Many plants now found in the Arctic are thought to be derived from ancestors that occurred at high altitudes in mountains to the south during the Tertiary (Hultén 1937, 1958; Tolmachev 1960; Weber 1965; Hedberg 1992). However, there have been few attempts to test this hypothesis (see Murray 1995). As mentioned previously, phylogenetic analysis of cpDNA variation in S. oppositifolia resolved a mainly ‘Eurasian’ clade and a mainly ‘North American’ clade. Interestingly, the two basal haplotypes in each of these clades, haplotypes D and E, respectively, co-occur in the Taymyr region of north Siberia (Fig. 2). A possible explanation for this distribution pattern is that the species first occurred in the Arctic in north Siberia before migrating in east and west directions to obtain a circumpolar distribution. S. oppositifolia is also distributed through the Sino-Himalayan region of central Asia where several other species of Saxifraga sect. Porphyrion subsect. Oppositifoliae occur, which like S. oppositifolia, have opposite leaves (Webb & Gornall 1989; Gornall, unpublished). It is conceivable, therefore, that S. oppositifolia is derived from ancestral stock located in the high mountains of central Asia, which migrated to the Arctic in north Siberia along mountain ranges that connect these two regions. The discovery of late Tertiary macrofossils of S. oppositifolia in the Canadian Arctic Archipelago (Matthews & Ovenden 1990) and in north Greenland (Bennike & Böcher 1990) suggests that migration of the species from north Siberia eastward to north Greenland would have been completed before the start of the Pleistocene. Further molecular phylogenetic analysis is required to confirm that the species with similar morphology which co-occur with S. oppositifolia in the Sino-Himalayan region are indeed its closest relatives, in which case the hypothesis of an origin of S. oppositifolia in this region would be greatly strengthened. Similar studies carried out on other arctic species will determine if the hypothesis of a southern origin at high altitudes during the Tertiary is correct.
The most prominent feature of arctic plant evolution during the Quaternary is complex reticulation. The recent history of the Arctic has been extremely dramatic; its biota have been shaped through numerous large-scale climate changes resulting in cycles of fragmentations, range expansions and reunions of previously isolated populations (Stebbins 1984, 1985). The early Quaternary flora was probably recruited from survivors from the arcto-Tertiary forests combined with immigrants from southern mountain ranges (Murray 1995). This floristic mixture has since been repeatedly spatially rearranged and remixed, and today the majority of arctic plants are hybrids, many of them between plants which themselves are, or were, hybrids. These hybrids have been stabilized by chromosome doubling — allopolyploidization. Successive cycles of divergent evolution among populations isolated in different glacial refugia, migration into deglaciated terrain, hybridization and chromosome doubling have built up increasingly intricate and increasingly high-ploid mixtures. The genes of their diploid or more low-ploid hybrid ancestors are combined into individual plants, each of which carries virtually the entire gene pool of its current population through the next climatic catastrophe. The packing of ancestral genes, originally diversified by divergent evolution, into highly fixed-heterozygous, duplicated genomes ensures that genetic diversity is maintained through periods of extreme inbreeding and bottlenecks, for example when a deglaciated area is recolonized by a single long-distance dispersed seed that establishes a selfing population (Brochmann & Steen 1999).
A recent review of the flora of the isolated arctic archipelago of Svalbard, which was almost completely ice-covered during the last glaciation (Andersen & Borns 1994; Landvik et al. 1998), showed that nearly 80% of the 161 native species are polyploid (Brochmann & Steen 1999). The average ploidal level is close to hexaploid, and quite a few species are very high-ploid. All of the polyploids examined for variation at isozyme loci are fixed-heterozygous, i.e. they are genetically allopolyploid. Most of the few extant diploid species in Svalbard are highly homozygous because of regular self-fertilization, whereas in the polyploids, the level of heterozygosity increases strongly with ploidal level. Thus, the post-Weichselian flora of Svalbard has low species richness in terms of taxonomic species, but these species represent considerable genetic diversity inherited from a much larger stock of diploid ancestral species.
It is well known that the frequency of polyploids is particularly high in the Arctic, but not all polyploids have been formed there. In some groups such as grasses and willows, the original immigrants to the Arctic were already polyploid (Murray 1995). However, many polyploids have certainly been formed in the Arctic throughout the glacial cycles. Because of the repeated cycles of reticulation, however, the detailed evolutionary history of arctic species complexes is extremely difficult to unravel, and it is in many cases impossible to establish a taxonomy and species circumscriptions that correctly reflect their evolutionary history (but see, e.g. Hansen et al. 2000; Fjellheim et al. 2001; Scheen et al. 2002).
A very simple example of polyploid speciation that must almost certainly have taken place after the last glaciation, less than 10 000 years ago, is represented by the allotetraploid, Scandinavian endemic S. osloensis (Brochmann et al. 1996). This species is distributed narrowly in a zone between the currently alpine Scandinavian distribution of the diploid S. adscendens and the lowland distribution of the diploid S. tridactylites. Whereas the cpDNA of S. osloensis is identical to that of S. adscendens, its nuclear multilocus genotypes (RAPDs) can be obtained almost to perfection by adding markers observed in the two diploid species. Thus, the fully sexual and autogamous tetraploid S. osloensis has originated most probably in situ by hybridization between these two diploid species after the last glaciation. The alpine diploid S. adscendens probably immigrated first, after the retreating ice, followed by the more thermophilous diploid S. tridactylites.
Two other narrow endemics in Saxifraga, S. svalbardensis and S. opdalensis, have also probably originated in situ after the last glaciation, but they merely represent the current reticulate endpoints of a far more complex history of reticulate evolution (Brochmann et al. 1998; Gabrielsen & Brochmann 1998; Kjølner et al. 2000; Steen et al. 2000; Brochmann & Håpnes 2001). These species are bulbil-reproducing high polyploids with low but variable levels of fertility, probably caused by aneuploid chromosome numbers. Evidence from a variety of molecular markers suggests that these two species have originated independently via postglacial hybridization between the same two circumpolar species, the octoploid S. rivularis and the variable polyploid S. cernua. Although the same parental species were involved, even with S. rivularis as the maternal parent in both cases, S. svalbardensis and S. opdalensis are morphologically distinct and commonly recognized as different taxonomic species. Their morphological differences probably reflect polymorphisms within one of their parental species, S. cernua; it is likely that two divergent lineages of this variable species were involved in the formation of S. svalbardensis in Svalbard and S. opdalensis in southern Scandinavia.
What, then, are the origins of the parental species themselves? Both of them are allopolyploids, and thus of hybrid origin, as evidenced by fixed heterozygosity at isozyme loci. Their large geographical ranges may indicate that they originated a long time ago. Molecular data suggest that one of the progenitors of the octoploid S. rivularis was S. hyperborea, a widespread arctic tetraploid (cited as diploid in Brochmann et al. 1998; but see Brochmann & Steen 1999). The situation in S. cernua is more complex. This species comprises a range of lineages with different chromosome numbers, some of them even co-occurring at small spatial scales, of which the low-ploid ones may have given and still continue to give rise to the high-ploid ones. Nevertheless, the current, post-Weichselian endemics S. svalbardensis and S. opdalensis have ultimately combined a number of ancestral diploid genomes with different evolutionary histories, inherited via repeated reticulations that have occurred throughout the complex Quaternary history of the arctic flora.
The situation in this group of arctic Saxifraga illustrates a central feature of polyploid speciation — the possibility for multiple origins of a species (see, e.g. Soltis & Soltis 1993, 1999, 2000). The process of hybridization and allopolyploid speciation is simple and rapid, and can be repeated easily in different places at different times, involving more or less divergent populations of the same or even different parental taxonomic species. The end-products of such repeated processes may remain more or less local and ephemeral or become widespread and long-lived, and more or less similar morphologically, so that they may or may not be referred to the same taxonomic species. In the genus Draba, there are some arctic diploids and numerous arctic allopolyploids, up to the 18-ploid level. In this genus, molecular data combined with evidence from morphology and crossing relationships indicate extremely complex evolutionary patterns, ranging from recent, independent and local formation of similar polyploids to multiple origins of widespread, high-polyploid taxa (e.g. Brochmann et al. 1992a, 1992b; summarized in Brochmann 1992). In the genus Saxifraga, the above-mentioned studies not only provided evidence suggesting recurrent formation of the widespread S. cernua itself, but also of polyploids derived from divergent populations of this species and S. rivularis (Brochmann et al. 1998; Steen et al. 2000). In addition to the well-studied endemic polyploids S. opdalensis in southern Norway and S. svalbardensis in Svalbard, there are reports of populations in northern Norway which probably have similar, but independent origins from hybrids between S. cernua and S. rivularis (Rune 1988; Øvstedal 1998; Steen et al. 2000).
Molecular evidence suggesting recurrent formation of arctic–alpine polyploid species has also been obtained in three other genera; Poa (Brysting et al. 1997; 2000), Cerastium (Brysting & Borgen 2000; Hagen et al. 2001), and Dupontia (Brysting et al. 2002). It is now believed that most polyploid plant species are polyphyletic in the sense that they have formed recurrently from genetically divergent diploid progenitors (Soltis & Soltis 2000). This is very evident in the arctic flora, where many diploids and polyploids have enormous and overlapping geographical ranges. Thus, the intriguing taxonomic complexity of the arctic flora — also noted by Hultén — can probably be explained to a large degree by recurrent polyploidizations and subsequent interbreeding of the resulting genotypes.
Divergent evolution at the diploid level in arctic plants can also be more complicated than previously envisioned. In the strongly autogamous diploid Draba fladnizensis, crossing experiments revealed that it consists of at least two sibling species in the North Atlantic (Brochmann et al. 1993). Crosses between morphologically indistinguishable populations of this species from Scandinavia and Svalbard resulted in F1 hybrids that were vigorous and luxuriantly flowering, but entirely sterile. Meiosis was regular, however, indicating that the sterility was caused by minor mutations. Further crossing experiments at the circumpolar scale within this species and within another autogamous diploid, Draba nivalis, have revealed numerous incompatible combinations (Grundt et al. 2001 and unpublished data). It is possible that sibling speciation is a rapid and common process in small and fragmented populations of highly selfing diploids in the Arctic, promoting divergent evolution also within currently recognized taxonomic species. Sibling speciation inhibits mixing of populations that have diverged in different glacial refugia but re-immigrated to the same deglaciated area. Furthermore, it is possible that sibling diploids may occasionally hybridize and form genetically allopolyploid, but taxonomically autopolyploid, derivatives, thus adding yet another level of complexity to arctic plant evolution.
The high proportions of high polyploids and repeated reticulations in many arctic plant groups have previously hampered rigorous analyses of organismal phylogenies based on sequencing of nuclear DNA regions. The best possibility in future projects will be to clone and sequence the various subgenomes of the polyploids and extant diploids, to produce robust phylogenies of all of the constituent diploid genomes present. This approach will make it possible to trace the deep evolutionary and biogeographical history of arctic plants, as well as to map in more detail their more recent reticulations in response to the dramatic climate changes of the Quaternary.