In the present study, we report that not only the predominant bacterial genera (Fig. 1e), but also their interactions between metadata covariates were common (Fig. 2c) in two independent cohorts and seemed to show the general characteristics of healthy gut microbiota in the Japanese. First, we investigated the overall structural characteristics by using enterotype analysis. Although one of the dominant genera, Faecalibacterium in the NIBIOHN cohort and Bifidobacterium in the MORINAGA cohort, seemed to be different between the two cohorts (Fig. 1a, 1b), integrated data indicated that these genera had a similar third directionality on gut microbial variations (Fig. S2c). Because the directionalities of the two genera did not exactly correspond, the third directionality was not chosen as an enterotype. Taken together with a previous report analyzing the relationship between enterotypes and assemblages [43], the gut microbiota could be divided into three clusters.
To clarify the features of gut microbiota in Japanese populations in detail, we focused on the dominant genus in each cohort, and showed 13 genera were common in the two cohorts (Fig. 1e). Compared with the dominant genera in a dataset containing a total of 2,186 North American and European individuals [44], most of our dominant genera were common, but the average relative abundance varied. The characteristic differences in abundance were that of Bifidobacterium, which was 4.4% in the NIBIOHN cohort, 8.4% in the MORINAGA cohort, while 1.4% in the previous report [44], and that of Blautia, which was 4.7% in the NIBIOHN cohort, 6.5% in the MORINAGA cohort (Fig. 1e), while 2.9% in the previous report [44]. The higher abundance of Bifidobacterium and Blautia is a characteristic feature of gut microbiota in Japanese individuals [32]. Altogether, the dominant genera seemed to be common regardless of the population, however, the abundance of genera showed population-specific features.
There were some cohort-specific features in gut bacteria (Fig. 1c) and alpha diversity indices (Fig. 1d) in the two cohorts, even though the same Japanese population. It was suggested that these differences between the two cohorts were due to subjects having different backgrounds as indicated in Table S2. Together with a previous study [38], which reported that consumption of dairy products increases Bifidobacterium abundance in Japanese population, the higher intake of normal fat dairy products in the MORINAGA cohort compared to the NIBIOHN cohort (Table S2) was thought to be one of the reasons for a significant difference in Bifidobacterium abundance. Most of the subjects in the MORINAGA cohorts were customers who purchased products from the Morinaga Milk Industry Co., Ltd.
The comparison of results by envfit analysis between the two Japanese cohorts revealed 18 common covariates, showing significant association with structural variations in microbiota (Fig. 2a). Seven common covariates including BSS, height, weight, BMI, alcohol consumption (beer, rice wine, and low alcohol liquor), age, and gender, were in agreement with previous reports of three different cohorts in Belgium [29], Netherland [30] and China [31]. It appears that regardless of the population, these seven covariates commonly associate with structural differences in gut microbiota. Significant common covariates involved in microbial variation in the three cohorts [29–31] of previous studies, such as smoking, disease state, medication, fruit and meat consumption, were not observed in the NIBIOHN and MORINAGA cohorts, while the significant association of cake and fatty fish consumption was not found in previous cohorts [29–31]. These differences may be attributed to distinctive dietary habits between populations. For instance, Japanese population show the highest intake of seafood in the world [45], which may be a contributing factor to gut microbial variations. Although Falony et al. [29] revealed that medication resulted in the largest total variance and interacted with other covariate-microbiota associations, the target of our study was different and the effects of medication showed no significant association with healthy gut microbiota.
A forward stepwise redundancy analysis showed a remarkable cumulative effect size on community variation, specifically 10.7% in the NIBIOHN cohort and 7.1% in the MORINAGA cohort (Fig. 2c); effect size in both Japanese cohorts were comparable to 7.7% reported in Belgian population [29]. These results indicated that the proportion of gut microbial variation explained by questionnaire-based covariates seemed to be approximately 10% regardless of the population or number of covariates, suggesting there were several additional intrinsic or extrinsic contributors such as immunity, host genetics, bacterial-bacterial interaction, as well as unknown factors [46]. Non-redundant determinants such as BSS, gender, and age were common in our two cohorts (Fig. 3c), in accordance with the previous report involving a Belgian cohort [29], indicating that these covariates are common among healthy populations independent of factors like country of origin. Falony et al. [29] reported that stool consistency showed the largest effect size on bacterial variations; BSS score for gut_stool_shape showed a higher non-redundant effect on total composition variation in Japanese cohorts (Fig. 2c). Gender [15] and age [9] intricately associated with life style factors such as dietary habits, therefore, the selection of gender and age as non-redundant covariates was justified. Interestingly, non-redundant analysis also showed population- or cohort-specific results. Beer consumption and related covariates were identified as non-redundant covariates in each cohort (Fig. 2c), presumably because beer was the most popular alcoholic drink in the Japanese cohorts (Table S2). This result is in line with a previous study [47] stating that alcohol affects the composition of gut bacteria.
Another cohort-specific feature with a high effect size was observed to be residential area in the NIBIOHN cohort, whereas no association was observed in the MORINAGA cohort (Table S3). The distribution of residential area was extremely different between the two cohorts; NIBIOHN cohort comprised of individuals from a limited area with a high concentration of subjects in each area (Fig S2), and MORINAGA cohort comprised of individuals from a dispersed and continuous area. Notably, a previous report, involving a 16S rRNA gene analysis of fecal samples collected from 516 healthy Japanese adults residing in various regions of Japan [33], was very similar to the MORINAGA cohort, and demonstrated no association between residential area and gut microbiota variation. However, the residential area was found to have the highest influence on gut bacterial variance in the NIBIOHN cohort. Further large-scale nationwide cohort studies are required to understand the effect of residential area on total gut microbial variance among Japanese populations.
Besides the common interactions of covariates and compositional variation, some associations between gut bacteria and covariates were also common between the two cohorts (Table S4). The negative association of BSS score with Alistipes was in accordance with a previous report [29], whereas the negative association with Bifidobacterium was not previously reported. Furthermore, the reported association of Christenellaceae, Mogibacteriaceae, and Rikenellaceae with bowel movement frequency in the Japanese population [33] was not observed in our study. Interestingly, regardless of the influence of BSS score and frequency of bowel movement on colon transit time, both covariates showed non-redundant associations with inter-individual gut microbial variation. This difference was represented by the fact that Subdoligranulum and Blautia only associated with defecation frequency. This difference highlighted the complex association between colon transit time, gut microbiota, and diet [27]. In relation to the common association of gender and gut bacteria, the higher abundance of Bifidobacterium in women and Prevotella_9 in men (Table S4) was in agreement with a previous report of Japanese gut microbiota [34], indicating the possibility of population-specific results. In contrast, the higher abundance of Faecalibacterium and Alistipes in women (Table S4) is first reported in the present study. The common association of age with the abundance of Blautia, Parabacteroides, and Roseburia was not shown in the previous report [29] and seemed to be population-specific. A decrease of Bifidobacterium abundance with age in the NIBIOHN cohort was in accordance with previous reports [8, 29], identifying Bifidobacterium as an adult-enriched bacteria. However, in the MORINAGA cohort, Bifidobacterium showed no significant association with age. One of the reasons for this discrepancy was presumed to be the unique dietary habits, which was high consumption of dairy products, and the high relative abundance of Bifidobacterium in the MORINAGA cohort.
In this study, we reported the characteristics of gut microbiota and comprehensively examined the major microbiome-associated variables in Japanese gut microbiota using two independent large-scale cohort data. The comparison between these large, distinct cohorts provided reliability and robustness to our study. One of the limitations of our study was that it was a cross-sectional study and did not show the causal relationships between gut microbiota and metadata. Prospective observational or interventional studies would be required to delineate these relationships. Another limitation was that all participants were applicants, which may be a source of potential bias in this study. Random sampling of subjects from all over the country or a larger sample size would be essential for overcoming this bias.