Conceived and designed the experiments: SM UR. Performed the experiments: SM. Analyzed the data: SM UR. Contributed reagents/materials/analysis tools: SM UR. Wrote the paper: SM KUK UR. Provided laboratory facilities and funding: UR. Contributed samples: KUK.
The authors have declared that no competing interests exist.
With only ∼3,000 wild individuals surviving restricted to just 7% of their historical range, tigers are now a globally threatened species. Therefore, conservation efforts must prioritize regions that harbor more tigers, as well try to capture most of the remaining genetic variation and habitat diversity. Only such prioritization based on demographic, genetic, and ecological considerations can ensure species recovery and retention of evolutionary flexibility in the face of ongoing global changes. Although scientific understanding of ecological and demographic aspects of extant wild tiger populations has improved recently, little is known about their genetic composition and variability. We sampled 73 individual tigers from 28 reserves spread across a diversity of habitats in the Indian subcontinent to obtain 1,263 bp of mitochondrial DNA and 10 microsatellite loci. Our analyses reveals that Indian tigers retain more than half of the extant genetic diversity in the species. Coalescent simulations attribute this high genetic diversity to a historically large population size of about 58,200 tigers for peninsular India south of the Gangetic plains. Furthermore, our analyses indicate a precipitous, possibly humaninduced population crash ∼200 years ago in India, which is in concordance with historical records. Our results suggest that only 1.7% (with an upper limit of 13% and a lower limit of 0.2%) of tiger numbers in historical times remain now. In the global conservation context our results suggest that, based on genetic, demographic, and ecological considerations, the Indian subcontinent holds the key to global survival and recovery of wild tigers.
Tiger range and numbers have collapsed globally despite substantial conservation efforts. Genetic data quantifying variation from 73 wild tigers in 28 reserves in the Indian subcontinent suggests historically high numbers for tigers, and simulations reveal a signature of a 200yearold, possibly humaninduced decline. Simulations suggest that only 1.7% of historical tiger numbers now persist in peninsular Indian. Our data also reveal that tigers of the Indian subcontinent retain most of the species' genetic diversity, besides this region harbouring maximum diversity of tiger habitats. Overall, the Indian subcontinent appears to be a global hotspot holding the key to any future recovery of wild tigers from both an ecological and genetic perspective.
As top predators, large carnivores strongly shape ecological interactions in biological communities, thus playing a critical role in maintaining their structure and diversity
The tiger (
Limited phylogeographic studies
In this paper we investigate (1) the proportion of global tiger genetic variation harboured by tigers in the Indian subcontinent and (2) the demographic history of tigers in the Indian subcontinent, with a synthesis based on historical population sizes and recent human impacts. We address these questions using 1.26 kb of mitochondrial DNA and 10 microsatellite loci surveyed in 73 individual tigers from 28 different populations, and compare our results to published data from 68 tigers outside the Indian subcontinent.
Our sampling strategy concentrated on tiger populations living in varied habitats throughout the Indian subcontinent. Using 71 (
(A) A haplotype network based on 1,263 bp of mitochondrial DNA reveals that 76% of global tiger genetic variation (32 haplotypes out of 42) is retained within the Indian subcontinent. Locations of each haplotype within India correspond only approximately to the site of sample collection. 71 samples are included from Myanmar, Nepal, and India. All samples from outside India are from Luo
Subspecies  Observed heterozygosity (S.D.)  Number of alleles (S.D.)  Allelic size range (S.D.) 
Bengal (P. tigris tigris)  0.70 (0.16)  12.4 (3.6)  32 (7.7) 
All other subspecies (IndoChinese, Malayan, Sumatran and Siberian)  0.53 (0.07)  7.2 (1.6)  16 (6.1) 
All SouthEast Asian subspecies (IndoChinese, Malayan and Sumatran)  0.56 (0.14)  7.2 (1.6)  16 (6.1) 
IndoChinese (P. tigris corbetti)  0.57 (0.27)  6.2 (1.5)  14.8 (4.8) 
Malayan (P. tigris jacksoni) and Sumatran (P. tigris sumatrae)  0.55 (0.05)  5.8 (1.5)  13.2 (6.1) 
Viewed alone, this higher genetic variability observed within the Indian subcontinent is concordant with India being the geographic source for tigers. However, fossil evidence suggests that the tiger evolved in southern China
The higher genetic diversity of Indian tigers could be explained by higher effective population size, due to (1) high levels of population differentiation between tiger populations within the subcontinent due to habitat variability and past fragmentation and/or (2) high historic abundance of tigers in the Indian subcontinent.
We investigated the impacts of population differentiation on our results. We divided our samples into those roughly from the North, Central and South of the Indian subcontinent. Our results reveal a strong signature of population structure for mitochondrial DNA (
North (n = 10) 
Central (n = 11) 
South (n = 18) 

North (n = 24) 

Central (n = 18) 
0.236* (p = 0.000)  
South (n = 26) 
0.298* (p = 0.000)  0.026 (p = 0.279) 
Mitochondrial DNA sample sizes.
Microsatellite sample sizes.
We estimated historical effective population size for Indian tigers. Our mitochondrial DNA data suggest population expansion within the Indian subcontinent (Fu's F = −26.33 (p = 0.000); LAMARC: MLE of g = 2859.7 (2092.67, 5549.68), indicating growth). In contrast, the microsatellite data from the same populations indicate population decline (M ratio 0.35 (s.d. 0.08), BOTTLENECK
Given that tigers from Central and Southern India do not reveal strong subdivision (low and nonsignificant pairwise F_{st}'s for both mitochondrial and microsatellite DNA), we investigate the demographic history of tigers in this region that we refer to as peninsular India (including the states of Madhya Pradesh, Chattisgarh, Maharastra, Karnataka, Tamil Nadu, Andhra Pradesh and Kerala) using coalescent simulations
(A) Posterior distributions for population size change based on coalescent simulations for peninsular Indian tigers based on 10 microsatellite loci and the Beaumont method. Red and green curves correspond to the posterior distributions under models of exponential and linear population size change, respectively. The prior distribution is represented by the flat dotted line. Irrespective of various models, there is no support for population increase. For peninsular Indian tigers, the results reveal about 10fold decrease. (B) Posterior distributions for population size change based on coalescent simulations for peninsular Indian tigers based on 10 microsatellite loci and the Storz and Beaumont method. The posterior distributions for ancestral (red curve) and present (green) effective population size are represented here. The priors are represented by the dotted line (present population) and dashed line (ancestral population). Results confirm that postdecline population size is much smaller than the predecline population size. (C) The posterior distribution for the time since the population decline started for Indian tigers (black curve). The priors are shown by the dashed lines. The distribution has a median value at around 200 years. The vertical red and blue lines represent the approximate time since bountykilling by the British and first written history of tiger hunting by Mughals respectively. (D) The joint posterior distribution of ancestral and present effective population size based on Indian tiger data. The 90%, 50%, and 10% highest probability density (HPD) limits are plotted for the joint distribution of ancestral and current population size on a logarithmic scale. The diagonal line corresponds to stable population size.
We investigated the sensitivity of our results on demographic history to the number of genetic loci. Coalescent simulations that included 5 wild caught tigers (data from 13) genotyped at 30 microsatellite loci also revealed a very similar extent and timing of demographic decline (
An assessment of genetic variation for tigers reveals tigers in the Indian subcontinent retain more than 60% of the genetic variability of the species. In this study, we have taken extreme care to sample Indian tigers in a spatially exhaustive way. However, our results are also conditional on adequate sampling of tiger variation outside the India. Future studies might also include additional sampling of wild tigers outside the Indian subcontinent. Additionally, data from captive tigers of known origin could also be used to investigate discrepancies in genetic variation.
Indian subcontinent tigers could retain higher genetic variation not only because they had high historical population size but also because other tiger subspecies declined more severely in the recent past. Single population models in LAMARC for Indochinese tigers using the microsatellite data also exhibit a signature of recent population decline (MLE of g = −0.881825 (−1.29, −0.560142)), although lower than the estimated decline for Indian tigers. We also quantified the timing and extent of demographic decline, and simulations revealed a relatively recent decline (158 years ago,
Data from mitochondrial DNA potentially reveal a signal of demographic expansion, while microsatellite data reveal a signal of a recent population decline. The strong population differentiation for mitochondrial DNA (
Our genetic data in combination with a series of simulation models suggests that prior to historical human impacts, the genetic effective population size for tigers from peninsular India was between 2,964 and 151,008, with a median value of 23,280. Converting this effective population size into a population size suggests that between 7,412 and 377,520, with a median of 58,202 (using N_{e}/N = 0.4
It is known that in the last ∼600 years, two major historical events
The differences between the observed patterns of population differentiation between mitochondrial and nuclear markers could be because of the lower effective size for mitochondrial DNA
Our results are important for global tiger conservation because they suggest that Indian tiger populations are critical for species recovery. However, because tiger habitats in India are often small, disjunct and fragmented, conservation options are limited. Ecological studies
Genetic diversity retains the history of a species
Samples were opportunistically collected from wild individuals living inside protected areas and national parks spanning all over India. We collected 71 fecal samples and two tissue samples from most of the tiger habitat in India. Tissue samples were collected from poached animals with permissions. All samples were collected in sterile vials and preserved in absolute alcohol until processed. To avoid the effects of inbreeding in our analyses, samples were collected spatially far apart within a protected area (at least 15 km apart). DNA extraction and species identification was performed by methods explained in Mukherjee
We designed tigerspecific mitochondrial DNA primers. Polymorphic regions were ascertained using tiger mitochondrial sequences across all subspecies based on Luo
We selected ten felidspecific microsatellite loci based on PCR success rate, amplicon size, number of alleles and the level of observed heterozygosity (H_{obs}) in Indian tigers (
The mitochondrial regions were amplified in 10 µl volume reactions, cleaned by ExoSap mixture (NEB) and sequenced from both ends on an ABI 3100XL capillary sequencer. To monitor possible contamination, PCR blanks were included in all experiments.
Amplification for all the microsatellite loci was done using a multiplex approach.
A modified multiple tube approach, combined with a quality index approval was used for data quality management to account for the varying quality and quantity of DNA obtained from noninvasive sources. The complete genotyping process was repeated three times for all samples, and only those loci with quality index ≥0.75 were included in the analysis
Genetic diversity statistics, population growth indicators (Fu's F, Tajima's D) and genetic difference (F_{st}) were calculated using ARLEQUIN 3.1
If the higher number of haplotypes observed in Indian samples was due to a higher sample size, sampling fewer samples from the Indian haplotype distribution should result in fewer haplotypes. We used the observed haplotypic distribution from Indian tigers, and sampled (with replacement) 57 (total number of samples outside India) individuals for which we tabulated the number of haplotypes. This process is repeated 10,000 times. We then compared the number of haplotypes sampled in the simulation to the observed number of nonIndian haplotypes.
To determine the presence of hidden population structure within India, we performed a Bayesian clustering approach, as implemented in program STRUCTURE. We performed the analysis for K values between one and ten, using 50,000 iterations and a burnin of 10,000 assuming correlated allele frequencies. The optimal value of K was selected based on Evanno
We used five different approaches to detect past population demography. The first two approaches use summary statistics to detect population size changes, whereas the other three are likelihood or Bayesian methods. The summary statisticbased methods used were the Ewens, Watterson, Cornuet and Luikart method implemented in program BOTTLENECK
This method uses two summary statistics of the allele frequency spectrum, number of alleles (n_{A}) and expected heterozygosity (H_{e}) to achieve the patterns expected for a demographically stable population. Simulations were performed under three mutation models: infinite allele model (IAM), single stepwise model (SMM) and twophase model (TPM) to obtain the distribution of H_{e} and the values are then compared to the real data values. For TPM model, 30% multistep mutation events were allowed during the simulations. This method can detect departures from mutationdrift equilibrium and neutrality, which can be explained by any departure from the null model, including selection, population growth or decline. More importantly, consistent results from independent loci could be attributed to demographic events over selection. Thus, this approach allows the detection of population size changes across different mutational models.
This method uses data on the frequency and the total number of alleles and the allelic size range to investigate population decline. In a reducing population, the expectation of the reduction of number of alleles is much higher than the reduction of allelic size range. Thus the ratio between the number of alleles and the allelic size range is expected to be smaller in recently reduced populations than in equilibrium populations.
This method provides maximum likelihood estimates of growth (or decline) rates using a MetropolisHastings Monte Carlo Markov Chain algorithm. As it is a time backward approach, a negative value of g indicates that the population has been shrinking (it was bigger in the past than it is now) and a positive value indicates that it has been growing (it was smaller in the past than it is now). Genetic data from ten microsatellite loci was used for these analyses.
This approach assumes that a stable population of size N_{1} started to change (either decrease or increase) t_{a} generations ago to the current population size N_{0}. This change in the population size is assumed to be either linear or exponential under stepwise mutation model (SMM). This bayesian approach uses the information from the full allelic distribution in a coalescent framework to estimate the posterior probability distribution of (i) r = N_{0}/N_{1} (rate of population size change), (ii) t_{f} = t_{a}/N_{0} (time since the population size change started, scaled by N_{0}), and (iii) θ = 2N_{0}μ. A Markov Chain Monte Carlo (MCMC) algorithm is used to generate samples from the posterior distribution of these parameters. Although this method allows the quantification of a population increase or decrease, N_{0} and N_{1} cannot be estimated independently. Similarly, it can only approximate t_{a} as a time scaled by N_{0}, with N_{0} being unknown. Thus, the population size change can be quantified, but it cannot be dated. This model was employed to test for a genetic bottleneck under different model of population size change (linear or exponential).
For each analysis, at least three independent runs were performed using different parameter configurations and starting values. Most importantly, all the runs were carried out with positive starting values of log(r) (see
Rectangular prior distributions were assumed for log(r), log(θ) and log(t_{f}). Wide bounds were chosen (between 10^{−3} and 10^{3} on a natural log scale) for minimum effects on posterior distributions.
This approach is an extension of Beaumont's method and allows quantification of effective population sizes N_{0} and N_{1}, rather than their ratio along with T, time since the population change. The method assumes an exponential model of population size change. In this model, prior distributions for N_{0}, N_{1}, T and μ are assumed to be log normal. The mean and the standard deviations of these prior log normal distributions are drawn from priors (or hyperpriors) distributions.
Wide “uninformative” priors were used to perform multiple runs for this approach (
The generation time for tigers is known to be about five years
Figure showing investigated mitochondrial regions and the variable positions (lines) and the associated heterozygosities of these positions for all tiger subspecies. The colored boxes highlight the regions amplified by our primers for this study.
(0.52 MB TIF)
(A) Population size change for the tigers in the Indian subcontinent, with black and green curves corresponding to the posterior distributions under models of exponential and linear population size change, respectively. The prior distribution is represented by flat dotted line. Irrespective of various models, there is no support for population increase. (B) The posterior distributions for ancestral (red curve) and present (green) population size are represented here. The priors are represented by the dotted line (present population) and dashed line (ancestral population). (C) The posterior distribution for the time since the population decline started for Indian tigers (black curve) is represented here. The priors are shown by the dashed lines. The distribution has a median value at around 270 years. (D) Joint posterior distribution of ancestral and current population size based on South and Central India tiger data. The 90%, 50%, and 10% highest probability density (HPD) limits are plotted for the joint distribution of ancestral and current population size on a logarithmic scale. The diagonal line corresponds to stable population size.
(1.34 MB TIF)
(A) Population size change for the Indian tigers (30 microsatellites, n = 5) with red and green curves corresponding to the posterior distributions under models of exponential and linear population size change, respectively. The prior distribution is represented by flat dotted line. (B) The posterior distributions for ancestral (red curve) and present (green) population size are represented here (30 microsatellites, n = 5). The priors are represented by the dotted line (present population) and dashed line (ancestral population). (C) The posterior distribution for the time since the population decline started for Indian tigers (black curve) is represented here. The priors are shown by the dashed lines. The distribution has a median value at around 258 years. (D) Joint posterior distribution of ancestral and current population size based on Indian tiger data. The 90%, 50%, and 10% highest probability density (HPD) limits are plotted for the joint distribution of ancestral and current population size on a logarithmic scale. The diagonal line corresponds to stable population size.
(1.28 MB TIF)
(A) Posterior distributions for rate of population size changes from independent MCMC runs for Indian tigers under linear change model (Beaumont method). Details of the models are given in
(0.68 MB TIF)
(A) Posterior distributions for current population size from independent MCMC runs for Indian tigers under Storz and Beaumont method. Details of the models are given in
(0.64 MB TIF)
(A) Population size change for the IndoChinese tigers (
(1.29 MB TIF)
(A) Posterior distributions for rate of population size changes from independent MCMC runs for IndoChinese tigers under linear change model (Beaumont method). Details of the models are given in
(0.66 MB TIF)
(A) Posterior distributions for current population size from independent MCMC runs for IndoChinese tigers under Storz and Beaumont method. Details of the models are given in
(0.64 MB TIF)
Information on samples used in this study.
(0.11 MB DOC)
Speciesspecific mitochondrial primers designed and used in this study.
(0.05 MB DOC)
Information of 10 microsatellite loci used in this study.
(0.05 MB DOC)
Linear models (Beaumont method).
(0.04 MB DOC)
Exponential models (Beaumont method).
(0.04 MB DOC)
Exponential models (Storz and Beaumont method).
(0.05 MB DOC)
We thank the forest departments from Kerala, Tamil Nadu, Karnataka, Chattishgarh, Maharastra, Madhya Pradesh, Andhra Pradesh, Rajasthan, West Bengal, Assam, and Orissa for permissions and assistance with sample collection. We thank N. Samba Kumar, A. J. T. Johnsingh, J. Krishnaswamy, N. Mukherjee, A. Kumar, and many field researchers for assistance in procuring samples. We thank T. Shivanand, S. Mukherjee, D. Chakraborty, and E. A. Hadly for extensive discussions and comments on the manuscript. V.V. Robin, V. Kolipakam, and D. Jathanna provided comments on the earlier versions of this manuscript. S. Koushika provided lab space and computing facility. S. Joseph provided technical assistance. We thank the editor Dmitri Petrov and two anonymous referees for their constructive comments and helpful suggestions, which greatly improved this manuscript.