Heat Stress Enhances the Accumulation of Polyadenylated Mitochondrial Transcripts in Arabidopsis thaliana

Background Polyadenylation of RNA has a decisive influence on RNA stability. Depending on the organisms or subcellular compartment, it either enhances transcript stability or targets RNAs for degradation. In plant mitochondria, polyadenylation promotes RNA degradation, and polyadenylated mitochondrial transcripts are therefore widely considered to be rare and unstable. We followed up a surprising observation that a large number of mitochondrial transcripts are detectable in microarray experiments that used poly(A)-specific RNA probes, and that these transcript levels are significantly enhanced after heat treatment. Methodology/Principal Findings As the Columbia genome contains a complete set of mitochondrial genes, we had to identify polymorphisms to differentiate between nuclear and mitochondrial copies of a mitochondrial transcript. We found that the affected transcripts were uncapped transcripts of mitochondrial origin, which were polyadenylated at multiple sites within their 3′region. Heat-induced enhancement of these transcripts was quickly restored during a short recovery period. Conclusions/Significance Our results show that polyadenylated transcripts of mitochondrial origin are more stable than previously suggested, and that their steady-state levels can even be significantly enhanced under certain conditions. As many microarrays contain mitochondrial probes, due to the frequent transfer of mitochondrial genes into the genome, these effects need to be considered when interpreting microarray data.


Introduction
Polyadenylation of RNAs has a decisive role in the regulation of RNA stability. In eukaryotes, it confers stability for nuclear mRNA, regulates export of processed mRNAs to the cytoplasm and promotes translation initiation [1]; [2]. In bacteria, in contrast, polyadenylation facilitates RNA degradation by attracting the degradosome, a complex containing phosphorylase (PNPase) [3]. As for prokaryotes, polyadenylation of mRNAs in chloroplasts serves as a RNA degradation signal [4] and it promotes mRNA degradation in plant mitochondria [5]; [6]. Plant mitochondrial PNPase degrades rRNA and tRNA maturation by-products, but also removes highly transcribed non-functional RNAs and antisense transcripts, following their polyadenylation. Polyadenylation therefore appears to be part of a RNA turnover system that counterbalances relaxed transcription in plant mitochondria [7].
Polyadenylation has different consequences on the mitochondrial transcripts in different organisms. In human, for example, polyadenylation is required for stabilisation of mitochondrial mRNA [8]. The classical view that polyadenylation of nucleusencoded RNA is always associated with enhanced RNA stability has also been challenged by the recent discovery of a nuclear exosome system, that includes nuclear polyadenylation and degradation of RNAs transcribed from intergenic regions [9]. It has been suggested that ancient poly(A)-linked degradation functions have been maintained in nuclei, while poly(A) tails acquired a new role in the cytoplasm. It therefore appears that RNA polyadenylation has different signal functions in different RNA control systems, whose effects on the stability of the involved RNAs depend on the origin and structure of the RNA and on the enzyme machinery that is present in subcellular compartments.
The analysis of mitochondrial gene transcripts in plants is complicated by the frequent exchange of DNA between nuclear and mitochondrial genomes. The Arabidopsis thaliana ecotype Columbia-0 (Col), for example, contains a 618642 kb insert of mitochondrial origin (the 'numt' insertion) in the pericentric region of chromosome 2 [10]. The insert contains complete copies of the mitochondrial regions A, B, C and D, representing the 367 kb mitochondrial genome, and an additional internal part representing two copies of a complete D region and two copies of a 41 kb section of the A region.
Owing to the presence of the mitochondrial genome in Col, microarrays that cover the Arabidopsis genome contain probes designed to detect transcripts from the mitochondrial genes.
Considering that polyadenylated mitochondrial mRNAs are rare, as they are subject to degradation, one would assume that selective labelling of polyadenylated transcripts would prevent crosshybridisation of mitochondrial mRNAs in microarray experiments. Much to our surprise, however, when we analysed a group of microarray experiments, we found significant signals for a gene cluster on chromosome 2 that corresponds to the mitochondrial insertion, showing significant correlation with heat shock response genes and up-regulation in response to heat treatment. This observation prompted us to investigate the origin of the corresponding transcripts. We don't find any evidence for expression of the nuclear copies of mitochondrial inserts. Instead, we detect a pool of polyadenylated transcripts of mitochondrial origin that increases after heat treatment.

Results
Co-expression of a heat shock transcription factor gene and mitochondrial gene transcripts As part of a general study of gene co-expression and transcriptional regulation, genes showing co-expression with the heat shock transcription factor At2g26150 were investigated in the Arabidopsis co-expression tool (ACT). These genes were further clustered with the CliqueFinder tool, and this revealed two main clusters. The first contained a variety of heat and general stress response genes, as would have been expected. Intriguingly, the second comprised 20 genes from the numt insertion. Within this second cluster, transcript levels showed an average pair-wise Pearson correlation of 0.80, and an average correlation of 0.70 with the transcription factor. Full details of the ACT clique clustering analysis are available as supplementary material (Fig. S1). Heat induction of the transcript signals from these genes on the Affymetrix ATH1 microarray was then investigated in a 38uC heat stress time series experiment. Figure 1 shows the fold change in expression signal in heat stressed plants (shoot tissues) compared to controls over the time course, for the genes of the numt insertion. Many of the genes show a signal that peaks strongly at the 3h time point, subsequently decaying back to normal levels. Of the 53 genes within the insertion for which the ATH1 array has probes, 48 show statistically significant (as judged by a Benjamini-Hochberg corrected p value for differential expression of ,0.0001) fold changes at this time point.

Mitochondrial origin of a heat enhanced polyadenylated transcript
The high similarity between the mitochondrial genome and its genomic insertion makes it difficult to define the origin of these transcripts. We therefore used a T-DNA insertion line (SALK_143190) to tag a mitochondrial model gene in the chromosome 2 cluster, and to sequence the nuclear copy of the gene. We labelled the transcript Mito1 as it was unclear if it derived from the mitochondrial gene rpl2 or its nuclear insert At2g07715. Sequence comparison of the nuclear insert At2g07715 with the corresponding mitochondrial gene rpl2 identified three polymorphisms, one of which created a DraI site in At2g07715 that was absent in rpl2 (Fig 2A). An RT-PCR with primers that included the DraI sequence failed to amplify any product, while rpl2-specific primers amplified Mito1 specific transcripts (Fig 2C), suggesting a mitochondrial origin of Mito1 transcripts. This was confirmed when we cloned cDNAs representing the 39 part of the Mito1 transcript, both from material harvested at 24uC and 40uC. All clones matched the mitochondrial rpl2 transcript, and all except three resembled the unspliced version of the rpl2 transcript, with its editing site [11] unchanged. Most cDNAs had different polyadenylation sites, characteristic for the variable polyadenylation of  copy of rpl2, and its nuclear counterpart integrated into the Arabidopsis genome, At2g07715. Arrows depict exon ORFs, T-DNA localises the transgene insertion in SALK_143190 Nucleotides mark differences between the two sequences and the C.U editing site of rpl2. (B) Vertical bars mark the distribution of polyadenylation sites detected in cDNA copies of Mito1 transcripts isolated at 40uC or 24uC, respectively. Only four sites were found more than once. P1 and P2 show the localisation of nested primers used for the RT-PCR. Nucleotide positions are labelled with reference to the annotated 59 end at position 1. The 39 end of rpl2 is located at position 3793. [28] (C) RT-PCR primers including the DraI site of the nuclear Mito1 insertion At2g07715 (40 cycles) and with primers lacking the DraI site and specific for the mitochondrial Mito1 gene rpl2 (32 cycles). Both at 24uC and at 40uC, only rpl2 specific transcripts could be amplified. EF1a was used to calibrate the amount of cDNA. Genomic wild type DNA was used as positive control (+); (-) indicates control reactions at 40uC without Reverse Transcriptase treatment. (D) RT-PCR analysis of the Mito1 transcript after sequential treatment with Alkaline Phosphatase (AP), Tobacco Acid Pyrophosphatase (TAP) and Terminator exonuclease (Ter). Lack of signal reduction, compared to the untreated sample (U) after combined treatment with AP, TAP and Ter (CAP) indicates absence of 59CAP structure. Sensitivity of the transcript to Ter treatment (P) suggests a 59phosphate structure for the Mito1 transcript. The Elongation Factor 1 (EF1a) transcript was assayed as a control. doi:10.1371/journal.pone.0002889.g002 mitochondrial transcripts (Fig. 2B, Fig.S2). Nuclease treatments suggested that the polyadenylated rpl2 transcripts did not contain a CAP structure (Fig 2D), as expected for a mitochondrial transcript. All results therefore suggest that the Mito1 transcripts represent polyadenylated transcripts of the mitochondrial rpl2 gene.
The heat-specific increase mainly affects polyadenylated rpl2 transcripts To differentiate between total and polyadenylated transcripts, we used random primers or oligo dT primers, respectively, to compare rpl2 transcript levels with and without heat treatment (Fig. 3). In randomly primed RNA pools, rpl2 transcript levels increased about 2-fold after heat treatment in total RNA preparations. A similar tendency was observed, although at a more pronounced level, for polyadenylated RNA pools where heat treatment resulted in a 30fold increase of rpl2 transcripts in total RNA preparation. This suggests that the heat-specific enhancement of rpl2 transcripts predominantly affects polyadenylated transcripts.

Recovery from heat-induced transcript enhancement
To assess if the accumulation of rpl2 transcripts was reversible, we measured transcript levels during and after two heat treatment sessions (Fig. 4). A maximum increase level was reached within one hour of heat treatment. After two hours of a recovery period rpl2 transcripts levels had almost reached pre-treatment levels. Very similar increase and recovery effects were detectable during a second heat phase. The data suggest that the temperature effect is reproducible over successive treatment phases, and that transcript levels quickly revert to normal during the recovery phase. The quick reversion of rpl2 transcripts levels after the end of the heat treatment implies that polyadenylated rpl2 transcripts are quickly degraded once their supply is reduced.

Turnover rates of rpl2 transcripts
To assess if the increased temperature affects the stability of polyadenylated rpl2 transcripts, their level was measured at 24uC and 40uC in the presence of 0.6mM cordycepin [12], which  inhibits transcript synthesis. Transcript levels were normalised to EF1a. Since EF1a is also affected by the treatment, the calculations performed in this way actually show whether the turnover of the tested gene is faster or slower than the turnover of EF1a. Over a period of four hours, rpl2 transcript turnover rates were comparable to EF1a and no changes were detectable for the two temperature treatments with respect to rpl2 transcript turnover (Fig. 5), which suggests that heat treatment does not significantly alter the turnover rate of polyadenylated rpl2 transcripts.

Heat-specific increases in other mitochondrial transcripts
To test if the effects that we observed for the Mito1 transcript equally apply to other mitochondrial transcripts, we tested the expression of three other genes, Mito2-4, in semi-quantitative RT-PCR experiments (Fig 6). For all three genes, we monitored a similar heat response as already observed for rpl2 transcripts, which suggests that the effects that we detect for rpl2 transcripts equally apply for a number of other mitochondrial transcripts. We also cloned polyadenylation sites of Mito3 and Mito4 transcripts (Fig. S3) and observed a very similar variability in polyadenylation as we had previously seen in rpl2 transcripts

Discussion
Considering that polyadenylation of mitochondrial transcripts is part of a degradation system, it was surprising to detect positive hybridisation results for so many mitochondrial genes in microarray experiments. It was even more surprising that for a large number of mitochondrial genes, polyadenylated transcripts became more abundant after moderate heat treatment. During evolution, mitochondrial genes have been integrated into nuclear genomes, where they may have developed into functional nuclear copies or from where they may have been lost again [13]. Functional mitochondrial gene transfer into the nucleus is a prominent feature in plants [14] and green algae [15].
Most microarray experiments use the Columbia-0 ecotype as experimental material, which represents an unusual example for a mitochondrial gene transfer event as it contains a 618642 kb insert in the pericentric region of chromosome 2 [10]. This insert  represents the 367 kb mitochondrial genome with complete copies of the mitochondrial regions A, B, C and D, and an additional internal part that contains two copies of a complete D region and two copies of a 41 kb section of the A region. Comparison of a 262 kb insert of the Arabidopsis nuclear mitochondrial DNA (numtDNA) with the mitochondrial DNA (mtDNA) region, revealed a 99.91% identity of the two genomes, and suggested that the mitochondrial DNA transfer into the nucleus occurred about 88,000 (44,000-176,000) years ago [16]. It is highly unlikely that the genes located on the numtDNA have undergone sufficient adaptation that would allow them to be expressed under the control of the nuclear compartment. Such adaptation would require the development of promoter sequences that allowed transcription by nuclear polymerases, and the establishment of 39signals that could be recognised by nuclear polyadenylation functions. It was therefore concluded that very little, if any, transcription occurs from numt genes [16].
Our analysis of a model transcript, Mito1, which was selected because it allowed us to differentiate between transcripts derived from the mtDNA and the numtDNA, confirms the assumption that the numt gene is not transcribed. The four lines of evidence that point towards a mitochondrial origin of the polyadenylated Mito1 transcripts are its sequence identity with the mitochondrial gene, the failure to amplify nuclear transcripts, the lack of a CAP structure and the presence of multiple polyadenylation sites, a feature characteristic for mitochondrial transcripts [6]. Enhanced polyA transcript levels quickly revert back to normal levels when the temperature is reduced to 24uC, which suggests that the heat treatment does not cause lasting damage This is also supported by reports about Arabidopsis cell cultures being protected from heatinduced cell death at 37uC and only becoming susceptible at 50uC [17]. We therefore have no indications for apoptotic effects from heat stress, which would have been surprising as plants frequently experience periods of high temperatures. A treatment with cordycepin, which inhibits transcription of mitochondrial DNA [18], does not reveal any temperature-specific differences in the turnover rates for the polyadenylated transcripts of the model transcript. This makes it less likely that the heat-specific enhancement in polyadenylated transcripts is the consequence of a reduced efficiency of transcript turnover. We must, however, be careful in our interpretation of cordycepin effects, because cordycepin has been shown to inhibit polyadenylation in some systems [19], and because we could only use a nuclear transcript to confirm the efficiency of the cordycepin treatment. The small but significant increase in randomly primed Mito1 transcript levels at 40uC suggests that heat stimulates mitochondrial gene transcription. Considering the much larger increase in polyadenylated transcripts, it is tempting to speculate that heat treatment not only increases transcription but also the level of faulty transcript synthesis that leads to enhanced polyadenylation activity.
Three other mitochondrial transcripts show a similar heatdependent increase of their polyadenylated transcripts, which suggests that the effect is not limited to a single gene but that it affects a number of mitochondrial transcripts. The practical consequence of this effect is that we have to be careful in interpreting the origin of poly(A)-specific transcripts in profiling experiments. In a recent study on chromosome 2 specific transcript variants, in which RNA had been prepared from a variety of tissues subjected to different treatments, including heat, a significant number of transcripts were cloned from mitochondrial insertion genes [20]. Exclusion of 59phosphorylated transcripts by Terminator nuclease would help to differentiate between nuclear transcripts and polyadenylated transcripts of mitochondrial origin.

Microarray data analysis
The Arabidopsis co-expression tool (ACT) [21] [22] was used for co-expression analysis of relevant genes over a large database of microarray experiments. The CliqueFinder tool within this software was used to identify groups of genes with mutually high levels of co-expression. The heat stress time series from AtGenExpress (Nover et al. Gene Expression Omnibus (GEO) Accession GSE5628) and the corresponding control time series (Townsend et al. GEO accession GSE5620) were used in detailed investigation of heat induction. These data sets were downloaded from GEO, expression summaries were obtained with the RMA algorithm [23], and differential expression analysis was carried out using linear models in the LIMMA package [24], implemented in the R statistical programming environment.

Genetic material
A SALK insertion line (ecotype Col-0) with a T-DNA insertion [25] in the coding sequence of At2g07715 (SALK_143190) was obtained from the Nottingham Arabidopsis Stock Center [26]. Plants carrying the insertion were identified using primers 59-GGAGAACCTGCGTGCAATCC-39 and 59-GCCCTAAAG-TATCTTGCCGC-39. The positive plants were selfed and the homozygosity status of the insertion allele was tested by screening its segregation in the progeny.
To initiate reverse transcription, random hexamers were used for random priming and a oligodT primer 59 GGCCACGCGTC-GACTAGTACTTTTTTTTTTTTTTTTT-39 for polyA-specific transcript priming. The qPCR products were quantified according to the 2(-Delta Delta C(T)) method [27].
Poly(A) analysis (39 RACE) 5 mg of total RNA from 10-day-old seedlings was reverse transcribed using Superscript II reverse transcriptase (Invitrogen) according to the manufacturer's instructions using the primer 59-GGCCACGCGTCGACTAGTACTTTTTTTTTTTTTTTTT-39. The 39 end of the rpl2 transcripts was amplified by semi-nested PCR, using primer 59-GGCCACGCGTCGACTAGTAC-39 with primer 59-GGAGAACCTGCGTGCAATCC-39 for the primary amplification and with primer 59-AGCATTCTCTGGGCACA-TAGG-39 for the secondary amplification. The PCR product was cloned into the TOPO-TA Vector (Invitrogen) according to the manufacturer's instructions and sequenced. Similarly, the 39 end of Mito3 was amplified using primer 59-GGCCACGCGTCGAC-TAGTAC-39 with primer 59-GATAATCAATTCGGTCGTT-GTGG-39 for the primary amplification and with primer 59-GGATTTCTGACCACATTCTCC-39 for the secondary amplification. The 39 end of Mito4 was amplified using primer 59-GGCCACGCGTCGACTAGTAC-39 with primer 59-CAGGA-GAATGAGTCCCGCCTCGA-39 for the primary amplification and with primer 59-AGATCGGTCGAGTGGTCTCAG-39 for the secondary amplification.

CAP analysis
15 mg of total RNA obtained from 10-day-old seedlings was split in three aliquots. An aliquot was left untreated (control). An aliquot was treated with Terminator enzyme (EpicentreH Biotechnologies) according to the manufacturer's instructions to selectively digest 59monophosphate transcripts. The last aliquot was incubated at 37uC for 30 minutes with Calf Intestinal Alkaline Phosphatase (Promega) in the provided buffer to remove 59 phosphate groups from transcripts. The RNA was then extracted with phenol:chloroform and precipitated with ethanol. The CAP structure was removed from the RNA with Tobacco Acid Pyrophosphatase (EpicentreH Biotechnologies) according to manufacturer's instructions. After extraction with phenol:chloroform and precipitation the RNA was finally treated with Terminator enzyme (EpicentreH Biotechnologies) according to manufacturer's instructions. The three aliquots were reverse transcribed with Superscript II reverse transcriptase (Invitrogen) using an oligo-dT primer. Figure S1 ACT clique clustering analysis that identifies genes showing co-expression with the heat shock transcription factor At2g26150.