Extensive profiling of the Korean gut archaeome
This study includes faecal samples collected from 897 healthy East Asian subjects living in South Korea [20]. Considering that approximately 50% of the human population possesses archaea in their gastrointestinal tracts [21], we first determined the presence of archaeal colonization in all samples using an archaea-specific primer set. The results showed that 381 out of 897 faecal samples (42.47%) were positive for archaeal colonization, and the positive samples were subsequently subjected to deep sequencing of the gut archaeome. In total, 275,909,328 reads were obtained from the Illumina HiseqTM X platform. After quality control, the remaining 193,370,457 reads (mean: 507,534 reads per sample; median: 240,156) were subjected to further analysis. The rarefaction analysis based on observed amplicon sequence variants (ASVs) showed that the sequencing depth had reached saturation (see Additional file 1: Supplementary Fig. S1a). Annotation of the archaeal 16S rRNA gene sequences to the SILVA database led to the prediction of 685 ASVs in the Korean gut archaeome (see Additional file 1: Supplementary Fig. S2).
The taxonomic classification of the gut archaeome revealed the predominance of the phylum Euryarchaeota, followed by the phylum Crenarchaeota (Figs. 1a and b). At the genus level, the Korean gut archaeome showed proportional abundance of sequences belonging to the methanogen group; the genera Methanobrevibacter and Methanosphaera in the family Methanobacteriaceae of the order Methanobacteriales were mostly proportionally abundant (54.89% and 25.68% relative abundance, respectively) with minor contributions from the genus Methanobacterium. In particular, the Korean gut archaeome contained haloarchaea-assigned sequences with 9.63% mean relative abundance; sequences belonging to the genera Halolamina, Haloplanus, Halorubrum, Halobacterium, Haloterrigena, Natronomonas, Halarchaeum, Haloarcula, Halonotius, and Halorussus were also detected. At the individual level, certain participants harboured the haloarchaea-dominated archaeal community structure (i.e., haloarchaea-assigned sequences showed 99.33% relative abundance; Fig. 1b). Next, we assessed the core ASVs, i.e., ASVs detected extensively in all faecal samples. The methanogen-assigned ASVs (e.g., ASV124, ASV066, and ASV130) were detected in > 97% of the total samples, whereas haloarchaea-assigned ASVs (e.g., ASV305 assigned to the genus Haloplanus) were detected in 95.53% of the samples (Fig. 1c), suggesting that members belonging to both methanogens and haloarchaea might be the soft core microbial components, i.e., over 95% detection rate but not 100%.
Abundance estimation of the human gut archaea
The abundances of bacterial cells (109 to 1011 bacteria g-1) in the human gut have been well described [22, 23]. We attempted to estimate the archaeal abundance by determining the archaea/(archaea + bacteria) ratios. We randomly selected 150 samples from the 381 archaeome-positive faecal samples and quantified both the archaeal and bacterial 16S rRNA gene copy numbers using real-time quantitative PCR. The results showed that the mean archaeal abundance was 9.89 ± 4.48% of the total bacterial and archaeal abundance (Fig. 2). The archaeal proportion varied largely among individuals, ranging from 1.22% to 22.55%.
Structural characterization of the Korean gut archaeome
One of the major features of the human bacterial microbiome is inter-individual variation commonly observed even in healthy individuals [24]. Our metataxonomic analysis revealed remarkable differences in the archaeal structure of individual Korean guts (Fig. 1). Hence, we assessed the structural features of the gut archaeome of 342 normal subsampled individuals (i.e., the number of sequences was evenly normalized at a sampling depth of 10,000 across the subjects; see Additional file 1: Supplementary Fig. S1b). Principal coordinate analysis (PCoA) of the weighted UniFrac distance matrix revealed that the proportional abundances of several key archaeal taxa, i.e., families Methanobacteriaceae, Haloferacaceae, and Methanomethylophilaceae, are the discriminant factors determining the distance between samples (Fig. 3a).
The gut bacterial communities in the human gastrointestinal tract have been partitioned into several clusters. Each cluster (enterotype) is overrepresented by a distinct set of bacterial genera [25]. As shown in the taxonomic and clustering analyses above, the community composition and structure of the Korean gut archaeome highlighted the importance of the several robust clusters that were prevalent across samples with different abundances. Next, we assessed the presence of enterotypes by applying partitioning around medoids (PAM) clustering analysis to the Bray-Curtis dissimilarity matrix generated from the family-level relative abundance profiles (Fig. 3b). The results showed four distinct clusters: Methanobacteriaceae as archaeal enterotype (MBA enterotype), Methanomethylophilaceae as archaeal enterotype (MMA enterotype), Haloferacaceae as archaeal enterotype (HFA enterotype), and the unclassified Euryarchaeota related to uncultured phylotypes in genus Methanosphaera or Haloplanus as archaeal enterotype (UEA enterotype; Figs. 3c and d). These suggested that the distinct set of overrepresented taxa (enterotypes) might be the consequence of a well-balanced symbiotic relationship between the host and archaea in the gut environment.
Diversity of haloarchaeal phylotypes in the human gut
Phylogenetic analysis of the methanogen-assigned sequences revealed three clusters of the methanogen phylotypes with the genera Methanobrevibacter (318 ASVs) and Methanosphaera groups (178 ASVs) and the family Methanomethylophilaceae group (44 ASVs; see Additional file 1: Supplementary Fig. S3). Studies on human gut archaeomes dominated by methanogens have been already described elsewhere [11, 16, 19, 26]. Therefore, we analysed the haloarchaea-assigned sequences, which have not yet been reported in the metataxonomic studies of the human gut archaeome. To evaluate the diversity of the haloarchaeal phylotypes in the human gut, we conducted a phylogenetic analysis based on the 139 haloarchaea-assigned ASVs with trimmed 16S rRNA gene sequences of the validated haloarchaeal species and public clonal sequences. We observed that the majority of haloarchaeal phylotypes are closely related to the genera Haloplanus, Halolamina, and Halorubrum (Fig. 4a). In particular, we observed three branched clusters of the haloarchaeal phylotypes, which were relatively distantly located from the validated haloarchaeal species. The Haloplanus subgroup consisted of 68 ASVs (Fig. 4b); Halolamina, 32 ASVs (Fig. 4c); and Halorubrum, 13 ASVs (Fig. 4d). We next assessed whether sequences assigned to haloarchaea occur in other sample cohorts by trawling the publicly available human metagenomic and metataxonomic datasets bases using the EBI MGnify. As shown in the Additional file 2: Supplementary Table S1, we found several studies possessing the metagenomic and metataxonomic sequences assigned to haloarchaea, implying that the human gut is capable of harbouring diverse and metabolically unknown haloarchaeal strains.
A positive correlation between the richness of the halophilic bacteria and faecal salinity has been recently reported [18]. To verify this association, we randomly selected 20 faecal samples with different proportional abundances of haloarchaea, i.e., 5.38 – 99.33% relative abundances, and measured faecal salinity using a salinity refractometer, yielding a mean salinity of 0.68%, ranging from 0.30–1.05% (see Additional file 1: Supplementary Fig. S4a). Correlation coefficient and linear regression analyses revealed no association between the relative abundance of haloarchaea and faecal salinity. Further, we analysed the inorganic elements mainly consisting the salt (e.g., sodium, potassium, magnesium, and calcium) of 20 selected samples using an inductively coupled plasma-mass spectrometer (ICP-MS). On average, each inorganic element possessed less than 1% of total faecal weight: 0.04%, 0.40%, 0.24%, and 0.50% for sodium, potassium, magnesium, and calcium, respectively (see Additional file 1: Supplementary Fig. S4b). Similar with the salinity data, no positive nor negative relationship was found between the relative abundance of haloarchaea and the relative amount of faecal inorganic elements.
Detection of haloarchaea in the human gut by fluorescence in situ hybridization
We attempted to verify the presence of haloarchaea in the human gut by a non-sequence-based approach using the FISH analysis. We designed a haloarchaea-specific probe (HALO775) and tested the specificity of the oligonucleotide probe. Neither the undesired match (i.e., in silico binding of the HALO775 with non-haloarchaeal taxa, such as methanogen, bacteria, and eukaryotes; see Additional file 2: Supplementary Tables S2 and S3) nor the unspecific binding of the HALO775cy3 with the cultured bacterium (Escherichia coli K12) and other archaeon (Methanobrevibacter smithii JCM 30028T) was found (Fig. 5a). A positive signal of the HALO775cy3 was only observed with the cultured haloarchaeon (Haloplanus salinus JCM 18368T). Based on both the total archaeal abundance and the metataxonomic data, we selected three faecal samples that possess a different proportion of the haloarchaea-assigned sequences (sample J0885, J0111, and J0054; Fig. 2). The FISH analysis successfully detected a positive signal for haloarchaea from the selected samples (Fig. 5b). These results collectively suggest the presence of haloarchaea in the human gut, which was robustly detected by sequence-based (i.e., metataxonomics) and non-sequence based (i.e., FISH) methods.
Correlation analysis of the gut archaeal profiles with host factors
We next evaluated the effect of host factors on the community structure of the Korean gut archaeome by conducting correlation coefficient analysis using two variables: relative abundance of the proportionally abundant archaeal taxa (> 0.1% of the mean abundance) at the genus level and host factors, including dietary nutrients, clinical phenotypes, and food categories. Details on the participants, faecal collection, and DNA extraction from faecal samples were previously described [20]. We observed strong negative/positive correlations (Spearman’s rank correlation analysis, P < 0.05) between several archaeal taxa and dietary nutrients: the genus Halorubrum correlated negatively with calcium, phosphorus, iron, potassium, vitamin A, vitamin B2, vitamin B6, vitamin C, folate, carotene, and fibre levels; the genus Methanosphaera correlated positively with all elements of the dietary nutrients (Fig. 6 left). In host clinical phenotypes, the genera Halorubrum and Halobacterium showed significantly negative correlation (P < 0.01 and P < 0.05, respectively) with several lipids, i.e., total cholesterol (TotalC) and low-density lipoprotein cholesterol (LDLC). In addition, we observed a positive correlation (P < 0.05) between the genus Halolamina and renal functions (estimated glomerular filtration rate using the MDRD formula, eGFR_MDRD, and estimated glomerular filtration rate using the CKD-EPI formula, eGFR_CKDEPI; Fig. 6, right). We also found several positive/negative correlations of the haloarchaeal taxa with food categories (see Additional file 1: Supplementary Fig. S5). However, these results implied no common (dietary) factor influencing the abundance of haloarchaea in the human gut.
Comparison of the human gut archaeome with the gut archaeomes of the great apes
We observed the assortative characteristics of the four distinct enterotypes (Fig. 3d). Because host diet might affect the formation of enterotypes [27], the archaeal enterotypes may have arisen only recently. However, Fig. 6 and Additional file 1: Supplementary Fig. S5 show weak (or no) relationship between archaeal abundances and host dietary factors. Next, we assessed whether these enterotypes are the results of more ancient features (such as host immune system and gut physiology) that might be derived before or during the diversification of the great ape species by comparing the Korean gut archaeome with other human gut archaeomes [26, 28] and those of the great apes, including orangutan, gorilla, chimpanzee, and bonobo [29]. PCoA of the Bray-Curtis dissimilarity matrix showed the typical horseshoe shape of all merged samples (see Additional file 1: Supplementary Fig. S6a), indicating high dissimilarity among the gut archaeomes of the great apes. Comparison of the human and non-human samples revealed that weighted PCoA (see Additional file 1: Supplementary Fig. S6a) and unweighted PCoA (based on the Jaccard dissimilarity matrix; see Additional file 1: Supplementary Fig. S6b) showed significantly separated community structure and composition of the gut archaeome, respectively (permutational multivariate analysis of variance [PERMANOVA], P = 0.001, comparison between the human and non-human samples). UPGMA clustering analysis showed that the community structure of the gut archaeome of the great apes (except those of humans) appeared to mimic the host phylogeny, as shown by the absence of any significant difference in the distance between chimpanzee and bonobo gut archaeome (see Additional file 1: Supplementary Fig. S6c). Neither the community structure nor the composition of the human gut archaeome was closely related to those of chimpanzees or bonobos (see Additional file 1: Supplementary Fig. S6c and d), suggesting that the human gut archaeome is distinct and not related to host phylogeny.