Targeted generation of polyploids in Hydrangea macrophylla through cross-based breeding

Up to now, diploid and triploid cultivars were reported for the ornamental crop Hydrangea macrophylla. Especially, the origin of triploids and their crossing behaviors are unknown, but the underlying mechanisms are highly relevant for breeding polyploids. By screening a cultivar collection, we identified diploid, triploid, tetraploid and even aneuploid H. macrophylla varieties. The pollen viability of triploids and tetraploids was comparable to that of diploids. Systematic crosses with these cultivars resulted in viable diploid, triploid, tetraploid and aneuploid offspring. Interestingly, crosses between diploids produced diploid and 0 or 1–94% triploid offspring, depending on the cultivars used as pollen parent. This finding suggests that specific diploids form unreduced pollen, either at low or high frequencies. In contrast, crosses of triploids with diploids or tetraploids produced many viable aneuploids, whose 2C DNA contents ranged between the parental 2C values. As expected, crosses between diploid and tetraploid individuals generated triploid offspring. Putative tetraploid plants were obtained at low frequencies in crosses between diploids and in interploid crosses of triploids with either diploid or tetraploid plants. The analysis of offspring populations indicated the production of 1n = 2x gametes for tetraploid plants, whereas triploids produced obviously reduced, aneuploid gametes with chromosome numbers ranging between haploid and diploid level. While euploid offspring grew normally, aneuploid plants showed mostly an abnormal development and a huge phenotypic variation within offspring populations, most likely due to the variation in chromosome numbers. Subsequent crosses with putative diploid, triploid and aneuploid offspring plants from interploid crosses resulted in viable offspring and germination rates ranging from 21 to 100%. The existence of diploids that form unreduced pollen and of tetraploids allows the targeted breeding of polyploid H. macrophylla. Different ploidy levels can be addressed by combining the appropriate crossing partners. In contrast to artificial polyploidization, cross-based polyploidization is easy, cheap and results in genetically variable offspring that allows the direct selection of more robust and stress tolerant polyploid varieties. Furthermore, the generation of polyploid H. macrophylla plants will favor interspecific breeding programs within the genus Hydrangea.


Background
Polyploid organisms contain more than two complete sets of chromosomes per cell nucleus. This often results in an increased fitness, higher yield, improved product quality and better tolerance to abiotic and biotic stresses in comparison with diploid relatives. Moreover, polyploid plants may produce larger and more beautiful flowers, tend to flower later or show elongated flowering periods compared to related diploids [1]. Thus, the production of polyploid plants is an important strategy of crop breeding, especially in ornamental crops.
Hydrangea macrophylla (THUNB.) SER. is an economically important ornamental crop, which belongs to the upmarket segment. This perennial, deciduous species produces attractive foliage and large, colorful inflorescences. Its cultivars are used for indoor and outdoor cultivation, for landscaping and for production of fresh and dried flowers for florists. Important traits in breeding of H. macrophylla are brightly colored flowers arranged in impressive inflorescences, attractive foliage, remontant flowering as well as compact growth. Further relevant traits are tolerance against biotic and abiotic stress and various characteristics that are important for efficient production. H. macrophylla is an obligate out-crossing species [2][3][4], and therefore highly heterozygous. Heterozygosity of the parental plants and genetic variability between the different crossing partners result in genotypic and phenotypic variation already in the F 1 generation. Thus, F 1 plants can be used for selection. Subsequent plant propagation occurs vegetatively through cuttings.
The existence of diploid and triploid cultivars in H. macrophylla has first been reported by Demilly et al. [5]. Diploid plants of H. macrophylla contain 2n = 2x = 36 chromosomes and show 2C DNA contents from 3.9 to 5.0 pg. In contrast, triploid plants have 2n = 3x = 54 chromosomes, and 2C DNA contents vary between 6.5 and 7.3 pg [6][7][8][9][10][11]. Recently, a cultivar has been found with a 2C DNA content of 8.9 pg, suggesting tetraploidy [12]. About 30 triploid cultivars were reported until now [7][8][9][10]. Some of these cultivars date back to a breeding program that has been undertaken in a research station in Wädenswil (Switzerland) in the middle of the twentieth century [13]. Based on flow cytometric and genotypic data, the Wädenswil pedigree of 26 H. macrophylla lacecap cultivars has been reconstructed [9]. This reconstruction showed that triploid cultivars were produced independently (and most likely unwittingly) in different crosses. Furthermore, this reconstruction revealed crosses between diploid and triploid plants that obviously resulted in viable offspring [9].
The breeding of polyploid H. macrophylla is of high interest. Artificial polyploidization has been successfully demonstrated for the diploid H. macrophylla cultivars 'Adria' and 'Libelle' and the triploid cultivars 'Blaumeise' and 'Nachtigall'. It resulted in tetraploid and hexaploid plants. While the tetraploid plants developed normally and were attractive as the diploid initial cultivars, the hexaploid plants showed growth depression, leaf deformation and hampered flower development [7]. Although artificial polyploidization was successfully applied in H. macrophylla, it is labor and cost intensive. Furthermore, it does not explain the high frequency of existing triploid cultivars of H. macrophylla.
Jones et al. [8] hypothesized that triploid H. macrophylla cultivars were obtained unwittingly by crossing diploids with a coincidentally existing tetraploid individual or by using triploids in crosses. The latter is possible, because triploid individuals of H. macrophylla are at least partially fertile in crosses when used as seed or pollen parent [8]. Spontaneous polyploidization was observed when the diploid H. macrophylla cultivars 'Princess Juliana' and 'Trophee' were crossed. This cross combination resulted in 94% triploids when 'Trophee' was used as pollen parent, whereas only diploids were produced in the reciprocal cross [11]. Due to the bimodal pollen size of 'Trophee', the author suggested that unreduced pollen might be causative for the production of these triploids.
The main process of triploid production is widely unknown, particularly with regard to interploid crosses between diploid and triploid plants of H. macrophylla. Furthermore, it is unknown whether triploids produce offspring populations that are suitable for cultivar selection. However, the knowledge about both of these issues will support the targeted breeding of triploid and even tetraploid H. macrophylla.
In this study, we realized extensive and systematic crosses between diploid, triploid and tetraploid individuals of H. macrophylla and analyzed the different F 1 populations by flow cytometry. In this way, we aimed to reveal i) to which extent crosses between cultivars of different ploidy levels are successful, ii) whether offspring from interploid crosses is viable and suitable for selection and further crossings, iii) which cross combinations allow the targeted generation of polyploids, iv) which ploidy (or aneuploidy) is existent in gametes produced by diploid, triploid and tetraploid crossing partners.
Based on our findings, we developed a systematic breeding strategy for the targeted generation of polyploids in H. macrophylla.
In order to reveal crossing strategies to generate polyploid H. macrophylla plants sexually, we selected 12 diploid, 6 triploid and 1 tetraploid H. macrophylla cultivar(s) for crossing experiments ( Table 1). The pollen viability of the parental plants ranged from 26 to 63% (mean 49 ± 13%) for diploids, from 17 to 58% (mean 36 ± 16%) for triploids and was 61% for the tetraploid cultivar (Table 1). Although pollen viability was more frequently reduced in triploid cultivars, significant differences between diploids and triploids were not observed (Student's t test, p = 0.096, α = 0.05). In total, all plants produced sufficient amounts of viable pollen and were suitable for subsequent crosses.
In total, we studied 44 cross combinations and 6 selfings. Each combination was represented by 1 to 7 hand crosses (120 crosses in total), which were performed in the years from 2013 to 2017. All crosses and selfings are summarized in Table 2. As expected, all selfings failed to produce seeds, which reflects the widespread self-incompatibility of H. macrophylla. In contrast, 25 out of 44 cross combinations between different cultivars succeeded, 8 were difficult and 11 combinations failed. Nine intraploid cross combinations between 6 different diploid cultivars were included as a control. Eight out of these 9 crosses (89%) were successful. In interploid crosses between diploids and triploids, 12 out of 16 cross combinations (75%) were successful irrespective of the cross direction. Crosses between diploid and tetraploid and between triploid and tetraploid plants were successful in 3 out of 7 (43%) and in 2 out of 4 (50%) combinations, respectively. In contrast, all 8 intraploid crosses between different triploids failed: 6 combinations yielded no seeds, and 2 combinations resulted in only 3 and 10 seeds (Table 2).
Subsequently, we determined the germination rates of seeds from 21 successful cross combinations. The results  Table 2 Summary of crosses. Crosses with more than 20 seeds were defined as successful, with 1-20 seeds as difficult and without seeds as not successful Cross type    Table 2. The germination rates of 7 intraploid crosses between diploids ranged from 45 to 81% (mean 59 ± 16%). For 11 interploid crosses between diploid and triploid plants, we determined germination rates from 28 to 82% (mean 49 ± 16%). Two crosses between diploid and tetraploid plants showed germination rates of 46 and 54%, while 34% germinating seeds were observed for a cross between a triploid and a tetraploid plant. The germination rates of intraploid and interploid crosses showed no significant differences (Student's t test, p = 0.137, α = 0.05). Hence, nearly all diploid, triploid and tetraploid cultivars used in these crosses were fertile. Interploid crosses between diploid, triploid and tetraploid plants were mostly successful and resulted in viable offspring. Thus, triploidy and tetraploidy caused per se no crossing barriers in H. macrophylla and triploid as well as tetraploid H. macrophylla cultivars can be used for crossings. However, general crossing barriers seem to exist for crosses between triploids.

Crosses between diploids result in diploid, triploid and tetraploid progenies
In order to study the ploidy level in plants of the next generation, we determined the 2C DNA content of F 1 plants. First, we analyzed 8 F 1 populations derived from intraploid crosses between diploids. These populations included 34 to 424 plants. The 2C values are summarized per population in Fig. 1 and presented for each plant per population in Figure S2. Two of these populations produced offspring plants with 2C DNA contents between 4.36 and 4.56 pg, on average 4.44 ± 0.04 pg (n = 54) and 4.48 ± 0.03 pg (n = 60), respectively. These 2C values indicated diploidy for all of these F 1 plants. In contrast, the other 6 diploid-diploid populations generated diploid plants as well as 1 to 319 plants with 2C DNA contents more than 6.0 pg, clearly exceeding the 2C DNA content of a diploid H. macrophylla plant. In these populations, the 2C DNA contents grouped into three discrete intervals: 4.10 to 4.67 pg (mean 4.44 ± 0.08 pg, n = 285), 6.35 to 6.89 pg (mean 6.57 ± 0.08 pg, n = 425) and 8.45 to 8.88 pg (mean 8.67 ± 0.22 pg, n = 3), indicating diploidy, triploidy and tetraploidy. As shown in Table 2, 2 out of these 6 populations included 1 to 2 putative triploid F 1 plants (1.4 and 3.3%), whereas 4 populations contained 32 to 317 putative triploid (50.0 to 91.2%) and sporadically even tetraploid F 1 plants ( Figure  S2). While 1 to 2 triploid and tetraploid F 1 plants might have arisen randomly, a fixed mechanism seems to be responsible for the development of high proportions of polyploid plants. These high rates of polyploid offspring were observed only when 'Dark Angel' or 'Hörnli' were used as pollen parents. In contrast, when these cultivars were used as seed parents, they produced diploid and 0 to 3.3% triploid offspring (Fig. 2, Figure S2). Hence, it seems that some H. macrophylla cultivars form partially unreduced gametes, primarily unreduced pollen, and that unreduced pollen is produced genotype-specifically either at low or high frequencies.
All progenies from crosses between diploids showed a normal development, independently of the ploidy level. Within populations, plants looked similar and differed less than between populations. A differentiation between diploid, triploid and tetraploid progenies was not possible by eye (Fig. 3a, b).

Crosses between diploids and tetraploids generate triploid offspring
In a wide range of plant species, triploid plants can be generated by crossing diploid with tetraploid individuals. Correspondingly, we studied the offspring of 4 cross combinations between 3 diploid H. macrophylla cultivars and the tetraploid cultivar 'Benelux'. Each population included 1 to 83 F 1 plants (Table 2). In total, 202 F 1 plants were analyzed by flow cytometry. The 2C DNA contents of these plants ranged from 6.08 to 7.18 pg, on average 6.59 ± 0.20 pg (Fig. 1, Figure S2). The majority of these plants had 2C values within the range of triploid H. macrophylla cultivars [9,12], and we suggested triploidy for all of these F 1 plants. Hence, crosses between tetraploid and diploid H. macrophylla individuals produced exclusively triploid plants as expected. Our results show that the tetraploid cultivar 'Benelux' performs a regular meiosis and produces maternally and paternally 1n = 2x gametes. Plant phenotyping was not done for these populations.

Crossing triploids with diploids or tetraploids result in predominantly aneuploid offspring
Triploid plants are often considered to be sterile due to disturbed meiosis. However, 75% of cross combinations between triploid and diploid H. macrophylla cultivars were successful. Subsequently, we studied the 2C DNA content of 217 and 373 F 1 plants, respectively, derived from 7 diploid x triploid and 8 triploid x diploid crosses. The 2C DNA contents of these F 1 plants ranged from 3.87 to 8.50 pg, on average 5.02 ± 0.55 pg. As summarized in Fig. 1 and shown for each plant per population in Figure S2, all combinations produced aneuploid progenies with differing 2C DNA contents. These 2C DNA contents ranged between the 2C values of the corresponding diploid and triploid parents. When the triploid cultivar was used as pollen parent, 2C DNA contents between 4.3 and 5.2 pg predominated. When the triploids were used as seed parent, less F 1 plants with 2C values between 4.3 and 4.7 pg and more F 1 plants with 2C values between 4.8 and 5.7 pg were generated (Fig. 1), suggesting an effect of the cross direction. Furthermore,  3 combinations generated also F 1 plants with 2C DNA contents above the triploid level (Fig. 1, Figure S2). Hence, triploids seem to perform an unbalanced separation of homologous chromosomes during meiosis, which obviously generates male as well as female gametes with varying numbers of chromosomes. 2C DNA contents above triploid level might be caused through partially produced, unreduced gametes or through gametes, which have enriched 2 and 3 homologous chromosomes by unbalanced separation. In contrast, 2C DNA contents below diploid level might be caused through gametes, which lack some chromosomes completely.
Although intraploid crosses between different triploids were generally not successful, we had obtained in total 10 F 1 plants from 4 hand crosses between 'Enziandom' and 'Zorro'. The 2C DNA content of these plants ranged from 5.15 to 7.03 pg (Fig. 1, Figure S2). Most plants showed 2C values below triploid level, suggesting that gametes with lower DNA contents might have a selective advantage in generating viable offspring. As mentioned above, triploid H. macrophylla individuals produce female as well as male gametes with varying numbers of chromosomes. Combinations of aneuploid gametes seem to result in even stronger genetic disturbance, which might explain the crossing barrier between triploid H. macrophylla plants.
Four different cross combinations between triploid and tetraploid plants resulted in total in 333 F 1 plants ( Table 2). The 2C DNA contents of these plants ranged from 5.51 to 9.05 pg, on average 7.22 ± 0.44 pg. Most of these F 1 plants showed 2C values between the triploid and tetraploid level indicating aneuploidy in the range between 54 and 72 chromosomes (Fig. 1, in detail Figure  S2). In accordance with the results observed for crosses between diploid and triploid plants, we suggest that aneuploidy is caused by the triploid parent due to gametes with varying numbers of chromosomes.
Most of the generated F 1 plants showed an abnormal growth, irrespective of whether the triploid plant was used as seed or pollen parent (Fig. 3c, d). Plants were stunted or showed abnormal elongated growth. The size, form, serration and color of leaves differed strongly between offspring plants. Some plants did not produce flowers, whereas other plants produced flowers with partially deformed floral organs. The phenotypic variation within these populations was clearly stronger compared to the F 1 populations derived from crosses between diploids or between diploids and tetraploids.

Interploid crosses generate fertile offspring
In order to test the fertility of offspring derived from interploid crosses, we performed test crosses of putative diploid, triploid and aneuploid F 1 plants with diploid H. macrophylla cultivars. For this purpose, we selected 23 F 1 plants from 5 (Table S1). The 2C DNA content of these F 1 plants ranged from 4.38 to 7.72 pg. Sixteen out of these 23 F 1 plants produced pollen, whose viability ranged from 4 to 43%, whereas 6 F 1 plants produced no pollen. All F 1 plants were used as seed parent. In case they produced pollen, also reciprocal crosses were performed. The cultivars 'Bläuling', 'Libelle', 'Baby Blue' and 'Sheila' were chosen as crossing partners. These cultivars are diploid and were successfully used in previous crosses. In total, we performed 62 crosses. As shown in Table S1, 31 (50%) crosses were successful, 14 difficult and 17 crosses failed. Crosses with plants derived from diploid-triploid and triploid-tetraploid crosses were at least partially successful, irrespective of whether or not the F 1 plants were euploid and independent of the cross direction. Neither variation in ancestry, 2C DNA content nor pollen viability could account for the success or failure of these crosses. Subsequently, we sowed between 24 and 250 seeds per successful cross. Each of these sowings produced viable seedlings. We determined germination rates ranging from 20.8 to 100.0% (Table S1). Although we did not phenotype theses progenies further, our results show that offspring plants from interploid crosses can be used successfully in further crossings.

Discussion
Until now, only the existence of diploid and triploid H. macrophylla cultivars was reported [5,[8][9][10]. The origin of these triploid H. macrophylla cultivars was unknown. The two general mechanisms for polyploidization are autopolyploidy and allopolyploidy. Autopolyploidy occurs through somatic doubling or more often by crosses within the same species involving unreduced gametes. In contrast, allopolyploid organisms result from hybridization between different species, followed by chromosome doubling [15]. Somatic doubling of diploid and triploid H. macrophylla cultivars was previously achieved through Trifluralin-mediated polyploidization, and tetraploid and hexaploid H. macrophylla plants were generated [7]. In the present study, we demonstrated that autopolyploid H. macrophylla plants can also be generated sexually following different crossing strategies. These strategies include tetraploid plants as well as plants that produce unreduced gametes. As reviewed by Kreiner et al. [16], the production of unreduced gametes has been found across widely disparate phyla. Most individuals of natural populations produce no or low frequencies of 0.1 to 2.0% of unreduced gametes, whereas a few individuals produce also high frequencies, i.e. more than 10%. In H. macrophylla, we made similar observations based on flow cytometric analyses of different offspring populations. We found either no, random or predominant polyploidization. Similar findings were also reported for crosses between the diploid H. macrophylla cultivars 'Trophee', 'Princess Juliana' and 'Zaunkönig' [11]. The cross 'Princess Juliana' x 'Trophee' generated 94% triploids, the reciprocal cross no triploids and the cross 'Zaunkönig' x 'Princess Juliana' 4% triploids [11]. Hence, the production of no, low or high frequencies of unreduced gametes seems to be widespread in cultivated H. macrophylla. Reciprocal crosses suggested that predominant polyploidization is depending on the pollen parent. This indicates the production of reduced female gametes and high frequencies of unreduced male gametes in these particular cultivars, namely 'Hörnli', 'Dark Angel' and 'Trophee'. However, the source for random polyploidization is unknown and might involve unreduced male and female gametes.
The molecular mechanism of unreduced gamete formation in H. macrophylla is completely unknown. As mentioned above, the production of high frequencies of unreduced gametes is cultivar-specific, and therefore genetically determined. Shared genetics between 'Trophee', 'Hörnli' and 'Dark Angel' seems to be unlikely. 'Trophee' was bred by the French breeder Lemoine in 1919, whereas 'Hörnli' was breed 1952 in Wädenswil, Switzerland, and 'Dark Angel' by the German breeder Engel in 2004. Thus, the phenomenon of unreduced pollen formation might have arisen several times independently or is more widespread in at least cultivated H. macrophylla.
Crosses with 'Trophee', 'Hörnli' and 'Dark Angel' as pollen parent resulted in different percentages of polyploidized offspring. Crosses with 'Dark Angel' as pollinizer yielded 50, 68 or 75% polyploid offspring in combination with different seed partners, whereas 'Hörnli' and 'Trophee' as pollinizers yielded 94% ( Fig. 2 and [11]). The reason for partial generation of diploid and polyploidized offspring is unknown. Self-fertilization of the seed partner might produce diploids. Although self-fertilization is unlikely due to the widespread selfsterility in H. macrophylla, we had excluded the risk of selfings by performing hand crosses with removal of the anthers of closed floral buds before pollination. Hence, 'Trophee', 'Hörnli' and 'Dark Angel' probably produce reduced as well as unreduced pollen simultaneously, resulting in diploid and triploid offspring. In order to determine whether the ratio of diploid and triploid offspring is affected by the pollen parent, the individual cross combination or by environmental parameters needs further investigations such as repeated crosses.
Besides diploid and triploid cultivars, we identified the first tetraploid H. macrophylla cultivar. The tetraploid cultivar 'Benelux' was bred in 1950 by D. Baardse [17]. This tetraploid cultivar might have developed by fusion of two unreduced gametes. We found 3 out of 6 cross combinations between diploids that produced single offspring plants with 2C DNA contents corresponding to the tetraploid level ( Figure S2). Furthermore, the crosses 'Zorro' x 'Choco Bleu' and 'Benelux' x 'Bela' showed that plants with tetraploid DNA level can also arise from crosses between triploid and diploid or tetraploid partners, although biased numbers of homologous chromosomes cannot be excluded in this offspring despite its clear tetraploid DNA content. A third way for the generation of tetraploids is the fusion of unreduced gametes of triploids with reduced gametes of diploids as found in other species [18], although we found no evidence for this possibility in our study. Hence, the tetraploid cultivar 'Benelux' might be the result of one of these seldom events with unwitting selection by the breeder.
In addition, we also found an aneuploid variety, 'Mücke II' [9], in our cultivar collection. This variety looked as normal as euploid cultivars. The origin of this aneuploid variety is unknown. Possibly, this variety originated from an interploid cross including a triploid parent and was selected unwittingly by the breeder. In contrast to many other plant species, triploid and even aneuploid H. macrophylla plants are fertile. No significant difference in pollen viability was observed between diploid and polyploid H. macrophylla individuals, although the pollen viability differed largely between genotypes. Similar findings were previously reported also for other diploid and triploid H. macrophylla cultivars. Jones et al. [8] determined pollen viabilities between 7 and 98%, on average 70%, for 42 diploid cultivars and between 25 and 85%, on average 63%, for 19 triploid cultivars. Alexander [19] found a range between 44 and 87% for 3 diploid and between 48 and 75% for 3 triploid cultivars, on average 65 and 61% for diploid and triploid cultivars. Viable offspring was obtained when triploids were used as seed parent or as pollen parent [8,19], indicating the production of viable male as well as female gametes in triploids.
Using triploid cultivars as parents, we produced a large number of aneuploid plants in this study. These aneuploids differed strongly in chromosome numbers according to their 2C DNA contents. This finding indicates that triploid individuals of H. macrophylla successfully perform meiosis, delivering one or two or -even more disturbed -no or three homologous chromosomes to a gamete as has been observed for example in maize [20]. Astonishingly, these aneuploid gametes were obviously viable and able to fertilize, resulting in the numerous aneuploid plants derived from different crosses. The aneuploid plants also were viable, grew for several years in the greenhouse and -to some extend -produced also seeds in crosses with diploid plants. It is still unknown, how these aneuploid plants balance the effect of biased chromosome numbers.
Most of the aneuploid plants showed an abnormal growth and were unattractive. From a breeder's point of view, crosses based on triploids and aneuploids are of limited use for cultivar selection. However, a few of these plants developed nearly normal and exhibited new, attractive characteristics of leaves or flowers, color and growth habit. Thus, interploid crosses can be a shortterm source for the creation of novelty. However, for complex breeding programs that comprise several generations, e.g. breeding for resistance or stress tolerance, the use and development of euploid plants with homogenous development seems more promising.
Moreover, the discovery of polyploid H. macrophylla plants and the knowledge about interploid crossing behaviors might support breeding programs regarding interspecific crosses. Interspecific crosses in the genus Hydrangea are highly interesting for breeders who aim to combine e.g. winter hardiness and brightly colored flowers. The genus Hydrangea includes more than 208 species varying in chromosome number (2n = 2x = 30, 34, 36 or 38 chromosomes) und 2C DNA content (1.95 to 5.00 pg). Most of these species are diploid, but some contain also triploid, tetraploid and/or hexaploid individuals [6,21,22]. Interspecific crosses using H. macrophylla as one parent succeeded already in combination with H. angustipetala and Dichroa febrifuga [23,24]. In contrast, crosses between H. macrophylla and H. paniculata, H. quercifolia or H. arborescens failed widely [25][26][27]. These failures might be caused by choosing unwittingly triploid individuals as parents [27] yielding in aneuploid offspring or by genetic or cytogenetic incompatibilities as suggested by Reed et al. [24,27]. However, crossing a tetraploid H. macrophylla plant with a tetraploid individual of another Hydrangea species might overcome this incompatibility and might result in allotetraploid hybrids.

Conclusions
In the present study, we demonstrated that diploid, triploid, tetraploid and even aneuploid H. macrophylla are mostly fertile and produce viable offspring in interploid crosses. Neither triploidy nor aneuploidy cause crossing barriers. Furthermore, we showed that targeted breeding of polyploid H. macrophylla plants is possible. We proved that triploid and even tetraploid offspring can be generated by crossing diploid individuals with diploids that form unreduced pollen, by crossing diploid with tetraploid individuals or by crossing triploids either with diploid or with tetraploid plants. Based on the flowcytometric data of the resulting F 1 populations of our study, we suggest that diploid H. macrophylla plants produce usually reduced female gametes, but reduced or unreduced male gametes depending on the cultivar. Triploid cultivars obviously form aneuploid male and female gametes with varying numbers of chromosomes, while tetraploid individuals generate reduced 1n = 2x female and male gametes.
Based on these findings, we conclude that targeted breeding of triploid H. macrophylla plants can be realized according to the above-mentioned crossing strategies. Furthermore, we assume that also controlled breeding of tetraploid H. macrophylla plants is possible by crossing either tetraploid with diploid individuals that develop unreduced pollen or later on by realizing crosses between tetraploids. Thus, targeted breeding of polyploid H. macrophylla offspring involving diploids with formation of unreduced pollen and tetraploid plants is possible. Tetraploid plants might be useful in interspecific crosses within the genus Hydrangea.

Plant material
In this study, we used 19 commercially available cultivars of H. macrophylla (Table 1) for crossing experiments. Plants of these cultivars were provided by the company Kötterheinrich-Hortensienkulturen, Lengerich, Germany. The plants were kept in 17 cm pots filled with Einheitserde® CL Hortensien blau and cultivated in frostfree greenhouses in Erfurt, Germany or in frost-free greenhouses of Kötterheinrich-Hortensienkulturen in Lengerich, Germany. Plants were fertilized with Univer-sol® blue 0.1% (Everris International BV) and irrigated as necessary. Each year in June or July, all plants were pruned. Plants in full bloom were used for crossings and pollen analyses.

Quantification of pollen viability
For each cultivar, 3 pollen mixes from non-decorative flowers were stained with fluorescein diacetate (FDA) solution as described by Behrend et al. [28]. Eighty-eight to 375 pollen clusters were analyzed per pollen mix. Green-fluorescent pollen kernels were recorded as viable. The pollen viability of diploids and triploids was statistically analyzed using two-tailed Student's t test with α = 0.05. internal standard. We analyzed 1 to 9 samples per parental plant and one sample per offspring plant.

Chromosome counting
Chromosomes of macerated root tips were prepared and stained with DAPI as described in detail by Hempel et al. [9]. Chromosome counts were made from at least 5 metaphase cells per cultivar.