Flavonoids are determinants of freezing tolerance and cold acclimation in Arabidopsis thaliana

In plants from temperate climates such as Arabidopsis thaliana low, non-freezing temperatures lead to increased freezing tolerance in a process termed cold acclimation. This process is accompanied by massive changes in gene expression and in the content of primary metabolites and lipids. In addition, most flavonols and anthocyanins accumulate upon cold exposure, along with most transcripts encoding transcription factors and enzymes of the flavonoid biosynthetic pathway. However, no evidence for a functional role of flavonoids in plant freezing tolerance has been shown. Here, we present a comprehensive analysis using qRT-PCR for transcript, LC-MS for flavonoid and GC-MS for primary metabolite measurements, and an electrolyte leakage assay to determine freezing tolerance of 20 mutant lines in two Arabidopsis accessions that are affected in different steps of the flavonoid biosynthetic pathway. This analysis provides evidence for a functional role of flavonoids in plant cold acclimation. The accumulation of flavonoids in the activation tagging mutant line pap1-D improved, while reduced flavonoid content in different knock-out mutants impaired leaf freezing tolerance. Analysis of the different knock-out mutants suggests redundancy of flavonoid structures, as the lack of flavonols or anthocyanins could be compensated by other compound classes.


Results
We used a collection of 20 mutants in two different Arabidopsis thaliana accessions (Col-0 and Ler) that cover all major regulatory and biosynthetic steps in flavonoid biosynthesis ( Fig. 1) to investigate the role of flavonols and anthocyanins in plant freezing tolerance and cold acclimation. Nineteen of these mutants are single (15 lines), double (three lines) or triple (one line) gene knock-outs. In addition, we included the activation tagging line pap1-D, which exhibits constitutive overexpression of the gene PAP1/MYB75 encoding the PRODUCTION OF ANTHOCYANIN PIGMENT 1 (PAP1) MYB transcription factor. The ko mutants showed no obvious phenotypic differences compared to the corresponding wild type plants under either non-acclimating or cold acclimating (14 days at 4 °C) conditions. This is in agreement with the finding that our cold acclimation conditions did not induce significant accumulation of reactive oxygen species (ROS) 29 or ROS-related chilling stress that might require flavonoids for protection 15,30 . The pap1-D mutant, on the other hand, exhibited a purple leaf phenotype due to the activation of the whole flavonoid biosynthetic pathway 31 . Effects of mutations on the expression of flavonoid biosynthesis genes. We used quantitative RT-PCR to characterize the effects of the mutations on transcript levels of 26 genes encoding transcription factors regulating flavonoid biosynthesis and enzymes in the flavonoid biosynthetic pathway ( Fig. 2; Suppl. Table 1). As described in detail recently, this analysis revealed a strong cold-induction of the pathway at the transcript level, with the exception of TTG1/WD40 and OMT1, which showed high constitutive expression 20 . The ko mutant lines generally showed a significant decrease in the abundance of the transcripts encoded by the affected genes compared to the wild type under both cold acclimated and non-acclimated conditions (Figs 3 and 4; Suppl. Table 4). Notable exceptions were the lines f3′ h, f3′ h 3gt and myb11 myb12 myb111 in Col-0, and both tt4-1 lines in Ler.
The T-DNA insertion in the F3′ H/TT7 gene is localized in the fourth exon immediately upstream of the 3′ UTR region, rendering the design of qRT-PCR primers that bind beyond the T-DNA impossible. This led to the detection of partial transcripts with our primers binding in the second exon. The MYB12 transcript in the triple mutant contains a single base pair deletion (myb12-1f) and is measurable by qRT-PCR, but results in premature termination of translation 32 . Similarly, the detected CHS/TT4 trancripts in the tt4-1 mutants in Ler are nonfunctional due to point mutations 33 . It remains unclear why the expression of UGT89C1/F7RT was significantly reduced in the 7rt3 3gt double mutant, but was almost unchanged in the 7rt3 single mutant line.
In general, mutations in the genes encoding enzymes had only minor effects on the abundance of all other investigated transcripts ( Fig. 4; Suppl. Table 4). On the other hand, knock-out of the MYB11, MYB12 and     Table 2). Three kaempferol and two quercetin derivatives, three major anthocyanins (A8, A9 and A11), four minor flavonols and one minor anthocyanin (A7) as well as seven unidentified flavonols and two unidentified anthocyanins could be quantified. A preliminary identification of these compounds based on MS/MS fragmentation analysis is provided in Suppl. Table 2. The significance of differences in flavonoid content between mutants and wild types ( Fig. 6; Suppl. Table 5) was evaluated by t-tests corrected for multiple testing errors 34 . Changes in Arabidopsis flavonoid content during cold acclimation have been reported recently 20 and will not be considered here. Figure 5 shows that most mutations resulted in the absence or a strong reduction of major flavonoids, while only pap1-D showed a strong accumulation, particularly of anthocyanins, but also of the two major quercetin derivatives. Additionally, kaempferol-3Glc increased massively in the acclimated state in this mutant. On the other hand, myb11 myb12 myb111 showed a strong reduction in the content of all flavonols, but not of anthocyanins. All four investigated tt4 lines in both accessions completely lacked flavonoids, while the tt5 and tt6-1 lines, also affected in the initial steps of the pathway, still contained low, significantly reduced levels of flavonols and anthocyanins, particularly under cold acclimated conditions. Similarly, the lines 75c1 and tt18 in Col-0 and tt3-1 in Ler were specifically reduced in anthocyanin content.
All other mutant lines showed more specific effects on the content of particular flavonoids. For example, the f3′ h lines lacked all major quercetin and anthocyanin derivatives, whereas the f3′ h 3gt line additionally showed reduced levels of glucosylated kaempferols that were also affected in line 3gt, while 3rt was specifically reduced in its kaempferol-3Rha-7Rha content. The line 7rt3 showed a strong reduction of all central 7-O-rhamnosylated flavonols. Noteworthy, 7rt3 and f3′ h single and double mutant lines as well as 75c1 accumulated mutant-specific flavonol or anthocyanin derivatives (see Suppl. Table 2 for preliminary structure assignments) which might be evidence of compensatory effects. These findings are consistent with previous studies under normal (non-acclimated) growth conditions 4,7,9,31,[35][36][37][38] .  Effects of mutations on leaf freezing tolerance. To assess whether flavonoids play a functional role in leaf freezing tolerance, we compared wild type and mutant lines both before and after cold acclimation (Fig. 7), using an electrolyte leakage assay 39 . Only the pap1-D line showed significantly improved freezing tolerance under acclimated conditions, which can be attributed to the accumulation of quercetin and anthocyanin derivatives, while all ko lines showed either reduced freezing tolerance or were not significantly affected.
Mutants in the three earliest steps of the pathway (tt4, tt5, tt6) which showed a massive reduction in both flavonols and anthocyanins, exhibited a strong reduction in freezing tolerance in the Ler background, in particular  Gray fields indicate that no t-test could be performed because the metabolite was undetectable in these samples (see also Suppl. Table 5).
after cold acclimation. Interestingly, while the tt5 mutant behaved similarly in Col-0, this was not found for the tt4 mutants in this accession, which only showed reduced freezing tolerance in the non-acclimated state. On the other hand, the tt3-1 mutant in Ler showed a significant reduction in cold acclimated freezing tolerance, although only small, non-significant differences in flavonoid content could be detected compared to the wild type (Figs 5 and 6).
No changes in freezing tolerance were observed in mutants where only anthocyanin (tt18, 75c1) or only flavonol contents (myb111 myb12 myb111) were reduced. However, when all flavonoids were strongly reduced (tt5, tt6-1 and 3rt 3gt) freezing tolerance was impaired at least under one condition. The reduction of all 3-O-glucosylated flavonoids in 3gt reduced freezing tolerance under both conditions, while a complete lack of only kaempferol-3Rha-7Rha in 3rt resulted in slightly impaired freezing tolerance only under non-acclimated conditions. The absence of all major flavonols in 7rt3 accompanied by the occurrence of flavonols with unknown structure at unchanged anthocyanin levels had no impact on freezing tolerance. The additional reduction of anthocyanins in the 7rt3 3gt double mutant led to impaired non-acclimated freezing tolerance. In addition, a complete lack of quercetin and anthocyanin derivatives with unchanged kaempferol levels in f3′ h had no significant effect on freezing tolerance. The f3′ h 3gt double mutant line was additionally reduced in kaempferol-3Glc-7Rha and kaempferol-3Glc2″ 7Rha-7Rha and therefore only unaffected in the content of kaempferol-3Rha-7Rha resulting in impaired freezing tolerance after cold acclimation.

Effects of mutations on primary metabolism. Cold acclimation is accompanied by a massive accu-
mulation of many primary metabolites such as sugars and amino acids that may act as compatible solutes to improve cellular freezing tolerance 27,28 . We therefore checked whether the mutations in secondary metabolism had additional effects on primary metabolism. We compared the relative pool sizes of 114 primary metabolites (34 organic, amino and hydroxyl acids, four N-compounds, six P-compounds, 15 sugars and 55 unknown putative primary metabolites) measured by GC-MS (Suppl. Table 6) between wild type and mutant plants under both non-acclimated and acclimated conditions. This analysis showed only very little effect of mutations in the flavonoid biosynthetic pathway on primary metabolism under either condition, with only eight significantly different pool sizes among 114 metabolites in 40 comparisons (Suppl. Table 6) in agreement with an earlier study that was conducted only under non-acclimating conditions 4 .

Discussion
Comprehensive characterization of the flavonoid mutants with respect to their freezing tolerance, gene expression and metabolite profiles indicates a role of flavonoids in freezing tolerance and cold acclimation. Previous work has implicated flavonoids in the defence against ROS (e.g. ref. 15), although the importance of flavonoids for ROS scavenging in vivo has been questioned 40 . This ROS-scavenging activity has been invoked mainly for plant chilling tolerance, i.e. tolerance against low temperatures above 0 °C, while here we investigated the tolerance of plants against freezing, i.e. the stresses associated with apoplastic ice formation at sub-zero temperatures. A clear physiological role for flavonoids, on the other hand, has been established for their UV-B screening function 18,30,41 . A function of flavonoids in plant freezing tolerance has been hypothesized in Arabidopsis due to the correlation of both metabolite and transcript abundances with freezing tolerance 19,20,23 , but a clear causal relationship had not been established previously.
In contrast to the overexpression line pap1-D with improved freezing tolerance, most of the ko mutant lines showed reduced freezing tolerance. In general, the complete loss or strong reduction of flavonoid compounds resulted in impairment of freezing tolerance or in the accumulation of other known or unknown flavonoids with potential compensatory effects, while the accumulation of predominantly quercetin and anthocyanin compounds in the pap1-D mutant enhanced freezing tolerance (see Fig. 8 for a summary of these results). This is in good agreement with our previous finding that the content of all three quercetin derivatives and of the anthocyanins A5 and A9 were correlated with acclimated freezing tolerance in a collection of Arabidopsis accessions 20 . Interestingly, the reduction of either flavonols or anthocyanins did not result in a decline in freezing tolerance, indicating that the two groups of compounds may have at least partially redundant functions. As the presence of kaempferols, flavonols in general or anthocyanin derivatives preserved freezing tolerance, the combined contribution of a set of different compounds is more likely than the specific action of a particular molecule.
Differences in flavonoid composition between Col-0 and Ler have been reported previously for both plants exposed to non-acclimating and cold acclimating conditions 20,42 . The effects of mutations in the flavonoid biosynthesis pathway on freezing tolerance were generally more pronounced in the Ler than in the Col-0 background, suggesting that the contribution of flavonoids to freezing tolerance may be genotype-dependent. This is supported by the fact that tt3-1 in Ler, with only a moderate reduction in anthocyanin levels, was affected in its acclimated freezing tolerance, while tt18 and 75c1 in Col-0, with much stronger reductions in anthocyanin contents, were not affected. In addition, the Ler tt5-1 and both tt4-1 lines were more strongly affected in their freezing tolerance than the corresponding lines in Col-0. This accession effect is particularly striking for the tt4 mutants, which showed a significant reduction in anthocyanin content after cold acclimation in Col-0, but not in Ler, while freezing tolerance was only reduced in Ler (compare Fig. 8). The physiological basis of this difference between the accessions is currently unclear but it strongly underscores the fact that plant freezing tolerance is a quantitative trait that is based on the concerted action of many proteins and metabolites. This makes it difficult to estimate the contribution of single components to the total phenotype. For example, while raffinose content correlates strongly with acclimated freezing tolerance in Arabidopsis accessions, a ko mutant in the raffinose synthase gene that is completely devoid of raffinose shows no measurable impairment in freezing tolerance 43 .
Due to their antioxidant activity, a possible protective mechanism of flavonoids during freezing may be the scavenging of ROS 44 . In our experiments, however, this is rather unlikely, as freezing of leaves was performed in the dark to exclude ROS formation in the chloroplast. In addition, the cold acclimation treatment was performed at a significantly lower light intensity than growth under control conditions, also reducing ROS accumulation 29 . Flavonoids could also modulate stress response pathways through phytohormones as they negatively regulate auxin transport 45 and jasmonate levels 46 . However, we find such a pleiotropic effect unlikely, as plant growth, development and primary metabolism were unaffected in all mutants used in this study.
The direct protection of membranes and/or proteins from freezing damage are other possible mechanisms through which flavonoids could increase leaf freezing tolerance, in particular since flavonoids have been localized in the vacuole, nucleus and chloroplast, which may all be protected by flavonoids during freezing 47 . Membranes have long been recognized as major sites of freezing injury 48 and soluble cold-induced proteins that specifically stabilize membranes during freezing in Arabidopsis have been characterized 49,50 . A similar function of flavonoids seems possible as they can partition into both model and plant membranes in vitro and influence their stability 51 . In addition, enzymes can be inactivated when Arabidopsis leaves are frozen to temperatures that also induce electrolyte leakage and the inactivation of photosynthesis 50 . Flavonoids are able to stabilize proteins in vitro by preventing their aggregation [52][53][54] . Clearly, further research is needed to understand the functional mechanisms that link flavonoid content to plant freezing tolerance.

Methods
Plant material. The mutant lines used in this study included one activation tagging line and 13 T-DNA insertion lines of which nine were single, three double and one triple gene knock-outs in Columbia-0 (Col-0). Five mutants in Landsberg erecta (Ler) were generated by ethylmethansulfonate (EMS), fast neutron or x-ray mutagenesis and one line in Col-0 by carbon ion induction. In addition we used one activation tagging line. Details to all lines are given in Fig. 1 Determination of leaf freezing tolerance. Freezing damage was determined as LT 50 (temperature of 50% electrolyte leakage) from electrolyte leakage measurements after freezing of detached leaves to temperatures ranging from − 1 to − 21 °C at 4 °C h −1 as described in detail in previous publications 39,56 . Per line and condition four temperature curves were analyzed in each experiment, using leaves from three to six individual plants each. Due to the number of mutant lines, several experiments had to be performed, which each contained samples from the respective wild type. Data were collected for every mutant line from two to four independent sowings.
qRT-PCR analysis of gene expression. Quantitative RT-PCR was performed according to ref. 20, where also the sequences of all primers can be found. Ct values for the genes of interest were normalized by subtracting the mean Ct of four reference genes and averages of three biological replicates were determined (Suppl. Table 1).

LC-MS analysis of flavonoids.
Secondary metabolite profiling by LC-MS was carried out as described recently 20,57 . Peak identification and annotation was performed with Xcalibur (Thermo Finnigan) based on previously published reference data 4,58,59 . Relative peak areas representing mass spectral ion currents were normalized to fresh weight and peak area of the internal standard. The data represent the averages of three or six (Col-0 wild type and single mutants) biological replicates (Suppl. Table 2). The relative peak area obtained for kaempferol-3Glc-7Rha also represents a minor contribution from quercetin-3Rha-7Rha, as these compounds were not completely separated.
Primary metabolite profiling by GC-MS. Polar metabolites were extracted from100 mg frozen plant material and analyzed by gas chromatography/electron impact ionization-time of flight mass spectrometry (GC/EI-TOF-MS) as described in detail recently 60 . For every metabolite, relative mean intensities were calculated for each genotype as the average of five biological replicates (Suppl. Table 3).