Structure and diversity of the polar microbiota
Sampling locations are shown in Fig. 1. A total of 39 Arctic seawater and 21 Antarctic seawater samples were collected for this project. Based on the sample depth, the samples were further divided into four groups: Arctic-Surface (0-100 m), Arctic-Deep (200-4000 m), Antarctic-Surface (0-100 m), and Antarctic-Deep (200-4000 m). Correspondingly, 60 metagenomes were obtained after Illumina sequencing. 16S rRNA gene sequences were extracted from the metagenomes for analysis and are subsequently referred to as 16S miTags. Classification of OTUs at 97% similarity resulted in a total of 24,504 OTUs. Classification of OTUs at the phylum level (class level for Proteobacteria) (Fig. 2) revealed 90 taxa, with a dominance by Alphaproteobacteria and Gammaproteobacteria in most of the metagenomes, followed by Bacteroidetes. In addition, Deltaproteobacteria, Actinobacteria, Cyanobacteria, Crenacrchaeota, OD1 (Parcubacteria), and SAR406 (Marinimicrobia) showed high relative abundances in certain metagenomes. When the OTUs were classified at genus level (Additional file 1: Figure S1), 2,121 genera were obtained, dominated by an unnamed Pelagibacteraceae group. A total of 12.63 ± 0.06% of the reads could not be classified to genus level.
To examine the similarity between the Arctic, Antarctic and the non-polar seawater microbial communities, 55 surface seawater (5 m) and 19 deep seawater metagenomes (100-1000 m) were downloaded from the Tara Oceans project [1] (http://ocean-microbiome.embl.de/companion.html) for comparison. Alpha-diversity analysis (Fig. 3a), based on rarefied 16S miTags (10,000 miTags per metagenome), revealed an overall higher Chao1 diversity, Shannon diversity, and Simpson diversity for the polar microbial communities; the alpha-diversities in the deep seawater samples tended to be higher than those in the surface seawater. Principal coordinates analysis (PCoA) (Fig. 3b) using rarefied 16S miTag data revealed clear separation of the polar from the non-polar seawater microbiota, especially for the surface microbiota, with PCo1 explaining 25.14% of the variability. Interestingly, the Arctic-Surface and the Antarctic-Surface seawater were clustered together, as were the Arctic-Deep and Antarctic-Deep seawater (Fig. 3b).
As found in previous studies based on 16S rRNA gene amplicon sequencing [23,24,25,26], the microbiota of Arctic and Antarctic seawater are more similar to each other than to seawater from other regions, and the dissimilarity between polar and temperate communities is more marked at the surface than in deeper water. Consistent with these findings, metagenomic analyses performed in the present study revealed a high similarity between microbiota in the two polar regions, which is likely to be attributed to environmental filtering (effect of environmental variables on community composition) and microbial dispersal, as the environmental conditions in the Arctic and the Antarctic are more similar to each than to the temperate regions. Higher diversity of microbes in deep polar seawater than in surface seawater was observed, implying that the deep waters are more of a reservoir of microbial species. Moreover, a higher similarity between the deep waters of temperate and polar area waters than between the surface waters of polar and temperate areas were found, which can be explained by the fact that surface communities have a high phylogenetic turnover rate [27] and minimal microbial connectivity by major currents [28], whereas the biogeography of deep water communities is largely controlled by ocean circulation [29].
Taxonomic and functional specificity
The 60 polar metagenomes were combined and the overlapping OTUs between all the polar seawater microbiota and all the Tara seawater microbiota (n = 243) were analyzed. The reads were mapped back to the OTUs; of the OTUs with more than 2 reads, 7520 were specific for the polar seawater; of the OTUs with more than 10 reads, 5176 were specific for the polar seawater; and of the OTUs with more than 50 reads, 3221 were specific for the polar seawater; this exceeded the number of OTUs specific for the Tara Oceans seawater (Fig. 4a). The specificity of the polar seawater (i.e., the ratio of polar-specific OTUs to the total number of OTUs present in the polar and non-polar seawater) consistently increased, whereas the specificity of the Tara Oceans seawater decreased with the number of minimum sequences possessed by the OTUs used for examination (Fig. 4b).
When the functions of the polar seawater microbiota were analyzed, a total of 16,638,499 orthologs were derived from the 60 metagenomic assemblies, resulting in a polar marine reference gene catalog (PM-RGC). In comparison with the OM-RGC using BLASTp, 3,903,052 (23.46%) of the PM-RGC orthologs were specific (e value > 1e-7 or similarity < 40%) (Fig. 4c). The orthologs specific for OM-RGC or PM-RGC were further searched against the NCBI-Nr database. It was found that more than 97% of the PM-RGC specific orthologs could not be identified (e value >1 e-7 or similarity < 40%) in the NCBI-Nr database, while about 61% of the OM-RGC specific orthologs had no hits (e value > 1e-7 or similarity < 40%) in NCBI-Nr (Fig. 4d).
The OM-RGC and PM-RGC specific orthologs were annotated using the COG database and compared with each other. While the PM-RGC orthologs received 3650 hits in the COG database, the OM-RGC orthologs had 4711 COG hits. Statistical analysis, based on the COG relative abundances (proportion of each COG in the total number of COGs), revealed that 2921 had significantly changed COGs (chi-squared test, p value < 0.05). The 40 most abundant COGs significantly enriched in PM-RGC are shown in Fig. 5. These COGs included functions for lipid and sugar metabolism, such as rarelipoprotein B (COG2980), CelD involved in cellulose biosynthesis (COG5653), and sugar phosphate permease (COG2271). Functions for cell membrane biosynthesis, such as the membrane proteins COG4291 and COG4648 without known functions, and membrane proteins involved in dissulfite bond formation (COG5061), were enriched in PM-RGC.
The 16S miTags extracted from the Tara Oceans metagenomes mapped to a total of 35,650 OTUs, with the rate of new gene detection at 0.01% by the end of sampling [1]. However, by sampling both surface and deep polar seawater, a large reservoir of microbial OTUs and functional genes were found that did not overlap with the Tara Oceans datasets. Overall, analyses of the polar seawater metagenomes increased the previously reported microbial diversity by more than 20%. Of the protein-coding genes derived from the polar microbiomes, between 20 and 40% showed a similarity with the OM-RGC generated by the Tara Oceans datasets. More importantly, 97% of the PM-RGC specific orthologs have unknown functions, and to our knowledge this is the first time to evaluate the overall functional specificity of polar seawater microta based on global seawater datasets. These findings suggest that there is an underestimation of the microbial taxonomic and functional diversity in the global oceans and the existence of specific and function-unknown genes.
Microbes enriched in polar microbiomes and their genomic features
To further explore novel functional potential of the polar microbiota through genomic analysis, 214 microbial metagenome assembled genomes (MAGs) (> 80% completeness and < 2% potential contamination) from the 60 polar metagenomes were recovered. These microbes belonged to 24 different microbial, including Alpha-, Beta-, Gamma-, and Deltaproteobacteria, Bacteroidetes, Actinobacteria, Chloroflexi, Verrucomicrobia, and Parcubacteria (Additional file 1: Figure S2). The distribution of these 214 microbes in polar and non-polar locations was further examined by mapping reads (10 million reads per metagenome) from the 60 polar metagenomes and the 74 Tara Oceans metagenomes; this led to the discovery of 32 microbes enriched in polar locations (average coverage in polar metagenomes was > 50 fold higher than that in Tara metagenomes) (Fig. 6). The microbes enriched in polar metagenomes included members of Alpha- and Gammaproteobacteria, Actinobacteria, Bacteroidetes, Chlamydiae, and Parcubacteria; notably, some taxa (e.g., Parcubacterium sp. Arctic_04) were widespread in most of the polar seawater metagenomes but were almost undetectable in the Tara Oceans metagenomes; moreover, within the Tara Oceans metagenomes, these MAGs tended to be enriched in the deep ocean and almost absent in the surface seawater metagenomes (Fig. 6).
The metabolic pathways of the 32 genomes enriched in polar metagenomes were reconstructed. Based on the KEGG pathway annotation (Fig. 7), several of these genome possessed pathways for nitrogen and sulfur cycling, such as periplasmic nitrate reductase NapA (K02567) and cytochrome c-type protein NapB (K02568) for nitrate reduction, nitrite reductase large subunit nitrite NirB (K00362) and small subunit nitrite NirD (K00363) for nitrite reduction, and sulfate adenylyltransferase subunit 1 CysN (K00956) and subunit 2 CysD (K00957) for sulfate reduction. The C4-dicarboxylic acid pathway was identified as the major pathway adopted by these microbes for carbon fixation. A number of genes for sugar and lipid biosynthesis and uptake were identified in most of the genomes, such as trehalose 6-phosphate phosphatase OtsB (K01087) for trehalose biosynthesis and lipoprotein-releasing system permease proteins (K09808 and K09810) for lipid biosynthesis.
Metagenomic and gneomic analyses in the present study have revealed the functional basis for environmental filtering of microbiota in the polar seawater. Saccharide and lipid biosynthesis genes, enriched in the polar metagenomes and MAGs, can be used by microbes to combat cold environments. For example, after a temperature decrease from 37 to 16 °C (“cold shock”), trehalose levels in Escherichia coli cells increased up to 8-fold [30]; biosynthesis of unsaturated fatty acids, which causes a decrease in membrane fluidity of bacterial cells, is crucial to the survival of microbes at low temperature [31,32,33]. The synthesized polysaccharides can be components of the bacterial cell membrane [34]; consistent with the result that many membrane synthesis related genes are enriched in the polar microbiomes, suggesting their roles in cold adaptation. The C4-dicarboxylic acid pathway also occurs with high frequency in the polar microbes; this pathway does not have as high a demand for coenzymes and metals as other carbon fixation pathways [35] and this may contribute to the success of these microbes in polar regions. In addition, the occurrence of several nitrate and sulfate reduction pathways in certain microbial genomes might be a strategy of metabolic versatility to adapt to environmental change.
Comparison between the Arctic and Antarctic microbiota
In total, 10,754 OTUs were only present in the Arctic microbiota, while 3034 OTUs were only present in the Antarctic microbiota (Additional file 1: Figure S3A). Dereplication of all the ORFs derived from the assembled Arctic metagenomes resulted in 11,806,833 orthologs, and all the assembled Antarctic metagenomes had 6,460,206 orthologs. BLASTp searching indicated that 3,268,783 (27.69%) and 1,160,189 (17.96%) orthologs are unique for the Arctic and Antarctic microbiota, respectively (Additional file 1: Figure S3B and S3C). After annotation, 4170 and 2125 COGs were identified in the Arctic specific and Antarctic specific microbiomes, respectively, and statistical analysis, based on the COG relative abundances, revealed that the two polar regions were enriched with different functions (Additional file 1: Figure S4 and Figure S5). For example, functions related to antibiotic resistance, including vancomycin resistance genes (COG2720), lantibiotic modifying enzyme (COG4403) and exporter of polyketide antibiotics (COG3559) were largely enriched in the Arctic specific orthologs (Additional file 1: Figure S4), whereas functions related to DNA recombination, DNA splicing and RNA transcription were enriched in the Antarctic specific orthologs (Additional file 1: Figure S5).
Despite the notable differences between the polar and temperate microbiomes, implying the existence of microbes that are adapted to the polar environments, comparison between the Arctic and Antarctic microbiomes identified considerable differences in terms of both taxonomy and function. This is consistent with a previous estimation that 78% of OTUs are unique to the Southern Ocean and 70% unique to the Arctic Ocean [9]. While more than 50% of the Arctic OTUs could not be found in the Antarctic microbiomes, only 27.69% orthologs are specific for the Antarctic microbiomes. This suggests that the Arctic and Antarctic microbiomes are more similar to each other at the functional level than from a taxonomic perspective. A further comparison suggests that antibiotic biosynthesis, DNA recombination, and DNA splicing are likely to be functions that contribute to the differences between Arctic and Antarctic microbiomes. The greater abundance of antibiotic resistance genes in the Arctic microbiomes might be a result of biological impacts, including human activity, which has already influenced environmental change in the Arctic [36]. DNA recombination and splicing are related to the repair of damaged DNA [37], which can be caused by exposure to ultraviolet radiation; this may be a molecular response to environmental change [38].