Novel phenotypes often arise in generations immediately following polyploidization. Previous studies have shown that separate lineages derived from a resynthesized Brassica napus allopolyploid rapidly evolved heritable differences in flowering time. These early-flowering and late-flowering polyploid lines were expected to be genetically identical because they were derived from a single, chromosome-doubled amphihaploid plant. In this study, we investigated the molecular genetic basis for these flowering time differences. We assessed the diploid B. rapa and B. oleracea parents and the early- and late-flowering B. napus lineages for changes in genome structure, and for changes in transcript levels of four sets of FLOWERING LOCUS C (FLC) genes. No such changes were observed for BnFLC1, but we detected chromosomal rearrangements (e.g. de novo non-reciprocal transpositions) and changes in transcript level for BnFLC2 and BnFLC3 between the early- and late-flowering B. napus. A chromosomal rearrangement of a genomic segment containing BnFLC3 was responsible for 29% of the phenotypic variation among the B. napus lines. Expression of BnFLC5 was silenced in all polyploids, although no changes in genome structure were detected. An ongoing investigation of 50 identical B. napus allopolyploids may further reveal the dynamics of changes in phenotype, genome and transcriptome at the early stages in polyploid evolution.
Polyploids often possess novel traits that are not present in their diploid progenitors (reviewed in Ehrendorfer, 1980; Levin, 1983; Ramsey & Schemske, 2002). These novel life-history traits, such as drought tolerance, apomixis (asexual seed production), increased organ size and biomass, and changes in pest resistance or flowering time, could allow polyploids to enter new ecological niches (Lumaret, 1988; Thompson & Lumaret, 1992; Segraves & Thompson, 1999; Husband & Schemske, 2000; see also Brochmann et al., 2004; Thompson, Nuismer & Berg, 2004 – both this issue). Some polyploids also vary in phenotypic plasticity relative to their diploid progenitors (Emery, Chinnappa & Chmielewski, 1994; Bretagnolle & Lumaret, 1995; Petit, Thompson & Bretagnolle, 1996; Bretagnolle & Thompson, 2001; Schranz & Osborn, 2004).
Despite the observation that newly formed polyploids differ from their diploid parents, the molecular genetic mechanisms by which polyploidization contributes to novel phenotypic variation are not well understood. One explanation for the success of polyploids was that the duplicate genes in polyploids may diverge and acquire new functions (neofunctionalization; Force et al., 1999; Lynch & Conery, 2000). Although functional divergence of duplicate genes might confer a selective advantage to polyploids over a long evolutionary time period, it is unlikely to provide any immediate advantage to new polyploids. Because immediate phenotypic effects are seen shortly after polyploid formation, other sources of molecular variation must be operating aside from neofunctionalization (reviewed in Soltis & Soltis, 1993; Matzke et al., 1999; Comai, 2000; Wendel, 2000; Liu & Wendel, 2002, 2003; Ramsey & Schemske, 2002; Osborn et al., 2003b). Osborn et al. (2003b) classified the mechanisms that can lead to novel forms of gene expression in polyploids into three categories: (a) increased variation for dosage-regulated gene expression, (b) altered regulatory interactions and (c) rapid genetic and epigenetic changes.
Investigating the molecular genetic changes that contribute to the difference between polyploids and their diploid progenitors, as well as among polyploids, requires knowledge of both the diploid progenitors and the pedigree (phylogenetic history) of the polyploids. As a result, investigating existing polyploids in their natural environment is problematic because the exact parental genomes are often unknown or have evolved since polyploid formation. In addition to uncertain parentage, untangling the history of natural polyploids can be complicated by multiple-polyploid events and subsequent hybridizations (Soltis & Soltis, 1993; Soltis & Soltis, 1999; see also Abbott & Lowe, 2004; Ainouche, Baumel & Salmon, 2004; Soltis et al., 2004– all this issue). To circumvent these problems, some polyploids can be resynthesized from current forms of diploid progenitors to obtain known pedigrees and genotypes. These synthetic polyploids allow for more precise comparisons with the exact diploid progenitors and a greater range of experimental options. Recent studies have demonstrated rapid genetic changes in resynthesized allopolyploids of Arabidopsis (Comai et al., 2000; Lee & Chen, 2001; Madlung et al., 2002; reviewed by Chen et al., 2004 – this issue), Brassica (Parkin et al., 1995; Song et al., 1995; Quijada, 2003; Udall, 2003; Schranz & Osborn, 2004; see also Lukens et al., 2004– this issue) and Triticum (Feldman et al., 1997; Ozkan, Levy & Feldman, 2001; Shaked et al., 2001; Kashkush, Feldman, & Levy, 2002, 2003; reviewed by Levy & Feldman, 2004 – this issue).
The genus Brassica is an excellent system for creating and studying resynthesized allopolyploids. In fact, the famous ‘Triangle of U’ (U, 1935), which illustrates the crossing relationships between three diploid Brassica species and their three corresponding allopolyploids, was one of the earliest uses of resynthesized polyploids to infer the evolutionary origins of the naturally occurring polyploids (Karpechenko, 1928; Morinaga, 1931; U, 1935). Since these early biosystematic studies, resynthesized Brassica species have been used to illustrate the rapid and extensive evolution of polyploid genomes shortly after polyploid formation (Song et al., 1995). Recently, we have used resynthesized B. napus lines to investigate the effects of polyploidy on Brassica life history traits, with a focus on flowering time (Schranz & Osborn, 2000, 2004).
Although many genes undoubtedly control flowering time in Brassica species, FLOWERING LOCUS C (FLC), a MADS-box gene isolated from Arabidopsis, is a key regulator of the autonomous flowering and vernalization pathways in Brassica (Kole et al., 2001). Arabidopsis has a single copy of FLC (AtFLC) at the distal end of the short arm of chromosome 5 (At5) that acts in a dosage-dependent manner to repress flowering (Michaels & Amasino, 1999; Sheldon et al., 1999). Diploid Brassica species are expected to contain three FLC copies, due to a triplication of this At5 region (Osborn et al., 1997; Axelsson, Shavorskaya & Lagercrantz, 2001; Parkin, Lydiate & Trick, 2002; Schranz et al., 2002; Parkin, Sharpe & Lydiate, 2003; Lukens et al., 2003; see Lukens et al., 2004). These three expected homologues of FLC were isolated from B. rapa (BrFLC1 on chromosome R10, BrFLC2 on chromosome R2 and BrFLC3 on chromosome R3) along with a fourth unexpected homologue, BrFLC5, that mapped to a region of R3 without homology to the top of At5 (Schranz et al., 2002). By contrast, only three homologues of FLC have been identified in B. oleracea (BoFLC1, BoFLC3 and BoFLC5), but other genotypes may possess a BoFLC2 sequence (Schranz et al., 2002). Given that B. napus is an allopolyploid derived from B. rapa and B. oleracea (U, 1935), Schranz et al. (2002) predicted that B. napus should have at least seven BnFLC loci, and eight loci were later mapped in B. napus with four loci in the B. rapa portion of the genome and four loci in the B. oleracea portion of the genome (Quijada, 2003; Udall, 2003). Allelic variation for flowering time has been associated with several BnFLC loci in crosses of B. rapa and B. napus (Schranz et al., 2002; Quijada, 2003; Udall, 2003; reviewed in Osborn & Lukens, 2003).
In this paper, we report new data on the molecular genetic basis for flowering time variation in early- and late-flowering lineages that were directly derived from a resynthesized B. napus allopolyploid (Schranz & Osborn, 2004). Specifically, we measured divergence in flowering time, assayed genotypes to find patterns of rapid genome structural changes (e.g. chromosomal rearrangements) and examined the gene expression of BnFLC loci. We also describe ongoing experiments involving the development and analyses of 50 resynthesized B. napus allopolyploids, which may provide further insight on the dynamics of genome changes associated with polyploidization.
MATERIAL AND METHODS
Plant materials and phenotypic analyses
The early-flowering and late-flowering lineages of Brassica napus (Schranz & Osborn, 2004) were derived from a single chromosome doubled amphihaploid from a cross between B. rapa cv. Reward TO1067 (female parent) and B. oleracea TO1434 (male parent, which is a selfed progeny of TO1000). Thus, all polyploid lines within this lineage were expected to be genetically identical because they were derived from a single B. napus ancestor (TO1141, see Fig. 1). Reward is an oilseed cultivar and TO1000 is an inbred (S5) rapid cycling plant derived from CrGC3-3 (Crucifer Genetics Cooperative, Madison, WI, USA). An individual amphihaploid plant (AC) was colchicine treated to generate an S0 amphidiploid (AACC) by the same process as described by Song, Osborn & Williams (1993). A single S0 plant (TO1141) was self-pollinated to obtain S1 seed. A sister lineage to this S0 resynthesized B. napus plant (TO1147) was used as a parent in a genetic mapping study and was found to contain complete genome complements derived from B. rapa and B. oleracea (Udall, 2003). Two S1 plants (TO1217 and TO1216) were self-pollinated to generate two pools of S2 seed. One hundred S2 plants from each pool were transplanted to a field in Arlington, Wisconsin, in rows spaced 1 m apart and 0.3 m between plants within rows. The hundred S2 plants were measured for flowering time and selected plants were self-pollinated to generate S3 seeds. The five earliest and five latest flowering plants for each of the two pools were selected and self-pollinated, giving rise to 20 lines of S3 seed. Five S3 seeds from each of the 20 selected lines were individually planted in 10-cm pots and grown in a greenhouse. Two lines of S3 seeds were utilized in this study: one early-flowering selection (ES98) and one late-flowering line (ES88). These S3 lines were also used as progenitors for the S4 generation.
Seeds from selected early- and late-flowering S4 plants (LL152 and LL149, respectively) were self-pollinated to generate two pools of S5 seed (Fig. 1). One hundred S5 plants from each pool were sown in 36-well flats in a growth chamber and grown for 3 weeks. The seedlings were then transplanted to a field in Arlington in double-rows, with 1 m between the double-rows and 0.3 m between plants within rows. All plants were measured for days to flowering when the first flower opened (results shown at bottom of Fig. 1). The five earliest plants from the early-flowering line and the five latest plants from the late-flowering line were individually enclosed within pollination bags.
Seeds from these selected S5 plants were grown in a growth chamber, and individual early-flowering (ES341) and late-flowering (ES342) S6 plants were utilized for genetic crosses. Two reciprocal crosses between these two lines were made (ES342 × ES341 and ES341 × ES342). Individual F1 seeds derived from both crosses were grown in 15.2-cm pots in a greenhouse. Two F1 plants, ES349 and ES352 (each from a separate reciprocal cross), were self-pollinated to generate F2 seeds.
Phenotypic analysis of F2generation
For phenotypic evaluation and segregation analysis, 50 F2 plants from each reciprocal cross, and eight seeds from each F1 cross, the S6 polyploid parents, and the original diploid progenitors were grown in PGW-132 growth chambers (Percival Scientific, Boone, IA, USA) under equal light intensity (550 µmol m−2 s−1), as measured at mid-canopy height. The reciprocal crosses were grown in flats that were distributed in the growth chambers in a completely randomized design and grown under long day growth conditions of 16 h of light and 8 h of dark. Light measurements were made using an LI-COR spectro-radiometer LI-1800 (LI-COR, Lincoln, NB, USA). The ratio of R:FR was calculated from the intensity of the light from wavelengths of 655–665 nm for red light and 725–735 nm for far-red light. Growth under 28 215-W fluorescent light bulbs and 12 60-W incandescent light bulbs provided the low R:FR ratio of 1.6. Temperature was maintained at a constant 21 °C. Plants were grown in 24-well flats using Jiffy mix– soil (Jiffy Products of America, Batavia, IL, USA) and were watered with 0.5× Hoagland's solution. All plants were measured for days to flowering when the first flower opened.
Genomic DNA was isolated from lyophilized leaf tissues collected from 12–24 individuals for each genotype using the CTAB procedure described by Kidwell & Osborn (1992) with some modifications (see http://osbornlab.agronomy.wisc.edu/research/protocols/dna.html). DNA was first isolated from six genotypes to be analysed by genetic markers: the two diploid Brassica parents, the S3 and S6 generations of the early-flowering lines (ES98, ES341), and the S3 and S6 generations of the late-flowering lines (ES88, ES342). For these six genotypes, the leaf tissues collected were pools of several individuals. For subsequent DNA isolation from the F2 population, the leaf tissues were collected from individual plants.
Genetic markers: RFLP analysis
The DNA clones included as probes for analyses by restriction fragment length polymorphisms (RFLP) were the same probes used in previous studies (Ferreira, Williams & Osborn, 1994; Teutonico & Osborn, 1994; Thormann et al., 1994; Kole, Vogelzang & Osborn, 1997; Butruille, Guries & Osborn, 1999; Quijada, 2003; Udall, 2003). The procedures used for restriction enzyme digestion, gel electrophoresis, Southern blotting, probe radiolabelling and membrane hybridization are as described by Ferreira et al. (1994). DNA clones were from three libraries: a PstI genomic library, a cDNA library and an EcoRI library (Thormann et al., 1994; Ferreira et al., 1994). The probe nomenclature of Parkin et al. (1995) and Sharpe et al. (1995) was used in which the clones from the PstI and EcoRI libraries were renamed ‘pW’ probes and the clones from the cDNA library were renamed ‘pX’ probes (see http://osbornlab.agronomy.wisc.edu/research.html for details on these probes). In addition, four homologues of FLC, corresponding to BrFLC1, BrFLC2, BrFLC3 and BrFLC5 from B. rapa (Schranz et al., 2002), were used as probes.
DNA from seven genotypes was analysed by Southern hybridization: the two diploid Brassica parents, a DNA mix of the diploid parents, the S3 and S6 generations of the early-flowering lines (ES98, ES341), and the S3 and S6 generations of the late-flowering lines (ES88, ES342). Blots probed with the four BrFLC probes had DNA digested with 12 enzymes (EcoRI, HindIII, MspI, HpaII, DraI, XbaI, EcoRV, PstI, PvuII, HaeIII, HhaI, BamHI). In addition to the FLC probes, 31 other RFLP probes, chosen to cover all the linkage groups of B. napus, were hybridized to blots that had DNA digested with four enzymes (EcoRI, HindIII, MspI, HpaII).
Genetic markers: microsatellite analysis
DNA from the same seven genotypes analysed by Southern hybridization was also used in microsatellite analyses. PCR of diluted genomic DNA samples with microsatellite or simple-sequence-repeat (SSR) primers and electrophoresis of amplified DNA was performed according to Senior & Heun (1993) with some slight modifications (de Leon, 2002). A total of 31 PCR primer sequences were chosen to cover all of the linkage groups of B. napus except N15 (A. Sharpe & D. Lydiate, pers. comm.).
RNA isolation and cDNA synthesis
RNA was isolated from the same six genotypes analysed by genetic markers: the two diploid Brassica parents, the S3 and S6 generations of the early-flowering lines (ES98, ES341) and the late-flowering lines (ES88, ES342). These six genotypes were grown in the same growth chambers and environmental conditions as described above for phenotypic analysis. Two RNA samples were extracted per genotype. These two replications were used in all the following experiments in parallel. Each RNA preparation was a pooled sample of 20–24 plants, gathered from six ‘four-packs’ of plants arranged in a randomized complete block design. Leaf tissue was collected at the four-leaf stage (2- to 3-week old plants) and immediately placed in liquid nitrogen.
Total RNA was extracted from 300 mg of frozen leaf tissue using TRIzol reagent (Sigma, T-9424) according to manufacturer's instructions and treated with RNase-free DNase using the DNA-free kit (Ambion ♯ 1906). Single-strand cDNAs were synthesized using the Superscript First-Strand synthesis system (Invitrogen, ♯11904–018) and 3.5 µg of total RNA as a template according to the manufacturer's instructions. The cDNA products were diluted 1 : 10 with water, and 5 µL of diluted cDNA was used as template for PCR amplification.
Primer design and gene amplification ofFLC
Four FLC genes were amplified using their corresponding specific primers (Table 1). Primers were designed based on the sequences of B. rapa and B. oleracea alleles so as to amplify two homoeologous targets that have nucleotide substitutions in the amplified fragment. Forward primers for BnFLC1 and BnFLC3 were the same as those used in Schranz et al. (2002). The other FLC primers were designed specifically for each FLC in order to prevent cross-amplification. Initially, FLC primer specificity was tested using genomic DNA as a template for PCR to make sure that we were amplifying the correct fragments and to optimize PCR cycling conditions. Subsequently, cDNA of the two diploid parental Brassica lines, a mixture consisting of equal amounts of quantified cDNA from the two diploid parents and cDNA from the four allopolyploid B. napus lines were analysed by RT-PCR to check the specificity of FLC expression levels and to determine the linear stage of amplification.
|FLC1||5′-CTTGAGGAATCAAATGTCGATAA (in exon 4)||5′-CCATCTGGCTAGCCAAAACAT (in exon 6)|
|FLC2||5′-AACATGCTGATGATCTTAAGGCTC (in exon 2)||5′-CCCTGGTTCTCTTCTTTCAGCATT (in exon 6)|
|FLC3||5′-GTGGAATCAAATGTCGGTGG (in exon 4)||5′-AGCCAAAGCCTGATTCTCTTC (in exon 6)|
|FLC5||5′-CCTCGTTGAGCTAGAAGATCA (in exon 4)||5′-GGAGATTTGTCCAGATGACATCTCT (in exon 7)|
|FLC1||5′-CTTGAGGAATCAAATGTCGATAA (in exon 4)||5′-CCATCTGGCTAGCCAAAACAT (in exon 6)|
|FLC2||5′-AACATGCTGATGATCTTAAGGCTC (in exon 2)||5′-CCCTGGTTCTCTTCTTTCAGCATT (in exon 6)|
|FLC3||5′-GTGGAATCAAATGTCGGTGG (in exon 4)||5′-AGCCAAAGCCTGATTCTCTTC (in exon 6)|
|FLC5||5′-CCTCGTTGAGCTAGAAGATCA (in exon 4)||5′-GGAGATTTGTCCAGATGACATCTCT (in exon 7)|
For RT-PCR analyses, a constantly expressed ubiquitin gene was used as a control for DNA contamination and PCR cycles. All the reactions were performed in an Eppendorf Mastercycler. PCR reactions were composed of 0.5 µm of each primer, 0.2 µm dNTP, 2 mm MgCl2 and TaqDNA polymerase with a total volume of 10 µL. The cycling profile was for 4 min at 96°C, followed by 30 s at 94°C, 45 s at 60°C, 1 min at 72°C and then a final 10 min extension at 72°C. Different PCR cycles were applied to each FLC, with 25, 30, 25 and 30 cycles for FLC1, FLC2, FLC3 and FLC5, respectively. The ubiquitin control had 20 cycles.
Single-strand conformational polymorphism (SSCP) analysis
Qualitative homoeologous gene expression was conducted by cDNA-SSCP analysis (Adams et al., 2003; Cronn & Adams, 2003). Briefly, the PCR reaction for SSCP was the same as for RT-PCR except that 0.1 µL 32P dCTP was added to the mixture. Fifty microlitres of loading dye was added to each 10 µL product and 1.7 µL was loaded after denaturing at 95°C for 8 min. SSCP gel was prepared using a 0.5× MDE (Cambrex Bio Science, ♯50620) gel solution. The gel was run at room temperature at a constant 7 W for 15–19 h to separate each FLC homologue. The gel was dried at 70°C for about 1 h and exposed to film for 5–24 h.
Segregation analysis was conducted in F2 populations derived from reciprocal crosses between an early- and a late-flowering line in the S6 generation. One hundred individuals (50 from each cross) were genotyped by PCR amplification using FLC2, FLC3 and FLC5 specific primers. PCR profiling for genomic DNA was the same as used for amplifying cDNA, but with 35 cycles. Primers from conserved regions of exon 2 and exon 5, which amplified both FLC1 and FLC2, were used for FLC2 amplifications to check PCR artefacts. Only FLC3 fragments, showing homoeologous expression patterns, were further analysed for their association with flowering time by analysis of variance (SAS Institute, 2000).
The distributions for days to flowering of the early- and late- flowering B. napus lineages were distinct and had very little overlap: 37–52 days to flowering, with a mean of 41.9 days for the early-flowering lines; 50–58 days, with a mean of 54.4 days for the late-flowering lines (Fig. 1). A two-tailed t-test indicated that the early- and late-flowering lines were significantly different for days to flowering (P < 0.01).
Each of the four BnFLC genes (BnFLC1, BnFLC2, BnFLC3, BnFLC5) was used for genome structure and gene expression analyses. In our materials, the four BnFLC genes occur at seven loci: four orthologues from B. rapa (BrFLC1, BrFLC2, BrFLC3, BrFLC5) and three orthologues from B. oleracea (BoFLC1, BoFLC3, BoFLC5), because BoFLC2 was not found in our genotype of B. oleracea. The map positions of these loci were determined in previous studies (Schranz et al., 2002; Quijada, 2003; Udall, 2003).
BnFLC1 in the allopolyploid B. napus is derived from homoeologous regions from the diploid parents: BrFLC1 is on chromosome 10 of B. rapa (R10) and BoFLC1 is on chromosome 9 of B. oleracea (O9; Fig. 2A). Southern blots probed with BrFLC1 revealed that RFLP fragments found in the diploid Brassica parents were also found in all of the allopolyploid B. napus lines (additive patterns, Fig. 2B). Based on these Southern hybridization results, we constructed ideograms of early-flowering and late-flowering B. napus lines for BnFLC1 on the N10 and N19 chromosome (Fig. 2C). We did not investigate any genotypic markers immediately flanking BnFLC1, but other markers on N10 and N19 showed similar additive patterns, including two SSRs on N10, and two SSRs and the RFLP probe pX140 on N19 (data not shown).
Expression analyses of BnFLC1 showed the same cDNA-SSCP pattern for B. rapa and B. oleracea (Fig. 2D). All B. napus lines also showed this pattern. Owing to the lack of polymorphisms between the diploid parents, we cannot rule out silencing of one of the parental genes in the allopolyploids; however, the expression patterns and the Southern hybridization results provide no evidence that BnFLC1 contributed to the differences in flowering time between the B. napus lines.
BnFLC2 in the allopolyploid B. napus is expected to be derived from homoeologous regions from the diploid parents: BrFLC2 is on chromosome 2 of B. rapa (R2) and BoFLC2 is on chromosome 2 of B. oleracea (O2; Schranz et al., 2002; Quijada, 2003; Udall, 2003). However, Southern blots probed with BrFLC2 revealed RFLP fragments only in B. rapa, the pooled diploid mixture and the early-flowering B. napus lines (Fig. 3A, B). Based on these Southern hybridization results, we constructed ideograms of early-flowering and late-flowering B. napus lines for the putative location of BnFLC2 on the N2 or N2 and N12 chromosomes of the early-flowering lines but the absence of BnFLC2 in late-flowering B. napus lines (Fig. 3C). The absence of FLC2 on N2 in the late-flowering line can be explained by the occurrence of a non-reciprocal transposition involving a segment of N12 to the homoeologous position N2. This explanation is supported by Southern blot results using the probe pW241 that detected a marker locus flanking FLC2 on R2, and in the homoeologous position on O2 (Fig. 3A). The late-flowering lines are missing the N2 locus and appear to have two copies of the N12 locus based on a visual inspection of fragment intensity (Fig. 3B). Results from pW241 also suggest that a non-reciprocal transposition occurred between the S3 and S6 generations of the early-flowering lines. The S3 early-flowering line contains both the B. rapa and the B. oleracea pW241 loci, whereas the S6 early-flowering line is missing the B. oleracea locus and appears to have two copies of the B. rapa locus (Fig. 3B). The S6 early-flowering line also appears to have two copies of BrFLC2, based on fragment dosage (Fig. 3B). Two microsatellites (SSRs) on N2 and one SSR on N12 that are well below the mapped locations for BnFLC2 showed an identical additive pattern for all polyploids (data not shown), indicating that the hypothesized rearrangements involved only an upper portion of the N2 and N12 chromosomes.
The results from expression analyses of BnFLC2 using RT-PCR were consistent with the Southern blot results. Gene expression was observed in B. rapa, the pooled mixture and the early-flowering B. napus lines, but not in B. oleracea or the late-flowering lines, which were missing copies of FLC2 (Fig. 3D). The S6 early-flowering line appeared to have more BnFLC2 transcripts, consistent with a double-dose of the BrFLC2 gene
If the early-flowering parent had two copies of BrFLC2 and the late-flowering parent had zero, we would expect a digenic segregation ratio with 1 : 16 having no FLC2. We analysed the F2 population for segregation of BrFLC2 and found zero F2 individuals with no copies of the gene. Thus, we could not test for an association of the BnFLC2 changes with flowering time. The lack of fit between the observed and expected segregation ratios may be due to unusual segregation associated with the chromosomal rearrangement.
BnFLC3 in the allopolyploid B. napus is derived from homoeologous regions from the diploid parents: BrFLC3 is on chromosome 3 of B. rapa (R3) and BoFLC3 is on chromosome 3 of B. oleracea (O3; Fig. 4A). Southern blots probed with BrFLC3 revealed an additive pattern for the late-flowering lines (both BrFLC3 and BoFLC3 present); however, the early-flowering B. napus lines have lost BoFLC3 and appear to have a double-dose of BrFLC3 (Fig. 4B). These results can be explained by a non-reciprocal transposition of an N3 segment to the homoeologous position on N13 in the early-flowering lineage (Fig. 4C). Further evidence for the rearrangement was provided by probe pX133, which identified loci adjacent to BnFLC3 on N3 and N13 (Fig. 4A, C) having similar patterns to BnFLC3 in the early- and late-flowering lines. Other markers on N3 or N13, including one SSR on the bottom of N3, two SSRs on the middle and bottom of N13, and several RFLP probes that hybridized to loci on N3 and N13 (including BnFLC5), detected additive patterns in all polyploids, indicating that the rearrangements involved only a distal portion of N13 (data not shown).
The results from expression analyses of BnFLC3 using cDNA-SSCP were consistent with the Southern hybridization results. The early-flowering lines did not express BoFLC3 and appeared to have higher levels of BrFLC3 expression (Fig. 4D). We also observed a decrease in expression of BrFLC3 between the S3 and S6 late-flowering lines. This may be explained by a change in gene dosage (from 2 BoFLC3 : 2 BrFLC3 to 3 BoFLC3 : 1 BrFLC3) owing to a non-reciprocal transposition event in the S5 generation that had not yet become homozygous. This explanation is supported by results from the Southern blot showing a dosage shift between the S3 and S6 generations (Fig. 4B).
We performed a segregation analysis of BnFLC3 on the F2 population. Figure 5A shows a model of inheritance for BnFLC3 based on the F2 from the cross between the S6 early- and late-flowering lines. The distributions of the two reciprocal F2 populations for flowering time were almost identical (means of 46.2 and 49.5 days). Thus, there does not appear to be a maternally inherited effect for the control of flowering time. Given this, and the fact that the flowering time–population interactions were not significant (P = 0.93), we combined the individuals from the two populations for all subsequent analyses (Fig. 5B–E). The distribution of days to flowering for the c. 100 individuals in the two F2 populations was mostly between the flowering times of the allopolyploid parents (Fig. 5B).
The segregation ratio for BnFLC3 on N13 was not statistically different from the expected 1 : 2 : 1 ratio for allelic segregation of BrFLC3 and BoFLC3 (17 Br/Br : 53 Br/Bo : 24 Bo/Bo; χ2 = 2.6, P = 0.27). The segregation of BnFLC3 was associated with flowering time. Plants with two B. rapa alleles flowered significantly earlier (mean of 43.9 days to flowering, Fig. 5C) than those with two B. oleracea alleles (mean of 52.1 days to flowering, Fig. 5E) (P < 0.0001). Thus, the expected effect of substituting one B. rapa allele for one B. oleracea allele was a delay in flowering of 4 days. Segregation of BnFLC3 explained 29% of the phenotypic variation for days to flowering in the ANOVA of the F2 population.
BnFLC5 in the allopolyploid B. napus is derived from homoeologous regions from the diploid parents: BrFLC5 is on chromosome 3 of B. rapa (R3) and BoFLC5 is on chromosome 3 of B. oleracea (O3; Fig. 6A). Southern blots probed with BrFLC5 revealed an additive pattern for the early- and late-flowering lines (both BrFLC5 and BoFLC5 present, Fig. 6B). Based on this result and the map positions of FLC loci from other studies (Quijada, 2003; Udall, 2003), the break point of the transposition on N13 in the early-flowering line appears to be between BnFLC3 and BnFLC5 (Fig. 6C).
RT-PCR analysis of BrFLC5 indicated expression only in the B. rapa parent. We did not detect BoFLC5 expression in B. oleracea or in any of the allopolyploid B. napus lines (Fig. 6D). Unlike other BnFLC loci, the expression patterns differed from the RFLP patterns seen on Southern blots (Fig. 6B). We might expect silencing of BoFLC5 in the allopolyploid lines, because this was not expressed in the diploid parent, but silencing of BrFLC5 in the allopolyploid line was unexpected. However, given additivity of the gene and absence of expression in both early- and late-flowering lines, we have no evidence that BnFLC5 expression differed among the polyploids or had an affect on the divergence in flowering time.
The results from our study show that selection among the progeny of a resynthesized allopolyploid B. napus was effective in creating lineages with divergence in flowering time. Within six generations, there was almost no overlap of these lineages in the time to initiation of flowering, and in nature this rapid evolution may be sufficient for ecological separation and sympatric speciation. The phenotypic differentiation between these lines was more fully assessed in another study (Schranz & Osborn, 2004), in which significant differences were found for eight life-history traits. What mechanisms could cause the phenotypic differences between these early- and late-flowering lineages in B. napus? Based on previous studies (Michaels & Amasino, 2000; Schranz et al., 2002; Schranz & Osborn, 2004), we chose four pairs of BnFLC genes as candidates to begin our investigation into the molecular mechanisms that could cause these rapid phenotypic changes. We found that three of the four loci (BnFLC2, BnFLC3, BnFLC5) had undergone structural and/or expression changes in the allopolyploids, and have strong evidence that genomic rearrangements at one locus (BnFLC3) were associated with the divergence of flowering time.
BnFLC1 has been shown to have a large impact on flowering time in B. napus (Tadege et al., 2001) as well as B. rapa (Kole et al., 2001). However, we found no qualitative differences for BnFLC1 between the early- and late-flowering lines for either gene structure or gene expression, and thus no evidence that BnFLC1 is a cause of differences in flowering time between the lines. However, we cannot rule out quantitative differences in expression levels of BnFLC1 or differences in post-transcriptional modifications.
For BnFLC2, a structural change occurred between the early- and late-flowering lineages that was associated with expression differences among the lineages. The early-flowering lines had BrFLC2 from N2 duplicated on N12 whereas the late-flowering lines had the chromosomal fragment from N12 (which lacks a corresponding BoFLC2) duplicated on to N2. Thus, BrFLC2 was deleted from the late-flowering lines, resulting in no detectable expression for this gene. This finding is surprising because the early-flowering lines show expression of FLC2, a repressor of flowering, and the late-flowering lines did not. One might expect the opposite pattern, in which the expression of FLC2 would be seen in the late-flowering lines and not the early-flowering lines. It is possible that this rearrangement does not affect the flowering time differences in these lines because their FLC2 genes are non-functional or have diverged to another function. It is also possible that BrFLC2 is non-functional but it interferes with the function of other FLC genes. In this case, BrFLC2 might act as a negative repressor of flowering and its presence would be associated with early-flowering as was observed in our study. However, we did not obtain data that would allow us to test for a genetic association between the changes involving flowering time.
For BnFLC3, gene structure and expression changes were discovered and these changes were related to differences in flowering time based on a genetic segregation analysis. In the early-flowering lines, BoFLC3 was lost and BrFLC3 was duplicated from N3 to N13. The late-flowering lines had both BrFLC3 and BoFLC3, although the S6 generation may have been heterozygous for a rearrangement on N3 (Fig. 4D). The segregation analysis confirmed that replacement of the BoFLC3 allele on N13 with BrFLC3 was associated with earlier flowering. Although the BoFLC3 gene came from a rapid-cycling line, the diploid B. rapa parent flowered earlier than the B. oleracea parent and it is possible that BoFLC3 expresses at higher levels or is a more effective inhibitor of flowering than BrFLC3.
The results for BnFLC5 are interesting because a parental gene is silenced within the allopolyploids. The patterns of gene expression clearly do not match the gene structure data. The RFLP results indicate that each of the diploid parents has FLC5 and that all the polyploid lines have both BrFLC5 and BoFLC5. The lack of BoFLC5 expression in the parent and polyploids may be due to a pre-existing mutation, but the silencing of BrFLC5 in the polyploids would have to be due to genetic or epigenetic changes that arose with or after polyploid formation. We did not observe any changes in DNA methylation of BrFLC5 in the polyploids compared with the B. rapa parent based on comparing restriction fragment patterns after digesting with methylation-sensitive and -insensitive enzymes (MspI and HpaII). It is possible that further analyses will reveal methylation changes or some other chromatin modification that silenced the gene. Alternatively, the process of polyploidization may have altered the regulatory network that allows expression of BrFLC5 (Osborn et al., 2003b).
Our results indicate that extensive genomic rearrangements (e.g. de novo non-reciprocal transpositions) can be an important source of novel phenotypic variation in resynthesized B. napus. De novo chromosomal rearrangements have been observed in previous studies of segregating populations derived from natural and/or resynthesized B. napus that involve the transposition of segments between homoeologous chromosomes (Parkin et al., 1995; Sharpe et al., 1995; Osborn et al., 2003a; Quijada, 2003; Udall, 2003). These studies suggest that there are ‘hot spots’ for homoeologous recombination in the B. napus genome resulting in three classes of transposition products found in mapping studies: segregating reciprocal transpositions (RTs), segregating non-reciprocal transpositions (NRTs) and de novo NRTs. Quijada (2003) and Udall (2003) found 50 de novo NRTs and several segregating NRTs in four mapping populations of B. napus DH lines, and some segregating NRTs were associated with variation in flowering time (e.g. FLC2 associated with an N2/N12 transposition). Our study found that a chromosomal rearrangement was associated with novel variation in flowering time. Specifically, we found that a de novo N3/N13 transposition that contained BnFLC3 accounted for a significant portion of the phenotypic variation between the B. napus lines. Although FLC is a likely candidate for the observed flowering time affects associated with this rearrangement, other flowering time genes that are linked to FLC3 on N3/N13 (e.g. CONSTANS, Axelsson et al., 2001) also could contribute to this effect. Changes in other flowering time genes (Levy & Dean, 1998; Mouradov, Cremer & Coupland, 2002; Simpson & Dean, 2002) in other genomic regions also may have contributed to the divergence between the B. napus lines.
The study of synthetic polyploids allows us to recognize the phenotypic changes and the molecular genetic components that may influence them following polyploidization. Despite the fact that we only examined a single trait, a single gene and two lineages derived from a single resynthesized polyploid, after a few generations we observed very large phenotypic changes, gene loss and possible epigenetic regulatory changes. Important questions remain: how broad are these phenomena? Are the same genes, genic regions and traits sensitive to polyploid-induced changes to the same degree? To address these questions, we are currently performing a survey of phenotypes, genomes and transcriptomes (see Chen et al., 2004, for detecting genome-wide expression changes in Arabidopsis and Brassica polyploids) for a set of 50 replicates of B. napus synthetic polyploids created de novo from better-known diploid parents. The progeny of the 50 independently derived S0B. napus should be genetically identical, and this gives a strong null hypothesis for observations and statistical tests. In theory, any changes observed should be due to the processes of hybridization or polyploidization.
Our study of early- and late-flowering lines has shown that genetic changes associated with polyploid formation can lead to divergence in a phenotype with important agricultural and ecological implications. Far from being a ‘dead end’ (Wagner, 1970), polyploidy is associated with remarkable rapid evolution and variation. Using a replicated set of 50 polyploid lines, we will broaden this view by examining the frequency of dynamic changes. Ultimately, our goal is to relate studies of resynthesized polyploids back to the mechanisms that occur in natural polyploids.
We thank Andrew Leitch, Z. Jeffrey Chen, Josh Udall and an anonymous reviewer for their thoughtful comments on the manuscript. Josh Udall assisted with probe selection, Ivan Maureira assisted with the segregation analysis and Amorntip Muangprom assisted with the gene expression analysis. Funding was provided by the NSF Plant Genome Program (♯0077774) and USDA-NRI (♯ 2002-35301-12071) to T.C.O.