Cold shock protein 1 chaperones mRNAs during translation in Arabidopsis thaliana


  • Piyada Juntawong,

    1. Department of Botany and Plant Sciences, Center for Plant Cell Biology, University of California, Riverside, California, USA
    2. Department of Genetics, Faculty of Science, Kasetsart University, Bangkok, Thailand
    Search for more papers by this author
  • Reed Sorenson,

    1. Department of Botany and Plant Sciences, Center for Plant Cell Biology, University of California, Riverside, California, USA
    Search for more papers by this author
  • Julia Bailey-Serres

    Corresponding author
    • Department of Botany and Plant Sciences, Center for Plant Cell Biology, University of California, Riverside, California, USA
    Search for more papers by this author

For correspondence (e-mail:


RNA binding proteins (RBPs) function post-transcriptionally to fine-tune gene regulation. Arabidopsis thaliana has four Gly-rich, zinc finger-containing RBPs called cold shock proteins 1–4 (CSP1–CSP4), that possess an evolutionary conserved cold shock domain. Here, we determined that CSP1 associates with polyribosomes (polysomes) via an RNA-mediated interaction. Both the abundance and polysomal co-fractionation of CSP1 was enhanced in the cold (4°C), but did not influence global levels of polysomes, which were minimally perturbed by above freezing cold temperatures. Using a polyclonal antiserum, CSP1 was co-immunopurified with several hundred transcripts from rosettes of plants cultivated at 23°C or transferred to 4°C for 12 h. CSP1-associated mRNAs were characterized by G+C-rich 5′ untranslated regions and gene ontologies related to cellular respiration, mRNA binding and translation. The majority of the CSP1-associated mRNAs were constitutively expressed and stable in the cold. CSP1 abundance was correlated with improved translation of ribosomal protein mRNAs during cold stress and improved maintenance of homeostasis and translation of mRNAs under water-deficit stress. In summary, CSP1 selectively chaperones mRNAs, providing translational enhancement during stress.


Abiotic stresses, including low temperatures and water deficit, pose a major limitation on plant growth and development. Some plants can acclimate to cold or freezing temperatures; however, this usually requires reprogramming of gene expression to accumulate freezing protective proteins and metabolites (Zhu et al., 2007). Cold stress-induced transcription networks, such as those regulated by the conserved CBF/DREB1 transcription factors, are well characterized (Liu et al., 1998; Fowler and Thomashow, 2002). Similarly, there is knowledge of transcription factor-based networks that reconfigure the transcriptome in response to water-deficit stress (Yamaguchi-Shinozaki and Shinozaki, 2006). However, gene regulation in response to environmental stimuli extends beyond the control of transcription. Regulation at the post-transcriptional levels during abiotic stresses includes alternative splicing, miRNA- and siRNA-mediated RNA turnover, and translational control (Kawaguchi et al., 2004; Kawaguchi and Bailey-Serres, 2005; Chinnusamy et al., 2007; Bailey-Serres et al., 2009; Filichkin et al., 2010; Zhu et al., 2011; Ambrosone et al., 2012; Park et al., 2012; Sunkar et al., 2012).

The mechanisms that orchestrate gene regulation at the post-transcriptional level involve hundreds of RNA-binding proteins (RBPs), which interact with mRNA via RNA binding domains (Bailey-Serres et al., 2009; Lorkovic, 2009). Among the plant RBPs is a small class of cold shock proteins (CSPs), containing a specialized DNA/RNA binding domain called a cold shock domain (CSD). This domain of approximately 70 amino acids includes two consensus RNA binding motifs, and is conserved across organisms (Graumann and Marahiel, 1998). Bacterial CSPs accumulate to high levels in response to cold shock, and function as RNA chaperones to destabilize nucleic acid secondary structure (Sommerville, 1999). They serve to reduce transcription anti-termination, enhancing ribosome biogenesis and mRNA translation during conditions of limited protein synthesis (Phadtare and Severinov, 2010). CSPs have been recognized in representative dicots, monocots and mosses (Karlson et al., 2002; Karlson and Imai, 2003; Chaikam and Karlson, 2008). The Arabidopsis thaliana Columbia (Col-0) genome has four genes that encode a protein with an N-terminal CSD domain, multiple CCHC zinc fingers and glycine-rich regions (AtCSP1–AtCSP4). CSP1 and CSP3 are similar in domain organization, as are CSP2 and CSP4. These genes are developmentally regulated, and their mutation and overexpression cause various phenotypes throughout the plant life cycle (Fusaro et al., 2007; Kim et al., 2007, 2009; Sasaki et al., 2007; Park et al., 2009; Yang and Karlson, 2011; Sasaki and Imai, 2012) . CSPs are also implicated in abiotic stress responses as AtCSP1AtCSP3 transcripts were enhanced by cold treatment (Karlson and Imai, 2003), and overexpression of AtCSP3 but not AtCSP1 or AtCSP2 improved freezing tolerance (Kim et al., 2007, 2009). Arabidopsis CSP1 binds RNA in vitro, leading to the proposal that it functions as an RNA chaperone (Kim et al., 2007; Park et al., 2009); however, it is not known if CSPs selectively bind individual mRNAs or facilitate specific post-transcriptional processes.

Dynamic changes in the transcriptome do not necessarily predict key alterations in the proteome of model organisms in response to external cues (Piques et al., 2009; Vogel et al., 2010; Lee et al., 2011). One factor that contributes to this discrepancy is differential translation of individual mRNAs (Bailey-Serres, 1999). Selective mRNA translation fine-tunes the expression of genes under standard growth conditions, and is often further altered by environmental stimuli. In Arabidopsis, cadmium exposure, dehydration, high salinity, high temperature, hypoxia, gibberellin, ozone-induced oxidative stress, sugar starvation and unanticipated darkness alter the selection of mRNAs for translation (Kawaguchi et al., 2004; Branco-Price et al., 2005, 2008; Nicolai et al., 2006; Mustroph et al., 2009b; Matsuura et al., 2010; Juntawong and Bailey-Serres, 2012; Puckette et al., 2012; Ribeiro et al., 2012). Moreover, conditions that reduce cellular ATP availability generally limit the translation of transcripts associated with cell growth, such as ribosomal protein (RP) mRNAs, but enable the translation of conditionally expressed mRNAs. The effect of cold treatment on translation mRNAs has not been evaluated at the genomic level in Arabidopsis.

Here we explored the function of Arabidopsis CSP1, demonstrating that it co-fractionates with polysomes and co-immunopurifies with a large number of cytosolic mRNAs. CSP1 levels increased in polysomes at low temperature, and its abundance was correlated with the translational activity of its target mRNAs under cold and water-deficit conditions.


CSP1 associates with polysomes via mRNA

To begin our study of CSP1, we transiently expressed carboxy-terminal yellow fluorescent protein-tagged CSP1 (CSP1-YFP) in leaf pavement cells of Nicotiana benthamiana. Confocal imaging localized CSP1-YFP to the nucleus and cytoplasm (Figure 1a–f). CSP1-YFP became localized in granular cytoplasmic foci when the leaf pieces were maintained under the coverslip (Figure 1d–f), as has been observed for several other mRNA-associated RBPs of plants (Weber et al., 2008). Next, we added a carboxy-terminal FLAG-His6 epitope tag to the full-length AtCSP1 cDNA and stably expressed the transgene in A. thaliana (Col-0). To determine whether or not CSP1-FLAG co-fractionated with ribosomes, we selected a 35S:AtCSP1-FH transgenic with moderate CSP1 overexpression and no effect on growth and development. 35S:AtCSP1-FH#11 seedlings were homogenized in a polysome-stabilizing buffer and centrifuged to obtain crude supernatant (S-16K), post-ribosomal supernatant (S-170K) and ribosome pellet (P-170K) fractions (Figure 1g). CSP1-FLAG was present in the S-170K and P170K cellular fractions, whereas cytosolic glyceraldehyde-3-P dehydrogenase (GAPDH) and the 40S ribosomal subunit protein RPS6 were limited to the S-170K and P-170K fractions, respectively (Figure 1h). The association of CSP1 with ribosomes was validated by centrifugation of the P-170K fraction through a 20–60% sucrose gradient to separate complexes by density (Figure 2a). However, the levels of ribosomal subunits, monosomes and polysomal complexes were similar in 35S:AtCSP1-FH#11 and Col-0 seedlings, indicating CSP1 overexpression did not alter global protein synthesis. CSP1-FLAG was limited to the fractions containing polysomes, whereas RPS6 was also in the less dense monosomes and 40S subunits.

Figure 1.

CSP1 is localized in the nucleus and cytosol, and co-purifies with polysomes. (a–c) Imaging of CSP1-YFP in epidermal cells of Nicotiana benthamiana leaves 48 h after transient transformation with Agrobacterium tumefaciens containing p35S:AtCSP1-YFP. Images were obtained with a Leica SP2 confocal microscope by fluorescence detection at 520–540 nm and 350 nm for YFP (a) and DAPI (b), respectively, at 63 ×  objective magnification. Transmitted light (c). (d–f) Representative confocal microscopic pictures of coverslip-induced CSP1-YFP granules. Images were taken at time zero (0 min) (d), within 2 min of sealing leaf pieces between a glass slide and coverslip, and 20 (e) and 40 min (f) later. Scale bars: 20 μm. (g) Cell fractionation scheme. (h) Two-week-old seedlings (35S:AtCSP1-FH#11) were extracted in a polysome stabilizing buffer and processed to obtain a crude supernatant (S-16), post-ribosomal supernatant (S-170) and polysome pellet (P-170) by centrifugation at the g force indicated. Proteins from each fraction were separated by 12.5% (w/v) SDS-PAGE and subjected to immunoblot analysis with an anti-FLAG epitope horseradish peroxidase-conjugated antibody, anti-RPS6 or anti-cytosolic GAPDH antisera. Molecular mass markers are indicated on the right.

Figure 2.

CSP1 associates with polysomes via RNA. (a) Comparison of polysome profiles from 10-day-old 35S:CSP1-HF#11 and Col-0 seedlings grown under control conditions. Following centrifugation of the P-170K fraction through a 20–60% (w/v) sucrose gradient, the UV absorbance at 254 nm was recorded and the proteins of 12 fractions were separated by 12.5% (w/v) SDS-PAGE and subjected to immunoblot analysis with anti-FLAG-HRP for CSP1-FLAG detection and anti-RPS6 antisera for RPS6 detection. (b, c) Polysome profile and immunoblot analysis of the P-170K fraction of 10-day-old 35S:AtCSP1-FH#11 and 35S:HF-RPL18 seedlings not treated and treated with RNaseA. Protein fractionation and detection was performed in the same manner as for (a). The FLAG-tagged protein in 35S:AtCSP1-HF#11 seedlings is CSP1-FLAG (34 kDa) and in 35S:HF-RPL18 seedlings is FLAG-RPL18 (22 kDa). The apparent molecular mass of RPS6 is 28 kDa.

To investigate whether CSP1-polysomal association was caused by an interaction with mRNA or a ribosomal subunit, the P-170K fraction was treated with RNaseA before centrifugation through a sucrose gradient. As expected, RNaseA reduced polysomes to monosomes and ribosomal subunits. The majority of CSP1-FLAG shifted to the low-density fractions, in contrast to RPS6, which remained in the ribosomal subunit and monosome fractions (Figure 2b). The experiment was performed in parallel on 35S:HF-RPL18 seedlings, a well-characterized line useful for monitoring RPL18 of the 60S subunit (Zanetti et al., 2005), as a control (Figure 2c). Altogether, these results demonstrate that CSP1 associates with polysomes by binding to mRNA rather than to a ribosomal subunit.

CSP1 levels increase at cold temperatures

A polyclonal antiserum was prepared against a peptide corresponding to a region of CSP1 that is not conserved in CSP2–CSP4 (Figure S1). In initial experiments, the antiserum was used to monitor steady-state levels of CSP1 in rosette leaves of plants maintained at 23°C or shifted to 4°C under similar light conditions (Figure S2). We found that CSP1 increased in abundance following 6 h of cold, and was sustained at higher levels than at the control temperature.

We sought T-DNA insertion alleles for AtCSP1 to facilitate the study of the molecular function of CSP1. Only one viable homozygote was successfully established, a knock-down line (atcsp1-1) with a 75% reduction in CSP1 mRNA relative to Col-0 (Figure S3). CSP1 levels were lower in atcsp1-1 rosettes at normal and low temperatures in comparison with the 35S:HF-RPL18 line used as a control (Figure 3a). As anticipated, the 35S:AtCSP1-FH#11 overexpression line accumulated CSP1-FLAG in addition to endogenous CSP1. We also confirmed that endogenous CSP1 and CSP1-FLAG increased in polysomes in response to cold (Figures 3b, S4a–c); however, we were unable to detect any significant difference in polysomal levels in the three genotypes at 23, 12, or 4°C in seedlings or rosettes for any duration tested (Figures 3b,c, S4d). Based on these results we conclude that neither knock-down nor overexpression of CSP1 or above freezing cold temperatures influenced global levels of polysomes in Arabidopsis rosettes.

Figure 3.

Cold elevates CSP1 abundance but does not alter global polysome levels. (a) Plants of three genotypes, 35S:HF-RPL18, atcsp1-1 and 35S:AtCSP1-FH#11, were grown for 25 days at 23°C and transferred to 4°C for 4 days before harvesting at ZT2. For CSP1 detection, SDS-solubilized proteins of rosettes were separated by 12.5% (w/v) SDS-PAGE and subjected to immunoblot analysis using an antiserum that detects endogenous CSP1 and FLAG-tagged CSP1. An asterisk identifies a 37-kDa non-specific band that was used as a loading control. The apparent molecular mass of CSP1 and CSP1-FLAG are 30 and 34 kDa, respectively. (b) Polysome profiles from 2-week-old plants grown at 23°C (black line) or exposed to 4°C for 9 h at ZT2 (red line). Proteins from each sucrose gradient fraction were precipitated, fractionated by 12.5% (w/v) SDS-PAGE and subjected to immunoblot analysis with anti-CSP1 (CSP1) or anti-RPS6 (S6) antisera. The apparent molecular mass of RPS6 is 28 kDa. This panel and data for atcsp1-1 and 35S:AtCSP1#11 are shown in Figure S4. (c) Quantitative sucrose density gradient analysis of ribosomal subunits and polysomes in rosettes of 25-day-old plants of the three genotypes, after maintenance at 23°C or transfer to 4°C at ZT2 and harvest after 12 h. Polysome peak areas were quantified after adjustment of the absorbance profiles to equivalent optical density units per gradient. Left panel: representative polysome profiles of 35S:HF-RPL18 seedlings. Right panel: mean percentage cellular RNA content in polysomes, calculated from three independent biological replicate experiments. Error bars represent SDs; samples are not significantly different based on a Student's t-test.

CSP1 co-immunopurifies with mRNAs

We explored whether the CSP1 antiserum could be used for immunopurification (IP) of CSP1 in association with mRNA from rosettes. We achieved CSP IP from formaldehyde cross-linked extracts using the same buffer developed for IP of ribosomes (Zanetti et al., 2005). Confirming the results in Figure 3a, the S-16K fraction in all three genotypes contained endogenous CSP1 (30 kDa), with the lowest level in atcsp1-1 rosettes (Figure 4a). Following the IP, CSP1 and CSP1-FLAG were present in the eluate (CSP1 IP), whereas RPS6 was limited to the S-16K and unbound fractions. The IP was efficient, as only a small quantity of CSP1-FLAG remained in the unbound fraction of the overexpression line. Moreover, the CSP1 IP was enriched in RNAs ranging in size from 200 to 2000 nt, and was depleted of rRNAs (Figure 4b). Neither IP with protein A agarose nor a control antibody yielded a significant quantity of RNA (Figures 4b, S5). The release of CSP1 from polysomes by RNaseA digestion, together with the lack of rRNAs and RPS6 in the CSP1 IP fraction, supports the conclusion that CSP1 binds directly to mRNAs.

Figure 4.

CSP1 co-immunopreciptiates mRNA. (a) Crude cell extracts (S-16K) prepared with polysome-stabilizing buffer from rosette leaves of three genotypes, 35S:HF-RPL18, 35S:AtCSP1-HF#11 and atcsp1-1 maintained at 23°C, were incubated with Protein A agarose beads that had been pre-incubated with CSP1 antisera. Following the removal of the unbound fraction (Unbound) and extensive washing of the agarose beads, the CSP1 immunoprecipitate (CSP IP) was eluted by the addition of 100 mm glycine, pH 2.5. Approximately 6% (v/v) of S-16K and unbound fractions, and 10% (v/v) of IP fractions, were heated in SDS loading buffer, fractionated by 12.5% SDS-PAGE and evaluated by immunoblot detection with anti-CSP1 or anti-RPS6 antisera. Asterisks indicate non-specific cross-reactive proteins. An open circle indicates IgGs present in the CSP IP. CSP1-FLAG and endogenous CSP1 are indicated. The mock IP lane was a 35S:HF-RPL18 sample incubated with Protein A agarose that was not coated with α-CSP1. (b) Comparative capillary electrophoretic analysis of total, sucrose gradient-purified polysomal (≥2 ribosomes), CSP IP and mock IP RNAs from rosette leaves of 35S:HF-RPL18 plants grown at 23°C. Nuclear rRNAs (28S, 18S, 5.8S and 5S) and plastid rRNAs (16S rRNA and 23S rRNA degradation products, marked with an asterisk) are indicated.

CSP1 binds a subpopulation of cellular mRNAs

Next we investigated whether CSP1 binds selectively to mRNAs by quantitatively monitoring total, polysomal and CSP1 IP mRNA populations using ATH1 microarrays (Figure 5a). The analysis compared the CSP1 IP from rosettes of plants cultivated at 23°C or transferred to 4°C for 12 h. The hybridization signal values were highly correlated between biological replicate samples (R2 values were 0.95–0.98 for total and polysomal mRNA, and 0.81–0.95 for the CSP1 IP; Table S1a). To quantitatively characterize CSP1-associated mRNAs, we first identified the probe pair sets (genes) with signal above the limit of detection across all samples, and then performed a series of pairwise signal log2 ratio (SLR) comparisons (n = 24). Of the 6998 genes with detectable signal in all samples, 5335 met our criteria for a differentially expressed gene (DEG) (|SLR| ≥ 1; false discovery rate, FDR < 0.05; Table S1b). The comparison of total and CSP1-enriched mRNA levels using the normalized signal values of the DEGs confirmed that CSP1 binds both high- and low-abundance mRNAs, with similar distribution in signal values under both non-stress and cold conditions in the Col-0 (WT) and CSP1 overexpression (OE) genotypes (Figure 5b).

Figure 5.

Identification of mRNAs associated with CSP1. (a) Experimental strategy used to isolate the three mRNA populations from 35S:HF-RPL18 (WT) and 35S:AtCSP1-FH#1 (OE) and two mRNA populations from atcsp1-1 (KD) under non-stress (23°C) or cold (4°C, 12 h) conditions. (b) Comparison of total RNA and CSP1-IP mRNA abundance of differentially expressed genes from WT and OE genotypes under two conditions. RMA-normalized signal values are plotted. (c) Fuzzy k-means clustering analysis was performed on 24 mean signal log2 ratio (SLR) comparisons of total, polysomal (Poly) and CSP1 IP mRNA levels, as indicated above each column. Genes included in the analysis were those with at least one P or M call across all hybridization samples, and significantly different in abundance in at least one comparison (|SLR| ≥ 1; FDR < 0.05, n = 5335 probe pair sets; Table S1b). The heat map shows median SLR values for each of 20 clusters of mRNAs with similar regulation. Color indicates increase (yellow), decrease (blue) or no change (black) in SLR value (Table S1c). Cluster ID, cluster number; no. genes, number of genes in cluster. Groups 1–4 are clusters of genes with similarity in CSP1 association or cold-responsive accumulation. (d) Heat map for mRNA encoding the 89 cytosolic RPs from clusters 15–19 (Table S1b).

We hypothesized that CSP1 may regulate translation of its targets because of its co-fractionation with polysomes (Figure 2a). Therefore, we applied Fuzzy k-means clustering to identify mRNAs that were co-regulated based on change in abundance, translation or CSP1 association under non-stress and cold conditions (k = 20; n = 5335 DEGs; Figure 5c; Table S1b,c). The DEGs fell into four groups, composed of between two and eight clusters. The heat map of mean cluster SLR values revealed that group-1 mRNAs were similarly induced in the transcriptome and translatome in response to low temperatures (n = 707 mRNAs; clusters 1–4; Total Cold/Total NS and Poly Cold/Poly NS comparisons of genotypes). Group-2 mRNAs were maintained at similar steady-state levels under both conditions, but were depleted in the CSP1 IP faction (n = 2537 mRNAs; clusters 5–12; CSP1 IP/Total or CSP1 IP/Poly comparisons of genotypes). The association of these mRNAs with CSP1 was greater in the overexpression line, particularly in the cold. Group-3 mRNAs were unstable in the cold, regardless of CSP1 association (n = 481; clusters 13 and 14). On the other hand, group 4 contained mRNAs that were preferentially associated with CSP1 under either conditions (n = 1609 mRNAs; clusters 15–20). Cluster15, -17 and -18 mRNAs were generally poorly translated, as indicated by the lower level of polysomal versus total mRNA under both conditions in all three genotypes. Differences between the translatome and transcriptome (Poly/Total mRNA comparisons) were limited when all values in a cluster were averaged, except for cluster 16, in which mRNAs were more poorly translated in the atcsp1-1 mutant in the cold. Comparable levels of DROUGHT INDUCIBLE 21 (DI21; cluster 17) and RUBISCO ACTIVASE (RCA; cluster 5) mRNAs were found in the total and CSP1 IP mRNA populations using qRT-PCR (Figure S6).

CSP1 targets mRNAs with specific function

To seek functional relationships among the CSP1-enriched or -depleted mRNAs, we performed a gene ontology (GO) analysis with the cluster data (Table S1d,e). As expected, mRNAs induced and translated in response to cold (group 1, clusters 1–4) were associated with abiotic stress and reactive oxygen species (cluster 1, response to abiotic stimulus, adjusted P = 6.16E–04; cluster 2, response to cold, 4.63E–17; response to oxidative stress, 2.03E–07). The mRNAs that were stable under cold stress but poorly associated with CSP1 (group 2, clusters 5–12) had diverse functionalities, including photosynthesis (cluster 6, 6.56E–11) and the cellular nitrogen compound biosynthetic process (cluster 10, 3.13E–19). The transcripts that were unstable and poorly translated in the cold (group 3) encoded proteins involved in various processes, including glucosinolate biosynthesis (cluster 13, 2.26E–08) and response to biotic stimuli (cluster 13, 1.02E–06). Finally, cold-stabilized and CSP1-bound mRNAs (group 4, clusters 15–20) were highly enriched in genes associated with energy-consuming processes, including ribonucleoprotein complex biogenesis (1.49E–17), structural constituents of ribosomes (4.53E–39) and components of the mitochondrial envelope (5.38E–08). The enrichment of cytosolic RP mRNAs in the CSP1 IP fraction led us to more carefully evaluate their association with polysomes. RP genes were represented on the ATH1 array by 219 probe pair sets, of which 135 were DEGs. Group 4 included 90 RPs, with 56 in cluster 16 alone (Table S1b). Focusing on group 4, we found that the atcsp1-1 mutant had the lowest level of association of RP mRNAs with polysomes of the three genotypes, especially in the cold (Figure 5d).

A survey of features of the mRNAs (i.e. untranslated region, UTR, length and di-nucleotide content) identified high G+C content in the 5′-UTR as prevalent in CSP1-associated mRNAs (clusters 16–20; P < 0.001; Figure 6a). Although the CSP1-associated mRNAs of cluster 15 did not have G+C-rich 5′-UTRs, they had a low predicted ΔG free-energy value, which is indicative of a propensity to form secondary structures (Figure 6b). Cluster-12 mRNAs also had G+C-rich 5′-UTRs, and like several group-4 clusters showed enhanced polysomal association in the overexpression line as compared with atcsp1-1 in the cold (Poly OE/Poly WT and Poly KD/Poly WT in the cold; Figure 6a). Thus, a 5′-UTR with features that promote double-stranded regions is a characteristic of CSP1 clients.

Figure 6.

CSP1-enriched mRNAs have distinct 5′-UTR features. (a) Analysis of 5′-UTR G + C content (%) in each of the 20 clusters represented in Figure 5c, the entire DEG data set and all 5′-UTRs from TAIR10 (ALL). (b) Analysis of 5′-UTR ΔG free energy at 23°C calculated from mRNAs with 5′-UTR length ≥ 60 nt. Values are means and standard errors. An asterisk indicates a significant difference as compared with the mean of all DEG mRNAs by the use of the Student's t-test (< 0.001).

CSP1 overexpression improves water-deficit tolerance

As an increase in CSP1 abundance was observed in polysomes at 4°C (Figures 3b, S3), we evaluated whether manipulation of CSP1 levels altered growth or survival at low temperatures. Four genotypes, including Col-0, two overexpression lines (35S:AtCSP1#11 and 35S:AtCSP1-FH#4) and the knock-down mutant (atcsp1-1) were evaluated at germination, seedling and rosette stages for altered growth, survival or molecular response to prolonged cold stress. Despite careful study, we found no statistically significant phenotypic differences between genotypes in these assays (data not shown). We went on to test whether these genotypes differ in response to water-deficit stress, as ectopic expression of CspB from Bacillus subtilis or CspA from Escherichia coli in maize (Zea mays L.) helps to maintain physiological homeostasis and yields under water-deficit stress (Castiglioni et al., 2008). Plants were cultivated under well-watered conditions until the 10-leaf stage (Figure 7a), and were deprived of irrigation for up to 11 days. All genotypes showed a decline in leaf relative water content (RWC) after 7 days (Figure 7b), with the least dramatic decrease in 35S:AtCSP1-FH#4 rosettes, the genotype that accumulated the highest level of CSP1 (Figures 7b,c, S7). By examination of selected mRNAs in the total and polysomal mRNA populations in Col-0 and 35S:AtCSP1-FH#4 rosettes, we found that ectopic expression of CSP1 promoted water deficit-induced polysome loading of DROUGHT INDUCIBLE 21 (DI21), two uncharacterized RBPs (At2 g43970 and At5 g46250) and RIBOSOMAL PROTEIN L39A (RPL39A) (Figure 7d). These mRNAs were CSP1-enriched under cold stress and better translated in the overexpression genotype in the cold, as compared with Col-0 and atcsp1-1 (cluster 17, Figure 5a; Table S1b). By contrast, there was no significant change in polysome association of RUBISCO ACTIVASE (RCA) mRNA, a transcript not found highly associated with CSP1 (cluster 5). These data suggest that overexpression of CSP1 enhances translation of its targets under water-deficit stress.

Figure 7.

CSP1 overexpression confers drought tolerance and improves the polysomal association of cellular mRNAs. (a) Photos of well-watered 10-leaf-stage plants grown under long-day conditions (scale bars: 1 cm). Genotypes: Col-0, 35S:AtCSP1-FH#4, 35S:AtCSP1-FH#11 and atcsp1-1. (b) Plot of percentage relative water content (RWC) after drought (n = 10 per genotype). (c) Mean RWC calculated from data points in (b). Error bars represent standard deviation. Letters represent significant differences calculated by one-way anova (< 0.05). (d) qRT-PCR analysis of selected transcripts using samples obtained from total and polysomal mRNAs of whole rosettes of well-watered control (NS) or 9-day drought-stressed (DS) plants. DI21, DROUGHT INDUCIBLE 21; At2 g43970 and At5 g46250 are uncharacterized RNA binding proteins; RPL39A, RIBOSOMAL PROTEIN L39A; RCA, RUBISCO ACTIVASE.


CSP1 is a polysome-associated mRNA chaperone

Here we demonstrate that the cold-induced RBP CSP1 co-purifies with polysomes and preferentially binds a subset of cellular transcripts. CSP1 was previously shown to bind double- and single-stranded RNA and destabilize secondary structures in vitro, leading to the suggestion that it acts as an mRNA chaperone under low temperatures (Kim et al., 2007; Park et al., 2009). By use of epitope-tagged CSP1 and an antibody that recognizes the endogenous protein, we demonstrated that much of the cellular CSP1 co-fractionates with polysomes (Figures 1h, 2, 3, S4). RNaseA digestion of single-stranded RNA released CSP1 from polysomes and ribosomal subunits (Figure 2b,c), indicating it interacts with mRNA.

The accumulation of CSP1 in seedlings and rosette leaves was enhanced after six or more hours of cold treatment (Figure S2), and this conditional increase correlated with higher levels of CSP1 in polysomes (Figures 3b, S4). Remarkably, cold treatment did not perturb global polysome levels (Figure 3c), growth or survival of seedlings or rosette-stage atcsp1-1 plants, as compared with Col-0 or two CSP1 overexpression lines (data not shown). This suggests that the level of CSP1 present in atcsp1-1 was sufficient to maintain homeostasis in the cold, and the elevated level of CSP1 in 35S:AtCSP1-FH lines was insufficient to discernibly alter the tolerance of above freezing cold temperatures. Consistently, we observed limited differences in the adjustment of the transcriptome and translatome following cold treatment of wild-type, 35S:AtCSP1-FH and atcsp1-1 rosettes (Figure 5c). We conclude that Arabidopsis is capable of maintaining polysomes and presumably protein synthesis in Arabidopsis seedlings at 4°C.

CSP1 IP and subsequent microarray analyses revealed that it binds over 6000 mRNAs without bias for transcript abundance. mRNAs with roles in photosynthesis, glycolysis and cell wall biogenesis were poorly associated with CSP1, relative to their abundance in the total or polysomal mRNA fraction (Table S1c). On the other hand, CSP1-bound mRNAs encoded proteins involved in energy-demanding processes, including ribosome biogenesis. These mRNAs possessed 5′-UTRs with a high G+C content or low predicted ΔG (Figures 5c,d and 6). A 5′-UTR with a high G+C content is unfavorable for translation during drought stress, hypoxia, ozone-induced oxidative stress and unanticipated darkness (Branco-Price et al., 2005; Kawaguchi and Bailey-Serres, 2005; Juntawong and Bailey-Serres, 2012; Puckette et al., 2012), and is typical of mRNAs encoding RPs (Kawaguchi and Bailey-Serres, 2005). Because Arabidopsis CSPs and bacterial CSD proteins can denature the secondary structure of double-stranded RNA in vitro (Kim et al., 2007; Park et al., 2009; Sasaki and Imai, 2012), it is reasonable to suggest that CSP1 facilitates the removal of secondary structure in the 5′-UTRs of its targets, assisting 43S pre-initiation complex scanning. In support of this, CSP1-bound mRNAs encoding RPs were better associated with polysomes at low temperatures in Col-0 and the overexpression line, than in atcsp1-1 seedlings (Figure 5d).

Evidence of functional redundancy among CSPs and other RNA chaperones

Studies indicate that Arabidopsis CSP1–CSP4, other Gly-rich proteins and RZ-1a-c are candidates for mRNA chaperone activity in the nucleus and/or cytosol (Kim et al., 2005, 2007, 2008, 2009, 2010; Kwak et al., 2005; Nakaminami et al., 2006, 2009; Sasaki et al., 2007). Several members of these families bind and relax double-stranded RNA, an activity associated with CSD proteins in bacteria (Sommerville, 1999). Indeed, plant CSPs and GRPs may function in a manner similar to bacterial CSPs, as AtCSP1 and AtGRP7 (At2g21660) complemented the cold-sensitive phenotype of an E. coli strain lacking four native CSPs (Kim et al., 2007). There is also evidence of functional redundancy among these RBPs in Arabidopsis, as the overexpression of AtCSP1 rescued atgrp7-1 from freezing damage (Kim et al., 2007). We surmise that our inability to detect quantifiable differences in growth or survival in the cold of the CSP1-deficient atcsp1-1 mutant may reflect sufficient residual CSP1 or compensation by a functionally redundant protein, such as GRP7 or the closely related CSP3.

CSPs, mRNA translation and abiotic stress

Arabidopsis can cold-acclimate, grow and reach maturity at low temperatures (4–6°C; Miquel et al., 1993; Schneider et al., 1995). CSPs related to AtCSP1–AtCSP4 are found in diverse plant species, although their elevation during cold stress may differ in cold-hardy versus cold-sensitive plants. For example, the CSPs of wheat gradually increase during cold acclimation in the crown tissue of seedlings (Karlson et al., 2002; Sasaki et al., 2007), whereas those of cold-sensitive Oryza sativa (rice) are not elevated under low temperatures (Chaikam and Karlson, 2008). Interestingly, both short- and long-term cold treatment had limited effect on global polysome levels in the cold-tolerant seedlings of wheat (Triticum aestivum; Fehling and Weidner, 1986; Perras and Sarhan, 1990) and rye (Secale cereale; Laroche and Hopkins, 1987), but significantly reduced polysome levels in cold-sensitive rice seedlings (Park et al., 2012). This decline in polysome levels in rice was concomitant with reduced levels of mRNAs encoding proteins associated with translation, ribosome biogenesis, secondary metabolite biosynthesis, transport and metabolism. In rice, levels of CSPs may be insufficient to maintain stability and translation of its clients at low temperatures, whereas in Arabidopsis and other cold-tolerant species it could prime cells to sustain translation of its targets in the cold. As many CSP1-associated mRNAs were associated with ribosome biogenesis, CSP1 may help maintain growth at above freezing cold temperatures.

Castiglioni et al. (2008) reported that the expression of bacterial CSPs in several plant species was beneficial during abiotic stress. In Arabidopsis they improved growth at 8°C, in rice they improved cold, heat and drought tolerance, and in maize they improved fitness under water-limited conditions in the field. The benefit in maize was abrogated by a point mutation of a key RNA binding residue of CspB, supporting the hypothesis that the functional activity involves RNA interaction. The finding that Col-0 and AtCSP1 overexpression genotypes displayed better association of RP mRNAs with polysomes during cold and water deficit (Figure 5d) supports the hypothesis that CSP1 acts to improve the translation of its clients, many of which are essential for growth. Remarkably, in bacteria, CSPs also enhance the translation of mRNAs required for ribosome biogenesis (Phadtare and Severinov, 2010).

In conclusion, the Arabidopsis mRNA chaperone CSP1 selectively binds to mRNAs in polysome complexes, and functions to maintain translation of its clients under conditions of abiotic stress.

Experimental procedures

Genetic material

Arabidopsis thaliana genotypes included the Col-0 accession, 35S:HF-RPL18 (12-2-4) (Zanetti et al., 2005), atcsp1-1 (SALK_048960; T-DNA insertion into the promoter of At4 g36020) and 35S:AtCSP1-FH transgenic (ectopic overexpression) lines, produced as described below.

Plant growth conditions

Plants were grown in soil (Sunshine Mix LC1; McConkey, containing 150 g of Osmocote 14-14-14 fertilizer (Scotts #90036; Scotts, and 75 g Marathon pesticide (Crop Production Services, per 0.12 ft3 of soil in a controlled environmental growth room (16 h at approximately 100 μE s−1 m−2 light/8 h dark, at 23°C).

For growth in sterile culture, seeds were surface sterilized, stratified at 4°C for 48 h and plated on solid Murashige and Skoog (MS) medium [0.43% (w/v) MS salts (Caisson Laboratories, containing 1% (w/v) sucrose and 1% (w/v) agar, pH 5.7] in 100 15-mm Petri dishes. Growth was in a vertical orientation in a chamber (model CU36L5C8; Percival Scientific Inc., under 16 h of approximately 50 μE s−1 m−2 light and 8 h of dark at 23°C.

For cold treatment, plants grown in 5.08-cm2 pots or seedlings grown on plates were placed at 4°C or 12°C (16 h approximately 50 μE s−1 m−2 light/8 h dark) for 0–48 h or at 23°C (16 h approximately 100 μE s−1 m−2 light/8 h dark) in a growth chamber. Cold treatment began 2 h following the initiation of the light period (ZT2). Rosette leaves were collected at specific time points, flash frozen and stored at –80°C. For water deficit treatment, plants were grown in 5.03-cm2 pots supplemented daily with water until they reached the 10-leaf stage. Plants were randomly positioned in a tray in replicates of 10 per genotype, and the relative water content (RWC) was measured on specific days (Kawaguchi and Bailey-Serres, 2005).

Isolation of the atcsp1-1 mutant homozygote

SALK_048960 (Arabidopsis Biological Resource Center, was used to generate a homozygous line, designated atcsp1-1, with a T-DNA insertion in the promoter region. Segregants were genotyped by polymerase chain reaction (PCR) amplification of genomic DNA with primers AtCSP1geneF, AtCSP1geneR and a T-DNA left border primer LBa1 (Table S2). To evaluate transcripts, total RNA was extracted from 100 mg of pulverized tissue of 10-day-old whole seedlings with TRIzol (Invitrogen, Approximately 400 ng of total RNA was used for cDNA synthesis with the oligo d(T) primer (Promega, and Superscript II reverse transcriptase (Invitrogen). Non-quantitative reverse Transcriptase-Polymerase Chain Reaction (RT-PCR) was performed using gene-specific primers for both AtCSP1 and ACTIN7, with annealing at 55°C and 28 PCR cycles.

Generation of 35S:AtCSP1-FH transgenic lines

The CSP1 (At4 g36020) full-length open reading frame was amplified by RT-PCR from RNA extracted from 10-day-old Col-0 seedlings with gene-specific primers AtCSP1codingF and AtCSP1codingR (Table S2). cDNA was synthesized, cloned into the pENTR/D-TOPO vector (Invitrogen), transformed into One Shot TOP10 Chemically Competent E. coli (Invitrogen) and selected with 50 μg ml−1 kanamycin. Verified clones were recombined into the p35S:GATA-FH vector (Mustroph et al., 2010), transformed into E. coli DH5α and transformants selected with 50 μg ml−1 chloramphenicol. The vector provides a CaMV 35S promoter, TMVΩ 5′-UTR leader, Gateway recombination site upstream of a FLAG-His (FH) tag [(G)7DYKDDDDK(G)3(H)6], and nopaline synthase terminator sequence in a Ti binary plasmid with a neomycin phosphotransferase II gene. After sequence confirmation, plasmid was electroporated into Agrobacterium tumefaciens GV3101 and colonies selected with 50 μg ml−1 chloramphenicol. Col-0 transformation was performed (Clough and Bent, 1998), T1 seeds were collected, seedlings resistant to 50 μg ml−1 kanamycin propagated, and homozygous single insertion events established.

Cellular fractionation of ribosome complexes by differential ultracentrifugation

Extraction of polysome by centrifugation through sucrose gradients was based on the procedure of Mustroph et al. (2009a). Briefly, 3 mL of pulverized tissue was thawed in 7 ml of polysome extraction buffer [PEB; 200 mm Tris-HCl, pH 9.0, 200 mm KCl, 36 mm MgCl2, 25 mm EGTA, 5 mm DTT, 1 mm PMSF, 50 μg ml−1 cycloheximide, 50 μg ml−1 chloramphenicol, 1% (v/v) Triton X-100, 1% (v/v) Brij-35, 1% (v/v) Tween-40, 1% (v/v) NP-40, 1% (v/v) PTE]. After removal of cell debris by centrifugation at 16 000 g, the supernatant (S-16K) was layered on top of an 8 ml 1.75 m sucrose cushion (400 mm Tris-HCl, pH 9.0, 200 mm KCl, 30 mm MgCl2, 1.75 M sucrose, 5 mm DTT, 50 μg ml−1 chloramphenicol, 50 μg ml−1 cycloheximide) and centrifuged at 135 000 g for 18 h at 4°C to obtain a ribosome pellet (P-170K) and ribosome-depleted supernatant (S-170K). The P-170K fraction was resuspended in 250 μl PB buffer (200 mm Tris–HCl, pH 9.0, 200 mm KCl, 36 mm MgCl2, 25 mm EGTA, 5 mm DTT, 50 μg ml−1 cycloheximide, 50 μg ml−1 chloramphenicol, 20 U ml−1 RNAseout; Invitrogen) and approximately 2000 units (OD260) was loaded on top of a 20–60% (w/v) sucrose gradient. For RNaseA treatment, 2000 units (OD260) of the P-170K fraction was incubated with 2.5 μg RNaseA for 10 min at room temperature (23°C) before loading onto the sucrose gradient. Gradients were centrifuged and analyzed as described previously by Mustroph et al. (2009a) and Williams et al. (2003).

Proteins from each fraction were precipitated with ethanol and resuspended in SDS loading buffer before electrophoresis. Total RNA and polysomal RNA (≥2 ribosomes per mRNA; typically gradient fractions 7–13) were extracted with TRIzol. Total RNA samples were treated with RNase-Free DNAse Set (Qiagen, Total and polysomal RNA were cleaned by the use of the RNAeasy Plant Mini Kit (Qiagen).

Quantitative real-time PCR

Approximately 1 μg of total RNA was used for cDNA synthesis, as described above. cDNAs were diluted fivefold. Quantitative real-time PCR (qRT-PCR) was performed using 2 μl of cDNA, 0.5 μm of each primer and 10 μl of SYBR Green supermix (Bio-Rad, by the use of CFX96 real-time PCR detection system (Bio-Rad). Primers used for qRT-PCR were listed in Table S2. Data were analyzed by the comparative ΔCt method (Livak and Schmittgen, 2001).

CSP1 antibody production and preparation of antibody coated beads

A peptide (C151AGNGDQRGATKGGN163) specific to A. thaliana CSP1 was synthesized and injected into a rabbit to generate a polyclonal antiserum against CSP1, which was subsequently affinity purified using the CSP1 peptide (anti-CSP1; Genscript, EZview Red™ Protein A Affinity Gel (200 μl; Sigma-Aldrich, or Protein G Dynabeads (Life Technologies, were coated with anti-CSP1 antiserum (ratio: 0.1 μg antibody to 1 μl suspended beads) in PB for 16 h at 4°C with gentle shaking. The beads were washed three times with 5 ml of PB and stored in 200 μl PB at 4°C until use.

Protein separation and immunodetection

Proteins were separated by 12.5% (w/v) SDS-PAGE and subjected to immunodetection with anti-FLAG horseradish-peroxidase (HRP) conjugated monoclonal antibody (1:1000; Sigma-Aldrich), affinity-purified anti-CSP1 (1:4000; 1 mg ml−1 stock), a polyclonal antiserum against maize ribosomal protein S6 (RPS6; 1:10 000; Williams et al., 2003) or Arabidopsis cytosolic GAPDH (1:10 000; kindly provided by Ming-Che Shih, Academia Sinica, Taiwan). HRP-conjugated goat anti-rabbit IgG (1 : 10 000; Bio-Rad) was used as a secondary antibody. Visualization was performed with the ECL Plus chemiluminescence system (Amersham, now GE Healthcare,

CSP1 immunopurification for RNA analyses

For CSP IP, rosette leaves were washed with ice-cold autoclaved milli-Q water and cross-linked by bathing in 1% (v/v) formaldehyde (Molecular Biology Grade; Fisher Scientific, for 15 min by vacuum infiltration (Niranjanakumari et al., 2002). Cross-linking was stopped by the addition of 2 m glycine (pH 7.0) to a concentration of 125 mm and vacuum infiltrated for 5 min. Leaves were washed twice with ice-cold double-distilled H2O, ground in liquid nitrogen, thawed in 2 ml PEB per millilitre tissue, mixed and centrifuged at 16 000 g at 4°C for 15 min to obtain a supernatant and then passed through Miracloth. The clarified extract was added to the anti-CSP1 antibody-coated Protein A beads (100 μl of antibody-coated beads per millilitre tissue) and incubated at 4°C for 2 h with gentle shaking in 15 ml Falcon tubes. Extracts were preincubated with non-antibody coated beads before incubation with anti-CSP1 Protein A to reduce non-specific binding (50 μl beads millilitre tissue). To control for non-specific RNA binding to IgGs, control IPs were performed with Protein A (or G) Dynabeads or with anti-HA mAb coupled to Protein G Dynabeads. The supernatant (unbound fraction) was removed and the beads were washed four times in 6 ml PEB with detergents modified to 1.125% (v/v) Triton X-100, 0.125% (v/v) Brij-35, 0.125% (v/v) Tween-40, 0.125% (v/v) NP-40, 40 U ml−1 SUPERase.In (Life Technologies). To elute CSP1 and associated molecules, 150 μl of elution buffer [100 mm Tris–HCL, pH 8.0, 25 mm EDTA, 1% (w/v) SDS] was added to the beads, incubated for 10 min at room temperature with shaking and then centrifuged. Elutions were repeated with 150 μl of elution buffer, followed by incubation at 65°C for 10 min. The eluate was combined and 4 μg Proteinase K (New England Biolabs, was added and incubated at 65°C for 1 h to reverse the cross-linking. RNAs were extracted with TRIzol according to the manufacturer's protocol, and further purified on an RNAeasy Plant Mini Kit column (20 μl elution).

Microarray hybridization and data analyses

RNA used for microarray hybridization was examined with a Bioanalyzer 2100 and RNA 6000 nano chip or RNA 6000 pico chip (Agilent Technologies, The total, polysomal, and CSP1 IP samples were hybridized to ATH1 (Affymetrix, GeneChips at the Genomic Core Facility, Institute for Integrative Genome Biology, University of California, Riverside, CA, USA. Amplified RNA synthesis and biotin-labeling was conducted with 150 total or polysomal RNA, and 3–6 ng CSP1 IP RNA using GeneChip® 3′ IVT expression (Affymetrix). Hybridizations were performed at 45°C for 16 h, with 10 μg of biotin-labeled aRNA for total and polysomal and 7–10 μg of biotin-labeled aRNA for CSP1 IP RNA. Microarray hybridization was performed on three completely independent biological replicate samples for 35S:HF-RPL18 (WT) and two each for 35S:AtCSP1-FH (OE) and atcsp1-1 (KD). The CSP1 IP was not evaluated for atcsp1-1 because of low yields in the IP. A mock IP sample hybridization was also performed and analyzed. Expression data obtained in .cel files was extracted by use of the Bioconductor package of the r statistical software. Low-level normalization followed using Robust Multichip Average (RMA), and included probe-specific and multichip background corrections. Present, marginal or absent (P/M/A) calls, describing signals for probe pair sets that were above or below the background default threshold, were obtained with Affymetrix mas 5.0.

The normalized data were used to calculate signal log2 ratio (SLR) values for total, polysomal and immunoprecipitated mRNA samples. False discovery rates (FDRs) for significant differences between mRNA in the samples compared were generated using P-value distributions (Smyth, 2004). These values were used to identify differentially expressed genes (DEGs). Normalized SLR data of probe pair sets (genes) detected with at least one P or M call across all comparisons, at least one comparison with an FDR < 0.05, and with at least one comparison with |SLR| ≥ 1 were used for the subsequent cluster and gene category analyses. The microarray data are accessible from the National Center for Biological Information Gene Expression Omnibus database (GSE38030).

Identification of co-regulated genes by clustering and ontology

Co-regulated genes were identified by the use of Fuzzy k-means clustering with Euclidean correlation for the distance measure, a membership exponent of 1.1, maximal number of iterations of 5000 and 20 clusters (Mustroph et al., 2009b). The mean SLR value for each cluster was determined for summary visualization. Clusters membership was analyzed for gene ontology (GO) category enrichment analysis by the use of the GOHyperGall function according to Horan et al. (2008). GO annotations were obtained from (TAIR, 08/03/2010 release).

Analysis of G + C content of 5′ untranslated regions

The 5′-UTRs were extracted from (TAIR 10) based on availability, and G + C nucleotide content was calculated using Microsoft excel and 5′-UTR ΔG free energy was calculated using mfold 2.3 (23°C, 1 m NaCl; Zuker, 2003).


We thank Jennifer Oki for technical assistance. This work was supported by the US National Science Foundation (IOS-0750811, MCB-1021969 to J.B.-S. and DGE 0504249 IGERT fellowship to R.S.) and a Royal Thai Ministry of Science and Technology scholarship to P.J.