To determine whether cells in stationary phase cultures could respond to stress by changes in transcript abundance, cells were harvested at 30 minute time intervals after the addition of 50 μM menadione (final concentration). For time course experiments, total RNA was isolated using the modified Gentra method (see Materials and methods). A carbon source was not added to ensure that metabolic activity would be low and constant and that any changes in transcript abundance would be due to stress and not to re-feeding.

Four patterns of change in mRNA abundance were observed over the eight hour time course (Figure 1; Additional data file 1). Similar patterns were identified by hierarchical clustering (Figure 1), K-means, and SOM(self-organizing maps) (not shown). By 30 minutes, two groups of transcripts (groups 1 and 2), comprising 1,090 mRNAs, showed 2- to 3-fold increases in abundance (Figure 1b, blue and red lines). Another group of transcripts (group 3) remained unchanged for about three hours and then increased three-fold (Figure 1b, black line). The fourth group of transcripts decreased in abundance (Figure 1b, green line). We concluded from these results that cells in stationary phase cultures could respond to additional environmental stress at the level of transcript abundance and that the response was rapid and relatively complex.

Groups 1 and 2, which increased by the first time point, differed in their patterns of expression over subsequent time points (Figure 1b, blue and red lines). Group 1 (616 transcripts) increased as much as three-fold by the first 30 minute time point and was relatively constant thereafter (Figure 1b, red line). A relatively large percentage (12%, p < 10^-4) of the mRNAs in this group encode proteins with functions related to stress responses, including the bZip transcription factor, YAP1 19 and other proteins associated with oxidative-stress resistance, such as the superoxide dismutase genes SOD1 and SOD2 20, thioredoxins TRX2, TSA1, PRX1, TRR1 and TRR2 21, and cytochrome-c peroxidase CCP1 22 (Additional data file 1). Transcripts encoded by DNA repair/damage response genes, including NTG1 23, DNA2 24, DIN7 25, were also in this group. Group 2 (474 transcripts) increased by the first time point and began to decrease within two to four hours (Figure 1b, blue line). A large and significant percentage of these transcripts (25%, p <10^-10) encode proteins required for ribosomal biogenesis and processing.

Group 3, comprising 475 transcripts, remained relatively unchanged for the first 4 hours and then gradually increased about 3-fold above T₀ levels (Figure 1b, black line). A significant number of mRNAs in this group (10%, p <10^-4) encode proteins associated with the proteosome, including ubiquitin (Ubi4p), polyubiquitin 26, Doa1p, which is involved with ubiquitin-mediated protein degradation 27, and the ubiquitin-conjugating enzymes Cdc34p 28 and Rad6p 29 (Additional data file 1). Also in this group are mRNAs that encode components of the proteasome core complex, including PUP2, PUP3, SCL1, and PRE1-10 30.

Group 4, comprising 170 transcripts, decreased approximately 4-fold in abundance within the first 30 minutes and remained constant thereafter (Figure 1b, green line). A significant number of mRNAs in this group (p <10^-2) encode proteins associated with DNA and phospholipid binding (Additional data file 1). We concluded from these results that there are coordinated changes in the abundance of transcripts encoding proteins with a variety of functions, including ribosome processing and biogenesis and response to stress, especially response to oxidative stress.

To determine the rate of the initial response to oxidative stress, shorter time intervals were needed. To sample more frequently while maintaining culture sterility, we designed and built a pneumatic device that can sample cells in culture at intervals as short as ten seconds 31. Using this device, cells were harvested at 1 minute time points for the first 35 minutes with an added time point at 1 hour. Surprisingly, a large group of mRNAs increased significantly by the first time point (1 minute) and a much smaller group of mRNAs decreased (Figure 2a; Additional data file 2). After the first time point, transcript abundance remained constant. Analysis of the first time points for both the 30 minute and 1 minute interval time courses revealed 508 transcripts that increased by 2-fold or more in both time courses. We concluded that the same response was detected in both experiments. Because the rate of transcription is known to be very low in cells in stationary phase cultures in the absence of an added carbon source 15 and this response has been shown not to be an artifact of automated sampling 31, the source of these transcripts was puzzling.

We hypothesized the rapid increase in transcripts in response to oxidative stress was due to one or more of three potential mechanisms. First, the apparent increase in transcripts could be due to new transcription. This seemed unlikely because the cells for these experiments were not given a carbon source during the oxidative stress and would be expected to have very low metabolic activity. Second, the transcripts might 'appear' to increase if they lacked a poly(A)⁺ tail because oligo-dT is used to prime cDNA synthesis for microarrays and, thus, non-polyadenylated transcripts in solution would not be detected. Rapid polyadenylation is known to occur in other systems, including during Xenopus oocyte development as well as during the dorsal ventral patterning of the Drosophila embryo 32. A third possibility was that transcripts might not be detected if they were not present in total RNA isolates because they were resistant to extraction, for example, by being in a complex with some structural component in the cell, but were solubilized or released after a stress. This would result in an apparent increase in abundance. If mature RNAs were sequestered in a protein complex, such as stress granules 33 or P-bodies 34, these RNAs would not typically be present in total RNA preparations because one of the first steps in all RNA extraction protocols is selective precipitation of proteins.

To determine whether new transcription was responsible for transcript increases, stationary phase cultures of a temperature-sensitive RNA polymerase II mutant (rpb1-1) 35, an RPB1 parental, and the wild-type S288c strain were incubated for three hours under non-permissive conditions for the rpb1-1 mutant (36°C) prior to oxidative stress. By two minutes after menadione exposure, transcript abundance increased dramatically in all three strains (Figure 3). Although there were some differences between these strains, 267 transcripts were identified that increased in all three (Additional data file 3). We concluded from these results that new transcription was an unlikely source for the rapid increase in mRNA after exposure to menadione.

To test the second hypothesis that transcripts were present in isolated total RNA but lacked poly(A)⁺ tails, we carried out quantitative RT-PCR analysis on cDNA samples synthesized using oligo-dT or random hexamer primers. Oligo-dT would not prime cDNA synthesis from non-adenylated transcripts. To determine whether partial transcripts were present, primer pairs that would amplify small fragments from either 5' or 3' ends of four transcripts (RPL38, TEF1, SOD1, and YAP1) that increase significantly after oxidative stress were used. Fold changes between random hexamer-primed and oligo-dT-primed cDNAs were less than one (indicating that more mRNAs were reverse transcribed using oligo-dT primers). In addition, fold changes were not significantly different at T₀ or T₂ nor were there major differences in fold change between the 5' or 3' end of any of the four transcripts (Figure 4). We concluded from this that all the transcripts seen at T₀ and T₂ were polyadenylated; thus, partial transcripts did not make up a significant percentage of all transcripts. Although a small number of transcripts were evaluated, these results provide evidence that the increase in transcript abundance detected by microarray analysis was not due to the presence of mRNA lacking a poly(A)⁺ tail but from real differences in mRNA abundance in the isolates.

To test the third hypothesis that the mRNA was present in the cell but resistant to extraction during RNA isolation, cell-free lysates obtained after breaking open cells and precipitating cell debris were tested. Lysates from T₀ and T₂ samples (at T = 0 and 2 minutes after menadione exposure) were incubated in buffer or with one of three different proteases, trypsin, proteinase K, or Qiagen protease, for 1 hour at 4°C prior to protein precipitation. Protease digestion was monitored by SDS-PAGE (Additional data file 4). Because the initial cell-free lysate in our RNA isolation protocol is aqueous, it was possible to use protease treatment to determine whether there was protease-labile RNA present. In RNA isolation protocols in which the initial lysis is done with phenol chloroform, protease treatment is not possible.

Approximately 2,100 transcripts from total RNA from T0 samples exhibited 2- to 128-fold increases after trypsin treatment (Figure 5, lanes 1 and 2: Additional data file 5). Many of these same transcripts increased after two minutes of exposure to menadione (without trypsin treatment) (Figure 5, lane 3). However, the increases in transcript abundance after exposure to menadione were less than after trypsin treatment. When T₂ samples were treated with trypsin, transcripts increased to the level seen in trypsin-treated T₀ samples (Figure 5, lanes 3 and 4). Digestion of T₀ samples with proteinase K resulted in similar increases of a larger number of transcripts (Figure 5, lanes 5 and 6), while treatment with Qiagen protease resulted in increases in fewer transcripts (Figure 5, lanes 7 and 8). We suspect that the lower efficiency of Qiagen protease may be a function of substrate specificity. We concluded from this result that RNAs were present in T₀ cells and by two minutes after menadione exposure most transcripts were in a soluble form. Because these transcripts are released by protease treatment it is likely that they are in complex with protein that is precipitated during most RNA isolation protocols. Finally, because the transcripts released by protease digestion of lysates were detected using oligo-dT primers for cDNA synthesis, we concluded that these transcripts were polyadenylated.

Because most studies of yeast are carried out using exponentially growing cultures, it was of interest to determine whether protease-labile mRNA was also present in dividing cells. Lysates incubated with trypsin, proteinase K, Qiagen protease or buffer only prior to RNA isolation revealed that few, if any, transcripts were in a protease-labile complex in these cells (Figure 5, lanes 9 to 14; Additional data file 5). Of the 16 transcripts that were common to trypsin- and proteinase K-treated lysates, six, YRB2, STV1, VPS28, DSS4, SRO7, and SGE1, encode proteins involved in intracellular transport and establishment of cellular localization. The small group of transcripts that increased after treatment with all three proteases (five genes), suggests that only a few genes are protease labile in these cells. The differences in transcripts that increase after treatment with the three proteases suggests that, if protease-labile, extraction-resistant mRNAs are present in dividing cells, the complexes may be more heterogeneous. Many transcripts show small decreases in abundance with protease treatment that could result from a relatively small change in the mRNA to total RNA ratio in these samples after protease treatment. We concluded from these results that extraction-resistant mRNA, which provides a major pool of mature mRNA in cells from stationary phase cultures, plays a small role, if any, in mRNA pools in non-stressed cells from exponentially growing cultures.

Although our RNA isolation protocol, which includes an initial aqueous lysis step, allowed us to treat the lysates with proteases prior to protein precipitation, it was possible that a potentially more disruptive RNA isolation technique might solubilize these transcripts in the absence of protease treatment. To test this hypothesis we isolated total RNA from unstressed cells from stationary phase cultures using hot phenol and compared these results with RNAs isolated after trypsin treatment. Transcript abundance was quantified using both microarray analysis (Figure 6) and quantitative RT-PCR (Additional data file 6). When compared with our RNA isolation protocol, replicate analysis mRNA extracted with hot phenol revealed that a small percentage of transcripts were solubilized more effectively with hot phenol while a larger percentage of mRNA was isolated more efficiently in our protocol. We concluded, after comparing mRNAs isolated by either protocol with transcripts isolated after proteinase K treatment, that neither protocol was as effective as the isolation including protease treatment. Quantitative RT-PCR corroborated these findings for a small number of transcripts (Additional data file 6). Although neither standard RNA isolation protocol was as effective as proteinase treatment, the differences between the two extraction protocols suggest that transcripts may bind to intracellular components with very different affinities and be extracted differentially by the two procedures.

To determine the overlap between transcripts released after oxidative stress and proteinase K treatment, we compared transcripts from 30 minute and 1 minute oxidative stress time courses and after proteinase treatment of T₀ samples (Figure 7). Although there were transcripts exclusive to each treatment, 65% of all transcripts that increased in response to oxidative stress also increased after proteinase K treatment. For both the 30 and 1 minute time courses, a significant number of transcripts that increased in response to oxidative stress also increased after proteinase K treatment (p <1 × 10^-15 for both). In contrast, only 45% of the transcripts that increased after proteinase K treatment of unstressed cells from stationary phase cultures also increased in response to oxidative stress. This is consistent with the earlier observation that many transcripts are present in a protease-labile form in cells two minutes after oxidative stress (Figure 5, lanes 3 and 4) and suggests that sequestered transcripts are present in these cells 30 minutes after oxidative stress.

Finally, to examine whether mRNA released from protein complexes is stress specific, we determined the overlap between transcripts that increased at least two-fold after oxidative stress (1 minute), temperature upshift (30°C to 39°C for 30 minutes), and proteinase K treatment (Figure 8) of stationary phase samples. As seen above (Figure 7), a significant percentage of transcripts that increased after oxidative stress also increased after proteinase K treatment. Likewise, a significant percentage (52%, p = 9.58 × 10^-5) of transcripts that increased after a temperature upshift (Additional data file 7) also increased after proteinase K treatment. Interestingly, only 45 transcripts were found to increase in response to both oxidative stress and a temperature upshift. This represented 16% and 5% of transcripts increased in response to temperature upshift and oxidative stress, respectively. The p value of 0.28 for this overlap indicated that transcripts released in response to oxidative stress and by temperature upshift showed no significant similarity. Consistent with earlier results, only 28% of proteinase K-labile transcripts increased in response to either stress; that is, 72% of the transcripts that increased in response to proteinase K treatment were not increased after either stress. From these results we concluded that: there is a large pool of protease-labile mRNA that are not released in response to either of the stresses tested here and that may be released in response to other environmental signals, such as re-feeding 14; and protease labile mRNAs appear to be released from protein-mRNA complexes in a stress specific manner.

Supplemental information and more detailed protocols are available at 46.

MATα S288c (his3 leu2-3, 112 lys2 trp1 ura3-52) cells were grown in YPD+A (1% yeast extract, 2% peptone, 2% D-glucose, and 0.04 mg/ml adenine) with aeration at 30°C for 7 days (to OD₆₀₀ 20-24). Rapidly dividing (exponential) cells were collected after overnight growth (OD₆₀₀ 1-2) under the same conditions.

Cells from stationary phase and exponentially growing cultures were collected to serve as a common reference for all experiments. Several cell samples were taken prior to exposure to menadione to serve as a T₀ reference. For oxidative stress menadione was added from a 500 μM stock to a final concentration of 50 μM. For the 30 minute time interval time course, cells were collected by pipette every 30 minutes for 8 hours. Cell viability was constant for at least 8 hours after exposure to this concentration of menadione (data not shown). For the second time course, an automated-sampling device 31 was used to collect cells at 1 minute intervals for 35 minutes, with a final time point at 1 hour. For both time courses, cells were harvested and analyzed in duplicate by microarrays. For temperature upshift, cultures were grown to stationary phase at 30°C, shifted to 39°C, and cells harvested by pipette every 30 minutes for 8 hours.

A random block design 47 was used to eliminate artificial sources of periodicity that may be introduced due to specific, constant ordering of time course samples during RNA isolation, cDNA labeling or hybridization, as well as to avoid confounding factors throughout the experiment. In a randomized block design, each time course is treated as a single block and samples within a time course were randomized for RNA isolation and re-randomized for both cDNA labeling and hybridization.

For RNA isolation (Additional data file 8), a modified Gentra protocol was used. Briefly, 20 OD₆₀₀ of cells from exponential phase or 30 OD₆₀₀ of cells from stationary phase cultures were lysed at 4°C in 300 μl of Cell Lysis Solution (Gentra, Minneapolis, MN, USA) using 0.5 mm glass beads (Sigma, St Louis, MO, USA) and a mechanical bead beater (Biospec Products, Bartlesville, OK, USA) at 4,800 rpm. Lysis was carried out in 6, 30-second bursts alternating with 30 seconds on ice. Samples were spun at 13,000 × g for 3 minutes at 4°C. After centrifugation, 100 μl of Protein-DNA Precipitation Solution (Gentra) was added to the supernatant and samples were incubated on ice for 5 minutes. Protein-DNA precipitate was pelleted at 13,000 × g for 3 minutes at 4°C. RNA was precipitated using 300 μl 100% isopropanol and pelleted by centrifugation. After centrifugation, RNA was resuspended in DEPC (diethylpyrocarbonate)-treated H₂O and a phenol-chloroform (5:1) 'back extraction' was performed (Ambion, Austin, TX, USA). RNA was precipitated overnight in 0.1 volume of 0.5 M NH₄OAc and 2.5 times the volume of 100% ethanol and subsequently purified using a Qiagen RNeasy kit (Qiagen, Alameda, CA, USA). RNA quality was evaluated using a Bioanalyzer 2100 (Agilent, Palo Alto, CA, USA).

For hot phenol extractions, 30 OD₆₀₀ of cells from stationary phase cultures were resuspended in 1 ml of sodium acetate buffer (50 mM sodium acetate, 10 mM EDTA, 0.1% SDS) and an equal amount of saturated phenol (Fisher Scientific, Pittsburgh, PA, USA) at 65°C. Samples were incubated at 65°C for 10 minutes, with vortexing every 1 minute for 10 seconds. Samples were spun for 10 minutes. The aqueous phase was then transferred to 1 ml of saturated phenol, vortexed for 1 minute and spun for 10 minutes. After the previous step was repeated the aqueous phase was transferred to 1 ml of chloroform (Sigma), vortexed for 45 seconds, and spun for 5 minutes. RNA was precipitated with 0.1 volume of 3 M Na Ac pH 5.2 and 2 times the volume of 100% ethanol at -20°C for 1 hour. RNA was pelleted by centrifugation, washed with 70% ethanol, and RNA was resuspended in DEPC-treated H₂O. RNA quality was evaluated using a Bioanalyzer 2100 (Agilent).

UltraGAPS slides (Corning, Corning, NY, USA) were printed using an OmniGrid 100 Arrayer (GeneMachines, San Carlos, CA, USA) with SMP4 printing pins (TeleChem, Sunnyvale, CA, USA). The yeast genomic oligonucleotide set (containing 70-mers corresponding to 6,307 open reading frames; Qiagen) was resuspended in 3× SSC to a final oligonucleotide concentration of 40 μM, and used to print the slides.

During printing, the relative humidity was maintained between 50% and 52% and the ambient temperature was maintained between 21°C and 23°C. After printing, slides were UV cross-linked at 90 mJ in a UV Stratalinker 1800 (Stratagene, La Jolla, CA, USA) and baked at 80°C overnight. For validation and quality control, slides were scanned after pre-hybridization treatment (see below) to screen for spot-localized contamination 48; SYBR green II staining (Invitrogen, Carlsbad, CA, USA) was used to test for DNA binding; and reproducibility experiments to test slide-to-slide reproducibility were carried out prior to and in each experiment. Typically, slide-to-slide standard deviation was in the range of 0.08 log₂ units for expression ratios.

A modified direct-labeling protocol 49 was used to fluorescently label cDNA with Cy3-dCTP or Cy5-dCTP (Amersham Biosciences, Piscataway, NJ, USA) using the Corning Microarray Technologies (CMT) Yeast Array 9/00 protocol (Corning; Additional data file 9). Total RNA (20 μg) from experimental samples were reverse transcribed to cDNA labeled with Cy5. A common reference sample RNA (20 μg of stationary phase and exponential RNAs combined at a 1:1 ratio) was reverse transcribed to cDNA labeled with Cy3 and pooled after labeling. A pooled sample of the common reference was optimized to maximize the number of genes hybridized and reduce variability throughout the array 50 and was used for all experiments. The use of a common reference allows all experimental information to be analyzed in relationship to the same reference, allowing better normalization 51.

For within-slide normalization, 10 ng of Arabidopsis thaliana CAB mRNA (Stratagene) was added to each labeling reaction to be used. Because we use a common reference and all the experimental information is in Cy5-labeled cDNA, even if the CAB mRNA labels slightly differently using Cy3 or Cy5, the normalization is consistent throughout the experiment.

Slides were pre-hybridized in a 250 ml glass Coplin jar for 1 to 2 hours at 42°C in a freshly prepared solution containing 50% formamide, 5× SSC, 0.1% SDS and 0.1 mg/ml bovine serum albumin fraction V (Sigma; Additional data file 10). The slides were rinsed several times with ddH₂O, dipped in 100% ethanol, and dried under a 30 psi stream of N₂ gas. Prior to hybridization, 22 × 30 mm Lifterslip coverslips (Erie Scientific, Portsmouth, NH, USA) were cleaned in a solution of 1 M KOH and 50% ethanol, rinsed with ddH₂O and dried with 30 psi of N₂ gas.

The hybridization buffer contained 50% formamide (Sigma), 5% dextran sulfate (Sigma), 5× SSC, 0.1% SDS, 0.1 mg/ml bovine serum albumin fraction V (Sigma), and 100 μg/ml salmon sperm DNA (Invitrogen). For all hybridizations, reference samples were labeled and pooled prior to hybridization. Each labeled experimental sample was combined with an aliquot of the labeled reference sample and dried down in a vacuum centrifuge.

The combined reactions for each slide were resuspended in 35 μl hybridization buffer, incubated at 95°C for 5 minutes, centrifuged for 30 seconds, and applied to the center of the coverslip that was subsequently positioned to cover the printed section of the slide. Slides were sealed in CMT hybridization chambers (Corning) and incubated at 42°C for at least 16 hours on a rocking platform. After hybridization, slides were washed as previously described 48 and subsequently dried using a 30 psi stream of N₂ gas.

All scans were performed with 100% laser power and photomultiplier tube (PMT) settings of 630 to 700 for the 635 nm laser and 430 to 500 for the 532 nm laser using Axon 4000B (Axon Instruments, Union City, CA, USA). These settings provided the maximum signal to noise ratio while reducing the number of saturated spots on each array. Grids used to define spot circumference, location, and identity were made and aligned using GenePix Pro 6.0 (Axon Instruments). All microarray data have been submitted to Gene Expression Omnibus series accession number GSE3729.

An ANOVA measurement model was used to determine variance between six independent biological and technical replicates of T₀ samples (Additional data file 11). The analysis showed a standard deviation for log₂ ratios of 0.08 for both biological and technical replicates, indicating that a change in gene expression greater than 1.6-fold could be viewed as statistically significant with a Type I error (or false positive rate) of ≤ 0.01.

Primers were designed using Primer Express (ABI, Foster City, CA, USA) and total RNA was reverse transcribed using either oligo-dT or random hexamers as primers. The two types of primers were used to determine whether transcripts lacking a poly(A)+ tail were present in total RNA isolated from unstressed, stationary phase samples. Quantitative RT-PCR reactions contained SYBR Green PCR Master Mix (ABI), forward and reverse primers (0.6 μM each), and a cDNA template (10 ng). Reactions were done in triplicate (technical replicates) for each sample and exogenous A. thaliana RNA (Stratagene) was used as a control (Additional data file 12). The cycle threshold (C_T) value for each reaction was determined using the ABI Prism 7000 SDS software package (ABI). C_T values were used to calculate the mean fold change of the reactions via the method, for which 1 indicates no change in abundance 52. For this experiment, the mean fold change was the difference in abundance in random hexamer-primed cDNA compared with the abundance in oligo-dT-primed cDNA.

After bead beating samples were centrifuged at 13,000 × g for 3 minutes at 4°C to remove cell debris. Cell-free lysates were divided into 2 tubes and incubated on ice for 1 hour with either 17 mg/ml trypsin in 0.9% (w/v) NaCl (Invitrogen), 14 mg/ml proteinase K in 10 mM Tris pH 7.5 (Qiagen), 11 mg/ml Qiagen protease in 10 mM Tris pH 7.5 (Qiagen), or 10 mM Tris pH 7.5 alone. After incubation, the RNA isolation for both samples was completed as described above. To visually evaluate the extent of protein digestion under these conditions, proteins isolated from protease-treated and control samples were separated by electrophoresis on a 10% TBE polyacrylamide gel (Bio-Rad, Hercules, CA, USA; Additional data file 4). All experiments were done in duplicate, using biological replicates. Microarray analysis showed that RNA isolated from samples incubated in buffer for 1 hour at 4°C were essentially identical to samples processed immediately (R² = 0.97).

MATa (ura3-52 rpb1-1) RNA polymerase II temperature-sensitive mutant 35, parental MATα (ura3-52), and wild-type S288c cells were grown to stationary phase (10 days) with aeration in 100 ml of YPD+A at 24°C. Prior to exposure to menadione, all strains were incubated for 3 hours at 36°C (non-permissive conditions for rpb1-1) prior to the collection of T₀ samples. Oxidative stress was induced using menadione as described above and samples from rpb1-1, parental and S288c cultures were harvested at 2 and 30 minutes after exposure.

A normal approximation to Fisher's exact test 53 was used to generate a p value for testing the association between membership on two sets of gene lists.