|Home | About | Journals | Submit | Contact Us | Français|
Heat stress affects epigenetic gene silencing in Arabidopsis. To test for a mechanistic involvement of epigenetic regulation in heat-stress responses, we analyzed the heat tolerance of mutants defective in DNA methylation, histone modifications, chromatin-remodeling, or siRNA-based silencing pathways. Plants deficient in NRPD2, the common second-largest subunit of RNA polymerases IV and V, and in the Rpd3-type histone deacetylase HDA6 were hypersensitive to heat exposure. Microarray analysis demonstrated that NRPD2 and HDA6 have independent roles in transcriptional reprogramming in response to temperature stress. The misexpression of protein-coding genes in nrpd2 mutants recovering from heat correlated with defective epigenetic regulation of adjacent transposon remnants which involved the loss of control of heat-stress-induced read-through transcription. We provide evidence that the transcriptional response to temperature stress, at least partially, relies on the integrity of the RNA-dependent DNA methylation pathway.
Plants frequently encounter unfavorable environmental conditions such as heat, cold, drought, and pathogen infections. Adaptive strategies of plants to cope with stress are complex and depend on the precise and timely regulation of stress-responsive gene networks (Chinnusamy et al., 2007; Larkindale and Vierling, 2008; Cramer et al., 2011; Mittler et al., 2012). In response to stress, genome-wide transcriptional reprogramming includes the induction of genes involved in direct stress protection (e.g. osmoprotectants, detoxifying enzymes) and of genes that encode regulatory proteins (e.g. transcription factors, protein kinases), whereas genes involved in growth are generally repressed. However, after a period of stress (e.g. heat), and when environmental conditions are again favorable, plants need to terminate the transcriptional stress response and to reinitiate the developmental program in order to successfully recover and resume growth and development.
The transcriptional capacity of genes is intimately connected to their epigenetic states (Bonasio et al., 2010; Gibney and Nolan, 2010). Triggered by developmental or environmental cues, genes can adopt alternate epigenetic states, which result in differences of gene expression without changes in the primary DNA sequence (Lukens and Zhan, 2007; Schmitz and Amasino, 2007; Chinnusamy and Zhu, 2009). The induction of these states is associated with modifications of DNA (e.g. cytosine methylation) and histones (e.g. acetylation, methylation, phosphorylation, ubiquitination), which together modulate chromatin architecture, and hence the accessibility of particular regions of the genome to the transcriptional machinery (Henderson and Jacobsen, 2007; Feng et al., 2010). In plants, chromatin states can be modulated by double-stranded RNA and derived 24 nucleotide (nt) short RNAs in a process called RNA-directed DNA methylation (RdDM) (Huettel et al., 2007; Pikaard et al., 2008; Law and Jacobsen, 2010; Zhang and Zhu, 2011). Briefly, in Arabidopsis, these short RNAs are produced from double-stranded RNA precursors generated by the combined action of the plant-specific DNA-dependent RNA polymerase IV (Pol IV) and RDR2 (RNA-DEPENDENT RNA POLYMERASE 2). DCL3 (DICER-LIKE3) mediates processing of dsRNA to siRNAs, in conjunction with their loading onto AGO4 (ARGONAUTE 4) effector complexes (Li et al., 2006; Pontes et al., 2006). Loaded AGO4, together with other numerous protein factors, subsequently assembles on nascent transcripts produced by RNA polymerase V (Pol V), resulting in the recruitment of chromatin modifiers like the de novo DNA methyltransferase DRM2 and histone-modifying activities (Wierzbicki et al., 2009).
There is growing evidence that stress responses in plants affect epigenetic regulation and require certain epigenetic regulators (Chinnusamy and Zhu, 2009): for example, UV, cold, and heat stress result in the reactivation of silent transgenes and endogenous transposable elements, albeit without the reduction of DNA methylation and repressive histone marks (Lang-Mladek et al., 2010; Pecinka et al., 2010; Tittel-Elmer et al., 2010; Ito et al., 2011). As a mechanism for reactivation during heat stress, a reduction in nucleosome occupancy was proposed. In tobacco, metal, cold, and salt stress leads to demethylation of coding regions, for which reactive oxygen species (ROS) formed in response to these abiotic stressors have been suggested as the cause (Choi and Sano, 2007). Similarly, salt stress leads to hypermethylation of CHG sites within satellite DNA of the facultative halophyte Mesembryanthemum crystallinum (Dyachenko et al., 2006). Low-temperature growth conditions result in demethylation of the Antirrhinum majus transposon Tam3 (Hashida et al., 2006). Roots of maize seedlings subjected to cold stress react by genome-wide demethylation (Steward et al., 2002). The RdDM factor NRPD1 is involved in the biogenesis of natsiRNAs, which are formed from overlapping transcripts of convergent gene pairs and are required for salt stress tolerance in Arabidopsis (Borsani et al., 2005). Responses of Arabidopsis to salt and cold stress are further impaired by mutations within components of histone deacetylase complexes (Zhu et al., 2008; Chen et al., 2010; To et al., 2011a).
It is well established that stressful conditions, including specific long-term heat regimes, interfere with and partially override epigenetic pathways that protect plant genomes from the deleterious action of transposable elements (Chinnusamy and Zhu, 2009; Lang-Mladek et al., 2010; Pecinka et al., 2010; Tittel-Elmer et al., 2010; Hauser et al., 2011; Ito et al., 2011; Grativol et al., 2012). However, it remained unclear whether epigenetic pathways play a vital role in successful plant stress responses. In this study, we analyzed a set of epigenetic mutants for their tolerance to heat stress and present evidence that the Pol IV/V pathway and the Rpd3-type histone deacetylase HDA6 are important for basal heat tolerance. Interestingly, the Pol IV/V pathway and HDA6 appear to function independently at different steps in the Arabidopsis heat-stress response. We also propose molecular mechanisms for how RdDM-controlled remnants of transposable elements may contribute to the expression of adjacent protein-coding genes during the heat response of Arabidopsis plants.
Previously, we showed that prolonged heat stress affects epigenetic silencing of transgenes and transposons in Arabidopsis (Lang-Mladek et al., 2010). To investigate whether epigenetic pathways are required for basic heat-stress tolerance, we first characterized mutants deficient in DNA methyltransferases, histone-modifying enzymes, chromatin-remodeling factors, and genes involved in small RNA biogenesis and function for their tolerance to heat treatment (Table 1 and Figure 1).
Heat-tolerance tests showed that loss of NRPD2, the second-largest subunit of both Pol IV and Pol V complexes (Herr et al., 2005; Onodera et al., 2005; Pontier et al., 2005; Ream et al., 2009), rendered plants hypersensitive to acute heat stress (Figure 1). To exclude any sporadic epigenetic variation (Becker et al., 2011; Schmitz et al., 2011), we analyzed an additional nrpd2 allele (nrpd2:G1072E, M. Matzke, unpublished) that was also hypersensitive to heat stress (Supplemental Figure 1A). Interestingly, the loss of neither the Pol IV-specific largest subunit NRPD1 (Herr et al., 2005; Onodera et al., 2005; Pontier et al., 2005) nor the Pol V-specific subunit NRPE1 (Kanno et al., 2005; Pontier et al., 2005) significantly affected survival rates of the respective mutants (Figure 1), suggesting that both Pol IV and Pol V must be simultaneously compromised for phenotypic consequence with regard to Arabidopsis heat tolerance.
The involvement of RdDM components in basal heat tolerance is further supported by heat hypersensitivity of hda6 mutants (Figure 1). HDA6 is an RPD3-type histone deacetylase known to be involved in gene repression by RdDM (Aufsatz et al., 2002; Probst et al., 2004; May et al., 2005; Earley et al., 2010). It should be noted, however, that HDA6 also has a role in gene silencing independently of RdDM, such as in jasmonate and ethylene signaling (Zhu et al., 2011), ABA and salt stress responses (Chen et al., 2010), freezing tolerance (To et al., 2011a), and in phytochrome B-dependent chromatin compaction (Tessadori et al., 2009).
While hda6 mutants were as sensitive to heat as nrpd2 mutants, ago4 mutants were slightly less sensitive (Figure 1). AGO4, loaded with 24-nt siRNAs, is a key component of RdDM effector complexes and physically interacts with Pol V at its target loci, leading to recruitment of DNA- and histone-modifying proteins (Matzke et al., 2007; Haag and Pikaard, 2011; Zhang and Zhu, 2011). Reduced sensitivity of ago4 to heat stress might be due to partially redundant functions of AGO4 and AGO6 in RdDM (Zheng et al., 2007). Loss of other RdDM components, DCL3 and RDR2 (Zhang and Zhu, 2011), also significantly decreased thermo tolerance (Figure 1). Heat sensitivity of dcl3 and rdr2 plants, however, was less pronounced than that of nrpd2, hda6, and ago4 mutants. This could be due to genetic redundancy within DICER and RDR gene families during the Arabidopsis heat-stress response. Taken together, our mutant analysis suggests that integrity of the RdDM pathway is required for heat-stress acclimation in Arabidopsis. In contrast, mutants of DNA methyltransferases (drm1/drm2; cmt3) (Bartee et al., 2001; Lindroth et al., 2001; Cao and Jacobsen, 2002), of chromatin-remodeling factors (ddm1; mom1) (Jeddeloh et al., 1998; Mittelsten Scheid et al., 1998; Amedeo et al., 2000), and of histone methyltransferases (kyp1; suvh2) (Jackson et al., 2002; Naumann et al., 2005) showed only minor, if any, heat-response phenotypes (Table 1). Interestingly, loss of functional DCL4, which is involved in the production of 21-nt tasiRNA in posttranscriptional gene silencing, rendered plants hypersensitive to heat stress (Supplemental Figure 1B and Table 1), whereas two other mutants involved in this pathway, dcl2 and rdr6 (Dalmay et al., 2000; Mourrain et al., 2000; Xie et al., 2004), showed a wild-type heat tolerance (Table 1).
The marked hypersensitivity of nrpd2 and hda6 mutants to acute heat stress points towards an involvement of small RNA-mediated transcriptional gene silencing in the stress-induced transcriptional reprogramming that is important for basic heat tolerance. To study the role of NRPD2 and HDA6 during the transcriptional heat-stress responses in Arabidopsis, we compared the transcriptional profiles of nrpd2 and hda6 mutants under heat stress and after recovery from stress with those of corresponding wild-type plants. RNA from soil-grown plants exposed to heat (42°C for 16 h) and after recovery (21°C for 2 d) was hybridized to the Arabidopsis ATH1 Affymetrix GeneChip (Figure 2).
Under control conditions, only a few genes showed an altered transcriptional response in nrpd2 and hda6 mutants, in line with the normal developmental phenotype of the mutants. Seventeen genes were differentially regulated in nrpd2 plants compared to wild-type plants (cut-off logFC ≥ 1 or ≤ –1, p-value < 0.05; eight genes up- and nine genes down-regulated, respectively) (Figure 2 and Supplemental Table 1). In hda6 mutants, 71 probe sets were up- and 13 genes were down-regulated under control conditions (Figure 2 and Supplemental Table 1). These data are in good agreement with recent results from tiling arrays that identified 81 transcriptional units misregulated in axe1-5 mutants of HDA6 (To et al., 2011b).
The direct heat response in nrpd2 mutants closely resembled that of wild-type plants (48 probe sets were up- and 93 down-regulated in the mutant, compared to wild-type plants), whereas hda6 mutants displayed a much broader disturbed transcriptional response (404 probe sets were up- and 752 down-regulated) (Figure 2 and Supplemental Table 1). Conversely, the transcriptional response upon 2 d of recovery from heat was compromised to a higher extent in nrpd2 mutants (810 probe sets were up- and 509 down-regulated) than in hda6 plants (48 probe sets were up- and 13 down-regulated), compared to wild-type plants (Figure 2 and Supplemental Table 1). There was only little overlap between misregulated gene sets in nrpd2 and hda6 mutants both within and across conditions (Figure 2, Supplemental Figure 2, and Supplemental Table 2). Thus, although nrpd2 and hda6 mutants exhibited a similar degree of hypersensitivity to acute heat stress (Figure 1), NRPD2 and HDA6 are required at different stages of the Arabidopsis heat response and directly or indirectly regulate distinct sets of target genes. This is in line with recent observations showing only a minor role of the RdDM pathway in HDA6-mediated gene silencing during normal development (To et al., 2011b). Cluster analysis of transcriptional changes between wild-type and mutants during heat and recovery further revealed that, in hda6 mutants, compared to nrpd2 mutants, differences in gene expression were rather quantitative than qualitative, indicating that hda6 mutants responded in a similar way to wild-type plants, albeit with an altered amplitude (Supplemental Figure 3).
GO analyses showed that genes up-regulated during recovery of nrpd2 mutants were primarily involved in abiotic stress responses, starch catabolism, and fatty acid beta oxidation, whereas down-regulated genes played roles in auxin and cytokinin signaling pathways (Supplemental Figures 4 and 5). Thus, the transcriptional pattern of nrpd2 plants during the recovery period resembled that of wild-type plants under stress conditions, indicating that nrpd2 plants are unable to properly terminate the transcriptional stress response and to reinitiate the developmental program during recovery. GO annotation of gene sets misregulated in hda6 during heat stress revealed that they belong to very diverse functional categories, including vegetative and reproductive development, hormone and stress response, signaling, metabolism, transport, and protein processing (Supplemental Figures 6 and 7). Two functional groups, ‘chromatin assembly/disassembly’ and ‘regulation of transcription’, were overrepresented within down-regulated probe sets. The proper control of gene activation or repression depends on chromatin organization, including appropriate nucleosome assembly and positioning, which regulates the accessibility of Pol II to particular regions of the genome (Chodavarapu et al., 2010; Zhu et al., 2012). Environmental changes induce a quick response of the cellular transcriptional machinery that is associated with reduced nucleosomal density at stress-specific cis-elements and rapid recruitment of Pol II to the 5′ regions of stress-responsive genes (Kim et al., 2012). Thus, a general dysfunction in nucleosome formation and transcriptional regulation might weaken and/or delay stress-specific transcriptional responses in hda6 mutants.
During normal plant development, the major function of RdDM is the control of transposable elements, thereby contributing to other epigenetic pathways in maintaining genome stability (Matzke et al., 2007; Hollick, 2008). The representation of transposon-derived probe sets on ATH1 gene chips is low (Supplemental Figure 8A), which hampered the analysis of ‘genuine’ RdDM targets during the heat-stress response. Within these limitations, we found an enrichment of transposon-derived probe sets only among genes up-regulated in nrpd2 mutants under control conditions and during heat stress, compared to their general representation on the ATH1 gene chip (Supplemental Figure 8B). The enrichment was more pronounced within probe sets up-regulated under control conditions than among those up-regulated upon stress. It has recently been reported that exposure to heat stress per se results in the activation of transposable elements, which is likely caused by a heat-induced reduction of nucleosome occupancy (Lang-Mladek et al., 2010; Pecinka et al., 2010; Tittel-Elmer et al., 2010; Ito et al., 2011). If heat-dependent activation overrode epigenetic regulation and occurred to a similar extent in wild-type plants and in nrpd2 as well as hda6 mutants, the identification of transposons additionally regulated by NRPD2 and/or HDA6 may have been masked in our analyses.
The involvement of HDA6 in stress responses and underlying molecular mechanisms of HDA6-dependent genome regulation have been previously studied in some detail (Chen et al., 2010; Lang-Mladek et al., 2010; To et al., 2011a, 2011b; Kim et al., 2012). However, the role of the Pol IV/V pathway in stress tolerance remains unclear to date (Kanno and Habu, 2011). Thus, we focused on mechanisms of NRPD2-mediated, temperature-responsive gene expression, exemplified by single genes either up- or down-regulated in nrpd2 mutants during recovery from heat stress.
At1g34220 (Figure 3A), encoding a regulator of vacuolar protein sorting 4 (Vps4) activity in the multivesicular body pathway, was up-regulated in stress-recovered nrpd2 mutants in our transcriptome analyses (Supplemental Table 1). We verified by qPCR that the transcription of At1g34220 was induced by heat in wild-type plants and, to a higher level, in nrpd2 mutants (Figure 3B). During recovery, transcript levels strongly decreased in wild-type plants but remained high in nrpd2 mutants (Figure 3B). Analysis of the genomic location revealed that the region upstream of At1g34220 contains remnants of transposable elements (AT1TE40565 and AT1TE40570). AT1TE40565 is a part of the pseudogene At1g34230, while AT1TE40570 corresponds to the promoter region of At1g34230 (Figure 3A). The whole region is associated with small RNA signatures, including 24-nt small RNAs, indicating that the epigenetic status of the transposon remnants is at least partially regulated by RdDM.
We compared DNA methylation of the transposon remnant in wild-type and nrpd2 plants from control, heat, and recovery conditions by Chop-PCR: genomic DNA was digested with methylation-sensitive restriction enzymes followed by PCR amplification with primers specific for the transposon. In wild-type control plants, the transposon remnant had CG, CHG, and CHH methylation (Figure 3C). In nrpd2 control plants, the transposon remnant showed a strong reduction in CHH methylation, which persisted upon exposure to heat stress and during recovery (Figure 3C). Since CHH methylation is primarily controlled by RdDM (Matzke et al., 2007; Lahmy et al., 2010) and depends on the RdDM factor NRPD2 (Figure 3C), the transposon remnant is a true target of the Pol IV/V pathway. However, CHH methylation of this transposon remnant is also modulated by heat stress, since wild-type plants had strongly reduced CHH methylation upon heat stress and during recovery (Figure 3C).
Since loss of DNA methylation promotes transposon reactivation (Miura et al., 2001; Singer et al., 2001), we analyzed whether the observed changes in DNA methylation had an impact on the expression of the At1g34230 transposon-related pseudogene (Figure 3A and and3D).3D). nrpd2 and wild-type control plants showed similar At1g34230 expression, indicating that loss of CHH methylation does not impact on transcription in the absence of a heat stimulus (Figure 3D). Heat stress resulted in a slightly higher transcriptional induction of At1g34230 in nrpd2 mutants compared to wild-type plants (Figure 3D). Moreover, while transcription of At1g34230 declined during recovery in wild-type plants, the induction persisted in recovered nrpd2 mutants (Figure 3D). Collectively, these results suggest that loss of NRPD2 and heat stress act synergistically on At1g34230 expression and that resilencing during recovery requires NRPD2. Since CHH methylation levels of the At1g34230-associated transposon remain low during recovery in both wild-type plants and nrpd2 mutants (Figure 3C), NRDP2-dependent resilencing does not seem to depend on the reacquisition of CHH methylation.
Transcriptional derepression of transposons that are dispersed in euchromatic regions of the genome can induce transcription of intergenic regions (IGRs) or neighboring genes (Huettel et al., 2006). Since the transcriptional pattern of the transposon-related pseudogene At1g34230 closely resembled that of the neighboring protein-coding gene At1g34220 (Figure 3B and and3D),3D), we scanned for read-through transcription originating in the transposon-related pseudogene. We performed RT–PCR analyses of read-through transcripts using a forward primer anchored within the 3′ end of the pseudogene and different reverse primers binding 400 bp, 700 bp, and 1100 bp downstream of At1g34230 (Figure 3E and Supplemental Figure 9). In wild-type plants, read-through transcription was only induced by heat stress and was terminated again during recovery. In nrpd2 mutant plants, low levels of read-through transcripts were already detected under control conditions. During stress, read-through transcription was enhanced compared to wild-type plants and transcripts were only partially reduced during recovery (Figure 3E and Supplemental Figure 9A). Thus, during heat stress and recovery, similar transcript profiles of a protein-coding gene and upstream RdDM-controlled transposon sequences correlate with the presence of read-through transcripts, and timely clearance of read-through transcripts during recovery appears to involve NRPD2. Intriguingly, when read-through transcription was initiated, repeats located directly upstream of At1g34220 (Figure 3A) lost some CG methylation in wild-type plants during heat stress (Supplemental Figure 9A), highlighting the possible potential of read-through transcripts from a remote transposon to influence the methylation status of neighboring promoter regions.
Another putative NRPD2-dependent heat-stress target identified in our microarray analyses is the pentatricopeptide repeat (PPR)-containing protein-coding gene At1g07590 (Figure 4A), which has previously been associated with responses to heavy metal stress (Sarry et al., 2006). Under control conditions, At1g07590 was expressed at similar levels in wild-type and nrpd2 plants (Figure 4B). Heat stress moderately induced the expression in wild-type plants and to higher levels in nrpd2 mutants (Figure 4B). During recovery, transcript levels did not change in wild-type plants, whereas they further increased in NRPD2-deficient plants (Figure 4B). Interestingly, the expression pattern of At1g07590 in nrpd1 was similar to that of nrpd2, whereas the transcriptional response in nrpe1 resembled that of wild-type plants (Supplemental Figure 10A), suggesting a Pol IV-specific involvement in regulation of PPR gene expression in response to heat stress.
The promoter region of At1g07590 contains a repetitive element with an EnSpm DNA transposon consensus sequence (Kohany et al., 2006) and is associated with small RNA signatures (Figure 4A). Chop–PCR on wild-type plants revealed that the repeat was methylated in CG and CHG but not CHH contexts under all conditions tested (Figure 4C). In nrpd2 mutants, both CG and CHG methylation was already strongly reduced under control conditions and remained low during heat stress and recovery (Figure 4C), demonstrating that repeat methylation is controlled by RdDM. Scanning of the At1g07590 upstream region for stress-responsive cis-elements revealed two heat shock elements (HSE), one drought-responsive element (DRE), and two abscisic acid response elements (ABRE) outside of the repeat, and one ABRE located within the repetitive sequence (Supplemental Figure 10B). Thus, the methylation status of the repeats and possibly associated stress-responsive cis-elements regulates the amplitude of PPR gene heat-stress induction and is dependent on the Pol IV/V pathway.
At1g29430, At1g29440, At1g29450, At1g29460, At1g29500, and At1g29510 represent clustered auxin-responsive genes with high sequence similarity that were found to be down-regulated in recovered nrpd2 mutants in our microarray analyses (Figure 5A, Supplemental Figure 11, and Supplemental Tables 1 and 3). In total, the cluster contains eight highly homologous genes (Figure 5A and Supplemental Figure 11). Expression of three selected genes, At1g29450, At1g29490, and At1g29500, was verified by qRT–PCR (Figure 5B). Upon heat stress, expression of these genes was repressed in both wild-type and nrpd2 plants. During recovery, expression was elevated to roughly the control condition levels in wild-type plants, whereas transcript levels increased only moderately in nrpd2 plants (Figure 5B). To investigate the individual role of Pol IV and Pol V in the transcriptional response, we analyzed the expression of the representative auxin-response gene At1g29450 in Pol IV and Pol V single mutants. The transcriptional response of At1g29450 in nrpd1 and nrpe1 resembled that of nrpd2 plants (Supplemental Figure 12A), suggesting an involvement of both Pol IV and Pol V in regulating the expression of At1g29450.
The cluster of auxin-responsive genes harbors the COPIA-like transposon At1g29475 (Figure 5A). We monitored the DNA methylation status and transcriptional activity of the transposon in wild-type and nrpd2 plants under control conditions, upon heat stress, and during recovery (Figure 5C and and5D).5D). In wild-type plants, the transposon was methylated in CG and CHH contexts across all tested conditions (Figure 5C). Methylation was strongly reduced in nrdp2 plants already under control conditions and remained low during heat stress and recovery (Figure 5C), demonstrating that the methylation status of the COPIA-like transposon is controlled by RdDM. Reduction of DNA methylation in nrpd2 mutants resulted in slightly elevated expression of the transposon under control conditions compared to wild-type plants (Figure 5D). Heat stress reactivated the transposon in wild-type plants to high levels, whereas transcription was reduced again during recovery (Figure 5D). Heat-dependent activation in wild-type occurred despite persisting DNA methylation and might be mechanistically similar to the reported activation of other silent repeats under prolonged exposure to acute heat stress (Lang-Mladek et al., 2010; Pecinka et al., 2010; Tittel-Elmer et al., 2010; Ito et al., 2011). In nrpd2 mutants, transcription of the COPIA-like transposon was induced by heat to levels comparable to wild-type; however, expression did not decrease during recovery (Figure 5D). Thus, the transposon responded to heat stress independently of NRPD2, but resilencing during recovery required NRPD2.
Next, we investigated whether read-through transcription from the transposon might be involved in regulating the expression of the auxin-responsive genes (Figure 5E). Indeed, read-through transcription into adjacent IGRs was observed in wild-type plants exclusively under heat-stress conditions and terminated again during recovery (Figure 5E and Supplemental Figure 12B). In nrpd2 mutants, read-through transcription during heat stress was observed at slightly higher levels than in stressed wild-type plants, and persisted during recovery (Figure 5E and Supplemental Figure 12). In summary, a failure to clear heat-induced read-through transcription from a transposon in nrpd2 mutants during recovery correlated with misregulation of an adjacent cluster of highly homologous auxin-responsive genes at the same stage of the heat-stress response.
The sorting of genes misregulated in nrpd2 mutants according to their ATG numbers revealed several cases where putative targets were part of the same gene family, clustered within the same genomic region, and regulated in a similar manner. We found four clusters of genes up- and eight clusters of genes down-regulated, respectively, during recovery of nrpd2 plants (Supplemental Table 3). Intriguingly, three of four clusters with up-regulated genes and six of eight clusters with down-regulated genes are associated with transposable elements, either in adjacent IGRs or within genes (data not shown). This might suggest a mechanistic role of transposable elements in the coordinate regulation of adjacent genes during stress conditions.
It has been reported recently that heat stress transiently affects epigenetic programs of plants. Specific long-term heat regimes triggered the alleviation of silencing of loci within constitutive heterochromatin (Lang-Mladek et al., 2010; Pecinka et al., 2010; Tittel-Elmer et al., 2010; Ito et al., 2011). This process is connected rather to heat-induced reduction of nucleosome occupancy than to the loss of DNA methylation and repressive chromatin marks. Resilencing occurs upon shifting plants back to normal temperatures and does not seem to depend on canonical epigenetic pathways.
On the other hand, stress responses in Arabidopsis require the integrity of epigenetic pathways. Pol IV, for example, is required for the generation of natsiRNAs that originate from and are involved in stress-dependent regulation of transcripts from overlapping gene pairs (Borsani et al., 2005). Components of histone deacetylase complexes play pivotal roles in freezing tolerance (Zhu et al., 2008; To et al., 2011a) and responses to salt stress (Chen et al., 2010).
In this work, we asked whether loss of epigenetic pathways affects tolerance of Arabidopsis to heat stress. Specifically, we tested mutants of DNA methyltransferases (drm1drm2; cmt3) (Bartee et al., 2001; Lindroth et al., 2001; Cao and Jacobsen, 2002), histone-modifying enzymes (hda6; kyp1; suvh2) (Aufsatz et al., 2002; Jackson et al., 2002; Naumann et al., 2005), chromatin remodelers (ddm1, mom1) (Jeddeloh et al., 1998; Mittelsten Scheid et al., 1998; Amedeo et al., 2000), and genes involved in small RNA biogenesis and function (nrpd1, nrpe1, nrpd2, dcl2, dcl3, dcl4, drd1, rdr2, rdr6, ago4) (Dalmay et al., 2000; Mourrain et al., 2000; Zilberman et al., 2003; Kanno et al., 2004; Xie et al., 2004; Herr et al., 2005; Kanno et al., 2005; Onodera et al., 2005; Pontier et al., 2005) for their tolerance to acute heat stress. The RdDM mutants nrpd2 and hda6 showed a strong and rdr2, dcl3, and ago4 a moderate reduction in survival rates. NRPD2 is the second-largest common subunit of Pol IV and Pol V involved in small RNA biogenesis and siRNA-dependent DNA methylation, respectively (Law and Jacobsen, 2010; Haag and Pikaard, 2011; Zhang and Zhu, 2011). HDA6, a histone deacetylase, functions as a repressor of several RdDM target genes (Aufsatz et al., 2002; Lippman et al., 2003; Probst et al., 2004; May et al., 2005; Earley et al., 2010), but clearly also has functions outside of RdDM (Tessadori et al., 2009; Chen et al., 2010; To et al., 2011a, 2011b; Zhu et al., 2011). Collectively, these results point towards an important function of the RdDM pathway in tolerance to heat stress.
Intriguingly, single knockouts of either Pol IV (nrpd1) or Pol V (nrpe1) had normal heat tolerance, suggesting an at least partly synergistic function of Pol IV and V in heat acclimation or pointing towards a specific requirement for Pol IV and Pol V at different loci (Borsani et al., 2005; Pontes et al., 2009). In line with a complex, locus-specific role of Pol IV and Pol V in regulating the transcriptional response to temperature stress, expression analysis of the NRPD2-dependent genes At1g0790 and At1g29450 in nrpd1 and nrpe1 mutants indicate a regulatory function of Pol IV in the transcription of At1g0790 and of Pol IV and Pol V in the expression of At1g29450 during recovery from heat.
ago4 mutants had a less drastic reduction in tolerance than either nrpd2 or hda6 mutants, which might be due to functional redundancy of AGO4 with AGO6 during RdDM (Zheng et al., 2007). dcl3 and rdr2 mutants, compromised in the production of RdDM-specific 24-nt siRNAs (Xie et al., 2004), were moderately affected, which might suggest a functional replacement of 24-nt siRNAs by siRNAs of other size classes. That siRNAs of different size classes can take over functions distinct from their normal roles has been described, for example with regard to inter-cellular silencing (Dunoyer et al., 2007).
ATH1 microarray transcriptome analysis of the two mutants with strongest hypersensitivity to acute heat stress, nrpd2 and hda6, under control, heat-stress conditions, and upon recovery clearly demonstrated that they are affected during different phases of the heat-stress response. GO analyses of misregulated gene sets showed that hda6 mutants exhibited a more general transcriptome dysregulation during acute heat-stress conditions, whereas the transcriptome profile during recovery did not differ extensively from wild-type plants. Conversely, nrpd2 mutants responded to acute heat stress similarly to wild-type plants, but were unable to exit the transcriptional stress program during recovery. There was little overlap between misregulated gene sets of both mutants within and across conditions, suggesting that the major function of HDA6 during the heat-stress response is not directly connected to RdDM.
All genes misregulated in recovered nrpd2 mutants that were further analyzed in this study were located adjacent to transposon remnants and/or siRNA producing clusters, suggesting that altered heat-responsiveness of protein-coding genes is brought about by a defective epigenetic regulation of nearby transposons in plants deficient in NRPD2. There are numerous examples of transposons affecting gene expression of protein-coding genes (Slotkin and Martienssen, 2007), but information about underlying mechanisms is scarce. Recently, it was shown that the retrotransposition of the COPIA-type retroelement ONSEN, which itself is heat-responsive, confers heat-responsiveness to genes close to the new insertion site, but the underlying mechanism was not elaborated (Ito et al., 2011).
We identified several possible mechanisms of how transposons can control gene expression during heat stress: the PPR locus At1g07590 was moderately heat-stress-responsive and has a repeat element with EnSpm DNA transposons consensus within its promoter. The repetitive element was methylated in wild-type plants under control conditions and during heat stress and recovery. In nrpd2 mutants, methylation was reduced already in non-stressed plants, and also during heat stress and recovery. However, demethylation of the repeat in nrpd2 mutants did not result in enhanced At1g07590 transcription under control conditions compared to control wild-type plants. Transcription during stress and recovery, however, was enhanced in nrpd2 mutants, indicating that demethylation of the repeat enhanced the heat-response capacity of At1g07590. Indeed, the repeat harbors a stress-responsive cis-element (ABRE), which might contribute to heat-dependent regulation of At1g07590 when demethylated. This is similar to mPing transposons in rice, which have stress-responsive cis-elements within control regions and confer stress-inducibility to adjacent genes (Naito et al., 2009).
As a potential mechanism of gene regulation by more distantly located transposon remnants, we detected heat-dependent read-through transcription from transposon remnants into the region of two loci misregulated during recovery of nrpd2 mutants. In the case of At1g34220, the read-through transcript correlated with gene up-regulation and, in the case of an auxin-responsive gene cluster, there was a correlation with a coordinated transcriptional down-regulation of homologous genes within the cluster. The transposon remnants that were the source of read-through transcripts lost DNA methylation in nrpd2 mutants already under control conditions, indicating that they are direct targets of NRPD2 and the RdDM pathway. The transposons associated with At1g34220 also exhibited reduction of DNA methylation in heat-stressed wild-type plants, demonstrating that the methylation status of certain loci is affected by stress. In contrast, methylation of the COPIA-like element associated with the auxin-responsive gene cluster remained unchanged across conditions in wild-type plants, which is in line with recent observations that heat stress can trigger transcriptional induction without reductions in repressive DNA methylation (Lang-Mladek et al., 2010; Pecinka et al., 2010; Tittel-Elmer et al., 2010). However, for both loci, read-through transcripts were induced to high levels in both wild-type and nrpd2 plants only upon heat stress. Thus, initiation of read-through transcription does not require NRPD2, and pre-stress demethylation of the transposons in nrpd2 plants is not sufficient to trigger efficient, high-level read-through transcription. The latter seems to require an additional stimulus brought about by heat stress, which is suggested to be a reduction in nucleosome occupancy (Lang-Mladek et al., 2010; Pecinka et al., 2010). Levels of read-through transcripts, however, were higher in nrpd2 plants than in wild-type plants. This is consistent with two recent reports showing that loss of epigenetic control enhances heat-dependent transcriptional induction of silent transgenes and endogenous transposons (Pecinka et al., 2010; Tittel-Elmer et al., 2010). Intriguingly, heat induction of the retrotransposon ONSEN was enhanced in several RdDM mutants (Ito et al., 2011). Importantly, the read-through transcripts disappeared in recovered wild-type plants, whereas they were still present at high levels in recovered nrpd2 plants, indicating that the attenuation of read-through transcription requires NRPD2. For both read-through loci, the methylation status of the transposon remnants in wild-type plants remained unchanged during recovery compared to heat stress. This excludes a causal role of DNA remethylation in the clearance of read-through transcripts during recovery and points towards a methylation-independent role of NRPD2 in this process. Whether siRNAs are involved remains to be determined.
How could read-through transcription from transposon remnants result in both up- and down-regulation of adjacent genes? At1g24220 is in sense orientation with regard to read-through transcripts. Thus, read-through from the adjacent transposon upon release from Pol IV/V-mediated silencing during heat stress could either result in higher mRNA levels per se or indirectly stimulate genic transcription by opening up the conformation of gene control regions. In support of the latter, promoter regions of At1g24220 lost DNA methylation during heat stress, when read-through transcription was initiated. In the case of the cluster of auxin-responsive genes, at least three homologous genes are coordinately down-regulated upon heat stress. Such negative co-regulation by a transposon-derived read-through transcript could be achieved by the formation of aberrant genic RNAs that are fueled into a posttranscriptional gene silencing (PTGS) pathway that erases homologous mRNAs within the cluster. Such a pathway would be in favor with regard to the observed rapid regain of transcriptional activity under recovery conditions when read-through transcription ceases, since PTGS of endogenous genes (in contrast to transgenes) is not necessarily associated with the acquisition of epigenetic marks on the DNA (Stam et al., 1998; Jones et al., 1999). In summary, we show that a successful heat response in Arabidopsis depends on the integrity of epigenetic pathways and provide evidence that heat-dependent gene expression is influenced by closely located transposon sequences. Such influence is expected to be even higher in crop plants that have larger genomes littered with remnants of transposable elements (Haberer et al., 2005; Sequencing Project International Rice, 2005; Paterson et al., 2009; Choulet et al., 2010).
The following Arabidopsis mutants and their corresponding wild-types were propagated simultaneously in the same climate chamber and used in this study: nrpd1a-4 (nrpd1), nrpe1 (SALK N517795), nrpd2a-1 (nrpd2), mom1-2, kyp1 (SALK 041474), suvh2 (SALK 079574), drd1-6, dcl3-1, dcl2-1, dcl4-1, rdr2-1, rdr6-1 (Col-0 background), ago4-1 (Ler/Col-0 background), ddm1-5 (Zh background), drm1/drm2, cmt3-11a (WS background), and rts1-1 (hda6) and nrpd2;G1072E (DT background). nrpd2;G1072E carries a C to T point mutation (Tair10) and was recovered in the lab of M. Matzke (unpublished). For basic heat-tolerance assays, seeds were germinated and grown for 8 d on 0.5 MS agar plates supplemented with 1% sucrose before transferring plants to soil (one plant/pot). Plants were grown at 21°C under long-day conditions (16/8 h day/night, 80 μmol m–2 s–1 light intensity) for 3 weeks (before bolting) and subsequently exposed to heat stress. Acute heat stress (42°C for 24–34 h, depending on mutant ecotype background) was administered in a PERCIVAL CU-36/L4 (USA). After heat exposure, plants were returned to growth chambers for recovery at 21°C and survival rate was assessed after 2 weeks.
For microarray analysis, leaf material was harvested from 3-week-old plants exposed to heat stress (42°C for 16 h) and after 2 d of recovery. Control plants were harvested in parallel. RNA was extracted from pools of 10 plants using TRIzol reagent (Sigma) followed by DNase treatment (QIAGEN) and purification with the RNasy Kit (QIAGEN). RNA concentration and quality were determined by nanodrop and 1% agarose gels. Hybridization was performed on the Affimetrix Gene Chip Arabidopsis ATH1 Genome Array (NASC). Data were obtained from three independent biological experiments. The probe-level intensities from the raw data were normalized and summarized with the standard method, GeneChip Robust Multi-array Averaging (GC-RMA) (Wu et al., 2004). For differential expression analysis, Limma (Linear model for microarray analysis), incorporating the empirical Bayes moderated t-test (Smyth, 2004) with linear modeling, was used. The Benjamini and Hochberg method was used for FDR (False Discovery Rate) multiple testing correction; the changes in expression with cut-off logFC ≥ 1 or ≤ –1 were called statistically significant if the adjusted p-value was <0.05. All statistical analyses were done using the Bioconductor solution (www.bioconductor.org) under the R platform (www.r-project.org). Microarray results were analyzed using the agriGO toolkit and database (http://bioinfo.cau.edu.cn/agriGO/). The epigenome database (Lister et al., 2008) was used to analyze target gene location, orientation, structure, and methylation status. The Repbase database was used for screening of repetitive elements (Kohany et al., 2006). Microarray data deposition on http://affymetrix.Arabidopsis.info/narrays/experimentpage.pl?experimentid=538.
For RT–PCR, 5 μg of total RNA was reverse-transcribed either with random or gene-specific primers using the Fermentas kit. For qRT–PCR, 1 μg of total RNA was reverse-transcribed using the qScript cDNA SuperMix (Quanta Biosciences) according to the manufacturer’s protocol. qPCR was performed on a Biorad iQ5 cycler with 25-μl reactions that were set up using the SensiMix Plus SYBR Kit and Fluorescein (Peqlab). Each PCR was performed in triplicate. Tubulin (At5g12250) was used as a reference gene. qRT–PCR data represent average values obtained from three independent biological experiments. Primer sequences for RT– and qRT–PCR are given in Supplemental Table 4.
Genomic DNA was extracted using the Phytopure DNA extraction kit (GE Healthcare) according to the manufacturer’s protocol. Genomic DNA linearization with BamHI and digestions with HpaII (SsiI), MspI, and HaeIII (DdeI) followed by PCR amplification were performed as described (To et al., 2011b). The primers used for Chop–PCR are listed in Supplemental Table 4. The number and contexts of the restriction sites are represented in Supplemental Figure 13.
Supplementary Data are available at Molecular Plant Online.
This work was supported by the Austrian Science Foundation (P22062-B16) and the GEN-AU program from the Austrian Federal Ministry of Science and Research.
We thank O. Mittelsten Scheid for cmt3, drm1/drm2, kyp1, suvh2, ddm1, and mom1 mutant lines, Z. Lorkovic for discussion and nrpd2:G1072E seeds, and M. Matzke for support and comments on the manuscript. No conflict of interest declared.