Another lesser-known example of component sharing is the differentiation switch triggered by alterations in pheromone concentration. Whereas yeast cells exposed to a high dose of pheromone undergo cell division arrest and shmoo formation, at lower doses they transiently arrest, become elongated and then divide, presumably in the direction of a pheromone gradient (Dorer et al., 1995; Erdman and Snyder, 2001; Paliwal et al., 2007; Segall, 1993) (Fig. 1A). The growth-arrest and elongated-growth pathways employ the same cell surface receptor, G protein transducer, and downstream protein kinases (Fig. 1A). Whereas either Fus3 or Kss1 will sustain the elongated cell morphology, Fus3 alone is required for growth-arrest and shmoo formation (Erdman and Snyder, 2001). It is not known if Fus3 or Kss1 contribute to a chemotropic-growth response.

In the past, stimulus-specific responses have been ascribed to the scaffolded-association of specific protein kinases. The term “scaffold” is commonly used to describe a protein that can assemble multiple distinct binding partners and promote their mutual interactions. In yeast the scaffold protein Ste5 has binding partners that include Ste11, Ste7, and Fus3. Whereas Ste5 is absolutely required for activation of Fus3 (Andersson et al., 2004; Breitkreutz et al., 2001; Flatauer et al., 2005; Kusari et al., 2004; Maleri et al., 2004), Ste5 binds poorly to Kss1 and is not required for the Kss1-mediated invasive-growth response (Choi et al., 1994; Kusari et al., 2004; Printen and Sprague, 1994). Based on these differences, it has been widely assumed that Ste5 insulates Fus3 from other MAP kinases that use the same upstream components, and thus prevents activation by irrelevant signals. While clearly required for Fus3 activity, however, Ste5 nevertheless allows activation of Kss1 by pheromone (Andersson et al., 2004; Breitkreutz et al., 2001; Choi et al., 1994; Flatauer et al., 2005; Maleri et al., 2004; Printen and Sprague, 1994). These observations have prompted a search for additional functions for Ste5.

More recently, Ste5 was reported to undergo a stimulus-induced conformational change and to allosterically regulate Fus3 kinase activity (Bhattacharyya et al., 2006; Sette et al., 2000). Ste5 stimulates Fus3 auto-phosphorylation at Tyr-182, one of two sites normally phosphorylated upon full activation of the kinase (Gartner et al., 1992). This mono-phosphorylated form of Fus3 is partially active. Moreover the interaction of Ste5 with Fus3 results in diminished overall signaling at the level of new gene transcription (Bhattacharyya et al., 2006; Sette et al., 2000). Thus Ste5 is essential for Fus3 signaling but also limits the overall output of the Fus3 signaling pathway.

Initially we investigated dose-dependent responses to pheromone, using a variety of functional assays. Pheromone-induced transcription has been shown to exhibit a graded dose-response profile (Poritz et al., 2001), whereas downstream processes such as G1 arrest and mating are switch-like. Switch-like behavior is an extreme example of ultrasensitivity, defined as any response with a Hill coefficient greater than one (Ferrell and Machleder, 1998; Goldbeter and Koshland, 1981). To determine if other aspects of the pheromone response exhibit ultrasensitivity, we monitored differentiation of cells exposed to a range of pheromone concentrations. Cells were classified into one of three categories: vegetative-growth, elongated-growth, or shmoo. Vegetative-growth occurs in the absence of pheromone. In this case cells divide rapidly and bud axially (new buds emerge from the daughter cell in the direction of the mother cell). At intermediate doses of pheromone the cells appear elongated, proliferate slowly, and divide in a bipolar fashion, opposite the mother cell (Dorer et al., 1995; Erdman and Snyder, 2001; Madden and Snyder, 1992). At high doses of pheromone the cells arrest and form a shmoo. Each of these morphologies is illustrated in Fig. 1. As shown in Fig. 2A, >85% of wild-type cells grow vegetatively at pheromone concentrations of 0.1 µM or less. At concentrations of 0.3 – 3 µM the population transitions from vegetative-growth to elongated-growth, even in the absence of a pheromone gradient. At 10 µM nearly all cells undergo growth-arrest and shmoo formation. The first transition occurs over a broader range of pheromone concentrations, as compared with the second transition. Thus the differentiation switch from elongated-growth to growth-arrest may exhibit ultrasensitivity.

We then investigated how the MAP kinases contribute to the two developmental transitions. It is well-established that mutants lacking Fus3 are unable to shmoo, but maintain the ability to transition from vegetative-growth to elongated-growth (Fig. 2A, second row). Conversely, cells lacking Kss1 still undergo both transitions (Fig. 2A, third row), but the dose-response profile for the transition from vegetative-growth to elongated-growth is significantly steeper than that of cells containing both MAP kinases or Kss1 alone. Signaling exclusively through Fus3 increases the pheromone concentration required for elongated-growth, thereby reducing the range of pheromone levels for which this behavior is observed. Kss1 broadens this range allowing cells to become elongated at lower doses of pheromone, as compared with cells that express both MAP kinases. Thus Kss1 and Fus3 confer distinct dose-response characteristics on cell growth: Kss1-mediated responses are graded while Fus3 responses appear ultrasensitive.

We next examined whether elongated-growth corresponds to chemotropic-growth, defined here as cell expansion and division in the direction of a stimulus gradient. It has been suggested that chemotropic-growth could allow yeast, which are otherwise non-motile, to orient new bud formation in the direction of a weak pheromone stimulus and thus towards a distant mating partner (Dorer et al., 1995; Erdman and Snyder, 2001; Madden and Snyder, 1992). To this end we constructed a microfluidic chamber capable of exposing cells to a precisely-controlled linear concentration gradient. The gradient is achieved by passive diffusion between two parallel microchannels containing either no pheromone or a dose of pheromone sufficient to induce cell division arrest (Fig. 1B and Supplementary Materials). There is no active flow within the growth chamber, thereby allowing the non-adherent cells to remain stationary during the course of the experiment. In this method the cells are at low density and the pheromone is constantly replenished, resulting in a lower dose-activity profile (see below).

As shown in Fig. 1C, Fig 2B, as well as in Supplementary Fig. S1 and the accompanying movies (Supplementary Information), cells within the growth chamber exhibit all three morphologies, depending on the local concentration of pheromone. Cells exposed to minimal doses of pheromone grow vegetatively and divide axially (Dorer et al., 1995; Madden and Snyder, 1992). Cells exposed to high doses of pheromone undergo cell division arrest and form a shmoo. At intermediate doses the cells appear elongated but continue to divide slowly and in the direction of increasing pheromone concentrations (Fig. 2B). This chemotropic-growth response generally entails just one round of elongated-growth, after which the daughter cell detaches from the mother cell and forms a shmoo. This is in contrast to invasive-growth, where cells undergo several rounds of polarized budding to form a long filament. In wild-type cells the vast majority of new buds and cell projections emerge within +/− 60° of the gradient, and very few are outside +/− 90° (Fig. 2C). Mutants lacking Fus3 or Kss1 retain the ability to undergo the elongated-growth response. Likewise, mutants lacking Kss1 are able to divide in the direction of the gradient, although Kss1 may contribute at very low doses of pheromone or in mutants that are supersensitive to pheromone (Fig. 2B and 2C) (Dorer et al., 1995; Erdman and Snyder, 2001; Paliwal et al., 2007; Segall, 1993). In cells lacking Fus3 however, newly formed buds emerge randomly with respect to the gradient and the cells no longer undergo growth-arrest and shmoo formation (Fig. 2B and 2C). Thus Fus3 is specifically required for chemotropic-growth but is dispensable for elongated-growth.

Cellular responses mediated by Fus3 appear to be ultrasensitive, while responses mediated by Kss1 appear graded (Fig. 2). Measurements of cell morphology are somewhat qualitative, however, so we also measured changes that occur at a molecular level. Specifically, we asked if there is ultrasensitivity at the level of protein kinase activity. To this end we monitored Fus3 and Kss1 activation in cells treated with a range of pheromone concentrations, using antibodies that recognize the dually-phosphorylated (fully activated) form of each protein. As shown in Fig. 3A and Fig. S2 the temporal-and dose-response profiles for Fus3 and Kss1 phosphorylation are distinct. At 1 µM pheromone (a dose sufficient to trigger chemotropic-growth but not shmoo formation) Fus3 is phosphorylated to 22% of maximum while Kss1 is phosphorylated to 48% of maximum. At 10 µM pheromone (a dose sufficient to trigger cell division arrest and shmoo formation) both kinases are maximally activated. Remarkably, we found that maximum kinase activity occurs at different times depending on the pheromone concentration; for example Fus3 activity peaks at 60 min in response to 10 µM pheromone whereas it peaks at 15 min in 1 µM pheromone (Fig. 3A). Thus two different doses may produce identical kinase activity at a single (early) time point, but nevertheless exhibit dramatic differences in the duration and final maximum level of kinase activation. Accordingly, we quantified peak kinase activity (Fig. 3D) and area under the curve (Fig. 3F) for each dose of pheromone. The latter incorporates the elements of duration and activity into a single parameter. Both of these methods accurately account for the observed differences in activity, and yielded similar results. As shown in Fig.3F, the effective Hill coefficient for activation of Fus3 is substantially higher than for activation of Kss1 (n_H = 2.2 and 1.3, respectively). This difference mirrors the dose-response profiles for Fus3-and Kss1-mediated cell differentiation responses, as reported in Fig. 2. Note that we measured the MAP kinase activation of a population of cells. The response might be even more switch-like in individual cells.

Fus3 and Kss1 are activated by the same upstream protein kinases, most immediately by the MAP kinase kinase Ste7. Nevertheless we have shown above that Fus3 and Kss1 exhibit very different temporal-and dose-dependent behaviors. Whereas Kss1 activity peaks quickly, Fus3 activity increases slowly at a constant rate (slope) that is independent of the pheromone level. In both cases, increasing the pheromone level delays the time at which MAP kinase activity begins to decline. One way to account for this unusual behavior is to assume that the predominant effect of increases in pheromone concentration is to prolong the duration of Ste7 activity (Fig. 3B, top panel). To address this possibility we constructed simple mathematical models of Fus3 and Kss1 activation and investigated each model’s response to varying durations of Ste7 activation. Both models assume that the phosphorylation and dephosphorylation reactions follow Michaelis-Menten kinetics. Each model further assumes that the ν_max for phosphorylation of the MAP kinase is proportional to the active Ste7 concentration. The difference between the models is that the rate constants describing phosphorylation and dephosphorylation of Kss1 are larger than those of Fus3 (Supplementary Materials). Fig. 3B shows the Ste7 activation profile used as the input signal for both MAP kinases (top panel), as well as the activity for Fus3 (middle panel) and Kss1 (bottom panel) predicted by the corresponding models. There is a very good qualitative agreement between these simulations (Fig. 3B) and the experimental data (Fig. 3A). The models also demonstrate how a purely kinetic mechanism (slow activation) can be used to convert a graded response to one that is more switch-like (see Discussion).

We then considered why Fus3 is activated more slowly than Kss1. This difference could be an inherent property of each kinase. Alternatively, slow activation of Fus3 could be due to binding of Ste5. Indeed, Ste5 was reported previously to diminish Fus3-mediated transcription responses (Bhattacharyya et al., 2006). To investigate the contribution of Ste5 to Fus3 dynamics, we monitored kinase activation in the absence of Ste5-Fus3 interaction. Deletion of STE5 blocks signaling altogether, so as an alternative we used a mutant form of Ste5 in which Fus3 docking is disrupted (“non-docking” allele, Ste5^ND). This mutant binds poorly to Fus3 (Maeder et al., 2007), yet produces an enhanced transcription-induction response (Bhattacharyya et al., 2006). Thus we investigated how Ste5^ND affects dose-and time-dependent signaling events including cell morphogenesis, gradient-sensing and kinase-activation. To eliminate any possibility that Kss1 can substitute for Fus3 or otherwise modulate Fus3 activity in the absence of Ste5, these experiments were conducted in a Kss1-deficient strain. As shown in Fig. 2A, cells that express Ste5^ND are able to undergo the transitions from vegetative-growth to elongated-growth and then to growth-arrest. However, the transitions occur at lower pheromone concentrations and the dose-response profile is more graded than that seen in wild-type cells. Indeed we observed considerable overlap in the distribution of all three categories of cells within the gradient chamber. Moreover, cells that express Ste5^ND exhibit diminished elongation and orientation towards the source of pheromone (Fig. 2B and 2C). Likewise, Fus3 becomes fully activated in the Ste5^ND strain, but activation occurs with faster kinetics (Fig. 3C). Additionally, the Hill coefficient for activation of Fus3 is reduced from n_H=2.2 in wild-type cells to n_H=1.5 in Ste5^ND mutant cells, a value similar to that normally observed for Kss1 (n_H= 1.3) (Fig.3F). Thus under conditions where it is no longer modulated by Ste5, the dynamic and dose-dependent behavior of Fus3 resembles that of Kss1.

We also considered the possibility that Ste5^ND might affect the activity of other signaling components upstream of Fus3, such as Ste11 or Ste7. To test this we examined whether Ste5^ND alters the activity of Kss1, which likewise requires Ste11 and Ste7. In accordance with the model, Ste5^ND (in the absence of Fus3 expression) had no effect on Kss1 activation (Fig. S3). Another upstream mechanism of regulation is pheromone degradation. The pheromone protease Bar1 is up-regulated upon prolonged pathway activation, and different degrees of induction by Fus3 and Kss1 might somehow underlie the differences observed for the two kinases. To address this concern, and as a further test of our computational model, we measured Fus3 and Kss1 activation in a bar1Δ mutant strain (Fig. S4A). As anticipated, differences in the dynamic and dose-dependent behaviors of Fus3 and Kss1 remain unchanged. Strikingly, by altering only the input signal profile the equations and parameters that govern kinase activation in wild-type cells could also be used to accurately describe kinase activation in the bar1Δ mutant cells (Fig. S4B).

Finally, we examined the physiological importance of the elongated-and chemotropic-growth behaviors in mating. Previous investigations employed the “pheromone-confusion” assay, which compares overall mating efficiency in the presence or absence of added pheromone. The underlying assumption is that exogenous pheromone will obscure natural pheromone gradients, and thereby diminish the ability of cells to detect a potential mating partner (Dorer et al., 1995; Nern and Arkowitz, 1998; Strickfaden and Pryciak, 2007; Valtz et al., 1995). One drawback to this method is that addition of excess exogenous pheromone is also likely to obscure or otherwise interfere with pheromone clearance by Bar1 protease, receptor occupancy, as well as adaptation/desensitization responses that further impinge on mating efficiency. As an alternative approach, we measured mating efficiency under conditions where potential mating partners are rare (10:1 ratio) and spatially segregated (plated on solid growth medium). As shown in Fig. 4, the Ste5^ND mutants mate substantially better than wild-type cells, presumably because they can better detect and respond to a distant mating partner. Conversely, Ste5^ND mutants mate more poorly under conditions where potential mating partners are more abundant (1:10 ratio) but still dispersed, presumably because they can no longer orient towards the closest of multiple potential mating partners. This reduction in mating efficiency occurs despite the increase in pheromone sensitivity.

Experiments conducted in a variety of other systems have revealed that MAP kinases can produce distinct cellular responses depending on the temporal behavior of the pathway. In one oft-cited example, epidermal growth factor causes transient activation of the ERK MAP kinase leading to cell proliferation, while nerve growth factor causes sustained ERK activation and cell differentiation (Marshall, 1995). Similarly in yeast, temporal differences in MAP kinase activation can dictate whether cells follow a mating or nutrient-driven invasive-growth program. Specifically, when upstream kinases Ste11 or Ste7 are mutated so as to be permanently active, Kss1 is activated in a sustained manner, in preference to Fus3, and this condition leads to invasive-growth but not mating (Andersson et al., 2004; Flatauer et al., 2005; Maleri et al., 2004).

Another way MAP kinases can produce distinct cellular responses is through changes in the dose-dependent behavior of the pathway. MAP kinase signaling pathways have long been known to exhibit ultrasensitivity (Koshland et al., 1982). For example, MAP kinase cascades in Xenopus oocytes use a variety of mechanisms to produce an all-or-none response leading to cell maturation (Ferrell and Machleder, 1998). On the other hand the pheromone-induced transcription response in yeast exhibits a more graded dose-response profile (Poritz et al., 2001), even when downstream processes such as G1 arrest and mating are switch-like. Our data reveal that the two pheromone-responsive MAP kinases exhibit distinct dose-dependent behaviors. Whereas activation of Kss1 is graded, activation of Fus3 is ultrasensitive. The ultrasensitivity exhibited by Fus3 is not as dramatic as that demonstrated in Xenopus, but is comparable to the benchmark example hemoglobin, which exhibits a Hill coefficient of 2.8.

Our data show that Ste5 is required for the unique time-and dose-dependent characteristics exhibited by Fus3. Ste5 was originally considered to function as a passive tether for protein kinases. More recently, Ste5 was shown to be required for Fus3 activation, but also impedes full activation of the signaling pathway (Bhattacharyya et al., 2006). That loss of signaling was demonstrated using a reporter transcription assay. To determine the contribution of Ste5 to MAP kinase function, we evaluated signaling in the absence of Fus3 binding (Ste5^ND). Signaling was monitored by cell differentiation, gradient sensing, and MAP kinase activation assays. By all of these measures it appears that Ste5 is necessary for the slow and ultrasensitive responses exhibited by Fus3. When Ste5 binding is abrogated, Fus3 behaves like Kss1.

A distinction between our analysis and prior investigations is that we considered the role of pheromone gradient sensing. Most studies of the yeast pheromone response have been done with uniform and saturating concentrations of pheromone. In a physiological setting however, yeast cells are likely to be exposed to a pheromone gradient emanating from a potential mating partner. When that partner is too distant for mating to occur, the gradient will be weak and some mechanism must exist to bridge the divide. Here we demonstrate that Ste5 promotes cell elongation in the direction of a pheromone gradient, thereby increasing the probability of mating with a distant partner.

Finally, our computational analysis reveals how a kinetic mechanism (specifically, slow activation) can potentially be used to convert a graded response to one that is more switch-like. We postulate that the main effect of increasing pheromone concentration is to increase the time the signal is on. In this scenario the fast activation kinetics of Kss1 allows it to faithfully track the input signal. Conversely, the slow activation kinetics of Fus3 means that it does not saturate while the incoming signal is present. Because Fus3 activity increases linearly with time, the response increases with time and thus with pheromone concentration.

01

02