plus (Finnigan, Bremen, Germany) continuous-flow isotope ratio mass spectrometer coupled to a Carlo Erba NA2500 elemental analyzer (Carlo Erba Reagenti, Milan, Italy) at the Institute of Hydrobiology, Chinese Academy of Sciences. Stable isotope ratios were expressed in δ, a deviation from the international standards of parts per thousand (% ) that was calculated according to the following equation: δ X = [(R sample /R standard ) −1] × 1000, where δ is the measure of heavy to light; the standards are Vienna Pee Dee Belemnite for carbon and atmospheric N2 for nitrogen [53]; X is 13 C or 15 N; and R is the corresponding ratio 15 N/ 14 N or 13 C/ 12 C. Three working standards (USGS40, USGS41 and UREA-Thermo) were employed to calibrate the analyzer. The average standard errors of replicate measurements for δ 13 C and δ 15 N were both less than 0.3% . After obtaining δ 13 C and δ 15 N, we corrected the values according to the fitting function obtained from the fitting of the measured and standard values of the three working standards.

Statistical Analysis
To eliminate any effect of size (length) on the dataset, the morphometric parameters were transformed according to Lahnsteiner and Jagsch [54]: Transformed morphometric parameter = measured morphometric parameter standard body length or head length × 100 . One-way analysis of variance (ANOVA) was firstly used to test whether those traits significantly differed between two species. Then a principal component analysis (PCA) was used to reduce the dimensionality of the variables, transform interdependent variables into significant and independent components [43,55], clarify the greater part of the variation, and extract new composite variables [56]. According to Kaiser-Meyer-Olkin (KMO) measures of sampling adequacy, and Bartlett's test of sphericity (Yakubu and Okunsebor, 2011), the factor analysis of the transformed morphological data set was valid (χ 2 = 1028.608; P < 0.01). To avoid super factorization, and to select variables that better represent morphology, only components with eigenvalue scores greater than 1.000 were considered, following the Kaiser-Guttman criteria [57]. The extracted principal components (PCs) were rotated using varimax rotation for simplifying factors, which could help interpret the factors or rotated PCs. Finally, an independent sample t-test was used to analyze traits comprising PC1, then the means of these traits were compared. The above procedures were all carried out using IBM SPSS Statistics (version 20.0, Armonk, New York, NY, USA).
In this study, catch per unit effort (CPUE) was expressed as g·boat −1 ·day −1 [58]. We investigated the CPUE of the two species based on the same sampling methods, using the same time for each location. Therefore, the CPUE results could be used to compare the population density of the two fish species in each location. The data for all five years were divided into spring and autumn seasons for each site, and then the CPUE data were compared between (a) different species from the same season and (b) different seasons in the same location.
One-way ANOVA was conducted to test whether there were significant differences in the δ 13 C and δ 15 N values between H. leucisculus and H. bleekeri. We also calculated the δ 13 C range (CR), the δ 15 N range (NR), the total area of the convex hull encompassing the data points (TA), and the corrected standard ellipse area (SEAc) for H. leucisculus and H. bleekeri [59][60][61], and then compared the trophic niches between these two species. All metrics were calculated using the R statistical computing package "siar" [62].
The principal component analysis of 21 morphometric parameters extracted six factors with eigenvalues > 1, explaining 75.59% of the variance ( Table 1). The first principal component (PC1) accounted for 26.00% of the variation, and the following PCs (PC2-6) accounted for 18.31%, 10.68%, 7.76%, 7.41% and 5.41%. For parsimony, only those factors with loadings above 0.60 were considered significant in this study.
Comparison of the mean and standard deviations of morphological traits in PC1 (Table 1) showed that H. leucisculus had greater values than H. bleekeri for BD/BL and DPrD/BL, but had smaller values for the measurements HD/HL, ED/HL and LAb/BL. Therefore, H. leucisculus had posteriorly located dorsal fins, whereas H. bleekeri had a slender body, bigger eyes, a longer anal fin base and a higher head. Table 1. Percentage of explained variance and weights of 21 morphometric characteristics (corrected based on standard body/head length) in six principal components (PCs), and the mean and standard deviation of morphometric characteristics in PC1. See Figure 4 for abbreviations. Sample number (n) = 58.

Spatial Niche
We identified 521 stream segments, according to the rules of river habitats classification [63], with lengths ranging from 0.216 to 337.904 km. Some sections of the river were short because there were many river junctions and the segmentation points were close to each other; these were mainly distributed in areas with good water systems. On the contrary, some rivers were poorly developed with fewer river intersections, yielding a longer river section.
According to the clustering results and the comprehensive characteristics of rivers, the river habitats in the upper Yangtze River were divided into six categories (Figure 1), containing 59, 60, 36, 104, 163 and 99 stream segments, respectively. The results of one-way ANOVA showed that river slope, sinuosity, drainage density, stream order and altitude were all significant in the six categories (P < 0.01). The six habitat types were named according to the differences in their index characteristics: sinuous headwater habitat, sinuous tributary habitat, steep headwater habitat, middle and upper sinuous river habitat, low altitude sinuous river habitat and upstream river habitat ( Table 2). The habitat classification in the upper Yangtze River, from the Yichang to Panzhihua reaches, and the distribution of H. leucisculus and H. bleekeri, suggested that H. leucisculus was distributed in types 2, 4, 5 and 6, mainly in types 4 and 5, and H. bleekeri was distributed in types 4 and 5 ( Figure 1).

Temporal Niche
Surveys over the entire sampling period, from 2012 to 2016, yielded 2460.2 g of H. leucisculus and 4870.6 g of H. bleekeri in total. The CPUE of H. leucisculus in the Mudong reach was 92.8 g·boat −1 ·day −1 in spring, and 137.7 g·boat −1 ·day −1 in autumn; the CPUE of H. bleekeri was 61.5 g·boat −1 ·day −1 in spring and 159.1 g·boat −1 ·day −1 in autumn. The CPUE of H. leucisculus in the Hejiang reach was 7.4 g·boat −1 ·day −1 in spring and 0.8 g·boat −1 ·day −1 in autumn; the CPUE of H. bleekeri was 58.9 g·boat −1 ·day −1 in spring and 34.8 g·boat −1 ·day −1 in autumn. The CPUE of H. bleekeri in autumn was 8.2 g·boat −1 ·day −1 in the Yibin reach. The stationary gill nets did not capture H. leucisculus or H. bleekeri in spring in the Yibin reach, or H. leucisculus in autumn in the Yibin reach.
The CPUE of H. bleekeri was higher than that of H. leucisculus in all reaches except the Mudong reach in spring. In the Mudong and Yibin reaches, the CPUEs of H. leucisculus and H. bleekeri in spring were lower than those in autumn ( Figure 5).

Trophic Niche
According to the one-way ANOVA, the δ 13 C and δ 15 N values were all significantly different between the two species (δ 13 C: F = 54.040, P < 0.001; δ 15 N: F = 31.897, P < 0.001). The δ 13 C range (CR), δ 15 N range (NR), corrected standard ellipse area (SEAc), and total area of the convex hull encompassing the data points (TA) of H. leucisculus and H. bleekeri at Mudong were all different ( Table 3). The CRs of H. leucisculus and H. bleekeri were 2.280 and 3.882, respectively. The NRs of H. leucisculus and H. bleekeri were 1.223 and 3.444, respectively. The SEAcs of H. leucisculus and H. bleekeri were 0.789 and 2.999, respectively. The TAs of H. leucisculus and H. bleekeri were 1.905 and 7.064, respectively (Table 3). Table 3. δ 13 C range (CR), δ 15 N range (NR), corrected standard ellipse area (SEAc), and total area of the convex hull encompassing the data points (TA) of H. leucisculus and H. bleekeri at the Mudong site.

Morphological Characteristics and Ecological Niches
Morphological characteristics provide evidence for important ecological characteristics, which might be indicators of life habits, resource utilization types or environmental adaptations of species [24,26,33,43]. Since recent studies have confirmed the close relationship between morphological features and ecological niches, it has proven useful to infer ecological information from morphological characteristics [25,26,33,34,43]. Morphological variations are tightly related to niche segregation.
First, body depth (BD) is related to swimming behavior. According to the hydrodynamic theory, fishes in lotic habitats have more slender bodies than those living in lentic habitats, as they reduce drag during steady swimming while searching for prey and during fast acceleration while attacking prey [54,64]. Generally, the location of the dorsal fin (DPrD) is associated with the fish's vertical position in the water. Fishes with posteriorly located dorsal fins adapt more easily to surface habitats in non-flowing water, and have the capacity to stabilize and brake when accelerating [56]. Eye diameter (ED) is assumed to be related to the visual sensitivity and foraging position of the fish in the water column [33]. Head depth (HD) is generally associated with food size, as fishes with larger heads always consume larger prey items [25,26,33]. Anal fin base length (LAb) is associated with maneuverability capacity and movement stabilization [65]. Fishes in lotic habitats usually have wider anal fins to improve their swimming and stability in a current [56].
In the present study, we found a clear morphological difference and strong relationship between morphology and ecology in these two species. The PCA results showed that H. leucisculus has a posteriorly located dorsal fin, whereas H. bleekeri has a more slender body, bigger eyes, a longer anal fin base, and a higher head than H. leucisculus (Table 2). Combining all of the above characters and their functions, we conclude that H. bleekeri can consume larger prey than H. leucisculus, and the former inhabits a part of the water column closer to the surface. Moreover, H. bleekeri swims better than H. leucisculus, and is better suited to living in a flowing water environment. As expected, morphological variation between these two species is indeed a reliable predictor of differences in their habitats and diet preferences [25,27,33,34].
In terms of the large distribution area, H. leucisculus and H. bleekeri, in the upper reaches of the Yangtze River, are only distributed in the lower and middle reaches of the main stream and tributaries at low altitudes, but it is difficult for them to live in the source and upper reaches. This may be related to the following factors: (1) Body size. Both H. leucisculus and H. bleekeri are compressiform, and fish with this body shape mostly inhabit areas with slow currents [25]. However, the river source and upstream areas have fast currents, high altitudes and large slopes [74][75][76]. Fishes living here are mostly anguilliform (e.g., Coreius guichenoti and Rhinogobio cylindricus) or depressiform (e.g., Homaloptera and Gastromyzon) [76]. (2) Feeding mode. The eutrophication degree of the water in the source and upstream areas is low. Most of the phytoplankton are sessile diatoms [77], and most of the aquatic arthropods and crustaceans are benthic. Therefore, the fish in these areas can either stabilize their bodies in the water flow and search for crustaceans in the bottom sand with their snouts, like Rhynchocypris does, or develop scraping habits [74], like Onychostoma and Schizothorax do. Therefore, fish in source and upstream areas are mainly adapted to the rapid current and benthic life, while H. leucisculus and H. bleekeri are pelagic fishes that struggle in these areas.
In the current study, H. bleekeri was usually distributed in areas of the upper Yangtze River basin with low altitude and slope, like habitat types 4 and 5. H. leucisculus, however, was distributed in habitat types 2, 4, 5 and 6, including the low altitude and slope areas (types 4 and 5)-which H. bleekeri preferred-and the slightly higher altitude and slope areas (types 2 and 6), which were difficult for H. bleekeri to occupy. Therefore, H. leucisculus is more widely distributed than H. bleekeri in the Yangtze River. Yibin had the highest H. leucisculus and H. bleekeri biomass in spring and autumn, followed by Hejiang and Mudong. Yibin is more upstream in the upper Yangtze River, Mudong is more downstream, and Hejiang is in between the two. This also shows that H. leucisculus and H. bleekeri are more suited to living in the lower reaches, with a lower altitude and slope.
When they coexist in a specific place, H. bleekeri is significantly more abundant than H. leucisculus. The results of the trophic niche analyses at the Mudong site showed that the food source (CR), trophic level (NR) and trophic niche width (TA and SEAc) of H. bleekeri were all greater than those of H. leucisculus, which indicates that H. bleekeri uses more resources and is more competitive than H. leucisculus when the two are sympatric. Therefore, when H. leucisculus and H. bleekeri are sympatric, to reduce the disadvantage of competition, H. leucisculus is forced to move to regions without H. bleekeri. According to the literature, H. leucisculus can survive in various habitats, such as reservoirs, lakes and rivers, even in other countries [78][79][80]. Some studies have also shown that the spawning habits of H. leucisculus are different in different regions. For example, it spawns adhesive eggs in Erhai [81], Fenhe Reservoir [82] and Dalai Lake [83], but pelagic eggs in Heilongjiang [84] and Erlonghu Reservoir [85]. In addition, H. leucisculus is invasive in some places (e.g., Erhai) [81]. All the above factors indicate that H. leucisculus has a strong adaptability, and this explains why it is more widely distributed in the upper Yangtze River.

Conclusions
In summary, niche segregation occurs between H. leucisculus and H. bleekeri in the upper Yangtze River, and such niche segregation and partitioning likely allows these two closely related species to co-exist. Of the three important dimensions of niche partitioning-space, time and food-food and space are the principal mechanisms of niche segregation between H. leucisculus and H. bleekeri in the upper Yangtze River, whereas we found no robust evidence of temporal segregation. H. bleekeri is more competitive, while H. leucisculus is more adaptive. In addition, there are morphological variations between H. leucisculus and H. bleekeri, and morphological characteristics are good indicators of a species' niche.
Supplementary Materials: The following are available online at http://www.mdpi.com/1424-2818/12/7/284/s1, Table S1: Comparison of the range of morphometric characteristics between H. leucisculus and H. bleekeri from reference "Fauna Sinica -Osteichthyes Cyperiniformes II" and this study, Table S2: Sampling location and sample information (body length, body weight, and their mean values) that used in stable isotope analysis, and the mean of all samples.