1 Introduction

Natural interspecific hybridization is an important evolutionary process and a common phenomenon in vascular plants implicated in ca. 40% of 282 families and 16% of 3212 genera (Whitney et al. 2010). Even though interspecific hybridization is common in vascular plants, it is concentrated in some groups, mainly polyploid, such as Orchidaceae and in some geographical locations (Atlantic, Mediterranean, and Alpine biogeographic regions) where hybridization hotspots were previously reported (Marques et al. 2017; Whitney et al. 2010). The evolutionary importance of this phenomenon resides in its possibility to enhance genetic diversity and species adaptation and may lead to speciation (Kerbs et al. 2017; Mallet 2007). Moreover, interspecific hybridization can arise without or with whole-genome duplication defined, in that second case as “allopolyploidization” and classified as one of the most common mechanisms of speciation in sympatric areas (Soltis and Soltis 2009).

In conifers, interspecific hybridization was shown to be more common than polyploidization except for Juniperus L., in which both phenomena are well represented (Ahuja 2005; Farhat et al. 2019a; Husband et al. 2013; Neale and Wheeler 2019).

Juniperus species are evergreen trees and shrubs of Cupressaceae family. This genus is considered as the most diversified inside its family. It is a monophyletic genus with approximately 75 species classified in three monophyletic sections: Caryocedrus Endlicher, Juniperus Spach, and Sabina (Miller) Spach (Adams 2014; Farjon 2005; Mao et al. 2010). They are widely distributed in the Northern Hemisphere except for J. procera Hochst. ex Endl., the only species distributed in the Southern Hemisphere (Adams 2014). This wide distribution of juniper species is mainly due to their tolerance to extreme environmental factors (Kukowski et al. 2013; Mathaux et al. 2016; Noble 1990; Rawat and Everson 2012), allowing them to adapt to diverse habitats. Indeed, species from this genus are found from the sea level until high altitudes, in forests and deserts, on rocky cliffs and sand dunes (Adams 2014; Díez-Garretas and Asensi 2014; Douaihy et al. 2012; Farjon 2005; Mathaux et al. 2016; Pinna et al. 2015). In addition, wide overlapping geographical distributions of Juniperus species and varieties occur (Adams 2014). Based on their morphology, potential hybrids between juniper species either with the same or different ploidy levels have been described in sympatric locations in Spain (Aparicio and Uribe-Echebarría 2006, 2008, 2009). They are supposed to be hybrids between J. sabina L. (2n = 2x) and J. phoenicea L. (2n = 2x) (Juniperus × herragudensis J.M. Aparicio & P.M. Uribe-Echebarría), between J. thurifera L. (2n = 4x) and J. phoenicea (2n = 2x) (Juniperus × palancianus J.M. Aparicio & P.M. Uribe-Echebarría), and between J. thurifera (2n = 4x) and J. sabina (2n = 2x) (Juniperus × cerropastorensis J.M. Aparicio & P.M. Uribe-Echebarría) (Aparicio and Uribe-Echebarría 2006, 2008, 2009). Besides, hybridization was commonly reported between closely related diploid juniper species in areas of sympatry in North America, as for example, between J. virginiana L. var. silicicola (Small) Silba and J. bermudiana L.; J. osteosperma (Torr.) Little and J. occidentalis Hook. (Adams and Wingate 2008; Terry et al. 2000). Recently, in the French Alps, based on morphological observations, hybridization between J. thurifera and J. phoenicea and between J. thurifera and J. sabina was suspected to occur (Garraud and Abdulhak 2014).

Six juniper taxa are native to the French Alps. Three taxa of section Juniperus: Juniperus communis L. subsp. communis and subsp. nana (Hook.) Syme, J. oxycedrus L., and the other three taxa belong to section Sabina: J. sabina, J. thurifera, and J. phoenicea (Lebreton et al. 2013). Their distribution in this region was well surveyed by the National Alpine Botanical Conservatory (CBNA). Interestingly, sympatric occurrence between the tetraploid J. thurifera and the diploid J. sabina was reported particularly in Saint Crépin forest (Garraud and Abdulhak 2014; Lebreton et al. 2013). These two dioecious species have a very different distribution and morphology. The Western Mediterranean J. thurifera is a tree with a pyramidal crown that often grows to 20 m in height, compared with the Eurasian J. sabina, which is a shrub less than 1 m in height except in Mongolia and Spain. However, in the Sierra Nevada, it forms a horizontal shrub (Adams 2014; Adams and Schwarzbach 2006; Perez-Latorre and Cabezudo 2009). In general, both species have slightly overlapping phenology at the end of the winter (Adams 2014). Remarkably, potential hybrids between these two species with atypical morphology were described in sympatry at Saint Crépin location (Garraud and Abdulhak 2014; Lebreton et al. 2013). However, no genetic data are yet available to confirm their hybrid status.

In this study, we aimed to investigate the occurrence of hybridization between the tetraploid J. thurifera and the diploid J. sabina in Saint Crépin forest.

Considering the ploidy level difference between the potential parental species, we determined by flow cytometry the ploidy level of putative hybrids and parents. Molecular DNA markers were also used to confirm the hybrid status as well as to gain a first insight into the possible occurrence of interspecific introgression through backcross. Since meiosis abnormalities are common in interspecific hybrids, we aimed to assess the regularity of the microsporogenesis in the putative hybrids. In this goal, an observation of pollen size and morphology was carried out. Pollen size was used as a proxy to discuss the potential diversity of male gamete cytotypes produced by the studied taxa.

2 Material and method

2.1 Studied area and plant material

The studied area was the Saint Crépin forest in the department of “Hautes-Alpes” in the South of France. The site’s substrate was limestone. Altitude ranged from 1100 to 1250 m a.s.l. with moderate slopes. Three putative hybrids (JTS8, JTS14, and JTS19) have been identified in this locality based on their shape. They are semi-prostrate shrubs with ascending branches: JTS8 is a female shrub of 50 cm tall, JTS14 is a male shrub of 80 cm in height, and JTS19 is a male shrub of 2 m in height (Fig. 1).

Fig. 1
figure 1

Putative hybrids at Saint Crépin forest. a Represents the semi prostrated shrub JTS8. b and c More focused pictures for JTS14 representing the erected edge of the shrub. d Shrub JTS19, showing its semi prostrated feature. e and f Global view of the tetraploid JT28 (e) and JT39 (f) with a clear “thurifera” architecture type

Leaf samples of 24 individuals of J. thurifera, 11 individuals of J. sabina, and the three putative hybrid individuals (JTS8, JTS14, and JTS19) were collected and directly preserved in silica gel before analysis. All samples were used for the AFLP study.

A sub-sampling was selected from the entire sampling for further detailed analyses comprising six individuals of J. thurifera (JT2, JT3, JT5, JT9, JT16, JT39), six individuals of J. sabina (JS1, JS4, JS13, JS17, JS22, JS33), and the three putative hybrids (Table 1; Fig. 2).

Table 1 Details of the individuals belonging to the sub-sampling, their genome size values (2C/pg), and the corresponding coefficient of variation (CV). Sample abbreviation corresponds to J. thurifera (JT), J. sabina (JS), and putative hybrids (JTS)
Fig. 2
figure 2

a, b Studied area at Saint Crépin forest in the department of “Hautes-Alpes” in the South of France. c Locations of the sub-sampling individuals at Saint Crépin, J. thurifera individuals represented by red circles, J. sabina individuals in blue rectangles, and their putative hybrids represented by green triangles

Male cones from three J. thurifera male trees (JT2, JT3, JT9), three J. sabina male shrubs (JS13, JS22, JS33), and the two male potential hybrids (JTS14, JTS19) were collected and dried.

Herbarium specimens of all the trees analyzed in this study have been conserved in the herbaria of the National Alpine Botanical Conservatory.

2.2 Genome size measurement

Preparation of samples belonging to the sub-sampling and flow cytometry analysis for genome size assessment were performed as described in Farhat et al. (2019b).

2.3 DNA extraction

Total genomic DNA was extracted by the cetyltrimethyl ammonium bromide (CTAB) method (Doyle and Doyle 1990) according to the modifications for conifers elaborated by Bou Dagher-Kharrat et al. (2007). Approximately, 30 mg of dried leaves were ground in a 2% CTAB solution. The DNA was quantified and diluted to a concentration of 50 ng/μL.

2.4 Polymerase chain reaction on cpDNA and nrDNA genetic markers

ITS and four chloroplast regions (petN/psbM; tmL/trnF; trnS/trnG; trnD/trnT) were amplified from extracted DNA of the sub-sampling. Each Polymerase chain reaction (PCR) reaction contained 10 pmol of each forward and reverse primer (Table 4 in Annex 2), 0.2 mM of each dNTPs, 1X DreamTaq Buffer (Thermo Scientific), 2.5 units of DreamTaq DNA Polymerase (Thermo Scientific), and 100 ng of genomic DNA in a final volume of 50 μL. PCR was performed using a Gene Amp PCR System 9700 set with an initial denaturation step at 95 °C for 3 min followed by 30 cycles at 95 °C for 30 s, annealing temperature depending on the primers (Table 4 in Annex 2) for 30 s, 72 °C for 1 min, and a final extension step at 72 °C for 10 min.

In order to determine the sequences of the haplotypes in the potential hybrids, PCR amplicons of ITS from these potential hybrids were cloned. The cloned amplicons were obtained by using the Taq DNA polymerase (MP Biomedicals). In each reaction, 100 ng of genomic DNA was added to 10 pmol of each ITS forward and reverse primer, 0.2 mM of each dNTPs, 1X Taq DNA polymerase buffer with 25 mM MgCl2, and 2.5 units of Taq DNA polymerase. The PCR was performed using the cycling parameters: initial denaturation at 94 °C for 1 min, followed by 27 cycles at 94 °C for 1 min, 61 °C for 1 min, 72 °C for 1 min, and a final extension at 72 °C for 5 min.

2.5 Cloning

Ligation of the fresh PCR product into pCR®2.1 vector (Thermo Scientific) was performed in a final volume of 10 μL following the manufacturer’s instructions. Then, 2 μL of the ligation product was added to 50 μL vial of frozen One Shot® Competent Cells (Thermo Scientific). The competent cells were transformed based on the manufacturer’s instructions. Transformed bacterial colonies containing the recombinant plasmid were grown overnight in 5 mL LB liquid medium with 100 μg/mL ampicillin at 37 °C on 225 rpm. Plasmid extraction of the cultured colonies was done using the MACHEREY-NAGEL Nucleospin Plasmid kit following the manufacturer’s protocol.

2.6 Sequencing and sequence analysis

Sanger sequencing was carried out at the GENEWIZ Genomics platform. All sequencing reactions were performed directly on PCR products except for the ITS clones of the three putative hybrids for which four clones of each individual were sequenced using M13 universal primers.

All sequencing reactions were performed on the forward and reverse directions. Consensus sequence construction and alignments were conducted under BioEdit software v7.2.6 using the global alignment algorithm.

The corresponding accession number of each studied sequence deposit in NCBI genebank is available in Table 5 in Annex 2.

2.7 AFLP analysis

The AFLP experiment was conducted according to Vos et al. (1995) with slight modifications. About 200 ng genomic DNA of each sampled individual was used for digestion and ligation. After pre-amplifications, two different primer combinations were used for the PCR selective amplifications. Adaptors and primers used for the AFLP study are given in Table 6 in Annex 2. Each forward primer used for the selective amplification step was labeled with 6-FAM modification at the 5′ extremity to allow fragment detection on the capillary sequencer (Applied Biosystems® 3730XL). AFLP migration profiles were analyzed using GeneMapper v.5 (Thermo Fisher Scientific) with the GS500 (− 250) LIZ control to scale for fragment size. Fragment selection was based on the calculation of the Bonin error rate (Bonin et al. 2004) using repetitions of control samples. All fragments that had an error > 0% were discarded.

The data were analyzed using AFLP-SURV v1.0 (Vekemans 2002) to estimate the within (percentage of the polymorphic locus (PLP)) and between (Fst) species genetic diversity. Allelic frequencies were computed on the basis of Hardy-Weinberg genotypic proportions expectations. The population genetic structure was analyzed using the population genetic model-based Bayesian clustering method implemented in STRUCTURE software v. 2.3.4. The estimated best value of K was determined by using ΔK statistics (Evanno et al. 2005) implemented on the web site STRUCTURE HARVESTER (http://taylor0.biology.ucla.edu/structureHarvester/). CLUMPP 1.1.2 (Jakobsson and Rosenberg 2007) was used in order to both homogenize label switching between runs and identify the possibility of multiple clustering solutions (due to multimodality in posterior probability distributions). Then, plots of individual genomic admixtures within inferred genetic clusters were built using STRUCTURE PLOT (http://omicsspeaks.com/strplot2/) (AFLP analysis was detailed in Annex 1).

AFLP raw data are available in Farhat et al. (2020) (https://doi.org/10.5061/dryad.h44j0zpgk).

2.8 Pollen analysis

Pollen staining according to Alexander (1969) was applied on pollen sampled from 10 male cones of each J. thurifera tree (JT2, JT3, JT9), from 5 male cones of each J. sabina shrub (JS13, JS22, JS33), and from 10 male cones of each putative hybrid (JTS14 and JTS19). Alexander stain was used in this study only as a way to ensure pollen hydration and not for the test of viability. Pollen grains were examined using a Zeiss Axiophot microscope and photographed using a highly sensitive CCD camera (RETIGA 200R, Princeton Instruments, Evry, France) and image analyzer (Metavue, Evry, France).

Around 300 pollen/individual were surveyed for their morphology and classified as normal or abnormal. Those classified as abnormal were collapsed, emptied, and aborted pollen grains. Owing to the spheroidal form of Juniperus pollen grains, with one aperture, only the diameter of the pollen grains was measured to determine their sizes. Diameters of around 100 pollen/cone for each studied individual were measured using the software ImageJ. One way ANOVA and Kruskal Wallis tests were carried out to test pollen size differences among J. sabina and J. thurifera.

The distribution of pollen size within each sample of J. sabina and J. thurifera trees as well as for JTS14 and JTS19 were used to assess the frequency of large pollen grains that may be considered as outliers relatively to a normal distribution. Because, the pollen size was continuously distributed, the following approach was used. First, the mean and standard deviation were estimated for each of the four observed distributions. Then, under the hypothesis of the normal distribution, a 1% threshold was used to identify pollen grains having a size bigger than expected. Finally, for each of the four analyzed pollen samples, the number of outlier pollen grains detected was compared with the expected one under the hypothesis of normal distribution at the 1% threshold by using a chi-square test (1 ddl).

Pollen size and morphology were used as criteria to assess the regularity of meiosis in the potential hybrids and their parental taxa.

3 Results

3.1 Genome size estimation of the parental species and their putative hybrids

The nuclear DNA content (2C value) of J. thurifera and J. sabina was 45.83 pg (σ = 1.04 pg) and 23.25 pg (σ = 0.25 pg), respectively (Table 1). Regarding the putative hybrid individuals, the 2C DNA values were very close to each other with 34.08 pg, 34.63 pg, and 35.35 pg for JTS8, JTS14, and JTS19, respectively, showing an intermediate genome size between J. sabina and J. thurifera (Table 1; Fig. 3).

Fig. 3
figure 3

Distribution of fluorescence intensity representing the DNA content on x-axis of J. sabina (a), the putative hybrid (b), and J. thurifera (c). The leaves of the three individuals belonging to J. sabina (JS13), J. thurifera (JT5), and the putative hybrid JTS14 were chopped simultaneously without the internal standard plant. Each peak corresponds to one of three ploidy level: 2n = 2x (a), 2n = 3x (b), and 2n = 4x (c)

3.2 Genetic differentiation of parental species and putative hybrids based on ITS and cpDNA sequences

No intra-specific polymorphism was detected in the chloroplast sequences studied within J. sabina and within J. thurifera. Chloroplast regions displayed a high number of fixed differences between the two species. However, the rate of nucleotide divergence between J. sabina and J. thurifera varied according to the chloroplast region (petN/psbM; trnL/trnF; trnS/trnG; trnD/trnT). All putative hybrids showed the same sequences as J. thurifera for the four chloroplast regions (Table 2).

Table 2 Molecular polymorphism observed between J. sabina, J. thurifera and the putative hybrids for the chloroplast sequences (petN/psbM, trnD/trnT, trnL/trnF, trnS/trnG)

Concerning the ITS region, 1008 bp were successfully sequenced. Very few intra-specific polymorphisms were detected (4 SNPs within J. sabina and 2 SNPs within J. thurifera). The ITS studied sequences displayed fixed differences between these two species: 13 SNPs and 3 indels. Each putative hybrid showed two haplotypes that correspond to non-recombinant haplotypes from J. sabina and J. thurifera (Table 3).

Table 3 Polymorphic sites observed within ITS sequences between J. sabina, J. thurifera and haplotypes found in the putative hybrids

3.3 Interspecific genetic admixture revealed by AFLP markers

In total, 147 polymorphic loci remained after the process of peak selection. Only one individual was discarded from the analysis because of its unreadable peak profile. A high polymorphism level was found for AFLP markers at Saint Crépin for both J. sabina (PLP = 55.1%) and J. thurifera (PLP = 82.31%). The genetic differentiation between the two species was relatively high, as expected for two different species (Fst = 0.17, p < 10−4) despite the fact that only one locus was fixed for alternative alleles. A unique population genetic pattern with two genetic clusters was inferred to be the best solution that fitted the observed data on the basis of the Bayesian clustering analysis (Fig. 7 in Annex 2).

The distribution of the dominant phenotype (fragment presence) frequency differences between J. sabina and J. thurifera was centered close to zero (mean = − 0.03; median = − 0.08; Fig. 8 in Annex 2) with a very slight bias towards a majority of locus having the dominant phenotype in J. thurifera. Inferences on individual genomic admixtures were therefore expected to be only slightly influenced (with an advantage to J. thurifera assignment), on average, by dominance effects. The genomic individual assignment pattern showed that the two inferred genetic clusters corresponded, as expected for distinct species, to J. sabina and J. thurifera (Fig. 4). However, the genomes of a few J. thurifera individuals were clearly admixed whereas admixture levels were nearly null for all J. sabina individuals. Besides, the three putative triploid hybrids were inferred to have an admixed genome. The mean proportion of their genome assigned to “sabina” genetic cluster was 11.7% for JTS8, 28.1% for JTS14, and 42.0% for JTS19. It was noticeable that two other individuals showed a clear admixed genomic pattern: JT39 and JT28 displayed 18.4% and 14.0% of their genome assigned to “sabina” genetic cluster, respectively. These two individuals were shown to be tetraploid, their morphology was “thurifera” like (Fig. 1) and they had J. thurifera haplotypes for both chloroplast and ITS sequences (data not shown).

Fig. 4
figure 4

Individual genomic assignment to each of the two genetic clusters (K = 2). JS stands for J. sabina and JT stands for J. thurifera. JTS is used to name the three putative triploid hybrids. Black arrows point, from left to right, to the two tetraploid individuals with high genomic admixture between “thurifera” and “sabina” genetic clusters and to the three putative triploid hybrids

3.4 Pollen characterization of Juniperus taxa

Pollen grains of juniper studied taxa are spherical and do not hold air sacs (Fig. 5). Morphological abnormalities of pollen grains have been observed in individuals of J. sabina and J. thurifera but at relatively low frequencies ca. 9.5% and ca. 0.9%, respectively. In contrast, pollen grain abnormalities were found more frequently in the two potential male hybrids: JTS14 (ca. 35.3%) (Fig. 5d) and JTS19 (32.5%).

Fig. 5
figure 5

Microphotographs of pollen grains (all pollen pictures are on the same scale size). a J. sabina pollen grains showing variability in pollen sizes. b J. thurifera pollen grains showing size variation representative of most pollen from this species. c J. thurifera pollen grain showing a large size, considered as possibly unreduced. d Aborted pollen grains (arrows) produced by the potential hybrid JTS14. e JTS14 pollen grains showing various sizes. f, g, and h show JTS19 pollen grains of different sizes. h Pollen grain with a large size and considered as possibly unreduced

A total of 1712 pollen grains was measured for J. sabina showing that the pollen sizes ranged from 13.6 to 35.6 μm with a mean of 21.3 μm (σ = 2.3 μm) (Fig. 6). Regarding J. thurifera, the size of 3171 pollen grains was measured and the pollen sizes ranged from 20 to 45.5 μm. The mean pollen size was 27.8 μm (σ = 3.2 μm) (Fig. 6). Both ANOVA (F = 48193, df = 1, p < 2e-16) and Kruskal Wallis (chi-squared = 2924, df = 1, p value < 2.2e-16) tests showed that the tetraploid species, J. thurifera, produced significantly bigger pollen grains (1.31 bigger in mean) than the diploid J. sabina. The pollen size distribution was clearly Gaussian for J. sabina but slightly skewed for J. thurifera due to large pollen grains (Fig. 9 in Annex 2). Indeed, the frequency of pollen grains showing large-diameter values in J. thurifera was approximately three times higher than expected (at 1% threshold) under a normal distribution (2.9%; χ2 = 113,65; 1ddl; p value < 10−7). Contrarily, the proportion of large pollen observed in J. sabina corresponded to the expectation under a normal distribution and at the 1% threshold (ca. 0.9%; χ2 = 0.82; 1ddl; not significant (NS)) (Figs. 5 and 6).

Fig. 6
figure 6

Box plot showing pollen grain size variability of J. sabina (N = 3 individuals and 1712 pollen grains), J. thurifera (N = 3 individuals and 3171 pollen grains) and the two male putative triploid hybrids JTS14 (686 pollen grains) and JTS19 (928 pollen grains)

For the two potential male hybrids, JTS14 and JTS19, 686 and 928 pollen grains were measured, respectively. For JTS14, pollen size varied from 17 to 44 μm (Fig. 6), the mean size was 25.4 μm (σ = 3.2 μm). The mean pollen size was therefore intermediate between those observed in the two parental species. JTS14 did not produce a significant frequency of pollen grains with large sizes at the 1% threshold (1.6%; χ2 = 2.31; 1ddl; NS). For JTS19, pollen size varied from 17.8 to 57 μm with a mean pollen size of 27 μm (σ = 5.5 μm) (Fig. 6), close to the value observed in J. thurifera. JTS19 presented a larger variance in pollen sizes than JTS14 and the two parental species (Fig. 9 in Annex 2). This is explained by the pollen size distribution that was clearly skewed in favor of large pollen (ca. 4.9%; χ2 = 145.4; 1ddl; p < 10−7) (Fig. 6). Pollen size measurements are summarized in Table 7 in Annex 2.

4 Discussion

4.1 Genome size and ploidy level diversity in the hybrid zone

In this study, estimated genome size showed that J. thurifera (45.83 pg, σ = 1.04 pg) studied trees had genome size approximately 2 fold bigger than the studied shrubs of J. sabina (23.25 pg, σ = 0.25 pg). This suggests that studied individuals of J. thurifera and J. sabina were tetraploid and diploid, respectively. These results broadly agree with previous studies of genome size and chromosome number for other populations of those two species reporting J. sabina as diploid (2n = 2x = 22 with ca. 23 pg) except for the tetraploid J. sabina var. balkanensis R. P. Adams and A. Tashev (Farhat et al. 2019a, 2019b) and J. thurifera as tetraploid (2n = 4x = 44 and ca. 41.2 pg/2C) (Romo et al. 2013; Vallès et al. 2015). The estimated genome sizes of the three putative hybrids were intermediate (mean 2C DNA content: 34.69 pg, σ = 0.64 pg) between those of the potential parents. It was demonstrated in Juniperus that genome size could be used as a reliable proxy for the ploidy level determination (Farhat et al. 2019a). Moreover, three classes of genome sizes were shown to correspond to three ploidy levels (diploidy, tetraploidy, and hexaploidy) (Farhat et al. 2019a). Based on these classes, the genome size estimated for the three putative hybrid individuals showed that they are most likely triploids. In this genus, few sporadic triploid cytotypes have already been found in some ornamental cultivars (Adams et al. 2019; Hall et al. 1973). However, this study brought the first evidence for natural triploid in the Juniperus genus and generally among all conifers found in the wild.

4.2 Hybridization patterns of Juniperus taxa in the population of Saint Crépin

It has been commonly assumed that hybridization between species with different ploidy levels is very limited (Coyne and Orr 2004). Nevertheless, if hybridization occurs, the establishment of those hybrids must overcome severe post-zygotic barriers such as unviability and sterility (Baack et al. 2015; Husband and Sabara 2004), particularly in triploids and odd ploidy levels.

Our results on ITS nuclear sequences, showing that the three morphologically atypical individuals (JTS8, JTS14, and JTS19) displayed two ITS haplotypes: one specific to J. sabina and the other specific to J. thurifera, proved that these three individuals are hybrids between these two species. This last conclusion was also strongly supported by AFLP markers that clearly showed the admixture genetic nature of these three hybrids. Obviously, estimated values of individual genomic admixture must be taken with great caution since estimation uncertainty is related to several factors: effect of locus sampling, estimation of allele frequencies within true genetic clusters, non-repeatability of detected peaks (since not all individuals were repeated in the experiment), dominance effects (although we showed in our case that dominance effects across locus were almost evenly distributed between the two species), and possible violations of assumptions underlying the genetic model used by STRUCTURE, especially the “no linkage disequilibrium” hypothesis. However, it was noticeable that the proportion of J. sabina genome, inferred from AFLP data, into two of the three triploid hybrids (JTS14 and JTS19) was relatively close to 33.3%. This value is expected in the case of a first-generation triploid hybrid and of an identical frequencies distribution of dominant phenotypes between the two species. Despite that JTS8 displayed only 11.8% of its genome assigned to “sabina” genetic cluster, it still may be a first-generation triploid hybrid. In that case, JTS8 had inherited by chance a majority of loci for which the dominant allele was inherited from J. thurifera parent, a situation that may bias the genomic assignment to “thurifera” cluster. In addition to these three triploid hybrids, we found two tetraploid trees (JT28 and JT39) displaying a large portion (> 10%) of their genome assigned to “sabina” cluster. This result suggested that these two individuals could be backcrossed progenies.

In conifers and more particularly in the Cupressaceae family, the chloroplast has been shown mainly to be paternally inherited (Adams 2019; Hipkins et al. 1994; Neale and Sederoff 1989; Sakaguchi et al. 2014). The fact that the three triploid hybrids had the same chloroplast DNA sequences of J. thurifera argues in favor that this species was the paternal parent, and therefore, J. sabina was the maternal parent. However, the paternal inheritance of the chloroplast genome in Juniperus should be checked in the near future to confirm this hypothesis.

Finally, the combination of genome size with molecular data supports that JTS14 and JTS19 at least, and possibly JTS8, are first-generation hybrids produced through the fertilization of a reduced female gamete (n = 1x = 11) of the diploid J. sabina by reduced pollen (n = 2x = 22) of the tetraploid J. thurifera.

Ancient hybridization between ancestral lineages of those two species was previously suggested leading to the tetraploid variety J. sabina var. balkanensis (Adams et al. 2016; Farhat et al. 2019b). Currently, J. thurifera has a West-Mediterranean range and is absent from the Balkans where J. sabina var. balkanensis mostly exists (Adams et al. 2018). Also, other putative hybrids displaying intermediate morphology between those two species were described in one sympatric area from Spain (Aparicio and Uribe-Echebarría 2009). These observations together with our results suggest that reproductive barriers between J. sabina and J. thurifera are incomplete, allowing interspecific hybridization, and even genetic introgression through further backcrossing, to occur. However, the question of whether gene flow between these two species at Saint Crépin is rare or has occurred regularly in the past is still opened. Moreover, overlapping in their flowering period is essential for gene flow occurrence. Unfortunately, no data is yet available on this issue in the French Alps. Investigations on the phenology of both species across years at Saint Crépin and other places of sympatry will enlighten this question.

4.3 Pollen production of the parental species and their triploid hybrids

In general, a strong positive correlation has been observed between pollen size and both genome size and ploidy level (Katsiotis and Forsberg 1995; Srisuwan et al. 2019). This was also the case in our study where the mean pollen diameter of the tetraploid J. thurifera was shown to be 1.3 fold bigger than the mean pollen diameter of the diploid J. sabina.

In this study, J. thurifera displayed pollen grains with a large diameter at significant frequencies (ca. 2.9%) which could witness the production of unreduced gametes in this tetraploid species. This was not the case for J. sabina. However, in this study, only three trees were included for this species. In conifers, unreduced gametes have been found only in the diploid Cupressus dupreziana A. Camus with a pollen diameter of ca. 38 μm (Pichot and El Maâtaoui 2000). Conversely, in angiosperms, the production of unreduced gametes has been noticed to be frequent and classified as a major mechanism leading to polyploidy (Otto and Whitton 2000; Ramsey and Schemske 1998). Interestingly, the potential production of unreduced gametes has been suggested to contribute to the high rate of polyploidy previously found in Juniperus (ca. 15%, the highest rate within conifers) (Farhat et al. 2019a). However, further work must be conducted to estimate more reliably the eventual presence of unreduced gametes in Juniperus species by measuring the genome size of pollen grains.

Yet, the large variance observed in pollen sizes of the hybrid JTS19, as well as the observation of pollen grains with a morphological abnormality in JTS14 and JTS19 (ca.35.3% and 32.5%, respectively), might reflect the irregular meiosis of these hybrids. Indeed, meiotic irregularities are well documented in interspecific hybrids, especially in those with odd ploidy levels such as triploids, which are often considered as sterile (Comai 2005; Giles 1941; Karlsdóttir et al. 2008). Interestingly, studied individuals of both parental species showed a relatively low frequency of morphological abnormal pollen grains, especially for the tetraploid J. thurifera (ca. 0.9%). In general, meiotic irregularities have been reported in polyploid individuals (Comai 2005; Van de Peer et al. 2017). However, the very low frequency of abnormal pollen grains found in this study for the tetraploid J. thurifera suggested that meiosis was regular and that this species may have undergone diploidization. This was consistent with the hypothesis that J. thurifera is considered as a paleopolyploid based on cytological approaches (Vallès et al. 2015). Further work must be conducted to confirm the production of partially reduced and unreduced gametes by the triploid hybrids using flow cytometry. Also, more research should be conducted to assess the viability of pollens produced by the hybrids and the parental species.

5 Conclusion

This study reported the first evidence of triploid hybrids between two junipers, J. sabina (2n = 2x) and J. thurifera (2n = 4x), that grow in sympatry at Saint Crépin forest of the French Alps. Also, it suggested that genetic introgression through backcrosses between triploid hybrids and parental species may have occurred. The considerable variation found in pollen size of studied individuals belonging to J. thurifera would propose their possibility for unreduced gametes production. The two male triploid hybrids showed evidence of abnormal pollen suggesting meiosis irregularities.

Additional investigations on genome size and genetic diversity of a larger sample of individuals in Saint Crépin forest and in other locations, where the two species have been reported to be in sympatry, are needed to reach a better insight into the existence of recurrent genetic introgressions. Further analysis of the phenology of J. sabina and J. thurifera in Saint Crépin is needed to strengthen assessments of opportunities for gene flow occurrence. This study opens new avenues towards studying the ecological and genetic consequences of genetic introgressions that might occur between those two species.