During the period of 2009 and 2019, 377 individuals of C. guichenoti were collected from the upper reaches of the Yangtze river basin. Tissue specimens of C. guichenoti were collected in strict accordance with relevant national laws and animal ethics requirements. During the collection process, a small amount of fish fin tissues was cut off and disinfected, and the fish were released into the river. Sampling sites cover most of the distribution range of the species. To ensure comparability, sample collection was considered using the same or similar as possible sampling points in the same water system or section. A total of five populations were collected in the Yanbian section of the Yalong River (YB), the Xufu section of the Jinsha River (SF), the Jiangan section of the Yangtze River (JA), the Luxian section of the Yangtze River (LX) and the Beibei section of the Jialing River (BB) in 2009. A total of seven populations were collected in the Xiqu section of Jinsha River (XQ), Hongmenchang section of Jinsha River (HMC), Longchi section of Minjiang River (LC), Caiba section of Yangtze River (CB), Xuzhou section of Yangtze River (XZ), Jiangyang section of Yangtze River (JY) and Liemian section of Jialing River (LM) in 2019 (Fig. 1).

To reduce the influence of different ages on the experimental error, 308 samples aged around 3 years were selected from 377 individuals, according to the relationship between age and body length (Ding 1994; Table 1). Total genomic DNA was extracted from the fin clips of C. guichenoti using PureLink Genomic DNA Mini Kit (Thermo Fisher Sentific, China).

Primers DL1 (ACCCCTGLCTCCCAAALC, Ta: 62 °C) and DH2 (ATCTTALCATCTTCAGTG, Ta: 62 °C) (Liu et al. 2014a) were set for the amplification of mtDNA control regions in all populations. The PCR conditions, electrophoresis and sequencing were as described in Liu et al. (2014b).

Haplotype number (H), Haplotype diversity (h), and nucleotide diversity (π) were used to estimate genetic diversity in mtDNA control regions (Nei 1987). Genetic diversity indices were calculated using DnaSP 3.2 (Rozas et al. 2003). Based on the mtDNA control region, a phylogenetic analysis was performed by Bayesian inference (BI) using Coreius heterodon (Genbank no. FJ376082.1) and Abbottina rivularis (GenBank no. AP011257.1) as outgroups. The median connection network (MJN) method (Bandelt et al. 1999) was used by Network v. 3.1 to describe the relationships among all haplotypes. Pair genetic differentiation (FST) among populations were estimated using Genepop v. 2.4 (Zaigham et al. 2012).

Population demography history was tested with mtDNA control region. First, Tajima’s D (Tajima 1989) and Fu’s FS (Fu 1997) were calculated for each population, under the recent population expansion model, using Arlequin v. 2.0. In addition, population history was studied by using DnaSP software to compare the mismatch distribution of each geographic sample with that of stable and expanded populations. Bayesian horizon map (BSP) was implemented in BEAST v. 1.5.4 (Drummond et al. 2005) and visualized in Tracer v. 1.5 (Drummond and Rambaut 2007). The final analysis was performed on 50 million generations, sampled every 2,000 generations, and aged for 5 million generations. The time for possible bottleneck or population expansion (T, calculated in substitution) is calculated from the relation S = 2UT (Rogers and Harpending 1992), where S is the pattern of mismatch distribution and U is the estimated mutation rate of the sequence.

Using the isolation with migration (IM) model in IMA2 program (Hey 2010), the migration rate is calculated based on the mtDNA control region. Running IMa2 required two steps (M mode and L mode). Firstly, in M mode, the model parameter function is estimated based on 6 ×10⁸ MCMC steps, according to the aging cycle of 200,000 steps. Secondly, the model parameter function estimated by M mode is used to estimate the offset rate in L mode.

To test the correlation between genetic differentiation and geographical distance, we performed Mantel test with mtDNA data. The significance of the correlation (r) between log-transformed genetic distance and log-transformed geographic distance was determined by using 2, 000 permutations of Distance Isolation (IBD) Web Service v. 3.1.6 (Jensen et al. 2005).

Genetic diversity analysis of 12 populations was performed based on mtDNA control region. Sequencing of 899 bp revealed 301 variable loci. The average base composition is A = 28.5%, T = 31.4%, C = 22.4% and G = 17.7%. A total of 65 haplotypes were identified from the control region sequences extracted from 308 individuals. Each population has some private haplotypes. However, with the exception of JY, Hap1 is shared in every population. LC, JA, XZ, HMC, and CB shared Hap3. YB, JA, SF, XZ, CB, and XQ share Hap4. All populations shared Hap5, except LC and YB. Each population shared Hap7, except for LX and JY. BB, JY, YB, JA, LC, and XZ shared Hap8 (Fig. 2). The haplotype diversity (h) varied from 0.869 to 0.901, and the nucleotide diversity (π) varied from 0.0031 to 0.0048. In both cases, the YB population had the highest number of individuals (Table 2). XZ and LM had the largest private number of haplotypes, 11 and 10, respectively. The numbers of private haplotypes were the least in BB (1) and JA (1). All differences of genetic variation (h and π) of 12 populations was not significant by paired t-test (P > 0.05).

Figure 2. 

Phylogenetic trees of the mtDNA control region haplotypes in C. guichenoti reconstructed with Bayesian inference. Numbers at nodes represent Bayesian posterior probabilities and neighbor-joining tree. At the right side of the figure, the numbers represent the total of individuals from different sampling locations in each haplotype.

According to the sampling sites, the 12 populations were divided into three groups: upper group (YB, XQ and HMC), middle group (LX, JA, SF, JY, LC, CB, and XZ), and lower group (BB and LM). AMOVA analysis of five populations (JA, YB, BB, SF, and LX) collected in 2009 showed no significant molecular differences in mtDNA among populations, between populations and within populations (P > 0.05; Suppl. material 1: Table S1). AMOVA analysis of seven populations (JY, XZ, LM, CB, HMC, XQ, and LC) from 2019 showed that the intrapopulation variance (97.05%) was significant (P < 0.05, Suppl. material 1: Table S2).

The FST values of five populations (LX, BB, YB, SF, and JA) collected in 2009 were not significant (P > 0.05). The genetic differentiation between XQ and JA, HMC, and BB was not significant. Two neighborhood link trees were constructed based on the FST values in five populations (2009) and seven populations (2019). BB and LX populations clustered together, and SF, JA, and YB populations clustered in adjacency trees (Suppl. material 1: Fig. S1a). LC and JY, and LM and CB, are clustered together in adjacency trees, respectively (Suppl. material 1: Fig. S1b). HMC, XZ, and XQ populations are clustered in adjacent connection trees (Suppl. material 1: Fig. S1b).

The results of mtDNA haplotype sequence analysis showed that there were four clades in the tree (Fig. 2), but no clear geographical distribution pattern was shown. Clades 1, 2, and 3, were composed of 12 population haplotypes. Clade 4 consists solely of 12 populations except for LX (Fig. 2). At the same time, the middle join network shows all single haplotypes from 12 populations that do not form a unique geographic structure of the haplotype network (Fig. 3). The four centers (Hap1, Hap3, Hap4, and Hap11), rich haplotypes, and other low-frequency-derived haplotypes were included in the haplotype network (Fig. 3). The frequency of Hap1 (n = 20), Hap3 (n = 32), Hap4 (n = 20), and Hap11 (n = 11) was different at different sites. Hap3 and Hap4 were not observed in the down group (BB and LM). The analysis of the aggregation method based on the IM model showed that the level of gene flow between populations was very limited (Suppl. material 1: Fig. S2). There was no significant gene flow in any of the five populations (Suppl. material 1: Fig. S2a; 2009) and seven populations (Suppl. material 1: Fig. S2B; 2019).

Figure 3. 

Median-joining network of the mtDNA control region haplotypes of C. guichenoti. The size of each circle indicates the relative frequency of the corresponding haplotype in the whole data set.

Mantel test showed that geographical distance was significantly associated with genetic differentiation in five populations (2009; Fig. 4a) and seven populations (2014–2019; Fig. 4b). However, in seven populations (2019; Fig. 4b), there was a weaker (than 2009) correlation between observed genetic differentiation and geographical distance (r = 0.1579, P = 0.2900).

The multichannel mismatch distribution, insignificant Tajima’s D and Fu’s FS values, and BSP analysis showed that SF, LX, BB, JA, YB, and XQ populations were relatively stable. However, LC and LM suffer genetic bottlenecks and XZ suffers dilatations. The bottleneck trend in the JY population and a weak population growth trend in the HMC and CB population.

Figure 4. 

Isolation by distance (IBD) relationship among C. guichenoti wild populations in five populations collected in 2009 a and seven populations collected in 2019 b.

In this study, we assessed genetic variation in twelve populations of C. guichenoti to understand the dynamics of the species’ genetic variability and its population structure and explore whether the rapid reduction of genetic variability will be observed in this species. Adaptive potential is related to genetic diversity, and the loss of genetic diversity can increase the possibility of population extinction. Previous work has highlighted the possible importance of genetic variability in determining the health of both individuals and, perhaps by implication, populations (Pritchard et al. 2000). Adults of C. guichenoti usually swim upstream when rivers are in flood, and after spawning they drift downstream (at lower elevations), suggesting that the upstream region may be the breeding and distribution center of the Yangtze River and its tributaries (Cheng et al. 2015). There was no significant difference in genetic diversity among the 12 populations by paired t-test (P > 0.05), indicating that there was no significant decrease in genetic diversity from 2009 to 2019, through the mean value of genetic diversity in 2009 was higher than that in 2019. Based on this result, it appears that the breeding of C. guichenoti has not been completely disrupted by dams. Some individuals may not need to go upstream to spawn and may be able to breed in scattered habitats, even if they are severely affected by dams.

The results of AMOVA analysis showed that the genetic differentiation between populations and between populations was limited (P > 0.05), and most of the total genetic variation occurred among individuals within groups, which was insignificant neither (P > 0.05). The FST values showed that the population structure of each sampling site was not significant. Moreover, these five groups do not form independent groups in terms of population structure, and each individual actually has an equal probability of belonging to any of these groups in any analysis. There was no obvious systematic geographic structure among the haplotypes of the samples from different geographical locations. The sequence differences within and between groups were small, indicating high genetic similarity among different groups and relatively shallow population history. The genetic differentiation of the five populations was significantly correlated with geographical distance. After spawning, this species usually flows downstream with floods, and when rivers are in flood, juvenile and adult fish subsequently swim upstream (Chen et al. 2002). These life history characteristics can facilitate gene communication between populations. Insignificant genetic differentiation among groups was also observed in Leptobotia elongata (Bleeker, 1870) in the Yangtze River, and this was also attributed to spawning of drifting eggs and larvae (Liu et al. 2020). This is consistent with the lack of genetic population structure due to the spawning and life history traits of C. guichenoti and also might be related to the uncertainty of the distance that adults swim upstream.

However, different population structures were found in seven populations collected from 2019. There were significant differences in FST values between some populations collected from 2019. The seven populations did not form an independent cluster in population structure, but the composition of each population was different. Our AMOVA analysis showed that although the genetic differentiation between populations was very limited, there was significant genetic variation among individuals within populations (P < 0.05). Significant differences among all populations have not yet developed, but differences among individuals within populations are emerging. To explain this, the dam restricted the genetic links between the upstream and downstream populations of C. guichenoti, splitting them into small, isolated populations. A strong, though occasionally permeable, barrier to transmission may exist to limit the flow of genes between sites. In this case, the species’ habitat becomes fragmented, which not only affects the spawning and development environment of C. guichenoti, but also impedes upstream and downstream migration and gene exchange. Thus, weak correlation between the observed genetic differentiations and geographical distances was found in seven populations (2019, P > 0.05) while it is significant negative correlation in five populations (2009, P < 0.05). As the hydrological environment changes, the individuals within the population become different. Then there may be a strong demographic structure in the future. Furthermore, compared with large populations, small populations are more prone to inbreeding and loss of genetic diversity due to genetic drift and bottlenecks (Hoffman et al. 2009). In addition, the formation of reservoirs in the upper reaches of the Yangtze River has greatly changed the hydrological environment, which may damage the habitat and spawning grounds of C. guichenoti. Due to the lack of running water, larvae may not develop successfully. As a result, the abundance and genetic diversity of wild populations will inevitably decline.

Though the decrease of genetic diversity was not significant from 2009 to 2019, the population structure of C. guichenoti showed a trend of change. If conservation and restoration measures will be not carried out, a loss of genetic diversity may occur in the future. Loss of genetic diversity will impair the ability of the population to respond to environmental changes, so conservation of genetic variation should be a priority in the recovery efforts of C. guichenoti, especially in light of recent habitat fragmentation due to dams.

Migration (increased gene exchange) can lead to gene rescue or recovery (Richards 2000). Increasing the population of a threatened population by culturing individuals may reduce the abundance of a wild population, but when properly applied, it can increase the genetic exchange of the species, which is a viable strategy for the conservation of the species. Various fish passage facilities have been developed to help species to cross these obstacles (Gow et al. 2011). Fishways are hydraulic structures that help fish to climb over obstacles such as dams and locks to spawn and migrate. Fishway plays an important role in fish conservation, but the corresponding design work is complex, which needs to consider the practical design criteria of hydraulic engineering, the path selection of fish swimming in fishway and the corresponding mortality evaluation of fish (Liu et al. 2012). Hydropower stations and researchers in China should devote great efforts in solving these problems, aiming for more precise quantitative predictions of efficient fish-passage and an optimum design of the hydraulic structure.

Furthermore, in order to maintain heterozygosity and genetic diversity, mature adults should be collected for artificial propagation from different locations across the distribution, especially considering the high genetic diversity among groups (An et al. 2016). Aquaculture conditions of hatcheries should be improved to increase the number of qualified parental fish and attention paid to the individual genetic relationship during artificial propagating. Restocking efforts also should strive to maintain genetic connectivity and exchange among local parental fish in the hatcheries to avoid inbreeding and increase genetic variation. Similarly, cultured juveniles and young fishes should be released into various habitats across the distribution as well (Clay 1995). Then, the population dynamics and genetic diversity of these wild populations should be monitored frequently, and protection measures be adjusted subsequently when needed.