Ecological differentiation, lack of hybrids involving diploids, and asymmetric gene flow between polyploids in narrow contact zones of Senecio carniolicus (syn. Jacobaea carniolica, Asteraceae)

Areas of immediate contact of different cytotypes offer a unique opportunity to study evolutionary dynamics within heteroploid species and to assess isolation mechanisms governing coexistence of cytotypes of different ploidy. The degree of reproductive isolation of cytotypes, that is, the frequency of heteroploid crosses and subsequent formation of viable and (partly) fertile hybrids, plays a crucial role for the long-term integrity of lineages in contact zones. Here, we assessed fine-scale distribution, spatial clustering, and ecological niches as well as patterns of gene flow in parental and hybrid cytotypes in zones of immediate contact of di-, tetra-, and hexaploid Senecio carniolicus (Asteraceae) in the Eastern Alps. Cytotypes were spatially separated also at the investigated microscale; the strongest spatial separation was observed for the fully interfertile tetra- and hexaploids. The three main cytotypes showed highly significant niche differences, which were, however, weaker than across their entire distribution ranges in the Eastern Alps. Individuals with intermediate ploidy levels were found neither in the diploid/tetraploid nor in the diploid/hexaploid contact zones indicating strong reproductive barriers. In contrast, pentaploid individuals were frequent in the tetraploid/hexaploid contact zone, albeit limited to a narrow strip in the immediate contact zone of their parental cytotypes. AFLP fingerprinting data revealed introgressive gene flow mediated by pentaploid hybrids from tetra- to hexaploid individuals, but not vice versa. The ecological niche of pentaploids differed significantly from that of tetraploids but not from hexaploids.


Introduction
Ecological differentiation is among the most important mechanisms of reproductive isolation among cytotypes of different ploidy (for simplicity termed "cytotypes" from here on) in heteroploid plant species (Levin 1983). It may arise as a direct consequence of genome duplication (Otto and Whitton 2000) or result from subsequent disruptive selection (Petit et al. 1999;Ramsey and Schemske 2002). The degree of ecological differentiation ranges from shifts in the relative abundance of accompanying species (Johnson et al. 2003) via different preferences along ecological gradients within the same habitat type (Raabov a et al. 2008) to separation of cytotypes into formations of different physiognomy (Lumaret et al. 1987). Allopolyploids are expected to show stronger ecological differentiation than autopolyploids due to the merging of two differentiated genomes (Soltis and Soltis 2009;Parisod et al. 2010), but there is also evidence for adaptive niche divergence in autopolyploids (Parisod et al. 2010, and references therein).
Contact zones of cytotypeswe use the term in a strict sense to encompass areas of close spatial proximity of individuals of different ploidycan be observed in many taxa. They provide a unique opportunity to assess isolation mechanisms governing coexistence of cytotypes, such as ecological differentiation (Petit et al. 1999). Major aspects include avoidance of competition and patterns of gene flow between parental cytotypes, potentially leading to long-term coexistence of cytotypes or the formation of new hybrids (Kol a r et al. 2009;H€ ulber et al. 2011). To date, niche differentiation among cytotypes has been assessed by comparing single-cytotype populations (e.g., Manzaneda et al. 2012;McIntyre 2012;Martin and Husband 2013) or by large-scale surveys of the distribution and ecological differentiation of parapatric (Hardy et al. 2000) and sympatric cytotypes (Sonnleitner et al. 2010;Sabara et al. 2013). Patterns of niche differentiation in areas of immediate contact allow inferring whether contact zones represent hybrid zones, that is, habitats suited for both cytotypes, or mosaic zones, that is, a microspatial mixture of habitats each suited for a single cytotype. In the first case, niche differences in contact zones are expected to be smaller compared to both adjacent pure populations and the entire distribution ranges of the cytotypes, whereas no such reduction in niche differences is expected in case of mosaic zones.
The degree of reproductive isolation of cytotypes, that is, the frequency of heteroploid crosses and subsequent formation of viable and (partly) fertile hybrids, plays a crucial role for the long-term integrity of lineages in contact zones (Barton and Hewitt 1985) in general, and for the local maintenance of ploidy variation in particular Madlung 2013). For instance, gene flow via individuals of intermediate ploidy may lead to introgression and thus an increase of genetic diversity, transfer of adaptations, or the emergence of new adaptations in the receiving lineage (Soltis and Rieseberg 1986;Rieseberg et al. 1996;Petit et al. 1999). As a consequence, introgressed lineages tend to have broader niches than their pure counterparts (Choler et al. 2004). In heteroploid systems, gene flow and introgression have so far mostly been observed in diploid/tetraploid contact zones (Neuffer et al. 1999;St ahlberg and Hedr en 2009), while studies on genetic interactions between lineages of higher ploidy are largely lacking for wild species. In two species of Rorippa (Brassicaceae), bidirectional introgression between diploids and polyploids (tetra-and hexaploids) was found (Bleeker 2003), but the consequences of introgression for niche evolution were not explored.
Hybrid cytotypes emerging in contact zones face competition with the parental cytotypes. Establishment, persistence, and genetic integrity of hybrid cytotypes will be affected by the magnitude of niche divergence from parental cytotypes, conferring spatial separation and, thereby, reducing competitive interactions and the incidence of heteroploid crosses. Although niche differentiation among cytotypes was documented even in narrow contact zones (Mr az et al. 2012) and odd-ploid hybrid cytotypes were found in many model systems (e.g., Sabara et al. 2013), little is known on niches of hybrid cytotypes and their ecological position relative to their parents. St ahlberg and Hedr en (2009) reported an intermediate position of triploid hybrids in mixed diploid/tetraploid populations of the Dactylorhiza maculata group, albeit without statistical evaluation due to the low number of triploids.
A well-suited system to study mechanisms of ploidy coexistence is the high mountain plant Senecio carniolicus (Asteraceae). This species comprises three main cytotypes (diploids, tetraploids, hexaploids) co-occurring in every conceivable combination across the distributional range in the Alps (Sonnleitner et al. 2010). Cytotypes are ecologically differentiated on a large scale (Sonnleitner et al. 2010); so far, only the frequently co-occurring diploids and hexaploids were shown to occupy different niches also on a local scale (Sch€ onswetter et al. 2007;H€ ulber et al. 2009). Cytotypes show low crossability (crosses between diploids and polyploids) or are interfertile (crosses between polyploids; Sonnleitner et al. 2013), although a range-wide survey of natural populations revealed only low frequencies (< 1%) of hybrid cytotypes (Sonnleitner et al. 2010). Here, we analyze the microspatial, ecological, and genetic structure of narrow contact zones. Specifically, we address the following questions: (1) Does the occurrence of hybrid cytotypes in contact areas correspond to patterns of crossability of cytotypes? (2) Can the ecological differentiation of main cytotypes observed at large spatial scales also be found in areas of immediate contact? Do ecological requirements of hybrids differ from those of the parental cytotypes? (3) What are the patterns of gene flow between the interfertile polyploid cytotypes? Is there evidence for the presence of F2 or later-generation individuals, suggesting at least partial fertility of F1 hybrids? (4) Is there indication for broadening of the ecological niche in introgressed individuals?

Study species
Senecio carniolicus Willd. (syn. Jacobaea carniolica (Willd.) Schrank) is a herbaceous perennial common on acidic bedrock in the alpine to subnival belt of the Eastern Alps and the Carpathians. It constitutes a polyploid complex comprising mainly diploids (2n = 2x = 40), tetraploids (2n = 4x = 80), and hexaploids (2n = 6x = 120) in the Eastern Alps and only hexaploids in the Carpathians (Suda et al. 2007;Sonnleitner et al. 2010). The chromosome number of 40 does not correspond to the diploid level when taking the entire tribe Senecioneae into account but rather represents the lowest number encountered in the "Incani Clade", where S. carniolicus belongs to (Pelser et al. 2003;Escobar Garc ıa et al. 2012). In contrast to the majority of heteroploid taxa, S. carniolicus does not form a single contact zone containing otherwise geographically well-separated cytotypes (Husband and Schemske 1998;Hardy et al. 2000;Mand akov a and M€ unzbergov a 2006;Spaniel et al. 2008); instead, various combinations of cytotypes occur through-out major parts of the Eastern Alps (Suda et al. 2007;Sonnleitner et al. 2010). Of 100 investigated sample sites, diploids and hexaploids, tetraploids and hexaploids, and diploids and tetraploids co-occur in 28, five, and three sites, respectively, and all three cytotypes co-occur in eight sample sites. Molecular genetic evidence suggests that the polyploid cytotypes are autopolyploid derivatives of a diploid lineage distributed in the easternmost Alps (M. Winkler, G. Pedro Escobar, R. Flatscher, M. Sonnleitner, J. Suda, K. H€ ulber, P. Sch€ onswetter, G.M. Schneeweiss, unpublished data). Strong genetic divergence between the ancestral eastern diploid lineage and its polyploid derivatives as well as weaker but consistent differentiation between tetraploids and hexaploids renders ongoing polytopic origin of the polyploids unlikely (M. Winkler et al., unpublished data), which is in line with consistent morphological differentiation (Flatscher 2010; Fig. 1). Despite substantial habitat segregation (Sonnleitner et al. 2010), individuals of different cytotypes commonly occur in close spatial proximity (less than one meter; H€ ulber et al. 2009), making in situ heteroploid pollination likely.

Field work
Three mountains with contact zones of two main cytotypes of S. carniolicus were selected: Grosser Rosennock carniolicus individual was determined with a laser distance meter (Leica DISTO D5, Leica Geosystems, Heerbrugg, Switzerland). The DNA-ploidy level of all individuals was determined from silica-dried leaf material using flow cytometry (see Sonnleitner et al. 2010 for details); highresolution histograms (with coefficients of variation of G0/G1 peaks of S. carniolicus samples below 3%) were achieved in more than 92% of analyses. Presence of vascular plant species occurring within a radius of 0.2 m around each Senecio individual was recorded; data for Sadnig were taken from H€ ulber et al. (2009).

DNA extraction, AFLP fingerprinting, and data analysis
Of all individuals sampled in the tetraploid/hexaploid contact zone, total genomic DNA was extracted from similar amounts of dried tissue (ca. 10 mg) with the DNeasy 96 plant mini kit (Qiagen, Hilden, Germany) following the manufacturer's protocol. The AFLP procedure followed Escobar Garc ıa et al. (2012). Six plants were extracted twice to test the reproducibility of AFLP fingerprinting (Bonin et al. 2004). From the restriction/ligation step onwards, 13 samples were replicated twice, and seven samples were used as replicates between PCR plates, and therefore replicated in every plate. Fragments were scored manually using Genographer 1.6 (version no longer available).
The error rate was calculated as the ratio of mismatches (scoring of 0 vs. 1) over phenotypic comparisons in AFLP profiles of replicated individuals (Bonin et al. 2004). Nonreproducible fragments were excluded from the analyses. Monomorphic fragments and those present/ absent in all but one individual were removed from the dataset to avoid biased parameter estimates (Bonin et al. 2004). Intercytotype gene flow was inferred with NewHybrids (version 1.1beta; Anderson and Thompson 2002;Anderson 2008), which allows for the accommodation of dominant multilocus markers such as AFLPs (Anderson 2008). The posterior probability that each sampled individual belongs to each of several classes (parents, F1 and F2 hybrids, backcrosses) is computed by Markov chain Monte Carlo (MCMC) in a Bayesian model-based clustering framework. The probability of class membership was computed with the default settings, without prior information on hybrid status, and using 1.3 million generations following a burn-in of 100,000 generations.

Analyses of ecological data
Characterization of environmental conditions around sampled individuals was achieved via unweighted mean Landolt indicator values (Landolt 2010) of all vascular plant species (except Senecio carniolicus) per circular plot of 0.2 m radius. Landolt indicator values describe ecological requirements of species in terms of temperature (T), light (L), soil moisture (F), soil reaction (R), nutrients (N) and soil humus content (H), and range from 1 (low) to 5 (high). Niche differences among cytotypes in contact zones were tested by comparing mean indicator values among cytotypes using a multivariate analysis of variance (MANOVA). A principal component analysis (PCA) using the same indicator values but standardized to zero mean and unit variance was applied to attain a graphical illustration of the cytotypes' niches. Spatial aggregation/ segregation of cytotypes was tested via Mantel tests correlating a pairwise cytotype "distance" among individuals (0 and 1 for the same and for different ploidy, respectively) with the geographic distances; Kendall's tau coefficient was statistically evaluated by 999 randomizations. A Monte Carlo randomization technique was applied to test whether the niche differences in the contact zones are smaller than those observed across the Eastern Alps (Sonnleitner et al. 2010). The empirical F-value of the MANOVA test for niche differentiation in the contact zone was compared against a null distribution of F-values generated from randomly chosen individuals of the corresponding cytotypes from the aforementioned survey; the sample size in each of the 9999 permutations equals the number of individuals per cytotype in the contact zone. All analyses were performed separately for each of the three contact zones.

Results
A total of 181, 275 and 190 individuals were recorded in the three contact zones Rosennock (diploid/tetraploid), Sadnig (diploid/hexaploid) and Hoazh€ ohe (tetraploid/ hexaploid), respectively. In the diploid/tetraploid and in the diploid/hexaploid contact zones, no individuals with the expected intermediate ploidy were found; one pentaploid plant found in the diploid/hexaploid contact zone most likely arose because of the involvement of an unreduced gamete of the diploid and was disregarded in further analyses. In contrast, within the tetraploid/hexaploid contact zone, 26 pentaploid individuals were observed. Mantel tests revealed highly significant spatial clustering of main cytotypes in the contact zones (Table 1; P = 0.001 for each pairwise comparison). Pentaploids were spatially significantly separated from tetraploids (P = 0.044), but not from hexaploids (P = 0.816). diploids and hexaploids showed stronger spatial clustering in their contact zone than diploids and tetraploids ( Fig. 2; Table 1). Tetra-and hexaploids formed largely pure clusters, but pentaploids were intermixed with both parental cytotypes (Fig. 2). Pentaploids were restricted to a narrow, 6-to 7-m-wide strip at the immediate contact of tetra-and hexaploids, where the three cytotypes were approximately equally abundant (Fig. 2).
The main cytotypes were ecologically highly significantly differentiated ( Table 1). The strongest contrast was found in the diploid/hexaploid contact zone followed by the tetraploid/hexaploid zone, while the weakest contrast was between diploids and tetraploids (Fig. 3). The niche of pentaploids was significantly different from tetraploids, but not from hexaploids. Niche differentiation between main cytotypes was significantly lower in the three contact zones than in their overall distribution (P = 0.004, P < 0.001 and P = 0.003 for the comparison of diploids/ tetraploids, diploids/hexaploids, and tetraploids/hexaploids, respectively; Fig. 4).
The three AFLP primer combinations yielded 131 unambiguous polymorphic fragments after the removal of 14 nonreproducible, four singular, and 101 homogeneous markers. Seven individuals with nonreproducible AFLP profiles were removed from the dataset, resulting in a total of 183 analyzed individuals. In the AFLP profiles from replicated samples, 451 differences were observed of 14,250 phenotypic comparisons, resulting in an error rate of 3.16%. Almost all tetraploid individuals were classified as Tetraploid Parents NH (hybrid classes suggested by NewHybrids are marked by capitalization and the superscript "NH" ), and pentaploids were predominantly classified as F2 NH . In contrast, less than two-thirds of the hexaploid individuals were Hexaploid Parents NH , the remaining ones falling into classes F2 NH and Backcrosses F1 9 Hexaploid Parents NH . None of the individuals was categorized as F1 NH ( Table 2). The mean posterior probability of class membership of an individual was highest for Tetraploid Parents NH . Penta-and hexaploid individuals showed a highly admixed class membership (Fig. 5).

Discussion
The spatial distribution of cytotypes provides valuable insights into evolutionary processes shaping polyploid complexes. In a mixed-ploidy populationindependent of its origin as primary or secondary hybrid zonethe frequency-dependent mating disadvantage (Felber-Girard et al. 1996) will progressively remove the less frequent cytotype unless a sufficient degree of assortative mating ensures its integrity (Levin 1975;Fowler and Levin 1984;Rodr ıguez 1996). Among others, lineage integrity may be fostered by microspatial segregation based on divergent ecological preferences. In accordance with these expectations, the three main cytotypes present in Eastern Alpine populations of the heteroploid mountain plant Senecio carniolicus, which exhibit individual biogeographic patterns across their distribution range (Sonnleitner et al. 2010) and spatial segregation at the population level Interfertile cytotypes are expected to be strongly clustered in contact zones, whereas incompatible cytotypes may be less segregated (e.g., Castro et al. 2012). Supporting these expectations, we found the strongest spatial separation for the fully interfertile tetraploids and hexaploids (Table 1).
For the incompatible diploids and polyploids, other factors such as the fine-grained mosaic of alpine habitats caused by steep abiotic gradients or spatial autocorrelation as a consequence of leptokurtic dispersal kernels skewed toward short dispersal (Nathan and Muller-Landau 2000) may be of greater relevance. However, we cannot exclude that these factors contribute to the strong segregation of tetra-and hexaploids. Diploid, tetraploid, and hexaploid S. carniolicus showed highly significant niche differences within the pairwise contact zones (Table 1; Fig. 3). Such ecological differentiation , as well as for diploid, tetraploid, and hexaploid Solidago altissima (Richardson and Hanks 2011). In contrast, Halverson et al. (2008) found a random spatial pattern suggesting no habitat preferences, and Hanzl et al. (2014) detected no ecological shifts between diploid and autotetraploid Knautia. Along the same line, Keeler (1992) identified no relationship between cytotype and grazing, burning regime, or water availability in Andropogon gerardii. Eco-logical differentiation among cytotypes of S. carniolicus was significantly weaker in the contact zones than across their entire distribution ranges in the Eastern Alps (Fig. 4). This is likely due to the more restricted amplitude of ecological gradients on this small spatial scale, where historic and biogeographic effects are expected to be marginal or lacking (Wiens and Donoghue 2004). Despite introgressive gene flow from tetraploids to hexaploids mediated by pentaploid hybrids (Fig. 5) direct gene flow is unlikely because the ploidy of unreduced and reduced gametes of the lower and higher ploid cytotype, respectively, do not match as in diploid/tetraploid systemswe found no evidence for niche convergence in the tetraploid/hexaploid contact zone. Niche differentiation between tetraploids and hexaploids was intermediate between the stronger and weaker differentiation of diploids and hexaploids, and of diploids and tetraploids, respectively; this pattern remained stable when only introgressed hexaploids (F2 NH , Backcrosses F1 9 Hexaploid Parents NH ; Fig. 5) were considered. Occurrence of hybrid cytotypes differed strongly among the contact zones and agrees well with patterns of cross compatibility (Sonnleitner et al. 2013). Not a single triploid or tetraploid plant was found in the diploid/tetraploid and diploid/hexaploid contact zones, whereas pentaploid hybrids were frequent in the tetraploid/hexaploid contact zone (Fig. 2). Reproductive barriers between diploid and polyploid plants appear to be common (e.g., Castro et al. 2012;M€ unzbergov a et al. 2013), whereas different polyploid cytotypes might be interfertile to some extent (Schneider 1958). Pentaploid individuals, which are vigorous, flower regularly, and set well-developed seeds (M. Winkler and M. Sonnleitner, personal observation), were (A) (B) (C) Figure 4. Niche differentiation among cytotypes of Senecio carniolicus compared between three narrow contact zones (A: diploid/tetraploid; B: diploid/hexaploid; C: tetraploid/hexaploid) and the overall distribution range of cytotypes in the Eastern Alps. The x-axis represents F-values derived from multivariate analyses of variance (MANOVA) testing for pairwise niche differentiation between cytotypes, while the y-axis represents their frequency. The dashed line and the histogram are the empirical value in the contact zone and the distribution of simulated values derived from a Monte Carlo randomization applying 9999 permutations. Differences were highly significant (P = 0.004, P < 0.001 and P = 0.003, respectively).  (13) Values represent the percentage of individuals (number of individuals in parentheses) for each cytotype with predominant posterior probability of membership to a specific hybrid class. The most frequent hybrid class in each cytotype is given in bold.
limited to a narrow, a few meters wide strip in the immediate contact zone of their fully interfertile parental cytotypes (Fig. 2C). This spatial restriction together with the scarcity of contact areas between tetraploids and hexaploids most likely explains the low frequency (~0.7%) of pentaploids in a survey of 100 populations throughout the Eastern Alps (Sonnleitner et al. 2010). Although meiotic irregularities should strongly restrict the reproduction of pentaploids as aneuploid seeds are frequently inviable or at least less viable than euploid ones (Comai 2005), AFLP fingerprinting data revealed introgressive gene flow mediated by pentaploid hybrids from tetraploid to hexaploid individuals, but not vice versa (Fig. 5). Presence of primary (F1 NH ) hybrids was not supported, probably indicating that their establishmentdespite the interfertility of the two polyploid cytotypesdid not occur recently and may be connected to specific ecological conditions or disturbances (Levin et al. 1996). In contrast, we detected a high number of second-generation hybrids, that is F2 NH (i.e., F1 9 F1 crosses) as well as Backcrosses F1 9 Hexaploid Parents NH (Fig. 5). Preliminary data from flow cytometric seed screening (J. Suda, unpublished data) revealed that pentaploids yield embryos with hexaploid and aneuploid (DNA contents being intermediate between tetraploids and pentaploids) genome size lending support to the pattern of gene flow suggested by genetics. Aneuploid offspring of pentaploids was frequently found in a putative hybrid zone of tetraploid Knautia arvensis and hexaploid K. dipsacifolia (Kol a r et al. 2009).
Due to the combination of diverged genomes, ecological amplitudes of homoploid hybrids tend to differ from those of their parents (Rieseberg 1997). The hybrid niche is not necessarily intermediate but might also transcend parental attributes due to transgressive segregation (Rieseberg et al. 1999). In heteroploid species, newly generated polyploids might be preadapted to occupy novel habitats (Levin 2003), but studies exploring the niches of hybrid cytotypes in sexual plants are scarce. Only St ahlberg and Hedr en (2009) estimated the niche of triploids found at low frequencies in diploid/tetraploid contact zones of Dactylorhiza maculata s.l. as intermediate between those of the parental cytotypes. The ecological niche of pentaploid S. carniolicus differs significantly from that of tetraploids but not from that of hexaploids, albeit being somewhat narrower, resulting in broad overlap with the hexaploid cytotype's niche (Fig. 3C). The nontransgressive state of the pentaploid cytotype's niche might contribute to its restricted spatial distribution in close proximity of the parental cytotypes. Moreover, the ecological resemblance of pentaploids to hexaploids indicates that introgression in S. carniolicus has not (yet) resulted in adaptive evolution, contrasting previous observations of transfer of adaptations through hybridization in various plant species (Arnold 2004;Martin et al. 2006;Kim et al. 2008;Whitney et al. 2010).

Data Accessibility
The AFLP data matrix and the matrix of accompanying species along with mean Landolt indicator values for each Senecio carniolicus individual are available at Dryad: doi: 10.5061/dryad.dh178.