Sympatric Speciation of Tibetan Loaches (Triplophysa) Driven by Dietary Niche Specialization Authors: Sijia Liu, Ruoyu Zhang, Zhengli Wang, Kemao Li, Wulong Ma, Jinliang Wei, Mingzhu Wang, Wen‐Jun Ma, Shengxue Chen, Kai Zhao, Guogang Li, Fei Tian Published: 2026 ABSTRACT ABSTRACTThe Qinghai‐Tibeta Plateau is a hotspot for endemic fish biodiversity. Multiple isolated glacial lakes provide an ideal scenario for verifying the speciation mode of highland ichthyofauna. Sympatric speciation, a controversial topic in evolutionary biology, is driven by ecological niche differentiation, including dietary specialization. The coexistence of Triplophysa leptosoma and Triplophysa longianguis in a small semi‐enclosed postglacial Sunmcuo Lake makes it an excellent model for testing speciation mechanisms. Phylogenetic and population genetic analyses revealed monophyletic groups of T. longianguis and T. leptosoma . Together with the estimated divergence time, our results confirmed the sympatric speciation of T. longianguis and T. leptosoma . Gut contents and stable isotope analysis demonstrated the overlap of food resources with species‐specific private diet items, which is consistent with the observed morphological variations in trophic structures. The sequencing of 16S rRNA genes and LEfSe analysis revealed similar compositions and different functions of the gut microbiota between T. longianguis and T. leptosoma , which might contribute to host adaptation to specialized diets. Overall, the results of the present study identified a new case of sympatric speciation of Triplophysa fish and provided novel insight into sympatric speciation underlying the synergistic mechanism of the gut microbiomes and dietary specialization, which improved our understanding on the formation of endemic fish biodiversity on the Qinghai–Tibeta Plateau. Introduction The Qinghai–Tibet Plateau (QTP) is characterized by chronic cold, high salinity and alkalinity, and limited nutritional resources and serves as a cradle for endemic fish. The Tibetan loach Triplophysa Rendahl 1933 represents the primary ichthyofauna of the QTP and is widely distributed in the lakes and rivers of the QTP and surrounding regions (Chen et al. 1996; Yuan et al. 2020; Qian et al. 2023; He et al. 2024). Phylogenetic and biogeographic evidence suggests that the speciation of Triplophysa fishes occurred during the uplift of the QTP with a complex evolutionary history (Qian et al. 2023). To cope with food shortages, Triplophysa species have evolved normal‐ and scrape‐feeding behaviors with morphological diversification in the trophic apparatus, allowing them to survive in food‐restricted environments (Feng, Tang, et al. 2019). Therefore, drainages on the QTP provide natural laboratories to test speciation modes under extreme environments. Sympatric speciation is one of the most controversial topics in evolutionary biology. It is defined as the origin of new species from a single species in the absence of geographic barriers (Elmer et al. 2010; Bird et al. 2012; Foote 2018; Richards et al. 2019). Four criteria have been proposed to approve sympatric speciation (Coyne and Orr 2004): (1) overlapping geographic distribution; (2) genetically based reproductive isolation; (3) sister taxa in phylogenetic relationships; and (4) no historical isolation. Although sympatric speciation is considered rare in nature, some compelling evidence, such as African cichlids in crater lakes (Barluenga et al. 2006), Howea palms of Lord Howe Island (Gavrilets and Vose 2007), and Darwin's finches (De León et al. 2014), has been widely regarded as the mode of speciation, which occurred under ecological conditions related to intraspecific competition for trophic resources (Barluenga et al. 2006; Bolnick and Fitzpatrick 2007; Getz et al. 2016; Sutra et al. 2019). The divergence of a panmictic population is caused mainly by disruptive selection through ecological competition for traits associated with assortative mating (Nosil et al. 2009; Getz et al. 2016; Recuerda et al. 2023). Diet is a key biotic component of the ecological niche and plays a critical role in ecological speciation and diversification. Diet differentiation has been hypothesized to be a driving factor for divergence in sympatric populations (Grant and Grant 1979; Zhao et al. 2009; Zhang et al. 2013; Galvez et al. 2022; Pillay et al. 2025). Analysis of dietary composition (i.e., stomach contents) and trophic position (i.e., stable isotopes) revealed that coexistent cichlid fishes, with diverse lip and mouth structures, shared broad dietary niches with species‐specific private resources. These findings indicated that selection favored individuals with the ability to explore alternative resources through phenotypic specialization (Galvez et al. 2022). Therefore, the collaboration between shifts in feeding preference and variations in trophic morphology facilitated disruptive selection and ecological speciation among phylogenetically closely related species (Todd Streelman and Danley 2003; Sun et al. 2022; Neves et al. 2025). The gut microbiomes play an important role in host adaptation to diverse nutritional resources, which expand host food resources and facilitate interspecific trophic differentiation (Greene et al. 2020; Kuang et al. 2022; Ren et al. 2025). In sympatric blind mole‐rats, the community composition of the host gut microbiota did not significantly differ, and their functional divergence was caused mainly by differences in host diet, suggesting the evolution of the gut microbiomes with dietary differentiation (Kuang et al. 2022). Studies on herbivorous and carnivorous coral reef fishes, mammals and benthic and limnetic ecotypes of three spine stickleback have demonstrated the concordance between gut microbiome shifts and host dietary specialization and have emphasized the influence of dietary selective pressures on gut microbiome assembly. These results indicate that the composition and function of gut microorganisms reflect host dietary niches and trophic morphologies and mimic host evolution and associated speciation (Rennison et al. 2019; Degregori et al. 2024). However, the role of the gut microbiota in the coexistence of fish on the QTP remains understudied. Triplophysa leptosoma (Herzenstein 1888) is widely distributed on the QTP and forms a sister relationship with Triplophysa longianguis (Wu and Wu 1984), which is endemic to Sunmcuo Lake, a semienclosed postglacial lake connected to the Yellow River through a torrential mountain stream. Previous study on Gymnocypris fishes in Sunmcuo Lake linked dietary differentiation to sympatric speciation (Zhao et al. 2009; Sun et al. 2022). These findings prompted our hypothesis that T. leptosoma and T. longianguis represented another case of sympatric speciation in Sunmcuo Lake, which was driven by diet specialization synergized with the divergence of the gut microbiomes. To test this hypothesis, we integrated mitochondrial genomes and Cyt b sequences to reconstruct the phylogenetic relationship and verify the speciation patterns of these two species. By combining gut content, stable isotope, and gut microbiome analyses, we aimed to infer the role of dietary differentiation in sympatric speciation and explore the influence of the gut microbiomes on dietary specialization in the two Triplophysa fishes. Our findings offer critical insights into the formation of ichthyofaunal biodiversity in the highland aquatic environment of the QTP. Materials and Methods Sample Collection Field investigations were performed from August 2021–2023 in Sunmcuo Lake, the Yellow River, the Yangtze River, the Heihe River, and the Chaidamu River systems on the QTP (Figure 1 and Table S1), which were approved and supervised by the Qinghai Provincial Bureau of Fishery. In total, 213 samples from T. leptosoma and T. longianguis were collected from the sampling area and recognized based on the description by Wu and Wu (1992) and He et al. (He et al. 2024). The dorsal fin of each sample was collected and stored at −80°C for DNA extraction. (a) Sampling locality. Black circles indicated the sampling sites. 1, 10, and 11 represented sampling sites in Hexi Basin; 2, 3 and 4 represented sampling sites in the Yellow River, Sunmcuo Lake and the Yangtze River; 6–9 represented sampling sites in the Chaidamu River; 12 represented a sampling site in Lake Qinghai Basin. (b) Triplophysa longianguis . (c) Triplophysa leptosoma . Both photos were taken by Dr. Sijia Liu. In addition, 24 samples of each species were transported to the laboratory and sprayed with 70% ethanol for sterilization. The fish were aseptically dissected, and the gut contents were gently squeezed from the intestine, which was disinfected with 70% ethanol and sterile PBS to remove transient bacteria. The gut contents were collected from 24 specimens of each species, of which 10 were subjected to microscopic examination and 14 were subjected to microbiome sequencing. Muscles of 10 samples from each species were collected from each species for isotope analysis. All animal procedures were approved by the Animal Care and Use Committee of the Northwest Institute of Plateau Biology, Chinese Academy of Sciences (Approval No. NWIPB‐202006). DNA Extraction and Mitochondrial Genome Sequencing Total DNA for mitochondrial genome sequencing and population analysis was extracted from the dorsal fin using the Tissue Blood DNA Extraction Kit (Tiangen, Beijing, China). The integrity of the resulting total DNA was tested via 1% agarose gel electrophoresis, and the purity and concentration were detected by an ultramicro spectrophotometer, Nanodrop NC2000 (Thermo, USA), and subsequently stored in liquid nitrogen for subsequent analysis. The extracted DNA from T. leptosoma and T. longianguis was broken by sonication into approximately 350 bp fragments, followed by end–repair and the addition of a sequencing adapter. PCR amplification was conducted, and the products were subsequently purified via magnetic beads for library construction. The sequencing library was checked on an Agilent Bioanalyzer using an Agilent High Sensitivity DNA Kit (Agilent Technologies, USA) and then quantified on a Promega QuantiFluor fluorescence quantification system (Promega, USA) using the Quant‐iT PicoGreen dsDNA Assay Kit (Thermo Fisher, USA). The qualified libraries were sequenced on the Illumina NovaSeq 6000 sequencing platform (Illumina, USA) at Personalbio Technology Co. Ltd. (Shanghai, China). Mitochondrial Assembly and Phylogenetic Analysis The raw data were processed for quality control using fastp (v0.20.0) to remove adapters and low‐quality reads. High‐quality reads were assembled into the complete mitochondrial genome using SPAdes (v3.10.1), which was annotated using Mitos2 (http://mitos2.bioinf.uni‐leipzig.de/index.py) (Bernt et al. 2013). The annotation results were further manually corrected according to published mitogenomes of Triplophysa fishes (Table S2). The secondary structures of tRNA were obtained from the annotation outputs of Mitos2. The mitochondrial genome map was generated using OGDRAW (https://chlorobox.mpimp‐golm.mpg.de/OGDraw.html) (Greiner et al. 2019). The base composition and amino acid contents were calculated using MEGA 11 software (Tamura et al. 2021), including A + T content, AT skew, and GC skew. Relative synonymous codon usage (RSCU) was analyzed using the RSCU module implemented in PhyloSuite (version 1.2.3) (Zhang et al. 2020). The complete mitochondrial genome of Triplophysa fishes was downloaded from a public database (Table S2), which was imported into PhyloSuite (version 1.2.3) (Zhang et al. 2020). MAFFT was used for sequence alignment, and MODELFINDER was subsequently used to select the best‐fit model. The phylogenetic tree was constructed using the maximum likelihood (ML) method and Bayesian inference (BI), which was illustrated via iTOL (version 6). Population Genetic Analysis The mitochondrial gene Cyt b was selected for PCR amplification using the universal primers L14724 (5′–GACTTGAAAAACCACCGTTG–3′) and H15915 (5′–CTCCGATCTCCGGATTACAAGAC–3′) (Xiao et al. 2001; Li et al. 2008). The PCR products were examined by electrophoresis and sent to General Biologicals (Anhui) Co. for Sanger sequencing on an Agilent 3730 platform. The sequences were assembled using SeqMan (DNASTAR, Madison, WI, USA); sequence alignment was performed via CLUSTALW embedded in MEGA X, and corrections were made by manual inspection (Thompson et al. 1994). DnaSP (version 5.1) was used to calculate the genetic diversity of the population (Librado and Rozas 2009), including the number of polymorphic sites (S), the number of haplotypes (H), haplotype diversity (Hd), and nucleotide diversity (Pi). Haplotypes of Cyt b sequences (Table S3) were used to construct the phylogenetic tree via PhyloSuite (version 1.2.1), with the same method used for phylogenetic analysis of the mitochondrial genome PhyloSuite (version 1.2.1). According to the lineages in the phylogenetic tree, K2P genetic distance and genetic differentiation (F st) were calculated using MEGA X and DnaSP (version 5.1). Molecular Dating The divergence time was estimated based on the Cyt b dataset using BEAST (version 1.10.4) (Suchard et al. 2018). The selected substitution model was HKY + I + G, with a lognormal uncorrelated relaxed clock model, combined with a Yule process tree prior and a piecewise linear with constant root population size prior. Given that no fossil record of Triplophysa fishes was available, two validated time calibrations were employed, including the split between Sinibotia superciliaris and Leptobotia zebra at 13.9 Ma (Kumar et al. 2022) and the estimated separation of Triplophysa bleekeri and Triplophysa leptosoma at 7.9 Ma (Feng et al. 2017). Four independent Markov chain Monte Carlo (MCMC) runs were conducted for 200 million generations each, with sampling every 1000 steps. The first 20% of the samples were discarded as burn‐in, and the remaining samples were combined using LogCombiner 1.10.4 in BEAST 1.10.4 (Suchard et al. 2018). Convergence was assessed by an average standard deviation of split frequencies < 0.01 and ESS values > 200 in Tracer v1.7.1 (Rambaut et al. 2018). The maximum clade credibility tree was constructed using the TreeAnnotator 1.10.4 module of BEAST 1.10.4 (Suchard et al. 2018) and edited using Figtree v1.4.3. Carbon and Nitrogen Isotope Analysis To determine the position of the trophic level and assess the feeding habits, carbon and nitrogen isotopes were determined in Triplophysa leptosoma and T. longianguis of Sunmcuo Lake. Dorsal muscle of 1 cm3 in size was cut from 10 samples of each species and dried in an oven at 55°C for 48 h until a constant weight was reached. The samples were ground into powder in a high‐throughput tissue grinder. Powder (0.6–1.0 mg) was coated with aluminum foil, and an elemental analyzer (Sercon Integra 2, UK) was used. IAEA‐600 caffeine was used as a standard (Letourneur et al. 2013). Carbon and nitrogen stable isotope analyses were carried out at Shanghai Paisenow Biotechnology Co. Ltd. Each sample was assayed three times and averaged to ensure the accuracy of the results. For the representation of carbon stable isotope abundance, we used the δ value (delta value), which is a widely accepted representation. Specifically, the δ value reflects the difference in isotope ratios between the sample and the laboratory reference material (i.e., the working standard) in terms of the difference in parts per thousand (‰, per mill). In the measurement, the difference in isotope ratios between the sample and the working standard is first calculated, and this difference is subsequently converted to a δ value relative to the international isotope standard to obtain the final representation of carbon stable isotope abundance. δ=RSA−RSTRST×1000=RSARST−1×1000 where δ represents 13C or 15N and RSA and RST are the isotope ratios (13C/12C or 15N/14N) of the measured sample and the reference standard sample, respectively. The error of sample analysis was within ±0.3‰. Student's t‐test was applied to compare δ13C or δ15N values between two species, with a p value of 0.05 set as the significance level. Examination of Intestinal Contents The intestines were collected from 10 samples of each Triplophysa species from Sunmcuo Lake and then were tied tightly at both ends to fix in 4% PFA for 24 h. The intestinal contents were rinsed in purified water three times and then placed in Petri dishes for stereoscopy examination (Nikon, Japan). Based on the morphological characteristics, the food items were identified as plant material and seeds, periphytic algae, insect pieces and larvae, and zoobenthos, which were weighed to determine their relative abundance (Figure S1). Unrecognized materials (less than 10% of total weight) were excluded from the diet analysis. DNA Extraction and 16S rRNA Gene Sequencing For 16S rRNA sequencing, DNA was extracted from the intestine using Mag Beads Fast DNA Kit for Soil (MP Biomedicals, CA, USA). The quality and quantity of the DNA were measured using 0.8% agarose gel electrophoresis with a Nanodrop NC2000 (Thermo, USA). The gut microbiomes of 14 samples of each species were investigated using 16S rRNA gene sequencing. The V3‐V4 region of bacteria was amplified via PCR using universal primers (Table S4). PCR with a negative control was performed, and the products were purified for library construction. PCR amplicons were quantified using Quant‐iT PicoGreen dsDNA Assay Kit (Thermo Fisher, USA) in a microplate reader (BioTek, FLx800). According to the manufacturer's protocol, the sequencing library was constructed using the TruSeqIM DNA Sample Prep Kit. After end‐repair, adapters were ligated with A bases, which were then purified using BECKMAN AMPure XP beads (BECKMAN, USA). After size selection, the quality and quantity of the sequencing library were determined using Agilent Bioanalyzer 2100 (Agilent, USA) and Promega QuantiFluor (Promega, USA). The constructed library was subsequently sequenced on the Illumina MiSeq PE300 platform (Illumina, USA) by Personalbio Technology Co. Ltd. (Shanghai, China). Data Processing The raw sequencing data were subjected to removal of sequencing adaptors and low‐quality reads by fastp (0.19.6) and then imported into Vsearch (v2.13.4) (Rognes et al. 2016) for OUT clustering and QIIME 22019.4 (Bolyen et al. 2019) for microbiome bioinformatics. Briefly, sequencing data were demultiplexed, followed by trimming of primers using the cutadapt plugin (Martin 2011). The sequences were then merged, filtered and dereplicated, and all the unique sequences were clustered at 98% followed by chimera removal. The non‐chimera sequences were re‐clustered at 97% to generate the OTU sequences and the OTU table. The representative sequences were aligned using mafft (Katoh et al. 2002) for phylogenetic tree construction using fasttree2 (Price et al. 2009). Taxonomy was assigned to the OTUs using the classify‐sklearn naïve Bayes taxonomy classifier in the feature‐classifier plugin (Bokulich et al. 2018) against Silva v132 at 99% of the OTU reference sequences (Quast et al. 2013). The microbial community of each sample was classified and recorded at the phylum, class, order, family, genus, and species levels. The alpha diversity of the gut microbiomes of the two species was evaluated based on the Shannon index, Simpson index, species richness index and Pielou index. The Mann–Whitney U test was applied to compare the differences in α diversity between the two species. Bray‐Curtis distances matrix was constructed using the adonis2 function in the vegan package (Dixon 2003) in R (v 4.4.2), which was applied to group difference comparisons using permutational multivariate ANOVA (PERMANOVA). The result was visualized using principal coordinates analysis (PCoA) with the ape package (Emmanuel et al. 2018) in R (v 4.4.2). Linear discriminant analysis effect size (LEfSe) was carried out to examine the variations in the gut microbiome between the two species (Segata et al. 2011). Taxa with absolute LDA scores > 4 were retained (Yang et al. 2025). The alterations in functional pathways were assessed based on the KEGG database for bacterial taxa via 16S rRNA sequencing (Langille et al. 2013). The neutral process in bacterial assembly was evaluated using the Sloan model in the R v 3.5.1 code (Sloan et al. 2006; Chen et al. 2019). Results Mitochondrial Genomes of T. leptosoma and T. longianguis The mitochondrial genomes (mitogenomes) of T. leptosoma and T. longianguis from Sunmcuo Lake were 16,565 and 16,571 bp in length, with GC contents of 43.02% and 42.61%, respectively (Table 1). Like Triplophysa species, the double‐stranded mitogenome contains 37 genes, including 13 protein‐coding genes (PCGs), 22 tRNA genes and two rRNA genes, among which NAD6 and 8 tRNAs (tRNAGln, tRNAAla, tRNAAsn, tRNACys, tRNATyr, tRNASer, tRNAGlu, and tRNAPro) are located in the light chain, and the rest of the genes are in the heavy chain (Figure 2). TABLE 1Statistics of the mitochondrial genomes of Triplophysa leptosoma and Triplophysa longianguis . T. leptosoma T. longianguis RegionLength (bp)A%C%G%T%A + T%G + C%Length (bp)A%C%G%T%A + T%G + C%Whole genome16,57128.6525.117.5128.7457.3942.6116,56528.1725.2317.7928.8156.9843.02atp668426.3225.8813.7434.0660.3839.6268429.6125.9223.2921.1850.7949.21atp816833.3326.7913.126.7960.1239.8816834.8221.6121.1222.4557.2742.73cob114126.3827.4316.329.8956.2743.73114123.7927.1818.5630.4654.2645.74cox1155125.3424.7618.7631.1456.4843.52155126.5127.5616.5629.3855.8944.11cox269128.5124.617.6629.2357.7442.2669119.3525.8135.4819.3538.7161.29cox378427.325.3816.9630.3657.6542.3578425.3424.8218.5731.2756.6143.39nad197526.0526.5616.5130.8756.9243.0897528.6525.1817.2228.9457.642.4nad2104527.0828.2316.1728.5255.644.4104533.3327.3813.6925.658.9341.07nad334926.0724.0716.0533.8159.8940.1134926.6125.5813.8933.9260.5339.47nad4138227.7124.4615.9231.9159.6240.38138226.6625.2617.4730.6157.2742.73nad4l29723.9127.9516.8431.3155.2244.7829724.6425.2117.1932.9557.5942.41nad5183927.9525.0715.3931.5959.5440.46183922.927.6117.8531.6554.5545.45nad652217.8215.930.4635.8253.6446.3652227.525.1116.0631.3358.8341.17OH74330.6921.814.2733.2463.9336.0774326.2625.8316.3731.5457.842.2OL3119.3525.8135.4819.3538.7161.293118.3915.5229.6936.454.7945.21rrnL165834.9221.6521.1122.3257.2442.76165725.7727.5216.6530.0655.8344.17rrnS95229.5825.6823.4721.2650.8449.1694930.8221.414.1333.6564.4735.53 Statistics of the mitochondrial genomes of Triplophysa leptosoma and Triplophysa longianguis . Mitochondrial genome structures of Triplophysa longianguis (a) and Triplophysa leptosoma (b). Blue blocks: CDSs of mitochondrial genes; red blocks: Transfer RNAs; yellow blocks: Ribosomal RNA; gray blocks: Others; black bars: GC contents; green bars: GC skew +; purple: GC skew –. Based on the mitogenomes of 35 Triplophysa species and outgroups ( Barbatula nuda , Lefua costata , Schistura balteata and Hedinichthys yarkandensis), the reconstructed phylogenetic tree classified Triplophysa fishes into three major clades (A, B and C), which to a certain extent reflected their geographic patterns (Figure 3). The cave‐dwelling loaches clustered in Clade A, which included Triplophysa rosa , Triplophysa nasobarbatula , Triplophysa baotianensis, and Triplophysa zhenfengensis , preliminarily on the southwestern QTP. Clade B1 included Triplophysa species mainly from the Yellow River and the Hexi River system at the northeast edge of the QTP. Clade C1 included mainly Triplophysa fish in the Tarim River, Yellow River, and Lake Qinghai basin. Furthermore, T. longianguis and T. leptosoma were grouped together with the closest phylogenetic relationship in Clade C1. In Clade C2, Triplophysa species on the central QTP, such as the Yarlung Zangbo River and lakes from Tibet, were classified as widespread species, such as Triplophysa stenura and Triplophysa orentalis. Phylogenetic positions of Triplophysa longianguis and Triplophysa leptosoma in the genus Triplophysa. ML and BI trees were constructed based on the sequences of the mitochondrial genomes, which demonstrated the same topology. The posterior probabilities and Bayesian support values are labeled on the branches. Triplophysa longianguis and Triplophysa leptosoma are indicated in red. Phylogenetic Analysis of T. leptosoma and T. longianguis in Sunmcuo Lake In total, we sequenced the Cyt b gene in 15 individuals of T. longianguis and 154 samples of T. leptosoma from Sunmcuo Lake, which generated 11 and 23 haplotypes, respectively. Additionally, 44 samples of T. leptosoma from the Yellow River, Lake Qinghai Basin, Changjiang River, Chaidamu River, and Heixi River yielded 25 haplotypes. By setting Leptobotia zebra and Sinibotia superciliaris as the outgroup, we constructed BI and ML phylogenetic trees based on 59 haplotypes from T. longianguis and T. leptosoma , which consisted of two clades. The first clade consisted of three lineages, including a lineage of T. leptosomas from the Yangtze River, a lineage of T. leptosomas from the Chaidamu River, and a lineage of T. leptosomas from the Hexi River, Lake Qinghai River, and Yellow River. The other clade included samples of T. longianguis and T. leptosoma from Sunmcuo Lake, which were grouped together in a paraphyletic relationship with T. leptosomas from other river systems (Figure 4). This phylogeny revealed that the monophyletic groups of T. longianguis and T. leptosoma in Sunmcuo Lake presented the closest genetic relationships. The estimated divergence time was 7.49 Mya (95% HPD = 5.525–9.436 Mya) between Triplophysa fish from Sunmcuo Lake and T. leptosoma from other water systems in the other clade. Additionally, the separation time of T. longianguis and T. leptosoma in Sunmcuo Lake was estimated to be 3.94 Mya (95% HPD = 2.781–6.608 Mya) (Figure 4). These results indicated that T. longianguis had a closer genetic relationship with T. leptosomas from Sunmcuo Lake than with those from other geographic populations. Phylogenetic tree of Triplophysa longianguis and Triplophysa leptosoma . ML and BI trees were constructed based on Cyt b sequences. The posterior probabilities and Bayesian support values are in black. The estimations of the divergence times and 95% HPD are labeled in red on the branches. Population Genetics of T. leptosoma and T. longianguis in Sunmcuo Lake Population genetics analysis revealed that the nucleotide diversity (Pi) was 0.0429 and 0.0088 and that the haplotype diversity (Hd) was 0.933 and 0.678 in T. longianguis and T. leptosoma in Sunmcuo Lake, respectively. Additionally, the nucleotide diversity (Pi) and haplotype diversity of T. leptosoma from water systems outside of Sunmcuo Lake were 0.0224 and 0.961, respectively (Table 2). The genetic distance and genetic differentiation between T. longianguis (Sunmcuo Lake) and T. leptosoma (Sunmcuo Lake) reached 7.762% and 0.80323, respectively, which were lower than those among geographic populations of T. leptosoma , ranging from 15% to 16% and 0.93–0.94, respectively (Table 3). Consistent with the phylogenetic tree, the lower genetic differentiation and genetic distance between T. longianguis and T. leptosoma in Sunmcuo Lake suggest that their genetic relationships are closer than those among geographic populations of T. leptosoma . TABLE 2Population genetics of Triplophysa fish in Sunmcuo Lake.SpeciesSequence numberIndexValue T. leptosoma (the Sunmcuo Lake)154S103H23Hd0.678Pi0.0088 T. longianguis (the Sunmcuo Lake)15S111H11Hd0.933Pi0.0429 T. leptosoma (out of the Sunmcuo Lake)44S94H25Hd0.961Pi0.0224Abbreviations: H, number of haplotypes; Hd, haplotype diversity; Pi, nucleotide diversity; S, number of polymorphic sites. Population genetics of Triplophysa fish in Sunmcuo Lake. Abbreviations: H, number of haplotypes; Hd, haplotype diversity; Pi, nucleotide diversity; S, number of polymorphic sites. TABLE 3K2P genetic distance (%) and genetic differentiation (F st) among Triplophysa fish populations. T. leptosoma (the Yangtze River) T. leptosoma (The Hexi River) T. leptosoma (The Yellow River) T. leptosoma (the Lake Qinghai Basin) T. leptosoma (The Chaidamu River) T. leptosoma (the Sunmcuo Lake) T. longianguis (the Sunmcuo Lake) T. leptosoma (the Yangtze River)0.880010.899970.950120.894600.940420.81361 T. leptosoma (The Hexi River)3.8500.860280.948820.845540.937690.74333 T. leptosoma (The Yellow River)4.7443.9590.556680.490430.940210.82107 T. leptosoma (the Lake Qinghai Basin)4.5543.7340.6650.760600.942810.84137 T. leptosoma (The Chaidamu River)4.6983.9471.0611.2080.939660.80105 T. leptosoma (the Sunmcuo Lake)16.20215.98016.19616.07116.2600.80323 T. longianguis (the Sunmcuo Lake)17.34116.14817.20116.96017.0987.762 Note: The lower left diagonal represents the K2P genetic distance (%) between two populations, and the upper right diagonal represents the degree of genetic differentiation. K2P genetic distance (%) and genetic differentiation (F st) among Triplophysa fish populations. Note: The lower left diagonal represents the K2P genetic distance (%) between two populations, and the upper right diagonal represents the degree of genetic differentiation. Dietary Analysis of T. leptosoma and T. longianguis It has been reported that closely related species are partitioned in resource usage to avoid competition, including food utilization. Therefore, we performed carbon and nitrogen isotope (δ 13C and δ 15N) analyses to explore the diet compositions of T. longianguis and T. leptosoma from Sunmcuo Lake, which revealed the separation of the two species in terms of diet preference (Figure 5a). No difference was detected in the stable N isotope (δ 15N) values, suggesting that T. longianguis and T. leptosoma have the same trophic levels. The relatively small stable C isotope (δ 13C) signature indicated that T. longianguis is a pelagic species that prefers to inhabit open water and forage with phytoplankton, whereas T. leptosoma, with relatively large δ13C values, tends to ingest benthic organisms (Figure 5b,c). Consistent with the isotope analysis results, plants and algae, as the major dietary components, accounted for 98% and 63.65% of the gut contents of T. longianguis and T. leptosoma , respectively (Figure 5d). Insects and zoobenthos were rarely identified in T. longianguis but accounted for 18.09% and 18.26%, respectively, of T. leptosoma (Figure 5d). Similarly, morphological differences in mouth structure were detected between T. longianguis and T. leptosoma in Sunmcuo Lake. The lips of T. leptosoma are wrinkled or milky, whereas smooth lips are observed in T. longianguis , which might reflect differences in feeding habits (Figure 5e,f). (a) Scatterplot of δ13C and δ15N isotopic values for Triplophysa longianguis (n = 10) and Triplophysa leptosoma (n = 10). Box plots of δ13C isotope values (b) and δ15N isotope values (c) for T. longianguis and T. leptosoma . Student's t‐test was applied to compare δ13C or δ15N values between the two species. ***p value < 0.005 in Figure 5b. (d) Profiles of the gut contents of T. longianguis and T. leptosoma . The color and length of each bar represent different dietary components and average proportions of food items. Lip structures of T. leptosoma (e) and T. longianguis (f). Intestinal Microbiomes of T. leptosoma and T. longianguis To confirm the difference in food preference, the gut microbiota was characterized in 13 samples of both T. longianguis and T. leptosoma from Sunmcuo Lake, which yielded 3,634,154 sequences. After data processing and quality control, 8667 unique OTUs were generated, which were classified into 667 groups at the genus level belonging to 33 phyla. Similar taxonomic compositions were observed for T. longianguis and T. leptosoma , and Cyanobacteria and Proteobacteria were the predominant phyla in both species, with a relatively high abundance of Actinobacteria in T. leptosoma and a relatively high abundance of Firmicutes in T. longianguis (Figure 6a and Table S5). At the genus level, Synechococcus was the most abundant bacteria in both species, and Rhodoplanes and Paulinella accounted for greater percentages of T. leptosoma , whereas Ruminococcus and Clostridium were more abundant in T. longianguis (Figure 6b and Table S6). Comparisons of alpha diversity revealed that the Simpson and Pielou indices were greater in T. leptosoma than in T. longianguis , suggesting that the gut bacteria of the two Triplophysa fishes in Sunmcuo Lake significantly differed (Figure 6c,d). Based on Bray‐Curtis distances, the gut microbiota structure significantly differed between T. longianguis and T. leptosoma at the species, genus, and family levels (PERMANOVA, p < 0.05) but not at the OTU, order and phylum levels (Figure 6e and Table 4). LEfSe analysis revealed that 24 species of the four phyla presented distinct abundances in the bacterial communities of T. leptosoma compared with those in T. longianguis . The abundances of Bacteroidales in Bacteroidetes, Actinomyces and C111 in Acidimicrobiia, and unclassified species and SC_I_84 in Betaproteobacteria were greater in T. leptosoma, and the abundance of the genus Ruminococcus in the phylum Firmicutes was greater in T. longianguis (Figure 7a,b and Table S7). The functional prediction suggested that the different bacterial communities might significantly influence fatty acid salvage in their hosts (Figure 7c). With low R2 values of 15.2% and 21.6%, the occurrence frequency of intestinal bacterial OTUs fit weakly to the Sloan neutral model in T. leptosoma and T. longianguis , suggesting that deterministic processes, for example, ecological selection, such as diet, might play an important role in microbial community assembly in T. leptosoma and T. longianguis (Figure 7d,e). Both PERMANOVA and Sloan neutral model suggested that the divergence in diet selection might have contributed to the differences in the composition and function of the gut microbiota, which further supported the hypothesis that the sympatric speciation between T. leptosoma and T. longianguis was driven by food preferences. Composition of the gut microbiota at the phylum (a) and genus (b) levels in T. longianguis (n = 14) and T. leptosoma (n = 14). Box plots of the Simpson (c) and Peilou (d) indices of the gut microbiomes of T. longianguis and T. leptosoma . The Mann–Whitney U test was applied to compare the differences in Simpson and Peilou indices between the two species. *p value < 0.05 in (c) and (d). (e) Bray–Curtis distance of PCoA representing the calculated distance between T. longianguis and T. leptosoma (permutational multivariate analysis of variance, PERMANOVA, p < 0.05). TABLE 4Summary of PERMANOVA results.LevelNumberMean of R 2 Mean of F p valueOTU2200.0601.650.065Species1590.0711.980.044Genus1380.0722.010.042Family1170.0701.960.049Order920.0712.000.047Class520.0681.880.061Phylum230.0782.190.065 Summary of PERMANOVA results. (a) Identification of significantly different taxa between the gut microbiota of Triplophysa longianguis and Triplophysa leptosoma from the phylum to the genus level via linear discriminant analysis effect size (LEfSe). (b) Cladogram of the phylogenetic position of the gut microbial lineages of T. longianguis and T. leptosoma . Each circle represents a classification (from the inner to outer rings, genus, family, order, class and phylum). (c) KEGG pathways significantly different between the gut microbiota of T. longianguis and T. leptosoma . Red and green in (a–c) indicate higher abundances in T. leptosoma and T. longianguis , respectively. Fit of the neutral community model (NCM) of community assembly in T. leptosoma (d) and T. longianguis (e). The solid line indicates the best fit to the NCM, and the dashed lines indicate 95% confidence intervals of the model prediction. The green and red circles represent OTUs with greater and lower frequencies than the prediction by NCM, respectively. Nm: Metacommunity size times immigration. R 2: The fit to the NCM. Discussion Origin of T. longianguis Through Sympatric Speciation Triplophysa leptosoma is widely distributed on the QTP, including Sunmcuo Lake, the Yellow River, the Lake Qinghai Basin, the Changjiang River, the Chaidamu River, and the Heixi River, whereas T. longianguis is restricted to Sunmcuo Lake, a semi‐enclosed lake that connects with the Yellow River through a small creek. The molecular phylogeny revealed the strong phylogeographic structure of T. leptosoma , with a monophyletic group of T. leptosoma and T. longianguis in Sunmcuo Lake. This phylogenetic structure implied that genetic differentiation within the geographic population of T. leptosoma might have occurred before the separation of T. leptosoma and T. longianguis in Sunmcuo Lake, which is consistent with estimated divergence times between T. leptosoma and T. longianguis (~3.94 Mya) as well as lineages between Sunmcuo Lake and the Yellow River (~7.49 Mya). Geological evidence has indicated that the formation of Sunmcuo Lake occurred after glacial retreat on the QTP (Lei et al. 2008). The Yellow River reached the present headwaters in the Zhaling‐Eling Basin until the Holocene (~0.03 Mya), suggesting that Sunmcuo Lake was maintained as an isolated inland lake throughout the late Pleistocene (Li et al. 2001). Therefore, the split between T. leptosoma and T. longianguis occurred in Sunmcuo Lake before it opened to the Yellow River. The phylogeny and divergence time reflected the development of historical drainage and supported the hypothesis that the sympatric speciation of T. longianguis from T. leptosoma in Sunmcuo Lake. Trophic Partitioning in the Sympatric Speciation of T. longianguis and T. leptosoma How divergence is achieved is the fundamental issue in sympatric speciation (Foote 2018). Compelling evidence suggests that feeding substrate preferences might facilitate trophic segregation and promote sympatric speciation through dietary niche partitioning (Zhao et al. 2009; Sun et al. 2022; Ramellini et al. 2024). In the present study, we detected a significant difference in the δ13C value between T. longianguis and T. leptosoma . Carbon stable isotopes are considered markers for feeding‐habitat axes (e.g., pelagic vs. benthic, shallow vs. deep) (Iken et al. 2001; Ñacari et al. 2023). Therefore, the lower δ 13C signature indicated the preference of T. longianguis to forage with phytoplankton, whereas the higher δ 13C value suggested the preference of T. leptosoma for benthic organisms, which was consistent with the examination of the gut content in the two coexistent Tibetan loaches. Both plants and algae were predominant in the food of T. longianguis and T. leptosoma , which might explain the lack of significant difference in the isotope analysis. Insects and zoobenthos were observed in the gut contents of T. leptosoma and were rarely identified in T. longianguis , which might result in higher δ 13C values in T. leptosoma . Species‐specific diets and stable isotopic signatures provide evidence of ecological isolation between two species, which reduces competition in coexistent species. Many studies have highlighted the importance of dietary niche partitioning in the divergence of coexisting species by avoiding interspecific competition for limited food resources (Correa and Winemiller 2014; Pillay et al. 2025). The sympatric speciation of Gymnocypris eckloni scoliostomus and G. eckloni eckloni in Sunmcuo Lake has been demonstrated to be driven by changes in their utilization of food resources (Zhao et al. 2009; Sun et al. 2022). The cichlid fish that coexist in Barombi Mbo crater Lake overlap in their dietary niches with the unique food resources of some species, which might contribute to adaptive radiation in sympatry (Galvez et al. 2022). These phenomena were consistent with our observations in T. longianguis and T. leptosoma , suggesting that the divergence in diet preferences might facilitate the sympatric speciation of T. longianguis and T. leptosoma . The current study identified evident differences in trophic morphology (e.g., mouth and lip structures) between T. longianguis and T. leptosoma , which was consistent with the distinct dietary niches inferred from stable isotope and gut content analyses. While these morphological variations likely contribute to diet partitioning, future studies integrating genome‐wide analysis, biomechanical modeling, and common‐garden feeding experiments will be needed to directly establish the causal links between trophic structures and food selection. Synergistic Mechanism of Dietary Specialization via the Gut Microbiomes The gut microbiomes influence host diet selection by mediating the metabolism of essential nutrients, alleviating food competition among coexistent species (Zhang et al. 2013; Perofsky et al. 2019; Trevelline and Kohl 2022; Yang et al. 2025). Comparisons between two sympatric blind mole‐rats revealed no difference in community composition but rather variations in the functions of gut microorganisms, revealing the correspondence of the intestinal microbiota with ecological dietary factors in sympatric speciation (Kuang et al. 2022). Additionally, significant differences in the structure and diversity of the gut microbiota were observed in two sympatric rodents, with alterations in dietary adaptation and carbohydrate metabolism (Ren et al. 2025). In the present study, we also observed a similar composition of intestinal microorganisms, which was attributed to the overlapping food compositions and habits between T. longianguis and T. leptosoma . Firmicutes are generally more abundant in herbivores, and the genus Ruminococcus in the phylum Firmicutes contributes to cellulose degradation (Cann et al. 2016). Triplophysa longianguis might benefit from the greater abundance of Firmicutes in the gut for the better digestion of the higher levels of plant materials and seeds in their food. The abundance of Bacteroides, Actinomyces and Betaproteobacteria, which are generally high in omnivores, is usually related to adaptation to diverse diets, with the ability to decompose proteins and carbohydrates, such as polysaccharides and chitin (Pepper and Gentry 2015). The relatively high abundance of these bacteria might contribute to the metabolism of insects and zoobenthos, the unique food resources of T. leptosoma . This result indicated the reciprocal influence of dietary specialization on gut microbiota assembly and function in that species‐abundant intestinal bacteria might further promote trophic specialization. Therefore, we propose that the ancestral population of T. longianguis and T. leptosoma likely harbored substantial genetic and morphological diversity (Figure 8). Upon isolation in Sunmcuo Lake, the limited food resources placed strong ecological pressure on the ancestral population and favored disruptive selection on heritable variation in trophic morphology (e.g., lip structure) and food choice. This diet bias led to population divergence, resulting in the formation of two lineages that specialize in phytoplankton/plants and benthic organisms. The different food resources further imposed selection on the gut microbiota composition, which promoted digestive efficiency and increased host fitness. This positive effect of the gut microbiomes on the host reinforced diet specialization, which in turn strengthened the divergence of the gut microbiota community. This host–gut microbiome interaction facilitated habitat partitioning and genetic differentiation and ultimately led to reproductive isolation and sympatric speciation of T. longianguis and T. leptosoma . Therefore, our findings suggest that the amplification effect of gut microbiota differentiation accelerated ecological and reproductive isolation, which promoted sympatric speciation of T. longianguis and T. leptosoma , which was initially mediated by diet bias. This synergy between the gut microbiota and diet highlights the involvement of multiple factors in sympatric speciation in the resource‐limited environment of the QTP. Conceptual diagram of the model of sympatric speciation in T. leptosoma and T. longianguis . The background photo was taken in Sunmcuo Lake by Dr. Kai Zhao (co‐author). Phylogenetics of Triplophysa Fishes The taxonomy of the genus Triplophysa has long been debated. Hedinichthys was classified as a subgenus of the genus Triplophysa in previous studies (Zhu 1989; Wu and Wu 1992) but was defined as an independent genus by He et al. (2024). The phylogenetic tree of the mitogenome revealed the monophyly of Triplophysa fish with clear separation from H. yarkandensis, which supported the taxonomic revision of Triplophysa and Hedinichthys as two genera. The genus Triplophysa is classified into four lineages, among which cave‐dwelling loaches are grouped into clade A, which is indicative of the potential adaptive mechanism of subterranean ecosystems. The phylogeography of Triplophysa on the QTP reportedly resulted from the occurrence of geological events during the uplift of the QTP (Wang et al. 2016; Jin et al. 2022). The geographic distribution of Triplophysa fishes in Clades B1, B2 and B3 was consistent with previous reports based on whole‐genome sequencing and mitochondrial genes, supporting the phased uplift process of the QTP (Wang et al. 2016; Feng, Zhou, et al. 2019; Qian et al. 2023). The geographic pattern of Triplophysa fish also indicated that allopatric speciation was the major speciation mode of the QTP. The paraphyly of T. leptosoma geographic populations might have resulted from population differentiation due to allopatric isolation. Geographic isolation is considered one of the major driving forces of population divergence among endemic fish on the QTP. A series of studies revealed that vicariance caused by the uplift of the QTP led to population differentiation in Triplophysa tenuis and Triplophysa robusta (Feng, Zhou, et al. 2019; Feng et al. 2026). Additionally, convergent evolution has been reported to mislead the taxonomic positions of geographic lineages in Triplophysa stoliczkae (Feng, Tang, et al. 2019). Multiple paraphyletic clades of T. leptosoma might be attributed to the phenotypic convergence of different ancestral species that evolved under the same eco‐environmental pressure on Sunmcuo Lake. Additional samples from all the populations might reveal the evolutionary modes of T. leptosoma . Conclusion In conclusion, our results confirmed that T. longianguis and T. leptosoma in Sunmcuo Lake originated through sympatric speciation, which was driven mainly by dietary niche partitioning as the key ecological factor. The differences in the composition and function of the gut microbiomes are consistent with diet specialization, which contributes to host adaptation to different diet niches. These findings expand our knowledge of the synergistic effects of dietary partitioning and the gut microbiomes on sympatric speciation and provide valuable insights into the formation and maintenance of fish biodiversity on the QTP. Further analyses are needed to elucidate the genomic basis of food specialization and its relationship with reproductive isolation. Author Contributions Sijia Liu: formal analysis (lead), funding acquisition (supporting), investigation (lead), writing – original draft (equal), writing – review and editing (supporting). Ruoyu Zhang: formal analysis (equal), investigation (equal), writing – original draft (equal), writing – review and editing (equal). Zhengli Wang: formal analysis (supporting), investigation (supporting), writing – review and editing (supporting). Kemao Li: data curation (equal), investigation (supporting), writing – review and editing (supporting). Wulong Ma: data curation (supporting), investigation (supporting), writing – review and editing (supporting). Jinliang Wei: formal analysis (supporting), investigation (supporting), writing – review and editing (supporting). Mingzhu Wang: investigation (supporting), writing – review and editing (supporting). Wenjun Ma: formal analysis (supporting), writing – review and editing (supporting). Shengxue Chen: investigation (supporting), writing – review and editing (supporting). Kai Zhao: funding acquisition (equal), writing – review and editing (equal). Guogang Li: funding acquisition (equal), project administration (supporting), supervision (equal), writing – review and editing (equal). Fei Tian: conceptualization (equal), formal analysis (supporting), funding acquisition (supporting), project administration (equal), writing – original draft (lead), writing – review and editing (lead). Funding This work was supported by the National Natural Science Foundation of China (No. 32071489), the Sino BON‐Inland Water Fish Diversity Observation Network, the Biological Resources Program (Chinese Academy of Sciences, No. CAS‐TAX‐24‐069), the 2023 Award Fund of Qinghai Provincial Key Laboratory of Animal Ecological Genomics (QHEG‐2023‐01), and the Youth Fund of Qinghai University (2024‐QNY‐6). Ethics Statement All animal procedures were approved by the Animal Care and Use Committee of the Northwest Institute of Plateau Biology, Chinese Academy of Sciences (Approval No. NWIPB‐202006). Consent The authors have nothing to report. Conflicts of Interest The authors declare no conflicts of interest. Supporting information Table S1: Sampling sites. Table S2: Mitochondrial genomes used in the study. Table S3: Haplotypes of Cyt b used in the study. Table S4: Sequencing statistics of gut microbiomes. Table S5: Bacteria compositon at phylum level. Table S6: Bacteria compositon at genus level. Table S7: LEfSe result of bacteria with significant differences between T. leptosome and T. longianguis . Figure S1: Examples of microscopic examination of gut contents in T. longianguis and T. leptosoma . (a) and (b) were zoobenthos; (c) and (d) were zooplankton; (e) and (f) were phytoplankton.