Introduction

Tunisia, situated in the northeast extreme of Africa, is open to the Mediterranean Sea all along its coast. The geographical location in the heart of the Mediterranean has allowed Tunisia to have an important role in the history of this region. The Neolithic Age was present in North Africa through three cultural periods.1 The first, Neolithic with Sudan origins, started about 9000 years before present (YBP) in the extreme south of North Africa, where the current Saharan regions of Algeria are, and is characterized by an ethnic contribution from Sudan. The second, Neolithic with Capsian origins, started about 7000 YBP, represents the continuity of the local Capsian culture (10 000–8000 YBP) and is considered to be the source of ‘proto-Mediterranean’ peoples. The third, Mediterranean Neolithic, embraces the northern coastal regions and very likely represents the continuity of the local Oranean culture (17 000–10 000 YBP), together with influences from some civilizations from the northern side of the Mediterranean.

Historians name ‘Berbers’ the people living in North Africa in the last 6000 years. This population represents, at least in part, the descendants of the aforementioned ancient peoples. As a result, the current Tunisian population is probably composed by an ancient Berber background together with influences from the different civilizations settled in this region in historical times: Phoenicians from Tyre (the present-day Lebanon), who founded the celebrated city of Carthage, Romans, Vandals, Byzantines and, finally, the substantial expansion of Arabs in the seventh century AD. During all these periods, local Berbers were opposed to the invaders, who were unable to dominate all Tunisian regions, with the exception of the Arabs, who settled permanently in Tunisia, as well as in other North African countries.2 In fact, although at first the Arabs met remarkable resistance from the Berbers, they persuaded them to adopt Islam, to learn the Arabic language and to accept intermarriages. Finally, Tunisia received additional, less important contributions, such as that of the Ottoman Turks during the sixteenth century.

In spite of the general adoption of the Muslim Arab culture, some large Berber groups in Algeria and Morocco have kept their Berber language and customs until now and avoided intermarriage. All the same, the original Berber and Arab populations in Tunisia were widely mixed except for few small Berber communities, more or less geographically isolated, such as Takrouna and Jeradou in the center; Douirete-Chenini in the extreme south; and a group living in the ‘Gallala’ region of Djerba island.3, 4 These small Berber groups, often comprising less than 5000 individuals, can be considered to be isolates, supported by several genetic studies showing a remarkable genetic heterogeneity among them.3, 4, 5, 6, 7 Small effective sizes, founder effects and variation of sub-Saharan African traces in their gene pool could be the main factors contributing to current heterogeneity of these Berber groups. Nevertheless, leaving these particular groups aside, the general Tunisian population shows a considerable degree of genetic homogeneity.8

Bearing in mind the historical data quoted, it could be said that the present-day Tunisian population has a uniform Muslim Arab culture, but in genetic terms it represents a mixture mainly composed by Berbers—the autochthonous population—and Arabs, with a relatively small contribution from other surrounding peoples that ruled North Africa in later periods. The aim of this study was to determine to what extent current Tunisians show traces of these mixed origins, taking advantage of the combined information derived from autosomal Alu/STR compound systems. These particular markers have been selected for two main reasons. The first is the widely contrasted informative nature of the Alu polymorphisms9 because of their stability, low mutation rate and known ancestral state. The second is the high degree of qualitative information provided by an STR linked to an Alu marker. Information from haplotype Alu/STR frequencies has been successfully used to estimate fine genetic relationships in the Mediterranean region.10, 11, 12

The main question of this work will be addressed through some specific objectives: (1) determining the degree of heterogeneity of the general Tunisian population through the comparison of two samples from different geographical areas; (2) analyzing the genetic relationships of these samples in a wider North African and Mediterranean context, in particular with those groups historically related to the Tunisian population and (3) deducing more about the mixed origin of the current general Tunisian population, particularly from the analysis of Alu/STR compound systems.

Materials and methods

Blood samples of 268 autochthonous Tunisian individuals were collected: 120 from north-center regions and 148 from the south. All individuals were healthy, unrelated donors, who signed an informed consent approved by the ethical committees of the Universities involved in the study. We have considered the north and the center in a unique north-center sample because preliminary results8 showed a high degree of homogeneity between these samples. The south of Tunisia, particularly the extreme south, is compared for the first time with the rest of the Tunisian population. The subjects from the southern sample are from the extreme southern regions of Gabès, Gbelli and Mednine.

DNA from all individuals has been genotyped for 16 Alu polymorphisms (DM (chromosome 19), HS4.69 (chromosome 6), HS4.32 (chromosome 12), Ya5NBC221 (chromosome 22), Sb19.3 (chromosome 19), HS2.43 (chromosome 1), Sb19.12 (chromosome 19), Yb8NBC120 (chromosome 22), Yb8NBC125 (chromosome 22), PV92 (chromosome 16), FXIIIB (chromosome 1), A25 (chromosome 8), CD4 (chromosome 12), TPA25 (chromosome 8), APOA1 (chromosome 11), ACE (chromosome 17)) and three STRs from the CD4, FXIIIB and DM loci. Supplementary Table 1 includes the GenBank/EMBL accession number of each Alu, together with information about the frequency population ranges that have been established using populations of Figure 1.

Figure 1
figure 1

Geographical location of the samples studied and other populations used for comparisons.

Full size image

Genomic DNA was extracted from blood by standard phenol-chloroform techniques. Alu genotyping was carried out by PCR followed by electrophoresis separation on 2% agarose gels. Three STRs, a pentanucleotide from the CD4 locus, a tetranucleotide from the FXIIIB gene and a trinucleotide from the DM locus, were determined by PCR amplification with fluorescent-labeled primers. PCR products were electrophoresed on ABI PRISM 3700 DNA sequencer (Applied Biosystems, Foster City, CA, USA). GeneScan and GeneMapper 3.0 programs (ABI PRISM; Applied Biosystems) were used to genotype individuals. Technical references on PCR and electrophoresis are extensively explained in our previous works11, 12 for both Alu and Alu/STR combinations. The nomenclature of Alu/STR combinations consists of a number indicating the size in base pairs of the corresponding STR allele, followed by a symbol + (presence of the Alu element) or − (absence of Alu).

Allele frequencies were calculated by direct counting and Hardy–Weinberg equilibrium was checked through an exact test.13 Heterozygosity by population and by locus was estimated according to Nei's formula.14 Maximum likelihood haplotype frequencies were computed using the EM algorithm. The geographical structure of the allele frequency variance was tested by hierarchical analyses of molecular variation using Wright's F statistics from populations clustered according to geographical criteria. These calculations were performed using the GenePop 3.3 (GenePop, Montpellier, France) and Arlequin packages (Arlequin, Berne, Switzerland).15, 16 Population genetic relationships for the Alu and Alu/STR data were also assessed by pairwise FST genetic distances,17 and represented by a multidimensional scaling (MDS) plot from the distance matrix. Levels of genetic admixture were estimated with the Leadmix program (Leadmix, London, UK).18 This Fortran program has been designed to obtain maximum likelihood estimates and 95% confidence intervals of several parameters of ancestral, parental and hybrid populations, together with time estimations since the split of populations. The method and program apply to the case of more than two parental populations contributing to the admixture. The samples can be analyzed for different markers (for example, DNA sequence, microsatellites) and are used to estimate admixture proportions and genetic drift.

For comparative purposes, 22 Mediterranean samples previously tested for the same genetic markers, and a sub-Saharan African sample from the Ivory Coast (95 sampled individuals) were selected.11, 12 The geographical location and sample sizes of the Mediterranean populations used in this study are indicated in Figure 1. The samples included eight continental north Mediterranean groups spanning from the Iberian Peninsula to Turkey; seven samples from the western Mediterranean islands of Majorca, Corsica, Sardinia and Sicily; and, finally, a set of seven North African populations composed by six Berber groups and one Arab-speaking Moroccan sample. Population ranges and pairwise population comparisons have been established using the complete set of 22 samples, but for genetic distance and analysis of molecular variance calculations some samples (Spanish Pas Valley, Sardinians and Siwa Egyptians) were excluded due to their extreme differentiation, discussed in previous reports.11, 12

Results

Alu and Alu/STR genetic variability in north-center and south Tunisian samples

The Alu allele frequencies for the 16 loci examined are shown in Table 1. All loci fit Hardy–Weinberg equilibrium after Bonferroni correction except for Alu HS4.32 and Ya5NBC221 in the southern Tunisian sample.

Table 1 Alu insertion frequencies and heterozygosity values in Tunisia north-center (Tunisia NC) and Tunisia south (Tunisia S)
Full size table

On average, the gene diversity value observed in the north-center (0.360±0.121) Tunisian sample was slightly higher than that of the south (0.348±0.133). Alu frequency comparisons, checked through the exact test of population differentiation, yielded significant differences between our two samples for DM (P=0.035), HS2.43 (P=0.028), PV92 (P=0.006) and APOA1 (P=0.013), and also across the 16 markers considered (P=0.011). As to population variation ranges (Supplementary Table 1) established through data from 22 Mediterranean populations, our populations occupied intermediate positions in all cases except south Tunisia for Alu APOA1, which showed the lowest value. Pairwise population differences revealed a remarkable degree of genetic heterogeneity in the whole Mediterranean context. Across the 16 loci, the exact test indicated that 260 comparisons of 276 were significant. The remaining 16 nonsignificant population comparisons included, among others, the comparisons of the north-center Tunisian sample with both northeast Atlas and Middle Atlas Moroccan Berbers.

Alu/STR haplotype frequencies are shown in Table 2 for the CD4, FXIIIB and DM loci. In all three compound systems, Alu and STR alleles were in linkage disequilibrium. Haplotype diversity values in south Tunisia for CD4 (0.809±0.013) and DM (0.765±0.023) were slightly higher than those reported for the north-center sample (0.836±0.016 and 0.780±0.026, respectively). The opposite trend was observed for the FXIIIB haplotypes (Table 2). In agreement with that, the number of different haplotypes (only those having frequency values >1%) in the north-center sample was slightly higher than that in south Tunisia for the CD4 (13 vs 9 haplotypes, respectively) and DM loci (9 vs 8), but not for F13 B (8 vs 9). Population differentiation was considerable in all three systems; from 276 pairs of comparisons, 75 for CD4, and 122 for both FXIIIB and DM were statistically significant. However, our Tunisian samples did not show any significant differences between them in any of the three compound systems.

Table 2 CD4, FXIIIB and DM Alu/STR haplotype frequencies in Tunisia north-centre (N=120) and Tunisia south (N=147)
Full size table

Population relationships in the Mediterranean

Overall genetic relationships among our Tunisian samples and other Mediterranean populations were assessed through FST-related genetic distances (depicted in MDS plots) for 16 Alu polymorphisms (Figure 2) and for 3 Alu/STR compound systems (Figure 3). In both matrices, distance values were significantly different from zero in more than 85% of the cases. The MDS representation of Alu-based genetic distances underlined the difference between the North African samples and the remaining ones, clustering together continental European Mediterraneans and the samples from Majorca, Corsica and Sicily islands. The genetic heterogeneity inside these two groups, estimated by a nonhierarchical analysis of molecular variance, showed similar values (average FST in North Africa of 1.2%, P<0.001; average FST in European Mediterraneans 1.4%, P<0.001). When we compared the genetic structure between North and South we found that, although presenting moderate values, average FCT between these two groups (2%, P<0.001) was higher than the average FSC inside each group (1.3%, P<0.001).

Figure 2
figure 2

Multidimensional scaling plot based on Reynolds distance matrix, data from 16 Alus in 20 populations. Stress of 0.117.

Full size image
Figure 3
figure 3

Multidimensional scaling plot based on Reynolds distance matrix, data from CD4, F13B and DM Alu/STR haplotypes in 20 populations. Stress of 0.087.

Full size image

The plot (Figure 3) based on Alu/STR compound systems revealed a similar pattern of population relationships with that found for the Alu markers. However, the hierarchical analyses of molecular variance indicated a considerably higher effect of the geographic structure as average FCT between groups (2.6%, P<0.001) accounted for more than four times the average FSC within groups (0.6%, P<0.001). In the second dimension of the plot, the particular position of our Tunisian samples and the Middle Atlas Berbers probably underpins the genetic differentiation between north and south Mediterraneans.

Sub-Saharan African and Berber contribution to the Tunisian genetic background

As stated in previous anthropological studies,11, 19 three CD4 Alu/STR haplotypes (100(+), 85(−) and 115(−)) present a clear sub-Saharan African origin. In fact, they are absent or present in negligible frequencies in Europeans and Asians, whereas they are often present in high frequencies in sub-Saharan African groups and in relatively low frequencies in some of their neighboring populations. These haplotypes are also present in our two Tunisian samples with a remarkable quantitative difference: the total number of different sub-Saharan African haplotypes in the north-center sample was four times higher (7%) than that of the south (1.7%). This difference was statistically significant (Fisher's exact probability=0.002). Moreover, two Mediterranean-specific combinations, particularly in the North Africans11, 12 (the CD4 haplotype 110(−) and the DM haplotype 107(−)) also showed higher frequencies in the north-center sample (2 vs 1% for the CD4 110(−) combination; 1.5 vs 0.4% for the DM 107(−) combination) than in the south.

Sub-Saharan African gene flow in Tunisians was tested through LEADMIX simulations under different parental groups. The only consistent results were those based on Alu/STR haplotypes taking as parental populations a sub-Saharan African sample on one hand, and a sample from continental south Europeans (described in the Materials and methods section) on the other. For both Tunisian samples, the overall sub-Saharan African contribution reached a similar value: 0.398 (−95% CI 0.228; +95% CI 0.617) for north-center Tunisia and 0.392 (−95% CI 0.190; +95% CI 0.632) for south Tunisia.

Discussion

Recent genetic studies dealing with uniparental20, 21 data have emphasized the genetic distinctiveness of the Mediterraneans, the heterogeneity of some populations from this region due to particular demographic histories and the differential sub-Saharan African gene flow received by both southern and northern shores. Concerning autosomal data,11, 12 the usefulness of the combined use of Alu and STR markers to detect fine population relationships has been clearly shown through the study of a wide population set of Mediterranean samples. However, some particular questions still remain to be addressed: the genetic heterogeneity in the Mediterranean seems to be closely related to the differentiation among Berber groups. In that respect, some typical sub-Saharan and Mediterranean Alu/STR combinations have been detected with noticeable frequencies in different Berber communities in Morocco and Algeria, but little is known about their presence in the rest of the present-day North African populations, particularly the Tunisian population having a uniform Arab Muslim culture (100% Arab speakers). Do Berber speakers and Arab speakers in the Maghreb share a common genetic background? Should we be talking about a ‘Berber’ or maybe a ‘North African’ genetic distinctiveness?

The 16 Alu markers and 3 Alu/STR compound systems considered here are the first to be jointly described in Tunisian samples from different geographical locations in the attempt to contribute to a better knowledge of the general Tunisian population, allowing us to address the questions stated above.

North-center and south Tunisians showed similar levels of gene diversity for both Alu (0.360 and 0.348 for north-center and south, respectively) and Alu/STR compound systems (0.814 and 0.813, respectively). These noticeable values are in agreement with those described in other Mediterraneans.11, 12 Genetic differentiation inside Tunisia was significant for Alu loci (P=0.011) but not for Alu/STR compound systems. Such discrepancy of results could be attributed to the effect of the different number of markers compared (16 vs 3). However, the nature of the markers involved, with remarkably different mutation rates, could also provide an explanation for the above observations: previous studies11, 12 have proposed that Alu loci are more suitable to detect ancient relationships, whereas Alu/STR haplotypes perform better in quantifying ancestral genes or gene flow thanks to some population specific combinations. Thus, the presence of Berber and sub-Saharan African-specific combinations in remarkably higher frequencies (10.5%) in north-center Tunisia, as compared with the southern sample (3.1%), suggests a certain degree of genetic heterogeneity also for the Alu/STR data.

The presence of a sub-Saharan component in the gene pool of the Tunisians was first shown by the GM and immunoglobulin Cγ gene polymorphisms.22 The autosomal markers analyzed here have allowed the quantification of sub-Saharan gene flow for the Alu/STR haplotypes. Sub-Saharan African contribution in our samples reached 39%. This value is comparable to, and even slightly higher than, other gene flow estimations previously described11 in several North African populations ranging from 16.8% in Moroccan northeast Atlas Berbers to 37.7% in Mozabite Berbers from Algeria. The presence of noticeable sub-Saharan African traces in present-day Tunisians is in agreement with mtDNA data23 reporting a higher number of sub-Saharan L lineages in Tunisia (48%) as compared with Morocco (25%).

The qualitative information provided by some particular Alu/STR combinations of the CD4 locus, such as 100(+), 85(−) and 115(−), could be another indication of sub-Saharan gene flow. In this case, north-center Tunisia attained a value (7%) considerably higher than that observed in the south of the country (1.7%). These frequencies range from 2.9% in northeast Atlas to 12.3% in Middle Atlas Moroccan Berbers, but they have also been found in Algerian Mozabites (5.8%). The observed fluctuations of sub-Saharan gene flow in North Africa could be related to particular demographic events that may have enhanced the effect of genetic drift on a single locus. Whatever the case, the existence of trans-Saharan African gene flow through the Maghreb is obvious, and has been reported by other genetic studies,12, 23, 24 as well as in archeological and historical records.1

Notwithstanding, it is important to ask whether this sub-Saharan gene flow is relatively recent or more ancient. Our results are compatible with the latter alternative. In fact, as mentioned above, we have found that the presence of three sub-Saharan Africa-specific CD4 Alu/STR combinations is considerably higher in the north-center Tunisian sample than in the one from the south. If the corresponding gene flow occurred in relatively recent times, we should find the opposite trend, because the south of Tunisia would naturally be the first to receive population movements from sub-Saharan countries. Moreover, about 5000 YBP, the immense Sahara desert already had the current severe climate that represents a considerable barrier to human migration, but it was more accessible to human migration25 before, due to a better climatic conditions. All these data considered together suggest that the sub-Saharan component found in Tunisia is rather ancient and could be traced back to the first stage of Neolithic Age (around 9000 YBP), characterized by an ethnic contribution from present-day Sudan.

When other Mediterranean samples were added to the comparisons, the distance matrices of the Alu and Alu/STR systems were positively correlated (P<0.001, Mantel test), emphasizing the consistency of the relationship pattern. In both cases, the first dimension classified populations into North Africans and Europeans, continental and insular samples included. This geographical differentiation was also evidenced by the overall values of the hierarchical analyses of molecular variance. The genetic variance attributable to the North–South differentiation was significant (Alu: FSC within groups=1.3%, FCT between groups=2%, P<0.001; Alu/STR: FSC within groups=0.6%, FCT between groups=2.6%, P<0.001). Our Tunisian samples did not show any close genetic affinity with either the Sicilians or the Turks, two Mediterranean populations that had historical ties with Tunisia.

The Tunisian samples cluster together with the Berber groups from Morocco and Algeria, in agreement with recent works based on other genetic data.26, 27, 28 The close genetic relationship of the two Arab-speaking populations with the Berber-speaking samples could be explained assuming a small number of Arabs coming from the Arabian Peninsula, as compared with that of the autochthonous Berbers, resulting in a weak Arab genetic influence in the current mixed North Africans.

In conclusion, the results discussed here allow us to postulate that the general ancient genetic profile of the native North Africans—the Berbers—is not very different from that of the present-day North African populations, despite some admixture with other peoples, particularly Arabs, during successive historical periods. The populations of the Maghreb seem to share a substantial genetic background, regardless of culture and geography.