Identifying a locus influencing petal area.
We measured the length, maximum width and area of petals of 272 Arabidopsis thaliana accessions collected from southern and northern Sweden (Additional File 1) that were grown in controlled conditions after vernalization. Additional File 2 shows petal phenotype measurements. All three petal parameters varied substantially within the sampled collection. For example, mean petal areas varied from 0.915 mm2 (Hov1-10) to 4.92 mm2 (Vår2-6), a difference of 537%. Additional File 3 shows representative petals from these accessions and from Död 1, an intermediate size for comparison. Figure 1A shows that mean petal area variation formed a normal distribution and was therefore suitable for association studies. GWAPP was used [16] with an Accelerated Mixed Model, incorporating information across 250,000 SNPs. This analysis identified a significant SNP association on chromosome 4 for petal area (Figure 1B). The most significantly associated SNP within this region was located at position 9471419 bp, within gene model At4G16830. The marker at this position was bi-allelic, with those accessions carrying an “T” allele at this locus exhibiting a ~15% increase in petal area relative to those carrying the alternative “A” allele (Figure 1C). The extent of Linkage Disequilibrium (LD) in the region was visualised using information for all SNP markers within +/-10 kb the 9471419bp position from each accession and colour-coded based on the allele present. These markers, in chromosome order, were then sorted by phenotype values. This identified a clear block of LD (Figure 1D), with accessions exhibiting larger petals carrying a distinctive set of alleles from those with smaller petals. This block of LD spanned six Arabidopsis gene models, from At4g16820 to At4g16850. Sequence variation altering the activities of any of these genes may explain the variation in petal size observed across accessions. Assessment of gene annotations revealed no known regulators of petal or organ size. The effect of genetic variation within the haplotype defined by LD on petal growth was assessed in a subset of three accessions with small petal areas and three accessions with large petal areas (Figure 1E). Petal cell areas and numbers were quantified using Scanning Electron Microscopy (SEM) and Image J. A significant increase in petal abaxial epidermal cell area was observed in accessions carrying the increasing T allele at position 9471419 relative to accessions carrying the decreasing A allele (Figure 1E). Therefore, the major effect of genetic variation in the haplotype was on petal cell area.
Expression levels of At4g16850 are correlated with quantitative variation in petal size
To assess the potential role of the 6 candidate genes in regulating petal growth, petal areas were measured in available potential loss-of-function T-DNA mutants in the accession Col-0. T-DNA insertion lines were available from stock centres for all genes found to be in high LD with associated markers with the exception of At4g16850, a small hypothetical gene of unknown function. We measured the expression of these 6 candidate genes in developing floral tissues in the six accessions with varying petal sizes. For At4g16820 to At4g16845 no differences in petal area were seen in the T-DNA insertion lines relative to Col-0 plants (Figure 2C). Furthermore, no differential expression of these genes in developing flowers was observed between the six accessions with small and large petals (Figure 2B). However, for At4g16850, a gene of unknown function, there was an increase in petal transcript levels in the three accessions with larger petals (P ≤ 0.001) (Figure 2B). At4g1650 expression was also increased in seedlings of these three high-expressing accessions (Additional file
Polymorphisms in the At4g16850 gene region
The relationship between increased petal areas and increased expression of At4g16850 in the selected accessions suggested that sequence variation between accessions may influence At4g16850 expression. Inspection of available genome sequence reads [13] from the larger petal accessions Dju-1, T1070 and T880 showed a limited range of SNP and possible small indel variation in the coding region and flanking sequences compared to the Col-0 assembly. To identify a wider range of promoter sequence variation, a region 2 kb upstream of At4g16850 in three accessions with the decreasing A allele (Roed-17, Lis-3, Had-1) and three with the increasing T allele (Dju-1, T1070, T880) were amplified by PCR, cloned and sequenced to identify the precise location and types of sequence variation in the putative promoter regions of At4g16850. Primers were designed in conserved regions of all accessions. The upstream regions were readily amplified from the three smaller petal accessions and were found to be very similar to the sequence of the Col-0 promoter (Additional File 4) carried the decreasing A allele (Figure 1C), consistent with its relatively small petal phenotype. Col-0 was therefore selected as the “small petal” reference genome due to the high level of sequence conservation between small petal accessions and Col-0 at the locus. However, no full-length promoter amplicon could be generated from any of the large petal accessions. We therefore generated whole genome assemblies from Illumina sequence of un-amplified DNA templates [17] made from three large petal accessions to access sequence variation in At4g16850. An ABYSS de novo assembly generated a large contig spanning the region upstream of the At4g16850 in line Dju-1. Comparison to the Col-0 small petal sequence identified multiple variants (Figure 3A and Additional File 4). Notably, the Col-0 and Dju-1 promoters had a common 23bp dA:dT-rich region that was extended by 30 bp in the Dju-1 promoter, making an approximately 50 bp dA:dT-rich region in Dju-1. It is likely that this dA:dT richness impeded PCR amplification of full-length upstream regions of large-petal accessions. There were also many other promoter polymorphisms, including another large dA/T-rich insertion in the intergenic region of Dju-1 compared to Col-0, and a deletion in Dju-1 compared to Col-0 in the 5’UTR intron (Figure 3A).
At4g16850 encodes a predicted 6-transmembrane domain protein with 3 non-cytoplasmic domains and 4 cytoplasmic domains (Figure 3B). Comparison of the Dju-1 and Col-0 assemblies revealed the predicted protein was highly conserved between these large- and small-petal accessions, with only two non-conservative amino acid changes in trans- membrane region 4 and in the C-terminal cytoplasmic domain (Figure 3B). To assess the predicted subcellular location of the protein encoded by At4g15850, its coding region was fused at its C-terminus with GFP and transiently expressed from the 35S promoter in Col-0 developing petal protoplasts, together with a known transmembrane receptor-like kinase TMK4 [18] fused to RFP. Confocal imaging showed that the At4g16850-GFP fusion protein co-localised with the RFP-tagged TMK4 plasma membrane protein (Figure 3C), demonstrating that it can be localised to the plasma membrane. At4g16850-GFP fusion protein was also observed in cytoplasmic structures.
Overexpression of At4G16850 increases petal size due to increased cell growth
Analysis of the expression of At4G16850 across accessions displaying high variation for petal area established that differential expression accounted for 76% of the variation in petal size in the analysed accessions (Figure 4A). To establish whether this variation in At4g16850 caused petal size variation, the coding region of At4g16850 from Col-0 was expressed from the constitutive 35S promoter in transgenic Arabidopsis Col-0 plants. Col-0 has relatively small petals and inherits the decreasing allele in the associated haplotype that segregates with low At4G16850 expression. Therefore Col-0 was an appropriate accession in which to observe any expected increase in petal size following overexpression of At4g16850. Comparison of petal areas in transgenic lines and untransformed Col-0 plants revealed that all transgenic plants overexpressing At4G16850 (lower panel) exhibited significantly increased petal size relative to WT plants (P ≤ 0.01) (Figure 4B). No other phenotypic differences between Col-0 and the transgenic lines were observed. Therefore, increased expression of At4g16850 leads to increases in petal area by approximately 125%, indicating that variation in At4g16850 expression among the accessions directly influences petal area. In the tested accessions exhibiting increased At4g16850 expression, cell areas were increased (Figure 1E). Petal cell areas were also increased in transgenic lines overexpressing At4g16850 by approximately 175% (Figure 4C). This suggested that there were fewer larger cells in these larger petal. Taken together, these results show that increased expression of At4g16850 promotes cell growth in Arabidopsis petals. To take account of this information about a previously unknown gene in Arabidopsis thaliana we named the gene KSK (KronbladStorleK, Swedish for petal size).
Increased expression of KSK reduces expression of genes that limit petal cell growth.
Previous studies have identified several genes that influence petal cell growth in Arabidopsis. BPEp [19] and ARF8 function together [20] to limit petal cell growth, and FRL1 [21] also represses petal cell growth. The expression of these genes in developing petals of three transgenic Col-0 lines over-expressing At4g16850 and in untransformed Col-0 was measured using Q-RT-PCR to assess whether KSK may influence petal cell growth through these genes. Although only one transgenic line showed significant reduction in BPE expression in petals (Figure 5A), consistent reductions in ARF8 and FRL1 expression in developing transgenic petals was seen (Figures 5B, C) compared to Col-0. This suggested that KSK may promote petal cell growth by reducing the expression of these petal cell growth genes. AGAMOUS reduces BPEp expression [19] and the ag1 loss of function mutant has larger petals [22], consistent with the larger cell and petal sizes in BPEp loss of function mutants. Although we did not observe consistent reduction of BPEp expression in all KSK overexpressing transgenic lines, we tested whether AG influences KSK expression. KSK expression was doubled in the ag1 loss-of-function mutant (Figure 5D) consistent with a model in which AG repressed KSK expression, leading to increased ARF8 and FRL1 expression and corresponding reduced petal cell size and overall petal area.
Overexpression of KSK leads to partial homeotic conversion of petals to stamenoid structures.
In addition to observing a significant increase in petal cell growth in the 35S::KSK over-expressing lines, we also observed partial organ identity changes in ~10% of flowers from all eight 35S::KSK transgenic plants. Flowers with organ identity changes had an additional petal-like structure in the second whorl. This developed in the outer margin of the second whorl and displayed varying extents of stamenoid features such as a partial pollen sac (Figure 6A) and stomata, a cell type not observed in Col-0 petals (Figure 6B). Stamen numbers and development in the third whorl were normal in these flowers. The presence of stomata on petal epidermal surfaces has also been seen in ant mutants deficient in the transcription factor AINTEGUMENTA (Krizek 2000). Using qRT-PCR, we assessed ANT expression in developing petals of the three 35S::KSK over-expressing lines. A significant decrease in ANT expression was observed in petals overexpressing KSK (Figure 6C). This suggests that KSK expression levels contribute to determining floral organ identity in a pathway involving ANT.
Reduced auxin responses in lines over-expressing KSK.
One common feature of ARF8 and FRL1 expression, which was suppressed by over-expression of KSK, is that the expression of both genes increases in response to auxin [20, 23, 24]. We therefore tested auxin responses in petals of wt Col-0 and 35S::KSK over-expressing lines. The DII-n3Venus reporter protein is a fusion of the auxin-dependent DII degron to a nuclear-localised Venus reporter coding region. Auxin levels are detected by reduced Venus fluorescence relative to a mutant form, mDII-ntdTomato, which is not degraded in the presence of auxin [25]. This dual auxin reporter, called R2D2, is suitable for single cell assays as different transformation efficiencies can be accounted for by the relative fluorescence of nuclear-localised Venus and Tomato fluorescent proteins. Protoplasts were isolated from developing petals from Col-0 and a 35S::KSK transgenic line over-expressing KSK and transfected with the R2D2 plasmid. After 16 hrs to allow for protein expression, protoplasts were treated with either 0 nM or 1000 nM IAA. Protoplasts were imaged between 1-2 hr after IAA treatment. The nuclear localisation of both fluorophores is shown in Figure 6D, while Figure 6E shows relative fluorescence of mDII-Tomato/DII-Venus. The ratio of mDII-Tomato to DII-Venus was significantly increased in Col-0 protoplasts, demonstrating that transiently-expressed petal protoplasts respond to added auxin similarly to stable transgenic plants [25]. In contrast, in 35S::KSK transgenic protoplasts, the ratio of Venus to Tomato fluorescence was not decreased to the same extent as Col-0. This indicated that auxin responses may be reduced in this transgenic line. This interpretation was tested by measuring the expression levels of two auxin- responsive genes (IAA1 and IAA9) in petals of Col-0 and KSK over-expressing lines. Their expression was reduced (Figures 6F, G), supporting the interpretation that auxin responses are decreased in KSK over-expressing lines.