MRF can be defined as the probability that two paired segments of non-sister chromatids recombine. To determine MRF in diploids and tetraploids, we used an MRF assay with visible markers previously established 9 . Arabidopsis thaliana line 3A [Col3-4/20 sensu 9] contains transgenic inserts coding for Green and Red Fluorescent Proteins (GFP and RFP), driven by a seed-specific promoter. Both inserts are located on the top arm of chromosome 3, GFP distal and RFP proximal, with a physical distance of 5.1 Mb and a genetic distance of 16.3 cM 9 . Seeds with both markers appear yellow upon UV excitation; the occurrence of seeds with exclusively green or exclusively red fluorescence indicates a meiotic recombination event between the markers. From the relative frequencies of these events, MRFs can be estimated 9 . Both markers were stably expressed over several generations 9 , and plants homozygous for both markers exclusively produced seeds with yellow fluorescence also during our experiments (n = 3177, progeny from three individual plants).

To analyze diploid MRF, line 3A, homozygous for both marker genes, was crossed with a non-transgenic wild type to produce diploid F₁ plants with only one non-recombined chromosome (Figure 1, grey panel). After selfing, the progeny plants indicated an MRF of 15.4% (Table 1 and Additional File 1), not significantly different (P > 0.1, unpaired t-test) from the previously published value of 16.3% 9 . Backcrosses of the heterozygous F₁ plants in reciprocal configuration indicated a female MRF of 7.4% and a male MRF of 20.2% (Table 1). The significant (P < 0.0001) 2.7-fold difference is in agreement with published genetic and cytological data 10 11 .

To determine MRFs in autotetraploids, we identified plants with a stable and complete tetraploid chromosome set of 4 × = 20 among the progeny of a colchicine-treated diploid, homozygous line 3A. Seeds from these plants (n = 3652, progeny from five individual plants, three generations after polyploidization) exclusively showed yellow fluorescence, indicating stable expression of both markers after polyploidization. The tetraploid derivatives, with four copies of the marker chromosome, were backcrossed to non-transgenic autotetraploid Arabidopsis thaliana (Figure 1, yellow panel). This results in plants carrying two copies of the transgenic chromosome 3. To exclude a dependence of MRF on the copy number and to strictly compare with diploids having a single marker chromosome, we backcrossed the double-copy lines once more to wild-type plants (Figure 1). The intensity of red and green fluorescence depends on the dosage of the marker genes. With the double-copy F1 plants as reference, plants with one gene copy each were selected and used for the final test crosses. We have taken into account that markers could have been separated by recombination already during the backcross, but nevertheless be transmitted through the same gamete. However, in this case, they would segregate like uncoupled markers in the selfed progeny, and we have retrospectively eliminated such plants from the analysis. The single copy plants were backcrossed once more to the tetraploid wild type, again in reciprocal orientation, or allowed to self-pollinate. Again, female and male MRFs were significantly different from each other (P < 0.0000001), but more importantly, single-copy autotetraploid MRFs were significantly higher in comparison to diploid MRFs (selfing MRF = 20.5%, P < 0.0001; female MRF = 15.0%, P < 0.001; male MRF = 28.0%, P < 0.001, Table 1 and Additional File 2).

We observed a slight tendency towards more red-only than green-only seeds [see Additional File 2], and potentially, this could have been caused by silencing of the GFP marker gene in the polyploids, as seen for other transgenes 12 13 14 . Therefore, we tested for the presence of the GFP marker in the genomic DNA of 12 plants grown from red-only seeds and derived from several individual parents, using PCR. All were negative, indicating that the green marker was indeed lost by recombination. Some bias towards red-only seeds could be the result of double reduction, a special type of meiotic recombination in the case of multivalent formation in autotetraploids in which sister chromatids end up in the same gamete 15 . Its frequency depends on the distance between marker and centromere and is therefore different between proximal and distal positions. However, double reduction cannot serve as an explanation in allopolyploids. Potentially, a bias can also originate from unequal fitness and transmission of gametes with a specific recombined chromosome, but this is rather unlikely in the case of the autopolyploids.

Increased MRF in autotetraploids could potentially be connected with multivalent formation, the major distinction of autotetraploid meiosis (described above), due to the homology of four chromosomes. Therefore, we generated an equivalent assay system for scoring MRF in tetraploids with only bivalent formation. Arabidopsis arenosa is a natural autotetraploid and close relative of Arabidopsis thaliana. Crosses between tetraploid Arabidopsis thaliana and Arabidopsis arenosa give rise to fertile allopolyploid hybrids that resemble Arabidopsis suecica, a species originating from an ancient hybridization event between these species 16 . We crossed the homozygous tetraploid line 3A with non-transgenic tetraploid Arabidopsis arenosa (4 × = 32) (Figure 1, blue panel) and identified plants with stable and full parental chromosome sets of 4 × = 26. Chromosomes derived from the two parental genomes in Arabidopsis suecica and in newly established hybrids were shown to pair in a diploid-like manner 6 . To verify the strict exclusion of multivalent formation or pairing between non-homologous chromosomes, we performed a cytological analysis of the resulting hybrids during the transition from meiotic metaphase I to anaphase I. Chromosomes from different parental origin were distinguished by fluorescence in situ hybridization with differentially labeled species-specific centromeric repeat probes. Among 55 metaphases analyzed, Arabidopsis thaliana and Arabidopsis arenosa chromosomes were found as homologous bivalents in 99.3% and 98.0%, respectively (Figure 2), with the few remaining chromosome pairs already separated due to the onset of anaphase I. No obvious case of pairing between homoeologs was detected. Thus, the newly synthesized hybrids, containing the meiotic tester chromosome, also showed a strictly diploid-like behavior of meiotic chromosomes. Subsequent crosses of the hybrids (this time to Arabidopsis suecica), identification of plants with one non-recombined chromosome, and the final cross in reciprocal orientation and with self-pollination, were performed as for autotetraploids (Figure 1, blue panel). Selfing, female, and male MRFs (24.1%, 13.0%, and 29.8%, respectively, Table 1 and Additional File 3) were again significantly higher than those in diploids (P < 0.001, P < 0.01, and P < 0.01, respectively) and in the same range as those in autotetraploids. These results suggest that the increase of MRF in tetraploids is independent of the formation of multivalents and pairing-partner switches.

Arabidopsis thaliana Col-0 line 3A [Col3-4/20, 9] harbors GFP and RFP transgenes on the top arm of chromosome 3, at nucleotide positions 256,516 and 5,361,637, respectively. All marker genes were homozygous in the diploid starting material. Seeds of autotetraploid Arabidopsis thaliana, autotetraploid Arabidopsis arenosa, and synthetic allotetraploid Arabidopsis suecica were obtained from L. Comai (University of California, Davis, USA). All plants were grown at 22°C under long-day (16 hour light: 8 hour dark) conditions.

Two week-old in vitro grown plants of line 3A were flooded with 0.1% colchicine dissolved in tap water for two hours, rinsed extensively with tap water, transferred to soil, and grown to set seeds. Nuclear DNA content of their progeny plants was analyzed by ploidy analyzer PAII (Partec, Münster, Germany). Tetraploid candidates were selected and the karyotype confirmed by counting mitotic chromosomes.

Meiotic or mitotic chromosomes were prepared from young inflorescences as described 30 . FISH probes specific for centromeric repeats were amplified from Arabidopsis thaliana and Arabidopsis arenosa genomic DNA and directly labeled with biotin-dUTP and digoxigenin-dUTP, respectively, during PCR. Primers for amplification of Arabidopsis thaliana repeats were CGT TCA ACG GAC CGG ATG ACA AA (180 bp-F) and GAC TTC AGC TAC ACT ATT AAA GT (180 bp-R), those for Arabidopsis arenosa repeats AGT CTT TGG CTT TGT GTC TT (ALU) and TGG ACT TTG GCT ACA CCA TG (ALR). Slide pre-treatment, hybridization and detection were performed as described 31 .

Mature, dry seeds were photographed using a MZ 16 FA stereo-microscope equipped with a set of GFP- and RFP-specific filters and a DFC 300 FX camera (all from Leica, Wetzlar, Germany). Microscopic analysis of chromosomes was done with an AxioImager.Z1 (Zeiss, Jena, Germany) equipped with a Spot Pursuit camera (Diagnostic Instruments, Sterling Heights, MI, USA). Images were assembled and analyzed in Photoshop (Adobe Systems, San Jose, CA, USA).