Diverse microbiome functions, limited temporal variation and substantial genomic conservation within sedimentary and granite rock deep underground research laboratories

Abstract

Background: Underground research laboratories (URLs) provide a window on the deep biosphere and enable investigation of potential microbial impacts on nuclear waste, CO₂ and H₂ stored in the subsurface. We carried out the first multi-year study of groundwater microbiomes sampled from defined intervals between 140 and 400 m below the surface of the Horonobe and Mizunami URLs, Japan.

Results: We reconstructed draft genomes for > 90% of all organisms detected over a four year period. The Horonobe and Mizunami microbiomes are dissimilar, likely because the Mizunami URL is hosted in granitic rock and the Horonobe URL in sedimentary rock. Despite this, hydrogen metabolism, rubisco-based CO₂ fixation, reduction of nitrogen compounds and sulfate reduction are well represented functions in microbiomes from both URLs, although methane metabolism is more prevalent at the organic- and CO₂-rich Horonobe URL. High fluid flow zones and proximity to subsurface tunnels select for candidate phyla radiation bacteria in the Mizunami URL. We detected near-identical genotypes for approximately one third of all genomically defined organisms at multiple depths within the Horonobe URL. This cannot be explained by inactivity, as in situ growth was detected for some bacteria, albeit at slow rates. Given the current low hydraulic conductivity and groundwater compositional heterogeneity, ongoing inter-site strain dispersal seems unlikely. Alternatively, the Horonobe URL microbiome homogeneity may be explained by higher groundwater mobility during the last glacial period. Genotypically-defined species closely related to those detected in the URLs were identified in three other subsurface environments in the USA. Thus, dispersal rates between widely separated underground sites may be fast enough relative to mutation rates to have precluded substantial divergence in species composition. Species overlaps between subsurface locations on different continents constrain expectations regarding the scale of global subsurface biodiversity.

Conclusions: Our analyses reveal microbiome stability in the sedimentary rocks and surprising microbial community compositional and genotypic overlap over sites separated by hundreds of meters of rock, potentially explained by dispersal via slow groundwater flow or during a prior hydrological regime. Overall, microbiome and geochemical stability over the study period has important implications for underground storage applications.

Keywords: Granite; Groundwater; Metagenomics microbiome; Sedimentary rocks; Stability; Underground research laboratory.

Fig. 1

a Map of the Mizunami and Horonobe URL locations in Japan. b Layout of boreholes in shafts and galleries in the Mizunami URL and (c) the Horonobe URL

Fig. 2

Phylogenetic tree of 956 representative sequences of ribosomal protein S3 from the Horonobe and Mizunami samples, along with reference sequences. Orange dots indicate individual sequences with < 70% rpS3 amino acid identity to sequences in NCBI. Numbers in squares show the average rank abundance of each phylum (or class in the case of Proteobacteria). Blue squares indicate organisms from the Horonobe URL and pink squares indicate samples from the Mizunami URL

Fig. 3

Overview of URL microbial diversity based on the 15 most abundant organisms in each sample, classified mostly at the phylum/class level. Samples are listed in order of increasing the distance between the sampling site and the closest access tunnel to seek evidence of perturbation due to the presence of the tunnel. (a) 7 Mizunami 0.2 µm-filter URL samples collected between 2014 and 2015. (b) 18 Horonobe URL 0.2 µm-filter samples collected between 2013 and 2016

Fig. 4

Phylogenetic tree for CPR bacteria constructed using ribosomal protein S3 sequences. Pink and blue shadings indicate sequences from the Mizunami and Horonobe URLs, respectively. Green branches indicate lineages with sequences from both the Mizunami and Horonobe URLs. The numbers after M (Mizunami) and H (Horonobe) indicate the number of sequences in each named lineage. Brown highlighted sequences are reference sequences. The long branches without color indicate Archaea, which were used as the outgroup

Fig. 5

Detection (blue bars) of organisms (columns) in samples (rows) listed in approximate order of decreasing distance from the access tunnels. Organisms lacking genomes are indicated by a dark gray box in the Organisms bar. (a) Almost one quarter of all organisms detected within the Horonobe URL were present in > 25% of the samples. 90.389.2% of all organisms are represented by draft genomes. Organisms lacking genomes were all detected in ≤ 3 samples. (b) Within the Mizunami URL, 45% of organisms were detected in at least 25% of the samples and 10% of all organisms detected were present in > 70% of the samples. 92.0% of organisms are represented by draft genomes. All organisms lacking genomes were detected in just one sample. For details see Table S3

Fig. 6

Key metabolisms across the Mizunami and Horonobe URLs. Presence/absence of each metabolic pathway based on the occurrence of indicative marker genes annotated with KEGG Orthology using Kofamscan (y-axis) in each of the recovered genomes (x-axis). The URL relative abundance (%) shows the proportion of the 265 Horonobe and 225 Mizunami genomes with that metabolism

Fig. 7

Comparison using the dereplicated genome sets from five genomically well sampled terrestrial subsurface ecosystems to seek instances where two ecosystems are similar or significantly different from each other. The highest scoring pairwise hit for each of 1472 sequences from genomes was assigned to an ecosystem comparison category and the aa ID (%) inventoried. For example, when all 503 sequences from Crystal Geyser were compared to all sequences from the four comparison datasets, there were 59 instances where the closest sequence was found in the Horonobe dataset. The 59 aa ID % values are represented by column 1. Of the comparisons, those that were significantly different are Horonobe compared to northern California (NorCal) vs. Rifle (p value of 1 × 10^–3), Crystal Geyser compared to Rifle vs. NorCal (p value of 5 × 10^–3), and NorCal compared to all other ecosystems (p < 1 × 10^–5. For details see Table S5