Although we know there is considerable variation in gut microbial composition within host species, little is known about how this variation is shaped and why such variation exists. In humans, obesity is associated with the relative abundance of two dominant bacterial phyla: an increase in the proportion of Firmicutes and a decrease in the proportion of Bacteroidetes. As there is evidence that humans have adapted to colder climates by increasing their body mass (e.g. Bergmann's rule), we tested whether Firmicutes increase and Bacteroidetes decrease with latitude, using 1020 healthy individuals drawn from 23 populations and six published studies. We found a positive correlation between Firmicutes and latitude and a negative correlation between Bacteroidetes and latitude. The overall pattern appears robust to sex, age and bacterial detection methods. Comparisons between African Americans and native Africans and between European Americans and native Europeans suggest no evidence of host genotype explaining the observed patterns. The variation of gut microbial composition described here is consistent with the pattern expected by Bergmann's rule. This surprising link between large-scale geography and human gut microbial composition merits further investigation.
There is growing appreciation that the gut microbial community (i.e. microbiota) may have played an important role in the evolution of host species . Although the gut microbial composition is quite distinctive between host species , there is also considerable variation within host species [2–7]. Nonetheless, how within-species microbial variation is shaped and why such variation exists remains largely unexplored.
One of the factors that is associated with gut microbial composition is host physiology [8–11]. For example, obese mice have distinct bacterial composition compared with lean mice characterized by relative abundance of two dominant phyla: increase in Firmicutes and decrease in Bacteroidetes . Lean mice transplanted with the microbiota from obese mice demonstrated increased fat storage without increased food consumption when compared with control mice . This suggests that relative increases in Firmicutes and decreases in Bacteroidetes can increase energy extraction and fat storage from a given diet. Interestingly, the pattern is consistent in humans where increases in Firmicutes and decreases in Bacteroidetes are associated with obesity  and also associated with increased energy harvest from the diet .
It seems plausible that an increase in energy extraction and fat storage may be more important for animals in colder regions compared with animals in warmer regions as an environmental adaptation to climate. It is generally accepted that humans follow ‘Bergmann's rule’  at least to some degree, whereby populations in higher latitudes tend to have larger body mass compared with populations in lower latitudes . We speculated that such large-scale, eco-geographical observations might be mediated in part by gut microbial communities that modulate energy extraction and fat storage. However, few studies have applied ecological theories to the geographical variation of gut microbial composition in humans or other animals across broad latitudinal scales. Here, we test whether the relative abundance of Firmicutes in the human gut increases with latitude and whether that of Bacteroidetes decreases (i.e. the pattern that might be expected given body mass variation with latitude) using six published human microbial studies [2–7], including more than 20 populations from a variety of geographical locations worldwide.
2. Material and methods
We searched human gut microbial studies that included population samples. We either requested the data for each individual used in the study from the original authors or acquired the data from MG-RAST . We were able to access six microbial studies in this way [2–7] including 23 population samples representing a total of 1020 healthy individuals (table 1; electronic supplementary material, table S8; figure 1). The relative abundance of Firmicutes (i.e. Firmicutes/total bacteria), the relative abundance of Bacteroidetes (i.e. Bacteroidetes/total bacteria), sex, age, sampling localities and bacterial detection methods were collected from the studies. Age was divided into five age classes: Z (<1), A (1–10), B (11–20), C (21–54) and D (>60). Latitudes were based on sampling locality of the population. The bacterial detection methods were divided into two categories: FISH-based or 16S-sequencing-based methods. Ethnicity information was also collected when available (see the full dataset in the electronic supplementary material, table S8). Spearman's correlations were used for all correlations and the Wilcoxon rank sum test was used for all pairwise comparisons.
3. Results and discussion
There was a highly significant positive correlation between Firmicutes abundance and latitude (ρ = 0.857, p < 0.0001) and a negative correlation between Bacteroidetes and latitude (ρ = −0.637, p = 0.001) using the population averages from all 23 population samples (see electronic supplementary material, figure S1). The pattern is robust even when considering individual data points (table 2 and figure 2). Although one might expect opposite correlations between the two phyla, as the majority of the human gut microbiome consists of Firmicutes and Bacteroidetes (i.e. if one phylum increases, the other phylum is likely to decrease), the correlation between Firmicutes and latitude is consistently greater compared with Bacteroidetes and latitude. This pattern can be explained either by taxa in Firmicutes mainly driving the pattern relative to Bacteroidetes or by taxa in minor phyla (e.g. Proteobacteria, Actinobacteria, Tenericutes, etc.) sharing some functional roles with Bacteroidetes. How individual bacterial taxa relate to the net effect of the microbiota remains an open question. Regardless, the correlations observed here are consistent with the pattern expected by Bergmann's rule  where increases in Firmicutes and decreases in Bacteroidetes are known to be associated with an increase in body weight  potentially owing to an increase in energy extraction and fat storage from a given diet [8,11]. However, the pattern might be explained not only by latitude but also perhaps by age, sex or bacterial detection methods. Therefore, we tested the potential factors that might be driving the observed pattern as described below.
First, it is unlikely that age is biasing the overall pattern. Although bacterial composition differed among different age classes (see electronic supplementary material, table S1), all of the significant correlations within all age classes were in the directions expected by Bergmann's rule except in more than 60 age class (see electronic supplementary material, table S2). Interestingly, elderly individuals showed significant correlations in the opposite direction for both bacterial phyla (see electronic supplementary material, table S2). However, elderly individuals (and infants) are known to have less stable and distinct gut microbial compositions compared with non-elderly adults ; we therefore analysed the data using not only all of the data (n = 1020) but also a subset of data representing non-elderly adults (age class C, n = 438; table 2), with similar results in each case.
Second, sex is apparently not biasing the results, because the relative abundances of two bacterial phyla between men and women are similar (see electronic supplementary material, table S3) and the correlation between the bacterial phyla and latitude remained significant within each sex (see electronic supplementary material, table S4).
Next, we tested potential biases between different microbial detection methods. The populations with FISH-based methods had higher Firmicutes and lower Bacteroidetes values compared with populations investigated with 16S-sequencing-based methods (see electronic supplementary material, table S5). This could cause a potential bias in our dataset, because FISH-based methods were concentrated in the higher latitudinal populations and 16S-sequencing-based methods were concentrated in the lower latitudinal populations (table 1). However, there were significant correlations within populations with 16S-sequencing-based methods alone (table 2 and electronic supplementary material, table S7). Although the Spearman's ρ values became weaker, the overall pattern cannot be explained by the potential method bias alone. This issue could in the future be resolved by characterizing gut microbial composition in European samples using consistent 16S-sequencing-based methods. Sampling an independent latitudinal transect in the Southern Hemisphere would also help to test this pattern in a more robust way.
Finally, to investigate whether there is an effect of host genotype on the variation of gut microbiota, we focused on four ethnic groups: native Africans (AF), native Europeans (EU), African Americans (AA) and European Americans (EA) [2,5,7]. All populations were studied using 16S-sequencing methods. Applying the concept of a common garden experiment, if there is a host genotype effect on gut microbial composition, we expect populations of African ancestry (i.e. AF and AA) to have similar composition to each other compared with populations of European ancestry (i.e. EU and EA) owing to genetic relatedness. The result did not support the prediction of host genotype effect: abundance of bacterial phyla was always different in native African populations (AF) from the other populations (i.e. EU, AA and EA; electronic supplementary material, table S7). Although quantitative measures of diet were unavailable from most populations, the pattern could potentially be explained by the effect of diet since at least AF-Burkina Faso had a low fat/high fibre diet and EU, AA and EA had high fat/low fibre diet or non-restricted typical western diet [2,5]. Also, climate and pathogen/parasite distribution might potentially explain the overall pattern. Dissecting the covarying factors of latitude may help identify the potential driver of gut microbial variation.
If gut microbial community is playing a role in Bergmann's rule (or, more generally, size or physiology differences between latitudes), there are three major models that could explain the observed pattern, and they are not mutually exclusive. (i) Host genes could maintain the observed microbial community variation as an environmental adaptation. For example, immune genes, such as antimicrobial peptides or behavioural genes, that can alter food preferences or amount of food intake could potentially regulate the microbiota. (ii) The environment itself could maintain the observed microbiota variation if a key environmental factor is covarying with latitude. For example, if food availability (e.g. food cultural variation in the case of humans) or microbial availability were correlated with latitude, environmental variation itself could explain the pattern without a fitness consequence of microbes to the host. (iii) Host plasticity could also maintain the observed microbiota variation. For example, cold environment increases food intake in animals . If different microbiota have different fitness consequence to the host (e.g. different microbiota add different amounts of fat), non-genetic vertical or horizontal transmission of microbes could maintain the observed pattern. Which of the models explains the observed variation of human gut microbiota remains an open question.
Although further investigation is necessary to characterize the geographical variation of gut microbial composition in humans, the robust pattern raises some interesting points in microbial ecology. (i) ‘Healthy microbiota’ may differ in different geographical regions. (ii) Independent latitudinal transects using consistent methods may help to identify environmental variables that shape gut microbial composition. (iii) Gut microbial composition could potentially help mediate the fit of host phenotype to its environment with or without a host genotype effect. (iv) We can generate novel hypotheses by understanding the geographical variation of host-associated microbial communities within species. Studying other species that show clinal variation in body size may help to establish whether the observations reported here are general.
We thank M. Blaut, D. Cavalieri, C. De Filippo, J. Gordon, C. Lay, S. Muller, L. Rigottier-Gois P. Turnbaugh, T. Wang and L. Zhao who generously provided individual data from their studies. We also thank M. Nachman for early discussions and comments on the manuscript.
- Received December 4, 2013.
- Accepted January 17, 2014.
- © 2014 The Author(s) Published by the Royal Society. All rights reserved.