Biogeographic variation in genetic variability, apomixis expression and ploidy of St. John's wort (Hypericum perforatum) across its native and introduced range

Abstract

Background and Aims

St. John's wort (Hypericum perforatum) is becoming an important model plant system for investigations into ecology, reproductive biology and pharmacology. This study investigates biogeographic variation for population genetic structure and reproduction in its ancestral (European) and introduced (North America) ranges.

Methods

Over 2000 individuals from 43 localities were analysed for ploidy, microsatellite variation (19 loci) and reproduction (flow cytometric seed screen). Most individuals were tetraploid (93 %), while lower frequencies of hexaploid (6 %), diploid (<1 %) and triploid (<1 %) individuals were also identified.

Key Results

A flow cytometric analysis of 24 single seeds per individual, and five individuals per population demonstrated opposite patterns between ploidy types, with tetraploids producing more apomictic (73 %) than sexual (24 %) seed, while hexaploids produced more sexual (73 %) than apomictic (23 %) seed. As hexaploids are derived from tetraploids, these data imply that gene dosage, in addition to the effects of hybridization, influences the switch from apomictic to sexual reproduction. No significant differences in seed production were found between Europe and North America. An analysis of population structure based upon microsatellite profiling demonstrated three major genetic clusters in Europe, whose distribution was reflective of Pleistocene glaciation (e.g. refugia) and post-glacial recolonization of Europe.

Conclusions

The presence of pure and mixed populations representing all three genetic clusters in North America demonstrates that H. perforatum was introduced multiple times onto the continent, followed by gene flow between the different gene pools. Taken together, the data presented here suggest that plasticity in reproduction has no influence on the invasive potential of H. perforatum.

INTRODUCTION

The relative ability of different plant taxa to invade new biogeographic regions successfully is dependent upon a number of biological and physical factors, one of which is the reproductive system, which directly influences population structure, gene flow and evolutionary potential. Considering seed formation, plants can reproduce through sex (selfing and outcrossing) or apomixis (asexual reproduction through seed; Nogler, 1984).

Apomixis has evolved from sexual reproduction, and has been reported from >40 mono- and dicotyledonous plant families, the most frequent being in the Asteraceae, Hypericaceae, Poaceae and Rosaceae (Asker and Jerling, 1992). Upon the origin of an apomictic lineage, a multitude of factors can lead to niche differentiation relative to its sexual ancestors, with effects on both the spread and relative competitiveness of either reproductive form. For example, sexuality is advantageous with respect to purging mutations (Muller, 1964; Kondrashov, 1982), generating genetic variance (Fisher, 1930; Crow, 1970) and adapting to changing environments and parasite interactions (Van Valen, 1973; Bell, 1982). Asexuality, on the other hand, may be a better strategy in stable environments as genetically identical offspring from a single mother are equally fit, or for range expansion when an asexual population is expected to grow more rapidly compared with a sexual one (i.e. the 2-fold cost of sex; Maynard Smith, 1978). In a number of plants (and animals), it has been shown that sexual and asexual taxa exhibit different ranges of distribution (Bierzychudek, 1987; Kearney, 2005), a phenomenon referred to as ‘geographical parthenogenesis’ (Vandel, 1928).

Interestingly, most apomictic plant and animal taxa are hybrid and/or polyploid in nature, and both phenomena have been linked to the origin and/or stability of apomixis (Suomalainen, 1950; Carman, 1997; Richards, 2003). In addition, polyploidy and heterosis-like responses to hybridization could lead to elevated fitness of apomicts in specific environments (Carman, 1997; Grimanelli et al., 2001; Richards, 2003; Kearney, 2005). Although apomixis by definition implies clonal reproduction and limited genetic variability, apomictic plants are typically facultative, and hence frequently retain the ability to produce seeds through a variety of mechanisms (Matzk et al., 2001; Halket et al., 2005; Aliyu et al., 2010). For example, Aliyu et al. (2010) have shown genotype-specific quantitative variation for seed developmental traits associated with facultative apomixis. Hence, facultative apomixis, in conjunction with hybridity and polyploidy, could provide a source of genetic and phenotypic variation upon which natural selection could act, and this would be reflected in an apomictic lineage's ability to invade a novel environment successfully. Baker's law (Baker, 1967) states that species with uniparental reproduction are more likely to establish in a new environment after long-distance dispersal (e.g. Hao et al., 2011). Consequently, not only selfers but also apomicts should in general be better colonizers (Hörandl, 2006; Hörandl et al., 2008).

St. John's wort (Hypericum perforatum) is such an invasive species which is indigenous to central and eastern Europe; it is self-compatible and can reproduce through sex or apomixis (Noack, 1939; Matzk et al., 2001). Hypericum perforatum has successfully invaded North America since the first record of introduction in Lancaster, Pennsylvania in 1793 (Sampson and Parker, 1930). Its high genotypic plasticity (Maron, 2006) in conjunction with variable levels of facultative apomixis (Matzk et al., 2001) are hypothesized to have contributed to its rapid spread throughout the continent. For example, in an analysis of multiple phenotypic traits, Maron et al. (2004) demonstrated that the introduction of H. perforatum into North America was accompanied by rapid climatic adaptation.

Hypericum perforatum reproduces mainly through aposporic parthenogenesis, whereby a nucellar cell changes its fate to become an aposporous initial (AI) cell which grows into an embryo that is genetically identical to the mother plant, while the original megaspore mother cell (MMC) degenerates (Galla et al., 2010). Fertilization of the AI typically does not occur, while pseudogamy (fertilization of the central cell to produce a functional endosperm) characterizes most apomictically derived seed (Matzk et al., 2001). A single individual can produce both sexual and apomictic seeds, although this can range naturally from nearly obligate apomixis to complete sexuality (Noack, 1939; Matzk et al., 2001).

Hypericum perforatum is typically tetraploid, although diploid, triploid and hexaploid plants have been identified (Matzk et al., 2001; Robson, 2002; Koch et al., 2013). It was long unclear whether H. perforatum is allotetraploid or autotetraploid, and diploid populations have not been described from wild populations until very recently (Koch et al., 2013). Hybridization between the diploids H. attenuatum and H. maculatum, followed by chromosome doubling, has been hypothesized on the basis of their distinct morphology and separate geographic distribution (Campbell and Delfosse, 1984; Robson, 2002). In contrast, cytogenetic analyses of different Hypericum species implied that H. perforatum may have originated through self-polyploidization from an ancestor closely related to the diploid H. maculatum (Brutovská et al., 2000). However, most recently it was demonstrated that no hybridization between different species gave rise to polyploid H. perforatum (Koch et al., 2013). Diploid populations were identified from glacial refuge areas in Southern Europe and the Balkans, making up two different ancestral H. perforatum gene pools from which polyploid H. perforatum evolved independently several times. However, secondary hybridization and introgression with H. maculatum is a very common and frequently observed phenomenon (Koch et al., 2013). This led to the recognition of three major gene pools within H. perforatum based on amplified fragment length polymorphism (AFLP) data and chloroplast DNA (cpDNA) sequence variation (Koch et al., 2013).

Taken together, hybridization, introgression or admixture (Ellstrand and Schierenbeck, 2000) are factors which could have had dual effects on H. perforatum, on the one hand being possibly associated with apomixis induction and stability, and on the other hand influencing invasiveness via heterosis-like effects. Considering the ability of facultative apomictic H. perforatum to produce variable levels of sexual seed (Matzk et al., 2001), their ability both to outcross and to self-fertilize, and their invasiveness and human-mediated introduction (Maron et al., 2007), gene flow between genotypes from different geographic origins is hypothetically possible. In the present study we examine the correlation between genetic variation, ploidy-level variation, gene flow and quantitative variation for facultative apomictic seed production as factors which may have influenced the ability of H. perforatum to invade and colonize North America. Using an analysis of a collection of European native and North American invasive accessions, we examine biogeographic differentiation in both natural and introduced populations, and test whether variation in apomixis traits is correlated with the propensity for H. perforatum to invade novel environments.

MATERIALS AND METHODS

Plant material and genome size (ploidy) analysis of leaf tissue

Relative ploidy levels of 2019 plants from 43 localities (Supplementary Data Table S1) were determined by measuring genome size using a high-throughput flow cytometric protocol developed in our laboratory, whereby 96 samples can be prepared and analysed within 3 h (see Aliyu et al., 2010). The protocol employs a two-step procedure consisting of: (1) lysis of nuclei into isolation buffer [0·1 m citric acid monohydrate and 0·5 % (v/v) Tween-20 dissolved in H₂O and adjusted to pH 2·5]; and (2) staining of the filtrate in staining buffer [0·4 m Na₂HPO₄·12H₂O dissolved in H₂O plus 4 µg mL⁻¹ 4′,6-diamidino-2-phenylindole (DAPI)] (see Otto, 1990; Dolezel and Gohde, 1995). Approximately 50 mg of fresh leaf sample per plant was homogenized in each well (PP-Masterblock 128·0/85MM, 1·0 mL 96-well plate, Greiner bio-one, www.gbo.com) containing a 3·22 mm stainless steel bead and 400 µl of isolation buffer using a Geno-Grinder 2000 (SPEX Certi-Prep). The filtrate was stained with a staining buffer in a 4:1 (staining buffer:filtrate; v/v) ratio, and immediately analysed on a Partec PAII flow Cytometer (Partec GmbH, Münster, Germany) connected to a temperature-regulated (4 °C) Robby-Well auto-sampler. To reduce sample degradation over the 2 h period required for the analysis of 96 samples, the sample plate was covered with aluminium sealing film (Nunc GmbH & Co. KG, Langenselbold, Germany). Leaf samples from a tetraploid sexual (4C) Hypericum perforatum (Matzk et al., 2003) were prepared in parallel and included as external reference at well positions 1 and 96 on each sample plate in order to correct for the histogram peak shifts over the analysis period.

Flow cytometric seed screen (FCSS) analysis of seed production

Single seeds were homogenized in each well of a 96-well deepwell plate (Masterblock 96-well plate) containing three stainless steel balls (2·3 mm diameter) and 80 µL of isolation buffer (0·1 m citric acid monohydrate, 0·5 % Tween-20, pH adjusted to 2–3 and β-mercaptoethanol) using a Geno-Grinder 2000 (SPEX Certi-Prep) at 50 strokes min⁻¹ for 2 min. An additional 250 µL of isolation buffer was added to the ground seed suspensions and the mixture was then filtered through a 30 µm nylon membrane. An 80 µL aliquot of the filtrate was stained in 80 µl of staining buffer (0·4 m Na₂HPO₄·12 H₂O, 2 ml of DAPI solution, pH adjusted to 8·5; Otto, 1990; Dolezel and Gohde, 1995). The fluorescence intensity of DAPI-stained nuclei was determined using a Ploidy Analyser PA II flow cytometer (Partec GmbH, Münster, Germany), and relative embryo and endosperm ploidy per seed were compared with a known H. perforatum sexual tetraploid used as internal reference standard using the Flomax software (Partec GmbH).

DNA extraction and microsatellite analysis

A total of 371 accessions of H. perforatum collected across 36 localities from Europe and North America were sampled for this study (Supplementary Data Table S1).

Approximately 200 mg of fresh leaf tissue was collected per plant into individual wells of a 96-well deepwell plate and the plates were then flash-frozen in liquid nitrogen and ground to a fine powder using a Geno-grinder 2000 (Spex CertiPrep). A two-step protocol was used for obtaining genomic DNA free from any contaminants. First, DNA was isolated using a NucleoSpin 96 Plant extraction kit (Macherey-Nagel) according to the manufacturer's protocol. This was followed by a second DNA purification step, whereby the first DNA extractions were subsequently purified using an Agencourt Chloropure kit (Beckman Coulter) following the standard plant extraction protocol.

We used a three-primer PCR amplification protocol to analyse 19 microsatellite loci (Supplementary Data Table S2), whereby each locus-specific forward primer was designed with an added 5' M13 sequence (5'TGTAAAACGACGGCCAGT3'), and a homologous- M13-labelled WellRED (Beckman Coulter) primer was added to each PCR prior to fragment size analyses (Missiaggia and Grattapaglia, 2006). Reactions of 10 µL were carried out using a mixture containing 15–30 ng of DNA, 0·2 µm M13 tailed (forward) primer, 0·4 µm primer (reverse), 0·36 µm WellRED M13 primer (the same M13 primer labelled in three WellRED colours), 25 mm each dNTP, 1·5 mm MgCl₂, 1× reaction buffer (standard MgC1₂ free) and 0·25 U of Taq polymerase. The PCR was performed using an initial denaturing step of 10 min at 94 °C followed by 35 cycles (94 °C for 30 s, 60 °C for 10 s and 72 °C for 20 s) and 15 min at 72 °C. The PCR products of multiple loci (labelled in different colours) were mixed (Table 1) and purified using an Agencourt AMPure kit (Beckman Coulter) on a Biomek 3000 pipetting robot, and fragment analysis was performed on a CEQ 8000 (Beckman Coulter).

Table 1.

Genetic diversity comparisons of 19 microsatellite loci for 372 European and North AmericanH. perforatum

Locus	na	HO	HS	FST	G'ST (Nei)
HP04	37	0·648	0·786	0·150	0·154
Hperf04	13	0·636	0·694	0·088	0·09
Hperf13	14	0·285	0·623	0·236	0·241
Hperf1	18	0·76	0·838	0·030	0·031
HP103	24	0·583	0·764	0·147	0·151
HP11	28	0·549	0·831	0·013	0·014
HP130	42	0·725	0·888	0·044	0·045
HP74	34	0·293	0·611	0·340	0·347
HP154	42	0·771	0·858	0·086	0·089
HP302	15	0·458	0·64	0·231	0·238
Hperf05	34	0·613	0·676	0·184	0·188
Hperf30	28	0·52	0·692	0·237	0·245
Hperf72	32	0·506	0·661	0·240	0·247
HP132	35	0·706	0·849	0·081	0·084
Hperf02	41	0·706	0·789	0	0
Hperf08	56	0·488	0·726	0·246	0·252
Hperf26	36	0·551	0·735	0·178	0·183
Hperf73	30	0·29	0·571	0·360	0·367
Hperf94	13	0·419	0·631	0·214	0·22

n_a, number of alleles; H_O, observed heterozygosity; H_S, expected heterozygosity; F_ST, fixation index; G'_ST(Nei), Nei's corrected fixation index.

Genetic data analysis

A multistep process was used to score microsatellite genotypes. Using the CEQ8000 Genetic Analyzer software (Beckman Coulter), the internal size standards of all individual runs were all compared to ensure proper size calling. Fragment (i.e. allele) size and peak area (i.e. PCR product dosage) were subsequently identified for each locus using the CEQ8000 Genetic Analyzer software, followed by manual verification of each call in order to ensure proper peak designation. The checked data were then exported into SPSS for Windows (release 11·5.0, SPSS Inc.), where fragment size distributions were generated using the complete data set for each locus. This final statistical step (i.e. binning) enables fragments which differ slightly in size (resulting from variability in genetic analyser run conditions) to be assigned to specific size groups, each of which represents a particular allele. The data for all loci were finally concatenated into a multilocus genotype for each individual. All samples were processed and analysed blindly, and in random order. Following the MAC-PR (microsatellite DNA allele counting-peak ratios) method, updated for polyploidy plants with unknown pedigrees (Esselink et al., 2004), we estimated the ratio of the peak area between all alleles of each locus to estimate the number of copies of each allele.

Pairwise genetic identity between individuals was calculated based upon the multilocus genotypes using the POPDIST program (Guldbrandtsen et al., 2000) and the algorithm of Tomiuk and Loeschcke (1991) which enables the comparison of genotypes of different ploidy. The pairwise genetic identity matrix generated from POPDIST was then imported into SPSS 14 for a principal component analysis (PCA).

The GENOTYPE and GENODIVE software packages (Meirmans and Van Tienderen, 2004) were used to assign genotypes to particular clonal lineages and to calculate various population genetics statistics. GenoDive was used to estimate G_ST (fixation index for multiallelic markers) and G_ST = (H_T – H_S)/H_T (Meirmans and Hedrick, 2011). Genodive was used to generate a pairwise differentiation matrix with the values of F_ST for all populations.

Individuals were assigned to genetic clusters using the model-based clustering algorithm implemented in the software STRUCTURE version 2·3.3 (Pritchard et al., 2000), which identifies related individuals and gene pools based on allele distribution (Koch et al., 2013). Genetic population structure was calculated separately for different ploidy classes (tetraploids and hexaploids), using all microsatellites (n = 19) and the admixture model. The runs were carried out with 10 000 burn-in iterations for tetraploids (100 000 for hexaploids) and 100 000 data collection iterations (200 000 for hexaploids), assuming correlated allele frequencies (Falush et al., 2003). To determine the number of possible genetic clusters (K), ten runs were carried out for each data set with a range of K explored from 1 to 20, and the value of K was then estimated using ‘delta’K (Evanno et al., 2005) with StructureSum 2·2-R (Ehrich, 2006). Since it is not always possible to determine the true value of K, we followed the recommendation of Hubisz et al. (2009) and took the lowest estimated value which gives the most reliable and conservative estimate of population structure. For visualization, both STRUCTURE analyses were merged into one figure. The corresponding colour codes (blue, red or green) were chosen using the cluster-colour permutation with the highest matching score for all individuals to best match the cluster colours of both ploidy levels.

Mantel tests to compare the relationships between genetic differentiation (F_ST; Weir and Cockerham, 1984) and geographic distances were performed on the data for Europe and North America separately, using the default parameters of the Isolation by Distance Web Service (Version 3·23; Jensen et al., 2005).

RESULTS

The ancestral range of H. perforatum is characterized by the highest genetic variability

In the analysis of leaf tissue from 2019 individuals, 93·5 % (n = 1886) were found to be tetraploid, with lower frequencies of hexaploid (6·3 %, n = 127), diploid (1 %, n = 2) and triploid (0·5 %, n = 1) individuals (Supplementary Data Table S1; note that some populations were not used due to sample dropout).

Of the 19 microsatellites (Supplementary Data Table S2) which were used to analyse the 372 H. perforatum individuals, 14 were highly polymorphic and showed >20 alleles (Table 1). A total of 506 alleles were identified from 17 populations in Europe, while 403 alleles were found in 18 North American populations (Tables 2, 3). For all populations, the number of alleles per locus ranged from 13 to 56, while it ranged from 12 to 49 and from 8 to 38 for the European and North American populations, respectively (Tables 2, 3). European populations showed higher allelic richness compared with those from North America, except for the loci HP130 and Hperf72 (Tables 2–4). Average within-population heterozygosity (H_S) ranged from 0·563 to 0·886 for European individuals, and for North American individuals H_S ranged from 0·506 to 0·900 (Tables 2, 3). North American populations nonetheless showed higher values of observed heterozygosity (H_O) in the following loci HP04, Hperf1, HP103, HP11, HP130, HP302, Hperf05, Hperf02, Hperf73 and Hperf94 (Tables 2, 3). North American populations showed lower levels of inbreeding, as is evidenced by G_IS values ranging from –0·004 to 0·335, compared with from 0·016 to 0·419 for European populations (Table 4).

Table 2.

Genetic diversity comparisons per locus for different geographic groupings of 219 individuals of EuropeanH. perforatum

Locus	Num	Eff_num	HO	HS	HT	GIS	FST	G'ST(Nei)
HP04	26	2·662	0·634	0·723	0·899	0·122	0·201	0·211
Hperf04	12	2·413	0·608	0·677	0·763	0·103	0·105	0·111
Hperf13	14	2·092	0·362	0·613	0·784	0·409	0·211	0·222
Hperf1	16	3·175	0·690	0·794	0·872	0·131	0·093	0·098
HP103	21	2·516	0·554	0·701	0·852	0·209	0·162	0·171
HP11	23	2·505	0·532	0·700	0·870	0·240	0·206	0·216
HP130	31	3·136	0·664	0·790	0·910	0·160	0·132	0·139
HP74	31	2·144	0·388	0·625	0·892	0·380	0·306	0·319
HP154	38	4·259	0·787	0·886	0·941	0·111	0·051	0·054
HP302	14	2·350	0·495	0·669	0·838	0·261	0·199	0·209
Hperf05	29	2·202	0·525	0·633	0·762	0·171	0·169	0·177
Hperf30	28	3·297	0·742	0·805	0·888	0·078	0·098	0·104
Hperf72	25	2·574	0·588	0·710	0·810	0·171	0·112	0·119
HP132	34	3·490	0·743	0·825	0·922	0·099	0·106	0·111
Hperf02	38	2·787	0·628	0·744	0·775	0·155	0·034	0·036
Hperf08	49	3·134	0·571	0·794	0·967	0·281	0·176	0·185
Hperf26	36	2·921	0·523	0·769	0·928	0·319	0·168	0·177
Hperf73	29	1·929	0·296	0·567	0·912	0·478	0·379	0·393
Hperf94	12	1·928	0·371	0·563	0·804	0·341	0·295	0·308
Overall	26·632	2·711	0·563	0·715	0·863	0·212	0·168	0·176

Num, number of alleles; Eff_num, effective number of alleles; H_O, observed heterozygosity; H_S, heterozygosity within populations; H_T, total heterozygosity; G_IS, inbreeding coefficient; F_ST, fixation index; G'_ST(Nei), Nei's corrected fixation index.

Table 3.

Genetic diversity comparisons per locus for different geographic groupings of 150 individuals of North AmericanH. perforatum

Locus	Num	Eff_num	HO	HS	HT	GIS	FST	G'ST(Nei)
HP04	24	2·678	0·653	0·762	0·912	0·144	0·164	0·172
Hperf04	8	2·353	0·601	0·699	0·746	0·141	0·063	0·066
Hperf13	9	1·917	0·317	0·594	0·832	0·467	0·286	0·298
Hperf1	14	3·438	0·800	0·859	0·857	0·068	–0·002	–0·002
HP103	19	2·587	0·675	0·744	0·901	0·092	0·174	0·183
HP11	21	2·572	0·639	0·743	0·796	0·140	0·066	0·069
HP130	38	3·861	0·790	0·900	0·928	0·122	0·030	0·031
HP74	28	1·781	0·310	0·543	0·910	0·429	0·403	0·417
HP154	32	2·938	0·734	0·800	0·923	0·082	0·133	0·141
HP302	11	2·586	0·662	0·745	0·816	0·112	0·087	0·092
Hperf05	24	2·531	0·699	0·731	0·851	0·044	0·141	0·147
Hperf30	25	2·272	0·565	0·683	0·868	0·172	0·213	0·224
Hperf72	26	2·029	0·522	0·617	0·880	0·154	0·298	0·31
HP132	24	2·801	0·689	0·781	0·908	0·118	0·139	0·147
Hperf02	22	2·951	0·774	0·799	0·814	0·031	0·018	0·02
Hperf08	35	2·278	0·504	0·688	0·931	0·268	0·261	0·272
Hperf26	16	2·087	0·577	0·632	0·822	0·087	0·231	0·242
Hperf73	15	1·699	0·357	0·506	0·868	0·295	0·417	0·431
Hperf94	12	2·097	0·526	0·638	0·812	0·176	0·214	0·224
Overall	21·211	2·498	0·600	0·709	0·862	0·154	0·175	0·183

Num, number of alleles; Eff_num, effective number of alleles; H_O, observed heterozygosity; H_S, heterozygosity within populations; H_T, total heterozygosity; G_IS, inbreeding coefficient; F_ST, fixation index; G'_ST(Nei), Nei's corrected fixation index.

Table 4.

Genetic diversity comparisons for different geographic groupings ofH. perforatum

Population ID (no.)	Num*	HO	HS	GIS
Tuscola IL, USA (1)	5·105	0·615	0·651	0·056
ESGR MI, USA (2)	2·632	0·746	nan	nan
Green Lake WI, USA (3)	3·158	0·532	0·529	–0·004
Point Beach WI, USA (4)	7·316	0·648	0·733	0·116
Kewaunee MI, USA (5)	3·158	0·526	0·513	–0·025
Gillett WI, USA (6)	4·158	0·576	0·568	–0·014
Afton MN, USA (7)	3·053	0·667	0·952	0·3
Rideau River ON, Canada (8)	8·158	0·633	0·772	0·179
Menominee MI, USA (9)	4·526	0·63	0·776	0·188
Wausaukee WI, USA (10)	5·053	0·623	0·698	0·108
Carney MI, USA (11)	6·421	0·573	0·709	0·193
Iron Mountain MI, USA (12)	5·105	0·628	0·649	0·033
Tecumseh MI, USA (13)	4·737	0·627	0·758	0·173
Cazadero CA, USA (14)	2·632	0·433	0·546	0·207
Bass Lake CA, USA (15)	3·579	0·515	0·58	0·113
Sierra Field Station CA, USA (16)	3	0·58	0·871	0·335
Winchester OR, USA (17)	4·947	0·601	0·693	0·132
Corvallis OR, USA (18)	7·474	0·688	0·78	0·117
Granera, Spain (19)	6·579	0·521	0·633	0·177
Centellas, Spain (20)	11·526	0·623	0·762	0·182
La Selva de Mar, Spain (21)	6·053	0·571	0·72	0·207
Bolzano, Italy (22)	3·579	0·494	0·502	0·016
Badia, Polesine Italy (23)	6·684	0·551	0·674	0·184
Assas, France (24)	4·579	0·658	0·893	0·263
Clapier, France (25)	7·737	0·529	0·732	0·277
Cerbere, France (26)	6·421	0·544	0·723	0·249
Adliswil, Switzerland (27)	6·158	0·493	0·596	0·172
Arlesheim, Switzerland (28)	3·368	0·504	0·868	0·419
DeSteeg, The Netherlands (29)	7·105	0·625	0·763	0·181
Velp, The Netherlands (30)	4·684	0·605	0·711	0·15
Silwood, UK (31)	4·000	0·586	0·632	0·074
Praha, Czech Republic (32)	4·842	0·649	0·662	0·019
Bonn, Germany (33)	5·632	0·479	0·642	0·254
Hamburg, Germany (34)	6·526	0·611	0·753	0·188
Suu Ravine, Kyrgizstan (35)	3·632	0·566	1·11	0·49
Neuves Maisons, France (36)	2·053	0·561	nan	nan

Num, number of alleles; H_O, observed heterozygosity; H_S, heterozygosity within populations; G_IS, inbreeding coefficient; nan means that a correction for bias due to sample size could not be performed for that sample.

The population number is also given with the population ID as shown in Figs 1 and 2.

*The number of alleles is averaged over populations and loci.

A PCA using the genetic identity values (Tomiuk and Loeschcke, 1991) calculated for tetraploids and hexaploids from both Europe and North America demonstrates no apparent differentiation between the continents (Supplementary Data Fig. S1). Mantel test correlations between geographic and genetic distances was significantly positive for samples collected in Europe (r = 0·319, P < 0·001) but not significant for samples from North America (r = 0·091, P = 0·248; Supplementary Data Fig. S2).

The structure analysis using Evanno's deltaK revealed an optimal K = 3 for both ploidy levels (Supplementary Data Fig. S3). The best-fitting adjustment of the colour code of the recognized genetic clusters from the hexaploid to the tetraploid genetic clusters shows a congruence of about 70 %, while any other colour code adjustment yielded congruence <50 %. More importantly, in any mixed population with tetraploids and hexaploids and only a single genetic cluster (‘colour code’), respectively, the congruence was of about 100 %. The results also indicate that almost any population showed admixture of the three different genetic clusters (Fig. 1). If genetic clusters are superimposed on the results from the respective PCA, a high degree of congruence is obviously seen, also indicating that the three genetic clusters are largely independent from ploidy-level variation (Supplementary Data Fig. S1).

Fig. 1.

Bayesian cluster analysis using STRUCTURE v. 2·3.3 (Pritchard et al., 2000). Tetraploid and hexaploid individuals were calculated separately in independent analyses. Genetic assignments to three genetic clusters (red, green and blue) indicated by bars (STRUCTURE output) were converted into pie charts for better comparisons of ploidy levels within and between populations. Assignment of colour codes comparing tetraploid and hexaploids was done optimizing the overlap of the two complete data sets.

For the most part, the STRUCTURE analysis revealed no significant difference between hexaploid and tetraploid individuals within and between populations. Nevertheless, a number of populations from Europe were characterized by genetically distinct tetraploid and hexaploid forms (e.g. populations 23, 32 and 34; Fig. 1). The three genetic clusters show evidence for geographic structure in Europe, as is further supported by the Mantel test (see above; Supplementary Data Fig. S2). For tetraploids, the green genetic cluster shows a central to south-western range; the red genetic cluster is found mostly in the north (Fig. 2). The European distribution is similar for hexaploids, although the red genetic cluster extends further eastward (Fig. 2). The blue genetic cluster is present in various tetraploids and hexaploids, with a tendency to be found more frequently in the East. No apparent structuring of the genetic clusters is evident in North America, and the Great Lakes region appears to be a centre of genetic diversity, having population representatives of all three European genetic clusters, on both the tetraploid and hexaploid levels (Fig. 2).

Fig. 2.

Geographic structuring of genetic diversity among populations of tetraploid and hexaploid Hypericum perforatum in native European and introduced North American populations. Microsatellite analysis: Program STRUCTURE, assignment scores for K = 3 clusters; the numbers indicate the population code shown in Supplementary Data Table S1.

Tetraploids are apomictic, while secondarily derived hexaploids are sexual

Twenty-four mature single seeds per individual and five individuals per population were analysed for a sub-set of the 372 H. perforatum which were genotyped using microsatellites (total 6695 analysed seeds; Table 5). Polyembryony was found in 141 (2 %) and 3 (<1 %) of all measured seeds in tetraploids and hexaploids, respectively (data not shown).

Table 5.

Overview of seed formation in tetraploid and hexaploid H. perforatum from Europe and North America, based upon an analysis of 6695 seeds

Origin	Mother plant ploidy	Embryo ploidy	Endosperm ploidy (%)
			4	5	6	7	8	9	10	11	12	13	14	15	16	18
Europe	Tetraploid	2	0·4	0·1	1·3	0·1	0·1	0·1	0·3				0·1
		3		0·1	0·1
		4			25·5	0·2	1·5	1·8	46·2	8·1	0·8		1·6
		5				0·3	0·1	0·1	0·1
		6					0·5	0·2	9·1	1	0·1
		8									0·1
Europe	Hexaploid	2			0·8
		3		0·3	1	0·5	9·2	1·3	0·3
		4			7·9	1·5	0·3		0·3
		5			0·5		34·7	0·3
		6					0·5	11·2	4·1		3·3	1·5	13·8	3·6	1·8	0·8
		8									0·3		0·5
North America	Tetraploid	2	0·1	0·1	1		0·3	0·1	0·5
		3		0·1			0·1
		4			17	0·2	1·5	2·2	62·4	4·5	0·6	0·1	1·4
		5				0·1		0·1		0·1	0·1
		6					0·1	0·1	7·3	0·2			0·1	0·1
North America	Hexaploid	2					0·7
		3		1		0·3	10·3	1·7	0·7		0·3
		4			1·4	2·4		0·3	0·3
		5				0·3	40·4	0·7	0·3
		6						13·7	2·7	0·7	2·7	2	9·9	5·1	0·7
		8									0·3		0·7

Percentage values are calculated based upon the total number of measured seeds for tetraploids and hexaploids from Europe and North America independently. Numbers in bold represent frequencies >5 %.

For the most part, tetraploids from Europe produced seeds characterized by 4x embryos with 6x, 10x or 11x endosperm, or 6x embryos with 10x endosperm (Table 5). Hexaploids from Europe produced seeds with 3x embryos and 8x endosperm, 4x embryos with 6x endosperm, 5x embryos with 8x endosperm, and 6x embryos with 9x or 14x endosperm (Table 5). Tetraploids from North America produced seeds with 4x embryos and 6x or 10x endosperm, and 6x embryos with 10x endosperm (Table 5). Hexaploids from North America produced seeds with 3x embryos and 8x endosperm, 5x embryos with 8x endosperm, and 6x embryos with 9x, 14x or 15x endosperm (Table 5). Taken together, these data show that tetraploids and hexaploids have the ability to produce both meiotically reduced and unreduced pollen.

For all measured seeds together, the frequency of meiotically reduced vs. unreduced female gametes showed opposing patterns, being 24 and 73 %, respectively, in tetraploids and 73 and 23 %, respectively, in hexaploids (Table 6). Embryo development also differed between ploidy types, with the highest frequency of seeds produced by parthenogenetic development of the unreduced female gamete in tetraploids (63 %), while fertilization of the reduced female gamete occurred mostly in hexaploids (59 %; Table 6). An embryonic ploidy increase via fertilization of the unreduced female gamete was higher (10 %) in tetraploids compared with hexaploids (2 %; Table 6).

Table 6.

Overview of seed formation in tetraploid and hexaploid H. perforatum from Europe and North America

Mother plant	No. of seeds	Female gamete (%)		Embryo development (%)
Tetraploid	5862	Reduced	1411 (24·0)	Fertilization	1200 (20·5)
				Parthenogenesis	131 (2·2)
		Unreduced	4314 (73·6)	Fertilization	586 (10·0)
				Parthenogenesis	3717 (63·4)
		No peak	137 (2·3)
Hexaploid	833	Reduced	608 (73·0)	Fertilization	490 (58·8)
				Parthenogenesis	119 (14·3)
		Unreduced	188 (22·6)	Fertilization	17 (2·0)
				Parthenogenesis	171 (20·5)
		No peak	37 (4·4)

Percentage (%) values in parentheses are calculated based on the total number of measured seeds for tetraploids and hexaploids independently, and ‘No peak’ refers to the number of analysed seeds which did not yield any peak information.

Variation for various seed formation components measured using flow cytometry demonstrated no obvious geographic pattern. Apomeiosis, parthenogenesis and polyembryony levels were all higher across tetraploids compared with hexaploids, regardless of continent of origin (Fig. 3A, B, D). For the same three traits, outlier and extreme values for tetraploids overlapped with the ranges of hexaploid individuals (Fig. 3A, B, D). Fertilization of reduced or unreduced female gametes was consistently higher in hexaploids compared with tetraploids (Fig. 3C). Finally, the range of variation for apomeiosis was consistently greater in hexaploids compared with tetraploids (Fig. 3A).

Fig. 3.

Variability of frequency of apomeiosis (A), parthenogenesis (B), fertilization (C) and polyembryony (D) for all tetraploids and hexaploids (as indicated in the key) for North America and Europe.

DISCUSSION

Apomixis has repeatedly evolved from sexual ancestors in a number of taxa (Grossniklaus, 2001), and the switch from sexual to asexual reproduction is characterized by concomitant effects on fitness, competitive ability (e.g. invasiveness), gene flow and population structure. In addition, facultative apomixis, which may involve quantitative variation for sexual and asexual seed production (Aliyu et al., 2010), represents a trait upon which natural selection can act. Hence, elucidating the relative advantages of apomixis vs. sex on a population level requires an understanding of variation at both the genetic and reproductive levels. Here we have analysed genetic and quantitative reproductive variation in 36 populations of H. perforatum from both their centre of origin (Europe) and the continent which they have recently invaded (North America; Sampson and Parker, 1930), in order to shed light upon possible links between introgression, apomixis and invasiveness.

Population sub-division and glacial refugia in Europe

The microsatellite data presented here demonstrate that European H. perforatum individuals can be conservatively assigned to three distinct genetic clusters which provide evidence of geographic structure for both tetraploid and hexaploid individuals (Figs 1, 2; see Mantel test; Supplementary Data Fig. S2). The distribution of individuals from the genetic clusters across known glacial refugia (e.g. Iberia and Italy; Fig. 2) is suggestive of geographic isolation followed by genetic differentiation, as has been shown for other plant taxa (Sharbel et al., 2000; Medail and Diadema, 2009). These data are consistent with those of Koch et al. (2013), which similarly calculated a small K-value based on genome-wide AFLP scans. In that study, for H. perforatum two major gene pools were detected which could be followed back to the respective diploid progenitors. In addition, a third gene pool was characterized as having its primary origin from H. maculatum, with subsequent ongoing intogression into H. perforatum (Koch et al., 2013). However, because the marker systems differed (AFLPs and cpDNA vs. microsatellites herein) between this study and that of Koch et al. (2012), a direct comparison of gene pools and genetic clusters is not possible. Nonetheless, the general trends of both data sets imply that the three gene pools and genetic clusters do correspond to one another to a large extent. For example, the two populations from Spain (population numbers 19 and 20; Supplementary Data Table S1) were analysed in both studies, and AFLP and cpDNA data are suggestive of introgression from the H. maculatum gene pool into both (Koch et al., 2013).

Although the microsatellite data do not indicate such a strong footprint of introgression, the data here similarly identified some introgression of a ‘blue’ genetic cluster in both populations, a result which is also consistent with the hypothesis that H. perforatum is widely affected all over Europe by introgression from H. maculatum, and that morphology is not a consistent indicator for reticulate evolution in this group of species (Koch et al., 2013). Consequently the microsatellite-based ‘blue’ genetic cluster probably corresponds to H. maculatum's genetic variation introgressed over time into polyploid H. perforatum, which would be fully consistent with AFLP and cpDNA data (Koch et al., 2013).

Multiple introductions in North America from different genetic backgrounds

Hypericum perforatum has its centre of origin around south-eastern Europe (Koch et al., 2013), and is an invasive species in North America after being introduced in the 18th century from Europe (Sampson and Parker, 1930). The analyses presented here demonstrate that North American H. perforatum originated from multiple introductions from all three European gene pools (Fig. 2). The lack of significant correlation between geographic and genetic distances in North American samples (see Mantel tests; Supplementary Data Fig. S2) is furthermore consistent with the multiple introduction hypothesis. Both tetra- and hexaploid individuals were found in almost all populations from Europe and North America (Figs 1, 2). In most cases, both ploidy types from the same population shared similar genotypes, but there were also populations in which the hexaploids were clearly of genetic origin different from that of their sympatric tetraploids (Fig. 1).

Introgression, as evidenced by the analysis of individuals with respect to the three identified genetic clusters (i.e. individuals with two or three colours), was more evident in tetraploids than in hexaploids (Fig. 1). Populations showing signs of introgression in Europe reflected potential contact zones between differentiated gene pools in both the north (e.g. populations 29 and 30) and south (e.g. populations 27 and 28; Figs 1, 2; Supplementary Data Table S1). In North America, introgressed populations were in a contact zone of all three genetic clusters around the Great Lakes in the north (e.g. populations 7–11 and 13) as well as in a zone between two genetic clusters in the west (Figs 1, 2). Because of the apparent association between contact zones and introgression, we hypothesize that secondary contact of introduced European gene pools, followed by gene flow between them, best explains the pattern in North America (Figs 1, 2). This is supported, furthermore, by the fact that North American populations showed higher levels of H_O at almost all loci (Tables 2, 3). Alternatively, the introduction of individuals which had already undergone introgression in Europe could explain this observation, although this is unlikely considering the fact that the genotypes of European introgressed individuals are different from those in North America (Fig. 1).

Due to factors including higher linkage disequilibrium, selective sweeps and background selection, asexual genetic variation is expected to decrease over time (Barrett, 2011). The data here demonstrate a higher percentage and relatively high genetic variability of apomicts in introduced populations (Table 4), an observation which is consistent with low levels of sexuality playing a role in the establishment of newly introduced populations. In inbreeders or asexual populations, F_ST values are expected to be significantly high and should reflect pronounced population sub-division (Barrett, 2011). The relatively low values characteristic of many samples here (Table 3) is suggestive of gene flow between populations, and provides additional support for the important role of residual sexuality in the facultative apomictic H. perforatum.

Reproductive patterns of different ploidy types is reflective of dosage of an apomixis factor

Apomixis in H. perforatum is aposporic and of the Hieracium type, whereby sexual embryo sac development is often terminated at the MMC or reduced megaspore stage, followed by one or more somatic AI cells differentiating from nucellar tissue to initiate unreduced embryo sacs through mitosis (Mártonfi et al., 1996; Galla et al., 2010). Endosperm formation can occur either via pseudogamy (i.e. fertilization of the central cell) or autonomously, and the variable modes of seed formation measured here (Tables 5, 6) are comparable with what has previously been found (Matzk et al., 2001).

In tetraploids, the most frequent class of seeds produced had a 4:10 embryo:endosperm genome ratio, and hence were derived from an aposporous 4x ovule whose central cell was fertilized by a 2x pollen cell (Table 6). Tetraploid apomictic H. perforatum thus have the capability to produce normal meiotically reduced pollen. Furthermore, the high frequency of tetraploids vs. hexaploids, in conjunction with the analyses of embryo to endosperm ploidy in seeds (e.g. mostly 10C endosperm in tetraploids, and 8C endosperm in hexaploids), demonstrates that hexaploids are derived from tetraploids via fertilization of an apomeiotically derived 4x egg cell with meiotically reduced 2x pollen. Considering that they are self-compatible (Matzk et al., 2001), both selfed and outcrossed progeny are possible, both types of which are reflected in the genotypic variation presented here (Fig. 1). The comparative genotype data of sympatric tetra- and hexaploid individuals (Fig. 1) are thus indicative of different origins of hexaploids, including: (1) generation of hexaploids from genotypes of the same population (i.e. identical ‘pure’ genotypes); (2) introduction of genotypically different tetra- and hexaploids into the same geographic area (i.e. a different ‘pure’ genotype characterizing each ploidy type); and (3) selection for particular genotypic combinations during hexaploidization.

Considering that hexaploid H. perforatum are secondarily derived from tetraploid apomictic H. perforatum, it is highly interesting that hexaploids were characterized by significantly higher levels of meiotically reduced MMC formation (i.e. sex) than were tetraploids (apomictic; Tables 5, 6). Hence, apomixis is highly penetrant in the tetraploid condition, but high levels of sexuality can apparently be ‘reconstituted’ in hexaploids. Furthermore, no differences between the various components of apomictic seed production and the origin of plants in Europe (ancestral range) and North America (invasive range) could be found (Fig. 3).

Although apomixis is hypothetically advantageous during colonization and invasiveness (Lavergne and Molofsky, 2007), no evidence for a shift in reproductive patterns with respect to colonization could be found for H. perforatum. Hybridization is hypothesized to be the trigger for apomixis (Carman, 1997; Grimanelli et al., 2001; Richards, 2003; Kearney, 2005; Paun et al., 2006), and gene flow between the differentiated gene pools of H. perforatum identified here, in addition to closely related H. maculatum (Koch et al., 2013), is consistent with the hybridization–apomixis correlation. The data presented here are additionally suggestive of a dosage effect whereby the number of copies of some factor, irrespective of genetic origin (Figs 1 and 2), is associated with the switch from sex to asexual reproduction.

Conclusions

To summarize, we show here that European Hypericum perforatum can be sub-divided into three distinct genetic clusters whose geographic distribution is reflective of glacial and post-glacial history and reticulate evolution, and that representatives from all three genetic clusters have taken part in the colonization of North America. Importantly, the data presented here imply that neither genotype, ploidy nor quantitative variation for sexual and apomictic seed formation is associated with the colonization of North America (and potentially invasiveness). A complex pattern of reticulation and hybridization between different genetic clusters on both continents has resulted from facultative apomixis, whereby residual gene flow through meiotically unreduced gametes has led to the repeated formation of hexaploids from tetraploid ancestors. Interestingly, the derivation of highly sexual hexaploids from highly apomictic tetraploids points to gene dosage in addition to hybridization as the underlying factors leading to apomictic seed formation. High plasticity in ploidy and reproductive system in conjunction with ancestral genetic variation are thus probably intertwined factors leading to the long-term success and adaptive potential of H. perforatum.

SUPPLEMENTARY DATA

Supplementary data are available online at www.aob.oxfordjournals.org and consist of the following. Table S1: accession data, sample sizes and geographical co-ordinates. Table S2: microsatellite loci and their details. Figure S1: principal component analysis of genetic data and superimposed information from ploidy level, genetic cluster analysis, origin and reproductive mode. Figure S2: scatterplots of genetic distance vs. geographic distance and reduced major axis (RMA) regressions for Hypericum perforatum populations. Figure S3: searching for the optimal K using Structure-Sum and Evanno's deltaK for the subsequent analysis of the microsatellite data.