|Home | About | Journals | Submit | Contact Us | Français|
Signaling through mitogen-activated protein kinase (MAPK) cascade pathways can show various input-output behaviors, including either switch-like or graded responses to increasing levels of stimulus. Prior studies suggest that switch-like behavior is promoted by positive feedback loops and nonprocessive phosphorylation reactions, but it is unclear whether graded signaling is a default behavior or if it must be enforced by separate mechanisms. Scaffold proteins have been hypothesized to promote graded behavior.
Here, we experimentally probe the determinants of graded signaling in the yeast mating MAPK pathway. We find that graded behavior is robust, as it resists perturbation by loss of several negative feedback regulators. However, the pathway becomes switch-like when activated by a crosstalk stimulus that bypasses multiple upstream components. To dissect the contributing factors, we developed a method for gradually varying the signal input at different pathway steps in vivo. Input at the beginning of the kinase cascade produced a sharp, threshold-like response. Surprisingly, the scaffold protein Ste5 increased this threshold behavior when limited to the cytosol. However, signaling remained graded whenever Ste5 was allowed to function at the plasma membrane.
The results suggest that the MAPK cascade module is inherently ultrasensitive, but is converted to a graded system by the pathway-specific activation mechanism. Scaffold-mediated assembly of signaling complexes at the plasma membrane allows faithful propagation of weak signals, which consequently reduces pathway ultrasensitivity. These properties help shape the input-output properties of the system to fit the physiological context.
Eukaryotic cells use mitogen-activated protein kinase (MAPK) cascades to respond to a wide variety of stimuli . Due to this diversity, the core MAPK cascade module may need to adopt different input-output signaling behaviors that can be tailored to the physiological context. For example, some MAPK pathways react to increasing levels of stimulus in an all-or-none, or "switch-like", manner (Figure 1A), where individual cells in the population are either "on" or "off", with no intermediate response . By converting the analog input into a binary output, this signaling behavior can facilitate all-or-none cellular decisions that control irreversible changes in biological states such as cell cycle progression [2, 3], as well as responses to stress .
Other MAPK pathways show “graded” responses (Figure 1A), where all cells in the population uniformly increase their signal output in proportion to the level of stimulus. This behavior is observed during pheromone response in the S. cerevisiae mating pathway [5, 6], despite its use of a four-kinase cascade that is expected to be intrinsically more switch-like than the usual three-tier cascade . Nevertheless, graded signaling suits the physiological characteristics of the mating response, including its reversibility (i.e., growth resumes when pheromone is removed) and the activation of distinct responses at different threshold doses over a wide range [8, 9]. Similarly, metazoan growth factors such as EGF and PDGF also trigger graded MAPK signaling [10, 11], and the intermediate responses permitted by graded signaling may allow individual MAPK pathways to trigger different cell fates during development . While seemingly simpler than switch-like signaling, graded signaling is not necessarily the default behavior of MAPK pathways. Furthermore, unique challenges may be posed by the need for accurate signal transmission over a wide range of amplitudes.
Theoretical and biochemical studies reveal that switch-like signaling can be promoted by positive feedback loops [2, 13] and by non-processive phosphorylation of kinases that must be phosphorylated multiple independent times [2, 3, 14]. Each mechanism contributes to a sigmoidal dose-response relationship termed “ultrasensitivity”, in which a small change in the level of input stimulus near a threshold value produces a dramatic change in signaling output. In principle, therefore, graded signaling could result from the simple absence of these mechanisms, or from the presence of separate mechanisms that counteract a default tendency toward threshold behavior.
Many MAPK pathways incorporate “scaffold” proteins that bind multiple kinases in the cascade [15, 16]. It has been hypothesized that scaffold proteins might counteract the switch-like tendencies of MAPK cascades, because assembly of a multi-kinase complex could allow multiple phosphorylations to proceed processively, without intervening steps of dissociation [16–19]. In the yeast mating pathway (Figure 1B), Ste5 is a prototypical scaffold protein that is essential for MAPK pathway signaling [15, 20, 21]. In addition to providing binding sites for multiple pathway kinases, Ste5 co-recruits these kinases to the plasma membrane in response to a pheromone-activated receptor and heterotrimeric G protein [22–24]. Of special relevance to the present study, membrane translocation of Ste5 promotes both the initial activation of the kinase at the top of the pathway (the MAPKKK Ste11) as well as the propagation of signal through the remainder of the kinase cascade . However, it is unknown whether Ste5, or any MAPK scaffold, plays a major role in promoting a graded mode of signaling. Furthermore, it has been hypothesized that scaffold proteins could equally likely make signaling more switch-like, rather than less so .
Here we dissect the determinants of graded signaling in the yeast pheromone response pathway, using single-cell assays and a variety of tools for modulating signaling input at different pathway steps. We find that graded behavior is robust, as it resists genetic perturbation. However, the core MAP kinase cascade module embedded within this pathway is not inherently graded. Instead, it is inherently ultrasensitive. Contrary to expectation, the mere presence of the Ste5 scaffold protein in the cytoplasm increased, rather than decreased, this ultrasensitivity. However, the opposite effect was observed when Ste5 functioned at the membrane, which promoted graded signaling. These findings suggest that scaffold-mediated assembly of signaling complexes at the plasma membrane helps enforce graded signaling by counteracting threshold behavior intrinsic to the MAPK cascade. This may result from enhanced signal propagation that converts low levels of input into a proportional level of output. These properties help shape the input-output behavior of the system in a physiologically appropriate manner.
Switch-like and graded signaling can be indistinguishable at the population level , and so single-cell analysis is essential. As in earlier studies [5, 6, 26, 27], we expressed GFP from the promoter of a pheromone-induced gene, FUS1 (Figure 1C), and then measured pheromone response in single cells by FACS. In wild-type cells, PFUS1-GFP fluorescence increased in response to pheromone in a gradual, uniform manner (Figure 1D). That is, despite some population variance at any given concentration , increased pheromone levels caused the majority of cells to gradually shift to greater intermediate levels of signaling. This behavior confirms that the mating pathway response is indeed graded, as described by Poritz et. al.  and subsequent studies [6, 26].
Because positive feedback loops can promote switch-like signaling , negative feedback could have the opposite effect and hence promote graded signaling. Therefore, we tested the role of three classes of negative regulators. First, we tested two proteins that act at the top of the pathway. Bar1 is a protease that degrades α factor, whereas Sst2 inactivates the heterotrimeric G protein . Because expression of each gene is induced by pheromone , they are part of a negative feedback loop. Nevertheless, despite clear shifts in the effective dose of pheromone, graded signaling was still observed in both bar1Δ and sst2Δ cells, and hence neither protein is required for this behavior (Figure 1E). Second, the MAP kinases Fus3 and Kss1 are positive components of the mating pathway, but they also participate in negative feedback loops that attenuate signaling at upstream steps in the pathway [28–30]. Despite some effects on variance within the population, signaling in both fus3Δ and kss1Δ cells remained graded (Figure 1F), consistent with results obtained via microscopy [6, 27]. Finally, the tyrosine phosphatases Ptp2 and Ptp3, as well as the dual-specificity phosphatase Msg5, negatively regulate mating pathway signaling by dephosphorylating kinases in the MAPK cascade . Yet mutants lacking one, two, or all three phosphatases still showed graded signaling (Figure 1G). Collectively, these results indicate that no single negative regulator of the pathway is solely responsible for the graded mode of signaling, and therefore the graded behavior is robust to genetic perturbation.
Previous studies theorized that scaffold proteins may promote graded signaling behavior by increasing phosphorylation processivity within the kinase cascade (see Introduction). Although the scaffold protein Ste5 is ordinarily critical for mating pathway signaling, this requirement can be bypassed via the phenomenon of “crosstalk” between the mating and the high osmolarity glycerol (HOG) MAPK pathways, which share some common components (Figure 2A). These two pathways are normally insulated from each other, but treatment of hog1Δ mutant cells with high osmolarity will activate the mating pathway, and this crosstalk requires the mating pathway kinases but not Ste5 . When hog1Δ ste5Δ mutants carrying the PFUS1-GFP reporter were treated with sorbitol, the response was heterogeneous: at intermediate concentrations the cell population bifurcated into responding and non-responding groups, visible as two peaks in the FACS profile (Figure 2B). Thus, crosstalk signaling is switch-like. Importantly, in hog1Δ STE5 cells, which can respond to either stimulus, signaling was graded in response to pheromone but not in response to sorbitol (Figure 2C). Therefore, the response behavior is dictated by the stimulus, not by the mere absence of Hog1.
Because the precise mechanism by which cells measure high osmolarity remains unclear , we cannot rule out that the switch-like behavior is generated at the level of the initial stimulus sensor. That is, if cells do not sense the magnitude of osmolarity change but only whether it exceeds a certain setpoint (e.g., like a thermostat), then the osmotic stimulus itself may be inherently switch-like. Indeed, the population-averaged response to sorbitol rises very suddenly over a narrow dose range, unlike pheromone response (Figure 2C, bottom). A similar caveat may apply to examples where mammalian cells show switch-like responses to stress stimuli . Nevertheless, our results emphasize two important points. First, our ability to observe switch-like signaling shows that the normal graded behavior does not simply reflect a technical limitation of the transcription-based assay system. Second, a single MAP kinase cascade can mediate either graded or switch-like responses, depending on the nature and mechanism of pathway input.
To further probe the critical determinants of graded signaling behavior, we developed a method for varying the level of input at different steps in the mating pathway by using a dose-dependent expression system (Figure 3A). Here, active pathway components were expressed from the GAL1 promoter, and expression levels were controlled by a hybrid transcription factor (Gal4DBD-hER-VP16) whose activity can be varied over a wide range by adding the exogenous hormone β-estradiol . We first established that increasing doses of β-estradiol gave proportional and uniform increases in gene transcription from the GAL1 promoter, as evidenced by graded expression of a PGAL1-GFP reporter (Figure 3B). Subsequently, we used the GAL1 promoter to express various activators of the mating pathway (Figure 3C), and then used the PFUS1-GFP reporter to measure pathway signaling as a function of β-estradiol dose, in the absence of added α factor. This method could trigger maximum signaling levels comparable to those induced by galactose (Figure 3D).
When signaling was triggered near the top of the pathway by graded expression of the Gβ subunit (Ste4) or a membrane-targeted form of Ste5 (Ste5-CTM), we observed a graded response (Figure 4A), similar to the normal behavior induced by pheromone. In contrast, when signaling was triggered by expression of pre-activated versions of the MAPKKK Ste11 (Ste11-4 or Ste11-Asp3), we observed a threshold-like response in which there was a sudden transition from very weak to very strong signaling output (Figure 4B), indicative of ultrasensitivity . Curve-fitting analysis of multiple trials showed that the Ste11-4 and Ste11-Asp3 experiments follow a sigmoidal dose-response profile with a Hill coefficient (nH) of ~4, which is close to the nH ~ 5 behavior observed for the Xenopus MAPK cascade in vitro . This contrasts with the nH ~1 behavior observed during pheromone treatment or graded expression of Ste4 and Ste5-CTM. Thus, the mating MAPK pathway can show either ultrasensitive or graded behavior, depending on the position in the pathway where input is modulated.
Additional experiments addressed possible explanations for these results. First, because overexpression of constitutively-active forms of Ste11 can bypass the requirement for Ste5 , it seemed possible that the ultrasensitive response to active Ste11 could be a consequence of scaffold-independent signaling [17–19]. In fact, although Ste5 was not required for signaling at the highest levels of Ste11-4 or Ste11-Asp3 expression, Ste5 increased signaling efficiency at lower expression levels (e.g., at 3–10 nM β-estradiol; Figure 4C). Contrary to expectations, however, the presence of Ste5 made signaling by the active Ste11 alleles more threshold-like rather than less so (nH ~ 4 vs. nH ~ 2). Thus, in these experiments Ste5 clearly affects the propagation of signal from active Ste11, but its effect on ultrasensitivity is opposite to prior theoretical predictions [17–19]. Note that these effects of Ste5 were likely mediated in the cytoplasm because no stimulus (e.g., pheromone) was added to recruit Ste5 to the plasma membrane. Indeed, the ability of cytoplasmic Ste5 to enhance signaling at mild levels of Ste11-4 expression (e.g., 3 nM β-estradiol) was independent of domains that regulate membrane and nuclear localization, or MAPK binding, but it required intact MAPKKK and MAPKK binding domains (Figure 4D); this may imply that signaling is limited by the efficiency of the Ste11 → Ste7 phosphorylation step(s), which in turn could explain why Ste5 is dispensable when activated forms of Ste11 are highly overexpressed. Despite the dispensability of its MAPK-binding domain, Ste5 was specifically required for activation of the MAPK Fus3, as in prior studies [36, 37], and this occurred at β-estradiol doses that coincided with the transition to strong signaling (Figure 4F and Figure S1A).
Next, we wondered if the threshold-like behavior seen during graded expression of active Ste11 might result from changes in the stoichiometric ratio of Ste11 to its substrate(s). However, this is unlikely because expression of a different form of Ste11, a membrane-targeted derivative (Ste11-Cpr) whose signaling requires membrane-localized Ste5 , produced a graded response (Figure 5A). Finally, we asked whether regulation of the very first phosphorylation step (Ste20 → Ste11) might be required for graded signaling, so that constitutively activating mutations in Ste11 are prohibitive. This possibility was ruled out by using ste20Δ cells that express Ste11-Asp3 at native levels, in which Ste20 is bypassed but pheromone is still required for efficient signal propagation ; in such cells, pheromone response remained graded (Figure 5B). Further controls showed that the graded signaling by Ste5-CTM was independent of both endogenous Ste5 and the N-terminal region of Ste5 involved in Gβγ-binding and dimerization (Figure 5C), although endogenous Ste5 did affect the level of Fus3 activation (Figure S1). In addition, we found that the ability of Ste11-4 and Ste11-Asp3 to activate the HOG pathway [39–41], which can antagonize the mating pathway [41, 42], contributed to the heterogeneity of their signaling responses in ste5Δ cells (i.e., broad FACS histograms in Figure 4B) but did not otherwise contribute to their threshold behavior (Figure 5D and unpublished observations).
Analysis of the collective results revealed that the involvement of membrane-localized Ste5 (rather than simply the presence of Ste5) was common to all experiments that yielded graded behavior. This includes signaling by pheromone (both in wild-type cells and in ste20Δ STE11-Asp3 cells) and by graded expression of Ste4, Ste5-CTM, or Ste11-Cpr. In contrast, membrane-localized Ste5 does not participate in settings that yield switch-like or ultrasensitive signaling, such as hyperosmotic crosstalk and graded expression of Ste11-4 or Ste11-Asp3. We previously found that active forms of Ste11 show relatively weak signaling on their own, but their signaling efficiency is substantially enhanced when Ste5 is recruited to the membrane . This enhancement effect may simultaneously promote a graded response by allowing signaling output to increase in direct proportion to the amount of active Ste11, rather than requiring a threshold level of Ste11 activity to be surpassed (see Discussion).
Lastly, because non-processive phosphorylation reactions promote ultrasensitivity [2, 3], we also analyzed signaling when Ste11-4 was directly fused to its downstream substrate, the MAPKK Ste7 . In principle, this could increase processivity of the Ste11 → Ste7 phosphorylation reactions by preventing their dissociation. Interestingly, ultrasensitivity was mildly reduced (from nH =1.9 to nH = 1.6), but most striking was the acquisition of relatively uniform intermediate responses at intermediate expression levels (Figure 5E). This result is consistent with the prediction that increasing processivity can reduce ultrasensitivity, though we could not directly monitor processivity. Nevertheless, the contrast between this behavior and the increased ultrasensitivity effect of cytoplasmic Ste5 may suggest that cytoplasmic Ste5 is incapable of promoting processive phosphorylation in the manner that had been predicted (see below).
This study probes how scaffold proteins and subcellular compartmentalization influence the input-output behavior of a common signaling module, the MAPK cascade. Our results, along with those from previous studies [2, 3], suggest that the MAPK cascade module is inherently ultrasensitive. In turn, specific mechanisms can either enhance or counteract this default tendency in order to generate switch-like or graded responses, respectively (Figure 6A). A general model that can explain the graded behavior of the yeast mating pathway relates to the effect of Ste5 membrane recruitment on signal propagation through the kinase cascade . Specifically, because the active form of the MAPKKK Ste11 on its own is relatively inefficient at signaling, it must accumulate to a high threshold level before any significant output occurs. But membrane recruitment of Ste5 enhances propagation of signal from active Ste11 through the kinase cascade (Figure 6B, top), thus allowing low levels of Ste11 activity to produce some output signal. This broadens the range of input levels that can yield a measurable output, making signaling less ultrasensitive and more graded (Figure 6B, bottom).
While it is reasonable to assume that the level of active Ste11 increases roughly in linear proportion to the dose of pheromone stimulus, this appears neither necessary nor sufficient to ensure a graded output. It is not sufficient because cells in which the level of constitutively-active Ste11 is gradually increased (e.g., PGAL1-STE11-Asp3 cells treated with β-estradiol) show an ultrasensitive response. It is not necessary because cells expressing a constant, low level of constitutively-active Ste11 still show a graded response to pheromone (e.g., ste20Δ STE11-Asp3 cells treated with α factor). Thus, an important determinant of graded behavior is the manner in which pheromone and Ste5 enhance the steps subsequent to Ste11 activation. The next phosphorylation reaction (Ste11 → Ste7) is likely to be the remaining rate-limiting step, because strong signaling is achievable by simply overproducing constitutively-active Ste11 (i.e., unlike upstream components such as Ste20 and Cdc42; see ). Furthermore, signaling methods that show graded behavior require domains in Ste5 that promote both membrane recruitment and binding to Ste11 and Ste7 (see Figure S2). Thus, it is likely that Ste5 membrane recruitment directly (and separately) promotes both the Ste20 → Ste11 and Ste11 → Ste7 steps. It is less clear whether the final Ste7 → Fus3 step is also stimulated by Ste5 membrane recruitment, as Fus3 activation in vivo is highly dependent on Ste5 [36, 37] and yet the Fus3 binding site in Ste5 is not required ( and Figure S2); hence, the precise role of Ste5 in Fus3 activation remains mysterious.
The surprising finding that Ste5 molecules in the cytoplasm cannot reduce ultrasensitivity may indicate that cytoplasmic Ste5 is incapable of promoting processive phosphorylation, contrary to most prior expectations [16–19]. Why would this be so? A simple explanation would be that the common view of scaffolds—in which they are fully occupied with kinases that efficiently interact with each other while bound to a single scaffold molecule—is incorrect. Instead, it may be the case that most Ste5 molecules in the cytoplasm are incompletely occupied with kinases (Figure 6C, left). This view was postulated previously based on both experiment and theory [25, 37, 43], and is supported by recent evidence using fluorescence cross-correlation spectroscopy [24, 44]. Hence, cytoplasmic scaffolds may largely influence kinases only one at a time, such as by directly modulating their activity , which would negate a role in fostering processivity.
Several possible molecular models could explain how the assembly of scaffolded signaling complexes at the membrane might reduce ultrasensitivity. Membrane recruitment could increase occupancy of the scaffold (Figure 6C, middle), which has been suggested by recent quantitative microscopy [24, 44], or it could promote signaling in trans between kinases bound to different scaffold molecules (Figure 6C, right), which has been detected indirectly by complementation between co-expressed Ste5 mutants . Either mechanism could now permit the scaffold to promote processive phosphorylation reactions largely as was previously assumed to occur on single, cytoplasmic molecules [16–19]. More complex alternative models are also possible (Figure S3). Although our results do not distinguish among these scenarios, they highlight the notion that the relevant molecular context in which scaffold-mediated signaling occurs is likely to be fundamentally different from the simplest models involving fully-occupied scaffolds in the cytoplasm.
A variety of theoretical studies considered the possible effects of scaffolding on the input-output behavior of MAPK cascades [16–19, 43, 46], but none of them predicted the experimentally-observed behavior in which pathway ultrasensitivity can be either increased or decreased depending on whether the scaffold is cytoplasmic or membrane-associated, respectively. However, recent mathematical simulations suggested that confinement of signaling proteins into membrane-localized “nanoclusters” may promote graded signaling through the mammalian Raf-MEK-ERK cascade . While it is unknown if analogous clustering structures exist in yeast, the broader impact of each set of findings is that the assembly of signaling complexes at the plasma membrane can have profound effects on the input-output behavior of a pathway. Modulation of these effects can allow cells to tune the systems-level properties of the signaling pathway in a manner that optimally suits the biological phenomenon being controlled.
Temporal dynamics could also influence input-output behavior. A recent study suggests that the duration of pheromone pathway signaling is sensitive to the dose of input stimulus . Although the responsible mechanism is unknown, negative feedback loops could contribute. Yet removing individual feedback loops did not eliminate either transient signaling  or graded responses (Figure 1). Moreover, signaling induced by β-estradiol-regulated pathway activators is persistent, not transient (Figure S1B), and yet the output response can still be graded (see Figure 4). Hence, multiple overlapping control mechanisms may make graded signaling robust.
Details regarding yeast strains, plasmids, and genetic methods are provided in the Supplemental Data.
Log phase cultures growing in YPD were mixed with an equal volume of YPD containing α factor, sorbitol, or β-estradiol at twice the desired final concentration, and incubated at 30°C for 2 hrs. Cells were harvested by centrifugation, resuspended in phosphate buffered saline, dispersed by sonication, and chilled on ice without fixation. GFP expression was measured using a Becton-Dickinson FACScan flow cytometer (10,000 cells per condition). Representative experiments are shown. To plot dose-response profiles combining results from multiple experiments, and to calculate Hill coefficients, the modal fluorescence from each stimulus dose was first normalized to that observed at the maximum (saturating) dose. The data were then analyzed by non-linear least-squares fitting to a modified Hill equation ;
where Y = fluorescence at stimulus dose X, Bottom = minimum fluorescence (without stimulus), Top = maximum fluorescence, X = stimulus concentration, K = stimulus concentration giving half-maximal response, and n = Hill coefficient (nH). Fitted response curves were overlaid onto the observed data points (mean ± SD) using Prism 4 software (GraphPad Software, Inc.).
To monitor FUS1-lacZ induction, cells growing in YPD or selective media were induced with α factor, sorbitol, or β-estradiol as described above, or by adding galactose (2% final) to cells growing in 2% raffinose medium. One ml of culture was collected by centrifugation, and β-galactosidase assays were performed on cell lysates [22, 25].
Cell extracts were prepared by glass-bead lysis of frozen cell pellets directly in trichloroacetic acid solution, and assayed for phosphorylated MAPKs using rabbit anti-phospho-p44/42 antibodies. Detailed procedures are provided in the Supplemental Data.
We thank Robert Deschenes, Wendell Lim, Jennifer Pinkham, and Jeremy Thorner for gifts of yeast strains and plasmids, Nick Rhind for use of his FACScan analyzer, Rachel Lamson and Matt Winters for technical assistance, Kazuaki Homma for help with curve-fitting analysis, and Danny Lew and Fabian Rudolf for comments on the manuscript. This work was supported by a grant from the National Institutes of Health (GM57769) to P.M.P.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.