The subjects were identified from a rural population in the Pune district of Western India. These participants belong to a cohort from Vadu Health and Demographic Surveillance System (Vadu HDSS) area who have been followed over years by the Vadu Rural Health Program, KEM Hospital Research Centre, Pune. The details of the sampling strategy and recruitment have been described earlier (Tiwari et al., 2017). Predominant Prakriti types, which comprises 8–10% of the population, were identified from randomly selected 10,100 individuals between the age group of 18–40 years. After preliminary screening using a questionnaire, 528 self reported healthy individuals were enrolled for detailed Prakriti evaluation using a questionnaire and methods developed in our earlier study (Prasher et al., 2008). Prakriti screening and clinical assessment was carried out by Ayurveda clinicians and a trained team of field research assistants. Assignment to Prakriti groups was carried out by two groups of physicians, one at field site and the other at CSIR-IGIB. Using unsupervised clustering approaches we have recently shown that these Prakriti groups form three natural clusters (Tiwari et al., 2017). The enrolled subjects were requested to provide fresh stool samples and it was ensured that they were not under any medication especially antibiotics. Field camps were organized in residential villages in the Vadu HDSS area. Two separate home visits were made by field teams—the first one, 8 days prior to camp to ensure availability of participants and the second, a day prior to camp to provide sterile containers along with the instructions to collect fresh stool samples. Standard operating procedures were strictly adhered to, while collecting samples, their storage at Vadu molecular lab, isolation of DNA, quality assurance and transportation of DNA aliquot to processing lab at CSIR-IGIB. A total of 135 extreme Prakriti individuals were identified, namely Kapha (n = 48), Pitta (n = 35), and Vata (n = 52).

The study population is relatively homogeneous in terms of ethnic and linguistic background as well as with respect to dietary and socio-cultural life style. In order to reaffirm the genetic homogeneity of the study population, we have earlier analyzed the genetic relatedness using unlinked and shared panel of 17675 SNP markers with the Indian Genome Variation Consortium (IGVC) diversity panel. Principal Component analysis (PCA) of the genotype data performed using EIGENSOFT 5.0 reaffirmed the homogeneity of the study population (Tiwari et al., 2017). This study has been carried out as per protocols approved by institutional ethics committee at CSIR-Institute of Genomics and Integrative Biology, Delhi and KEM Hospital Research Centre, Pune, India. Fresh stool samples were collected from subjects belonging to predominant Prakriti groups (Supplementary Table S1). Metagenomic DNA from stool samples was isolated using QIAamp DNA stool mini kit (Qiagen, Cat. No. 51504, USA).

We amplified and sequenced V2-V6 region of 16S rRNA gene using metagenomic DNA of 135 individuals, inclusive of 70 females and 65 males using Roche GS FLX+ sequencing technology. At highest stringency of Q40, we got approximately 580 Mb per sequencing run with median read length of 800 bps.

The Quantitative Insights into Microbial Ecology (QIIME) software package version 1.8.0 (Caporaso et al., 2010) was used to process and analyse raw sequencing data, separately for the male and female datasets. The split_library.py script was used in QIIME as a quality filtering step in which each sample was pre-processed with the maximum allowed one barcode error and two ambiguous bases (Ns). Sequences shorter than 300 or longer than 800 nucleotides with average quality <30 were removed from downstream analysis. In first pass of quality filtering, chimeric sequences were identified and removed by USEARCH 6.1 through QIIME's chimera processing scripts. Reads were clustered into operational taxonomic units (OTUs) with a sequence similarity threshold of 97% using UCLUST v1.2.22q (Edgar, 2010), within QIIME. Reads were assigned to OTUs using a closed reference OTU picking workflow (Caporaso et al., 2010) against GreenGenes 16S rRNA gene database (version 13_8), filtered at 97% sequence identity. In a closed-reference OTU picking, input sequences are aligned to pre-identified taxonomic clusters in a reference database. The input sequence is excluded if it does not match any reference sequence at user defined identity threshold. In further analysis, GreenGenes reference tree and taxonomic assignments were used. Microbial abundance were normalized to generate relative abundance of taxa present in each sample.

Alpha diversity (Shannon diversity and Observed species) for all samples were calculated using QIIME, to estimate species diversity, richness and evenness. Overall taxonomic differences and beta diversity were estimated through Principal Coordinates Analysis (PCoA) based on Bray-Curtis distances, using LabDSV (Roberts, 2013). Results of analyses were visualized using ggplot2 (Wickham, 2016) on R. The taxa of two highly abundant phyla, viz., Firmicutes and Bacteroidetes were removed in QIIME. The alpha diversity (Shannon diversity, richness and evenness) was calculated using “Vegan” package (Dixon, 2003) in R and beta diversity was calculated as mentioned above.

Considering the variable nature of metagenomic compositional data, we also performed further analysis only for conserved taxons. Toward this, we estimated core microbial group within the samples with presence in at least 50% of the study samples. Taxonomic classification using alpha and beta diversity were analyzed for the core microbiome as explained above. LEfSe (Segata et al., 2011) was used to identify the microbiological markers associated with Prakriti by linear discriminate analysis (LDA) effect size of 2, and for multiclass analysis one-against-all option was used with default parameters (Goecks et al., 2010). Differentially abundant taxons were annotated to their genus and species level through manual Blast (NCBI web server) against the “refseq_rna” database, retaining the highest scoring hit. Since our objective was to identify signature taxa pertaining to Prakriti groups, we systematically removed redundancies keeping the OTU with highest LDA score and selected only those differential taxa that were exclusively present in a particular group.

qPCR based validation of Prakriti specific dominant microbial enterotypes, Prevotella, Roseburia hominis, Eubacterium rectale, Blautia torques, and Blautia obeum was performed by absolute quantification of 16S rRNA gene copy number using genus specific primers (Supplementary Table S2). Total of 48 samples was used with proportional representation of Prakriti types. DNA template concentration for each sample was adjusted to 25 ng/μl. Amplification and detection were performed in a 10 μl reaction [5 μl 2x KAPA SYBR Green PCR Master Mix, 1 μl of each specific primer (10 μM), 1 μl template (25 ng/μl) and 2 μl molecular biology grade water] in triplicate using LC480 Real time PCR system (Roche, Switzerland). Amplification condition include one cycle of activation at 50°C for 2 min and denaturation at 95°C for 3 min, followed by 45 cycles at 95°C for 30 s, 58/64°C for 30 s, 72°C for 40 s; followed by extension at 72°C for 3 min. Melting curve was analyzed for non-specific products at 95°C for 5 s, 65°C for 1 min and 72°C continuous, followed by cooling at 37°C for 3 min. Group specific standard curves was generated from 10-fold serial dilutions of a known concentration of genomic DNA. Average Ct-values of the triplicate was used for estimating 16S rRNA gene copy numbers for each group using standard curves. Percentage abundance of each genus was obtained by calculating ratio of copy number of that genus to that of total bacteria (Eubacteria). Throughout the qPCR, efficiency was maintained above 90% with a correlation coefficient of >0.99. The statistical significance differences for Prakriti specific microbial enterotypes were determined by “Wilcoxon Rank Sum” test in R package.

Using high throughput sequencing, we obtained 2,699,584 and 2,266,514 reads of 16S rRNA gene from 65 male and 70 female predominant Prakriti individuals respectively. Quality filtered raw sequences were taken forward for OTU identification and analyses. QIIME assisted chimera identification using USEARCH resulted in removal of 63,973 and 66,965 sequences from the male and female respectively, with remaining quality filtered set of 2,054,437 and 1,697,708 reads. GreenGenes based OTU identification through closed reference OTU picking protocol of QIIME resulted in identification of 3363 unique OTUs among males and 2882 within females. For stringency, reduced noise in the data and to ensure proper coverage of the entire gut microbial flora, we discarded samples with <4,000 reads. Our final dataset had 1,870,897 and 1,445,312 sequences from 50 male and 63 female subjects. We detected on average, 32,570 and 21,586 OTUs across male and female samples, indicating good coverage of microbial flora, although with variability (σ_male = 25621.7 and σ_female = 14529.2). To overcome this variability, we normalized the absolute abundance counts to reflect relative abundances.

Taxonomic summary shows Bacteroidetes and Firmicutes to be the majority constituents at phylum level for both the male and female groups across Prakriti groups, together accounting for more than 98 percent of total abundance (σ_male = 4.5%, σ_female = 1.9%) (Figure 1). To summarize diversities in individuals and Prakriti groups, we used the alpha and beta diversity metrics as introduced by R. H. Whittaker. Alpha diversity captures the richness, evenness and diversity of a sample. Our analyses showed that individuals have non-homogenous composition within their gut microbial assemblage, however with comparable alpha diversity variations across different Prakriti (Figure 2). Similarly, our beta diversity analysis to investigate community differences revealed that there was no clustering of individuals with respect to Prakriti (Supplementary Figure S1). We made similar observations across both genders. This is in consonance with previous findings of genomic and transcriptomic heterogeneity in healthy individuals (Cho and Blaser, 2012; Schwartz et al., 2012; Zhang et al., 2014). However, it needs to be highlighted that Bacteroidetes and Firmicutes are the two overwhelmingly dominant phyla in gut communities, which may have a masking effect on contributions of specific organisms of different phyla.

In order to access the roles of the lesser abundant taxa in Prakriti classifications, we excluded the top two abundant phyla, viz., Bacteroidetes and Firmicutes, and re-estimated the alpha and beta diversity. The alpha diversity within females showed that among the different Prakriti, Pitta individuals had lesser diversity and richness but higher evenness than the Vata and Kapha (Supplementary Figure S2). However, the diversity, richness and evenness was found to be similar in all the three Prakriti in the male samples (Supplementary Figure S2). The beta diversity analysis showed no Prakriti specific separation of male and female samples in PCoA plot even after the removal of the most highly abundant phyla (Supplementary Figure S2). Despite inter-individual heterogeneity, we observed overlapping variation across the Prakriti classes, which most likely is an outcome of similar genetic makeup and lifestyle habits.

Stable members of a microbial community often modulate physiology of the host-microbial symbiotic system (Tschop et al., 2009; Shade and Handelsman, 2012; D'Ainsworth et al., 2015). The dysbiosis or differential abundance is one of the primary determinants of health and disease spectrum. To investigate this, we estimated core microbiome by qualifying OTUs as core only if their presence was consistent across 50% of all samples. Using in-house custom scripts, we filtered taxons from the male and female groups, to identify 209 and 224 OTUs respectively (Supplementary Table S3). Taxonomic analysis of core groups showed that the phyla Bacteroidetes and Firmicutes follow similar trends of composition as that of the total microbiome. However, the core in female group was comprised of only Bacteroidetes, Firmicutes and Proteobacteria, whereas the core in males had additional members of the family Coriobacteriaceae of phylum Actinobacteria (Figure 3). We also observed abundant bacterial species to be present across all healthy individuals, thereby occupying majority space within the core microbiome group. Core microbiome also exhibited comparable alpha and beta diversity trends as that of the total microbiome.

To control for sparseness associated with gut microbiome profiling, we looked for microbial organisms that were part of the core microbiome while being differentially abundant in a particular Prakriti using LEfSe. We performed statistical tests at multiple taxonomic levels and discovered 49 and four taxons across female and male respectively, to be significantly enriched in specific Prakriti categories. To control for potential false positives, we manually curated these differentially abundant species to arrive at a final set of 15 and two taxons that were associated with Prakriti and had no taxonomic overlap with other Prakriti specific bacteria (Table 1). To identify the correct taxonomic details of the signature OTUs, we used sequence alignment methods to identify the most probable candidate at the species level. It is important to note that these specific abundant species are present across most individuals in all Prakriti groups, however they are significantly enriched in a particular Prakriti, hence termed as signature species (Table 1, Figures 4A–F).

The Pitta females have over representation of seven species, out of which three belong to the genus Blautia—Blautia luti, B. obeum, and B. torques (Figure 4D). Blautia is a newly classified genera in the order Clostridiales, and its species constitute a major fraction of the gut flora, often responsible for conversion of carbon and hydrogen to acetate (Liu et al., 2008). Blautia, a commensal group of bacteria associated with nutrition processing for the host (Eren et al., 2015) is also associated with protection from graft versus host disease (Jenq et al., 2015) and restricts colonization of Vibrio cholera, thereby aiding in recovery from disease (Hsiao et al., 2014). Few Blautia species have increased levels during diseases like irritable bowel syndrome (IBS), though their role in the pathology is yet to be ascertained (Rajilic-Stojanovic and de Vos, 2014; Taverniti and Guglielmetti, 2014). Pitta females also showed overabundance of Butyricicoccus pullicaecoruma, a butyrate producing beneficial bacteria (Figure 4E), which has a protective effect against IBS (Eeckhaut et al., 2013) and is being considered as a potential probiotic (Geirnaert et al., 2014). Another bacterium specifically enriched in Pitta females was Gemmiger formicilis (Figure 4F), a beneficial bacteria which has been shown to be induced by a Chinese traditional medicine treatment for type 2 diabetes (Xu et al., 2015). Additionally, Incertae Sedis Mahella and Lachnospira eligens were significantly overabundant in female subjects of Pitta Prakriti, however their role in the human gut remain unclear (Supplementary Figure S3).

Kapha females were characterized by overabundance of Prevotella copri (Figure 4C), a species of bacteria commonly associated with a plant rich diet (Wu et al., 2011). It is also shown to be strongly correlated with inflammation and rheumatoid arthritis (RA) (Scher et al., 2013). Very recently, P. copri was found to induce insulin resistance in humans, resulting in increased glucose insensitivity through biosynthesis of branched-chain amino acids (BCAA) (Pedersen et al., 2016).

In Vata females, we observed significant abundance of Bacteroides vulgatus, which along with Prevotella copri, is shown to induce insulin resistance (Pedersen et al., 2016). Interestingly, studies in mice have also highlighted its protective effect in obesity and related metabolic disorders (Ridaura et al., 2013). Signature species of Vata females included E. rectale and R. hominis (Figures 4A,B), butyrate producers. They are believed to be beneficial for the gut health as they have been observed to be depleted during ulcerative colitis in independent studies, similar to Faecalibacterium prausnitzii (Vermeiren et al., 2012; Machiels et al., 2014; Cockburn et al., 2015; Riviere et al., 2015). Vata females also show enrichment of Butyrivibrio crossotus which was found to be depleted in chronic kidney disease patients (Barros et al., 2015). Relative overabundance of these species suggest a healthy intestinal flora in the Vata subjects, however we also note excess of Oscillibacter valericigenes which belongs to the genera Oscillibacter that have been reported in a case of bacterimia (Sydenham et al., 2014). We identified two other signature species, Blautia stercoris and Clostridium indolis (Supplementary Figure S3), whose role in the human gut is not clear, though Clostridium indolis is believed to be involved in carbohydrate metabolism (Biddle et al., 2014). This indicates the importance of this species in gut and its abundance dynamics can have a protective as well as an adverse effect.

Though we detected more OTUs among male subjects (Supplementary Figure S4), we found very few Prakriti specific signature microbes that qualified our stringent analysis threshold. We observed two signature bacterial species, Roseburia inulinivorans and Fusicatenibacter saccharivorans belonging to Pitta and Vata subjects respectively. Though R. inulinivorans has been characterized to be a butyrate bacteria of potentially beneficial contribution to the intestinal health (Scott et al., 2006), there is limited information with respect to mechanism underlying as to how overabundance of these two species might affect the human gut. Our analysis suggests that the male gut communities are relatively more homogenous as compared to female counterpart, which may have resulted in detection of fewer differentially abundant species.

To validate the 16S rRNA amplicon based identification of Prakriti specific microbial enterotypes, we carried out qPCR assays for absolute quantification of 16S rRNA gene copy number of differentially abundant microbial enterotypes in the study subjects (Supplementary Table S4). Relative abundance analysis of microbial enterotypes in different Prakriti has re-affirmed the insights from the 16S rRNA gene sequencing analysis (Figure 1). The differential abundance of signature species within each Prakriti type also show statistically significant difference based on qPCR. Prevotella group was significantly abundant in Kapha females compared to Pitta (p-0.0000635) and Vata (p-0.007), while Eubacterium rectale & Roseburia group was found significantly enriched (p-0.002 and p-0.035) within Vata females. Simultaneously, Blautia sp. was statistically enriched (p-0.006, p-0.02) within Pitta females (Figure 5). It would be important to understand and explore the functional role of these microbes vis-à-vis Prakriti types in subsequent studies.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.