Report

Signaling and Circuitry of Multiple MAPK Pathways Revealed by a Matrix of Global Gene Expression Profiles

See allHide authors and affiliations

Science  04 Feb 2000:
Vol. 287, Issue 5454, pp. 873-880
DOI: 10.1126/science.287.5454.873

Abstract

Genome-wide transcript profiling was used to monitor signal transduction during yeast pheromone response. Genetic manipulations allowed analysis of changes in gene expression underlying pheromone signaling, cell cycle control, and polarized morphogenesis. A two-dimensional hierarchical clustered matrix, covering 383 of the most highly regulated genes, was constructed from 46 diverse experimental conditions. Diagnostic subsets of coexpressed genes reflected signaling activity, cross talk, and overlap of multiple mitogen-activated protein kinase (MAPK) pathways. Analysis of the profiles specified by two different MAPKs—Fus3p and Kss1p—revealed functional overlap of the filamentous growth and mating responses. Global transcript analysis reflects biological responses associated with the activation and perturbation of signal transduction pathways.

Mitogen-activated protein kinase (MAPK) cascades control changes in gene expression, cytoskeletal organization, and cell division (1). Genome-wide DNA microarrays are emerging as a powerful tool for broad correlation of gene activity with alterations in physiological or developmental states (2–4). Global expression profiles assembled from diverse experiments have the potential to reveal the hierarchical connections between signal transduction pathways, the patterns of gene expression underlying complex biological responses, and the activity of specific pathways and effectors. To elucidate the signal circuitry and gene activity that specifies MAPK-regulated states, we used global transcript analysis to study activation and genetic perturbation of the budding yeastSaccharomyces cerevisiae pheromone response MAPK pathway.

During mating, haploid MAT a and MATαS. cerevisiae cells communicate with each other through secretion of peptide pheromones a- and α-factor, respectively, to which the opposite cell type responds. Pheromone stimulates yeast cells to increase the expression of mating genes, arrest cell division in the G1 phase of the cell cycle, and form polarized mating projections directed toward the pheromone source (5). These responses are initiated by a cell surface receptor that couples to a heterotrimeric G protein and downstream MAPK kinase cascade (Fig. 1A). Three additional MAPK modules are potentially linked to the pheromone response pathway through functional relationships or shared components (Fig. 1A). The protein kinase C (PKC) pathway is activated by cell surface stress, which occurs during formation of the mating projection (6). The HOG pathway responds to hypertonic stress (7). The filamentous growth pathway responds to environmental and nutritional cues to promote haploid invasive growth on rich medium and diploid pseudohyphal development on medium low in nitrogen (8). Although components of the pheromone response pathway, such as Ste20p and Ste11p, form part of the HOG and filamentous growth MAPK pathways (7–9), their activation by high osmolarity or nutritional cues does not initiate a mating response in wild-type (wt) cells, which suggests that these signaling components assemble into functionally independent complexes (7, 9) (Fig. 1A). The pheromone and filamentous growth pathways both activate the transcription factor Ste12p (Fig. 1A). During pheromone signaling, Ste12p activates the promoters of mating genes either as a homomultimer bound to pheromone response elements (PREs) or as a heteromultimer with Mcm1p (5). During signaling of the filamentous growth pathway, Ste12p activates promoters that contain filamentation response elements as a heteromultimer with Tec1p (8).

Figure 1

Simplified schematic of four MAPK pathways in yeast. (A) During mating, peptide pheromones α-factor anda-factor activate the receptors Ste2p and Ste3p, respectively. Receptor stimulation releases free Gβγ (Ste4p/Ste18p) and leads to activation of the MAPK cascade. MAPK Fus3p activates both gene expression, by phosphorylation of the Ste12p transcription factor and its associated negative regulators Rst1p and Rst2p (21), and G1 arrest, by phosphorylation of the cyclin-dependent kinase inhibitor Far1p (19). Cdc42p controls polarized morphogenesis through interactions with Bni1p and other polarity proteins linked to the actin cytoskeleton (20, 24). In the PKC pathway, signaling by Wsc1p/Wsc2p/Wsc3p and Mid2p (30) activates Rho1p, Pkc1p, and a MAPK module that culminates in Mpk1p (and probably Mlp1p), which control at least two transcription factors—Rlm1p and Swi4/Swi6p (30). Activated Rho1p also interacts with Bni1p (20). In the HOG pathway, hypertonic shock activates Sho1p and negatively regulates Sln1p, activating a MAPK module leading to Hog1p-dependent activation of transcription factors including Msn2/4p and Mcm1p (31). In the filamentous (diploid pseudohyphal and haploid invasive) growth pathway, Sho1p (9) and other sensors feed into two parallel pathways, one of which is a MAPK module that activates Kss1p, resulting in phosphorylation of Rst1p/Rst2p and activation of the Ste12p/Tec1p transcription factor complex (8). Not shown is a parallel branch of the filamentous growth pathway involving Ras2p/cAMP signaling (32). (B) Transcriptional reporter genes for activation of the MAPK pathways. FUS1, FUS3, andFIG1 are known pheromone-induced mating genes (14); GPD1 and GPP2 encode proteins required for intracellular production of glycerol (13). Asterisks mark reporter genes identified in this study.MPK1, MLP1, RLM1, andYLR194C were induced under conditions that activate the PKC-regulated MAPK pathway; KSS1, YLRO42C,PGU1, and SVS1 are potential filamentous growth genes induced preferentially by Kss1p signaling.

To characterize the genome-wide changes in transcription that accompany pheromone signaling, DNA microarrays consisting of >97% of the known or predicted genes of S. cerevisiae were probed with differentially labeled (Cy3, green; Cy5, red) cDNA pools derived from pheromone-treated or mock-treated cells (10, 11). The relative abundance of each RNA message was measured by directly comparing the signals from competitive hybridizations, resulting in a red/green expression ratio for each gene. We first examined the pheromone-induced changes in gene expression (response profile) after exposure of MAT a cells to α-factor concentrations ranging from 0.15 to 500 nM for 30 min. The response profile was relatively invariant above the 15.8 nM pheromone treatment, which suggests that the response was saturated; at 50 nM pheromone, expression of more than 200 genes was increased and expression of more than 200 genes was repressed (Fig. 2A) at a 99% confidence level (i.e., <1% chance the expression change was due to measurement errors alone) (4, 11). A genome-wide comparison of gene expression changes resulting from saturating treatments of 15.8 or 50 nM pheromone showed a strong correlation, which indicates that the results were highly reproducible (Fig. 2B) (ρ = 0.98) (12). We also examined the response ofMATα cells to a-factor. Unlike α-factor,a-factor contains a COOH-terminal farnesyl modification and is secreted through a nonclassical pathway (5). Apart from mating-type–specific genes (e.g., ASG7 in acells and MFα2 in α cells) (13), thea- and α-factor response profiles were highly correlated (11) (ρ = 0.92). Pheromone treatment caused increased expression of genes involved at each stage of the mating response (5). A large set of novel pheromone-induced genes was identified (14), and gene deletion experiments have demonstrated that some of these genes are required for efficient mating. For example, mutations that alter the highly conservedHYM1 product cause defects in mating, pheromone-induced polarized morphogenesis, and basal expression of mating genes (15). Expression of genes that promote cell cycle progression, DNA replication, budding, and mitosis appeared to be repressed by pheromone treatment (14).

Figure 2

DNA microarray analysis ofS. cerevisiae pheromone response. (A) α-Factor response profile for the wt MAT a strain R276 (10, 28). Fluorescently labeled cDNAs derived from wt cells either mock treated or treated with 50 nM α-factor for 30 min were hybridized competitively to four DNA microarrays, each containing >97% of all yeast open reading frames (4). Results from the four chips were averaged. Data are displayed as mean intensity [log10(intensity)] versus the red/green ratio [log10(expression ratio)] for each probe (4). Gray dots represent genes that showed no change in expression; genes that were induced or repressed by α-factor treatment, at a 99% confidence level (4,11), are plotted as red stars (237 in total) or green stars (248 in total), respectively (11, 14). Genes labeled in the plots include previously identified (FUS1,FIG1) and novel (YML047C, YPL192C) pheromone-induced genes and mitotic cell cycle S phase–induced genes (HHF1, histone H4, and RNR1, ribonucleotide reductase large subunit). (B) Correlation plot of two different concentrations of α-factor (4, 11). Fluorescently labeled cDNAs derived from wt mock treated or treated with 15.8 nM α-factor for 30 min and wt mock treated or treated with 50 nM α-factor were analyzed on pairs of microarrays. Genes induced by the pheromone treatment are plotted as positive values. Log10 of the expression ratio for each gene derived from the 15.8 nM α-factor treatment is plotted versus log10of the expression ratio in the 50 nM α-factor treatment. Genes whose expression did not change significantly in the two experiments are shown as gray dots; genes whose expression was induced or repressed, at a 99% confidence level, are plotted as colored stars. Red stars correspond to genes whose expression changed, either induced or repressed, in both experiments; green and blue stars correspond to genes whose expression changed only in a response profile plotted on the x and y axis, respectively. Spread- sheets listing the data for all genes shown as colored stars in the correlation plots within Figs. 2 and 3 are provided with the supplementary material (11). Response profile derived from the 30-min, 50 nM α-factor treatment of wt is plotted against the following competitive hybridizations. (C) R1269 (ste2Δ) mock treated versus treated with 50 nM α-factor for 30 min. (D) R418 (ste12Δ) mock treated versus treated with 50 nM α-factor for 30 min. (E) R496 (far1Δ) mock treated versus treated with 50 nM α-factor for 30 min. (F) Kinetic analysis of 50 nM α-factor treatment. Fluorescently labeled cDNAs derived from strain R276 (wt) mock treated versus treated with 50 nM α-factor for 0, 15, 30, 45, 60, 90, and 120 min. At each time point, samples of cells were examined microscopically to quantify the pheromone-induced changes in cell morphology (18). Genes whose expression ratio changed >2-fold in two or more experiments, for which data were obtained at a 99% confidence level in three or more experiments, are plotted (156 genes induced and 100 genes repressed). Most genes are plotted as blue lines and genes of interest are highlighted in red. (G) R276 (wt) mock treated versus treated with 50 nM α-factor for 120 min. Brown stars represent genes whose expression is anticorrelated (i.e., induced in one experiment but repressed in the other or vice versa). Several genes preferentially induced (MPK1, MLP1,RLM1, YLR194C) or repressed (CYK2, EGT2) at 120 min of pheromone treatment are labeled. (H) R994 (bni1Δ) mock treated versus treated with 50 nM α-factor for 120 min. (I) Expression profile derived from Y1469 (wt, carrying empty vector pRS316) versus Y1469 (carryingGAL1/10pr-RHO1-Q68H), both strains treated with 2% galactose for 3 hours, is plotted against the expression profile derived from Y1469 (carrying pRS316) versus Y1469 (carryingGAL1/10pr-PKC1-R398A), both strains treated with 2% galactose for 3 hours.

Analysis of ste2Δ cells, which carry a deletion of the gene encoding the α-factor receptor, revealed that the entire pheromone response profile was specified by the pheromone receptor (Fig. 2C). Several observations indicate that signaling, down to the level of Fus3p, takes place independently of Ste12p activation (6, 16). ste4Δ, ste18Δ,ste5Δ, ste11Δ, ste7Δ,fus3Δ kss1Δ, and ste12Δ mutant cells were all completely defective for pheromone-induced changes in transcription (Fig. 2D) (17, 18). Thus, the entire transcriptional response appears to be derived from pathway-dependent activation of Ste12p. This finding indicates that all functional links between upstream signaling molecules of the pheromone response pathway and other transcription factors must be regulated by a Ste12p-dependent circuit (6) and that disruption of signal transmission at intermediate steps does not result in cross-talk activation of other MAPK pathways (9).

To determine the effect of G1 cell cycle arrest on pheromone-regulated gene expression, we examined cells lacking the cyclin-dependent kinase inhibitor Far1p. far1Δ cells have normal pheromone-dependent activation of Ste12p and increased transcription of FUS1 but do not show G1 cell cycle arrest (19). Comparison of the pheromone response profile of far1Δ cells with the response profile of wt cells showed that essentially all gene repression required Far1p (Fig. 2E). Thus, Ste12p controls repression through increased transcription of FAR1, whose activated product arrests cells at Start in late G1 phase (19). Indeed, most of the pheromone-repressed genes we identified are subject to cell cycle regulation and are expressed preferentially outside G1phase (14).

After transcriptional activation and G1 arrest, the pheromone signal initiates polarized cell growth to form a mating projection or “shmoo.” These morphological changes result from localized cell surface expansion and depend on reorganization of the cortical actin cytoskeleton (20). To determine the changes in gene expression that correlate with projection formation, we exposed wt cells to 50 nM α-factor for up to 2 hours (Fig. 2F). Comparison of the response profile after 120 min with the profile after 30 min revealed a set of transcripts induced to a greater extent at 120 min (Fig. 2G). This set included transcripts from genes encoding MAPKs associated with the PKC pathway—MPK1 andMLP1 (YKL161C)—and a downstream transcriptional effector, RLM1 (Fig. 1A), which suggests that transcription of these genes might be induced by PKC pathway signaling during polarized morphogenesis. To test this possibility, we examinedbni1Δ mutant cells, which are defective in projection formation (20) and substantially delayed for pheromone-mediated activation of the PKC-regulated MAPK Mpk1p (6). Comparison of the pheromone response profile for bni1Δ cells at 120 min with wt cells at 30 min indicated that bni1Δ cells activated Ste12p and initiated G1 arrest but did not show increased transcription ofMPK1 and other late-induced genes (Fig. 2H). Genes that showed increased transcription by extended pheromone treatments also showed increased transcription after overexpression of genes encoding constitutively active forms of Pkc1p and Rho1p from the inducibleGAL1/10 promoter (Fig. 2I). Moreover, many of these genes, including MLP1 and RLM1, are pheromone-induced in an Mpk1p-dependent manner (17). Thus, temporal analysis of pheromone-dependent transcriptional changes identified genes induced by signaling of the PKC-regulated MAPK pathway during mating projection formation (6).

Two partially redundant negative regulators of Ste12p, called Rst1p (Dig1p) and Rst2p (Dig2p), prevent transcription of Ste12p target genes in the absence of signaling (21). Double-mutant cells, rst1Δ rst2Δ, that lack Rst1p Rst2p exhibit constitutive FUS1 expression and invasive growth, which suggests that they express both mating and filamentation genes (Fig. 1A). The transcript profile of rst1Δrst2Δ cells in the absence of pheromone (vegetative profile) showed abundant expression for mating genes without the concomitant repression of cell cycle genes (Fig. 3A). This profile is consistent with constitutive activation of Ste12p below the cell cycle control point at Fus3p (Fig. 1A). A gene set corresponding to potential filamentation genes was identified through their increased expression in the rst1Δ rst2Δ profile when compared with the pheromone response profile and includesPGU1, YLR042c, SVS1, andKSS1 (Fig. 3A). PGU1 and YLR042c also showed increased expression after pheromone treatment of wt cells (Fig. 2B), which suggests overlap between the mating and filamentation responses.

Figure 3

Derepression of mating and filamentation genes in cells defective for Ste12p regulation. Correlation plots of expression profiles for two different microarray experiments are shown in (A), (C), and (D); in each case, the response profile derived from the 50 nM α-factor treatment of wt strain R276 is plotted on the x axis. Data representation is described in Fig. 2legend. (A) Vegetative profile of R4063 (rst1Δ rst2Δ) cells; genes induced in therst1Δ rst2Δ cells relative to wt cells are plotted as positive values. (B) Vegetative profile of R500 (fus3Δ) cells; changes in gene expression are plotted against the mean intensity of each spot (genes labeled as inFig. 2A). (C) R426 (wt) cells mock treated versus R500 (fus3Δ) treated with 50 nM α-factor for 30 min. (D) R426 (wt) cells mock treated versus Y1787 (fus3Δ tec1Δ) treated with 50 nM α-factor for 30 min.

The MAPK that mediates activation of filamentous growth, Kss1p, should induce expression of bona fide filamentation genes regulated by Rst1p and Rst2p. In the absence of Fus3p, Kss1p is subject to activation by both basal and pheromone-induced signaling of the pheromone response pathway (8). To define gene sets induced specifically by Fus3p and Kss1p, we examined profiles derived from mutants defective for either MAPK. Both the vegetative and pheromone response profiles for kss1Δ cells were very similar to those of wt cells (11). Thus, global transcriptional changes directed by Fus3p resemble those observed for wt cells. The vegetative and pheromone response profiles for fus3Δ cells demonstrated that expression of PGU1, YLR042c, SVS1,KSS1, and other potential filamentation genes were induced preferentially by basal (Fig. 3B) or pheromone-induced Kss1p signaling (Fig. 3C) and that increased expression of these genes was Tec1p-dependent (Fig. 3D). The promoter sequences of PGU1,YLR042c, SVS1, and KSS1 directedlacZ expression that was co-regulated with the Kss1p MAPK pathway-specific reporter FG::Ty1-lacZ(11). Thus, a set of filamentation genes defined from therst1Δ rst2Δ profile appear to be under the control of Kss1p-activated Tec1p.

The S288c genetic background used in these experiments carries aflo8 mutation that renders cells defective for both haploid invasive growth and diploid pseudohyphal development in response to environmental stimuli (22). However,rst1Δ rst2Δ mutant cells bypass theFLO8 requirement and invade agar constitutively (21). This invasive phenotype may depend on increased transcription of the filamentation gene set, expression of the correlated pheromone-responsive gene set, or both. Apart from increased expression of the filamentation gene set, the pattern of constitutive gene expression of rst1Δ rst2Δ cells (Fig. 3A) resembles that of pheromone-stimulated far1Δ cells (Fig. 2E). To determine whether the filamentation genes are responsible for the haploid invasive growth of rst1Δrst2Δ cells, we examined whether pheromone could stimulate this process. Surprisingly, pheromone signaling caused cell adhesion and agar invasion of far1Δ cells (Fig. 4). Similarly, for wt cells, agar invasion was stimulated by nonsaturating pheromone doses that were below the threshold for G1 arrest (Fig. 4). Pheromone-induced invasion was also observed for kss1Δ andtec1Δ mutants but not for pheromone response pathway mutants such as ste12Δ and ste2Δ cells (Fig. 4) (23). Thus, it appears that partial activation of pheromone-responsive genes in the absence of G1 cell cycle arrest initiates a developmental program that resembles haploid invasive growth. Because pheromone response is sensitive to gradients (24) and low levels of pheromone cause cells to switch from an axial to a bipolar budding pattern (25), pheromone-induced invasion may enable elongation and growth of dividing cells toward a diffuse pheromone source and thereby facilitate mating among haploid cells within dispersed populations.

Figure 4

Pheromone-induced agar invasion. One microliter of α-factor (1 mM) was applied to a lawn of MAT a bar1Δ cells (R426, wt; R418, ste12Δ; R496,far1Δ; Y1543, tec1Δ) (28) spread on the surface of agar plates containing rich medium. Before washing, plates were incubated for 2 days and then imaged to record the extent of pheromone-induced growth inhibition. After washing under running tap water, the plates were imaged to record the cells that adhered to one another and invaded into the agar substrate.

We performed two-dimensional hierarchical clustering analysis of 46 experiment profiles and 383 genes (represented by 400 microarray probes), whose expression changed more than threefold, at a 99% confidence level, in at least two different experiments. This method identified sets of experiments that lead to similar global patterns of gene expression and sets of genes subject to similar regulation. In the resultant two-dimensional plot (Fig. 5A), the gene cluster tree is represented on the horizontal axis and the experimental cluster tree is on the vertical axis. Patches of red (gene induction) and green (gene repression) identify gene sets that show co-regulation across multiple correlated experiments.

Figure 5

Two-dimensional clustering analysis of experimental expression profiles and gene behavior. (A) Three hundred eighty-three genes whose expression was induced or repressed greater than threefold, at a 99% confidence level, in any two microarray experiments, were clustered by the correlation of their response signatures (33). Forty-six experiments were clustered by the correlation of their signatures. Genes are plotted along the horizontal axis, with the gene cluster tree above. Experiments are plotted along the y axis, with the experiment cluster tree on the outer leftmost side of the plot. Increases in mRNA levels are represented as shades of red, and decreases in mRNA levels are represented as shades of green. (B) Sections of (A) are expanded to allow visualization of specific gene clusters (labeled I to VIII). Experiment names are listed on the left and gene names are at the bottom. Experiment clusters are labeled on the right. Seven experiments that show changes in expression of Kss1p-regulated genes are labeled with arrowheads on the right (27).

Experimental profiles 1 to 31 encompass conditions that activate the pheromone response pathway. Most of these experiments showed increased expression of a pheromone-induced gene set (genes 150 to 265) and decreased expression of a cell cycle–regulated gene set (genes 1 to 121). However, experiments 20 to 22, which include the vegetative profile (unstimulated) of rst1Δ rst2Δ cells and the pheromone response profile of far1Δ cells, showed increased expression of pheromone-induced genes with little or no gene repression. Experiments 26 to 31, which include the response associated with nonsaturating pheromone doses, short pheromone treatments, and mutants compromised for signaling (fus3Δ, fus3-K42R, ste20Δ,rst1Δ rst2Δ), showed increased expression of a subset of pheromone-responsive genes. Experiments 36 and 37 correspond to the overexpression of dominant-activated alleles ofPKC1 and RHO1 and define a candidate cluster of PKC-regulated genes (genes 289 to 378). Experiments 38 to 44 include the vegetative profiles of pheromone response pathway signaling mutants and identify genes regulated by basal signaling of the pathway (26).

Expanded views of selected gene clusters are shown in Fig. 5B. The genes in cluster I were repressed by brief pheromone treatments and include several genes activated early in the cell cycle during the G1-to-S transition (3). In general, the genes in cluster II were repressed by longer pheromone treatments and activated late in the cell cycle during the M-to-G1 transition. Genes in clusters III and V were induced by brief pheromone treatments andSTE12 overexpression but were repressed in pheromone response pathway mutants. The pheromone-responsive genes in cluster VII displayed a more complex pattern of regulation and appear to be activated by Ste12p indirectly through PKC pathway signaling. These genes were induced by extended pheromone treatments (60 to 120 min) and overexpression of the dominant-activated alleles of PKC1 andRHO1, but they were not induced in response toSTE12 overexpression or pheromone treatment ofbni1Δ cells. Consistent with these observations, clusters III and V, but not VII, were enriched for potential Ste12p binding PREs (11). Several different clusters contain genes that appear to be regulated by Kss1p signaling (Fig. 3, B and C). Cluster IV contains PGU1, cluster VI contains YLR042c, and cluster VIII contains KSS1. Indeed, promoter analysis revealed that the genes within these clusters were enriched for Tec1p binding sites (11). The absence of coherent clustering by these genes indicates that each set is subject to additional, unique modes of regulation; however, co-regulation of these genes was observed in seven experiments (Fig. 5B, black arrowheads) (27). Several of the Kss1p-regulated genes may participate in cell surface remodeling during filamentous growth (27).

Genome-wide transcript analysis simultaneously monitors the gene expression programs of all signal transduction pathways. In addition to identifying the gene sets regulated by particular pathways, internal feedback regulatory loops also may be uncovered. For example, expression of genes encoding four of the six yeast MAPKs (Fus3p, Kss1p, Mpk1p, and Mlp1p) is increased by stimulation of their cognate pathways (Fig. 1B). Global profiling also reveals higher order signaling circuits, such as sequential activation of the yeast pheromone response and PKC-regulated MAPK pathways during mating projection formation. Genome-wide transcriptional profiling thus provides a means to trace the signaling mechanisms and circuits that underlie complex biological responses.

  • * These authors contributed equally to this work.

  • To whom correspondence should be addressed. E-mail: boonec{at}biology.queensu.ca and sfriend{at}rii.com

REFERENCES AND NOTES

View Abstract

Navigate This Article