Genetic identification and diversity of stocks of the African bonytongue, Heterotis niloticus (Osteoglossiformes: Arapaiminae), in Nigeria, West Africa

Inland fisheries are an important source of protein and income for people in Africa. Their sustainable management can greatly benefit from identification of regional genetic stocks and characterization of their genetic diversity, but such information is lacking for most African freshwater fisheries. The African bonytongue, Heterotis niloticus, is an important component of inland fisheries in West Africa. Nigeria has the largest fishery for African bonytongues, representing ~ 86% of the global total. Recent declines in yields at some Nigerian locations, however, suggest current levels of exploitation may be unsustainable. Habitat degradation also may be impacting some stocks. Despite its commercial and nutritional importance, the African bonytongue has been the subject of scant genetic research to support management. We examined patterns of genetic diversity in natural populations of H. niloticus at four locations in Nigeria, including Kainji Lake, a reservoir on the Niger River in north-central Nigeria, and three southern localities (Ethiope River, Igbokoda River, and Epe Lagoon), as well fish from the Ouémé River delta near Porto Novo, Benin. Eighty-five specimens were genotyped for nine microsatellite-loci. Genetic diversity estimates were highest at Kainji Lake, and substantially lower at southern localities. High levels of genetic differentiation were detected between samples from Kainji Lake and those from southern localities. Low, yet significant FST values were observed among samples from southern Nigerian localities that were more differentiated from the sample from nearby coastal Benin. We thus recommend that African bonytongues from the five locations be considered distinct genetic stocks and managed accordingly.


Methods
Sampling and DNA extractions. Samples of wild H. niloticus specimens were obtained from four inland waters in Nigeria: Kainji Lake (including samples from an area just above the lake and a location below the dam), Igbokoda River, Epe Lagoon and Ethiope River (Fig. 1). Additional specimens were obtained from the Ouémé River floodplain near Porto Novo, Benin. Dead fish were purchased from local commercial fishermen, and in some cases only pectoral fin clips were obtained from fish at the point of sale to consumers. Animal ethics approval was not required for this study because live fish samples were not collected. A pectoral fin clip from each specimen was stored in an Eppendorf tube with 70% ethanol. Genomic DNA was extracted from the pectoral fin clips of the tissue samples using the Qiagen DNeasy Blood & Tissue Kit (Valencia CA).
Microsatellites amplification and genotyping. Samples were genotyped for nine microsatellite loci (Table SM1) reported for the African bonytongue 20,22 . Amplification of the extracted DNA was accomplished through multiplexing using Qiagen's Multiplex PCR kit (Multiplex PCR Kit, QIAGEN, Valencia, CA) carried out in 10 µl reaction volumes containing 50-1000 ng genomic DNA, 5  www.nature.com/scientificreports/ cent labelled) and reverse primer mix and 3 µl of nuclease-free water (Qiagen Sample and Assay Technologies, Germany). Amplifications were carried out using a BIO-RAD Thermal Cycler (T100TM) programmed according to standard multiplex PCR protocols as follows: initial activation step: 15 min at 95 °C, 35 cycles of denaturation at 94 °C for 30 s, annealing at 57 °C for 90 s, extension at 72 °C for 60 s, and a final extension step 60 °C for 30 min. PCR products were run on 2% agarose gel electrophoresis to check for amplification of all loci prior to genotyping, and positive PCR products were then run on an ABI 3130xl automated sequencer using Genescan 400 HD ROX as size standard (Applied Biosystems). Allele calling and sizing for each sample was obtained using Genemarker software (version 1.75).
Genetic diversity and population differentiation analyses. To 32,33 was conducted to visualize levels of genetic differentiation among the samples from the different locations, defining sampling locations as a priori groups. DAPC is a non-model-based method that maximizes the differences between groups while minimizing variation within groups 34 ; and distances between genetic clusters using a priori groupings reflect underlying F ST 35 . STRU CTU RE 2.2.3 36 , which performs modelbased clustering with a Bayesian approach, was also used to examine population subdivision. Two models were used: admixture with correlated allele frequencies, and no admixture with correlated allele frequencies. K values from 1 to 5 were tested in ten iterations, with 500,000 steps and a burn-in of 125,000 steps, and all other settings were set to default. GenAlEx was used to construct a genetic distance matrix, from which a principal coordinates analysis (PCoA) was performed to identify population clusters.
Finally, we analyzed the relationship between genetic differentiation and geographic distance by testing for isolation by distance (IBD) using ISOLDE in GENEPOP on the Web 4.7. Two kinds of analyses were performed: (1) between localities; and, (2) between individuals. We used Google Earth Pro to measure distances between sample locations. For the distances between locations within the southern rivers, we measured the shortest straight-line distance. For distances between the Kainji Lake and all other localities, we followed the contour of the Niger River to the closest point to the Ethiope River, and from there we used straight lines. Statistical significance based on the Spearman's rank correlation coefficient was evaluated using Mantel tests 37 .

Results
Genetic diversity. Genotyping scores for all individuals are shown in Dataset S1. Complete genotypes for the nine microsatellite-loci were obtained for 83 specimens: 23 from Kainji Lake, 19 from Ethiope River, 15 from Igbokoda, 20 from Epe Lagoon, and six from Porto Novo (Table 1). No linkage disequilibrium was detected among loci. A total of 98 individual alleles were observed across these 83 specimens; and the number of alleles per locus ranged from 4 in locus Hn45 to 19 in Hn30 (Table SM2). Departures of Hardy-Weinberg equilibrium were detected for locus Hn47 in Kainji Lake, and locus Hn11 in Igbokoda. According to MICROCHECKER, only locus Hn11 in the Igbokoda sample shows signs of a null allele, which is suggested by a general excess of homozygotes for most allele size classes. Excess of homozygotes at this locus, however, may reflect inbreeding at this locality (see below).
The sample from Kainji Lake revealed the highest genetic diversity among the five populations examined (Table 1): average number of alleles (Na) = 8.33; average number of effective alleles (Ne) = 4.56; Shannon's information index (I) = 1.6; observed heterozygosity (Ho) = 0.73; expected heterozygosity (He) = 0.70; and unbiased expected heterozygosity (uHe) = 0.72. The highest number of private alleles was also observed at Kainji Lake: 21 out of 75 alleles in this locality. This sample had a low inbreeding coefficient (F IS = -0.03). Genetic diversity in the samples from the three southern Nigerian populations were substantially lower: Na ranged between 4.89 (Ethiope River) and 5.67 (Igbokoda); Ne between 2.55 (Igbokoda) and 2.65 (Epe Lagoon); I between 0.97 (Epe Lagoon) and 1.11 (Igbokoda); Ho between 0.44 (Epe Lagoon) and 0.50 (Ethiope River); He between 0.47 (Epe Lagoon) and 0.54 (Igbokoda); uHe between 0.48 (Epe Lagoon) and 0.56 (Igbokoda). In these localities, F IS Table 1. Genetic diversity estimates for H. niloticus from five localities in Nigeria and Benin. N, number of individuals; N A , number of alleles; N P , number of private alleles; Na, average number of alleles per locus; Ne, average number of effective alleles per locus; I, Shannon's information index; Ho, mean observed heterozygosity; He, mean expected heterozygosity; uHe, mean unbiased expected heterozygosity; F, inbreeding coefficient (fixation index).  (Fig. 2), with the greatest distances observed between fish from Kainji Lake and fish from the southern localities. When the sample from Kainji Lake was removed, the greatest distances observed were between fish from the Ouémé River locality in Benin and fish from the three southern Nigeria localities. Fish from Epe Lagoon appear to be more genetically differentiated from conspecifics obtained from the other two southern Nigerian localities, and some overlap is observed between individuals from Igbokoda and Ethiope River. Analyses in STRU CTU RE and PCA (Figs. 3 and 4) also show a clear separation between fish from Kainji Lake and fish from the other localities; however, they do not show a separation among fish samples from the different southern localities. The STRU CTU RE plot in Fig. 3 includes two additional specimens from the Kainji Lake sample for which partial genotypes were obtained (one with four and one with five microsatellite-loci), and they grouped with all other specimens from Kainji Lake. The sample from Kainji Lake contained 3-4 individuals that were collected at Faku, just below the dam. For analyses excluding the Kainji Lake sample, STRU CTU RE and PCA plots do not show evidence of genetic differentiation among fish from the southern localities. IBD analysis using F ST values between populations produced an R 2 = 0.65 with a p value = 0.049 for the analysis including all localities (Fig. 5a), whereas the analysis excluding Kainji Lake resulted in an R 2 = 0.22 with a p value = 0.21 (Fig. 5b). IBD analyses using genetic differentiation between individuals resulted in an R 2 = 0.09 with a p value < 0.0001 for the analysis including all localities (Fig. 5c), whereas the analysis excluding individuals from Kainji Lake resulted in an R 2 = 0.005 with a p value = 0.21 (Fig. 5d).

Discussion
Even though genetic evidence has long been recognized as crucially relevant for sustainable management of fisheries 11,38 , such information is lacking for most African inland fisheries 12 . The present study advances understanding of population genetics of the African bonytongue, an important fish for commerce and subsistence throughout a large region of Africa. We documented patterns of genetic diversity and structure of African bonytongue in Nigeria, where > 85% of the global capture of this fish occurs. The knowledge generated by the present study is important for guiding management and conservation strategies for this species.
The greatest genetic differentiation was observed between fish from Kainji Lake and those collected in the southern localities (average F ST = 0.18). Although high, this level of differentiation is smaller than that observed (average F ST = 0.27) between fish from Malanville, also in the Niger River, and southern Benin localities in Mono River and the Ouemé-Sô river floodplain system 20 . Differentiation between fish from Kainji Lake and those Table 2. Pairwise F ST estimations with (above the diagonal) and without correction for null alleles (95% confidence intervals in brackets; p values below brackets; * = < 0.0001). www.nature.com/scientificreports/ collected in the southern localities was supported by results from all genetic differentiation analyses, as well as the high number of private alleles (21) observed in Kainji Lake. The large geographic distance between Kainji Lake and southern populations is probably an important factor restricting gene flow of the African bonytongue between the two regions. A pattern of IBD was detected only when analyses included the sample of Kainji Lake.
In addition, movements of the African bonytongue must be affected by the presence of the Kainji Dam, completed in 1968 39 . The dam constitutes an effective barrier for upstream movement of fishes from the river below, although fish can move downstream from the lake through the spillway. Nonetheless, the dam appears to have impacted downstream populations after its construction. Bonytongues were rarely caught at Faku, just below the dam, and Awuru, ~ 25 km downstream from the dam, in 1973 and 1974 39 . Our sample from Kainji Lake included a few specimens from Faku, and our genetic analyses grouped them in the same cluster as fish obtained from upstream of the lake, suggesting gene flow occurs from the lake to reaches just below the dam. The African bonytongue also became less frequent in landings in the second half of 1968 and in 1969 at Jebba and Pategi, ~ 100 km and ~ 220 km below the dam, respectively 40 . Between 1970 and 1972, however, bonytongues comprised a large percentage of the fish biomass landed at these locations, as well as at Lokoja, ~ 400 km downstream the dam. www.nature.com/scientificreports/  www.nature.com/scientificreports/ This proportion dropped substantially in these three below-dam localities between 1974 and 1975 39 , which may be due to local exploitation. It is important to investigate genetic differentiation of African bonytongues at these below-dam localities, where this species has historically been a valuable fishery component. Comparisons between microsatellite results for bonytongues from Malanville in Benin 20 and Kainji Lake (this study) suggest restricted gene flow between these two localities on the Niger River separated by a river distance of ~ 230 km. Microsatellite datasets for these samples were not analyzed together because of potential bias when combining datasets from different studies 41 . Marked differences in mean expected heterozygosity (He) for these samples, however, suggest highly restricted gene flow between these two areas of the Niger River. He was 0.57 in Malanville (n = 12), whereas He was 0.73 in the sample from Kainji Lake (n = 23). Mean He is expected to be fairly insensitive to potential bias due to differences in sample size 42,43 , and can be useful for comparisons between different studies 44 .
Among southern locations, gene flow appears to be more restricted between the lower Ouémé River in Benin (Porto Novo) and the southern Nigeria rivers sampled (average F ST = 0.11). This level of differentiation is higher than the one reported in southern Benin between fish from Mono River and localities in the Ouemé-Sô river floodplain system (average F ST = 0.09). Average divergence among southern Nigeria localities (average F ST = 0.05) is higher than the reported between localities within the Ouemé-Sô river floodplain system (average F ST = 0.03). A pattern of IBD was not detected among southern localities in this study. An ocean connection and the presence of brackish waters in Lagos Lagoon, which is connected to the western portion of Epe Lagoon, may represent effective barriers to gene flow for H. niloticus between Porto Novo and Epe Lagoon. According to our results, gene flow between upstream areas of the Ouémé River Basin and rivers connecting to Epe Lagoon also appears restricted. Significant pairwise F ST values among the southern Nigerian localities suggest restricted gene flow among them, despite potential surface water connections provided by a complex network of waterways and annual flooding. Igbokoda River and Epe Lagoon are connected through a water channel that joins the Lekki Lagoon, which is connected to the eastern portion of Epe Lagoon. Thus, it appears that although movement of fish between the two localities may occur, it is not enough to prevent subtle genetic differentiation. Igbokoda and Ethiope rivers are separated by > 150 km. Flooding and water channels may facilitate movement of fish between www.nature.com/scientificreports/ both localities, although the lower portion of the Igbokoda River has brackish water that may act as a barrier to dispersal. STRU CTU RE analyses did not detect subdivision among fish from southern locations, even though simulations have shown that this method can detect substructure in cases where F ST values are as low as 0.03 when ten highly variable microsatellite loci are used 45 . Thus, it is possible that the microsatellite loci we used in this study are not variable enough for STRU CTU RE to detect substructure with F ST values of 0.11. In the previous study from Benin, most STRU CTU RE analyses did not detect genetic differentiation between fish from the Mono River and those from localities in the Ouemé-Sô River floodplain system, even though all F ST pairwise values were significant (average F ST was 0.09), and other methods also clearly showed this differentiation 20 . Studies of large-scale patterns of genetic structure of other important capture fisheries are limited in West Africa. A recent study of the Nile tilapia Oreochromis niloticus analyzed samples collected from 23 localities across eight West African countries, representing the major catchments of the Volta, Niger, Senegal and Gambia River basins 46 . That study found a pattern of IBD among all localities, and significant spatial genetic structure that largely corresponds to major river basins and, to a lesser extent, sub-basins. Within the Volta Basin, a significant, yet much weaker relationship between genetic and geographic distances was observed, suggesting IBD was a relatively minor factor shaping genetic differentiation amongst populations. This is similar to our findings for the African bonytongue, for which IBD was significant at a large scale (i.e., when Kainji Lake samples were included), but not at smaller scales (i.e., when only samples of southern populations were included). Most of the pairwise F ST values in the Nile tilapia study were significant, except for pairwise comparisons among localities separated by some of the shortest distances. Significant genetic differentiation in the Nile tilapia was usually observed between populations separated by more than ~ 90 km. Interestingly, Nile tilapia samples from the only two localities sampled in the Niger River, Malanville and Mopti, which are separated by ~ 1400 km, show high genetic similarity, suggesting high levels of gene flow. This contrasts with what we observed for the African bonytongue in the Niger River, for which high genetic differentiation appears to occur at comparatively shorter distances within this river, i.e., between Malanville and Kainji Lake (~ 230 km), and between Kainji Lake and the lower Niger portion (~ 700 km).
An effect of floodplain connectivity and geographic scale has also been reported in the African bonytongue's closest living relative, Arapaima gigas 47 . This species, which is distributed throughout the Amazon River Basin, is the only other living member of the Arapaiminae 48 , and both species construct nests where they lay and protect eggs. The predatory arapaima protects free-swimming larvae much longer (several weeks) than the omnivorous African bonytongue (a few days), and grows much larger (over 2 m and 100 kg compared to Heterotis at 1 m and 10 kg). At a fine scale (e.g. within the same floodplain system; < 25 km in Arapaima and ~ 75 km in Heterotis), the two species tend to exhibit genetic homogeneity 20,47 . At a meso-scale (e.g. in separate floodplain systems; ~ 100 km in Arapaima and 69-400 km in Heterotis), both species exhibit low but significant values of genetic differentiation. Finally, at the largest scale (e.g. > 1300 km in Arapaima and > 510 km in Heterotis), the highest levels of genetic differentiation were observed in both species. Thus, despite occupying two separate continents these two sister taxa exhibit similar patterns of genetic differentiation.
A broad range of genetic diversity values are observed in African bonytongue populations of Nigeria and Benin. Na values for this fish are similar or much lower than the average (Na = 9.1) reported for freshwater fishes 49 . Heterozygosity includes values considerably higher or lower than the average heterozygosity (H = 0.54) reported for freshwater fishes 49 . The sample from Kainji Lake had the highest heterozygosity (Ho = 0.73; He = 0.70) and second highest allelic diversity (Na = 8.33) among all African bonytongue populations examined to date. The sample from the Oueme-Sô river-floodplain system in Benin has the highest Na (9.25) for African bonytongues, but with a much larger sample examined (n = 184), and the second highest heterozygosity (Ho = 0.60; He = 0.69) 20 . Genetic diversity values for fish from the southern Nigeria localities are much lower (Na range = 4.89-5.67; Ho range = 0.44-0.50; He range = 0.47-0.54), and one of these localities, the Igbokoda River sample, shows high levels of inbreeding (F IS = 0.18). The high genetic diversity in Kainji Lake contrasts with the low genetic diversity observed at Malanville, Benin (Na = 3.50; Ho = 0.34; He = 0.43), the only other Niger River locality sampled to date, and where high inbreeding was also observed (F IS = 0.20) 20 . Low genetic diversity at Malanville could be associated with intense fishing pressure 20 . Although we observed low allelic diversity (Na = 3.56) for the sample from the Oueme-Sô River at Porto Novo, Benin, Ho (0.57) was more similar to the value previously reported for the Oueme-Sô river-floodplain, and He (0.49) was similar to that observed for samples from southern Nigeria. The smaller Na and He values obtained for Porto Novo compared to the Oueme-Sô river-floodplain system are likely due to the small (n = 6) sample size 43 . Values of Na more similar to the one obtained in Porto Novo were previously obtained for some localities with small sample sizes within the Oueme-Sô river floodplain system 20 . For a sample of n = 6 in the Ouemé River channel, values were Na = 3.38, Ho = 0.48 and He = 0.66; whereas for a sample of n = 10 in the Sô River channel, values were Na = 3.62, Ho = 0.55 and He = 0.61. Implications for conservation and management. The observed patterns of genetic differentiation indicate that African bonytongue from each of the localities examined in this study correspond to differentiated populations (i.e., genetic stocks), and should thus be treated separately for conservation and management. The Kainji Lake population is highly differentiated from the southern populations. In the south, the sample from Porto Novo is highly differentiated from southern Nigerian populations. Porto Novo is part of the Ouemé-Sô river floodplain system in Benin, for which Hurtado et al. 20 report low levels of genetic differentiation among fish collected from this system, with sampling localities separated by up to ~ 75 km (average F ST = 0.03). In that study, the fish from the Ouemé-Sô river floodplain system were highly differentiated from those collected in the Mono River and the Niger River locality of Malanville. The southern Nigerian populations sampled in this study show subtle, albeit statistically significant differentiation; we thus recommend that they be managed as local stocks. These genetically distinct populations could constitute valuable genetic resources for future use in aquaculture. www.nature.com/scientificreports/ Multiple activities threaten the sustainability of the African bonytongue stocks we have identified in Nigeria. Overfishing compromises the sustainability of fishery resources in Kainji Lake 50,51 , and illegal fishing activities, including the use of prohibited gear (e.g. small mesh size nets and destructive fishing gear), fish poisoning, and explosives, have exacerbated overfishing in this lake 52 . Environmental pollution also impacts fish and people who consume them; and high levels of heavy metals have been detected in fish from the area 53 . The Ethiope River flows through a densely populated area of Nigeria's Delta state, where pollution has impaired water quality in some stretches, with documented effects on macroinvertebrates 54 . Effects of degraded water quality in this river have not yet been shown to impair fish survival directly 55 . The Igbokoda River is an important fishing location in the Ilaje local government at Ondo State, a major oil-producing state. Unsafe levels of heavy metals have been found in water samples and fish in this river 56 . Consumption of fish is also considered unsafe in other rivers located in the Ilaje local government due to oil and industrial pollution 57 . Given the low quality of its riverine environment, and high levels of inbreeding detected in this study, urgent attention needs to be paid to the African bonytongue in Igbokoda River. A decade ago, Epe Lagoon was reported to have diverse and abundant fish stocks threatened by a rapidly growing population in the Lagos metropolitan area 58 . Anthropogenic activities appear to largely contribute to pollution in Epe Lagoon. Recent studies of Epe lagoon reported the presence of benzene, toluene, ethylbenzene and xylene (BTEX) and unsafe levels of Polycyclic Aromatic Hydrocarbons (PAH) in H. niloticus; as well as high levels of heavy metals in sediment of the lake, and a high prevalence of an intestinal parasite in H. niloticus, which may be attributable to the water pollution 59,60 .

Conclusion and recomendations
Consistent with a previous study conducted in Benin, we found significant genetic structure among African bonytongue samples from locations in Nigeria. The highest values of genetic differentiation were observed between Kainji Lake and the southern localities examined. We found evidence of gene flow from Kainji Lake to the area just below the dam, but the large distance between Kainji Lake and southern populations probably restricts gene flow between the two regions. Relatively low but significant F ST values were obtained for pairwise comparisons among fish from the southern localities. Based on these results, we suggest that populations of the African bonytongue in each of the Nigerian localities examined in this study be treated separately for conservation and management. Given the observed levels of genetic differentiation, understanding patterns of genetic diversity in the African bonytongue throughout the rest of its range is needed for adequate delineation of management stocks. We recommend that populations located in different rivers that are not part of the same floodplain system and are separated by > 70 km, or populations within a major river separated by > 200 km, be treated as separate management units pending genetic evidence. www.nature.com/scientificreports/