The Kalash genetic isolate: ancient divergence, drift, and selection

Abstract

The Kalash represent an enigmatic isolated population of Indo-European speakers who have been living for centuries in the Hindu Kush mountain ranges of present-day Pakistan. Previous Y chromosome and mitochondrial DNA markers provided no support for their claimed Greek descent following Alexander III of Macedon's invasion of this region, and analysis of autosomal loci provided evidence of a strong genetic bottleneck. To understand their origins and demography further, we genotyped 23 unrelated Kalash samples on the Illumina HumanOmni2.5M-8 BeadChip and sequenced one male individual at high coverage on an Illumina HiSeq 2000. Comparison with published data from ancient hunter-gatherers and European farmers showed that the Kalash share genetic drift with the Paleolithic Siberian hunter-gatherers and might represent an extremely drifted ancient northern Eurasian population that also contributed to European and Near Eastern ancestry. Since the split from other South Asian populations, the Kalash have maintained a low long-term effective population size (2,319-2,603) and experienced no detectable gene flow from their geographic neighbors in Pakistan or from other extant Eurasian populations. The mean time of divergence between the Kalash and other populations currently residing in this region was estimated to be 11,800 (95% confidence interval = 10,600-12,600) years ago, and thus they represent present-day descendants of some of the earliest migrants into the Indian sub-continent from West Asia.

Figure 1

Population Structure and Isolation of the Kalash (A) Geographic location of the three Pakistani villages where the Kalash samples were collected. (B) Principal-component analysis (PCA) of Eurasian populations shows the first two components superimposed with the spatial kriging interpolation of the admixture coefficient of the Kalash genetic cluster. The proportion of admixture is indicated by color: orange represents the maximum level of admixture, and black represents the lowest. There is no gradient into the proportion of admixture with the Kalash cluster, suggesting a low level of gene flow between nearby populations and a high degree of isolation. (C) Admixture analysis in which the lowest cross-validation error (k = 7) shows the unique Kalash cluster (dark green).

Figure 2

Kalash Demographic History (A) PSMC analysis shows a low effective population size for the Kalash. (B) Kalash effective population size estimated from LD analysis. (C) MSMC analysis of the time of the split between the Kalash and African genomes (YRI, LWK, and MKK) and non-African genomes from East Asia (CHB and JPT), Europe (CEU and TSI), South Asia (GIH), and America (MXL). (D) A UPGMA (unweighted pair group method with arithmetic mean) dendrogram shows the LD-estimated time of divergence between populations. The mean time of divergence between the Kalash and other populations from the Indian sub-continent is estimated to be 11,800 years ago (dashed red line).

Figure 3

Shared Genetic Drift with Ancient Genomes (A) Proportion of shared genetic drift (measured as f3 statistics) between extant world-wide HGDP-CEPH populations (including the Kalash) and the ancient Siberian hunter-gatherer (MA-1). The magnitude of the computed f3 statistics is represented by the graded heat key. The proportion of genetic drift shared between the Kalash and MA-1 is comparable to that shared between MA-1 and the Yakut, Native Americans, and northern European populations. (B) Ternary plot of shared genetic drift with three ancient genomes: MA-1 (left), La Braña 1 (middle), and Ötzi, the Tyrolean Iceman (right). The high proportion of genetic drift shared between the Kalash and MA-1 is comparable to that shared between MA-1 and Native Americans. In comparison with other populations from South Asia, the Kalash also share a higher proportion of genetic drift with La Braña 1 and Ötzi.

Figure 4

Consequences of Drift and Selection in the Kalash (A) A nonsense variant in ACTN3 (rs1815739) is present at a higher frequency (left) in the Kalash than in their neighbors in Pakistan. Forward-time simulations (right) show that such a high frequency of the derived allele in the Kalash (dashed blue line) is only observed in a scenario that considers positive selection acting on the variant. The lower line represents the observed mean frequency of the derived allele in the Pakistani population, the orange lines represent the simulated allele frequency of the derived allele in each replicate in the scenario without selection, and the dark red lines represent each replicate in the scenario with positive selection. The observed frequency of the derived allele in Kalash population is reached only in the scenario with selection and only after 400 generations of drift (∼10,000 or 11,200 years ago if we assume a generation time of 25 or 28 years, respectively), suggesting that the observed pattern for this stop gain on ACTN3 can best be explained by selection acting in ancient times and not by any recent population split. (B) The Kalash are fixed for the ancestral allele of the MCM6 intronic variant (rs4988235) that is associated with lactose intolerance. The derived allele that is associated with lactase persistence is present at moderate frequency in populations from Pakistan (left panel and upper dashed line in the right panel). Forward-time simulations (right panel) suggest that recent isolation and genetic drift cannot explain the observed pattern for this functional polymorphism in the Kalash population. Only 1/1,000 replicates (represented by orange lines) reach fixation after 500 generations of drift (∼12,500 years ago if we assume a generation time of 25 years).