MtDNA and Y-chromosome Variation in Kurdish Groups

Introduction

Kurds are an Indo-European speaking group that inhabit the highlands in the border area of Turkey, Syria, Iran and Iraq. This region lies astride the Zagros Mountains of Iran and the eastern extension of the Taurus Mountains in Turkey, and extends in the south across the Mesopotamian plain to include the upper reaches of the Tigris and Euphrates rivers (Figure 1). Several languages and/or dialects of Kurdish are recognized, and are classified as belonging to the northwestern branch of Iranian languages (Ethnologue, 2000).

The first mention of the Kurds in historical records is in cuneiform writings from the Sumerians from around 3,000 B.C. (Wixman, 1984), who talked of the “land of the Karda”. In the 7th century A.D., the Arabs conquered the area and in time converted everyone in it - including the Kurds - to Islam. In the centuries that followed, the Kurds withstood invasions from Central Asia which brought the Turkic peoples as far west as Asia Minor (now Turkey), probably because they occupied an area that was too difficult for outsiders to reach. As the Ottoman Empire rose to power in the 13th through to the 15th centuries, it extended its territory to what is roughly now the border between Iran and Iraq. From then until World War I, the area inhabited by the Kurds was under the dominion of the Ottomans and Persians.

There have been several migration events involving Kurds. An extensive resettlement of Kurds from Turkey and Iran into the Caucasus began during the late 19^th century and continued through World War I (Wixman, 1984). Since that time Kurds have formed compact settlements in Georgia and Armenia, and have kept their cultural and linguistic identity. At the time of the 1979 census there were 51,000 Kurds in Armenia and 26,000 in Georgia (USSR Census 1979). Kurds from Central Asia (Turkmenistan and Kazakhstan) speak the Kurmanji language (Ethnologue, 2000). They are also descendants of Kurds from Eastern Anatolia and Iran.

Kurmanji is a North-West Iranian language spoken in Southeast Anatolia in Turkey. This language belongs to the Kurdish linguistic branch, along with the Behdini, Herki, Kurdi, Shikaki and Surchi languages, all of which are spoken in Northern Iraq (Ethnologue, 2000). Zaza (or Dymli) is also a northwest Iranian language spoken in Southeast Anatolia, northwest of the Kurdish speaking regions. Although first thought to be a Kurdish dialect, since the beginning of the 20th century Zazaki has been accepted as a language of its own (Paul, 1998). Virtually nothing is known about the origin of Zazaki-speaking people, as they do not possess a written language and therefore lack a recorded history. Based on some structural similarities of the Zazaki language with the Talyshi, Gilani and Mazandarani languages spoken in northern Iran, linguists have hypothesized that their origin lies in the mountainous region of the Southern Caspian Sea area (MacKenzie, 1962).

Only a few genetic studies have been carried out on Kurdish groups. Previous genetic studies of classical markers (Cavalli-Sforza et al. 1994) indicated an overall genetic similarity of Kurds with other Middle Eastern populations. Comas et al. (2000) studied mtDNA HV1 sequence variability among Kurmanji-speaking Kurds living in Georgia (Caucasus), and found close European affinities for Kurdish mtDNA lineages. Richards et al. (2000) studied mtDNA HV1 sequence variability among 53 Kurds from Eastern Turkey and found that some mtDNA haplotypes found in Kurdish samples presumably originated in Europe, and were associated with back-migrations from Europe to the Near East. Wells et al. (2001) investigated Y chromosome SNP haplogroup distributions among Central Asian groups, including a group of Kurmanji-speaking Kurds living in Turkmenistan, but no specific conclusions were made regarding the history of the Kurdish group. Nebel et al. (2001) studied Y chromosome SNP and short tandem repeat (STR) loci among different groups from the Middle East, including a group of 95 Kurds from northern Iraq, and found close affinities for the Kurdish group to other Middle Eastern groups. Finally, Quintana-Murci et al. (2004) studied 20 Kurds from Western Iran and 32 Kurds from Turkmenistan, among other groups from Iran, Pakistan and Central Asia, but did not come to any specific conclusions concerning the Kurds.

None of these previous studies were focused specifically on Kurdish groups and their origins or relationships. Rather, these previous studies targeted a larger geographic scale and/or more general questions concerning the genetic relationships of different populations. In order to investigate the genetic relationships between different Kurdish groups, we present here data on mtDNA and Y chromosomal variation in 139 individuals from Zazaki and Kurmanji speaking groups (from Turkey), and from Kurds living in Georgia (Caucasus) who are also Kurmanji speakers.

The new genetic data presented in this study were combined with previously-published data on mtDNA HV1 sequence variation in 29 Kurds living in Georgia (Comas et al. 2000) and with other previously-published data from Kurds from Iran, Turkmenistan and Turkey (Richards et al. 2000; Quintana-Murci et al. 2004) to address the following questions: (1) how closely related genetically are the Kurdish groups living in different areas and/or speaking different languages?; (2) what is the genetic relationship between Indo-European speaking Kurdish groups and other West Asian Indo-European and non-Indo-European speaking populations?; (3) can a source population for the Kurdish group from Georgia be identified?; (4) was there any subsequent genetic exchange between the Kurdish group from Georgia and the surrounding Georgian population?

Materials and Methods

Results

MtDNA HV1 Sequence Variability

A total of 377 bp of the mtDNA HV1 region, comprising nucleotide positions 16024 to 16400 (Anderson et al. 1981), were determined for 78 individuals from the Zazaki speakers and Kurmanji speakers from Turkey (hereafter referred to as Zazaki-T and Kurmanji-T, respectively). MtDNA HV1 sequence data for the Kurmanji speakers from Georgia (Kurmanji-G) and for Kurds from Iran (Kurds-I), Turkmenistan (Kurds-Tm), and Turkey (Kurds-T) were previously published (Comas et al. 2000; Richards et al. 2000; Quintana-Murci et al. 2004). As a check on the accuracy of the HV1 sequences, we used the network method to search for so-called “phantom” mutations (Bandelt et al. 2002). No such artifacts were found in the Kurdish HV1 sequences (analysis not shown). The sequences have been deposited in the HVRbase database (http://www.HVRbase.de) and are also available upon request from the authors.

Parameters summarizing some characteristics of the mtDNA HV1 sequence variability in Kurdish groups are presented in Table 1. The haplotype diversity and mean number of pairwise differences (MPD) were both lower in the Kurmanji-G than in the other samples, but not significantly so. Tajima's D was negative and significantly different from zero in five of the six Kurdish groups, suggesting population expansion (Table 1).

Table 1. MtDNA HV1 sequence variability among Kurdish populations

Population	N	no. of haplotypes	Haplotype diversity and SE	MPD	Tajima's D	Source
*P < 0.05; **P < 0.01
Kurmanji_T	51	43	0.988+/−0.008	6.95	−2.05**	present study
Zazaki_T	27	23	0.986+/−0.015	4.95	−1.81*	present study
Kurmanji_G	29	22	0.958+/−0.028	4.29	−2.15**	Comas et al. 2000
Kurds_I	20	19	0.995+/−0.018	6.13	−1.57*	Quintana-Murci et al. 2004
Kurds_Tm	32	20	0.970+/−0.015	6.55	−1.20	Quintana-Murci et al. 2004
Kurds_T	53	40	0.979+/−0.012	5.06	−1.94**	Richards et al. 2000

The mean number of pairwise nucleotide differences (Table 1) are in the range of values observed in the Caucasus (4.40–5.87; Nasidze et al. 2004), but lower for the Zazaki-T and Kurmanji-G than in West Asia (5.38 - 7.08; Nasidze & Stoneking, 2001). Haplotype diversity in these groups (Table 1) falls within the range observed in the Caucasus and West Asia (0.948–0.999; Nasidze et al. 2004).

Pairwise F_st comparisons between different Kurdish groups showed that the Kurmanji-T and Kurds-Tm are most different from the other Kurdish groups, whereas the other four Kurdish groups all show lower F_st values with each other (Table 2).

Table 2. Pairwise Fst values between three Kurdish groups, and Caucausian, European, and West and Central Asian populations. Below diagonal - paiwise F_st values based on Y-SNP haplogroups; above diagonal - pairwise F_st values based on mtDNA HVI sequences

	Kurmanji-T	Zazaki-T	Kurmanji_G	Kurds_I	Kurds_T	Kurds_Tm	Caucasus	West Europe	East Europe	Centr. Asia	West Asia
*p < 0.05; **p < 0.01
Kurmanji-T		0.010	0.017	0.013	0.008	0.019*	0.032*	0.011	0.013	0.076**	0.012
Zazaki-T	0.014		0.001	0.004	0.001	0.015	0.032*	0.015	0.012	0.078**	0.008
Kurmanji_G	0.091**	0.228**		0.001	0.002	0.018*	0.030*	0.019*	0.020*	0.070**	0.018
Kurds_I	N/A	N/A	N/A		0.002	0.011	0.030*	0.029*	0.030*	0.059**	0.009
Kurds_T	N/A	N/A	N/A	N/A		0.012	0.021*	0.010	0.013	0.053**	0.010
Kurds_Tm	0.073*	0.137**	0.189**	N/A	N/A		0.034*	0.020*	0.025*	0.037**	0.012
Caucasus	0.087	0.145	0.217	N/A	N/A	0.127		0.023*	0.026*	0.021*	0.020*
West Europe	0.221	0.224	0.362	N/A	N/A	0.218	0.278*		0.009	0.092**	0.025*
East Europe	0.137	0.081	0.333	N/A	N/A	0.234	0.259*	0.364**		0.076**	0.025*
Centr. Asia	0.094**	0.120	0.205	N/A	N/A	0.160	0.225*	0.376**	0.153		0.056**
West Asia	0.019**	0.089	0.144	N/A	N/A	0.074**	0.251*	0.227*	0.169*	0.117

Overall, all six Kurdish groups show basically the same pattern: the smallest F_st values are for West Asia, next smallest are Europe [combining West and East European groups, as they do not differ significantly with respect to pairwise F_st values with Kurdish groups (t-test; data not shown)], then the Caucasus, and then Central Asia (Table 2).

An MDS plot (Figure 2a) based on the pairwise F_st values illustrates these patterns. Most of the Kurdish groups are located near each other, whereas the Kurmanji-T and Kurds-Tm are further away. Moreover, the Kurdish groups all tend to fall within West Asian groups and near European groups, with the Caucasus and Central Asian groups being further away.

Y-Chromosome STRs

Because of the unusually high frequency of haplogroups P1(M124) and J2*(M172) in the Kurmanji-G, we typed nine Y-STR loci in individuals with these Y-SNP haplogroups. Since haplogroup I*(M170) was found at relatively high frequencies in the Zazaki-T and Kurmanji-T, we also typed the same set of Y-STR loci in individuals with this Y-SNP haplogroup (Y-STR haplotypes are available from the authors upon request). We compared the results with the same set of loci on the same Y-SNP background, typed in other groups from the Caucasus, Iran and Turkey (Nasidze et al. 2003b). The average variance in number of repeats per locus on the background of haplogroup P1(M124) is reduced in Kurmanji-G compared with Kurmanji-T and groups from the Caucasus, but higher than in Iran, although the latter is based on just two individuals (Table 4). The average variance in number of repeats per locus on the background of haplogroup J2*(M172) is also reduced in Kurmanji-G compared with Kurmanji-T, and roughly equal to those for their geographic neighbours, groups from the Caucasus and Iran. The average variance in number of repeats per locus on the background of haplogroup I*(M170) is roughly equal in Zazaki-T and Kurmanji-T, as well as groups from the Caucasus, Iran and Turkey (Table 4).

Table 4. Number of Y-STR haplotypes (n) and average variance in number of repeats per locus (avnr) on the background of Y-SNP haplogroups P1(M124), J2* (M172) and I* (M170) observed among Kurdish, Caucasian, Iranian and Turkish individuals. Data from Caucasus, Iran and Turkey are from Nasidze et al. (2003) (2004)

population	haplogroups
	P1		J2*		I*
	n	avnr	n	avnr	n	avnr
Kurmanji-G	10	0.99	8	0.72	0	-
Kurmanji-T	4	1.24	15	1.23	9	1.16
Zazaki-T	0	-	0	-	19	1.26
Caucasus	8	1.27	77	0.75	42	1.15
Iran	2	0.61	17	0.72	32	1.04
Turkey	0	-	1	-	10	1.15

A median network (Figure 3) based on Y chromosome STR haplotypes on the background of haplogroup P1(M124) shows close relationships between haplotypes found in the Kurmanji-G and Kurmanji-T, although there was no haplotype sharing between these two groups. The median network shows a clear distinction of Kurdish Y-STR haplotypes and haplotypes found in the Caucasus and Iran. Furthermore, clustering of the latter is much more compact, suggesting that haplogroup P1(M124) may have arisen in a population ancestral to present-day Kurds. For the median network of Y-STR haplotypes on the background of haplogroup J2*(M172), individuals from different groups are intermixed (data not shown). The median network based on Y-STR haplotypes for haplogroup I*(M170) similarly did not reveal any differences between the Zazaki and Kurmanji and their geographic neighbours (data not shown).

Discussion

The Kurdish groups in this study come from different regions and speak different languages: the Kurmanji-T and Zazaki-T reside in geographical proximity, but speak different languages, whereas the Kurmanji-G and Kurds-Tm speak the Kurmanji language (Ethnologue, 2000) but are located in different geographic areas; the original publications from which the data on the Kurds-I and Kurds-T were obtained do not specify the languages spoken by these groups. How closely related genetically are the Kurdish groups living in different areas and/or speaking different languages? MtDNA data indicates close relationships among four Kurdish groups (Zazaki-T, Kurmanji-G, Kurds-I, and Kurds-T), while the Kurmanji-T and Kurds-Tm differ from these and from each other (Figure 2a). No clear pattern in terms of either geography or language is evident from our results (i.e., two Kurdish groups from Turkey are different from each other, and Kurmanji speakers from Turkey are not similar to Kurmanji speakers from Georgia). However, the Y chromosome shows somewhat different patterns, indicating some effect of geography. The Kurmanji-T and Zazaki-T, who are geographic neighbours in Turkey, are closer to each other compared with the Kurmanji-G and Kurds-Tm (Figure 2b). The latter two groups also showed significantly reduced haplogroup diversity (Table 3), suggesting a bottleneck effect in these groups.

The Zazaki (or Dymli) language used to be classified as a Kurdish dialect, but is now considered to be a separate Iranian language (Paul, 1998) It has been suggested, on linguistic grounds, that the origins of the Zazaki people may be in northern Iran (MacKenzie, 1962). However, the Zaza people self-define themselves as Kurds (Donald Stilo, personal communication), so it was of interest to examine the genetic relationships of the Zazaki speakers with Kurdish-speaking groups. Genetically, it turned out that Zazaki speakers overall group quite closely with other Kurdish groups, and with their geographic neighbours from the South Caucasus for mtDNA, and with Kurmanji-T for the Y-SNP haplogroups. The previous hypothesis of a close relationship of the Zaza people to populations from northern Iran (MacKenzie, 1962) therefore does not gain genetic support, although the genetic evidence of course does not preclude a northern Iranian origin for the Zazaki language itself.

Kurdish languages belong to the Iranian branch of the Indo-European language family. What is the genetic relationship between Indo-European speaking Kurdish groups and other West Asian Indo-European and non-Indo-European speaking groups? For both mtDNA and the Y-chromosome, all Kurdish groups are more similar to West Asians than to Central Asian, Caucasian, or European groups, and these differences are significant in most cases. However, for mtDNA, Kurdish groups are all most similar to European groups (after West Asians), whereas for the Y-chromosome Kurds are more similar to Caucasians and Central Asians (after West Asians) than to Europeans. Richards et al. (2000) suggested that some Near Eastern mtDNA haplotypes, among them Kurdish ones from east Turkey, presumably originated in Europe and were associated with back-migrations from Europe to the Near East, which may explain the close relationship of Kurdish and European groups with respect to mtDNA. Subsequent migrations involving the Caucasus and Central Asia, that were largely male-mediated, could explain the closer relationship of Kurdish Y-chromosomes to Caucasian/Central Asian Y-chromosomes than to European Y-chromosomes.

Kurds migrated into the Caucasus at the end of the 19^th and beginning of the 20^th centuries from Turkey and/or Iran (Wixman, 1984). However, the source population for these migrants is unknown; does the genetic evidence suggest a source population for the Kurds from Georgia? Both the mtDNA and Y-chromosome data indicate reduced diversity in the Georgian Kurds, and the consequently increased effect of drift makes it more difficult to infer the origins of this group. The mtDNA data are equivocal, in that the smallest pairwise F_st values involving the Georgian Kurds are with Kurds from Turkey and from Iran. Interestingly, the Kurmanji speakers from Turkey exhibit a larger pairwise F_st value with the Georgian Kurds than the Zazaki speakers from Turkey, even though the former speak the same language as the Georgian Kurds. The Y-chromosome data do suggest that the Kurdish group in Georgia was founded by Kurmanji speakers from Turkey, although the presence of haplogroups P1 and J2 are at unusually high frequencies in the Georgian Kurds, which is an indication of genetic drift, making conclusions based on the Y-chromosome suspect. However, analysis of Y-STR haplotypes on the background of Y-SNP haplogroups P1 and J2 provided some additional insights. Y-STR haplotypes found in the Kurmanji-G are closely related to those found in the Kurmanji-T, whereas Y-STR haplotypes found in other Turkish or Iranian groups are distinct, and more distantly related to the Y-STR haplotypes found in the Kurds. Furthermore, haplogroups P1 and J2 are absent in the Zazaki-T (Nasidze et al. 2004), indicating an even more distant relatedness with the Georgian Kurds.

The Georgian Kurds have lived in the Caucasus for more than 100 years since their migration from Turkey and/or Iran. Was there any subsequent genetic exchange between the Kurds from Georgia and the surrounding Georgian population? MtDNA data suggest that there was very little, if any, gene exchange between these two groups, as the only HV1 sequence type shared between these two groups is the Cambridge reference sequence (Anderson et al. 1981), which is widespread in this region (Nasidze et al. 2004). The absence of any other shared mtDNA sequences between Georgians and the Kurmanji-G suggests very little (if any) maternal admixture between these groups. For the Y-chromosome, the strong founder effect in the Kurmanji-G, as evidenced by the unusually high frequencies of haplogrooups P1 and J2, makes it difficult to draw conclusions. However, analysis of the relatedness of Y-STR haplotypes on the background of these Y-SNP haplogroups shed some additional light on this question. A median network (Figure 3) clearly shows the distinctiveness of Y-STR haplotypes in the Kurmanji-G from those in the Georgians, suggesting an absence of genetic exchange between these two groups. On the contrary, Kurmanji-G Y-STR haplotypes are closely related with Kurmanji-T haplotypes, suggesting their common origin.

In conclusion, the genetic results indicate a close relationship between the Kurmanji-G and Kurmanji-T groups, and a clear distinction of the former from Georgians, their geographic neighbours. Moreover, the genetic results indicate that during the Kurdish migration into the Caucasus they experienced a bottleneck effect, and since that time Kurds have not undergone detectable admixture with Georgians. Our results also do not support a hypothesized northern Iranian origin for the Zazaki people; instead, genetically they are closely related to Kurds. Thus, genetic studies of even such recent events as the origins and migrations of Kurds can provide additional insights into the circumstances surrounding such migrations.

Acknowledgments

We are grateful to the original donors for providing DNA samples. We thank Donald Stilo for useful discussion. This research was supported by funding from the Max Planck Society, Germany.