mtDNA haplogroup and single nucleotide polymorphisms structure human microbiome communities

Humans are remarkable hosts to microbes, and we have in fact co-evolved as highly plethoric communities. The NIH Roadmap initiative, known as the Human Microbiome Project (HMP) [1], enabled sequence-based comprehensive characterization of the adult human microbiota. Human-associated microorganisms are present in numbers exceeding the quantities of human cells by at least 10-fold beginning in the neonatal period, and the collective microbiome metagenome exceeds our human genome in terms of gene content by orders of magnitude (>150 fold) [2–4]. We now appreciate that the microbiota are a metabolically and antigenically vibrant diverse community, which may function as mutualists (symbiotically beneficial), commensalists (of neither harm nor benefit), or pathogens (detrimental to the host). However, we do not yet understand how the host genomic content influences its establishment.

The HMP Consortium [2–5] established just such a population-scale framework with which to characterize the relationship of microbial communities with their human hosts. The signature frameworks of as many as 18 body sites of 242 screened and phenotyped adults from the target population of 300 subjects have been described. To minimize exogenous and environmental exposure that may influence taxonomy abundance, rigorous clinical standards were applied to screen subjects to assure that the cohort were similar in baseline health status [4]. The majority of the subjects were non-vegetarian, non-smokers, non-obese and yet diverged with respect to race/ethnicity, and parental country of origin. Of interest, although no taxa were universally present among all body habitats and individuals, the carriage of metabolic pathways was surprisingly alike, with a greater degree of similarity observed among related race or ethnic groups [2, 3]. These carriage patterns were functionally relevant, and genomic variation in microbial strains (gains, losses, and polymorphisms) underscored inter-individual variation in the microbiome. Taxonomic profiling associating both clades and metabolism with host covariates (namely age, gender, BMI, blood pressure, race and ethnicity, etc) demonstrated that most microbial variations are not well explained by examined clinical covariates other than race/ethnicity [2].

Race and ethnicity exert their effects through innate or genetically determined biologic mechanisms, and have broader implications with relation to socioeconomic status, diet habit, life style, etc. Therefore, it is not surprising to see the strong association of race/ethnicity with the microbiome. However, self-defined race/ethnicity is not always accurate and further complicated by secondary associations of race and ethnicity with diet, birth country, etc. Thus, further investigation down to the molecular level is essential to gain more knowledge on the underlying mechanism of association and to prevent potential misclassification bias.

One recent study has reported on both the commonality and the distinctions in the gut metagenome when compared among children and adults from rural Venezuela, Malawi, and the urban U.S. Of note, the study cohort was comprised of 531 subjects from a limited number of families (151). While relatively few distinctions in the gut microbiome were observed across all cohorts through the first 3 years of life, pronounced differences in the gut microbiome and functional gene repertoires were noted among geographic locations. However, there was no clustering observed among adult Malawians, Amerindians, nor regional U.S. populations. The investigators concluded that the host age/stage of development and geography served as primary determinants of the gut microbiome [6]. However, it bears mention that the potential for shared or divergent host ancestral genomic variation (mitochondrial or nuclear DNA) among their study cohorts was not investigated, and thus, this study may have failed to recognize an independent contribution of the host genome by examining likely homogenous populations (e.g., 34 Malawian and 19 Venezuelan families with a significant number of mono and dizygotic twin pairs). Moreover, when analyzed by the degree of familial relatedness (mono and dizygotic pairs, siblings, and unrelated children and adults) and cohabitation, significant UniFrac distance metrics was observed with greater heterogeneity. Similar studies have not been duplicated in larger genetically divergent population-based cohorts living in the same region with relatively common diets and exogenous exposures [7].

In contrast, there are several lines of evidence suggesting that the host genomic ancestry may structure the microbiome. A large murine advanced intercross line detected a core measurable microbiota (CMM) consisting of 64 conserved taxonomic groups. In this cohort, 13 murine genomic regions and five quantitative trait loci significantly associated with 26 CMM taxa at both the genus and species level of operational taxonomic unit (OTU) projections [8]. With respect to humans, previous studies have largely utilized twin and unrelated sibling pairs as surrogates for host genomic identity [7, 9, 10]. Turnbaugh et al [7] employed 16S rRNA based analysis and observed that the gut microbial community structures of adult monozygotic (haplotype identical) twin pairs had a degree of similarity comparable to dizygotic pairs, and only slightly more similar to their mothers. Of note, there was a roughly comparable degree of covariation between adult monozygotic and dizygotic twin pairs and deviations from this “core” gut microbiome were associated with adult obesity [7, 9, 10]. Total community DNA and RNA was relatively deep sequenced from one set of obese monozygotic twins, and comparative analysis indicated that the majority of species-level projections were shared with relatively significant variations in abundance [10]. These findings expanded upon earlier unrelated infant and dizygotic twin pair studies across the first year of life, which similarly demonstrated that a higher degree of host relatedness was associated with akin gut microbiome profiles [11]. However, these studies were limited to associations with gut microbiota, failed to consider whether the host nuclear or mitochondrial genome (which is solely maternally inherited) might be driving these associations, nor controlled for the potential impact of perinatal co-morbidities known to accompany human twin gestations [12]. This is of potential importance, as twinning in human is an independent risk factor for the development of common perinatal morbidities and mortalities as prematurity, growth discordance, and twin-twin transfusion syndrome, all of which are recognized to render risk of adult obesity [13–15]. Despite these potential limitations, such early studies at relatively lower microbial fingerprinting resolution collectively suggest that that the structure of microbial communities is in some part influenced by their human host’s genome.

Human populations can be divided into mtDNA haplogroups based on SNPs scattered throughout the mitochondrial genome, reflecting mutations accumulated by the maternal lineage. Although there is association between ethnic groups and haplogroups, mtDNA variants are most broadly representative of ancient ancestral roots from tens of thousands of years ago. As people migrated to form isolated groups, small changes accumulated over generations to set genetic diversity. Ergo, haplogroups may be regarded as broadly representative of genomic ancestry but are neither the sole nor definitive ancestral tags and may exist in linkage with other nuclear DNA markers [16, 17]. Given that race and ethnicity are delineated by haplogroup, and that mtDNA is more susceptible to DNA damage and acquires mutations across an individual’s lifetime at a higher rate than nuclear DNA, we hypothesized that mtDNA haplogroup and polymorphisms may be associated with variations in the human microbiome [2].

There are additional inherent characteristics of mtDNA, which make it an attractive candidate for host-microbiome association studies. Mitochondria derive from ancestral endosymbiont bacteria and have 16.5 kb circular double-stranded DNA molecules in multiple copies per cell (heteroplasmy). Consistent with their role as the generators of cellular ATP by oxidative phosphorylation, they play crucial roles in energy metabolism and apoptosis [18], and serve as the primary cellular source of reactive oxygen species [19]. The central role of mitochondrial proteins in cellular energy also makes mtDNA an ideal system for rapid human adaptation to new climate and dietary conditions [20]. The mtDNA genome encodes 13 protein-coding genes involved in respiration and oxidative phosphorylation, alongside two rRNAs and a complete set of 22 tRNAs that are important for protein synthesis (http://www.mitomap.org). Likely due to high levels of reactive oxygen species, lack of protective histones, and a limited DNA repair capacity, somatic mtDNA mutations (primarily in the D-Loop) are known to accumulate in individuals over time [21]. However, the majority of mtDNA variants are benign polymorphisms related to SNPs in the nuclear genome. The frequencies of these variants differ among populations and serve as the foundation of haplogrouping. The mutations in mtDNA have been related to a series of diseases related to neurological, muscular or metabolic disorder [22, 23]. Recent population genetics studies also indicate the association of mtDNA variants with complex human diseases, such as Alzheimer’s, Parkinson disease and cancer [24, 25]. Of interest to our studies, mitochondria have been recently suggested to play a pivotal role in the innate immune response [19, 21, 26–29].

In this study, we sought to provide initial evidence regarding the contribution of the host genome to its microbiome through robust interrogation of sequencing data and clinical metadata from the HMP [2–4]. Our aim was to determine whether there exists a significant association between haplogroups and mtDNA variants with microbiome taxonomic abundance and functional profiling.

By enabling concomitant analysis of host genomic variants with their respective microbiome community profiles, we can begin to address a number of crucial gaps in our understanding. In this study, we observed significant associations among both mtDNA haplogroups and mtSNPs with microbiota taxa alongside their carriage patterns. These studies serve as robust initial population-based interrogations into host genome-microbiome associations, and further clarify previously described microbiome-ethnicity associations at the molecular level.

First, mtSNPs were identified for individuals in the HMP cohort with high confidence. Haplogroups were identified for each individual based on mtSNPs. Given the fundamental role of the mitochondrial genome in cellular metabolism, evidence has already accumulated that different human mtDNA lineages are functionally different [48, 49]. mtDNA haplogroups are also thought to modify mitochondrial function because there is evidence that they are formed by adaptation to thermal environments [50] and subsequent natural selection [20]. It is clear that mitochondria haplogroup is more than a genomic marker. It reflects the ancestral difference in the human genome among the population. In publications from HMP consortium, association of clades and metabolic pathways with host properties, such as age, gender, BMI and race, have been preliminarily examined and reported [2, 3]. A wide variety of taxa and metabolic pathways were differentially distributed by virtue of subject ethnicity at limited body sites. Although age was shown to be associated with several metabolic pathways and one skin clade, most other metadata (BMI, gender, etc) are generally modest and non-significant [2]. Ergo, other factors, such as host genetics, may play roles in shaping the host microbiome profile [2]. We acknowledge as a potential limitation to our study that while we observed an association between certain mtDNA SNPs and haplogroups with community microbial profiles, these associations may be more broadly representative of ancestral differences, which are in linkage with the mitochondrial genome.

Of interest to our data, Ravel et. al.[41] previously reported on the relationship of racial or ethnic background to vaginal bacterial community composition. Using a cohort of self-sampled subjects, they have previously reported on the proportion of vaginal subtypes, which significantly varied among Caucasian, Asian, African American and Latino subjects. Similarly, other investigators with the MetaHIT have shown that stool enterotype 1 is observed in strong association with self-identifying Japanese subjects [40]. Given that mtDNA haplogroups are the maternally derived genomic markers of race/ethnicity, our results have confirmed and extended these observations of others to molecular level.

Nevertheless, having defined mtDNA haplogroups existing in significant association with specific taxa, mtSNPs across mitochondria genome were further tested to identify regions with higher association. We observed several findings of potential relevance.

T14798C is a non-synonymous SNP that encodes an amino acid substitution of phenylalanine (Phe) to leucine (Leu). This mtSNP maps to cytochrome b and was observed in strong association with differential abundance of Dialister in the Veillonellaceae family specifically in the vaginal posterior fornix. This is of potential biological importance as the posterior fornix of the vagina is most proximal to the cervix and uterus and Dialister has been found in both amniotic fluid and placental tissue of women with preterm premature rupture of the amniotic sac [51]. There is long-standing racial and ethnic disparity in the risk of preterm birth [52].

Turning to the gut microbiome, A13434G is a synonymous mtSNP on ND5 and T15784C is a synonymous mutation also on cytochrome b. Although synonymous mutations are usually referred to as “silent”, increasing evidence demonstrates significant effects on transcription, splicing, and mRNA transport or translation, all of which would alter phenotype [53]. Both of the above synonymous mutations were observed in strong association with Eubacterium (belonging to Clostridium cluster IV) and Roseburia (Clostridium cluster XIVa). Clostridial clusters IV and XIVa are highly oxygen-sensitive anaerobes and produce butyrate along the GI tract, which is excreted in feces [54]. Notably, decreased abundance in members of the butyrate-producing Clostridial clusters IV and XIVa have been reported in the gut of inflammatory bowel disease (IBD) patients, another prevalent human disease which displays variable risk by virtue of race and ethnicity [55]. Functionally extending these observations to the microbiota metabolic profile, several mtSNPs mapping to the mitochondrial encoded NADH dehydrogenase genes were observed in significant association with pathways in amino acid metabolism and protein secretion system from stool samples. It has long been debated whether amino acid derived molecules produced by intestinal bacteria affect host health by regulating host immunity and cell function, or alternately by varying microbial composition and community metabolism [56]. Our findings suggest that the host mtDNA variants in key redox pathways inherently define the gut microbiome, which in turn will structure their community and carriage patterns. This is consistent with long standing demonstrations that in microbiota, protein secretion transport effector molecules function from the interior to exterior. Moreover, protein secretion is known to play a central role in modulating the interactions of bacteria with their environments, particularly when symbiotic bacteria interact with host cellular constituents [57].

The primary role of the mitochondria is to produce energy for the cell. Overall, we observed more clades and metabolic pathways to occur in significant association with mtSNPs mapping to genomic loci encoding for electron transport. The mitochondrial respiratory chain, which is comprised of five multi-subunit protein complexes, carries out oxidative phosphorylation. During oxidative phosphorylation, mitochondrial enzyme complexes generate an electrical charge on either side of the inner mitochondrial membrane through transferring of electrons (membrane potential). This potential difference in electrical charge provides the energy for ATP production. NADH dehydrogenase genes encoded by mtDNA comprise complex I, which is responsible for the first step in the electron transport process. Mitochondrial cytochrome b is a subunit of complex III, which passes electrons to cytochrome c, which relays them to complex IV; three of protein subunits of complex IV are encoded by mitochondrial originated cytochrome c oxidase.

Our observed mutations on mitochondrial genes are known to disrupt normal activity of the electron transport chain, which will affect the production of ATP and increase the production of reactive oxygen species (ROS)[58]. ROS have important roles in cell signaling and homeostasis, but excessive amount of ROS could also cause significant damage to cell structures [59]. In human disease studies, oxidative stress is involved in the development of many diseases, including cancer [60], Parkinson’s disease, Alzheimer’s disease [61], and inflammatory bowel disease [62]. In addition to generating cellular energy, mitochondria are also involved in cell differentiation, cell death, as well as the control of cell cycle and growth [63]. It is also well established that the mitochondrial genomes are dynamic structures whose quantity and quality alter in response to cellular oxidative and metabolic demands. Although the mitochondrial genome is small, mtDNA encodes genes essential to perform the above functions and maintain symbiotic host-microbiome relationships.

In sum, the regulation of metabolic function is highly conserved in higher eukaryotes and is essential to the establishment of microbial communities. Although pathogenic mutations on mtDNA result in severe disease, milder mutations may have subtle phenotypic consequences. Ergo, the mtDNA variations among individuals in the HMP cohort may still render different microbiome profiles, possibly through inflammation response to different levels of ROS activity [58, 64]. Although the detailed mechanism awaits further analysis, individuals who have mtSNPs associated with haplogroups discovered in our study may show differences in their capability of energy metabolism and thus affect their own microbiome profile. We speculate that larger population-based and prospective analyses, which would enroll both healthy and disease-afflicted subjects, will provide further robust evidence for our initial associations proposed herein.

We have described our approach to leverage sequencing data from the HMP to demonstrate significant association between the human microbiome with host mtDNA variants (haplogroup and nucleotide polymorphisms). We observe that among stool and vaginal posterior fornix samples, several haplogroups and mtDNA variants show significant association with specific microbiota and their community structure traits. Delving deeply into both mtSNPs and metagenomic-derived carriage patterns, we have further shown that our observed associations between host and microbe are of likely functional relevance. Given the long-standing described role of mitochondria in cellular metabolism and oxidative stress, and emerging data describing its role in innate immunity, our findings may be of likely high significance. In sum, these data provide initial evidence for the host mitochondrial genome influencing the structure of its microbiome, and underscore the capacity for metagenomics to explore host-microbe interactions.