Molecular and morphological evidence of hybridization between two dimorphic sympatric species of Fuchsia (Onagraceae)

Abstract Hybridization is commonly reported in angiosperms, generally based on morphology, and in few cases confirmed by molecular markers. Fuchsia has a long tradition of ornamental cultivars with different hybrids produced by artificial crosses, so natural hybridization between sympatric Fuchsia species could be common. Natural hybridization between F. microphylla and F. thymifolia was tested using six newly developed microsatellites for F. microphylla in addition to other molecular markers with codominant and maternal inheritance. Geometric morphometrics of leaves and floral structures were also used to identify putative hybrids. Hybrids showed a different degree of genetic admixture between both parental species. Chloroplast DNA (cpDNA) sequences indicated that hybridization occurs in both directions, in fact, some of the hybrids showed new haplotypes for cpDNA and ITS (internal transcriber spacer of nuclear ribosomal RNA genes) sequences. The morphology of hybrid individuals varied between the two parental species, but they could be better identified by their leaves and floral tubes. Our study is the first to confirm the hybridization in natural populations of Fuchsia species and suggests that hybridization has probably occurred repeatedly throughout the entire distribution of the species. Phylogeographic analysis of both species will be essential to understanding the impact of hybridization throughout their complete distribution.


Introduction
Defining a species has always been a subject of debate (Aldhebiani 2018), especially for plants due to the frequent reports of hybridization and the enormous phenotypic variation exhibited by many groups of plants (Rieseberg and Willis 2007).The most used definition of species corresponds to the biological species concept as 'groups of interbreeding organisms that can successfully interbreed and produce fertile offspring and are reproductively isolated from other groups of organisms' (Mayr 1942;Aldhebiani 2018;Gustafsson et al. 2022).Reproductive isolation is a process that prevents individuals from different populations to mate and produce fertile offspring through pre-zygotic or post-zygotic barriers.Together these barriers are critical for driving speciation and maintaining species identity (Ramsey et al. 2003;Ouyang and Zhang 2018).Species that co-flower, share pollinators, and have incomplete reproductive barriers, may transfer interspecific pollen facilitating natural hybridization.This means that divergent lineages may meet, reproduce, and form at least some offspring of mixed ancestry (Keller et al. 2020).Hybridization may occur in different spatial and temporal contexts, for example, secondary contact after a period of independent evolution or primary intergradation with divergent selection, both cases occurring before species are fully reproductively isolated from each other (Abbott et al. 2013;Abbott 2017).
The initial recognition of hybrids can arise in different ways such as (i) artificial hybridization of cultivated plants, (ii) spontaneous hybridization of plants in gardens, (iii) from plants that had previously been regarded as species, or as infraspecific variants that were later recognized as hybrids and finally, (iv) some hybrids had been discovered by taxonomic specialists that subjected a genus to more detailed and critical study than they had hitherto received (Preston and Pearman 2015).The use of molecular markers helps to know the evolutionary consequences of hybridization which will always depend on its frequency and extent, as well as on the fitness of hybrids (Rieseberg et al. 1993;Pinheiro et al. 2010;Whitney et al. 2010).When the F 1 hybrids are not reproductively isolated from their parents, they can promote introgression by recurrent backcrossing (Burgess and Husband 2004;Kirk et al. 2005;Field et al. 2009;Kamiya et al. 2011;Abbott 2017).In those cases, when reproductive isolation is incomplete and introgressive hybridization occurs, complex hybrid zones may be produced (Rhymer and Simberloff 1996;Abbott 2017).
Natural hybridization is a common phenomenon in plants and reports of spontaneous hybridization ranges from 25% at the species level to 40% at the family level; however, hybridization is not uniformly distributed among plant taxa (Ellstrand et al. 1996;Mallet 2007;Whitney et al. 2010).
Onagraceae stands out as one of the families with the highest propensity to hybridization; nonetheless, research with molecular markers in this family remains relatively limited (Whitney et al. 2010).From the 22 genera of the Onagraceae family, natural hybridization has been reported in species of Ludwigia, Epilobium, Clarkia, Oenothera, and Fuchsia; the first three genera include cases of polyploid species with the probable hybrid origin, while Oenothera has evolved a specialized system of permanent translocation heterozygosity (PTH), apparently as a product of hybridization.Finally, the success of artificial crosses in species of Epilobium and Fuchsia, suggests the possibility that natural hybridization is taking place (Wagner et al. 2007).
For Fuchsia (Onagraceae), a genus with over 110 species of mesic shrubs confined to cool, moist habitats, the phylogeny indicates a rapid initial diversification, due to its poor basal resolution, whose modern lineages are estimated to have diverged 31 mya (Berry et al. 2004).Fuchsia is well known for its ornamental cultivars developed from artificial crosses of some species of section Quelusia and Ellobium (Wagner et al. 2007;Talluri 2012).In addition, artificial hybridization has been performed between several species and different cultivars of Fuchsia, observing that the success of the crosses depends on the taxonomic distance between the species, indicating weak reproductive barriers (Talluri 2009(Talluri , 2012;;Talluri and Murray 2014).Natural hybridization has also been suggested among sympatric species of Fuchsia (Breedlove 1969;Talluri 2009).For example, F. colensoi was first described as a species but later it was discovered that F. colensoi individuals can be produced by artificial crosses between F. excorticata and F. perscandens.Therefore, Fuchsia colensoi changed its name to F. x colensoi although no differences were found in the chloroplast genome among the three species, which suggests a high level of introgression (Sytsma et al. 1991;Godley and Berry 1995).Natural hybridization between F. regia and several other sympatric Fuchsia species has also been suggested based on the apparent intermediate morphology (Berry 1989), but artificial crosses between F. regia and F. campos-portoi, did not produce fruits (Archer et al. 2021).In a similar way, the hybridization between F. microphylla and F. thymifolia, two closely related species from the section Encliandra (Berry et al. 2004) has been suggested (Breedlove 1969;Talluri 2011) but there is no evidence to demonstrate their hybridization in natural populations.
Here, we tested the hypothesis that individuals with intermediate morphology, found in a sympatric population of F. microphylla and F. thymifolia are hybrids.We used geometric morphometry and molecular markers to answer the following questions: Are F. microphylla and F. thymifolia hybridizing in a sympatric population?if so, what is the direction of hybridization? and finally, how is the morphology of the intermediate individuals compared to their parents?

Study species
F. microphylla and F. thymifolia are two morphologically gynodioecious but functionally subdioecious species of shrubs with female and hermaphrodite individuals, but hermaphrodite individuals function as males (Arroyo and Raven 1975;Cuevas et al. 2014).Both species occupy very similar habitats and almost the same range of distribution in Mexico, but F. microphylla extends to Panama; while F. thymifolia only reaches Guatemala (Breedlove 1969).The two studied species have sympatric and allopatric populations throughout their distribution and share some relevant features: hermaphrodite flowers are larger than female flowers, each flower has four petals and four sepals, as well as a tetra lobed stigma, a cylindrical or obconical floral tube, with eight stamens or staminodes in hermaphrodites and females, respectively.The flowers of F. microphylla are always purplish-red, while the petals of F. thymifolia are white when the flower opens and then turn purplish (two or three days after, Cervantes and Cuevas 2023).The leaves of F. microphylla are serrate from the apex to the middle, and those of F. thymifolia are sub entire.The literature reports different pollinators for each species (i.e.hymenopterans for F. microphylla and dipterans for F. thymifolia, Breedlove 1969;Arroyo and Raven 1975), but recent observations suggested that they may share at least one pollinator (Bombus ephippiatus) in the sympatric population of Garnica where some overlap in the flowering season has been observed (Cuevas et al. 2014;Cervantes et al. 2018).

Study area
The studied populations are located mainly in the Trans-Mexican Volcanic Belt (Fig. 1).In general, the climate of these mountains is temperate with rainfalls in summer with a mean annual temperature between 10 and 20 °C and between 400 and 1000 mm of annual precipitation (Hernández and Carrasco 2007).In general, both species are found on slopes, ravines, and humid areas of Pinus, Abies, Quercus, and mesophilic forests at elevations ranging from 2500 to 3400 m a.s.l.(Breedlove 1969;Arroyo and Raven 1975;Rzedowski et al. 2005).
Leaf tissue was collected from 212 individuals belonging to two allopatric populations of each species and two sympatric populations (Fig. 1; Table 1).Twenty to 32 individuals per species/per population, separated at least 3 m from each other, were randomly collected.The leaves were stored at −72°C or dried in silica until processed.Given the limited number of flowers during the leaves collection for the genetic analyses, the flowers, and leaves for the morphometric analyses were collected in different flowering seasons from different individuals and a smaller number of populations (one sympatric and one allopatric for each species; Table 1).
In the sympatric population of Garnica, the individuals of each species were collected at random, while 32 individuals with intermediate morphology between F. microphylla and F. thymifolia were collected and considered as putative hybrids.The sympatric population of Amecameca was included in the study because Breedlove (1969) reported one hybrid in this population.However, at the time of collection there were no flowers to try to identify the hybrids, so the collection of leaves was at random.Allopatric populations for each species were included as a reference since it is likely that 'pure' individuals may not be found in sympatric populations due to possible hybridization.

Nuclear simple sequence repeats (nSSRs)
Simple sequences repeats (SSR) were characterized de novo by Genetic Marker Services (Brighton, United Kingdom; www.geneticmarkerservices.com,see protocol and characterization of 15 loci from F. microphylla in the Supporting Information 1 (SI1)).
From the 15 developed loci, only seven were scorable and polymorphic for 40 individuals from F. microphylla and then were further tested using the PCR conditions specified in (Supporting Information 1; Table S1 and S2), in other 40 individuals of F. microphylla, 80 from F. thymifolia and 32 intermediate individuals from Garnica (see Table 1).Locus Fus 33 only amplified for F. microphylla and intermediate individuals, therefore it could not be included in the microsatellites analyses.

Nuclear and chloroplast DNA sequences
Genomic DNA was chosen from 3 to 4 individuals per population, plus the 32 hybrids from the Garnica population.Chloroplast DNA (cpDNA) was amplified with the trnL-c and trnL-d primers (Taberlet et al. 1991), and for the cpDNA, rpL16 intron we used the primers F: 71 and R: 1661 (Kelchner and Clark 1997).The nuclear rDNA (ITS) region was amplified with the primers ITS-p4: R and ITS-p5 F (Cheng et al. 2016; see Supporting Information: Table S3).
All PCR reactions had a total volume of 50 μL, containing 5 μL of diluted DNA (~40 ng), 0.2 μM of each primer, and 25 μL 1x Qiagen Multiplex PCR Master Mix.In the case of trnL and ITS reactions 2.5 μL of dimethyl sulfoxide (DMSO) were added.The PCR products were examined via electrophoresis in 1.5% agarose gels and with automated capillary electrophoresis with QIAxel (QIAGEN).The PCR products were purified using purification pearls (AMPure XP, Beckman Coulter).Finally, samples were sequenced in both directions on an ABI 3720xl System (Macrogen Korea).

Nuclear microsatellites
To detect genotyping errors caused by null alleles, stuttering, or large allele dropout we used Micro-checker 2.2.3 (Van Oosterhout et al. 2004).We looked for private alleles of each species to determine whether the hybrids possess private alleles of each parental species.
We evaluated genetic structure in the sampled populations with a principal coordinates analysis (PCoA) in GENALEX 6.5 (Peakall and Smouse 2012).STRUCTURE 2.3.4 (Pritchard et al. 2000) was used to calculate an admixture coefficient (q) for every individual.In the runs, the admixture model was employed with allele frequencies correlated, with K values from 1 to 10, with 10 runs for each K.In each run, a burn-in of 500 000 iterations followed by 1 000 000 Markov chain Monte Carlo (MCMC) iterations were used and then STRUCTURE HARVESTER (Earl and vonHoldt 2012) was used to determine the most likely K value by measuring the ΔK.Individuals with q ≥ 0.90 were considered as belonging to F. microphylla, while individuals with q ≤ 0.1 were considered of F. thymifolia.Individuals with 0.1 ≤ q ≤ 0.9 were classified as hybrids.Given the low number of private alleles for F. thymifolia, if we only consider the admixture coefficient (q) we could misclassify most hybrids as pure F. microphylla individuals.Therefore, the private alleles and the homozygous fixed alleles of F. thymifolia in Fus 31 and Fus 34 were key to identify hybrids.We used NewHybrids version 1.1 beta (Anderson and Thompson 2002), to estimate the posterior probability of each individual belonging to categories such as parental purebreds, F 1 , F 2 and backcross, assuming that samples are composed of pure parental species and hybrids.We used the default genotype categories for first and second generations of crossing and ran 100 000 MCM with burn in a period of 50 000 with Jeffrey-type prior.Also, the allopatric populations were used as reference samples of pure individuals.The individuals were assigned to one of the six genotypic classes if P ≥ 0.90.
Finally, GenAlEx 6.501 (Peakall and Smouse, 2012) was used to estimate the number of alleles (A), number of private alleles (A p ), observed (H o ), and expected (H e ) heterozygosity, departure from Hardy-Weinberg equilibrium, inbreeding coefficient (F is ) and pairwise F st among the parental taxa and the hybrids, and between all the populations with 999 permutations.

Chloroplast and nuclear sequences
In total, 60 individuals were sequenced, except for the ITS ith only 38 individuals amplified despite repeated attempts.All sequences from each locus were assembled and inspected using Sequencher 4.1.4and Phyde 0.9971 (2010) and then were aligned with Muscle 3.8.31(Edgar 2004).Adjacent multiple base gaps were treated as a single indel because a single deletion event is the most parsimonious explanation for contiguous alignment gaps (Garrick et al. 2004).The trnL and rpL16 sequences were concatenated and then haplotype networks were constructed separately for chloroplast and nuclear sequences using statistical Parsimony in TCS v1.21 (Clement et al. 2000).ITS sequences were reconstructed with PHASE in DnaSPv.5.10 (Librado and Rozas 2009).We also calculated the number of haplotypes (h), haplotype diversity (H d ), and nucleotide diversity (π), for F. microphylla, F. thymifolia and hybrids using DnaSP v.5.10 (Librado and Rozas 2009) always considering gaps as the fifth state and excluding gaps only in pairwise comparison.

Geometric morphometry
The leaves and flowers collected for geometric morphometry were taken from a sympatric population (Garnica) and an allopatric population for each species (FmDL and FtT, see Table 1).Between 1-3 flowers and between 3-4 leaves from 10 plants per morph (female and hermaphrodite) per population were collected and then photographed with a reference scale in different positions to visualize the floral tube, the corolla, the calyx, and the leaves obtaining a total of 377, 391, 394 and 433 photographs, respectively.A different configuration of landmarks was constructed for each set of photographs to adequately represent each structure.Only one petal or sepal was used from each flower (see Supporting Information 2 and Fig. S1 for description and example of landmarks positions).
For all structures, we built a tps file from images with tpsUtil 1.33 (Rohlf 2004) and then the landmark configurations were digitized using the program tpsDig 2.31 (Rohlf 2017).Indira Cervantes-Díaz et al. -Fuchsia hybridization in natural populations A generalized procrustes analysis (GPA) was performed for each landmark configuration of each morphological structure with the geomorph 4.0 package of the RStudio software 1.4.1106(R Studio Team 2018).The GPA consists of minimizing the sum of squared distances between corresponding landmarks to extract shape data by removing information on size, location, and orientation (Klingenberg et al. 2012;Savriama 2018).The landmarks configuration from flowers and leaves of the same individual were averaged after superimposition, and then a GPA was done.To maximize the differences between groups relative to the variation within groups, a canonical variates analysis (CVA) was performed independently using the procrustes coordinates of leaves, floral tubes, petals, and sepals (Klingenberg et al. 2012).The statistical significance of pairwise differences in mean shapes was assessed with permutation tests using Mahalanobis distance (10 000 permutations per test).

Nuclear microsatellites (nSSRs)
After exhaustive amplification tests, we were able to amplify six nuclear microsatellite loci (Fus 27, Fus 52, Fus 31, Fus 47, Fus 29, and Fus 34) in both parental species and in the 32 putative hybrids.Micro-checker 2.2.3 (Van Oosterhout et al. 2004) suggested possible null alleles in three populations for Fus 34 (FmDL, FmA, and HG), and in FtT, FtG, and FmG for Fus 52, Fus 31, and Fus 47, respectively.Finally, Fus 52 could have probable errors due to stuttering in FtT.In total, we found 86 alleles, and 29 (34%) are shared among F. microphylla and F. thymifolia while 39 (45%) and 17 (20%) are private alleles, respectively.Almost all hybrids had at least one private allele from each of the parental species, as evidence of their hybrid origin except for five intermediate individuals, who only had private alleles from F. microphylla (HG4 and HG25) or from F. thymifolia (HG23, HG29, and HG30).Specifically, hybrids showed four private alleles from F. thymifolia: one in the Fus 52 and Fus 31 loci, and two in the Fus 47 locus.In contrast, hybrid individuals shared 14 private alleles with F. microphylla at all the loci except Fus 29.All the individuals of F. thymifolia, except three (FtA7, FtG9, FtG16), are homozygous at the Fus 31 and Fus 34 loci, with the alleles 131 and 124, respectively.For F. microphylla, four individuals (FmDL11, FmDL30, FmA1, and FmA8) have one 124 allele, while the FmDL10, FmDL19, and FmDL26 individuals are homozygous for this allele.
The PCoA clearly separated the individuals of F. microphylla from those of F. thymifolia and placed the putative hybrids in an intermediate position, although with more variation (Fig. 2).The first axis accounted for 25% of the genetic variance and separated the two parental species, while the second axis explained 9% and mainly separated the F. microphylla populations in two groups, one including FmDL-FmA and the other FmCB-FmG.Interestingly, the 32 putative hybrids were represented by only 19 points, and individuals (Fig. 2) HG4 and HG6 grouped with F. microphylla individuals; while HG29 grouped with F. thymifolia.Furthermore, six individuals from FmDL population (FmDL6, FmDL10, FmDL11, FmDL19, FmDL20, FmDL30) and two from FmA (FmA1, FmA8) grouped with the putative hybrids from Garnica, as well as three F. thymifolia individuals (FtA7, FtG9, FtG16) from the sympatric populations (Fig. 2).
STRUCTURE identified two main genetic clusters, one corresponding to F. microphylla and the other to F. thymifolia.Ten out of 32 putative hybrids had an admixture coefficient between (0.1 ≤ q ≤ 0.90) corresponding to the hybrid category.This analysis also identified 18 hybrids as pure Fm individuals and only four as Ft pure individuals.With such a threshold value, five individuals from the allopatric population FmDL, and three from FtL had an admixture coefficient corresponding to the hybrid category.In contrast, according to the NewHybrids program the putative hybrids were classified into 18 pure Fm individuals, three pure Ft individuals, and 11 hybrids that could not be assigned to any of the six categories because they had a posterior probability between 0.45 and 0.82 (Fig. 3).In addition, the NewHybrids classification, rejected the FtA7, FtS10 and four individuals from FmDL, as pure individuals.
Genetic diversity was higher in F. microphylla than in F. thymifolia across all populations and parameters (Table 2).Even though the hybrids of Garnica only had one private allele, they showed greater genetic diversity than all the populations of F. thymifolia and some of the F. microphylla populations.Genetic differentiation was significant between all pairs of F. microphylla, F. thymifolia and hybrid populations, with F st values ranging from 0.086 to 0.497 (Table 3).With respect to pairwise differentiation between parental species and hybrids, the highest F st corresponds to F. microphylla-F.thymifolia comparison (0.299, P < 0.001), followed by Ft-HG (0.208, P < 0.001), and finally lower genetic differentiation was observed for Fm-HG (0.105, P < 0.001).

Chloroplast sequences
We found 11 haplotypes on the concatenated cpDNA sequences (1575 bp) with 43 variable sites including polymorphic/indel/missing sites after a codification to adjacent multiple base gaps.The software TCS with a 95% parsimony limit (connection limit = 17) produced one haplotype network    Hap7).The haplotype diversity for chloroplast sequences by taxa ranged from 0.47 (Fm) to 0.86 (Ft) and the nucleotide diversity ranged from 0.001 (Fm) to 0.007 (HG; Table 5).
Although FmG15 was initially collected as a 'pure individual' of F. microphylla, it shows a unique haplotype that is much closer to F. thymifolia.In contrast, FtA3 has a unique haplotype more similar to F. microphylla than to F. thymifolia, and both cases indicated their hybrid origin.

Nuclear sequences
We found 24 haplotypes on ITS sequences (625 pb) with 20 variable sites from 38 individuals (Fig. 4b).The TCS software with a 95% parsimony limit (connection limit = 10) produced one haplotype network from the reconstructed sequences (76 in total).Eight haplotypes were found in F. microphylla individuals but only Hap8 and Hap10 were shared with one (HG02) and two (HG23, HG30) of the putative hybrids, respectively.The allopatric population of FmCB had one private haplotype (Hap11), while the sympatric population of FmG had four (Hap12-15), and FmDL shared the Hap9 with one individual from FmA.In F. thymifolia populations, we found seven haplotypes and three were exclusive to FtL and FtT (Hap5,Hap6,and Hap7); the Hap1 and Hap2 correspond to the individuals of the sympatric population of FtA.In Garnica, the FtG-HG individuals shared Hap3 and Hap4.Specifically, the hybrids had nine exclusive haplotypes, four very similar to Fm (Hap17,Hap18,Hap21,and Hap23) and the other five to Ft (Hap16, Hap19, Hap20, Hap22, and Hap24; Fig. 4b).However, only 14 hybrids could be sequenced for ITS (Table 4).The ITS haplotypes observed in four hybrids were more closely related to F. microphylla; seven hybrids showed an opposite pattern and only the hybrids HG23 and HG32 had haplotypes mixed from the two parents.Unfortunately, we were not able to obtain data for some putative hybrids.The haplotype diversity for ITS  sequences (Table 5) by taxa ranged from 0.84 (Fm) to 0.92 (HG) and the nucleotide diversity ranged from 0.0040 (Ft) to 0.0057 (HG; Table 5).

Geometric morphometrics
According to CVAs, the leaves, floral tubes, and petals showed a clear morphometric distinction between the two Fuchsia species and the putative hybrids (Fig. 5).Populations within F. thymifolia only showed moderate overlap in the shape of the leaves.In contrast, the CVA of floral structures displayed a greater intraspecific separation in shape.For F. microphylla populations the shape of the floral tubes and petals were more uniform, while leaves and petals showed more separation between populations.Mahalanobis distances range from 6.74 (FtG vs FtT) to 14.32 (FtT vs FmG) for leaves, 9.2 (FmG vs FmDL) to 26.5 (FtG vs FmDL) for floral tubes, and 9.73 (FmG vs FmDL) to 16.18 (FtT vs FmDL) for petals, and all permutation tests indicated that mean shapes differ significantly among populations (P < 0.01 in pairwise permutation tests between populations).

Discussion
The results of the present study provided the first strong evidence of natural hybridization in Fuchsia, supported by Table 4. Classification of hybrid plants found in Garnica population.Structure (q) shows the genetic proportion Fm/Ft.The NewHybrids show the genetic class assigned and its posterior probability.Chloroplast and ITS showed the haplotype(s) for each hybrid and the species to which each haplotype belongs is indicated in parenthesis.Fm = F. microphylla, Ft = F. thymifolia, HG = hybrids from Garnica.The hybrids in parentheses have the same SSR genotype and the same admixture coefficients and posterior probabilities.NS = sequences could not be obtained.*Only have private alleles from one parental species.analyses of floral and leaf morphology, and genetic markers.

Hybrid
The results of this study support the previously pointed-out weakness of the reproductive barriers in this genus, based on a large number of artificial hybrids between other Fuchsia species (Berry et al. 2004;Talluri 2012).

Hybrid identification with molecular markers
The different molecular markers used in this study confirmed that several intermediate individuals are the result of hybridization between F. microphylla and F. thymifolia.The nuclear microsatellites allowed us to find private alleles of both species in most hybrids.In addition, the SSRs confirmed a high genetic differentiation between F. microphylla and F. thymifolia, and showed that backcrosses to both species are occurring.Microsatellites also reduced the 32 initial putative hybrids to only 19 different genotypes, and from those, structure identified nine as purebread individuals, five from F. microphylla and four from F. thymifolia and the remaining hybrids had different degrees of genetic admixture (Fig. 3).In addition, according to NewHybrids using a threshold value of 0.90, eight putative hybrids were classified as pure individuals, but the program could not assign the rest of the putative hybrids from Garnica to the F 1 hybrid category, so they could be considered as hybrids of later generations.In other hybrid systems, the threshold values used for NewHybrids are decided under two main criteria: (i) the threshold value is applied to each category (pure species, F 1 hybrids, backcrosses) separately, by assigning only the individuals with q ≥ Tq and leaving the other individuals unassigned or, (ii) a threshold value above 0.9 or 0.94 is used for pure individuals and a lower threshold for hybrid categories that may range from 0.5 to 0.80 (Zha et al. 2008;Burgarella et al. 2009;McIntosh et al. 2013;Tsy et al. 2013;Vega et al. 2013).The loss or gain of the homozygous condition on Fus 31 and Fus 34, and the results found in the PCoA, the structure and the NewHybrids analysis indicated the presence of other individuals from the allopatric populations with genetic admixture in FmDL, FtA, and FtL populations.In contrast, individuals from sympatric populations initially collected as 'pure' but with a certain degree of genetic admixture could suggest introgressive hybridization as in other hybridizing genera as Silene (Minder et al. 2007), Quercus (Lee et al. 2014) and Vincentoxicum (Li et al. 2016).This scenario may happen when hybrids backcross more frequently with one parental species, providing a greater genomic contribution in the hybrids, which may be confused as pure individuals, as in the case of hybrids found in the sympatric populations of FtA and FtG.On the other hand, hybrids identified in the allopatric populations might be indicating past events of introgressive hybridization, or incomplete lineage sorting, but distinguishing between them is a difficult task.Although hybridization is expected to leave traces only under sympatric conditions, incomplete lineage sorting leads to a spread-out pattern of shared genetic variation (Goetze et al. 2017).
In angiosperms, the DNA of chloroplast is generally maternally inherited, potentially allowing to identify the maternal progenitor of hybrids (Burgess et al. 2005;Zhou et al. 2008;Yuan et al. 2010;Matos et al. 2016).The hybrids of the Garnica population share chloroplast haplotypes from both species, meaning that hybridization may occur in both directions.Specifically, from the 19 hybrid genotypes eight have F. microphylla haplotypes, five have F. thymifolia haplotypes, and six have new haplotypes closer to F. thymifolia (Table 4).The individual FmG15 has a unique haplotype very similar to the Hap2 of FtG, even though it was assigned as pure F. microphylla according to the q-value.In turn, FtA3 shows a unique haplotype very similar to the Hap9 of F. microphylla but with a q-value of F. thymifolia.
Furthermore, only the allopatric populations FmCB, FtL, and FtT showed private haplotypes, contrasting to the allopatric FmDL population which shares haplotypes with the two sympatric populations of F. microphylla.This could reinforce the hypothesis that the population of Desierto de los Leones has undergone more recent introgression compared to the other allopatric populations.
For biparentally inherited markers such as the ITS region, we expected hybrids to have both maternal and paternal genetic components as well as their own copies resulting from genetic recombination (Matos et al. 2016).The ITS region is a universal species-specific marker for plants and fungi, typically having several hundred copies within plant genomes, meaning that the signature of autopolyploidization, recent hybridization, or introgression, might be detected unless enough generations have passed for concerted evolution to homogenize all ITS copies within the genome (Nazre 2014;Wan et al. 2014).The ITS region has been used to detect hybrids in many different hybridizing species (Fuertes-Aguilar et al. 1999;Hardig et al. 2000;Tsukaya et al. 2003;Wu et al. 2009;Korpelainen et al. 2010;Nazre 2014).In Fuchsia, individuals from allopatric populations of both species exhibited only one allele at the ITS region, while some pure individuals from sympatric populations and most hybrids were heterozygous, confirming their hybrid origin.For example, the hybrids HG7 to HG11 shared one allele with F. thymifolia individuals from Garnica and the other allele was very similar to F. thymifolia.By contrast, HG12 hybrid showed two different alleles very similar to individual FmG15 previously identified as pure F. microphylla but with cpDNA haplotype more similar to F. thymifolia.
The different molecular markers indicated that there are different classes of hybrids within the sympatric population of Garnica.However, many intermediate individuals from Garnica previously classified as pure individuals, show chloroplast or nuclear haplotypes that do not correspond to the species they were classified to.Thanks to this variety, the hybrid lineage has been able to persist within the population, since surely some genotypes have been able to reproduce successfully, functioning as a bridge for gene flow between both parental species.

Morphological identification
Hybrids have been shown to display a mosaic of intermediate, parental, and transgressive characters, rather than just being intermediate (Rieseberg et al. 1993).Traditionally, morphometric comparisons between populations were based on the analysis of differences in their linear dimensions (Toro et al. 2010), but new methods that generate quantitative descriptions of shapes are quite useful for comparing shapes within-and among-species (Jensen et al. 2002).Geometric morphometric analyses have been applied to discriminate between parental species and their hybrids in plants, mainly using leaf shape (Jensen et al. 2002;Lexer et al. 2009;Viscosi et al. 2009;Peñaloza-Ramírez et al. 2010;Liu et al. 2018) and to a lesser degree flower shape (Shipunov and Bateman 2005;McCarthy et al. 2016;Pisano et al. 2018).In this study, in addition to leaves, we included various floral structures and found that the shape of leaves and the floral tube differentiated the two species to a greater extent, however, F. microphylla has less intraspecific variation in floral tube shape while F. thymifolia has less intraspecific variation in leaf shape.The putative hybrids have leaves more similar to those of F. thymifolia and flower tubes are intermediate between their parents.
Although the petals and sepals also allowed to differentiate both species, these structures showed greater variation in their shape, and the putative hybrids had petals and sepals more similar to those of F. microphylla.In addition, in the sympatric population of Garnica it is easy to recognize both parental species, because F. microphylla is a small shrub (less than 1 m) with purplish flowers, while F. thymifolia may reach 2-3 m and their flowers change with age from white to purplish (Cervantes and Cuevas 2023).The different hybrids found show different flower colors as well as different reproductive fitness (C.Cervantes, unpubl.data).

Does gynodioecy facilitate hybridization?
Fuchsia microphylla and F. thymifolia, are two morphologically gynodioecious but functionally dioecious species and both are visited by Bombus eppiphiatus (Cuevas et al. 2014;Cervantes et al. 2018).Although the degree of overlap in flowering time varies every year, hermaphrodite plants of F. thymifolia begin to flower earlier than female plants.In F. microphylla the opposite pattern is observed; female plants blooms earlier than hermaphrodites (Cinthya Cervantes, unpubl. data).This pattern suggests that the direction of hybridization could be from F. thymifolia as a pollen donor to F. microphylla as a recipient; however, the genotypes found in the hybrids indicated a symmetric bidirectional hybridization.Some hybrids of later generations were misidentified as individuals belonging to the parental species, two of them with the maternal progenitor of F. thymifolia.Therefore, other factors might be promoting bidirectional hybridization.For example, the little or no nectar production in the hermaphrodite flowers of F. thymifolia (Cervantes et al. 2018) could encourage pollinators to visit these flowers in search of pollen and later, when searching for nectar, they probably visit the female flowers of both species as well as the hermaphrodite flowers of F. microphylla.
Finally, clonal reproduction and possible apomixis occur in F. microphylla and F. thymifolia (Cuevas et al. 2014;Cervantes et al. 2018;Cervantes and Cuevas 2023) and it might be present in hybrids (Cinthya Cervantes, pers comm.).For example, hybrids that showed identical genotypes and haplotypes were growing near each other.In the Garnica population, these phenomena may help to maintain the hybrid genotypes until they can reproduce successfully.Furthermore, it will be important to assess the reproductive biology of hybrids because most hybrids in Garnica are female individuals.

Conclusion
Natural hybridization between F. microphylla and F. thymifolia was confirmed with molecular and morphometric studies within the sympatric population of Garnica, and it seems that most of the hybrids are from later generations and that hybridization is bidirectional.Future phylogeographic analysis of both species will be important to understand more the hybridization between these closer species of Fuchsia.Indira Cervantes-Díaz et al. -Fuchsia hybridization in natural populations Table S3.Name and sequences of the forward and reverse primers of each of the markers used.Description of the PCR conditions for each reaction and references to the primers.
Indira Cervantes-Díaz et al. -Fuchsia hybridization in natural populations

Figure 1 .
Figure 1.Map showing the location of sampled populations of Fuchsia microphylla and F. thymifolia in the Transmexican volcanic belt.The red dot shows the allopatric populations of F. thymifolia, and the blue dots the allopatric populations of F. microphylla.Sympatric populations are represented by two colors (blue and red).The putative hybrids were collected in the Garnica population indicated with an asterisk.The acronyms used for each species and population are indicated in parentheses.

Figure 2 .
Figure 2. Principal coordinates analyses (PCoA) using microsatellite data of F. microphylla, F. thymifolia and their hybrids from different populations.A clear separation of the parental species and the intermediate position of the putative hybrids from the Garnica population is shown.

Figure 3 .
Figure 3. Bayesian assignment (K = 2) with structure for all the individuals of F. microphylla (blue), F. thymifolia (red) and the putative hybrids from Garnica population (between black bars); each vertical line represented one individual.The posterior probability for each individual was calculated with NewHybrids which indicates the possibility of belonging to the following categories: Pure species, F 1 , F 2 and backcrosses.For both analyzes the threshold value was ≥0.90.

(
Fig. 4a).For F. microphylla, only three haplotypes were present, Hap9 the most common and shared with some hybrids.Hap10 was exclusive to FmCB individuals and Hap11 only was found in FmG15 individuals.The other seven haplotypes were found in F. thymifolia individuals.Hap2 was the most common and the only haplotype shared between some hybrids and individuals of the FtG population.The allopatric populations FtL and FtT had one (Hap1) and two (Hap3 and Hap4) exclusive haplotypes, respectively.Each one of the three FtA individuals had a different haplotype (Hap5, Hap6,

Figure 4 .
Figure 4. Haplotype network constructed from (A) concatenated chloroplast sequences and (B) ITS nuclear sequences with F. microphylla (blue), F. thymifolia (red), and hybrids in green.Branches represent mutations, the filled circles represent inferred unsampled or extinct haplotypes, and the frequency of each haplotype is represented by the size of the circle.

Figure S1 .
Example of landmarks position in (a) petals, (b) floral tube, and (c) leaf.

Table 2 .
Genetic Indira Cervantes-Díaz et al. -Fuchsia hybridization in natural populations diversity parameters (means ± 1 SEM) for the studied populations of F. microphylla and F. thymifolia.Fm: F. microphylla, Ft: F. thymifolia; N: number of analyzed individuals; A: average number of alleles; A p : number of private alleles; H o : average observed heterozygosity; H e : average expected heterozygosity; HWE: loci in Hardy-Weinberg equilibrium; Fis: inbreeding coefficient.

Table 5 .
Genetic Indira Cervantes-Díaz et al. -Fuchsia hybridization in natural populations diversity in F. microphylla, F. thymifolia and putative hybrids.N: number of individuals analyzed for chloroplast and ITS respectively; h: number of haplotypes; H d: haplotype diversity; π: nucleotide diversity.