Hybrids and fruit set in a mixed flowering-time population of Gymnadenia conopsea (Orchidaceae)

Transcription

1 Hereditas 143: (2006) Hybrids and fruit set in a mixed flowering-time population of Gymnadenia conopsea (Orchidaceae) MIKAEL LÖNN 1, RONNY ALEXANDERSSON 2 and SUSANNE GUSTAFSSON 3 1 Södertörn University College, School of Life Sciences, Huddinge, Sweden 2 Biology Education Centre, Uppsala University, Uppsala, Sweden 3 Dept of Evolutionary Functional Genomics, EBC, Uppsala University, Uppsala, Sweden Lönn, M., Alexandersson, R. and Gustafsson, S Hybrids and fruit set in a mixed flowering-time population of Gymnadenia conopsea (Orchidaceae). * Hereditas 143: Lund, Sweden. eissn Received April 25, Accepted October 3, 2006 We have recently found that the morphologically determined subspecies Gymnadenia conopsea ssp conopsea in Sweden includes early and late flowering individuals. We were interested in the interactions between the flowering time groups; if there were gene flow between them and if so this was detrimental or advantageous. A spatially mixed population of early and late flowering individuals was studied using three microsatellite loci. We measured patterns in genetic differentiation and inferred occurrence of hybridisation and introgression. Variation in flowering time, fertility and relative and absolute fruit set was measured. The pattern of introgression between flowering-time groups differed between loci. In two of the three investigated loci, allele separation was distinct between early and late flowering plants and one genetically obvious hybrid was infertile. In the third locus, several alleles were shared between the two flowering time variants. The degree of introgression was associated to fruit set failure, which was higher in the late flowering plants and lower in early flowering plants. A small group of early flowering individuals with somewhat delayed flowering compared to the main group was genetically distinct and had lower relative and absolute fruit set. This group was not genetically intermediate, but rather constituting an independent group, with lower fruit set possibly caused by absence of pollinators. There seem to be a strong barrier against introgression into the late flowering group which is kept genetically distinct and less diverse. The early flowering group is diverse, includes two subgroups and seems to benefit from gene flow. Mikael Lönn, Södertörn University College, School of Life Sciences, SE Huddinge, Sweden. mikael.lonn@sh.se Hybridization between taxa is evolutionary important in several respects. Hybridisation may spread genes between taxa, in this way supplying new variation to populations. This variation can be detrimental or favourable in the current situation, but will in any case supply raw material for further evolution. Advantageous alleles can spread readily despite low gene flow levels (MORJAN and RIESEBERG 2004). Low fitness in hybrids, on the other hand, is also seen as a major mechanism keeping taxa separated through reinforcement (SERVEDIO and NOOR 2003). In situ hybridisation has been detected between species with over-lapping flowering time, e.g. in Eucalyptus species (BARBOUR et al. 2003), Silene species (RUNYEON-LAGER and PRENTICE 2000). It has also been shown that differentiation in flowering time can be built up at a 100-year timescale, adjacent Anthoxanthum odoratum populations in the Rothamstead Park Grass Experiment have diverged in flowering time and the division was reinforced by natural selection (SNAYDON and DAVIES 1976; SILVERTOWN et al. 2005). The taxonomy within Gymnadenia conopsea is intriguing. Traditionally, two varieties with distinct morphological separation have been identified, one early flowering, var. conopsea, and one late flowering, var. densiflora. In the study by GUSTAFSSON and LÖNN (2003) individuals morphologically similar to var conopsea, but late flowering as var densiflora were observed. Flowering time was found to clearly divide the species into two groups, using ITS and microsatellite markers. The early flowering G. conopsea was close to Gymnadenia odoratissima, with identical ITS, but those two species were distinctly separated from late-flowering G. conopsea. The lateflowering G. conopsea show low variability in microsatellite loci and have smaller habitat amplitude with preference for wetter habitats than the early-flowering type. GUSTAFSSON and LÖNN (2003) found microsatellite alleles to be diagnostic for the earlyflowering group of plants in 17 localities in southern Sweden, indicating a general pattern of isolation between the early and late flowering groups. This study revealed two certain hybrids and five individuals with partly hybrid alleles indicating introgression, at a low rate. No fitness measurements were made. There is obviously strong gene flow barriers between the flowering time groups, but GUSTAFSSON and LÖNN (2003) could not evaluate the fitness aspects of hybridisation and introgression.

2 Hereditas 143 (2006) Hybrids and fruit set in Gymnadenia 223 In this study we investigated a population with spatially mixed flowering-time types; we collected data on fruit set and characterised individual multi-locus genotype using microsatellite markers. The aims were to study the direction and effect of introgression between the flowering-time types divided into four flowering time groups (early, medium, late and, and one group of infertile plants). We show that introgression and hybridization influence flowering time and fruit set. MATERIAL AND METHODS Gymnadenia conopsea ssp conopsea, the fragrant orchid, is a perennial, terrestrial orchid. It is associated with calcareous areas and occurs in grasslands like grazed meadows and close to marshes and fens. The flowers are heavily scented and pollinators, different species of Lepidoptera, are rewarded with an abundant amount of nectar (PROCTOR et al. 1996). The geographic distribution covers most of Europe and parts of Asia. Sampling was made close to the lake Horsan, in the north-western part of the Baltic island of Gotland, Sweden. The site is mesic and inhabits a spatially mixed population of early- and late flowering Gymnadenia conopsea ssp conopsea (GUSTAFSSON and LÖNN 2003). The two flowering forms are about equally common and spatially intermingled. The population was visited at five occasions and these occasions were used to delimit the flowering time groups. The idea from the beginning was to identify only early- and late flowering individuals, but the variation in flowering time observed motivated a further subdivision. At the first occasion, 18 June 2003, all individuals (93) within an 30/30 m area were marked and a 2/2 mm leaf part was sampled in silica gel. At the second occasion, the 22 June, individuals in flower were recorded and considered early flowering. Individuals that were flowering at 26 June were considered EAR- LYmedium. At 16 July all individuals were flowering and those that had started to flower later than the 26 June were considered flowering. Two individuals started to flower after the 26 June but were in fruit the 16 July. They were considered late. Four individuals did not flower, the flower buds wined before flowering, they were considered. In August the numbers of filled and unfilled capsules were counted. Microsatellite loci Three microsatellite loci were used, Gc29, Gc31 and Gc51. The methods are given in GUSTAFSSON and LÖNN (2003). The repeated sequences in all three loci consisted of dinucleotide repeats. Statistical analyses Linear discriminant analysis was used to evaluate the genetic differences between flowering time groups. A diffusion matrix (VENABLES and RIPLEY 2002) was used to reassigned individuals to flowering-time classes and the proportion of correct reassignment was used as a measure of the level of discrimination between the classes. We used the procedure lda in the library MASS (VENABLES and RIPLEY 2002) within the program package R (R DEVELOPMENT CORE TEAM 2004). One analysis was made using all three loci, and one analysis was made for each separate locus to display differences in discrimination between the loci. To measure fitness aspects of flowering-time and genotype the number of filled capsules out of all capsules were used in a generalized linear model with a binomial error, and total number of filled capsules assuming a normal error distribution, using the program package R (R Development Core Team 2004). Explanatory variables were flowering time class and microsatellite allele frequencies. To measure introgression the diffusion matrix from the linear discriminant analysis on locus 29 was used, because alleles in this locus are generally most evenly spread between early and late flowering G. conopsea (GUSTAFSSON and LÖNN 2003). Individuals that did not classify with their original flowering time category were considered to be more introgressed. The alleles at the two other loci were almost totally separated between groups, a general pattern in Swedish populations (GUSTAFSSON and LÖNN 2003), and were considered to contain little information on introgression, but those two loci are then useful in detecting recent hybridization. The effect of introgression and flowering time category was tested in generalized linear models using filled capsules out of all capsules, filled capsules and unfilled capsules as response variables. Minimal adequate models (CRAWLEY 2002) were created by omitting all explanatory variables (if not involved in a significant interaction) and merging of factor levels that did not cause a significant change in deviance when omitted from the model or merged with other factor levels. The elimination of variables and factor levels was made stepwise, starting with the less influent explanatory variables and merging the most similar factor levels. Genetic differentiation between flowering time groups was calculated as the Euclidean distance based on mean allele frequencies.

3 224 M. Lönn et al. Hereditas 143 (2006) We used the program TFPGA (MILLER 1997) to calculate WRIGHT S (1978) modification of Roger s genetic distance and F ST values according to WEIR and COCKERHAM (1984). RESULTS Genetic differentiation Allele frequencies are given in Table 1. The mean number of filled and unfilled capsules for the flowering-time groups are given in Table 2. The numbers of flowers is the sum of filled and unfilled capsules. Twenty-eight individuals were eaten by animals before the fruit set could be recorded. The measure of F ST between the four flowering time groups was 0.14 for locus 29, 0.45 for locus 31 and 0.49 for locus 51 (average 0.34). Rogers genetic distances (WRIGHT S 1978 modification) between the three larger groups were: early vs / 0.66, early vs medium /0.16 and medium vs /0.65. The results from the linear discriminant analyses, based on three microsatellite loci, are given in Fig. 1a. The analysis using all three loci clearly distinguish the early, medium and groups, while three of the individuals group with. The fourth seem to be intermediate between and early. The two late individuals group with the early individuals. The diffusion matrix correctly classify all individuals to their original flowering-time class, except for three which are classified as. The single-locus linear discriminant analysis, for locus 29 (Fig. 1b), reveals several groups containing mixed flowering-time categories and one group of only early individuals. In the diffusion matrix 25 out of 35 early individuals are correctly classified, the proportion for medium is 4 out of 9, late 1 out of 2, 19 out of 24 and infertile 0 out of 4. The linear discriminant analysis for locus 31 identify four clusters (Fig. 1c), the cluster to the right contains genotypes 0707 which is the group, the upper left is genotype 0404 and the lower left There is one medium individual with the genotype The diffusion matrix correctly classify all EAR- LYearly and individuals, other flowering time categories are classified as early, except for the 0407 individual which is misclassified as. The linear discriminant analysis for locus 51 (Fig. 1d) reveals a pattern similar to the linear discriminant analysis using all loci, but one individual each from medium, late and INFER- TILE are placed in the early-flowering class. In locus 51 the alleles 18 and 19 (Table 1) are common in the group, but in the early-flowering groups they are absent. In the same allele-size interval in the earlyflowering groups we find the alleles 16.5, 17.5, 19.5 and The genetic differentiation between groups calculated as Euclidean distance between group centroids in the linear discriminant analysis, was 2.55 between the early group and the group is between medium and. The genetic distance between the medium and EAR- LYearly group is 0.78, so the medium group is closest to the early group but there is no indication of medium being intermediate between early and. Fruit set The mean number of filled and unfilled capsules in the flowering-time groups is given in Table 2. When the number of filled capsules are analysed as a proportion of the total number of capsules in a generalized linear with quasibinomial error distribution (quasibinomial to account for overdispersion) (Table 3), the EAR- LYmedium flowering-time category has a lower relative set than the other categories. There is also a positive effect on fruit set of having a genotype containing allele 24 in locus 51 and a negative effect of having allele 7 in locus 29. The absolute number of unfilled capsules was explained by an interaction between flowering time and degree of introgression using a generalized linear Table 1. Microsatellite allele frequencies in flowering-time classes of Gymandenia conopsea from an allopatric population on the island of Gotland, Sweden. Only alleles with a total frequency of more than 10 are shown. Flowering Time N Locus 29 Allele Locus 31 Allele Locus 51 Allele early medium late

4 Hereditas 143 (2006) Hybrids and fruit set in Gymnadenia 225 Table 2. Summary statistics for the Gymnadenia conopsea individuals analysed for filled and unfilled capsules. Flowering Time N Filled capsules Unfilled capsules Mean (SE) Mean (SE) early (1.8) 7.9 (1.4) medium (3.8) 10.0 (3.5) late (10.0) 3.0 (2.1) (3.7) 6.8 (1.6) model with a Gamma error distribution (Table 4, the initially assumed Gaussian error distribution was changed to Gaussian after inspection of diagnostic plot). The combinations without introgression and early/medium/ late with introgression showed similar and lower numbers of unfilled capsules. Unfilled capsules were more common in the other two combinations. DISCUSSION Differences between loci, hybridisation and introgression There is a main division based on private microsatellite alleles (Table 1) and allele frequencies (Fig. 1) between the flowering group and the cluster of early, medium and late, the (a) All three loci (b) Locus (c) Locus 31 (d) Locus 51 S Fig. 1. Plots of the results of linear discriminant analyses of flowering time categories of Gymnadenia conopsea, one analysis using all three microsatellite loci and three separate analyses for each locus. In the plots of separate loci a small random value is added to make points separated, individuals with the same genotype are else overlayed.

5 226 M. Lönn et al. Hereditas 143 (2006) Table 3. A minimal adequate generalised linear model using seed set measured as filled capsules out of all capsules in flowering time categories and microsatellite allele counts in Gymnadenia conopsea as the response variable, assuming a binomial error distribution. The explanatory variables are flowering time and microsatellite allele counts. The levels of flowering time has been have been merged so that levels with similar response values are merged if this is not causing a significant difference between models with and without the merging. In this model the levels early, and late are merged, while the level medium is kept. The direction of the effect of the alleles are given in parenthesis). Values given in the predictions are calculated when the values for the two alleles are kept at their mean values. Explanatory variable DF Deviance F P Flowering time Allele (/) Allele 29.7 ( /) Residual Predicted values of seed set from the model: Flowering time category Seed set SE early//late medium latter belonging to the early flowering type in GUSTAFSSON and LÖNN (2003). The somewhat later flowering individuals within the early flowering group (medium) is genetically distinct, but closer to the early group than to the group (Fig. 1a). Three of the four plants are classified together with the plants, while one is placed by the linear discriminant analyses between early and and is likely to be of hybrid origin. In locus 51 alleles 18 and 19 are common in the group, while the early-flowering plants only have the alleles 16.5, 17.5, 19.5 and 20.5 in this length range. This pattern points to an occasion when the two main groups of early and late flowering were separated from each other, where a half-repeat have been added or removed, or that base-pairs have been added or removed in the base-pair sequence between the primer site and the repetitive sequence. The same strong pattern is seen in locus 31, where allele 7 occurs in late flowering plants and alleles 1 and 4 in early flowering plants. Several alleles in locus 29 are shared. The F ST value for locus 51 is 0.49, so half of the genetic diversity is contained between groups. The F ST value for locus 31 is 0.45, close to the value for locus Table 4. An analysis of deviance using unfilled capsules in flowering time categories of Gymnadenia conopsea as the response variable, assuming a gamma error distribution. The explanatory variables are flowering time and introgression individuals that do not cluster with their original flowering time category in the diffusion matrix from the linear discriminant analysis using locus 29 are considered introgressed. The levels of flowering time has been have been merged so that levels with similar response values are merged if this is not causing a significant difference between models with and without the merging. In this model the levels early, medium and late are merged, while level is kept. Explanatory variable DF Deviance F P Flowering time Introgression Flowering time: introgression Residual Predicted values of unfilled capsules from the model, standard errors are given in parentheses: Flowering time category Introgression Yes No early/late/medium 5.2 (1.6) 9.5 (1.5) 10.2 (3.6) 5.6 (1.7)

6 Hereditas 143 (2006) Hybrids and fruit set in Gymnadenia Other alleles than the separating ones in locus 51 as well as most alleles in locus 29 are shared. The F ST value for locus 29 is the lowest, The separating alleles in loci 51 and 31 have not been spread between the late and early flowering plants, despite the fact that gene flow between the groups seem to occur. We found one obvious hybrid using the information from locus 31. Alleles in locus 29 occur in both groups, indicating higher gene flow between groups in this locus. Introgression must not occur at the same rate in all loci, different introgression rates between chromosome blocks are shown in a Helianthus natural hybrid zone by LEXER et al. (2004). The different patterns indicate that there might be incompatibility connected to the microsatellite alleles or to genes they act as markers for. A possible explanation is that the alleles in loci 31 and 51 are markers linked to genes which make hybrid individuals intermediate in some important character that drastically lower their fitness. Two Silene species hybridizing in a contact zone were kept morphologically distinct despite genetic introgression, the suggested mechanism being low fitness in either of the parent species habitats (RUNYEON-LAGER and PRENTICE 2000). Fruit set The group of somewhat later flowering individuals within the early flowering group (medium) has significantly lower relative fruit set. This could be due to low pollination rate or to genetic incompatibility due to hybridization with the flowering group to which they are closer in flowering time than the early group. The explanation involving low pollination is supported by the fact that the EAR- LYlate flowering individuals do not have lower fruit set; they might be pollinated by the same pollinators as the flowering group. Although, the EAR- LYlate group consists of only two individuals so no firm conclusions can be drawn. The second explanation involving hybrid origin is supported by the fact that one of the medium individuals obviously is a hybrid in locus 31 where the alleles 4 and 7 else are strictly bound to the main earlylate division. The intermediate flowering time also suggests a hybrid origin as a possible explanation, e.g. intermediate morphology was found in Asclepas hybrids by KLIPS and CULLEY (2004). However, when the total genetic composition is taken into account, the medium group is not genetically intermediate between the early and group (Fig. 1) and the genetic distance between medium and (0.65) is similar to the distance between early and (0.66). The suggested scenario is that the medium group has become genetically distinct by isolation in flowering-time from the early group and that the lower fruit set might not be a long term fitness problem it might vary between years or being compensated by other life-history events. It may also be that the medium group is declining, but we have no records between years. The positive and negative effect of individual microsatellite allele frequencies on relative fruit indicates that there are groups of genetically similar individuals that do differently well in producing fruit, suggesting that natural selection have acted on quantitative traits connected to these groups (LEXTER et al. 2005). Hybridisation between the late-flowering and the early-flowering groups do occur, but we cannot trace the potential effects on fruit set and flowering time. Introgression between the groups is different in different microsatellite loci, suggesting that the less introgressive loci are connected to genes that are incompatible or give effects that lower fitness in hybrids. One of the four infertile plants is actually a hybrid between late and early flowering groups. Among the early-flowering plants there is a genetically distinct group with delayed flowering and a lower relative and absolute fruit set, but the intermediate flowering time is not accompanied by an obvious intermediate genetic composition. The measurement of introgression taken from locus 29 is associated with number of unfilled capsules, indicating detrimental effects of introgression into the group, but positive effects on fruit set of introgression into or within the flowering group. This result offers an explanation where natural selection keep the group distinct and less diverse through reinforcement, while the group benefits from gene flow. REFERENCES Barbour, R. C., Potts, B. M. and Vaillancourt, R. E Gene flow between introduced and native Eucalyptus species: exotic hybrids are establishing in the wild. Aust. J. Bot. 51: Crawley, M. J Statistical computing. John Wiley & Sons. Gustafsson, S. and Lönn, M Genetic differentiation and habitat preference of flowering-time variants within Gymnadenia conopsea. Heredity 91: Klips, R. A. and Culley, T. M Natural hybridization between prairie milkweeds, Asclepias sullivantii and Asclepias syriaca: morphological, isozyme, and handpollination evidence. Int. J. Plant Sci. 165: Lexer, C., Heinze, B., Alia, R. et al Hybrid zones as a tool for identifying adaptive genetic variation in outbreeding forest trees: lessons from wild annual sunflowers (Helianthus spp.). For. Ecol. Manage. 197:

7 228 M. Lönn et al. Hereditas 143 (2006) Lexter et al Miller, M. P Tools for population genetic analyses (TFPGA) 1.3: a Windows program for the analysis of allozyme and molecular population genetic data. Computer software distributed by author. Morjan, C. L. and Rieseberg, L. H How species evolve collectively: implications of gene flow and selection for the spread of advantageous alleles. Mol. Ecol. 13: Proctor, M., Yeo, P. and Lack, A The natural history of pollination. Harper Collins Publishers. R Development Core Team (2004) R: a language and environment for statistical computing. ject.org. Runyeon-Lager, H. and Prentice, H. C Morphometric variation in a hybrid zone between the weed, Silene vulgaris, and the endemic, Silene uniflora ssp. petraea, on the Baltic island of Öland. Can. J. Bot. 78: Servedio, M. R. and Noor, M. A. F The role of reinforcement in speciation: theory and data. Annu. Rev. Ecol. Evol. Syst. 34: Silvertown, J., Servaes, C., Biss, P. et al Reinforcement of reproductive isolation between adjacent populations in the Park Grass Experiment. Heredity 95: Snaydon, R. W. and Davies, M. S Rapid population differentiation in a mosaic environment. IV. Populations of Anthoxanthum odoratum L. at sharp boundaries. Heredity 36: 925. Venables, W. N. and Ripley, B. D Modern applied statistics with S. Springer. Weir, B. S. and Cockerham, C. C Estimating F- statistics for the analysis of population structure. Evolution 38: Wright, S Evolution and the genetics of populations, vol. 4. Variability within and among natural populations. Univ. of Chicago Press.