It was hypothesized that any variation in the cellular GSH contents of individual yeast cells could cause heterogeneity in GSH-dependent phenotypes. To test this, heterogeneity in a gsh1Δ-deletion strain (defective for the rate-limiting step of glutathione synthesis) was compared with that of wild-type cells. Heterogeneity was compared according to the relative gradients of dose–response curves, as described previously (Sumner et al., 2003; Bishop et al., 2007). Dose–response plots for the gsh1Δ mutant were shifted left relative to the wild type (Fig. 1). This indicated a culture-averaged sensitisation of gsh1Δ cells to Cd and H₂O₂, consistent with previous results (Wu and Moye-Rowley, 1994; Grant et al., 1996). Moreover, loss of Gsh1p was also associated with decreased cell-to-cell variability. Thus, loss of viability of mutant cells occurred over a narrower range of stressor doses (producing a steeper dose–response curve) than wild-type cells. An approximate 15-fold decline in viability of gsh1Δ cells (from 35% to 2.4%) resulted from only an ∼11% increase in Cd concentration (from 90 to 100 μM), whereas a similar loss of viability in wild-type cells required a > 70% increase in Cd concentration (Fig. 1A). No viable gsh1Δ cells could be detected at > 100 μM Cd. This degree of homogeneity resulting from a single gene knockout [equating to a heterogeneity ratio (HR) value ∼0.41; see Experimental procedures] was unprecedented in our experience, and was highly reproducible (further plots for wild-type and gsh1Δ cells are shown in Fig. 4B and D). GSH also contributed slightly to heterogeneous H₂O₂ resistance (Fig. 1B), but this effect (i.e. the difference in kill gradient for gsh1Δ versus wild-type cells) was considerably less marked than for Cd.

GSH-dependent Cd detoxification in yeast involves the formation of bis(glutathionato)cadmium (Cd·GS₂) adducts (Li et al., 1997). These adducts are transported to the vacuole by Ycf1p. To see whether vacuolar uptake of Cd·GS₂ (rather than adduct formation alone) was required to establish GSH-dependent heterogeneity, heterogeneous Cd resistance was also tested in a ycf1Δ mutant. Dose–response plots for the gsh1Δ and ycf1Δ mutants were superimposed (Fig. 1A). This indicated that the vacuolar detoxification pathway is the route through which heterogeneity is perpetrated. This was substantiated with a vam1Δ mutant which is defective for vacuole formation and maintenance. The vam1Δ mutant exhibited similarly decreased heterogeneity in Cd resistance (data not shown). As well as the Ycf1-mediated Cd detoxification mechanism, GSH also supplies reducing power for general antioxidant defence. This role is independent of Ycf1p. Accordingly, deletion of YCF1 influenced neither culture-averaged nor heterogeneous H₂O₂ resistance (Fig. 1B). Because GSH could affect heterogeneity independently of Ycf1p (in the case of H₂O₂), the data suggested that the relationship between Ycf1p and heterogeneity seen with Cd (Fig. 1A) was more likely to be a consequence of heterogeneous GSH rather than heterogeneous Ycf1p levels.

To confirm that the single-cell phenotypes were not inheritable (i.e. not genotypic), a number of ‘resistant’ colonies isolated after an initial exposure to the stressor on agar were subcultured to YPD broth in the absence of stressor. These were grown for 24 h in the broth before resistance was retested (non-inheritable stress resistance is normally lost within 24 h of out-growth) (Bishop et al., 2007). Cadmium was selected for these tests because it yielded the strongest phenotype in gsh1Δ cells and because of its genotoxicity (Jin et al., 2003). This should enhance detection of any genotypic variation. The initial isolation of ‘resistant’ wild-type and gsh1Δ cells (via colony formation on Cd) was at Cd concentrations (see Fig. 2 legend) that gave either 100% viability (minus-Cd control), ∼50% viability or ∼10% viability. Resistant isolates subcultured from these plates (see above) were retested for resistance at each Cd concentration. Wild-type isolates that had not previously been exposed to Cd (‘control’ isolates) exhibited the anticipated ∼50% and ∼10% colony formation when tested at the relevant Cd concentrations (Fig. 2). With one exception out of six colonies tested, the same was true of cultures derived from Cd-resistant wild-type colonies (‘50’ and ‘10’ isolates in Fig. 2). This indicated that these cells did not retain their resistances during the intervening growth period in the absence of Cd. Therefore, the variable Cd resistances of wild-type cells was primarily due to phenotypic rather than genotypic (inheritable) heterogeneity. In contrast, occasional Cd resistance in gsh1Δ cultures appeared to be primarily genotypic rather than phenotypic: five out of six resistant isolates exhibited inheritable Cd resistance (Fig. 2) (note that the higher Cd concentration gave closer to 0% than the normal 10% viability in some gsh1Δ retests, reflecting the difficulty of reproducing equivalent dosages in successive experiments where the kill gradient is very steep; Fig. 1A). The results indicated that much of the residual detected variability in gsh1Δ cells was probably of genotypic origin. Therefore, the relative gradients shown in Fig. 1A actually underestimate the true impact of GSH on phenotypic (non-genotypic) heterogeneity. It is concluded that elimination of GSH synthesis eliminates most of the non-genotypic heterogeneity in Cd resistance, i.e. there is no GSH-independent mechanism that makes a substantial contribution to this heterogeneity phenotype.

The GSH-specific fluorescent dye monochlorobimane (mBCl) (Stevenson et al., 2002) was used to substantiate that the above effects on heterogeneity were due to single-cell GSH content. The specificity of mBCl for GSH was confirmed by the severe decrease in fluorescence in gsh1Δ cells versus wild-type cells (Fig. 3A). The intensity of mBCl fluorescence in individual wild-type cells varied markedly, and some cells exhibited negligible discernible fluorescence (Fig. 3A). This heterogeneity in GSH content was reflected by the dispersion of mBCl fluorescences among wild-type cells when analysed by flow cytometry (Fig. 3B). Note that in contrast to certain other forms of heterogeneity (Smits et al., 2006), we found no evidence for bistability (distinct subpopulations) in mBCl fluorescences among cells. Our flow cytometry data indicated a graded phenotype, and this correlated with the continuous gradients seen in dose–response curves. To verify that single-cell resistance was related to single-cell GSH content, wild-type cells were sorted into subpopulations comprising cells with the lowest ∼20% or highest ∼20% GSH contents (mBCl fluorescences). The sorted subpopulations were then tested for Cd resistance at a range of Cd concentrations (Fig. 3C). Percentage resistance to Cd was up to 10-fold greater in high-GSH cells than in low-GSH cells. The results show that single-cell GSH content is a key determinant of single-cell Cd resistance. Furthermore, the gradients of the plots (Fig. 3C) indicated that the degree of heterogeneity in the low- and high-GSH subpopulations was decreased markedly compared with that in a total-cell population (HR ∼0.57 and ∼0.49 respectively). This substantiated that the extent of heterogeneity in GSH content of cell populations determines the extent of heterogeneity in Cd resistance.

It should be emphasized that the results shown in Fig. 3C are with cells sorted according to their ‘basal’ GSH contents, i.e. before any exposure to Cd. Therefore, the results demonstrate that it is this basal GSH content that determines single-cell Cd resistance. This supports the hypothesis (Sumner and Avery, 2002; Bishop et al., 2007) that it is the initial (pre-induction) state of cells on first contact with stressor that is the major determinant of heterogeneous resistance in this type of experiment. Individual survivors may subsequently mount an adaptive response (e.g. GSH1 induction in response to Cd) which should confer longer-term resistance and continued growth in the presence of stressor (Sumner and Avery, 2002), but this response does not determine the initial heterogeneity.

Levels of GSH are known to cycle during, and to help regulate, short-period metabolic oscillations in continuous cultures of S. cerevisiae (Murray et al., 1999). This could account for the cell-to-cell heterogeneity in GSH content. We confirmed such an oscillation among cells from the asynchronous batch populations used here. This involved sorting the high-GSH cells from cultures, to give a GSH-synchronized (homogeneous) subpopulation. Subsequently, changes in GSH content of these cells were monitored over time according to mBCl fluorescence. We observed a ∼40 min oscillation in GSH content, which was sustained for at least two cycles before the waveform became dampened as the subpopulation became less synchronous (Fig. 4A). This ∼40 min oscillation matched that observed elsewhere in oscillation-synchronized continuous cultures (e.g. Adams et al., 2003; Klevecz et al., 2004; Murray et al., 2007). The product of the GTS1 gene is required for generating normal oscillations (Wang et al., 2001; Adams et al., 2003; Xu and Tsurugi, 2007), and Gts1p also binds to Ycf1p (Kawabata et al., 1999). Therefore, it was hypothesized that the GSH/Ycf1p-dependent heterogeneity in Cd and H₂O₂ resistance described above could be established by the Gts1p-dependent oscillatory behaviour of individual cells. Consistent with this, cells of a gts1Δ mutant were less heterogeneous than wild-type cells in their resistances to Cd and H₂O₂ (Fig. 4B and C). This effect was strongest for Cd, particularly when it is taken into account that the wild-type strain starts to decrease in viability at a slightly lower Cd dose than the mutant, which accentuates the heterogeneity difference between the strains (Fig. 4B). In the case of Cd, the effect of GTS1 deletion on heterogeneity (HR ∼0.69) was less marked than that of GSH1 (or YCF1) deletion (HR ∼0.41; Fig. 1A). In contrast, these genes had equivalent effects on heterogeneity in H₂O₂ resistance (HR ∼0.78 in both cases). Examination of heterogeneity in a gsh1Δ,gts1Δ double mutant revealed that the effects of GSH1 and GST1 deletion on heterogeneous Cd resistance were not additive (Fig. 4D). This suggests that the gene products modulate heterogeneity via the same pathway.

The influence of GTS1 on metabolic oscillations in S. cerevisiae is known to be regulated at the level of GTS1 transcription. Oscillations are markedly dampened when GTS1 is expressed under the control of a constitutive promoter (Tonozuka et al., 2001; Wang et al., 2001). Consequently, we postulated that expressing GTS1 behind an alternative promoter could provide a means to manipulate the degree of heterogeneity in stress resistance within cultures. This approach would avoid GTS1 deletion, an important prerequisite for assigning a fitness effect specifically to heterogeneity (versus a culture-averaged change). For the same reason, we would need to express GTS1 at a culture-averaged level equivalent to that in the wild type (albeit with altered heterogeneity). Therefore, we replaced the genomic GTS1 promoter with the tet-regulated promoter system. This system gives constitutive gene expression at an averaged level that is scaleable with the concentration of doxycycline supplied (Belli et al., 1998a,b). To verify that P_TET-regulated GTS1 expression had the anticipated effects on heterogeneity, the constructed strain (P_TETGTS1) was stained with mBCl and analysed for cell-to-cell variation. The GSH contents of individual cells were more homogeneous in P_TETGTS1 cultures (and in gts1Δ cultures; not shown) than in the wild type, as reflected by the narrower range of cellular mBCl fluorescences (Fig. 5A). The derived coefficients of variation (CVs) for mBCl fluorescence were 114.5 and 70.2 for the wild-type and P_TETGTS1 cultures respectively. In contrast, the culture-averaged GSH contents were similar. Mean mBCl fluorescences from independent replicates were 33.7 ± 0.9 and 35.1 ± 0.2 respectively. Based on the calculations of Drakulic et al. (2005), mean intracellular GSH was estimated as ∼10.4 and ∼10.8 mM respectively. A comparable dampening of heterogeneity in GSH content was observed also in gts1Δ cells with a P_TETGTS1 construct integrated at the HO site (data not shown). This suggested that the observed effects on heterogeneity were not dependent on genomic locus. The ∼40 min oscillation seen in sorted high-GSH cells of the wild type was strongly dampened in the P_TETGTS1 strain, which exhibited a more uniform GSH content over time after sorting (Fig. 5B). A similar dampening was seen in a gts1Δ mutant (data not shown). Therefore, the decreased heterogeneity in GSH content (Fig. 5A) could be attributed to suppression of Gts1p-dependent metabolic oscillations.

To equalize mean Gts1p expression in wild-type and P_TETGTS1 cultures, GTS1 transcript levels were measured with quantitative real-time polymerase chain reaction (qRT-PCR) over a range of doxycycline concentrations (Fig. 6A). As expected, the levels of GTS1 mRNA in the P_TET-regulated strain decreased relative to those in wild-type cells with increasing doxycycline concentration. GTS1 expression was not affected by doxycycline in wild-type cells, consistent with observations elsewhere (Wishart et al., 2005). Doxycycline at 1.0 μg ml⁻¹ gave a GTS1 expression level that was just lower than that of wild-type cells (Fig. 6A), and no further repression of the P_TETGTS1 construct was observed at higher doxycycline concentrations. A concentration of 0.8 μg ml⁻¹ doxycycline was selected for further experiments. This gives equivalent averaged-expression of GTS1 in the two strains (Fig. 6A), but with altered consequences for heterogeneity (Fig. 5). We also employed Western blotting to confirm that the averaged level of the Gts1 protein was the same between the two strains at 0.8 μg ml⁻¹ doxycycline, in either the absence or presence of Cd(NO₃)₂ (Fig. 6B). Moreover, GSH levels were also equalized between the strains, as described above (Fig. 5A).

Dose–response assays were used to test whether the effects of P_TET-regulated GTS1 expression on variability in cellular GSH content (Fig. 5) resulted in effects on heterogeneous stress resistance. This prediction was borne out, as the P_TETGTS1 strain exhibited decreased cell-to-cell heterogeneity (steeper dose–response plots) compared with the wild-type when tested over a range of Cd or H₂O₂ concentrations (Fig. 6C). This result was reflected by comparison of the Cd resistances of sorted high/low-GSH subpopulations (see Fig. 3) with the P_TETGTS1 strain. Whereas the wild-type subpopulations exhibited a > 50% difference in survival with Cd (Fig. 3C), equivalent tests with the corresponding P_TETGTS1 subpopulations yielded only a ∼30% survival difference (data not shown). Another key observation with the dose–response curves (Fig. 6C) was that the P_TETGTS1 plots were shifted left relative to the wild-type plots. This indicated culture-averaged sensitization to Cd and H₂O₂ in the P_TETGTS1 strain. Because culture-averaged GTS1 expression and Gts1p levels (Fig. 6A and B), as well as mean GSH content (Fig. 5), were similar in the two cultures, it could be inferred that the advantage in wild-type cells was attributable specifically to heterogeneity.

To test this inference more rigorously, the cell division rates of the P_TETGTS1 and wild-type strains during exponential-phase growth in broth were determined over a range of Cd concentrations. The two strains exhibited very similar growth in the absence of Cd and at non-inhibitory Cd concentrations (< 12.5 μM Cd). However, the inhibitory effect of higher Cd concentrations on relative division rate was more marked in P_TETGTS1 than in wild-type cultures (Fig. 7A). Thus, at 75 μM Cd(NO₃)₂ the division rate of the wild type was still ∼30% of that measured in the absence of Cd, whereas the corresponding determination for the P_TETGTS1 strain was only ∼11%.

The hypothesis was tested further in direct competition assays. These were conducted under alternating control and Cd-stress conditions. This was because it has been speculated that non-genotypic heterogeneity might be particularly important in allowing survival of intermittent environmental stresses, in which a subset of resistant cells might survive and then re-establish populations under less stressful conditions (Tolker-Nielsen et al., 1998; Avery, 2006). The P_TETGTS1 and wild-type strains were initially mixed in 1:1 ratios and then co-cultured for 10 days, with daily subculture to maintain exponential phase. The test incubations comprised 24 h without Cd, then 24 h with Cd, over five cycles. The fitness-neutral KanMX4 marker in P_TETGTS1 cells was exploited to discriminate the strains for enumeration on agar, with or without selection by G418. Control incubations, with daily subculture in the absence of Cd throughout, confirmed that neither strain had a competitive advantage over the other under non-stress conditions (Fig. 7B). This was evident also from the control growth periods (preceding the arrows) in the alternating control/Cd incubations, during which the strain ratios did not change (Fig. 7B). In contrast, each 24 h growth period in the presence of Cd was associated with a marked shift in the relative proportions of the two strains, favouring the wild type. After five growth cycles ± 37.5 μM Cd, the ratio of P_TETGTS1 to wild-type cells had shifted from 1:1 to 1:3. Consistent with the previous data (Fig. 6A and B), semiquantitative RT-PCR as well as Western blot analyses confirmed that culture-averaged Gts1p expression did not differ significantly between the two strains (these analyses were with the strains growing separately under the same conditions as used in the main experiment, either in the absence or in the presence of Cd). Thus, out-competition of P_TETGTS1 cells by wild-type cells during stress can be linked to the enhanced phenotypic heterogeneity of the wild-type cells.

Yeast strains were maintained and grown in YPD broth [2% (w/v) bacteriological peptone (Oxoid), 1% yeast extract (Oxoid), 2% d-glucose] or in YNB medium [0.69% yeast nitrogen base without amino acids (Formedium), 2% (w/v) d-glucose] supplemented as required with amino acids or uracil (Ausubel et al., 2007). When selection was needed, either hygromycin B (Invitrogen) or G418 (Sigma) was added to media to a final concentration of 250 μg l⁻¹. Where necessary, media were solidified with 2% (w/v) agar (Sigma). Experimental S. cerevisiae cultures were inoculated from overnight starter cultures grown from single colonies, and cultured to exponential phase (OD₆₀₀∼2.0) in liquid medium at 30°C, 120 r.p.m.

Saccharomyces cerevisiae strains BY4741 (MATa; his3Δ1; leu2Δ0; met15Δ0; ura3Δ0), BY4742 (MATα; his3Δ1; leu2Δ0; lys2Δ0; ura3Δ0), BY4743 (MATa/α; his3Δ1/his3Δ1; leu2Δ0/leu2Δ0; lys2Δ0/LYS2; MET15/met15Δ0; ura3Δ0/ura3Δ0) and derivative single deletion mutants were obtained from Euroscarf (Frankfurt, Germany). A gts1Δ,gsh1Δ double deletion strain in the BY4743 background was created after disruption of GSH1 in gts1Δ single mutants from each of the haploid BY4741 and BY4742 backgrounds, using the HphNT1 cassette (Janke et al., 2004) to replace the entire GSH1 ORF by short flanking homology (SFH) PCR (Wach et al., 1997). Transformation was by the lithium acetate method (Gietz and Woods, 2002), and appropriate integration of the cassette was confirmed by diagnostic PCR (Wach et al., 1997). The haploid double mutants were mated by co-incubation overnight on YPD agar without selection, followed by two washes in sterile distilled water and incubation for 5 days on YNB agar supplemented with appropriate amino acids for selection.

To construct a cassette giving tet-regulated GTS1 expression, plasmid pCM225 containing P_TET (consisting of KanMX4 as the selectable marker, the tetracycline-responsive tTA activator gene and the tetO₇ promoter) (Belli et al., 1998a) was cut with PvuII and BglII, recessed DNA ends were filled with Klenow and ligated to produce plasmid pMS01. A 1.2 kb PCR fragment containing the GTS1 ORF was amplified from yeast genomic DNA with addition of terminal BclI and HpaI sites, and cloned between the BamHI and HpaI sites of pMS01 to yield plasmid pMS02 containing P_TETGTS1. A 1 kb region upstream of the GTS1 ORF (and preceding the GTS1pr.183 sequence required for oscillatory GTS1 transcription) (Tonozuka et al., 2001) was amplified using primers which incorporated terminal NotI and BstEII restriction sites and ligated between the corresponding sites in pMS02. The resulting plasmid, pMS06, was digested with NotI and SfoI to release a 6.2 kb cassette containing P_TETGTS1 and targeted for integration into the genome at the native GTS1 locus via the cloned 1 kb upstream fragment and the GTS1 ORF. This P_TETGTS1 cassette was transformed into BY4741 and BY4742 cells, G418-resistant transformants were screened for appropriate integration and haploid strains mated as described above to create the diploid P_TETGTS1 strain. To produce an equivalent cassette targeted to the HO locus, pMS02 was cut with NotI and SapI and recessed ends filled with Klenow to generate a blunt KanMX4, P_TETGTS1 fragment. This fragment was ligated into pBSTHO at the Klenow-filled EcoRI site, which lies in between ∼500 bp regions cloned from the 5′ and 3′ ends of the S. cerevisiae HO ORF (Payne, 2006). The product, pMS04, was digested with NotI and AarI to generate a 6.4 kb P_TETGTS1.HO cassette, which was transformed into BY4741 and BY4742 cells, in which the GTS1 ORF had previously been deleted by SFH PCR using the HphNT1 cassette, as described above. The resultant strains were mated to create a diploid P_TETGTS1.HO strain. All DNA cloning and genetic manipulations were performed in Escherichia coli strain DH5α (Invitrogen). Restriction digests, DNA ligations, sequencing and PCR were carried using standard protocols (Ausubel et al., 2007). All primer sequences are available upon request.

Exponential-phase experimental cultures (see above) were diluted in fresh YPD, and cells (∼200 per plate) were spread plated on YPD agar supplemented or not with the specified concentrations of Cd(NO₃)₂ or H₂O₂. All media for experiments involving P_TET-regulated GTS1 expression additionally were supplemented with 0.8 μg ml⁻¹ doxycycline. Colonies were enumerated after incubation at 30°C for up to 8 days, to allow for any slow growth under the stress conditions. Percentage resistances were calculated with reference to control incubations in the absence of stressor. HRs were calculated according to Sumner et al. (2003), to provide a measure of mutant versus wild-type gradients (heterogeneity) in dose–response curves. (In the sorting experiment of Fig. 3, HR was calculated from comparison of sorted versus unsorted subpopulations.) HR was defined as the ratio of the log increases in stressor concentrations required to give 1-log decreases in viability for mutant cultures relative to wild-type cultures (we used a 50% to 5% decrease in viability as a convenient 1-log standard). Complete elimination of heterogeneity would give an HR value of 0, whereas HR ∼1.0 would indicate that the deleted gene has no effect on heterogeneity.

Growth rates for P_TETGTS1 and BY4743 strains (indirect competitions) were determined separately using a BioTek Powerwave XS microplate spectrophotometer. Experimental cultures (OD₆₀₀∼2.0) grown in YPD containing doxycycline at 0.8 μg ml⁻¹ were diluted (OD₆₀₀∼0.1) into pre-warmed fresh medium and 300 μl samples aliquoted into 48-well plates (Greiner Bio-One) before addition or not of Cd(NO₃)₂ at the specified concentrations. Cultures were incubated at 30°C with shaking for 40 h, with the OD₆₀₀ measured every 30 min. Cell division rates (the number of generations per hour) in the presence of Cd(NO₃)₂ were derived from a minimum of at least four OD₆₀₀ determinations obtained during exponential growth, and were calculated as a percentage of the control growth rate measured in the absence of Cd(NO₃)₂.

Direct competition assays were performed by mixing exponential-phase P_TETGTS1 and BY4743 strains in a 1:1 ratio (based on OD₆₀₀) into fresh YPD supplemented with 0.8 μg ml⁻¹ doxycycline, and Cd(NO₃)₂ as required. Cultures were grown for 10 consecutive 24 h periods at 30°C, 120 r.p.m., with daily subculture to maintain exponential growth and allow media switches. At the time of each subculture, samples of cells were diluted in YPD before spread plating (∼200 cells per plate) on YPD agar supplemented or not with G418. Colonies were enumerated after incubation at 30°C for 3 days. The ratios of P_TETGTS1 to wild-type cells in cultures were calculated by comparison of the numbers of G418-resistant (P_TETGTS1) and G418-sensitive (wild type) colonies.

RNA was extracted from cells using the hot acid phenol method (Schmitt et al., 1990). Residual DNA was removed by treatment with RNase-free DNase (Promega), and treated RNA was recovered using an RNeasy Mini kit (Qiagen). The absence of protein or DNA contamination was confirmed according to A₂₆₀/A₂₈₀ ratios and standard PCR tests respectively. RNA was quantified using a NanoDrop spectrophotometer (NanoDrop Technologies) and integrity confirmed by formaldehyde agarose gel electrophoresis. Samples were snap-frozen in liquid nitrogen and stored at −20°C until use. Reverse transcription reactions were performed using Superscript III reverse transcriptase (Invitrogen) with 2 μg of RNA template per sample. Residual RNA was removed by treatment with E. coli RNase H (Invitrogen).

The relative abundance method was used to determine resultant cDNA levels by quantitative RT-PCR. Reactions (in triplicate) comprised 15 pmol each of gene-specific primers (HPLC purified, Sigma–Genosys), 2 μl cDNA template (10⁻¹ dilution), 12.5 μl of 2× QuantiTect SYBR Green PCR Master Mix, made up to 25 μl with RNase-free water. PCRs [95°C for 15 min (95°C for 30 s, 52°C for 30 s, 72°C for 30 s) for 40 cycles] were monitored using a MX4000 RT-PCR thermocycler (Stratagene). Initial template concentrations were calculated using the MX4000 software, and results were normalized using ACT1 as reference mRNA.

For semiquantitative RT-PCR, 1 μl of 10⁻¹, 10⁻², 10⁻³ and 10⁻⁴ dilutions of cDNA were used as templates in 50 μl PCRs [95°C for 5 min (95°C for 30 s, 52°C for 30 s, 72°C for 30 s) for 30 cycles; 72°C for 5 min] containing 50 pmol of each gene-specific primer. Red Hot Taq polymerase (AB Gene) was used for amplifications, and products were examined by agarose gel electrophoresis. The intensities of products at template dilutions in which the reaction had not progressed to saturation were estimated with ImageJ software (National Institutes of Mental Health, Maryland, USA) using ACT1 as reference mRNA. All primers were designed using Primer 3 software (Rozen and Skaletsky, 2000).

Preparation of S. cerevisiae lysates and Western blotting were performed using standard procedures (Ausubel et al., 2007). Proteins isolated from 2 × 10⁶ cells were separated by SDS-PAGE (10%) using an Invitrogen Nu-PAGE electrophoresis system, and blotted to PVDF membrane (Westram). Blots were probed with rabbit polyclonal anti-Gts1p primary antibody (a gift from Dr Kunio Tsurugi, Yamanashi Medical University) (1:2000 dilution), and alkaline phosphatase-conjugated anti-rabbit IgG secondary antibody (Sigma) (1:3000 dilution). Gts1p was detected with BCIP-NBT (Sigma) and quantified by densitometry with a Quantity One system (Bio-Rad).

Samples of exponential-phase cells (2 × 10⁷ cells) in YPD were stained with the reduced-glutathione (GSH)-specific dye monochlorobimane (mBCl; Molecular Probes) (Stevenson et al., 2002) at 100 μg ml⁻¹ for 10 or 15 min at 30°C. Cells were washed and resuspended in 2 ml PBS, briefly sonicated (Sanyo Soniprep 10 s, 2 μm), and analysed using a Coulter Epics Altra flow cytometer (Beckman) equipped with a UV laser. Data for fluorescence from mBCl via a 450 DF65 filter (PMT2) were collected for 50 000 cells in each sample. Cell-to-cell heterogeneity was determined as the CV [(standard deviation/mean) × 100%] after correction for autofluorescence using Weasel software (The Walter and Eliza Hall Institute of Medical Research, Australia). For Cd-resistance tests with culture subpopulations, cells were gated according to GSH content, sorted and spread plated onto YPD agar supplemented or not with Cd(NO₃)₂. Plates were incubated as described above to determine % viability. For analysis of GSH oscillations, sorted cells were incubated in YPD with shaking, and samples were removed at intervals for staining with mBCl and analysis as described above.

For fluorescence microscopy, cells were examined using a Zeiss Axioscope MS fluorescence microscope fitted with a HB050 illuminator (Carl Zeiss, Thornwood, NY), and images were captured with a Zeiss Axiocam digital camera fitted with a 470 DF20 filter.