Population genetics of the African snakehead fish Parachanna obscura along West Africa's water networks: Implications for sustainable management and conservation

Abstract An essential factor for aquatic conservation is genetic diversity or population divergence, which in natural populations reflects the interplay between geographical isolation with restricted gene flow and local evolution of populations. The long geological history of Africa may induce stronger among‐population divergence and lower within‐population divergence in fish populations of African watersheds. As an example, we studied population structure of the African snakehead fish Parachanna obscura. Our study aimed: (1) to develop a set of highly polymorphic microsatellite markers suitable for the analysis of genetic diversity among P. obscura and (2) to study the genetic diversity and structure of P. obscura populations from the West Africa region and mostly from Côte d'Ivoire, with respect to the effects of climate region and geographical distance on the genetic differentiation. A total of 259 specimens from 15 locations of P. obscura were collected over the West Africa region reflecting a high variability of pairwise geographical distances and variability of hydrological connectivity of the area. We developed a set of 21 polymorphic microsatellite markers for studying population genetics of the fish. The results showed relatively low intragenetic diversity for all the 15 locations, certainly attributable to confinement of fish in segregated catchments, resulting in limited gene flow. We also found evidence for local adaptation processes, suggested by five out of 21 microsatellite loci being under putative selection, according to BAYESCAN analysis. In turn, there was strong genetic differentiation (F ST > 0.5) among fish from most locations, reflecting the allopatric development in watersheds without hydraulic connectivity. Neighbor‐joining dendrogram, Principal Coordinate Analysis, and analysis of ancestral groups by STRUCTURE suggested that the 15 locations can be divided into three clusters, mainly matching the dominant climate zones and the segregation of the watersheds in the region. The distinct genetic structure of the fish from the 15 locations obtained in this study suggests that conservation and sustainable management actions of this fish resource should avoid genetic mixing of potentially locally adapted populations.


| INTRODUC TI ON
Tropical regions host the world's richest freshwater fish faunas.
According to estimates, there are over 3000 species of freshwater fish in Africa, which is comparable to the number of species in Asia (over 3600) and South America (over 4200; Lévêque et al., 2008).
Complex climatic and geological events have caused a long history of geographical isolation followed by diversification for some and extinction for other populations, ultimately resulting in the diversified fauna of the African freshwater ecosystems (Amoutchi et al., 2021;Darwall et al., 2011). Since the division of Gondwanaland in the early Cretaceous, many African rivers went through complex geological histories involving changes in the structure of their catchments and river beds (Goudie, 2005). The evolution of Africa's fauna and landscapes has been influenced by the cycles between the Pleistocene and Pliocene dry and rainy eras (deMenocal, 2014;Maslin et al., 2014;Van Steenberge et al., 2020). Along with fluctuations of water levels in the African Great Lakes, these climate cycles have caused alternating expansions and contractions of savannah and forest-like habitats (Malinsky & Salzburger, 2016). According to Tedesco et al. (2005), these climatic shifts led to migration, extinction, and allopatric divergence, which resulted in the current diverse fish faunas. In addition, the long geological history of Africa also affects the interplay between the contemporary distribution of populations and their genetic diversity, with longer periods of spatial segregation and allopatric development than those typically found for the much younger post-glacial landscapes of the Palearctic or Nearctic.
An essential factor for aquatic conservation is genetic diversity or population divergence, which in natural populations reflects population history and the evolutionary potential of a species (Jaisuk & Senanan, 2018). Within a fish species, population subdivision results from the interaction of distinct genetic changes within isolated populations and restricted gene flow among them (Hedrick, 2011).
Conspecific populations typically diverge from one another in the absence of gene flow due to mutation, natural selection, and genetic drift (Freeland, 2005). In addition, the degree to which landscape shapes patterns of genetic variation among populations is determined by life-history features linked to migration and fish dispersal capabilities (Pilger et al., 2017). Geographical factors favoring population division include geographical distance between locations (Beneteau et al., 2009;Crookes & Shaw, 2016), the presence of barriers (Neville et al., 2006;Yamamoto et al., 2004), the complexity of a river network (Pilger et al., 2017), and habitat fragmentation (Sterling et al., 2012).
The freshwater ecosystem in Côte d'Ivoire in Africa is characterized by a large and complex system consisting of four major river basins: Sassandra, Cavally, Bandana, and Comeo ranging in length from 650 to 1160 km and rising in geographically wideranging areas beyond the Côte d'Ivoire. In addition to these systems, there are many coastal rivers such as the Tabou, San Pedro, Niouniourou, Boubo, Agneby, Bia, and Me Rivers as well as two tributaries of the Niger River and many lakes (Girard et al., 1971).
These habitats are endowed with numerous economically important fish species.
Among the economically important fish, Parachanna obscura (Gunther, 1861), is the most popular and widespread African fish species from the Channidae family. It is commonly known as African snakehead fish and has great economical and commercial values for local African communities. This species is benthopelagic and a strict freshwater habitant. It is generally distributed in the intertropical convergence zone where the water temperature ranges from 26°C to 28°C principally in West Africa. Nevertheless, it is also found in the upper course of the White Nile, the Lake Chad basin, and the Congo River basin (UA, 2013). Given the complexity and geological history of the region, the resulting patterns of genetic variation merit investigation in P. obscura. For example, Bezault et al. (2011) suggested that paleo-geographic history, climatic events of Africa, and geographic barriers have induced strong genetic differentiation among Oreochromis niloticus populations from different parts of Africa. Strong genetic differentiation was also detected among African freshwater river and lake populations of Lates niloticus, reflecting the complexity of freshwater systems originating from the geological history of the continent (Basiita et al., 2018). Furthermore, the long-time isolation of populations can favor adaptations to local environmental conditions with the occurrence of particular alleles.
The adaptive capacity of populations depends on microevolution, e.g., the selection of local genotypes better adapted to changing environmental conditions (Canale & Henry, 2010;Hoffmann & Sgrò, 2011). These adaptation processes will likewise influence the amount and distribution of genetic diversity among populations (Pauls et al., 2013). Thus, assessment of the genetic diversity of P. obscura is necessary for understanding the evolutionary patterns of this species, and its capacity to cope with future environmental conditions, as well as for planning conservation policy for sustainable management of this species as fisheries resource.
Previous studies on P. obscura have mostly focused on biology (Bolaji et al., 2011;Odo et al., 2012), reproduction (Vodounnou et al., 2017), aquaculture potential (Azrita & Hafrijal, 2015), and phylogenetic range of the fish (Adamson et al., 2012;Conte-grand et al., 2017;Li et al., 2003). However, there is no information about the genetic diversity among populations of this species. Since no previous studies have been conducted on the genetic diversity of K E Y W O R D S genetic differentiation, genetic diversity, hydrological networks, paleogeographic history, snakehead fish

T A X O N O M Y C L A S S I F I C A T I O N
Population genetics P. obscura, genetic markers, for example, microsatellites, were not available. Accordingly, the objectives of our study were (1) to develop a set of highly polymorphic microsatellite markers suitable for the analysis of genetic diversity among P. obscura fish and (2) to study the genetic diversity and structure of P. obscura fish from 15 locations from the West Africa region and mostly from Côte d'Ivoire. According to the geological history of the region with potentially long periods of geographical isolation, we expected low genetic diversity and high genetic differentiation for this species.
Furthermore, we expected that the genetic differentiation was affected by geographical isolation caused by the interplay between connectivity barriers, landscape structure, and geographical distances among locations.

| Study area and collection of samples
A total of 259 specimens of P. obscura from 15 locations were collected over the West Africa region from Côte d'Ivoire (14 locations) and Benin republic (one location; see Figure 1 and Table 1). The sampling sites were selected to represent a high variability of pairwise geographical distances. In Côte d'Ivoire, individuals were sampled from Bia river (KRINA, and KRINB, locations about 7 km distant from each other), Wayadji stream (SIK), Sassandra river (SBR), San-Pedro Lake (SANP), and Abengourou lake (ABIN). This region is characterized by a sub-equatorial climate (Guinean climate zone) including two rainy and dry seasons, with an estimated annual precipitation of more than 1500 mm (Bernard, 2014). Specimens were also collected from Kan Lake (KAN), Baho (BAHO) and Glo (GLO) streams, Buyo Lake (SASA and SASB, sampling distance of about 26 km), and Nzo River (NZOA and NZOB, sampling distance of about 13 km), in the center and centre-western Cote d'Ivoire, characterized by an equatorial transition climate (Sudano-guinean climate zone), with two rainy and dry seasons, and an annual precipitation of 1200 to 1500 mm per year. Finally, individuals were collected in Bagoue river (BAG), situated in the north Cote d'Ivoire, characterized by a tropical climate (Sudanean climate zone) with a very hot and dry season from November to March and a rainy season from April to October. Specimens sampled from Benin republic were collected in Nokoue lake (BIN), located in the southern part characterized be a sub-equatorial climate with two rainy and dry seasons.

| Molecular genetics analyses
Genomic DNA from fin clippings was isolated using the DNeasy Blood & Tissue Kit (Qiagen) following the manufacturer protocol.
The development of microsatellite markers suitable for the analysis of population genetics of P. obscura was conducted by the commercial company GenoScreen. Their procedure consisted of two steps: (1) GenoSat library preparation using 5 μg DNA from an equimolar pool of 10 DNA samples followed by high throughput DNA sequencing run on Nano 2x250 v2-MiSeq Illumina and bioinformatic analysis and primer design. (2) Biological validation of 142 primer pairs on eight P. obscura DNA samples from different populations including PCR amplification and analysis of the obtained profiles on QIAxcel (Qiagen). GenoScreen usually considers as validated those primer pairs with a specific PCR product at the expected size for at least 5 samples. This was the case for 120 (or 84.5%) of the tested 142 primer pairs. Based on the delivered list of validated primers with comments for each pair and migration profiles generated by QIAxcel ScreenGel 1.6.0, the number of potentially suitable primer pairs was further reduced by excluding those with weak, very weak, or no visible PCR product for one or more samples and focusing on those indicating polymorphism on the migration profiles. This screening resulted in 29 primer pairs selected for testing on a larger number of individuals using the PCR protocol described below. However, only 21 primer pairs (Appendix A) turned out to be suitable for routine genotyping based on the observed polymorphism.
Each forward primer of the 21 loci was synthesized with either DY751 or Cyanine 5 or BMN-6 fluorescent dyes attached to its 5′ end. A set of eight multiplex reactions were conducted (Appendix B).

| Statistical data analyses
The majority of the analyses and graphical output were created using R 4.1.2 (R Core Team, 2021). Allele polymorphism at each of the 21 microsatellite loci and the intra-population genetic diversity metrics, such as number of alleles per location and allelic richness A R , were quantified by PopGenReport 3.0 package (Adamack & Gruber, 2014). Potential existence of individuals with missing genotypes among the loci was evaluated using poppr 2.9.3 (Kamvar et al., 2014) package in R. Null allele frequencies per locus were calculated in the genepop 1.1.7 package (Rousset, 2008).
The inbreeding coefficient (probability that the two alleles at one locus of an inbred individual are identical alleles per descent (Gazal et al., 2014); F IS ) was estimated in genepop_in_R. The significance of this coefficient was evaluated based on 95% confidence intervals calculated using bootstrapping (N = 100) in hierfstat 0.5-11 TA B L E 1 Information on samples sizes, basins, country, and climate of locations of 15 samples Parachanna obscura (with total sample size of 259 individuals).  (Guo & Thompson, 1992) per locus per location were corrected by the false discovery rate for multiple tests (Benjamini & Hochberg, 1995), supplemented by U-tests on excess homozygotes. We conducted an Analysis of Molecular Variance (AMOVA, Excoffier et al., 1992) using the R-package pegas 1.1 (Paradis, 2010) to compare within-and between-location variance for the 15 locations.

River basin Water body Location
Through the genepop v1.1.7 package in R, genetic differentiation was estimated by F-statistics between locations (F ST ; Weir & Cockerham, 1984), with the significance of differentiation assessed by exact conditional contingency table tests for genotypic differentiation. Using the poppr v2.9.3 package in R, a neighborhood tree (Saitou & Nei, 1987) based on Prevosti's pairwise genetic distance (Prevosti, 1974)  were performed by the vegan v 2.6-2 package (Oksanen et al., 2022) in R.
To examine non-neutral evolutionary forces acting on the microsatellite loci, a scanning analysis was realized using the BAYESCAN v2.1 software (Foll & Gaggiotti, 2008) to detect candidate loci under selection. BAYESCAN was run with a sample size of 5000, a number of pilots runs of 20, length of pilot runs of 5000, a burn-in of 50,000, and the false discovery rate (FDR) threshold of 0.05.

| Genetic structure within populations
In total, 0.3% missing genotypes were obtained among all microsatellites for the 259 individuals from the 15 locations (Appendix C).  (Dempster et al., 1977), relatively high null allele frequencies were estimated for loci Para038 (16%) and Para059 (59%). However, since no difference was observed in the estimates of genetic differentiation among locations after removing these loci (pairwise F ST values; Appendix E), we included all 21 loci in the analyses.
The genetic diversity was low in most of the locations ( Table 2).

| Genetic structure between populations
The pairwise F ST was high and significantly differentiated all location pairs of P. obscura, except the few locations from the same watershed only few kilometers apart (  Table 4).
The neighbor-joining dendrogram obtained based on Prevosti's genetic distance separated the locations into three major clusters

| Detection of loci potentially under selection
The BAYESCAN analysis indicated five among the 21 loci with a q-value of less than 5%, suggesting that they are under selection

| DISCUSS ION
We characterized the genetic structure and diversity of P. obscura  Diptera, Lehmann et al., 1997;and Orthoptera, Chapuis et al., 2005) and mollusks (Astanei et al., 2005;Li et al., 2003). It is generally not advised to include null allele loci in population genetics because they may affect estimates of population differentiation (Chapuis & Estoup, 2007). However, we did not exclude any locus from our study because it has been demonstrated that the influence of null alleles in studies of population genetics may be minimal compared to other parameters such the number of loci (Carlsson, 2008).
Moreover, for our data, the estimation of the populations differen- with them (Via, 2009;Via & West, 2008). Genetic hitchhiking is a process that allele frequencies change without being under natural selection, because these genes are located on the same DNA chain near to another gene that is undergoing a selective sweep.
The hitchhiking process may affect larger regions of the genome in particular in small populations (Charlesworth et al., 1997). As a result, genetic diversity of populations may decrease fast in response to environmental changes, lowering the population's ability  (Faulks et al., 2010;Gomez-Uchida et al., 2009;Loxterman & Keeley, 2012;Pfrender et al., 2004;Taylor et al., 2003). The results are in line with the hypothesis developed by Meffe and Vrijenhoek (1988) for explaining population genetic patterns for aquatic organisms inhabiting stream networks. Indeed, these authors predict that genetic isolation will occur among stream networks that do not have hydrological connections, resulting in an imbalance between drift and gene flow (where gene flow is effectively zero). A study on Channa argus, another Channidae species, revealed significant genetic differentiation among populations related to the structure of the river system (Yan et al., 2018). (2012) (Bezault et al., 2011;Drake et al., 2011;Lévêque, 1997 (Egger et al., 2007).

Loxterman and Keeley
Although there was a significant and positive trend between genetic and geographical distances, some location pairs strongly deviated from the expected linear pattern. This pattern suggests that genetic diversity may additionally be shaped by insurmountable barriers resulting from landscape characteristic between the different sampling locations that prevent any exchange of genetic material among populations from non-connected water bodies. The population most strongly differentiated from all others was from Bagoue River (BAG), situated in the north Cote d'Ivoire, characterized by a tropical climate. As seen on the map, the watersheds of Bagoue River are completely disconnected from the watersheds of the other populations analyzed, making geographical separation very likely.
This separation is likely induced by the geography of the area, which is characterized by a mountain ridge south of Bagoue River that prevents the river flowing to the south, as most of the other rivers in the southern part. Similar observations were taken on salmonid fishes for which it has been suggested that landscape characteristics such as the complexity of the drainage network and differences in channel gradients between habitats are likely to limit dispersal between their populations (Angers et al., 1999;Castric et al., 2001;Guy et al., 2008;Hebert et al., 2000). regions. The population structure characterized by the ancestral groups inferred by admixture coefficients in STRUCTURE has given the same classification (K = 3) but showed that ABIN shared a higher admixture with those populations from the sub-equatorial climate areas, which form cluster I. Therefore, it is more likely that ABIN location also belongs to cluster I. The partly match between clusters and climate zones suggests that regional climate conditions may have contributed to the patterns of genetic diversity of P. obscura in West Africa. However, more appropriate types of markers (e.g., SNP data obtained through RAD-seq) should be applied in future for confirming this hypothesis of regional climate as driver of the pattern of genetic diversity in P. obscura. In contrast, the population structure at higher resolution with classification into nine ancestral groups (K = 9) reflects the geographical configuration of the network of watersheds with populations from the same watersheds belonging to the same cluster. Hence, hydraulic connectivity or isolation have shaped gene flow among the habitats, facilitating differing evolution of the populations.

| CON CLUS IONS
Using the collection of polymorphic microsatellite markers developed in the present work, we characterized for the first time, the F I G U R E 6 Plot representing the BAYESCAN results searching candidate loci under selection. The vertical line represents a false discovery rate (FDR) threshold of 0.05. Points to the right of the vertical line represent loci under selection. Q-value: Minimum false discovery rate at which a locus may become significant; F ST , Coefficient to measure the difference in allele frequency between the common gene pool and each population, calculated as a posterior mean using model averaging.
genetic diversity, and structure of P. obscura populations from West Africa, representing an important baseline for further exploration of the population dynamics in this species. Understanding the genetic diversity of wild populations can help establishing aquaculture breeding programs as well as conservation initiatives to preserve fish stocks and their unique genetic identities (Yan et al., 2018). The low genetic diversity of P. obscura demonstrated that particular attention has to be paid for conservation and sustainable management of this fish resource. Regarding the high genetic differentiation between populations attributable to habitat heterogeneity and local adaptation, in situ conservation will be required in order to maintain genetic integrity. These population may serve as reservoirs or stocks for future selection programmes in aquaculture as well as improving population fitness and ability to respond to future environmental disturbance. writing -review and editing (equal).

ACK N OWLED G M ENTS
We are grateful to Leibniz Institute of Freshwater Ecology and Inland Fisheries (IGB) and WASCAL (supported by German Federal Ministry for Education and Research) for sponsoring this study, which is part of the PhD project of Amien Isaac Amoutchi. Open Access funding enabled and organized by Projekt DEAL.

CO N FLI C T O F I NTE R E S T
The authors have no conflicts of interest to declare.

DATA AVA I L A B I L I T Y S TAT E M E N T
The sequences of microsatellite markers developed and used in this study have been submitted to Genbank (https://www.ncbi.nlm.nih. gov/genba nk/).