This study was conducted on individuals of Nile tilapia O. niloticus from five different localities in Senegal. The terms “strain” or “population” are used in this study to designate these localities but there is no evidence that they are genetically different. Among these strains, four (Mbodiene, ANA, Sauvage, ITACA) originate from the experimental hatchery of Institute of Research for Development (IRD) in Hann Bel Air in Dakar, Senegal. The fifth strain (Richard-Toll) was collected in the fish farm of the University Gaston Berger (UGB). The Mbodiene strain is from broodstock of natural population maintained in captivity for several generations in the Mbodiene private fish farm, located in the department of Mbour in the region of Thiès. The ANA strain is also from initial broodstock of a natural parental population maintained for several generations in the ANA fish farm of Richard-Toll (Dagana Department, Saint-Louis Region). The individuals of this population analyzed in this study were maintained in captivity in the IRD experimental hatchery for one and a half years. The individuals of the Richard-Toll strain also come from the ANA fish farm but were collected at the UGB fish farm where they were maintained in captivity for five months prior to the experiment. For the ITACA population, the initial broodstock were collected at Technopole, Dakar. This population was then maintained in captivity for several generations in the IRD fish farm. As for the Sauvage strain, it is from the Senegal River, Saint-Louis. It was then maintained in captivity in the IRD Bel-Air fish farm for experimental studies on sex reversal and the production of male mono sex population.

Fish from these five populations (Richard-Toll, ANA, ITACA, Mbodiene and Sauvage), were sampled. Four populations (ANA, ITACA, Sauvage, and Mbodiene) were analyzed for the SCARII marker whereas all five populations were studied with the SSR markers.

DNA from each sampled individual was extracted using the commercial “Quik DNA Miniprep Kit”. Thus, a 20 mg piece of tissue was taken from each pelvic fin sample. It was then cut into small pieces and placed in a mini mortar with quartz and crushed until a homogeneous powder was obtained. The powder from each sample was placed in a 2 ml eppendorf tube containing 95 μl of solid tissue buffer (tissue buffer that causes degradation of cell connections), 10 μl of proteinase K (for protein degradation) and 95 μl of pure water. This mixture was homogenized for 15 seconds with a vortex and incubated in a water bath at 55˚C for 3 hours.

After incubation, the tubes were centrifuged at 12,000 g for 1 minute to remove insoluble debris. The supernatant was then transferred to a new clean tube and measured. The volume of each tube was then doubled by adding genomic binding buffer and homogenized for 15 seconds using a vortex. After thorough mixing, the content of each tube was poured into a zymo spin column placed on a collection tube and centrifuged at 12,000 g for one minute. The collection tubes were discarded with their content. Each spin column was taken up and placed on a new collection tube. A total volume of 400 μl of DNA prewash buffer was added to each column and centrifuged at 12,000 g for 1 minute. A 700 Μl volume of G-DNA wash buffer was then added to each spin column, and the tubes were centrifuged at 12,000 g for 1 minute and the collection tube and its contents discarded.

A volume of 200 μL of G-DNA wash buffer was added to the spin column, which was replaced on the same collection tube and centrifuged at 12,000 g for one minute. After centrifugation, the column was transferred to a 1.5 ml eppendorf tube and dried to remove any traces of wash buffer. After drying, 32 μl of DNA elution buffer was added directly to the matrix and incubated for 5 minutes at room temperature. The tubes were centrifuged at 12,000 g for one minute to elute the DNA. The resulting DNA solution was then stored in the refrigerator at +4˚C. Quality control and quantification of the extracted DNA were performed by electrophoretic migration through a 1% agarose gel to ensure its integrity (quality) and to estimate its concentration. The estimation was done by comparing the band of each sample with the bands of known concentrations of the control marker or the Smart Ladder.

A dozen genetic markers were initially selected for the genetic characterization analyses. These markers were selected based on data from the literature of studies that have been conducted on tilapia and other fish species [12] [15] [26] [27] [28] [29]. Preliminary analyses were then performed to ensure their quality and suitability for use in Nile Tilapia O. niloticus. As a result of these tests, six markers including one SCARII (sequence characterized amplified regions) marker and five microsatellite markers were selected for this study. The microsatellite markers were analyzed for five populations (Mbodiene, ANA, Sauvage, ITACA and Richard-Toll) while for the SCARII marker analyses were conducted only on four of the five populations (Mbodiene, ANA, Sauvage and ITACA). The Richard-Toll population was not analyzed for this marker.

The SCARII marker is one of the most stable markers, usually derived from RFLP, RAPD, AFLP markers [15] [20] [30]. The basic principle of this genetic marker is to design specific primers of about 20 bp in length from the acquired sequence information and then reveal a polymorphism by ordinary PCR. Because of its speed, simplicity of application, and low cost, the SCARII marker is particularly suitable for analysis of large numbers of samples Li et al., [22]. However, it requires minimal information about the composition of the sequence to be amplified and enormous effort and additional expense in defining locus-specific primers.

For PCR amplification of the SCARII marker, the conventional amplification protocol that requires a pair of primers (one forward and one reverse primer) was used (Table 1). PCR amplification was performed in one well of a 96-well PCR plate with a total reaction volume of 25 μl containing 1 μl of template DNA, 1 μl for each primer at 10 μM (forward and reverse primer), 12.5 μl of 2× Master Mix (MGCL2 buffer mix, Taq, DNTPs), 9.5 μl of ultrapure molecular free water nuclease.

Primer testing and selection of SSR markers (Table 1) were performed by acrylamide gel electrophoretic migration after PCR. This PCR amplification assay selects primers that correctly amplify the correct DNA fragment and eliminates those that did not work. Thus, out of seven pairs of SSR primers tested, only one (G12288) was eliminated as monomorphic.

AT: annealing temperature; bp; base pair; GC: guanine-cytosine.

Amplification of microsatellite markers or SSRs was performed using a conventional PCR protocol but all direct primers were labeled with a fluorescent dye (M13-Primer) for detection of the amplified fragment. The PCR amplification reaction consisted of a 5 ng/l DNA template and a master mix composed of TP 10 × (85 μl), DNTP 2000 (85 μl), MgCl₂ 50 mm (34 μl), AMF-M13 10 μm (7 μl), AMR 10 μm (7 μl), labeled M13 (Dye 700 or Dye 800): 10 μm (7 μl), Taq 2 U/μL (8 μl), and H₂O (191 μl). Amplification was performed in a thermal cycler according to the following program: initial denaturation at 94˚C for 5 minutes, denaturation cycles at 94˚C for 45 s, hybridization at 54˚C - 58˚C for 1 min, elongation at 72˚C for 1 min, final extension at 72˚C for 5 min, hold at 4˚C.

Sequencing of the DNA fragment of each individual amplified by the SCARII primers was performed by the South African genomics company “Inqababiotec”. The first step of the SSR genotyping consists in depositing the DNA on an acrylamide gel to produce an electrophoresis on the Li-Cor system (sequencer). Preparation of the acrylamide gel consisted of pouring 20 ml of cold acrylamide (LR) into a small beaker and adding 175 μl of 10% APS (ammonium persulfate) and 25 μl of temed (tetra-methyl-ethylene-diamine). After gel polymerization (approximately 1.5 hours), 2 μl of PCR product from each sample was plated onto a 96-well plate into which 8 μL of urea blue is added. PCR-derived microsatellite amplicons migrated through the gel mesh using an ultrahigh-voltage electric field (1500 V). The fragments marked during amplification emit fluorescence when excited by laser diodes at 2 different wavelengths (680, 780 nm). An IR camera detects these signals and the system retranscribes a migration profile on the computer interface connected to Li-Cor. At the end of the migration, the images are recorded and processed with the XNView software.

Data obtained from SCARII sequencing and SSR genotyping were cleaned to correct any errors. They were then used to assess genetic variability (within- and between-population genetic diversity) and the genetic structure of populations.

Genetic diversity was assessed using GENALEX 6.501 [30] by analyzing the following parameters: polymorphism rate, allele frequency at each locus, number of alleles per locus (NA), average number of alleles or allelic richness, number of effective alleles (NE), observed heterozygosity (Ho) and expected heterozygosity (He). The Fis index (differentiation of individuals within populations) or inbreeding coefficient was used to assess happlotypic diversity and to assess the degrees of inbreeding within each population using DNASP 6. The typical Haplex Network was generated by the Haplotype Network.

Total genetic diversity (HT), within-population genetic diversity, between-population genetic diversity (Dst), and genetic differentiation coefficient (Gst) were calculated from the allelic frequencies of microsatellite loci for the five populations. Genetic differentiation among populations was assessed using AMOVA, Fst, (the correlation between alleles within a population versus all subpopulations), differentiation index (differentiation of individuals from the total), and discriminant analysis of principal component (DAPC).

The DAPC was used to illustrate the genetic structuring of populations. In the DAPC approach, data are first analyzed using principal component analysis (PCA) and then clusters are identified using discriminant analysis (DA). The performance of this method was evaluated using simulated data, which was also analyzed using the structure as a reference.

Nei’s genetic distance was used on the SCARII marker to assess intra- and inter-population genetic structure [31]. Tajima’s neutral test was used to illustrate the allelic frequency distribution of nucleotide sequence data [32]. Tajima’s D was used to test the null hypothesis of mutation-drift equilibrium and constant population size by considering the difference of nucleotide sequence pair (π) and the number (s) of segregating sites.

After extraction, all individuals showed good profiles (Figure 1(A), Figure 1(B)), indicating that the quality and quantity of DNA extracted from these individuals was sufficient for all PCR amplifications that were to be performed in this study. The PCR amplification of the SCARII marker (Figure 1(C)) and SSR (Figure 1(D)) also showed that all individuals were amplified and that the amount of amplified DNA as well as its quality was good enough to perform all future analyses.

A total of 59 haplotypes were determined, with 15 phaplotypes per population (Table 2). All haplotypes are different from each other with happlotypic diversity

H: number of haplotypes; Hd: haplotypic diversity; S: polymorphic sites; Pi: nucleotide diversity; K: nucleotide differences.

(HD) equal to 1 for each population. The highest number of polymorphic sites (S) was observed for ANA, followed by Mbodiene (188 and 178 polymorphic sites, respectively) and the lowest were noted for the ITACA and Sauvage populations (99 and 111 polymorphic sites, respectively) (Table 2). As for nucleotide diversity (Pi), it ranged from 0.07158 for the ITACA population to 0.10544 for the Mbodiene population. Thus, it was higher for ANA and Mbodiene populations compared to ITACA and Sauvage. The nucleotide difference (K) was higher ANA and Mbodiene populations (42.8381 and 44.18095, respectively) and lower for of ITACA (29.99048) and Sauvage (33.83516) (Table 2).

For the five microsatellites or SSR markers used, a total of 52 different alleles were identified in the five populations. All markers showed a high rate of polymorphism (100%). The average number of alleles per locus (Na) varied from 5.000 to 7.600 with an average value of 6.920 (Table 3, Figure 2). The most polymorphic marker was G12362 with 14 alleles and the least polymorphic was G12330 with six alleles. The most diverse populations were ANA and Mbodiene (Na = 7.600), followed by ITACA and Sauvage (Na = 7.200) and the least genetically diverse population was Richard-Toll (Na = 5.000). The number of effective alleles (Ne) varied from 3.405 to 5.404 (Table 3, Figure 2). The number of effective alleles (Ne) varied from 3.405 (Richard-Toll) to 5.404 (IACA) (Table 3, Figure 2). The Ne was higher for ITACA (5.404), followed by ANA (4.393) and Mbodiene (4.222) and lower for Richard-Toll (3.405). The number of unique allele to a single population (No. private allele) was higher for the ANA (4 alleles) compared to Richard-Toll, ITACA and Sauvage (Figure 2). The No was null for the Mbodiene population. The number of different alleles with a frequency ≥ 5% (Na Freq. ≥ 5%) was not significantly different between populations (Figure 2). The No. LComm alleles (≤50%), number of locally common allele (Freq. ≥ 5%) found in 50% or fewer populations was higher Mbodiene, Sauvage and ITACA and lower for ANA and Richard-Toll (Figure 2). As for the number of alleles with a frequency ≥ 5% (Na Fq), it was higher for ITACA, ANA and Mbodiene and lower for Sauvage and Richard-Toll (Figure 2). The Shannon-Weaver index (I) varied between 1.341 and 1.743 (Figure 2). It was slightly higher for ANA, ITACA and Mbodiene and lower for Richard-Toll and Sauvage.

Na: average number of alleles; Ne: effective number of alleles of PRLI-MS01, UNH104, G12373, G12330, G12362 markers.

The values of expected heterozygosity (He) and observed heterozygosity (Ho) are presented in Table 4. The Ho values ranged from 0.267 for UNH104 (ITACA) to 1.000 for PRLI-MS01 (ANA) and G12373 (all populations). As for He values, they ranged from 0.473 for G12330 (Mbodiene) to 0.885 for G12362 (ITACA). The mean Ho values range from 0.68 to 0.81 for Richard-Toll and ANA, respectively. The mean values of He range from 0.71 for Richard-Toll population to 0.82 for ITACA. The mean He was significantly higher for ITACA compared to the other populations (Table 4, Figure 2).

N: number; Ho: observed heterozygosity; He: expected heterozygosity.

The allelic composition (10 alleles in total) and the distribution of their frequencies for the PRLI-MS01 locus showed that the ANA and Mbodiene populations have the highest number of alleles (9 alleles in total) (Figure 3). They are followed by ITACA, Sauvage and Richard-Toll (7, 5 and 4, respectively). Alleles 2, 3 and 4 have the highest frequencies (20% to 40%) in these populations. Allele 8 is rarer and was only found in the ANA population, followed by alleles 10 and 11 which are present in two populations (ANA/Mbodiene and ITACA/Mbodiene, respectively), but with low frequencies (<10%) (Figure 3).

For the UNH104 marker, 12 alleles were identified in total, two of which (alleles 3 and 4) are present in all populations (Figure 4). The most polymorphic

populations are Sauvage and Mbodiène (9 alleles each), followed by ANA (7 alleles) and ITACA (6 alleles). The least diverse population is the Richard-Toll population with three alleles in total. Allele 11 is present only in the Sauvage population while alleles 2, 5, 7, 12 and 13 are all found in two of the five populations analyzed (Figure 4). Alleles 1 and 8 are present in four populations while alleles 6 and 9 are found in three of the five populations (Figure 4). In terms of frequency, allele 3 has the highest frequency in all populations (40% - 60%), in contrast to the other alleles that are present with a frequency < 20%.

A total of 10 alleles were identified for locus G12373 across the five populations studied (Figure 5). The most polymorphic population for this locus is ITACA (8 alleles), followed by ANA and Sauvage (7 alleles each) and finally

Mbodiene and Richard-Toll (6 and 5 alleles, respectively). Alleles 6, 8 and 9 are present in all populations but allele 9 has the highest frequency. In contrast, alleles 1 and 2 are rarer because they are present in only one population (ITACA and Richard-Toll, respectively). Allele 4 is found in only two populations (Richard-Toll and Mbodiene) while allele 3 is present in three of the five populations (ANA, ITACA and Sauvage). The remaining alleles (5, 7 and 10) are found in four of the five populations, but with low frequency (Figure 5).

Locus G12330 has six alleles, two of which (alleles 1 and 2) are found in all populations, followed by allele 5 which is present in four of the five populations (Figure 6). Alleles 4 and 6 are rarer as they are found in only two of the five populations. Allele 3 is present in three populations (Richard-Toll, Mbodiène and Sauvage). The frequency of allele 1 was high in all populations (40% - 60%) while that of the other alleles 2, 3, 4, 5 and 6 was low (Figure 6). The most diverse population for this locus is Sauvage with a total of five alleles, followed by the Richard-Toll, ANA, ITACA and Mbodiene populations, which have four alleles each (Figure 6).

Locus G12362 has 14 alleles of which 2 (10 and 12) are present in all populations with relatively high frequencies (Figure 7). For this locus, the ANA and ITACA populations are more diverse (11 alleles), followed by Mbodiene and Sauvage (10 alleles). The least diverse population is Richard-Toll with 7 alleles (Figure 7). Allele 14 is rarer because it is present only in the ANA population. It is followed by alleles 2 and 5 found in the ITACA and Sauvage populations with the lowest frequencies. Alleles 3, 4, 9, 10 and 11 are present with high frequencies (>20) in the ANA, Richard-Toll, Sauvage and Mbodiene populations, respectively, while the other alleles are present with relatively low frequency (<20) in all populations.

Total genetic diversity (HT), within-population genetic diversity (HS), between-population genetic diversity (Dst), and coefficient of genetic differentiation (Gst) are shown in Table 5. The HT varied from 0.638 to 0.911 with an average of 0.779 whereas the HS ranged from 0.621 to 0.875 with an average of 0.759. The Dst and Gst are low as evidenced by their values that ranged from 0.012 to 0.036 (average: 0.020) and from 0.019 to 0.049 (average: 0.032), respectively (Table 5). The contribution of within-population genetic diversity (average HS = 0.759) to total genetic diversity (average HT = 0.779) is greater than the contribution of genetic diversity between populations (Dst = 0.020).

HT: total genetic diversity; HS: intra population genetic diversity; Dst: inter population genetic diversity; Gst: gene differentiation coefficient.

The AMOVA based on the five populations showed that the vast majority of genetic diversity (94%) occurs within populations (Table 6), while diversity between populations accounted for only 6% of total genetic variation. Multiple comparisons of Wright’s F statistics from all microsatellite data showed that the index of genetic differentiation Fst was 0.059 (P value = 0.0001) whereas the Fis and Fit were 0.026 and 0.051, respectively.

Although genetic differentiation between populations was significant, genetic variation within populations was maintained. The SSR results confirmed that the observed genotypic frequencies in O. niloticus are consistent with the expected frequencies, indicating that there was no significant deviation from the Hardy-Weinberg hypothesis. They also showed that panmictic interbreeding within populations is relatively high. The estimated migration rate between populations was 4.74, suggesting low gene flow between populations.

Genetic distance was higher for the ITACA/Richard-Toll pairwise populations (Nei genetic distance = 0.312) which are more genetically distant compared to the other populations (Table 7). The lowest genetic distance was noted for the Mbodiene/ANA pairwise populations (Nei genetic distance = 0.126). Comparison of genetic distances also indicated that the ITACA/ANA, Sauvage/ANA pairwise populations have equivalent genetic distances, which are slightly lower than that of Sauvage/ITACA, Mbodiene/Sauvage and Mbodiene/ITACA.

The haplotype network (Figure 8) showed that all haplotypes appear with equivalent frequency. The network did not show a haplotype representing a center of differentiation. Otherwise, there is no haplotype variant shared by two or more individuals belonging to different populations or localities. All haplotypes are unique to different populations. The Tajima neutrality test was obtained from the total number of polymorphic sites (278) and the total number of nucleotide differences K (41.44). The value of D obtained is negative equal to −1.72.

The principal component discriminant analysis (Figure 9) performed on the basis of allelic positions indicates that axes 1 (43.9%) and 2 (32.36%) contribute approximately 92.04% of the total variance. The analysis reveals four population groups (Figure 9). Group 1 is made up of individuals from the ANA, ITACA, Mbodiene and Sauvage populations, while group 2 includes individuals from the Richard-Toll, ITACA, Mbodiene and Sauvage populations. As for group 3, it is formed by individuals from Richard-Toll, ANA, ITACA, Mbodiene and Sauvage

and group 4 is made up of individuals from the Richard-Toll, ANA, ITACA, Mbodiene and Sauvage. These results show that there is heterogeneity in the populations studied.

Molecular analysis by the SCARII marker showed a high nucleotide diversity (polymorphism) ranging from 0.0715 for the ITACA population to 0.105 for the Mbodiene population. These results are in agreement with those of Ndiwa et al. [29] who found nucleotide diversities ranging from 0 to 0.051 for the Senegal River strain inhabiting the Chelaba source. However, they revealed higher polymorphism at nucleotide sites compared to the results of Ndiwa et al. [29]. The SSR markers also showed a high genetic polymorphism with a total of 52 alleles for the five populations studied and a number of alleles per locus ranging from 6 (locus G12330) to 14 (locus G12362). This high number of alleles is sufficient to reduce the standard error committed for an estimation of genetic distances according to Takezaki and Nei [33], who suggested at least 4 alleles per locus. The levels of allele variability observed in this study are similar to those of Moreira et al. [28] who found a high level of polymorphism with a total of 37 alleles ranging from 8 alleles for loci UNH104 and UNH118 to 7 alleles for loci UNH108, UNH222 (G12373) and UNH231. Melo et al. [27] found similar results in the same species with a total of 39 alleles ranging from 6 alleles for loci UNH108 and UNH169 to 11 alleles for locus UNH160. Similarly, when studying 4 strains of Nile tilapia, Rutten et al. [15] obtained a lower number of alleles per locus (5 to 7.5 alleles), which can be explained by the fact that they worked only on farmed populations. On the contrast, Romana-Eguia et al. [16] observed a higher number of alleles (10 to 14.8 alleles per locus on average). The difference in the average number of alleles per locus between these two species with similar reproductive systems could be explained by the fact that the individuals sampled by Romana-Eguia et al. [16] belonged to domestic strains including genetically improved breeding strains and Nile hybrid tilapia strains. The number of alleles in this study is also lower than that observed in wild O. mossambicus strains (ranging from 8.937 to 15.751. These differences may be explained by the plurality of rivers sampled for Mozambique (Limpopo, Umbeluzi, incomati and sabié) that promote high genetic diversity [34].

The estimating of Ho and He provides information on the degree of genetic diversity within populations. In this study, the average He ranged from 0.71 to 0.83 while that of Ho ranged from 0.68 to 0.81. Consistent with these results, Rutten et al. [15] in a genetic characterization study conducted on four strains of Nile tilapia, found a relatively higher expected heterozygosity (He = 0.624 to 0.711) than observed heterozygosity (Ho = 0.612 to 0.696). These results are of the same order of magnitude but significantly lower than those found in the current study, which can be explained by a reduction in genetic variability under hatchery conditions due to a limited number of broodstock. Indeed, Rutten et al. [15] sampled captive populations whereas the individuals analyzed in this study were from both captive and wild populations. This is consistent with the high He and Ho (0.819 to 0.911 and 0.900 to 1.000, respectively) reported in wild populations of O. mozambicus from the Limpopo, Umbeluzi, incomati, and Sabie rivers [34].

The mean He obtained in the present study is higher than the Ho, suggesting a heterozygosity deficiency in the populations analyzed. This reduction in Ho relative to He may be explained by gene flow, allele reading error, or inbreeding. In addition, the observed genotypic frequencies are consistent with expected frequencies, suggesting a lack of significant deviation from Hardy-Weinberg imbalance (Fit = 0.051). The low Fis (0.026) found in this study also suggests the absence of Hardy-Weinberg deviation in the five populations analyzed and indicates that the tilapia populations studied are not affected by inbreeding. In addition, the migration rate between populations was 4.74, suggesting low gene flow between populations. Thus, the high heterozygosity and allelic frequency found in this study may be due to a common origin of some populations (ANA, Richard-Toll, Mbodiene, and Sauvage), which may promote random mating within a Hardy-Weinberg parental population.

Tajima’s D parameter obtained with the neutral test compares the average number of differences per nucleotide pair with the number of segregation sites [32]. The neutral test gave negative Tajima values ranging from −1.508 to −0.540 for the ANA, ITACA, Mbodiene and Sauvage populations. The lowest Tajima D value was noted for the ANA and Mbodiene pairwise populations and the highest for the ITACA and Sauvage. SSR results showed moderate genetic differentiation among populations (Fst = 0.059; p = 0.0001), consistent with the strong contribution of intra-population genetic diversity (HS) to total genetic diversity (HT). In agreement with our results, Simbine et al. [34] revealed moderate differentiation (STF = 0.006, p < 0.001) between O. mossambicus strains from different rivers. Discriminant analysis of principal component indicates two axes that contribute to approximately 92.04% of the total variance. It reveals four groups gathering 80% to 100% of the individuals of the studied populations, which are discriminated by alleles 1, 2, 11, 14 and 8. However, alleles 1, 2 and 11 have a stronger discriminating power, followed by allele 14 and then allele 8. These four groups are very heterogeneous, which can be explained by a common origin and a recent geographical separation, probably by the establishment of aquaculture farms and the isolation of the breeders of these farms. These results are consistent with those obtained with the genetic diversity, which indicated high within-population genetic diversity, but low among-population genetic diversity. The negative Tajima values indicate an excess of rare mutations in the studied populations, a result of recent population expansion [35]. Thus, the low genetic differentiation observed in this study could due to a limited number of initial breeders and their isolation in the hatcheries. They have then expanded in the hatcheries during their years of captivity.