Genetic isolation and metabolic complexity of an Antarctic subglacial microbiome

Abstract

Microbes inhabiting and evolving in aquatic ecosystems beneath polar ice sheets subsist under energy-limited conditions while in relative isolation from surface gene pools and their common ancestral populations of origin. Samples obtained from beneath West Antarctic Ice Sheet (WAIS) allowed us to examine evolutionary relationships of and identify metabolic pathways in microbial genomes recovered from the Mercer Subglacial Lake (SLM) ecosystem. We obtained 1,374 single-cell amplified genomes (SAGs) from individual bacterial and archaeal cells that were isolated from samples of SLM's water column and sediments. These genomes reveal that a diversity of microorganisms including Patescibacteria exists in SLM. Comparative analyses show that most genomes correspond to new species and taxonomic groups, with phylogenomic and functional evidence supporting their genetic isolation from marine and surface biomes. Genomic data reveal diverse metabolisms in SLM that are capable of oxidizing organic and inorganic compounds via aerobic or anaerobic respiration. Distinct metabolic guild structures are observed for the subglacial populations, where trophic shifts from organotrophy to chemolithotrophy may depend on oxygen availability. Our SAG data suggest versatile metabolic capabilities in the characterized microbial assemblage, reveal key energy-generating strategies in the subglacial aquatic ecosystem, and provide a framework to assess microbial evolution beneath WAIS.

Fig. 1. Location and physicochemical characteristics of Mercer Subglacial Lake (SLM).

a Geographical setting of SLM beneath West Antarctic Ice Sheet (WAIS), showing its position relative to glacial ice streams and the grounding line. The background satellite image is from the MEaSUREs MODIS Mosaic of Antarctica 2013-2014 (MOA2014) Image Map, Version 1 (NSIDC). Subglacial lake polygons (navy) are based on the Antarctic Active Subglacial Lake Inventory from ICESat altimetry (NSIDC). The light blue lines indicate the subglacial water flow paths beneath WAIS, and the grounding line (light grey line) is derived from MEaSUREs Antarctic Grounding Line from Differential Satellite Radar Interferometry, Version 2. Image/photo courtesy of the National Snow and Ice Data Center, University of Colorado, Boulder. Subglacial lake names: Upper Subglacial Lake Conway (USLC), Subglacial Lake Conway (SLC), Subglacial Lake Whillans (SLW), Lake 8 (L8), and Lake 10 (L10). b Physicochemical properties of SLM lake water, including lake depth, surface area, overlying ice thickness, basal ice temperature, and concentrations of key nutrients^,,,,.

Fig. 2. Phylogenomic trees of bacteria (1367 SAGs) and archaea (7 SAGs) of Subglacial Lake Mercer.

Two maximum likelihood trees were reconstructed using RAxML from the concatenated GTDB protein sequence alignment (5035 aligned sites) of 120 bacterial marker genes and from that (10,100 aligned sites) of 53 archaeal marker genes, respectively. The bacterial tree was rooted by the most recent common ancestral node of Patescibacteria. The archaeal tree is unrooted. From the innermost to the outmost, colors indicate the most abundant genera, phyla, the depth of sediment, and microbial cell sizes of the water column, respectively. Bootstrap support values were indicated on the nodes using black, grey, and open circles. An asterisk indicates the unclassified genus Burkholderaceae incertae sedis. Source data are provided as a Source Data file.

Fig. 3. Genetic isolation of SLM’s SAGs.

a A phylogenomic tree reconstructed using the concatenated GTDB protein sequence alignment that represents 122 SLM SAGs and 81 GTDB species representative genomes belonging to the family Gallionellaceae. The tree was rooted by an outgroup affiliated with the family SG8-39 of the order Burkholderiales. The 107 out of 108 SLM SAGs belonging to the genus Nitrotoga are cohesively clustered (see a circle in pink). The Shimodaira-Hasegawa support values for node confidence are presented next to the branches. b A histogram representing genomic average nucleotide identity (ANI) between SAGs and their closest GTDB genomes. The dotted vertical line at 95% sequence identity separates known and unknown bacterial species. When less than 10% of the SAG sequence aligned with any GTDB genomes, the SAG was considered a’No match‘. The ANI values for the 1374 SLM SAGs in orange and for 64 SAGs isolated from seawater beneath Ross Ice Shelf (RIS) in blue. c Histograms represent sequence similarity between query proteins and BLAST top hits of the NCBI NR database. The query proteins were derived from 537,441 and 45,895 orthologs that were clustered from the 1374 SLM SAGs (distribution in orange) and from the 64 RIS SAGs (distribution in blue), respectively. Two overlapped vertical bars on left represent the number of orthologs without significant BLAST hits under the E-value threshold 10. Source data are provided as a Source Data file.

Fig. 4. Metabolic potential inferred from SLM SAGs.

The proportion of SLM SAGs involved in a metabolic reaction were calculated to correct for incomplete genome recovery and correlate with the arrow thickness. Note that proportion values exceeding 100% are due to the presence of multi-copy genes in the SAGs. The upper and bottom sections represent metabolic potential of the water column (background in light blue) and sediments (in light brown), respectively. Details on the enzymes involved in the pathways are provided in Supplementary Data 6-7.

Fig. 5. Taxonomic distribution and network representation of metabolic potential.

a The darker the circle shading, the higher the proportion of SAGs responsible for a metabolic potential. The SAGs in a genus were divided into two sets: the water column- (blue horizontal bar for # of SAGs) and sediment (red horizontal bar for # of SAGs). The sets with <9 SAGs were omitted in this figure to simplify presentation. Networks were constructed using co-occurrence of every possible pair of metabolic genes across the SLM SAGs. The upper and bottom networks were derived from SAGs of the water column (b) and sediment (c). The node size and edge thickness correlate with the proportion of metabolic potential and with the inverse magnitude of the hypergeometric E-value for the co-occurrence, respectively. Note that edges are shown only when their E-values are less than 0.1. The networks were visualized using CoNet and ClusterViz in Cytoscape. Further details on the chemical reactions of the metabolic pathways are provided in Supplementary Data 6 and 7.

Fig. 6. Schematic representation of metabolic potential for Patescibacteria in SLM.

The presence or absence of metabolic pathways was determined based on KEGG annotations and the criteria outlined in Supplementary Data 6 and 7. Green-filled sectors represent pathways detected in one or more SAGs of the genus UBA1550, which is the most abundant genus of Patescibacteria in SLM. Red-filled sectors indicate pathways present in one or more SAGs of the genus C7867-001, the second most abundant genus. Blue-filled sectors denote pathways found in one or more genomes of the remaining Patescibacterial SAGs. Unfilled circles indicate pathways that were not detected in any patescibacterial SAGs. An inset provides a detailed view of the non-oxidative pentose phosphate pathway, spanning from fructose-6-phosphate to xylulose-5-phosphate. Sedo-7P stands for sedoheptulose 7-phosphate. Source data are provided as a Source Data file.