Spatial distribution of freshwater crustaceans in Antarctic and Subantarctic lakes

Antarctic and Subantarctic lakes are unique ecosystems with relatively simple food webs, which are likely to be strongly affected by climate warming. While Antarctic freshwater invertebrates are adapted to extreme environmental conditions, little is known about the factors determining their current distribution and to what extent this is explained by biogeography or climate. We explored the distribution of freshwater crustaceans (one of the most abundant and diverse group of organisms in Antarctic and Subantarctic lakes) across four biogeographic provinces (Continental Antarctic, CA; Maritime Antarctic, MA; Subantarctic islands, SA; and Southern Cool Temperate, SCT) based on the literature, predicting that species distribution would be determined by biogeography, spatial autocorrelation among regions (in relation to dispersal) and climate. We found that variation in species composition was largely explained by the joint effect of spatial autocorrelation and climate, with little effect of biogeography – only regions within the SA province had a clearly distinct species composition. This highlights a plausible main influence of crustacean dispersal – mainly through migratory seabirds – and suggests that some regions will be more affected by climate warming than others, possibly in relation to the existence of nearby sources of colonists.

Ak, Auckland Island. Underlined species: Endemic from one biogeographic province; *: Endemic from two or more biogeographic provinces; **: Endemic from one region within a province.  Table 1 footnote.
www.nature.com/scientificreports www.nature.com/scientificreports/ more similar species composition than regions from different provinces), (ii) spatial autocorrelation among regions (i.e., regions closer to each other will have more similar species composition than more distant regions due to higher dispersal among them) and (iii) climate (due to species-specific environmental constraints).  Table 1 footnote.   Table 3. Similarity Percent analysis (SIMPER) to identify the contribution (%) of each species to the similarity and dissimilarity of each group.

Results
We extracted a list of 66 crustacean taxa (59 species and 7 genera/mosphospecies; hereafter species for simplicity) representing 8 orders (Table 1). Species were distributed mainly across the Subantarctic islands (SA) and Southern Cool Temperate province (SCT) (46 and 26 species, respectively). SA showed at least 1 species from each of the 8 crustacean orders, and all SA islands except Prince Edward Island (Pe) contained species that were endemic of our study area. Iles Kerguelen (Kr) showed the greatest richness (19 species, 5 endemic), followed by South Georgia (Sg; 17 species, 5 endemic), Macquarie Island (Mc; 14 species, 7 endemic including Iais sp., the only isopod recorded at these latitudes) and Iles Crozet (Cr; 11 species, 2 endemic). For SCT, Falkland/Malvinas Islands (Fa) concentrated most of the species (25) allocated across 7 orders and a high number of endemic species (10). Campbell (Ca) and Auckland Islands (Ak) had one endemic species, Chiltonia mihiwaka (Amphipoda). The cladoceran Ovalona weinicki 11  The latter group was further divided into 3 sub-groups: (3a) Cr, Pe, Kr, Heard Island (Hd) and Mc from SA, the latter without significant support; (3b) MA, Sg from SA and Fa from SCT, the latter without significant support; and (3c) En and Wi from CA (Fig. 1). The NMDS produced the same groups as cluster analysis (Fig. 2).

Discussion
Our results showed that spatial autocorrelation among Antarctic and Subantarctic lakes and climate were key determinants of crustacean distribution, while biogeography had a secondary role. Multivariate analyses revealed that only the Subantarctic biogeographic province had a distinct crustacean fauna. This province contained 46 species belonging to the 8 crustacean orders described for Antarctica. The species that most contributed to the distinctness of the Subantarctic province was the cladoceran O. weinecki 11 , which is the only Antarctic representative of a genus of mainly tropical and subtropical distribution 13 . Van Damme and Dumont 14 re-described this species from a complex of Alona sp. (principally, A. weinecki) described for Subantarctic islands 6,10 . Geographic distance may explain the fact that Macquarie island, which is separated ~6,000 km from the other Subantarctic islands, shared only a few species with them; and could also explain the high incidence of endemic species of crustaceans and other freshwater organisms in Macquarie island 15 . The relevance of distance was also revealed by our partial redundancy analysis (which showed that spatial autocorrelation among regions explained a large amount of variance), and it is most likely related to patterns of dispersal. Dispersal among nearby islands occurs mainly via migratory seabirds, which can transport resistant eggs within the gut or in mud adhering to feet 1,6,10,16,17 . Distance among regions is often a key determinant in the distribution of freshwater fauna 18 , which may help explain some inconsistencies in the definition of biogeographic provinces. The Maritime Antarctic province had similar species composition to South Georgia island from the Subantarctic province and the South American Falkland/Malvinas islands from the Southern Cool Temperate province. All these regions are separated by less than 2,000 km, so geographic distance could again be important in their similarity. It has also been proposed that Antarctic and South American crustacean fauna could have a common origin, as both continents were separated ~30 Mya 19,20 , thus being vicariant faunas 21,22 . However, this is unlikely, because most crustacean species in Continental and Maritime Antarctic provinces are Holocene immigrants, having arrived within the last 11 ka 10 .
Campbell and Auckland Islands (New Zealand), from the Southern Cool Temperate province, had a distinct fauna and mainly shared the unique amphipod species C. mihiwaka 23 and the widespread B. poppei. The separation of the Scott sector from the Continental Antarctic province from other regions was related to the Cyclopoida D. joycei 24 , which is the only species that inhabits this region. The other two sectors of the Continental Antarctic province (Enderby and Wilkes) have other species of this genus: D. mirnyi (present in both regions), D. walker (in Enderby) and D. kaupi (in Wilkes). This group of Diacyclops species is known as the "michaelseni group" 25 , a circum-Antarctic assemblage that shares some morphological characteristics and originated in Antarctic freshwater lakes in late Pliocene, prior to the onset of glaciation 24 . Lastly, the wide distribution of some species such as O. weinecki or B. poppei could be due to recent colonization events from northern latitudes, of anthropogenic www.nature.com/scientificreports www.nature.com/scientificreports/ origin in some cases 10 . Other authors have suggested an ancient origin for these species, which may have survived during Pleistocene glaciations in refugia [26][27][28] , such as Kerguelen Island 14 .
The lack of consistent climatic data for different Antarctic and Subantarctic regions precluded a more robust assessment of the influence of climate on freshwater crustacean distribution. However, our analyses using latitude as surrogate for climate suggested that climate affects distributional patterns, and its effect is variable among regions, depending on their location. Thus, some regions of Antarctica are likely to be more affected by climate warming than others, and this variation could be related to geographic distance to other sources of colonists. These differences may be further enhanced by the fact that some parts of Antarctica are experiencing greater temperature increases than others; the increase is particularly large in the Antarctic Peninsula, which has registered an increase of 0.67 °C per decade in the last 50 years [29][30][31] . Further studies are needed in order to improve our knowledge on biodiversity patterns and their main drivers in this continent that is experiencing some of the most rapid environmental changes on Earth 32 .

Methods
Study area. The Antarctic continent can be divided into 3 biogeographic provinces which differ considerably in climatic conditions 3,32-34 : the CA, which is the largest and coldest region with temperature rarely above freezing 35 , comprising the continent landmass south of 72°S and the Balleny Islands; the MA, which includes the western side of the Antarctic Peninsula north of 72°S and experiences seasonal snowmelt 35 ; and the SA, which comprises a series of islands and small archipelagos in the Southern Ocean proximate to the zone of Antarctic Polar Front (APF), with temperatures that on average are above freezing point year-round 36 . Besides, we considered a fourth biogeographic  www.nature.com/scientificreports www.nature.com/scientificreports/ province, north of the APF and influenced by low temperatures: the SCT province, which is formed by several islands from New Zealand and South America 10 , with cool to cold temperate climate 37 (Fig. 4).

Data collection. We elaborated a presence/absence matrix of all freshwater crustacean species reported for
Antarctic and Subantarctic lakes, based on two major literature reviews 6,10 , which contained all the available information to date. We divided each biogeographic province into regions following the above two reviews: CA comprised the En (30°E-90°E), Wi (90°E-150°E) and Sc (150°E-150°W) sectors; MA included the Pa, Ss and So; SA included Sg, Pe, Mc, Hd, Cr and Kr; and SCT included Ca and Ak from New Zealand and Fa from South Atlantic ocean (Table 1, Fig. 4). We excluded suspect records from the dataset, ruled out possible synonymies, and updated scientific names. We assumed that sampling effort of different taxa was similar across sites, although potential differences may have some influence on our results. Data analysis. We explored the influence of biogeography on regional species composition using hierarchical cluster analysis integrated with similarity profile analysis in SIMPROF 38 and metric multidimensional scaling, MDS 39 based on a similarity matrix using the Jaccard index. We tested for significance of the different groups of regions generated by cluster analysis using one-way ANOSIM, with biogeographic province as factor 40,41 , followed by pairwise tests. Further, we identified the main species associated with each group through SIMPER based on the presence/absence matrix of crustacean species. These analyses were done using Primer v.6 software 42 .
We explored the separate and joint influence of spatial autocorrelation among regions and climate using pRDA. The amount of variation explained by each factor and by their shared contribution was calculated by variance partitioning analysis, which is based on adjusted R 2 (R 2 adj ), and their statistical significance tested through permutation tests (999 randomizations). Species composition data was Hellinger-transformed prior to analysis to provide an unbiased estimate of variance partitioning based on RDA. Spatial autocorrelation was obtained with the eigenfunction analysis known as Principal Coordinates of Neighbor Matrix PCNM 43 , which created 10 spatial variables (PCNM vectors) based on a matrix of Euclidean distances between regions calculated using the geographic coordinates. These vectors allow the representation of different spatial relationships among regions at different spatial scales and can be treated as independent variables 44 . As we were not able to obtain consistent climatic data for all the study regions -there are relatively few meteorological stations in Antarctica and any gross estimate based on different data sources could be misleading -, we used decimal latitude as surrogate for climate. To eliminate any effect caused by different elevations, we used the residuals of a linear regression with latitude (as a response variable) against elevation (as a predictor) in the analysis 45,46 . Elevation was obtained from www.gps-coordinates.net based on latitude and longitude. These analyses were performed on R v. 3.5.1 47 , using the functions rda, varpart, anova.cca and pcnm from vegan package 48 .

Data Availability
Data will be available on the Open Science Framework online repository.