2018 Volume 87 Issue 2 Pages 264-273
Double flower and hortensia (mophead) hydrangea (Hydrangea macrophylla (Thunb.) Ser.) traits are recessively inherited. Cross breeding of these traits in hydrangea is difficult because it takes about two years from crossing to flowering. In this study, we aimed to obtain DNA linkage markers that would allow accelerated selection of these traits. We used next-generation sequencing to comprehensively collect DNA sequences from the ‘Kirakiraboshi’ with a double flower and lacecap inflorescence and the ‘Frau Yoshimi’ with a single flower and hortensia inflorescence, and designed simple sequence repeat (SSR) primer pairs for map construction. We screened 768 SSR primer pairs in 93 F2 progeny derived from ‘Kirakiraboshi’ and ‘Frau Yoshimi’. We identified 147 loci, which were expanded to 18 linkage groups with a total map length of 980 cM. Linkage analysis identified that both the double flower trait from ‘Kirakiraboshi’ (dKira) and the hortensia trait from ‘Frau Yoshimi’ (hFrau) were located on linkage group KF_4. Detailed linkage analysis using 351 F2 progeny revealed a 34.8 cM map length between the two loci and identified two tightly linked SSR markers, STAB045 for dKira and HS071 for hFrau. Genetic analysis suggested that double flower and hortensia traits are each controlled by a single recessive gene. Together, the linkage map, SSR markers, and genetic information obtained in this study will be useful for future hydrangea breeding.
Hydrangea macrophylla (Thunberg) Seringe (= Hortensia macrophylla (Thunb.) H. Ohba and S. Akiyama f. macrophylla), commonly known as hydrangea, is native to Japan and has a long history as an ornamental garden plant. In recent years, hydrangea has become a popular Mother’s Day gift and demand has steadily increased. Cultivars with different types of flowers are produced every year. Hydrangeas have two types of flowers, decorative and non-decorative. These differ in the shape of the sepals, number of floral organs, positions on the inflorescence axes, and pedicel morphology (Uemachi et al., 2004, 2006). Decorative flowers have large ornamental sepals that attract pollinators, whereas non-decorative flowers have inconspicuous perianths and instead play a major role in seeds (Uemachi and Okumura, 2012). Decorative flowers have single and double types. Single flowers have only petaloid sepals, while in double flowers petals and stamens, as well as the petaloid sepals, are mutated to form petaloid organs.
The shapes of the inflorescences in hydrangeas can also be divided into two types: lacecap and hortensia (mophead). These are divided according to the position and number of decorative flowers in the inflorescence (Uemachi and Okumura, 2012). Lacecap is composed mainly of non-decorative flowers, with several decorative flowers located only on the periphery of the inflorescence. In contrast, the hortensia inflorescence is composed of many decorative and non-decorative flowers (Uemachi and Okumura, 2012). Because of their ornamental value, double flower and hortensia inflorescences are the most important breeding targets in hydrangeas.
‘Kirakiraboshi’ was obtained by Kodama et al. (2015) from a selfed progeny between line HK-01 with a double flower and lacecap inflorescence as the maternal parent and ‘Frau Yoshiko’ with a single flower and hortensia inflorescence as the paternal parent. HK-01 is a double flower breeding line derived from the double flower ‘Sumida-no-hanabi’ (= ‘Hanabi’). ‘Frau Yoshiko’, registered in 1993, is a cultivar derived from the cross between ‘Prima’ with a single flower and hortensia inflorescence and ‘Silver Edge’ with a single flower and hortensia inflorescence.
‘Kirakiraboshi’ with lacecap inflorescence produces large decorative picotee flowers, in which the base color of the sepals is reddish purple and the edge color is purple to white. The flower has 14 double petaloid sepals and the edge of the sepals is serrated (Fig. 1A; Kodama et al., 2015). We are currently trying to introduce the hortensia trait into ‘Kirakiraboshi’ by crossing hortensia inflorescence cultivars.
Parental cultivars used in this mapping study. A: ‘Kirakiraboshi’ (double flower and lacecap inflorescence). B: ‘Frau Yoshimi’ (single flower and hortensia inflorescence).
A genetic study suggests that the double flower trait in hydrangeas is controlled by a single recessive gene (Suyama et al., 2015). Crossing experiments using the double flower cultivars ‘Sumida-no-hanabi’ and ‘Jogasaki’ suggest that the genes controlling the double flower trait are different between cultivars (Suyama et al., 2015). Therefore, there are at least two genes associated with double flower formation in hydrangeas.
In the case of the hortensia trait, a single mutation in a lacecap cultivar leads to the replacement of partial inflorescences with decorative flowers on the upper nodes of the inflorescence axes (Uemachi et al., 2006). Uemachi and Okumura (2012) reported that hortensia inflorescence was a recessive characteristic controlled by a single major gene.
H. macrophylla is diploid (2n = 36; Cerbah et al., 2001), with a 2C value of 4.5 pg and an estimated genome size of 2.3 Gb (Zonneveld et al., 2005). Linkage maps are available for most ornamentals, including roses, chrysanthemums, carnations, and petunias (Rout and Mohapatra, 2006; Yagi, 2015). For hydrangeas, some simple sequence repeat (SSR) DNA markers for phylogenic analysis are available (Reed and Rinehart, 2007, 2009; Rinehart et al., 2006); however, no linkage maps have been reported to date. SSRs allow detection of polymorphisms in the number of repeats in 2- to 5-bp monomeric repeat units (Staub et al., 1996). Because SSRs are stable, hypervariable and behave in a co-dominant manner, they are ideal genetic markers for applications such as plant genome mapping linkage studies (Morgante and Olivieri, 1993). Moreover, because the hydrangea is perennial and takes about two years to flower, producing new cultivars is time- and labor-consuming. Introducing recessive traits, such as double flower and hortensia inflorescnece, is particularly difficult and obtaining markers associated with these traits would be very useful.
In this study, we sequenced ‘Kirakiraboshi’ and ‘Frau Yoshimi’ genomic DNA using next generation sequencing (NGS). ‘Frau Yoshimi’ with a single flower and hortensia inflorescence, derives from the cross between ‘Coral’ with a single flower and hortensia inflorescence and ‘Silver Edge’ and is the half-sib cultivar of ‘Frau Yoshiko’. ‘Frau Yoshimi’ was used as the mapping parent because sufficient F2 progeny were obtained. We designed new SSR primers and used these, together with previously reported primers, to genotype 93 F2 progenies derived from crosses of the two cultivars. We then constructed a linkage map for hydrangeas and used this map to identify markers associated with double flower and hortensia inflorescence traits.
F1 crosses were made between ‘Kirakiraboshi’ (seed parent) and ‘Frau Yoshimi’ (pollen parent; Fig. 1) at Tochigi Prefectural Agricultural Experiment Station in May 2011 and 2012, resulting in a total of 12 F1 progeny with single flowers and lacecap inflorescence. In May 2013, one of these progeny was selfed to produce the 351 F2 progeny. In April 2015, phenotypic observations were made and 93 of these F2 plants were used for map construction. Young leaves from both parents, and F1 and F2 progenies were sampled in June 2014. Samples were stored at −80°C until required. Total DNA was extracted using the cetyl-trimethylammonium bromide (CTAB) method (Murray and Thompson, 1980) and used as PCR templates.
NGS library construction and DNA sequencingNGS libraries were constructed using 2 μg of DNA and a TruSeq DNA PCR-Free Sample Prep Kit (Illumina, USA), according to the manufacturer’s protocol. DNA was sheared using an ultrasonicator (Covaris M220; Covaris, USA) to produce libraries with an average size of 550 bp. A MiSeq system (Illumina) was used to obtain 300-bp sequence reads for ‘Kirakiraboshi’ and 250-bp paired-end sequence reads for ‘Frau Yoshimi’. All sequence data were assembled using CLC Genomics Workbench ver. 6.5 (CLC bio, Denmark) and default settings.
SSR primer designSSR motifs were detected using the MISA tool (http://pgrc.ipk-gatersleben.de/misa/). SSRs were identified as motifs containing 2–5 nucleotides and a minimum of five contiguous repeat units. For the identification of compound SSRs, 100 bp was selected as the minimum distance between any two SSRs. All other parameters were set to default. BatchPrimer3 (You et al., 2008) was used to design primer pairs; expected PCR products were limited to 120–200 bp, and all other parameters were set to default. Markers identified in this study are prefixed with “HS”. Twenty-five H. macrophylla SSR motif-containing sequences registered in NCBI (accession numbers FJ971640–FJ971642, FJ971644–FJ971659, DQ521439, DQ521440, DQ521442, DQ521444, DQ521445, and DQ521450) were also used to design SSR primers. These markers are prefixed with “STAB”.
Fragment analysisPCR was performed using the M13-tailed primer method (Schuelke, 2000), with minor modifications. SSR forward primers were modified by 5'-concatenation with the M13 reverse sequence (caggaaacagctatgacc). We also labeled the M13 reverse sequence primers with FAM or NED fluorescent dyes (M13r-FAM, M13r-Ned) and the reverse primer was extended with a pigtail (gtttctt) sequence for amplification stabilization (Brownstein et al., 1996). PCR amplification was carried out in 10 μL of solution containing 10 ng of the extracted genomic DNA, 0.4 pmol of the M13-tailed primer, 1.6 pmol of M13r-FAM or M13r-Ned, 1.6 pmol of each unlabeled reverse primer, 200 μM dNTPs, 1× reaction buffer (Takara Bio, Japan), and 0.2 U Ex Taq polymerase (Takara Bio). Amplification was performed as follows: 94°C for 5 min, followed by 30 cycles at 94°C for 30 s, 56°C for 90 s, and 72°C for 90 s, followed by 8 cycles at 94°C for 30 s, 53°C for 90 s, and 72°C for 90 s, with a final extension at 72°C for 10 min. Markers showing polymorphisms were genotyped using a Multiplex PCR kit (Qiagen, Germany) for F2 populations. The allele sizes obtained using this primer method contained about 18 bp of the tailed sequences. Amplified PCR products were separated and detected using an ABI PRISM 3730 Genetic Analyzer (Applied Biosystems, USA). The sizes of the amplified bands were determined using a DNA internal standard (400HD-ROX; Applied Biosystems) and Genemarker v2.6.3 software (Soft Genetics, USA).
Linkage map constructionLinkage analyses were conducted using JoinMap 4.0 software (Kyazma, Netherlands). The Kosambi map function was used to calculate genetic distances between markers, and the ML (MaximumLikelihood) method was used as a mapping algorithm. Minimum LOD scores of 4 were used for map construction. The linkage groups were displayed using version 2.2 of MapChart (Voorrips, 2002). Mapped SSR markers and the neighboring sequences we found were registered in DDBJ as LC225346–LC225572.
Fitting rates for the nearest markers in F2 populationsTo evaluate the fitting ratios of adjacent markers in the F2 populations, PCR products were analyzed by denaturing polyacrylamide gel electrophoresis. The gel composition was 6% acrylamide (acrylamide:N,N′-methylenebisacrylamide = 19:1), 1 × TBE buffer, 7M urea, 30% formamide, 0.1% ammonium persulfate, N,N,N′,N′-tetramethylethylenediamine (TEMED). Electrophoresis was performed at a constant voltage of 250 V in 1 × TBE buffer. Gels were stained with SYBR gold and visualized using a Pharos FX system (BioRad, USA).
Using a Miseq system, 50 million paired-end (PE) reads, corresponding to 13.1 Gb, were obtained for ‘Kirakiraboshi’. Sequences were assembled, and were equivalent to 96061 contigs. The N50 values of the contigs were 1788 bp and the longest contigs were 19 kb. A total of 5.3 million PE reads were obtained for ‘Frau Yoshimi’ (1.6 Gb), which was insufficient for computational assembly.
SSR primer designThe ‘Kirakiraboshi’ contigs were screened for SSR motifs. A total of 1121 SSR primer pairs, having 2- to 5-bp repeat units and more than five repeats, were selected using the MISA software. Of these 1121 primers, 71 primer pairs were used for subsequent analysis. To increase the likelihood of the SSR primer pairs identifying polymorphisms between the parent cultivars, ‘Frau Yoshimi’ sequences were mapped to ‘Kirakiraboshi’ contigs as reference sequences using CLC Genomics Workbench ver. 6.5 (CLC bio, Denmark) as the default setting. In total, 672 SSR primer pairs were computationally predicted to detect polymorphisms between the parent cultivars, and were used for subsequent screening. A further 25 SSR primer pairs, designed from registered sequences in NCBI, were also used. Of these 25 primers, five primer pairs (STAB045, STAB061, STAB125, STAB321, and STAB429) had been reported previously (Reed and Rinehart, 2009; Rinehart et al., 2006). In total, 768 SSR primer pairs were used for screening the mapping population in this study.
Screening of SSR primer pairsOf the 768 SSR primer pairs, 226 SSR markers detected polymorphisms between the parent cultivars. The remaining primer pairs either detected no polymorphisms, detected an identical allele between the parents that was not inherited in the F1 line, detected multiple loci, or resulted in poor product amplification. Using the 226 primer pairs that detected polymorphisms between the parent cultivars, 157 loci were genotyped in the F2 population. Of these, 110 showed co-dominant inheritance and the other 47 were dominant because of their unclear fragment peaks or because of overlapping peaks caused by multiplexed PCR.
Linkage map constructionWe constructed a hydrangea linkage map using minimum LOD scores of 4. The map consisted of 147 loci including 104 co-dominant and 43 dominant loci expanded to 18 linkage groups (Fig. 2; Table 1; Table S1). Ten loci were ungrouped in this study. The total map length was 980.0 cM, with an average distance of 6.8 cM. The lengths in each linkage group ranged in size from 128.0 cM (LG KF_1) to 3.9 cM (LG KF_18), with an average length of 54.4 cM. The largest number of loci was 27 in LG KF_1, and the smallest was two in KF_17 (Table 2). The largest distance between loci was 37.8 cM, on LG KF_9 between HS373 and HS141 (Fig. 2). Each of the following loci pairs (HS213 and HS535 in LG KF_1, STAB239 and HS042 in KF_8, HS330 and HS300 in KF_12, HS269, and HS633 in KF_13) were located on the same locus. The average loci interval was ranged from 17.1 (LG KF_9) to 1.0 (LG KF_18). Ninety-three loci segregated according to expected 1:2:1 or 3:1 ratios. Another 6 (P < 0.05), 3 (P < 0.01), and 45 (P < 0.001) loci deviated from these ratios.
Genetic linkage map of hydrangeas derived from the 93 F2 progeny of ‘Kirakiraboshi’ × ‘Frau Yoshimi’. Genetic distances (cM) and SSR loci are listed on the left and right sides of each linkage group, respectively. Distorted segregation is indicated by a significant P value in the chi-squared test: *P < 0.05; **P < 0.01; ***P < 0.001. Dominant-loci are underlined. The loci for double flowers and hortensia inflorescence are shown as dKira and hFrau, respectively.
List of mapped SSR loci on KF_4.
Summary of mapped SSR loci.
Of the 93 mapped F2 progeny, 82 plants were evaluated for their phenotypic appearances after flower opening. Of these plants, 60 had single flowers and 22 had double flowers, while 77 had lacecap and 5 had hortensia flower architecture. The remaining 11 plants died and could not be examined. Segregation of the double flower trait was fitted to the expected ratio (single:double = 3:1, P = 0.70 > 0.05). Previous studies suggested that both traits are controlled by a single recessive gene (Suyama et al., 2015; Uemachi and Okumura, 2012); however, segregation of the hortensia trait in our F2 population did not fit the expected ratio (P = 0.000 < 0.01). We first mapped both traits as markers according to the previous studies and using linkage analysis identified that both traits were located on LG KF_4. The double flower locus of ‘Kirakiraboshi’ (dkira) was mapped between the HS350 (furthest from 11.3 cM) and STAB045 markers (furthest from 1.1 cM; Fig. 2). The hortensia inflorescence locus of ‘Frau Yoshimi’ (hFrau) was located between the HS314 (furthest from 2.4 cM) and the HS527 markers (furthest from 11.8 cM; Fig. 2). To examine the suitability of neighboring markers as selection markers, genotyping and phenotyping were conducted in the remaining F2 progeny.
Validation of linked markersOf the 351 F2 progeny, including the 93 mapping progeny, 217 (61.8%) had a single flower and lacecap inflorescence phenotype, 104 (29.6%) had a double and lacecap phenotype, 24 (6.8%) had a single and hortensia phenotype, and 6 (1.7%) had a double and hortensia phenotype (Table 3). Categorizing these plants by flower phenotype, they segregated into 241 (68.7%) plants with single flowers and 110 (31.3%) plants with double flowers (Table 3). In terms of inflorescence architecture, 321 (91.5%) progeny segregated into the lacecap phenotype and 30 (8.5%) into the hortensia phenotype. However, chi-squared analyses of the hypotheses that each trait is controlled by a recessive gene found that the expanded F2 progeny did not meet the expected 3:1 phenotype ratios (chi-squared scores obtained were 7.5 and 50.6 for double flowers and hortensia, respectively).
Segregations of decorative flower and inflorescence types in the 12 F1 and 351 F2 progeny derived from ‘Kirakiraboshi’ and ‘Frau Yoshimi’.
To examine the effectiveness of markers adjacent to the locus of interest, markers linked to both traits were used for genotyping (HS436, HS350, STAB045, HS224, HS527, and HS071 shown in Table 1). The HS314 and HS423-2 markers were excluded because they had dominant, complicated fragmental patterns and unstable amplification. The STAB045 marker, which is most adjacent to the dKira locus, detected 108 of the 110 progeny with double flowers that had the ‘Kirakiraboshi’ homozygous allele (Table 4). Only 3 of the 241 single progeny were homozygous for the double flower ‘Kirakiraboshi’ allele, and these were “recombinants” that had differences between their genotypes and phenotypes. The STAB045 recombinant rate was lower (1.4%) than the HS224 (13.1%), HS350 (18.2%), and HS436 (21.1%) marker recombination rates. As for the inflorescence type, lacecap was the dominant trait. Only 1 of the hortensia progeny was heterozygous for the ‘Frau Yoshimi’ hortensia allele, at the HS071 marker locus. The HS071 marker recombination rate was 3.7%, which is lower than that of the HS527 marker (6.6%).
Fitting of adjacent markers to double flower and hortensia traits in the 351 F2 progeny.
Segregation of double flower and hortensia inflorecence in the F2 populations was associated with the presence or absence of a single marker. Therefore, we concluded that double and hortensia traits are each controlled by a single recessive gene. We considered both traits as trait markers and constructed a detailed linkage map using the genotyping data from the 351 F2 progeny (Fig. 3). The map length of 70.6 cM from HS436 to HS071 on LG KF_4 was reduced to 58.8 cM by detailed mapping, and the map positions of STAB045 and dKira were exchanged (Fig. 3). The map length between HS071 and hFrau (32.3 cM) was reduced to 5.1 cM. The double flower trait (dKira) was mapped in the center of LG KF_4, separated from STAB045 by 1.3 cM (Fig. 3). Hortensia trait (hFrau) was also mapped to LG KF_4, separated from HS071 by 5.1 cM. All markers showed significantly skewed segregation ratios of 1:2:1 at the 0.01% level.
Detailed linkage map of group KF_4 derived from the 351 F2 population of ‘Kirakiraboshi’ × ‘Frau Yoshimi’. Genetic distances (cM) and SSR loci are listed on the left and right sides of each linkage group, respectively. The loci for double flowers and hortensia inflorescence are shown as dKira and hFrau, respectively.
In this study, we constructed the first linkage map for the Hydrangea. The linkage map consisted of 147 loci expanded to 980 cM, and comprised 18 linkage groups (Fig. 2). The average interval between loci was 6.8 cM. The number of mapped loci ranged from 2–27, and the length in each linkage group ranged from 3.9 cM to 128.0 cM. Many large gaps existed within this linkage map. Eighteen linkage groups were consistent with the basic chromosome number of Hydrangeas (x = 18), but also included many small linkage groups (LG KF_9 to LG KF_18). Many more markers would be needed to produce a linkage map covering the entire genome. However, we believe that we mapped an extremely large number of SSR markers in hydrangeas, especially given the dearth of available genomic sequence data (including SSRs) before we began the present study. This is the first step in hydrangea genomic analysis. Linkage analysis using NGS technology is rapidly conducted these days; however, genomic research in ornamental plants has fallen behind that of other crop plants. Nonetheless, transcriptome and whole genome sequence reports are still steadily increasing for ornamentals (Yagi, 2015). Recently, we successfully constructed a detailed genetic linkage map using restriction site associated DNA (RAD)-sequence analysis using NGS technology in the carnation (Yagi et al., 2017). We are currently applying the same approach to hydrangeas. Saturated linkage maps are more useful for QTL analysis and map-based gene cloning for important traits such as flowering times and disease resistance.
In this study, the double flower and hortensia traits did not fit the expected ratio for monogenic traits (3:1), in contrast with previous reports (Suyama et al., 2015; Uemachi and Okumura, 2012). In the F2 population, the ratio of double to single flowers was about 1:2, while for hortensia and lacecap architectures the ratio was about 1:10 (Table 3). Linkage analysis revealed that double flower and hortensia traits were mapped on the same linkage group, KF_4, with a map length of 34.8 cM between loci (Figs. 2 and 3). We also identified tightly-linked co-dominant SSR markers (STAB045 and HS071) for 2 important traits. The recombinant ratios of the nearest identified markers were very low (1.4% and 3.7%; Table 4). These results suggest that segregation of the double flower and hortensia traits are each associated with a single marker. We concluded that double flower and hortensia traits are controlled by a recessive single gene, in agreement with previous reports (Suyama et al., 2015; Uemachi and Okumura, 2012). The reason for the difference in the segregation ratio compared with previous reports is less clear. Both traits are relatively distinguishable, so the methods of observation and criteria for trait classifications would not have such large effects on the results. More than 35 progeny were analyzed for each cross in previous reports (Suyama et al., 2015; Uemachi and Okumura, 2012), so the tested numbers were also not so large. The reason for the skewed segregation in our study is unclear. Segregation distortion is a common phenomenon in many plants and is recognized as a potentially powerful evolutionary force (Li et al., 2010; Li et al., 2015); however, its underlying mechanism is not understood (Cai et al., 2015). Suggested causes include aneuploidy, chromosomal translocation, competition among gametes, and the inheritance of alleles affecting the viability of the zygote, embryo, or seedling (Lashermes et al., 2001). Another possibility is that a transcription factor controls both traits or induces a factor for recombination in this region. Segregation distortion can also arise because of conscious or unconscious selection by researchers during the mapping of populations (Li et al., 2015). Further analysis is required to clarify the cause of the skewed segregation in our study. We also intend to test other F2 populations and perform backcrosses to examine the heredity of double flower and hortensia traits.
All 12 F1 progeny obtained by crossing ‘Kirakiraboshi’ and ‘Frau Yoshimi’ had single flowers and lacecap architecture (Table 3), while more than 300 F1 progeny obtained from crossing ‘Kirakiraboshi’ with three single flower and hortensia cultivars also had the single flower and lacecap phenotype (data not shown). Combined, these results show that single flower and lacecap traits are dominant traits, while double flower and hortensia traits are recessive in hydrangeas. This study also revealed that ‘Kirakiraboshi’ is homozygous for double flower and lacecap alleles, and ‘Frau Yoshimi’ is homozygous for single flower and hortensia alleles. We are currently trying to introduce the hortensia trait into ‘Kirakiraboshi’, which has large double flowers, by crossing it with ‘Frau Yoshimi’ to develop a hydrangea cultivar with an enhanced appearance. The availability of markers near the loci of interest will be very useful for early marker-assisted selection. The percentage of progeny with the hortensia trait was low in this study because of skewed segregation, and only 1.7% of the F2 progeny (recombinants) had both double flower and hortensia traits (Table 3). These low numbers illustrate the value of tightly-linked DNA markers for selection of desired traits. We are currently attempting to determine the availability of tightly-linked markers in other cultivars and breeding lines. Suyama et al. (2015) suggested that at least two genes controlling double flowers exist in H. macrophylla. We are developing segregation populations using another major double flower, ‘Jogasaki’, to evaluate the availability of the STAB045 marker.
Double flowers, characterized by excessive development of petals, are among the most important traits of ornamental flowering species. Double flowers are often preferred by consumers because they are larger, more floriferous, and more showy than single flowers. A double flower is a dominant trait in carnations (Scovel et al., 1998; Yagi et al., 2014), petunias (Liu et al., 2016), and roses (Crespel et al., 2002). In Matthiola incana (stock), a double flower is a recessive trait, and the allele for the double-flower phenotype is closely linked to a lethal allele (Ecker et al., 1993). In ornamentals, 4 morphological changes that induce double flowers are recognized: (1) conversion of the stamens and carpels into petals; (2) a simultaneous increase in petals and other floral organs including stamens, sepals, and carpels (carnations, petunias, and garden balsam); (3) an increase in the number of ray florets in the capitulum (Asteraceae plants including chrysanthemums, zinnia, and dahlias); and (4) paracorolla development (snapdragons; Nishijima, 2012). Conversion of the stamens and carpels into petals (1) is the most common morphological change and is observed in various floricultural plants, including roses, stock, peonies, and the Japanese morning glory (Saito, 1959). The hydrangea ‘Kirakiraboshi’ has petaloid sepals and stamens that account for its double flower phenotype. Molecular and genetic studies have been carried out using the model plants Arabidopsis thaliana and Antirrium majus to explain this floral development. A classical ABC model has been proposed (Coen and Meyerowitz, 1991; Weigel and Meyerowitz, 1994), in which a combination of 3 gene functions specify the 4 floral organs. The class C mutant in A. thaliana produces double flowers, in which the stamens are converted to petals and carpels are converted to a new flower. Petunias express two such C-class genes, FLORAL BINDING PROTEIN6 (FBP6) and PETUNIA MADS BOX GENE3 (PMADS3). Reduced expression of these genes by PMADS3-RNAi in a mutant fbp6 background results in a double-flowered phenotype (Heijmans et al., 2012). These facts suggest that the gene responsible for controlling dKira is associated with C-class mutations. Linkage mapping and genetic information, including the STAB045 marker, would be useful for the isolation and functional analysis of C-class genes.
There is little information about the physiological and genetic mechanisms underpinning the hortensia phenotype. Uemachi and Okumura (2012) suggested that the appearance of the hortensia phenotype in lacecap populations is not through a stepwise increase in the numbers of decorative flowers, but is rather a qualitative change. They reported a hortensia-inducing mutation in the lacecap inflorescence of the ‘Blue Sky’ hydrangea cultivar. This mutation arose through the insertion of a long terminal repeat (LTR) retrotransposon into the locus controlling inflorescence type.
To reveal the genes associated with double flower and hortensia traits, further studies are required. Such studies include constructing a high density genetic linkage map of hydrangeas, constructing a bacterial artificial chromosome library, and gene expression analyses. Identification of the genes responsible for desired floral traits will provide new technological advances for flower enhancement. The availability of mapped SSR markers and genome sequences such as those obtained in this study will help with such analyses. Moreover, Hydrangeas have many interesting traits such as the extent of the flower color change (response to soil pH or metal compounds in the flower sepal), flower differentiation and development (as seen in the perpetual flowering of Hydrangea arborescens ‘Annabelle’), the shape of the inflorescence (such as the “pyramid” shape of Hydrangea quercifolia), flower greening (natural greening in decorative flowers after flowering or infection with phytoplasma), climbing traits (as seen in Hydrangea petiolaris and Schizophragma hydrangeoides), fragrance (as seen in Hydrangea quercifolia), and hydrangea tea. The SSR markers developed in this study will be useful for the genetic analysis of these important traits, particularly in plants of the genus Hydrangea.
