Transient Activation of the HOG MAPK Pathway Regulates Bimodal Gene Expression

See allHide authors and affiliations

Science  06 May 2011:
Vol. 332, Issue 6030, pp. 732-735
DOI: 10.1126/science.1198851


Mitogen-activated protein kinase (MAPK) cascades are conserved signaling modules that control many cellular processes by integrating intra- and extracellular cues. The p38/Hog1 MAPK is transiently activated in response to osmotic stress, leading to rapid translocation into the nucleus and induction of a specific transcriptional program. When investigating the dynamic interplay between Hog1 activation and Hog1-driven gene expression, we found that Hog1 activation increases linearly with stimulus, whereas the transcriptional output is bimodal. Modeling predictions, corroborated by single-cell experiments, established that a slow stochastic transition from a repressed to an activated transcriptional state in conjunction with transient Hog1 activation generates this behavior. Together, these findings provide a molecular mechanism by which a cell can impose a transcriptional threshold in response to a linear signaling behavior.

Mitogen-activated protein kinase (MAPK) cascades orchestrate many cellular processes including cell growth, division, and differentiation (1). In Saccharomyces cerevisiae, the high osmolarity glycerol (HOG) pathway is needed to reestablish the balance between internal and external pressures upon osmotic shock (2). Osmosensors at the cell membrane activate either the MAPK kinase kinases (MAPKKKs) Ste11 or Ssk2,22, which converge on the MAPKK Pbs2. In turn, Pbs2 doubly phosphorylates the MAPK Hog1, leading to rapid translocation into the nucleus to launch a transcriptional program. Although increased transcription is essential to survive very high osmotic stress (0.8 M NaCl), it is not required for milder stress conditions (0.4 M NaCl) (3), under which Hog1 kinase activity alone is sufficient to drive cellular adaptation. By contrast, in the yeast mating MAPK pathway, transcription and new protein expression are required for cell cycle arrest and mating (4).

Transcriptional activation of mating genes occurs with linear kinetics and high fidelity (5, 6), and the observed cell-to-cell variation in protein expression is governed by the ability of cells to express proteins (expression capacity) (5). Whereas the mating pathway can be compared to a cell-fate decision system with sustained MAPK activity, the HOG pathway is an adaptation response, which is only transiently induced like other stress-activated pathways (7). We therefore investigated whether this transient response would trigger different expression behavior.

To quantify the transcriptional output induced by osmotic stress, we engineered a reporter system based on a quadruple Venus (qV) fluorescent protein expressed under the control of specific osmostress-inducible promoters dependent on the three main transcription factors orchestrating the transcriptional response to osmotic stress (Hot1 and Sko1: pSTL1; Msn2,4: pALD3; or Msn2,4 and Hot1: pHSP12) (8). Flow cytometry revealed a Pbs2-dependent 20-fold increase in pSTL1-qV reporter expression when 0.4 M NaCl was added to the growth medium (Fig. 1, A and B). No expression was detected at low salt concentrations (below 0.05 M), while above 0.15 M, all cells expressed the reporter and the amount of fluorescence increased linearly with stress. However, at intermediate concentrations, we observed histograms with two distinct subpopulations representing nonexpressing cells with basal autofluorescence levels and expressing cells with higher intensities. These distributions are termed bimodal. The pALD3-qV and pHSP12-qV reporters displayed a similar bimodal expression behavior (Fig. 1B and fig. S1A). Induction of the mating pathway for 45 min with α-factor also generated a bimodal expression output of the Ste12-specific reporter pFIG1-qV. However, signaling in the mating pathway is prevented from “start” through S phase (9), and expression output became unimodal after relieving this cell cycle–dependent restriction (Fig. 1B and fig. S1B).

Fig. 1

Bimodal expression of fluorescent reporters upon osmotic stress. (A) Dose response of the quadruple-Venus (qV) fluorescence reporter driven by the STL1 promoter (pSTL1) measured by flow cytometry. (B) Dose responses of wild-type or pbs2Δ cells harboring the indicated osmostress-inducible reporters driven by the STL1, ALD3, or HSP12 promoters, or α-factor–inducible reporter pFIG1-qV in a cdc28-as background with 1-NM-PP1 inhibitor (10 μM) or dimethyl sulfoxide (DMSO). The mean of the log-normal distribution fitted to the flow cytometry histograms is plotted. The best fit between single or double log-normal distributions was selected for each curve. The open and closed circles represent, respectively, the population of nonreacting and reacting cells. Circle size is proportional to the population under each distribution. (C and D) Intrinsic and extrinsic noise revealed by microscopy in a strain that contains both CFP (cyan fluorescent protein) and YFP (yellow fluorescent protein) expression reporters driven by the pSTL1 promoter stressed with 0.1 M (red) or no (black) NaCl (C), or in cdc28-as cells inhibited by 1-NM-PP1 with expression reporters driven by the pFIG1 promoter treated with 0.03 μM (red) or no (black) α-factor (D). (E) Percentage of intrinsic (gray) and extrinsic (white) noise over total noise quantified for osmotic stress or α-factor treatment in cells bearing two pSTL1 or pFIG1 expression reporters, respectively.

To investigate the source of the HOG pathway bimodal expression behavior, we integrated two reporters driving the expression of a quadruple cyan fluorescent protein (qCFP) and a qV construct in the same cell. Correlation of the cyan and yellow intensities measures the contribution of cell-to-cell (extrinsic) and intracellular (intrinsic) variability to the overall expression noise (5, 10). The two pFIG1 reporters induced by α-factor demonstrated that the mating pathway is governed by extrinsic noise. By contrast, we observed a lack of correlation between the two pSTL1 reporters (Fig. 1, C to E, and fig. S2), demonstrating that the bimodal expression behavior of the HOG pathway is independent of cell-to-cell variability caused by extrinsic factors such as expression capacity or cell-cycle stage.

To assess the observed bimodality and Hog1 signaling simultaneously, we combined a Hog1-relocation assay (11, 12) with the pSTL1-qV expression reporter (Fig. 2A). Because nuclear accumulation of Hog1 is linked to its kinase activity (13), this assay allows one to correlate in each cell the signaling and expression outputs (Fig. 2, B and C). When cells were stressed with increasing salt concentrations, Hog1 nuclear accumulation gradually augmented both in magnitude and retention time. At the single-cell level, a clear discrepancy was apparent between the linear increase in signaling output versus the bimodal behavior observed in the expression output (Fig. 2, D to F, and fig. S3). We conclude that Hog1 activation as measured by its nuclear translocation is not sufficient to induce a defined transcriptional output, indicating that an unknown intracellular factor (or factors) sets a threshold for gene expression.

Fig. 2

Comparison between Hog1 nuclear relocation and pSTL1-qV reporter expression in single cells. (A) Wild-type cells expressing the nuclear marker Hta2-CFP, Hog1-mCherry, and the pSTL1-qV expression reporter imaged before or after addition of 0.4 M NaCl. (B and C) Quantification of single-cell traces; the integral below the nuclear accumulation curve is used as a measure of the signaling output (B). The difference between final and initial average intensity allows quantification of the expression output (C). (D) The signaling and expression outputs were quantified in single cells exposed to 0.1 M (red) or no (black) NaCl. (E and F) Traces of three single cells marked in (D) displaying almost identical Hog1 relocation dynamics (E) but different expression outcomes (F).

Hog1 has been implicated at various steps in the complex mechanisms leading to gene transcription (14, 15). First, Hog1 associates with transcription factors that bind at specific promoters (16). The MAPK then recruits RNA polymerase II (Pol II) as well as chromatin-remodeling complexes such as SAGA (Spt-Ada-Gcn5 acetyltransferase) and RSC (chromatin structure remodeling), which evict nucleosomes (1719). During active transcription, the INO80 complex and histone chaperones are involved in redeposition of histones, and therefore help in silencing these genes once stress has been overcome (20).

To test if bimodality is reflected in chromatin remodeling, we used chromatin immunoprecipitation (ChIP) to monitor the occupancy of histone H3 on the STL1, HSP12, and ALD3 promoters. Histone eviction occurred in a Hog1-dependent manner (fig. S4A) (18) and was complete from 0.15 M (STL1, ALD3) or 0.2 M NaCl (HSP12) (Fig. 3A). The partial eviction observed at low stress levels suggests that only a fraction of the population could remodel chromatin to allow for efficient transcription. In contrast to other transcription regulators such as Asf1, Cyc8, or Htz1, the bimodal behavior of the pSTL1-qV stress reporter was already present in the absence of stress in cells deleted for the INO80 subunit Arp8 (Fig. 3B; fig. S5, A and B; and table S1). Conversely, bimodality was markedly increased in cells with impaired SAGA or RSC function (Fig. 3B and fig. S6), indicating that chromatin remodeling activity affects the threshold of gene expression. We verified that eviction of histone H3 is incomplete in gcn5∆ cells (Fig. 3C and fig. S4B), reinforcing the notion that the partial histone eviction observed at the population level is linked to the bimodal expression measured in single cells.

Fig. 3

Influence of chromatin remodeling on reporter expression. (A) Dynamics of histone H3 eviction at the STL1, ALD3, and HSP12 promoters was measured by quantitative ChIP experiments after addition of the indicated NaCl concentration. Histone H3 binding was normalized to a TEL2 sequence control. The error bars correspond to the standard deviation of three independent measurements. (B) Mean of the log-normal fit of flow cytometry histograms of pSTL1-qVenus expression quantified in wild-type (wt), arp8∆, and gcn5∆ cells after addition of NaCl. (C) Histone H3 occupancy in wt and gcn5∆ cells was quantified as in (A) 5 min after osmotic stress.

Deletion of either of the two transcription factors Sko1 or Hot1 strongly reduced pSTL1-qV expression and led to a bimodal expression pattern at high stress levels (fig. S5C). This behavior could be partially rescued by the additional deletion of ARP8. Moreover, cells grown at low glucose concentration (0.05%), where glucose repression is alleviated, display a bimodal transition around 0.05 M NaCl (fig. S7, A to C). The bimodal transition shifts to higher stress levels as glucose repression increases with the amount of glucose in the medium. Together, these results suggest that the bimodality depends on a number of dynamic processes cooperating at stress-induced promoters to regulate the activation of the transcription.

To better understand the dynamics of gene activation, we designed a simple stochastic model of Hog1-induced transcription (fig. S8 and SOM Text), which identified the formation of an active gene complex (i.e., a gene with open chromatin, which can be efficiently transcribed) as the crucial step governing the bimodal expression pattern. As predicted by the model, a transient activation of the HOG pathway for specific lengths of time with high stresses revealed a bimodal distribution of pSTL1-qV reporter expression at early time points (Fig. 4, A and C). Microscopy analysis confirmed that these conditions result in short-lived nuclear relocation of Hog1 in all cells and a bimodal expression response of the population (Fig. 4, D and E). Conversely, sustained activation of Hog1 with low stresses with a ramping protocol resulted in a transition from nonexpressing to fully expressing cells going through a bimodal stage (Fig. 4, B to E). We conclude that both the retention time and concentration of Hog1 in the nucleus are critical parameters that control bimodality of the transcription of stress-activated genes.

Fig. 4

Duration and intensity of Hog1 nuclear accumulation controls bimodality. (A to C) Flow cytometric measurement of pSTL1-qV expression and mean of log-normal fits of the histograms (C) upon transient or sustained activation with a pulse of 0.2 M NaCl (A) or a ramp from 0.05 to 0.4 M NaCl (B) Cycloheximide (CHX) is added after 45 min to block protein synthesis. (D) Modulation of the Hog1 activation pattern in flow chambers studied by microscopy. Mean Hog1 nuclear accumulation upon transient (0.4 M NaCl for 3 min, green) and sustained (from 0.075 to 0.4 M NaCl in 20 min, red) activation of the pathway. Stepwise activation (dashed lines) for 0.075 M (red), 0.1 M (blue), 0.2 M (cyan), and 0.4 M NaCl (green) are shown for comparison. (E) Average signaling (red) and expression outputs (blue) for different activation patterns. The open and closed circles represent, respectively, the population of nonreacting and reacting cells. Circle size is proportional to the population under each distribution. The error bars show the standard deviation of 40 to 130 single cells.

Stress genes must fulfill two contradictory requirements, which may explain their large noise in expression (21, 22). First, under normal growth conditions these genes are silenced, although basal pathway activity can be present (23, 24). Second, upon stress, these genes must be expressed at a high rate to contribute to the adaptation of the cell during the short period of activity of the pathway. Notably, bimodal gene expression may be a general feature of stress-induced genes, because oxidative or heat stresses also generated a bimodal expression pattern (fig. S9).

If the approximately 300 genes induced by osmotic stress (8) were expressed in a stochastic fashion, a unique set of 150 genes would be present in each cell. Fluorescence reporter expression can therefore not be linked to increased resistance to osmotic stress, and we failed to detect a pattern in reporter expression in cells subjected to two subsequent mild osmotic stresses (fig. S10). However, this broad diversity in expression pattern will result in a large variability in protein content of the cells, which could be advantageous to survive in changing environments (25). Thus, noise in stress gene expression may confer an evolutionary advantage to a population by increasing fitness to face a large range of stress events.

Supporting Online Material

Materials and Methods

SOM Text

Figs. S1 to S11

Tables S1 to S5


References and Notes

  1. Acknowledgments: We thank R. Dechant, C. Schüller, G. Ammerer, A Colman-Lerner, and A. Smith for strains, plasmids, and helpful discussions, and S.-S. Lee and H. Koeppl for help with flow chamber experiments and model design, respectively. We are grateful to C. Rupp and D. Condé for technical assistance. This work was supported by QUASI, UNICELLSYS, the organization, and the Competence Centre “Systems Physiology and Metabolic Disease” (CC-SPMD). M.N.-R. is supported by ISCIII, and F.P is recipient of an ICREA Acadèmia (Generalitat de Catalunya) award. Work in the Posas and de Nadal laboratories is funded by the Fundación Marcelino Botín (FMB) and the Ministerio de Ciéncia y Innovación (BFU2008-00530 to E.N. and BIO2009-07762 to F.P.). The Peter laboratory is supported by the Swiss National Science Foundation and ETHZ. The authors declare that they have no competing financial interests.
View Abstract

Navigate This Article