Introduction

The interest in hybrid zones has increased over the last two decades, as hybridization between genetically divergent individuals of related taxa is considered to be an important factor in the evolution of both plants and animals (Arnold, 1997). Hybridization and introgression are believed to increase genetic diversity within species, transfer genetic adaptations between species, break down or reinforce reproductive barriers between closely related groups and lead to the emergence of new ecotypes or species (Barton and Hewitt, 1989; Abbott, 1992; Rieseberg, 1997). Hybrid zones between cytotypes with different ploidal levels are especially interesting, as they allow studies of the mechanisms involved in the early stages of polyploid speciation and how reproductive isolation mechanisms affect the establishment of polyploids in diploid populations (Thompson and Lumaret, 1992; Petit et al, 1999).

Stable secondary hybrid zones between allopolyploids (ie interspecific polyploids) and their diploid parent species are seldom reported in nature (Petit et al, 1999). Theoretical models investigating the pathways of polyploid speciation are thus rarely discussed with the basis in empirical data from allotetraploid populations. Allopolyploidy is assumed to be more common than autopolyploidy, as the autotetraploids tetrasomic inheritance pattern is considered more unstable than the disomic inheritance pattern of the allopolyploids (Wendel and Weeden, 1989a; Ramsey and Schemske, 1998, 2002). Nevertheless, evolutionary mechanisms leading to speciation are more often investigated among autotetraploids (Petit et al, 1999). Deviations from the fixed heterozygous pattern displayed by divergent alleles at homoeologous loci in allopolyploids could, however, be used as an indication of a ‘triploid bridge’, gene flow and introgression. If one also considers events such as the exchange of chromosomal fragments between homoeologous chromosomes and the dynamics behind the diploidization process in neopolyploids, further studies of allopolyploids ought to give us insight into valuable new information (Sybenga, 1996; Wu et al, 2001).

The genus Dactylorhiza (Neck. ex Nevski) includes several taxa that are endemic to North-western Europe. Most of the species, subspecies and varieties described have a rather plastic morphology and a genome that seems to lack strict reproductive barriers (Hedrén, 1996a; Pedersen, 1998a). Since most taxa display sympatric distribution patterns, both primary and secondary hybrids occur frequently (Hedrén, 1996a; Pridgeon et al, 1997). This conclusion is supported by morphologic and cytological data, which indicates that the evolution in Dactylorhiza tends to be highly reticulate (Hedrén, 1996a; Pridgeon et al, 1997; Hedrén et al, 2001). Thus, infraspecific systematics in Dactylorhiza is controversial, and well-defined species require wide circumscription (Pedersen 1998a).

Repeated hybridization events between the diploid taxa Dactylorhiza incarnata (L.) Soó (2n=40) and D. fuchsii (Druce) Soó (2n=40) are considered to be the origin of the allotetraploid complex D. majalis sensu lato (2n=80) (Hedrén, 1996a, 2003; Devos et al, 2003), which includes D. lapponica (Læst. ex Hartman) Soó. Alternatively, D. maculata (L.) Soó (2n=80), the autotetraploid derivative of D. fuchsii, can by hypothesized as a contributor to the D. majalis sensu lato genome (Hedrén, 1996a).

D. incarnata ssp. cruenta (hereafter called D. cruenta) and D. lapponica are rather common in areas with calcareous or base-rich mineral soils in the upper boreal vegetation zones of Scandinavia. D. cruenta grows mainly in open, flat fens with high groundwater level, whereas D. lapponica occurs mainly in sloping fens. Hybrids between D. incarnata sensu lato and D. lapponica are known from few localities in Norway. At the managed Nature Reserve at Sølendet in central Norway, plants of intermediate morphology are especially frequent in sloping fens influenced by spring water (cf. Moen, 1990). It is hypothesized that clearing as the result of mowing has created new potential habitats, bringing the species into contact and giving hybrids an opportunity to establish. Traditional haymaking at Sølendet ceased around 1950, and natural regrowth processes then prevailed for 25–30 years until a management plan including mowing of the fens started. Monitoring in permanent plots for more than 20 years (Moen, 1990; Øien and Moen, 2002) has shown that individual specimens of the parental species are long-lived (more than 20 years in some cases). Although experimental studies have shown that intensive scything (every year, every second year) reduces the longevity and prolongs the nonflowering periods of orchids (Øien and Moen, 2002), it is assumed that these conditions are favorable for both the parental species and their hybrid.

Here, we investigate whether morphologically intermediary individuals are of hybrid origin and to what extent hybridization and interploidal gene flow occur. In particular, we wish to address whether different ploidal levels are morphologically distinct or if hybridization entails breakdown of morphological boundaries between taxa.

Materials and methods

Sampling

Fieldwork was carried out at Sølendet nature reserve, Røros (62°40′ N, 11°50′ E) in the period July 9–17, 2000 and on July 12, 2001. Plants were sampled from one allopatric population of each species and two populations where the species were found in parapatry. D. lapponica is easily distinguished from D. cruenta, as the latter displays spotting on the abaxial surface of the leaf. D. lapponica is the only representative of the D. majalis complex in the area (Moen, 1990). Plants that were intermediary between the parent species in several other morphological characters were classified as hybrids. Plants were sampled along transects varying in length from 30 to 110 m. Flowering plants within a range of 20 m from the transect were identified, marked with small plastic rods and assigned to one of the groups ‘D. cruenta’, ‘hybrids’ or ‘D. lapponica’. Sampling proceeded until the desired number of plants in each category was reached. From each parapatric subpopulation, 53 assumed hybrids, 29 D. lapponica and 29 D. cruenta were collected. Similarly, 24 plants of D. cruenta and D. lapponica were collected from each of the two allopatric populations. As the diploid D. fuchsii is considered to be the other putative progenitor of the D. lapponica (cf. Hedrén, 1996a, 2003), analyses of 20 plants were carried out to obtain an impression of the genetic variation in this species. For the same reason, five D. maculata individuals were analysed. Both species are abundant in the area.

Photometric quantification of nuclear DNA content

The level of polyploidy was determined from a subset of 68 selected plants by video-based Feulgen-cytophotometric assessment of nuclear DNA contents (ie C-values) using the CIRES (Kontron, Munich, Germany; cf. Vilhar et al, 2001; Greilhuber et al, 2003). A flower bud from each plant was fixed in methanol–acetic acid (3:1), stored in absolute ethanol at −20°C and sent to Johann Greilhuber's lab in Vienna for further analysis. Quantitative staining of nuclear DNA followed an established Feulgen protocol (Greilhuber and Ebert, 1994). Root tip meristems of Pisum sativum cv. Kleine Rheinländerin (2C=8.84 pg) were coprocessed for internal calibration of the staining result. A minimum of 10 prophases and 10 telophases or interphase nuclei in G1 and G2 from each plant was measured. Additional chromosome counts were carried out from Feulgen slides in two or three individuals per taxon and the hybrid D. lapponica × D. cruenta to ascertain the assignment of DNA contents to ploidal levels.

Enzyme assays

The apical part of the two uppermost leaves of each plant were detached (Hedrén, 1996a) and kept in a container with blue ice during fieldwork. For long-term storage, the material were kept in an ultracold freezer at −80°C. Approximately 1 cm2 of each analysed leaf was ground with a small amount of sand in 60 μl of a Tris-HCl grinding buffer (Soltis et al, 1983) on a cold porcelain plate. The extracts were absorbed onto paper wicks and stored in microtitre dishes at −80°C until analysis.

Isoelectric focusing on polyacrylamide gels with pH 4–9 (cf. Westermeier, 1993) was used to analyse phosphoglucoisomerase (PGI, E.C. 5.3.1.9), phosphoglucomutase (PGM, E.C. 5.4.2.2), menadione reductase (MNR, E.C. 1.6.99.–) and phosphogluconate dehydrogenase (6-PGD, E.C. 1.1.1.44). Staining recipes followed Wendel and Weeden (1989b) and Cronberg (1995), with some minor modifications. PGI and PGM could be stained simultaneously, as visible loci had distinctly separate pI. Isoelectric focusing was chosen as an analysing method since it provides sharp banding patterns that simplify visual dosage interpretation.

Interpretation of banding pattern and the distribution of alleles in polyploids

Ploidal level should in theory be identifiable by relative banding pattern intensity. Strict allotetraploids show fixed heterozygosity at loci where the alleles from the parents are fixed for different alleles. They will thus produce three bands with a balanced 1:2:1 intensity ratio when stained for dimeric enzymes. Triploids, which inherit two chromosome sets from the tetraploid parent and one chromosome set from the diploid parent, will have an unbalanced banding pattern with an expected 1/9:4/9:4/9 intensity ratio. The dimeric enzymes PGI and MNR displayed sufficient resolution and enzyme activity for interpretation of relative banding intensity at each allele. Alleles representing chromosome sets found in D. cruenta are referred to as ‘C’ and the homoeologous alleles found in D. fuchsii referred to as ‘F’. A strict allotetraploid D. lapponica is hence expected to only display a ‘CCFF’ allele distribution. Should, however, the inheritance pattern in D. lapponica be tetrasomic (ie an autopolyploid), the different combinations of the four chromosomes will have an expected distribution of genotypes equal to (r+q)4=r4+4r3q+6r2q2+4rq3+q4=1, where r and q are the frequency of C and F alleles in the population, respectively. A single-classification test for goodness of fit was performed to assess independent segregation in putatively homoeologous chromosomes (G-test; Sokal and Rohlf, 1995).

Morphology

From each plant, a fully developed flower from the lowest part of the spike, the belonging bract and the lowermost leaf were collected for morphological analysis. The material was kept in a freezer at −20°C to conserve colour intensity and shape until further analysis. All plants were measured for 18 morphological characters (Table 1), mainly taken from the literature (Dufrêne et al, 1991; Pedersen, 1998a).

Table 1 The quantitative and qualitative measured (M1–M18) and derived (M19–M28) morphological characters
Full size table

A discriminant analysis was performed, using the measured characters as well as 10 derived characters (cf. Table 1), to address what linear combination of morphological measurements is best able to classify D. lapponica, D. cruenta and the hybrids. Crossvalidation was used for estimating the rate of error conditioned on the given data set (Venables and Ripley, 2003). In the crossvalidation procedure, each observation is systematically left out, the discriminant function re-estimated and the excluded observation classified based on the re-estimated function.

Equality of means among ploidal levels was tested with Wilks’ lambda and Pillai's trace. Pairwise equality of means was tested by Hotelling's T2. Mahalanobis distance measure was used to quantify the relative morphological divergence between ploidal levels and populations. Manual forward selection was used to identify the characters that contribute significantly to the model. Each character was analysed separately for conditional effects, and its contributed significance was calculated by Monte Carlo tests (ter Braak and Šmilauer, 2002; Lepš and Šmilauer, 2003). Statistical computations were performed with the program package S-PLUS 2000 (Mathsoft, 1999). The discriminant plot and forward selection procedure were performed using the Canonical Variates option in the program package CANOCO (version 4.5; ter Braak and Šmilauer, 2002).

Results

Assessment of ploidal level from DNA C-values

Nuclear DNA content quantification in combination with chromosome counting was found to be a suitable method for ploidal level determination and recognition of triploids and aneuploids (Table 2). Variation between individuals within taxa was low (coefficient of variation 2–3.5%). All specimens of D. cruenta and D. fuchsii had C-values corresponding to the expected diploid level, whereas D. lapponica and D. maculata samples had C-values corresponding to the tetraploid level (Table 2). One exception was a hypertetraploid D. lapponica (2n∼5.4x; Table 2). The 25 plants initially classified as hybrids from morphology turned out to be triploid (11), aneuploid (2) or tetraploid (12) based on DNA content (Table 2). Chromosome counts of the aneuploids also confirmed that they had 2n∼70.

Table 2 Genome size data in the investigated Dactylorhiza taxa
Full size table

Allozyme variation

D. cruenta was monomorphic and differed from D. fuchsii at all the allozyme loci. The only exception was a D. cruenta allele (6-PGD ‘10’) found in one of the D. fuchsii genotypes. D. fuchsii was monomorphic for PGI, whereas two alleles were present in 6-PGD, MNR and PGM (Table 3, Figure 1). D. maculata contained the same alleles as D. fuchsii. The alleles found in D. lapponica matched the alleles in D. cruenta and D. fuchsii, and the two genomes were readily separated in the tetraploid (Table 4, Figure 1). However, one allele (6-PGD ‘8’), commonly found in D. lapponica, was lacking in the diploid species.

Table 3 Electrophoretic phenotypes of D. lapponica, D. cruenta × lapponica and phenotypes that are mixtures of triploids and tetraploids are indicated
Full size table
Figure 1
figure 1

A graphical illustration of the band intensity pattern in the different species and ploidal levels in the dimeric enzymes PGI and MNR. The monomorphic allele originating from D. cruenta is denoted 10′ at both loci, whereas the faster migrating alleles are denoted 11′ and 12′. Plants heterozygous for all three alleles in MNR displays five bands since the heterodimer between 10′ and 12′ seems to overlap with the band expressed by 11′.

Full size image
Table 4 Number of plants of tetraploid D. lapponica and triploid D. cruenta × D. lapponica displaying different dosages of MNR and PGI alleles
Full size table

All triploids had excess of D. cruenta alleles, as expected from crosses between D. lapponica and D. cruenta. The majority of the tetraploids were balanced heterozygotes, but a large fraction displayed excess of alleles originating from either D. fuchsii or D. cruenta in PGI (Table 4). However, the hypothesis that the distribution of genotypes was equal to the distribution expected under tetrasomic inheritance was rejected (Table 5). Similar patterns were also found in the plants that had not been analysed for DNA content (not shown). These plants were classified as tetraploid if they were balanced heterozygous at one or both of the loci, MNR and PGI. Triploids were identified as plants that displayed excess of alleles originating from D. cruenta at both loci.

Table 5 Observed frequencies of PGI and MNR genotypes of the tetraploid D. lapponica (CCCC, CCCF, CCFF, CFFF, FFFF; N=27)
Full size table

Morphology

Triploids were morphologically intermediary between D. lapponica and D. cruenta (Figure 2), and the three groups had discriminant scores with significantly different means; Wilks’ lambda (F56,30=6.32, P<0.001) and Pillai's trace (F56,32=4.06, P<0.01). Mean discriminant scores were also different for all pairs of groups (Hotelling's T2 2x−3x: F28,15=5.42, P<0.001, 2x−4x: F28,15=18.64, P<0.001, 3x−4x: F28,15=4.65, P<0.001). A number of characters contributed significantly (P<0.05) to the variance between the different ploidal levels in the forward selection analysis. These were: area of lowermost leaf covered with spots; pattern on labellum; contrast of colour in pattern on labellum; length from spore entrance to the tip of the middle lobe; length from spore entrance to the tip of the lateral lobe; and shape of teeth and the number of teeth/cells pr. 1.5 cm on leaf margin and midbract (Figure 2).

Figure 2
figure 2

Triplot presenting the discriminant functions based on 28 morphological characters in 68 plants with ploidal level confirmed by DNA photometry. Diploid, triploid and tetraploid plants are denoted 2x, 3x and 4x, respectively. All ploidal levels differed significantly (P<0.01). Vectors represents the morphological variables that contributed significantly to the model (P<0.05) by forward selection. The denotations of the different characters follow Table 1. The hyperploid plant and the aneuploids are not included in the plot. However, the predict function in S-PLUS classifies them as tetraploids.

Full size image

The crossvalidation estimates indicated that misclassifications occur more often between triploid and tetraploid than between diploid and triploid plants (Figure 2, Table 6). Similarly, Mahalanobis distance showed that diploids and tetraploids are morphologically most divergent, whereas triploid individuals were more similar to tetraploids than diploids (Table 7). D. lapponica and D. cruenta did not differ morphologically more in allopatry than in parapatry (Table 7).

Table 6 Crossvalidation performed on (A) plants with confirmed DNA content and (B) all analysed plants
Full size table
Table 7 Mahalanobis distance between the different and populations ploidal levels (2x, 3x and 4x; distance measures presented with bold numbers), all individuals included
Full size table

Discussion

Allelic segregation in tetraploid D. lapponica deviated from the balanced fixed heterozygous banding patterns expected under strict allopolyploidy. Similar unbalanced patterns were found in the allotetraploid D. purpurella displaying excess alleles of D. fuchsii (Hedrén, 1996a, 1996b). One explanation is that the tetraploid originated from less divergent ancestral diploid genotypes than those observed today. However, in our study, there is little reason to believe that the unbalanced banding patterns in D. lapponica originated from polyploidization events in ancestor populations where D. fuchsii and D. cruenta had alleles in common. Rather, the close correspondence between the diploid and tetraploid genomes may indicate a relatively recent origin of the D. lapponica populations from diploid populations genetically similar to today's populations.

The staining intensities of the triploids were consistent with the expression of three alleles, except for one plant lacking the F allele. This indicates that unbalanced tetraploid genotypes are not due to silenced alleles, as such null alleles are rare or absent in the hybrids. This conclusion is also consistent with observations from similar studies of D. purpurella (Hedrén, 1996b). Unbalanced banding patterns may, instead, be due to chromosomal rearrangements caused by rare pairing events between homoeologous chromosomes originating from the different parental genomes. Strict allopolyploidy and autopolyploidy are considered to be two extremes, whereas most polyploids represent intermediate stages, displaying both auto- and allopolyploid pairing behaviour (Jackson and Jackson, 1996; Ramsey and Schemske, 2002). Our results suggest that D. lapponica has a relatively high frequency of genomic rearrangements at the PGI and MNR loci. Although the distribution of genotypes in the populations still indicates a high degree of preferential pairing of homologous chromosomes, genomic reshuffling seems to be an additional source of genetic diversity in D. lapponica. The low frequency of CC and FF gametes (inferred from genotype frequencies) could indicate that CCCC and FFFF genotypes are lacking in the populations because of sampling effects. Genotype frequencies imply an even lower frequency of gametes with CC and FF composition in MNR (compared to PGI). This could be explained if MNR is located closer to the centromere than PGI, a region less exposed to rearrangements. On the other hand, PGI seems to be located in an area that is more exposed to chromosome fragmentation (Hedrén, 1996a). Controlled crosses are necessary in order to fully corroborate these assumptions (cf. Wendel and Weeden 1989a; Hedrén, 1996b; Wu et al, 2001; Ramsey and Schemske, 2002).

Another potential explanation for unbalanced tetraploids is that production of viable gametes by hybrids may lead to introgression of D. cruenta alleles into tetraploid plants. Some backcrossing between the triploid and the tetraploid is indicated by the finds of two aneuploids with chromosome number 2n∼70. Such patterns were also found in a hybrid swarm between D. fuchsii and D. purpurella on the British Isles (Lord and Richards, 1977). If triploids produce gametes with n∼2x or n∼x, both tetraploid and triploid individuals may have originated from backcrosses of hybrids to D. lapponica. Triploids are known to produce fertile hyperdiploid (2x+1) and hypotetraploid (4x−1) offspring occasionally, and the mean fertility in 18 different ‘allotriploid’ hybrids was as high as 23.7% (Ramsey and Schemske, 1998). Triploid hybrids seem, however, to produce less viable gametes than polyploids with odd numbers of chromosome sets of a higher ploidal level (eg 5x, 7x; Brochmann et al, 1992). D. fuchsii could very likely be an ancient polyploid, as much lower chromosome numbers are found in the related genus Orchis (Lord and Richards, 1977). In addition, partly compatible homoeologous chromosomes may compensate for the loss of the homologous chromosome set during meiosis, especially if the unpaired chromosome set is so unevenly distributed that the gametes closely resemble diploid and haploid gametes.

No backcrosses between the hybrid and D. cruenta were discovered. Pedersen (1998b) likewise found no clear indication of introgression from D. majalis sensu lato or D. maculata sensu stricto into D. incarnata sensu lato in sympatric populations. In contrast, Lord and Richards (1977) concluded that backcrossing between triploids and diploids where more common than between triploids and tetraploids in the hybrid zone between D. fuchsii and D. purpurella. Generally, backcrossing is thought to occur more often between a triploid hybrid and the diploid ancestor, rather than the tetraploid ancestor (Ramsey and Schemske, 1998). Why our study deviated from this pattern is difficult to explain.

Analyses of morphology also show that D. lapponica and D. cruenta are well differentiated in a number of characters. The triploid hybrids are morphologically intermediate, but resemble D. lapponica more closely than D. cruenta. Together with the occasional findings of aneuploid backcrosses between the eutriploid hybrids and the tetraploids, this may indicate that interploidal gene flow occurs, which may in turn weaken morphological boundaries between the primary hybrids and D. lapponica. Allopatric populations, on the other hand, are no more divergent morphologically than the populations of the parent species that grow parapatrically. Thus, the allopatric populations may not be isolated enough to prevent gene flow between and within taxa.

A more plausible explanation is that morphology is unaffected by rare instances of backcrossing between triploids and tetraploids. Instead, the hybrid nature of both the triploids and the allotetraploids entails that loci of both D. fuchsii and D. cruenta are expressed in these plants, whereas only a single genome is expressed in the diploid. This qualitative difference in genes that are expressed may result in a clearer morphological difference compared to the dosage difference of D. fuchsii genes between triploids and tetraploids (Lord and Richards, 1977; Guo et al, 1996; Wang et al, 1999).

Concluding remarks

Despite evidence of backcrossing, the pattern of genotypic variation does not indicate widespread introgressive hybridization of D. cruenta into D. lapponica. Rather alleles originating from D. fuchsii seem to be over-represented in D. lapponica. Although rarer, putative hybrids between D. lapponica and D. fuchsii or D. maculata have been detected in the area, based on morphology. As the autotetraploid D. maculata produces diploid gametes, potentially viable hybrids with D. lapponica could produce tetraploid offspring directly. Even if this hybrid is rarer, introgression may be more effective via this pathway as it does not depend on triploid viability. The significance of this process for explaining the predominance of F alleles in D. lapponica remains to be studied. Lack of genetic variation in D. cruenta renders it difficult to distinguish between introgression and genomic rearrangements within tetraploids, such as those resulting from pairing between homoeologous chromosomes. More variable markers, such as single- or low-copy nuclear DNA, are required to gain a better understanding of the dynamics of genomic exchange between ploidal levels (cf. Hoot et al, 2004).

Although genomic recombination and hybridization between and within ploidal levels has made identification of a strictly defined D. lapponica difficult, this study gives little indication that widespread hybridization entails breakdown of species boundaries. Rather, the hybrid zone may be a local transient phenomenon due to extensive mowing opening habitats, and thus bringing the parental species into close contact. This conclusion would be in agreement with the general findings of Pedersen (1998b) concerning Dactylorhiza taxa with different genome compositions. However, we have only begun to understand the genetic consequences of such extensive hybridization processes in these ecosystems, and whether genetic plasticity in polyploids is an evolutionary adaptation strong enough to eliminate the traditional species boundaries.