1Besides the value of a comparative approach between different markers, an additional importance of the island of Jerba is related to its location in North Africa. This article follows a previous study concerning Egypt (Manni et al. 2002) within the wider framework of a project on the genetic patterns of variation in North Africa. The study of the north-eastern part of Africa is likely to provide a deeper understanding of i) remote human history as in the farming/language dispersals hypothesis (Krings et al. 1999; Manni et al. 2002; Lucotte, Mercier 2003) as well as ii) more recent migrations around the Mediterranean basin. We have shown that a) the genetic history of Egypt is very complex with sub-Saharan, Middle Eastern and European contributions to its genetic pool and confirmed that b) there are no genetic signatures of admixtures between NW Africa and southern Europe (Manni et al. 2002). Despite the particular value of North Africa, only a few studies have defined the genetic background prevailing in modern populations from this region. We focus here on Jerba, which on a smaller scale appears to reflect this complexity. There are four distinct human groups which have cohabited on the island for centuries, and who identify themselves as having a different ethnic origin: Arabs, Berbers, Blacks and Jews. Moreover, since Jerba is an island, possible local admixtures are likely to be traced back more successfully because of the absence of geographic continuity with surrounding populations.

2According to North African history, Berbers are believed to have their ancestors among Capsians (7000-5000 B.C.) (Cavalli-Sforza et al. 1994), and several other historical records can be useful in tracing back the demographic history of this island, thus providing a strong historical background for interpretation of the genetic variation of the samples (Mourali 2000). The first Arab settlement on the island occurred in the 7th century A.D. when Islam and the Arabic language were imposed upon the Berbers. It has been proposed that the black Africans were descendants of Sudanese or Nigerians slaves who came in the 18th century A.D. (Tlatli 1967). The origin of the Jews appears to be more complex since different migrations have been proposed, one represented by Jews that migrated to the island after the destruction of the Temple in Jerusalem in the 6th century B.C., a second one during the 15th century A.D. Furthermore, additional migrations were suggested (Auctores). This picture is further complicated by the fact that between the 9th century B.C. and the 19th century A.D., Jerba was a pivotal area in the migration routes of several populations: Phoenicians, Romans, Byzantines, Bedouins, Germans, Sicilians, Turkish and French (Tlatli 1967).

3Currently, Berbers, Blacks and Jews live in different areas of Jerba, whereas only Arabs are distributed over the whole island. The main language spoken by all islander groups is Arabic, even though Berbers also speak a dialect called “Chelha” that belongs to Shilha (a cover term for Berber languages in Marocco and Tunisia). Religious differences may have played a role in subdividing populations; Arabs and Blacks are orthodox Muslims, Berbers are heterodox Muslims while the Jewish group follows Judaism.

4To portray the genetic differences and variability of the present Jerban population, we present the allelic profile of eleven specific Unique Event Polymorphisms (UEPs) meaning that each mutation defines a group of evolutionary related Y-chromosomes (haplogroups). This study is intended to provide a better understanding of the origin of the four ethnic groups living in Jerba.

Materials and methods

Sample

5Genetic analyses were performed on a sample of 127 Y-chromosomes representing 47 Jerban Arabs, 46 Jerban Berbers and 34 Jerban Jews. Since a large majority of Jerban Jews migrated to Israel in 1948 A.D., we sampled such recent immigrants from Jerba to the Israeli towns of Safed, Haifa, Ashdod, Gedera, Ashkelon, Brekhia, Eitan and Tlamim. Only Jews with a Jerban ancestry of at least four generations were included in the study. This criterium also applies the DNA donors belonging to the other ethnic groups. Overall, our sample represents almost 1‰ of the total population of 114,170 individuals. Each Y-chromosome comes from different unrelated families. Fully informed consent was obtained from all participants in this study.

Y-chromosome biallelic polymorphisms typing

6DNA was extracted from whole blood by means of a standard phenol/chlorophorm protocol. Subsequent purified DNA was analyzed in order to identify the 11 possible Y-chromosome haplogroups (Hgs) defined by 10 UEPs (Unique Event Polymorphisms) (Y-Chromosome Consortium 2002; Jobling, Tyler-Smith 2003). The markers used are the as follows: SRY-2627 (Veitia et al. 1997), SRY-10831a (Kwok et al. 1996), SRY-4064 (Whitfield et al. 1995), 92R7 (Mathias et al. 1994), Tat (Zerjal et al. 1997), YAP (Hammer, Horai 1995), M2 (Seielstad et al. 1994), LLY22g (E. Righetti, C. Tyler-Smith—unpublished), M9 (Underhill et al. 1997) and 12f2q (Casanova et al. 1985). The typing was carried out as described in Rosser et al. (2000). The deep-rooting markers SRY-10831a, M9, 92R7, YAP and 12f2q were typed in all samples and, in some cases, the remaining markers were typed hierarchically; e.g. SRY-4064 and M2 only on YAP⁺ chromosomes, LLY22g and Tat only on those chromosomes which are M9 derived and 92R7 ancestral, and SRY-2627 only on those chromosomes which are 92R7 derived.

Measures of genetic diversity and differentiation

7Two genetic diversity estimators were computed by means of the Arlequin version 2.1 software (Schneider et al. 2000). Intra-population diversity was measured through the gene diversity index (Nei 1987) as shown in table I. Inter-population diversity was calculated through conventional F-statistics (Wright 1965). Two measures of differentiation are provided: FST accounts for genetic divergence between samples (table II); FCT accounts for genetic divergence between different groups of samples (table III). A group of reference samples (table III, IV) was designated according to geographic criteria as follows: NW-Africa (North African composite sample [includes different Berber and Wahabi populations], Morrocan Arabs and Berbers, Algerians); Middle East (Egyptians, Druses, Palestinians, Lebanese, Saudi Arabians and Syrians); SE-Europe (Turks, Greeks, Georgians); western Europe (Spanish, French, Italian). A matrix of FCT values concerning the four Jerban samples and these groups was computed to identify the putative mother population of each Jerban sample (table III).

Multivariate analysis

8Principal component analysis (PCA) was applied to the data to graphically identify possible patterns of genetic similarity between the four Jerban ethnic groups and a reference sample of 19 populations around the Mediterranean basin (table IV). The PCA method involves a mathematical procedure that transforms a number of (possibly) correlated variables into a (smaller) number of uncorrelated variables called principal components. The first principal component accounts for as much of the variability in the data as possible, and each succeeding component accounts for as much of the remaining variability as possible.

9Multidimensional relationships between items can be seen in a bivariate plot (if two or three axes are plotted one against each other). The analysis was performed with the Excel applet GenAlEx by Peakall and Smouse (2001) freely available at the following Internet address: http://www.anu.edu.au/​BoZo/​genAlEx

Results

Y-chromosome biallelic profiles

10Allelic profiles of the 11 Y-chromosome lineages in the total sample of 165 Jerban Y-chromosomes are shown in table I. The first result concerns the total homogeneity of the Jewish group where all the 34 typed individuals share the same haplogroup Hg J that is also present, at a very low frequency (4.3%), in the Arab group. This haplogroup presents its highest frequencies in the Fertile Crescent and has been suggested to be a genetic signature of migrations from the Middle East in Neolithic and post-Neolithic times (Semino et al. 1996, 2004; Rosser et al. 2000; Quintana-Murci et al. 2001; Arredi et al. 2004; Di Giacomo et al. 2004).

11The only haplogroup spread between the remaining two ethnic groups is the 21. High frequencies of this lineage are recorded in the Jerban Berbers (80.4%) and Arabs (93.6%). This haplogroup shows its highest frequencies in North Africa. Hg P(xR1a, R1b3f) lineage has been found predominantly in Europe, showing increasing frequencies from the Middle East to north-western Europe (Semino et al. 1996; Rosser et al. 2000) and was found at a very low frequency in Berbers (2.2%). Hgs B, C and F (xK, J) is found at significant frequency in Berbers (13.2%) but are absent in Arabs. However, little information can be deduced from this latter lineage since it is present in both Europe and Asia and does not show any informative geographic variation in the Eurasian landscape. We recorded the presence of Hg K (xN, P) in a single individual in the Arab sample. Given the little information conveyed by a single Y-chromosome we will not discuss it further in this context. Hg A and Hg E3a are found at high frequency in sub-Saharan African lineages and they have been suggested to be genetic signatures of Khoisan and Bantu language families, respectively. They have never been recorded outside Africa (Hammer et al. 1998; Rosser et al. 2000; Jobling, Tyler-Smith 2003) and their presence is unusual in North Africa (Bosch et al. 1999); they were found to be absent in all three groups.

Measures of genetic diversity

12Nei’s measures of genetic diversity (table I) mirror the homogeneity of Jewish population and the very low variability of the Arab sample. Berbers show an intermediate level of genetic diversity with a fairly low level of variability. FST distance measures (table II) give an overall view of the genetic difference existing among the four Jerban samples. The Jewish group is the most divergent from all the others (Jewish-Berbers FST = 0.827; Jewish-Arabs FST = 0.898) while Arabs and Berbers are found very similar (FST = 0.099).

13Based on a list of reference populations (table IV), we computed FCT distances (table III) between our four samples and four groups of populations designated according to their geographic location, that is N Africa, Middle East, SE Europe and western Europe (see table IV). Results suggest a north African origin for the Arab and Berber samples. Concerning the Jewish sample, the Middle Eastern origins can be inferred and are much more obvious when Egypt is withdrawn from the north African group of samples (FCT = 2.7). Nevertheless, since we had no intermediate populations between Egypt and Algeria, we decided to include Egypt with the Middle Eastern group since the sample comes from the Cairo area.

14Voir document Annexe n°1

Principal component analysis

15In the Principal Component Analysis (PCA) plot of figure 1 the Jerban Berbers overlap the north African composite sample [that includes Berber and Wahabi populations] and the Moroccan Berber samples, leading to the conclusion that these three groups could be considered as part of the same population. Similarly, the Jerban Arabs fall close to Berber populations and to the Moroccan Arab sample. This close correspondence between some of the reference populations and our samples is lost for the Jerban Jewish samples. The Jerban Jews appear to cluster together with other Middle Eastern populations (Palestine, Lebanon, Syria), nevertheless they are plotted far apart from them. This feature could be explained by the fact that Jerban Jews encountered a significant bottleneck that is mirrored by allelic profiles and diversity values.

16Voir document Annexe n°2

17The topology of samples in the PCA plot suggests a Gutman (horseshoe) effect, meaning that the second axis is curved and twisted relative to the first, and is likely not to represent a true secondary gradient. It can be noted that, if we only sampled a small enough section of the gradient, the data might be linear enough to allow the use of PCA. The horseshoe effect happens because genetic variables, like most environmental variables, can be monotonically related to underlying factors and to each other as is expected in close populations exchanging migrants. To provide an alternative representation, we visualize only the first principal component at the bottom of figure 1. The third axis still describes 11% of the total genetic variance, but we were unable to depict any pattern in it. A large proportion of studied populations cluster together with two major exceptions represented by Basques and Jerban Jews, who occupy a very lateral position on the axis. This axis just confirms the deviant allele profile of the latter sample; therefore, to provide a more readable plot, we decided to not display the third PCA axis.

Discussion

Diversity and differentiation of Jerban samples

18The present-day Jerban population is composed of four different ethnic groups that have been cohabiting on the island for centuries. Ethnological and anthropological studies suggest that each group maintained and developed a specific identity (Benoît, Kossovitch 1930). Furthermore, a number of historical records, as well as local traditions, provide information on the different geographic origin for these communities (Stablo 1941; Tlatli 1962, 1967). Our results clearly point to the very low level of diversity of both the Jewish and the Arab population (table I) and reflects a severe bottleneck together with the effects of drift. This conclusions can apply also to the Berber sample, which exhibits a low diversity index which may be explained by a more rapid genetic drift due to longer isolation on the island and/or by any patrilinear social structure as recently shown by Chaix et al. (2005). Since our results contradict those of Yacoubi Loveslati et al. (2001) based on restriction polymorphisms of IGHG genes that are known to be under selective pressure, future research will probably give more definite answers concerning this point.

19These findings are confirmed by results from a genealogical study on marriages that provides clear evidence of high levels of endogamy in the island in recent times (Mourali 2000). According to such genealogical results, the Jewish group was found to be 100% endogamous (in the island) and mixed Black-Berber marriages were not recorded. Other possible mixed marriages were recorded at very low frequencies. Such a picture suggests that gene flow from the Berber to the Black group (and vice-versa) is possible only through preliminary mixed unions with Arabs. The genetic profiles of the Jewish group from Jerba is very different from the ones of Berbers and Arabs. Further, migration levels are very low. Overall, the results suggest that each community may have been founded by a different migration event to the island.

Genetic profiles of the Jerban samples and their geographic origin

20The genetic profiles of the four ethnic groups provide further evidence of their separate origin; in fact some haplogroups are specific of each sample: (i) Hg P (xR1a, R1b3f) was found at a very low frequency only in the Berber group, (ii) Hg J is specific to Jews while (iii) Hg R1a is recorded only among Berbers. The most widespread Y-chromosome lineage is Hg E (xE3a), absent only in the Jewish sample. These specificities highlight the genetic differences found in the Jerban samples and point to different ancestral populations as is also shown by FCT distances (table III). These differences are presented in details in the next subsections. We would like to note that we also sampled and typed the Black minority for the purposes of an early version of this article but, according to a new nomenclature (Cruciani et al. 2002), it was shown that the two different haplogroups “R1-M173*” and “R1-M269” (usually considered as a single one Hg P [xR1a, R1b3f]) are respectively distributed in Europe and sub-Saharan Africa. When our samples were typed, only the previous nomenclature and typing method were available; therefore we decided to widhraw this sample from analyses.

The Jewish population

21The genetic profile of the Jewish sample is related to populations in the Middle East where Hg E3a has its highest frequencies. Moreover, the Middle Eastern origin of the Jewish group is confirmed by preliminary results on Y-chromosome microsatellites of 15 individuals sampled in Tlamim; 13 of them carried for the Cohen Modal Haplotype (CMH). This haplotype is frequent in the Jewish diaspora and it has been suggested that it might be particularly frequent in families which trace patrilineal ancestry to the Jewish priesthood (Thomas et al. 2000). While the CMH is not specific of Jews, being widespread in several Middle Eastern populations (Brinkmann et al. 1999, Zoosman-Diskin 2000), such a datum seems to support two different scenarios for the Jewish colonization of Jerba: a) a founding group that migrated to Jerba in an early dispersal of Jewish population from the Middle East or b) a recent migration of families of ancient Jewish ancestry. The first scenario seems unrealistic because microsatellite markers are rapidly evolving, therefore we would expect more variable microsatellite haplotypes in the case of an ancient founding event, unless some bottlenecks occurred in the male lineage of Jerban Jews.

Arab and Berber populations

22The Arab and Berber samples are almost totally characterized by Hg E (xE3a), which is present at its highest frequencies in North Africa, frequencies that decrease in a gradient shape towards the south. Hg E (xE3a) is also observed at moderate frequencies in the northern Mediterranean basin, thus relating Jerban Arabs and Berbers to North African populations as confirmed by FCT values (table III).

Distances between Jerban populations

23The population of the island of Jerba has attracted the interest of several scholars, therefore it is possible to compare our results on Y-chromosome UEP with a wide range of studies based on different markers. To this end, we have compared our FST distance matrix (table II) with other distance matrices accounting for the variability of Y-chromosome microsatellites (Khodjet El Khil et al. 2001), GM hallotypes (Yacoubi Loveslati et al. 2001), and ABO blood groups (Mourali 2000). Our Jewish sample could not be compared because it is absent in all previous published studies. According to FST distances based on UEPs Y-chromosome profiles, we show that there is almost no differentiation between the Berber and Arab samples. This data is strongly confirmed by all the systems with the exception of ABO, a result that is not surprising since this system is unlikely to portray genetic differences at fine geographic scale. Moreover, in a previous study based on mitochondrial DNA (Fadhlaoui-Zid et al. 2004), no significant differences between Arabs and Berbers were found, thus suggesting that maternal and paternal lineages in Jerba may lead to a similar demographic scenario.

24This paper shows that an accurate portrayal of the inter-group Y-chromosome variability of three ethnic groups living in Jerba was possible through accurate sampling of the population of the island. When an ethnological and historical approach in the definition of populations is lacking, there is a risk of unrepresentative results, as shown by the position of a grouped sample of Jerban inhabitants on the Principal Component Analysis (PCA) of figure 1. Such a sample falls very close to Algeria and Egypt, thus mirroring the geographic position of the island. Actually this topology just reflects a “centre of gravity” of the genetic variability of the four ethnic groups, which are shown to have very different geographic origins.

Acknowledgements

25We wish to thank all the DNA donors who made this study possible. Indeed we are indebted to Soufia Mourali for providing comments on the manuscript and to two anonymous reviewers for their useful suggestions. This work was made possible through French-Tunisian cooperation and financial support from the INSERM.