Insights into the adaptive response of Arabidopsis thaliana to prolonged thermal stress by ribosomal profiling and RNA-Seq
- 2k Downloads
Environmental stress puts organisms at risk and requires specific stress-tailored responses to maximize survival. Long-term exposure to stress necessitates a global reprogramming of the cellular activities at different levels of gene expression.
Here, we use ribosome profiling and RNA sequencing to globally profile the adaptive response of Arabidopsis thaliana to prolonged heat stress. To adapt to long heat exposure, the expression of many genes is modulated in a coordinated manner at a transcriptional and translational level. However, a significant group of genes opposes this trend and shows mainly translational regulation. Different secondary structure elements are likely candidates to play a role in regulating translation of those genes.
Our data also uncover on how the subunit stoichiometry of multimeric protein complexes in plastids is maintained upon heat exposure.
KeywordsTranslation Ribosome profiling Transcription RNA-Seq Secondary structure G-quadruplexes Heat stress response
Environmental stress or suboptimal growth conditions reduce cell viability and require an immediate but specific response in order to maximize the survival of the whole organism. Particularly, plants are constantly exposed to changing environmental conditions and are under threat of severe adverse conditions. On the subcellular level, heat exposure changes membrane fluidity [1, 2] and protein stability [3, 4] which consequently alter photosynthesis  and central metabolic activities . Plants are highly sensitive to temperature stress and respond over different time scales [7, 8, 9, 10]. One of the most potent steps to regulate heat stress response has been suggested to occur at the level of transcription . Long heat exposure triggers epigenetic changes, some of which are conserved between yeast and plants indicating that these stress response mechanisms are evolutionarily conserved among organisms . Ultimately, proteins mediate stress response and their levels have to be rapidly adjusted to ensure cell adaptability and survival particularly under prolonged stress.
Gene expression is subject to extensive regulation, including transcription, mRNA degradation, translation and protein degradation, each of which operates on a different temporal regime [12, 13, 14]. Translation is a downstream process of transcription and provides the opportunity to rapidly adjust protein concentration in response to external stimuli . Although transcriptional reprogramming upon heat exposure has been addressed in plants, little is known for the role of translation. Does translation complement transcription in shaping the heat stress response?
Advances in massively parallel sequencing platforms and approaches to capture ribosomal position with nucleotide resolution, i.e. ribosome profiling , precisely capture gene expression at the level of translation. Combined with RNA-Seq to measure changes in mRNA population , the transcriptional and translational responses can be deconvoluted. Ribosome profiling has been successfully applied in mammalian systems, for example to study the effect of heat , oxidative  and proteotoxic stress  on translation in mammalian systems. The depth of those approaches revealed unprecedented aspects in the stress response programs which were not detected with a single sequencing method. The applicability of ribosome profiling technology in plants has been recently demonstrated by two studies assessing the global expression reprogramming in A. thaliana during dark to light transition  and the response to hypoxia . We used the combined approach of ribosome profiling (Ribo-Seq) and RNA-Seq to assess the response of A. thaliana to prolonged heat stress. Our study reveals a complex picture of adaptive response in plants and provides a rich resource for future hypothesis testing.
A subset of genes shapes the plant adaptation to thermal stress
Overall, for the majority of genes that are translationally active under heat (i.e., for which RPFs were detected), we found a positive linear log-log correlation with changes in their mRNA reads (Additional file 1d) suggesting that the adaptive response is shaped in a coordinated manner between transcription and translation.
However, a sizeable fraction of genes differed in their expression (i.e. exhibited disproportionate ratios of the mRNA to RPF reads) (Additional file 2a, b). Those gene groups may provide candidates whose expression is controlled either transcriptionally or translationally. Hence, we used differential expression analysis (DESeq) to compare the mRNA and RPF counts of each gene expressed in control plants grown under ambient temperature to that in the plants exposed for 3 h to 37 °C (Fig. 1b). The confidence intervals for the fold-change analysis were set based on the reproducibility of the biological replicates for the control plants (Additional file 1a, b). DESeq analysis considers as expression level the sum of all RNA or RPF reads over a transcript but is insensitive to the distribution of reads along a transcript. If translation of a gene is enhanced, we expect increased RPF reads along the entire open-reading frame (ORF) length. We reasoned that if a gene is uniformly translated with no detectable heat-induced stalling over certain position(s) within the CDS, the counts of the RPF reads between the two halves of a gene should be equal. Notably, RPFs were nearly symmetrically distributed between the two halves of a coding sequence (CDS) of genes expressed under heat exposure and resembled the uniform distribution between the two halves of the mRNA (Additional file 2c, d), suggesting that higher total RPF reads truly report on enhanced expression of those genes under heat stress.
The DESeq analysis revealed co-directional changes in the mRNA and RPF counts for 579 out of 14,246 genes (525 upregulated and 54 downregulated; green designated, Fig. 1b). A sizeable fraction of genes showed only changes in the mRNA (723 genes, blue designated, Fig. 1b) or in RPF (1150 genes, red designated, Fig. 1b). For each of the groups with altered RNA expression or translatability (i.e., altered RPF reads), we performed enrichment analysis using DAVID (Additional file 3). The most prominent groups among those upregulated at transcriptional and/or translational level (i.e. significantly higher mRNA and/or RPF read counts) were genes involved in the heat stress response and protein folding (e.g. chaperones and heat-shock proteins). Interestingly, although the plants were exposed to heat for 3 h, which should elicit the adaptive response to heat stress, the mRNA of key heat shock proteins was very high (Additional file 4a, b). In contrast, groups comprising genes related to the chromatin structure, cytoskeleton organization, cell wall synthesis, cell cycle, and anabolic process were mostly down-regulated at transcriptional and/or translational levels (Additional file 3). Together, prolonged exposure to heat stress resulted in large changes in gene expression and reprogramming of both transcriptional and translational activities of the plants that are likely to shape their survival under sub-optimal growth conditions.
Genes with lower secondary structure propensity in 5′ start vicinity are translated under thermal stress
To address the question as to whether RNA-binding proteins contribute to translational regulation through binding the 5′ and 3′ UTRs, we performed a motif search in the UTR sequences of the genes which were translationally upregulated (i.e. changed RPF reads, but unchanged mRNA reads) upon the heat stress. In total, 55 genes (out of 895 heat upregulated) exhibited an increased number of RPF in their 5′UTR and in 23 of them we detected a conserved A/G-rich motif (Fig. 2c). Similarly, in 82 genes the RPFs in their 3′ UTRs increased upon stress and in 23 of them we also detected the A/G-rich motif (Fig. 2d).
G2-quadruplexes in the UTRs may also control gene expression under stress
The effect of the G2 quadruplexes in the 3′UTRs of genes upregulated under heat exposure is unclear; they did not contribute to the stability of the mRNA under heat stress, i.e. the mRNA reads for the two halves of a gene remained unchanged under stress (Fig. 3e).
Detection of an alternative transcript and ORF upon stress exposure
Within the region 1 kb upstream of this alternative ORF and of the heat shock responsive genes we performed a sequence motif search to extract putative sequences that may serve as putative transcription factors binding sites. Interestingly, we identified motifs which share conserved features with motifs found in the promoter regions of the known heat-stress responsive genes (Fig. 4c) which were transcriptionally induced upon heat stress in our data set. The presence of motif 2 bears significant resemblance to the heat shock promoter element AGAAnnTTCT recognized by heat shock factors in Arabidopsis  supports the idea of this alternative transcript being heat-responsive. This alternative transcript with a start at 2195 nt encodes an 88 amino-acid long peptide/protein with high overrepresentation of positively charged amino acids; proteins with overrepresented charged amino acids may play a protective role under stress, e.g. scavenging reactive oxygen species. Although it remains to be determined whether the expression of this alternative transcript generates a viable protein or peptide, our results underline the potential of Ribo-Seq in determining alternative ORF or proteins resulting from alternative, independent translation initiation which differ from the start of the main transcript.
Stoichiometry of protein complexes in chloroplasts under heat exposure
We next used this measure to evaluate the production of stable multiprotein complexes with known stoichiometry (Fig. 5b and Additional file 6a). Remarkably, for the ATP synthetase complex, which has the most complex stoichiometry, the protein production of each subunit quantitatively reflected its stoichiometry within the complex (Fig. 5b). The ribosome density of each ORF was different despite comprising the same polycistronic mRNA (atpA/E/I/H and atpB/F are the two operons, Fig. 5c). The mRNA levels of these two operons were similar as confirmed by RNA-Seq analysis, further validating that differences in the stoichiometry might be controlled at the level of translation. For some complexes, which are encoded mostly on different polycistronic mRNAs, the ratio of protein production of some subunits differed from their stoichiometry (Additional file 6) and suggests an additional regulation mechanism at the level of degradation.
The expression of protein-coding genes in chloroplasts changed under heat exposure and for the majority of the ORFs changes in mRNA levels were co-directional with changes in transcription (i.e. mRNA reads) and translation (i.e. RPF reads) (Fig. 5d). Strikingly, the production of the subunits within one protein complex changed disproportionately, even for those upregulated under heat stress (Fig. 5e and Additional file 6b) which could suggests that the different susceptibility to degradation of various subunits may additionally change the abundance of the subunits under stress.
Here we analyze the adaptive response of A. thaliana to prolonged heat exposure (3 h at 37 °C) at both transcriptional and translational level using RNA-Seq and deep sequencing of RPFs of nuclear- and chloroplast-encoded genes. The plant habitat suggests that a typical heat exposure is long, for example for several hours in a summer midday. The expression changes of the majority of nuclear-encoded genes are modulated in a coordinated manner at the transcriptional and translational levels. While at early time points of heat exposure, i.e. between 15 and 45 min, translation is globally downregulated and stress response is counteracted mainly by transcriptional programs [32, 33, 34], our results show that prolonged exposure to stress (3 h) activates translational programs which shape the adaptive response. At prolonged exposure to heat stress the majority of the genes are transcribed and translated in a coordinated fashion, but a sizeable set of genes opposes this trend and instead shows only changes at the level of translation. Among those translationally regulated transcripts we detected several shared features which are likely candidates to regulate their expression. The A/G-rich motifs in the 5′ or 3′ UTRs of the translationally upregulated genes resemble sequences identified as RNA-protein binding motifs . The presence of relatively conserved A/G-rich motifs to which most likely the same RNA-binding protein binds would allow coregulation of the expression of those transcripts . Another common feature among the genes translationally upregulated under heat stress is their lower propensity to form secondary structure, likely to facilitate ribosome binding and enhances translation [35, 36]. Furthermore, some of the transcripts preferentially translated under heat contain a putative G2 quadruplex in their 5′UTR. The increased RPF reads over the quadruplex structures correlate with the enhanced expression in the downstream CDS, suggesting a role in activating translation of the downstream ORF with a yet unclear mechanism. This is in line with an earlier observation that transcripts with highly structured 5′UTRs are enriched upon heat exposure . Although the type of the secondary structure in the 5′UTR is not specified , the authors suggest a mechanism to sense heat in a similar fashion to the riboswitches in bacteria .
The duration of heat stress has different effects on transcriptional and translational programs. The sequence of response seems to follow a conserved pattern in microorganisms and mammalian systems. An initial reaction upon acute heat stress comprises global translational downregulation [18, 20] and a quick transcriptional activation of heat-shock proteins . This first transcriptional burst is followed by an adaptive response which includes reprogramming of many cellular activities with a prominent activation of the heat-stress response, relating to protein folding and degradation, at both translational and transcriptional levels . Previous studies in A. thaliana addressing short term (10–45 min) heat exposure provide evidence that translation is greatly inhibited [33, 34, 40]. By contrast, our data show that under prolonged heat exposure (3 h) translation is fully active (the polysomal fraction is only marginally changed, Additional file 1e), suggesting a common pattern of stress response between A. thaliana, mammalian cells and microorganisms. Despite the global translational repression under short term heat stress, some transcripts are selectively translated, which includes genes involved in transcriptional regulation, chromatin structure rearrangements, mRNA degradation, salicylic acid-mediated signaling and protein phosphorylation are activated under short term heat exposure [9, 33, 34, 40]. In contrast, extended heat stress (3 h) activates genes involved in heat-stress response and protein folding and deactivates genes related to the chromatin structure, cytoskeleton organization, cell wall synthesis, cell cycle, and anabolic processes. Strikingly, the most prominent gene groups translated in both short term  and extended heat stress (Additional file 3) as also observed for mammalian systems  suggesting a common features in maintaining heat stress among organisms.
Translation in chloroplasts shares many features with bacteria, including Shine-Dalgarno-driven initiation and polycistronic mRNAs. The prevalence of genes encoded in polycistronic transcripts in prokaryotes has been suggested as a mechanism to couple translation and control the stoichiometry of the single subunits in multisubunit complexes or to control the level of proteins with related functions in metabolic pathways . Although the premise might be true for some examples, it does not explain how a higher number of subunit copies can be achieved downstream in the operon (the example with ATP synthetase, Fig. 5c). Furthermore, in bacteria translation rates among genes within the same operon are only weakly correlated  and the architecture of several metabolic pathways is robust against variations in the single proteins, suggesting that a precise translational coupling may not be crucial for their performance [42, 43]. In plastids of A. thaliana, we detected a clearly decoupled translation of single genes within an operon; each single ORF within one polycistronic message is initiated in an independent manner with distinct yield (the example with ATP synthetase, Fig. 5c). Similar observation has been made in plastids of maize  and in bacteria . Rather than coupling of their translation, the differential synthesis rates within one message [45, 46] and/or degradation rates of the subunits  determine the precision to achieve the balanced stoichiometry of subunits. Under thermal stress, we observed variations in the translation rates of single subunits (Fig. 5e and Additional file 6) which lead to alterations in the production of the single subunits that deviated from the expected stoichiometry. Translation rates may change disproportionately because of temperature-dependent variations in the diffusion properties of the different translation components . Precise control of the stoichiometry of protein complexes at elevated temperatures could also be established by differential degradation of the subunits at elevated growth temperatures [48, 49, 50, 51].
In summary, our data unravel new aspects in the adaptive response of A. thaliana to heat stress at the level of translation. The adaptation to heat exposure is fine-tuned by a sizeable set of genes whose translation is most likely regulated by different secondary structure elements. Furthermore, the Ribo-Seq and RNA-Seq data provide a vast resource of the Arabidopsis transcriptome and translatome at permissive temperature (i.e. control growth conditions) and under heat stress that can inform future experiments focused on understanding transcriptional and translational regulation of nuclear and plastid-encoded genes.
Plant growth and heat treatment
Wild-type Col-0 plants (The European Arabidopsis stock center NASC, ID N1092) were grown on soil in a greenhouse in a long-day condition (16 h/8 h, lamps Philips Master HPI-T Plus, 400 Watt Philips SON-T Agro, 400 Watt, light intensity ~140 μmol.m−2.s−1, humidity 60 %). The leaves from 15 pre-bolting, 3-week-old plants (stage 3.50 according to ) were exposed for 3 h at 37 °C with constant humidity of 60 % (heat stress; ), or at 22 °C and served as control. The choice of the duration of the thermal stress was also driven by the availability of data from a previous study addressing transcriptional changes using microarray technology . Plants were pooled, leaves harvested and immediately frozen in liquid nitrogen and stored for further treatment. The total RNA was extracted using TRIzol reagent (Invitrogen), cDNA was synthesized with random hexamers and RevertAid™ H Minus First-Strand cDNA Synthesis Kit (Fermentas) and analyzed with qRT-PCR using Power SYBR Green Master Mix (Life Technologies) and with the following primer pairs: for HSP70 (At3g12580) 5′-CCGTCTTCGATGCTAAGCGTCT-3′ and 5′-AACCACAATCATAGGCTTCTCACC-3′; HSP101 (At1g74310) 5′-ATGACCCGGTGTATGGTGCTAG-3′ and 5′-CGCCTGCATCTATGTAAACAGTG-3′; HSFA2 (At2g26150) 5′-TCGTCAGCTCAATACTTATGGATTC-3′ and 5′-CACATGACATCCCAGATCCTTGC-3′; UBQ10: 5′-AAAGAGATAACAGGAACGGAAACATAGT-3′ and 5′-GGCCTTGTATAATCCCTGATGAATAAG-3′; At1g76880 main transcript 5′-ACGATGATGCAACTGGGTGGTG-3′ and 5′-AGCAGTTGTGACGGTTGTAGCC-3′; At1g76880 alternative transcript 5′-TCGGAGCAGAACTTTGATGATGA-3′ and 5′-GCTCGAACTCACCTCCTTCCTC-3′.
Polysomes were isolated according to  with some modifications. Briefly, 10 g of leaf material was thawed in polysome extraction buffer (0.2 M Tris pH 7.4, 0.2 M KCl, 0.025 M EGTA, 0.035 M MgCl2, 1 % Brij-35, 1 % Triton X-100, 1 % Igepal CA 630, 1 % Tween 20, 1 % DOC, 1 % PTE, 5 mM DTT, 1 mM AEBSF, 100 μg/mL cyclohexamide, 100 μg/mL chloramphenicol), homogenized using a glass homogenizer, filtered through four layers of sterile cheese cloth and two layers of sterile Miracloth (Calbiochem) and incubated on ice for 10 min. The supernatant after centrifugation (4 °C, 16,000xg for 15 min) was additionally filtered through Miracloth and transferred onto a sucrose cushion solution (0.4 M Tris pH 7.4, 0.2 M KCl, 0.005 M EGTA, 0.035 M MgCl2, 1.75 M sucrose, 5 mM DTT, 100 μg/mL cyclohexamide, 100 μg/mL chloramphenicol) and centrifuged at 4 °C, 170,000xg for 3 h. The ribosome-containing pellet was gently resuspended in ice-cold resuspension buffer (0.2 M Tris pH 7.4, 0.2 M KCl, 0.025 M EGTA, 0.035 M MgCl2, 5 mM DTT, 100 μg/mL cyclohexamide, 100 μg/mL chloramphenicol) and applied onto 15 – 60 % sucrose gradient (0.04 M Tris pH 7.4, 0.02 M KCl, 0.01 M MgCl2, 100 μg/mL cyclohexemide, 100 μg/mL chloramphenicol) and centrifuged at 4 °C, 237,000xg for 1.5 h.
Preparation of RPF and total mRNA libraries
Purified polysomes were digested with RNAse I (1.5U/1OD/1 μl) at 22 °C for 10 min, loaded directly onto 15–60 % sucrose gradient and centrifuged at 4 °C, 237,000xg for 1.5 h. The amount of loaded sample was normalized according to the input material, hence the variations in the samples mirror the different RNA content of each sample. The monosome fraction was concentrated with an Amicon-Ultra4 Centrifugal Unit (MWL 100 kDa) and RPFs were released by adding the release buffer (20 mM HEPES-KOH pH 7.4, 100 mM KCl, 1 mM EDTA, 2 mM DTT, 2 μl/ml Ribolock) and incubated for 10 min on ice and centrifuged at 4 °C, 1900xg for 30 min. RNA was extracted using the hot acid phenol method and depleted of rRNA using RiboMinus Plant Kit (Ambion). Samples were normalized on the input material.
Total RNA was extracted using TRIzol reagent and spiked with ERCC RNA Spike-In Mix (Ambion). rRNA was depleted using RiboMinus Plant Kit (Ambion), and randomly fragmented by alkaline lysis in alkaline fragmentation solution (2 mM EDTA, 12 mM Na2CO3, 87 mM NaHCO3) at 95 °C for 40 min. The randomly fragmented RNA was recovered by precipitation in the presence of glycogen.
The sequencing libraries were prepared according to . Briefly, both randomly fragmented total RNA and RPF were loaded onto a 15 % TBE-polyacrylamide gel (containing 8 M urea). RNA fragments with a size of 25–35 nucleotides, which size corresponds to a nucleotide sequence covered by the ribosomes , were cut out of the gel and isolated by centrifugation at 17,000xg for 5 min to crush the gel, eluted by incubating with 3 M Na acetate buffer (pH 5.5) containing glycogen and RiboLock (Thermo Fischer Scientific) for 4 h at 4 °C and purified by precipitation with isopropanol. To those fragments 5′ and 3′ adaptors were ligated and subjected to deep sequencing on the Illumina Hiseq2000 platform.
Mapping of the sequences and reads distribution analysis
The sequencing data was mapped against the TAIR 10 annotated A. thaliana genome (downloaded from ENSEMBL, version 21) using Bowtie 1.0.0. Perfectly mapped reads (i.e., without any mismatch) to an rRNA reference were discarded after the first mapping round. The remaining reads were the mapped to the genome, the parameters were adjusted according to the properties of very short reads (-v 2 -m 1 --strata --best -y) and only uniquely mapped reads were kept for further analysis. The number of raw reads unambiguously aligned to ORFs in both RNA-Seq and ribosome profiling data sets from two biological replicates were used to generate gene read counts, which were normalized as reads per million of the total mapped reads (rpm) or reads per kilobase per million of the total mapped reads (rpkm) .
Differential expression and enrichment analysis
Mapped read counts were applied to protein coding transcripts using the longest annotated transcript for each AGI identifier. The detection limit (2 rpkm) in each experiment was determined from the linear range of detection of the Spike-In Mix and the selected reliably expressed transcripts were subjected to differential expression analysis by means of DESeq (version 1.16.0; ) using a false discovery rate of 0.1. Enrichment analyses were performed using DAVID [55, 56].
Secondary structure analysis, motif search and analysis for putative G-quadruplex structures
Secondary structure of the 5′ UTRs and coding mRNAs was computed with RNAfold program (2.1.7; default parameters) from the ViennaRNA Package 2.0  using a sliding window of 39 nt and assigning the minimal free energy to the middle nucleotide . Average profiles for different gene groups were generated by taking the mean of their per-base folding energy contributions.
To identify conserved motifs in the sequences, we used MEME version 4.10.0_4 (available online at http://meme-suite.org/tools/meme). The parameters were set as motif minimal width of 6, motif maximal width of 20, maximal numbers of motifs 10 and ‘zero or one per sequence’.
The position of potential G-quadruplexes were predicted using custom R-scripts which analyzed the sequences for presence of G2-, G3- and G4-quadruplexes sequences matching the pattern G2N1-7G2N1-7G2N1-7G2, G3N1-7G3N1-7G3N1-7G3 and G4N1-7G4N1-7G4N1-7G4, respectively.
Analysis of the protein production in chloroplasts
Protein production was determined using only the normalized RPF reads in the CDS normalized to the mRNA reads as described by Li et al. . The protein abundance data we obtained were compared with the protein production in chloroplasts determined with mass spectrometry by Baginsky and colleagues . The data from the two measurements  were averaged and only proteins present in both were used.
We thank A. Bartholomäus for the help with some of the statistical analysis.
This work was supported by the grants of the Deutsche Forschungsgemeinschaft (to ZI).
Availability of data and materials
The data sets supporting the findings in this article are included within the article and its additional files. The sequencing data are deposited in the Gene Express Omnibus (GEO: GSE69802) database.
RL and ZI conceived and designed the experiments. RL performed experiments. RL, PF and ZI analysed and interpreted the data. RL and ZI wrote the manuscript. All authors read and approved the manuscript.
The authors declare that they have no competing interests.
Consent for publication
Ethics approval and consent to participate
- 7.Kilian J, Whitehead D, Horak J, Wanke D, Weinl S, Batistic O, D’Angelo C, Bornberg-Bauer E, Kudla J, Harter K. The AtGenExpress global stress expression data set: protocols, evaluation and model data analysis of UV-B light, drought and cold stress responses. Plant J. 2007;50:347–63.CrossRefPubMedGoogle Scholar
- 29.Jackson RJ, Kaminski A, Pöyry TAA. Coupled termination-reinitiation events in mRNA translation. In: Matthews MB, Sonenberg N, Hurshey JWB, editors. Translational control in biology and medicine. New York: Cold Spring Harbor Laboratory Press; 2007. p. 197–223.Google Scholar
- 31.Bischof S, Baerenfaller K, Wildhaber T, Troesch R, Vidi PA, Roschitzki B, Hirsch-Hoffmann M, Hennig L, Kessler F, Gruissem W, et al. Plastid proteome assembly without Toc159: photosynthetic protein import and accumulation of N-acetylated plastid precursor proteins. Plant Cell. 2011;23:3911–28.CrossRefPubMedPubMedCentralGoogle Scholar
- 32.Merret R, Descombin J, Juan YT, Favory JJ, Carpentier MC, Chaparro C, Charng YY, Deragon JM, Bousquet-Antonelli C. XRN4 and LARP1 are required for a heat-triggered mRNA decay pathway involved in plant acclimation and survival during thermal stress. Cell Rep. 2013;5:1279–93.CrossRefPubMedGoogle Scholar
- 53.Mustroph A, Zanetti ME, Jang CJ, Holtan HE, Repetti PP, Galbraith DW, Girke T, Bailey-Serres J. Profiling translatomes of discrete cell populations resolves altered cellular priorities during hypoxia in Arabidopsis. Proc Natl Acad Sci U S A. 2009;106:18843–8.CrossRefPubMedPubMedCentralGoogle Scholar
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.