In a time of rapid environmental change, the study of biogeographic patterns on global and local scales, and the identification of drivers behind the spatial distribution of species, has become increasingly important (Walther et al., 2002). Biogeographic zones are formed for most ecosystems and associated taxa, as increasing geographic distance leads to a decrease in the exchange of individuals between two localities (distance decay theory, Morlon et al., 2008). Consequently, distant and isolated habitats are expected to have a specific diversity with a high level of endemism, whereas geographically closer habitats have more exchange and a higher proportion of shared species with lower endemism. Such patterns have been observed in both macro- (Astorga et al., 2012) and micro-organisms (Tamames et al., 2010; Bahram et al., 2013; Hahn et al., 2015). The main drivers behind these processes are thought to be intrinsic dispersal limitations (Mittelbach and Schemske, 2015) in combination with environmental filtering by spatial gradients of physical, chemical and biological traits (Chong et al., 2015). However, microorganisms (including prokaryotes), in contrast to many macroorganisms, are considered to be less confined by geographical barriers due to more efficient dispersal capacities (Bell, 2010; Chong et al., 2015), so that even remote and isolated habitats should be under constant exchange with both near and distant habitats. In this case, local as well as more cosmopolitan species are expected to be present and species distributions are likely to reflect environmental filtering rather than dispersal limitations (global ubiquity hypothesis, Baas-Becking, 1934; O'Malley, 2008).

High Arctic islands and the Antarctic continent are the most geographically distant terrestrial bioregions on Earth. It has been speculated that, because of the isolated character of the Antarctic continent and the environmentally extreme conditions in both polar regions, “the limits of species' abilities to disperse and survive may become more apparent” (Cox et al., 2016). Yet, both regions are similar in their physical-chemical characteristics, showing continuously low temperatures, limited freshwater availability, extreme seasonality in light conditions and localized nutrient availability. The harsh conditions prevent the survival of most macroorganisms (Convey et al., 2014) but allow diverse microbial life to thrive and dominate (Vincent, 2000a; Vincent and Quesada, 2012). A common group of microbial communities found in the terrestrial and lacustrine environments of both polar regions are those forming “macro-biofilms,” which are highly organized structures which can have vertical differentiation of functional and physiological assemblages (Fernández-Valiente et al., 2007; Mohit et al., 2017).

Both the Arctic as well as large parts of the West Antarctic, including the Antarctic Peninsula, are facing rapid environmental changes such as increasing temperatures due to both regional and global processes (Turner et al., 2009, 2014; Stocker et al., 2013). It has been proposed that climate change has already led to a weakening of biogeographic boundaries and the merging of ecological niches in the polar regions (Walther et al., 2002; Post et al., 2009; Nielsen and Wall, 2013; Pointing et al., 2015). The polar regions therefore provide a unique opportunity to test biogeographic patterns of microbial diversity, providing predictors/indications of the future consequences of environmental change. The survival strategies and dispersal limitations, genetic identity and endemic character of polar microorganisms and, thus, their resilience and sensitivity are still under discussion (Vincent, 2000b; Taton et al., 2006; Jungblut et al., 2010; Vyverman et al., 2010). A comprehensive exploration of these microbial features using consistent methodology and sampling has been difficult to date due to the remoteness and inaccessibility of the polar regions.

We here evaluated the roles of two opposing mechanisms, geographic distance effects (distance decay theory) vs. environmental filtering (global ubiquity hypothesis), in structuring bacterial biogeography in the polar regions. To do so, we undertook a large-scale microbial community analysis using standardized methods. We applied this approach to 89 freshwater microbial mat samples from multiple Arctic and Antarctic locations, including glacier runoffs and small streams, meltwater ponds, and larger lakes, as well as cryoconite holes, which have all been recognized as hotspots for microbial diversity (Vincent, 2000a; Quesada et al., 2008). Similar reference sites were also examined in temperate regions from alpine or marine influenced habitats.

This study provides the first large-scale comparative study of a unique set of microbial communities of the Arctic, Antarctic, and temperate areas using high-throughput sequencing. Due to the extreme remoteness and harsh environmental conditions, most polar microbial habitats and regions remain under-sampled and, comparative studies are limited due to the use of heterogeneous methodologies (Pearce et al., 2012). Thus, total polar bacterial biodiversity is currently underestimated, as is also suggested by the rarefaction analysis in the present study. The number of OTUs found in this study exceeded the number of OTUs and species detected using more traditional methods such as clone libraries or microscopic identification as summarized by Pearce et al. (2012). The high sample number combined with the high throughput sequencing used in this study enabled a more realistic and in-depth analysis of true (bi)polar microbial diversity.

Despite the heterogeneity of the sampled habitats, Arctic, Antarctic, and non-polar microbial mat communities in this study had a similar OTU richness and composition in terms of the dominant phyla. They do not show a declining latitudinal diversity gradient as described for most macro-organisms (Chown et al., 2004). The detection of the same bacterial OTUs (32% of all) in both polar regions as well as in temperate sites suggests a potentially ubiquitous distribution for at least some bacterial taxa, which supports the global ubiquity theory. Cosmopolitan distribution of prokaryotes and eukaryotes has been discussed for organisms smaller than one millimeter and usually decreases with increasing organism size (Finlay and Fenchel, 2004; Smith and Wilkinson, 2007; Yang et al., 2010). Finlay and Fenchel (2004) state that the cosmopolitan fraction of organisms in the bacterial size range is usually about 2.5-fold higher than detected in the present study. Cox et al. (2016) also found a percentage of ~35–60% cosmopolitan OTUs in Antarctic fungi and Petz et al. (2007) showed that 13% of freshwater ciliate species were shared between the Arctic and the Antarctic. For larger organisms Griffiths et al. (2009) suggest a general level of endemism of 50% in Southern Ocean benthos, and even greater levels of endemism (up to possibly 100% at species level) are reported across a range of Antarctic terrestrial biota (Convey et al., 2008; Pugh and Convey, 2008). This indicates that other mechanisms, such as geographical and climatic isolation and environmental selection play a particular role in structuring biogeography in the polar regions (Chown et al., 2004).

Since even 98% 16S rRNA gene similarity can represent >10 million years of bacterial evolution (Kuo and Ochman, 2009), we tested this result using a more stringent single linkage clustering sequences with less than two nucleotide differences (Mahé et al., 2014), that results in more OTUs with less average dissimilarity each, which confirmed the cosmopolitan and bi-polar distribution of 19% of the swarmOTUs. The high percentage of shared OTUs at a high 16S rRNA gene similarity between both polar regions suggests at least a periodic dispersal and exchange of individuals between both regions. This shows that geographical distance and barriers are not rigid and can be overcome by many microbial species, which is in line with the ubiquity hypothesis (Baas-Becking, 1934; O'Malley, 2008; Biersma et al., 2017). The underlying mechanisms could include efficient dispersal strategies for microorganisms via aerial (Hughes et al., 2004; Pearce et al., 2010) or vector-based transport including animals. Beyond these natural dispersal mechanisms, anthropogenic mediated dispersal for example via ships and planes (Cowan et al., 2011; Chown et al., 2012) could facilitate the dispersal of microorganisms. The high cosmopolitan fraction of cyanobacteria, in contrast to other bacterial taxa, suggests this phylum employs the above mentioned dispersal strategies, likely profiting from broad tolerances to extreme physical and chemical conditions (Vincent, 2000a; Vincent and Quesada, 2012).

In contrast to the cosmopolitan fraction, in our dataset more than half of all OTUs were uniquely present in one of the three regions (Arctic, Antarctic, or non-polar) and there was a decay in community similarity with increasing distance on a local and regional scale as described previously in Antarctica for microorganisms (Yergeau et al., 2007) and macroorganisms (Griffiths et al., 2009) supporting the distance decay theory. The percentage of these uniquely present OTUs was highest in non-polar samples in our dataset and likely reflects the higher similarity of the polar compared to non-polar regions. The somewhat larger fraction of endemic OTUs in the Antarctic compared to the Arctic may partially be explained by more extreme environmental conditions in Antarctica (Vyverman et al., 2010), but may also reflect the geographical situation of the two regions. The terrestrial Arctic has broad north-south connections over continental land masses, ocean sectors and ecoclimatic zones, whereby wind, water currents and migrating animals may serve as vectors for species dispersal. In contrast, the Antarctic continent is oceanographically as well as meteorologically isolated from the rest of the globe (Clarke et al., 2005; Barnes et al., 2006; Pointing et al., 2015). This geographic isolation has been speculated to cause a higher degree of genetic separation and thus be the driver for microbial endemism on the Antarctic continent (Vincent, 2000b; Vyverman et al., 2010; Hahn et al., 2015) as well as for Southern Ocean benthos (Griffiths et al., 2009). Isolated microhabitats in combination with local and regional distance-decay on the wide-ranging Antarctic continent may additionally provide ecological niches for a high microbial diversity and a high level of endemism (Yergeau et al., 2007; Vyverman et al., 2010). The uniqueness of Antarctic habitats and the underrepresentation in public databases of OTUs found here in combination with the high percentage of potentially endemic microbial OTUs detected in the south polar region, highlights the importance of specially protected areas for the conservation of microbial diversity in the Antarctic (Hughes et al., 2015). The rapid disappearance of unique habitats in the High Arctic (Vincent, 2010) and the potential for loss of endemic northern taxa indicate the need for a concerted global effort in protecting polar microbial ecosystems.

Antarctic as well as Arctic microbial communities were more similar to each other than to the temperate reference sites, and more than half of the potential ubiquitous swarmOTUs occurred exclusively in the polar regions and were not detected in temperate sites. Even though temperate habitats from alpine areas in New Zealand and marine influenced sites in Europe are not fully comparable to polar habitats, these results are corroborated by studies on eukaryote biogeography finding bi-polar distributions of fungal OTUs (Cox et al., 2016), testate amoebae (Yang et al., 2010), and bryophyte species (Biersma et al., 2017). These findings suggest that the global dispersal of certain (microbial) species is combined with environmental selection shaping bacterial communities in the polar regions (Vyverman et al., 2010; Chong et al., 2015; Cox et al., 2016), though temporal (Rochera et al., 2010) or spatial (Villaescusa et al., 2013) variability in environmental conditions can cause differences among prokaryotic meta-communities in polar habitats. Highly seasonal conditions in the polar regions (Chong et al., 2015), such as the absence of light in winter (Alonso-Saez et al., 2012) freezing and limited freshwater availability (Convey et al., 2014; Mohit et al., 2017) and low temperatures (Yergeau et al., 2007, 2012; Kleinteich et al., 2012) may cause the observed differences between polar and non-polar communities, since non-polar regions exhibit more moderate environmental conditions. In fact, large differences in temperature and annual solar radiation but not regional precipitation explained community differences in the present dataset. A broad scale phylogenetic meta-analysis recently suggested the presence of cold-tolerant cyanobacteria in all cyanobacterial taxonomic clades (Chrismas et al., 2015). The latter is also reflected in our dataset, in which bacterial species and strains surviving in polar environments belong to cosmopolitan groups at higher taxonomic levels. Cold tolerance, among other features, therefore seems to be a characteristic that has evolved several times in diverse taxa, and thus appears to be a trait of polyphyletic adaptation, given the complex optimizations required across the bacterial genome (Amico et al., 2006). We propose that, in addition to endemic taxa, there is a pool of cosmopolitan bacteria with widespread dispersal capacities, from which certain species and strains are selected for their capability to sustain themselves in polar environments.

Polar communities may therefore be disproportionately influenced by future physical, chemical and biological environmental changes (Stocker et al., 2013). Both temperature and precipitation (frequency, intensity and duration of snow or rain deposition (Convey et al., 2014), as well as the length of the ice melting period) are predicted to increase with climate change in polar regions (Turner et al., 2009, 2014; Stocker et al., 2013) providing more liquid freshwater but also less predictable conditions in some polar habitats (Vincent et al., 2017). Based on our results and those of others, both, the availability of liquid freshwater (Convey et al., 2014) and temperature (Yergeau et al., 2007, 2012; Kleinteich et al., 2012) can induce a shift in the community composition of polar microbial communities. Communities with a higher diversity often show a higher resilience toward perturbations, as functional redundancy lowers the chance for key species to vanish during perturbations (Yachi and Loreau, 1999). Similarly, increased richness might also be beneficial to a community by enabling a community to react with several distinct responses to a perturbation, increasing the resilience (Elmqvist et al., 2003; Folke et al., 2004). Higher temperatures in the polar regions will probably directly influence microbial communities by an elongation of the growth season, the availability of freshwater and daily temperature cycles with less freeze-thaw events. Lower environmental filtering induced by climate change as described here may support the establishment of cosmopolitan taxa, capable of long-distance dispersal, over more specialized and endemic taxa and may support the invasion of species on other trophic levels which could impact benthic microbial biofilm biology and biodiversity (Nielsen and Wall, 2013). In the longer term it can be expected that multicellular organisms such as mosses and grasses will replace microbial primary producers in currently microbial dominated habitats (Convey et al., 2003; Biersma et al., 2017), with microbial-dominated communities increasingly restricted to the highest latitudes (Nielsen and Wall, 2013).

We conclude that the microbiota of both polar regions may be more connected than previously assumed and our data suggest a biogeographical connection between both poles, consistent with some previous studies (Pearce et al., 2007; Jungblut et al., 2010; Chrismas et al., 2015; Cox et al., 2016; Biersma et al., 2017; Mohit et al., 2017). The polar microbial communities in our study are an assemblage of ubiquitously-distributed and potentially endemic taxa. This suggests a combined model of global dispersal of bacterial species in combination with habitat-specific selection pressure in cold environments (global ubiquity hypothesis) and distance-related biogeography in isolated Arctic and Antarctic communities (distance-decay theory). In both cases, the severe climate of the polar regions as well as their spatial and climatic isolation from the rest of the globe selects for a specialized community and leads to a high level of potentially endemic species. These organisms are potentially vulnerable to ongoing climate change in both polar regions, which is likely to favor cosmopolitan generalist species at the expense of distinct taxa that currently occur in specific polar habitats.

JK, FH, and DD designed the study. Laboratory research and experiments were performed by JK, FH, and AV with the methodological and material support of DD, PB, SW, and CZ. FH and MB performed the bioinformatics and statistical analyzes. Field work and sample collection were carried out by JK, SW, AJ, FK, AQ, and AC. Field trips were organized by WV, AC, AQ, DP, and PC. JK, FH, and DD wrote the first draft of the manuscript with substantial help of CZ, MB, and PB, and all other authors contributed substantially to revisions.