MAP Kinase Pathways in the Yeast Saccharomyces cerevisiae

Activation of the MAPK Cascade

Gβ activation of the mating-pheromone MAPK cascade is mediated primarily by Ste5p (51, 235, 446, 487) and Ste20p (236, 237, 499). Other proteins, such as Ste50p (387, 502) and Bem1p (204, 234, 243, 270), may also play a role in transducing signals from the Gβ Ste4p, but their functions are not essential. Ste5p and Ste20p appear to be necessary and limiting for MAPK cascade activation by Gβ (3, 61, 164, 218, 235). The way in which these proteins cooperate is not yet understood, but for each mediator there is some information about the transduction mechanism.

Signaling Pathways and the Cytoskeleton

Cellular localization and activation of the pheromone-activated MAPK cascade appears to involve proteins that are functionally connected to the cytoskeleton. This is a common observation in eukaryotic signal transduction. For example, tethering of signal transduction proteins to particular regions of the cell is mediated in part by the cytoskeleton. Also, some signaling pathways regulate the function of the cytoskeleton and, in certain situations, the cytoskeleton participates in transmitting signals to the nucleus. One system in which these different cytoskeleton-signal transduction relationships have been well explored is the pheromone response pathway in yeast. Two proteins in particular, Cdc42p and Bem1p, connect the pheromone response pathway to the actin cytoskeleton.

Cdc42p is required to orient the actin cytoskeleton to form a bud, to divide the cell during cytokinesis, and to form mating projections (1, 110, 254, 527). Cdc42p therefore interacts with a variety of different proteins that regulate actin cytoskeleton function. Cdc42p in cells exists in a dynamic equilibrium between the GDP-bound and GTP-bound forms. Exchange of GDP for GTP on Cdc42p is activated by Cdc24p (525), and the hydrolysis of the Cdc42p-bound GTP to GDP is predicted to be regulated by the GTPase-activating proteins (GAPs) Bem3p (525) and Rga1p (445). Cdc24p, like Cdc42p, is an essential protein required for polarized cell growth during bud formation and formation of mating projections during conjugation (60, 439).

Several observations suggest that Cdc42p plays an important role in the pheromone response pathway. Temperature-sensitive cdc24 or cdc42 mutants, when grown at a nonpermissive temperature, do not show an increase in FUS1-lacZ expression (FUS1 is a pheromone-induced gene [303, 466]) in response to pheromone treatment (435, 521) and cannot mate (393). Strains with the Cdc42-GAP Rga1p deleted show increased pheromone-induced transcription (445). Indeed, overexpression of a mutant Cdc42p locked in the GTP-bound state activates FUS1 expression (435, 521), even in a strain carrying a dominant negative mutant of the Gβ Ste4p (435). The increased FUS1-lacZ expression in cells expressing an activated Cdc42p does require the presence of pheromone, suggesting that Cdc42p acts to modulate signaling by the pheromone response pathway.

Cdc42p appears to have multiple functions in the mating response, at least one of which does not involve the MAPK pathway. Yeast cells form mating projections in response to pheromone treatment. The growth of these projections is spatially oriented toward the source of pheromone and is therefore called chemotropic growth (416). This process involves activation of Cdc42p (335) but, importantly, does not require the protein kinases of the MAPK cascade (410). As discussed above, Gβ Ste4p interacts with Ste5p and Ste20p and, by mechanisms yet unclear, activates the MAPK cascade. Ste4p also interacts with Cdc24p (335, 521), the guanine nucleotide exchange factor for Cdc42p. Mutations in Cdc24p that block the interaction with Ste4p also block chemotropic growth but have no effect on other responses to pheromone including MAPK cascade-mediated growth arrest and FUS1-lacZ expression (335). Because the function of Cdc24p is to activate Cdc42p and Cdc42p mediates polarized cell growth, the interaction of Gβ with Cdc24p may provide a mechanism to locally activate Cdc42p and Cdc42p-dependent growth in the vicinity of pheromone-occupied receptors.

Bem1p, like Cdc42p, interacts with several proteins important for the function of the actin cytoskeleton in polarized growth (28, 57, 59, 110). Bem1p associates with actin and with the pheromone response pathway-signaling proteins Ste5p and Ste20p (243, 270). The Bem1p-bound Ste5p is complexed to the Ste11p-Ste7p-Fus3p MAPK cascade (270). Interaction of Ste20p with Bem1p is required for association of Ste20p with actin (243). The fraction of these signaling proteins associated with macromolecular complexes in the cell is considerable. At least half of the cellular Ste5p, Ste20p, and Bem1p localizes to a particulate fraction of the cell and remains there after extraction of membrane proteins with nonionic detergents (243). Bem1p interacts in cells with other signaling proteins: the Cdc42p guanine nucleotide exchange factor Cdc24p (298, 370); Far1p (270), a protein needed for pheromone-induced cell cycle arrest (55) (see below); and Boi1p and Boi2p (29, 298), proteins involved in the regulation of the Rho-type GTPase Rho3p and Rho3p-dependent growth-related processes.

Bem1p-associated proteins can have more than one function. For example, Far1p has two functional parts, a COOH-terminal domain required for chemotropism (107, 473) and an NH₂-terminal domain required for pheromone-induced cell cycle arrest (473). The observation that the MAPK cascade is required for cell cycle arrest (113) but not chemotropism (410) provides further confirmation that these are mechanistically separate responses to pheromone. The mechanism by which Far1p performs two very different functions is unknown. Thus, the multitude of interacting partners for Bem1p and their functional diversity raise the question whether a single Bem1p molecule can complex simultaneously with all potential partners or whether different Bem1p molecules form separate complexes with different protein partners.

So far, it has not been possible to detect an effect of pheromone on the extent of interaction between Bem1p, Ste20p, and Ste5p as assayed by coimmunoprecipitation experiments (243, 270). Thus, Bem1p might just simply tether the signaling pathway to the cytoskeleton. Bem1p does, however, facilitate signaling by the pheromone pathway. Deletion of BEM1 decreases the pheromone-induced transcription of FUS1 (204, 270). In addition, overexpression of BEM1 stimulates the kinase activity of the MAPK Fus3p (270) and suppresses the mating defect of a dominant negative STE4 mutant (234). These data suggest that Bem1p is involved not only in cross-linking the Ste5p-MAPK cascade complex to the cytoskeleton but also in transmitting signals to the MAPK cascade either directly or by facilitating its association with an upstream activator.

Sending Signals to the Nucleus: a Role for the MAPK Cascade

The pheromone-activated signaling pathway containing the Ste11p-Ste7p-Fus3p MAPK cascade is required for sending signals from the pheromone receptors in the plasma membrane to gene targets in the nucleus. There are no known second messengers relaying signals on the pathway. Therefore, some protein or protein complex must leave the cytoplasm and move across the nuclear membrane. In animal cells, MAPK moves from the cytoplasm into the nucleus following stimulation by growth factor (58). This movement involves dissociation of MAPK from its cytoplasmic complex with MEK (135). MEK appears to be in the cytoplasm and to remain there after growth factor treatment (522). However, more recent experiments suggest that MEK can also be induced to move from the cytoplasm to the nucleus following growth factor stimulation if its nuclear export signal (134) and catalytic site are inactivated by mutation (191). Disruption of the nuclear export signal in MEK strongly stimulates MEK-dependent morphological changes and malignant transformation (133). Thus, the apparent cytoplasmic localization of MEK in growth factor-stimulated cells may reflect transient nuclear entry followed by rapid export from the nucleus (133, 191). A leucine-rich sequence near the NH₂ terminus of MEK acts as the nuclear export signal (134); it is interesting that the yeast MEK Ste7p has a very similar sequence near its NH₂ terminus.

In the case of the yeast pheromone response pathway, it is still a mystery how the signal actually gets to the nucleus. Of the proteins on the MAPK cascade, the MAPK Fus3p appears to be present in the cytoplasm and nucleus (62). The MAPK Kss1p of the filamentation-invasion pathway (see below) is mostly in the nucleus (271). These MAPK locations change little after pheromone treatment. Due to their apparent low abundance, the locations of Ste11p and Ste7p in the cell have been more difficult to determine and are not known with certainty at present.

Ste5p does seem to change location after pheromone treatment, although whether nuclear entry of Ste5p is required for signaling has not yet been determined. At different times and under different conditions, Ste5p is alternatively found at or near the plasma membrane, in the cytoplasm, or in the nucleus. Microscopic analysis shows Ste5p to be present in both the cytoplasm and the nucleus in vegetatively growing cells (283). After pheromone treatment, Ste5p moves from the nucleus to the cytoplasm and becomes associated with the plasma membrane in mating projections (97, 283). Interaction of Ste5p with Ste4p is required for the association of Ste5p with the plasma membrane (97). The association of Ste5p with the plasma membrane appears to be a critical step in signal transduction, because fusion of membrane-targeting signals to Ste5p induces activation of pheromone responses in the absence of added pheromone (384). A striking result is that Ste5p with an NH₂-terminal truncation removing the Gβ-binding domain is nonfunctional unless fused to membrane-targeting signals (384). Thus, plasma membrane localization of Ste5p is sufficient for signaling.

Ste5p localizes to the nucleus when untethered from Gβ (97, 283). Thus, Ste5p may be part of the signaling machinery that shuttles signals to the nucleus, perhaps released from Ste4p in pheromone-activated cells. It should, however, be pointed out that nuclear localization of Ste5p is not sufficient for signaling (97, 283). In addition, the situation may not be as simple as a single protein or protein complex shuttling signals to the nucleus: there may be multiple mechanisms acting in parallel. Deletion of the MEK gene STE7 enhances Ste5p-Ste5p interaction in the two-hybrid system, suggesting that Ste7p-Ste5p and Ste5p-Ste5p complex formation might be mutually exclusive, i.e., that Ste5p dimerization might lead to Ste7p ejection (505). Ste5p preferentially interacts with the underphosphorylated, preactivated form of Ste7p, suggesting that phosphorylation of Ste7p might induce its release from the complex with Ste5p (61). Perhaps Ste7p, like the animal cell MEK (191), also carries signals to the nucleus. The tight complex formed between Ste7p and Fus3p (19, 21) suggests that instead of individual kinases, a complex of MEKK and MAPK may be the molecular species that carries signals to the nucleus. Movement of a protein or protein complex from the cytoplasm to the nucleus will require its dissociation from other cytoplasmic proteins. This may require more than one regulatory event or cooperative changes in protein conformation, especially in the case of ternary or higher-order complexes, where a protein must dissociate from more than one binding partner before it can break free of the complex (372).

Induction of Cell Cycle Arrest

The MAPK pathway plays another important role in mediating cell cycle arrest in response to pheromone (494). Conjugation of two haploid mating partners is accompanied by the synchronization of the cell cycles of the two cells such that they both contribute 1N content of DNA to the zygote product of their union. Thus, mating pheromone-treated cells arrest at a position in the cell cycle prior to bud formation and initiation of DNA synthesis: they arrest as unbudded cells with a 1N DNA content. Growth of the G₁-arrested cell is not inhibited but redirected into the formation of mating projections. This pheromone-induced cell cycle arrest in G₁ involves signaling through the MAPK cascade (112, 113, 132, 469) and the cell cycle inhibitor Far1p (55, 141, 469).

To explain the mechanism of cell cycle arrest and how the MAPK pathway is involved, we first review the mechanisms that regulate cell cycle progression at the G₁/S transition in yeast (331). Formation of a bud, initiation of DNA synthesis, and duplication of the spindle pole body mark the progression of a yeast cell into S phase, past a G₁/S transition point called START. These post-START events require the activation of cyclin–cyclin-dependent kinase complexes consisting of the kinase Cdc28p and one of three G₁ cyclins: Cln1p, Cln2p, or Cln3p. An active G₁ cyclin–Cdc28p complex is needed to induce the degradation of a cyclin-dependent kinase inhibitor that is specific for B-type cyclin–Cdc28p complexes (413). This protein inhibitor, called Sic1p (344) (also called Sbd25p [106]), blocks the activity of Cdc28p in complex with the B-type cyclins Clb5p and Clb6p but not the activity of G₁ cyclin–Cdc28p complexes. The B-type cyclin–Cdc28p complex, freed of its inhibitor protein, activates DNA replication (414). The mechanism responsible for activation of bud initiation by the G₁ cyclin–Cdc28p complex is independent of Sic1p (408, 468).

Cell cycle arrest by mating pheromone involves Far1p-dependent (367) and Far1p-independent processes (468). Far1p expression is normally restricted to the G₁ phase (305) by mechanisms of cell cycle-dependent transcription and protein turnover (167, 306, 347). Results from early studies indicated that Far1p is a cyclin-dependent kinase inhibitor that inhibits the activity of G₁ cyclin–Cdc28p complexes, but not that of B-type cyclin–Cdc28p complexes (196, 368). However, a more recent study could not detect a pheromone-induced reduction in the activity of the Cln2p-associated Cdc28p kinase, even though these complexes retain Far1p (141). Nevertheless, Far1p is required for pheromone-induced inhibition of G₁ cyclin–Cdc28p-dependent responses such as the expression of CLN1 and CLN2 (472). The MAPK Fus3p (but not Kss1p) is also required for cell cycle arrest in response to mating pheromone (112, 113). The functions of Fus3p and Far1p are linked, because pheromone induces the Fus3p-catalyzed phosphorylation of Far1p (56, 119, 141, 367, 469). G₁ cyclin–Cdc28p also phosphorylates Far1p (167, 367, 469) and thereby stimulates its degradation by a ubiquitin-dependent mechanism (167). Fus3p-catalyzed phosphorylation appears to have the opposite effect of stabilizing the Far1p protein (unpublished results cited in reference 167).

Far1p is a bifunctional molecule, required not only for cell cycle arrest but also for chemotropism (107, 473). This latter function is not connected to the function of the MAPK Fus3p or to that of the rest of the MAPK cascade (410). The way in which these two functions of Far1p are coordinated is not yet clear. Interestingly, the mechanism by which Far1p mediates cell cycle arrest is also not well understood at present (141).

The effects of pheromone on the cell cycle may be more complex than altering the activity of Far1p. In the absence of Cln1p, Cln2p, and the cyclin-dependent kinase inhibitor Sic1p, pheromone induces a Far1p-independent arrest of the cell cycle (468). In cells that have reduced activity of the Cln class of cyclin, another type of cyclin–cyclin-dependent kinase complex containing the Cdc28p-related protein Pho85p becomes critical for cell cycle progression (121, 308). The mRNA level for one of the Pho85p-associated cyclins, Pcl1p, is rapidly down-regulated by pheromone treatment (309) with a time course similar to that of the pheromone-induced decrease in Cln1p and Cln2p mRNA (495). Perhaps the Far1p-independent cell cycle arrest induced by pheromone treatment in cln1Δ cln2Δ sic1Δ cells (468) reflects parallel regulation of the Pho85p kinase through transcriptional control of expression of the Pcl1p cyclin associated with Pho85p. The Far1p-Cdc28p paradigm also suggests that the Pho85p inhibitor Pho81p (409) might be a target of regulation by the pheromone pathway. However, Pho85p interacts with many different cyclins, and its physiological functions appear to be very complex (9). For example, although Pcl1p and Pcl9p mRNAs are decreased by pheromone treatment, other Pho85p cyclins show no change or an increase in mRNA expression after addition of pheromone (309). Finally, the role of the MAPK pathway in the Far1p-independent cell cycle arrest by pheromone has not yet been determined.

As discussed above, the pheromone response pathway regulates the cell cycle but the converse is also true. For example, the basal level of protein kinase activity of MEK Ste7p and MAPK Fus3p fluctuates during the cell cycle, reaching a peak in early G₁ (481). The activity of the MAPK cascade—high in early G₁ and low in late G₁—correlates well with the amount of mRNA for different pheromone-dependent genes (346, 347, 517). These molecular changes in the absence of pheromone may allow the cell to be maximally responsive to pheromone in early G₁, a cell cycle position close to the pheromone arrest point in late G₁. The cell cycle regulation of the MAPK cascade and pheromone-dependent genes appears to be mediated through the G₁ cyclins Cln1p and Cln2p, cyclins that reach their peak expression level in late G₁ (392). Hence, overexpression of CLN2 represses the mating pathway (346, 481). Analyses of various mutants that either allow the Cln2p repression or block its effect suggest that the target of Cln2p repression is downstream of the Gβ Ste4p. The MEKK Ste11p or one of the proteins involved in activating the MAPK cascade are the current candidates for the target of Cln2p repression (481). Another potential connection between Cln2p and the mating pathway is at the level of the MAPK substrates Dig1p and Dig2p, repressors of the transcription factor Ste12p (68, 457). Cln1p and Cln2p each show specific interactions with Dig1p and Dig2p in the two-hybrid system (457). While the functional significance of this interaction has not yet been determined, it is tempting to speculate that positive regulation of Dig1p and Dig2p by Cln2p-Cdc28p or Cln1p-Cdc28p might repress Ste12p and shut off the pheromone response, thereby enhancing recovery. In summary, there is a reciprocal relationship between the activities of the pheromone response pathway and the G₁ cyclin-Cdc28p complex, regulator of the G₁/S cell cycle transition, with each inhibiting the other. This situation allows the cell to make a clean switch from one function to the other, from budding to mating or vice versa.

One MAPK—One Pathway

Kss1p is the MAPK for the filamentation-invasion pathway (69, 277, 278). Historically, the observation that formation of pseudohyphae is not blocked by deleting any or all of the MAPKs in yeast led to an initial hypothesis that the filamentation-invasion pathway does not use a MAPK for signaling (257). For example, cells with or without the MAPK Kss1p show diploid pseudohyphal development on low-nitrogen medium, haploid invasive growth, and expression of FRE-lacZ (69, 257, 278, 395). However, cells with an inactivated Kss1p (with STE7 deleted or expressing a nonphosphorylatable mutant Kss1p in a kss1Δ background) do not undergo pseudohyphal development or haploid invasive growth and have reduced FRE-lacZ expression (20, 69, 278). These findings indicate that the unactivated form of the Kss1p kinase inhibits the pseudohyphal response. The haploid invasive growth response is inhibited not only by Kss1p but also by Fus3p (69, 278).

Induction of the pseudohyphal response by the MEK Ste7p appears to involve two effects. Ste7p-catalyzed phosphorylation of Kss1p relieves inhibition of the pseudohyphal response by Kss1p. Expression of wild-type Kss1p or a catalytically inactive but phosphorylatable mutant of Kss1p allows a kss1Δ fus3Δ strain to show invasive growth and normal levels of expression of FRE-lacZ (20). In contrast, a nonphosphorylatable Kss1p does not allow these responses (20, 69, 278). The mechanism by which nonphosphorylated Kss1p inhibits invasive growth and FRE-dependent transcription appears to be mediated by binding of the unactivated MAPK to the transcription factor Ste12p. For example, a mutant of Kss1p that binds normally to Ste7p and to the Ste12p-repressors Dig1p and Dig2p but not to Ste12p was isolated. This mutant Kss1p can no longer inhibit the pseudohyphal response (20).

Ste7p-catalyzed phosphorylation of Kss1p not only removes a repressor (unphosphorylated Kss1p) but also appears to generate an activator (phosphorylated Kss1p). This dual role of Kss1p can be appreciated by comparing wild-type and kss1Δ strains. Although cells lacking (the repressor) Kss1p show some invasiveness and expression of FRE-lacZ, the levels of each are significantly lower than that observed for KSS1⁺ cells (69, 278). Expression of hyperactive forms of either the MEKK Ste11p or the MEK Ste7p induces a strong pseudohyphal response and greatly increased FRE-lacZ expression (257, 276, 278). Cells with KSS1 deleted show no response to expression of these hypermorphic mutants (278), providing further support for the idea that Kss1p in its phosphorylated, active state is a positive regulator of the pseudohyphal response.

Kss1p has also been proposed to be part of the pheromone response pathway MAPK cascade. Kss1p, in the absence of Fus3p, allows near-wild-type levels of mating (112, 142), suggesting that Kss1p may also play a part in signaling by the pheromone response pathway. A fus3Δ kss1Δ strain is thus completely sterile. Further support for Kss1p as a mediator of mating pheromone responses is that pheromone treatment increases Kss1p kinase activity (19), although the fold increase is much lower than that for pheromone stimulation of Fus3p kinase activity (114). Furthermore, Kss1p interacts in the two-hybrid system with the pheromone pathway scaffold protein Ste5p (61), although whether this interaction is mediated through the MEK Ste7p was not tested.

However, it has been recently argued that Kss1p is not normally part of the pheromone response pathway and fills that role only when Fus3p has been deleted. Rather, it was proposed on the basis of several observations that Fus3p is the MAPK for the pheromone response pathway (277, 278), just as Kss1p is the MAPK for the filamentation-invasion pathway (69, 278). Kss1p cannot fully cover for the loss of Fus3p. For example, pheromone-induced cell cycle arrest requires Fus3p and Kss1p cannot mediate this response (112). As mentioned above, pheromone does increase Kss1p kinase activity but the increase is much lower than that for the Fus3p kinase. Pheromone effects on Kss1p kinase activity were also tested under conditions of Kss1p overexpression (19), in which Kss1p could artifactually compete with Fus3p.

Deletion of FUS3 may thus allow Kss1p to perform new functions; e.g., pheromone induces a Kss1p-dependent increase in FRE-lacZ expression but only in fus3Δ cells (278). A fus3Δ strain shows increased haploid invasive growth; kss1Δ or ste4Δ suppresses this phenotype. Haploid invasive growth of a wild-type FUS3 strain is not inhibited by ste4Δ (278). These observations suggest that in the absence of Fus3p, Ste4p inappropriately signals to Kss1p and therefore activates FRE-dependent transcription of invasive growth genes. One mechanism to explain the fus3Δ phenotype is that the absence of Fus3p allows Kss1p to bind to the MAPK binding site on Ste5p and receive signals from pheromone. The observation that a strain expressing a catalytically inactive mutant Fus3p in a fus3Δ strain is more sterile than a fus3Δ strain (278) is consistent with this possibility. The inactive Fus3p mutant had no effect when expressed in a wild-type FUS3 strain (278), showing that the mutant is not acting as a dominant negative mutant to Fus3p and, by extension, to Kss1p. Although many of these data support a model in which Fus3p is the MAPK for the pheromone response pathway, additional experimental tests are needed to fully resolve this point. For example, it is important to know whether addition of mating pheromone induces the activation of Kss1p phosphorylation or kinase activity with similar kinetics to the observed activation of Fus3p, particularly under conditions where both proteins are present at wild-type expression levels. In addition, it is important to know whether Kss1p is physically associated with Ste5p in cells under conditions of normal expression levels for both proteins.

Signaling Proteins Shared by two MAPK Pathways

Yeast cells use the same signaling proteins (Ste20p, Ste11p, Ste7p, and Ste12p) in two different pathways that receive different input signals and generate different outputs. Pheromone induces mating, and nitrogen starvation induces filamentation and invasion. Three factors are important in matching input signal to output response by using the same signaling proteins for the central part of two different pathways. Cell-type-specific gene expression is one such factor. To respond to mating pheromone, cells need receptors for the pheromone (Ste2p and Ste3p) plus a G protein (Gpa1p-Ste4p-Ste18p), a MAPK cascade scaffold protein (Ste5p) to transmit the signal from the receptors to the MAPK cascade, and a MAPK (Fus3p) to induce cell cycle arrest. Diploid cells do not express these components and therefore cannot respond to mating pheromone (113, 163, 235, 326, 366, 442, 486). However, haploid cells can activate either a pheromone response pathway or a filamentation-invasion pathway (395). A second factor important for determining pathway specificity is a protein complex that allows specific input signals to the MAPK cascade and then to the transcription factor. One pathway-specific protein complex has been identified for the pheromone response pathway (e.g., the Ste4p-Ste5p-MAPK cascade) but the corresponding complex for the filamentation-invasion pathway is unknown. How pathway specificity is generated at the steps involving the PAK Ste20p and the transcription factor Ste12p, respectively, has not yet been determined. The final factor important for generating specificity is input from one or more additional pathways, a critical factor for the filamentation-invasion pathway (146, 223, 259, 264). Although the molecular details of transcriptional regulation during the pseudohyphal response are sketchy at present, it seems reasonable to expect that the combination of signals from different pathways dictates which genes to turn on and which to keep off.

Activation of the Pathway

Interconnections of the Pathway

The cell integrity pathway MAPK cascade, together with the signaling proteins that input signals to the pathway and the targets of the cascade, must form specific complexes that allow rapid signaling and control cross talk with other pathways. A cell integrity pathway equivalent of the pheromone response pathway Ste5p, a scaffold protein that holds together the different protein kinases, has not yet been uncovered. Such a scaffold may not be required. The cell integrity pathway components are unique to that pathway, whereas the pheromone response pathway MEKK and MEK are shared with other pathways, suggesting the need for a scaffold protein to keep their functions isolated. By using the two-hybrid system, a series of interactions have been observed between kinases of the cell integrity pathway (358, 440). For example, Pkc1p interacts with the MEK Mkk1p independently of the MEKK Bck1p. The MEKK Bck1p interacts with Mkk1p and Mkk2p through the COOH terminus of the MEKs. Mkk1p and Mkk2p interact with the MAPK Slt2p. Interestingly, as in the HOG pathway (279, 378) (see below), the MEK of the cell integrity pathway appears to play a central role in organizing that MAPK cascade. Although there are important limitations of the two-hybrid system that counsel against overinterpreting its results (372), it is intriguing that Pkc1p shows a two-hybrid interaction with the MEK Mkk1p and no other protein kinase of the cascade (358). Spa2p and Sph1p are two nonessential proteins that localize to growing regions of the cell (14, 438); Spa2p is required for bud site selection and for formation of mating projections (144). Both of these proteins interact in the two-hybrid system with the MEKs Mkk1p and Mkk2p, through the NH₂-terminal nonkinase domain of the MEK (400, 422). Whether these interactions with Spa2p and Sph1p localize the MAPK cascade or mediate signal transmission to the protein kinases is not yet known. In this context, an interesting and potentially important effect of a spa2Δ mutation is that it delays pheromone induction of Slt2p kinase activity but blocks the later decline in activity of this MAPK (50). Without stimulation by pheromone, spa2Δ cells show increased Slt2p kinase activity and elevated phosphorylation of the Slt2p substrate Swi6p over that observed in SPA2⁺ cells (422). Spa2p thus appears to act as a negative regulator of the cell integrity pathway, but the physiological meaning of this regulation remains mysterious.

Transcriptional Regulation

Several proteins have been identified as downstream substrates of the Slt2p MAPK. One of these, Rlm1p, a transcriptional regulatory protein, is discussed below. Two others, Swi4p and Swi6p (275), together form the transcription factor SBF (10, 42, 382, 432). Swi4p is the DNA binding subunit and transcriptional activator of SBF and is required for normal expression of the G₁ cyclin genes CLN1, CLN2, PCL1, and PCL2 at the G₁/S transition (79, 212, 308, 332, 348, 362, 448). Swi6p is more of a regulatory subunit, because loss of Swi6p leads to constitutive intermediate levels of CLN1 and CLN2 expression (101, 266). Cln1p and Cln2p are G₁ cyclins that complex with the cyclin-dependent kinase Cdc28p and thereby activate the G₁/S transition (78, 394). The complex of Pcl1p and Pcl2p with a different cyclin-dependent kinase, Pho85p, also promotes entry into S phase (121, 308). SBF is therefore a key regulator of the G₁/S transition.

The cell integrity pathway, like SBF, appears to be important for the G₁/S transition. One mechanism by which this pathway may promote this transition is through regulation of the SBF transcription factor. Phenotypic similarities between cell integrity pathway mutants and SBF mutants suggest this connection. For example, strains with mutations in SLT2 or PKC1 show a reduction in expression of several SBF-regulated genes expressed in late G₁ or early S phase, including the G₁ cyclin genes PCL1 and PCL2 (275) and several cell wall genes (181). In addition, the temperature-sensitive, osmosis-remedial cell lysis phenotype of a swi4Δ strain in some strain backgrounds (275, 348) is similar to that of slt2Δ strains. The cell integrity pathway appears to regulate SBF through Slt2p-catalyzed phosphorylation of the SBF subunits Swi4p and Swi6p (275). The amount of Swi6p phosphorylation in cells correlates well with the activity of Slt2p under a variety of conditions, e.g., changes in Slt2p expression level and growth temperature (275). This phosphorylation is likely to be the direct result of Slt2p-catalyzed phosphorylation because Slt2p is associated with SBF in cells and Slt2p phosphorylates the SBF subunits in vitro (275). Not all SBF-regulated genes require the MAPK cascade for proper expression; for example, CLN1 and CLN2 expression is unaltered by slt2Δ (275). Cell integrity pathway mutations are lethal when combined with swi4Δ (182), indicating that SBF and the MAPK cascade have some nonoverlapping functions.

SBF plays an important role in cell cycle progression at the G₁/S transition and is therefore the target of several regulatory pathways. SBF function is regulated by the G₁ cyclin Cln3p-Cdc28p kinase (100, 449). Because the cell integrity pathway is activated by Cdc28p (see above), it may be that this pathway mediates part of the cell cycle regulation of SBF activity. However, as discussed above, the MAPK cascade is required for expression of only a subset of SBF-regulated genes. Different DNA sequences appear to mediate SBF-dependent transcription (79, 362, 448), and so it seems reasonable to expect that SBF interacts with other, still unknown proteins. These proteins would help determine not only which subset of SBF targets is regulated by specific stimuli but also whether the gene expression is repressed or activated. The response to DNA damage illustrates this last point. DNA-damaging agents also induce the phosphorylation of Swi6p (174, 433), but the effects on SBF function are complex, inhibiting the SBF-dependent expression of CLN1 and CLN2 (433) and activating the SBF-dependent expression of a subset of the DNA damage-inducible genes (174). The DNA damage-induced phosphorylation of Swi6p appears to be mediated by the casein kinase I isoform Hrr25p (174). The way in which phosphorylation of SBF by Slt2p or other protein kinases affects its function is unknown. One possibility, translocation to the nucleus of Swi6p in late G₁, does not appear to be stimulated by phosphorylation but is instead activated by dephosphorylation (434).

Another target of the cell integrity pathway is Rlm1p, a member of the MADS box family of transcription factors (430). Rlm1p and the related yeast protein Smp1p are most similar to the mammalian MEF2 subfamily of MADS box proteins (515). This family of transcription factors generally form heterodimers with other transcription factors and are known to be MAPK targets in mammalian cells (159, 206) as well as in yeast cells (103, 483, 484). Overexpression of a hyperactivated cell integrity pathway MEK (Mkk1p) inhibits the growth of yeast. Mutations in rlm1 were initially isolated as suppressors of the toxicity of activated Mkk1p overexpression (483); rlm1 cells are also resistant to the growth inhibition induced by overexpression of an activated allele of RHO1 (103). A similar strategy, in which the toxicity of pathway hyperactivation is used to isolate suppressors in pathway genes, has also been used with great success in studies of the HOG pathway (281) and a related pathway in fission yeast (312, 428).

Rlm1p appears to be a downstream substrate of the cell integrity pathway MAPK Slt2p. Rlm1p is a substrate for the MAPK in vitro and shows heat stress-induced, SLT2-dependent phosphorylation in vivo (484). Rlm1p also interacts with Slt2p in the two-hybrid system (483). Expression of a reporter construct consisting of a MEF2-consensus DNA target element fused to lacZ is both RLM1 and SLT2 dependent (103). The COOH-terminal two-thirds of Rlm1p is sufficient to mediate SLT2-dependent transcriptional induction. This was demonstrated (484) by replacing the NH₂-terminal, putative DNA binding domain of Rlm1p with that of LexA and expressing this fusion protein in a strain containing a LexA promoter-lacZ construct. Strong expression of the lacZ-encoded β-galactosidase was induced by the LexA protein-Rlm1pΔN fusion; mild heat stress induced expression by two- to threefold. Deletion of SLT2 or BCK1 reduced the expression to ∼5% or less of control. Rlm1p has three functional domains: the NH₂-terminal MADS box DNA binding domain, a central domain that is the target for phosphorylation and regulation by the MAPK Slt2p, and a COOH-terminal region that is required for transcriptional activation (484).

Slt2p is not the only protein kinase to interact with Rlm1p. A two-hybrid screen for Rlm1p-interacting proteins identified Mlp1p (484), a putative protein kinase with strong sequence similarity to Slt2p and placed by sequence alignment analysis in the MAPK family (179). Mlp1p binds to the same region on Rlm1p as does Slt2p (484). Deletion of MLP1 has no phenotype by itself except for its effect on the caffeine sensitivity of mutants with different cell integrity pathway mutations: mlp1Δ enhances the caffeine sensitivity of slt2Δ, and overexpression of Mlp1p suppresses the caffeine sensitivity of bck1Δ. This suppression of part of the bck1Δ phenotype requires Rlm1p, because overexpression of Mlp1p does not suppress the caffeine sensitivity of a bck1Δ rlm1Δ strain (484).

Let us divert our attention briefly and consider the structure of Mlp1p and its significance for MAPK function. Mlp1p is a bit unusual in that one of the two consensus phosphorylation sites, a conserved threonine, in the MAPK activation domain of Mpl1p is replaced with lysine and a consensus lysine in the catalytic site region is replaced by arginine. These substitutions have been introduced into other yeast MAPKs to make them inactive (142, 241, 271, 411)! The Lys in the active site is needed for phosphotransfer from ATP, and phosphorylation of the Thr (and Tyr) by a MEK is needed for binding protein substrate; phosphorylation of tyrosine is less important. Perhaps like Kss1p (69, 278), Mlp1p has a function in the inactive state and has been subject to selective pressures that favor a less functional protein kinase. The tyrosine in the MAPK activation domain of Mlp1p is required for suppression of the caffeine sensitivity of bck1Δ, suggesting that the function of Mlp1p might be regulated by a MEK-catalyzed phosphorylation of this residue.

Genetic analysis of Rlm1p function makes it clear that there are two outputs from the cell integrity pathway MAPK cascade and that they involve different transcription factors and regulate different sets of genes involved in different cellular responses. Mutants lacking pathway kinases have a pleiotropic phenotype that includes temperature-sensitive growth (73, 242) and sensitivity to caffeine (73). The way caffeine affects cells is not known at the molecular level, but it enhances the cell lysis phenotype of slt2 mutants (292). Caffeine probably affects some aspect of cell wall synthesis, because a large subset of other mutants with altered cell wall construction are also caffeine sensitive (267, 376, 389). In any case, Rlm1p appears to mainly affect the caffeine sensitivity of pathway mutants. Overexpression of RLM1 suppresses the caffeine sensitivity of the bck1Δ MEKK mutant but not its temperature sensitivity (484). Cells with a rlm1Δ mutation are caffeine sensitive like bck1Δ but are not temperature sensitive (483). The temperature-sensitive phenotype of cell integrity pathway mutants may reflect decreased signaling to the SBF transcription factor, since swi4 mutants are temperature sensitive (275, 348). To determine whether SBF and Rlm1p are the only targets of Slt2p, it would be interesting to determine whether the phenotype of a swi4Δ rlm1Δ double mutant copies that of a slt2Δ mutant.

Under conditions of mild heat stress, where the activity of the Slt2p kinase is increased by ∼100-fold or more, there is either a 2- to 3-fold (484) or no (103) increase in Rlm1p-dependent expression of a reporter construct. It would thus appear that Rlm1p provides a strong, relatively constant induction to a subset of genes regulated by the MAPK cascade. A computer-assisted analysis of the genome for genes with a MEF2 consensus target sequence in their promoter uncovered several cell wall-related genes including those that encode the β-glucan synthesis regulator Hkr1p, the mannosyltransferase Ktr2p, and the flocculation protein Flo1p (103). However, whether the expression of these genes is cell integrity pathway-regulated has not yet been determined.

Two closely related proteins, Nhp6Ap and Nhp6Bp, also appear to act downstream of the MAPK cascade and may play a role in gene regulation (74). Nhp6p is a nuclear DNA binding protein related to mammalian HMG1 chromatin-associated proteins (216, 217). Cells lacking both NHP6A and NHP6B have a phenotype (74) similar to that of slt2Δ mutants (241), and a double mutation, nhp6AΔ nhp6BΔ, does not enhance the phenotype of slt2Δ mutants (74), indicating that Nhp6p acts in the same pathway as Slt2p. Overexpression of NHP6A or NHP6B suppresses a MEKK mutation (bck1Δ), but overexpression of BCK1 does not suppress nhp6AΔ nhp6BΔ, demonstrating that Nhp6Ap and Nhp6Bp may act downstream of Slt2p (74). The way in which these two nuclear proteins help mediate cell integrity pathway responses is not yet known, but direct interactions with the MAPK Slt2p could not be detected (74).

Regulation of a MAPK Cascade by a Three-Component System

As previously summarized, one of the upstream branches of the HOG pathway contains a phosphorelay system of three components, Sln1p, Ypd1p, and Ssk1p (379), that regulates the activity of two MEKKs Ssk2p and Ssk22p (279) (Fig. ​(Fig.6).6). These MEKKs in turn regulate the activity of the downstream MEK Pbs2p and MAPK Hog1p. A relatively linear signaling mechanism connects the activity of Sln1p, a putative osmosensor for the HOG pathway (281), to the activity of the MEKKs. Thus, Ssk1p is required for activation of the MEKK Ssk2p, a role that appears to involve its binding to the NH₂-terminal regulatory domain of Ssk2p (279, 377). In addition, it is the unphosphorylated form of Ssk1p that appears to activate the HOG pathway, as measured by increased phosphorylation of the MAPK Hog1p (279). These findings are particularly important because they show that one major class of signaling device, a three-component system, is physically and functionally connected to another major class of signaling device, a MAPK cascade.

Open in a separate window

FIG. 6

Three-component system of the HOG pathway. Phosphate is transferred from ATP to a histidine residue on the histidine kinase domain (octagon) of Sln1p and from there to an aspartate residue on the receiver domain (triangle) of a separate molecule of Sln1p. Whether phosphate can be transferred from the histidine kinase domain to a receiver domain on the same Sln1p polypeptide chain as the former remains to be determined. Phosphate is then transferred from Sln1p to the histidine kinase Ypd1p, to either of two receiver domain proteins Ssk1p or Skn7p, and then to water. The phospho and dephospho forms of Skn7p have different functions. The dephospho form of Ssk1p is an activator of the HOG pathway MEKKs Ssk2p and Ssk22p; the phospho form of Ssk1p has no known function.

Modulation of the activity of the MEKK regulator Ssk1p through changes in its phosphorylation state is mediated through the three-component system of Sln1p, Ypd1p, and Ssk1p functioning as a phosphorelay system. Sln1p appears to contain three functional domains (353, 379). Starting at its NH₂ terminus, the putative osmosensor domain of Sln1p has two predicted membrane-spanning hydrophobic segments with an intervening extracellular hydrophilic segment. Moving toward the COOH terminus, the putative sensor domain is followed by a histidine kinase domain and then a receiver domain containing a conserved aspartate residue. The postulated role of Sln1p as an osmosensor for the HOG pathway can be traced in part to the similarity between Sln1p and the Escherichia coli osmosensor EnvZ. The overall structure of EnvZ is similar to that of Sln1p, except that EnvZ lacks a COOH-terminal receiver domain (381).

The following mechanism for transferring phosphate from ATP to the MEKK regulator Ssk1p was revealed by an elegant series of genetic and biochemical experiments carried out by the Saito group (379). Figure ​Figure66 shows the route taken by that phosphate in the three-component system of the HOG pathway. First, Sln1p catalyzes the transfer of phosphate from ATP to a conserved histidine residue in its histidine kinase domain. In vitro reconstitution experiments with two mutant Sln1p proteins show that phosphate is transferred from the histidine of one Sln1p to the conserved aspartate residue in the COOH-terminal receiver domain of a second Sln1p. The intermolecular nature of this phosphotransfer reaction is confirmed by genetic analysis. Expression of a Sln1p with a defective or absent receiver domain or a defective or absent histidine kinase domain does not complement sln1Δ, but coexpression of both mutant proteins does complement sln1Δ (126, 353, 379). Whether phosphotransfer can occur between histidine kinase and receiver domains on the same Sln1p molecule is unknown.

In vitro experiments (379) show that phosphate attached to the aspartate residue in the receiver domain of Sln1p is transferred to a histidine residue on Ypd1p and then from there to a conserved aspartate residue in the receiver domain of Ssk1p. These three proteins therefore form a phosphorelay system that moves phosphate from ATP to Sln1p-His to Sln1p-Asp to Ypd1p-His to Ssk1p-Asp. Mutation of any one of these four conserved amino acids in the three proteins completely blocks the negative regulation of the HOG pathway and results in hyperactivation of the MAPK cascade and subsequent toxicity (279, 281, 353, 379). Because the same phosphate is being transferred, there is no signal amplification in this process. The multistep nature of this phosphorelay system potentially allows regulation at any one of the steps. Moreover, observations of bacterial two-component signaling pathways (361, 447, 501) instruct us that both phosphorylation and dephosphorylation reactions are potential targets of regulation. Whether and how other signaling proteins in yeast regulate the activity of Sln1p-Ypd1p-Ssk1p is an important issue about which there is no current information.

From the model described above, Sln1p and Ypd1p are predicted to act as negative regulators of the HOG pathway MAPK cascade. Deletion of SLN1 or YPD1 is lethal unless HOG pathway signaling is blocked by deletion of SSK1, SSK2, PBS2, or HOG1 (281, 379). Expression of Ssk1p with its receiver domain aspartate changed to a nonphosphorylatable asparagine is also toxic unless the downstream HOG pathway is blocked by mutation (279). Overexpression of the MEKKs Ssk2p or Ssk22p without their regulatory NH₂ terminus also induces HOG pathway hyperactivation and lethality unless the MEK Pbs2p or MAPK Hog1p is deleted. On the other hand, deletion of the upstream regulator SSK1 cannot suppress the lethality of Ssk2pΔN overexpression (279). Negative regulation of MAPK cascades is therefore important in yeast because in three separate examples, i.e., the pheromone response pathway, the cell integrity pathway, and now the HOG pathway, hyperactivation is lethal. The HOG pathway is also negatively regulated by the action of the tyrosine-specific protein phosphatases Ptp2p and Ptp3p (153, 195, 352), which dephosphorylate the MAPK Hog1p (193, 500). Another negative regulator of the HOG pathway is the protein phosphatase Ptc1p (197, 280, 281), whose substrate is unknown. Removal or inactivation of these negative regulators causes reduced cell growth through hyperactivation of the HOG pathway. For example, ptp2Δ is lethal in combination with ptc1Δ (280). Overexpression of a downstream negative regulator Ptp2p can in turn suppress the lethal phenotype of a mutant lacking Sln1p, an upstream negative regulator (281).

Unphosphorylated Ssk1p activates the MAPK cascade (279), and Sln1p-Ypd1p relays phosphate from ATP to Ssk1p (379). The phosphorylation of Ssk1p is then predicted to inhibit its ability to activate the MEKKs Ssk2p and Ssk22p. This model makes several predictions. Increasing the osmolarity of the medium should inhibit the upstream phosphorelay proteins, leading to dephosphorylation of Sln1p, Ypd1p, and Ssk1p. Although the effect of hyperosmotic stress on the phosphorylation state of Sln1p or Ssk1p has not yet been reported, increasing the osmolarity of the medium induces a decrease in Ypd1p phosphorylation (379). This response provides stronger evidence that the upstream three-component is actually sensing a change in osmolarity. One part of the model that seems incomplete is whether all the signaling in the system is mediated through modulation of phosphorylation reactions or whether phosphatase activity is being modulated by hyperosmotic stress. Another major unanswered question is how Sln1p senses a change in osmolarity. This is a tough question to answer. The function of the related osmosensor EnvZ has been studied for about 10 years (381), and its sensory mechanism is still under investigation (246, 360, 461–463, 506).

Activation of the MEKK Ssk2p by Ssk1p appears to be mediated by modulation of an Ssk2p autophosphorylation reaction similar to that used to induce MEKK activation in other MAPK cascades (92). Addition of unphosphorylated Ssk1p in vitro induces Ssk2p autophosphorylation. Autophosphorylation and activation of Ssk2p in the unstimulated cell appear to be blocked by an inhibitory domain at the NH₂-terminus (377), similar to that found in other MEKKs (314, 467). The receiver domain of Ssk1p binds close to this region of Ssk2p (377) and is therefore in a position to relieve the intramolecular inhibition of Ssk2p kinase activity. The binding of Ssk1p to Ssk2p is critical because increasing the external osmolarity induces autophosphorylation of wild-type Ssk2p but not of a Ssk2p mutant lacking the Ssk1p binding site (377).

A Second Osmosensor and the Role of Pbs2p as a Scaffold Protein

A second upstream branch of the HOG pathway (Fig. ​(Fig.5)5) contains the putative membrane protein Sho1p and Ste11p, a MEKK found on other yeast MAPK cascades. As discussed above, the downstream protein kinases Pbs2p and Hog1p are regulated by a cascade of signaling proteins including a three-component system and two redundant MEKKs, Ssk2p and Ssk22p. However, single or double mutants with this part of the HOG pathway blocked upstream of Pbs2p are not osmosensitive like a pbs2Δ mutant (279, 281). Moreover, a ssk2Δ ssk22Δ mutant still shows induction of Pbs2p phosphorylation by hyperosmotic stress (378), showing that another MEKK can phosphorylate Pbs2p. To uncover genes in the missing upstream part of the HOG pathway, mutations that confer osmosensitivity only when combined with ssk2Δ ssk22Δ were isolated. This screen yielded mutations in SHO1, encoding a putative membrane protein with a COOH-terminal SH3 domain (279). The SH3 domain of Sho1p binds a proline-rich segment in the NH₂ terminus of Pbs2p (279). A point mutation in this segment of Pbs2p blocks its interaction with Sho1p and has an osmosensitive phenotype in combination with ssk2Δ ssk22Δ but not by itself (279).

Increasing osmolarity induces the phosphorylation of a catalytically inactive form of Pbs2p, even in an ssk2Δ ssk22Δ strain (378). Based on its structure, Sho1p cannot be the missing protein kinase. Rather, Ste11p was identified as the third MEKK of the HOG pathway (378). An important observation is that a ste11Δ mutation confers osmosensitivity only when combined with ssk2Δ ssk22Δ (378); in fact, ste11 was isolated by using the same screen that yielded sho1. Ste11p is required for in vivo hyperosmotic stress-induced phosphorylation of Pbs2p in a strain lacking the MEKKs Ssk2p and Ssk22p (378). In addition, Ste11p lacking its NH₂ terminus but retaining its kinase domain can phosphorylate Pbs2p in vitro (378).

The MEKK Ste11p is thus a signaling device that is used in three functionally distinct MAPK cascades in yeast (Fig. ​(Fig.1)1) (257, 378). Expression of a hyperactivated form of Ste11p (Ste11pΔN), made by deletion of its NH₂-terminal regulatory domain, is toxic to cells (51, 446). Deletion of the downstream pheromone response pathway and filamentation-invasion pathway MEK Ste7p or deletion of the downstream HOG pathway MEK Pbs2p does not suppress the lethality of Ste11pΔN expression. However, deletion of both MEKs, Ste7p and Pbs2p, completely suppresses the lethality of Ste11pΔN expression (378). The ability of a ste11Δ mutation to confer osmosensitivity to an ssk2Δ ssk22Δ strain is not the result of reduced signaling by the other Ste11p-dependent pathways. Deletion of STE5 specifically blocks the pheromone response pathway; deletion of STE20 or STE7 blocks the pheromone response and filamentation-invasion pathways. These mutations, unlike ste11Δ, do not show synthetic osmosensitivity with ssk2Δ ssk22Δ (378).

The ability of Ste11p to function in separate pathways requires stable associations with pathway-specific proteins. Some of these interactions have been identified. For example, Ste11p interacts with the pheromone response pathway-specific scaffold protein Ste5p (61, 187, 287, 383). In the HOG pathway, the MEK Pbs2p serves as a scaffold protein, interacting with Ste11p (378) and Sho1p (279). A Ste11p binding scaffold protein specific for the filamentation-invasion pathway may exist but has not yet been identified. The pheromone response pathway and HOG pathway scaffold proteins may help bring together proteins that do not interact or interact only weakly but that are required for pathway signaling. For example, Ste11p shows no or very weak interaction with the MEK Ste7p in the two-hybrid system (61, 287, 383) and Ste5p binds both proteins (61, 187, 287, 383). Also, interaction between Sho1p and Ste11p is undetectable (378) and Pbs2p binds both proteins (279, 378).

The mechanism by which Ste11p is activated by hyperosmotic stress is unknown. For other MAPK cascades in yeast, there is an upstream activating protein kinase: Ste20p, the cell integrity pathway Pkc1p, or the spore wall assembly pathway-specific Sps1p. Ste20p (which acts on Ste11p in other pathways) and the Ste20p-related protein kinases Cla4p and Skm1p are not required for high-osmolarity growth of an ssk2Δ ssk22Δ mutant and are therefore not part of the Sho1p-Ste11p upstream arm of the HOG pathway. One of the important unsolved problems concerning the HOG pathway is therefore how the MEKK Ste11p is activated in response to hyperosmotic stress.

Mechanisms of Osmosensing

The mechanism by which the HOG pathway senses a change in extracellular osmolarity remains a major mystery. Defining which protein is an osmosensor, which physical parameter is being sensed, and how the sensor detects changes in that parameter are therefore important goals. Genetic and biochemical analysis of the upstream regulators of the MAPK cascade (279, 281, 378, 379) suggests that there are at least two different osmosensors, with Sln1p and Sho1p as the most likely candidates. One would expect that an osmosensor would span the plasma membrane, and Sln1p and Sho1p are each putative membrane proteins. However, the predicted structures of Sln1p and Sho1p are quite different from each other, suggesting a different role or function in sensing hyperosmotic stress. Measurements of the time and osmosis dependence of Hog1p phosphorylation in different mutant strains suggest that the two upstream branches of the HOG pathway may sense different physical parameters or react differently to the same parameter. Cells with SHO1 and SSK22 deleted show hyperosmotic stress-induced, Ssk2p-dependent changes in Hog1p phosphorylation (279). These changes are similar to those observed in wild-type cells (44, 279). In both cases, Hog1p phosphorylation appears within 1 min after addition of NaCl and is induced by 100 mM NaCl, with a maximum response at ∼300 mM. In cells lacking SSK2 and SSK22, hyperosmotic stress-induced, Sho1p- and Ste11p-dependent Hog1p phosphorylation cannot be detected until up to 300 mM NaCl is added and then only after 5 min since the addition of salt (279). The Ssk2p-dependent HOG pathway activation in sho1Δ ssk22Δ cells is thus faster and more sensitive than the Sho1p- and Ste11p-dependent response in ssk2Δ ssk22Δ cells.

Whether Sln1p and Sho1p are bona fide osmosensors remains uncertain. They are clearly required for hyperosmotic stress-induced phosphorylation of the HOG pathway MEK and MAPK. However, these two putative membrane proteins could function to correctly position other signaling proteins in the pathway to receive signal from the real osmosensor(s). To determine whether Sln1p or Sho1p are osmosensors, it will eventually be necessary to reconstitute the system from purified components and show that hyperosmotic stress regulation still occurs. One would predict that an osmosensor would alter its conformation or be chemically modified in response to osmotic stress. This has not yet been shown for Sln1p or Sho1p. However, the Ypd1p component of the Sln1p-Ypd1p-Ssk1p phosphorelay system in the HOG pathway does show, as predicted by the current model (379, 501), a decrease in phosphorylation after exposing cells to an increase in osmolarity (379). Turning to Sho1p, this putative membrane has an SH3 domain that interacts with Pbs2p (279). SH3 domains are involved in protein-protein interactions in yeast (29, 130, 298) and many other eukaryotes (67, 363). SH3 domains can regulate the catalytic activity of associated protein molecules (150, 168). Whether changes in osmolarity alter the interaction between the SH3 domain of Sho1p and Pbs2p has not yet been determined.

The physical parameter that is actually sensed by the HOG pathway has not yet been determined. This is a very difficult area in which to formulate specific molecular hypotheses. Increasing the osmolarity of the medium should reduce turgor pressure and decrease cell volume. Large-scale cell volume changes are not likely to be involved, because Hog1p phosphorylation reaches a maximum upon addition of NaCl to a final concentration of ∼300 mM (44, 279), or approximately 500 mOsm. This change in osmolarity is too small to cause large changes in cell volume (15). The internal osmolality of nongrowing yeast cells is estimated at ∼600 mOsm (15), and growing cells have an even higher internal osmolarity (437). The osmolality of normal rich growth medium (YEPD) is approximately 250 mOsm. Thus, maximal activation of the HOG pathway appears to occur when the external osmolarity (∼750 mOsm) is roughly similar to the internal osmolarity (>600 mOsm), suggesting the possibility that the HOG pathway is activated by a loss of turgor pressure.

How might a membrane protein such as Sln1p or Sho1p sense a change in turgor pressure? Turgor pressure is a stress that creates strain or tension in the cell surface. Thus, changes in tension in any of the components of the cell surface—the cell wall, plasma membrane, or plasma membrane-associated actin cytoskeleton—are potential signals. The cell wall is a complicated structure with many connections to underlying plasma membrane proteins. Changes in tension on these connections may cause conformational changes in cell wall-interacting plasma membrane proteins acting as osmosensors. Turgor may also create tension in the plasma membrane bilayer that can be a signal. Mechanosensitive ion channels in the yeast plasma membrane are activated by such tension (155). The cell integrity pathway is activated by chlorpromazine (201), a drug that intercalates into one leaflet of the lipid bilayer and creates tension (294). Perhaps the loss of this tension after an increase in external osmolarity may induce conformational changes in integral membrane proteins acting as osmosensors for the HOG pathway.

Cortical patches of actin filaments could be important for osmosensing. Electron micrographs of the yeast plasma membrane reveal many places where the membrane is invaginated (327). These membrane invaginations are wrapped with a helical bundle of actin filaments (327), forming a cortical actin patch (41). Changes in turgor could cause these invaginations to become deeper or shallower as the external osmolarity decreases or increases, respectively. The helical bundle of actin filaments may become correspondingly stretched or relaxed, leading to changes in conformation of actin-associated proteins acting as osmosensors. Whether Sln1p and/or Sho1p is present in these areas of the plasma membrane has not yet been tested.

Regulation of Gene Expression

An increase in external osmolarity is stressful to yeast and induces many physiological changes (176, 282). These include collapse of the osmotic gradient, temporary cessation of growth, loss of an organized actin cytoskeleton (63), temporary decrease in protein synthesis (282), induction of a subset of the heat shock proteins (33, 282, 411, 474), and increases in the concentration of intracellular solutes and macromolecules. If NaCl is used to increase the external osmolarity, there is additional stress due to the specific toxicity of this salt (418, 419). Increasing the osmolarity of the medium induces increased expression of many genes as measured by monitoring changes in mRNA or protein levels. This is a complex process in which the panoply of genes expressed depends on the solute used to increase the osmolarity and the extent of the increase in osmolarity. For a given osmotic stimulus, different genes show different time-dependent patterns of expression and extent of induction (33, 172). This suggests that the activities of many different cis- and trans-acting mechanisms are involved in determining the pattern of gene expression under conditions of hyperosmotic stress. Based on comparison of wild-type and HOG pathway mutants, the HOG pathway is required for the increased expression of many but not all of these genes (4, 5, 172, 343, 411). This analysis is not complete, but with new technology for examining the expression of all yeast genes now available (95, 232, 496), it should be completed soon.

Among the genes that show high-osmolarity-induced, HOG1-dependent expression, there appear to be at least three different regulatory mechanisms. One class of genes includes those that are induced by other stresses such as heat shock and that have a DNA sequence element called STRE in their promoter (123, 402, 431). STRE is different from the classical heat shock element (374). The catalase gene CTT1 is a member of the STRE-regulated class of stress genes (411). Another class consists of genes whose induction does not involve STRE and appear to respond more specifically to hyperosmotic stress, for example, the glycerol-3-phosphate dehydrogenase gene, GPD1 (116). Finally, the aldehyde dehydrogenase gene ALD2 has no STREs and, in contrast to GPD1, is induced by a derepression mechanism (315). The transcription factors responsible for the induction of the last two classes of hyperosmotic stress-induced genes have not been identified. STRE-mediated gene expression appears to be mediated in part by Msn2p and Msn4p (148, 295, 407). Whether Hog1p directly interacts with Msn2p and Msn4p is not known. Cells lacking MSN2 and MSN4 show a reduction in the magnitude of hyperosmotic stress induction of several STRE-regulated genes including CTT1; however, the fold induction was relatively unchanged and some PBS2-dependent expression remained (295). Increasing osmolarity induces Msn2p to concentrate in the nucleus (148). Osmotic stress-induced nuclear translocation of Msn2p is unaffected by deletion of Hog1p. Instead, translocation of Msn2p to the nucleus appears to be negatively regulated by cyclic AMP and protein kinase A (148). Other proteins that appear to be required for hyperosmolarity-induced, HOG1- and STRE-dependent gene expression are Rox3p, a transcription factor, and Rts1p, a protein with homology to the regulatory subunit of phosphatase 2A (123). As with Msn2p and Msn4p, the mechanisms of interaction of Rox3p and Rts1p with the MAPK cascade need further investigation.

The regulation of gene expression in osmotically stressed or salt-stressed cells is clearly complex, with the HOG pathway cooperating with other pathways to regulate gene expression. For example, STRE-dependent gene expression is negatively regulated in many cases by the RAS-cyclic AMP-protein kinase A pathway (27, 39, 149, 286, 411, 474, 489, 492). This negative regulation appears to be mediated by cyclic AMP and protein kinase A-mediated inhibition of the STRE-binding transcription factors Msn2p and Msn4p (148, 436). Regulation of a STRE-lacZ reporter gene by the cyclic AMP pathway appears to be additive with that by the HOG pathway. For example, deletion of the Bcy1p regulatory subunit of protein kinase A causes constitutive activation of the protein kinase A (299) and reduces the expression of a STRE-lacZ reporter gene in response to hyperosmotic stress (411). However, bcylΔ cells still show a relatively strong (10- to 15-fold) HOG1-dependent induction of the STRE-lacZ reporter gene by hyperosmotic stress (411). Increasing the osmolarity of the medium decreases cellular cyclic AMP levels (289), suggesting that a combination of a decrease in protein kinase A activity plus an increase in the MAPK Hog1p activity mediates the hyperosmotic stress induction of certain STRE-containing genes.

Another gene that shows combined regulation by the HOG pathway and a second pathway is ENA1, a gene encoding a P-type Na⁺-ATPase responsible for Na⁺ efflux from yeast (138, 161, 488). At low salt concentrations, induction of ENA1 expression requires HOG1; at higher salt concentrations, the calcium-activated phosphatase calcineurin is required for ENA1 induction (289).

A major unsolved question about the budding yeast HOG pathway is how Hog1p mediates the hyperosmotic stress-induced changes in gene expression. Hog1p (like the MAPK Slt2p [440]) is a strong transcriptional activator when fused to the Gal4p DNA binding domain (154). This activation may reflect an innate property of Hog1p or the recruitment of still-unidentified transcriptional activators to Hog1p. No substrates that act downstream of the MAPK Hog1p have so far been identified.

Regulation of the Pheromone Response Pathway

Each of the MAPK cascades in yeast responds to a specific signal, e.g., pheromone, an increase in osmolarity, nitrogen limitation, an increase in temperature, or induction of sporulation. One function of the HOG pathway MAPK cascade is to negatively regulate the pheromone response pathway and thereby inhibit activation by an inappropriate signal: hyperosmotic stress. The pheromone pathway MAPK Fus3p shows a small increase in tyrosine phosphorylation in response to high osmolarity, and deletion of HOG1 or PBS2 greatly enhances this response (158). Deletion of genes that code for members of the pheromone response pathway suppresses the osmosensitive growth phenotype of hog1 and pbs2 mutants (88). This suggests that an inappropriately activated pheromone pathway is deleterious for growth in high-osmolarity medium. Additional evidence for a role for HOG pathway regulation of the pheromone response pathway is the observation that cells lacking Hog1p or Pbs2p have an increased phosphorylation, transcription, and growth arrest response to exogenous pheromone addition (158, 445). The way in which the HOG pathway regulates the pheromone pathway and the mechanism of induction of pheromone response pathway by hyperosmotic stress are both unknown.

Regulation of Cell Growth

Transferring wild-type cells from normal growth medium to high-osmolarity medium induces a temporary growth arrest that is correlated with a loss in the normal distribution of the actin cytoskeleton (63). Actin filament patches, abundant in the growing bud (41), become more uniformly distributed between mother and daughter cells immediately after an increase in osmolarity and then later return to the original location upon resumption of growth (63). Mutants with mutations in the HOG pathway MAPK cascade show a similar initial response to hyperosmotic stress but later fail to restore actin filaments to the original bud. Instead, hog1Δ or pbs2Δ cells initiate the formation of a new bud, localize actin filaments to this new bud, and abandon the growth of the previous bud (45). This phenotype suggests that one function of the HOG pathway is to correctly position cell growth and division after hyperosmotic stress. The molecular mechanism by which Hog1p regulates this transition from arrested growth to resumption of growth is not known, but the HOG pathway appears to be required for sustaining cell cycle arrest in G₂ after an increase in osmolarity (7). Thus, the initiation of a new bud in the HOG pathway mutants may reflect cell cycle progression from S or G₂ to G₁ in a growth-arrested cell and formation of a new bud when growth resumes.

Turning off the MAPK Cascade

Addition of NaCl or other solutes to increase the extracellular osmolarity induces an increase in Hog1p tyrosine phosphorylation followed by a decrease back to near prestimulus levels (44, 193, 500). When cells contain a catalytically inactive Hog1p mutant in place of the wild-type protein, this mutant Hog1p does not show the later decrease in tyrosine phosphorylation after hyperosmotic stress (411, 500). This mutant is also osmosensitive (411). Thus, Hog1p-activated processes are required not only for adaptation to hyperosmotic stress but also for down-regulation of the HOG pathway. Constitutive activation of the pathway is lethal (279, 281, 500), even under hyperosmotic conditions (281), and so turning down the pathway activity is a vital function for yeast.

Hog1p could negatively regulate itself or the upstream part of the HOG pathway in several different ways. It could phosphorylate and inhibit upstream regulatory proteins. Alternatively, it could activate metabolic responses such as glycerol synthesis that restore turgor and turn off the osmosensors for the pathway. These two possibilities, involving a reduction in the rate of Hog1p phosphorylation, have not yet been investigated. However, one mechanism proposed (193, 500) for negative-feedback control of the HOG pathway is through Hog1p stimulation of the rate of Hog1p dephosphorylation by regulation of the expression or activity of Hog1p-specific phosphatases. This type of mechanism is known to mediate down-regulation of other MAPK cascades (70, 93, 152, 312). Mutational analysis (411) indicates that, like other MAPKs, Hog1p activity requires dual phosphorylation on a threonine residue and a tyrosine residue. Ptp2p and Ptp3p mediate the tyrosine dephosphorylation of Hog1p (193, 500); the threonine-specific phosphatase for Hog1p has not yet been identified. Thus, the late decrease in Hog1p phosphorylation after hyperosmotic stress is partially blocked by deletion of the tyrosine phosphatase Ptp2p and fully blocked by deletion of both Ptp2p and the Ptp2p-related phosphatase Ptp3p; a ptp3Δ mutation by itself has little effect on Hog1p phosphorylation (193, 500). Ptp2p thus appears to be the major tyrosine phosphatase for the HOG pathway MAPK, with Ptp3p playing a minor role; this situation is reversed in the pheromone response pathway, where Ptp3p is more important than Ptp2p in regulating Fus3p activity (519).

Whether Hog1p down-regulation is mediated through regulation of Ptp2p and Ptp3p expression or activity remains an open question. For example, hyperosmotic stress does induce HOG1-dependent increases in PTP2 and PTP3 mRNA levels (193), but the increases are relatively small and, in a separate study, barely detectable (500). Hog1p has been proposed to regulate Ptp2p and Ptp3p activity (500), but this possibility has not yet been experimentally investigated.

We thank Ed Winter and Elaine Elion for their helpful comments on the pheromone response pathway and the spore wall assembly pathway, respectively, and members of the Gustin laboratory for helpful comments on all the sections.