An Ethnolinguistic and Genetic Perspective on the Origins of the Dravidian-Speaking Brahui in Pakistan

Abstract

Pakistan is a part of South Asia that modern humans encountered soon after they left Africa ~50 – 70,000 years ago. Approximately 9,000 years ago they began establishing cities that eventually expanded to represent the Harappan culture, rivalling the early city states of Mesopotamia. The modern state constitutes the north western land mass of the Indian sub-continent and is now the abode of almost 200 million humans representing many ethnicities and linguistic groups. Studies utilising autosomal, Y chromosomal and mitochondrial DNA markers in selected Pakistani populations revealed a mixture of Western Eurasian-, South- and East Asian-specific lineages, some of which were unequivocally associated with past migrations. Overall in Pakistan, genetic relationships are generally predicted more accurately by geographic proximity than linguistic origin. The Dravidian-speaking Brahui population are a prime example of this. They currently reside in south-western Pakistan, surrounded by Indo-Europeans speakers with whom they share a common genetic origin. In contrast, the Hazara share the highest affinity with East Asians, despite their Indo-European linguistic affiliation. In this report we reexamine the genetic origins of the Brahuis, and compare them with diverse populations from India, including several Dravidian-speaking groups, and present a genetic perspective on ethnolinguistic groups in present-day Pakistan. Given the high affinity of Brahui to the other Indo-European Pakistani populations and the absence of population admixture with any of the examined Indian Dravidian groups, we conclude that Brahui are an example of cultural (linguistic) retention following a major population replacement.

Introduction

Pakistan is located in south Asia, a region with an estimated population of around two billion people. The Pakistani population constitutes 10% of this mass of humanity and its population of approximately 200 million is represented by many diverse ethnicities and linguistic groups. Although modern humans are likely to have entered this region in their expansion out of Africa, and there is abundant archaeological evidence of Neolithic expansions that started around 9,000 years before present (YBP), best represented by the Harappan culture, it is not clear when, or by which route, modern humans first arrived here. Since then, the region has witnessed several additional migrations that have influenced the pattern of genetic diversity observed in the populations that currently reside there.

Studies of DNA variation in indigenous populations from this region have been greatly assisted by the wide availability of DNA samples of ethnic populations from Pakistan through the Foundation Jean Dausset’s HGDP-Centre d’Etude du Polymorphisme Humain (CEPH) Human Genome Diversity Cell Line Panel. Eight diverse ethno-linguistic groups from all over Pakistan are the sole South and Central Asian population representatives in this valuable collection (Cann et al., 2002). These samples have been analysed using a variety of genotyping platforms (Illumina HumanHap 550K and 650K bead chips and Affymetrix GeneChip Mapping Array 500K) and are currently being sequenced at high coverage using next-generation sequencing technology (Rosenberg, 2006; Rosenberg et al., 2002). The Punjabis, who form the bulk of Pakistan’s population and were not included in the HGDP panel, have recently been sequenced by The 1,000 Genomes Project, a project that also included several other populations from South Asia, including two expatriate Indian populations, speaking an Indo-European (Gujarati from Houston, Texas) and a Dravidian (Telugu from Great Britain) language (The 1000 Genomes Project Consortium, 2012).

Studies utilising markers on the human Y chromosome and mitochondrial DNA in selected populations provided initial glimpses into their male and female ancestry, respectively (Qamar et al., 2002; Quintana-Murci et al., 2004). Since then, many subsequent studies have confirmed that overall in Pakistan genetic relationships are more accurately predicted by geographic proximity than linguistic origin. One notable exception is the Kalash population, who have recently been shown to represent an ancient drifted population somewhat related to Paleolithic Siberian hunter-gatherers (Ayub et al., 2015). Other interesting groups include the Parsi, relatively recent migrants from Persia, who show evidence of South Asian admixture (Qamar et al., 2002; Quintana-Murci et al., 2004).

Y-chromosomal markers have identified male lineages associated with specific sources outside Pakistan, particularly in the Pathan and Hazara populations. A few Pathans (~2%) share haplotypes common to the Balkans and Greece that have been attributed to Alexander’s invasion of the sub-continent, whereas a common Asian ‘star haplotype’ cluster directly links the Hazara population of Pakistan to Genghis Khan (Firasat et al., 2007; Zerjal et al., 2003). Uniparental markers have also revealed signatures of East African slave trade in populations from southern Pakistan (Quintana-Murci et al., 2004).

Indo-European languages (Balochi, Punjabi, Pushto, Sindhi) predominate in this region. Other linguistic groups found in Pakistan include a language isolate, Burushaski, several Tibeto-Burman speaking groups, and a group of Dravidian-speakers. Burushaski is spoken by populations residing in the Karakoram Mountain ranges that form the western flank of the Himalayan Mountain Range, separating Pakistan from China. An initial analysis based on microsatellite or short tandem repeats (STRs) showed that this language isolate group was genetically related to their neighbours and this has been borne out by several subsequent analyses (Ayub et al., 2003; Li et al., 2008). Several Tibeto-Burman groups such as the Balti reside in northern Pakistan, but there are scant genetic data available for these populations (Qamar et al., 2002).

The Brahui are the only Dravidian-speaking population in Pakistan, where they are surrounded by Indo-European speakers, and are well separated from all other extant Dravidian-speaking populations who currently reside in southern India and parts of Sri-Lanka. This Dravidian-speaking population resides in the Balochistan province in south western Pakistan, in the vicinity of Mehrgarh, the earliest example of sub-continental Neolithic culture. Y-chromosomal, mitochondrial DNA and more recent high-resolution surveys of DNA variation have shown them to be genetically similar to their neighbours in Pakistan, who show varying proportions of west and south asian ancestry that extends in a continuum along a north-east to south-west gradient (Li et al., 2008). In the current study, we specifically explore the genetic relationship between the Brahui population, other populations within Pakistan, and several additional Indo-European- and Dravidian-speaking populations from India. We propose that an elite dominance model of Indo-European speakers can explain the genetic similarities observed between the Brahui Dravidians and Indo-European populations from Pakistan.

Materials and Methods

Data sets

Whole-genome SNP chip data were obtained from previously published work for 18 unrelated samples each (Figure 1) of the following populations: Yoruba from Nigeria; Han from China; Adygei from Russia; Baloch, Brahui, Burusho, Hazara, Kalash, Makrani, Pathan and Sindhi from Pakistan (all from the HGDP-CEPH panel) (Li et al., 2008) and three groups of mixed Indo-European (IEI), Dravidian (DRI) and Austroasiatic (AAI) Indian speakers (Chaubey et al., 2010; Metspalu et al., 2011). The data were merged using PLINK (Purcell et al., 2007) and single nucleotide polymorphisms (SNPs) lacking calls in >10% of the samples were removed. The resulting dataset featured 469,706 genotyped SNPs in 237 samples. In order to remove SNPs in close linkage disequilibrium (LD) for downstream analyses, the PLINK command “—indep-pairwise 50 10 0.1” was used to prune the dataset and obtain 67,036 SNPs with low or no LD (Alexander et al., 2009).

Figure 1.

Population locations and language affinities. Black circles indicate Indo-European and red Dravidian-speaking populations. We analysed 18 unrelated samples from each population, including 3 groups of mixed Indo-European (IEI), Dravidian (DRI) and Austroasiatic (AAI) Indian speakers.

Principal component analysis and ADMIXTURE

The pruned dataset was processed with Eigenstrat (Price et al., 2006) with five iterations of outlier removal (default parameters) and the first three components plotted. The same dataset was also processed with the ADMIXTURE software (Alexander et al., 2009) using an increasing number of K values (K = 2-10). The best fit K, i.e. the one that minimised the cross validation error, was K = 5.

Population tree models tested

We applied the three (f3) and four (f4) Population Tests (Moorjani et al., 2011; Reich et al., 2009) on the pruned dataset to formally check the topologies of three population and four population phylogenetic trees including Brahui and a set of related populations. We used the f3 test to explore all the possible three-population trees featuring Brahui, Dravidian Indians and a third population (Table 1). The rationale was that if Brahui and Dravidian Indians were not independent populations (i.e. if they shared a close genetic ancestry) the f3 Z-score is expected to be strongly negative. Similarly, we also used the f4 test to estimate the Dravidian contribution to each of the analysed Pakistani populations to see whether the Brahui had an excess of such a component. The four-population tree we tested was in the shape of (Yoruba, (Adygei, (Dravidian Indian (Han Chinese)))) with Brahui, or any other Pakistani population (X), as a putative admixture between Adygei and Dravidian Indian. To estimate the proportion of Dravidian Indian admixture, we computed the quantity: [Yoruba-Han Chinese).(X–Adygei)]/[(Yoruba-Han Chinese).(Dravidian Indian-Adygei)] as described earlier (Moorjani et al., 2013).

Table 1.

Z-scores for the f3 test to assess significance of admixture between pairs of population. Each cell reports the Z-score for the f3 test Brahui; Pop1, Pop2. Pop1 and 2 refer to the two populations listed in the first column and the top rows.

f3 test for Brahui as a mixture of:
Yoruba	Yoruba
Adygei	-3.5	Adygei
Burusho	4.1	17.7	Burusho
Makrani	20.6	12.2	11.0	Makrani
Han	23.6	7.2	39.5	6.9	Han
Kalash	-1.8	17.8	24.0	9.7	16.6	Kalash
Balochi	3.4	10.8	11.4	10.8	7.6	9.0	Balochi
DR-lndians	13.8	-0.4	34.5	9.3	48.3	18.9	10.1	DR-lndians
Hazara	13.3	19.9	36.2	10.4	72.1	19.8	10.1	36.1	Hazara
IE-Indians	5.6	10.9	33.9	10.2	38.9	22.9	11.5	46.9	33.1	IE-Indians
Sindhi	14.1	11.2	25.2	11.9	24.2	18.6	11.4	31.3	23.1	30.2	sindhi
AA-lndians	22.5	-2.2	37.5	7.9	71.7	17.7	9.4	65.0	49.2	49.5	32.1	AA-lndians
Pathan	0.0	20.3	29.6	11.8	26.8	22.1	13.0	28.4	28.8	30.1	23.1	28.8

Split times from LD/F_ST

The full dataset was used to estimate the decay of LD over different genetic length bins in each population. This measure was compared with the pairwise genetic distance (F_ST) as described by McEvoy et al., with an in-house script to estimate the divergence time between pairs of populations (McEvoy et al., 2011). Split times were used to discriminate between the two trees shown in Figure 2.

Figure 2.

Alternative models to explain the origins of the Brahui in Pakistan. (A) A model showing that the Brahui are a Dravidian-speaking group whose ancestors were present in the area before the arrival of the Indo-European speakers (represented by Adygei). This tree shows that the Brahui split (b) from the main Eurasian lineage, predates the one (c) between the Adygei and Han Chinese. The first branch (a) reflects the split between Africans and non-Africans. (B) Another model proposes that the Brahui were replaced by the incoming Indo-European speakers, while retaining their original language. This tree indicates that the Brahui originated from the aforementioned Indo-European expansion and that their split (c) took place after the Adygei-Han Chinese split.

Results and Discussion

Principal component analysis (PCA) and ADMIXTURE consistently show Brahui as a population typical of the Pakistani genetic landscape (Figure 3). The Brahui localise together with most Pakistani populations along a PC gradient that stretches from the Caucasian Adygei (light blue in Figure 3A and Figure 3B) to South Indians (yellow) and is present in all Indo-European speaking groups, except for the Kalash. Additionally, the ancestry composition of Brahui (Figure 3C) is comparable to that of the Balochi, Makrani and Pathan, all populations of Indo-European speakers in Pakistan. In contrast, the Dravidian Indians are distinguished from all Pakistani populations examined here by their PC3 values (Figure 3B) and (with the exception of the Kalash who show a unique pattern) their high yellow ancestry component at K = 5 (Figure 3C).Therefore, both analyses concur in showing that the linguistic affiliation of Brahui is not a good predictor of their genetic position and that they are genetically similar to Indo-European speakers from Pakistan. The blue African component seen in populations from southern Pakistan are traces of recent genetic admixture consistent with the documented slave trade from East Africa.

Figure 3.

Principal components and ADMIXTURE analysis. (A) Plot of the first two principal components (PC1 and PC2). (B) Plot of PC2 and PC3. Each individual is coloured according to their population of origin. The proportion of total variance explained by each component is reported in parenthesis next to the axis labels. (C) ADMIXTURE analysis of the same samples using a number of K ancestral clusters between 2 and 5, each represented by a different colour along the y-axis.

It is interesting to note that the Hazara (grey circles in Figure 3B), who speak an Indo-European language, are more similar to Han Chinese (red circles) and share a higher amount of East Asian ancestry (red in Figure 3C) than any other Pakistani population. This has been attributed to the Mongol invasion of the sub-continent in the 13th century, and indeed Y-chromosomal data indicate that ~8% of Asian and 60% of Hazara men share a male lineage linked to Genghis Khan (Hellenthal et al., 2014; Zerjal et al., 2003).

To test whether the Brahui are the result of admixture between a pre-existing Dravidian-speaking population and Indo-Europeans, we applied the f3 statistic (Moorjani et al., 2013; Reich et al., 2009) to any trio of populations including the Brahui. We examined a range of Indo-European- and Dravidian-speaking groups from the Indian sub-continent, including three Indo-European-speaking ethnic groups from Pakistan (Balochi, Makranis and Sindhi) who live in the vicinity of the Brahuis. As shown in Table 1, the only combination of populations for which f3 yielded a significant Z-score (< -2 in cells highlighted in grey), thus indicating that they could have played a role in the genetic makeup of the Brahui, were the Brahui; Adygei, Yoruba and the Brahui; AA Indians, Adygei combinations, therefore, showing no significant Dravidian Indian (DRI) contribution to the genome of the Brahuis. Furthermore, the proportion of Dravidian Indian genome present in the Brahui was estimated with the f4 statistic and shown to be comparable to that estimated for other Pakistani populations (results not shown, as the f3 test for admixture was not significant). Together, these results show that the genomes of the Brahui do not share any excess component with Dravidian-speakers from South India. They are comparable to most of the Indo-European populations currently inhabiting Pakistan and the only traces of genetic admixture (with Yoruba and AA Indians) are consistent with the documented slave trade from East Africa and close contacts with other Indian populations, traces which are also visible in the PC and ADMIXTURE analyses (Figure 3B and 3C).

The final test we devised to assess whether Brahui were present in the area when the Indo-European speakers arrived from West Asia was to estimate the genetic split time between pairs of populations. The LD/F_ST based method showed (Table 2) that the Brahui genetic splits from Yoruba, Adygei and Han Chinese are consistent with the topology reported in Figure 2B, and hence consistent with the conclusion that the Brahui, while retaining their Dravidian language, experienced a total replacement of their genetic component with the arrival of Indo-European speakers.

Table 2.

Population split times estimates (using an average generation time of 25 years) from F_ST and linkage disequilibrium.

Populations	FST	Average Time ± SEM (generations)	Average Time ± SEM (years)
Adygei-Brahui	0.023	351 ± 5	8,767 ± 129
Adygei-Dravidians	0.040	617 ± 20	15,418 ± 503
Adygei-Han	0.081	1,437 ± 31	35,924 ± 781
Adygei-lndo-Europeans	0.028	465 ± 14	11,625 ± 357

SEM = Standard error of mean.

Conclusions

In summary, we show that the Brahui, the only Dravidian-speaking population of Pakistan, do not show a higher genetic affinity with Dravidian Indians than any of their neighbouring Indo-European-speaking Pakistani populations. While this is still compatible with an ancient Dravidian genetic substrate (putatively, the yellow component in Figure 3C) shared by all the Pakistani populations, this does not highlight a preferential link between Brahui and the other Dravidian-speaking populations. Furthermore, a shared Dravidian substrate in all the Pakistani populations would not explain why most speak Indo-European languages, and only one a Dravidian language. Therefore, the most plausible scenario to explain the presence of Dravidian-speakers in Pakistan is a genetic or linguistic replacement. Formally, two models could be considered. In model 1, the ancestors of the Brahui people were a pre-existing Dravidian-speaking group in Pakistan, who were gradually assimilated by the Indo-European migrants, who arrived ~3,000 years ago, while their language was preserved. In model 2, the Brahui ancestors were Indo-European speakers, who later adopted a Dravidian language. No historical or linguistic data support model 2, so model 1 provides the best explanation for the unique characteristics of the Brahui. Interestingly, a similar phenomenon might have occurred in the Hazaras after their arrival in this region. They speak an Indo-European language and show genetic affiliation with other Pakistani populations, but retain considerable evidence of an East Asian genetic signature, consistent with their relatively recent Mongolian origin (Hellenthal et al., 2014).

The outstanding linguistic and genetic diversity reported for the populations currently inhabiting Pakistan is a legacy of the intricate migrations and admixtures that contributed to its gene pool. Availability of high-coverage whole-genome sequences, that are currently being generated, will further assist in understanding fine-scale population stratification and origins in this region.