Soil bacterial diversity is positively associated with air temperature in the maritime Antarctic

Terrestrial ecosystems in the maritime Antarctic experienced rapid warming during the latter half of the 20th century. While warming ceased at the turn of the millennium, significant increases in air temperature are expected later this century, with predicted positive effects on soil fungal diversity, plant growth and ecosystem productivity. Here, by sequencing 16S ribosomal RNA genes in 40 soils sampled from along a 1,650 km climatic gradient through the maritime Antarctic, we determine whether rising air temperatures might similarly influence the diversity of soil bacteria. Of 22 environmental factors, mean annual surface air temperature was the strongest and most consistent predictor of soil bacterial diversity. Significant, but weaker, associations between bacterial diversity and soil moisture content, C:N ratio, and Ca, Mg, PO43− and dissolved organic C concentrations were also detected. These findings indicate that further rises in air temperature in the maritime Antarctic may enhance terrestrial ecosystem productivity through positive effects on soil bacterial diversity.


Results
Associations between latitude and environmental factors. We observed a significant increase (r 2 = 65%, P < 0.001) in MASAT between southeast Alexander Island at 72°S (MASAT −11 °C) and Signy Island at 60 °S (MASAT −4 °C), with a 0.58 °C increase in air temperature for each degree decrease in latitude ( Fig. 1; Supplementary Table S2). In addition, we found that the ratio of soil organic C to N decreased at lower latitudes (r 2 = 41%, P < 0.001; Fig. 1; Supplementary Table S2), which was principally owing to the highest C:N ratios being recorded in four soils from the most southerly location along the gradient (Fig. 1). An association was also found between latitude and soil Mg concentration, with higher concentrations of the element in soils from lower latitudes (r 2 = 27%, P < 0.001; Fig. 1; Supplementary Table S2). None of the other environmental factors were significantly associated with latitude (Supplementary Table S2).

Factors influencing soil bacterial alpha diversity.
Stepwise multiple regression modelling indicated that MASAT was the strongest and most consistent predictor of the numbers of observed and predicted (Chao1) bacterial taxa, as well as the phylogenetic diversity of each community (Table 1; Fig. 2). For every degree Celsius  (Table 1). All three measures of soil bacterial alpha diversity were also positively, but less closely, associated with soil moisture concentration (Table 1). Lastly, the number of observed OTUs was positively associated with Mg concentration, and the two other measures of alpha diversity were negatively associated with concentrations of water extractable PO 4 3− , or dissolved organic C and Ca (Table 1). Soil pH value, which ranged from pH 5.1-7.9 (Supplementary Table S1), was not found to influence bacterial alpha diversity.

Factors influencing soil bacterial beta diversity.
Stepwise PERMANOVA analyses indicated that MASAT was the strongest predictor of differences in bacterial community composition between soils (Table 2). Significant associations were also detected between changes in soil bacterial community composition and the concentrations of Mg, moisture, dissolved organic C and soil C:N ratio (Table 2). Soil pH had no influence on beta diversity.
Univariate regression analyses showed that the relative abundances of members of the Acidobacteria (OTU 6), Chitinophagaceae (OTU 19, Bacteroidetes), Gemmatimonadetes (OTUs 34 and 36) and Comamonadaceae (OTU 44, Betaproteobacteria) were positively associated with MASAT ( Fig. 3a-e), and that members of the Chloracidobacterium (OTUs 1 and 2, Acidobacteria) and Dechloromonas (OTU 46, Betaproteobacteria) were negatively associated with MASAT ( Fig. 3f-h). The relative abundances of four bacterial taxa were found to be related to soil C:N ratio, with a member of the Chitinophagaceae (OTU 14) being positively associated with C:N ratio (Fig. 4a), and a Comamonadaceae (OTU 43), a Flavobacterium (OTU 26, Bacteroidetes) and a Zymomonas (OTU 39, Alphaproteobacteria) population being negatively associated with C:N ratio ( Fig. 4b-d). Similarly, another member of the Chitinophagaceae (OTU 18) and a representative of the Xanthomonadaceae (OTU 48, Gammaproteobacteria) were found to be positively associated with dissolved organic C concentration (Fig. 4e,f). Members of the Acidobacteria (OTU 5) and the Flavobacteriaceae (OTU 25, Bacteroidetes) were relatively more abundant in soils with high Mg concentrations (Fig. 4g,h). soil bacterial community composition. In addition to members of the Acidobacteria, Bacteroidetes, Gemmatimonadetes and Proteobacteria, representatives of the Actinobacteria and Cyanobacteria were abundant in maritime Antarctic soils (Fig. 5). Those of the Candidate phylum AD3, Chloroflexi and Firmicutes were also abundant in soils at several locations but were more sporadically distributed (Fig. 5). The bacterial taxa detected were closely related to those observed in other surveys of soils from Antarctica or other cold and arid environments (Supplementary Figs S1-S5).

Discussion
Our analyses, based on 40 soils sampled from along a 1,650 km climatic gradient, indicate that air temperature is the strongest and most consistent factor explaining latitudinal changes in maritime Antarctic soil bacterial diversity. In accordance with other studies in the region showing positive associations between air temperature and soil fungal diversity 9 , decreased bryophyte and lichen diversity in more southerly habitats 17,18 and increased plant growth rates in response to warming in the latter half of the 20th century 4,6 , the data reported here indicate that the warming which is predicted to occur towards the end of the current century under moderate greenhouse gas emission scenarios 8 is likely to lead to increased diversity of soil bacterial communities across the region. These predicted changes to soil bacterial diversity are likely to be effected by bird 19 and human vectors 5 , as well as by wind currents, which are thought to be capable of transporting bryophyte spores, pollen and microbes from South America to the maritime Antarctic, and southwards along the Antarctic Peninsula [20][21][22] .
Although bacterial diversity is probably more closely associated with soil temperature than air temperature, the inaccessible nature of the majority of the sites studied here -some of which had not previously been visited -precluded , C and Ca concentrations in mg kg −1 in these analyses, respectively. Error degrees of freedom were 35 (observed and predicted nos OTUs) and 36 (Phylogenetic Diversity).
www.nature.com/scientificreports www.nature.com/scientificreports/  www.nature.com/scientificreports www.nature.com/scientificreports/ the long-term measurement of soil temperatures and their inclusion in regression models as predictors for bacterial diversity. However, previous research indicates that, despite air and soil temperatures becoming decoupled during winter in maritime Antarctica owing to the insulating effects of snow and ice cover 23 , air temperature is the best predictor for soil temperature in the region 24 (Supplementary Fig. S6). Rising air and soil temperatures might also affect the frequency of freeze-thaw cycles at soil surfaces, which short-term laboratory studies have found to have marginal effects on maritime Antarctic bacterial community composition 12 . However, rises in temperature will increase the availability of liquid water, a key driver of biological diversity in Antarctic terrestrial ecosystems 25 . Indeed, in this study, soil moisture was positively associated with the number of bacterial taxa in soil and was a significant predictor of changes to the relative frequencies of taxa between soils. In these frigid and usually arid habitats, growth is typically limited to periods between late spring and late summer, when soil temperatures rise above 0 °C and liquid water becomes available 23,26 , with the increased number of bacterial taxa in warmer www.nature.com/scientificreports www.nature.com/scientificreports/ and wetter soils likely reflecting the enhanced potential for metabolic activity for a wider range of taxa during late austral spring and summer 9,18 . Although rising air temperatures across the region might lead to accelerated soil thaw and increased meltwater and summertime precipitation 27 , amplifying the effects of climate warming on soil bacterial diversity, we cannot discount the possibility that changing atmospheric circulation patterns, such as increased frequencies of easterly winds bringing cold, dry air from continental Antarctica 28 , will force the drying of maritime Antarctic soils, limiting the effects of rising air temperatures on soil bacterial diversity.
Increased soil bacterial diversity associated with rising air temperatures can be expected to enhance terrestrial ecosystem productivity by accelerating the mineralisation rates of C and N and other elements in soil that limit www.nature.com/scientificreports www.nature.com/scientificreports/ the growth of other organisms, in particular higher plants and bryophytes 12,13,29 . The analyses here suggest that future rises in air temperature in the region will result in increases in the abundances of specific bacterial taxa, notably members of the Acidobacteria -a frequent phylum in more northerly maritime Antarctic soils 10 -and the Gemmatimonadetes, Chitinophagaceae and Comamonadaceae. Increases in the abundances of these taxa might be expected to enhance the decomposition of chitin and other soil organic compounds 14,30 and, given that a relative of OTU 34, Gemmatimonas sp. AP64, is capable of photosynthesis 31 , perhaps also to increase the fixation of atmospheric C into soil. However, of the three taxa that were found here to decrease in abundance in warmer soils, two (OTUs 1 and 2, both Chloracidobacterium) are close relatives of the photoheterotroph Candidatus Chloracidobacterium thermophilum 32 , suggesting decreased C fixation into warmer maritime Antarctic soils. Reductions in these taxa may hence potentially counteract any changes arising from increases in the abundances of members of the Gemmatimonadetes, representing a shift in the structure of the microbial photoautotroph community.
Increasing air temperatures in maritime Antarctica are thought not only to enhance plant growth and soil fungal diversity 4,6,9 , but also to accelerate mineralisation rates and soil nutrient cycling 33 . Maritime Antarctic soils in which there are greater availabilities of inorganic nutrients might be expected to support more active 34 and perhaps more diverse bacterial communities. This expectation was only weakly supported by the multiple regression analyses here, which indicated a positive effect of soil Mg concentration on the observed number of bacterial taxa, with univariate regressions also indicating increased abundances of members of the Acidobacteria and Flavobacteriaceae in soils containing higher concentrations of the element. In contrast, the multiple regressions indicated that the concentrations of dissolved organic C, water extractable PO 4 3− and Ca were all negatively associated with the predicted numbers of soil bacterial species and phylogenetic diversity, suggesting that the positive effects of warming on maritime Antarctic soil bacterial diversity may be counteracted by increased concentrations of soil nutrients associated with more productive soil and plant communities 33 . These findings corroborate observations from the Arctic, showing reductions in bacterial diversity and evenness in nutrient-amended soils associated with increased relative abundances of copiotrophic Alphaproteobacteria and Betaproteobacteria 35,36 . The substantial decline reported here in phylogenetic diversity associated with increasing dissolved organic C concentrations in soil is likely to have arisen, at least in part, from increases in the abundances of specific taxa such as members of the Bacteroidetes and Gammaproteobacteria, close relatives of which exhibit copiotrophic behaviour 29 . Univariate regressions also indicated that soils with low C:N ratios, which form in warmer environments owing to faster organic matter decomposition 37 , and may be indicative of nutrient-rich habitats, were preferred by Zymomonas, Flavobacterium and a member of the Comamonadaceae, representatives of phyla frequent in soils with high rates of C turnover 29 .  www.nature.com/scientificreports www.nature.com/scientificreports/ In agreement with previous studies in the maritime Antarctic that have found no influence of soil pH on soil bacterial alpha diversity or community composition 10,15 , or in which pH has been identified as an only marginally significant predictor (P > 0.06) for the frequencies of C and N cycling genes 14 , soil pH had no influence on bacterial alpha or beta diversity in the present study. This is in stark contrast to continental scale studies across North America and South America, which report highly significant first order polynomial relationships between soil pH and bacterial diversity, and no effects of mean annual temperature on diversity, in 88-98 soils sampled from Argentina through to northern Alaska 38,39 . This disparity is probably due to the absence from the analyses here of strongly acidic soils, in which there are highly significant linear increases in bacterial diversity between pH 3.5-5.0 in the Americas 38,39 . In contrast, no changes in bacterial diversity are recorded in North American and South American soils over the pH range of the soils studied here (pH 5.1-7.9) 38,39 . Nevertheless, why temperature influences soil bacterial diversity in maritime Antarctica and not in the Americas remains to be resolved. One possible explanation is that the much harsher environmental conditions encountered in maritime Antarctic soils, such as desiccation, low temperatures (<0 °C for approximately eight months each year and annual minima of between −15 °C and −40 °C) and wide temperature fluctuations (annual ranges 35-65 °C) 23 exert stronger selection pressures on bacterial taxa than in the soils sampled from across South America and North America, of which only six are exposed to annual mean temperatures of <0 °C 38,39 . This view is supported by the finding here that the taxa of bacteria present in maritime Antarctic soils -which are phylogenetically similar to those previously recorded in the soils of the region 10,40-42 -have affinities with taxa encountered in other cold and arid environments, such as ice from the McMurdo Dry Valleys in Continental Antarctica 43 , a Lake Vostok ice core 44 and hyper-arid soils from the Chilean Andes 45 .

Conclusions
The data reported here indicate that temperature is the predominant factor determining the diversity of bacterial communities in maritime Antarctic soils. Future rises in air temperature, predicted to occur later this century under moderate greenhouse gas emission scenarios 8 , are thus likely to lead to increased numbers of bacterial species in the soils of the region, enhancing terrestrial ecosystem productivity. However, increased concentrations of soil nutrients such as PO 4 3− , Ca and dissolved organic C, which are negatively associated with bacterial diversity, may counteract the positive effects of rising air temperatures on the numbers of bacterial species in soil. Air temperature data. Mean annual surface air temperature (MASAT) data for each site, gridded at a horizontal resolution of 55 × 55 km for the year 2007, were derived from the Regional Atmospheric Climate Model 16 . Although long-term soil temperature measurements for the majority of the sites studied here are unavailable, an analysis of five years of air and soil temperatures at Mars Oasis, the southernmost site sampled, shows that there is a close correlation between daily mean air temperature measured at 1 m above ground level and soil surface temperatures at 0-5 cm depth ( Supplementary Fig. S6).

Methods
DNA extraction, PCR amplification and 454 pyrosequencing. Total DNA was extracted from soils as previously described 9 . PCRs were performed on 2 µl DNA extracts, in 1× High Fidelity PCR Buffer (Invitrogen), with 100 nM of each dNTP (Invitrogen), 2 mM MgSO 4 (Invitrogen), 1 unit of Platinum ® Taq High Fidelity (Invitrogen), and 400 nM of each universal bacterial primer, made up to 30 µl total volume with molecular biology grade water. The forward primer was 27F (5′ AGAGTTTGATCCTGGCTCAG), 5′-labelled with the 454 FLX sequencing primer adapter B sequence. The reverse primer was 338R (5′ TGCTGCCTCCCGTAGGAGT), 5′-labelled with a sample specific barcode sequence 46  processing of sequence data. Sequences were quality filtered and dereplicated using the QIIME script split_libraries.py with the homopolymer filter deactivated 47 and then checked for chimeras against the October 2013 release of the GreenGenes database using UCHIME 48 v. 3.0.617. Homopolymer errors were corrected using Acacia 49 v. 1.48. Using QIIME, sequences were then clustered at 97% similarity using UCLUST 50 and cluster representatives were randomly selected. GreenGenes taxonomy 51 (October 2013 release) was then assigned to the cluster representatives using BLAST, and tables with the abundances of each Operational Taxonomic Unit (OTU) and its taxonomic assignment in each sample were generated. In addition, full length sequences that were identified as the nearest BLAST matches for each OTU were aligned and a midpoint rooted phylogenetic tree was generated.
www.nature.com/scientificreports www.nature.com/scientificreports/ A plot of the observed number of taxa relative to the number of sequences per sample showed that the sequencing did not account for all of the taxa present in the soils (Supplementary Fig. S7). Rarefaction 52 was hence used to calculate the expected number of taxa for an equal number of sequences per sample. The numbers of reads were rarefied to the nearest multiple of 50 sequences below the minimum number of sequences per sample (5,734) by re-sampling the OTU table. All comparisons of diversity were subsequently based on rarefied datasets comprising 5,700 sequences per sample. Inspection of the 95% confidence intervals for each rarefied sample ( Supplementary Fig. S7) showed that there was considerably more between-than within-sample variation, indicating that the mean alpha diversity values used in subsequent analyses adequately represented the within-sample diversity to facilitate robust comparisons between samples. The mean number of observed OTUs, predicted OTUs (Chao1) 53 and Faith's Phylogenetic Diversity Index values 54 were calculated using QIIME. statistical analyses. Relationships between latitude, mean annual surface air temperature (MASAT), altitude and the soil physicochemical characteristics were identified using Pearson's correlations. The influence of MASAT and the soil physicochemical factors on the alpha diversity metrics (i.e., the numbers of observed and predicted [Chao1] OTUs, and Faith's Phylogenetic Diversity Index values) were assessed using stepwise multiple regression analyses. The influence of MASAT and the soil physicochemical parameters on the composition of bacterial communities (beta diversity) was assessed using Permutational Multivariate Analysis of Variance 55 (PERMANOVA) as implemented in the Vegan R package 56 . Parsimonious PERMANOVA models were built by forward selection of significant predictors and the OTU relative abundances were Hellinger transformed prior to analysis. All analyses were implemented using R version 3.2.3 57 . phylogeographical analyses. Bacterial 16S rRNA gene amplicon sequences from the 2003-2005 survey of maritime Antarctic soils 10 (Genbank ACC: EF219488 -EF221599) were downloaded and aligned with the GreenGenes database (October 2013 release) and the most dominant OTUs from the present study using PyNAST 58 . All sequences were then inserted into the full GreenGenes phylogeny (October 2013 release) by maximum parsimony using Arb 59 . Sequences neighbouring the OTU sequences from this study were selected and then aligned in MEGA6 60 using MUSCLE 61 . Phylogenetic trees were inferred in MEGA6 using the maximum likelihood method based on the Jukes-Cantor model 62 .

Data Availability
The 16S rRNA gene amplicon sequences associated with this study have been deposited in the NCBI SRA under accession: PRJNA213362.