Asexual organisms usually have larger and more northern distributions than their sexual relatives. This phenomenon, called geographical parthenogenesis, has been controversially attributed to predispositions in certain taxa; advantages of polyploidy and/or hybrid origin; better colonizing abilities because of uniparental reproduction; introgression of apomixis into sexuals; niche differentiation of clones; or biotic interactions. This review on apomictic plants demonstrates that each of these factors alone has not been able to explain the observed distributions. Establishment of the complex regulatory system of apomixis requires taxonomic and geographical predispositions; hybridization and/or polyploidization do create diversity, but they do not necessarily result in large distributions; colonizing abilities depend on clonal diversity and are outweighed by sexuals by self-compatibility and higher potentials for speciation; niche differentiation, ploidy levels and selfing keep sympatric sexuals and apomicts separated; and the impact of biotic interactions on distributions is uncertain. In conclusion, the distributional success of apomicts has a complex causality and depends on certain circumstances and combinations of factors. The rare establishment of apomixis may help to explain the predominance of sexuality on the large scale.
Apomixis (here defined as agamospermy, i.e. asexual reproduction via seeds) frequently results in considerable abundance and distributional success of some groups, such as dandelions (Taraxacum), blackberries (Rubus) and grasses (e.g. Poa, Elymus), challenging the general question of the evolutionary potentials of asexuality. The phenomenon that related asexual and sexual taxa have different distributional patterns has long been recognized, but it is still not well understood. Since Vandel (1928) coined the term ‘geographical parthenogenesis’, several authors have described the phenomenon both in animals and plants (Bell, 1982; Bierzychudek, 1985; Asker & Jerling, 1992; Van Dijk, 2003; Haag & Ebert, 2004; Kearney, 2005). For higher plants, Bierzychudek (1985) provided the most comprehensive evaluation of the previous literature, and concluded that apomictic plant groups (1) have larger distributional ranges, often considerably exceeding that of their sexual relatives; (2) tend to range to higher latitudes and altitudes than their sexual relatives; and (3) tend to colonize previously glaciated areas. Later studies add the aspect that sexual relatives often have distributions centred within much larger ranges of apomictic complexes (Antennaria, Bayer, 1990; Stevia, Soejima et al., 2001; Paspalum, Urbani, 2002; Taraxacum and Chondrilla, Van Dijk, 2003; Ranunculus, Hörandl & Paun, In press; earlier studies reviewed by Asker & Jerling, 1992). Asker & Jerling (1992) noted that abundance of apomicts decreases from the south to the north, but increases in northern regions only relative to the rest of the flora. They also discussed the difficulty of direct floristic comparisons because neither agamic complexes nor agamospecies are taxonomically equivalent to sexual species. Richards (1997) pointed out that geographical parthenogenesis applies only to taxa with gametophytic apomixis. Sporophytic apomixis, which is frequently observed in multiseeded tropical plants, and has completely different developmental and control mechanisms, is connected with other ecological and geographical features. This mode of reproduction, and also vegetative propagation, will not be discussed further here.
A broad evaluation of hypotheses to explain possible causes of geographical parthenogenesis led Bierzychudek (1985) to the conclusion that distributional success might be mainly caused by advantages of polyploidy, which is found in almost all apomictic plants. The most recent review by Kearney (2005), based mainly on data from asexual animals, reports basically the same features of distributions, but stresses the importance of advantages connected with hybrid origin as the main causal factor for distributional patterns. Several other nonexclusive hypotheses have been proposed for the causality, relying either on intrinsic features of asexual reproduction such as uniparental reproduction (Stebbins, 1950; Baker, 1967; Mogie et al., In press), or interactions between sexuals and apomicts (Mogie, 1992; Mogie et al., In press); or on external factors such as the production of generalists (general-purpose genotypes, Baker, 1965; Lynch, 1984) vs specialized genotypes (Vrijenhoek, 1984), or biotic interactions with other organisms (Maynard Smith, 1978; Bell, 1982; Glesener & Tilman, 1987).
In higher plants, considerable information about the functionality and genetic control of apomixis has been accumulated in recent years, factors that have not often been considered geographical questions. Because of high frequencies of hybridization and polyploidy, plants are ideal model systems for discussion of these two factors. Because of some features of plant reproduction, asexual reproduction appears to be less advantageous in plants than in animals. (1) Because of the predominance of hermaphroditism, most plants do not have the frequently cited ‘cost of sex’ from producing male individuals (Maynard Smith, 1978); the costs of the male function are only of male organs. A resource allocation from male to female reproduction has been observed in rare cases of male sterility (Meirmans et al., 2006), but the majority of apomicts maintain the male function for fertilization of primary endosperm nuclei, which is required for formation of viable seed (pseudogamy). (2) Vegetative reproduction and self-fertilization outweigh expected advantages of uniparental reproduction for apomicts. In hermaphroditic plants, both selfing and apomixis avoid the ‘cost of outcrossing’, that is, selfers and apomicts can serve as pollen parents of progeny of outcrossers, but not vice versa (Holsinger, 2000).
Previous authors have emphasized that geographical parthenogenesis is probably caused by a combination of factors (Haag & Ebert, 2004; Pound et al., 2004; Mogie et al., In press). Consequently, the main focus of this review is to evaluate the causality for the phenomenon in angiosperms by discussing all possible factors shaping distributions: to what extent a factor can act alone and is compensated by modifications of the sexual pathway; how combinations of factors are effective; and the implications of conclusions for the evolutionary potentials of asexuality.
Geographical parthenogenesis and the origin of asexual reproduction in plants
For an understanding of the biogeographical histories of apomictic complexes, the locations, circumstances and mechanisms of de novo origins of apomixis are of interest. Geographically, sexual relatives of apomictic complexes have predominantly temperate to boreal distributions. Apomixis is thought most likely to arise in areas and time periods influenced by glaciations of the Pleistocene, where range fluctuations of plants because of climatic oscillations offered the most frequent opportunities for interspecific hybridization (Lynch, 1984; Carman, 1997). Recent molecular studies confirm a relationship of the origin of apomictic lineages with the Pleistocene or postglacial time periods (Bayer, 1990, 1991; Dobešet al., 2004a; Paun et al., 2006b).
Increasing understanding of the mechanisms of apomixis has given insights into not only functionality, but also the frequencies and circumstances of de novo origins of apomixis. It is well established that apomixis is heritable and under control of one or more Mendelian factors, and that it is a modification of the sexual pathway, rather than an independent trait. Three independent processes, not controlled by the same loci (Mogie, 1992; Grimanelli et al., 2001; Spielmann et al., 2003; Bicknell & Koltunow, 2004), must act together for the production of functional apomictic seeds: development of (1) an unreduced embryo sac bypassing meiosis (apomeiosis, via diplospory or apospory); (2) the egg cell without fertilization (parthenogenesis) to avoid increase of ploidy levels; and (3) a functional endosperm (Fig. 1).
Several different hypotheses have been proposed for the origin of apomixis (Koltunow & Grossniklaus, 2003): (1) that apomixis is caused by mutations in one or a few linked genes, or in a master gene that initiates the different elements in a cascade-like fashion; (2) that apomixis-specific factors in supernumerary chromatin would control apomixis; (3) that stable and heritable epigenes could arise after hybridization and/or polyploidy through changes in DNA methylation or chromatin structure; and (4) the hybridization-derived floral asynchrony (HFA) theory, that the combination of genomes of two parental ecotypes differing in timing of developmental pathways would lead in the duplicated genome to an asynchronous expression of wild-type genes that regulate the steps from meiosis to postfertilization. This would result in precocious embryo sac initiation and parthenogenetic embryogenesis (Carman, 1997, 2001). The idea is supported by observations that the timing of developmental steps influences frequencies of apomictic phenotypes (Grimanelli et al., 2003).
Models 1 and 2 would require at least one mutation for apomeiosis and a coordination with parthenogenesis (Mogie, 1992; Grimanelli et al., 2001), or two independent mutations for both traits, both being individually deleterious (review in Van Dijk & Vijverberg, 2005), plus overcoming imprinting barriers for endosperm development (Spielmann et al., 2003). In fact, a complex of closely linked genes may control apomixis (Van Dijk & Vivjerberg, 2005), which is supported by recent findings that in Poa pratensis five independently segregating major genes control apomixis (Matzk et al., 2005). It is unlikely that such complex changes for the origin of apomixis have happened frequently. In fact, apomixis occurs multiple times within angiosperms (Van Dijk & Vivjerberg, 2005), within genera and also within complexes (Sharbel & Mitchell-Olds, 2001; Fehrer et al., 2005), usually in evolutionarily advanced groups, but restricted to certain orders and families (Van Dijk & Vivjerberg, 2005). Epigenetic changes (models 3 and 4) are heritable, and might also arise easily after hybridization or polyploidization (Koltunow & Grossniklaus, 2003), often altering temporal, spatial and abundance patterns of gene expression (Rapp & Wendel, 2005). Nevertheless, the processes resulting in simultaneous establishment and appropriate timing of the three components mentioned above, and the connection to polyploidy and hybridization, are still not well understood. Apomixis can arise in autopolyploids within the same species (Yahara, 1990; Bayer, 1991), and can be induced by artificial autopolyploidization without any combination of different genomes (Quarin et al., 2001). Apomixis also occurs rarely in diploids (e.g. in Boechera holboellii, Sharbel & Mitchell-Olds, 2001), although this may, in fact, be a paleopolyploid (Bicknell & Koltunow, 2004). In contrast, apomixis does not appear in F1 hybrids of sexual relatives of apomictic complexes (Asker & Jerling, 1992). Apomixis is perhaps most frequently (but not necessarily) connected with hybridization and polyploidy, because the combination of both traits increases the likelihood of the required complex changes of genetic control (U. Grossniklaus, personal communication).
Despite considerable efforts and intensive research over decades, induction or transfer of apomixis into crops has so far had limited success (reviews in Savidan et al., 2001), confirming a high complexity of regulatory mechanisms. This is probably a main reason for its rare establishment, arising in less than 1% of angiosperm species (Mogie, 1992), despite the fact that that polyploidy and hybridization are frequent phenomena in higher plants (Table 1). Approximately 75% of apomictic taxa belong to only three families (Asteraceae, Poaceae and Rosaceae; Richards, 1997), which has been attributed to predispositions of some families either to overcome imprinting barriers in the endosperm, or to bypass fertilization of the endosperm via autonomous apomixis (Spielmann et al., 2003; Bicknell & Koltunow, 2004). Another hypothesis suggests common ancestry and introgression of apomixis genes at (sub)tribal level within these families (Van Dijk & Vivjerberg, 2005). A third idea relates apomixis to predominance of single-seeded fruits in these families (e.g. achenes, caryopses), which could infer selective fitness advantages of asexually produced, separated seeds and seedlings (Richards, 1997).
Table 1. Summary of advantages and disadvantages of reproductive systems and evolutionary traits in angiosperms: estimates of frequencies among angiosperm species (in parentheses) demonstrate the rarity of establishment of apomixis
Reduction of genotypic diversity, low adaptive potential, difficult to form and establish, less than 1%§
Concluding, apomixis arises not randomly, but requires the simultaneous establishment of complex genetic and/or epigenetic regulatory mechanisms, which require taxonomic, geographical and ecological predispositions. Considering the rather young geological age of apomicts, their higher abundance at higher latitudes/altitudes may simply reflect preferences for macroclimatic conditions similar to those of their ancestors and their main areas of historical origin.
Polyploidy and/or hybridization as causal factors of geographical parthenogenesis
Bierzychudek (1985) hypothesized that the distributional success of apomicts is simply caused by advantages of polyploidy. Multiple copies of the genome provide pathways for development of new gene functions, and combination of genetic information from different ancestors, thus providing more physiological and ecological flexibility compared with their diploid parents. Recent surveys have confirmed extensive effects on restructuring of genomes following polyploidization because of massive gene silencing, diversification in gene function, and different expression of parental genes (Adams & Wendel, 2005). Nevertheless, polyploidy also has negative effects, such as disrupting impact on nuclear and cell enlargement; the propensity of polyploids to produce aneuploid cells; and epigenetic instability in transgressive (nonadditive) gene regulation (Comai, 2005). Polyploids can be formed rapidly in higher plants by unreduced gametes, either via one-step formation or via the triploid bridge, with or without interspecific hybridization; estimates of frequencies among angiosperm species range between 40 and 70% (Ramsey & Schemske, 1998).
Nevertheless, polyploid taxa do not necessarily show broader distributions than diploids. After examination of 75 polyploid complexes in European genera, Stebbins & Dawe (1987) observed that, in 69 genera, the ratio of widely distributed to narrowly distributed species is not significantly higher in polyploid species than among diploid species of the same genus. They concluded that broad distributions are not a universal tendency of polyploids, but must also be attributed to other features. Autopolyploid sexual cytodemes may have much smaller distributional areas than the diploids of the same species, as shown for example in Ranunculus cassubicifolius (Hörandl & Greilhuber, 2002). Even in northern areas, where polyploidy is increasing in frequency and levels towards the north (Brochmann et al., 2004), polyploid species do not necessarily have wide distributions. Diploid sexual endemic vascular plants of the North Atlantic area do not show significant differences from polyploid sexual species in the size of distributions, estimated as the number of floristic regions occupied (Fig. 2; Brochmann et al., 2003).
It is difficult to test whether a diploid apomict has a narrower distribution than a polyploid, as would be expected, because such cases are extremely rare. In Boechera holboellii (=Arabis holboellii), apomixis occurs both in diploid and polyploid cytodemes, sexuality only in diploids; surprisingly enough, diploids show broader distributions and are substantially more frequent within previously glaciated areas than polyploids (Dobešet al., 2004a, 2004b). These findings suggest that polyploidy alone, although an important mechanism for speciation and diversity in the north, cannot explain the striking range expansions of apomictic complexes compared with their sexual relatives.
Excluding interspecific hybridization, apomictic autopolyploid cytodemes of two Antennaria species (Bayer, 1991) and Boehmeria spicata (Yahara, 1990) have larger and, at least in Antennaria, more northern distributions than diploid sexual cytodemes of the same species. That means a combination of polyploidy plus apomixis, without interspecific hybridization, can be effective for geographical parthenogenesis.
Excluding polyploidy, homoploid hybridization results in sexual plants not frequently in distribution areas separated from that of the parents. Floristic surveys such as that of Stace (1976) show that homoploid hybrids occur in the majority of cases just locally among their parents, as a result of multiple local origins rather than dispersal of single hybrids. Introgressive hybridization may form hybrid zones that are often not well defined geographically, frequently at ecotones or boundaries between habitats in the sympatric range of the parental species (Harrison, 1993). Introgression may occasionally result in replacement of one or both parents, but rarely in distributions distinct from the parents. A review of 165 ‘noteworthy’ case studies of widespread diploid, introgressive hybridization (definitely excluding only local hybrid swarms) demonstrates that only a minority of studies (c. 40%) suggest evolution of new ecotypes or species (Rieseberg & Wendel, 1993). Diploid hybrid speciation is, in general, regarded as rare (Wolfe et al., 1998a, 1998b and literature therein), and is often attributed to transgressive segregation and ecological speciation (Seehausen, 2004). In well documented cases of homoploid hybrid speciation, such as in North American irises (Arnold & Bennett, 1993); sunflowers (Rieseberg et al., 2003); and Penstemon (Wolfe et al., 1998a, 1998b), homoploid hybrid species show a differentiation of habitats or pollinators, but their distribution areas remain much smaller than those of the parental species; a northern tendency is not observed.
Concluding, hybridization alone is not correlated with significant range expansions outside the distribution area of the parents and towards the north, as is typical for geographical parthenogenesis. Hybridization creates genomic novelty, but requires a stabilizing mechanism providing reproductive isolation, such as polyploidy, chromosomal recombination, or external barriers (Grant, 1981). Hybrids have to overcome disadvantages of reduced fertility and of a minority cytotype among the parents (Levin, 2002). Apomixis provides a threefold solution for these problems: first, it prevents fertilization and introgression by sexuals (but allows introgression of apomicts into sexuals, see below and Fig. 1); second, potential meiotic disturbances following hybridization and/or polyploidy are not effective; and third, it avoids the necessity to build up populations to maintain certain genotypes and allows dispersal of single clones (see below). In this combination, apomixis can be more successful than sexual allopolyploidy, which has similar potential: large pools of genotypes and phenotypes with novel evolutionary trajectories (Arnold, 1997), and fixed heterozygosity (Brochmann et al., 2004). But the strong correlation with large range expansions compared with diploid sexual relatives, as observed in apomictic complexes, is not evident in sexual allopolyploid species. Therefore geographical parthenogenesis is better explained by a combination of novel genotype/phenotype, reproductive isolation provided by apomixis, plus uniparental reproduction promoting extensive range expansions. The latter is discussed in the following section.
Advantages of uniparental reproduction for colonization
Compared with vegetatively reproducing plants, apomicts benefit from advantages of reproduction via seed such as specialized dispersal mechanisms, protection, seed dormancy, and freedom from disease in vascular strands (Richards, 2003). Apomicts have often been thought to be better colonizers than sexual outcrossers, especially after long-distance dispersal, because of uniparental reproduction, similar to selfing plants (Stebbins, 1950; Baker, 1967, ‘Baker's law’), because of the potential to found populations with single individuals (Fig. 3a). Sexual relatives of apomicts are usually self-incompatible (Asker & Jerling, 1992). In contrast, apomicts reproduce either independently from pollination (via autonomous apomixis), or via pseudogamy. In the latter case, self-pollination enables the development of functional seed even without pollinators (observed in Ranunculus, Rutishauser, 1954; Hörandl, unpublished data; and in Rubus, Nybom, 1986; Kollmann et al., 2000). For both autonomous and pseudogamous apomicts, numerous molecular studies, even using highly sensitive markers, have confirmed the existence of populations consisting of only one lineage, suggesting that single individuals may indeed have founded such populations (Gornall, 1999; Hörandl & Paun, In press). Moreover, independence from cross-pollination provides reproductive assurance in environments where pollinators are less frequent. As apomicts are usually insect-pollinated, uniparental reproduction may contribute to the success of apomicts in regions with colder climates and shorter vegetation periods.
In contrast to selfing plants, apomicts bypass meiosis and avoid increased homozygosity in the offspring. Increased levels of observed heterozygosity have been documented in many apomicts in numerous molecular studies, and are attributed to allopolyploid origin, fixed heterozygosity in duplicated loci, and independent mutations in different alleles (reviewed by Gornall, 1999; Hörandl & Paun, In press). Hybrid vigour and buffering of inbreeding depression are important effects of heterozygosity; fixed heterozygosity in polyploids may suggest important advantages for colonization (Brochmann et al., 2004). Nevertheless, apomicts of autopolyploid origin do not have significantly higher levels of heterozygosity than their diploid sexual relatives, but do express geographical parthenogenesis (Yahara, 1990; Bayer, 1991; Hörandl & Paun, In press). Theoretically, apomicts are also unaffected by small population sizes and founder effects during initial phases of colonization, in contrast to outcrossing or selfing sexual species (Barrett & Pannell, 1999). In the long term, apomicts are expected to need a longer time than selfers to recover genetic variation after a bottleneck, because the latter maintain segregation at heterozygous loci during meiosis (Fig. 3a), which will create more variability posed to selection than multiple chromosomes (Burt, 2000). In sexual selfers, selection against homozygous deleterious alleles may lead to purging effects and subsequent increase of fitness after inbreeding (Carr & Dudash, 2003).
Selfing occurs as dominant in c. 20%, and as casual in c. 40% of angiosperms (Richards, 1997), offering a more frequent and flexible mode of uniparental reproduction. Polyploidization frequently causes a breakdown of self-incompatibility (Marble, 2004; Comai, 2005). Selfing is often associated with an annual life form, which infers an advantage for pioneer plants, but also occurs frequently in perennial colonizers of previously glaciated areas and in island plants (Richards, 1997). Comparisons with self-compatible species considerably restrict generalizations about the superior colonizing abilities of apomicts. In the North Atlantic area, despite the considerable dispersal capacity of endemics, apomictic vs selfing endemics show no significant differences in the number of subregions occupied (Fig. 2, data from Brochmann et al., 2003); but only among selfers some widespread species, occurring in up to seven of the nine subregions, are reported.
Better founder abilities of apomicts are also questioned by habitat preferences and distribution patterns. Despite expected advantages, even to selfing plants, a high frequency of apomicts in open pioneer habitats, such as ruderals, is not observed; instead, in northern Europe, apomictic plants occur most frequently in pastures and on forest floors, in moderately disturbed habitats, where they are significantly more abundant than the rest of the flora (Asker & Jerling, 1992).
High colonizing abilities after geographical isolation are also questioned by the scarcity of records of apomicts on islands (Bierzychudek, 1985), where long-distance dispersal is required to overcome geographical barriers. The assumed advantages of uniparental reproduction are confirmed by observations that self-compatibility and self-fertilization are common in island plants, and also occur frequently in island endemics (Barrett, 1996; Anderson et al., 2001). Nevertheless, species diversity of native apomicts on islands compared with the continents is low, even in well documented flora and climatic regions where apomicts are expected to be frequent. In the Flora Nordica region, agamospermous taxa are by far the largest element among Nordic endemics; but among 39 selected endemic agamospecies of the region, only two (5.1%) occur on islands (Jonsell & Karlsson, 2004). In the North Atlantic area, the hugely widespread apomictic complexes Alchemilla sect. Vulgares, Hieracium and Pilosella are missing in subregions 1, 2 and 5, respectively, isolated by the Atlantic (Brochmann et al., 2003, Appendix). Only Taraxacum and the Ranunculus auricomus complex occupy all subregions. The latter group comprises c. 600 agamospecies in the Flora Nordica region, and is abundant on the whole continental area (Ericsson, 2001). In contrast, just two agamospecies are certainly island endemics, and abundance of R. auricomus s.l. on the Faroer Islands, Svalbard, and most parts of Iceland is drastically reduced to fewer than 10 localities per subregion (Ericsson, 2001). Biogeographical studies on Hieracium in northern Europe confirm that all water surfaces provide important dispersal barriers for agamospecies (Tyler, 2000). On New Zealand, apomixis has been assessed so far for just five native species in three genera, whereby this reproductive system has not been widely examined in the native flora (Heenan et al., 2002, 2003). In the well studied flora of the Juan Fernández Islands, agamospermy has been found in only two native dandelions, whereas self-compatibility is reported for 18 species (Bernardello et al., 2001). Among endemics of this island, 81% are self-compatible, but no case of apomixis could be ascertained (Anderson et al., 2001). Nevertheless, rapid range expansions of apomicts on islands are reported after human introductions (Carino & Daehler, 1999; Houliston & Chapman, 2004). Multiple introductions and cases of hybridization of previously isolated species triggered the evolution of new genotypes and rapid speciation on islands (Chapman et al., 2000; Amsellem et al., 2001; Trewick et al., 2004).
The origin of agamospecies is caused by hybridization and/or polyploidy, which is not a common mode of speciation on islands (Stuessy et al., 2006), and is more likely on the continents because of the higher pool of potential sexual parents and the impact of glaciations. In contrast, island endemism often results from anagenetic speciation (Stuessy et al., 2006), which is not effective in apomicts because of low levels of genetic drift and genotypic variation on which selection can act. Contrary to sexuals, apomictic lineages have a low potential to speciate after strong geographical isolation. Their ability to respond to different environments probably relies on phenotypic plasticity rather than on adaptation.
To summarize: a theoretical advantage of apomixis for colonization can be outweighed by the more flexible system of self-fertilization. The low representation of native apomictic plants in pioneer habitats and in strongly geographically isolated areas contradicts the idea that apomixis suggests a general advantage for founder events. Colonizing success of apomictic complexes appears to depend on the potential to maintain or create a diversity of lineages, which is better achieved on the continent of their origin. Apomixis is probably helpful for range expansions and colonization, but only under certain conditions: (1) self-incompatibility of the sexual competitor; (2) a certain level of initial clonal diversity involved in colonization events; (3) opportunities for recurrent formation and stepwise dispersal; and (4) availability of certain types of habitat.
The male function in asexuals and replacement models
As apomictic plants arise from sexual ancestors, they are expected to occur initially sympatrically with sexual relatives. Most apomictic plants are hermaphrodites, and maintain microsporogenesis via meiosis and production of pollen, which is usually partly, but almost never completely, aborted. In a mixed sexual/apomictic population, a hermaphroditic apomict will also fertilize sexual individuals and transfer apomixis to sexuals via pollen, but will not be fertilized by the sexuals; as mother plants, apomicts have maternal, apomictic offspring. Thus theoretically there is a 1.5-fold advantage to apomixis per generation, and genetic factors controlling apomixis will soon increase in frequency and come to fixation (Mogie, 1992; Fig. 3b). This theoretical assumption will hold if male and female fitness of sexual and apomicts and all other factors are equal. In fact, crossing barriers because of different ploidy levels and/or ecological differences of apomicts and sexuals, facultative sexuality and aborted pollen in apomicts, reduced fitness of F1 hybrids, and variance in female fitness of apomicts may influence the actual frequencies of reproductive systems. During comigration, extinction of sexuals via introgression may build up a front of apomicts impermeable for sexuals (Mogie, 1992). Immigration of maladapted genotypes into a new environment could result in a loss of fitness in sexuals because of random mating, whereas this is not effective in apomicts (Peck et al., 1998).
Experimental studies in Hieracium (Krahulcova et al., 1999) and Taraxacum (Morita et al., 1990; Tas & Van Dijk, 1999; Brock, 2004) have shown that self-incompatibility of sexual taxa can break down when accompanied by additional pollen from another sexual or apomictic species (mentor effect, Richards, 1997). Thus sexuals coexisting with apomicts or other closely related species may reproduce via selfing as well, and avoid genetic assimilation by apomicts (Fig. 3b). In mixed populations, selfers also avoid the ‘cost of outcrossing’ and are expected to reproduce as successfully as apomicts (Holsinger, 2000). A study on mixed populations of native sexual Taraxacum ceratophorum and introduced apomictic T. officinale suggests partly genetic assimilation of sexuals because of asymmetrical hybridization, but also partly selfing because of a mentor effect (Brock, 2004).
The HFA theory (Carman, 1997) requires hybridization of ecotypes with different preferences of altitudes/latitudes for the origin of apomixis. Apomixis is probably advantageous in high altitudes/latitudes because of reproductive assurance when pollinators are rare, and a shortening of reproductive pathways (Mogie, 1992). Thus, during colonization of such areas, the apomict competes directly, mainly with the high-altitudinal/latitudinal sexual parent. In the case of missing niche differentiation, the apomict may replace the high-altitudinal/latitudinal sexual because of unidirectional hybridization. Only the more southern/lowland parental sexual would persist, which would fit with observed distribution patterns. This would explain the high relative frequency of apomixis in northern regions without the necessity of colonizing extremely high latitudes/altitudes or open pioneer habitats; in temperate regions with longer vegetation periods, the advantage of apomixis may be less important.
The big problem of replacement models is to explain the frequently observed coexistence of sexuals and apomicts, in other words, why sexuals have not gone extinct completely. Simulation studies on small-scale spatial dimensions in the form of patchy occupancy models of mixed sexual–asexual populations suggest that one of the reproductive modes comes to fixation in the landscape because of formation of clusters, whereby initial conditions are probably important. Nevertheless, examples of coexistence are also found in theoretical modelling (Carillo et al., 2002). Empirical studies in mixed apomictic/sexual populations are rare, but suggest coexistence because of niche differentiation (see below), and reproductive isolation via different ploidy levels (Vavrek et al., 1998; Hörandl et al., 2000; Meirmans et al., 2003; Verduijn et al., 2004; Fig. 3c). To conclude, replacement of sexuals by apomicts via introgression is probably limited by (1) the strength of reproductive barriers; and (2) the capacity of sexuals to shift to selfing.
The general-purpose genotype vs the frozen niche variation model
Lynch (1984) proposed a hypothesis that temporally variable environments select for lineages with broad ecological tolerances (general-purpose genotypes, GPG; Baker, 1965), and regards this as an important factor for geographical parthenogenesis. More recent experimental studies on parthenogenetic animals (Gade & Parker, 1997; Robinson et al., 2002; Vorburger et al., 2003) and apomictic plants (de Kovel & Jong, 1999) do not support the GPG hypothesis. The fact that clones are predominantly restricted to single localities indirectly contradicts the GPG hypothesis, because a general-purpose genotype would be expected to occupy broader distribution ranges than a specialized one. Nevertheless, as the GPG hypothesis is based on temporal heterogeneity of environments, snapshot-like diversity patterns may not show the potential of maintenance of advantageous genotypes over longer time periods.
The frozen niche variation (FNV) model (Vrijenhoek, 1984, 1994) suggests that arrays of apomictic lineages are regularly produced by sexual taxa by hybridization, and each lineage ‘freezes’ a portion of the genotypic variation of the sexual parent. Natural selection acts on the lineages and leads to niche partitioning among lineages; single lineages have a narrower ecological niche than sexuals, but all lineages together use the resource space more effectively than sexuals (Fig. 3d).
Niche partitioning among apomicts and/or sexuals vs apomicts has been observed in many studies on asexual animals (Vrijenhoek, 1994; Jokela et al., 1997; Semlitsch et al., 1997; Gray & Weeks, 2001) and plants (Vavrek et al., 1998; Hörandl et al., 2000; Meirmans et al., 2003; Verduijn et al., 2004; Paun et al., 2006a). Reviews of molecular population studies in plants support the importance of clonal diversity for range expansions (Gornall, 1999; Hörandl & Paun, In press). Niche differentiation may even prevent extinction of sexual reproduction, as shown by modelling the FNV model against the negative effects of accumulation of deleterious mutations in apomicts (Pound et al., 2004). For geographical parthenogenesis, the findings probably suggest a higher capacity of apomicts to colonize spatially heterogeneous environments better than their sexual relatives. Avoidance of genetic load and inbreeding depression may be an important initial advantage of asexual organisms in marginal and subdivided habitats, where populations exist as ‘metapopulations’ with high degrees of subdivision, local extinction and recolonization (Haag & Ebert, 2004). Moderately disturbed, patchy, man-made environments, which provide novel habitats such as meadows and pastures, may thus promote range expansions of apomicts. This could be an important factor for range expansions of apomicts when they occur in same latitude/altitude as sexuals, where advantages of apomixis related to colder climates are less plausible.
The problems of the model are that (1) niche differentiation is not a general feature, because a number of genera have co-occuring sexual and apomictic taxa; and (2) the model should be effective mainly within or near the area of sympatric sexuals, where arrays of new clones are regularly produced by hybridization. Increase of clonal diversity of apomicts in sympatric areas is probably caused by backcrossing to sexuals (reviewed in Hörandl et al., 2001; Hörandl & Paun, In press; Fig. 3d). But if the apomict colonizes areas distant from sexual ancestors, or if distribution areas of sexual species become separated, the only remaining sources of genotypic variation are facultative sexuality and mutations within lineages, thus drastically reducing the potential to create novel genotypes. That is, the observed extreme range expansions of apomictic complexes outside the sympatric sexual/apomictic range, and outside overlapping ranges of sexuals, are not easily explained by the FNV model alone. Here a combination of niche differentiation plus uniparental reproduction for stepwise colonization offers the best explanation (Fig. 3d).
Biotic interactions – the Red Queen hypothesis
The northern tendency of apomictic complexes has often been related to a lower capacity of asexuals to keep pace with coevolving parasites and pathogens. Asexual organisms are thought to have fewer disadvantages than sexuals in environments with few biotic interactions, such as in high altitudes, high latitudes and arid habitats (Bell, 1982; Glesener & Tilman, 1987).
The expected relation of biotic interactions to geographical patterns of apomicts is not supported in regions where apomicts and sexuals coexist in the same longitudinal range, where even mixed populations may occur (see examples above). Other contradictions are the facts that, for example in Fennoscandia, the abundance of apomictic species themselves clearly increases from north to south, and in Europe is highest in the temperate zone, but decreases southwards only relative to sexual species. Moreover, frequencies of apomicts become low in extremely high latitudes and altitudes, where biotic interactions are expected to be lowest (Asker & Jerling, 1992). In the European Alps, the phanerogamic flora in the glacier regions above 4000 m includes no apomictic taxa (Ozenda, 1988).
If biotic interactions are responsible for limitation of asexual distributions, it would be expected that high pressure of predators can drive asexuals to extinction. Long-term studies over >10 yr on parthenogenetic animals and apomictic weeds suggest that parasites drive host populations through cycles, in which the most common lineages decrease and rare lineages become frequent (Burdon et al., 1981; Lively & Jokela, 2002; Jokela et al., 2003). Nevertheless, the Red Queen model was not supported in other long-term studies on parthenogenetic freshwater snails (Ben-Ami & Heller, 2005), and in a clonal-sexual fish complex (Weeks, 1996). Resistance against rust fungi was observed in eight of 19 phenotypes of apomictic Chondrilla in Turkey (Espiau et al., 1998). In the grass Miscanthus sinensis most populations were infected by endophytic fungi, except apomictic M. sinensis var. condensatus (Chou et al., 2000).
The few studies cited above suggest that predators may indeed select for genetic variation, but on the other hand none of the studies supports the idea that predators can drive asexual organisms to extinction. Once apomixis has arisen, the recombination rate, and thus the ability to adapt to predators with a short turnover of generation, is expected to decrease, but this may be overcome by maintenance of clonal diversity.
Another general caveat is needed because of a likely change of response against predators resulting from changes in composition and expression of secondary metabolites in hybrids and polyploids. Hybrids may express all the secondary chemicals of the parental taxa; may fail to express certain parental chemicals; or may express novel chemicals that are absent in each parent. Hybridization may diversify chemical compounds, lead to introgression of resistance traits, facilitate host shifts, or act as sinks for herbivores (Orians, 2000). Differential expression and quantitative changes of metabolites have already been assessed in F1 hybrids (Kirk et al., 2005), and variation of chemical compounds tends to increase with the number of hybrid generations (Orians, 2000). From this information on sexual hybrids, it is to be expected that an apomict of hybrid origin has a different expression of chemical compounds than its sexual parents, and would also react differently to predators. Empirical studies of metabolomic profiles in apomictic plants, their response to predators, and how this could influence distribution patterns, are largely missing.
In summary, understanding of the impact of biotic interactions on geographical distributions of apomictic vs sexual plants is currently limited, and it remains uncertain whether biotic interactions are an important causal factor for geographical parthenogenesis.
Conclusion – a synthetic approach
The main conclusion of this review is that various factors contribute, to a certain extent, to geographical parthenogenesis. As discussed above, none of the factors alone is unequivocal and effective only within a certain context. Here I suggest a combination theory in a spatial and temporal context.
The tendency of apomictic plants to colonize previously glaciated areas is seen in a predominant origin of apomicts on the margins of former ice sheets, where extensive range fluctuations offered opportunities for hybridization and polyploidization. These processes probably triggered the genetic or epigenetic changes that have altered the reproductive pathway to make apomixis functional. This does not preclude origins of apomixis elsewhere, but makes the origin of apomixis in more southern regions less likely. A young age of apomictic complexes, plus a limited capacity for colonization after long-distance dispersal, restricts distribution areas mainly to their continents of origin.
In the temporal course of development, polyploidy and the hybrid nature of apomicts are important for the creation of initial basic genotypic diversity and genomic novelty. The shift to apomixis causes a strong reproductive barrier against the sexual parents, and helps to maintain and multiply apomictic lineages. In initial periods of coexistence of apomicts and sexuals, unidirectional hybridization in mixed sexual–apomictic populations may decrease frequencies of sexual individuals. During subsequent colonization of new areas and/or certain habitats, apomictic lineages may disperse more quickly than their self-incompatible sexual relatives because of uniparental reproduction. Higher initial frequencies of apomicts in new areas prevent sexual individuals from establishment. In the long run, apomicts may suffer from a lower potential to maintain and increase genetic diversity, especially in regions distant from the sexual area. Nevertheless, for the sexual relative the advantages of sexuality may be manifest too late to compensate for the initial distributional success of apomicts.
In spatial dimensions, advantages of apomictic plants in higher latitudes/altitudes over their sexual relatives are manifold: (1) a shortening of the reproductive pathway and faster seed development in shorter vegetation periods; (2) reproductive assurance if pollinators and mating partners are rare; and (3) potentially less pressure from predators because of reduced biotic interactions from colder climates. Together with the factors above, high altitudinal/latitudinal sexual relatives may be replaced by apomicts. This combination of factors may also explain an overall increase in apomicts relative to the rest of the sexual flora in northern regions (Asker & Jerling, 1992). In temperate or more southern regions, advantages related to a colder climate are perhaps not effective enough to explain the distributional success of apomicts. Here, niche differentiation of apomicts and sexuals may help to maintain sympatric areas. Clonal diversity allows niche differentiation among lineages and optimal use of resource space by apomictic lineages. This is probably important for range expansions in spatially and temporally heterogeneous environments, and may explain the tendency to man-made habitats.
Each factor has its limitations, and can be compensated by features of sexuality. Indirectly, these conclusions may shed light on the still frequently discussed question of why sex is predominant in nature and not replaced by asexual reproduction. (1) At least in angiosperms, the shift to apomixis requires complex changes in reproductive pathways. A predisposition of certain taxa, plus an opportunity for polyploidization and/or hybridization as trigger for the required genetic or epigenetic changes, must coincide to make apomixis functional; this happens much more rarely than other modifications of the sexual system (Table 1). (2) For establishment of apomixis, the newly arisen apomict has to compete with a large pool of sexual allopolyploid lineages with features similar to the apomict: genomic novelty plus an effective reproductive isolation mechanism. (3) Uniparental reproduction and maintenance of heterozygosity offer short-term advantages for colonization of new areas, but this can be compensated by self-fertilization of sexuals, which is more frequently established in angiosperms (Table 1). Sexuals can probably re-establish genetic variation more effectively after a bottleneck, and speciate more rapidly after strong geographical isolation. Apomicts, in contrast, need longer time periods and/or multiple colonizations for the creation of clonal diversity. (4) Selfing, niche differentiation, and isolation by different ploidy levels prevent total replacement of sexuals by introgression of apomixis via pollen. Exploration of novel niches and range expansions of apomicts probably require a certain level of clonal diversity that is best maintained near the sympatric sexual area or via stepwise migration. (5) Apomictic reproduction is advantageous mainly in colder or more extreme climates and/or certain ecological conditions: heterogeneous, moderately disturbed environments. Consequently, apomicts are perhaps more successful than sexuals because of not only a single factor, but a combination of factors acting together. This might not occur frequently enough to replace the well established and more flexible system of sexuality.
I wish to thank Peter Van Dijk, Michael Mogie, Hilde Nybom, Ovidiu Paun and Franz Hadacek for many stimulating discussions on the topic; and Tod Stuessy and two anonymous reviewers for valuable comments on the manuscript. The study was supported by the Austrian Research Foundation (FWF), project P-15975.