Moors and Saracens in Europe: estimating the medieval North African male legacy in southern Europe

Abstract

To investigate the male genetic legacy of the Arab rule in southern Europe during medieval times, we focused on specific Northwest African haplogroups and identified evolutionary close STR-defined haplotypes in Iberia, Sicily and the Italian peninsula. Our results point to a higher recent Northwest African contribution in Iberia and Sicily in agreement with historical data. southern Italian regions known to have experienced long-term Arab presence also show an enrichment of Northwest African types. The forensic and genomic implications of these findings are discussed.

Introduction

After the collapse of the Roman Empire in Europe, the Arab dominance across the Mediterranean was one of the most impressive historical events that occurred in this region. Arabs appeared on the southern shores of the Mediterranean in the early seventh century and quickly conquered North Africa. They spread their language and religion to the native Northwest (NW) African Berber populations, which represented the bulk of the Muslim army that later conquered southern Europe.^{1, 2} Referred to either as Moors (in Iberia) or Saracens (in South Italy and Sicily), their arrival in Europe dates to 711 AD, rapidly subduing most of Iberia and Sicily (831 AD). Among European kingdoms their presence was seen as a constant danger, and only by the fifteenth century was the Iberian reconquest completed.³ In the thirteenth century Frederick II destroyed Arab rule in Sicily and between 1221 and 1226 he moved all the Arabs of Sicily to the city of Lucera, north of Apulia.³ Lucera was later destroyed by Charles II (1301) but an Arab community was recorded in Apulia in 1336.³ Guerrilla warfare was still conducted by Arabs in Sicily even after Frederick II's actions.³

So far, Y chromosome studies attempting to estimate the medieval North African (MNA) contribution to southern Europe have focused almost exclusively on the North African haplogroup E3b1b1b-M81, and have only partially taken into consideration the evolutionary relationships among haplotypes.^{4, 5, 6, 7} To generate a more comprehensive view of the genetic legacy of the MNA dominance in Europe, we systematically screened for Y chromosome haplotypes within three NW African specific haplogroups, across multiple southern European populations, and performed additional genotyping to refine the available genetic data. Our results confirm a general correlation between historical and genetic data: Iberia and Sicily are the regions with the highest MNA male legacy.

Materials and methods

Identification of recently introgressed NW African haplotypes

Given the historical indication of a prevalently Berber origin for the Arab groups invading southern Europe,^{2, 3} we focused on NW African specific haplogroups as markers of MNA contribution to this region. Haplogroups E1b1b1b (M81 derived), E1b1b1a-β (M78 derived chromosomes showing the rare DYS439 allele 10) and a subset of J1 (M267 derived) were identified in the literature as being NW Africa specific, together accounting for between 58 and 90% of males in populations from this area, but never above 13% in Europe.^{8, 9, 10, 11} We note that the other lineages present in these populations would also have been brought over to Europe, and any account of the total MNA contribution to present day Europe should take these into consideration.

Given a number of investigated loci n, and a mutation rate μ (estimated using locus specific data as in reference¹²), it is possible to obtain the posterior distribution of the Time to the Most Recent Common Ancestor for any pair of haplotypes differing at k loci, using the approach implemented in reference.¹³ The selected method is based on the infinite alleles model, a reasonable approximation when few mutations are expected to occur, as in the temporal framework evaluated here. So, considering 9 loci and 40 generations (approximately 1200 years ago with a 31-year generation length¹⁴), either 0 or 1 mutational difference is the most likely consequence. Two mutations are only slightly less likely, but overlap with other much more ancient events, for example 80 generations or 2400 years ago. Posterior distributions for more ancient events have probability peaks centred on a higher number of differences, with 0–1 mutations being extremely unlikely (data not shown). Therefore, following this, European Y chromosomes within the three haplogroups identical to, or with one mutational difference from, NW African STR haplotypes were considered compatible with an MNA ancestry. In Iberia and peninsular Italy, they account for 90, 78 and 42% of the E1b1b1b, E1b1b1a-β and J1 chromosomes respectively.

Samples

A NW African database was constructed for haplotype comparisons including more than 400 samples genotyped at nine STR loci (DYS19, DYS389 I–II, DYS390, DYS391, DYS392, DSY393, and the bi-allelic DYS385). The database included 127 Berbers from Tunisia;^{15, 16} 102 South Tunisians;¹⁷ 109 Moroccan Arab and Berber speakers;¹⁸ 50 Moroccan and 52 Tunisians (unpublished data). NW African specific haplogroups were identified by further genotyping of samples that were previously described elsewhere.^{5, 6, 7, 19, 20, 21} We also included a Basque dataset^{22, 23} and two novel Italian samples (Lucera and Veneto; Table 1). Within these populations, all E1b1b1a chromosomes were scored for the DYS439 locus to identify the E1b1b1a-β cluster⁹ and the M267 marker was investigated in those chromosomes previously identified as J^*(xJ2). Alternatively, the DYS458 .2 allele was used to identify the J1 types within J^*(xJ2) chromosomes.²⁴ All the individuals within E1b1b1b, E1b1b1a-β and J1 were also genotyped for the same nine STRs as the NW Africans (DYS19, DYS389 I–II, DYS390, DYS391, DYS392, DYS393 and DYS385). The DYS385 bilocal locus was considered as two different loci, the smaller allele assigned to locus DYS385a and the larger to DYS385b. A previous investigation²⁵ showed that misassignment would influence only a minimal fraction of the haplotypes and so this can be assumed to have a negligible effect on our estimates. A Sicilian population was also included (samples overlapping in references^{26, 27}). Sicilian genotypes were screened for E1b1b1^* and J^*(xJ2) lineages, and did not include DYS439. Within the E1b1b1^* and J^*(xJ2) haplogroups, 8 and 3 chromosomes, respectively, were found close to NW African types. These samples were then made available for further genotyping, to include DYS439, M78, M81 and M267. We note that because of partial sampling across NW Africa, a subset of the European chromosomes with true MNA ancestry could potentially fail to be identified. However, given the general homogeneity observed across NW Africa, the number of populations included, and the large dataset used, we believe that this is unlikely to influence our results.

Table 1. Historically introduced NW African types in Italy and Iberia.

	Sample	n	E1b1b1b	E1b1b1a-βa	J1	Total %
1	Val Badia	34	0.0	0.0	0.0	0.0
2	Veneto	55	1.8	0.0	0.0	1.8
3	Central Emilia	62	0.0	0.0	0.0	0.0
4	Central-Tuscany	41	0.0	0.0	2.4	2.4
5	Tuscany-Latium border	79	0.0	0.0	0.0	0.0
6	North-East Latium	55	1.8	0.0	0.0	1.8
7	Marche	221	0.0	0.5	0.9	1.4
8	South Latium	51	0.0	0.0	0.0	0.0
9	East Campania	84	2.4	1.2	1.2	4.8
10	North-West Apulia	46	4.3	0.0	2.2	6.5
11	Lucera	60	1.7	1.7	0.0	3.3
12	West Calabria	56	0.0	0.0	0.0	0.0
13	South Apulia	71	0.0	0.0	1.4	1.4
	Peninsular Italy	915	0.8	0.3	0.7	1.7
14	Sicily	93	2.2	2.2	3.2	7.5
15	Portugalb	659	5.0	0.3	1.8	7.1
16	Galiciac	292	4.1	0.7	2.1	6.8
17	Cantabriac	161	13.0	3.1	2.5	18.6
18	Basquesd	168	0.6	0.0	0.6	1.2
19	Basquese	43	2.3	0.0	0.0	2.3
20	Catalanse	16	0.0	0.0	0.0	0.0
21	Andalusianse	37	5.4	0.0	0.0	5.4
	Total Spain	717	5.2	1.0	1.5	7.7
	Total Iberia	1376	5.1	0.7	1.7	7.4

Frequencies of E1b1b1b, E1b1b1a-β and J1 chromosomes with 0-1-steps neighbour chromosome within the NW African dataset; the first column refers to the geographic location in Figure 1.

^aE1b1b1a-β chromosomes were identified as M78 derived bearing the DYS439 allele 10⁹.

^bOverlapping with Beleza et al.⁷

^cSamples from Brion et al;⁶ a subset of the J and E samples have been further tested with M81, M78 and DYS439 and used to estimate frequencies. J1 samples have been identified as J samples with the 0.2 DYS458 allelic variant.²²

^dCombined data from Alonso et al;²² Garcia et al.²¹

^eDYS439 and DYS385 were genotyped in the relevant samples from Bosch et al,⁵ except for one Basque sample, not included.

Results and discussion

To address the degree of historical NW African contribution, we used a combined SNP-STR approach. The coalescent times for the three NW African specific haplogroups ranges between 5000 and 24 000 years, spanning a number of historical scenarios each potentially explaining their presence on the Northern Mediterranean shores.^{9, 10} It follows that estimating MNA genetic legacy on the basis of haplogroups' occurrence only would be misleading. To avoid this limitation, we have extended our analysis to include STR data whose high mutation rate allows one to focus on more recent events. We screened more than 2300 South European samples (Figure 1; Table 1) to identify those haplotypes which are evolutionary close to NW African chromosomes. Total frequencies for these chromosomes range between 0 and 19% across southern Europe, the highest being in Cantabria and comprising a sample from the Pas Valley, previously shown to have an extremely high frequency of the North African haplogroup E1b1b1b.⁹ Our estimates of NW African chromosome frequencies were highest in Iberia and Sicily, in accordance with the long-term Arab rule in these two areas.³ The chromosome frequencies in the two samples were not significantly different from each other (Fisher's exact test P=0.83) but were both significantly different from the peninsular Italy sample (P<0.01). An inspection of Table 1 reveals a non-random distribution of MNA types in the Italian peninsula, with at least a twofold increase over the Italian average estimate in three geographically close samples across the southern Apennine mountains (East Campania, Northwest Apulia, Lucera). When pooled together, these three Italian samples displayed a local frequency of 4.7%, significantly different from the North and the rest of South Italy (P<0.01), but not from Iberia and Sicily (P=0.12 and P=0.33, respectively). Arab presence is historically recorded in these areas following Frederick II's relocation of Sicilian Arabs.³ In Iberia, a non-random distribution might also potentially be present, as suggested by our lower estimates in the northeast (Basque region and Catalans), but more samples across the peninsula will be required to properly address this issue. Assuming that a large population in regions such as Iberia, Sicily and Italy was present in the past, the ratio between Y chromosomes with a MNA ancestry and other types will have stayed approximately constant across time. Smaller areas, however, would have been influenced by drift, in the Pas Valley for example. Consistent with historical data,³ no population in Central Europe or the Balkans shows the presence of recently introgressed NW African types^{9, 10, 28} besides a few chromosomes in Albania and Romania.²⁹

Figure 1.

Geographical location of the investigated southern European samples. Numbers are same as in Table 1.

The increasing use of highly structured distributions of Y chromosome types to investigate the ethnic/geographic origin of unknown samples³⁰ gives the identification of regions in Italy enriched with recently introgressed NW African types forensic relevance. We found that more than 56% of the Italian individuals identified here as having a recent NW African do not have a match in a large Italian Y chromosome dataset comprising almost 1200 individuals.³¹ Of these, 31% instead perfectly overlap with types from NW African populations, potentially providing misleading advice to investigators. Such results are also of interest in the light of the expanding business of genealogical services offering Y chromosome analysis to identify an individual's ethnic ancestry. Our results clearly confirm that conclusions based on single chromosomes should be taken very cautiously.³² What are the expected genomic consequences of this historically recent admixture event? Suppose that 40 generations ago there was a 5% male introgression of African DNA into the European gene pool, corresponding to a total contribution of 2.5% of genetic material. Immediately after the admixture event, a fraction of chromosomes within Europe would have African ancestry. Recombination since this event will have substantially reduced the size of the fragments of African ancestry within European haplotypes, and with these parameters we would today expect to see an approximately exponential distribution (measuring size using genetic distance) of fragment sizes, with a mean value of roughly 2.6 cM. Assuming a genome-wide average recombination rate of 1.3 cM/Mb,³³ 2.5% of a typical present day southern European genome would consist on average of 2 Mb regions of African DNA. We therefore believe that signatures of this event would be correctly identified using modern dense genotype data.³⁴ By using northern Italian and Mozabite samples recently genotyped for a large SNP autosomal dataset³⁵ as the best available proxy of Italian and northern African populations, we estimated that about 41.5% of more than 640 000 genotyped SNPs showed an absolute allele frequency difference of at least 10% between the two groups. Such frequency differences (and sometimes even smaller) between cases and controls characterized the vast majority of the inferred disease-causing SNPs in a recent genome-wide investigation.³⁶ In general then, it is critical to take population structure into account so as to avoid false positives in case–control association studies.³⁷ Thus, an understanding of similar historical admixture events is likely to aid researchers conducting such studies.