Populations genetically rifting within a complex geological system: The case of strong structure and low genetic diversity in the migratory freshwater catfish, Bagrus docmak, in East Africa

Abstract The complex geological history of East Africa has been a driving factor in the rapid evolution of teleost biodiversity. While there is some understanding of how macroevolutionary drivers have shaped teleost speciation in East Africa, there is a paucity of research into how the same biogeographical factors have affected microevolutionary processes within lakes and rivers. To address this deficiency, population genetic diversity, demography, and structure were investigated in a widely distributed and migratory (potamodromous) African teleost species, Ssemutundu (Bagrus docmak). Samples were acquired from five geographical locations in East Africa within two major drainage basins; the Albertine Rift and Lake Victoria Basin. Individuals (N = 175) were genotyped at 12 microsatellite loci and 93 individuals sequenced at the mitochondrial DNA control region. Results suggested populations from Lakes Edward and Victoria had undergone a severe historic bottleneck resulting in very low nucleotide diversity (π = 0.004 and 0.006, respectively) and negatively significant Fu values (−3.769 and −5.049; p < .05). Heterozygosity deficiencies and restricted effective population size (N eLD) suggested contemporary exposure of these populations to stress, consistent with reports of the species decline in the East African Region. High genetic structuring between drainages was detected at both historical (ɸST = 0.62 for mtDNA; p < .001) and contemporary (microsatellite F ST = 0.460; p < .001) levels. Patterns of low genetic diversity and strong population structure revealed are consistent with speciation patterns that have been linked to the complex biogeography of East Africa, suggesting that these biogeographical features have operated as both macro‐ and micro‐evolutionary forces in the formation of the East African teleost fauna.


| INTRODUCTION
East African freshwater systems possess a diverse teleost fauna shaped by a complex geological history, including large-scale tectonic movements, volcanic activity, and significant uplifting (Danley et al., 2012;Sturmbauer, Baric, Salzburger, Rüber, & Verheyen, 2001;Verheyen, Salzburger, Snoeks, & Meyer, 2003). The East African Rift (EAR) Valley, which was formed by tectonic uplift, is the major geological structure that has forged the general hydrographical network in Africa (Giddelo, Arndt, & Volckaert, 2002;Pinton, Agnèse, Paugy, & Otero, 2013) and created various freshwater habitats. For instance, major continental river systems were particularly impacted by the regional uplift, including the Nile, Congo, and Zambezi Rivers (Baker & Wohlenberg, 1971;Roberts et al., 2012). Evolutionary and geological processes such as fragmentation, hydrological connectivity, river reversal, and desiccation, among others (Danley et al., 2012;Johnson et al., 1996;Russell & Johnson, 2001), were responsible for the creation and maintenance of these habitats in which substantial numbers of aquatic taxa were isolated.
Freshwater lakes and rivers within the EAR (i.e., Lake Edward, Lake Albert, Lake George, Lake Tanganyika, Lake Malawi), and the largest tropical freshwater body, Lake Victoria (which lies outside the EAR), are habitats to one of the world's most biologically diverse aquatic faunas. For instance, prior to the introduction of the Nile perch, Lates niloticus, into Lake Victoria, the lake had between 350 and 600 endemic cichlids (Helfman, 2007;Turner, Seehausen, Knight, Allender, & Robinson, 2001). Geological evidence suggests a former connection of Lake Edward to Lake Victoria by late Pleistocene rivers, which were subsequently truncated by uplifting causing river reversal and a break in connectivity of lake systems (Lévêque, 1997). Presently, connectivity of these lakes is restricted to Lake Edward, which is connected to both Lakes George and Albert (via the Kazinga Channel and Semliki River, respectively). The Lake Edward-George system provides a biogeographic confluence between the Victorian and Albertine freshwater fauna (Thieme et al., 2005). Despite this hydrological connectivity between lakes in the EAR through rivers and channels, biogeographic barriers are evident including the Semliki rapids and falls, which descends 300 m from Lake Edward to Lake Albert (Lowe-Mcconnell, 1993, although Greenwood (1966) considers these Semliki rapids as inefficient barriers. To the right of Lake Albert are the Murchison Falls along the Victoria Nile River, another biogeographical barrier separating the Albertine rift system (includes Lakes Edward, George and Albert) from Lake Victoria. This barrier has been documented as an effective obstacle in preventing, for instance, Nile perch stocks in Lake Albert from migrating into Lakes Kyoga and Victoria (Basiita et al., 2011;Hopson, 1972).
Africa's freshwater systems are degrading at a very high rate with over 80 species listed as critically endangered, 116 species endangered and up to 103 threatened (Thieme et al., 2005). As elsewhere in Africa, the unique East African teleost faunas in both riverine and lacustrine freshwater habitats are currently under threat due to natural and anthropogenic pressures, with many species experiencing rapid population declines. Relative to other environments, biodiversity declines are at their highest in freshwater lacustrine bodies owing to the level of compartmentalization that naturally exists among these large water systems (Ricciardi & Rasmussen, 1999). The natural divides of freshwater bodies over evolutionary timescales seemingly deem them important as far as defining management units for biodiversity. Any interventions such as aquaculture developments, restocking existing water bodies, fishing zoning, and breeding grounds that are geared toward mitigating the declines resulting from natural and anthropogenic pressures need to take into account evolutionary significant units in the region.
Although numerous studies have looked at fish diversity, composition, and endemism, to understand biogeographic processes and speciation in Africa (Craig, 1992;Elmer et al., 2009;Pinton et al., 2013;Sato et al., 2003), very few fish studies have looked at genetic signatures left by biogeographical processes below the level of species (i.e., among populations). Unraveling antecedent genetic signatures among populations may not only help understand the processes that have led to their evolution and adaption in recent timescales, but more importantly, will assist with the identification of genetically divergent populations and/or evolutionally significant units that can be integrated into the formation of contemporary management plans.
Additionally, the species' widespread distribution throughout Africa, along with its potamodromous migratory life history, identifies it as an ideal candidate to examine how the complex geological history of East Africa shapes evolution of the fish fauna biodiversity at below the species level.

| Study area
Samples of B. docmak were collected from five water systems across the species' distribution in central East Africa (from Lakes Albert, Edward and Victoria; and Rivers Victoria Nile and the Kazinga Channel) ( Figure 1). Lakes Albert and Edward are located in the western arm of the East African Rift Valley commonly referred to as the Albertine rift, while Lake Victoria is outside.

| Sample collection and laboratory procedures
Fin clips of B. docmak were obtained from commercial fishers at each of the five sampling locations (Figure 1). Samples were collected under animal ethic approval number A1824 issued at James Cook University (JCU). All fin clips were preserved in 20% dimethyl sulfoxide (DMSO) saturated with sodium chloride salt (Amos, 1991;Dawson, Raskoff, & Jacobs, 1998) and transported to the Molecular Ecology and Evolution Laboratory (MEEL) in Townsville, Australia, where they were stored at −20°C until extraction. DNA extractions and polymerase chain reaction (PCR) assays were also carried out at MEEL.
Total genomic DNA was extracted using a modified CTAB protocol (Wilson, 1990) and later using the Bioline Isolate II Genomic DNA kit. The CTAB protocol involved a digestion step using a CTAB buffer with 200 μg of Proteinase K incubated at 55°C for 2 hr, followed by a 24:1 chloroform: isoamyl alcohol purification (700 μl of chloroformisoamyl and centrifugation at 13,200 rpm for 20 min), and an ethanol precipitation (2.5× 100% EtOH: 1:10× 5 mol/L NaAce and centrifugation at 16,000 g for 30 min, 1× 70% EtOH at 16,000 g for 20 min).
Genomic DNA was re-suspended in 25 μl of 1× TE (10 mmol/L Tris-HCl, 1 mmol/L EDTA, pH 8.0). DNA quality was estimated based on 0.8% agarose gel electrophoresis, and quantity was assessed using a ND-1000 Spectrophotometer (Nano-Drop ® Technologies). For some inhibited samples that failed to amplify during PCR, DNA was reextracted using a column based Bioline Isolate II Genomic DNA kit following manufacturer's protocol. Briefly the protocol involved a prelysis stage using 180 μl Lysis buffer with 25 μl of protease K incubated at 56°C. 200 μl of G-3 lysis buffer was added and further incubated for 10 min at 70°C. 210 μl of 100% EtOH were then added to alter the buffer conditions prior to two (GW1 and GW2) buffer washes. The eluted DNA was stored at −80°C prior to downstream PCR.
DNA from 175 individuals was genotyped at 12 polymorphic microsatellite loci, Bd04, Bd18, Bd01, Bd02, Bd12, Bd09, Bd06, Bd20, Bd05, Bd03, Bd14, and Bd10 (Appendix 3) (Basiita et al. in Goossens, 2015). Forward primers were fluorescently labeled using the 5-dye system (6-FAM, VIC, NED, PET, and LIZ GS-500 size standard), and all reverse primers were pigtailed (Brownstein, Carpten, & Smith, 1996) F I G U R E 1 Map showing lakes and rivers (sampling locations within East Africa) and number of individuals of Bagrus docmak collected from each location. The inset at the top right corner denotes the distribution of B. docmak across Africa to ensure consistent amplification and minimize stuttering. A total of 20 μl PCR reactions were run on Biorad C1000 thermocycler under the following conditions: an initial denaturation at 95°C for 5 min, 6 cycles of 95°C for 30 s (denaturation)/59°C for 90 s (annealing)/72°C for 30 s (extension), 10 cycles at reduced annealing temperatures of 57, 55, and 53°C, prior to a final extension at 60°C for 30 min (Basiita et al. in Goossens, 2015). Visualization of PCR product was performed on an ABI-3730 instrument (Applied Biosystems) using a 5-standard dye system (6-FAM, VIC, NED, PET, and LIZ GS-500 size standard) at the Georgia Genomics Facility, USA. Alleles were scored using  (Chenoweth & Hughes, 1997) were used to amplify the D-loop mitochondrion region to obtain B. docmak sequences that were then used to design more robust species-specific primer pairs. Primers specific to Bagrus spp. were designed using a free online software, primer3 (Rozen & Skaletsky, 2000). A forward

| Mitochondrial DNA sequence editing and alignment
Individual D-loop sequences were aligned and consensus sequences generated (Bast, 2013) in GENEIOUS 8.02 (Biomatters Ltd). Following generation and alignment of all consensus sequences, a 400 bp size region was trimmed and exported to MEGA 5.2 (Tamura et al., 2011). The best substitution model for analyzing sequence divergence and population structure among the different populations was selected as T92+G (Nei & Kumar, 2000;Tamura et al., 2011) using MEGA 5.2 according to Schwarz (1978).
The F is was used as a measure of inbreeding and/or population subdivision. Additionally, the average level of relatedness was calculated within each population using ML-Relate (Kalinowski, Wagner, & Taper, 2006). Input files were converted for use between programs using the free software, PGDspider 2.0.5 (Lischer & Excoffier, 2012).
All microsatellite loci were polymorphic in all populations sampled, except for locus BD18 that was monomorphic in the Victoria Nile River and BD20 in the Lake Victoria populations. Micro-checker 2.2.3 was used to detect genotyping errors, the presence of null alleles and allelic dropouts (Morin et al., 2009;Van Oosterhout et al., 2004).
Diversity indices calculated for the mitochondrial data included; haplotype number (n), haplotype diversity (Hd), and nucleotide diversity (π). These parameters were used as a measure of genetic divergence at the mitochondrial D-loop region within and among the sampled populations. All computations were implemented in DnaSP (Librado & Rozas, 2009) and Arlequin 3.5 (Excoffier et al., 2010).
Historical demography was investigated at the mitochondrial D-loop by estimating Fu's F statistic for each of the five locations (Fu, 1997). Neutrality tests to estimate Fu's F statistic were carried out in Arlequin (Excoffier et al., 2010). Furthermore, mismatch analysis distribution was performed for the demographic analysis in which pairwise difference distributions and the frequency of segregating sites were analyzed in DnaSP (Librado & Rozas, 2009;Rozas, Sánchez-Delbarrio, Messeguer, & Rozas, 2003).

| Population structure
Genetic population structure was investigated at the D-loop region of the mtDNA and at 12 polymorphic microsatellite loci. Using both datasets, analysis of molecular variance (AMOVA) and pairwise comparisons between locations were completed in Arlequin 3.5 (Excoffier et al., 2010). For microsatellite data, the AMOVA was based on allelic frequencies, and for mtDNA sequence data the AMOVA was based on a genetic distance matrix of pairwise differences between pairs of populations. Significance was estimated at 10,000 permutations.
Furthermore, using mitochondrial sequence data organized in DnaSP (Librado & Rozas, 2009) and exported to Network 4.611 (Flexus Technology, as reported by Bandelt et al., 1995), the population genetic structure and geographical distribution of haplotypes were visualized. Calculation and drawing of a minimum spanning network based on haplotype distribution in sampled individuals from all the five populations were completed in Network4.611 Flexus Technology.
Bayesian clustering analysis conducted in the program, STRUCTURE 2.2 (Earl & von Holdt, 2012), was used to assign individuals from the five locations into distinct genetic clusters. The analysis was based on microsatellite data at 12 polymorphic loci, and runs were conducted on putative populations (K) set from 1 to 10 iterations with 10,000 burn-ins followed by 100,000 Markov-Chain Monte Carlo (MCMC) steps for each run. Potential clusters were determined based on an MCMC approach, both with and without a priori definition of structure, and also assuming independent frequencies of alleles. Using the web-based Structure Harvester software (Earl & von Holdt, 2012), the best K value was then selected.
Additionally a multivariate method, the Discriminant Analysis of Principal components (DAPC), was performed in the R package adegenet v1.4.2 (Jombart, 2008), to explore a finer scale structure of the populations based on a two-step procedure. Firstly, the genetic data are transformed using a Principal Component Analysis (PCA), and then clusters are identified by Discriminant Analysis (DA) without assuming panmixia (Jombart, 2008;Jombart, Devillard, & Balloux, 2010).

| Genetic diversity
Mitochondrial DNA D-loop variation among 93 B. docmak individuals from the five locations revealed high haplotype diversity, Hd, with a narrow range across populations (Hd range of 0.698 for Kazinga Channel to 0.857 for Lake Albert). The nucleotide diversity was highest in individuals from Lake Albert and the Victoria Nile river, both with π = 0.010, and least in Lakes Edward and Victoria, π = 0.004 and 0.006, respectively (Table 1). Overall there were 25 distinct haplotypes, with 31 polymorphic sites and a total of 33 mutations (Appendices 1 and 2). Haplotype sequences were deposited into GenBank with accession numbers MF118537 to MF118561.
All populations were in Hardy-Weinberg equilibrium, and there was no detection of large allelic dropouts. Null alleles were not detected in the Lake Victoria population; however, they were suggested at locus BD02 for individuals sampled from the Victoria Nile River, BD01 in the Lake Albert and Lake Edward populations, and BD20 in the Kazinga Channel population. There was no locus that consistently exhibited null alleles, and as such all microsatellite loci were polymorphic across populations with number of alleles up to seven alleles per locus and overall mean allelic diversity, mean N a ± SE = 4.05 ± 0.192 (Table 1). The allelic diversity was highest in Lake Albert (mean N a = 4.67 ± 0.497) and lowest in the Lake Victoria population (mean N a = 3.25 ± 0.494). Similarly observed and expected heterozygosities were highest in the Lake Albert population (mean H o of 0.55 ± 0.034 and H e 0.66 ± 0.031) and lowest in the Lake Victoria population (mean H o = 0.199 ± 0.046 and H e = 0.213 ± 0.053) ( Table 1). Bottleneck analysis under the SMM and TPM models revealed significant heterozygosity deficiencies (with p values <.005) for B. docmak populations from Lake Victoria and Edward. These two populations additionally displayed an L-shaped allelic distribution characteristic of populations undergoing expansion following contractions (data not shown). The effective population sizes, N eLD , for Lake Victoria, Lake Edward and the Nile River were finitely restricted (Table 1). On the contrary, the results showed that the sampled populations were mating randomly with the F is indices generally low and showing no significant deviation from zero (p > .05), with the exception of Lake Victoria having the highest F is of 0.1101 (p = .054), but still statistically insignificant (  in all populations and highly significant for B. docmak individuals from Lakes Victoria, Edward, and the Kazinga channel (see Table 1).

| Demographic history
Overall, all populations sampled exhibited heterozygosity deficiencies except for the Lake Albert population that did not show  Channel-Lake Albert" and "Lake Edward-Lake Albert" showed moderately low ɸ ST of 0.046 (p < .05) and 0.0968 (p < .001), respectively (Table 3). In both cases, there was relatively weak genetic structuring among these populations situated within the Albertine Rift valley.

| Population structure
There was no significant structuring between Lake Edward and the  The bottleneck probabilities reported for IAM, SMM, and TPM models above were Wilcoxon probability 2-tail tests for heterozygosity deficiency and excess as implemented in the program Bottleneck (Cornuet & Luikart 1996;Piry et al. 1999).
T A B L E 2 Genetic population bottleneck tests including inbreeding coefficient F IS and relatedness smaller haplotypes that were branching out in the periphery. The most common haplotype was shared by 33 individuals restricted to the Albertine rift (Lake Edward, Lake Albert and the Kazinga channel) and representing up to 35% of the all sampled individuals. The second largest haplotype had 16 individuals restricted to the Nile River and Lake Victoria. From these two major haplotypes span other smaller haplotypes ( Figure 3). These results suggest the presence of two fairly distinct genetic groups spanning the five geographical locations sampled.

| Genetic diversity, population structure, and connectivity of Bagrus docmak
The complex geological history of East Africa has acted as a biogeographical evolutionary driver of the diverse freshwater fish species fauna seen in the region (Danley et al., 2012;Sturmbauer et al., 2001;Verheyen et al., 2003). Similarly, genetic analyzes of B. docmak populations from lacustrine and riverine habitats of the EAR show evidence for vicariance leading to genetic divergence below the species level. Highveld (Johnson, Kelts, & Odada, 2000;Johnson et al., 1996;Morris et al., 2016;Russell & Johnson, 2001). Although there is hardly any information on genetic variation among other bagrid catfishes in African freshwater systems for direct comparison, the divergence of populations in the B. docmak was found to be higher than what has been reported in another Asian freshwater bagrid catfish, Leiocassis longirostris (Yang, Xiao, Yu, & Xu, 2012).
This study also highlighted evidence for admixture among Lake Albert B. docmak (Figure 2a). This admixture may be indicative of limited connectivity between Lakes Victoria and Albert through the Victoria Nile and is consistent with historical exchanges of fauna to have occurred previously following geomorphological events, especially in the Nile system and its associated basins (Stewart, 2009).
Murchison Falls likely acts as a one way biogeographical barrier inhibiting fish stocks moving upstream from Lake Albert into Lake Victoria; however, it may not act as a barrier to fish moving downstream from Lake Victoria to Lake Albert. The biogeographical separation of Lake Albert from Lake Victoria by the Murchison Falls has been well documented, and it is this barrier that prevented the Nile perch from crossing upstream from Lake Albert into Lakes Kyoga and Victoria (Hopson, 1972 Despite similar geological history of formation by tectonic movements and their connectivity through the Semliki River, moderate, but significant structuring was detected between populations of Lake Edward and Lake Albert based on both mitochondrial and microsatellite data. The divergence of B. docmak populations between these two lakes is possibly a result of the biogeographical barrier of the Semliki F I G U R E 2 (a) STURCTURE bar plot of Bagrus docmak populations depicting two genetic clusters (K = 2) as per the genome ancestry assignment revealed. X axis represents individuals from the five locations (1-Lake Victoria, 2-Victoria Nile River, 3-Lake Albert, 4-Lake Edward, and 5-Kazinga channel) assigned to two major stocks/populations; Victoria basin populations (

F I G U R E 3 Haplotype network for
Bagrus docmak from five geographical locations as drawn in Network 4.6.1. Each circle denotes a single haplotype whose size is proportional to the frequency of the haplotype. The colors represent the geographical source of the haplotype. Each branch indicates a single mutational event except where indicated by lines that correspond to the total number of mutations rapids which may restrict levels of gene flow between these two lakes (Thieme et al., 2005). This observed genetic structuring pattern is consistent with earlier species composition studies, where unique fish fauna have been identified in Lake Albert, but not in Lake Edward (Devaere, Jansen, Adriaens, & Weekers, 2007).

Correspondently, the weak and insignificant F ST values obtained
between lake-river systems (i.e., Lake Edward-Kazinga Channel River and Lake Victoria-Victoria Nile River) is consistent with the direct connection of these lake-river systems which have no known biogeographic barriers. In addition, B. docmak being potamodromous (Chapman et al., 2012;Manyala, Bolo, Onyango, & Rambiri, 2005), implies that gene flow between lakes and river systems will be enhanced where there are no barriers, but inhibition of gene flow will be accelerated in the presence of geological features that interrupt migration and consequently restrict gene flow. Connectivity would foster the exchange of genes of populations between the lacustrine and riverine habitats.

| Historical and contemporary genetic signatures
Understanding biogeography using genetics is important to de-  (Table 1), is a pattern consistent with one which has been documented in another catfish species, L. longirostris (Yang et al., 2012). This pattern has been associated with recent population expansions following population bottlenecks (Grant & Bowen, 1998;Yang et al., 2012). Evidence of a historical bottleneck and a possible recent colonization was also confirmed by the negative and significant Fu's statistic obtained in the Lake Victoria, Lake Edward, and Kazinga channel populations (Fu, 1997). The results are consistent with reports of mass extinctions in these lakes and recolonization events particularly in Lakes Edward, George, and Victoria (Beadle, 1974;Thieme et al., 2005). It should be noted that long droughts characterized the East African region during the late Pleistocene coupled with the tectonic uplifts that were responsible for drying up and river reversal between the lacustrine environments (Day et al., 2013).
Additional evidence to the Pleistocene events of river reversal is the complete desiccation of Lake Victoria following a long drought period (Johnson et al., 1996). B. docmak being potamodromous was most likely susceptible given its migration patterns between the riverine and lacustrine habits.
Corroborative evidence was obtained using the microsatellite data in which restricted N eLD estimates were recorded in addition to significant heterozygosity deficiencies in all the populations except for Lake Albert. The low N eLD (Table 1), relatively high inbreeding coefficient, and significant heterozygosity deficiencies in the Lake Victoria and Edward populations, further indicate that these populations are constrained and have not had enough time to recover from the severe historic bottleneck. As pointed out in earlier studies the population in Lake Victoria is additionally constrained by anthropogenic pressures such as heavy fishing and exotic species introductions (like Nile perch) that have been clearly documented in the region (Hauser et al., 1998;Kudhongania et al., 1992;Ogutu-Ohwayo, 1990. Evidence highlights that B. docmak was formally a dominant and higher-order carnivorous fish in the Lake Victoria basin, but it currently faces direct competition for prey and is also preyed upon by the introduced Nile perch (Frans Witte, 1997).
Furthermore our study reveals a recent expansion amidst high, but insignificant F IS , thus rejecting the hypothesis of possible inbreeding. It is, therefore, predicted that the large surface area of Lake Victoria (68,000 km 2 ) provides an advantage of a large habitat in which the species will survive for a while if proper management strategies are put in place. Conversely, Lake Edward which is a small (2,235 km 2 ) EAR lake, has been considered hydrologically and chemically sensitive to climatic changes as it lies in the intersection of the Indian and Atlantic air masses (Russell & Johnson, 2001). The truncated gene flow between Lakes Edward, Victoria, and Albert due to geological and biogeographic barriers largely exposes Lake Edward populations to risk, especially in the event where they are exposed to stochastic events (such as disease outbreaks) and anthropogenic pressures (including heavy fishing and habitat degradation).
The exception of Lake Albert and the Victoria Nile River in showing negative, but nonsignificant neutrality tests, reveals the possibility of a long-term demographic stability. The microsatellite data confirms the stability of the Lake Albert system following a balanced heterozygosity consistent with a stable population as assessed by the bottleneck test. Furthermore, the Lake Albert population was characterized by an infinite N eLD estimate coupled with the highest level of heterozygosity and number of private alleles. The evidence of stability shown in the Lake Albert population suggests the presence of undisturbed habitats within this lake. Hence, it is not surprising that Lake Albert has also been referred to as an abyss with deep waters compared to the shallow Lake Victoria (Mwanja et al., 2014).
The evidence for genetic stability suggested in the Lake Albert population in the current study does not in any way warrant populations of B. docmak from this particular lake to be neglected for conservation; however, the findings do provide a baseline on the genetic status of the species following historical processes (Johnson et al., 1996(Johnson et al., , 2000Russell & Johnson, 2001). The current anthropogenic pressures such as aquaculture establishments (Dickson et al., 2012) on the lake, as well as proposed oil extraction within the Albertine region (Kathman & Shannon, 2011;Vokes, 2012), could potentially accelerate habitat destruction and loss of genetic biodiversity. Therefore, a genetic baseline is needed to monitor and design appropriate management measures for the lake and the region at large.
Information provided by the current study will facilitate the comprehensive management of B. docmak and related taxa for sustainable harvest in the wild and/or culture under captive conditions. The strategy should consider river and lake regulations together with their associated developments, such as, damming, hydropower generation, fishing, transport, among others. Replenishment of fish stocks through restocking and aquaculture need to take into account the current and historic genetic diversity and population structure of the identified stocks. These stocks potentially translate into majorly two evolutionary significant units (evidence from mitochondrial data) and two management units (based on STRUCTURE analysis). These findings are important in developing appropriate conservation and management strategies which depend majorly on our ability to correctly assign genetically distinct populations (Latch, Dharmarajan, Glaubitz, & Rhodes, 2006).

| CONCLUSION
The low genetic diversity and strong population structure patterns Strategically, the management of aquatic fauna in the East African region should initially take into account the two major management units identified, that is the Albertine rift valley region and the Lake Victoria basin. However, further research is required for the Lake Albert population which was singled out as a discrete group at a finer scale with DAPC analysis to make a total of apparently three management units comprising of Lake Edward and Kazinga Channel cluster, Lake Victoria-Nile River cluster and Lake Albert.

ACKNOWLEDGMENTS
Sincere thanks also go to Ondhoro Constantine, Matthew Mwanja,