India is currently the second most populous country in the world after China with a population of 1,087 million that is predicted to expand to 1,628 million by the year 2050 1. The high prevalence of endogamy and relatively low admixture present in the population distinguishes Asian Indians (from the subcontinental region comprising India, Pakistan, Bangladesh and Sri Lanka) from most other populations presently used in genetic studies 2. As the country is becoming more westernized both in diet and lifestyle, the prevalence of diseases associated with these lifestyle changes is increasing.

The languages of India belong to four major families: Indo-Aryan (a branch of the Indo-European family), Dravidian, Austroasiatic, and Sino-Tibetan, with the overwhelming majority of the population speaking languages belonging to the first two families 3. The 1981 census – the last census to tabulate languages – reported 112 mother tongues with more than 10,000 speakers and almost 1 million people speaking other languages. The Indian constitution recognizes 22 official Indian languages. They are Assamese, Bengali, Bodo, Dogri, Gujarati, Hindi, Kannada, Kashmiri, Konkani, Maithili, Malayalam, Manipuri, Marathi, Nepali, Oriya, Punjabi, Sanskrit, Santhali, Sindhi, Tamil, Telugu, and Urdu. Of the official languages, about 43 percent of the population speaks Hindi as their mother tongue. Telugu, Bengali, Marathi, and Tamil rank next, each the mother tongue of about 4 to 5 percent; Urdu, Gujarati, Malayalam, Kannada, and Oriya are claimed by between 2 and 3 percent; Bhojpuri, Punjabi, and Assamese by 1 to 2 percent; and all other languages by less than 1 percent each. Pertinent to this study, with the exception of Hindi, which is the primary language spoken in several north Indian states, each of the other languages is spoken by individuals with origins from a distinct Indian state 3 and hence, each corresponds to a geographically defined region within India.

The Parsis are a member of the close-knit Zoroastrian community found primarily in the Indian state of Gujarat. Although the Parsis of India are descendents from Persian Zoroastrians who emigrated to the Indian subcontinent from Persia over 1,000 years ago 456, they no longer have social or familial ties to Persians, and do not share language or recent history with them. Mitochondrial and Y-chromosome analyses suggest that since their arrival in India, the Parsis have integrated themselves into Indian society through a male-mediated migration of their ancestors who admixed with local females 78. However, they have simultaneously maintained their own ethnic identity, making the Parsi community both Indian in terms of national affiliation, language and history, but not typically Indian in terms of consanguinity or cultural, behavioral and religious practices.

Large inter-ethnic variation is known for a number of common conditions, which may be due, in part, to differences in risk allele frequencies that are the result of either population structure, migration, or the action of natural selection on the risk variants themselves 9. Because selective pressures may vary with geography and other environmental variables, detectable patterns of geographic variation may sometimes be observed in disease alleles. Recent population structure analysis using 1200 genome-wide polymorphisms in 432 individuals from 15 Indian language groups has shown that Indians constitute a distinct population cluster and that despite the geographic and linguistic diversity of the groups, they exhibit a low level of genetic differentiation 10.

Metabolic disorders have been found to have a disproportionately high prevalence in the Indian population, a phenomenon which is likely associated with the increasing westernization of India. For example, many studies have shown that the prevalence of coronary artery disease (CAD) in individuals of Indian origin is currently much higher than in other ethnic groups and is even increasing 11121314151617. It is also currently estimated that 18% of the Indian population suffer from hypertension, one of the major risk factors associated with CAD, and that the prevalence of hypertension is also increasing within this population 18. However, this prevalence is comparable to that in other worldwide populations (15.0–40.0%) 19. Type-2 diabetes is now reaching epidemic proportions in India, most prominently in urban Indians where there has been a very steep increase in prevalence over the last decade from 8.3% in 1992, 11.6% in 1997, to 15.5% in 2005 20212223. Other diseases are also found at varying frequencies in the Indian population including Hirschsprung disease 2425, age-related macular degeneration (AMD) 2627, prostate cancer 2829, and type-1 diabetes 3031. However, despite the high prevalence of common disorders and the large population within which to study them, most modern genetic studies have not incorporated Indian populations.

In the present study, we have investigated the prevalence of common polymorphisms that have recently been reported to be risk factors for atherosclerosis, hypertension, diabetes, prostate cancer, Hirschsprung disease, and age-related macular degeneration (Table 1) in a cohort of 576 India-born Asian Indians sampled in the United States. This cohort consists of individuals whose mother tongue is one of 14 of the 22 "official" languages recognized in India, as well as individuals whose mother tongue is Parsi. We have also investigated variation in the TAS2R38 bitter taste receptor, associated with the ability to taste phenylthiocarbamide (PTC), and in the SLC24A5 gene, associated with variation in human skin pigmentation (Table 1). The prevalence of these diseases/traits 323334353637383940414243 and the risk-associated variant alleles of these polymorphisms (see Additional file 1) have been reported to vary between populations/ethnicities, suggesting that their influence may also vary between populations. The etiology of most cases of CAD 44, type-1 45 and type-2 46 diabetes, hypertension 47, prostate cancer 48, and AMD 49 are complex and are likely caused by the combined effect of genes and environment. However, the relative contribution of genes versus environment to the diseases/traits and the degree of genetic heterogeneity are likely very different for the different diseases/traits considered in this paper. Whilst we have investigated the prevalence of these minor alleles within this Indian cohort and compared the frequencies to those of other populations, we have not addressed the potential impact of environmental factors that may differ between these populations. This study will hopefully aid in the design of future genetic studies investigating the underlying causes of these diseases in India.

We have examined the variation of 19 common polymorphisms (Table 1) associated with diseases/traits in a collection of 576 individuals of Indian descent sampled in the United States that represent 14 of the 22 official or scheduled Indian languages as well as one additional cultural group (Parsi; Figure 1). These individuals are part of an expanded cohort that represents 15 groups defined by language and that was used in a previous population-genetic study 10. The allele frequencies of these 19 common polymorphisms in the Indian cohort can be found in Tables 2, 3, 4; for genotype frequencies see Additional file 2, Tables S1–3.

We have reported here the frequencies of common polymorphisms associated with an increased risk of several diseases or phenotypes, within a cohort of 576 Indian-born Asian Indians. It is important to note that because the study participants were born in India, we refer to the individuals and populations as being "Asian Indian" or "Indian"; however, it is important to note that because the Asian Indian individuals were sampled in the United States, and therefore do not represent a random sample from India or all of the different castes, tribes, or language groups, some biases may be introduced when extrapolating the results to India as a whole 10. It is also important to note that because Hindi is spoken by 43 percent of the Indian population, primarily in the states of Northern India, heterogeneity will be present between its various constituent population groups. Our Hindi language group is a sample of individuals who emigrated from a restricted region of India, and therefore biases may be introduced when extrapolating their results to Hindi speakers as a whole.

All polymorphisms, with the exception of ALOX5 g.20C>T, were found to be in HWE. The deviation away from HWE for the ALOX5 g.20C>T SNP is most likely due to chance, but given its low frequency we cannot exclude the possibility of genetic drift, differing selection between the groups, or the small sample size of all but one (Gujarati) of our language groups. Genotyping error could be another possible explanation, but the extremely high accuracy of the TaqMan genotyping assay 61 and the 100% concordance of replicated genotypes makes this explanation unlikely.

No significant population subdivision was identified by F_ST between the different Indian language groups (Tables 2, 3, &4). However, variation in the MAF of some of the polymorphisms between the different language groups was observed. Furthermore, the ALOX5 g.8322G>A and g.50778G>A SNPs, as well as the PTPN22 g.36677C>T SNP, were found to be present in some but not all of the Indian language groups investigated here. There was no discernible pattern to presence/absence of the minor allele of these SNPs with regard to the location of these groups within India (Figure 1), and it is possible that their presence or absence is the result of either the different origins of the language groups during the peopling of India and strengthened by their practice of endogamy 6263, or admixture introduced by the differential migration of non-Indian populations into regions of India during the course of its history 6465. With the exception of the Gujaratis, the small sample size of our individual language groups could also be responsible for this observation as we may be under powered to detect the minor allele in these groups. However, the absence of the ALOX5 g.8322G>A, ALOX5 g.50778G>A, and PTPN22 g.36677C>T SNPs in the Gujarati language group, for which we have around seven times the number of individuals of the other groups, would suggest that their absence in the other groups may not necessarily be due to a lack of a sufficient sample size for their detection. If this observed variation in the MAF of some of the polymorphisms between the different language groups is confirmed, then this could impact the design of future case-control studies in India as it would suggest that the inclusion of a disproportionate number of individuals from a particular language group in either the case or control groups could lead to missed or spurious associations.

Interestingly, the MAF of the atherosclerosis (ALOX5P, ALOX5 g.20C>T, g.8322G>A, and g.50778G>A), type-1 and type-2 diabetes (CAPN10 g.4834T>C, TCF7L2 g.98386G>T, and PTPN22 g.36677C>T), and Hirschsprung disease (RET g.9349G>A) associated polymorphisms were found to be similar to those of European populations. Similarly, the frequencies of variants associated with AMD (CFH g.37989T>C and LOC388715 g.205G>T) and prostate cancer (DG8S737 and rs1447295 were similar to those of African populations. However, none showed a high similarity to the populations of Eastern Asia. This could reflect a greater degree of non-Asian compared to Asian gene flow into India during the course of its history; however, if higher caste is correlated with a European or western Asian component of ancestry, this could be a reflection of our sampling of Asian Indian immigrants in the United States that are likely biased towards the higher castes.

The frequency of the CYP3A5 g.6980G>A, AGT g.802C>T, and GNB3 g.825C>T polymorphisms have been reported to vary significantly between different populations/ethnicities in a cline that correlates with variation in the latitudinal position of the populations 505556. We have shown that whilst the MAF of the CYP3A5 g.6980G>A SNP in the Indian groups agrees with the trend identified in other world populations, the inclusion of the Indian data reduces the previously reported correlation between the frequency of the AGT g.802C>T and GNB3 g.4423C>T SNPs and latitude. The association of the MAF of these three SNPs with changes in latitude, purportedly due to their effects on sodium homeostasis, suggests that the relevant selective pressures within India are not themselves strongly correlated with latitude.

An association between changes away from the common allele (5 repeats) in a repeat-polymorphism (ALOX5P) comprised of Sp1 binding sites in the promoter region of ALOX5 and atherosclerosis has been previously reported 57. Analysis of ALOX5P in the Indian cohort identified six alleles; 56.4% of individuals were homozygous for the common allele of five Sp1-binding site repeats, and 92.5% of individuals had at least one copy of the common allele. The remaining 7.5% of individuals had two variant alleles, which is much lower than the incidence of CAD in the Indian population 1666. This is higher than the 5.96% reported for a cohort of individuals of mixed-ethnicity, who reported 94.0% of individuals with at least one copy of the common allele 57. The same study also reported that variant alleles were more common in blacks (24.0%) and Asians or Pacific Islanders (19.4%) than in Hispanic (3.6%) and non-Hispanic whites (3.1%). The frequency of variants in this Indian cohort would suggest that their frequency is closer to Caucasians than Asians.

Three coding SNPs have also been identified in the ALOX5 gene (g.20C>T, g.8322G>A, and g.50778G>A) and these appear to be in partial or complete LD with the alleles of ALOX5P (HA; unpublished data). Haplotype analysis of the four ALOX5 polymorphisms in this Indian cohort found that there is LD between three alleles of ALOX5P – the major and minor alleles of the g.20C>T SNP (5 and 4 repeats, respectively), and the major allele of g.50778G>A SNP (3 repeats). Haplotypes containing variant alleles of ALOX5P are present in almost half of the cohort, with 8.1% of the cohort homozygous for haplotypes carrying a variant allele.

Analysis of the type-2 diabetes-associated SNPs CAPN10 g.4834T>C and TCF7L2 g.98386G>T, found their MAFs (23.1% and 24.8%, respectively) to be almost twice as common as type-2 diabetes in the Indian population (11.6%) 22. Two recent studies have reported that the TCF7L2 g.98386G>T variant is strongly associated with type-2 diabetes risk in a Asian Indians 6768, and our frequency of this variant (24.8%) in this Indian cohort is close to that reported for control subjects (22%) 68. The PTPN22 g.36677C>T SNP has a low MAF (0.009) in our Indian cohort, comparable to that previously reported (0.012) 69 and almost one hundred fold higher than the reported prevalence of type-1 diabetes in the Indian population (0.011–0.026%) 3031. The minor allele of PTPN22 g.36677C>T SNP is absent in other Asian populations, but present in those of European origin (see Additional file 1, Table S9), suggesting that its presence in Asian Indians is through the introduction of admixture by migrating individuals of European origin into India during the course of its history.

Interestingly, the most common allele of the DG8S737 microsatellite in this Indian cohort has been associated with an increased risk of prostate cancer in other populations (-1 allele) 58. However, whilst this allele has been reported to be in strong LD with the minor A-allele of SNP rs1447295 in other populations, we have found that a different allele (-4 allele) is in strong LD with the minor A-allele of SNP rs1447295 in this Indian cohort. The frequency of these polymorphisms is also still far higher than the frequency of prostate cancer in the Asian Indian population (10 per 100,000) 2829. Further studies on Asian Indian prostate cancer cases may allow for the true causative variant to be identified.

The moderate frequency of the Hirschsprung disease-associated minor allele of RET g.9349G>A in the Indian cohort is in contrast with the relatively low frequency of the disease in India (2.8 per 10,000 births) 25. However, the MAF of the CFH g.37989T>C and LOC387715 g.205G>T SNPs in this Indian cohort is in agreement with the low frequency of AMD in the Indian population 27.

Previous investigation suggests that the PAV "taster" haplotype is dominant over the other recessive haplotypes including the "non-taster" AVI haplotype as PAV/AVI individuals were largely able to taste PTC 5970. We would therefore expect, based upon the frequencies of the different haplotypes in this Indian cohort, that 57.4% of the population would be able to taste PTC (Table 16). This value is very close to the percentage of the population identified as able to taste PTC in previously reported studies: Gujarat (57.1%), Kerala (51.6%), Maharashtra (54.9%, 57.6% and 63.7%), Tamil Nadu (59.2%), and Uttar Pradesh (69.1%) 7172737475. These data would suggest that the lower than expected number of Indians able to taste PTC when compared with the average of 75% in worldwide populations 7677 is due to an increased presence of the non-taster haplotype in this population. In agreement with this notion, we observed that the frequency of the haplotype of the three SNPs in the TAS2R38 gene associated with the ability to taste PTC in this Indian cohort is almost 20% lower than other world populations.

Interestingly, the MAFs of the three TAS2R38 SNPs were found to vary in the Indian groups in a cline that correlated with the latitudinal variation of the groups. The MAFs of all three SNPs were higher in the more northerly Indian groups than in the more southerly groups, suggesting a greater frequency of tasters in the former than the latter. However, this trend was only observed in the other world populations for the TAS2R38 g.144G>C SNP, with the trend reversed for both the TAS2R38 g.784C>T and g.885A>G SNPs.

One possible explanation for the differences between the trends in the Indian groups and other world populations could be the differential migration of non-Indian individuals into different parts of India during the course of its history. This could result in variation in the degree and source of the admixture introduced into the different language groups such that their frequencies deviate from the trends of other populations. There could also be selective pressures that are unique to, more important in, or different in action in, India than the other world populations, and that also share a strong correlation with latitude.

It has been previously reported that balancing selection is acting on the TAS2R38 receptor gene, possibly due to the different taster and non-taster alleles encoding receptors that recognize different, but equally important, ligands 6077. The latitudinal cline we have identified in our Indian language groups with regards to the taster/non-taster allele frequencies could therefore be the result of the strength of the selective pressures exerted by these ligands affecting individuals in the north and south of India differently, but for reasons that are distinct from other world populations. Another possible selective pressure relates to bitter taste perception and the acceptance of foods containing increasing amounts of spice 78. This is of particular interest given that patterns of spice use have been reported to both vary between populations and regions, and also to have a significant latitude correlation, where more spices tend to be used in warmer climates purportedly due to their powerful antimicrobial action 79. The climate differences between the more arid south of India and the more mountainous north, and the incorporation of a lot of spices in Indian cooking, makes this a possible explanation for our latitudinal correlation. It has also been proposed that the local adaptation in human bitter taste receptor genes is common and has been driven by the fitness advantages of avoiding toxic, bitter compounds found in plants 80. If the composition and abundance of ligands to which the different taster and non-taster alleles of the TAS2R38 gene vary geographically in India in a cline associated with latitude, this could also explain our observed latitudinal correlation.

The MAF for the SLC24A5 g.13242G>A polymorphism (0.114) in the Indian population is closer to that of Caucasian European populations (0.000–0.020) than to that of East Asians (0.979–0.989) and Africans (0.730–0.980) (see Additional file 1, Table S14). This is surprising given that, as a whole, people of Indian origin have darker skin tone compared to Europeans. The East Asian and African populations share a similarly high frequency of the minor allele of the SLC24A5 g.13242G>A SNP, which in itself is surprising as East Asians are typically of much lighter skin than Africans. It is also interesting that the t-test suggested a greater similarity in MAF between Indians and those of populations in the Americas (Puerto Rican, Mexican, and Amerindian), who are of a similar skin tone (Table 4). The presence of the minor allele of the SLC24A5 g.13242G>A SNP cannot therefore, completely explain Indian skin color, making the investigation of polymorphisms in additional genes implicated in skin pigmentation 81 essential to understanding the intricacies of their skin color.

We have found that, with the exception of the TAS2R38 haplotypes associated with the ability to taste PTC, the frequency of the disease/trait-associated minor alleles of the polymorphisms investigated in this Indian cohort did not correlate with the reported disease/trait prevalence in the Indian population. This is unsurprising given that, with the exception of the ability to taste PTC, all are known to be complex diseases/traits that are the result of the combined effect of variants at multiple genetic loci and environmental factors. Variation in genetic backgrounds and environmental heterogeneity between populations will likely have a modifying effect on the correlation of the allele frequency of polymorphisms with disease/trait prevalence as their contribution to the presentation of the disease/trait varies. The degree of complexity between these disease/traits will also vary, as some, such as CAD, are known to be highly complex whereas others, such as AMD, may be less complex. One might therefore not expect the correlation of factors associated with CAD to be as strong as those associated with AMD, and therefore the lack of a strong correlation between many of these polymorphisms and their respective complex disease/trait prevalence most likely indicates that the environmental and/or other genetic factors make a stronger contribution to the presentation of the disease/trait in Indians than do these variants alone. Interestingly, the two polymorphisms associated with AMD (CFH g.37989T>C &LOC387715 g.205G>T), whilst not showing strong correlation between their disease-associated MAF and reported disease prevalence in the Indian population, did show a frequency that correlated with the MAF of these SNPs in other populations and the prevalence of AMD in those populations, suggesting that they may share similar risk factors.

The Asian Indian population represents a large population within which many complex-trait disorders are found at a high frequency. The high prevalence of endogamy and relatively low admixture present in this population distinguishes it from most other populations presently used in genetic studies as it represents a distinct genetic background that is largely unaffected by outside admixture. In some cases, such as coronary artery disease, Asian Indians exhibit unique phenotypic characteristics that distinguish them from the other populations, suggesting that unique causative factors underlie this and possibly other related diseases. Our results suggest that caution should be used when treating the Asian Indian population as a single population in genetic studies using SNPs. The overall difference in minor allele frequency of the polymorphisms investigated was small between the different language groups, but some inter-group variation was observed. However, caution is warranted in interpreting our results as whilst our Indian cohort is likely to be reasonably representative of first-generation individuals of Asian Indian descent currently residing in the United States, individuals from relatively mobile populations and those of higher caste and socioeconomic status are likely overrepresented within this cohort. Such individuals likely do not provide a random sample of the source populations in India and therefore if variables such as caste and socioeconomic status do have a significant effect on genotype frequency in India, this will not be reflected in these results. Additionally, if higher caste is correlated with a European or western Asian component of ancestry, a sample in the United States may be biased towards finding a greater similarity of populations from India to those of the Europe/Middle East rather than to those of East Asia 10, as we have discovered for many of the polymorphisms reported here. We must also use caution when interpreting the variation at the PTPN22 g.1825C>T and ALOX5 g.50778G>A loci between the 15 language groups, as the small sample size for each group, with the exception of the Gujarati, is a limiting factor in our ability to detect polymorphisms with a low minor allele frequency. However, our results do support the inclusion of the Indian population in modern genetic association studies as these data suggest that Asian Indians likely exhibit unique genotype as well as phenotype characteristics that may yield new insight into the underlying causes of diseases that are not available in other populations.

Age-related macular degeneration (AMD); Coronary artery disease (CAD); Hardy-Weinberg equilibrium (HWE); Linkage disequilibrium (LD); Minor allele frequency (MAF); Phenylthiocarbamide (PTC).

The author(s) declare that they have no competing interests.

TJP and PIP conceived the study. TJP and NM performed the genotyping. TJP performed the data analysis with assistance from DVC AD and DW assisted with the analysis of the CYP3A5 and AGT polymorphisms, and HA assisted with the genotyping of the ALOX5 polymorphisms. The paper was written primarily by TJP and PIP, with assistance from DVC, HA, and AD.