Fisheries and fishery products are vital in the developing world but heavily threatened through various anthropogenic activities which may compromise the continuity of the resources (35). One aspect of the anthropogenic threats is the change or alteration of the natural genetic structure of fish stocks through admixture (36, 37). To understand the admixture of stocks, it is only possible if the source populations can be differentiated using genetic markers. We show the importance of SSR-GBS for a deeper understanding of population dynamics, in particular, the East African O. niloticus, towards the alignment of management and conservation strategies. In this study, we investigated the phylogeographical patterns and we found large differences between lakes e. g. Lake Tana and also differences between natural water catchments that allows identifying populations. Here, we discuss the current state of O. niloticus in reference to phylogeographical patterns and anthropogenic activities.
Phylogeography of East African O. niloticus
In all analyses, we found a clear differentiation among all three African regions included in this study (East Africa, Burkina Faso, and Ethiopia), indicating a low degree of connectivity amidst them and highlighting the high level of differentiation between regions. Tana was completely distinct from the remaining populations. This applies not only to the Ethiopian populations but also to the East African ones. So, the genetic distance in Ethiopia is higher than between the East African and West Africa populations, indicating a divergence higher than we would expect within a species. These results are consistent with the previous reports (18), rather than the contradicting findings of the subspecies treatment based on the traditional morphometric and meristics (2). This high level of differentiation was confirmed in concurrence to these past studies and so the revision of the species delimitation for these populations is suggested.
Lake Tana lies in the Ethiopian mountains and is isolated from the Lakes in the Rift valley (38). This might explain the high degree of differentiation of this lake because of the lack of connectivity and divergent ecological conditions. Contrary, Lake Hashenge which is also in the Ethiopian mountains is related to the Rift Valley lakes. Lake Hashenge is reported to have been stocked with O. niloticus following mass mortalities of the native species (39). The native status of this lake is unclear since it could have been restocked with O. niloticus that originated from the Rift Valley Lakes. Besides that, we see a slight differentiation in PCoA between Lake Hashenge and the Rift Valley Lakes in Ethiopia, which may reflect an unsampled source of stoking or differentiation accumulated because of the high degree of isolation of the lake.
In East Africa, genetic structure reflected different catchments. The population from Lake Turkana was genetically distinct from the Ugandan populations which are expected given its high geographical isolation (40). Our findings concur with the previous works that treated the Turkana population as a different subspecies (O. vulcani) (2). The high diversity and number of private alleles found in Lake Turkana can be a consequence of this isolation. The East African arid Lake naturally is also characterized by a remarkable genetic diversity. One factor might be introgression perhaps from anthropogenic activities or influx of gene flow from River Omo (Ethiopia). However, this is not clear and a better sampling from the region needs to be included to evaluate the extent of the observed current genetic structure of the population.
In Uganda, despite the high degree of connectivity and proximity between the water bodies, O. niloticus populations were clearly structured. These reflected three main groups: 1) (Lakes George and Edward, as well as Kazinga Channel, 2) Lake Albert, River Nile, and Kyoga and 3) Lake Victoria system. The 2nd and 3rd groups are discussed in more detail under anthropogenic activities subsection. The 1st group, Lakes George and Edward are connected via the Kazinga channel which also explains the high natural migration rates between these populations. The different genetic structure between the western Rift Valley Lakes (Edward-George-Kazinga Channel and Albert) was conserved despite being connected through River Semliki that flows from Lake Edward and Albert (41). The strong rapids and falls present in this river (41, 42), might constitute a strong barrier to gene flow, which maintains these systems apart. These findings are congruent with recent work on O. niloticus geometric morphometrics (43) but do not concur with past studies (2, 20). This incongruity might be associated with different methodological approaches utilized between the earliest studies and the current one. For example, using morphometric and meristics methods, O. niloticus from the Edward-George system and Albert was treated as one subspecies; O. niloticus eduardianus (2). However, the use of traditional morphometrics has been strongly labeled as weak in identifying species due to the lack of informative characters (18). Similarly, while we used SSR-GBS techniques, (20) employed random amplified polymorphic DNA (RAPD) markers, which due to their dominance genotypic nature, provide only part of the information content (22).
Anthropogenic activities-fish translocations
In East Africa, we know that O. niloticus was introduced into several water bodies through stocking activities and here we show their genetic impacts. Referring to both genetic structure and migration rate analyses, the two Ugandan groups (the George-Edward complex and Lake Albert) contributed to the stocking of different water bodies. Apparently, O. niloticus from the southwestern Ugandan high-altitude Lakes; Mulehe and Kayumbu, originated from the Western Rift Valley Lakes – Edward and George. For the 2nd group, Lake Kyoga and River Nile (Victoria Nile) are genetically analogous to Lake Albert, suggesting that, the latter population might have contributed genes to the gene-pool of the former systems. Although Lake Kyoga is connected to Lake Albert via River Nile, their genetic similarity is unlikely related to the consequence of natural migration via water flow. The main reason here is the natural occurrence of Murchison Falls on the River Nile that acts as a barrier between the systems (3, 41). For this matter, the genetic similarity between River Nile, Lakes Kyoga, and Albert populations may have resulted in stocking regimes using the latter as source (3).
Fish farms seem to have sourced fish seed from multiple populations, resulting in admixed stocks. Our results show that Lakes Albert, and Kyoga, as well as River Nile, contributed to the gene pool of the farmed populations. Based on genetic distance, Lake Albert was the main contributor to Rwitabingi and Bagena farms while Kyoga to Sindi farm. However, we also observed a high amount of gene flow from Kyoga to Rwitabingi and all these farms appeared to be admixed with other populations including Lake Victoria. Apart from farms, evidence of admixture was probable in the East African natural populations, which seems to have been promoted by anthropogenic activities. This is supported by the fact that when non-native populations were unconsidered in the STRUCTURE and PCoA analyses, signals of admixture were minimal, and clear genetic structure assignments could be observed. In East African, admixture in O. niloticus populations may stem from three main processes: 1) translocation from multiple sources into the non-native water bodies, 2) back translocation from non-native to native populations, and 3) hybridization of O. niloticus with congeneric species promoted by translocations.
The first and third processes may explain partly the genetic variation found in the 3rd group; Lake Victoria (see above the three Ugandan groups). Although O. niloticus in Lake Victoria is generally isolated, based on the distance neighbor Network tree, the population occupied an intermediate position between the above described; 1st and 2nd, Ugandan groups. Thus, it is clearly possible that multiple stockings might have contributed to the gene-pool indicated by the Lake Victoria population. For example, (2) suggests that introductions into Lake Victoria may have originated from Lake Edward, with other authors suggesting multiple sources (4, 5, 12, 44), which support our results. The highly diverse and differentiated gene-pool in Lake Victoria could have originated from the admixture of several lineages due to multiple sources.
On the other hand, possible hybridization of the introduced O. niloticus with the indigenous relative species (O. variabilis and O. esculentus) in Lake Victoria may explain some of the genetic variation patterns found in this lake. First, this lake together with Turkana showed values of private alleles up to four times higher than the remain populations. This genetic variation could have originated from introgression with other species that have not been included in the analysis. Similarly, the probable hybridization may explain the high genetic diversity and divergent gene-pool detected in the system. Within Lake Victoria, the Sango Bay subpopulation appears to be an extreme case from this by showing the highest degree of genetic divergence. Remarkable genetic differentiation in Sango Bay was not only noticed when compared with the remaining subpopulations within the Lake, but also with the other East African populations. In this case, during the boom of the O. niloticus population in Lake Victoria (3-5, 45), a larger portion of the native species' genetic materials may have been introduced into O. niloticus gene-pool. This is just a hypothesis since, in this study, we cannot directly test for hybridization because we did not include samples of O. niloticus congenerics. However, hybridization involving O. niloticus and other tilapiines has been reported to be relatively frequent and it needs to be considered (9, 33, 46, 47).
In case admixture/hybridization shaped the gene-pool of Lake Victoria, it may have adaptive consequences and compromise the sustainability of O. niloticus. Although hybridization may lead to heterosis/hybrid vigor (48, 49), the admixture is usually reported to have negative consequences (37, 50). Introgression can contribute to outbreeding depression either by the introduction of maladaptive alleles or through the dilution of alleles important for local adaptation (51). In more drastic scenarios, hybridization can result in genomic incompatibilities contributing to a fast reduction of population fitness (51). Alternatively, the hybrids may potentially exhibit more fitness and subsequently extirpate the parental lines (46). The observed genetic structure of O. niloticus populations in Lake Victoria was unexpected and has not been reported before, which calls for further investigations for taxonomic recognition.
Evidence for the second process of admixture was only found in Lake Albert. In the structure analysis, this population showed to be admixed with Lake Kyoga. Besides we found significant migrations from Lake Kyoga to Lake Albert. These results indicated that admixture with respect to translocations does not only contribute to non-native populations but also to native ones. The sequence of gene flow from Lake Kyoga to Albert is not clear as none of the previous reports have indicated this. However, it is likely that aquaculture activities might be contributing to the observed gene flow between Lakes Kyoga and Albert.
Anthropogenic activities-Consequences of overfishing
Some water bodies, especially Lake Kyoga and River Nile showed signals of genetic erosion. These were indicated by low genetic variability and evidence of bottleneck with respect to G-W estimations. In a previous study, (20) had already reported this to be the case for Lake Kyoga. High loss of genetic diversity among populations, particularly, in fishes has been attributed to over-exploitation (52). This might be the case for the L. Kyoga population, but other factors may also prevail. The low diversity in River Nile could be linked to the hydro-electric power dams that have been constructed along the river (the upper Nile of the Ugandan side), which might be restricting gene flow and increasing genetic drift. However, this needs to be assessed in further analyses, especially when additional samples are collected in sections of the lower Nile (below Murchison falls), where apparently there no dams. Nevertheless, based on the introduction of Nile tilapia in these water bodies, the low genetic diversity might also be a consequence of the founder effect.
Implications for management and outlook
Overall, we found evidence that anthropogenic activities affected the gene-pool of the East African O. niloticus. The main consequence might have been admixture and apparently hybridization between different stocks and species respectively. In the long term, this may have negative effects on population fitness due to outbreeding depression and genetic swamping. Thus, management measures should inhibit any form of unauthorized spread of fish in the aquatic ecosystems. The Western or Albertine Rift Valley lakes (Edward-George) may be ideal broodstock sources for subsequent breeding programs and aquaculture, as these systems seem un admixed from external sources. To avoid an influx of feral populations and other infrastructure developments, a proper environmental impact assessment should be prioritized before implementation. We also found evidence of genetic erosion that in the long term might lead to inbreeding depression and loss of adaptive potential from these populations. Some of the potential causes were overfishing and the construction of hydropower dams, which should also be taken into consideration in future management practices.