A total of 436 flowers from 145 specimens were crossed in our intra- and interpopulation crossing experiments with a median number of five flowers per specimen (minimum 2, maximum 18). Forty-one flowers from eight specimens of the Pliezhausen (Pl) population did not produce any capsules after intra-population crossings.

Populations were assigned to crossing groups when crossings between populations did not produce seeds (dashed lines; Figure 2) but did so when crossed with populations from other crossing groups (solid line; Figure 2). By this method, 19 populations were assigned to five different crossing groups represented by different colors (with non-capitalized color names) in Figure 2.

The largest crossing group consisted of thirteen populations from Baden-Württemberg to Bavaria (green colour, Figure 2). Three populations from the most western part of our transect (Tübingen private garden (Tp), Tübingen old botanical garden (Tb), and Pl) formed a second crossing group (blue color, Figure 2). The other three populations formed crossing groups of their own: Tübingen Lustnau (Tl, red, Figure 2), Münsingen (Mü, pink, Figure 2) and Mehrstetten (Mt, violet, Figure 2). The Großaitingen (Gß) population could not be grouped since individuals never produced flowers in the greenhouse and no flower was observed in the field during the two collecting trips.

Among the ten investigated populations of V. filiformis in Germany (eight from the transect and one each in Bonn and Mainz; Figure 3), population Il had the highest percentage of mutant flowers (47 in 103 observed flowers; 45.63%), including 36 flowers that lacked filaments (Additional file 3C2), nine that had only one stamen (Additional file 3C3), one flower that had two stamens of different sizes (Additional file 3C1), and the unique case observed of one flower that had three stamens (Additional file 3C5). The Mt population had the lowest percentage of mutant flowers with 5.56% from a total of 54 counted flowers (Figure 3). Among the seven native populations investigated, only one had androecium-mutant flowers at low frequency (1.48%; Kz8). The populations of native V. chamaedrys (1.85%; N_flower = 379) or introduced V. persica (0.20%; N_flower = 507) from Mainz present a markedly lower percentage of mutants than the neighboring population of introduced V. filiformis (15.19%; N_flower = 953) and any other introduced populations but not the native populations. Significant differences exist between the native and introduced populations (p=0.01).

Number of pollen grains per flower differed significantly between areas (flowers in the native area had more pollen grains; p<0.001), crossing groups (p=0.038) and populations (Additional files 4, 5). Within the native area the population from Cross Pass (Kz8) produced significantly less pollen (p<0.001) than the other native populations. Ovule numbers differ significantly only between crossing groups (p<0.001; Additional files 4, 5) with green and violet crossing groups producing more ovules than the rest. Plants from the native area appear to produce less ovules, but this is not significant (p=0.130).

Capsules produced in the greenhouse after hand-pollination contained more seeds than capsules collected in the field after open-pollination (Kruskal-Wallis-test: H = 5.064; p-value = 0.024; Additional file 6). Variation in seed number per capsule was also found between the native and introduced area and within the introduced area between populations, crossing groups and genetic clusters (Additional file 6). Seeds from the introduced area were also smaller than seeds of the native area (Figure 5A; Kruskal-Wallis-test: H = 88.053; p-value < 0.001). Significant differences were found between Georgian populations for seed size (H = 169.202; p-value < 0.001; Additional file 6). In the REML analysis for seed number, area and population were the most important factors as indicated by AIC and BIC (699.66, 719.61 respectively), which means that the number of seeds per capsule is different between areas and between populations. For seed size using the mean width of seeds per capsule, the model with area and number of seeds per fruit as fixed factors and population as random factor proved to be the best (AIC −172.69; BIC −156.51). The next best model had an AIC difference of 2 and included more parameters, and was thus discarded 58 . For the introduced area alone, again, population had the strongest effect (AIC −60.99; BIC −50.61), whereas crossing group and genetic group were less important. The model was again better for the above mentioned criteria. For seed number the model with area and population was also the best (AIC 699.66; BIC 719.61).

For seed viability, a significantly lower number of dead seeds was found in the native versus the introduced plants (H = 1.84; p-value = 0.033; Figure 5B). Within the Tl-Pl-Re triangle of the introduced area, seeds produced by the widespread green crossing group and the Pink genetic cluster (Figure 3) have a significant lower viability (22% viable seeds) than the seeds produced by the two other crossing groups or genetic clusters (red crossing group and Green genetic cluster: 44% viable seeds, t = −30.98; P-value = 0.001 and blue crossing group and Blue genetic cluster : 36% viable, U = −3.24; p-value = 0.042) irrespective of the pollen donor (see below). A maternal effect was also found for seed size with plants belonging either to the blue crossing group or the Blue genetic cluster producing significantly bigger seeds than the ones from the other crossing groups or genetic clusters (data not shown). No paternal effect was found by comparing the results of seed viability and size between the same combinations ♀ - ♂: Tl-Pl/Pl-Tl; Tl-Re/Re-Tl and Pl-Re/Re-Pl (data not shown). Moreover, the seeds from the native area germinated more frequently than the ones from introduced populations during the three repeats (90% vs. 58%, s.d. 1.15) and this difference in germination rate is statistically significant (H = 3.97; P-value = 0.046; Figure 5C).

Positive correlations were observed between pollen production per flower and a) the number of polymorphic AFLP-fragments (r = 0.695; p-value = 0.019; Additional file 7C1), b) Nei’s gene diversity (Hj; r = 0.722; p-value = 0.019, data not shown) and c) private AFLP-fragments per population (r = 0.612; p-value = 0.040; data not shown). Negative correlations were found between pollen production and the number of fixed AFLP-fragment (r = − 0.793; p-value = 0.005; Additional file 7C2) and the percentage of clones per population (r = − 0.744; p-value = 0.011; Additional file 7D1). Negative correlations exist between the percentage of androecium-mutant flowers and a) pollen production per flower (r = − 0.508; p-value = 0.046; Additional file 7A and with b) number of polymorphic AFLP-fragments per population (r = − 0.579; p-value = 0.040; Additional file 7E1) and c) number of private AFLP-fragments per population (r = − 0.592; p-value = 0.036; data not shown). Conversely, ovule production is not correlated with flower mutation frequency (Additional file 7B) or with population genetic data (data not shown). It should be mentioned that the genetic diversity (Hj) is not correlated with the number of DNA-samples or specimens investigated per population (r = − 0.067; p-value = 0.380), but is instead significantly and positively correlated to the proportion of non-clonal specimens per population (N _{no clone}/N _{DNA sample}; r = 0.765; p-value < 0.001; Table 1), and negatively correlated with the proportion of clones per population (r = − 0.800; p-value < 0.001).

Ever since the first record of V. filiformis was documented growing around the roots of Georgian vine shoots near Marseille (Aubagne, southern France, 1893; 39 ), European populations of this species have been considered to be sterile by botanists because no seed set has been observed in the introduced area 39 45 48 . Bangerter & Kent 42 reported three plants with capsules out of more than 300 specimens checked, and Lehmann 39 summarized similar observations from other botanists.However, capsule formation alone is not proof of sexual reproduction. In our crossing experiments, capsules formed occasionally but did not necessarily contain seeds, an observation previously made by Lehmann 39 . In the course of our study, and despite our intensive investigations, we have neither observed capsules in the field nor in herbaria. A report of seed set in V. filiformis in Switzerland 59 has been shown by us to be due to misidentification (pers. obs.). Based on this observation and previous unsuccessful crossings by Lehmann 40 in one population from Tübingen Botanical Garden and Thaler 60 in three populations from Graz, we decided to limit our intrapopulation crossing to a single, population that contained many flowering individuals. This population likewise did not produce seeds. Subsequent investigations in populations from Oldenburg further support this view (Marszalkowska, unpubl.).

Two other aspects make sexual reproduction in the introduced populations of V. filiformis improbable. First, pollinators observed by Thaler 46 in introduced populations are small Diptera and Hymenoptera, which are territorial and rarely fly more than one meter 61 62 . Thus, long distance pollinator movement between clones with different S-alleles or between populations is unlikely. Even in the three geographically close populations (Tl-Pl-Re, minimum distance four kilometres), which were demonstrated to produce seeds in the inter-population crossing experiments (Figure 2), no seed production was found in the field. Second, the majority of introduced populations are located in public or private gardens, turfs or parks, which are mowed during the flowering period 54 . Consequently, even rarely produced fruits would be removed prior to seed maturation, which occurs at least one month after successful pollination (pers. obs.).

Although the absence of seed set after intrapopulation crossings could be explained by general sterility, the ability to set seed successfully was demonstrated by our interpopulation crossing experiments in the greenhouse. This means that the reproductive system of European V. filiformis is still functional (Figure 2). Consequently, the invasion process was not facilitated by the breakdown of the SI system or the disruption of the sexual reproduction as found in other species 10 63 64 65 . Instead, clonal reproduction either has substituted sexual reproduction, or is at least prolonging the phase until sexual reproduction can occur again, leaving the possibility of future sexual reproduction.

In the absence of sexual reproduction, V. filiformis relies on vegetative reproduction for colonization of a new area. In our study, AFLP-fingerprinting data revealed widespread clonality in the introduced area. Our genetic analyses show that clones are common in the introduced populations and indeed there are even several monoclonal populations (Figure 3 & Additional file 1), whereas no clonal individuals have been found in populations of the native area. Similar results, where clonal reproduction is restricted to introduced populations, have been found in other invasive weeds such as Rubus alceifolius 66 or Fallopia japonica 14 . The presence of clones in V. filiformis is also verified by the absence of a relationship between the number of genetically investigated individuals per population and the level of intra-population genetic diversity. However, such a correlation appears when clonal samples are deleted from the analysis (data not shown). Moreover, the significant difference between the distribution of molecular variance among and within native vs. introduced populations and the difference in the fixation index (φ_ST; Table 2) support the hypothesis of a difference in population structure, with predominant to exclusive clonal reproduction in the introduced area. Indeed, the molecular variance is distributed equally among and within the native populations (among: 25.48%/within: 74.52%) as well as among and within the genetic clusters of the introduced area (among: 26.09%/within: 73.90%) compared with that among and within the introduced populations (among: 45.47%/within: 54.53%; Table 2). Although we have identified extensive clonal reproduction, intra-population genetic diversity within introduced population is not zero (Table 2) and may be due to either i) the accumulation of somatic mutations within clones as is the case for the self-incompatible and asexually reproducing Oxalis pes-caprae 36 and the single clone of dioecious Sphagnum palustre on Hawaii 67 , or ii) the presence of several clones of multiple introductions, which either share the same S-alleles or do not produce seeds for other reasons (Figure 3). At present we are not able to distinguish between these. Comparing the genetic diversity in the populations from different parts of the native range with that in the introduced populations (Figure 4) it is obvious that a substantial amount of genetic diversity has been introduced.

It is tempting to hypothesize that our five genetic clusters identified by the assignment tests correspond to five different introductions, but the situation is certainly more complex. Our AFLP-fingerprint data suggest that two genetic clusters (Red cluster, thirteen populations between Kt and A/Pink cluster, eight populations between Tp and Gß) have been dispersed 130 km across the study area (Figure 3). These two genetic clusters could easily correspond to separate introductions, but this is likely to be an underestimate for the number of separate introductions, as separate introductions of closely related plants cannot be distinguished. The large zone covered by individuals belonging to the same genetic cluster in this non-native environment is accompanied by low frequency or absence of private fragments (Table 1). The successive introductions during the colonization of southern Germany must have produced a succession of founder effects or bottlenecks as observed in Fallopia sp. 14 or Aster furcatus 65 . However, we were not able to distinguish subsequent bottlenecks from the initial introduction bottleneck, indicated by fewer polymorphisms, lower intrapopulation genetic diversities (Table 1) and by a reduction of the frequency of rare fragments (f < 0.15; data not shown) in introduced populations compared to the native ones. Since specimens belonging to the Red, Blue and Pink genetic clusters cannot be significantly discriminated in the PCoA analysis, they are inferred to be closely related. This suggests that these genotypes came from a single region, as previously implied by Lehmann 39 based on the information on plant traders that introduced the species. In the extreme case, it is even possible that they correspond to a single introduction and fixation of different S-alleles and subsequent somatic mutations in different populations.

Several incongruencies exist between crossing and genetic data, for example: i) populations Tp and Bz are assigned to the same genetic clusters (individuals either in Blue or Pink clusters) but are inferred to contain individuals with different S-alleles (blue and green crossing groups; Figure 3); ii) individuals in Mü and Mt are also in the same genetic clusters (Red and Green clusters) but are again inferred to have different S-alleles (pink and violet crossing groups; Figure 3) and none is the green one. Such divergent S-alleles could either be introduced or evolved post-introduction. Alternatively, some post-introduction sexual reproduction could have allowed genetic exchange, especially at the S-locus, but the small population size would have led to a gradual fixation of two S-alleles in a particular population (S-Allee effect; 68 69 ).

The opposite case, genetically divergent individuals in the same crossing group, can be explained by the same S-allele being derived from a genetically differentiated introduction and fixed in different populations. One case may be the Pink and Red genetic clusters, which largely correspond to the green crossing group (Figure 3). Another example may be the population Hw, which is incompatible to other green populations although it is alone with the Tl population in the Green genetic clusters (Figure 3).

Thus, between three and five genetically distinct introductions can be inferred within the introduced area but the actual number cannot be estimated. The correlation found between geographic and genetic distances (Additional file 2) is characteristic for multiple introductions as already observed by others 30 66 . Moreover, three introduced populations (Kt, We and A), which are 120 kilometres apart from each other and have the highest genetic diversity of the German populations, are grouped together with the native population located in the Greater Caucasus (Kazbegi-Kobi) in the same genetic cluster (Turquoise cluster; Figure 3 and Table 1). This indicates a possible north-eastern Georgian origin of these German populations, as opposed to the two other native regions (north-eastern Turkey and central Georgia) as putative sources. However, detecting precisely the sources of each introduced population would require a phylogeographic analysis with a denser sampling in the native distribution area.

Since sexual reproduction no longer occurs in the introduced area, selection pressure on primary sexual floral characters has ceased compared to the native area. Thus, mutations affecting the reproductive parts fail to have a serious fitness effect and are free to accumulate 12 . This accumulation of deleterious mutations without the chance to reduce them again in organisms without meiotic recombination has been termed Muller’s ratchet 70 71 , and has been shown to lead to an increased extinction risk in asexual organisms 72 . Veronica filiformis is able to tolerate deleterious mutations affecting reproductive organs (in the androecium) better than other species since it can reproduce clonally in an effective manner. Nevertheless, mutations affecting the reproductive parts are considered deleterious since the lack of recombination prevents the reduction in mutation load. Mutations in flowers within introduced populations can be observed frequently (between 5% and 45% of flowers are affected within fifteen populations investigated; Figure 3) compared to their rare occurrence in the native range (~1.5% within one of seven populations). Similarly, they are much less frequent in other species of the genus 44 , including the self-compatible introduced Veronica persica (0.20%) and the self-incompatible native Veronica chamaedrys (1.85%) studied here, which are both able to reproduce clonally but not exclusively so 73 74 .

Our investigations indicate that the androecium-mutations in introduced populations of V. filiformis also affect pollen production. First, introduced populations produce a highly significantly lower number of pollen grains (30% less; Additional file 4). Second, the single native Georgian population possessing androecium-mutant flowers (Cross Pass Kz8) has a significant lower pollen production than the other native populations (40% less; Additional file 4). Third, a negative correlation was found between the percentage of androecium-mutant flowers counted per population and the pollen produced by these populations (r = − 0.5076). Since we assumed an origin of the German populations from the Greater Caucasus, where the single native population with androecium-mutant flowers (Kz8) was found, it is possible that the genetic basis for the androecium mutations was inherited from the source area. However, another possibility is that this population itself has a history of extensive vegetative reproduction with relaxed selection on floral reproductive characters since a partial alteration of the environment has been observed on the site (pers. obs.).

Although we observed mutations only in the male reproductive parts (stamen, pollen) in V. filiformis, female traits involved in sexual reproduction should be affected similarly and were inferred by the observed maternal effect on seed size and seed viability. Unfortunately, we were unable to bring living plants from the native area to the greenhouse and, therefore, we were only able to compare seeds produced in the greenhouse from introduced populations to those produced in the field in the native area. Differences between seed characters produced in the field versus greenhouse have commonly been observed, with seed germination rate being lower 75 and seed size being smaller in field-reared seeds 76 , likely due to better nutrient supply and less competition in the greenhouse. Thus, a seed number-seed size trade-off, for example, could be invoked to explain this difference, based on differences in ovule production between genotypes (Additional file 4) or pollen limitation in the native habitat as indicated by fewer seeds per capsule in native plants compared to those from capsules produced in our artificial crossings. However, we demonstrated that even when we controlled for seed number by estimating average seed size per capsule, seed size still differs between areas, which also explains why our observed differences in germination rate and seed size (Figure 5) are opposite to those expected (see above). Furthermore, the detected differences in seed viability between introduced populations of different crossing groups and different genetic groups on seed viability demonstrate that genetic factors influencing seed viability exist even within the introduced area. Finally, differences in seed size and germination rate parallel those observed in embryo size (data not shown) and number of dead seeds (Figure 5B; without a difference in ovule-seed ratio) between seeds of the two areas. Therefore, we consider the best explanation for these phenotypic mutations to be the genetic consequences of deleterious mutations in the genome accumulated by Muller’s ratchet, which occurred after the cessation of sexual reproduction and affects various life stages.

Similar to our observations, a relationship between pollen viability and other morphological characters and the age of the population were detected previously in other species that reproduce vegetatively 77 78 . The observed relationship between age of meadows, pollen viability and petal number in those studies suggests a similar effect of accumulation and finally fixation of mutations deleterious to sexual reproduction. Unfortunately, for most of the populations in our study, the time of colonization is not known and thus the gradual accumulation of mutations suggested by Muller’s ratchet cannot be inferred directly. However, the analysis of correlations between the proportion of androecium-mutant flowers, pollen production and genetic diversity offers some indirect evidence for the gradual accumulation of mutations. The results of correlations taken together (Additional file 7) suggest that the more clonal the populations and the lower the number of polymorphic AFLP fragments, the lower the number of flowers per individual, the lower the pollen quantity, the higher the frequency of dead seeds per capsule, and the higher the frequency of mutant flowers. Whereas the underlying common cause for these observations is not demonstrated per se, the most likely explanation is accumulation of mutations affecting various life stages and impairing sexual reproduction. Thus, our preferred scenario for the colonization of new populations of V. filiformis involves the introduction of a few but genetically divergent individuals that first lost the ability to reproduce sexually because there were not enough compatible S-alleles in the populations (Allee effect). Subsequently, the lack of sexual reproduction led to an erosion of genetic diversity and increasing importance of clonal reproduction in the populations. This erosion of genetic diversity was paralleled by the accumulation of mutations leading to a decrease in quantity and viability of pollen and seeds, as well as a reduction in the number of flowers per individual and an increase in phenotypically discernible mutants. The predicted endpoint of this seems to be reached in Großaitingen (Gß), a population apparently lacking the ability to flower altogether.

Between March and April 2007, plants from eighteen populations of Veronica filiformis were collected along a transect between Tübingen (Baden-Württemberg), Ulm (Baden-Württemberg) and Augsburg (Bavaria) with a maximum distance of 150 km (T-U-A; Figure 1; Additional file 1). The sampling was guided by the first observations in Germany based on records by Lehmann 39 and Thaler 49 in Tübingen Lustnau (1909), old cemetery of Ulm at Wiblingen (1936), and unspecified public parks of Augsburg (1939). Since three populations (Tubingen Lustnau Tl – Pliezhausen Pl – Reutlingen Re) separated by distances betweenfour (minimum) and ten kilometers (maximum) showed seed production in crossing experiments of 2007, two supplementary populations (Kirchentellinsfurt Kt – Betzingen Bz) were collected within the triangle formed by these populations. At each locality at least ten putative individuals of V. filiformis (with a maximum distance between each other and without connection between them) were collected in plastic bags with wet paper tissue in order to keep them alive during the collecting trip. Afterwards, these plants were transplanted to a greenhouse of the botanical garden of the Johannes Gutenberg-University, Mainz, Germany. Field trips were taken to five regions of the native range of the species to collect seeds and silica-gel dried material, including northern Turkey (2005), and eastern (Kakheti, Lg), central (Samtskhe-Javakhéti, Bk), northern (Kazbegi, Kz) and south-western (Adjara, Bt) Georgia (2008) (Figure 1; Additional file 1). Georgia is the most likely source of origin for introduced populations 40 and northern Turkey represents a distinct different part of the distribution range. Voucher specimens are deposited at the herbarium of Staatliches Museum für Naturkunde Stuttgart (STU).

Plants from each locality were covered collectively by tissue or plastic lids to prevent unintentional cross-pollination and each putative individual (10–12 per population) was potted and labeled separately. Fertilizers (10ml Hakaphos soft Spezial per pot; 16%N, 8% P₂O₅, 22% K₂O, 3% MgO with <0.2% boron, copper, iron, manganese, molybdenum, zinc) were used at the start of the experiment once to improve the success of the transplantation. Flowers are slightly protogynous and the virginity of the stigma was checked by binocular microscopes before crossing. Typically one anther with mature white pollen was used only once to fertilize a single style using sterile forceps in order to deposit enough pollen per style and prevent other crossings. Labeled threads with a code were attached to the pedicels after crossing in order to identify the flowers later. The code, date of crossing, the father individual and its population together with the mother individual and its population were noted. Presence or absence of developing fruits was checked regularly after each crossing. To confirm the absence of capsule production within introduced populations, eight specimens from the population Pliezhausen Pl were crossed with each other (n _flowers = 41). In inter-population crosses, priority was given to crossings of neighboring populations and only surplus flowers were used to cross geographically distant populations to check if these belong to the same crossing group (Figure 2).

A total of 108 DNA-samples considered to be putative individuals from twenty introduced and three native populations (Table 1) were sampled for AFLP fingerprints. DNA-samples do not correspond to the samples used for the crossing experiments although they come from the same German populations. Fresh leaves of the individuals were dried in silica gel and used to extract total genomic DNA by DNeasy^TM plant minikit (Qiagen) following the manufacturer’s instructions. DNA concentration was measured spectrophotometrically with a GeneQuant RNA/DNA calculator (Pharmacia, Uppsala, Sweden), or estimated visually by ethidium bromide staining on agarose gels. The AFLP^TM procedure followed Vos et al. 79 with the following modifications. Genomic DNA was digested with the two restriction endonucleases EcoRI and MseI and ligated to double-stranded EcoRI (5’ – CTCGTAGACTGCGTACC – 3’; 3’ – AATTGGTACGCAGTC - 5’) and MseI (5’ – GACGATGAGTCCTGAG – 3’; 3’ – TACTCAGGACTCAT - 5’) adapters in one step at 37°C for 3 hours. The reaction mix for circa 0.5 μg template DNA contained 1.1 μl T4 DNA ligase buffer (Genecraft, Cologne, Germany), 1.1 μl 0.5 M NaCl, 0.55 μl bovin serum albumin (BSA, 1mg/ml, New England Biolabs, Beverly, Massachusetts), 1.0 μl 50 μM MseI-adapters (Metabion, Martinsried, Germany), 0.1μl MseI (10U/μl, New England Biolabs), 1.0 μl 5 μM EcoRI-adapters (Metabion), 0.25μl EcoRI (20U/μl, New England Biolabs), 0.002 μl T4 DNA ligase (10WU/μl, Genecraft) and 0.9 μl ddH2O. Ligated DNA fragments were diluted 10-fold. Preselective and selective amplifications were performed in a thermocycler (GeneAmp PCR System 9700, Applied Biosystems, Foster City, California) with PCR protocols following Vos et al. 79 with modifications. The reaction mix for preselective amplification contained 2.5 μl 10× NebTaq PCR reaction buffer (New England Biolabs), 0.1 μl NebTaq PCR reaction mix (5U/μl, New England Biolabs), 0.25 μl 10 mM dNTPs (Applied Biosystems), 0.50 μl 5 μM preselective primers E01 (5’ – GACTGCGTACCAATTCA – 3’) and M02 (5’ – GATGAGTCCTGAGTAAC – 3’), 6.65 μl ddH2O and 2 μl diluted product of the restriction-ligation reaction. The PCR product was diluted 10-fold. The reaction mix for the selective amplification contained 1.667 μl 10× NebTaq PCR reaction buffer (New England Biolabs), 0.055 μl NebTaq PCR reaction mix (5U/μl, New England Biolabs), 0.167 μl 20 mM dNTPs (Applied Biosystems), 0.833 μl 50 mM MgCl2, 0.28 μl 5 μM MseI-primer (Metabion), 0.20 μl 1 μM EcoRI-primer (Metabion), 7.92 μl ddH2O and 5 μl diluted product of the preselective amplification. The six primer combinations for the selective PCR were E38 - Hex (5’ Hex - GACTGCGTACCAATTCACT - 3’) combined with M56 (5’ - GATGAGTCCTGAGTAACGC - 3’); E38 - 6 - Fam/M50 (5’ - GATGAGTCCTGAGTAACAT - 3’) and E37 - Ned (5’ Ned - GACTGCGTACCAATTCACG - 3’)/M54 (5’ - GATGAGTCCTGAGTAACCT - 3’); E38 - Hex combined with M62 (5’ - GATGAGTCCTGAGTAACTT - 3’); E38 - 6 - Fam/M53 (5’ - GATGAGTCCTGAGTAACCG - 3’) and E37 - Ned/M51 (5’ - GATGAGTCCTGAGTAACCA - 3’). 2.5 μl of 6-Fam - 3.75 μl of Hex and 3.75 μl of Ned labeled products of each sample were combined, and 2 μl of this multiplex product was run with 7.75 μl HiDi formamide (Applied Biosystems) and 0.25 μl internal size standard GeneScan ROX (Applied Biosystems) on an ABI 3130x automated capillary sequencer. Raw AFLP data were collected and aligned with the internal size standard using ABI Prism GeneScan analysis software 3.7 (Applied Biosystems). Peaks (i.e. fragments) were scored manually as present (1) or absent (0) in a readable region of bands from 75 to 500 bp in length with GeneMarker version 1.5 (GeneMarker, SoftGenetics, State College, Pennsylvania). Each peak with a signal intensity of more than 1000 was selected and checked for presence in each sample.

First, the mismatch error rate based on nine (8.3%) supplementary re-extracted samples (or replicates) belonging to one native and six introduced populations was calculated to evaluate the quality of our analyses and determine the clonality threshold as the number of genotype mismatches divided by the number of replicate pairs and the number of loci 56 80 Table 1. The presence/absence matrix generated with the six primer combinations was imported in FAMD v. 1.1 81 to assess genetic diversity between areas, regions, populations, crossing groups or genetic clusters. The genetic diversity is represented, within these different levels of analysis, by the mean of AFLP fragments present per individual, the number of polymorphic fragments (Frag. polym.) and its percentage (% Frag. polym.), the number of fixed fragment (Frag. fix.), the number of private fragment (Frag. priv.), the number of private and fixed fragments, and finally Nei’s gene diversity (Hj). Analyses of molecular variance (AMOVA) were conducted using the same software. To estimate if differences exist between samples (area, population, crossing groups and genetic clusters) within the genetic data, the non-parametrical Wilcoxon-Mann–Whitney-test (W/U) for two groups or the Kruskall-Wallis-test (H) for more than two groups were conducted using SPSS v 15.0 82 . Genetic structure of populations was analyzed using the software STRUCTURE vers. 2.3 83 . First, most of the parameters were set up as recommended in the user’s manual of STRUCTURE 83 and then, the option of “correlated allele frequencies” between populations was tested as recommended by Falush et al. 84 in case of subtle population structure. From a pilot study, we found that a burn-in period and MCMC (Markov chain Monte-Carlo) of 10000 and 50000 iterations, respectively, were sufficient. A series of 10 runs was conducted for each value of K populations between 2 to 23 (maximum number of localities) and the maximum value of Ln (Prob) obtained for each K was used to represent the best population structure acquired with the no-admixture and admixture models. The no-admixture model was chosen since the amount of admixture was found to be negligible. A Principle Coordinates Analysis (PCoA) was performed using Nei&Li’s coefficient 57 and the coordinates of each individual were extracted to statistically test their distributions by the same Wilcoxon-Mann–Whitney and Kruskall-Wallis tests using SPSS v. 15.0 82 .

In the course of our crossing experiments, normal flowers (Additional file 3A & 2B) and flowers with different types of mutations in the androecium (Additional file 3C1-5) were observed in the greenhouse and in the field. These different types of androecium-mutant flowers were counted in ten introduced and seven native populations (Figure 3 and Additional file 1) in the field or greenhouse. To check if these flower mutations are specific to V. filiformis (V. subgenus Pocilla), these particular mutant flowers were also counted in populations of two other Veronica species: the closely related V. persica (invasive in Europe; self-compatible) and V. chamaedrys (native in Europe; self-incompatible), both of which were growing both spontaneously near one of the V. filiformis populations (arboretum of the botanical garden of Mainz; Mz). The number of mutations in the androecium were compared with a generalized linear model (glm) function glm in R 85 . The differences between native and introduced regions were tested with a glm for binomial data with area and populations as factors.

To estimate the investment in gametophytes, at least eight buds per population in the five native and eight introduced populations of V. filiformis were sampled for a total of 135 mature buds (Additional file 1). Mature buds were collected in the field or in the greenhouse of the botanical garden immediately prior to anthesis and preserved in 0.9 ml of 70% ethanol. Each mature bud came from one separate individual. To count pollen grain number, the two anthers of one flower were dissected carefully under the binocular microscope. Two drops of methylene blue MTT (3 g of methylene blue in 1 l of 20% ethanol) and glycerine were added to the suspension for a total of 1 ml and stirred together for 30 sec with a vortex mixer. Then, 1 μl was transferred onto a slide and pollen grains were counted with the aid of a microscope. This was repeated ten times per flower. The total number of pollen grains was calculated as the mean of the counted pollen grains in the ten drops multiplied by 1,000 (factor of dilution). Ovaries were carefully dissected and the number of ovules per flower was counted under the binocular microscope. The pollen/ovule ratio (P/O) was scored as the total number of pollen grains in the two anthers divided by the number of ovules in the ovary of one flower.

Both the pollen and ovule production were tested for normality and homogeneity of variance by checking the residuals as stated useful by Quinn & Keough 58 . A histogram plot of each parameter showed a normal distribution and also residuals did not increase with increasing parameter values. For both parameters a nested ANOVA was conducted in R 85 with the lm function. Area (introduced vs. native), crossing group and populations were used as factors.

Viability of seeds from crossings was examined in the populations of the triangle Tl-Pl-Re. These populations belong to different crossing groups (red, blue and green crossing groups; Figure 1) and different genetic clusters (Green, Blue and Pink genetic clusters; Figure 2). These seeds were used to study maternal and paternal effects on seed size and viability within the introduced range. At least twenty seeds were tested in each combination of maternal and paternal parent (♀ - ♂: Tl-Pl, Pl-Tl, Tl-Re, Re-Tl, Pl-Re and Re-Pl) to have an equal number of seeds from the same mother and father population. Fifty seeds from three regions in Georgia (Adjara, Samtskhe-Javakhéti and Kakheti) were used to compare seed viability between introduced and native ranges. A total of 150 seeds from the introduced area were compared with 150 seeds from the native area by tetrazolium staining 87 88 89 . Seeds were sliced longitudinally through the midsection of the distal half, then placed in petri dishes with 2.5 ml of 1% 2,3,-triphenyl tetrazolium chloride. After staining for 24h, seeds that displayed a completely red-stained embryo and endosperm were classified to be “viable” 87 88 89 . Other distribution of color were considered to be “abnormal seeds” (i.e. seeds with mixtures of red and white in one tissue, likely staining artifacts of otherwise viable seeds) or “dead seeds” (i.e. seeds with one or both organs white). Seed viability was calculated as the number of viable seeds in the total of tested seeds for each area (introduced/native) or each mother ♀/father ♂ group (Tl/Pl/Re) and statistically tested by the non-parametrical Wilcoxon-Mann–Whitney-test with SPSS v. 15 82 .

A seed germination test 90 91 was conducted and adapted to V. filiformis. 150 seeds from two of the three previously cited native Georgian regions (Samtskhe-Javakhéti and Kakheti) and 150 seeds from a random sample of cross-compatible introduced populations from inside and outside the German transect (Baden-Württemberg and Bavaria) were investigated in three batches of 100 seeds each. Seeds were stored at 5°C for two weeks before being placed in wet Petri dishes (with fungicide added). The number of germinated and non germinated seeds was counted after a temperature treatment of 30°C - 8 hours/15°C - 16 hours during a period of 28 days. For the same reason as for seed size, we pooled the seeds and compared results only by the general non-parametrical Wilcoxon-Mann–Whitney-test using SPSS v. 15 82 .

After observing the differences between introduced and native plants we wanted to test whether Muller’s ratchet acted as a response to lower genetic diversity and increased clonality of populations. Our initial hypothesis was therefore that the percentage of clonality was positively correlated with androecium mutants and fixed AFLP markers, but negatively correlated with number of polymorphic loci, gene diversity (Hj), number of private AFLP fragments, seed, ovule, pollen and flower number. Furthermore, we wanted to test whether populations had spread from single origins or whether the dispersal is more erratic by comparing geographic and genetic distances following the approach by Genton et al. 30 . To examine these relationships, Spearman correlations were investigated using SPSS v. 15 82 .