Gene-environment interactions underlie the etiology of the majority of recognized diseases in the human population. Recognition of this causality led to the establishment of the Environmental Genome Project which has goals to identify genetic susceptibility factors for environmentally responsive diseases and to establish a database for quantitative modeling of susceptibility gene-environment interactions [1]. Exposure to environmental hazards perturbs cellular homeostasis and induces a set of conserved cellular stress responses that serves to protect the cell from damage and to restore normal cellular functions. Assay of these conserved cellular stress responses in model systems provides a function-based approach to drug and chemical safety evaluation, which is finding increasing applications in drug discovery and in risk assessment of exposure to environmental chemicals. Gene expression microarray analyses of Saccharomyces cerevisiae cells exposed to a broad spectrum of chemicals and xenobiotics have revealed an ordered global "Environmental Stress Response" network that responds to protect yeast cells from external insult [2]. Toxicant- or drug-specific profiles, diagnostic gene expression signatures, have also been identified from these microarray studies [3]. In general, the yeast system has been the genomics platform of choice for both methods development and genetic network modeling in systems biology [4].

With increasing frequency, large sets of molecular profiling data derived from genomics, proteomics, and metabolomics assays are being collected from chemical- and drug-exposed cultured cells in vitro, whole animal models in vivo, and clinical material in situ [5]. Interpretation of these molecular profiling data is limited by the frequent lack of a quantifiable "phenotypic anchor" that would provide a biological context and frame of reference for systems-toxicology-based data mining of these data [6]. A major contribution of the yeast system has been to provide quantitative phenotypic data to validate microarray methods and data modeling procedures. The availability of a complete set of approximately 6000 gene deletion strains for the S. cerevisiae genome provides a unique resource for systematic analysis of the functional role of individual genes in a conserved model eukaryote [7]. To exploit this unique resource, high-throughput assays are needed to provide a phenotypic profile of the functional role of individual genes in the life and behavior of the organism [8 and 9].

Measurements of cell viability and growth provide versatile and sensitive assays for characterization of toxicological agents and conditions that adversely effect eukaryotic cells including yeast [10 and 11]. Cytotoxicity and cell proliferation assays are well-established, essential components of both drug safety evaluation for pharmaceutical drug discovery and chemical hazard evaluation for environmental risk assessment. However, the performance of individual cell-based assays is relatively slow and laborious in comparison to high-throughput assay techniques of enzymatic and gene expression profiling. In addition, traditional cellular assays frequently provide categorical data rather than the discrete quantitative data that are required to develop pharmacological and risk assessment models. Thus, there is a need for quantitative cell-based assays that can be automated for high-throughput analyses. To address this need, in this report we describe an automated microculture assay system for yeast cell growth and demonstrate its utility for the quantitative investigation of gene-environment interactions in the S. cerevisiae model system.

The cellular response to oxidative stress encompasses a highly conserved set of pathways that protect an organism against both endogenous and environmental insult from a broad spectrum of chemicals and xenobiotics [12 and 13]. Recently, we have developed a yeast-genomics-based model that exhibits a graded concentration-dependent response to oxidative stress: cell growth and adaptation, cell cycle checkpoint delay, cell growth arrest/apoptosis, and cell lysis/necrosis [14]. The phenotypic profiling assays that we describe in this report will add biological context to our corresponding gene expression profiling studies of the cellular network that mediates sensitivity to oxidative stress. These yeast-cell-based assays are validated with hydrogen peroxide, a direct-acting oxidant. As a proof of concept for analysis of more complex gene-environment interactions, we also describe a combinatorial assay design for phenotypic profiling of the cellular responses to tert-butyl hydroperoxide, a complex oxidant that is actively metabolized by the target cells.

 

MATERIALS AND METHODS

Reagents and growth media

Standard YPDA growth medium for yeast was made with deionized water and Bacto-grade components from DIFCO (Detroit, MI), adjusted to pH 5.8, and subsequently sterilized by autoclaving [15 and 16]. Dilution and washing of yeast cells were performed with sterile YPDA or with sterile distilled water. Hydrogen peroxide (H₂O₂) was purchased as a 30% (w/w) solution and was Trace-select grade (catalog 95321; Fluka Chemical, Milwaukee WI). This grade is free of chemical stabilizers and has reduced levels of metal contamination, both of which are known to have toxic effects on cells. tert-Butyl hydroperoxide was purchased as a 70% (w/w) solution and was reagent grade (catalog 45,813-9; Aldrich Chemical, Milwaukee, WI). Oxidants were stored at 4 °C in a hazard-class safety refrigerator and protected from exposure to light to maintain chemical stability and to minimize the formation of peroxide contaminants. Intermediate dilutions from the oxidant stocks were made using sterile distilled water; final concentrations for exposure to cells were made immediately before use and directly into YPDA growth medium. Under our standard culture conditions of YPDA medium and 30 °C, both hydrogen peroxide and tert-butyl hydroperoxide are chemically stable and show negligible loss of cytotoxicity over the time course of exposure (24-60 h), as measured by direct bioassay (data not shown).

Yeast strains and culture conditions

The wild-type strain of S. cerevisiae used in this study, BY4741, has the genotype mat a, his3 Δ1, leu2 Δ0, met15 Δ0, ura3 Δ0. BY4741 is the haploid parental yeast strain used to construct the systematic mutant collection of the Yeast Genome Deletion Project [7]. Strain CG3050 is the yap1 gene deletion derivative from this collection and has the same genotype as BY4741 with the additional yap1Δ::kanMX knockout mutation. Rich YPDA medium was used for both plate and liquid culture of yeast strains, which were incubated at 30 °C and handled by standardized methods of yeast genetics [15 and 16]. Permanent stocks of yeast strains are maintained in 50% (v/v) glycerol at −80 °C. For routine use, an aliquot of the frozen stock was streaked onto a YPDA agar plate and incubated for 2 days to form single colonies. The genotype of four independent colonies was tested by replica plating to selective drop-out media and a single verified clonal isolate to be used as a working stock was maintained on a YPDA agar plate at 4 °C for no longer than 1 week.

Yeast strains were grown to mid-logarithmic phase as starter cultures in shake flasks for inoculation of microculture experiments. A loop full of cells was taken from the working stock (agar plate), resuspended in a small volume of medium, and titered by hemacytometer count for inoculation of a shake flask culture. Shake flask cultures were incubated with orbital shaking at 300 rpm in an Innova Model refrigerated, shaking incubator from New Brunswick Scientific Co. (New Brunswick, NJ) in a volume that was 1/10th of the total volume of the Erlenmeyer growth flask; these conditions provide maximal aeration for cell growth. Inoculum titers were chosen to allow overnight incubation (16-18 h) and harvesting of cells in mid-logarithmic phase; balanced growth was attained at a titer of 1 × 10⁷ cells/ml. Under these shake flask conditions, a culture of strain BY4741 (YPDA medium, 30 °C) has a doubling time of 80 min in logarithmic growth (with approximately 75% budded cells), maintains logarithmic phase until a titer of 1-2 × 10⁸ cells/ml, and reaches a final plateau phase titer of 4-5 × 10⁸ cells/ml (with approximately 47% budded cells). Titers of yeast cultures were determined by direct counting in a hemacytometer chamber and visualized by phase-contrast light microscopy (320× magnification). Only those starting flask cultures that had been in balanced exponential growth for at least three generations and that had a budding frequency of 65-75% were used to inoculate microculture experiments.

Bioscreen C microculture growth of yeast

Microculture growth of yeast was performed in the Bioscreen C incubator/plate reader from Thermo Labsystems (Franklin, MA) using the honeycomb plate format (Fig. 1B). The Bioscreen C accommodates two 100-well honeycomb microtiter plates and provides automated incubation/reciprocal shaking and optical density measurement within the instrument; these features enable the high-throughput analysis of yeast growth [17]. We observe that the strict environmental control of temperature and shaking provided by the Bioscreen C instrument results in significantly less variation in growth parameters in microculture replicates than can be achieved in comparable shake flask culture ( Fig. 1). Routinely, we performed a set of 10 replicate microcultures for each growth assay condition. A shake flask starter culture was used to provide yeast cells in balanced growth to inoculate microcultures in the Bioscreen C honeycomb wells at a titer of either 1 × 10⁵ or 5 × 10⁶ cells/ml. For most assays, the lower cell titer was used to minimize any possible effects of cellular metabolism on the effective concentration of the added oxidant. The higher cell titer was used to provide a greater OD₆₀₀ signal for more sensitive detection of immediate changes in growth behavior, such as for monitoring the adaptation (Fig. 4) or cell lysis ( Fig. 6) responses.

(10K)

Fig. 1. Optimized microculture growth of yeast in a Bioscreen C honeycomb plate format. Yeast strain BY4741 was inoculated at an initial titer of 1 × 10⁵ cells/ml into 350 l of YPDA per well in a honeycomb microtiter plate and grown at 30 °C with reciprocal shaking in the Bioscreen C. Microcultures were grown either for (A, B) 22 h under shaking conditions that yielded optimized growth curves or for (C) 18 h under shaking conditions that yielded poor, irregular growth curves. In (A), uncorrected OD₆₀₀ measurements were recorded automatically every 30 min and are plotted for 10 replicate cultures. The mean average time to reach the half-maximum increase in growth, Tmean half-max, for this set of 10 replicate microcultures is 12.9 ± 0.21 h with a corresponding mean average OD₆₀₀ half-maximum value of 1.00; the dotted lines in (A) indicate these values graphically. In (B) and (C), a section of the honeycomb plate was photographed immediately at the termination of growth to display the distribution pattern of cells in the bottom of the flat wells during incubation. Optimal shaking (A, B) results in a relatively even distribution of cells in the well, whereas excessive shaking (C) results in reciprocal sidewall pelleting of the cells in the well. Bioscreen C control settings in Growth Curves software that yield optimal growth of yeast microcultures (A, B) were continuous, low shaking speed, and step INTERVAL=4. The excessive shaking settings that resulted in poor growth of yeast microcultures (C) were continuous, extra intense speed, and step INTERVAL=140.

Operation of the Bioscreen C and collection of optical density measurements (OD₆₀₀)² were computer automated with Growth Curves version 2.28 software from Transgalactic, (Helsinki, Finland) and Excel 2002 software from Microsoft, Inc. (Seattle, WA). To stably operate the Bioscreen C and to collect OD₆₀₀ measurements at short time intervals, it was necessary to dedicate a computer specifically for this purpose. Therefore, we interfaced the Bioscreen C to a Dimension 2350 Model computer from Dell Computer (Round Rock, TX) with a 1.8-GHz Pentium 4 chip and 512 MB of RAM and running the Microsoft Windows 2000 Pro (Service Pack 3) operating system. Our optimized operating parameters for yeast growth assays in the Bioscreen C format were determined to be YPDA medium in a 350-l well volume and a 30 °C incubation temperature with a preheat step of 10 min. The best growth responses were obtained by reciprocal shaking in the continuous mode with continuous duration, a shaking speed of low, and a shaking step of 4 (Fig. 1). OD₆₀₀ measurements were routinely collected at 30-min intervals and growth curves were displayed with Excel/Growth Curves software.

Quantitative analysis of yeast microculture growth curves

Under our standardized assay conditions of yeast growth in the Bioscreen C honeycomb format, we are able to quantify the growth curve response by use of a single metric, which we define as T half-max, the time to reach half of the maximum increase in growth of the culture (Fig. 1A). We define the OD₆₀₀ half-maximum increase in growth as (final OD₆₀₀ of the plateau growth − starting OD₆₀₀ of the inoculum)/2. The T half-max value for each microculture is calculated by linear interpolation between the time values that correspond to the OD₆₀₀ values of the two time points that bracket the OD₆₀₀ half-maximum increase in growth. Note that to allow examination of the quality of the raw data collected by the Bioscreen C, we have displayed the growth curves with uncorrected absorbance data; i.e., the starting background due to the initial cell inoculum has not been subtracted in the figures. The mean average T half-max value and its associated SE, standard error of the mean, are calculated for a set of 10 replicates of the same exposure condition. Thus each exposure condition is quantitatively described with a single metric value, "Tmean half-max ± SE." As an example, the control value of Tmean HALF-MAX=12.9 ± 0.21 h is graphically displayed in Fig. 1A. Since the OD₆₀₀ value measured in the Bioscreen C is nonlinear at high cell titers, the true T half-max value is underestimated in our calculation [17]. The empirical correspondence between the actual cell titer measured by direct counting in a hemacytometer and the OD₆₀₀ measured in the Bioscreen C plate reader is graphed in Fig. 2. Conveniently, the use of a single metric, measured under our defined assay conditions, enables a direct and simple quantitative comparison between cell growth responses to different concentrations of toxicant exposure.

(7K)

Fig. 2. Empirical calibration of cell titer versus OD₆₀₀ for yeast microculture growth in the Bioscreen C. Strain BY4741 was grown in YPDA medium in shake flask culture to a mid-logarithmic phase growth titer of 1 × 10⁷ cells/ml, as determined by direct counting in a hemacytometer. Aliquots of cells were then either diluted or concentrated (by centrifugation and resuspension in a smaller volume) in YPDA to give a series of defined cell titers that incrementally cover the range of 10⁵-10⁹ cells/ml. Each defined sample was then directly counted twice in a hemacytometer and the average titer was used for the Y axis value; 350 l of the titer-verified stocks was put into each of 10 honeycomb wells and their corresponding OD₆₀₀ values were measured in the Bioscreen C. The average of the readings from the 10 replicate wells was used as the X axis value. In this calibration, the background absorbance of YPDA medium without yeast cells, similarly determined to have a mean average OD₆₀₀ value of 0.131 ± 0.003, is not subtracted from the plotted OD₆₀₀ values. Thus, an uncorrected OD₆₀₀ value measured in the Bioscreen C can be directly converted to the corresponding cell titer by use of this empirical calibration curve.

 

RESULTS

Automated high-throughput assay of yeast cell growth in microculture

Asynchronous growth in batch culture of the budding yeast S. cerevisiae, is well characterized both in small-scale shake flasks and in large-scale fermentors [18]. We have utilized the Bioscreen C (Thermo Labsystems), a combination shaking incubator and optical plate reader, to establish optimal conditions for automated microscale batch growth of yeast cultures ( Fig. 1). This instrument accepts two honeycomb plates, each of which contains 100 wells ( Fig. 1B), and thus enables the simultaneous optical monitoring of 200 individual yeast microcultures. Growth conditions were optimized for the yeast reference strain BY4741, which is the haploid parent strain that was used to construct the Saccharomyces Genome Deletion Project mutant collection [19]. We first determined that 350-l microcultures in Bioscreen C honeycomb wells provide the largest volume that avoids cross-contamination of adjacent wells under conditions of vigorous shaking. Under our optimal growth conditions, this well volume remains constant over at least 60 h of incubation at 30 °C. The tightly fitted seal of the lid and bottom of the honeycomb plate and the heated lid applied during plate incubation in the Bioscreen C eliminate the problems of condensation and loss of volume that we routinely observe with comparable incubation of yeast microcultures in other 96-well standard-format microtiter plates.

To guide our optimization, we compared the growth of yeast in 350-l microcultures in Bioscreen C honeycomb wells against our standard shake flask conditions of a 100-ml culture in a 1-L Erlenmeyer flask with orbital shaking at 300 rpm. In shake flask culture in YPDA-rich medium at 30 °C, strain BY4741 exhibits a logarithmic phase doubling time of 80-85 min and a stationary phase plateau titer of 4-5 × 10⁸ cells/ml. The Bioscreen C, when computer automated by the Growth Curves version 2.28 software (Rein Metssalu, Transgalactic), provides reciprocal shaking with adjustable parameters of shaking pattern and duration (continuous or on/off discontinuous), speed (from low to extra intense), and step length (from 4 to 140). We varied these parameters in combinatorial fashion until we obtained the following optimal growth conditions in YPDA medium at 30 °C for 350-l microcultures of yeast: continuous shaking, low speed, and a step length of 4. Under these optimal conditions of microcultivation, strain BY4741 exhibits a highly reproducible growth curve with a logarithmic phase doubling time of 85-90 min and a stationary phase plateau titer of 5 × 10⁸ cells/ml, which is comparable to growth curves observed with our standard shake flask cultures (Fig. 1A). Under these optimal conditions, yeast cells are evenly distributed over the surface of the flat-bottomed honeycomb wells when the plate is removed for examination ( Fig. 1B). In contrast, if reciprocal shaking is too extensive, then pelleting of the yeast cells against the sides of the wells is observed ( Fig. 1C) and microculture growth is both less efficient and less reproducible. Thus, yeast growth in microculture in the Bioscreen C is comparable to growth in traditional flask shake culture.

Growth curves of 10 replicate yeast microcultures, inoculated at an initial titer of 1 × 10⁵ cells/ml from the same starter flask culture in logarithmic phase growth, are highly reproducible and have a mean average time to reach half-maximum growth of 12.9 ± 0.21 h (Fig. 1A). This reproducibility allows precise quantitative measurement of growth parameters and facilitates their comparison in yeast cultures grown under varying environmental conditions. Similarly, we have examined the growth variability among independent clonal isolates of the same genetic strain of yeast. Growth curves of 10 replicate microcultures, inoculated from starter flask cultures of each of five independent clonal isolates of strain BY4741, were similarly reproducible with logarithmic phase doubling times between 85 and 90 min and reached stationary phase plateau titers of 5 × 10⁸ cells/ml (data not shown). Thus, microcultivation in the Bioscreen C enables comparison of growth parameters in yeast strains of varying genotype when grown under the same environmental conditions.

To determine the relationship between cell density and microculture growth curves, we calibrated the OD₆₀₀ measurements of a series of samples of growing yeast at varying cell densities in the Bioscreen C against their corresponding actual cell titers as determined by direct counting in a hemacytometer (Fig. 2). From this calibration curve, we determine that the minimum titer of yeast cells that can be measured above the background absorbance of the YPDA growth medium is 1 × 10⁵ cells/ml for strain BY4741. As is typical for OD measurements of microbial cell growth, the uncorrected readings become nonlinear at higher cell densities where optical scatter becomes significant. For our yeast strains, nonlinearity of growth measurements is observed at OD₆₀₀ values greater than approximately 0.75. This result is consistent with similar results for yeast strains reported by Warringer and Blomberg [17], who determined an empirical correction factor that maintains linearity of OD₆₀₀ measurements throughout a growth curve in the Bioscreen C.

Quantitative assay of the cellular response to oxidative stress in yeast microcultures

In shake flask experiments, yeast cells exhibit a graded concentration-dependent response to chemically induced oxidative stress: continued growth, cellular adaptation, checkpoint arrest/growth delay, apoptosis, and necrosis [13, 20 and 21]. To quantify these cellular responses to environmental stress and to determine their genetic basis, we have adapted our oxidative stress protocol to a high-throughput microculture format. Wild-type yeast cells, growing in balanced exponential growth in a shake flask starter culture, were inoculated into the wells of a Bioscreen C honeycomb plate and each set of 10 replicate wells received an increasing concentration of hydrogen peroxide ( Fig. 3). In this experiment, the growth delay response, which is difficult to measure reproducibly in shake flask cultures, is precisely quantifiable in the set of 10 replicate microcultures in Fig. 3C, which exhibits a delay of 3.2 h versus the unexposed control microcultures in Fig. 3A. In this microculture assay, the growth delay response increases with the concentration of hydrogen peroxide over the range of 0.2-2 mM exposure (data not shown). In contrast, at exposure concentrations equal to or greater than 3 mM hydrogen peroxide, the challenged cells respond by growth arrest and apoptosis (Fig. 3D). Thus, the microculture format provides a quantitatively reproducible and precise assay system to characterize concentration-dependent cellular responses to oxidative stress.

(11K)

Fig. 3. Growth responses of wild-type yeast exposed to hydrogen peroxide. Wild-type strain BY4741 was inoculated at 1 × 10⁵ cells/ml in Bioscreen C honeycomb wells and incubated under standard conditions with increasing concentrations of hydrogen peroxide for 22.5 h. Each panel describes the growth of a set of 10 replicate microcultures. (A) Unexposed control, continued growth response (0 mM H₂O₂) with a Tmean HALF-MAX=10.8 ± 0.08 h; (B) adaptation response (0.25 mM H₂O₂) with a Tmean HALF-MAX=10.9 ± 0.10 h; (C) growth delay response (1.0 mM H₂O₂) with a Tmean HALF-MAX=14.0 ± 0.19 h; (D) growth arrest response (4.0 mM H₂O₂). Dotted lines represent the Tmean half-max values graphically.

Quantitative microculture assay for cellular adaptation to oxidative stress

At low concentrations of hydrogen peroxide, cells undergo an adaptation response and become more resistant to subsequent challenge with a high "apoptosis-inducing" concentration of oxidant [13]. Measurement of cellular adaptation is laborious when assayed in shake flask cultures by the standard protocol of determining quantitative killing curves for varying amounts of oxidative stress [12]. Therefore, we established a high-throughput protocol to quantify the adaptation response in a microculture format ( Fig. 4). A shake flask culture of wild-type yeast was grown in YPDA-rich medium in balanced growth and was harvested in mid-logarithmic phase at a titer of 1 × 10⁷ cells/ml. This starter culture was split into two flasks, one of which (unadapted control; Figs. 4A and B) continued shaking with no addition, while the other (adapted; Figs. 4C and D) was primed by shaking in the presence of 0.4 mM hydrogen peroxide. Shaking and growth was continued for 60 min, after which the two cultures were processed in parallel and washed by centrifugation to remove spent medium and the hydrogen peroxide. Subsequently, these cells were used to inoculate sets of 10 replicate microcultures in the Bioscreen C, either in the presence (Figs. 4B and D) or absence ( Figs. 4A and C) of a challenge concentration of 2.0 mM hydrogen peroxide. For this experiment, the microcultures were inoculated at a starting titer of 5 × 10⁶ cells/ml, which increases the sensitivity of detection of initial changes in cellular behavior in response to the oxidative stress.

(13K)

Fig. 4. Adaptation responses of wild-type yeast challenged with hydrogen peroxide. In a two-step protocol, wild-type strain BY4741 was first grown to mid-logarithmic phase in shake flask culture, either with (C, D) or without (A, B) a 1-h exposure to a low priming concentration of 0.4 mM H₂O₂. Next, these flask cultures were used to inoculate sets of 10 replicate microcultures in a Bioscreen C honeycomb plate at an inoculum titer of 5 × 10⁶ cells/ml, which then either were left unchallenged (A, C) or were exposed to a high challenge concentration of 2.0 mM H₂O₂ (B, D). (A) the unadapted and unexposed cultures have a Tmean HALF-MAX=5.8 ± 0.04 h; (B) the unadapted and exposed cells have a Tmean HALF-MAX=20.0 ± 0.26 h; (C) the adapted and unexposed cultures have a Tmean HALF-MAX=7.2 ± 0.04 h; (D) the adapted and exposed cultures have a Tmean HALF-MAX=9.0 ± 0.20 h. Dotted lines represent the Tmean half-max values graphically.

As demonstrated in Fig. 4, the growth of the unadapted control culture is severely delayed in the presence of the challenge concentration of 2.0 mM hydrogen peroxide. The mean average half-maximum growth time for the set of control microcultures is 5.8 ± 0.04 h (Fig. 4A) and is increased by 14.2 h to give a mean average of 20.0 ± 0.26 h when challenged (Fig. 4B). In dramatic contrast, the mean average half-maximum growth time for the set of primed microcultures is 7.2  ± 0.04 h (Fig. 4C) and is increased by only 1.8 h to give a mean average of 9.0  ± 0.20 h when challenged (Fig. 4D). Thus, prior adaptation to a low level of oxidant confers a 7.9-fold reduction in the time of growth delay upon acute exposure to a high level of oxidative stress. Our implementation of a microculture yeast assay enables high-throughput quantification of the adaptation response in this defined model genetic system.

Phenotypic profiling for toxicogenomics: genetic sensitivity to oxidative stress in yeast

As a proof of concept for development of toxicogenomic applications, we used the microculture assay to establish a functional profile for the role of YAP1, a known antioxidant defense gene, in cellular sensitivity to hydrogen peroxide [22 and 23]. Yap1p is a redox-sensitive transcription factor that activates the promoters of several environmentally responsive antioxidant defense genes in response to increased cellular levels of oxidative stress [24]. Thus, a yap1 gene deletion strain is unable to fully respond to oxidative stress and is consequently hypersensitive to oxidant challenge.

CG3050, a yap1 gene deletion strain, was grown and processed in a hydrogen peroxide stress response profile (Fig. 5) by the same microculture protocol that was used for its wild-type parent strain BY4741 (Fig. 3). The oxidant hypersensitivity of CG3050 is clearly demonstrated in Fig. 5 by the quantitatively reduced concentration dependence of its growth responses to hydrogen peroxide. A low nontoxic concentration of 0.2-0.25 mM hydrogen peroxide has little if any affect on the wild-type strain (Fig. 3B) but causes a significant growth delay of 9.2 h for the yap1 mutant (Fig. 5B). An intermediate concentration of 1.0 mM hydrogen peroxide causes a modest 3.2 h growth delay in the wild-type strain (Fig. 3C) but results in complete growth arrest/apoptosis in the yap1 mutant ( Fig. 5C). At a high cytotoxic concentration of 4.0 mM hydrogen peroxide, the wild-type strain exhibits a growth arrest/apoptosis response (Fig. 5D), while the yap1 mutant strain exhibits an extreme response of necrosis/cell lysis ( Fig. 5D). Mean average OD₆₀₀ values for the replicate set of 10 CG3050 microcultures exhibiting the apoptosis response in Fig. 5C increased by 0.011 over the 46.5-h duration of this experiment, whereas there was a corresponding mean average decrease by 0.018 for the set of microcultures exhibiting the necrosis response in Fig. 5D. This slight difference is reproducibly observed in other microculture experiments and correlates with diagnostic differences in cell morphology that we observe by cytological examination for these two conditions of exposure (data not shown). We and others [20 and 21] have previously shown that this extreme necrosis/cell lysis response occurs in the wild-type strain BY4741 at hydrogen peroxide exposures of greater than 7.5-10 mM. Thus, the microculture honeycomb plate assay fully supports quantitative functional profiling of yeast gene deletion strains and enables toxicogenomic applications.

(12K)

Fig. 5. Growth responses of a sensitive yap1 mutant strain exposed to hydrogen peroxide. Similar to the wild-type parent strain described in Fig. 3, a yap 1 mutant strain, CG3050, was inoculated at 1 × 10⁵ cells/ml in Bioscreen C honeycomb wells and incubated under standard conditions with increasing concentrations of hydrogen peroxide for 35 h. Each panel describes the growth of a set of 10 replicate microcultures. (A) Unexposed control, continued growth response (0 mM H₂O₂) with a Tmean HALF-MAX=13.6 ± 0.14 h; (B) growth delay response (0.20 mM H₂O₂) with a Tmean HALF-MAX=22.9 ± 0.22 h; (C) growth arrest response (1.0 mM H₂O₂); (D) growth arrest/cell lysis response (4.0 mM H₂O₂). Dotted lines represent the Tmean half-max values graphically.

Phenotypic profiling of complex cell-environment interactions in yeast

Hydrogen peroxide is a cell-permeable and direct-acting oxidant; these useful properties facilitated our development of a quantitative assay system to functionally profile gene-environment interactions in yeast. However, most chemical or drug interactions with cells are complex and involve active transport and metabolism of the environmental agent by the target cells. To extend our phenotypic profiling approach to characterize more complex gene-environment interactions, we have used tert-butyl hydroperoxide, which is actively metabolized by its target cells, as a test oxidant in our microculture growth response assay system.

For an extensively metabolized oxidant, it is expected that the concentration of metabolizing cells will itself influence both the nature and the concentration of the toxic reactive oxygen species produced by a specific concentration of oxidant in the microculture assay system. This complexity adds to the difficulty of determining the molecular mechanisms by which oxidant challenge produces a specific cellular stress response [25]. As an initial means to investigate the toxicant action of such a complex, metabolized oxidant in our microculture assay system, we have adopted a two-factor combinatorial assay design that independently varies both the concentration of initial oxidant and the titer of metabolically active target cells ( Fig. 6). The toxicant effects of tert-butyl hydroperoxide-induced oxidative stress are visualized by the time course growth patterns of yeast microcultures in response to specific combinations of oxidant concentration and cell titer.

(20K)

Fig. 6. Two-factor experimental design for tert-butyl hydroperoxide exposure in microculture. Wild-type strain BY4741 was grown to mid-logarithmic phase (1 × 10⁷ cells/ml) in shake flask culture and used to inoculate sets of 10 replicate Bioscreen C honeycomb wells at (A-D) 1 × 10⁵ cells/ml, (E-H) 4 × 10⁵ cells/ml, (I-L) 1.25 × 10⁶ cells/ml, or (M-P) 5 × 10⁶ cells/ml. Growth was monitored for 60 h in the presence of an increasing concentration series of tert-butyl hydroperoxide: (A, E, I, M) unexposed control at a concentration of 0 mM, (B, F, J, N) growth delay response at a concentration of 0.5 mM, (C, G, K, O) mixed growth delay/growth arrest responses at a concentration of 0.75 mM, and (D, H, L, P) growth arrest/cell lysis responses at a concentration of 2.0 mM. Dotted lines represent the Tmean half-max values graphically: (A) 12.3 ± 0.03 h, (B) 17.3 ± 0.14 h, (C) 52.3 ± 2.24 h, (E) 9.5 ± 0.03 h, (F) 13.8 ± 0.26 h, (G) 48.7 ± 1.35 h, (I) 7.9 ± 0.02 h, (J) 9.6 ± 0.08 h, (K) 32.9 ± 0.95 h, (M) 4.7 ± 0.01 h, (N) 5.9  ± 0.05 h, and (O) 13.2 ± 0.95 h.

The growth behavior of microcultures exposed to tert-butyl hydroperoxide is qualitatively similar to the concentration-dependent series of cellular responses to oxidative stress that we and others [20 and 21] have observed for exposure to hydrogen peroxide ( Fig. 3): unrestricted growth ( Figs. 6A, E, I, M, and N), growth delay ( Figs. 6B, F, J, and O), growth arrest ( Figs. 6D, H, and L), and cell lysis ( Fig. 6P). However, the specific oxidant concentration at which these effects are observed is highly dependent on the titer of yeast cells in the microculture assay. For example, as the inoculum cell titer increases, the delay in culture growth that is induced by either 0.5 or 0.75 mM tert-butyl hydroperoxide is decreased and is almost eliminated at an initial cell titer of 5 × 10⁶ cells/ml (Figs. 6B, F, J, and N). Similar results were observed for hydrogen peroxide (results not shown) and thus may reflect a reduction in the concentration of active peroxide in the medium due to its elimination by the increased levels of peroxidase activity associated with the greater cell numbers [26]. At a concentration of 0.75 mM tert-butyl hydroperoxide, the 10 replicate microcultures exhibit considerable variability of response, with some cultures exhibiting a growth delay and others exhibiting a growth arrest (Figs. 6C, G, K, and O). This variability in replicate response can be minimized by choice of the initial cell titer; as the cell titer is increased, the standard error of the Tmean half-max value decreases from 2.24 ( Fig. 6C) to 0.95 ( Fig. 6O). Thus, a two-factor (cell titer by chemical concentration) experimental design enables both quantification of the cellular responses to a metabolized oxidant and evaluation of the relative contributions of oxidant concentration versus cell titer to these responses.

 

DISCUSSION

The oxidative stress response is among the most highly conserved of all the cellular responses of a eukaryote to environmental challenge. This conservation reflects an underlying genetic network which is dynamic in time and allows transient adaptation to local environmental conditions [2 and 14]. To characterize and model the gene-environment interactions that are mediated by this conserved genetic network, we have developed a quantitative cell growth assay system in S. cerevisiae. Yeast cells in the Bioscreen C honeycomb microculture format maintain uniform growth conditions that enable highly precise and reproducible assays of chemical sensitivity (Fig. 2). The precision of this assay supports quantitative measurement of the concentration dependence of chemical challenge under defined environmental conditions ( Fig. 3 and Fig. 4). In addition, both qualitative and quantitative differences in genetic sensitivity of strains with defined genotypes are readily determined ( Fig. 3 and Fig. 5). This capability to separately quantify both toxicant concentration dependence and genetic susceptibility differences is critical for toxicogenomic network modeling of gene-environment interactions [27 and 28].

The precision of the yeast growth curve assay in the Bioscreen C is demonstrated by the highly uniform behavior of a set of 10 replicate microcultures. Under fixed inoculation and growth conditions, the time taken to reach the half-maximum increase in OD₆₀₀ (Tmean half-max) value provides a convenient single metric to quantify the growth behavior of a yeast microculture. For a set of 10 replicate control microcultures that is not exposed to chemical stress, the standard deviation of the Tmean half-max value is in the range of 0.04-0.2 h for a growth duration of 20 h (Fig. 2, Fig. 3 and Fig. 4). In contrast, for microcultures that are exposed to oxidative stress, the standard deviation of Tmean half-max is diagnostically greater than 0.2 h. In addition, this growth variability exhibited by replicate microcultures is maximal at the exposure concentration that immediately precedes the transition from the growth delay response to the growth arrest response (Fig. 4 and Fig. 5). This phenomenon of increased variability of T half-max at specific concentrations of oxidant challenge is particularly evident for the condition of 0.75 mM tert-butyl hydroperoxide exposure in Figs. 6C, G, K, and O). Empirically, we observe that increased variability in growth behavior for a set of 10 replicate microcultures can be used as a quantitative indicator to identify the toxicant concentration that immediately precedes a change in the type of cellular response to oxidative stress. We suggest that this increased variability of replicate cultures reflects a stochastic aspect in the underlying cellular regulatory process that mediates the choice between cellular responses to oxidative stress.

Under the assay conditions established in this report, the growth curve patterns observed in yeast microcultures accurately reflect the stress responses of individual cells and thus provide a means for automated quantification of these cellular responses. Most significantly, the cell cycle checkpoint response is easily quantified in microculture assay as a growth delay that is diagnostically characterized by an increased Tmean half-max value (Fig. 3, Fig. 4 and Fig. 5). In contrast, in shake flask cultures yeast cell cycle progression and delay are typically assayed by sampling sequential time points for fluorescence-activated cell sorting analysis; this requires significant sample preparation and is thus incompatible with real-time analysis. Cell cycle delay is a response to low concentrations of chemical exposure that are problematic to quantify reproducibly in shake flask culture. The use of a low cell inoculum of 1 × 10⁵ cells/ml in yeast microculture increases the sensitivity of the growth assay at low concentrations of oxidant exposure and allows the concentration-response curve to be extended to lower concentrations than is practical for comparable assays in shake flask culture (Figs. 6B, F, J, and N). The cellular stress response of growth arrest/apoptosis is qualitatively reflected in microculture assay as a static OD₆₀₀ value that is invariant over the time course of the assay (Fig. 3, Fig. 5 and Fig. 6). At extremely high concentrations of oxidant, the cellular response of lysis/necrosis occurs and is evident in microculture growth assay as a small but steady reduction in the OD₆₀₀ value over the time course of the assay (Fig. 5 and Fig. 6). This modest effect is more quantitatively reproducible if a relatively high cell inoculum of 5 × 10⁶ cells/ml is used in the yeast microculture (Figs. 6D, H, L, and P). The inoculum size of the starting microculture can be optimized to increase the assay sensitivity of the cell delay, cell arrest, and cell lysis responses to environmental stress.

The cellular response to low concentrations of oxidative stress results in cell proliferation and adaptation which provides increased resistance to oxidant exposure [12 and 13]. This protective response is assayed in a split-dose protocol in which exposure to a low (nontoxic), priming concentration is subsequently followed by exposure to a higher (toxic), challenge concentration of oxidant. We have implemented an adaptation assay in microculture format by first performing the priming exposure in a shake flask culture and then using this primed culture to inoculate microcultures that are exposed to the challenge concentration ( Fig. 4). The microculture adaptation assay is easily quantifiable by the reduction in the T half-max value of the adapted culture. The higher throughput and increased precision of yeast growth assays in the microculture format enable a quantitatively reproducible characterization of the cellular adaptation response that would be difficult or impossible to perform using shake flask cultures. Thus, all of the graded cellular responses to hydrogen peroxide-induced oxidative stress can be readily distinguished and quantified in yeast microculture assays.

Hydrogen peroxide provides a simple, direct-acting oxidant for validation of our yeast microculture assay format. Because it is membrane permeant and direct acting, its effects on cells are quickly observable and the exposed cells exhibit clear concentration-dependent responses. However, cellular antioxidant defense activities, such as catalase and other peroxidase enzymes, are able to metabolize and thus to neutralize the oxidizing capacity of hydrogen peroxide. This confounding effect of cellular metabolism on hydrogen peroxide toxicity becomes appreciable at extended times of incubation in excess of 40 h under our microculture assay conditions and is enhanced by cell titers above 10⁸ cells/ml (data not shown). We have intentionally established microculture assay conditions that are not subject to these metabolic complications. However, it is common for an environmental toxicant or a drug to have immediate and complex dynamic interactions with its cellular targets, both enhancement of its toxicity by active cellular uptake and metabolic activation and reduction of its toxicity by metabolic processing and cellular elimination. For such complex cell-toxicant interactions, the titer of cells in the assay is itself a variable that can influence the cellular stress responses.

As a proof of concept for the capability of the yeast microculture format to assay complex cell-toxicant interactions, we have evaluated the cell titer dependence of tert-butyl hydroperoxide toxicity in this system. For this purpose, we used a two-factor combinatorial assay design that independently varies inoculum cell titer and chemical concentration (Fig. 6). As is evident in Fig. 6, all four of the graded cellular responses to oxidative stress are visualized by this two-factor design. Transitions between the cell delay and the cell arrest responses ( Figs. 6C, G, K, and O) and between the cell arrest and the cell lysis responses ( Figs. 6D, H, L, and P) are also evident. Since 200 wells can be simultaneously monitored in the Bioscreen C microplate incubator/reader, the large number of combinations that are required to evaluate complex cell-toxicant interactions are accommodated by this microculture assay format. Thus, we conclude that the yeast microculture assay format is suitable for the characterization of cellular responses to complex metabolically responsive chemical species.

The wild-type yeast strain used in this study, BY4741, is the parent strain from which the Yeast Genome Deletion Project mutant collection was constructed [7]. Parallel phenotypic profiling of these collected mutant strains enables a systematic screening of the functional role of each individual gene in the life cycle and behavior of the organism [8 and 19]. This functional genomics approach to the dissection of complex gene-environment interactions requires high-throughput assays for quantitative phenotypic profiling of environmental conditions [9, 17 and 29]. The microculture growth assays that we describe in this report facilitate high-throughput toxicogenomic profiling of the cellular responses to oxidative stress. For example, functional assay of the yap1 deletion mutant strain for sensitivity to hydrogen peroxide demonstrates that the entire wild-type pattern of cellular responses is recapitulated in the mutant strain but at lower threshold of response. Thus the yap1 mutant strain is hypersensitive to oxidative stress ( Fig. 3 versus Fig. 5), which is consistent with the known biochemical activity of the Yap1 protein as a redox-sensitive transcription factor that induces the expression of several antioxidant defense pathways. This specific example illustrates the use of cellular stress response assays to provide a phenotypic anchor for interpretation of gene expression changes in response to environmental challenge [6 and 14]. To provide quantitative phenotypic data for toxicogenomic modeling of gene-environment interactions, we have developed an automated high-throughput growth assay for microcultures of yeast exposed to environmental toxicants [1]. In addition to toxicogenomic applications, we anticipate that the phenotypic profiling assays described in this report will prove useful for quantitative risk assessment of chemical mixtures and of chronic low-concentration chemical exposures.

 

ACKNOWLEDGEMENTS

We thank Dr. Anders Blomberg for sharing his experience with yeast growth in the Bioscreen C with us and Rein Metssalu (Transgalactic Ltd.) for advice on use of the Growth Curves software. This research was supported by ONR Research Grant N00014-01-1-0220 to C.N.G. and by a pilot project grant to C.N.G. from NIEHS Center Grant ES 06639.

 

REFERENCES

1. M. Waters, G. Boorman, P. Bushel, M. Cunningham, R. Irwin, A. Merrick, K. Olden, R. Paules, J. Selkirk, S. Stasiewicz, B. Weis, B. Van Houten, N. Walker and R. Tennant, Systems toxicology and the Chemical Effects in Biological Systems (CEBS) knowledge base. EHP Toxicogenomics 111 (2003), pp. 15-28.

2. A.P. Gasch and M. Werner-Washburne, The genomics of yeast responses to environmental stress and starvation. Funct. Integr. Genomics 2 (2002), pp. 181-192.

3. T.R. Hughes, M.J. Marton, A.R. Jones, C.J. Roberts, R. Stoughton, C.D. Armour, H.A. Bennett, E. Coffey, H. Dai, Y.D. He, M.J. Kidd, A.M. King, M.R. Meyer, D. Slade, P.Y. Lum, S.B. Stepaniants, D.D. Shoemaker, D. Gachotte, K. Chakraburtty, J. Simon, M. Bard and S.H. Friend, Functional discovery via a compendium of expression profiles. Cell 102 (2000), pp. 109-126.

4. T. Ideker, T. Galitski and L. Hood, A new approach to decoding life: systems biology. Annu. Rev. Genomics Hum. Genet. 2 (2001), pp. 343-372.

5. F.S. Collins, E.D. Green, A.E. Guttmacher and M.S. Guyer, A vision for the future of genomics research. Nature 422 (2003), pp. 835-847.

6. R. Paules, Phenotypic anchoring: linking cause and effect. Environ. Health Perspect. 111 (2003), pp. A338-A339.

7. G. Giaever, A.M. Chu, L. Ni, C. Connelly, L. Riles, S. Veronneau, S. Dow, A. Lucau-Danila, K. Anderson, B. Andre, A.P. Arkin, A. Astromoff, M. El-Bakkoury, R. Bangham, R. Benito, S. Brachat, S. Campanaro, M. Curtiss, K. Davis, A. Deutschbauer, K.D. Entian, P. Flaherty, F. Foury, D.J. Garfinkel, M. Gerstein, D. Gotte, U. Guldener, J.H. Hegemann, S. Hempel, Z. Herman, D.F. Jaramillo, D.E. Kelly, S.L. Kelly, P. Kotter, D. LaBonte, D.C. Lamb, N. Lan, H. Liang, H. Liao, L. Liu, C. Luo, M. Lussier, R. Mao, P. Menard, S.L. Ooi, J.L. Revuelta, C.J. Roberts, M. Rose, P. Ross-Macdonald, B. Scherens, G. Schimmack, B. Shafer, D.D. Shoemaker, S. Sookhai-Mahadeo, R.K. Storms, J.N. Strathern, G. Valle, M. Voet, G. Volckaert, C.Y. Wang, T.R. Ward, J. Wilhelmy, E.A. Winzeler, Y. Yang, G. Yen, E. Youngman, K. Yu, H. Bussey, J.D. Boeke, M. Snyder, P. Philippsen, R.W. Davis and M. Johnston, Functional profiling of the Saccharomyces cerevisiae genome. Nature 418 (2002), pp. 387-391.

8. B.R. Bochner, New technologies to assess genotype-phenotype relationships. Nat. Rev. Genet. 4 (2003), pp. 309-314.

9. K.J. Rieger, M. El-Alama, G. Stein, C. Bradshaw, P.P. Slonimski and K. Maundrell, Chemotyping of yeast mutants using robotics. Yeast 15 (1999), pp. 973-986.

10. D. Deere, J. Shen, G. Vesey, P. Bell, P. Bissinger and D. Veal, Flow cytometry and cell sorting for yeast viability assessment and cell selection. Yeast 14 (1998), pp. 147-160.

11. A. Lentini, A review of the various methods available for monitoring the physiological status of yeast: yeast viability and vitality. Ferment 6 (1993), pp. 321-327.

12. D.J. Jamieson, Oxidative stress responses of the yeast Saccharomyces cerevisiae. Yeast 14 (1998), pp. 1511-1527.

13. K.J. Davies, The broad spectrum of responses to oxidants in proliferating cells: a new paradigm for oxidative stress. IUBMB Life 48 (1999), pp. 41-47.

14. C.N. Giroux, A. Weiss, J. Delproposto and S. Stapels, A dynamic genetic network mediates dose-dependent oxidative stress responses. Toxicologist 72 S-1 (2003), p. 92.

15. F. Sherman. In: C. Guthrie and G.R. Fink, Editors, Guide to Yeast Genetics and Molecular BiologyMethods in Enzymology vol. 194, Academic Press, New York (1991), pp. 3-21.

16. D. Burke, D. Dawson and T. Stearns, Methods in Yeast Genetics: A Cold Spring Harbor Laboratory Course Manual. , Cold Spring Harbor Laboratory Press, Plainview (2000).

17. J. Warringer and A. Blomberg, Automated screening in environmental arrays allows analysis of quantitative phenotypic profiles in Saccharomyces cerevisiae. Yeast 20 (2003), pp. 53-67.

18. G.M. Walker, Yeast Physiology and Biotechnology. , Wiley, New York (1998).

19. E.A. Winzeler, D.D. Shoemaker, A. Astromoff, H. Liang, K. Anderson, B. Andre, R. Bangham, R. Benito, J.D. Boeke, H. Bussey, A.M. Chu, C. Connelly, K. Davis, F. Dietrich, S.W. Dow, M. El Bakkoury, F. Foury, S.H. Friend, E. Gentalen, G. Giaever, J.H. Hegemann, T. Jones, M. Laub, H. Liao, R.W. Davis et al., Functional characterization of the S. cerevisiae genome by gene deletion and parallel analysis. Science 285 (1999), pp. 901-906.

20. C.N. Giroux, Y. Wang and A. Weiss, A yeast functional genomics model for the cellular response to oxidative stress. Toxicologist 66 1-S (2002), p. 19.

21. F. Madeo, E. Frohlich, M. Ligr, M. Grey, S.J. Sigrist, D.H. Wolf and K.U. Frohlich, Oxygen stress: a regulator of apoptosis in yeast. J. Cell Biol. 145 (1999), pp. 757-767.

22. S.T. Coleman, E.A. Epping, S.M. Steggerda and W.S. Moye-Rowley, Yap1p activates gene transcription in an oxidant-specific fashion. Mol. Cell. Biol. 19 (1999), pp. 8302-8313.

23. J. Lee, C. Godon, G. Lagniel, D. Spector, J. Garin, J. Labarre and M.B. Toledano, Yap1 and Skn7 control two specialized oxidative stress response regulons in yeast. J. Biol. Chem. 274 (1999), pp. 16040-16046.

24. A. Delaunay, A.D. Isnard and M.B. Toledano, H2O2 sensing through oxidation of the Yap1 transcription factor. EMBO J. 19 (2000), pp. 5157-5166.

25. K.J. Davies, Oxidative stress, antioxidant defenses, and damage removal, repair, and replacement systems. IUBMB Life 50 (2000), pp. 279-289.

26. E.J. Collinson, G.L. Wheeler, E.O. Garrido, A.M. Avery, S.V. Avery and C.M. Grant, The yeast glutaredoxins are active as glutathione peroxidases. J. Biol. Chem. 277 (2002), pp. 16712-16717.

27. T.E. Ideker, V. Thorsson and R.M. Karp, Discovery of regulatory interactions through perturbation: inference and experimental design. Pac. Symp. Biocomput. (2000), pp. 305-316.

28. R. Somogyi and L.D. Greller, The dynamics of molecular networks: applications to therapeutic discovery. Drug Discov. Today 6 (2001), pp. 1267-1277.

29. B.R. Bochner, P. Gadzinski and E. Panomitros, Phenotype microarrays for high-throughput phenotypic testing and assay of gene function. Genome Res. 11 (2001), pp. 1246-1255.

* Corresponding author. Fax: 1-313-577-6200

¹ Present address: Pfizer Global Research and Development, Ann Arbor, MI, USA.

² Abbreviations used: OD₆₀₀, optical density at 600 nm, Tmean half-max, mean average time for a set of cultures to reach half of their maximum increase in growth, as measured by OD₆₀₀.