- © 2014 American Society of Plant Biologists. All rights reserved.
Abstract
Plants are sessile organisms that gauge stressful conditions to ensure survival and reproductive success. While plants in nature often encounter chronic or recurring stressful conditions, the strategies to cope with those are poorly understood. Here, we demonstrate the involvement of ARGONAUTE1 and the microRNA pathway in the adaptation to recurring heat stress (HS memory) at the physiological and molecular level. We show that miR156 isoforms are highly induced after HS and are functionally important for HS memory. miR156 promotes sustained expression of HS-responsive genes and is critical only after HS, demonstrating that the effects of modulating miR156 on HS memory do not reflect preexisting developmental alterations. miR156 targets SPL transcription factor genes that are master regulators of developmental transitions. SPL genes are posttranscriptionally downregulated by miR156 after HS, and this is critical for HS memory. Altogether, the miR156-SPL module mediates the response to recurring HS in Arabidopsis thaliana and thus may serve to integrate stress responses with development.
INTRODUCTION
To cope with changes in their physical environment, plants have evolved physiological responses allowing them to tolerate stressful conditions such as extreme temperatures, drought, and high salinity (Jenks and Hasegawa, 2005). These abiotic stresses are a major limiting factor for the geographical distribution and productivity of plants; hence, stress tolerance is a central goal in crop breeding. So far, research has focused on the immediate responses to abiotic stress, leading to important insights into the molecular and genetic basis of stress tolerance. However, how plants respond to enduring and recurring stress conditions has received little attention, although this arguably constitutes the more “natural” stress scenario. Plants can be primed by stress, such that their responses to repeated stress exposure or their further development are modified (Pecinka et al., 2010; Ito et al., 2011; Ding et al., 2012; Sani et al., 2013). Priming precedes a memory phase, during which information on a past stress exposure is stored. A related phenomenon in animals is hormesis, where a low-level stress treatment enhances resistance against the same and other stresses and increases growth, fecundity, and longevity (Gems and Partridge, 2008).
Here, we studied priming by high-temperature stress in Arabidopsis thaliana as a model case for environmental stress memory. A plant acquires thermotolerance through a moderate (priming) heat stress (HS) and is then able to survive a severe HS that is lethal to an unadapted plant (Mittler et al., 2012). Acquired thermotolerance is maintained during several days, even if the plant is not exposed to elevated temperatures after the priming HS (Yeh et al., 2012). We call this maintenance of acquired thermotolerance short HS memory; it is an active process that can be separated genetically from the acquisition of thermotolerance (Charng et al., 2006, 2007; Meiri and Breiman, 2009). Across all kingdoms, HS induces the expression of various types of heat shock proteins (HSPs) that protect cellular proteins from denaturation (Larkindale et al., 2005; Richter et al., 2010). Generally, HSP expression is activated by HEAT SHOCK TRANSCRIPTION FACTORs (HSFs), peaks early after HS, and declines within a day to base levels (Scharf et al., 2012). HSFA2 is required specifically for HS memory (Charng et al., 2007). Mutants in two additional heat-induced genes (HEAT-STRESS-ASSOCIATED32 [HSA32] and ROTAMASE FKBP1) are specifically impaired in HS memory (Charng et al., 2006; Meiri and Breiman, 2009). Thus, while a few players have been identified, no mechanistic model of HS memory has emerged.
MicroRNAs (miRNAs) have been reported to modulate plant stress responses. miRNAs are ∼21-nucleotide-long single-stranded RNA molecules that guide effector proteins, most commonly ARGONAUTE1 (AGO1), to complementary mRNAs, causing either mRNA cleavage or translational inhibition (Voinnet, 2009; Axtell, 2013). miRNAs are generated from primary transcripts that are transcribed by RNA polymerase II and fold into a hairpin structure that is cleaved by DICER-LIKE1 (DCL1) to release the active miRNA (Rogers and Chen, 2013). miRNAs play important roles during plant growth, developmental transitions, and determination of cell identity (Jones-Rhoades et al., 2006; Voinnet, 2009). miRNAs are also involved in the modulation of plant responses to biotic and abiotic stresses (Sunkar and Zhu, 2004; Fujii et al., 2005; Navarro et al., 2008; Khraiwesh et al., 2012; Kruszka et al., 2012; Sunkar et al., 2012). It has been suggested that plant miRNAs that regulate growth and development may also mediate modified growth and development after stress exposure (Sunkar et al., 2012).
Here, we report a function for the miRNA pathway and more specifically for miR156 in the adaptation to recurring HS in Arabidopsis. We show that miR156 is induced by HS and that miR156 levels after HS are critical for HS memory and affect expression of HS memory–related genes through SPL genes.
RESULTS
The miRNA Pathway Is Specifically Required for HS Memory
To test whether miRNAs are involved in the maintenance of acquired thermotolerance (i.e., HS memory), we first tested whether ago1-25 and ago1-27 seedlings had a defect in this process. The ago1-25 and ago1-27 alleles are hypomorphic (i.e., they retain some residual activity) and display only mild developmental alterations in comparison to a full ago1 knockout (Morel et al., 2002). To assay HS memory (Charng et al., 2007), we induced thermotolerance in 4-d-old seedlings by a priming HS treatment, consisting of 1 h at 37°C, 1.5 h recovery at 23°C, and 0.75 h at 44°C (acclimation [ACC]; Figure 1). Two or three days after the priming HS, a tester HS (44°C for 80 to 115 min) was applied that is lethal to a plant lacking the priming HS (Figure 1C and Methods). A temperature of 44°C is commonly used as severe HS in Arabidopsis stress tolerance studies (Yeh et al., 2012). Arabidopsis does encounter such temperatures in nature where full insolation causes leaf temperatures to rise much higher than air temperatures (Salisbury and Spomer, 1964). In consequence, Arabidopsis is well adapted to withstand exposure at 44°C after HS priming. Following the described tester HS after priming, both ago1 mutants showed reduced seedling fresh weight and a higher proportion of highly damaged seedlings compared with the Columbia-0 (Col-0) wild type and only slightly better survival than the memory-deficient hsfa2 mutant, suggesting a defect in HS memory (Figures 1C, 1D, and 2A to 2C).
Representation of Different Types of Thermotolerance, Their Assay Conditions, and Schematic Representation of the Thermotolerance Profile of a Mutant with a Specific Defect in the Maintenance of Acquired Thermotolerance (i.e., HS Memory).
(A) Basal thermotolerance. ‘, minutes; t, time.
(B) Acquisition of thermotolerance.
(C) Maintenance of acquired thermotolerance (HS memory) is assayed by a tester HS of 80 to 120 min 2 or 3 d after a priming HS (ACC), which induced thermotolerance.
(D) Schematic representation of the temporal profile of acquired thermotolerance in the wild type and a mutant with a specific defect in its maintenance (such as hsfa2). There is no defect in the acquisition of thermotolerance (b), but a clear difference (see double-pointed arrow) in its maintenance (c).
The miRNA Pathway Is Required for the Maintenance of Acquired Thermotolerance.
(A) ago1-25 and ago1-27 mutants were assayed for the maintenance of acquired thermotolerance by applying a tester HS two (+2, 110 min) or three (+3, 90 min) days after a priming HS (ACC); hsfa2 was included as a control. Photographs were taken 14 d after ACC. All plants of one treatment were grown on the same plate. One representative replicate of at least three independent biological replicates is shown. For quantification, see (B) and (C). ACC, priming HS.
(B) Quantification of data shown in (A); average fresh weight from 59 ± 14 seedlings was determined 14 d after ACC. Due to very low weight, d+3 samples were not analyzed. Instead, ACC + HS d+2 90 min samples were included, which were omitted in (A) due to space constraints.
(C) Quantification of data shown in (A); the percentage of seedlings in different phenotypic classes was determined for different genotypes and heat treatments. Gray arrows indicate the corresponding plates from (A). Panels are (top to bottom): ACC, ACC + HS d+2 110’, ACC + HS d+3 90’. Two photographs with representative seedlings of each class are shown to illustrate the damage classes. Asterisk, significantly different distribution compared with Col-0 (Fisher’s exact test, *P < 0.05; **P < 0.005).
(D) dcl1-9 mutants show reduced maintenance of acquired thermotolerance (Fisher’s exact test, P < 0.001). Individual seedlings were phenotypically categorized 14 d after a ACC + HS d+2 90’ treatment as green, weak (bleached out cotyledons but green hypocotyl and meristem), and dead and genotyped for dcl1-9 (n = 78).
To test whether the observed phenotype was specific, we assayed acquisition of thermotolerance and basal thermotolerance (Figures 1A and 1B). The acquisition of thermotolerance is reflected in the amount of severe HS that a plant survives directly after a priming HS (Figure 1B). Basal thermotolerance reflects the dose of severe HS that an unprimed plant will survive (Figure 1A). Mutants with specific defects in the maintenance of acquired thermotolerance are expected to display normal acquisition of thermotolerance and normal basal thermotolerance (Figure 1D). ago1 mutants acquired thermotolerance to a level comparable to the wild type (Supplemental Figure 1A), and basal thermotolerance was similar to the wild type (Supplemental Figure 1B), indicating that ago1 is specifically required for the maintenance of acquired thermotolerance.
We next analyzed the hypomorphic dcl1-9 mutant that is specifically required for the biogenesis of miRNAs (Jacobsen et al., 1999). As homozygous dcl1-9 mutants are sterile, we used progeny of a heterozygous plant. Two weeks after the tester HS, individual seedlings were classified according to their phenotype and genotyped. We found more individuals with severe damage among dcl1-9 homozygous seedlings than their wild-type siblings, and this altered distribution of phenotypes was statistically highly significant (Figure 2D; P < 0.001, Fisher’s exact test). Thus, dcl1-9 is defective in HS memory, suggesting that miRNA biogenesis is required for HS memory.
Global Changes in the Transcriptional Response to HS in ago1 Mutants
We next investigated the transcriptome of ago1 during the maintenance of acquired thermotolerance. To this end, we performed ATH1 GeneChip analysis of Col-0 wild type and ago1-25 RNA extracted from seedlings at 4 and 52 h after a priming HS and from nontreated control (ntc) seedlings harvested at the same time as the 4-h HS seedlings. In the wild type, 401 genes were changed at 4 h at least 4-fold, and 96 genes at 52 h after HS (Figure 3A; Supplemental Data Set 1). Of these 96 genes, 87 (91%) were induced by HS and 41 (43%) were induced also at 4 h. We classified these 41 genes as HS memory related. Of the genes differentially expressed at 4 h, only 42% were induced. Selected memory-related genes (APX2, HSA32, HSP17.6A, and HSP22.0; log2 fold change [ATH1, 52 h] 5.1, 2.4, 6.6, and 5.3, respectively) were further analyzed using quantitative RT-PCR (qRT-PCR) and a more detailed time course (Figure 3B). This revealed that these genes reached their maximal induction 1 d (24 h) after HS and remained activated for 2 d (48 h) and to a lesser extent 3 d (72 h) after HS. This distinguishes the class of HS memory–related genes from classically studied HSP genes, such as HSP70, which were found in the “4 h only” class and were not analyzed further during this study. Of note, Gene Ontology term analysis for the three classes (“4 h only,” “52 h only,” and “4 h and 52 h”) revealed strong enrichment of HS-related terms (GO0009266, 0009408, and 0010286) for both the “4 h only” and the “4 h and 52 h” classes (Supplemental Figure 1C).
The Effect of ago1-25 on the Transcriptome Is Most Pronounced during the Late Stages of the HS Response.
(A) Venn diagram representation of ATH1 GeneChip transcriptome profiling of Col-0 samples after a priming HS (ACC). Samples were harvested 4 or 52 h after the end of the ACC; the number of genes that are up- or downregulated at either 4 or 52 h or at both time points relative to a 4 h ntc are shown.
(B) Experimental validation of the indicated HS memory–related genes in Col-0 (black bars) and ago1-25 (white bars) over a detailed time course after ACC or mock treatment using qRT-PCR. Expression values were normalized to TUB6 and to Col-0 0 h ntc [(GENE OF INTEREST/TUB6)x/(GENE OF INTEREST/TUB6)Col0 h ntc]. Error bars are se of the mean of three replicates. One representative of at least three independent experiments is shown.
(C) Percentage of genes whose induction or repression differed at least 2-fold in ago1-25 compared with Col-0 in the classes described in (A) or among all remaining genes on the microarray (all others), respectively.
We next identified HS-responsive genes whose expression pattern after HS depended on AGO1 activity. To this end, we searched for genes whose expression fold changes after HS were at least 2-fold changed in ago1 compared with Col [(Col HS/Col ntc)/(ago1 HS/ago1 ntc) >2 or <2]. Strikingly, 21% of upregulated and 28% of downregulated genes in the “4 h only” class fulfilled this criterion. This fraction increased to 51 and 100%, respectively, in the “4 h and 52 h” class and to 93 and 60%, respectively, in the “52 h only” class (Figure 3C). Compared with this, only 3% of all other genes were accordingly affected in ago1 mutants. Thus, ago1 seedlings show a defect in the dynamics of the response to HS at the transcript level, and this effect is most pronounced during later stages. These results were confirmed for selected genes over a more detailed time course (Figure 3B). Thus, ago1 is required for the later stages of the HS response.
Identification of HS-Responsive miRNAs
The requirement for AGO1 and DCL1 in HS memory suggests that miRNAs function in this process. We next identified miRNAs that were specifically induced by HS at early or late stages using a qRT-PCR platform (Pant et al., 2009) for 204 known primary miRNA (pri-miRNA) transcripts (Supplemental Data Set 2). Comparing samples from 4 and 52 h after a priming HS with ntc samples, we identified a number of heat-responsive pri-miRNAs (Figure 4A). Using log2 fold change >2 as a cutoff, 22 pri-miRNA transcripts were upregulated and 17 downregulated at 4 h, while at 52 h, seven pri-miRNAs were upregulated and four downregulated. Four upregulated pri-miRNAs and one downregulated pri-miRNA were common to both time points. Focusing on upregulated pri-miRNAs, we selected the three pri-miRNAs most strongly upregulated at either time point and the four pri-miRNAs upregulated at both time points and analyzed their induction profile in more detail. pri-miR156h and pri-miR831 transcript levels peaked at 4 h after HS, while pri-miR391 peaked only after 28 h (Figure 4B; Supplemental Figure 2A). We then tested whether induction of the pri-miRNA corresponded to elevated levels of the mature miRNA (Figure 4C). miR156a-f peaked at 4 h but was elevated up to 52 h after HS (relative to the respective ntc; the decline of miR156a-f expression between 4 and 52 h ntc samples is developmentally regulated; Wu and Poethig, 2006; Wang et al., 2009); miR156h was increased between 0 and 76 h; miR391 was increased at 52 and 76 h; and miR831 was increased between 18 and 76 h. For other miRNAs, confirmation of HS induction at either the pri-miRNA level or the mature miRNA level could not be obtained (possibly due to their low abundance close to the detection limit) and they were not analyzed further. Thus, for miR156h, miR391, and miR831, mature miRNA levels generally followed pri-miRNA levels, yet appeared more sustained, possibly because of the higher stability of miRNAs compared with their precursor transcripts in the cell.
Primary MIR Profiling and Transcript Levels of HS-Responsive miRNAs during Maintenance of Acquired Thermotolerance.
(A) Venn diagram representation of primary MIR transcript levels after a priming HS (ACC); samples were taken 4 or 52 h after the end of ACC. Up- or downregulated pri-miRNAs at either 4 or 52 h or at both time points are displayed. In blue are the miRNAs further analyzed in (B) and (C).
(B) qRT-PCR of the indicated primary MIR transcripts.
(C) qRT-PCR of the indicated mature miRNAs.
Expression values were normalized to TUB6 and the respective 4 h ntc [(GENE OF INTEREST/TUB6)x/(GENE OF INTEREST/TUB6)4 h ntc]. Black and white bars indicate HS-treated (ACC) and control (ntc) samples, respectively. Error bars are se of the mean of three replicates. One representative of at least three independent experiments is shown.
[See online article for color version of this figure.]
The HS induction was most robust for miR156. The miR156 clade is extremely well conserved among flowering plants and is also found in mosses (Arazi et al., 2005; Axtell et al., 2007). It plays important roles in the endogenous timing of leaf initiation and developmental transitions by targeting SQUAMOSA-PROMOTER BINDING-LIKE (SPL) transcription factors (Cardon et al., 1999; Rhoades et al., 2002; Wang et al., 2008, 2009; Wu et al., 2009; Huijser and Schmid, 2011). The miR156 clade in Arabidopsis is encoded by eight genes producing three different miRNA isoforms: miR156a-f are identical, miR156g has a single nucleotide substitution at the first position, and miR156h has two nucleotide substitutions at positions 11 and 14 relative to miR156a-f (Jones-Rhoades and Bartel, 2004). miRNAs of the ath-miR156h isoform have also been found in Arabidopsis lyrata and Sellaginella moellendorfii (Fahlgren et al., 2010; Ma et al., 2010; Kozomara and Griffiths-Jones, 2011). As the miR156a-f could be derived from any of six MIR156(a-f) genes, we also tested other MIR156 transcripts and found a HS-induced upregulation for MIR156c and d (Figure 4B). For these reasons, we focused further analyses on miR156.
SPL Transcripts Are Transiently Downregulated by HS through miR156
We next searched our transcriptome data for putative miR156 targets with HS-dependent expression. The transcripts of four putative target genes were downregulated at least 2-fold 4 h after HS (SPL2, SPL9, SPL11, At5g38610 [predicted pectin methylesterase inhibitor]), and one gene was downregulated at 52 h after HS (At5g40510, predicted protein kinase). Of these, SPL2, SPL9, and SPL11 were likely targeted by miR156a-f and miR156h, while At5g38610 and At5g40510 were predicted to be specific for miR156h (Supplemental Table 1 and Supplemental Figure 3D). Cleavage of At5g38610 mRNA by miR156 is supported by degradome data (Addo-Quaye et al., 2008). HS-regulated expression of SPL2, SPL9, SPL11, At5g38610, and At5g40510 was confirmed in a more detailed time course (Figure 5A; Supplemental Figures 2B and 3A). SPL2, SPL9, SPL11, and At5g38610 expression was strongly reduced immediately (0 h) and 4 h after HS and transcript levels recovered by 24 h, while At5g40510 transcripts were reduced at 48 h after HS. At5g38610 encodes a predicted pectin methylesterase inhibitor. Pectin methylesterase activity is required to loosen the cell walls at the sites of organ initiation at the shoot apical meristem (Peaucelle et al., 2008). Inhibiting this activity blocks organ initiation, suggesting that At5g38610 acts similarly to SPL proteins to limit organ formation.
Levels of Mature (Spliced) and Unspliced SPL Transcripts during the Maintenance of Acquired Thermotolerance and Their Dependency on Functional AGO1 and the miR156 Binding Site.
(A) Mature (top panel) and unspliced (middle panel) SPL2, SPL9, and SPL11 transcript levels relative to TUB6 (SPLn/TUB6)x in Col-0 and ago1-25 at the indicated times after a priming HS. Transcript level for the mature but not the unspliced SPL2, 9, and 11 was increased in ago1-25. Bottom panel: fold reduction of SPL transcript levels relative to the 0 h control. The repression depends on functional AGO1 and is evident for the mature mRNAs, but not the unspliced mRNAs, indicating that regulation occurs at the posttranscriptional level. The color code below the third panel applies to all panels in (A). Error bars are se of the mean of three replicates. The experiment was repeated three times independently, and one representative is shown.
(B) Reduction of SPL2 and SPL11 mRNA levels after HS is dependent on an intact miR156 binding site. ProSPL2:rSPL2 and ProSPL11:rSPL11 plants were subjected to a priming HS or no treatment (ntc) and harvested directly after the end of the treatment. Levels of the indicated endogenous SPL or rSPL transcripts were determined and the ACC/ntc ratio calculated. Error bars are se of the mean of three replicates. The experiment was repeated three times independently, and one representative is shown.
(C) GUS protein activity after a priming HS in ProSPL9:SPL9-GUS and ProSPL9:rSPL9-GUS as determined by MUG activity assay. Activity relative to the respective ntc and normalized to the 0-h time point averaged over three independent experiments ± se indicates that miR156 represses SPL9-GUS protein levels after HS.
If downregulation of target transcripts after HS was dependent on the miRNA, this would be expected to be reduced in ago1 mutants. Indeed, for all three examined SPL genes, downregulation was lost in ago1-25 mutants (Figure 5A, top and bottom panels). For SPL2 and SPL11, this loss was incomplete, which could be due to either residual AGO1 activity in the hypomorphic ago1-25 mutant or to an additional miRNA-independent mechanism of repression. miRNAs typically regulate their target genes posttranscriptionally through transcript cleavage. Thus, we investigated whether the HS-dependent repression occurred at the transcriptional or posttranscriptional level by quantifying unspliced SPL transcripts as a proxy for transcriptional activity (Bäurle et al., 2007; Le Masson et al., 2012). Unspliced transcript levels of SPL2 and SPL11 were not repressed by HS in the wild type or ago1-25 mutants (Figure 5A, middle and bottom panels; for primer positions, see Supplemental Figure 2C). For SPL9, a slight reduction of the unspliced transcript was observed, representing a small fraction of the reduction observed for the spliced transcript (Figure 5A, middle and bottom panels). Thus, HS-mediated repression of the tested SPL genes occurs largely or exclusively at the posttranscriptional level.
To directly test whether the miR156 binding site of SPL genes is required for the HS-dependent reduction of their transcript levels, we generated transgenic plants expressing a miRNA-resistant SPL2 or SPL11 gene under the control of its own promoter (i.e., ProSPL2:rSPL2 and ProSPL11:rSPL11, respectively) and compared transcript levels from the endogenous SPL gene and the transgenic rSPL gene by specific qRT-PCR. Comparing the native and miRNA-resistant version of SPL2 and SPL11, respectively, we found that the miR156-resistant transcript escaped HS-dependent downregulation (Figure 5B), demonstrating that an intact miR156 binding site is required for HS-mediated repression of SPL2 and SPL11.
We next analyzed whether GUS activity of a miRNA-resistant ProSPL9:SPL9-GUS line (ProSPL9:rSPL9-GUS) would be affected by HS (Yang et al., 2012). Indeed, relative GUS activity was progressively increased after HS in rSPL9-GUS compared with SPL9-GUS seedlings in a quantitative 4-methylumbelliferyl β-d-glucuronidase (MUG) assay (Figure 5C). Accordingly, suo-2 mutants, which are defective in miRNA-mediated translational repression (Yang et al., 2012), had reduced HS memory (Supplemental Figure 3B), suggesting HS-dependent miRNAs repress their targets in part through translational repression. Taken together, our results indicate that the HS-dependent downregulation of SPL2, 9, and 11 is controlled by miR156.
miR156 Is Required for HS Memory and High miR156h Levels Enhance the Memory
To investigate whether the induction of miR156 after HS correlated with a functional role in the HS response during the later stages, we generated a knockdown line of miR156a-f (Pro35S:MIM156a-f) (Franco-Zorrilla et al., 2007; Todesco et al., 2010). Pro35S:MIM156a-f plants were deficient in the maintenance of acquired thermotolerance (Figures 6A and 6B compared with Figures 1C and 1D), while acquisition of thermotolerance and basal thermotolerance were not affected (Supplemental Figures 4A and 4B compared with Figures 1A and 1B). Efficiency of the knockdown was confirmed by determining miR156a-f and miR156h levels in Pro35S:MIM156a-f after HS (Supplemental Figure 4C). Levels of miR156a-f were severalfold reduced, while miR156h levels were only slightly lower, suggesting a high specificity of the construct and of the stem-loop qRT-PCR. However, we cannot exclude a weak effect of Pro35S:MIM156a-f on miR156h. Plants expressing a mimicry construct against miR156h were also defective in HS memory (Pro35S:MIM156h; Supplemental Figure 4D). Thus, miR156 isoforms are required for HS memory.
miR156 Regulates the Maintenance of Acquired Thermotolerance.
(A) Knockdown of miR156a-f in Pro35S:MIM156 impairs HS memory, assayed by applying a tester HS two (+2) or three (+3) days after a priming HS (ACC). Photographs were taken 14 d after ACC.
(B) Quantification of data shown in (A); the percentage of seedlings in different phenotypic classes was determined for different genotypes and heat treatments. Gray arrows indicate the corresponding plates from (A). Panels are (top to bottom): ACC, ACC + HS d+2 110’, and ACC + HS d+3 90’. Damage classes are described in (D) and Figure 2C.
(C) Overexpression of miR156h enhances HS memory, assayed by applying a tester HS two (+2) or three (+3) days after a short ACC (37°C, 60’). Photographs were taken 14 d after ACC.
(D) Quantification of data shown in (C); the percentage of seedlings in different phenotypic classes was determined for different genotypes and heat treatments. Gray arrows indicate the corresponding plates from (C). Panels are (top to bottom): short ACC, short ACC + HS d+2 60’, and short ACC + HS d+3 45’. Damage classes are described below panels and in Figure 2C.
(E) Quantification of (C); average seedling fresh weight was determined from 27 ± 1 seedlings per genotype and treatment.
All plants of one treatment were grown on the same plate. The experiments were repeated at least five times with at least two independent transgenic lines with similar results. Asterisks indicate significantly different distribution compared with Col-0 (Fisher’s exact test, *P < 0.05; **P < 0.005).
We then asked whether overexpression of MIR156h was sufficient to enhance HS memory by generating transgenic plants expressing MIR156h under the control of the 35S promoter (Supplemental Figure 4C). Pro35S:MIR156h plants displayed an enhanced maintenance of acquired thermotolerance as evidenced by their increased survival and growth after a tester HS 2 or 3 d after the priming HS (Figures 6C to 6E). Conversely, acquisition of thermotolerance and basal thermotolerance were unaffected (Supplemental Figures 4A and 4B). In summary, miR156 levels are critical for HS memory.
As expected from previous reports on miR156a-f (reviewed in Huijser and Schmid, 2011), manipulating the levels of miR156h correlated with changes in growth and development. Reduction of active miR156h in Pro35S:MIM156h caused slightly early flowering with fewer leaves and a reduced leaf initiation rate (Supplemental Figure 4E and Supplemental Table 2). Overexpression of MIR156h caused an increased leaf initiation rate and very late flowering with respect to leaf number and days to flowering (Supplemental Figure 4E and Supplemental Table 2). Thus, miR156a-f and miR156h have largely overlapping functions during development and HS memory.
Repression of SPL2 and SPL11 by miR156 Is Required for HS Memory
To test whether the effect of miR156 on HS memory was mediated by downregulation of SPL function, we analyzed the HS memory of the miR156-resistant ProSPL2:rSPL2 and ProSPL11:rSPL11 lines. Indeed, HS memory was compromised in both lines similar to what was observed in the miR156 knockdown lines (Figures 7A and 7B; Supplemental Figure 3C). The reduced HS memory of ProSPL2:rSPL2 corresponded to a reduced maintenance of HS memory–related genes, particularly during the later stages after a priming HS (Figure 7C). Comparison of relative transcript levels at 48 and 72 h after HS indicated a 2- to 100-fold difference for APX2, HSA32, HSFA2, HSP17.6A, HSP22.0, and DIN2. Thus, SPL2 and SPL11 repression is required for HS memory and SPL2 directly or indirectly acts as a repressor of HS memory–related gene expression. SPL9, a close homolog of SPL2, can act as a transcriptional repressor in certain contexts by interfering with the assembly of activation complexes (Gou et al., 2011). A similar function could be envisioned for SPL2 and possibly SPL11 during the maintenance of acquired thermotolerance.
Downregulation of SPL2 Is Required for the Maintenance of Acquired Thermotolerance and Sustained Expression of HS Memory–Related Genes.
(A) ProSPL2:rSPL2 seedlings were assayed for the maintenance of acquired thermotolerance by applying a tester HS (HS) two (+2) or three (+3) days after a priming HS (ACC). Pictures were taken 14 d after ACC. All plants of one treatment were grown on the same plate. The experiments were repeated at least three times with two independent transgenic lines with similar results.
(B) Quantification of data shown in (A); the percentage of seedlings in different phenotypic classes was determined for different genotypes and heat treatments. Gray arrows indicate the corresponding plates from (A). Panels are (top to bottom): ACC, ACC + HS d+2 115’, and ACC + HS d+3 80’. Damage classes are illustrated in Figure 2C. Asterisks indicate significantly different distribution compared with Col-0 (Fisher’s exact test, **P < 0.005).
(C) Expression of selected HS memory–related genes during maintenance of acquired thermotolerance determined by qRT-PCR in Col-0 (black) and ProSPL2:rSPL2 (gray). Expression values were normalized to TUB6 and 48 h ntc [(GENE OF INTEREST/TUB6)x/(GENE OF INTEREST/TUB6)48 h ntc] and are displayed on a log10 axis for better resolution. Error bars are se of the mean of three technical and two biological replicates.
Expression of miR156 after HS Is Critical for HS Memory
Because of the developmental effects in the miR156-manipulating lines, we sought to separate the functions of miR156 during development and HS memory. To do so, we constructed transgenic lines that expressed either MIR156h or MIM156h under the control of the heat-inducible HSP21 promoter (Supplemental Figure 5A). In those transgenic lines, manipulation of miR156h levels occurred only after HS as evidenced by quantification of miR156h, while miR156a-f levels were not or only very slightly affected (Figure 8G). However, we cannot rule out that ProHSP21:MIM156h also affects miR156a-f, especially since target miRNAs may be inactivated by sequestration with no effect on miRNA levels (Todesco et al., 2010). ProHSP21:MIM156h and ProHSP21:miR156h plants were phenotypically indistinguishable from nontransgenic plants under control conditions and after a priming HS (Figures 8A and 8D; Supplemental Figure 5B). However, ProHSP21:MIM156h plants showed growth defects if the priming HS was followed by a tester HS 2 or 3 d later, indicating that HS memory was reduced (Figures 8A to 8C). This was confirmed by quantifying seedling fresh weight and the fraction of seedlings in different damage classes (Figures 8B and 8C). Conversely, ProHSP21:miR156h plants showed better growth and survival under those conditions (Figures 8D to 8F), which was confirmed by quantifying seedling fresh weight and the fraction of seedlings in different damage classes (Figures 8E and 8F). Thus, induction or repression of miR156h is critical for HS memory only after the priming HS, and a modified shoot architecture (as seen in plants expressing the 35S-driven constructs) is not responsible for this effect.
miR156 Is Required after HS to Regulate the Maintenance of Acquired Thermotolerance.
All plants of one treatment were grown on the same plate. The experiments were repeated at least three times with two independent transgenic lines with similar results. Asterisks indicate significantly different distribution compared with Col-0 (Fisher’s exact test, *P < 0.05; **P < 0.005).
(A) Heat-inducible knockdown of miR156h (ProHSP21:MIM156h) impairs the HS memory, assayed by applying a tester HS (HS) two (+2) or three (+3) days after a priming HS (ACC). Photographs were taken 14 d after ACC. For quantification, see (B) and (C).
(B) Quantification of data shown in (A); the percentage of seedlings in different phenotypic classes was determined for different genotypes and heat treatments. Gray arrows indicate the corresponding plates from (A). Panels are (top to bottom): ACC, ACC + HS d+2 110’, and ACC + HS d+3 80’. Damage classes are illustrated in Figure 2C.
(C) Quantification of (A); average seedling fresh weight was determined from 32 ± 3 seedlings per genotype and treatment.
(D) Heat-inducible overexpression of MIR156h (ProHSP21:MIR156h) enhances the HS memory, assayed by applying a tester HS two (+2) or three (+3) days after a priming HS (ACC). Photographs were taken 14 d after ACC. For quantification, see (E) and (F).
(E) Quantification of data shown in (D); the percentage of seedlings in different phenotypic classes was determined for different genotypes and heat treatments. Gray arrows indicate the corresponding plates from (D). Panels are (top to bottom): ACC, ACC + HS d+2 110’, and short ACC + HS d+3 80’. Damage classes are illustrated in Figure 2C.
(F) Quantification of (D); average seedling fresh weight was determined from 24 ± 3 seedlings per genotype and treatment.
(G) Mature miRNA levels of miR156a-f and miR156h in ProHSP:MIR156h and ProHSP:MIM156h determined by qRT-PCR relative to TUB6 (miR156h/TUB6)x. A several hundred fold induction of miR156h is observed in ProHSP:MIR156h plants after ACC, while no or only a very slight effect is observed for miR156a-f. Error bars are se of the mean of three replicates. Three independent experiments were performed with one representative shown.
(H) Expression of HS memory–related genes determined by qRT-PCR is increased after a priming HS in ProHSP:MIR156h. Expression values were normalized to TUB6 and 48 h ntc [(GENE OF INTEREST/TUB6)x/(GENE OF INTEREST/TUB6)48 h ntc]. Error bars are se of the mean of three replicates. Three independent experiments were performed with one representative shown.
To molecularly characterize the role of miR156 in HS memory, we analyzed transcript levels of HS memory–related genes in ProHSP21:miR156h plants. Expression of memory-related genes was more sustained after HS in ProHSP21:miR156h than in control plants, particularly during the later stages (Figure 8H). Thus, the enhanced maintenance of acquired thermotolerance due to MIR156h overexpression is likely caused by the increased expression of HS memory–related genes. Taken together, our findings suggest that by increasing memory-related gene expression and modifying development and its timing miR156 enhances HS memory.
DISCUSSION
While plant responses to acute stress are relatively well understood, those to chronic and recurring stress exposures are poorly characterized, despite their obvious importance for organisms in their natural environment. Here, we demonstrated that AGO1 acting through miR156 and SPL2 and 11 is an essential component of the HS memory in Arabidopsis. Despite not being the only miRNA induced after HS, miR156 is not only necessary, but its overexpression is also sufficient to boost HS memory, indicating its central role. miR156 is one of the most highly conserved miRNAs in plants, as it is found from higher plants to bryophytes (mosses), and its function in regulating the timing of developmental transitions appears to be conserved (Cho et al., 2012). Interestingly, miR156 induction after HS has also been reported in field mustard (Brassica rapa) and wheat (Triticum aestivum; Xin et al., 2010; Yu et al., 2012), suggesting that not only its developmental functions but also its functions in HS memory may be conserved. This directly suggests a way to enhance HS memory in two economically important crop species.
MIR156h showed the strongest upregulation after HS in our experiments. MIR156c and d also showed significant induction that could account for the observed increase in miR156a-f after HS. The nucleotide substitutions in miR156h compared with miR156a-f may have occurred as an evolutionary adaptation to functioning at elevated temperatures. It is conceivable that these substitutions may shift the balance of transcript cleavage and translational inhibition toward the latter, rendering the repression of target genes reversible and thus possibly better suited for HS conditions.
Similar to other stresses, HS transiently slows down growth and development including organ initiation (Skirycz and Inzé, 2010). As a predicted pectin methylesterase inhibitor, At5g38610 activity may be required to loosen the cell walls at the sites of organ initiation at the shoot apical meristem (Peaucelle et al., 2008). Inhibiting this activity blocks organ initiation, suggesting that At5g38610 acts similarly to SPL proteins to limit organ formation. It is conceivable that by promoting organ initiation shortly after HS (through repression of SPL and At5g38610), miR156 counterbalances the negative effect of HS on growth and helps the plant to resume growth after it has subsided. A more direct effect of miR156 appears to be the induction of HS memory–related genes, involving the HSFA2 transcription factor and release of SPL2-mediated repression. The SPL2-mediated repression of HS memory–related genes may be direct or indirect, as it has been shown that SPL9 can act as a transcriptional repressor (Gou et al., 2011). The genome-wide determination of direct SPL2 target genes during HS and development will be required for mechanistic insight into the extent to which the different functions involve the same or different downstream targets. It has been hypothesized that miR156 target genes act as licensing factors for the transition to flowering (Wu et al., 2009). Thus, stress-mediated induction of miR156 may delay the transition to flowering by prolonging the juvenile phase and serve to avoid flowering during a hot spell. The effect of a transient HS on the transition to flowering has not been systematically studied and may be different from the effect of chronic elevated ambient temperatures that promote the transition to flowering (Posé et al., 2013; Wigge, 2013). Because miR156 expression was not controlled by one of the well-described flowering time pathways (Bäurle and Dean, 2006; Fornara et al., 2010) and was found to gradually decrease with plant age, miR156 was grouped into the autonomous pathway and suspected to report the endogenous plant age for integration with other pathways. Recently, MIR156 expression was reported to be repressed by sugars, suggesting a role in the integration with endogenous energy levels (Yang et al., 2013; Yu et al., 2013). miR156 was also reported to delay flowering under cool ambient temperatures (Kim et al., 2012). Our findings add an additional dimension to this view in that plant age appears to be recorded not simply as chronological time, but rather as chronological time minus the time spent under adverse conditions and, thus, as “time well spent.”
Taken together, miR156 modulates plant development and response to HS. Several other miRNAs involved in plant development have also been found to be regulated by environmental stresses (Khraiwesh et al., 2012; Kruszka et al., 2012; Sunkar et al., 2012). Thus, a general pattern seems to emerge where plant development and its modulation in response to the environment is mediated by the same miRNAs. miRNAs may be particularly suitable for modulating longer term stress effects for two reasons: (1) regulation by a miRNA is reversible in the case of translational inhibition, and (2) miRNAs are relatively stable and are thus able to integrate signaling over longer periods of time. Interestingly, developmental timing in Caenorhabditis elegans is regulated by the miRNAs lin-4 and let-7. The induction of dauer larvae depends on environmental cues and both miRNAs are involved in its regulation (Karp and Ambros, 2012). Thus, environmental modulation of development through miRNAs appears to be a general mechanism in plants and animals.
It cannot be ruled out that the function of AGO1 during HS memory is in part mediated by additional miRNAs, such as other heat-induced miRNAs identified in this study. The elucidation of this question is an interesting field of further study. Here, we report the functional role of a plant miRNA in the temporal modulation of abiotic stress adaptation. Studies in many different plant species have shown that miR156 levels do not only change in response to heat stress but also after various other stresses such as cold, salt, and drought stress, as well as UV-B radiation, hypoxia, and biotic stress. The detailed analysis of the regulation of miR156 remains an important research question, as it will be interesting to see whether miR156 also integrates stress memory and development for other stresses.
METHODS
Plant Material, Growth, and Stress Treatment Conditions
Arabidopsis thaliana plants were grown in GM medium (1% [w/v] glucose) at 23°C/ 21°C in 16-h-light/8-h-dark cycles unless otherwise indicated. ago1-25, ago1-27, and dcl1-9 were obtained from N. Baumberger and C. Dean, respectively. ProSPL9:SPL9-GUS, ProSPL9:rSPL9-GUS, and suo-2 (Yang et al., 2012) were obtained from S. Poethig. Pro35S:MIM156 and Pro35S:MIMIPS1 Agrobacterium tumefaciens stocks, hsp101 (N566394), and hsfa2 (N508978) were obtained from the Nottingham Arabidopsis Stock Centre.
HS treatments were performed in a water bath or incubator. For the priming HS (ACC), 4-d-old seedlings were subjected to a heat regime of 1 h, 37°C; 1.5 h, 23°C; and 45 min, 44°C, except in Figure 6C, where a short ACC was applied omitting the 44°C step. All priming treatments were started 6 h after dawn to ensure comparability and minimize diurnal effects. Nine hours after dawn, a tester HS (44°C only) was applied 2 d later for 90 to 115 min or 3 d later for 80 to 90 min, as indicated in the figures. After the short ACC (Figure 6C), the tester HS was applied 2 d later for 60 min or 3 d later for 45 min. For acquisition of thermotolerance, the regime was 1 h, 37°C; 1.5 h, 23°C; and 0 to 180 min, 44°C; basal thermotolerance was assayed by incubation for 0 to 60 min at 44°C without prior treatment at 37°C. After heat treatments, plants were returned to normal growth conditions until analysis. For analysis of gene expression during maintenance of acquired thermotolerance, 4-d-old seedlings were subjected to ACC or mock treatment and harvested at the indicated time points. Samples of time-course experiments were harvested at the same time of day throughout the experiment to minimize diurnal effects wherever possible.
Construction of Transgenic Lines
For ProSPL2:rSPL2, a 3.9-kb promoter fragment was amplified with primers (all primer sequences are given in Supplemental Table 3) introducing NcoI and SacI restriction sites. The coding sequence and 3′ region (2.5 kb) was amplified, mutating the miR156 binding site while preserving the amino acid sequence with primers introducing SacI and XmaI sites (Wang et al., 2008). After sequencing, the ProSPL2:rSPL2 construct was assembled in pUC-ML939 and transferred via AscI into pBarMAP-ML516. For ProSPL11:rSPL11, a 2.6-kb promoter fragment was amplified with primers introducing NcoI and SacI restriction sites. The coding sequence and 3′ region (2.5 kb) was amplified, mutating the miR156 binding site while preserving the amino acid sequence with primers introducing SacI and XmaI sites. After sequencing, the ProSPL11:rSPL11 construct was assembled in pUC-ML939 and transferred via AscI into pBarMAP-ML516. Transgenic plants were generated using the floral dip method (Clough and Bent, 1998). For Pro35S:MIR156h, a 0.5-kb fragment encompassing the pri-miR156h sequence was amplified introducing XhoI and BamHI restriction sites and subcloned in ML595. MIM156h was generated using specific primer sequences analogous to MIM156 (Franco-Zorrilla et al., 2007). ProHSP21:MIR156h and PRoHSP21:MIM156h were generated by replacing the HSP21 coding sequence in a 2.8-kb genomic fragment with the MIR156h and MIM156h fragments described above. All experiments were performed with multiple independent homozygous lines carrying a single insertion (determined by segregation analysis).
RNA Preparation, Microarray Analysis, and qRT-PCR
RNA was extracted from seedlings using a Trizol-based protocol (Pant et al., 2009) or a hot-phenol RNA extraction protocol (Bäurle et al., 2007). For microarray analysis, RNA from three biological replicates per genotype and treatment (4 h after ACC, 52 h after ACC, and 4 h ntc) was purified over an RNeasy Plant RNA extraction column (Qiagen) and processed for hybridization of Affymetrix ATH1 GeneChips (NASC microarray facility). The times were chosen to harvest all samples at the same time of day. For qRT-PCR, total RNA was treated with TURBO DNA-free (Ambion), and 2 μg was reverse transcribed with SuperScript III (Invitrogen) according to the manufacturers’ instructions. cDNA was diluted 1:100 into qPCR reactions with GoTaq qPCR Master Mix (Promega) and an Applied Biosystems AB7300 or Roche LightCycler 480 instrument. Expression was normalized to TUBULIN6 using the comparative CT method. Primer sequences are listed in Supplemental Table 3. Time of sampling was chosen to match either the 0- or 4-h time point (all harvesting was done at the same time of day), without noticeable difference between experiments. Robust expression at low levels was detected in ntc controls for all HS-responsive genes in microarray and qRT-PCR experiments. Gene Ontology term enrichment was calculated with the AmiGO term enrichment tool (Boyle et al., 2004).
Primary miRNA Profiling, miRNA Target Prediction, and Mature miRNA Detection
The pri-miRNA profiling was performed as described (Pant et al., 2009). The miRNA target prediction was performed using online search tools with the default settings (Moxon et al., 2008; Ossowski et al., 2008; Alves et al., 2009; Dai and Zhao, 2011) and published degradome data (Addo-Quaye et al., 2008). For detection of mature miRNA, a modified stem-loop qRT-PCR protocol was used with SYBR Green as a dye (Chen et al., 2005; Pant et al., 2008).
MUG Assay
To quantify GUS protein level, MUG activity was determined with a fluorescence spectrophotometer according to standard procedures (Jefferson et al., 1987).
Accession Numbers
Sequence data from this article can be found in the Arabidopsis Genome Initiative database under the following accession numbers: AGO1 (AT1G48410), DCL1 (AT1G01040), HSFA2 (At2G26150), APX2 (AT3G09640), HSA32 (At4G21320), HSP17.6A (AT5G12030), HSP22.0 (AT4G10250), HSP70 (At3G12580), At5G38610, At5G40510, SPL2 (AT5G43270), SPL9 (AT2G42200), SPL11 (AT1G27360), MIR156A (AT2G25095), MIR156B (AT4G30972), MIR156C (AT4G31877), MIR156D (AT5G10945), MIR156E (AT5G11977), MIR156F (AT5G26147), MIR156G (AT2G19425), MIR156H (AT5G55835), HSP101 (AT1G74310), SUO (At3G48050), and HSP21 (At4G15802).
Supplemental Data
The following materials are available in the online version of this article.
Supplemental Figure 1. Acquisition of Thermotolerance and Basal Thermotolerance in ago1-25 and GO Term Frequency Analysis of ATH1 Transcriptome Profiling.
Supplemental Figure 2. MIR156h and SPL2, 9, 11 Expression Levels (Additional Time Points) and Primer Positions.
Supplemental Figure 3. Expression Profile of New miR156h Target Genes after a Priming HS and Requirement of SUO and the miRNA Binding Site of SPL11 for Maintenance of Acquired Thermotolerance.
Supplemental Figure 4. Further Characterization of Pro35S:MIM156 and Pro35S:miR156h.
Supplemental Figure 5. Further Characterization of Heat-Inducible miR156-Manipulating Lines.
Supplemental Table 1. miR156a-f and miR156h Targeted Genes Predicted by at Least Two Search Tools and Their Gene Description and Expression.
Supplemental Table 2. Average Flowering Time (Days and Terminal Leaf Number) and Leaf Initiation Rate of miR156-Manipulating Transgenic Lines.
Supplemental Table 3. Sequences of Oligonucleotides Used in This Study.
Supplemental Data Set 1. Transcriptome Profiling in Col and ago1-25 at 4 and 52 h after HS.
Supplemental Data Set 2. Profiling of Heat-Responsive Primary miRNAs.
Acknowledgments
We thank N. Baumberger, C. Dean, R.S. Poethig, and the European Arabidopsis Stock Centre for materials and D. Walther (MPI Molecular Plant Physiology) for help with the microarray analysis. We thank Y. Zhang and B. Pipke for technical assistance and M. Lenhard and members of our laboratory for helpful comments. I.B. acknowledges support from a Royal Society University Research Fellowship, a Sofja-Kovalevskaja-Award (Alexander-von-Humboldt-Foundation), CRC 973 Project A02 (Deutsche Forschungsgemeinschaft), and funds from the John Innes Centre, Norwich, UK.
AUTHOR CONTRIBUTIONS
A.S., S.A., W.S., and I.B. designed research. A.S., S.A., K.H., B.D.P., and I.B. performed research. A.S., B.D.P., W.S., and I.B. analyzed data. A.S. and I.B. wrote the article.
Footnotes
The author responsible for distribution of materials integral to the findings presented in this article in accordance with the policy described in the Instructions for Authors (www.plantcell.org) is: Isabel Bäurle (isabel.baeurle{at}uni-potsdam.de).
↵1 Current address: Plant Biology Division, The Samuel Roberts Noble Foundation, 73401 Ardmore, OK.
↵[C] Some figures in this article are displayed in color online but in black and white in the print edition.
↵[W] Online version contains Web-only data.
Glossary
- HS
- heat stress
- miRNA
- microRNA
- ACC
- acclimation
- Col-0
- Columbia-0
- ntc
- nontreated control
- qRT-PCR
- quantitative RT-PCR
- pri-miRNA
- primary miRNA
- MUG
- 4-methylumbelliferyl β-d-glucuronidase
- Received February 1, 2014.
- Revised March 27, 2014.
- Accepted April 4, 2014.
- Published April 25, 2014.