Floral isolation is an important component of pollinator-driven speciation. However, up to now, only a few studies have quantified its strength and relative contribution to total reproductive isolation. In this study, we quantified floral isolation among three closely related, sympatric orchid species of the genus Ophrys by directly tracking pollen flow. Ophrys orchids mimic their pollinators’ mating signals, and are pollinated by male insects during mating attempts. This pollination system, called sexual deception, is usually highly specific. However, whether pollinator specialization also conveys floral isolation is currently under debate. In this study, we found strong floral isolation: among 46 tracked pollen transfers in two flowering seasons, all occurred within species. Accounting for observation error rate, we estimated a floral isolation index ≥0.98 among each pair of species. Hand pollination experiments suggested that postpollination barriers were effectively absent among our study species. Genetic analysis based on AFLP markers showed a clear species clustering and very few F1 hybrids in natural populations, providing independent evidence that strong floral isolation prevents significant interspecies gene flow. Our results provide the first direct evidence that floral isolation acts as the main reproductive barrier among closely related plant species with specialized pollination.

Floral isolation is a form of prepollination reproductive isolation that can play an important role during the process of plant speciation (Grant 1994; Lowry et al. 2008; Kay and Sargent 2009; Schiestl and Schlüter 2009). Floral isolation can be mediated through the behavior of pollinators (ethological isolation) or the morphology of the flower (mechanical isolation) and work in concert with other, later-acting reproductive barriers (Grant 1994; Fulton and Hodges 1999; Schemske and Bradshaw 1999; Ramsey et al. 2003; Aldridge and Campbell 2007; Schiestl and Schlüter 2009). The determination of the relative importance of different types of reproductive barriers among species has become a central topic in the study of speciation (Ramsey et al. 2003; Coyne and Orr 2004; Cozzolino and Scopece 2008; Lowry et al. 2008; Widmer et al. 2009). Previous studies have shown that in many plants, prezygotic isolation contributes more to total isolation than postzygotic isolation (Rieseberg and Willis 2007; Lowry et al. 2008; Widmer et al. 2009). In the absence of geographic barriers to gene flow (i.e., among sympatric species), floral isolation can be the most important prezygotic barrier. However, the relative strength of prezygotic and postzygotic isolation may differ between species, and may depend on the pollination system (Cozzolino et al. 2004; Cozzolino and Scopece 2008). In orchids, floral isolation has been suggested to be strong, because their associations with pollinators are often highly specific (Schiestl and Schlüter 2009).

Ophrys L. is a genus of sexually deceptive orchids, which mainly occurs in the Mediterranean area. To attract pollinators, these orchids mimic the olfactory, visual, and tactile signals of the females of their associated pollinator insects, and thereby induce so-called pseudocopulations in males, leading to pollination (Kullenberg 1961; Kullenberg and Bergström 1976; Paulus and Gack 1990a,b; Schiestl et al. 2000). In this pollination system, floral odor is the key factor for specific pollinator attraction (Schiestl et al. 1999, 2003; Mant et al. 2005a,b; Peakall et al. 2010). One of the major characteristics of sexual deception is its high specificity, with each species of Ophrys only attracting one or very few species of male insects as pollinator(s) (Paulus and Gack 1990b). Therefore, different Ophrys species, which are mostly genetically compatible and crossable, are potentially isolated from each other due to ethological floral isolation, that is, the nonsharing of pollinator species (Ehrendorfer 1980; Paulus and Gack 1990b; Schiestl and Ayasse 2002; Scopece et al. 2007; Schiestl and Schlüter 2009). Mechanical floral isolation is also present between some Ophrys species, mainly between the sections Pseudophrys and Ophrys (Kullenberg 1950; Ågren et al. 1984; Borg-Karlson 1990; Cortis et al. 2009). Among these groups, different Ophrys species can be pollinated by the same pollinator in sympatry, because pollinia are attached to different parts of the pollinator's body (head or abdomen), thus preventing pollen transfer between species. A recent study by Cortis et al. (2009), however, showed that cross-pollination can occur in natural population despite mechanical isolation, which indicates that mechanical isolation in Ophrys may not be a very strong barrier to gene flow.

Understanding the process of speciation and diversification in Ophrys orchids is challenging due to their high morphological variability, which can mean that it is often difficult to reliably identify species in the field. This is further complicated by the multiple and often highly divergent taxonomic treatments of the group. For example, the number of species in Ophrys listed by different authors ranges from 17 species (and 44 subspecies) (Sundermann 1980) or 19 species (Pedersen and Faurholdt 2007) to 250 species (Delforge 2006). Moreover, recent genetic and molecular phylogenetic studies of Ophrys showed low genetic divergence among species (Soliva et al. 2001; Soliva and Widmer 2003; Devey et al. 2008). The pattern of low genetic differentiation among species can be explained by two (nonexclusive) hypotheses: (1), the genus Ophrys may have undergone (a) recent radiation(s), or (2), there is frequent gene flow among species. Under the first scenario, Ophrys species-diversification is either due to pollinator shifts mediated by a change in key floral traits (such as floral odor bouquets; Mant et al. 2005b; Schlüter et al. 2009; Vereecken et al. 2010) or habitat adaptation. However, under scenario 1, if the time since species diversification is short, neutral genetic structure would not yet be expected to show a clear separation among species (Harrison 1991; Klein 1998). Under scenario 2, it is assumed that the strength of reproductive isolation among sympatric Ophrys species is weak, perhaps due to low fidelity of pollinators, therefore resulting in frequent gene flow among species, which reduces the genetic differentiation among species after their initial divergence (Soliva and Widmer 2003; Devey et al. 2008). One fundamental difference between these two scenarios is the assumed strength of floral isolation among sympatric Ophrys species: the first scenario assumes strong floral isolation, whereas second scenario assumes relatively weak floral isolation.

The absolute pollinator specialization (i.e., the number of pollinators visiting each species) in Ophrys has previously been investigated (e.g., Paulus and Gack 1990b; Mant. et al 2005b). However, the relative pollinator specialization (pollinator sharing, ethological floral isolation) and the resulting proportion of interspecific pollen transfer are still unknown. In this study, we directly tracked pollen flow within and among three sympatric and co-flowering, closely related Ophrys species, and quantified floral isolation as well as components of postmating reproductive barriers among these three species. Additionally, the genetic structure was investigated among species. Specifically, the following questions are addressed in this article: (1) How strong is floral isolation among sympatric Ophrys species? (2) What is the contribution of prepollination, postpollination prezygotic, and postzygotic isolation barriers to the total reproductive isolation among sympatric Ophrys species? (3) What is the proportion of hybrids in natural populations?

Materials and Methods


To most effectively address the question of the relative importance of the different putative isolation mechanisms in Ophrys, a set of species with the following criteria are needed: (1) Species should occur and co-flower in sympatry; (2) species should have the same ploidy level; (3) species should be closely related. According to these criteria, the species Ophrys sphegodesMiller, O. exaltata subsp. archipelagi (Gölz & H.R.Reinhard) Del Prete, and O. garganicaNelson Ex O. & E. Danesch were chosen in this study, because these three species co-flower and co-occur sympatrically in Southern Italy, phylogenetic analysis indicates that these species are closely related (Devey et al. 2008), and ploidy levels of these species are expected to be the same and confirmed in this study (D’Emerico et al. 2005).


The species O. sphegodes, O. exaltata, and O. garganica were identified based on floral morphology, according to criteria described by Mant et al. (2005b). At Capoiale (CAP: 41°54′N, 15°40′E), where all three species co-occur and co-flower, samples of these three species were collected in 2008 and 2009 for both scent and genetic analysis; at Marina di Lesina (MDL: 41°54′N, 15°20′E), where mostly O. exaltata and O. garganica co-occur and co-flower (and only very few individuals of O. sphegodes were found), these two species were collected in 2008 and 2009 only for floral scent analysis; at the more distant Foce Garigliano (FCG: 41°13′N, 13°46′E), where O. exaltata and O. sphegodes co-occur and co-flower (and O. garganica does not occur), these two species were collected in 2008 for both scent and genetic analysis. Each study area was about five hectares in size, and was estimated to contain 2000–3000 flowering plant individuals (counting all three species). For each sampled plant individual, a piece of leaf tissue was collected, and placed in a plastic bag filled with silica gel (Sigma, Buchs, Switzerland) for subsequent molecular analysis, and one labellum of an unpollinated flower was cut, placed in a 2 mL vial (Supelco, Sigma Buchs, Switzerland) and rinsed in 500 μL hexane (Fluka, Sigma Buchs, Switzerland) for 1 min while gently shaking. Thereafter, the labellum was removed and all scent samples were stored at −28°C until being analyzed by gas chromatography (GC). In total, 73 O. sphegodes (49 from CAP, 24 from FCG), 72 O. exaltata (48 from CAP, 24 from FCG), and 26 O. garganica (all from CAP) were sampled for genetic analysis; 100 O. exaltata (35 from CAP, 34 from MDL, 31 from FCG); 94 O. sphegodes (62 from CAP, 12 from MDL, 20 from FCG) and 56 O. garganica (30 from CAP, 26 from MDL) were sampled for floral odor analysis.


We used an experimental approach with a plot design to measure floral isolation. The plots were set up in the field as follows, at the same localities as naturally occurring plants. Two individuals of each species were randomly positioned in each plot (six plants for each plot in CAP and MDL, four plants for each plot in FCG where O. garganica was absent from the natural populations). The distance between neighboring plants was 0.5 m. For each experiment, 20 plots were set up along a transect through the habitat of the orchids. The distance between neighboring plots was 20 m, because the average pollinia-carrying distance of Colletes pollinators was estimated to be around 5 m (Peakall and Schiestl 2004).

Plants for the plot experiments were picked from natural populations, each flower was checked for pollinia removal or pollen deposition, and pollinia were stained alternately with the dyes brilliant green, methylene aniline blue, orange G, and trypan red as described previously (Peakall 1989). The color used for each species was randomized between experiments to reduce potential effects of staining color on pollinator behavior. The inflorescence was put in a water-filled 15-mL plastic tube placed in the ground. Pollinia removal and deposition of massulae were recorded three days after setting up the plots. Because Ophrys massulae are relatively small, and the assessment of their presence requires some experience in the field, there is a potential for observation errors to happen. Thus, to assess the observation error rate, in a subset of the experiments, plants were checked at several time points. For about half of the plant individuals used in the plot experiments, one unpollinated flower labellum was removed from the inflorescence to collect floral odor as described above.

In 2008, two replicates of the experiment were performed at each of the following locations: CAP, MDL, and FCG (in total six experiments, 120 plots). In 2009, two replicates of the experiment were performed at locations CAP and MDL (in total, four experiments, 80 plots). The experiment at FCG was not repeated in 2009 due to the relatively poor overall pollinator activity at this location in 2008 (see Table 1). At each location where experiments were performed, the pollination success of naturally occurring plants in the surrounding area (within 20 m of the transect) was recorded at the end of the flowering season.

Table 1.  Summary of plot experiments. Plot experiments were set up in 2008 and 2009. CAP, MDL, and FCG refer to locations where experiments were set up. Number of total flowers refers to the total number of flowers from three species used in each plot experiment; Number of flowers with stained pollinia refers to number of flowers carrying stained pollinia—there were no flowers with unstained pollinia at the outset of the experiment, but some flowers without pollinia (either due to pollinators or handling by experimenter). Number of plants refers to total number of plants from three species in each plot experiment; pollinated flowers refers to number of flowers receiving pollinia; flowers pollinated by stained pollinia refers to number of flowers that received stained pollinia; pollinia removal refers to number of pollinia removed; interspecies pollinia transfer refers to number of interspecies pollinia movements observed in plots. Number of visited plants was counted as number of plants which lost or received pollinia.
 2008 2009SUM
Number of total flowers1531682022321151411922252052221855
Number of flowers with stained pollinia146154165205961191822081952181686
Number of plants58565960383960605759546
Pollinated flowers1472551331027207131
Flowers pollinated by stained pollinia52843001011346
Pollinia removal38167613141016464822299
Interspecies pollinia transfer00000000000
Visited plants201432812713252515171
Stained pollinia on O. sphegodes flowers201020063115
Stained pollinia on O. exaltata flowers015410017221
Stained pollinia on O. garganica flowers312000031010


Manual crosses were performed in spring 2010 in the greenhouse of the Department of Structural and Functional Biology, University of Naples Federico II. All crossed plants were collected from sympatric natural populations of the three investigated species at CAP, where we set up the plot experiments.

To prevent uncontrolled pollinations, plants were placed in cages covered with a thin net prior to flowering. Pollination experiments were performed by removing pollinia by touching the viscidia with a plastic toothpick and placing them on the stigmas of other plants of the same species (intraspecies pollinations), or of a different species (interspecies pollinations). Care was taken to pollinate no more than two flowers per individual to prevent the potential negative effects of over-pollination on fruit set and seed viability.

All possible crossing combinations among O. exaltata, O. garganica, and O. sphegodes were performed bidirectionally (yielding a total of 78 crossings, see Table 2).

Table 2.  Fruit set formation ratio after hand pollination within and between species pairs. The ratio (%) was calculated by 100 × number of fruit sets/number of crosses performed. No significant differences were found between interspecies and intraspecies crosses.
Crossing typePollen donor (♂)Pollen receiver (♀)Number of crossesFruit setsRatio
InterspeciesO. sphegodesO. garganica1010100
 O. garganicaO. sphegodes1010100
O. exaltataO. sphegodes 8 8100
 O. sphegodesO. exaltata 6 6100
O. exaltataO. garganica10 9 90
 O. garganicaO. exaltata 9 9100
IntraspeciesO. sphegodesO. sphegodes 8 6 75
 O. garganicaO. garganica10 9 90
O. exaltataO. exaltata 7 7100

Ripe fruits were collected and stored in silica gel. Seeds were then observed under an optical microscope with 100× magnification and assigned to two mutually exclusive categories: viable and inviable seeds, based on the presence or absence of embryos, respectively.


Differences in ploidy can provide an important barrier to gene flow. Although we expected all three study species to be diploid, with a chromosome number of 2n = 36 having been reported from O. sphegodes and O. garganica (Greilhuber and Ehrendorfer 1975; D’Emerico et al. 2005), the ploidy level of O. exaltata has not previously been reported. Furthermore, ploidy may also be variable within species and/or among different populations. Therefore, we investigated the ploidy levels of all three species in our study populations. Ploidy levels of the three species were analyzed using pollinia. Two pollinia of a single flower per individual were collected in spring 2010 from CAP and MDL populations. In total, 86 O. sphegodes, 71 O. exaltata, and six O. garganica samples were analyzed. Using flow cytometry, we analyzed the relative ploidy level for each individual separately. For sample preparation and analysis, we followed a two-step protocol (Doležel et al. 2007). Two pollinia were chopped and mashed together with approximately 25-mm2 leaf material of Phaseolus coccineus (2n, 1C = 1.01 ± 0.4 pg; Bennett and Leitch 2005), which served as internal standard (IS), with a sharp razor blade in 1-mL ice-cold Baranyi's solution (0.1 M citric acid, 0.5% Triton X-100; Baranyi and Greilhuber 1995). After filtering the suspension through a 30 μm CellTrics® disposable filter (Partec GmbH, Münster, Germany), the filtrate was centrifuged (5 min, 380 ×g, room temperature) using a Sorvall® RMC 14 centrifuge (Kendro Revco Lindberg Heraeus Sorvall, Asheville, NC). After removal of supernatant, nuclei were resuspended in 40 μL of ice-cold Baranyi's solution. One hundred and sixty micro liters of Otto II solution (0.4 M Na2HPO4) supplemented with DAPI (4′, 6-diamidino-2-phenylindole; final concentration: 4 μg mL−1) were added and relative fluorescence intensity was recorded using a Cell Lab Quanta™ SC-MPL flow cytometer (Beckman Coulter, Fullerton, Canada) with a mercury arc lamp. Only samples with pollinia peaks of at least 1000 counts and a coefficient of variation of less than 10% were analyzed. To determine relative ploidy level of the three species, the ratio between the median of pollinia peaks and the median of IS peaks was calculated.


GC analysis was performed as described by Mant et al. (2005b) with 300 ng n-octadecane (C18) added to the floral extracts as an IS. One micro liter of each sample was injected into an Agilent 6890 GC at 50°C, followed by opening of the split valve and heating to 300°C at rate of 4°C/min. An HP-5 column and flame ionization detector (FID) were used, and hydrogen was used as a carrier gas, with nitrogen as the make-up gas. For identification of compounds, several samples were re-analyzed by GC with a mass selective detector (GC/MSD; Agilent 5975) using the same oven and column parameters. Spectrum and retention time of compounds were compared with those of synthetic standards, that is, alkanes: nonadecane (C19), henicosane (C21), docosane (C22), tricosane (C23), tetracosane (C24), pentacosane (C25), hexacosane (C26), heptacosane (C27), octacosane (C28), nonacosane (C29); and alkenes: (Z)-7-heneicosene [(Z)-7-C21], (Z)-9-heneicosene [(Z)-9-C21], (Z)-7-tricosene [(Z)-7-C23], (Z)-9-tricosene [(Z)-9-C23], (Z)-7-pentacosene [(Z)-7-C25], (Z)-9-pentacosene [(Z)-9-C25], (Z)-11-pentacosene [(Z)-11-C25], (Z)-12-pentacosene [(Z)-12-C25], (Z)-7-heptacosene [(Z)-7-C27], (Z)-9-heptacosene [(Z)-9-C27], (Z)-11-heptacosene [(Z)-11-C27], (Z)-12-heptacosene [(Z)-12-C27], (Z)-7-nonacosene [(Z)-7-C29], (Z)-9-nonacosene [(Z)-9-C29], (Z)-11-nonacosene [(Z)-11-C29], (Z)-12-nonacosene [(Z)-12-C29], where (Z)-number indicates the cis double-bond position. For sources of standard compounds see Mant et al. (2005a). It is noted that the discrimination of (Z)-11- and (Z)-12-alkenes was not possible with the GC parameters used. The relative amount of each odor compound was calculated as the proportion of the total amount of all alkenes and alkanes of a chain length between 18 and 30 carbons.


Genomic DNA was extracted using GenElute Plant Genomic DNA Miniprep Kit (Sigma-Aldrich, Italy). The AFLP procedure was performed as described by Vos et al. (1995), with modifications as reported in Moccia et al. (2007) using fluorescent dye-labeled primers. An initial trial using 14 different primer combinations on four individuals each of O. sphegodes and O. exaltata was conducted to identify those primers that yield the highest number of easily detectable polymorphic peaks that were different between the two species. After the screening, six selective primer combinations were chosen: FAM-EcoRI-AGC/MseI-ACAC, NED-EcoRI-ACC/MseI-ACTG, HEX-EcoRI-AGC/MseI-ATCG, FAM-EcoRI-ATG/MseI-CGG, NED-EcoRI-AAC/MseI-CGC and HEX-EcoRI-AGC/MseI-CCAA. For the restriction digestion, the enzymes EcoRI and MseI were used on a total of 250 ng of genomic DNA. Ligation of EcoRI and MseI adapters took place in the same reaction. Two microliter of the restriction-ligation product were used for a preselective PCR with primers having one selective base. For the successive selective PCR, 1 μL of a 1:10 dilution of the PCR product was used. Primers were the same as in the preselective PCR, but with three or four selective bases. Fragment separation and detection took place on a 3130 Genetic Analyzer (Applied Biosystems, Foster City, CA). GeneScan-500 LIZ (Applied Biosystems) was used as IS. Processing of the raw data and sizing of the fragments were done with Genemapper 3.7 software (Applied Biosystems). Absence or presence of AFLP bands was carefully scored by eye. To avoid artifacts, only AFLP markers that could be unambiguously scored over the whole dataset were included in the binary matrix.

AFLP analysis was performed as two experiments at different dates and runs, and scored independently, preventing us from merging the two AFLP datasets. These two separate datasets were therefore analyzed separately: the first dataset contains 58 O. sphegodes (34 from CAP, 24 from FCG), 55 O. exaltata (31 from CAP, 24 from FCG), and 26 O. garganica (all from CAP) individuals; the second dataset contains 30 O. sphegodes (15 from CAP, 15 from FCG) and 32 O. exaltata (17 from CAP, 15 from FCG). In the second dataset, both species from population FCG were the same individuals as in the first dataset.


The strength of each type of reproductive isolation barrier was calculated based on the quantitative approach suggested by Lowry et al. (2008) and Martin and Willis (2007). The floral isolation index was calculated based on the following formula: RIfloral= 1 − (observed/expected interspecies pollen flow)/(observed/expected intraspecies pollen flow).

The precision of this estimate of RIfloral is limited by the number of observed events in our experiments and any potential observation error introduced when checking plants for pollination or pollinia removal. Therefore, RIfloral was recalculated so as to account for these errors. Three of our experiments were checked twice, any error recorded, and these data were used to estimate the observation error rate. For example, on the same flower, massulae might have been recorded at the first but not on the second inspection, indicating that one of these two datapoints was probably erroneous. This error rate followed a pattern, that is the best fitted curve decreased exponentially with the number of experiments we performed, concordant with our expectation that observation error rate decreases as the observers’ experience increases [ln(Error rate) =−0.25 ×i± SD, where i refers to the number of experiments performed by that time]. Based on this formula, the error rate for each experiment was estimated, and the floral isolation index recalculated by allowing observation error to occur at the estimated error rate ± SD. We repeated this estimation procedure 1000 times to obtain a simulated distribution of floral isolation index values, and used this distribution to obtain mean and 95% confidence values for RIfloral.

Because in orchids, female gametophyte development and fruit set formation usually happens after successful pollination with compatible pollen (Zhang and O’Neill 1993), the postmating prezygotic isolation index can be estimated as the proportion of fruit set (i.e., capsules) formed following interspecies pollinations, relative to the proportion of fruit set formed following intraspecies pollinations with each parental species (Scopece et al. 2007): RIpostmating-pre zygotic= 1 − (ratio of fruit set formed in interspecies crosses)/(average ratio of fruit set formed in parental intraspecies crosses). Similarly, the postzygotic isolation index was estimated by viable seeds and quantified based on the following formula (Scopece et al. 2007): RIpostzygotic= 1 − (proportion of viable seeds in interspecies crosses)/(average proportion of viable seeds in parental intraspecies crosses).


Linear discrimination analysis (LDA) was used for analysis of floral scent based on relative amounts of hydrocarbons. Comparison of fruit set formation ratios among interspecies crosses and intraspecies crosses was performed by Fisher's exact tests. The significance of different seed viability among interspecies and intraspecies crosses was assessed using Student's t-test, after normality testing of the data distribution by the Shapiro test (Royston 1982). Statistical analysis of AFLP data was performed in FAMD 1.25 (φST and PCoA) (Schlüter and Harris 2006) and Hindex 1.41 (hybrid index; Buerkle 2005). Principal coordinate analysis (PCoA) was based on Jaccard's similarity coefficient.

Correlations between pairwise floral odor and genetic distance was assessed in a Mantel test (10,000 permutations). Here, floral odor distance was calculated as a Euclidean distance, and genetic distance was calculated as 1 − Jaccard's similarity. Except for the analysis of AFLP data, all statistical analyses in this study were carried out in R 2.11.0 (R Development Core Team 2010).



Among the natural populations, the average pollination success (defined as the percentage of pollinated flowers) was 10.8% in O. sphegodes, 18.8% in O. exaltata, and 3.2% in O. garganica, based on two flowering seasons (2008 and 2009). Furthermore, within each species, pollination success varied among populations (Fig. 1). No significant differences were found between the two years of observations. The overall patterns of pollination success in our plot experiments were similar to the natural populations. Pollination success in plot experiments was about 7.8%, 9.2%, and 4.1% for O. sphegodes, O. exaltata, and O. garganica, respectively. The population variation in pollination success was similar for natural populations and plot experiments, except for O. sphegodes in MDL where pollination success was much lower than in CAP (3.9%), whereas in plot experiments, the pollination success was similar to CAP (8.9%).

Figure 1.

Pollination success of the three Ophrys species in different populations in 2008 and 2009. (A) natural population; (B) plot experiments. Ophrys exaltata, O. sphegodes and O. garganica are shown in a gray scale as indicated in the figure inset. Names on the axis refer to populations (see Materials and Methods). The number of plants surveyed is indicated on the top of each bar.


The flowering time of O. sphegodes and O. exaltata was similar, whereas the peak of O. garganica blooming was 1–2 weeks later (Xu, field observations). However, there was a broad overlap in flowering time, with around 70% of O. garganica flowers, and around 95% of O. sphegodes and O. exaltata flowers open during the experimental period.

Of all 1855 flowers and 1686 stained pollinaria used in the experiments, 131 flowers (7.1%) were pollinated. Among these pollinated flowers, 46 flowers were pollinated with stained pollinia. Fifteen O. sphegodes flowers, 21 O. exaltata flowers, and 10 O. garganica flowers received stained pollinia (Table 1). All of the 46 pollination events with stained pollinia were within species, and we did not observe a single interspecies transfer. Thus, because no interspecies pollen flow was observed, the floral isolation index equals 1. Likewise, the simulated data incorporating observation error rates (see Methods) showed strong prepollination reproductive isolation among each species, RIfloral≥ 0.98 (Table 3 and Fig. S1).

Table 3.  Components of reproductive isolation among Ophrys species (numbers are isolation indices; mean ± SD). Indices of both postmating prezygotic, and postzygotic isolations are not significantly different from zero.
Species ASpecies BFloral isolationPostmating prezygoticPostzygotic
O. sphegodesO. exaltata0.980±0.0015−0.14−0.22±0.53
O. garganicaO. exaltata0.982±0.0015 0.05 0.16±0.41
O. garganicaO. sphegodes0.980±0.0015−0.21−0.04±0.47


Postmating, prezygotic isolation was estimated as the fruit set ratio after hand pollination. Most of the inter- and intraspecies crosses led to the development of capsules (Table 2). The lowest fruit set ratio was found for intraspecies crosses in O. sphegodes, although this was not statistically significant when compared to other crosses. Thus, the postmating prezygotic isolation index among each species was estimated to be very low. For species pairs O. sphegodes/O. exaltata and O. sphegodes/O. garganica, the isolation indices were negative (however, not significantly different from zero), which might indicate that interspecies crosses performed better than intraspecies crosses, and for O. garganica/O. exaltata, the isolation indices was close to zero (Table 3).

The proportion of seeds with embryos (viable seeds) was used to estimate the postzygotic isolation index. The number of seeds analyzed for each capsule was 324 ± 103 (mean ± SD). Among all fruits, the average percentage of viable seeds was 46.8 ± 21.8% (mean ± SD). We did not find a significant difference between any inter- or intraspecies crosses (Fig. 2). Similar to postmating prezygotic isolation, the mean postzygotic isolation index was also negative between species pairs O. sphegodes/O. exaltata and O. sphegodes/O. garganica, whereas for O. garganica/O. exaltata, the mean value index was slightly higher (Table 3). However, statistical analysis showed that none of these values were significantly different from zero.

Figure 2.

Presence of embryos for crosses among each species (E: O. exaltata; G: O. garcanica; S: O. sphegodes). No difference was found between interspecies and intraspecies crosses based on Student's t-test, (P > 0.05). Error bars depict the standard error of the mean.


No difference in the ploidy level was detected among the three species. The ratios of the relative inflorescence intensity between the pollinia and the IS are shown in Figure S2. All samples of these species showed similar relative genome size. It is most likely that all three species in our study are diploid, because previous studies showed that O. sphegodes and O. garganica are diploid (Greilhuber and Ehrendorfer 1975; D’Emerico et al. 2005).


The differences in floral odor bouquets among studied species were similar to those reported previously (Mant et al. 2005b). The major floral odor difference among the species was the proportion of different alkenes. By LDA, 236 of 250 samples (94.4%) were classified as the same species as they were identified in the field based on floral morphology according to the criteria described by Mant et al (2005b). The morphological/chemical identification mismatch rate between O. sphegodes and O. exaltata was 4.1%, between O. sphegodes and O. garganica 3.3% and between O. garganica and O. exaltata 0.64%.


AFLP datasets one and two contained 242 and 322 markers, respectively. Genetic divergence among population pairs, as estimated by pairwise φST, was relatively low. For the first dataset, where all three species from CAP were analyzed, the lowest φST value (0.044) was found between O. sphegodes and O. garganica, and the highest (0.064) was found between O. garganica and O. exaltata (see Table 4). However, the differences among species pairs were very small. For the second dataset, where O. sphegodes and O. exaltata from both CAP and FCG were analyzed, the highest φST (0.074) was found between O. sphegodes in CAP and O. exaltata in FCG, whereas φST values between O. sphegodes in CAP and O. sphegodes in FCG, and between O. sphegodes in CAP and O. exaltata in CAP were lowest (0.059) (Table 4). Overall, φST values were low, within-species φST values being slightly lower than between-species values.

Table 4.  Pairwise population differentiation (φST) and overlap based on floral odor bouquet and AFLP among each species for two datasets. Dataset 1 contains all three species from CAP, whereas dataset 2 contains O. exaltata and O. sphegodes from CAP and FCG. The number of matches in floral odor was assigned based on the comparison between linear discrimination analysis of floral odor bouquet and species identification in the field. Matches in AFLP were assigned based on visual comparisons between principal coordinate analysis of AFLP markers and species identification in the field.
Species ASpecies BφSTNumber of matches in floral odor/Total number of samplesNumber of matches in AFLP/Total number of samples
Dataset 1
 O. sphegodes CAPO. garganica CAP0.0440/173=0%2/75=2.67%
 O. sphegodes CAPO. exaltata CAP0.0552/209=0.95%1/97=1.03%
 O. garganica CAPO. exaltata CAP0.0641/138=0.72%1/74=1.35%
Data set 2
 O. exaltata FCGO. sphegodes CAP0.0745/68=7.35%0/39=0%
 O. exaltata FCGO. exaltata CAP0.063121/126=96.03%1/41=2.44%
 O. exaltata in FCGO. sphegodes FCG0.0685/68=7.35%3/49=6.12%
 O. sphegodes CAPO. exaltata CAP0.0592/209=0.95%0/32=0%
 O. sphegodes FCGO. exaltata CAP0.0734/116=3.44%0/42=0%
 O. sphegodes FCGO. sphegodes CAP0.059150/151=99.34%5/40=12.5%

Genetic structure among species was investigated by PCoA (Fig. 3). These analyses suggest that the genetic similarity between O. sphegodes and O. garganica is higher than between the species pairs O. sphegodes/O. exaltata or O. garganica/O. exlatata. Although a few outliers were found for both CAP and FCG populations, the three species formed genetically separable clusters (Fig. 3A,B,C and Table 4). For the O. sphegodes/O. exaltata species pair, floral odor showed significant correlation with genetic distance for population FCG (r= 0.42, P = 0.0001), but not for population CAP (r=−0.11, P = 0.69). For species pairs O. sphegodes/O. garganica and O. garganica/O. exaltata in CAP, significant correlations were found in both (r= 0.28, P= 0.0026 and r= 0.17, P = 0.036, respectively).

Figure 3.

Plots of genetic structure (PCoA) and floral odor discrimination analysis for three species. (A) genetic structure of O. exaltata, O. garganica, and O. sphegodes in population CAP. Axis x and y represent 10.5% and 5.4% of variance, respectively; (B) genetic structure of O. exaltata and O. sphegodes in population FCG. Axis x, y, and z represent 12.1%, 8.1%, and 7.1% of variance, respectively; (C) genetic structure of O. exaltata and O. sphegodes in populations CAP and FCG. Axis x, y, and z represent 9.1%, 8.6%, and 5.9% variance, respectively; (D) floral odor bouquet of O. exaltata, O. garganica, and O. sphegodes in CAP, LD1, and LD2 representing 58.5% and 41.2% of trace, respectively; (E) floral odor bouquet of O. exaltata and O. sphegodes in FCG, LD1 and LD2 representing 55.9% and 44.1% of trace, respectively; (F) floral odor bouquet O. exaltata and O. sphegodes in FCG, CAP and MDL, LD1 and LD2 representing 55.9% and 44.1% of trace, respectively. An asterisk (*) indicates outliers for which both AFLP and floral odor bouquet were analyzed. Different colors indicate species (red: O. exaltata; green: O. garganica; blue: O. sphegodes).

For population CAP, no obvious F1 hybrids were found among the three species, as defined by a mean maximum-likelihood hybrid index estimate between 0.4 and 0.6. In contrast, for population FCG, two samples were classified as potential F1 hybrids between O. sphegodes and O. exaltata according to the same criteria (Fig. 4). Both samples were classified as O. sphegodes based on floral odor discrimination analysis (Fig 3F). Overall, for samples from both populations (146 individuals), the percentage of F1 hybrids was very low (1.37%).

Figure 4.

Hybrid index (with 95% confidence intervals) for individuals of each sympatric species pair. The two dashed lines indicate hybrid index h= 0.4 and h= 0.6 cutoff values for assigning putative F1 hybrids. Sample names of two putative hybrids are indicated. (A) Individuals of O. exaltata and O. garganica in population CAP; (B) Individuals of O. exalatata and O. sphegodes in population CAP; (C) Individuals of O. exaltata and O. sphegodes in population CAP; (D) Individuals of O. exaltata and O. sphegodes in population FCG.


Reproductive isolation has been a central topic in the study of speciation (Coyne and Orr 1998; Moyle et al. 2004; Rieseberg and Willis 2007; Scopece et al. 2007, 2008; Schiestl and Schlüter 2009; Widmer et al. 2009). Here, we quantified three different kinds of reproductive barriers (floral isolation, postpollination prezygotic isolation, and postzygotic isolation), as well as ploidy level among three sympatric sexually deceptive Ophrys orchids using experimental approaches. Among these potential barriers, floral isolation was found to be very strong (RIfloral≥ 0.98), whereas later-acting barriers were effectively absent in our study species. Furthermore, population genetic analysis showed a clear separation between species despite low genetic divergence, with few hybrids within natural populations. Our results shed light on the role of plant–pollinator interactions in the evolution of reproductive isolation and plant speciation. We suggest that pollinator adaptation, which conveys strong floral isolation, is the main driver of speciation in this plant group with highly specialized pollination.


Floral isolation has been found in many plant–pollination systems, such as Ipomopsis (Grant 1992), Mimulus (Schemske and Bradshaw 1999; Ramsey et al. 2003), Nicotiana (Ippolito et al. 2004), Petunia (Hoballah et al. 2007), and Silene (Goulson and Jerrim 1997; Wälti et al. 2008), and meta-analyses indicate that floral isolation acts as a strong reproductive barrier in various families of flowering plants (Grant 1994; Lowry et al. 2008; Schiestl and Schlüter 2009; Kay and Sargent 2009; Schiestl, in press). However, there are few examples where floral isolation alone is sufficient to maintain species differentiation in sympatry (Kay and Sargent 2009; Schiestl, in press). In most studied cases, floral isolation acts together with other isolation barriers (postpollination isolation, ecogeographic isolation etc.; Lowry et al. 2008). Co-occurrence of floral isolation with other isolation barriers may be due to two reasons: (1) in most plant systems, it is unlikely that floral isolation has initially evolved in sympatry, because the shift to completely new pollinators may require changes in many floral traits (but see Bradshaw and Schemske 2003 and Hoballah et al. 2007). Therefore, geographical or habitat-associated barriers would often be involved in the evolution of floral isolation; (2) once floral isolation is established, secondary isolation barriers can build up over time (Via and West, 2008; Matute et al., 2010; Moyle and Nakazato, 2010). However, to better understand the contribution of floral isolation to plant speciation, as well as its evolutionary patterns, cases in which only floral isolation is involved in the speciation process are particularly valuable.


As shown in our study, Ophrys may represent a case in which floral isolation is the most important barrier to gene flow among species with a large geographic overlap, sympatric occurrence, and flowering time overlap in their given habitats. Here investigated Ophrys species showed strong ethological floral isolation and a lack of postpollination isolation barriers. Our conservative estimation of the strength of floral isolation, which took into account the number of trackable pollination events and the observation error, showed that the floral isolation index among each Ophrys species pair was higher than 0.98. This estimation is consistent with our AFLP data, which indicate 1.37% (two of 146 samples) putative F1 hybrids between O. sphegodes and O. exaltata. Our finding of strong floral isolation is also consistent with expectations from pollinator-behavior studies often indicating little pollinator sharing among co-flowering species pairs (Kullenberg 1961; Paulus and Gack 1990b; Mant et al 2005a, 2005b; Schlüter et al. 2009). Although we have only partially quantified postzygotic isolation barriers, it is clear from our results that the early acting floral isolation is the major (if not the only) reproductive barrier among closely related Ophrys species.

Recently, floral isolation in Ophrys has come under scrutiny, because studies based on genetic markers argued that floral isolation in Ophrys might be weak and allow for considerable gene flow across species boundaries (Soliva and Widmer 2003; Devey et al. 2008). However, the data presented in these studies only allow for an indirect inference on floral isolation. The genetic pattern observed by Soliva and Widmer (2003) estimated gene flow among species based on FST, however, such estimation can be misleading and should be interpreted cautiously (reviewed by Whitlock and McCauley 1999). In the study by Devey et al. (2008), based on phylogenetic analysis, the authors suggested that cross-pollination is common among species because no clear phylogenetic patterns were found based on DNA sequences (ITS and plastid markers) and AFLP markers. However, gene flow cannot be assessed based on phylogenetic analysis without proper population genetic data (Slatkin 1985). Interestingly, Devey et al. (2008) found no significant genetic differentiation between O. exaltata and O. sphegodes (based on their ITS, plastid markers, and AFLP), whereas we found a clear clustering pattern with AFLP markers and even more so with floral-odor bouquet analyses. This discrepancy is likely because our study provides a more fine-grained resolution through the analysis of multiple populations and large sample sizes. We suggest one should be careful in drawing any conclusions on gene flow from investigations using only molecular markers or morphological data (see also discussion on hybridization in Ophrys below). As a consequence of the suggested gene flow across putative species boundaries, and the typically high variability among individuals in Ophrys, some authors have lumped several species together, resulting in a classification of few species and many subspecies (Pedersen and Faurholdt 2007). In Ophrys, species identification based on morphological characters alone can indeed be difficult. Floral scent, however, often shows a specific pattern among closely related, and morphologically very similar species (Mant et al. 2005b; Stökl et al. 2009), and should thus be taken into consideration when assigning individuals into species categories or testing such assignments. To better understand reproductive isolation and the implicated taxonomic consequences in Ophrys, we suggest that a complementary approach should be taken. This approach should incorporate the quantification of floral isolation and later-acting reproductive barriers (Ramsey et al. 2003), and combine these results with the analysis of the traits under selection (e.g., floral odor) as well as neutral molecular markers.


In the studied species only ethological isolation contributes to floral isolation, because all three species attach pollinia to their pollinators’ heads, and hence there is no evidence of mechanical isolation. In Ophrys, floral odor acts as a key trait for specific pollinator attraction (Schiestl et al. 1999; Mant et al. 2005b) and is therefore likely responsible for ethological isolation among species. Our discriminant function analysis of floral odor bouquets showed a clear separation among each species (Fig. 3 D,E,F). It has been shown that O. sphegodes attracts males of Andrena nigroaenea by emitting a hydrocarbon mixture with high proportions of (Z)-9 and (Z)-11/12 alkenes (Schiestl et al. 1999) (Fig. S3), whereas O. exaltata attracts males of Colletes cunicularius (Fig. S4) by emitting high proportions of (Z)-7 alkenes (Mant et al. 2005a). The pollinator of O. garganica has been reported to be Andrena carbonaria (Paulus and Gack 1990b) (Fig. S5); O. garganica emits high proportions of (Z)-9 and (Z)-11/12 alkenes with different carbon chain lengths (typically longer than in O. sphegodes), however, the active compounds for its pollinator have not yet been identified. Behavioral tests showed that C. cunicularius was only attracted by the floral odor of O. exaltata, but not by the other two species (Mant et al. 2005b). Recently, Vereecken and Schiestl (2009) showed that floral color differences between O. sphegodes and O. exaltata do not contribute to species-specific pollinator attraction. Collectively, these data suggest that strong ethological isolation in Ophrys is primarily due to different floral odor bouquets produced by each species, which are linked to the attraction of different, highly specific pollinators.


Pollinator adaptation may drive floral diversification and speciation in Ophrys, however, hybridization has sometimes been considered to be common among Ophrys species (Devey et al. 2008). In contrast, putative Ophrys F1 hybrids (as identified by morphology) were often found to be solitary, with large number of plants from the parental species surrounding them (Stebbins and Ferlan 1956), suggesting hybridization may not happen frequently. In accordance with this, we found only two putative F1 hybrids among 146 samples (1.37%) of O. sphegodes and O. exaltata. Both were found in one population (FCG). Those two putative F1 hybrids produced a floral odor bouquet similar to O. sphegodes. Possible (nonexclusive) reasons for hybrids found in natural population could be the following: (1) the strength of floral isolation may be variable among populations due to variable specificity in the responses of pollinators to floral odor bouquets; (2) Changes of floral odor in Ophrys may happen through occasional changes in scent genes, leading to a break-down of floral isolation. To test the first possibility, floral isolation should be assessed in various populations. We found consistently strong floral isolation in two adjacent populations, but could not precisely estimate floral isolation in the more distant population FCG, because the total number of pollination events observed in this population was small (only three in total). A study by Vereecken et al. (2010) suggested that floral isolation among Ophrys species pollinated by C. cunicularius and Andrena nigroaenea can break down in some populations, although the frequency of hybrids in that study was always much lower than that of parental species. Break down of floral isolation was also found in some populations among Ophrys species pollinated by other Andrena species (Stökl et al. 2008; Cortis et al. 2009). Varying strengths of floral isolation would suggest a geographical mosaic, with merging of populations through hybridization in some areas and divergence through strong floral isolation in other areas. This geographic mosaic may help to explain the phylogenetic pattern of Ophrys species observed in previous studies (Soliva et al. 2001; Devey et al. 2008). However, further investigations about geographical variation in pollinator behavior and floral isolation are needed to evaluate this hypothesis.

A second reason for hybrids occasionally found in nature may be the genetic basis of floral odor changes in Ophrys. Because changes in floral odor production in Ophrys may be brought about by few genetic changes (Schlüter and Schiestl 2008, Schlüter et al. 2011), one would expect that some individuals of one species could stochastically evolve the same floral odor as another species through mutation or recombination. This would eventually lead to hybridization in natural populations, considering that floral odor is the major attractant for specific pollinators in this system and postpollination barriers are effectively absent. This hypothesis is consistent with the genetic and floral odor analyses in this study. Among the samples investigated here, a few plant individuals showed mismatches among their assignments from genetic and odor data (samples S07, G10, G14, S47 in Fig. 3, summary in Table 4). In other words, these samples have the neutral genetic background of one species, but an odor phenotype of another species, possibly due to changes in few genes controlling floral odor production. However, to further test this hypothesis, detailed studies on the genetic basis of floral odor components in Ophrys, and their consequence for pollinator attraction are needed.


As a scenario for speciation in Ophrys, we propose that incipient Ophrys species adapt to different pollinators by changing floral traits, especially floral odor, that convey strong floral isolation and induce the speciation process. In Ophrys, pollination success is relatively low due to pollen limitation (compare pollination success of natural populations and fruit set rate from hand pollination); a sexually deceptive pollination mechanism may thus induce negative density-dependent selection: high population density may lead to low pollination success because pollinators are more likely to learn and avoid the deceptive flowers. Therefore, a shift in pollinators mediated by a change in floral scent genes may convey a selective advantage by increasing pollination success in the initially few novel genotypes. Furthermore, as shown in our study, different pollinators in Ophrys are associated with strong floral isolation, which is sufficient to prevent significant gene flow in sympatry. Changes of floral odor bouquets may be based on changes in few genes involved in the biosynthesis (or regulation) of pollinator-attractive floral compounds (Schlüter and Schiestl 2008; Schlüter et al. 2011). Therefore, speciation in Ophrys could happen rapidly, even in sympatry. For example, Vereecken et al. (2010) showed that novel floral odor bouquets in Ophrys could evolve rapidly (after only one generation of hybridization), and directly lead to pollinator shifts in sympatry. The remarkable plant–pollinator interaction in Ophrys orchids provides a particularly interesting system to study pollinator adaptation directly involved in species divergence, a process that may be important in several other, highly specific pollination systems (Schiestl, in press).


By in situ tracking pollen flow and experimental hand pollination, we found floral isolation to be very strong among closely related, sympatric Ophrys species, whereas later-acting barriers to gene flow were effectively absent. Our results provide direct evidence that the reproductive barrier among these closely related plant species with specialized pollination consists mostly of floral isolation. In such a system, pollinator adaptation could directly lead to floral isolation and speciation. This offers a particular opportunity to study the role of floral isolation during the evolution of reproductive isolation and speciation. However, further studies that systematically combine neutral traits (such as molecular markers), traits under selection (such as floral odor) and their genetic basis, pollinator behavior, as well as quantification of floral isolation in natural habitats will be helpful to better understand speciation processes in plants with specialized pollination systems.

Associate Editor: M. Burb


We wish to thank M. Menz for his help in the field and E. Connor for assistance with the GC analysis as well as for providing comments on the manuscript. We thank two anonymous reviews for their constructive comments. We are grateful for financial support by the ETH Zürich (TH0206–2 to FPS), the Austrian Science Fund (FWF, J2678-B16 to PMS), and PRIN program 2010 (to SC).