ddRAD-seq Reveals Cryptic Phylogeographic Structure and Adaptive Population Differentiation of Hyporhamphus Intermedius in Heterogeneous Environments from China Mainland

poor the the freshwater is an excellent adaptation sampled in from aquatic habitats to comprehend environmental and genetic contributions to phenotypic divergence.


Background
Elucidating the origins of phenotypic diversity and the processes that different forms of dispersal generate spatial patterns and structure of genetic variation remain a challenging endeavor for evolutionary biologists and ecologists (Avise 2000; Saenz-Agudelo 2015). Species occupy a heterogeneous environment on a large geographical scale, which may shape patterns of gene ow among populations across the distributing region and drive population divergence (Slatkin 1987;Riginos and Liggins 2013). Alternatively, when different populations experience different environmental conditions, local selective pressures can modulate the underlying genomic architecture and drive phenotypic divergence, thereby promoting local adaptation and novel natural lineage (George 2000;Via 2001; Nosil et al. 2009). Local adaption is of critical importance in the divergence and novel speciation, which occurs due to differential pressures of natural selection on populations from different environments result in allele frequency shifts, and advantageous variations accumulate over many generations to maximize tness in local environment, forming distinctive genetic and phenotypic traits eventually (Savolainen et al. 2013). Studies investigating the genomic basis of adaptation often compare genomic information and environmental variables of populations for the comprehension of the evolutionary forces driving local adaptation (Holt and Gaines 1992; Sultan and Spencer 2002; Wang et al 2016). Local selective pressure counterparts to certain geographical and environmental gradients, e.g. latitude, longitude, and altitude. Genome-environmental association analysis expects to manifest the prints of selection with increasing geographical scale owing to larger environmental gradient differences (Fraser et al. 2011). Overall, discerning relative contribution of geographical environment factors and identifying factors such as natural selection, genetic drift, and dispersion that modify patterns of genomic variation is critical in interpreting the mechanism of how organisms adapt to the local environment and initiate evolutionary diversi cation.
Fish are ideal organisms for studying local adaptive divergence and phenotypic plasticity (Rajkov 2018). Given the complex geomorphic conditions in China with dramatic climate change at high altitude and latitude, the hyperdiverse freshwater systems (e.g., Pearl and Yangtze rivers and Plateau lakes) exhibit remarkably high levels regarding sh diversi cation and peculiarly prone to phylogenetic endemism and novel species, as it is probable that these independent areas separated by long distance also facilitated differentiation and genetic landscape formation of shes (Stoddart 1978;Tang et al. 2017). Reported studies have highlighted that sh diversity assessments are based almost entirely on morphological differentiation, so that knowledge of the true diversity of different regions, as well as the underlying mechanisms shaping local adaptation and biogeography within the system, remains rare and limited depth (Wiens 2011).
Population genomic studies of sh in heterogeneous environments have thus far been restricted to model species for evolutionary mechanisms comprehension including salmonid (Narum et al. 2017), African cichlids (Brawand et al. 2014), and three-spined stickleback (Jones et al. 2012). Indeed, the phenomena can be broadly recognized in other similar species. In contrast to well-known the model organism, in few studies, Hyporhamphus intermedius (Family Hemiramphidae) is an anadromous halfbeak as forage sh that exhibits a large geographic dispersal within the estuarine area of China Mainland, spaning along the coasts from the East China Sea to the South China Sea (Collette and Su 1986;Li and Zhang 2011). Notably, the south-north divergence of its morphological traits is prominent. For example, the number of dorsal and anal ns, vertebra in the northern population is signi cantly higher than that in the southern population (Collette and Su 1986), which are notable for putatively adaptive features. Additionally, some halfbeaks can also be found in the upper reaches of the Pearl River, hundreds of miles away from shore, which involves a recent introduction intensively conducted since the 1980s (Zhu and Chen 1989), however, where the introduced populations originated from is unclear. The evolutionary dynamics remains another challenging aspect that Plateau Lakes created a unique set of environmental conditions including an altitudinal gradient of wide temperature range, low dissolved oxygen, probably leading to a number of rapids-adaptive features of introduced populations (Cui et al 2008; Sihai et al 2013). The species tolerate different selective pressures from increasing geographical scale and exhibit tangible evidence of strong adaptive capacities. The whole populations distributed in dissimilar habitats typically display broad phenotypic plasticity and environmental tolerance. Preliminary investigations imply extensive morphological diversi cation across this range, cryptic lineages within whole groups are incapable to identify (Collette and Su 1986). It is unknown whether the selective pressure resulting from environmental changes can be observed from large-scale sampling with extensive geographical coverage.
It is important to combine genetic, ecological components and phenotypes in interpreting adaptation and phylogeographical patterns we observed from empirical studies (Gamboa & Watanabe 2019). Under the rapid genetic technological development, taking advantage of molecular genetic approaches to access genetic diversity can enhance understanding of the process of underlying speciation in heterogeneous freshwater systems (Nosil et al. 2017). Double digest restriction-site associated DNA (ddRAD-seq) involves genome-wide sequencing reads into thousands of orthologous loci, when no reference genome is available (Andrews et al. 2016, Peterson et al. 2012, and has been reported as the most robust approaches to disentangling the evolutionary history of strong dispersal species. For instance, the high rates of mutation experienced by the whole genome allow more rapid accumulation of mutations, and its richer polymorphism loci could result in a better resolution when trying to differentiate populations and identify close relatives and local adaptation with its potential dispersal events among populations (Maroso 2018;Kyriakis et al. 2019). Entirely based on morphological assessments, our previous estimates of H. intermedius diversity within the group demonstrated phenotypic differentiation of intraspeci c biological traits and NGS data has been scarcely used (Zhu and Chen 1989;Collette and Su 1986). The increasing integration between morphological and molecular approaches to assess the diversity, evolutionary history and phylogeography of the exceptionally diverse ichthyofauna will foster a better understanding of the heterogeneous freshwater of speciation (Alter 2016; Fišer 2018).
As living within distributed heterogeneous environments, local abiotic parameters may have stronger effects on genetic diversity and drive the spatial distribution of freshwater species. Genetic structure and the relative strength of selection are hypothesized to in uence the distribution of genomic variation along with different spatial scales. Here, an integrated approach combining phenotypic, environmental and genomic data was used to elucidate the drivers and processes of local divergence of H. intermedius. The aim of this research was therefore to probe the genetic diversity, demographic history and environmental adaptation of H. intermedius using the ddRAD approach derived SNPs at a genome-wide scale by. Speci cally, we plan to address the following questions: (1) characterizing the spatial population structure of H. intermedius and genetic diversity at the genomic level; (2)  For these loci, a signi cant difference was detected in genotype frequencies and the number of these

Genetic diversity and population divergence
Ampli ed from 156 individuals, 61,898,184 bp region of genomics gave rise to 32,744 SNP sites. Based on all these variant SNPs, the evaluations of genetic diversity, including Ho, He, π and F IS , were estimated (Table 1). There was still a marginal but distinct difference in genetic diversity among groups for different watershed regions. Among the 8 populations, the DM, GM and SS showed slightly lower nucleotide diversity (0.1937 ~ 0.1954) and expected heterozygosity (0.1850 ~ 0.1891) than the others. High genetic diversity was observed in the northern and the plateau group. In addition, the value of F IS within the plateau populations was lower than other groups (0.1113 ~ 0.1147), which might be connected with random genetic drift and the founder effect after transplantation.
AMOVA tests based on 32,744 SNPs was employed to show that the majority of the molecular variances derived from the differentiation among groups (48.8%) or within populations (50.31%) rather than within groups (Table 2). Furthermore, pairwise F ST values of the southern group from the Pearl River System appeared a high level of differentiation (0.296 ~ 0.322) when compared to the remaining groups, coupling with low gene ow (0.527 ~ 0.595) ( Table 3), thus corroborating on a distinct cluster in the molecular phylogenetic tree (Fig. 4). However, genetic differentiation between each pair of populations within groups was lower. Due to the transplant event, the genetic divergence of both the northern and the plateau groups did not reach a signi cant level in a short time Genetics structure, phylogeography and migration Genetic structure and phylogenetic analysis determined that the 8 populations were divided into two clades three independent clusters, respectively. The sampled individuals from the Yangtze and Plateau Lakes Systems were assigned to one admixture by phylogenetic analysis, suggesting the close relationship between the two systems (Fig. 3). The southern group formed a distinct cluster that was well separated from the rest of populations in the phylogenetic relationship, lacking gene ow (0.527 ~ 0.595) and promoting genetic isolate. Similar phylogenetic topology was recognized by PCA which clearly divided the populations into three distinct lineages (Fig. 4).
Bayesian clustering analyses with high-density SNP markers inferred the optimal number of subgroups (K) by using Evanno's ΔK method and the likelihood at each K (Evanno et al. 2005). As the most parsimonious partitioning of individuals within species, the estimates for ΔK fall into two distinct patterns: the highest ΔK values at K = 3, followed by secondary values K = 4 ( Fig. 5a, b). Nevertheless, these analyses achieve the lowest likelihood at K = 3 and the ancestral barplots showed that the plateau groups and the northern were grouped into one cluster, which was not consistent with previous phylogenetic analysis. The barplot at K = 4 reasonably re ected the complex relationship that individuals can be attributed to 4 genetic clusters, identifying to the ancestral components, which they denoted as northern group (orange and pink), plateau group (green and pink), and the southern group (blue and orange) with no admixed genotype (Fig. 5c).
The analysis of population splitting patterns, the direction and magnitude of migration was conducted by TreeMix based on our SNP data (Fig. 6), and the tree rooted using SNPs variants of H. quoyi, one sister genus of H. intermedius, which revealed two key results: First, within two main clades, GM, SS, and DM populations, comprising of the southern groups, were well separated from the remaining sites, while the Plateau and northern groups were admixtures with strong support. 7 historical migration events were added to the tree sequentially. For all SNPs, the migration events were inferred for several pairs of taxa, including gene ow from the whole ve population except southern group into CM population (0.320), from northern and plateau groups into southern group (0.440), and from the northern group into Yangtze River System (SZ and CM; 0.345), as well as another 4 weak migration events (0.004 ~ 0.055). Noteworthily, the total direction of migration was non-random, moving from higher to lower altitude and higher to lower latitude.

Demographic history of ancestry populations
To assess deviations from neutrality across whole groups, the results for the overall SNPs dataset of H. intermedius showed that the signi cant positive value (mean Tajima's D = 2.773; Fig. 7), which revealed that the medium-frequency loci were polymorphic and dominant in the genomes, indicating effects of balancing selection or population subdivision, which maintained the higher genetic variance and contributed to local adaptation. There were fewer negative Tajima's D values and low-frequency loci, proving no strong positive selection.
Demographic history analysis based on posterior probabilities was conducted by ABC analysis to interpret the history of the genetic clusters detected in the phylogeography. Prior to this, we hypothesized that 4 scenarios were consistent with the demography of H. intermedius (Fig. 8c). The analysis results indicated that support for scenario 2 reached the highest level with a probability of 0.34 in direct approach (Fig. 8a) and 0.7577 in logistic approach (posterior probability, Table 4, Fig. 8b), whereas other models exhibited lower probability. Based on the demographic model, the scaled model values of effective population sizes varied greatly among regions: the population groups of PL had large population sizes (6.58e + 3), whereas the northern populations (NP) were smaller (6.53e + 3), and the southern population (SP) sizes was only 6.40e + 2 (Table.4).

Morphology and phenotype-environment association
Analysis distinguished pronounced changes of shape associated with the bases of the dorsal and anal ns. Furthermore, the numbers of dorsal and anal n rays showed a signi cant difference between the south groups and the remaining groups (P < 0.05), indicating the appearance of dramatic phenotypic difference which coincided with the genomic differentiation.
To explore the relationships between morphological traits and environmental heterogeneity, all the measured phenotypic and environmental parameters were conducted by RDA analysis (the redundancy analysis). After the Monte Carlo tests and the forward selection from RDA, water transparency, water temperature, and daily temperature range were selected as the key environmental variables which signi cantly in uence morphological traits of H. intermedius (P < 0.05; Fig. 9a, b), but dissolved oxygen content also contribute to certainly in uence on the traits with no signi cant p-value(p = 0.094 > 0.05).
Ordination diagram of RDA displayed that the rst two axes cumulatively explained 86.71% of the variance of the phenotype-environment relationship (Axes 1 and 2 modeled 53.7% and 33% of phenotype data respectively). Of the four variables, only water temperature was highly positively paralleled to the axis 1 of redundancy analysis, whereas daily temperature range, DO (dissolved oxygen) and transparency was negatively correlated with both axes. Axis 1 explained 53.7% of the total variability on the phenotypeenvironment relationship, while the second axis explained 33% of this variance in morphological traits (Table 5).
One main gradient along axis 1 was observed, linking to morphological data (Fig. 9a) and sites (Fig. 9b) being separated from three groups. The phenotypic traits such as dorsal and anal ns related to DTR and DO. While the UJAW(upper jaw length) and many other traits generally preferred warmer water, BW(body weight) and SL(standard length) were correlative with high transparency (Fig. 9a). In terms of sites, the plateau group performed a strong relationship with dissolved oxygen content, daily temperature range, while the southern group distribution was associated with the water temperature and JN population was admixed with the plateau group.

Discussion
Population structure and phylogeographic relationship Signi cant spatial genetic structure was evident in ddRAD-seq analyses and re ected the geographic arrangement of the mainland, intraspeci c lineage of species is often reported to be differentiated along latitude (north-south axis) and altitude (plateau-plain axis) ( It is generally supposed that long-distance dispersal events and the geographical distance isolation promote increases in genetic differentiation and decreases in genetic diversity (Wright 1943; Saastamoinen et al. 2017). However, we conclude that our original hypothesis of distance sequential diversi cation for differentiation was not supported. In this research, the mantel test results were insigni cantly correlated with the geographical distance between the different populations (R 2 = 0.2011, p = 0.096 > 0.05) and there appears no evidence of the impact of IBD (Isolation by distance model), running counter-response to the conclusions of previous studies. Taken together, the complex phylogeographic relationship consistent with ABC results suggested that H. intermedius diversi ed throughout the system and experienced subsequent range shifts that led to the current biogeographic distribution and complicated demographic history.

Genetic differentiation and cryptic lineages
In general, the levels of genetic diversi cation were observed in relation to spatial variation, especially along latitude. When compared with genetic variants, we observed that the number of polymorphic loci was present in the northern and plateau populations, signi cantly higher than the southern from Pearl River Systems. Meanwhile, the degree of polymorphism within populations at the genome level ranged from 0.1937 to 0.2376, in which the southern populations almost exhibited lower levels, in conjunction with the number of null loci in genome library was dominant. Consequently, strong geographical clustering of populations was observed, corresponding with our AMOVA estimation (variance among groups occupying up 48.8%). Furthermore, pairwise F ST analysis revealed signi cant genetic differentiation between the native and the introduced populations, but also detected the presence of high differentiation for the southern group, even regard as cryptic subspecies (F ST >0.25, Nm < 0.5), within the whole mainland. It is proved that the strong divergence of H. intermedius has appeared along north-south axis based on genetic analysis whereas Collette and Su (1986) resolved phylogeographic relationships within using morphological data. In the light of these causes, and given the presence of river barriers for populations' dispersion, we hypothesize that these lineages may represent genetically incipient species. Such differentiation could potentially be caused by selection or local adaptation and was reported in the sister taxon (Lim et al. 2016). The south of China mainland has been described as genetically differentiation and more diversi cation when compared to northern mega-regions related to the climate difference in the northern hemisphere (Zhang et al. 2012;Wen et al. 2004), a pattern referred to as southern richness to northern purity phenomenon (Hewitt 1996;. Such con ict, known as "northern genetic richness and southern purity", of genome variation decreasing from north to south alternatively interpreted as a result of northern origin of H. intermedius (Fig. 8c), addressed in several other animals' investigations (Nuñez et al. 2011).
The heterozygosity of the genome also determines the degree of inbreeding and genetic diversity of the population, in which the observed heterozygosity re ect the true levels of genetic variation within populations (Courtneë, et al. 2019). We found that Ho and He among whole populations were fairly low on a genomewide scale, which agrees with the previous nucleotide diversity. Meanwhile , coping with the detrimental effect of the decrease in tness. However, it is worth noting that positive F IS is not equivalent to necessary inbreeding, and it is presumably related to population structure effect, or allele problems such as loci is under selection pressure (either natural selection or breeding selection). Genomic variation plays a potentially complicated role in population adaptive differentiation and genetic polymorphism of H. intermedius, which is known to be an important predictor of plastic and adaptive potential.

Origins of the early introduced populations
The phylogenetic pattern observed from numerous SNPs at a genome-wide scale manifested the extreme hydrology of the Yunnan-Guizhou Plateau appears to have acted a central role in genetic differentiation and structuring diversi cation of Plateau group following initial colonization. Speci cally, individuals from the northern group serve as a lineage that is resolved as native populations to the genetically proximate the introduced group, which was evident by phylogeographic analysis and demographic inference and demonstrated the history record (Zhu and Chen 1989). Clear evidence of admixture as re ected in phylogenetic tree is present in whole genomes, although the northern are more distant from the Plateau Lakes and isolated by various continental barriers. According to the historical records, it was shown that H.intermedius was rst introduced into the Plateau Lakes in Yunnan Province under no planned programs in the 1980s, without a clear recording of migration routes during that time (Xiong et al. 2006  Both statistical genetic differentiation and an obvious geographic pattern to the population distribution in the diverse aquatic environment were expectedly identi ed in a background of migration. Introduced populations in the Plateau Lakes were implemented into an independent genetic cluster, though few individuals were an admixture with the northern group. Noteworthily, we identi ed the genetic divergence of the Plateau group and the northern group did not reach a high level, which is rather unexpected. The Plateau Lakes is geographically more adjacent to the Pearl River System than the Yangtze River System, which was hypothesized reasonably that exited one or more introgressive hybridization events, but its scenario was the lowest support among 4 competing scenarios by ABC analysis. Both pairwise F ST and phylogenetic analyses exhibited that all introduced populations were expected primarily to shape patterns across recent divergence, which was also supported by the inference of ancestral alleles. Similar

Local diversi cation in heterogeneous environments
The combination of morphological characters and genomic data together with environmental heterogeneity elucidates a scenario of diversi cation in H. intermedius. Besides genetic differentiation at the molecular level, we also found that the phenotypic plasticity of morphological traits: dorsal, anal ns rays and traits relative to the bases of the dorsal and anal ns perform local differences in the southern populations, indicating that local diversi cation along the north-south axis. This morphological polymorphism is in congruence with phylogenetic patterns that initial position is the southern group outside the remaining clades (Fig. 3) and its relatively old age is correspondent to splitting time and branch length while no large phenotypic difference occurs between plateau colonization and native population. Actually, phenotypic differentiation among watersheds detected in the research appears to be driven by selection pressure related to spatial heterogeneity in transparency, water temperature, daily temperature range and DO, which have been shown to be important determinants of phylogeography structure in many other species (Selleslagh et al. 2008; Guo et al. 2015). Despite, these associations with morphological polymorphism and ecological environment do not con rm causality necessarily, as the relatedness may derive from interaction or some unrecognized factor that is strongly associated with one detected parameter (Guo et al. 2016), suggesting that increased sample sizes or greater numbers of species may be needed to detect the effect. Interestingly, sampling sites within a geographical region were located at different altitudes and the plateau group perform strong relationship with dissolved oxygen content, daily temperature range, which re ect the plateau climate intensity on temperature

Conclusions
Our phenotypic and environmental results indicate that hydrological barriers and abiotic factors (water temperature, daily temperature range and transparency) perform a critical function in molding phylogeography and diversi cation on large scales. Genomewide SNP data broaden insights into the complex spatial genetic structure (two clades and 3 clusters) and demographic history of H. intermedius in China Mainland, especially uncovering genetic differentiation from three heterogeneous hydrobiological regions and serving as an essential source of genetic polymorphisms for divergence and adaptation. Furthermore, we also demonstrated complex demography that the Plateau populations of halfbeak originated from the northern groups with strikingly pairwise FST increasing, whereas the signi cantly high level of divergence early occurred in both genotype and phenotype to the southern group, which was proposed on an isolated lineage that may develop to incipient species. Furthermore, we demonstrated hypothesis that H. intermedius underwent balancing selection which is also likely indispensable to survival during adaptation to new habitats, respectively. Although IBD was not detected among remote groups, the latitude and altitude distribution of H. intermedius likely has an adaptive genetic basis. We offer robust validation that variable migratory behaviors, genetic diversity, population structure of H. intermedius and their adaptive divergence in heterogeneous environment arisen from the interaction of balancing selection and strong gene ow at microscales, as well as strong impact of abiotic factors.

Study sites and sample collection
The total specimens were collected from sh docks or local markets in 8  The ligation products were ampli ed in PCRs using a Veriti 96-well thermal cycler (Life Technologies) and the protocol consisted of initial denaturation at 98 °C for 30 s, 14 cycles (98 °C for 15 s, 65 °C for 30 s and 72 °C for 30 s), followed by a nal step at 72 °C for 5 min. DNA libraries were quanti ed using a 2100 Bioanalyzer (Agilent Technologies). Finally, pools were combined in equimolar concentration to form a single genomic library and sequenced on an Illumina HiSeq 2500 using 150 bp pair-end reads.

Population genetic and statistical analyses
All raw sequences from the Illumina HiSeq lanes were checked for initial quality using FastQC, then quality ltering, assembly, SNP discovery of per individual was conducted for bioinformatic analysis. The population script was applied to calculate population genetic diversity parameters such as nucleotide diversity (π), expected and observed heterozygosity ( The relationship between geographic coordinates and genetic structure was also identi ed in principal component analysis (PCA) using the R package "adegenet" (Jombart 2008, Jiang et al. 2018, which was de ned as the rst two components of the PCA. Phylogenetic relationships of H. intermedius among locations were constructed using maximum likelihood (Neighbor-Joining tree) methods by bootstrapping over loci for 1000 times. We choose Hyporhamphus quoyi (its closely related species as outgroups) to root the tree. The results were performed on iTOL (https://itol.embl.de/). Then, the maximum-likelihood tree to infer gene ow and the historical admixture events amongst populations was generated by TreeMix (Pickrell and Pritchard 2012) on SNPs data. We rooted the population graph with Hyporhamphus quoyi, and tested the signi cance of migration events with various migration rates to identify genetic drift amongst populations.
In order to investigate past demographic expansion, and neutrality tests, Tajima's D values (Tajima 1989) was calculated with the 20000 bp sliding windows. Historical changes in population size on the basis of approximate Bayesian computation (ABC) methods were conducted in DIYABC v2.1.0 from a subset of the genetic data (Cornuet et al. 2014). We eliminated variants with low minor allele frequencies (MAFs < 0.1) or the loci that completely miss in at least one regional population. 400 thousand simulated datasets including four different demographic competing scenarios were drawn from prior distributions with the consequences of population genetic structure. All 8 population were divided in advance into the following three regional groups based on the morphometric and population structure: Pop1-he Southern populations ( nd the most plausible one, ABC procedure was employed to infer the most probable evolutionary history. We characterized 4 scenarios by demographic and historical parameters that were marked as the number of generations back in time (t1, t2). The effective population sizes of three regional groups and the ancestral population were represented by ST, PL, NT, NA respectively. All uniform was chosen for prior distributions and effective population sizes be constant in time. In short, both DIYABC's direct and logistic regression to assess posterior probabilities were chosen for comparison to enable a ranking of scenarios. Finally, each scenario posterior probability was computed based on 10% simulated data closest to the observed data using a logistic regression procedure. The summary statistics were scaled by the mean effective population size of the present four populations because the mutation rates of the SNPs were unclear.

Phenotype-environment association
To evaluate associations between the phenotype and environment in driving spatial genetic differentiation, the morphological data were relevant to environmental variables using redundancy analyses (RDA) (Rao, 1964), which utilize matrices of dependent and independent (explanatory) variables. The dependent matrix consisted of the 13 standardized phenotypic variables (SL, Body Weight, Dorsal n, Anal n, PAL, PL, UJAW, ED, HL, CL, CD, BD, CFL), while 15 explanatory abiotic factors were regarded as the explanatory variables in one explanatory matrix to assess the phenotypic variation (Table.S2 Animal Use Protocol within the project "Population Adaptive Differentiation and Genetic Polymorphism of Hyporhamphus intermedius in Heterogeneous Environments" was approved by the committee (IACUC, SYSU). During the sampling, H. intermedius was obtained from living or dead individuals. As H. intermedius is not endangered or protected, we don't need any permission to collect the animals used in our study. All surgery for living samples was performed under MS-222 anesthesia, and every effort was made to minimize suffering.

Consent for publication
Not applicable.
Availability of data and material DNA SNP data and other analysis data have been deposited at supplementary data.

Competing interests
The authors declare that they have no competing interests.      The principal components analysis of H. intermedus in the 8 populations distributed in China.  Admixture graph constructed with TreeMix using SNP data, 156 individuals of 8 H.intermedius populations, and three other relative individuals. Branch lengths are proportional to the evolutionary change (the drift parameter) and terminal nodes were labeled with population codes (see Table 1). The scale bar represents 10 times the average standard error (s.e.) of the values in the covariance matrix, and the migration weight represents the fraction of ancestry derived from the migration edge. Migration edges are observed between the DM and SS, although gene ow in the other direction is also observed.  Competing scenarios were designed for inferring the demographic histories for ABC analysis. a.
Considering the results of STRUCTURE, and phylogenetic analyses, we classi ed the 8 population into the following three regional groups: