Continental Antarctica is a cold, polar desert [55], home to a few vast ice-free areas, the largest of which—the McMurdo Dry Valleys of southern Victoria Land—occupies around 40,000 km² [21]. Much of the phylogeographic work on the continent has centered on Victoria Land (Figure 2), and will subsequently form the bulk of our focus here.

Springtails are found throughout Victoria Land (Figure 1), where the distribution of individual species is generally constrained by distinct biogeographic breaks [26,56]. The northern and southern parts of the region are separated by the Drygalski ice tongue (Figure 2), which appears responsible for maintaining isolation between species in the two regions [26,56]. Northern Victoria Land has four springtail species: Desoria klovstadi (Carpenter), Gressittacantha terranova Wise, Friesea grisea Schäffer, and Cryptopygus cisantarcticus Wise [26]. Of these, the former two have largely non-overlapping distributions that are heavily partitioned by the presence of glaciers, while the distributions of the latter two species are mostly sympatric and scattered across northern Victoria Land. Southern Victoria Land has three species of springtail, the most widely distributed of which is Gomphiocephalus hodgsoni Carpenter [10,57,58]. Two other species of springtail (Neocryptopygus nivicolus Salmon and Antarcticinella monoculata Salmon) have more restricted distributions in the region and have been found at a few sites near Granite Harbour (GH; Figure 2) [12,59–61]. Of the seven springtails in Victoria Land, four have been examined phylogeographically. This includes D. klovstadi [31,61], Gr. terranova [62,63] and F. grisea [64] from northern Victoria Land, and Go. hodgsoni from southern Victoria Land [10,55].

For D. klovstadi, cytochrome c oxidase subunit II (COII) analysis focused on 69 individuals from five geographic locations (Figure 2; Table 1). Divergence among 26 identified haplotypes ranged up to 1.6%, placing the genealogical history of the populations for this species potentially within the last million years [61]. This analysis revealed striking population structure associated with geography, with just one instance of haplotype sharing among locations (between locations Football Saddle (FS) and Cape Jones (CJ); see Figure 2) [61]. This degree of segregation suggests a history of isolation between the locations while the presence of higher diversity at two of the sites (Daniell Peninsula (DP) and Cape Hallett (CH); Figure 2), highlights them as potential refugia from which the other lower diversity satellite populations may have been founded in more recent times. If this is the case, it is unusual that there are no remaining haplotype links between the two source populations and their ‘sinks’. Hence, it is also possible that each of the five populations are in fact refugial remnants or are the result of independent founding from as yet unsampled refugia. It would be particularly interesting to see if such heterogeneity among populations is a common feature throughout the northern area of the distribution range of D. klovstadi that is currently unsampled (see Figure 1 in [61]).

From the southern parts of northern Victoria Land, Gr. terranova has been examined using both allozymes [60] and mtDNA cytochrome c oxidase subunit I (COI) gene [63] (Table 1). Fanciulli et al. sampled 22 populations across the distributional range of Gr. terranova and found high levels of allelic variability among populations, even those in close geographical proximity [62]. These authors also found that populations could be arranged into three groups among which little gene flow was inferred from allelic frequencies. These groups corresponded to northern, central and southern parts of the distributional range, and were bounded in each case by major glacial systems (see Figure 2). While no time frame was provided by Fanciulli et al. [62], the high degree of differentiation among populations was taken as a possible indicator of ancient divergence for this species. Hawes et al. took an mtDNA (COI) approach, examining Gr. terranova samples taken from two altitudinal transects and a third more distant population in the Terra Nova Bay region (Figure 2) [63]. Here, the authors again found high genetic variability among samples despite close geographic proximity, with an upper divergence of 10% (i.e., corresponding to Pliocene divergence around 4-5 Mya) separating two distinct haplotype groups [63]. This suggests that Gr. terranova may be undergoing sympatric speciation, as has been suggested previously for Go. hodgsoni (e.g., [57]). Alternatively, two major genetic clades that were previously geographically isolated may have undergone recent (and now over-lapping) expansion from glacial refugia. Hawes et al. also found that, while Gr. terranova haplotypes showed little geographic structuring, sites with freshwater bodies seemed to be centers of haplotype diversity and, in several cases, links between geographically-isolated populations seemed to be accounted for by stochastic colonization events [63].

Finally, Torricelli et al. [64] examined F. griesea from northern Victoria Land using two mtDNA genes (ATP6 and COI). These authors found a total of 10 haplotypes among 55 individuals, which were separated by 10.9% COI sequence divergence (i.e., pre-Pleistocene origin) (Table 1). This high divergence resulted from a distinct group of Cape Hallett (CH; Figure 2) individuals. Within the southern portion of their sampling range, divergence among haplotypes was lower (<2.9%), although a population at Crater Cirque (CC; Figure 2) also remained distinct from the other populations [64]. The differentiation between northern and southern parts of the continental range of this species is in agreement with findings for other springtails in the region [61–63]. This suggests that the major glacial systems in the area have strongly impacted contemporary population structure by maintaining isolation among all refugial populations even as conditions have ameliorated.

In southern Victoria Land, Go. hodgsoni is arguably the most extensively studied Antarctic springtail in phylogeographic [10,49,57,65,66], ecophysiological (e.g., [67–69]) and now phylogenetic [58] contexts. This species is widespread throughout southern Victoria Land (Figure 2) and has been examined from both macro- and micro-scales across its range using both allozymes and mtDNA (COI and COII) (Figure 2, Table 1). In the most comprehensive analysis to date, McGaughran et al. sampled 289 and 191 Go. hodgsoni individuals for COI and COII, respectively, employing both descriptive haplotype distribution and diversity index analyses as well as nested cladistic analysis and a variety of demographic approaches (including maximum-likelihood based analyses) [49]. The demographic analyses in this study (using the programs ARLEQUIN, FLUCTUATE and MDIV; see [49]) all placed the genealogical history of Go. hodgsoni into timescales relevant to the last million years. In addition, haplotype distribution analysis showed an overall population structure for Go. hodgsoni that was compatible with the presence of historical glacial refugia maintaining isolated populations over this time. In essence, this resulted in a very fragmented haplotype pattern, with several putatively ancestral haplotypes present in high frequency. These haplotypes were rarely shared among populations for both COI (ca. 29%) and COII (21%) and most sharing was within large proposed refugial regions. However, there were a few isolated instances of haplotype sharing among regions (e.g., the McMurdo Dry Valleys (DV, sDV), Granite Harbour (GH); Figure 2). This suggests that isolated, stochastic, often long-distance colonization events are responsible for maintaining links across the gene pool for this species, and that occasional founding events have successfully established new populations (e.g., on Ross Island (RI); Figure 2) [49].

In southern Victoria Land, genetic structure has also been examined for the endemic mite, Stereotydeus mollis Strandtmann, which has an almost identical geographic distribution to Go. hodgsoni and often coexists in the same location [60,66]. In general, population genetic patterns of S. mollis have been shown to be congruent with those demonstrated for Go. hodgsoni and it is clear that S. mollis populations have also been dominated by contraction into refugia during glacial expansions [10,66]. In comparison to Go. hodgsoni however, S. mollis has shown far greater levels of COI sequence divergence (up to 14.4%, compared to 2.1% for G. hodgsoni) and therefore, further work is needed to clarify taxonomic issues for this species [66,70–73].

Another non-springtail study for comparison focused on the nematode Scottnema lindsayae Timm, which is found throughout southern Victoria Land as well as in regions to the north and south of the Transantarctic Mountains. In this species, no evidence has been found for geographical structuring of either mtDNA, 28s ribosomal DNA (rDNA) or ribosomal RNA (rRNA) haplotypes [27,74], and it has been suggested that the comparative ease of dispersal in nematodes prevents heterogeneity among populations from establishing [27,74,75]. However, though highly capable of long-distance dispersal, Antarctic nematode fauna are almost entirely endemic [26,76], suggesting that, while dispersal within Antarctica may be prevalent, dispersal into Antarctica offers more significant barriers to success.

In contrast, the moss Sarconeurum glaciale Card. & Bryhn, was found to have patterns of genetic differentiation (RAPD's) between Taylor Valley and Ross Island populations (DV and RI, respectively; Figure 2), while within Taylor Valley, the same authors detected high genetic (RAPD) variability between two isolated populations of the moss Bryum pseudotriquetrum (Hedw.) [77]. In the case of mosses, asexual reproduction coupled with limited propagule dispersal likely enhances genetic structuring. More recent work examining nuclear rRNA loci (18S–26S (ITS) [78], and ITS+phy2 [50]) of continental Antarctic populations of the moss Bryum argenteum Hedw., highlighted this species' isolation in Victoria Land (compared to sub-Antarctic and temperate populations). It also supported a disjunction between southern and northern locations within Victoria Land [50,78], which is very similar to the genetic breaks described above for springtails in the same region (e.g., [61]).

Thus, throughout continental Antarctica, common patterns within biogeographically distinct sub-regions may reflect a common history of isolation, persistence over Mya timescales, and re-colonization for terrestrial flora and fauna. Studies in the region have detected spatial genetic structure over both small (<1 km) and large (hundreds of kms) geographic distances (e.g., [10,49,61]) that span either: (1) Mio-Pliocene (4 – 5 Ma; Gr. terranova), or (2) Pleistocene (< 2 Ma; D. klovstadi and Go. hodgsoni) timescales, and recent work suggests that closer examination may reveal similar long-term isolation and population sub-structuring for other Antarctic biota (e.g., [50]). These data indicate minimal or no present-day gene flow among populations of the same putative species, despite their sometimes close geographic proximity. Thus, while re-colonization from refugia has played a significant role in determining contemporary patterns of population structure throughout continental Antarctica, this is likely to have taken the form of rare stochastic dispersal events. Such stochastic colonization processes are also likely to have largely driven species distributions in other Antarctic taxa (e.g., [79,80]), although these have not yet been the focus of studies employing molecular analyses.

In contrast to continental Antarctica, with its large expanse of rocks and well-developed coastal and inland terrestrial ecosystems, the maritime Antarctic consists of several small islands, and the western half of the Antarctic Peninsula (Figure 1). These areas are largely covered by snow and ice and peppered with terrestrial oases that are generally restricted in location to coastal areas [21]. Here, the springtail Cryptopygus antarcticus is the most widespread and abundant terrestrial arthropod [22].

The subspecies C. antarcticus antarcticus Willem, has been examined using mtDNA (COI and COII) sequences for 379 individuals across 24 locations throughout the Antarctic Peninsula and associated South Shetland Island archipelago [49]. Haplotype distribution analysis has found strong evidence of population structure and high levels of genetic diversity among locations [49]. Collapsing these locations into Northern, Central and Southern regions found little evidence of population linkage, with haplotypes shared in just a few instances, most of which were within these regions. There was also a striking excess of haplotypes present in low frequency, particularly in the mtDNA COI dataset (Table 1), suggesting recent population expansion for this species that has nearly obliterated ancestral signal [49].

To put their results into a greater biogeographic context, McGaughran et al. compared the patterns found for C. a. antarcticus on the Peninsula to findings for Go. hodgsoni on the continent (see previous section) [49]. In particular, these authors were interested in whether the two species, whose life history strategies and evolutionary persistence in Antarctica are postulated to be similar, would show similar contemporary patterns of population structure [48]. This theory was tested using demographic analyses (e.g., in the programs ARLEQUIN, FLUCTUATE and MDIV) to complement haplotype distribution analysis. With this approach, the strong genetic subdivision found in both species was shown to support a pattern in which multiple glacial refugia have allowed post-glacial expansion during the Pleistocene (including the last LGM), and subsequent population growth in isolation. However, there were distinct, and informative, differences in the finer scale of these patterns. For example, while both species showed evidence for population expansion, in C. a. antarcticus this appears to have been of a much greater magnitude and suggests that successful founding events on the Peninsula have been much more frequent through time [49].

In a more recent study, Torricelli et al. examined F. grisea from locations in continental Antarctica (see above) and the Antarctic Peninsula [64]. From a mtDNA (ATP6/COI) dataset of 80 individuals and nine populations, these authors obtained seven Peninsula haplotypes (<2.7% divergent) that differentiated into unique northern and southern groups. When compared to the Victoria Land dataset, COI divergence was 21.3% at its greatest, thus the taxonomic status of F. grisea has been questioned [64]. Indeed, work based on complete mitochondrial DNA sequences has shown that rather than being one widespread species, F. grisea is best described as separate species in their respective disparate locations [47], and this has recently been confirmed using morphological characters and nuclear DNA sequences [48]. The sequence divergence observed here corresponds to a speciation event initiated well into the Miocene [47], and is consistent with the hypothesis that Victoria Land (and continental Antarctica in general), has been isolated since that time [11].

The distinct intra-regional northern and southern genetic breaks found in F. grisea mirror the findings outlined above for C. a. antarcticus and Go. hodgsoni. Thus, while life in different areas of Antarctica is likely to be similar in certain respects, the different regions have imposed different genetic legacies upon their inhabitants, and this is consistent with the evidence outlined above supporting both macro- and micro-distinctions among the different Antarctic biogeographic elements (e.g., [21,32,33,40,47,49,51,69]). An interesting distinction between population bottlenecks/expansions of benthic relative to pelagic Antarctic shelf fish [81] shows that such differentiation among regions may not be restricted to the terrestrial realm.

Finally, in a non-springtail example, the southern maritime Antarctic hairgrass Deschampsia antarctica Desv. provides yet another illustration of isolation of populations among regions. This species showed overall low genetic diversity in an AFLP analysis. However, no genotypes were shared among populations from Signy Island to the north and from populations further south (Figure 2). Even within these two areas, significant genetic differentiation was found to exist between subpopulations separated by as little as 4 km [82]. Other differences in the fauna of the northern and southern parts of the maritime Antarctic have also been observed [20,83].

Thus, the pattern of isolation among subpopulations and refugial legacies prevalent in the continental Antarctic is recapitulated in maritime Antarctica. Because much of the presently available habitat in maritime areas is suggested to have been generally inundated by ice sheets at the LGM [84], many refugial locations have been difficult to identify [85]. However, pre-LGM refugia are beginning to be described [49,76] and these are likely to have been extremely important in creating maritime Antarctica's unique contemporary terrestrial fauna.

In comparison to the continental and maritime Antarctic, some islands of the sub-Antarctic have been well characterized (e.g., [86,87]). One interesting feature is that the circum-polar islands, of different age and in some cases, different geology, show an east-west division in species distribution [83]. Thus, species in the eastern parts of the zone (e.g., Macquarie Island and Îles Crozet) generally do not occur in the western (e.g., South Georgia) parts of the zone, and vice versa (Figure 2). Like the rest of Antarctica, the sub-Antarctic has a history of landscape subdivision through glaciation and/or volcanic activity [88–93], and indigenous fauna have histories extending in some cases to multi-million year timescales [11,32,94,95].

All of the sub-Antarctic islands are characterized by two major biotopes [96]. These correspond generally to an older epilithic biotope and a younger vegetated biotope dominated by eurytopic colonists [96,97]. Most sub-Antarctic islands are therefore inhabited by a mixture of ancient native taxa and young invading species that have successfully established, particularly following human exploration [32]. This highlights a key difference between the sub-Antarctic and other Antarctic zones: immigration into (and within) the sub-Antarctic is a much more common occurrence than at other less penetrable maritime and continental sites. Thus, many elements of the sub-Antarctic fauna are recent immigrants, originating either from temperate locations outside the polar frontal zone, or from the maritime Antarctic [32,98]. Such colonization in the case of springtails is likely to have been aided by their ability to raft for weeks at a time on both sea and fresh water [15,99,100], and may have been aided in some part by the ‘west wind drift’ and other oceanic currents as new island habitats have become available (e.g., [95]). The mixture of young and old taxa in the sub-Antarctic is particularly exemplified by springtails, which have been shown to have a varied and diverse evolutionary history in the circum-Antarctic that consists of both ancient and recent elements [11,95].

At island-scales, the mix of ancient and recent species is expected to be reflected in a pattern of substantially greater population structure in indigenous species (due their survival in isolated refugia over time), compared to introduced species whose residency time has been shorter. One study employing haplotype distribution analysis looked at this in detail and found exactly this pattern [96]. Myburgh et al. examined mtDNA (COI) genetic substructure in two indigenous and two introduced springtails on Marion Island [96]. They found considerable genetic structure for the indigenous species (e.g., 35 and 13 haplotypes from the springtails Cryptopygus antarcticus travei Deharveng and Tullbergia bisetosa Börner, respectively) and no structure (i.e., a single haplotype) in both of the introduced species (Table 1). For the two native species, the genetic patterns were also compatible with the geological and glaciological history of the island [96].

This result was supported by a second haplotype distribution analysis of C. a. travei from Marion Island [101]. This work used a dataset containing 113 individuals to obtain 39 unique COI haplotypes, of which 30 were singletons. This high degree of genetic diversity and haplotype distribution among populations supports long-term isolation and limited gene flow in C. a. travei [101]. Analysis of this species [101], and also of the mite Eupodes minutus [102] supported a north-east/south-west divide among the Marion Island populations that also corresponded to life history traits (e.g., patterns of population metabolic rate structure) in the case of C. a. travei [101]. This finding of geographical demarcation is consistent with the occurrence of multiple volcanic events on the southern side of Marion Island in conjunction with glaciation events over time.

Thus, a diverse genealogical history impacted strongly upon by events resulting in isolation of populations in refugia, is a pattern not only common to the deeper continental Antarctic, but also prevalent at its most northern (sub-Antarctic) locations.