Separation in flowering time contributes to the maintenance of sympatric cryptic plant lineages

Sympatric cryptic lineages are a challenge for the understanding of species coexistence and lineage diversification as well as for management, conservation, and utilization of plant genetic resources. In higher plants studies providing insights into the mechanisms creating and maintaining sympatric cryptic lineages are rare. Here, using microsatellites and chloroplast sequence data, morphometric analyses, and phenological observations, we ask whether sympatrically coexisting lineages in the common wetland plant Juncus effusus are ecologically differentiated and reproductively isolated. Our results show two genetically highly differentiated, homoploid lineages within J. effusus that are morphologically cryptic and have similar preference for soil moisture content. However, flowering time differed significantly between the lineages contributing to reproductive isolation and the maintenance of these lineages. Furthermore, the later flowering lineage suffered less from predispersal seed predation by a Coleophora moth species. Still, we detected viable and reproducing hybrids between both lineages and the earlier flowering lineage and J. conglomeratus, a coexisting close relative. Flowering time differentiation between the lineages can be explained by neutral divergence alone and together with a lack of postzygotic isolation mechanisms; the sympatric coexistence of these lineages is most likely the result of an allopatric origin with secondary contact.


Introduction
The discovery and description of so far unrecognized biological diversity are fundamental to evolutionary biology (Hendry et al. 2010). Molecular analyses identifying lineages that are morphologically and taxonomically indistinguishable and hence are called "cryptic," have become frequent in some taxonomic groups such as animals or fungi, but have been stated to be surprisingly rare in higher plants (Bickford et al. 2007). Obviously, the presence of cryptic lineages has major impact on management, conservation, and utilization of plant genetic resources that need to take into account the specific evolutionary history and current genetic structure (Tellier et al. 2011).
The investigation of cryptic diversity, in particular cases of sympatrically occurring cryptic lineages, is challenging because of two fundamental questions that are tightly interlinked. First, how did these lineages evolve and second, how are they ecologically and functionally maintained?
Although recently cryptic diversity has been revealed by DNA bar coding in several plant species (e.g., Ragupathy et al. 2009;Ni et al. 2012;Yang et al. 2012), studies providing detailed insights into the evolutionary mechanisms and ecological causes or consequences, respectively, still remain scarce. Here, we report on sympatric cryptic lineages in a common wetland plant and ask how reproductive isolation between lineages is maintained by phenological and ecological mechanisms in the face of hybridization and putative introgression. Furthermore, we seek for indications for an allopatric or sympatric evolutionary origin of these lineages.
The diversification of evolutionary lineages and the accumulation of functional and ecological divergence among groups of individuals are facilitated by barriers reducing effective gene flow (Rieseberg and Willis 2007). However, spatial separation is not always necessary to explain the formation of reproductive isolation between groups because even in sympatry both neutral processes and divergent selection can result in population divergence (Devaux and Lande 2008;Flaxman et al. 2014). Eventually, a combination of different isolation mechanisms often leads to an almost complete separation even between closely related groups (e.g., Ramsey et al. 2003), which consequently may be accompanied by a divergence in neutral genetic variation. However, reproductive isolation between lineages is not necessarily accompanied by morphological trait divergence and thus reflected by taxonomy. The incongruence between morphology and taxonomic species boundaries in cryptic lineages has been explained by two general motives (Bickford et al. 2007). First, uniform selection may promote morphological stasis or phenotypic convergence between genetically and reproductively well-isolated lineages. Second, more recently diverged and therefore morphologically similar lineages can be genetically differentiated due to reproductive isolation. Mechanisms of reproductive isolation can be generally categorized depending upon whether they act before or after zygote formation: For example, prezygotic isolation can result from extrinsic barriers such as spatial or temporal separation (e.g., Ellis et al. 2006;Savolainen et al. 2006) or from intrinsic barriers via conspecific gamete precedence (Howard 1999;Fishman et al. 2008). Postzygotic isolation can be mediated via ovule abortion or lower hybrid fitness, which can be expressed, for example, in reduced hybrid seed weight or fertility (see examples in Levin 2012). In cryptic lineages, such isolation mechanisms can affect reproductive interaction directly via acoustical and chemical communication in animals (Henry 1994;Esselstyn et al. 2012). In higher plants, however, reproductive isolation may be driven more indirectly by pollinator preferences (Bower and Brown 2009) or by temporal separation of flowering time (Silvertown et al. 2005). Often, cryptic lineages in higher plants are the result of ploidy heterogeneity (e.g., Sch€ onswetter et al. 2007) which can instantly lead to reproductive isolation via pre-and postzygotic barriers and which has been stated to be a major driver for speciation processes in flowering plants (Otto and Whitton 2000).
In allopatric lineages, prezygotic isolation is obvious (Borsch et al. 2011;Govindarajulu et al. 2011), and with time after divergence, additional pre-and postzygotic barriers are expected to accumulate. In sympatric lineages, the nature of reproductive isolation is often more difficult to elucidate as both pre-and postzygotic mechanisms have to be considered jointly (Whittall et al. 2004). Moreover, the selective potential of biotic interactions in a shared environment might alter the strength of potential isolating mechanisms, as, for example, shared predispersal seed predators are putatively a strong selective agent on inflorescence and flower morphology as well as the timing of reproduction (Elzinga et al. 2007). Thus, the under-standing of sympatric coexistence of cryptic and possibly functionally similar lineages needs to take into account multiple nonexclusive factors such as reproductive and other biotic interactions, but also resource partitioning, microenvironmental heterogeneity, or even neutral processes (Chesson 2000;Hubbell 2005;Leibold and McPeek 2006), which all might contribute to maintain such sympatric relationships.
The common or soft rush Juncus effusus L. is a selfcompatible, wind-pollinated species with an almost worldwide distribution and exhibits a wide ecological tolerance and morphological variability (e.g., Weimarck 1946;Agnew 1961;Kirschner et al. 2002). Because of its frequency in wetland communities and infestation potential in agriculturally used areas (e.g., McCorry and Renou 2003), J. effusus has a significant ecological as well as economic importance. Furthermore, the species has been established as a model plant in phytoremediation (e.g., Gruber et al. 2008). A preliminary molecular analysis revealed two sympatric but genetically strongly differentiated groups (S. G. Michalski, unpubl. data) giving rise to the hypothesis of cryptic lineages within the common rush. However, the evaluation of differentiation patterns in J. effusus is complicated by putative hybridization with its closest relative J. conglomeratus L. Most recent literature acknowledges the frequent occurrence of intermediate forms of putative hybrid origin (K r ısa 1962;Agnew 1968;O'Mahony 2002;Wilcox 2010). Consequently, when analyzing cryptic lineages within J. effusus and their ecology, the closely related J. conglomeratus has to be included. Here, using a combination of molecular, morphometric, karyological, phenological, and ecological analyses, we test the following hypotheses: 1 In Central Europe, Juncus effusus consists of two genetically distinct lineages representing cryptic species. Specifically, we test whether both nuclear and plastid genomes and morphology are differentiated among lineages and their closest relative, J. conglomeratus. 2 Sympatric cryptic lineages within J. effusus are ecologically differentiated and reproductively isolated. In particular, we hypothesize that lineages (1) have different soil moisture preferences and are differently affected by predispersal seed predation and (2) are reproductively isolated by different flowering times. A distinct ecological and functional differentiation in combination with reproductive isolation between the putative cryptic lineages and a relatively recent genetic divergence might favor the hypothesis of a sympatric origin of these coexisting lineages. On the other hand, strong genetic differentiation coupled with incomplete reproductive isolation and low ecological differentiation might indicate an allopatric origin with secondary contact leading to recent sympatric coexistence.

Study species
The most recent taxonomic treatment of the genus Juncus recognizes a number of regional subspecies for J. effusus emphasizing its high morphological variability (Kirschner et al. 2002). In contrast to North America (H€ amet-Ahti 1980;Zika 2003), for Central Europe, only J. effusus ssp. effusus is known. As all analyses were performed in that region, we will refer to this taxon with the shorter "J. effusus" in the following. No taxonomic units below species level are recognized for Juncus conglomeratus (Kirschner et al. 2002). Both species are perennial herbs growing mainly in wet, open to shady habitats. Although the species are able to reproduce vegetatively by short rhizomes, generative reproduction is likely to be more important, accomplished by the production of numerous tiny seeds with high potential for long-distance dispersal (e.g., Ervin and Wetzel 2001). Juncus effusus is known to be self-compatible (Edgar 1964) and like other Juncus species putatively predominantly selfing (Buchenau 1892;Richards and Clapham 1941b). Juncus effusus and J. conglomeratus have been cited to differ in a range of morphological characters, flowering time, and environmental requirements (Richards and Clapham 1941a,b;Tweed and Woodhead 1946;Agnew 1968;Kirschner et al. 2002). Additional morphometric diagnostic characters for the hybrid J. 9 kern-reichgeltii Jansen & Wacht. ex Reichg. have been described (O'Mahony 2002;Wilcox 2010); however, the parental species are extremely plastic in some of the diagnostic features (Agnew 1968;O'Mahony 2002). Hence, the distinction between parental species and hybrid based on morphology alone is often difficult. For European J. effusus and J. conglomeratus, only diploids are known with reported chromosome counts 2n = 40 and 42 for both species (see references in Dr abkov a 2013).

Sites and sampling
Our main study was conducted near Halle (Saale), Germany (51.511°N, 11.927°E) where Juncus effusus and J. conglomeratus co-occur on a former military training area characterized by a high edaphic heterogeneity comprising very dry as well as waterlogged areas. To assess how the genetic constitution at the main study site compares to a larger spatial scale, we additionally collected 85 specimens of both species from 27 locations across Germany (N = 21), Austria (N = 1), Denmark (N = 1), France (N = 3), and Scotland (N = 1) (Table S1).
In May 2011, before start of flowering, a total of 271 Juncus individuals not a priori assigned to either J. effusus or J. conglomeratus were haphazardly selected at the main study site in an area of 100 m 9 600 m. For each individual, a single stem was marked and observed every other day until start of flowering, and the date of first flowering was recorded. In the field, the following morphological traits were measured for these stems: absolute stem height, absolute length of the lower bract (spathe length) and relative to stem height (spathe: stem), absolute length of the upper cataphyll (cataphyll) and relative to the stem height (cataphyll: stem), inflorescence volume computed using length (l), width (w) and depth (d) of the inflorescence and assuming an ellipsoid shape (V = 4/ 3plwd), and flower density as the ratio of the number of flowers in the inflorescence and its volume. These quantitative traits are relatively easy to assess in the field and additionally have been frequently used for diagnostic purposes in this species complex. Furthermore, after fruit ripening, stem and inflorescence were harvested for laboratory analyses. First, the number of stem ridges 2-3 cm below the inflorescence was counted using a binocular, and second, seed length and width were measured by optical scanning with high resolution and applying image analysis implemented in WinSeedle (Regent Instruments Inc., Qu ebec, Canada). Dried material from each stem was used for molecular analyses.
Chromosome counts were obtained from fresh root tips of plants raised in the greenhouse. Pretreatment and maceration followed Schwarzacher et al. (1980). Squash preparations were DAPI stained with Vectashield medium (Axxora, L€ orrach, Gemany) and screened for metaphase chromosome spreads using an Axioskop2 plus (Zeiss, Jena, Germany), which were photographed for later chromosome counting.
In Juncus spp., herbivory by the moth genus Coleophora feeding on seeds has been reported frequently (H ard av Segerstad 1940;Randall 1986;Ellison 1991). Thus, predispersal seed predation may contribute to ecological differentiation between sympatric lineages. Hence, for each inflorescence, the proportion of capsules damaged by herbivore activity was recorded as an estimate for putative selective pressure by predispersal seed predation. Soil moisture content for each individual location was measured to assess potential small-scale habitat differentiation: After 2 weeks without rain, a sample of the upper 15 cm soil was taken as close as possible to each individual and soil moisture content was estimated by comparing fresh and dry weight.
To explicitly quantitate overlap in flowering time between the two J. effusus lineages at the study site, in June 2014, at the start of flowering, a total of 40 individuals were marked and all stems with open flowers were censused two or three times a week until the end of flowering.

Genetic data analysis
For the microsatellite data set, we used principal coordinate analysis (PCoA) of genotypic pairwise distances to visualize the similarity among individuals as implemented in GenAlEx v6.5 (Peakall and Smouse 2012). To test for genetic structuring, we used the Bayesian clustering approach implemented in STRUCTURE v.2.3.3 (Pritchard et al. 2000;Falush et al. 2003Falush et al. , 2007 (see Supporting information for detailed methods). Accounting for the fact that self-fertilization and inbreeding are very likely in the lineages studied, we also used InStruct (Gao et al. 2007) as an alternative to STRUCTURE which includes selfing in the respective model. Genetic diversity within gene diversity H E , and allelic richness corrected for sample size (A R ) and genetic differentiation (F ST ) among lineages (without hybrids, see below) were assessed using FSTAT v. 2.9.3 (Goudet 1995). Chloroplast sequences were aligned in Geneious Pro 5.6.5 (Biomatters 2012) using the MUS-CLE algorithm (Edgar 2004) and default options. A haplotype network was created using statistical parsimony as implemented in TCS 1.21 (Clement et al. 2000). The connection limit between haplotypes was fixed to seven steps.

Hybrid detection
Based on the genotypic data, hybrid status to each sample was assigned using NewHybrids 1.1 beta (Anderson and Thompson 2002). The program uses Bayesian inference to compute posterior probabilities for each sample to belong to genotype frequency classes. Here, posterior probabilities were calculated for six different classes that can arise after two generations of crossing between two parental populations (Anderson and Thompson 2002): Pure individuals of either population, F1 hybrids, second-generation hybrids (F2), and backcrosses between F1 hybrids with either parent (see Supporting information for detailed methods).

Morphometric, ecological, and karyological analyses
For all pure individuals, we tested for differences in morphological traits, date of first flowering, and soil moisture content among groups identified with STRUCTURE by one-way ANOVA. The degree of herbivory was compared using a generalized linear model with quasibinomial errors to account for overdispersion. Multiple comparisons were carried out using post hoc Tukey tests. A principal component analysis (PCA) as well as a linear discriminant analysis on all standardized traits and individuals was performed to assess the morphological differentiation among genetically defined groups. Diagnostic traits were identified by correlating logit-transformed posteriors of the discriminant analysis against individual traits. For all analyses, inflorescence volume and flower density were log-transformed to ensure approximate variance homogeneity and normal distribution of residuals. Morphometric analyses were performed in the R environment v. 2.15.1 (R Core Team 2012) and using the package "MASS" (Venables and Ripley 2002).
Flowering overlap within and between J. effusus lineages, with individuals genetically assigned to either group as described above, was first quantified as among-individual synchrony (mean r), calculated as mean of all Pearson correlation coefficients of the proportion of flowering stems per individual and census across all pairwise individual comparisons (Koenig et al. 2003;Michalski and Durka 2007b). Synchrony was first calculated within lineages, second between groups only considering correlations between pairs from different lineages, and third treating all individuals as one group. Differences in synchrony were assumed to be significant if the 95% confidence interval derived from bootstrapping individuals 1000 times did not overlap. Second, overlap was quantified as lineage-specific probability of hybrid formation by the following approach (cf. Martin and Willis 2007): Assuming random mating, for each census n and lineage i, the probability of mating events with the other lineage j was calculated as t n  A 95% confidence interval for the estimates was derived from bootstrapping individuals within lineages 1000 times.
Chromosome counts were analyzed using linear mixedeffect models with genetically defined groups as fixed and individual as random effect using "lme4" for R (Bates et al. 2013). Significance of the group effect was tested by comparing models with and without groups as fixed effect.

Hybrid fitness
To test whether first-and second-generation hybrids suffer from reduced fitness due to lower seed quality, we compared seed weight, approximated by seed volume, and germination percentage between selected hybrid samples (posterior probability to be either a F1 or F2 hybrid >0.95; N = 7) and arbitrarily selected pure individuals of all parental groups (N = 6-10). For the germination trial, between 15 and 126 seeds (mean # of seeds = 63) per individual were sown in plastic Petri dishes on filter paper soaked with tap water. Dishes were placed in a climatic chamber with a cycle of 18 h of light at 25°C followed by 6 h of darkness at 15°C known to provide optimal germination conditions (S. G. Michalski, unpubl. data). Germination success was assessed four times with a 7-day interval between counts. We tested for significant differences in seed volume between groups by one-way ANO-VA. Germination success was compared using a generalized linear model with quasibinomial errors.

Molecular analyses
In a PCoA of the microsatellite data for 271 individuals from the study site and 85 individuals from other European sites, 75% of the total genetic variation could be explained by the first two axes (Fig. 1A). Samples formed a triangular pattern formed by three larger groups, two of which corresponded to Juncus effusus, which formed two groups (eff1, eff2), separated mainly along the first axis. Genotypes putatively of J. conglomeratus (cong) grouped together well separated from eff1 along the first and from eff2 along the second axis. This pattern was evident for genotypes from the main study site as well as for the additional samples (Fig. S1). Similar to the PCoA, the Bayesian cluster analysis with STRUCTURE clearly indicated the presence of three clusters (K = 3, DK = 8703, mean S = 0.999, Fig. S2), reflecting the PCoA groups. The analysis with InStruct led to the same conclusion (data not shown). More than 80% of samples were unambiguously assigned to one of the clusters (mean individual posterior probability >0.95).

Morphometric and karyological analyses
For the three genetically identified groups (cong, eff1, and eff2), ANOVA found significant differences among groups for all morphological traits measured (F > 5.78, P < 0.01). Pairwise comparisons between pure groups showed that cong consistently differed from both eff1 and eff2 in all measured traits except for seed width. Furthermore, eff1 differed significantly from eff2 in eight of 12 traits assessed (Fig. 2). However, eff1 showed often intermediate values between or similar values to either eff2 or cong. Only absolute cataphyll length was on average lower in eff1 than in the two other groups (P < 0.03). Consequently, in a PCA of all morphological traits with the first two axes explaining 58.9% of the total variation, J. conglomeratus separated relatively well from both eff1 and eff2 which overlapped substantially (Fig. 3). Corroborating these results, linear discriminant analysis was able to assign 99% of pure individuals correctly to J. conglomeratus but only 86% and 59% of individuals to groups eff1 and eff2, respectively. Juncus conglomeratus was best separated from both J. effusus groups by the number of ridges and relative cataphyll length (r > 0.78, P < 0.001), and eff1 and eff2 were best distinguished by spathe length and the number of flowers per inflorescence (r > 0.64, P < 0.001). Chromosome counts were obtained for 12 individuals (Table S1: N = 2, 4, and 6 for cong, eff1 and eff2, respectively) from six to 38 chromosome spreads per individual (mean 20.8). Clean spreads were difficult to obtain; hence, counts varied between 38 and 44, but did not differ significantly among groups (mean 2n AE SE = 40.94 AE 1.19, ANOVA P > 0.94).

Site conditions, phenology, and herbivory
Soil moisture content did not differ among the three groups. Also, no obvious spatial clustering was evident, with individuals from all groups scattered across the main study site (Fig. S4). Groups differed significantly in the mean date of first flowering with J. conglomeratus blooming first, followed 7 days later by eff1 and another 12 days later by eff2 (Fig. 4). Similarly, in 2014, peak flowering of eff2 was 10-14 days later than that of eff1. Flowering lasted more than 20 days for eff1 but only approximately 17-20 days for eff2 with some overlap between the lineages (Fig. S5). Among-individual synchrony within the J. effusus groups was high (mean r = 0.43, 95% CI: 0.31-0.62 and mean r = 0.80 (0.73-0.90) for eff1 and eff2, respectively) and significantly different from overall synchrony (mean r = 0.17 (0.14-0.27)). A significant asynchrony was found in the comparison between both groups (mean r = À0.22 (À0.34 to À0.07). Flowering overlap measured as linage specific probability of hybrid formation under the assumption of random mating was P eff1 = 0.08 (0.06-0.11) and P eff2 = 0.07 (0.05-0.09).
The degree of herbivory on seeds measured as percentage of damaged capsules ranged from 0 to 88%, with the earlier flowering cong and eff1 being similarly (P = 0.495), but significantly more heavily affected than the later flowering eff2 (P < 0.001; 19; 17 and 4%, respectively; Fig. 4

Hybrid seed quality
Hybrids pooled into one group did not differ in seed volume from pure lineages (P = 0.62). This result did not change when hybrids were separated in two groups (J. conglomeratus 9 J. effusus and eff1 9 eff2, P = 0.50). All hybrids produced viable seeds (>30% individual germination percentage). Mean germination percentage after 4 weeks was >70% in all groups. In neither week, germination success differed significantly among groups. This result did not change when hybrids were separated in two groups.

Genetic differentiation
Our results show that Juncus effusus in Europe is composed of two genetically well-differentiated lineages (eff1 and eff2). In higher plants, cytotypes might constitute a significant component of cryptic diversity (e.g., Halverson et al. 2008;Koch et al. 2013;Muenzbergova et al. 2013). For example, in Juncus biglumis L., two genetically separated and occasionally co-occurring linages were found to differ in their ploidy levels (Sch€ onswetter et al. 2007). However, karyotyping individuals representing both lineages of J. effusus and J. conglomeratus did not reveal significant differences suggesting a chromosome number of 2n = 42 common for all groups.
The two linages within J. effusus differed not only at nuclear marker level, but also showed different chloroplast haplotypes. At the nuclear marker level, we found lineages eff1 and eff2 to show similar levels of diversity. At the chloroplast level, however, eff1 harbored four distinct haplotypes contrasting to eff2 with only one Figure 2. Differences in nine measured morphological traits between genetically defined groups: Juncus conglomeratus (cong), J. effusus 1 (eff1), and J. effusus 2 (eff2). Putative hybrids between groups were not considered. Different letters above boxes indicate significant differences between groups. haplotype which is, moreover, shared with J. conglomeratus. Such haplotype sharing between species has been reported from numerous taxa (Petit et al. 1997;Guti errez Larena et al. 2002;Gardner et al. 2004;Jakob and Blattner 2006). Horizontal gene transfer via hybridization as well as incomplete lineage sorting may account for such patterns (Maddison 1997).
Surprisingly, the genetic variation detected in samples across Europe was almost completely represented in the samples from the main study site (Fig. S1). Although wind-pollinated, extensive gene flow by pollen is unlikely to result in the observed lack of geographic genetic structuring because inbreeding coefficients for individual lineages were large (F IS > 0.85) indicating high selfing rates, supporting earlier results Durka 2010, 2012). Selfing does not necessarily result in low withinpopulation genetic diversity (Chauvet et al. 2004;Abbott et al. 2007;Michalski and Durka 2007a;Leger et al. 2009). Instead, high seed production and high seed dispersal potential mediated by wind, water, and animals (Richards and Clapham 1941b), and long-term persistent seed banks (Thompson et al. 1997) may account for the low spatial genetic structuring found.

Morphological differentiation
The three genetic groups were found to differ significantly in a number of morphological traits (Fig. 2). In line with other studies, the number of stem ridges discriminated best between J. effusus and J. conglomeratus (e.g., Stabbetorp 1989). Also relative cataphyll length provided a relatively good diagnostic certainty in the field (O'Mahony 2002). All other traits, despite significant differences in the mean values, showed a substantial overlap among lineages. In particular, eff1 often showed intermediate values between J. conglomeratus and eff2, for example, for inflorescence traits like flower density. Thus, a distinction between the two lineages in J. effusus based on the assessed morphological traits was not possible due to the large overlap of trait values, justifying them as "cryptic" lineages. In J. effusus, it has been shown that flower density, and hence the compactness of the inflorescence, is to some extent genetically controlled in contrast to spathe  length (Agnew 1968 Fernald and Wiegand 1910;Tweed and Woodhead 1946;Fern andez-Carvajal 1982), possibly corresponding to the genetically defined lineage eff1 within J. effusus.

Ecological differentiation
Ecological differentiation due to abiotic conditions, phenology, or biotic interactions may contribute to lineage divergence (Givnish 2010). In wetland ecosystems, soil moisture content describes an important environmental parameter and in our main study population indeed reflected a substantial gradient (range of soil moisture content: 7%-29%). However, although J. conglomeratus has been described as more tolerant to drier soil conditions than J. effusus (Strelkova 1928;Richards and Clapham 1941a), they did not differ in local soil moisture content, neither did the two J. effusus lineages. Still, we cannot rule out resource partitioning for other soil parameters such as soil pH or nutrient availability.
The impact of premating barriers and local individual density distributions on reproductive isolation is well known from a number of studies (e.g., Lepais et al. 2009;Field et al. 2011). Here, we demonstrated a strong differentiation in flowering time among the lineages with J. conglomeratus flowering first, followed by eff1 and much later by eff2 (Fig. 4, Fig. S5).
Assuming a flowering duration of 14-20 days for J. conglomeratus (cf. Michalski and Durka 2007b), our results suggest that the overlap in flowering between eff2 and J. conglomeratus is almost nonexisting, which is consequently reflected in the lowest number of putative hybrids detected between these lineages (N = 2). Flowering time is less probable to provide a reproductive barrier for the temporally intermediate eff1, which indeed contributed most to hybrid formation. Although both J. effusus lineages showed some degree of overlap in flowering time, the probability of hybrid formation was <10% even under the unlikely assumption of random mating. As outcrossing seems the exception rather than the rule in this species group, flowering time differentiation successfully contributes to reproductive isolation between the lineages.
Predispersal seed predation by Coleophora spp. may strongly affect seed production in Juncus (H ard av Segerstad 1940;Randall 1986). Thus, in addition to providing a reproductive barrier, flowering time may be under strong disruptive selection pressure. Although the two early flowering lineages cong and eff1 suffered on average similarly strong from seed predation, both the earliest flowering individuals of cong and the later flowering eff2 were less affected (Fig. S6).

Hybridization and hybrid fitness
Incomplete isolation or the breakdown of reproductive barriers can lead to hybridization which in turn may counteract lineage formation (Garcia et al. 2011). Indeed, we found evidence for hybridization between all three studied genetic lineages, mostly between eff1 and J. conglomeratus.
Assuming that the number of hybrids in our sample is representative for the natural hybridization frequency, rates are relatively high with 7% for the study site. There was no evidence for cytoplasmic introgression into either group. However, eff1 9 eff2 hybrids were biased toward chloroplast haplotypes of eff1 (Fig. 1), suggesting a stronger maternal contribution of eff1 than eff2 to hybrid formation. As flowers are wind pollinated and homogamous and other unidirectional acting barriers are not obvious, such asymmetric pollen flow must be caused by differences in local abundance of parental lineages (Lepais et al. 2009). In fact, at the two sites covered by ≥30 samples, eff1 was more common than eff2 (Table S1).
All investigated putative hybrids produced seeds of similar quality and had similar germination rates compared to the parental lineages. On the other hand, hybrids between the J. effusus lineages showed a higher predispersal seed loss than the later flowering parental lineage suggesting that fitness might be negatively affected by hybridization. The potential risk of extinction of these sympatric lineages may depend on whether other additional components of fitness are significantly decreased in hybrids. Maintenance of sympatric lineages might be fostered by life-history traits such as a long generation time and/or a selfing mating system (Wolf et al. 2001).
Evolutionary origin of cryptic lineages in J. effusus Sympatric coexisting and morphologically similar but genetically differentiated lineages can be explained evolutionary by either lineage differentiation in sympatry or by a secondary contact between lineages formed in allopatry. Our results on genetic and ecological differentiation patterns for the two J. effusus lineages can be interpreted in favor of the secondary-contact hypothesis. First, only little evidence could be found for a clear ecological and functional separation between the lineages except for a pronounced differentiation in flowering time and correlated herbivore pressure. Differences in flowering phenologies among closely related sympatric plant species have been reported frequently (McIntosh 2002;Martin and Willis 2007;Ferriol et al. 2009;Pascarella 2011). Although disruptive selection imposed by seed predation may contribute to flowering time differentiation and reproductive isolation (Elzinga et al. 2007), it not necessarily needs to be invoked to explain sympatric differentiation. Devaux and Lande (2008) modeled the evolution of flowering time using a multigenic model with assortative mating and mutation showing that sympatric formation of allochronic-isolated lineages is favored even without selection under conditions that are partly met in our study species, such as low inbreeding depression (Edgar 1964) or the absence of pollen limitation Durka 2007b, 2010). However, differentiation in flowering time between lineages may similarly arise in allopatry because of stochastic and/or selective divergence. Indeed, the degree of quantitative differentiation for the start of flowering (P ST computed with c = 1 and h 2 = 0.5, sensu Brommer 2011) between the two J. effusus lineages is comparable with that of differentiation at nuclear marker level (P ST = 0.79 vs. F ST = 0.72), suggesting that a simple stochastic mode of lineage divergence alone can explain the observed pattern. Second, the absolute magnitude of this neutral genetic differentiation between the J. effusus lineages, despite a lack of postzygotic reproductive isolation, is unlikely to be the result of a speciation process and a prolonged coexistence in sympatry.
In conclusion, while J. effusus and J. conglomeratus form morphologically distinct groups, they in fact represent a system of three genetically differentiated, homoploid lineages with J. effusus showing evidence for a deep genomic split between two morphologically cryptic groups suggesting a complex evolutionary history within the species. While not completely preventing hybridization among sympatric lineages, differences in flowering time and a selfing mating system are likely to promote reproductive isolation. Juncus effusus is a model plant for research on wetland ecosystem functioning and remediation approaches and studied with respect to pest control or secondary metabolites. Hence, future work on that species will need to include and address the impact of the distinct within-species diversity described here.

Supporting Information
Additional Supporting Information may be found in the online version of this article: Methods S1. Detailed methods for genetic marker analysis and hybrid detection. Table S1. Information on samples and locations included in the genetic analyses. Table S2. Information on newly developed microsatellite markers for Juncus effusus. Table S3. Summary statistics for microsatellite loci. Figure S1. Comparison of diversity patterns at microsatellite and chloroplast sequence level for samples from the main study site and other locations. Figure S2. STRUCTURE summary statistics. Figure S3. Posterior probabilities for different genotypic frequency classes for hybrid individuals. Figure S4. Spatial distribution of samples at the main study site. Figure S5. Proportion of flowering inflorescences against time. Figure S6. Proportion of damaged capsules against date of first flowering.