Five individuals of R. canina were sampled from a natural population "Himmelreich", Jena, Germany (plants: H13, H17, H19, H20, H21), and two individuals of R. canina were taken from the dogrose collection at the Botanical Garden Gießen, which were originally collected at the natural population "Einzelberg", Groß Schneen, Germany (plants 194, 378). Voucher specimens have been deposited at the Herbarium Gießen (GIE).
Flow-cytometry was conducted according to the method described in  using a Cell Counter Analyzer CCA II (Partec, Münster, Germany) and Rosa arvensis Huds. (2n = 2x = 14) as an internal diploid standard. A minimum of 10,000 nuclei giving peaks with a coefficient of variation of approximately 10% were counted.
DNA and RNA Extraction
DNA was extracted from young leaf material according to . Total RNA was obtained from small and large floral buds using RNeasy Plant Mini Kit (Qiagen, Hilden, Germany) following the manufacturer's protocol and its modifications described by . First strand cDNA was synthesized by RevertAid™ H Minus M-MulLV Reverse Transcriptase (Fermentas, St. Leon-Rot, Germany) using an oligo-dT primer.
Sequences of LEAFY, cGAPDH and nrITS-1 were obtained from genomic DNA to identify polymorphisms between alleles located on different chromosome sets. Primers for the amplification of LEAFY were designed from an alignment of cDNAs of LEAFY of different species of Rosaceae: LFYex1-fwd (5'-CAAGTGGGACCTACGAGGCATGG-3') and LFYex3-rev (5'-TCGGCGTGACAAAGCTGACGAAG-3'). Primers for the amplification of cGAPDH were designed from cDNAs of Rosa chinensis Jacq. and Fragaria × ananassa (Weston) Rozier taken from Genbank: GPDex2-fwd (5'-GCCAAGATCAAGATCGGAATCAACG-3') and GPDex11-rev (5'-CTCGTTCAATGCAATTCCAGCCTTG-3'). Primers for amplification of nrITS were taken from . PCR was performed in 50 μl containing 2 μl of undiluted or diluted genomic DNA, 2 units Taq -Polymerase (Fermentas, St. Leon-Rot, Germany), 5.0 μl 10-fold polymerase buffer (Fermentas), 4.0 μl MgCl2 (25 mM), 2 μl of each primer (10 μM), 5.0 μl dNTPs (2 mM). The following PCR protocol was performed: initial denaturation cycle of 150 s at 94°C, followed by 30 cycles of 30 s denaturation at 94°C, 60 s annealing [annealing temperature (TA): TA = 58°C for LEAFY, TA = 51°C for cGAPDH and TA = 48°C for nrITS-1], 180 s extension at 72° C and a final extension for 10 min at 72°C. Purified PCR-products (Wizard SV Gel and PCR clean up system, Promega, Mannheim, Germany) were cloned into the vectors pGEMT (Promega) or pJET1 (Fermentas). Ligation products were electroporated into E. coli JM109 or DH5α. Twenty positive clones of at least two PCR products were sequenced in both directions using the same primers as for amplification and additional internal primers for LEAFY (LFYex2-fwd: 5'-CAAGAGAAGGAGATGGTTGGGAG-3'and LFYex2-rev: 5'-GCTGCTTGGCAATGTTCTGGAC-3') and cGAPDH (GPDex6-fwd: 5'-GTCAATGAGCATGAATACAAGTCC-3' and GPDex6-rev: 5'-GACTTGTATTCATGCTCATTGAC-3'). Sequences of the alleles LEAFY-4, cGAPDH-2 and cGAPDH-4 were only sampled in some plants. To test for the presence of these alleles in the remaining plants we performed allele-specific PCRs according to the conditions described above using the forward primers (LFYin1-al4-fwd: 5'-GGACATGTAAATAGGTCGAGAATATAT-3', GPDin2-al2-fwd: 5'-AGTTTTCGGATTTTGGTTTCGATC-3' and GPDin3-al4-fwd: 5'-ATCTTTGATGTTTTCGGAGTTATATG-3', respectively) spanning over allele-specific indels in introns. Resulting sequences were assembled and aligned using Bioedit . New sequence information generated within this study was deposited at the EMBL sequence archive under accession IDs FR725963 - FR725973.
To estimate the copy numbers of LEAFY and cGAPDH 30 μg of genomic DNA of plant sample H20 was digested with either Eco RI, Hin cII, Hin dIII, Kpn I, Pst I or Xba I, separated on 1% agarose gels and blotted onto positively charged nylon membranes (VWR, Darmstadt, Germany). Membranes were hybridized with 32P-αdATP-labelled LEAFY or cGAPDH fragments according to NEBlot Kit (NEB, Frankfurt, Germany). Hybridization probes were prepared from pJET1 plasmids by PCR using the primers LFYex2-fwd, LFYex3-rev and GPDx7F , GPDex11-rev, respectively, under same conditions as above. Gene fragments of LEAFY produced under these conditions have an expected length of 1200 bp and those of cGAPDH a length of 850 bp.
The best fitting model according to the corrected Akaike Information Criterion for each alignment was estimated for exon and intron sequences separately with MrModeltest v. 2.3 . The parameters of the best model for each partition were employed to reconstruct phylogenies of LEAFY and cGAPDH with MrBayes v.3.1.2 , additional file 5). We ran the analyses over 10,000,000 generations, sampling every 100th generation and discarding the 100,000 trees as burn-in resulting in a 50% majority rule consensus tree showing all compatible partitions supported by posterior probabilities (PP) for each node. The phylogeny of LEAFY was rooted with cDNA sequences of Fragaria vesca L. and other species of Rosaceae, phylogeny of cGAPDH with a cDNA sequence of Fragaria × ananassa and Arabidopsis thaliana L. (Heynh.). Alignments and phylogenies were deposited in Treebase [http://www.treebase.org (study accession: TB2:S11025)].
A phylogenetic network was calculated with TCS v. 1.2.1  for the nrITS-1 sequence data including also consensus sequences of different Rosa nrITS alleles detected in a former study  under 95% connection limit and gaps treated as missing data.
Previous studies on microsatellite alleles demonstrated that alleles with two or more copies are involved in bivalent formation [16, 17] and thus undergo recombination during meiosis. Therefore, we wanted to investigate whether alleles of LEAFY and cGAPDH with two or more copies evolve differentially from alleles with one copy. We conducted Maximum Likelihood pairwise Relative Rate Tests (RRT) implemented in the program HyPhy  using the Muse-Gaut model (MG94W9 in HyPhy  of codon substitution to estimate the relative rates of substitutions between different alleles of LEAFY and cGAPDH, respectively, and out-group sequences from Fragaria. The resulting parameter estimates were compared by a series of Likelihood Ratio Tests (LRT). To control for the False Discovery Rate we corrected original P -values with the Benjamini-Hochberg  formula as recommended by the HyPhy online discussion forum.
To test whether coding regions of LEAFY and cGAPDH alleles with two or more genomic copies are under other selective regimes than alleles with one copy we estimated the ratio (ω) of the rate of non-synonymous substitutions at non-synonymous sites (dN) to synonymous substitutions at synonymous sites (dS). The estimates of ω indicate whether an allele is under purifying selection (ω < 1), positive selection (ω > 1) or evolves neutrally (ω = 1). We conducted the analyses based on an alignment of consensus sequences of the coding region of LEAFY and cGAPDH alleles and an unrooted topology of them using the program codeml from the PAML package [66, 67]. LRT was employed to test, whether the model assuming different ω's for the allele with two or more copies than alleles with one copy (alternative hypothesis) fits better to the data than the model assuming the same ω for all sequences (null hypothesis).
Allele-specific single nucleotide polymorphisms (SNPs) in the coding region were used to estimate the frequency of the different alleles in cDNA pools by pyrosequencing as a measure for their specific transcription. Suitable pyrosequencing templates containing allele-specific SNPs were deduced from alignments of LEAFY, cGAPDH and nrITS -1, respectively. We analysed the same SNPs in genomic DNA to control for the copy number of alleles in the plants. The expected frequency in genomic DNA of an allele-specific base at a SNP is 0.2 for an allele with one copy, 0.4 for an allele with two copies and 0.6 for an allele with three copies in pentaploid individuals. These expected frequencies represent the null hypothesis of equal transcription of all alleles referring to their copy number. Three PCR products from cDNA pools of small and large flower buds of the individuals H13, H19 and H20 were amplified with primers presented in additional file 6 according to the cycling programs mentioned above. To control for contamination of RNA extracts with genomic DNA we performed PCR reactions using RNA extracts directly. Additionally, two PCR products from genomic DNA of the same plants were generated.
Template generation was done as described previously . Briefly, purified PCR products were ligated into the vector pCR2.1-TOPO (Invitrogen, Karlsruhe, Germany). The recombinant DNA was used as template in a second PCR using universal biotin-labelled primers bt-f or bt-r and sequence specific pyrosequencing primers (additional file 7). Purification of biotin-labelled ssDNA was done using streptavidin Sepharose (Biotage, Uppsala, Sweden). Sequencing reaction and allele frequency determination was carried out on a PSQ96 MA machine (Biotage) following the manufacturer's instruction.
Statistical tests were performed with SPSS v. 17.0. To test the influence of bud age and the investigated individual on transcription levels of alleles we performed Univariate ANOVA for each SNP in each locus. We detected a significant impact of the individuals but no significant impact of bud age on transcription level (data not shown). Thus we performed General Linear Model (GLM) analysis with "individual" as random factor to test whether allele frequency measured in genomic DNA differs significantly from allele frequency measured in cDNAs for each SNP in each locus. In cases where genomic allele composition differed between individuals we performed the tests separately.