Data collection and summarizing

We first gathered reports already known to us with evidence for interspecific gene flow and polyploidy between peatmoss species. References from those papers were checked to identify other possible papers. Then, we conducted a literature search using the Web of Science Core Collection online database (ver. 5.26.2, Web of Science, Clarivate Analytics 2017) and in Google Scholar for articles published between 1960 and 2017. We searched for different combinations of such terms as ‘Sphagnum’, ‘peatmoss’, ‘introgression’, ‘hybridization’, ‘hybrid’, ‘polyploid’ and ‘admixture’. We performed the search in November 2017 but did not find new papers in addition to those we had collected earlier. To differentiate between reported cases of hybridization, we assigned the cases to the following groups: allopolyploid hybrids, admixed haploid individuals (hereafter – admixed individuals) and homoploid hybrids (see Glossary in Box 1).

Based on the collected information, we calculated a minimum coefficient of hybridization for each subgenus as the ratio of the number of identified hybridizing species to the total number of species within the subgenus. We also collected information about certain life-history traits of the identified parental species: mating system type, observed frequency of sporulation, spore colour, position of a species along the mire water table (‘hummock–hollow’) and the nutrient (‘poor–rich’) gradients (Eddy 1985, McQueen and Andrus 2007, Flatberg 2013, Johnson 2013, Kyrkjeeide et al. 2018), and maximum and minimum spore size (Suzuki 1958, McQueen and Andrus 2007, Kyrkjeeide et al. 2018). Mire is here used in a wide sense, also including moist heaths and forests.

Estimation of phylogenetic distance between the parents

We found no published phylogenetic tree including all species of interest that matched our objectives. In order to summarize the phylogeny of parental species, we constructed a composite cladogram using all available published phylogenetic trees containing species of interest. The cladogram was visualized with Dendroscope (ver. 3.5.8, Huson and Scornavacca 2012). In several cases, one or both parents could not be unambiguously identified. For example, there are 22 observed cases of species from subgenus Subsecunda being involved in hybridization, but parental species were only determined to subgenus in 11 cases, and to a species complex in two cases. As no information about the specific position in the phylogeny is available in these cases, we combined them for each subgenus into one ‘unknown species’ and placed it within the subgenus on the composite cladogram in order to visualize all reported evidence of hybridization simultaneously. Assuming, that all these unknown species represent different species, we added these cases to the number of unequivocally identified parental species in a subgenus. We then calculated a maximum hybridization coefficient by dividing this number by the total number of species in a subgenus. In this way, we can get an overview of the possible upper extent of hybridization for each subgenus.

The composite cladogram unites a broad spectrum of species from different subgenera, but it does not include many species within subgenera. Parental species might thus seem to be more related within subgenera than they actually are. Due to this, we counted the number of nodes separating parental species in each possible tree, and calculated the average number for each parental species pair as a measure of phylogenetic distance between the parents (Vellend et al. 2010). Because not all parents were identified, we estimated phylogenetic distances for 22 parental species pairs. The portion of the inter-subgeneric hybrids among these species pairs was the same as in the general sample.

Comparative analysis

We were interested in testing for associations between occurrence of hybridization on one hand, and life-history traits of the parental species on the other. Hence, we combined a dataset with seven life-history traits of parental species listed above with two additional factors: subgenus and intensity of hybridization, which corresponds to the number of hybrid species produced by each individual parental species. Sphagnum australe, S. irritans, S. ‘sp-3’ (Shaw et al. 2015) and S. ‘sp-4’ (Shaw et al. 2015) were excluded from the analysis because most of their life-history traits are unknown. As our data set contained both categorical and continuous variables, we explored it with factorial analysis of mixed data (FAMD) using the FactoMineR package in R (Lê et al. 2008), and used the missMDA package in R (Josse and Husson 2016) to account for missing trait values within some species. All analyses were conducted and visualized in the R statistical environment (ver. 3.4.1, < www.r-project.org>).

Occurrence of interspecific hybridization

There are 36 documented allopolyploid hybrids, seven cases of genetic admixture and one case of homoploid hybridization in our data set, respectively (Table 1, Fig. 1). Peatmoss autopolypoids have not been registered, although Shaw et al. (2012b) discuss potential autopolyploidy of diploid Sphagnum tescorum based on registered genetic admixture between parental S. fimbriatum and S. girgensohnii, which in turn might be the second evidence of homoploid hybridization for the genus. So far, we treat the latter as an example of admixture. The parentage is completely unknown for five allopolyploid hybrids and partially unknown for 18 allopolyploid and admixed hybrids. In total, there are 37 parental species, each producing from one to four hybrids (Fig. 2A, Table 2). Some species are involved in both admixture and polyploid hybridization (Table 2). Hybridization events often occur within subgenera, but as many as 13 out of 39 hybrids are the results of inter-subgeneric crosses (Table 1, Fig. 1).

Hybridization is common in all subgenera, and the fraction of species hybridizing varies from 9 to 20% (Fig. 2B). The maximum coefficient of hybridization is of the same magnitude and reaches 27% in Cuspidata, while the lowest coefficient is observed in subgenus Sphagnum, but here parentage is unknown for half of the registered hybrid species. In general, up to one out of five (21%) of peatmoss species potentially hybridize.

Table 1.

Polyploid hybrids and admixed haploid individuals registered in Sphagnum. Phylogenetic distance calculated as a number of nodes separating species in published phylogenetic trees (see 6, References). We refer to all studies which include information about 1) taxonomic status, 2) parentage, 3) ploidy level and/or 4) phylogenetic trees with the species of interest, which were used to obtain phylogenetic distance estimates. ^* homoploid hybrid, NA — data is not available.

Table continude.

Figure 1.

Hybridization in Sphagnum. All cases of allopolyploid and homoploid hybridization (species to the right) when at least one parent was identified (species to the left). Other cases, including admixture, are listed in Table 1. Coloured bars indicate subgenera as showed in the lower right corner. Dashed lines correspond to maternal parents, solid lines – to paternal or unidentified parents. Ploidy level is stated in brackets and otherwise is In, ^* – homoploid hybrid. Intersubgeneric hybrids are in bold, some of these hybrids are assigned to that subgenus, which their maternal plant belongs to. For references see Table 1.

Phylogenetic distance of parental species and life-history trait analysis

The majority of allopolyploid hybrids are produced by nonsister species, with a mean phylogenetic distance of 4.68 (SE = 0.56) nodes between parental species (Fig. 3). Distributions of phylogenetic distances between parents do not deviate from normal distribution (Shapiro test, p = 0.43 and p = 0.36 for parents of the allopolyploid hybrids and the admixed individuals, respectively). The average phylogenetic distance between parents of admixed individuals seems lower than between parents of allopolyploids, albeit insignificantly so (2.88 (SE = 0.41) versus 4.68 (SE = 0.56) nodes, respectively, Student's t-test, p = 0.06).

The FAMD based on the collected life-history trait information (Table 2) shows that subgenus and spore colour contribute the most to the variation in characters between parental species, followed by maximum spore size, mating system and sporulation frequency (Table 3). The contribution of spore colour to axis 1 is 23.1%, and contribution of the maximum spore size variable to axis 2 is 23.6% (Table 3). Contribution of subgenus variable to axis 1 and axis 2 is 23.4% and 16.7%, respectively (Table 3). The intensity of hybridization explains 2.6% of variance between species. Although we distinguish several groups, species producing different number of hybrids are scattered evenly across all of them (Fig. 4A). The FAMD individual factor plot shows that intensity of hybridization tends to be associated with polyoicous and monoicous reproductive systems, high sporulation frequency, poor habitats (low pH and few minerals), small spore sizes and high position along the water table (Fig. 4B). Data on spore sizes and sporulation frequency is unavailable for many parental species, because they are rarely or never observed with sporophytes.

Occurrence and a potential role of interspecific gene flow in speciation of peatmosses

The majority of hybrids in peatmosses result from allopolyploid hybridization (82%). Polyploidy represents a very important mechanism of speciation in these plants, and this seems primarily related to immediate postzygotic reproductive isolation between hybrid progeny and parents (Ricca and Shaw 2010, Abbott et al. 2013). Despite this, complete reproductive isolation is sometimes not established, and polyploids are able to backcross with their parents, preventing the establishment of new ‘distinct evolutionary lineages’ and also increasing the genetic diversities of both polyploid and parental species. In peatmosses, several allopolyploids are reported to undergo interploidal backcrossing with haploid parents (e.g. Sphagnum russowii, S. troendelagicum, S. missouricum, Flatberg et al. 2006, Ricca et al. 2011, Stenøien et al. 2011a).

It has been suggested that high levels of fixed heterozygosity can increase ecological amplitudes in hybrid plant taxa, even beyond the habitat and niche limitations of the parents (Levin 2002). Well-established allopolyploid species are relatively often found in habitats which parental species do not occupy. From this perspective, it seems that environmental heterogeneity can promote and contribute to the establishment of new hybrid species (Brochmann et al. 2004, Abbott et al. 2013). This might be the case for several peatmoss allopolyploids, for example S. missouricum and S. australe (Shaw et al. 2012a, Karlin et al. 2014). It could also be the other way around, when ecological divergence originates after the establishment of new species (Abbott et al. 2010). Polyploid plants are often characterized by greater vigour compared to their diploid progenitors, possibly facilitating their dispersal to and establishment in new habitats (Grant 1981). Brochmann et al. (2004) suggested that polyploid arctic plants might have been more successful in colonizing areas after the last glaciation than diploids, because fixed highly heterozygous duplicated genomes of polyploids contain much of the ancestral diversity In peatmosses, an increase of genetic diversity through polyploidization has been hypothesized to facilitate successful colonization of new habitats in allotriploid S. falcatulum (Karlin et al. 2014). Genetic diversity and frequency of polyploids in populations can also be increased via recurrent polyploidization, which thus may play an important role in the successful establishment of polyploid lineages (Ricca and Shaw 2010). This is probably the case for several peatmoss polyploids which have originated more than once, e.g. S. russowii (Shaw et al. 2003), S. jensenii (Såstad et al. 1999), S. carolinianum (Ricca et al. 2008), S. falcatulum and S. australe (Karlin 2014).

Figure 2.

Characteristics of hybridization across Sphagnum subgenera. (A) Intensity of hybridization across Sphagnum subgenera. Intensity of hybridization corresponds to the number of hybrids produced by an individual parental species. Bars on the X-axis represent subgenera, coloured according to the number of hybrid species produced by individual parental species as shown on the right. Y-axis represent number of parental species of each category within subgenera. Inter-subgeneric hybrid parents are assigned to that subgenus, which their maternal plants belongs to. (B) Occurrence of hybridization across Sphagnum subgenera. Maximum CH – the maximum coefficient of hybridization counted as the ratio of the maximum possible number of parental species to the total number of species in a subgenus. Minimum CH – the minimum coefficient of hybridization counted as the ratio of the number of the identified parental species to the total number of species in a subgenus.

Registered admixture between peatmoss species might result from hybrid speciation sensu stricto, but introgression in itself is not evidence of successful speciation. In order for speciation to take place, more complete reproductive isolation between newly formed admixed lineages and parental species must subsequently develop (Abbott et al. 2013). Nonetheless, extensive genomic admixture clearly indicates an early phase in speciation. In peatmosses, one case of admixture (S. girgensohnii and S. fimbriatum) presumably resulted in polyploidization and subsequent establishment of a separate species, S. tescorum (Shaw et al. 2012b). In other cases, however, observed admixture is an ongoing process in zones of contact, which does not affect distinctiveness of parental gene pools, as for S. capillifolium × S. warnstorfii (Cronberg 1997), S. capillifolium × S. quinquefarium (Cronberg and Natcheva 2002), S. capillifolium × S. rubellum (Cronberg 1996b), S. angustifolium × S. flexuosum (Szurdoki et al. 2014), or is a consequence of secondary contact, as for S. magellanicum expanse × S. magellanicum margin (cf. Yousefi et al. 2017, S. divinum and S. medium, respectively, in Hassel et al. 2018). There are also examples of past hybridization events, e.g. S. rubellum × S. capillifolium (Natcheva and Cronberg 2003) and S. austinii × S. affine (Thingsgaard 2001). The latter has been suggested to represent an example of past adaptive introgression (Thingsgaard 2001). Comparing to allopolyploid hybridization, admixture and homoploid hybridization in peatmosses might be underestimated since hybrid individuals remain undetected because of their morphological resemblance to one of the parents.

The age of admixed taxa can be important for distinguishing hybrid speciation from more or less “neutral” admixture (Abbott et al. 2013). Yet, there is modest information available about the age of hybrid taxa in peatmosses, particularly for admixed hybrids. Several allopolyploid species seem to have originated before the last glaciation maximum, e.g. S. troendelagicum (Stenøien et al. 2011a); S. guwassanense, S. triseriporum (Shaw et al. 2013); S. alaskense (Kyrkjeeide et al. 2016a), while others probably are of more recent origin, e.g. S. jensenii (Såstad et al. 1999), S. falcatulum (Karlin et al. 2013) and both the diploid and triploid S. australe cf. Karlin et al. (2009). For admixed hybrids the age is only estimated for those formed between S. divinum and S. medium, which seem to be around 20 000 years old (Yousefi et al. 2017).

Table 2.

Reproductive and microhabitat characteristics of identified parental species. Key to the columns abbreviations: 5 – Number of hybrid species produced by the parental one (Table 1); 6 – Reproductive system: m – monoicous, d – dioicous, p – polyoicous (including andropolyoicous species sensu Kyrkjeeide et al. 2018, i.e. which are reported with separate male plants, but with certainty not with pure female plants); 7 – Sporulation frequency: F – frequent, R – rare; 8 – Position along the water table (‘hummock–hollow’) mire gradient: H – hummock, L – lawns/carpets, I – intermediate (both in hummocks and lawns/carpets); 9 – Position along the nutrient (‘poor–rich’) mire gradient: R – rich fen habitat, P – poor fen and bog habitat, I – intermediate fen; 10 – Spore colour, sources: Eddy 1985, McQueen and Andrus 2007, Flatberg 2013, Johnson 2013, Hassel et al. 2018, Kyrkjeeide et al. 2018; 11 – Maximum spore size, µm; 12 – Minimum spore size, µm, sources: Suzuki 1958, McQueen and Andrus 2007, Kyrkjeeide et al. 2018; ^* – homoploid hybrid; ♂ – male parent of the hybrid, ♀ – female parent of the hybrid; NA – data is not available. Mire is here used in a wide sense, also including moist heaths and forests.

Abbott et al. (2013) point out that admixed individuals resulted from secondary contact can evolve into a separate species by occupying a niche not yet occupied by its parents. Early homoploid hybrid plant lineages are characterized by rapid and sometimes large-scale changes in gene expression patterns, which might increase phenotypic novelty and facilitate differentiation into new species (Abbott et al. 2010). Even if reproductive barriers are not complete at the initial stage of speciation, environmental based (exogenous) selection can maintain the distinctness of hybrid species. It is, however, not clear whether admixed peatmoss hybrids occupy different ecological spaces and are reproductively isolated from parents at any level. In addition, their morphological distinctiveness has not been examined in the majority of cases. Further research is thus needed to assess evolutionary significance of admixture observed in peatmosses and to define their taxonomical status.

Figure 3.

Phylogenetic distance between parental species in Sphagnum. X-axis represents averaged rounded number of nodes separating identified parental species pair in published phylogenetic trees, Y-axis represents counts of parental species pairs with the corresponding number of nodes. Dark grey bars represent counts for parents of admixed haploid individuals, light-grey bars – parents of allopolyploid hybrids as showed on the right. n = 22 species pairs.

Does phylogenetic distance explain occurrence of interspecific hybridization within the genus?

It has been suggested that certain degrees of genetic divergence between parents is required in order for allopolyploidization to occur as a result of impaired meiotic chromosome pairing in peatmosses (Natcheva and Cronberg 2007, Karlin et al. 2014). Although the only recorded homoploid hybrid species in peatmosses (S. contortum) is not included in our analysis because of unknown parentage, we do observe that parents of allopolyploids are less related compared to parents of haploid admixed individuals. At the same time, several allopolyploid hybrids are formed by closely related species, while several distantly related species are involved in admixture, which is not expected assuming that divergence ultimately leads to problems in meiosis during spore production. Patterns observed in allopolyploids and admixed species indicates that phylogenetic distance in itself does not define the success of interspecific crosses. It is worth noting that the homoploid S. contortum is thought to be an inter-subgeneric species, whose parents are rather distantly related. Sphagnum contortum also hybridizes inter-subgenerically with S. subsecundum, producing admixed individuals. But it cannot be ruled out that instead of being a homoploid hybrid, this species could have originated through polyploidization followed by chromosome number reduction, or also through introgression of genes between the subgenera and subsequent divergent speciation (Shaw et al. 2016).

Table 3.

Contribution of the variables to the axes in the FAMD. ^* – Number of species produced by individual parental species.

Levin (2013) argues that low divergence between parents leads to formation of homoploids, whereas strong and modest divergence result in strict and segmental polyploids (i.e. allopolyploids whose chromosomes are partially homologous), respectively. These patterns are observed in vascular plants (Chapman and Burke 2007), and might also explain formation of allopolyploid hybrids by closely related parents in peatmosses. It is unclear whether these hybrids are strict or segmental polyploids since strict disomic inheritance usually serves as a null-hypothesis in revealing allopolyploids (Karlin and Smouse 2017). So far, evidence of recombination between parental genomes has only been registered in two allopolyploids: S. tescorum (Shaw et al. 2012b) and S. palustre, the latter a hybrid species with unknown parentage (Stenøien et al. 2014). Otherwise, recombination between parental genomes has only been reported for admixed haploid individuals (Natcheva and Cronberg 2007).

Do life-history traits explain intensity of interspecific hybridization?

We show that intensity of hybridization tends to be associated with polyoicous and monoicous reproductive systems, high sporulation frequency, poor habitats (low pH and low concentration of minerals), small spore size and light spore colour. The latter is in line with theoretical expectations, as small spores and high sporulation frequency can give an advantage of dispersal to longer distances. More frequent hybridization found in monoicous and polyoicous species might result from their higher reproductive and dispersal success, and hence colonization ability, compared to dioicous species. Consequently, we can expect the portion of dioicous species to be smaller among hybridizing species than among non-parental species. This is indeed what we observe, even though we find that the difference is insignificant. Yet, one of the two species that have produced the highest observed number of hybrids is S. cuspidatum, a dioicous species which have the largest spores within the genus and rarely sporulate (Glime 2017b). According to the conducted FAMD, intensity of hybridization only explains about 2.6% of the total variance in life-history traits between species. Thus, other factors might certainly be more important in explaining the observed contradictions.

Figure 4.

Two-dimensional FAMD plots for 33 parental species. Based on seven life-history variables (Table 2, column 6–12), as well as subgenus and intensity of hybridization as factors (Table 2, column 3 and 5, respectively). Percentage of the variance explained by the axis is given along each axis. (A) Individual species plot. Size and colour of point markers corresponds to the number of hybrids produced by the species and to the subgenus, respectively, as showed on the upper right corner. Key to the species abbreviations: aff - S. affine, ang - S. angustifolium, ann - S. annulatum, aur - S. auriculatum, aust - S. austinii, balt - S. balticum, cap - S. capillifolium, cont - S. contortum, cusp - S. cuspidatum, div - S. divinum, fimb - S. fimbriatum, gir - S. girghensonii, inc - S. incundum, inx - S. inexspectatum, kush - S. kushiroense, les - S. lescurii, lin - S. lindbergii, med - S. medium, miss - S. missouricum, miya - S. miyabeanum, ort - S. orientale, pla - S. platyphyllum, quin - S. quinquefarium, rec - S. recurvum, rub - S. rubellum, rus - S. russowii, slv - S. slooveri, str - S. strictum, sub - S. subnitens, subs - S. subsecundum, warn - S. warnstorfii, ten - S. tenellum. (B) Individual factor plot for quantitative and qualitative variables. Shape and colour of point markers correspond to different variables used in analysis as showed on the upper right corner. Key to the factor abbreviations: Reproductive system: di – dioicous, mono – monoicous, poly – polyoicous (including andro-polyoicous species sensu Kyrkjeeide et al. 2018, i.e. which are reported with separate male plants, but with certainty not with pure female plants); Spore frequency: freq_sp – frequent sporulation, rare_sp – rare sporulation; Water gradient (i.e. position along the water table (‘hummock– hollow’) mire gradient): hummock – hummock, low – lawns/ carpets, intermediate – intermediate (both in hummocks and lawns/ carpets); Nutrient gradient (i.e. position along the nutrient (‘poor– rich’) mire gradient): rich – rich fen habitat, poor–poor fen and bog habitat, interm – intermediate fen. Mire is here used in a wide sense, also including moist heaths and forests. Black arrows correspond to maximum and minimum spore sizes, and the red arrow – to the number of species produced by individual parental species, according to the labels. (C) A scree plot from the FAMD: X-axis represents Axis 1–5 according to the labels, Y-axis represents the percentage of variance explained by the axis.

Johnson et al. (2015) showed a considerable phylogenetic signal for interspecific variation along the hummock–hollow gradient in peatlands. Based on this finding, we treat the position along the water table as a phylogenetically constrained trait and do not interpret how it is associated with intensity of hybridization. The same apparently applies to spore colour (Flatberg 2013). Phylogenetic constraints imply that traits of related species cannot be considered as independent and identically distributed, by that violating assumptions of most of the statistical tests (Sober and Orzack 2001). Because of the amount and incompleteness of the available data, we were not able to account for relatedness in the FAMD. In light of this, it should be kept in mind that the relatedness between species might contribute to the patterns we observe.

Nevertheless, the association between more intensive hybridization and monoico us and polyoicous reproductive system, high sporulation frequency and smaller spores in peatmosses makes biological sense and provides a rationale for further testing. Unfortunately, there is lacking knowledge about mating system and other life-history traits in many parental and non-parental species. Available floras (Chien et al. 1999, McQueen and Andrus 2007, Flatberg 2013) list 104 Sphagnum species out of 289 species known worldwide (McQueen and Andrus 2007, Michaelis 2011), mostly describing the peatmoss diversity in the Northern Hemisphere. In particular, mating system is only known for 89 of the listed species, and the fraction of polyoicous species might be generally underestimated (Kyrkjeeide et al. 2018). To that end, it is likely that with more knowledge about the reproductive biology of peatmosses, we will be able to show the importance of these life-history traits for the occurrence of hybridization.

What other factors can potentially affect interspecific gene flow in Sphagnum?

There are other factors that might explain occurrence and intensity of interspecific hybridization in peatmosses, including levels of intraspecific gene flow. Generally, interspecific gene flow has been viewed insignificant compared to intraspecific rates of gene exchange (Mayr 1942, Ehrlich and Raven 1969). However, recent metaanalyses of a range of different organism groups show that distribution of interspecific and intraspecific rates of gene flow sometimes overlaps (Hey and Pinho 2012). Substantial intraspecific gene flow and frequent mating could in itself lead to frequent introgression between highly dispersing species in plants (Levin and Kers ter 1974, Levin 1979), leading to a positive correlation between intra- and interspecific gene flow rates. Despite the high number of studies linking patterns of inter- and intraspecific gene flow and speciation in plants (Currat et al. 2008, Zhou et al. 2010), this has not been studied in bryophytes. Because of high ability of long-distance dispersal, peatmosses have high levels of intraspecific gene flow between populations, located even on different continents (Kyrkjeeide et al. 2016b, Désamoré et al. 2016). Therefore, intraspecific gene flow can potentially be important in explaining levels of interspecific introgression between species.

It was shown, that interaction of genetic and demographical factors, such as population size, time of season and relatedness, is important for explaining gene flow rates in plants (Goodell et al. 1997). Thus, studying speciation by gene flow in peatmosses primarily requires clarification of the relationships between inter- and intraspecific gene flow using genomic data and accounting for possible interaction between different factors.