Crop‐to‐wild hybridization in cherries—Empirical evidence from Prunus fruticosa

Abstract Crop cultivation can lead to genetic swamping of indigenous species and thus pose a serious threat for biodiversity. The rare Eurasian tetraploid shrub Prunus fruticosa (ground cherry) is suspected of hybridizing with cultivated allochthonous tetraploid P. cerasus and autochthonous diploid P. avium. Three Prunus taxa (447 individuals of P. fruticosa, 43 of P. cerasus and 73 of P. avium) and their hybrids (198 individuals) were evaluated using analysis of absolute genome size/ploidy level and multivariate morphometrics. Flow cytometry revealed considerable differentiation in absolute genome size at the tetraploid level (average 2C of P. fruticosa = 1.30 pg, average 2C of P. cerasus = 1.42 pg, i.e., a 9.2% difference). The combination of methods used allowed us to ascertain the frequency of hybrids occurring under natural conditions in Central Europe. The morphological evaluation of leaves was based upon distance‐based morphometrics supplemented by elliptic Fourier analysis. The results provided substantial evidence for ongoing hybridization (hybrids occurred in 39.5% of P. fruticosa populations). We detected homoploid introgressive hybridization with alien P. cerasus at the tetraploid level. We also found previously overlooked but frequent triploid hybrids resulting from heteroploid hybridization with indigenous P. avium, which, however, probably represent only the F1 generation. Although both hybrids differ in ploidy, they cannot be distinguished using morphometrics. Hybrids are frequent and may endanger wild populations of genuine P. fruticosa via direct niche competition or, alternatively or in addition, via introgression at the homoploid level (i.e., genetic swamping). The cultivation of cherries thus substantially threatens the existence of genuine P. fruticosa.

Hybridization as an evolutionary process (together with polyploidization) significantly contributes to the diversity of vascular plants (Soltis & Soltis, 2009). It may lead to evolutionary novelties and the establishment of new species. On the other hand, when reproduction barriers leak, hybridization followed by backcrossing may lead to the extinction of parental species (Rhymer & Simberloff, 1996). The production of hybrid seeds increases, and the reproduction success of parental species is significantly reduced (Levin et al., 1996). Hybrids with the same or greater fitness as their parental species can significantly affect the populations of their parents (genetic swamping; Todesco et al., 2016). Last but not least, even the mere production of sterile hybrid individuals may lead to the extinction of rare parents through the wasteful production of maladapted hybrids, which decreases the number of potential mating partners, and by competition for resources and suitable niches (i.e., demographic swamping; Todesco et al., 2016).
Crop-to-wild gene flow has been documented in several indigenous plant species and may lead to the establishment of aggressive weeds or even the extinction of rare species (Ellstrand et al., 2013(Ellstrand et al., , 1999. So far, only a few human-induced (i.e., with the participation of crop plants) cases of hybridization have been reported. Spontaneous introgression of wild Prunus orientalis (Duhamel) by cultivated Prunus dulcis (Mill.) D. A. Webb in south-west Asia (Delplancke et al., 2012) and genetic erosion of the rare wild species Malus sylvestris (L.) Mill.
in Belgium by domesticated apple (Malus domestica Borkh.; Coart, van Glabeke, de Loose, Larsen, & Roldán-Ruiz, 2006) often serve as model examples. One extreme case of crop-to-wild gene flow is the genus Aegilops L. in the Mediterranean, where more than one quarter of some wild populations bear signs of introgression from wheat (Arrigo et al., 2011). Besides conservation consequences, genetic swamping of wild relatives via hybridization with crops can lead to tremendous economic losses because wild taxa serve as an essential gene pool resource for breeding programmes (Barać et al., 2017;Ganopoulos, Aravanopoulos, & Tsaftaris, 2013).
The main goal of this study was to examine the extent of interspecific hybridization of the rare species Prunus fruticosa with wild and cultivated cherries (P. cerasus and P. avium) and to evaluate the im-  Table S1). Samples of the putative parents Prunus cerasus (43 individuals from 12 locations) and Prunus avium (73 individuals from 38 locations) were also collected in the study area for a better understanding of ongoing microevolutionary processes. Each population sample (usually 5-10 individuals, depending on population size) was represented by a branchlet with vegetative short-shoot leaves. Sampled individuals were as distant from each other as possible to avoid the collection of clonally emerged individuals. Individuals growing together in one place obviously separated from another place were considered a discrete population. As regards P. cerasus and P. avium, about three individuals were sampled from each location because these cultivated taxa are scattered in the landscape instead of constituting numerous populations.
The taxa were determined based on their ploidy level (indicating triploid Prunus × mohacsyana and diploid P. avium). Tetraploids were differentiated based on the presence of hairs on the abaxial surface of the lamina (glabrous P. fruticosa vs hairy P. ×eminens and P. cerasus) and growth form (shrubby P. fruticosa and P. ×eminens vs tree-like P. cerasus).
In total, plant material from 761 individuals of Prunus taxa (447 P. fruticosa, 99 Prunus × mohacsyana, 99 P. ×eminens, 43 P. cerasus and 73 P. avium) were used for three types of analyses-absolute genome size analysis using flow cytometry (FCM), distance-based morphometrics and elliptic Fourier analysis. Dry plant material was used (shortshoot leaves taped on to sheets of cardboard) for morphometrics, and fresh plant material was necessary for flow cytometric analysis.

| Flowcytometry(FCM)
Ploidy levels/absolute genome sizes of 761 individuals (see Supporting information Table S1 for samples details) were F I G U R E 1 Sample locations of Prunus fruticosa and its hybrids in Central Europe estimated using a Partec CyFlow instrument (Partec GmbH, Münster, Germany) equipped with a green solid-state laser (Cobolt Samba, 532 nm, 100 mW). A slightly modified procedure following Doležel, Greilhuber, and Suda (2007)  Because of the significant amounts of secondary metabolites contained in Prunus material (typical of the whole Rosaceae), which complicate FCM analyses, certain optimization steps had to be carried out (for details, see Macková et al., 2017). Although most of the samples were measured at one time point only, we checked the stability of FCM measurements over a long time period (from May to August, 18 individuals from three locations). Variation between two different measurements did not exceed 4% (for information on the stability of FCM measurements over short periods, see Macková et al., 2017). The whole range of measured absolute genome size values was calibrated by chromosome counts (standard karyological methodology; e.g., Lepší, Vít, Lepší, Boublík, & Suda, 2008).
Resulting FCM histograms were analysed using FloMax (version 2.4d, Partec, Münster, Germany). Absolute genome size values were visualized as boxplots in PAST 2.17c (Hammer, Harper, & Ryan, 2001) and as scatter plots in Microsoft Excel 2010. One-way ANOVA followed by Tukey's HSD test in PAST 2.17c (Hammer et al., 2001) was used to ascertain the significance of absolute genome size differences between species.

| Distance-basedmorphometrics
To examine morphological variation of the Prunus taxa under study, 17 characters (13 primary, four ratio)-eight vegetative and nine generative (see Table 1)-were selected based on the literature  (Lepší et al., 2011;Wójcicki, 1988Wójcicki, , 1991Wójcicki & Marhold, 1993) and own field observations. Well-developed short-shoot leaves (two leaves per individual) and flowers were measured using a digital calliper (accuracy 0.01 mm) and a stereo microscope (Olympus SZ51; magnification 40 × ). Most of the time, only shortshoot leaves were observed (because of their narrower range of variation; Marhold & Wójcicki, 1992). Plant height was measured in the field. Abaxial hairs were measured on at least four leaves per individual and then averaged. Plant height, shape of laminar tip, adaxial hairs and abaxial hairs were evaluated using semiquantitative scales (see Table 1). In total, 1,422 leaves (see Supporting information Table S1 for samples details) and only 84 flowers were measured because the flowering period was very short. Because P. fruticosa scarcely bears fruits (Chudíková, Ďurišová, Baranec, & Eliáš, 2012), no fruits were included in the study.
The data matrix was evaluated using multivariate statistical methods in PAST 2.17c (Hammer et al., 2001

| EllipticFourieranalysis
Shape contours of 1,407 leaves (see Supporting information Table S1 for samples details) were investigated using elliptic Fourier analysis.
Only well-developed leaves were included in the analysis (15 partly damaged leaves were excluded). Two leaves of each individual were taped on to a sheet of cardboard paper and scanned (scanner Canon MP270 series Printer; 300 dpi). For leaf shape analysis based on elliptic Fourier descriptors (Kuhl & Giardina, 1982), the SHAPE 1.3 package (Iwata & Ukai, 2002) was employed. The leaf shapes were converted into chain codes using ChainCoder, and the CHC2NEF programme converted these chain codes into coefficients of elliptic Fourier descriptors (using 20 harmonic axes). These coefficients were used to calculate the scores of principal components using the PrinComp function. The PrinComp routine also allowed the reconstruction of the leaf shape, corresponding to values of +2 and −2 standard deviations on the first and second component axes (see Lepší, Vít, Lepší, Boublík, & Kolář, 2009;and Macková et al., 2017, for details). The first and second component axes were visualized using Microsoft Excel 2010.

| AbsolutegenomesizeandDNAploidylevel
Ploidy levels and absolute genome sizes of 761 Prunus accessions were ascertained by flow cytometry (see Supporting information

| Distance-basedmorphometrics
Morphometric variation of 1,422 leaves (see Supporting information Table S1 for samples details) was analysed using distance-based morphometrics (for descriptive statistics, see Supplementary Table S3 Table 1 for character abbreviations; Supporting information Figure S3B). Moreover, a significant correlation between leaf morphology (represented by PC1 scores) and absolute genome size was found (r = 0.729; t = 35.3, df = 1097, p < 0.001), explaining 53% of the overall variation ( Figure 5).

| EllipticFourieranalysis
Variation in the shape contours of 1,407 leaves (see Supporting information Fourier analysis. The groups of Prunus taxa under study overlapped more in comparison with distance-based morphometrics ( Figure 6).
Prunus avium formed the most differentiated cluster, while the P. fruticosa cluster was distinguished only partly. Nevertheless, both overlapped with other Prunus taxa in the principal component analysis.
On the contrary, P. cerasus and both hybrids were scattered between these two clusters and formed a linked and completely overlapping cluster ( Figure 6). The first component axis (68.8%) explained the most variation but was not taxonomic specific (variation in relative leaf width), while the second component axis (13.9%), describing variation in the shape of the leaf base and the shape of the leaf tip, reflecting differences between the taxa studied. The most differentiated groups, P. avium and P. fruticosa, had elliptic leaves with an aristate apex and obovate leaves with an obtuse apex, respectively.
Prunus cerasus, P. ×eminens and P. ×mohacsyana clustered together and tended to form elliptic leaves with a broadly acuminate apex, never obtuse or with an aristate apex ( Figure 6). Thus, leaf shape represents a suitable additional character for the determination of parental Prunus taxa; however, it fails to distinguish hybrids (similar to distance-based morphometrics).

| Frequencyofhybridsundernaturalconditions
Our multidisciplinary approach has revealed that only 60.5% of popu- included both hybrids. At last, 9.2% of the populations analysed were mixed (i.e., composed of P. fruticosa and one of its hybrids).

| D ISCUSS I ON
Absolute genome size/ploidy level estimation coupled with morphometrics allowed us to identify the Prunus species and hybrids concerned (which occurred in 39.5% of populations under study).

Homoploid hybridization between the tetraploid parental taxa
Prunus fruticosa and P. cerasus produces tetraploid hybrids (P. ×eminens). By contrast, heteroploid hybridization between P. fruticosa with P. avium generates triploids (P. ×mohacsyana). The frequencies of the two hybrids turned out to be almost equal in the study area. In contrast to previous attempts to assess the rate of hybridization, which were based solely on morphometrics, our multidisciplinary approach revealed a continuous pattern, pointing to introgression.
Flow cytometry has been employed in several descriptive or local studies of Prunus (Bennett & Leitch, 1995;Dickson, Arumuganathan, Kresovich, & Doyle, 1992;Macková et al., 2017) and published genome size values fall within the range of measured values presented here. The morphological pattern is also analogous to those found in previous studies (Lepší et al., 2011;Wójcicki, 1991;Wójcicki & Marhold, 1993). Traditionally used morphological characters (abaxial hairs and plant height; Lepší et al., 2011;Wójcicki, 1988) have turned out to be more suitable in the field than the first three characters identified by morphometrics (i.e., distance from the widest part of the lamina to the laminar tip, laminar width and laminar length).
Until now, almost all hybrids had been suggested to be tetraploid (P. ×eminens; Lepší et al., 2011;Wójcicki, 1991;Wójcicki & Marhold, 1993), but our data show that the frequency of triploid hybrids, which is roughly 50%, had been considerably underestimated. Leaf shape (elliptic Fourier analysis) seems to be a useful complementary trait for distinguishing between pure Prunus species and hybrids, and a similar pattern was also detected in one local study of P. fruticosa (Czech Republic; Lepší et al., 2011). Thus, based on DNA ploidy level knowledge, the results of previous studies (Chudíková et al., 2012;Lepší et al., 2011;Wójcicki, 1991;Wójcicki & Marhold, 1993) might have to be substantially reevaluated.

| Identityofhybrids
Due to the broad range of absolute genome sizes possessed by the parental species and their hybrids, it is almost impossible to distinguish cytometrically between F1 hybrids and their more complex backcrossed counterparts at the homoploid level (i.e., 4× ).

Moreover, an intermediate genome size does not necessarily indi-
cate an F1 hybrid. To draw the conclusion that a plant is an F1 hybrid, one has to rule out the possibility that it is a higher or even backcrossed hybrid. Continuous patterns of absolute genome size are nevertheless usually accompanied by enormous morphological variation, and a continuous pattern of data distribution in both absolute genome size and morphology is usually indicative of introgressive hybridization (e.g., Hanušová et al., 2014;Suda et al., 2007;Šmarda & Bureš, 2006). In addition, our correlation analysis and RDA revealed that hybrids with an absolute genome size similar to that of one of their parental taxa are also morphologically close to that parent, which indicates that they are almost certainly backcrossed. A high probability of backcrossing at the tetraploid level is further supported by the substantial fertility of P. ×eminens (based on embryology; Macková et al., 2017). By contrast, heteroploid hybridization (i.e., 4× × 2× ) produces comparatively straightforward results due to the existence of an effective triploid block, which constrains backcrossing; this has been proved in the case of triploid P. ×mohacsyana (Macková et al., 2017).

| Crop-to-wildstudiesandtheirlimitations
Human-induced hybridization (or even introgression) affects wild plant species in different ways, and there are several cases that are analogous to that of Prunus fruticosa. While hybridization of cultivated Saccharum L. or Brassica L. with wild counterparts does not pose any risk to their wild relatives, hybridization of cultivated Oryza L. and Gossypium L. has been implicated in the near extinction of certain wild species of rice and cottonseed (Ellstrand et al., 1999).
Studies dealing with crop-to-wild gene flow rely on the ability to unequivocally distinguish between wild and cultivated plant forms. In most cases, however, this discrimination is not possible based solely on morphological grounds (e.g., Malus Mill. Coart et al., 2006;Vitis L. De Andrés et al., 2012). Plant sex might serve as another suitable and con-   Vítová et al., 2015) and by decreasing the number of potential mating partners (i.e., demographic swamping; Todesco et al., 2016).

| Conservationimplications
The potential for the displacement of P. fruticosa is further enhanced by the fact that the two hybrids tend to outgrow it. In contrast to sterile triploid hybrids (Macková et al., 2017), fertile tetraploid hybrids can directly endanger genuine P. fruticosa by introgression (i.e., genetic swamping; Todesco et al., 2016). Still, however, some isolated triploid hybrid populations could represent old, partly fertile, spontaneous hybrids with autochthonous P. avium (Lepší et al., 2011). Introgression involving triploid hybrids has also been documented in other genera (e.g., Betula L. in Iceland; Thórsson, Pálsson, Sigurgeirsson, & Anamthawat-Jónsson, 2007), so the potential risk that triploid F1 hybrid could participate in further backcrossing cannot be ruled out.
The main practical implication of our results is the necessity to limit the cultivation of both sour and sweet cherries in the vicinity of wild populations of genuine P. fruticosa (within a perimeter of at least 1.5 km, as recommended by Boratyński et al., 2003). To this end, it is first necessary to select populations to be protected with high priority (i.e., those which are the most genetically variable-see below).

| Genomesizeanalysisasasuitabletoolfor detectingintrogression
The continuous absolute genome size values at the homoploid level, together with the wide morphological variation, suggest repeated backcrossing between parents and hybrids (e.g., Hanušová et al., 2014;Suda et al., 2007;Šmarda & Bureš, 2006 and, particularly, the importance of clonal growth (genetic variation of populations).

| CON CLUS IONS
In the wild, genuine Prunus fruticosa frequently hybridizes both at the homoploid level (with cultivated P. cerasus) and at the heteroploid level (with P. avium). Our direct identification and quantification of interspecific hybridization/introgression under natural conditions has confirmed the serious risk of ongoing demographic and genetic swamping, as 39.5% of the populations we studied are of hybrid origin. Moreover, homoploid introgressive hybridization poses a substantial conservation threat because P. cerasus is alien to the European flora. Maintenance of a diverse and heterogeneous P. fruticosa gene pool is essential for Prunus breeding programmes as well as for the species' protection. A future conservation genetic investigation should focus on the identification of the most valuable (i.e., the most genetically variable) populations of genuine P. fruticosa.

ACK N OWLED G EM ENTS
We are grateful to Michael Macek, Martin Ševc, Pavlína Hrdinová,

Martina Tůmová and Petr
Glonek for their assistance in the field. Agency.