Introduction

Rose is one of the most commonly cultivated ornamental plants in the world. Rose production, encompassing cut flower, garden and potted plants, is increasing worldwide. Most rapidly expanding among these is the cut rose production, which has grown from 5,370 ha in 1994 (Gudin 2000) to about 8,500 ha in 2003 (Blom and Tsujita 2003). Despite its economical value, little is known about the inheritance of phenotypic traits in the genus Rosa, to which all roses belong. Nevertheless, this genus presents a wide range of phenotypic variability and a high level of genetic heterozygosity. The genetic basis of cultivated roses is narrow in comparison with the vast spectrum of genetic resources existing in the wild (Wylie 1954). Although there are more than 140 Rosa species (Wissemann 2003), it is believed that only 11 of them have contributed to the origin of modern cultivars (Gudin 2000).

Genus Rosa, along with Malus, Prunus, Fragaria and Rubus, is a member of the family Rosaceae. The basic chromosome number (x) of Rosa is 7, and 2C DNA amount of diploid roses varies from 0.78 to 1.29 pg (Yokoya et al. 2000).

Rose breeding is a long process, and a large number of phenotypic traits are observed in full-grown plants. The ability of rose breeders to select interesting parents for promising crosses and to identify progenies with favourable combinations of characters is still empirical and hampered by the limited knowledge of the genetic basis of important traits such as flowering, pathogen resistance, fragrance and plant architecture. Thus, the development of genetic maps locating genes controlling phenotypic characters can improve the speed and precision of rose-breeding programs by marker-assisted progeny selection.

Genetic maps have been previously constructed in diploid (Debener 1999; Debener and Mattiesch 1999; Crespel et al. 2002; Dugo et al. 2005) and tetraploid (Rajapakse et al. 2001) roses. These maps were constructed with dominant amplified fragment-length polymorphism (AFLP) and randomly amplified polymorphic DNA markers. Yan et al. (2005) constructed a genetic map in a diploid rose family with simple sequence repeat (SSR) markers developed by Esselink et al. (2003). Recently, the tetraploid rose maps have been expanded with a new set of SSR markers (Zhang et al. 2006).

Several major rose traits have been located on rose genetic maps. These include flower color and double corolla (Debener and Mattiesch 1999), prickles on the petiole (Rajapakse et al. 2001), single seasonal blooming and double corolla (Crespel et al. 2002), black spot resistance (Linde et al. 2004; Malek et al. 2000), and quantitative traits prickles on the stem (Crespel et al. 2002), leaf size, days to flowering and resistance to powdery mildew (Dugo et al. 2005).

In an attempt to make marker technology widely available for rose, easily transferable, highly polymorphic, codominant markers such as SSRs must be developed. SSRs have been detected in higher plants and are widely distributed throughout the plant genomes (Morgante and Olivieri 1993), in contrast to AFLP markers that are preferentially localised in centromeric regions and in clusters (Vuylsteke et al. 1999). SSR frequency in plants has been assessed by oligonucleotide hybridization, and the presence of SSR was detected every 8–60 kb of the genome (Panaud et al. 1996; Echt and May-Marquardt 1997). Genomic SSR markers have been previously developed in rose (Esselink et al. 2003; Yan et al. 2005; Zhang et al. 2006). SSRs appear to be abundant in rose, with an estimated density of one microsatellite present every ∼19 kb (Zhang et al. 2006).

SSR markers have been shown to be highly polymorphic, a property that would permit their application as molecular markers in rose population genetics, Rosaceae systematics and genome mapping. SSR markers have been used to estimate genetic variation between hybrid tea and rootstock rose varieties (Esselink et al. 2003), to understand reproduction in dog roses (Nybom et al. 2004) and to test the phylogenetic relationships between oil-bearing Rosa damascena Mill. accessions (Rusanov et al. 2005).

In the present study, we developed two sets of new rose microsatellites, one from a Rosa wichuraiana Crépin genomic library (Zhang et al. 2006) and the other from expressed sequence tags (ESTs) generated from cDNA libraries of rose vegetative and floral buds (Foucher et al., submitted). Latter SSRs represent the first report of SSR isolated from EST in rose. Both kinds of SSRs were used to expand the AFLP genetic map previously reported (Crespel et al. 2002) and to obtain a comprehensive genetic linkage map of the diploid rose genome. Using codominant SSR markers, we were able to identify homologous linkage groups (LGs) between the two parental maps. Furthermore, using the new maps, we investigated the genetic control of number of petals (NP) and blooming date (BD) in the diploid progeny.

Materials and methods

Mapping population and DNA isolation

The F1 mapping population used in this study has been previously used to develop an AFLP genetic map (Crespel et al. 2002). This interspecific population of 91 diploids was generated from a cross between dihaploid rose, H190, obtained from haploidisation of tetraploid Rosa hybrida cv Zambra (Meynet et al. 1994) and a hybrid of diploid R. wichuraiana, Rw, originating from Jardin de Bagatelle (Paris, France). H190 and Rw were chosen as parents because they differ in important horticultural traits; for instance, H190 is a recurrent blooming, double flowered, prickle-free bush susceptible to black spot and powdery mildew, whereas Rw is single seasonal blooming, single flowered, prickled, and ground cover partially resistant to black spot and powdery mildew. Moreover, H190 has 20–25 petals (a mean of 23 petals for five flowers counted) and flowers later than Rw in the field. The latter two traits, NP and BD, were scored in the 91 hybrids grown on their own roots and cultivated outdoors in pots, at INRA in Angers, France.

‘Félicité et Perpétue’ and ‘New Dawn’ cultivars used in the measurement of the nuclear genome size were obtained from the Nursery Loubert (http://www.rosesloubert.com). Young unfolded leaves (100 mg) were collected from all roses, and DNA was extracted from fresh leaves (Dneasy®96 Plant Kit, QIAGEN).

Nuclear genome size

The nuclear DNA content of diploid parents of the mapping population, diploid ‘Félicité et Perpétue’ and triploid ‘New Dawn’ cultivars, were measured using a cytometer (ISV, CNRS, Gif-sur-Yvette, France). Fresh leaves (1–2 cm2) were finely chopped with a sharp razor blade in 600 μL of Galbraith buffer (MgCl2 45 mM, NaNO3 30 mM, MOPS acid 20 mM pH 7.0) supplemented with 0.1% (w/v) Triton X-100 and 1% (w/v) Polyvinyl-pyrolidone 10.000, in the presence of same amount of Petunia hybrida cv PxPc6 [2C = 2.85 pg DNA/nucleus, GC ratio = 41.0% (Marie and Brown 1993)]. The extract was then filtered through a 48-μm mesh and treated with 40 μg/ml of RNase A (Roche) for 10 min on ice and stained with 50 μg/ml propidium iodide. The samples were analysed on an EPICS Elite ESP cytometer (Coulter, Hialeah, FL). Excitation of fluorochrome was performed by an air-cooled Argon laser (Uniphase) at 488 nm, 20 mW.

Development of microsatellite markers

The new rose SSR markers developed in this study are originated from two different sources: rose genomic DNA and rose EST. A genomic library from R. wichuraiana had been previously constructed, and about 800 SSR-positive clones had been identified by screening the library with 17 different SSR probes (Zhang et al. 2006). In the present study, 384 SSR-positive genomic clones from this R. wichuraiana library were sequenced by Ouest-Genopole®. The two cDNA libraries have been constructed from vegetative and floral buds of R. wichuraiana and R. hybrida Black Baccara®, respectively. Five thousand ESTs from these libraries have been sequenced and found to be clustered in 2,556 unigenes (Foucher et al., submitted). The EST sequences are available at the URGI public web server (http://www.urgi.versailles.inra.fr/GnpSeq/).

In the two sets of sequences (384 genomic clones and 2,556 unigenes from the EST bank), SSR detection and primer design were carried out using Ouest-Genopole® web server (http://genoweb.univ-rennes1.fr/Serveur-GPO/outils_access.php3?id_syndic=1//). This pipeline for SSR detection and primer design was developed by combining two softwares: Sputnik (http://www.cbi.labri.fr/outils/Pise/sputnik.html) for SSR detection and Primer3 (http://www.frodo.wi.mit.edu/cgi-bin/primer3/primer3_www.cgi) for primer design (Rozen and Skaletsky 2000). The search for SSR was carried out with selection criteria of repetition length greater than 20 bp (base pairs) for di-, 15 bp for tri- and 16 bp for tetra-nucleotides. Primers were synthesised by Sigma Genosys. For those EST sequences found to contain microsatellites, search for open reading frame (ORF) was performed using ORF Finder (http://www.ncbi.nlm.nih.gov/gorf/gorf.html) and Blastx (http://www.ncbi.nlm.nih.gov/BLAST) tools.

In addition to the new rose SSRs identified from genomic and EST clones in this study, SSRs already developed from other Rosaceae genera such as Prunus (Dirlewanger et al. 2002), Malus (Liebhard et al. 2003; Sosinski et al. 2000; Stam 1993) and Fragaria (B. Denoyes-Rothan, unpublished results) were also used to screen the mapping population.

Polymerase chain reaction amplification and visualisation of microsatellites

Some of the SSRs were analysed by gel electrophoresis coupled with silver staining, whereas others were analysed through a capillary sequencer. For those SSRs analysed by gel electrophoresis, polymerase chain reaction (PCR) was performed in a 10-μl volume containing 1X AmpliTaq buffer, 0.15 mM of each deoxyribonucleotide triphosphate (Promega), 2 mM MgCl2, 0.15 μM of each primer, 0.5 units of RedTaq DNA polymerase (Sigma) and 10 ng of genomic DNA. General PCR profiles used were: classical profile [94°C 2 min, (94°C 30 s, annealing temperature 15 s, 72°C 30 s) for 35 cycles, 72°C 7 min] and touchdown profile [94°C 2 min, (94°C 30 s, annealing temperature + 5°C 15 s, 72°C 30 s) for ten cycles with decrease of 0.5°C by cycle, (94°C 30 s, annealing temperature 15 s, 72°C 30 s) for 25 cycles, 72°C 7 min]. PCR products were mixed with an equal volume of 2× gel loading buffer (80% deionized formamide, 10 mM ethylenediamine tetraacetic acid pH 8.0, 1 mg/ml xylene cyanol, 1 mg/ml bromophenol blue) and denatured at 95°C for 5 min. Products were separated on 5% denaturing polyacrylamide sequencing gels by electrophoresis at 60 W for 2 h. An improved procedure described by Creste et al. (2001) was used for silver nitrate staining of gels.

For those SSRs analysed through a capillary sequencer, an economic fluorescent labelling method of PCR fragments was used (Schulke 2000). SSR forward primers were modified by 5′ concatenation with the 18-mer 5′-CACGACGTTGTAAAACGAC-3′ primer, which permitted concurrent fluorescent labelling of PCR products by a third primer (M13) incorporated with a fluorophore (ROX). The PCR mix was the same as the previously described mix except for the primers, extended forward and reverse primers (0.04 μM) and M13 primers (0.15 μM). The PCR profile used was 94°C 2 min, (94°C 30 s, 61°C 15 s) for two cycles, (94°C 30 s, 59°C 15 s) for two cycles, (94°C 30 s, 57°C 15 s) for two cycles, (94°C 30 s, 55°C 15 s, 72°C 30 s) for 25 cycles, 72°C 7 min.

PCR products were resolved on an ABI 3100 capillary electrophoresis array (Applied Biosystems) using GeneScan-500 [ROX] as the size standard (Applied Biosystems) on Ouest-Genopole® platform. The Genotyper software (Applied Biosystems) was used to analyse electrophoregrams.

Linkage map construction

Map construction in each parent was carried out using JoinMap version 3.0 (Stam 1993). Parental linkage maps were constructed using AFLP markers described in Crespel et al. (2002) and SSR markers developed in this study. Markers segregating from one or both parents were scored separately. For both maps, LGs were determined using a minimum of logarithm of odds ratio (LOD) threshold of 5.0, and map construction was performed using Kosambi mapping function with the following JOINMAP parameters: Rec = 0.45, LOD = 1.0, Jump = 5, first and second runs. The resulting linkage maps were drawn using MapChart software.

Evaluation of phenotypic traits and statistical analysis

In the present study, number of petals, designated NP, was counted from three flowers on each of the 91 F1 hybrids and analysed as both a major gene and a quantitative trait. To analyse NP as a major gene, flowers having less than ten petals were considered single, and flowers with ten or more petals, corresponding to two complete whorls, were considered double.

BD, the date of first appearance of a floral bud with sepals taken off corresponding to the stage 1 described by Put and Clerkx (2003), was recorded in the mapping population during one growing season.

Statistical analysis of the phenotypic data was carried out using Statistical Analysis System (SAS) software (SAS Institute, 1989). Analysis of variance (ANOVA) was performed using the general linear model (GLM) procedure to test the significance of the genotype effect. Broad-sense of heritability of genotypic means within the progeny was calculated using the following formula: \( h^{2} = {\sigma ^{2} g} \mathord{\left/ {\vphantom {{\sigma ^{2} g} {{\left( {{\sigma ^{2} g + \sigma ^{2} e} \mathord{\left/ {\vphantom {{\sigma ^{2} g + \sigma ^{2} e} n}} \right. \kern-\nulldelimiterspace} n} \right)}}}} \right. \kern-\nulldelimiterspace} {{\left( {{\sigma ^{2} g + \sigma ^{2} e} \mathord{\left/ {\vphantom {{\sigma ^{2} g + \sigma ^{2} e} n}} \right. \kern-\nulldelimiterspace} n} \right)}} \), where n is the mean number of replicates per genotype, σ 2 g is the genetic variance (i.e. inter-genotype variance) and σ 2 e is the residual error variance.

Additive quantitative trait locus (QTL) effects were detected using the simple interval mapping procedure of MapQTL software (Van Ooijen and Maliepaard 1997). QTLs with a maximum LOD score greater or equal to 3.0 were declared significant. For each significant QTL, confidence intervals corresponding to a LOD score drop of 1 or 2 on either side of the likelihood peak were calculated.

Results

Identification of new microsatellite markers in rose

In the present study, 156 new rose microsatellites were developed. Sixty-three were from rose genomic clones (designated genomic-SSR) and 93 from rose EST clones (designated EST-SSR). Through sequencing 384 SSR-positive clones from the R. wichuraiana genomic library of Zhang et al. (2006), we have identified 63 new genomic-SSRs. Fifty (79%) are di-nucleotide repeats, 12 tri-nucleotide repeats and one was a mono-nucleotide repeat. Among the di-nucleotide repeats, 39 are of the AG/GA/TC/CT repeat motif, 10 are of AC/CA/TG/GT and one is a TA/AT repeat. Twelve tri-nucleotide repeats are distributed as AAG (5), CCG (4), AAC (2) and GCA (1). The single mono-nucleotide repeat was (A).

From the 2,556 EST unigene sequences developed in our laboratory (Foucher et al., submitted), 93 SSRs have been selected. Among these EST-SSRs, 52.2% were di-nucleotide repeats, 44.6% were tri-nucleotide repeats and 3.2% were tetra-nucleotide repeats. The only di-nucleotide motif found was AG/GA/CT/TC, whereas a variety of tri-nucleotide repeats were found, of which the most common is AAG/AGA/GAA/CTT/TTC/TCT (22%). The localisation of microsatellites in the transcribed regions was studied. No ORF was detected in 5 of the 93 sequences. Di-nucleotide repeats were predominantly located in the 5′ untranslated regions (UTR; 89%) and less frequently in ORF (6.5%) or in the 3′ UTR (4.5%). Tri-nucleotide repeats were distributed across transcribed regions, principally in ORF (55%) and also in 5′ UTR (37%). Of the three tetra-nucleotide microsatellites, two are located in the coding sequence and one in 5′ UTR.

Characterization of SSR loci

In addition to these 156 new rose SSRs, previously described SSR from other Rosaceae genera (Sosinski et al. 2000; Dirlewanger et al. 2002; Liebhard et al. 2003; B. Denoyes-Rothan unpublished results) and from rose genomic DNA (Zhang et al. 2006) have been characterized. Therefore, we have analysed a total of 231 Rosaceae microsatellites (Table 1) in the mapping population.

Table 1 SSR markers used in this study, grouped by source and characterisation in the mapping population
Full size table

Among the 63 genomic-SSR tested, 43 (∼68%) failed to give any product or gave complex profiles of amplified products and were not further studied. The 20 (∼32%) remaining primer pairs amplified fragments in the range of the expected size for each microsatellite. Fourteen out of 20 genomic-SSRs (Table 2) showed polymorphism between parents H190 and Rw (Table 1). The six markers monomorphic in the mapping population are not presented in Table 2. Primer sequences of these markers are available upon request, as they could be polymorphic in other populations. From the SSRs previously characterized from the same genomic library (Zhang et al. 2006), 16 SSRs (Rw1F9, Rw8B8, Rw10J9, Rw10M24, Rw11E5, Rw14A5, Rw14H21, Rw17I17, Rw22A3, Rw22B6, Rw23H5, Rw27A11b, Rw29B1, Rw55C6, Rw55D22 and Rw61F2) were also tested and only one (Rw23H5) was monomorphic between the parents of our population (Table 1).

Table 2 Properties of rose polymorphic SSR primer pairs (F, forward; R, reverse) developed from genomic DNA
Full size table

Of the 93 EST-SSR, 26 (∼28%) failed to give any product, and 23 (∼25%) gave rise to a PCR product that had a larger size than the expected, suggesting the presence of introns between the primer sites in the genomic sequence (Table 1). Eighteen (∼19%) are monomorphic in the mapping population; their sequences are not included in Table 1 but available upon request. Twenty-six (28%) displayed polymorphism between parents of the population (Tables 1 and 3). Two EST-SSR primer pairs amplified multiple loci, two in C139 and three in H9B01 (Table 3).

Table 3 Properties of rose polymorphic SSR primer pairs (F, forward; R, reverse) from EST library
Full size table

Among the 59 heterologous Rosaceae microsatellites tested, only five have been used to construct the genetic map. Indeed, as shown in Table 1, of the 17 Prunus SSRs tested, only 3 [pchgms3 (Sosinski et al. 2000), BPPCT28 (Dirlewanger et al. 2002) and pchgms40 (Wang et al. 2002)] are able to amplify rose gDNA and are used to screen the mapping population. From 25 Malus SSR primers, two gave expected-size amplified fragment, but only one [CH02C11 (Liebhard et al. 2003)] is polymorphic. Transferability of Fragaria SSRs was high, as half of the Fragaria SSRs tested displayed polymorphism between parents of the mapping population (Table 1). Nevertheless in this work, only one (F3-7C, B. Denoyes-Rothan) has been mapped. Analysis of the six remaining Fragaria SSRs is in progress in a synteny study between Fragaria and Rosa.

Map construction

Segregation data obtained from 91 F1 progeny for AFLP (Crespel et al. 2002), genomic-SSRs, EST-SSRs and heterologous Rosaceae SSRs were used to generate the parental maps. A total of 126 loci derived from AFLP (73), genomic-SSR (24), EST-SSR (24) and heterologous Rosaceae SSR (5) were analysed to construct the map of maternal parent H190 (Table 4, Fig. 1). The genetic linkage map of the H190 has seven LGs (identified as A1–A7) and covers 432 cM of its genome (Table 4, Fig. 1). Of the 126 loci analysed, 106 were placed on the maternal parent map at a LOD score of 5.0. Of the 106 mapped markers, 27 (∼26%) departed from the expected Mendelian segregation (p < 0.05); these are distributed primarily along three LGs A2, A3 and A6. These distorted markers are identified on the map with asterisks (Fig. 1).

Table 4 Distribution of SSRs and AFLP markers in the linkage groups of the maternal (A1–A7) and paternal (B1–B7) maps
Full size table
Fig. 1
figure 1

Linkage maps of the dihaploid H190 (A1–A7) and the hybrid of diploid species R. wichuraiana (B1–B7) developed from 91 hybrids of an F1 population. EST-SSRs are indicated in bold, and genomic-SSRs are underlined. AFLP markers have been previously located by Crespel et al. (2002). Map distances in cM are listed on the left and loci on the right of each linkage group. SSRs common between maternal and paternal maps are linked by dotted lines. Level of distorted markers are indicated with an asterisk (***p < 0.01, ****p < 0.001, *****p < 0.0001). Localization of the major gene controlling number of petals (NP) is indicated by an arrow, and the corresponding gene studied by Crespel et al. (2002) (d6) is indicated by a dotted arrow. Additive QTLs for blooming date (BD) and NP are represented by a box prolonged with lines, which respectively stretch over the “LOD 1” and “LOD 2” confidence intervals

Full size image

A total of 158 loci derived from AFLP (106), genomic-SSR (26) EST-SSR (22) and heterologous Rosaceae SSR (4) were used to construct the paternal map of R. wichuraiana hybrid Rw. Of the 158 loci, 137 were assigned on seven LGs (identified as B1–B7) at a LOD score of 5.0 and covering 438 cM of its genome (Table 4, Fig. 1). Fourteen markers (∼10%) have segregation ratios significantly different from the expected Mendelian segregation (p ≤ 0.05); these were placed on two LGs B2 and B3 and identified on the map with asterisks (Fig. 1).

Nuclear genome size

To evaluate the physical genome size of the parents of the mapping population, nuclear 2C DNA amounts have been measured for both parental roses as well as that of two cultivars, ‘Félicité et Perpétue’ and ‘New Dawn’. The DNA amounts of the two cultivars were 1.27 pg (SD ± 0.02 pg) and 1.75 pg (SD ± 0.01 pg) for ‘Félicité et Perpétue’ and ‘New Dawn’, respectively. These values were comparable with the 2C DNA amounts previously reported, i.e. 1.29 pg (SD ± 0.02 pg) and 1.76 pg (SD ± 0.08 pg) for ‘Félicité et Perpétue’ and ‘New Dawn’, respectively (Yokoya et al. 2000). The 2C DNA amounts of the parental roses measured in this study were similar, 1.18 pg (SD ± 0.02 pg) and 1.16 pg (SD ± 0.01 pg), respectively, for H190 and Rw. The genomic size of Rw measured in this study is comparable with the size of R. wichuraiana (1.13 ± 0.08 pg) obtained by Yokoya et al. (2000).

Anchoring of the maternal and paternal parental maps

SSR markers that showed heterozygosity in both parents (types abxac and abxcd) were used as anchor points between maternal and paternal genetic maps (Table 4, Fig. 1). The homology of LGs A1/B1, A4/B4 and A7/B7 was established by two common SSR markers for each group: Rw25J16 and Rw34L6 for A1/B1, F3-7C and Rw16E19 for A4/B4 and CTG623 and Rw55D22 for A7/B7. Three SSRs, CTG329, Rw27A11b and Rw59A12, allowed us to define the homology between A2 and B2 groups. The greatest number of anchoring SSR markers, 5, was observed between A3/B3, A5/B5 and A6/B6. SSR markers C139, Rw55E12, CL3881, Rw20I17 and H2F12 anchor A3 and B3 groups, whereas SSR markers BPPCT28, Rw10M24, Rw15D15, H4F06 and Rw5G14 link A5 and B5 groups. Maternal LG A6 is homologous to paternal LG B6 based on SSR markers Rw52D4, CL2845, H24D11, H20G01 and H22C01.

Phenotypic data analysis

We analysed two phenotypic traits, NP and the BD, in the F1 population (Fig. 2). This study expands the results for NP previously named double flower and designated d6 and performed only on recurrent blooming hybrids (Crespel et al. 2002). A strong genotype effect (p value < 0.001) was observed for NP by ANOVA. The ratio 1:1 for double versus single flowers (40:39) in the mapping population suggests the effect of a major gene for doubleness as described by Debener et al. (2001). Thus, the maternal parent H190, characterized by double flowers (>10 petals) would be heterozygous NP/np, whereas the paternal parent Rw, derived from R. wichuraiana, characterized by five petals in the corolla, would be double recessive np/np. For single flowers corresponding to a number of whorls lower than two, the distribution ranged from mean values of 5 to 8.33. The double flowers displayed a distribution of NPs ranging from mean values of 10 to 180. The major gene for NP is located in the LG A4. Tight linkage between the major gene NP and AFLP markers E2M2.07 (LOD = 13.7) and E2M5.06 (LOD = 16) could be detected on LG A4. Difference in map positions observed between major genes NP and d6 identified by Crespel et al. (2002) is due to three hybrids that have a mean of petal number between 7.33 and 8.33, previously considered as double flowers but considered as single flowers in this study, as flowers with less than ten petals corresponding to less than two complete whorls were considered single in this study.

Fig. 2
figure 2

Distribution in the 91 F1 progeny of a the number of petals per flower recorded as the average of 3 flowers per plant. b The blooming date, recorded as the date of first appearance of a floral bud from April 26, 2005. Parental values are indicated by arrows (Rw for R. wichuraiana hybrid and H for H190). The number of recurrent/non-recurrent genotype for each bloom period i indicated above the graph

Full size image

According to Debener et al. (2001), the distribution of NP in double-flowered group (Fig. 2a) could suggest a polygenic inheritance of this trait. Broad-sense heritability of the NP was 0.54 for single-flowered sub-group and 0.97 for double-flowered sub-group. Effects of additive QTL was studied with MapQTL software for double-flowered sub-group. A QTL was detected with R 2 of 29% and a LOD score of 3 on B4 in Rw map (Fig. 1). As A4 and B4 are anchored by two SSRs, both the major gene (NP) and additive QTL controlling NP in the double flowers were localised on this A4/B4 consensus LG.

For BD in the F1 progeny (distribution shown in Fig. 2b), a strong QTL (LOD = 6.38), explaining 34% of the phenotypic variance, was detected on the B4 LG (Fig. 1). As the locus controlling recurrent blooming located by Crespel et al. (2002) and QTL affecting BD were located on the same LG, we analysed their interaction by the GLM procedure. The recurrent blooming group flowered significantly earlier than the non-recurrent single seasonal blooming group (R 2 = 0.16, as shown in Fig. 2b).

Discussion

One of the major goals of this study was to develop new easily transferable SSR markers in rose. A total of 64 new rose SSR markers were generated from this study, 44 from EST sequences and 20 from a genomic DNA bank. Out of these, 40 SSRs have displayed polymorphism between the parents of the mapping population and have been mapped. The remaining 24 SSRs, although not polymorphic in the mapping population, gave rise to clear PCR products, and they might be useful in other rose populations and studies. In addition to the new rose SSRs generated in this study, 15 SSRs developed from rose genomic DNA (Zhang et al. 2006) and 4 SSRs developed from other Rosaceae species have also been mapped in the diploid population.

EST as a good source of SSR markers

EST databases are increasingly becoming available for a wide range of plants and plant tissues. Bioinformatic tools allow easy searching for SSRs in these EST databases. In this study, a set of new rose SSRs was developed using such strategy. We searched a rose bud EST database of 2,556 unigenes obtained after analysis of 5,000 EST sequences (Foucher et al., submitted). In the rose bud EST database, 3.6% of the sequences contained repeats of 20 nucleotides for di-, 15 for tri- and 16 for tetra-nucleotide repeats. This frequency is comparable to those obtained for Vitis (2.5%, Scott et al. 2000), maize and rice (1.5 and 4.7%, respectively, Kantety et al. 2002) and recently for the Rosaceae species (4%, Jung et al. 2005) but lower than those reported in Capsicum (10.5%, Huang et al. 2001) and wheat (7.5%, Nicot et al. 2004). These differences may be explained by SSR length criteria used in these different studies.

Jung et al. (2005) showed that frequency of SSRs in rose is 59% for di-, 35% for tri- and less than 5% for tetra-nucleotides. This analysis for SSR sequences was performed on rose petal EST databases of Channeliere et al. (2002) and Guterman et al. (2002). In our study, this order of repeats remains the same with slightly different frequencies (52.2, 44.6 and 3.2%, for di-, tri- and tetra nucleotides, respectively). In both studies, the most common type of repeat motif is (CT)n for di-nucleotides and (GAA)n for tri-nucleotides.

The ESTs used in this study were obtained through 5′-end sequencing of directionally cloned full-length cDNA (Foucher et al., submitted). This could explain the over-representation of SSRs in the 5′ UTR. Although the di-nucleotide SSRs are over-represented in the 5′ UTR, an equal frequency of tri-nucleotide SSRs was found between 5′ UTR and ORF. Moreover, tri-nucleotide microsatellites were more abundant in ESTs (44.6%) than in genomic sequences (19%). Tri-nucleotide repeats do not lead to frame shift mutation and, as proposed by Morgante et al. (2002), tri-nucleotide microsatellites in the coding region could be a result of mutation pressure and positive selection for specific amino acid stretches.

Fewer SSRs derived from ESTs displayed polymorphism in our inter-specific mapping population (59%) compared to the SSRs derived from genomic sequences (70%; Table 1). Similar results have been found in other plants (Blair et al. 2003; Cho et al. 2000; Schloss et al. 2002), as EST sequences are generally more conserved than genomic sequences. Although SSRs derived from EST are less polymorphic than genomic SSR, the EST-SSRs could be more transferable inter-genus or intra-genus than anonymous genomic sequences. As a large number of SSRs can be very easily developed from the available EST databases, they can be effectively used in the construction of a consensus genetic map for rose as well as synteny studies with other Rosaceae species.

The use of SSRs to construct genetic maps

Microsatellite markers generated were used to enrich a genetic map previously constructed with AFLP markers (Crespel et al. 2002). Genome coverage and LGs in the original maps were considerably improved by the addition of SSRs. As a result, seven LGs covering a total length of 438 cM for the paternal parental map and 432 cM for the female parental map were obtained, showing an increase in the genome coverage of 81 and 52.5%, respectively, for maternal and paternal maps, compared to the previous maps (Crespel et al. 2002).

Recently, genetic maps for roses have been constructed with different types of molecular markers such as AFLP, SSR, PK, RGA, RFLP, SCAR and morphological markers (Yan et al. 2005). The genome coverage for Rosa multiflora was statistically estimated to be about 500 cM (Yan et al. 2005), whereas the two maps in the present study cover an average of 435 cM of parental genomes. The 2C amount of R. multiflora, estimated to be 1.66 pg (Dickson et al. 1992), is larger than the genomes of both maternal and paternal roses mapped in the present study (average of 1.16 pg). This difference in physical genome size might explain the difference in the length of genetic maps obtained by Yan et al. (2005) and in the current study, although genetic and physical map distances are not always correlated.

The density of markers in maps of the present study is lower (one marker every 4 and 3 cM for maternal and paternal maps, respectively) than the density (0.4–0.7 markers per cM) of the maps developed in Yan et al. (2005). Moreover, several large gaps are present in the maps of the present study. The largest gap (27 cM) is located on LG A5 (Fig. 1). Adding new SSR markers to the maps, such as those from other studies on Rosa (Yan et al. 2005), from Rosa EST (Channeliere et al. 2002; Guterman et al. 2002) and/or other Rosaceae species, will help fill these gaps and further saturate the maps.

Some markers, 26 and 10% of maternal and paternal parental markers, respectively, exhibited segregation distortion. These markers are located on several LGs. Markers with distorted segregation have already been reported in Rosaceae species (Aranzana et al. 2003; Liebhard et al. 2003; Sargent et al. 2004) and in rose (Crespel et al. 2002; Debener and Mattiesch 1999; Yan et al. 2005). Gametophytic selection by sub-lethal genes conferring “low viability” of zygote, embryo or seedling, could be located near the distorted markers in maps. In rose, the average rate of seed germination is around 40% (Meilland-Richardier Int., personal communication). Moreover, the origin of female dihaploid H190 (Meynet et al. 1994) could be explained by the large proportion of distorted markers in this genotype. This hypothesis is discussed in detail by Crespel et al. (2002).

The maternal and parental maps developed in this study were integrated using the codominant SSR markers. For this integration, 24 SSRs shared by the two maps were used.

Localization of genes controlling important rose traits

This map was constructed to locate major genes as well as QTL controlling interesting rose characters. We tested the integrated map for the localization of genomic regions involved in NP and BD. We have shown that petal number is controlled by a major gene and a QTL as previously proposed by Debener (1999) and located on A4 and B4 integrated LG. These results have been obtained from small number of flowers into 91 hybrids. An extended study could be realised to obtain additional QTL with smaller effects on a more significant number of flowers and/or on a higher population. Petal number has been previously located on other genetic maps on LG D10-2 (Dugo et al. 2005) and LG B3 (Debener and Mattiesch 1999). Based on NP gene, we can suggest that A4/B4 LGs (this study) would be homologous to LG D10-2 (Dugo et al. 2005) and LG B3 (Debener and Mattiesch 1999). This hypothesis can be tested by developing common markers (as SSR) on these different LGs.

A strong QTL controlling BD was located on the same integrated LG A4/B4. Dugo et al. (2005) detected two QTLs, Df1 and Df2 controlling flowering time on two LGs different from D10-2. Thus, the process or the variability of the BD could be different in our population compared to this previous study; indeed, date of flowering is known to be a complex process involving numerous genes. Moreover, in Dugo et al. (2005), flowering time was recorded when 50% of flowers were fully opened, whereas in our study, flowering time was recorded when the first flower reached a specific stage.

The integrated A4/B4 group locates important major genes and QTLs controlling several flowering-related genes. These include recurrent blooming, BD, NP and presence of prickles (Crespel et al. 2002). Thus, this integrated LG could contain genes with pleiotropic effect or cluster of genes implied in important ornamental traits, which have been selected by breeders.

Anchoring with other rose maps

Some of the SSR markers provide anchor points between maps constructed in the present study and the tetraploid rose maps (Zhang et al. 2006). The apple SSRs CH02C11 and Rw55D22, mapped to A7 in the present map, are placed on the second integrated LG of the tetraploid map that also carry SSRs Rw22A3, Rw11E5 and Rw5D11 (Zhang et al. 2006). Indeed, Rw22A3 and Rw11E5 are located on the B7 of the present map. Thus, SSRs Rw55D22 and CH02C11on LG A7 and Rw22A3 and Rw11E5 on B7 are homologous to the second integrated LG of the tetraploid map (Zhang et al. 2006). Furthermore, Rw10M24, one of the four SSRs that established homology between A5 and B5 groups (this study), is mapped to the first consensus group of the tetraploid consensus map (Zhang et al. 2006). Three SSRs, present in the consensus tetraploid map (Rw1F9, Rw8B8a and Rw22B6), are located on A5 (Rw8B8a) and B5 (Rw1F9 and Rw22B6), suggesting that A5/B5 is homologous to the first consensus group of the tetraploid maps.

Several important qualitative and quantitative rose traits have been located in different rose maps (Debener 1999; Debener and Mattiesch 1999; Crespel et al. 2002; Dugo et al. 2005; Yan et al. 2005; Zhang et al. 2006). SSR markers developed in this and other studies will provide tools for the establishment of a consensus linkage map for roses that combine traits and markers in the various maps. This consensus map locating important qualitative and quantitative traits will allow new marker-assisted approaches to breeding complex traits in rose.

The use of SSR markers to study synteny in Rosaceae species

A large number of SSR markers have now been characterized in Rosaceae; notably in Malus (Liebhard et al. 2003), Prunus (Dirlewanger et al. 2002; Sosinski et al. 2000), Fragaria (Sargent et al. 2004; B Denoyes-Rothan, unpublished results) and recently in rose (Yan et al. 2005; Zhang et al. 2006; this work). The development of molecular markers shared by the members in this family is an important tool for comparative genetics and to study synteny in Rosaceae family. Although SSR markers are attractive for this purpose, development of markers that would be shared by at least some genera is difficult and time consuming. In this study, the transportability of microsatellites developed from Fragaria to rose was rather good (50%); however, the transportability of SSR markers from Malus and Prunus to rose was poor (less than 10%). To date, only four markers developed from other Rosaceae species have been located on the rose maps of the present study. Only one marker among them, BPPCT028, is known to be transportable to Prunus, Malus and Fragaria species (Dirlewanger et al. 2002). Thus, SSRs appear to be good transportable markers for synteny study within the Rosoideae tribe of the Rosaceae that include strawberry, raspberry and rose but not suitable for synteny between different tribes as shown with the low transportability between Malus/Prunus and Rosa. Markers more suitable for synteny studies between distantly related genera could be conserved ortholog set markers (Fulton et al. 2002).