Transcripts Associated with Non-Acclimated Freezing Response in Two Barley Cultivars

MATERIALS AND METHODS

Cold Acclimation and Freeze Tests

To determine the phenotypic responses to NAF, control and NAF tests were conducted on 12 plants from each variety with four treatment conditions × three plants in a completely randomized design. Control plants were maintained under greenhouse conditions, while NAF plants were transferred directly from the greenhouse into a freezing chamber at −3°C for 24, 30, or 48 h, without light. After freezing, plants were returned to the greenhouse. Phenotypic measurements were taken on maturation, and comprised culm length, spike length, spikes per plant, seeds per spike, seed number, and seed yield. Least square means of each trait and significances of variety, treatment, and variety by treatment interaction were obtained using a general linear model in SAS (SAS, 2003). Two varieties and four NAF time points were tested for each phenotypic response.

For plant treatments before RNA extraction, additional plants from each variety were grown to the same jointing stage and then subjected to one of four treatments. Control plants were maintained under the greenhouse conditions. Acclimated (A) plants were transferred to a cold room at 4°C for 54 h with 8h/16h hrs light/dark; acclimated plus freezing (AF) plants were acclimated for 48 h as above and then transferred to the growth chamber at −3°C for 6 h without light, and non-acclimated freezing (NAF) plants were transferred directly from the greenhouse to the growth chamber at −3°C for 6 h without light. Plants in all treatments and the control were harvested at the same time. Upon completion of each treatment, leaf blades to the ligules were harvested, flash frozen in liquid nitrogen, and stored at −80°C.

RNA Extraction and Microarray Processing

RNA was extracted with Trizol reagent (Invitrogen, Carlsbad, CA), followed by column purification using the Qiagen RNeasy Midi kit (Qiagen, Valencia, CA). Quantity of RNA was estimated on a spectrophotometer, and quality of each sample was confirmed using an Experion bioanalyzer (Bio-Rad, Hercules, CA). Microarrays used the Affymetrix barley22K GeneChip (Close et al., 2004), performed on three individual plants × four treatments × two varieties, for a total of 24 chips. Ten μg total RNA was converted to cDNA using the Affymetrix one-cycle target labeling and control reagents (Affymetrix, Santa Clara, CA). Poly-A controls were spiked into the total RNA at a 1:5000 dilution before first and second strand cDNA synthesis, and half (6 μL) of the purified second strand cDNA was used in the synthesis of biotin-labeled cRNA. Twenty μg of labeled cRNA was fragmented in 1X fragmentation buffer by incubating at 94°C for 35 min, and quality of the labeled, fragmented cRNA was confirmed using agarose gel electrophoresis and an Experion bioanalyzer. Hybridization to the barley22K GeneChip arrays occurred for 16 h according to the manufacturer's protocol. GeneChips were washed and stained with a Fluidics Station 450 and scanned with a GeneChip Scanner 3000 (Affymetrix, Santa Clara, CA), both housed at the Center for Integrated Biosystems (Utah State University, Logan, UT).

Cell intensity (cel) files were analyzed using ArrayAssist version 5.1 (Stratagene, La Jolla, CA), using Robust Multi-array Average (RMA) normalization (Irizarry et al., 2003), followed by filtering for presence/absence calls with the MAS5 algorithm. For MAS5 filtering, probesets were removed from the study if they had an absence call in greater than 10 of the 12 chips when comparing A, AF, and NAF treatments to controls within the varieties. Thus only probesets with an absence call in the majority of chips from all treatments were discarded. When the two varieties were tested together, probesets with absent calls in greater than 22 of the 24 chips were removed from further analyses.

To assess clustering of replicates, principle component analysis (PCA) was conducted at a global transcriptome level for all 24 chips using ArrayAssist version 5.1. Significance gene lists were generated with ArrayAssist version 5.1 from unpaired t tests where a corrected P < 0.05 (Benjamini and Hochberg, 1995) and a twofold difference criteria were used for average transcript levels. Tests were conducted for A, AF, or NAF probesets compared to their control, and then Keunal NAF vs. Dicktoo NAF. For genes hypothesized to be associated with NAF tolerance, further t tests were conducted with the null hypothesis that the Keunal control average expression values were equal to the Dicktoo control expression values, at the corrected P < 0.05 but without a twofold criterion. Annotations for all probesets were procured from the Netaffx website (Affymetrix, Santa Clara, CA; http://www.affymetrix.com/analysis/index.affx; verified 25 Mar. 2008), and HarvEST version 1.62 (http://harvest.ucr.edu/ verified 25 Mar. 2008).

Gene lists of known barley cold-responsive genes were obtained from Skinner et al. (2005 and 2006) and Tondelli et al. (2006). Gene sequences for those genes were obtained from NCBI, and microarray probeset IDs were found through BLAST searches in the Netaffx website, keeping probesets with homology scores lower than e⁻¹⁰. Expression values of the resulting probesets were extracted from the microarray data using ArrayAssist version 5.1.

RESULTS

Non-Acclimated Freezing Tolerance of Dicktoo and Keunal

Dicktoo and Keunal barley varieties, vernalized and treated in the jointing stage of development, showed different phenotypic responses to non-acclimated freezing (NAF) conditions (Table 1). The main effects of variety and treatments were significant for all measurements with the exception of variety for the number of seeds per spike. Additionally, there was no treatment by variety interaction for any of the traits, indicating that the mean response of the Dicktoo variety was superior to Keunal on exposure to the NAF conditions. Non-acclimated freezing deleteriously affected both varieties, but Dicktoo was not affected to the same magnitude as Keunal. Keunal plants were unable to survive after 24 h of NAF, and the differences between Dicktoo and Keunal yield phenotypes were apparent at 24 h treatment, with the number of spikes per plant and the seed yield parameters showing a greater than 50% reduction in yield in Keunal. Dicktoo, however, had a less severe response to freezing at 24 h, with only yield per plant showing a marked reduction (Table 1).

Transcriptome Effects of Non-Acclimated Freezing

Dicktoo and Keunal array hybridizations resulted in 14,947 and 14,666 probesets with present calls from MAS5 filtering, respectively. These numbers represent approximately 65% of the probesets on the barley1 22K GeneChip. Principle component analysis of the 24 Genechips supported good repeatability among biological replicates based on clustering by genotype and treatment (Fig. 1), with the first two components explaining 75% of the variance. Non-acclimated freezing and control treatments were applied and two-fold differentially expressed transcripts at a corrected P value < 0.05 was used as a criteria for difference. With this criteria, 37 gene probesets had differential transcript levels in the Keunal NAF vs. control comparison (Comparison 1; Table 2), and 674 in Dicktoo (Comparison 2). In Dicktoo NAF 457 of the 674 variable probesets were downregulated and 217 upregulated with respect to control, while Keunal had 18 down- and 19 upregulated (Table 2). The complete microarray files from both varieties and treatments are available in the Plant Expression Database (http://www.plexdb.org/index.php; Accession BB56; verified 25 Mar. 2008).

Based on the differential sensitivity of Dicktoo and Keunal to NAF, we hypothesized genes with greater differential expression in Dicktoo than Keunal would either be directly involved in NAF tolerance or could serve as biomarkers of the differential response to freezing. Probesets were identified that were differentially expressed in Dicktoo NAF vs. Keunal NAF (Comparison 3; Table 2), Dicktoo NAF vs. Dicktoo control (Comparison 2), and Keunal NAF vs. control (Comparison 1) (Fig. 2). All 313 significant probesets are listed in Supplemental Table 1, while 71 probesets with a fourfold increase or decrease in Comparisons1 to 3, and with available annotations, are shown in Table 3. Of the 71 with annotations, 22 had Dicktoo NAF values upregulated with respect to controls, and 49 had Dicktoo NAF values downregulated with respect to controls (Table 3).

The 22 probesets upregulated in Dicktoo NAF (Table 3) also showed Keunal NAF and Keunal control transcript levels approximately equal to or less than Dicktoo control transcript levels, except for HD05F08r_at. Six of the probesets had variable transcript levels between Keunal control and Dicktoo control (bold face characters in Table 3), while the remainder were statistically equivalent—representing genes in which Dicktoo had equal or higher basal transcript levels and an increased capacity to induce expression in response to NAF treatment. The 49 probesets that were downregulated in Dicktoo NAF compared to Dicktoo control, Keunal NAF, and Keunal control expression values represent genes whose downregulation may be advantageous for NAF tolerance. All of the probesets downregulated in Dicktoo NAF (Comparison 2) had no transcript differences in Comparison 1 using our criteria (Table 3).

Among the probesets induced specifically in Dicktoo NAF were transcription factors involved in photoperiod response and circadian response, CCA1 and LHY (Hayama and Coupland, 2004). For both genes, quantitative qRT-PCR estimates of actin normalized expression values for three additional plants of each treatment and variety, confirmed the array results (Fig. 3). One of the additional three Keunal control plants had higher expression values compared to the other two replicates, raising the mean value and standard errors of the Keunal control, yet correlation coefficients between array and qRT-PCR expression values for the two were still 85% each (P < 0.01). Probesets with transcripts reduced specifically in Dicktoo NAF included several involved in water and sugar transport, the secretory pathway, mRNA binding proteins, and chaperone molecules (Table 3). Three of these genes were also analyzed on the additional plants with qRT-PCR, with the correlation between array and qRT-PCR values ranging from 75 to 85% (all P < 0.01) (Fig. 3).

Transcriptome Differences among NAF, A, and AF Treatments

To compare the effects of NAF with acclimated freezing responses, additional arrays were used to test transcriptome responses of cold acclimation at 4°C (A), and cold acclimation followed by freezing at −3°C (AF). Over 3500 probesets measured differential transcript levels in both the Keunal and Dicktoo control vs. A treatment, Comparisons 4 and 5 (Table 2). The majority of the differentially expressed probesets were downregulated with respect to the control conditions. Dicktoo hybridizations had more downregulated probesets and less upregulated probesets than Keunal, with an overall larger number detected. The AF treatment hybridizations, Comparisons 6 and 7, showed similar results; with more down- than upregulated probesets in both varieties and with more probesets detected in Dicktoo than Keunal (Table 2). Substantially more probesets were detected in the Comparisons 4 to 7 arrays compared to Comparisons 1 and 2 (Table 2). The array files of Comparisons 4 to 7 are also provided in the Plant Expression Database (http://www.plexdb.org/index.php; Accession BB56).

A large number of probesets with variable expression levels were shared among A and AF treatments for both varieties (Fig. 4). In Dicktoo 3992 probesets were shared, representing 75% of the Comparison 5 probesets and 88% of the Comparison 7 probesets. In Keunal, the 2327 probesets shared among A and AF treatments comprised 66% of each of the Comparison 4 and 6 variable probesets. A smaller number of probesets was shared among NAF-treated arrays and the A or AF treatments. Of the 674 Dicktoo NAF probesets with significant changes in transcript levels in Comparison 2, 426 (63%) were shared with Comparisons 5 and 7, while the remaining 248 (37%) were unique to Comparison 2 (Fig. 4). Of the 37 Keunal NAF probesets in Comparison 1, 29 (78%) were shared with Comparisons 4 and 6. Eight were unique to Comparison 1.

Expression Values of Known Low-Temperature Responsive Genes

Barley genes associated with cold-responsiveness were recently cloned and genetically mapped (Skinner et al., 2005; Skinner et al., 2006; Tondelli et al., 2006). Twenty four of those genes had matching probesets in the barley1 GeneChip, and expression levels are listed in Table 4. MYB and ICE-like probesets did not show significantly higher transcript levels in any of the cold temperature treatments in either variety. The dehydrin genes Dhn5 and Dhn13 were induced in Comparisons 4 to 7, A and AF treatments, in both varieties, while Dhn3 was only induced in Comparison 6. All HvCBF encoding genes identified, except HvCBF2, had significant transcript level differences in Comparison 7, while HvCBF1, HvCBF3, and HvCBF4 showed significant transcript level differences in Comparison 6. Fifteen of the 24 probesets were induced in at least one of the three low-temperature treatments and in at least one of the varieties, but only blt101 was induced significantly in a NAF treatment (denoted with asterisks in Table 4).

DISCUSSION

Dicktoo and Keunal are both winter barley types, yet the former exhibited less damage from NAF stress in its jointing stage. Dicktoo is considered a winter-hardy variety, and was the source of QTLs for winter hardiness and cold tolerance detected in a doubled haploid population (Hayes et al., 1993). Although both barley cultivars in this study are winter-types, Dicktoo is facultative in that it is low-temperature tolerant but lacks the ZCCT-H vernalization gene and the associated vernalization requirement (von Zitzewitz et al., 2005). Keunal is a winter-type variety selected for early maturity to be rotated with rice production in the southern part of Korea (Kweon et al., 1994). Recently, several winter-type barley cultivars were compared to spring-types and Dicktoo for vernalization and photoperiod responses (Limin et al., 2007). Among winter-types, variation was found for initial rates of cold acclimation and mechanisms of low-temperature response, suggesting likelihood that comparisons of other winter-types with Dicktoo could yield variable responses.

As Dicktoo showed greater tolerance to NAF than Keunal, and is winter-hardy, it is possible that the two traits are correlated among winter-barley varieties. Few studies have addressed NAF tolerance, but among them a study of reproductive frost tolerance in Australia found a QTL in several populations and environments in chromosome 5H (Reinheimer et al., 2004) that was coincident with the major winter hardiness QTL (Hayes et al., 1993; Francia et al., 2004). However, the two traits were not correlated in backcross families of Solanum species (Stone et al., 1993), such that the authors suggested independent selection for the two tolerances in breeding programs. The AF treatment in the current study is similar to the acclimation based freezing before LT₅₀ tests used to approximate winter hardiness (Fowler et al., 1981). When the AF vs. control and NAF vs. control probeset lists (Comparisons2 and 7) were tested in Dicktoo, only 382 were shared—representing 57% of the 674 NAF probesets and 8% of the 4557 AF probesets (Fig. 4A). Substantially fewer probesets were detected for Keunal Comparison 1, and 24 (63%) of Keunal NAF probesets and less than 1% of Keunal AF were shared. Thus the AF treatment transcriptome response likely would not predict the NAF response in the jointing stage. Whether or not extrapolation can be made for winter hardiness between transcript differences when sampled at the jointing stage in this experiment, to transcript differences at the seedling stage, is unclear. To better address commonalities in transcriptome responses to NAF and winter hardiness, additional studies using the same germplasm and treatment conditions are necessary at both stages.

Our criterion hypothesized to identify genes associated with NAF tolerance in Dicktoo found probesets with either higher or lower expression values in Comparisons 1 and 2, and also with higher or lower Dicktoo expression values in Comparison 3 (Table 2; Fig. 2). These were further classified into probesets with two-fold higher expression in Comparison 2, and probesets two-fold lower in Comparison 2 (Table 3).

The probesets with high expression levels in Comparison 2 were also tested for significant difference between Dicktoo control expression levels and Keunal control expression levels. Five of the 22 probesets were coincident with higher expression levels in Dicktoo control compared to Keunal control, one had lower expression levels in Dicktoo control, while the remaining probesets had equivalent expression values, suggesting that the higher expression levels under control conditions could provide preparation for, or an advantage to, Dicktoo plants on exposure to NAF. Consistent with this possibility were studies using transgenic, constitutive overexpression of CBF (CRT-repeat binding factor) transcription factors in Arabidopsis. When AtCBF1 or AtCBF3 were constitutively overexpressed, downstream CBF regulon genes were also constitutively expressed and transgenic plants survived NAF temperatures that wild-type plants could not (Jaglo-Ottosen et al., 1998; Gilmour et al., 2000). Similarly, the increased ability of Dicktoo in this study to survive and produce under NAF could result from the higher expression levels of upregulated genes before the cold treatment. Across the complete list of 313 hypothesized variable genes in the current study, only 13 had Dicktoo control expression levels lower than Keunal control while 121 had higher expression levels in Dicktoo control (Supplemental Table 1). From a selection perspective for NAF tolerance, increased tolerance may be selected for through–transcript based markers with plants tested under standard growth temperatures rather than freezing temperatures.

Probesets with low-expression values in Comparison 2 did not suggest a ‘prepared’ transcript state apparent in upregulated genes, as most Dicktoo control expression values were not lower than Keunal control. However, Dicktoo NAF values were still at least two-fold lower than Keunal NAF, suggesting that Dicktoo genotypes have an enhanced downregulation response that could allow for more tolerance to the freezing conditions. Prominent among the downregulated probesets were a glycine-rich RNA binding protein, an arginine methyltransferase, and several molecular chaperone genes. The RNA binding proteins are involved in RNA metabolism (Sachetto-Martins et al., 2000), and are common substrates for activation by arginine methyltransferases (Pahlich et al., 2006). The chaperone proteins DNAJ and DNAK are responsible for maturation of transcription factors and are all involved in stress response (Hendrick and Hartl, 1993). Other probesets detected as candidates, with previous reported involvement in low-temperature responses, were the tonoplast integral membrane proteins and aquaporins of maize (Chaumont et al., 2001). Interestingly, all of those genes were reported to be induced in response to low-temperature stresses, contrary to the downregulated expression effects seen in Dicktoo in this study. Perhaps, as these genes are members of gene families, they may be a family member that shows variable temporal, spatial, or developmental expression patterns. Alternatively, low-temperature stress response studies mainly occur at the seedling stage in plants, which may respond differently than the vernalized, jointing stage used in this study.

A previous report has shown several low-temperature responsive transcription factors to also be affected by circadian oscillations (Fowler et al., 2005), and among the candidates detected for NAF tolerance in the current study were probesets involved in photoperiod and circadian responses. The MYB-like transcription factors CCA1 and LHY, and TOC1-like pseudo response regulator (PRR), operate in Arabidopsis as a feedback loop essential for maintaining the circadian oscillations (Hayama and Coupland, 2004). The CCA1, LHY genes have maximal transcript expression in subjective mornings, while PRR has maximal expression in subjective evenings (Mizuno and Nakamichi, 2005). In the NAF tests in this study, plants from both varieties were transferred from the greenhouse to a freezing chamber in the mornings and subjected to freezing, without light for 6 h, before sampling in the afternoons. Therefore the predicted transcript levels of CCA1 and LHY would be reduced while the PRR would be increased, similar to Keunal but not Dicktoo (Table 3).

One possibility for the variable transcript levels of circadian genes in Dicktoo vs. Keunal would be genetic background responses to light rather than temperature, as the NAF treatment occurred without light while controls remained under greenhouse conditions. However if CCA1, LHY, and PRR were affected by light rather than temperature, the expectation would be a similar response of transcript levels in both Dicktoo and Keunal, even if the magnitude were different. Instead, we detected contrasting transcript levels in Dicktoo compared to Keunal (Table 3). Additionally, the Comparison 3 comprised plants treated under the exact same respective light and daytime conditions and harvested at the same time. Thus the reversal of transcript levels in Dicktoo NAF compared to Keunal NAF suggests that the NAF treatment, in addition to light, altered the circadian gene transcript levels. This alteration of transcript levels of Dicktoo circadian responsive genes by temperature may provide additional tolerance to freezing stress in Dicktoo. Interestingly, these circadian clock component genes are involved with photoperiod CONSTANS (CO) genes, which in turn induce a FLOWERING TIME (FT) gene coincident with a transition to flowering in Arabidopsis (Hayama and Coupland, 2004). The induced transcript levels for these photoperiod genes was substantially higher in Dickto, raising the possibility of an onset to flowering in response to NAF stress that was not seen in Keunal. Additionally, Gray et al. (1997) found that expression of Wsc cold response genes was affected by both light and low temperature, and suggested that both must be accounted for in attempts to increase levels of freezing tolerance.

Of the 24 genes in our arrays that were recently described in barley as involved in low-temperature responses (Skinner et al., 2005; Skinner et al., 2006; Tondelli et al., 2006), 15 were detected with significant expression differences using the comparisons and threshold we chose. Two genes with significant differences under A and AF conditions were dehydrins (Table 3), which are known to be correlated with a decrease in leaf water potential and an increase in acclimation-based freezing tolerance (Fu et al., 2000). Interestingly, the dehydrins were not upregulated in NAF treatment. Several CBF genes were also upregulated in A and AF treatments relative to control for Keunal as well as Dicktoo (Table 4) consistent with their role in cold acclimation.

The larger number of differentially expressed probesets in Dicktoo Comparison 2 compared to Keunal Comparison 1 (674 vs. 37; Fig. 2), and the low number of shared probesets between NAF and acclimation-based treatments for either variety (Fig. 4), suggests that the increase NAF tolerance of Dicktoo may also result from an increased number of genes detected in, and unique to, the NAF treatment. It was reported that freezing treatment following cold acclimation elicited additional transcriptome responses over those detected solely on acclimation (Herman et al., 2006). Our data support that report and show that additional unique probesets were affected by NAF vs. control compared to the AF vs. control treatment.

We have shown that Dicktoo is less susceptible of NAF stress than Keunal, and that its tolerance is coincident to two-fold transcript differences unique to a suite of Dicktoo NAF treated microarray probesets. The identification of these probesets provides initial insight into the genetic mechanisms for NAF response and tolerance: with sugar transport, water transport, RNA metabolism and transport, and circadian and photoperiod genes involved. Further studies, including time-course transcript quantification assays, will be necessary to elucidate the roles of these candidate genes. However, a starting point for selection or breeding of NAF-tolerant barley germplasm might focus on the probesets upregulated in Dicktoo NAF (Table 3; Supplemental Table 1). These probesets also showed a trend of higher transcript levels of Dicktoo control compared to Keunal control, such that selection could occur for base-expression level in untreated plants.