Material and methods
Population samples
Samples from one hundred and seventy-two unrelated Macedonians in Skopje (Institute of Blood Transfusion, Tissue Typing Laboratory), the Republic of Macedonia capital, were used for HLA geno-typing and phylogenetic calculations. All were Macedonian language speakers and their ancestors did not belong to a country minority group (detailed above). The origin of all other populations used for comparisons is given in Table 1.
______________________________________________________________
Table 1.
Populations used for the present work
______________________________________________________________
Identification Region and population n (1 References
numbers
______________________________________________________________
1 Macedonians 172 Present study
2 Moroccans (El Jadida) 98 22
3 Berbers (Souss) 98 29
4 Moroccan Jews 94 30
5 Spaniards 176 9
6 Basques 80 9
7 Portuguese 228 15
8 French 179 16
9 Algerians (Algier) 102 8
10 Sardinians 91 16
11 Italians 284 16
12 Jews (Ashkenazi) 80 31
13 Jews (non-Ashkenazi) 80 31
14 Cretans 135 10
15 Greeks (Aegean) 85 2
16 Greeks (Attica) 96 2
17 Greeks (Cyprus) 101 2
18 Lebanese (NS) (2) 59 2
19 Lebanese (KZ) (3) 93 2
20 Iranians 100 32
21 Turks 228 Arnaiz-Villena et al.
(unpublished. results)
22 Armenians 105 16
23 Egyptians (Siwa) 101 2
24 Oromo 83 2
25 Amhara 98 2
26 Fulani 38 2
27 Rimaibe 39 2
28 Mossi 42 2
29 San (Bushmen) 77 16
30 Senegalese 192 16
31 South-African-Blacks 86 16
______________________________________________________________
(1) n = number of individuals analysed for each population;
(2) NS = Niha el Shouff (town);
(3) KZ = Kafar Zubian (town)
______________________________________________________________
HLA genotyping, DNA sequencing and statistics
Generic HLA class I (A and B) and high-resolution HLA class II (DRB1 and DQB1) genotyping was performed using a reverse dot-blot technique with the Automated Innolipa system (Innogenetics N.V., Zwijndrecht, Belgium). HLA-A, -B, -DRB1, and -DQB1 allele DNA sequencing was only done when indirect DNA typing (reverse dot-blot) yielded ambiguous results (11). Statistical analysis was performed with Arlequin v1.1 software kindly provided by Ex-coffier and Slatkin (12). In summary, this program calculated HLA-A, -B, -DRB1 and -DQB1 allele frequencies, Hardy-Weinberg equilibrium and the linkage disequilibrium between two alleles at two different loci. Linkage disequilibrium (D¿; also named LD, see ref. 13) and its level of significance (P)for2À2 comparisons were deter-mined using the formulae of Mattiuz and co-workers (14) and the 11th International Histocompatibility Workshop methodology (13). In addition, the most frequent complete haplotypes were deduced following a methodology used in the 11th International Histocompatibility Workshop: 1) the 2, 3, and 4 HLA loci haplotype frequencies (2, 15, 16); 2) the haplotypes previously described in other populations (2, 16); and 3) haplotypes which were assigned if they appeared in two or more individuals and the alternative haplotype was well defined. In order to compare allelic and haplotype HLA frequencies with other populations, the reference tables used were those of the 11th and 12th International HLA Workshops (2, 16; see also Table 1). Phylogenetic trees (dendrograms) were constructed with the allelic frequencies by applying the Neighbor-Joining (NJ) method (17) with the genetic distances between populations (DA, 18) and using DISPAN software containing the programs GNKDST and TREEVIEW (19, 20). A three-dimensional correspondence analysis and its bidimensional representation was carried out using the VISTA v5.02 computer program (21, http:/forrest.psych.unc.edu). Correspondence analysis comprises a geometric technique that may be used for displaying a global view of the relationships among populations according to HLA (or other) allele frequencies. This methodology is based on the allelic frequency variance among populations (similarly to the classical principal components methodology) and on the display of a statistical projection of the differences.
Results
Characteristic HLA allele frequencies of the Macedonian population compared to other Mediterraneans. The expected and observed allele frequencies for HLA-A, -B, -DRB1 and -DQB1 loci do not significantly differ and the population sample is in Hardy-Weinberg equilibrium. Table 2 shows the HLA allele frequencies found in the Macedonian population. Fourteen different HLA-A and twenty-eight different HLA-B alleles were observed in the Macedonian population. Six HLA-A alleles and seven HLA-B alleles had frequencies higher than 5% (A*01, A*02, A*03, A*11, A*24, A*26, B*07, B*08, B*18, B*35, B*38, B*44 and B*51) and these are characteristic of Mediterranean populations (8–10, 22).
______________________________________________________________
Genetic distances between populations (DA) between Macedonians and other populations (À10 2) obtained by using HLA-DRB1 allele frequencies (see Table 1 for populations identification) HLA-DRB1 (DA)
______________________________________________________________
Cretans 8.38
Italians 10.45
French 14.41
Sardinians 17.66
Spaniards 17.76
Moroccan Jews 17.78
Non-Ashkenazi Jews 17.83
Lebanese (KZ) 20.98
Ashkenazi Jews 21.87
Algerians (Algiers) 22.37
Lebanese (NS) 23.29
Greeks (Attica) 23.69
Moroccans 25.47
Berbers (Souss) 28.50
Spanish-Basques 30.50
Greeks (Cyprus) 33.28
Greeks (Aegean) 37.52
South African Negroids 38.22
Senegalese 41.76
Oromo 43.26
Amhara 51.74
Mossi 53.46
Rimaibe 55.95
San (Bushmen) 57.78
Fulani 61.01
______________________________________________________________
Table 3
______________________________________________________________
With regard to the HLA class II alleles, thirty-one different DRB1 alleles were found and only six had frequencies higher than 5%; DQ allele frequencies reflect the DRB1 locus allele distribution due to the strong linkage disequilibrium between these two loci. Two types of analyses were carried out to compare Macedonian HLA frequencies with other Mediterranean population frequencies: 1) with DRB1 data, which is probably a more informative and discriminating methodology; and 2) with generic (low-resolution) DR-DQ data. These two types of analysis were both performed because some of the populations used for comparison lacked HLA-A and -B data [Berbers (from Souss, Agadir area, Morocco), Jews (Ashkenazi), Jews (Morocco), Jews (non-Ashkenazi), Lebanese (NS and KZ), see Table 1], or high resolution HLA-DQ data [(Greeks (Attica), Greeks (Cyprus), Greeks (Attica-Aegean), see Table 1]], or only generic HLA-DR and -DQ data were available [Portuguese, Turks, Iranians, Armenians and Egyptians, see Table 1]. These partially HLA-typed populations should have been ignored, but they could be analyzed conjointly taking into account only either DRB1 or generic DR and DQ frequencies (Tables 3, 6, Figs 1–3). Analyses using DRB1and DQB1 conjointly were made but are not shown because only a few populations could be used and the results are concordant with the DRB1 analysis. Finally, it should be pointed out that class I generic typing tends to homogenize the comparisons based on DRB1 high-resolution typing (see ref. 22); one class I allele obtained by generic DNA typing may contain several class I alleles, while this is not the case for most DRB1 alleles. Fig. 1 depicts an HLA class II (DRB1) neighbor-joining tree. Populations are grouped into three main branches with high boots-trap values: the first one groups both eastern (including Macedonians, Cretans, Jews, Lebanese) and western Mediterraneans (Europeans and North Africans; Sardinians are also included in the first group). The second branch is formed by African Negroid populations and the third one includes Greek and sub-Saharan populations. This distribution is also confirmed in the correspondence analysis (Fig. 2): the three groups are clearly delimited and a west to east Mediterranean gradient is shown. The Macedonian population shows the closest genetic distance with Cretans (Table 3) and no discontinuity is observed with eastern and western Mediterraneans reflecting the genetic similarity among these populations. It is evidenced that Cretans-Greeks distance is high. These results are con-firmed using DR and DQ generic typings (see Fig. 3 and data not shown) which were used in order to include other Mediterranean populations (Iranians, Armenians, Egyptians and Turks, see Table 1). A DR-DQ neighbour-joining tree (data not shown) maintains the West to East Mediterranean gradient and also the group formed by Greeks and sub-Saharan populations. Turks (old Anatolians), Kurds, Iranians and Armenians have been shown specifically to cluster with the eastern Mediterranean groups (Arnaiz-Villena et al., submitted). On the other hand, genetic distances obtained by using DR-DQ generic typing allele frequencies (data not shown) show that Iranians (1.10À10 »2 ) and Cretans (1.54À10 »2 ) are the two populations closest to the Macedonians followed by the other Mediterranean populations. A discontinuity is found between Berbers (Souss) and Greeks (Attica) (9.59À10 »2 vs. 12.42À10 »2) showing that the latter have a distant relationship with Mediterranean populations as previously described (10, 22) and cluster together with the sub-Saharan populations.
______________________________________________________________
Most frequent HLA-A, -B, -DRB1, and -DQB1 extended haplotypes in the Macedonian population and their possible origin
______________________________________________________________
Haplotypes HF (%) Possible origin
______________________________________________________________
A*01-B*08-DRB1*0301-DQB1*02 a 4.9 Pan-European
A*02-B*18-DRB1*1104-DQB1*0301 b 4.1 Mediterranean
A*02-B*51-DRB1*1601-DQB1*0502 c 3.2 Macedonian
A*03-B*18-DRB1*1601-DQB1*0502 d 2.6 Macedonian
A*01-B*52-DRB1*1502-DQB1*0601 e 1.7 North African-
Mediterranean
A*24-B*18-DRB1*1104-DQB1*0301 f 1.5 Central-South-
Eurasian
A*03-B*18-DRB1*1104-DQB1*0301 g 1.5 Macedonian-
Italian
A*25-B*18-DRB1*1501-DQB1*0602 h 1.2 Iberian-
Macedonian
A*26-B*38-DRB1*0402-DQB1*0302 i 1.2 Macedonian-
Turkish-Jewish
______________________________________________________________
HF: Haplotype frequency. a Also found in Basques (2.4%), Spaniards (3.4%), Britons (2.9%), Danes (3.4%), Cretans (1.1%), Germans (4.8%), Austrians (5.3%) and Yugoslavs (7.7%) (2, 9, 10, 15, 16). b This haplotype has been found in Albanians (3.9%), Italians (2.1%), Yugoslavs (3.5%), Turks (1.1%), Spaniards (1.1%) and Greeks (4.0%) (2, 16 and our own unpublished results). c and d Present only in Macedonians. e Partially (B52-DRB1*1502-DQB1*0601) found in Moroccans (1.5%), Cretans (2.5%), Spaniards (1.1%) and Italians (0.8%) (2, 16, 22). f Haplotype found in Armenians (2.1%) and Italians (0.7%) (2, 6). g Only found in Italians (0.8%) (2, 16). h Haplotype found only in Iberians, Portuguese (1.5%) and Spaniards (0.3%) (15). i Present in Turks (0.9%) and in Jews (our own unpublished results and 33). Other low frequency haplotypes present in Macedonians are also shared with central Europeans (A*03-B*07-DRB1*1501-DQB1*0602, HF: 0.8; A*02-B*13- DRB1*0701-DQB1*02, HF: 0.8; A*02-B*44-DRB1*0701-DQB1*02, HF: 0.6), western Europeans (A*02-B*07-DRB1*1501-DQB1*0602, HF: 0.6), north Africans (A*02-B*07-DRB1*1001-DQB1*0501, HF: 0.6) and Mediterranean-Europeans (A*23-B*44-DRB1*0701-DQB1*02, HF: 0.6) (2, 8–10, 16 and our own unpublished results)
______________________________________________________________
Table 4
______________________________________________________________
HLA-A, -B, -DRB1, and -DQB1 linkage disequilibria in Macedonians
Extended HLA haplotypes were determined in Macedonians and compared with those previously reported in other populations (Table 4). HLA-A-B and DRB1*-DQB1* two-loci linkage disequilibrium data (not shown) show that the most frequent combinations are characteristic of European and Mediterranean (western and eastern) populations (B*18-DRB1*1104, Haplotype Frequency (HF): 9.0; A*02-B*18, HF: 8.1; A*01-B*08, HF: 5.5; B*08-DRB1*0301, HF: 5.2; A*24-B*35, HF: 4.9 and B*07-DRB1*1501, HF: 4.1). The HLA-A-B-DR-DQ extended haplotypes found in the Macedonian population (Table 4) reflect common characteristics with the other ‘‘older’’ Mediterranean background (see footnote to Table 4). These haplotype results are concordant with those obtained by the allele frequency analyses (genetic distances, neighbor-joining trees and correspondence, see above).
Common alleles of Greeks with sub-Saharan Africans
In order to study the possible origin of the Greeks who remain outliers among Mediterraneans (10, 22), specific DRB1 alleles present in Greeks and not present in the other Mediterranean populations were searched in other geographically not very distant populations. Our own data, the 11th and 12th International Histocompatibility Workshops reference panels (2, 16, 23) and other previously described data were used (see Table 1). Table 5 shows the presence of these Greek alleles mainly in sub-Saharan populations from Ethiopia (Amhara, Oromo), Sudan (Nuba) and West Africa (Rimaibe, Fulani, Mossi). Some of these alleles are sporadically present in other populations without any relationships among them (see footnote to Table 5). It may be deduced from these data that sub-Saharans and Greeks share quasi-specific HLA-DRB1 alleles. The neighbor-joining tree (Fig. 1) and the correspondence analyses (Figs 2 and 3) confirm this Greek/sub-Saharan relatedness. The HLA-DRB1 genetic distances between Greeks and other Mediterraneans are shown in Table 6 and also support a sub-Saharan/Greek relatedness; genetic distances with HLA-DR and -DQ generic typings (not shown) give essentially the same results. No relationship of Greeks is seen with the Senegalese and South African Blacks (Bantu and people coming from the Guinea Gulf after the Bantu expansion, respectively (24)), nor with the present day Bushmen (24). Two different types of problem regarding the obtained data are discarded: 1) mistakes in the HLA typings and 2) mistakes in the assignation of these specific alleles (DRB1*0417, *1112, etc, see Table 5). These problems are not likely to exist in the present work because; 1) HLA typings have been made by genetic technologies in three different Greek populations (2, 23) and 2) similar results are obtained when generic typing is used (DR-DQ analysis in Fig. 3; see also ref. 22).
Discussion
Macedonians
Our results show that Macedonians are related to other Mediterraneans and do not show a close relationship with Greeks; however they do with Cretans (Tables 3, 4, Figs 1–3). This supports the theory that Macedonians are one of the most ancient peoples existing in the Balkan peninsula, probably long before arrival of the Mycaenian Greeks (10) about 2000 B.C. Other possible explanation is that they might have shared a genetic background with the Greeks before an hypothetical admixture between Greeks and sub-Saharas might have occurred. The cultural, historical and genetic identity of Macedonians is established according to our results. However, 19th century historians focused all the culture in Greece ignoring all the other Mediterranean cultures present in the area long before the classical Greek one (25).
Greeks are genetically related to sub-Saharans
Much to our surprise, the reason why Greeks did not show a close relatedness with all the other Mediterraneans analyzed (Tables 5, 6 and Figs 1–3) was their genetic relationship with sub-Saharan ethnic groups now residing in Ethiopia, Sudan and West Africa (Burki-na-Fasso). Although some Greek DRB1 alleles are not completely specific of the Greek/sub-Saharan sharing, the list of alleles (Table 5) is self-explanatory. The conclusion is that part of the Greek genetic pool may be sub-Saharan and that the admixture has occurred at an uncertain but ancient time. The origin of the West African Black ethnic groups (Fulani, Mossi and Rimaibe sampled in Burkina-Fasso) is probably Ethiopian (26, 27) (Fig. 4). The Fulani are semi-nomadic hunters and gatherers and one of the few people in the area to use cows’ milk and its by-products to feed themselves and to trade; their facial parameters show a Caucasian admixture. The Rimaibe Blacks have been slaves belonging to the Fulani and have frequently mixed with them (27). The Nuba people are now widespread all over Sudan, but are descendants of the ancient Nubians that ruled Egypt between 8th–7th centuries B.C. (28) and later established their kingdom at Meroe, North Khartoum. Two kinds of Nubians were described in ancient times: Reds and Blacks, probably reflecting the degree of Caucasian admixture. Both the Oromo and Amharic peoples live in the Ethiopian mountains (27). They obviously have in common a genetic back-ground with the west-African groups mentioned above. Linguistic, social, traditional and historical evidence supports an east-to-west migration of peoples through the Sahel (southern Sahara strip), although this is still debated (26, 27). Thus, it is hypothesized that there could have been a migration from southern Sahara which mixed with ancient Greeks to give rise to a part of the present day Greek genetic background. The admixture must have occurred in the Aegean Islands and Athens area at least (Figs 1 and 2). The reason why this admixture is not seen in Crete is unclear but may be related to the influential and strong Minoan empire which hindered foreigners establishment (10). Also, the time when admixture occurred could be after the overthrown of some of the Negroid Egyptian dynasties (Nubian or from other periods) or after undetermined natural catastrophes (i.e.: dryness). Indeed, ancient Greeks believed that their religion and culture came from Egypt (4, 25).
.jpg)
