Unstable Transcripts in Arabidopsis Allotetraploids Are Associated with Nonadditive Gene Expression in Response to Abiotic and Biotic Stresses

Genome-wide analysis has documented differential gene expression between closely related species in plants and animals and nonadditive gene expression in hybrids and allopolyploids compared to the parents. In Arabidopsis, 15–43% of genes are expressed differently between the related species, Arabidopsis thaliana and Arabidopsis arenosa, the majority of which are nonadditively expressed (differently from mid-parent value) in allotetraploids. Nonadditive gene expression can be caused by transcriptional regulation through chromatin modifications, but the role of posttranscriptional regulation in nonadditive gene expression is largely unknown. Here we reported genome-wide analysis of mRNA decay in resynthesized Arabidopsis allotetraploids. Among ∼26,000 annotated genes, over 1% of gene transcripts showed rapid decay with an estimated half-life of less than 60 minutes, and they are called allotetraploid genes with unstable transcripts (AlloGUTs). Remarkably, 30% of alloGUTs matched the nonadditively expressed genes, and their expression levels were negatively correlated with the decay rate. Compared to all genes, these nonadditively expressed alloGUTs were overrepresented 2-6-fold in the Gene Ontology (GOSlim) classifications in response to abiotic and biotic stresses, signal transduction, and transcription. Interestingly, the AlloGUTs include transcription factor genes that are highly inducible under stress conditions and circadian clock regulators that regulate growth in A. thaliana. These data suggest a role of mRNA stability in homoeologous gene expression in Arabidopsis allopolyploids. The enrichment of nonadditively expressed genes in stress-related pathways were commonly observed in Arabidopsis and other allopolyploids such as wheat and cotton, which may suggest a role for stress-mediated growth vigor in hybrids and allopolyploids.


Introduction
Polyploidy (whole genome duplication) is common in many plants and some animals [1,2,3]. The polyploid plants may represent over 70% of angiosperms [4,5], and .75% of which are allopolyploids [1,6,7]. Most important crops have a polyploidy origin. Wheat, cotton and canola are allopolyploid that consist of two or more divergent genomes [3], and some plants and animals exist as interspecific hybrids [8,9]. The common occurrence of plant polyploids indicates an advantage of being allopolyploids in response to selection and adaptation [6,10,11,12]. Stable allopolyploids with meiotic transmission of homoeologous genomes lead to permanent fixation of heteozygosity and hybrid vigor.
Stable Arabidopsis allotetraploids (Arabidopsis suecica-like) are resynthesized by crossing tetraploid Arabidopsis thaliana with Arabidopsis arenosa [13]. The resynthesized A. suecica allotetraploids show morphological and physiological differences from their progenitors [13,14,15]. Microarray analysis of gene expression indicated that 5-38% of the genes are nonadditively expressed (different from mid-parent value, MPV) in resynthesized allotetraploids [14]. Nonadditive expression suggests activation or repression of a gene through genetic and epigenetic mechanisms [16,17,18]. Gene activation or silencing in allotetraploids can be caused by transcriptional regulation through changes in chromatin structure. However, many expression assays including microarrays and RNA sequencing measure the steady levels of mRNA transcripts. It is unknown whether mRNA stability affects nonadditive expression of genes between allotetraploids and their progenitors, A. thaliana and A. arenosa.
Posttranscriptional regulation of mRNA plays an important role in gene expression through modulation of mRNA stability and turnover. For example, c-myc and c-fos transcripts in mammalian cells and mating-type transcripts in yeast are unstable [19,20,21]. The decay rate of mRNA is correlated with biological functions of the genes and the length of cell cycles in mammals and yeast [22,23,24]. There is a role for mRNA stability in shaping the kinetics of the transcriptome in response to environmental changes in yeast and mammals [25,26]. Approximately 50% of stress responsive genes in human cells are controlled at the level of mRNA decay [27]. In yeast, mRNA turnover is a major mediator in the response to heat shock, nutrition deprivation, oxidative stress, and osmotic stress [22,28,29].
Genome-wide analysis of mRNA decay was studied in Arabidopsis [30,31]. It is shown approximately 1% of Arabidopsis thaliana genes with unstable transcripts (AtGUTs) [30]. These AtGUTs include the genes that are induced by mechanical stimulation and circadian rhythms. Moreover, transcripts from intronless genes and microRNA target genes generally have short half-lives [31]. Unstable transcripts are also identified in photolabile phytochrome genes [32] and auxin inducible genes [33,34], which facilitates rapid turnover of mRNA, leading to tight and effective control of gene expression [35]. To test a role of mRNA stability in nonadditive gene expression in allotetraploids, we examined mRNA turnover rates in resynthesized A. suecica allotetraploids relative to the mid-parent value using oligo-gene microarrays [14]. The candidates of allotetraploid genes with unstable transcripts (AlloGUTs) were compared with the genes that are nonadditively expressed, as identified in a previous study [14]. We further analyzed Gene Ontology (GOSlim) classifications and compared alloGUTs with stress responsive genes in published microarray datasets. Finally, we validated expression for a subset of genes using qRT-PCR analysis. These comprehensive analyses revealed mRNA instability of the genes related to stresses and clock-mediated pathways, suggesting a role for rapid mRNA decay in regulating stress and circadian responses to growth vigor in allopolyploids.

Results
Genome-wide analysis of unstable transcripts in allotetraploids using oligo-gene microarrays Spotted oligo-gene microarrays were used to evaluate mRNA turnover in allotetraploids relative to the MPV ( Figure 1A). The 70-mer oligos were designed from annotated genes and could crosshybridize with A. thaliana and A. arenosa genes. This may underestimate changes in allotetraploids in which transcripts from one homoeologous locus display rapid decay, whereas the transcripts from the other locus are relatively stable. After data normalization and statistical tests (see Methods), the qualified values with reproducibly normalized intensity ratios of $2 were used for comparative analysis of gene expression. Approximately ,1% (208) of genes in the Arabidopsis allotetraploid (Allo733) were identified as unstable transcripts with a half-life time of less than 60 min using the same criteria as previously defined for unstable transcripts [30]. These are hereafter referred to as allotetraploid genes with unstable transcripts (AlloGUTs) ( Table S1). The percentage of AlloGUTs identified in allotetraploids is consistent with ,1% of AtGUTs identified in A. thaliana using cDNA microarrays [30]. To confirm the microarray data, a subset of 10 genes was validated using quantitative RT-PCR (qRT-PCR) analysis, and expression patterns of four AlloGUTs were determined ( Figure 1B, C). The four genes are: At5g26030 (ATFC1) encoding ferrochelatase 1, At4g17500 (ERF1) encoding ethylene responsive element binding factor 1, At1g27730 (ZAT10) encoding salt tolerance zinc finger, and At4g11280 (ACS6) encoding 1-aminocyclopropane-1-carboxylic acid (ACC) synthase 6. In qRT-PCR analysis, the gene encoding translation initiation factor 4A2 (eIF4A2, At1g54270) was used as a control because its transcripts were reasonably stable and did not depredate during the time course of chemical treatments [30]. The estimated turnover rate was 24-40 minutes by qRT-PCR ( Figure 1C), which was comparable to 28-47 minutes estimated from microarray data ( Figure 1B). A total of ,208 AlloGUTs and ,95 AtGUTs were classified into 20 GOSlim groups ( Figure 2) using TAIR release 9 (November 2009) and PEDENT (http://mips.gsf.de/proj/thal/db/index. html). The proteins encoded by AlloGUTs were distributed among all GOSlim biological classifications. Compared with the proportion of the whole-genome annotated genes, AlloGUTs were overrepresented by a factor of two in the groups of cell cycle and DNA processing, transcription, regulation of metabolism and protein, and cellular communication and biogenesis of cellular component ( Figure 2). In particular, the genes with unstable transcripts were enriched in four groups, namely, cell rescue, defense and virulence, interaction with environment, and systemic interaction with environment. They had significantly (2.6-3.2-fold) higher proportions of GUTs than that of all genes (dashed lines) [30] (Figure 2).

Rapid mRNA turnover and nonadditive gene expression in allotetraploids
Unstable transcripts are associated with mRNA stability and possibly steady levels of transcripts in allotetraploids. To test this, we compared the gene list of nonadditive expression identified in a previous study [14] with that of AtGUTs in A. thaliana diploid [30] and AlloGUTs in this study. Sixty-one of 208 (,30%) AlloGUTs matched the nonadditively expressed genes in both allotetraploids (Allo733 and Allo738) (Table 1, Figure 3A). The correlation was statistically significant (degree of freedom = 1, Chi-square = 460.945, and p#0.01) (Table S2). Hierarchical cluster analysis suggested a strong negative correlation between the expression levels of nonadditively expressed genes and mRNA decay rates of AlloGUTs ( Figure 3B). The data suggest that these nonadditive expressions of the genes in the allotetraploids tested are related to rapid mRNA turnover. We also examined an enrichment of these genes in biological processes compared with the whole genome annotation. As in the nonaddi-tively expressed genes [14,36], the nonadditively expressed AlloGUTs were enriched 2-8 fold in GOSlim groups of stress response pathways, followed by response to abiotic or biotic stimulus, signal transduction, and transcription ( Figure S1).

A relationship between nonadditively expressed AlloGUTs and abiotic stress and phytohormone responses
To test how nonadditively expressed AlloGUTs are affected by physiological and environmental conditions, we performed hierarchical cluster analysis with the Genevastor program [37,38], using the genes extracted from multiple expression datasets, including global stress [38] and response to hormones and hormone-like chemicals [37]. The comparative data analysis indicated that nonadditively expressed AlloGUTs matched the genes whose expression was highly induced by cold, drought, osmotic, oxidative, salt and wounding ( Figure 4). In addition, the nonadditively expressed AlloGUTs were also highly induced by ozone stress ( Figure 5) and tend to alter expression levels in response to the treatments of phytohormones and plant hormone substances, ABA and Methyl Jasmonic acid, cyclohexamide and AgNO 3 ( Figure S2).
Interestingly, under normal growth conditions these stress responsive genes had a fast decay rate, and nonadditively expressed genes were down-regulated in the allotetraploids. This suggests stress-related genes are temporally repressed in allotetraploids under normal conditions and may be inducible under stress conditions.

Nonadditively expressed AlloGUTs also show a nonadditive mRNA decay rate
To test whether nonadditively expressed AlloGUTs are subjected to nonadditive mRNA decay in the allotetraploids, we measured  the difference in mRNA decay rates between allotetraploids and their progenitors, A. thaliana and A. arenosa, using qRT-PCR ( Figure 6). Among 61 genes tested, 14 (,25%) of nonadditively expressed AlloGUTs were significantly more unstable than the MPV (Figure 6, Table1). The majority of 14 genes tested, except for At3g44260, had a shorter half-life time in the allotetraploid than MPV ( Figure 6). According to functional classification by PEDENT, eight genes (,57%) (At1g27730, At1g68840, At2g28200, At3g15210, At3g15500, At3g19580, At3g44260, and At4g17500) were predicted with a role in transcription. For example, AtNAC3 (At3g15500) is one of the nonadditively expressed AlloGUTs, and its expression is sensitive in response to a wide range of stresses and phytohormones. This transcript had a much shorter half-life in the allotetraploid than in the progenitors. Furthermore, hierarchical cluster analysis indicated that nonadditively expressed AlloGUTs are highly correlated with the genes in response to abiotic stress and phytohormones. The data suggest that nonadditive gene expression in the allotetraploids is associated with the nonadditive rate of mRNA decay and stress and phytohormone responses.

Different decay rates of circadian clock gene transcripts in allotetraploids
Nonadditively expressed genes in allotetraploids include those encoding circadian clock regulators, including CIRCADIAN CLOCK ASSOCIATED 1 (CCA1), LATE ELONGATED HYPOCOTYL (LHY), TIMING OF CAB EXPRESSION 1 (TOC1), and GIGANTEA (GI) [14,39,40,41,42]. The regulation of CCA1 transcripts by light and its mRNA instability play a role for accurately entraining the oscillator in response to environmental changes [43]. Indeed, we found CCA1 binding motifs and evening elements within 1000-bp upstream region of nonadditively expressed AlloGUTs. Among 14 AlloGUTs, 7 genes (50%) each contains at least one CCA1 binding site (CBS: AAAAATCT) or evening element [(AA)AATATCT] [39,44], suggesting that these genes are likely the targets of CCA1 and LHY. We measured the stability of CCA1, LHY, TOC1, and GI mRNA in allotetraploid and its progenitors (Figure 7). Transcripts of CCA1 and LHY were relatively unstable, as previously reported [43], whereas the transcripts of their reciprocal regulators, TOC1 and GI, were relatively stable. The decay rate of CCA1 mRNA in the allotetraploid is relatively faster than MPV, and LHY mRNA decreased faster in the allotetraploid than in one of the progenitors. These two nonadditively expressed clock regulator genes also showed a nonadditive rate of mRNA decay.

Posttranscriptional regulation of cell cycle and stressrelated genes in Arabidopsis allotetraploids
Transcriptome divergence between allotetraploids and their parents are controlled by transcriptional and/or post-transcriptional regulation [16,18]. At the transcriptional level, chromatin modifications and DNA methylation mediate activation and silencing of homoeologous genes in allotetraploids or interspecific hybrids [14,15,45,46,47]. Changes in transcript abundance can also be regulated at the posttranscriptional level. In synthesized wheat allotetraploids, activation of retrotransposon affects the expression of neighboring genes [48]. RNA-mediated gene expression may be associated with the equilibrium between mRNA accumulation and stability during growth, development, and physiological transitions [49,50,51,52,53]. Here we provided the first evidence for mRNA instability study in nonadditive gene expression in Arabidopsis allotetraploids. Our data indicated that the genes with unstable transcripts are correlated with nonadditive levels of steady mRNAs in the allotetraploids. Therefore, a proportion of nonadditively expressed genes in the allotetraploids is caused by unstable transcripts.
The relatively small number of alloGUTs identified in the study can be due to several factors. One possibility is that unstable mRNAs have low steady state levels, which may not be detectable by microarrays. For example, circadian clock genes such as CCA1 and LHY have unstable transcripts [43] but were not identified by microarrays. Alternatively, cross-hybridization of the probes between species could be another factor that have under-estimated the number of genes with unstable transcripts. One might expect that additional GUTs will be identified using high-resolution nextgeneration sequencing technology (RNA-seq) [54]. Alternatively, the noise associated with quantification of mRNA stability may also account for some differences observed.
GOSlim classifications show that like AtGUTs, AlloGUTs are found in all plant cellular processes. The proportion of AlloGUTs is enriched 2-fold or more than that of all genes in several GO groups, including transcription, cell cycle and DNA processing, mRNA Instability in Arabidopsis Allopolyploids cellular communication, and biogenesis of cellular components. The majority of these genes are involved in cell cycle regulation and DNA metabolism, a process that requires rapid and cyclic regulation. For example, in yeast, cell cycle and mating-type genes have unstable transcripts [22,24,55,56]. Significantly, AlloGUTs were enriched 3-fold more transcripts that are related to exogenous stimulus such as cell rescue, defense and virulence and interaction with the environments (Figure 2), suggesting that these genes are regulated by the rate of mRNA decay. Rapid decay of mRNA is a mechanism for fast response to environmental and external signals such as biotic and abiotic stresses [30,31]. The data are consistent with high retention rates of duplicate genes and high expression divergence between duplicate genes than singlecopy genes in the process of polyploidization [36,57]. Nonadditive expression of stress-related genes is also overrepresented in other allopolyploids such as wheat [58,59] and cotton [60,61,62]. This may also suggest that AlloGUTs are involved in these processes to facilitate faster mRNA adjustment in allotetraploids and hybrids than in diploids [21,24,30,35]. During evolution, rapid mRNA turnover of the transcripts made from additional genes in allopolyploids is likely to facilitate selection and adaptation for environmental niches [6].
AlloGUTs are associated with nonadditive gene expression in allotetraploids in response to stresses and phytohormones A significant finding is that , 30% of AlloGUTs overlap with the nonadditively expressed genes in allotetraploids [14] ( Figure 3A). Moreover, there is a statistically significant negative correlation between the expression levels of AlloGUTs and nonadditively expressed genes in the allotetraploids. The unstable mRNA transcripts tend to be nonadditively repressed in allopolyploids, suggesting a role of mRNA instability in nonadditive gene expression in allopolyploids. In other words, different transcriptome profiles of nonadditively regulated genes can be generated by different sensitivities of the mRNA degradation rate during allopolyploidization.
The AlloGUTs that overlap with nonadditively expressed genes are enriched 2-7-fold in GOSlim groups in response to stress, abiotic or biotic stimulus, signal transduction and transcription. There is also a 1.5-fold enrichment of the GO groups in transcription, developmental processes, other cellular processes and other metabolic processes. Compared to additively expressed AlloGUTs, nonadditively expressed AlloGUTs are significantly correlated with the genes in GOslim groups in response to cold, drought, osmotic, salt, and wounding stress. In response to environmental stimuli, an increase or decrease in mRNA abundance might be attained either by adjusting the rate of transcription or the rate of degradation, as demonstrated in many studies [25]. In addition, nonadditively expressed AlloGUTs are correlated with the genes induced by Methyl jasmonic acid and plant hormone substances, such as AgNo3, cycloheximide, and 2,3,5-triiodobenzoic acid. In cotton allopolyploids, expression ratios of AdhA homeologs vary considerably during seedling and fruit development [62]. Homeologous gene expression is altered by abiotic stress treatments, including cold, dark, and water submersion. These data collectively suggest that nonadditive gene expression and unstable transcripts provide a general mechanism in response to external signals such as biotic and abiotic stress as well as to internal signals such as phytohormones and intergenomic interactions.
Nonadditively expressed AlloGUTs also show nonadditive rate of mRNA decay in both parents. Among them, 25% of AlloGUTs have a faster decay rate in the allotetraploid than in the parents, 57% of which encode transcription factors. One gene, AtNAC3 (At3g15500), encodes one member of the NAC transcription factor family. The expression of AtNAC3 is highly induced by drought, high salinity, and abscisic acid [63]. Genome-wide analysis of AtNAC3-overexpressing transgenic plants showed a significant increase in drought tolerance, which was correlated with upregulation of several stress-inducible genes. The transgenic plants overexpressing AtNAC3 also show significantly higher survival rate (100%) under drought stress [63]. Another gene, AtERF1 (At4g1750), encodes EHTYLENE RESPONSE FAC-TOR1, a transcription factor, which controls the expression of pathogen responsive genes and prevents disease progression.  The analysis was carried out using publicly available microarray database (http://www.arabidopsis.org/portals/expression/ microarray/ATGenExpress.jsp) and Genevastor. Each row represents log2 expression values of the genes in a given microarray experiment, and each column represents a gene. The color represents relative expression levels of the genes (from 23 green to +3 red). doi:10.1371/journal.pone.0024251.g004 Overexpressing AtERF1 in the transgenic plants increases resistance to pathogens [64], as well as complements defense response defects of coronative insensitive 1 (coi1) and ethylene insentive2 (ein2) mutants [65]. Interestingly, the transgenic plants overexpressing AtERF1 are significantly smaller than wild-type plants.
The above data are consistent with the observation that many stress responsive genes are generally repressed in the allotetraploids under normal conditions [14]. Allotetraploids might have developed a mechanism that requires low maintenance at the transcript level for these genes in response to environmental changes. Under stress conditions, adjusting the ratio of unstable and stable mRNA transcripts may help rapidly facilitate the proper mRNA levels in response to stress. Whether or not the regulation of stress responsive genes is correlated with increased growth vigor in allopolyploids remains to be investigated.

Circadian clock genes produce unstable transcripts in allotetraploids and their progenitors
Plants are sessile and need to prepare for and respond to all environmental changes including biotic and abiotic stresses. Recent studies suggested that 10-90% genes in Arabidopsis are governed by circadian clock regulation [40,44,66,67]. CCA1 and LHY are members of a small family of single Myb domain transcription factors and have redundant but non-overlapping functions. CCA1 and LHY bind to the promoter of TOC1 and downstream genes in photosynthesis and starch metabolism [39,44]. CCA1 expression is negatively regulated by TOC1 and CHE [42,68]. Accurate entrainment of circadian rhythms is controlled by a combination of CCA1 transcripts level and mRNA degradation by light [43]. We found that CCA1 and LHY transcripts are unstable and show a relatively high rate of mRNA turnover. Compared to CCA1 and LHY, TOC1 and GI transcripts are relatively stable. Interestingly, these clock genes are nonadditively expressed in allotetraploids [14,39], and their expression levels are negatively correlated with their decay rates (this study). Specifically, CCA1 and LHY mRNA has a relatively lower level of stability in the allotetraploid than in both parents. The data suggest in addition to transcriptional regulation, some circadian clock genes such as CCA1 and LHY are regulated through mRNA decay at the posttranscriptional level. Together, the available data indicate that transcriptional and posttranscriptional regulation of circadian clock genes, stress responsive genes, and many other GUTs provides a general mechanism for rapid changes in gene expression induced by ''genome shock'' [69] in response to interspecific hybridization and allopolyploidization. Posttranscriptional regulation of mRNA transcripts offers a unique mechanism to discriminate the transcripts between homoeologous loci in the allopolyploids. It will be interesting to investigate how these homoeologous transcripts are discriminated and degraded.

Conclusions
The current data support a role of mRNA instability in nonadditive expression of genes in Arabidopsis allopolyploids. We found ,1% of genes with allotetraploids unstable transcripts (AlloGUTs) in Arabidopsis and a half-life time of less than 60 minutes, ,30% of which matched the nonaddtively expressed genes that are repressed in the allotetraploids. All AlloGUTs and Figure 5. Hierarchical cluster analysis showing the AlloGUTs that matched the genes whose expression is induced by ozone stress. The analysis was carried out with publicly available microarray database (http://www.arabidopsis.org/portals/expression/microarray/ ATGenExpress.jsp and Genevastor). Each row represents log2 expression values of the genes in a given microarray experiment, and each column represents a gene. The color represents relative expression levels of the genes (from 23 green to +3 red). doi:10.1371/journal.pone.0024251.g005 nonadditively repressed AlloGUTs are significantly enriched with GOSlim classifications of response to stress, environment, and phytohormones. These AlloGUTs overlap with the genes that are highly induced by abiotic stress, including cold, drought, osmotic, salt, and wounding, and plant hormones and substances, including Methyl jasmonic acid and cycloheximide. Expression of circadian clock genes such as CCA1 and LHY is controlled by transcriptional regulation as well as posttranscriptional regulation through mRNA decay. The data collectively suggest that transcriptional and posttranscriptional regulation affects expression of homoeologous genes related to stress response, light signaling, and hormone interactions in allopolyploids, which may provide a flexible and rapid response to external and internal stimuli as a consequence of polyploidization.
The plants were incubated in a beaker containing incubation buffer for 30 min [32]. After the incubation, 39-decoyadenosine (cordycepin) was added to a final concentration of 200 mg/ml for all seedlings. Cordycepin was used as a transcription inhibitor because it works better than other inhibitors such as Actinomycin D and it is more effectively to penetrate into leaf tissues [30,32,70].
Treated leaves were harvested at each time point, 0, 60, and 120 min, after inhibitor treatment, and immediately frozen in liquid nitrogen. Total RNA was isolated using Plant RNA reagent (Invitrogen, Carlsbad, CA). Each RNA sample was quantified by measuring the 260/280 ratio using Nano drop and by agaroseformaldehyde gel electrophoresis. Quantified total RNA was subjected to mRNA isolation by Micro-fast Track 2.0 mRNA isolation kit (Invitrogen). Isolated mRNA was reverse-transcribed to synthesize cDNA using Cy3-or Cy5-labelled dCTPs (Amersham Biosciences, Piscataway, NJ) that was used as probes for microarray analysis, as previously described [14].

Quantitative RT-PCR
Approximately 5 mg of total RNA treated with DNase I was reverse-transcribed using SuperScriptII (Invitrogen), following the manufacturer's recommendation. The synthesized cDNA was diluted 1:5 in DEPC water and subjected to quantitative RT-PCR (qRT-PCR) analysis using SYBR Green Supermix (ABI Biosystems, Foster City, CA) in an ABI 7500 instrument (ABI Biosystems). The qRT-PCR conditions were optimized to maximize the amplification efficiency as previously described [71]. In this study, the gene eIF4A2 was used as an internal control to estimate the relative transcript level of the gene tested. The primers were provided in Table S3.

Spotted oligo-gene microarrays and data analysis
Spotted oligo-gene microarrays with 26,090 annotated genes were used for monitoring mRNA stability in allotetraploids according to the published protocol [14,71,72]. A total of 16 slides were used for two-time course comparisons (0 min vs. 120 min), including two biological and two technical replications. Each biological replicate consists of two dye-swaps (four slides) that were hybridized with reverse-labeled probes using one mRNA sample from the allotetraploid and another mRNA sample from a mix of RNAs from two parents (mid-parent value, MPV) ( Figure 1A). For microarray hybridization, 500 ng of mRNA were used to synthesize cDNA in each labelling reaction using Cy3-or Cy5-dCTP (Amersham Biosciences, Piscataway, NJ). Probe labelling, slide hybridization, and washing were according to a published protocol [72]. After the slides were scanned using Genepix 4000B, raw data were collected using Genepix Pro4.1. The replicated data from the dye-swap of each comparison were extracted and normalized using Acuity 4.0 software (Molecular Devices, http://www.moleculardevices.com). Quantified values in each dataset were analyzed by t-test using a statistical significance level of a = 0.001. The common genes between the two datasets with reproducibly normalized intensity ratios of $2 were used for further analysis. The difference of the transcript abundance between each time point can be used to calculate the half-life by the equation ln (Normalized Ratio) = -k decay , with t 1/2 = 0.693/ k decay , since mRNA degradation generally obeys first-order kinetics, which is common in fast mRNA degradation [30,73]. Gene Ontology (GOSlim) of AlloGUTs was classified using TAIR release (November 2009) (ftp://ftp.arabidopsis.org/home/tair/ Ontologies/Gene_Ontology/ATH_GO_GOSLIM.txt) and PED-ANT (http://mips.gsf.de/proj/thal/db/index.html). Clustering analysis of gene expression was performed using CLUSTER and TREEWIEW [74] (http://rana.lbl.gov/EisenSoftware.htm). Our microarray data were comparatively analyzed with the published data [14] to determine if nonadditively expressed AlloGUTs were enriched in GOSlim groups in response to stress [38], hormones [37], and circadian regulation [75]. The data were downloaded from http://www.arabidopsis.org/portals/expression/microarray/ ATGenExpress.jsp. Visual display of expression profiles from multiple experiments was performed in Genevestigator [76] (https://www.genevestigator.com/gv/index.jsp).
All microarray data are MIAME compliant http://www.ncbi. nlm.nih.gov/geo/info/MIAME.html and deposited in GEO http://www.ncbi.nlm.nih.gov/geo/with the accession GSE26065. Figure S1 The percentages of the genes in GOSlim category that are nonadditively expressed AlloGUTs (61 genes), compared to all AlloGUTs. The ratios in the y-axis were calculated using the observed percentage of the nonadditively expressed AlloGUTs divided by the expected percentage of all annotated genes in the Arabidopsis genome. The dashed line shows the observed percentage of the genes that are AlloGUTs in a microarray experiment equal to that of the expected genes in the whole genome (100%). (TIF) Figure S2 Hierarchical cluster analysis showing Allo-GUTs that matched the genes whose expression is induced by hormone and chemical treatments. The analysis was carried out with publicly available microarray database (http://www.arabidopsis.org/portals/expression/microarray/ ATGenExpress.jsp and Genevastor). Each row represents a microarray experiment and each column represents a gene. The color represents the relative expression level of each experimental group (Red, up and green, down). (TIF)