|Home | About | Journals | Submit | Contact Us | Français|
The yeast histidine kinase, Sln1p, is a plasma membrane-associated osmosensor that regulates the activity of the osmotic stress MAP kinase pathway. Changes in the osmotic environment of the cell influence the autokinase activity of the cytoplasmic kinase domain of Sln1p. Neither the nature of the stimulus, the mechanism by which the osmotic signal is transduced nor the manner in which the kinase is regulated is currently clear. We have identified several mutations located in the linker region of the Sln1 kinase (just upstream of the kinase domain) that cause hyperactivity of the Sln1 kinase. This region of histidine kinases is largely uncharacterized, but its location between the transmembrane domains and the cytoplasmic kinase domain suggests that it may have a potential role in signal transduction. In this study, we have investigated the Sln1 linker region in order to understand its function in signal transduction and regulation of Sln1 kinase activity. Our results indicate that the linker region forms a coiled-coil structure and suggest a mechanism by which alterations induced by osmotic stress influence kinase activity by altering the alignment of the phospho-accepting histidine with respect to the catalytic domain of the kinase.
Recent characterization of the osmotic response pathway in Saccharomyces cerevisiae has uncovered a complex pathway consisting of a prototypical MAP kinase pathway regulated by a two-component phosphorelay system (Posas et al., 1996). Two-component signal transduction pathways were first characterized in bacteria and have since been identified in fungi, slime moulds and plants (Swanson and Simon, 1994; Loomis et al., 1997; Mizuno, 1998). In S. cerevisiae, there is a single two-component-type phosphorelay pathway in which the sensor kinase, Sln1p, working as a dimer, autophosphorylates a conserved histidine in response to changes in the osmotic environment (Posas et al., 1996). The phosphoryl group is transferred in trans to a conserved aspartate in the Sln1p receiver domain, then to a histidine in the phosphorelay molecule, Ypd1p, and finally to an aspartate on the receiver domains of the response regulators, Ssk1p and Skn7p (Posas et al., 1996; Li et al., 1998). The HOG1 MAP kinase pathway is repressed under normal growth conditions by Sln1p-dependent phosphorylation of the Ssk1p response regulator. Hypertonic conditions reduce Sln1p kinase activity, allowing Ssk1p to accumulate in the dephospho form to activate the HOG1 pathway (Posas et al., 1996). In contrast, hypotonic conditions stimulate Sln1p kinase and apparently cause increased phosphorylation of the Ssk1p and Skn7p response regulators (Tao et al., 1999). Mutants lacking Sln1p or the phosphorelay protein, Ypd1p, are inviable or very sick (depending on the strain background) because of the accumulation of Ssk1p in the dephospho form and toxic constitutive activation of the Hog1p MAP kinase (Ota and Varshavsky, 1993; Maeda et al., 1994). Mutations in SSK1 or HOG1 or overexpression of the Hog1p phosphatase, PTP2, rescue the lethality of sln1Δ or ypd1Δ mutants by reducing the level of phosphorylated Hog1p (Maeda et al., 1994).
The Sln1 histidine kinase of yeast is a hybrid kinase consisting of both kinase and receiver domains (Ota and Varshavsky, 1993). A pair of transmembrane domains (TM1 and TM2) at the N-terminus are important for association with the plasma membrane (Ostrander and Gorman, 1999; C. L. Malone and R. J. Deschenes, unpublished data). Downstream of TM2 is the so-called ‘linker’ region, the function of which is not well characterized. Immediately after the linker is the kinase domain, which can be subdivided into several homology blocks common to all histidine kinases (Ota and Varshavsky, 1993). The H box contains the phospho-accepting histidine. The N, G1, G2 and F boxes located downstream of the H box are involved in ATP binding and autophosphorylation (Dutta and Inouye, 2000). Autophosphorylation of sensor kinases typically occurs through a bimolecular transphosphorylation reaction (Yang and Inouye, 1991; Uhl and Miller, 1994; Surette et al., 1996). In some cases, such as EnvZ and CheA, dimerization domains have been identified (Park et al., 1998; Bilwes et al., 1999; Tomomori et al., 1999). Unlike many eukaryotic receptors that are induced to dimerize by ligand binding (Heldin, 1995; Hunter, 2000), sensor kinases pre-exist in the dimeric form, and the signalling mechanism for this type of receptor is proposed to involve a conformational change within the dimer (Milligan and Koshland, 1988; Kim, 1994; Cochran and Kim, 1996). Genetic analysis is consistent with Sln1p functioning as a dimer, as mutations of the phosphorylated histidine (H576Q) or the receiver domain aspartate (D1144N) are individually lethal, but the heterozygote, sln1-H576Q/sln1-D1144N, is viable (Maeda et al., 1994), implying that Sln1p kinase autophosphorylation occurs in trans.
Recent determination of the structure of the H box-containing portion of the histidine kinase domain of the Escherichia coli osmosensor, EnvZ, revealed a dimer consisting of two antiparallel helices from each of two EnvZ monomers (Tomomori et al., 1999). This four-helix bundle mediates both inter- and intramolecular interactions. Alignment of the corresponding region of Sln1p with this region of EnvZ shows sequence conservation not only through the H box, but also throughout helix 1 and, to a lesser extent, Helix 2, suggesting that the H box region in Sln1p dimers may assume a four-helix bundle structure similar to EnvZ (Fig. 1). Although there is no direct structural information for sequences upstream of Helix 1, secondary structure analysis of this region of EnvZ and histidine kinases with related topology suggest the presence of two amphipathic helices separated by a loop region (Park and Inouye, 1997; Butler and Falke, 1998). In EnvZ, this so-called ‘HAMP domain’ extends from the second transmembrane domain to within five residues of Helix 1 (amino acids 180–229). In Sln1p, the HAMP domain is separated from helix 1 by 25 amino acids, thus accounting for the gap in the alignment shown in Fig. 1.
A mutation just upstream of the Helix 1 homology in Sln1p (T550I) causes a kinase active (constitutive) phenotype (Fassler et al., 1997). We refer to this and similar mutations in SLN1 as sln1* mutations to emphasize their activated phenotype. Mutations in the linker region of other histidine kinases likewise have activating phenotypes (Kalman and Gunsalus, 1990; Collins et al., 1992; Miller et al., 1992; Tawa and Stewart, 1994). The isolation of linker mutants with altered activities has led to the suggestion that the linker plays a role in signal transduction or kinase regulation and is not simply a means of connecting the transmembrane domains to the kinase domain. Using the learncoil algorithm (Singh et al., 1998) designed to detect coiled-coils motifs in histidine kinases, a putative coiled-coil was identified just upstream and overlapping with Helix 1 and just downstream and overlapping with the second helix of the HAMP motif (Fig. 1, shaded box) in both Sln1p and EnvZ. The T550I sln1 activating mutation maps within the learncoil predicted coiled-coil region (Fassler et al., 1997). The analysis of sln1* and other kinase active mutants is expected to provide insight into the mechanism by which the Sln1p osmosensor senses and responds to changes in the osmotic environment. The goal of the current study was to assess the importance of the Helix1/2 similarity and predicted upstream coiled-coil region in Sln1p kinase activity and signal transduction. Our results confirm the presence of a coiled-coil in the Sln1p linker and demonstrate its requirement for Sln1p dimerization and kinase regulation. Our data lead us to reject a model for osmotic control of kinase activity that involves changes in coiled-coil interaction strength and to suggest an alternative model in which changes in coiled-coil rotation modulate accessibility of the phospho-accepting histidine.
The T550I mutation in SLN1 causes an activated phenotype that shifts the equilibrium between phospho Sln1p and dephospho Sln1p under ordinary growth conditions, so that the phosphorylated form of Sln1p is favoured (Fassler et al., 1997). To investigate the importance of the isoleucine substitution in the regulation of the Sln1p kinase, additional mutations were generated at this position, as described in detail in Experimental procedures. In addition to the isoleucine mutation that was reisolated in this screen, substitution of T550 with leucine and phenylalanine also activated the reporter several-fold, whereas substitution of T550 with aspartate did not increase reporter activity compared with wild type (Fig. 2A). In a separate screen (W. Tao and J. S. Fassler, unpublished data), the nearby E543V mutation was also identified as a sln1* mutation.
Although the linker domain has not been characterized structurally in any histidine kinase, the existence, position and register of a coiled-coil region is predicted to be a general feature of this domain of histidine kinases (Berger and Singh, 1997; Singh et al., 1998). The learncoil algorithm for predicting histidine kinase-type coiled-coils (Singh et al., 1998) identifies a possible coiled-coil in the linker of Sln1p overlapping helix 1 and terminating just before a conserved proline within the H block consensus (Fig. 1). In the predicted coiled-coil, the phospho-accepting histidine occupies position f of the coil heptad, where positions a and d represent the hydrophobic core of the coil that mediates interaction (Fig. 2B) (Singh et al., 1998). T550 and E543, two positions in which mutations are known to cause a sln1* phenotype, are predicted to occupy the a position in the hydrophobic core of the coiled-coil (Fig. 2B). This suggests the possibility that the effect of activating mutations such as T550 to I, L and F and E543 to V might be to increase the hydrophobicity of the coiled-coil interface, causing changes in the coil that mimic the response to osmotic stress.
Standard coil detection algorithms such as newcoils (Lupas et al., 1991) and paircoil (Berger, 1995) that are trained against dimeric, parallel coils from α-helical proteins such as myosin do not efficiently recognize the unusual subclass of coiled-coils present in histidine kinases (Singh et al., 1998). A high frequency of charged residues in the core a and d positions is apparent in the predicted Sln1p coil and in the equivalent region of EnvZ, portions of which are represented in Fig. 2B. As the presence of charged residues at core positions of the coil is predicted to lead to destabilization of coil-mediated interactions (Tripet et al., 2000), it was important to assess the functional importance of the putative coiled-coil region in sln1p. Two deletions were created in an otherwise intact sln1p gene. Deletion C1 (Δ527–540) and deletion C2 (Δ544–568) are both upstream of the H box (H576), and neither affects Sln1 protein levels detectably (data not shown). The deletion constructs were introduced into a sln1Δ strain (RJY1428) whose viability was maintained with a URA3-based plasmid overexpressing the PTP2 gene (Maeda et al., 1994). The ΔC1 construct was found to support viability; however, ΔC2 was inviable (Fig. 3A). As the viability of sln1 mutants depends on the ability to phosphorylate pathway intermediates, these results imply that the region covered by the C2 deletion is critical for kinase activity. To test this assumption, kinase activity was compared in constructs containing or lacking the predicted coiled-coil domain. The recombinant Sln1p kinase fragment used in this and other (below) in vitro assays included sequences between amino acids 537 and 947. This fragment contains neither the transmembrane domains nor the receiver domain. The 537–947 kinase fragment is referred to as Sln1K. The Sln1K and the Sln1KΔCC (570–947) lacking the coiled-coil domain were purified from bacteria as glutathione S-transferase (GST) fusions; however, the GST moiety was cleaved using thrombin before kinase activity assays. Reactions containing equivalent amounts of each kinase were incubated with [γ-P32]-ATP and sampled over time. Figure 3B shows that only Sln1K, the form of Sln1p kinase containing the predicted coiled-coil, has detectable kinase activity.
One possible role for the predicted coiled-coil region of the Sln1p linker is to mediate Sln1p dimerization. We used a co-immunoprecipitation assay to determine whether cytoplasmic derivatives of sln1p do in fact associate. As we had shown previously that the sln1p receiver domain does not interact (W. Tao and J. S. Fassler, unpublished data), the SLN1 constructs used in this assay included only the sequences (amino acids 356–1077) between the second transmembrane domain and the receiver domain. One construct was tagged with the myc epitope and was under the control of the GAL1 upstream activating sequence (UAS) (Fig. 4, lanes 1 and 2); the second construct was tagged with the HA epitope and was expressed under the control of the constitutive ADH1 UAS (Fig. 4, lane 3). The two constructs were co-expressed in the SLN1 wild-type strain, FY251. The presence of Sln1-HA in α-myc immunoprecipitates (Fig. 4, lanes 4 and 5) demonstrates an interaction between Sln1p monomers mediated by the kinase and/or linker domains.
Having established that the cytoplasmic Sln1p molecules do interact, two-hybrid assays were used to localize the interaction and evaluate the possibility that the coiled-coil region in the linker mediates the interaction (Fig. 5A). A comparison of the two-hybrid interaction was made between DNA-binding domain (DBD) and activation domain (AD) fusions to SLN1 (amino acids 356–1077) containing the coiled-coil region and comparable constructs lacking both C1 and C2 sequences in the coiled-coil region (amino acids 356–1077; Δ527–568). Expression of all fusions was roughly equivalent (Fig. 5B), but only the coil-containing constructs could be shown to interact. The absence of an interaction between constructs lacking the predicted coiled-coil region (Fig. 5A) indicates that the coiled-coil is needed for Sln1p interaction.
If the function of the C2 region was simply to form a coiled-coil, unrelated sequences known to mediate a coiled-coil interaction might restore viability to the sln1ΔC2 mutant. The coiled-coil region of the mammalian transcription factor, C/EBP, was ligated into the C2 deletion junction in an otherwise full-length and functional SLN1 gene, such that the register of the coil formerly present in the C2 sequences was preserved (Bustos and Schleif, 1993). Each construct (see Experimental procedures) contains six copies of the myc epitope at the 3′ end of the SLN1 open reading frame (ORF) to allow inspection of protein levels. To determine whether the sln1-C/EBP chimera is capable of supporting viability, the construct was introduced into the sln1Δ strain (RJY1428) carrying the URA3 PTP2 overexpression plasmid (Maeda et al., 1994), and growth was assayed on 5-FOA media as before. Figure 6A shows that the C/EBP sequences are able to substitute functionally for the sequences removed in the sln1ΔC2 mutant, thus suggesting that the importance of the C2 region of the Sln1 linker domain lies in its ability to form a coiled-coil structure. (Results with sln1-C/EBP + X constructs in Figs 6B and C are described below.)
Whether the C2 region is a simple zipper or a more sophisticated regulatory region was tested by assaying other SLN1 phenotypes. As the C/EBP zipper has a more hydrophobic interface than the native Sln1p coil, we anticipated that the substitution mutant might exhibit hyperactive phenotypes like the sln1* mutants. sln1* mutants are osmosensitive in the absence of Sho1p, an osmosensor that functions independently of Sln1p (Fassler et al., 1997). Osmosensitivity is taken to reflect impaired ability of the mutant Sln1p kinase to respond to salt because of either increased kinase or phosphorelay activity or the inability to turn off the kinase. The accumulation of phospho-Ssk1p prevents cells from activating the Hog1p MAP kinase pathway and inducing the genes needed for osmotic adaptation. A full-length sln1-C/EBP construct was expressed in the Δsln1 Δsho1 double mutant (RJY1525), and cultures were tested for osmosensitivity on media containing 0.9 M NaCl. Like the sln1* mutants, strains expressing sln1-C/EBP grew less well than wild type on high-salt medium (Fig. 7).
To determine whether the basis for the sln1* and sln1-C/EBP osmosensitive phenotype is increased kinase activity, we performed autokinase assays in vitro. Figure 8A shows the rate and extent of autophosphorylation by Sln1K*, which contains the T550I activating mutation in the coiled-coil region, and Fig. 8B shows autophosphorylation by Sln1-C/EBP kinase, in which the coiled-coil region has been replaced by the C/EBP zipper. Both mutant kinases are roughly comparable with Sln1K. In the representative experiments shown here, Sln1K* phosphorylation peaked at 3264 c.p.m. incorporated versus 3196 for wild-type kinase, and Sln1-C/EBP kinase phosphorylation peaked at twice the level of wild type (1375 versus 650). Sln1-C/EBP kinase activity fluctuated between one- and twofold of wild-type kinase activity depending on the amount of glycerol in the reaction. The lack of change in Sln1K* and the modest change in Sln1-C/EBP kinase activities are unlikely to explain fully the osmosensitive phenotypes of the sln1* and sln1-C/EBP mutants. Alternative explanations for the phenotype of these mutants including elevated phosphorelay activity and the inability to turn off kinase activity in response to osmotic stress will be considered in the Discussion.
A possible mechanism to account for the hyperactive phenotypes of sln1* mutants and the sln1-C/EBP chimera is an increase in the strength of the coiled-coil-mediated interaction. If, as suggested by the unusual features of histidine kinase-type coiled-coils, the coiled-coil interaction between Sln1p monomers is normally weak in order to allow kinase regulation by osmotic stress, an increase in interaction strength might interfere with the osmotic response. This was tested using the two-hybrid assay. Figure 9 is a representative experiment showing that there is no detectable effect of the sln1* mutation or C/EBP substitution on interaction strength. If the interaction between kinase monomers is not increased, what then accounts for the changes in pathway activity?
We explored the possibility that the sln1* and C/EBP alterations might cause other types of changes in the coiled-coil interaction. For example, the coils might take on a different orientation with respect to one another, thereby changing the solvent accessibility of the phospho-accepting histidines (Fig. 2B). To evaluate the sensitivity of the Sln1p kinase to the precise position of each residue within the heptad, we used the Sln1-C/EBP chimera, in which the coiled-coil register is unambiguous, to construct kinase mutants in which one, two, three or four amino acids were inserted at the interface between C/EBP and the downstream Sln1 sequence to create Sln1-C/EBP+1 (T), Sln1-C/EBP+2 (KT), Sln1-C/EBP+3 (AKT) and Sln1-C/EBP+4 (EAKT). These additions are presumed to have the effect of rotating the register of the predicted coil. Each construct was introduced into a sln1Δ strain (RJY1428) and assayed for viability. Strains expressing sln1-C/EBP and sln1-C/EBP+3 were viable, whereas sln1-C/EBP+1, +2 and +4 were inviable (Fig. 6B). The inviability in the sln1-C/EBP+1, +2 and +4 mutants is not a consequence of protein instability, as α-myc Western blots show that Sln1 protein levels in the mutants are comparable with wild type (Fig. 6C).
The expected correlation between viability and kinase levels was also investigated. GST fusion proteins corresponding to the C/EBP+1 and C/EBP +3 kinase domains were purified and assayed in vitro for autokinase activity (Fig. 8B). Like Sln1KΔCC, which lacks the coiled-coil (Fig. 3), the Sln1-C/EBP+1 kinase exhibited very low in vitro autophosphorylation activity (Fig. 8B), thus accounting for its failure to sustain viability. Interestingly, the autokinase activity of the Sln1-C/EBP+3 kinase was higher than that of Sln1-C/EBP+1 kinase but substantially lower than that of Sln1-C/EBP kinase (Fig. 8B). This low level of autophosphorylation activity is apparently sufficient to sustain viability.
Cellular responsiveness to signals from the environment depends on the conversion of extracellular cues into intracellular events. The mechanism for signal transduction across the membrane to the cytoplasm is proposed to involve conformational changes; however, the precise nature of the changes and how such changes result in altered function have been difficult to decipher. Various models have been proposed for ligand-induced movement in transmembrane receptors. For bacterial receptors, there is abundant experimental support for a subtle piston-like movement of a signalling helix towards the cytoplasm (Chervitz and Falke, 1996; Ottemann et al., 1999); reviewed by Falke and Hazelbauer, 2001). Rotational movement may also play a role (Cochran and Kim, 1996). However, transducing the signal across the membrane is just the first step. Subsequently, the conformational change must be propagated through the linker to the catalytic domain. The structural basis for this step is not known.
In many histidine kinases and in the structurally related chemoreceptors, the linker is predicted to consist of helical elements named ‘HAMP motifs’ (Aravind and Ponting, 1999; Williams and Stewart, 1999), followed by the structurally characterized four-helix bundle (Kim et al., 1999; Tomomori et al., 1999), which includes the H box in the histidine kinases. The importance of the linker region in regulating signal transduction is well documented in the prokaryotic literature (reviewed by Singh et al., 1998; Williams and Stewart, 1999). In one model for linker domain function, the HAMP motifs are proposed to mediate alternative interactions with the membrane and with one another to generate changes in conformation that transduce the signal to the kinase domain (Williams and Stewart, 1999). The present work concerns a region of the linker downstream of the HAMP motifs and upstream and overlapping helix 1 of the four-helix bundle, and proposes another potential mechanism for linker-mediated regulation of signal transduction. These models are not mutually exclusive, and both may be relevant to the control of kinase activity in response to environmental signals.
Our attention was drawn to the linker domain of the Sln1p histidine kinase by a set of activating mutations that cluster there. Based on computational predictions of secondary structure, the linker region of histidine kinases had previously been postulated to form a coiled-coil. However, the unusual features of the computationally predicted coiled-coils in the linker domains of histidine kinases left open the possibility that the resemblance to coiled-coils might be fortuitous and that other features unrelated to this aspect of the predicted secondary structure might account for the functional importance of this region. Several experimental lines of evidence support the computational hypothesis that the coiled-coil nature of these sequences is an important feature of this region of histidine kinases. First, mutations could be isolated that increase both the activity of the pathway and the hydrophobicity of the predicted hydrophobic core of the coil. Mutations such as T550D that did not increase hydrophobicity of the coil also did not affect pathway activity. Secondly, we demonstrate that the native SLN1 sequences encoding the putative coil could be replaced with a heterologous leucine zipper. Finally, we demonstrate (discussed below) that these sequences are required for Sln1p interaction.
Recent studies (Ostrander and Gorman, 1999) have shown that constructs lacking the extracellular domain (ECD) of Sln1p cause constitutive phosphorylation of Hog1p, implying that the domain is required for normal kinase activity. As normal activity could be restored by the addition of the C/EBP leucine zipper cassette, and mutations known to prevent dimerization of the leucine zipper reduced the activity of the Sln1p kinase, it was postulated that dimerization is mediated by the extracellular domain (Ostrander and Gorman, 1999). However, the viability of the ΔECD construct (in spite of the elevated Hog1p phosphorylation) suggests that kinase activity and dimerization are not fully abolished in the mutant. Our studies of SLN1 coiled-coil mutants indicate that the linker domain also makes an important contribution to dimerization and kinase activity.
Our analysis of SLN1 linker domain deletion mutants revealed that the sequences deleted in the ΔC2 mutant are important for kinase activity. The inviability of sln1 mutants stems from the accumulation of dephospho-Ssk1 and the resulting activation of the Hog1p MAP kinase pathway. As the 20-amino-acid deletion in the ΔC2 mutant causes inviability, the construct is presumed to reduce kinase and/or phosphotransfer activity below the threshold required to repress the MAP kinase cascade effectively. This experiment also suggests that the sequences encoding the predicted four-helix bundle (567–627), which are retained in the ΔC2 construct (Δ544–568), are not sufficient for Sln1p dimerization. Interestingly, the ΔC1 deletion was viable. This may imply that the portion of the coiled-coil distal to the H box is less important than the proximal portion that appears to be sufficient to carry out the vital function of the region. As the C1 sequences overlap the second helix of the HAMP motif, these results imply that the HAMP helices are distinct from the C2 coiled-coil.
Interestingly, sln1-C/EBP, although viable, is not wild type with respect to pathway activity. The osmosensitive phenotype of this mutant suggests that phospho-Ssk1p levels in the mutant may be increased relative to wild type. Thus, the C/EBP leucine zipper is sufficient for some but not all aspects of SLN1 function. This suggests that the unusual features of the coiled-coil in SLN1 and other histidine kinases may be a specialized adaptation for regulating kinase activity.
Although a substantial body of work in bacterial systems indicates that dimerization is necessary but not sufficient for histidine kinase activation (Yang and Inouye, 1991; Uhl and Miller, 1994; Surette et al., 1996), we nonetheless tested the possibility that changes in osmotic environment might modulate interaction strength between cytoplasmic Sln1p monomers. This idea is attractive because the hydrophobic core of the coiled-coil region of native Sln1p contains a total of four hydrophobic residues, whereas the sln1* mutations T550I and E543V each create one additional hydrophobic interaction and might therefore be expected to mediate a somewhat stronger interaction. However, this was not the case. There were no detectable differences in kinase–kinase interactions despite alterations in the hydrophobicity of the coiled-coil in SLN1+, sln1* or sln1-C/EBP constructs. Thus, as proposed for prokaryotic histidine kinases, conformational changes may play an important role in regulating Sln1p activation.
Despite insights from the recently determined individual structures of two EnvZ kinase subdomains (Tanaka et al., 1998; Tomomori et al., 1999) (the substrate phosphotransfer domain, A, whose dimerization yields a four-helix bundle including the phosphorylatable histidine, and the catalytic domain, B), the spatial relationship of these domains in a histidine kinase active centre remains unknown. Efforts to reconstruct a functional monomeric kinase by fusing two A and one B domains in tandem have led to an elegant model of kinase domain organization, in which the ATP binding site of domain B of one subunit is situated adjacent to the phosphorylatable histidine residue on Helix 1 of domain A of the other subunit in a homodimer (Qin et al., 2000, Fig. 7). Using this model as a starting point, we propose that regulation of Sln1p kinase activity might involve changes in the relative orientation of the phosphorylatable histidine with respect to the catalytic domain (Fig. 10). Although a prototypical coiled-coil may not accommodate flexibility in the register of the coiled-coil, the special nature of histidine kinase coiled-coils, including the weak ‘hydrophobic’ character of the a–d core that holds the conserved histidine in the exposed f position in native Sln1p, might allow sufficient range of motion to allow changes in the absolute location of the f position within the coil. Alternatively, the histidine kinase coiled-coil could be designed to allow shifts in coil register, such that a different set of residues occupies the a and d positions of the heptad repeat under low versus high osmotic conditions. In this model, the T550I and E543V mutants would favour the coil register that leads to pathway activation.
We tested the idea that the relationship between the coiled-coil domain and the phosphorylatable histidine is important by introducing small insertions into the coil region of the Sln1-C/EBP chimera. We chose the C/EBP chimera for these experiments because the robust hydrophobic interface in this coil creates a situation in which the coiled-coil register is unambiguous. The effect of modification to the C/EBP coiled-coil could therefore be predicted with confidence. The position of the insertion was designed to maintain the C/EBP a–d interface. However, the phosphorylatable histidine should rotate with respect to the coil as a result of the insertion. The results of this analysis indicate that kinase activity is sensitive to insertion in the heptad repeat of the coil. Of the constructs we tested (+1, +2, +3 and +4), only +3 was viable. The 3-amino-acid insertion is presumed to move the histidine from the exposed f position to the adjacent b position, which is also reasonably exposed (Fig. 2B). Interestingly a +4 insertion is inviable, although it is expected to move the histidine to c, a position that appears superficially to resemble b in the extent to which it is solvent exposed. The difference between the +3 and +4 kinases may reflect the relative accessibility of the b and c positions to the catalytic domain.
Although the sequences upstream of Helix 1 (amino acids 223–235) appear to be disordered in the EnvZ structure (Tomomori et al., 1999), this may result from the loss of secondary structure elements that reside further upstream. Our data suggest that sequences upstream of Helix 1 are also helical, and we have tentatively depicted them as continuous with the upstream half of Helix 1 (Fig. 10). The predicted Helix 1 of Sln1p and EnvZ (and approximately two-thirds of all known histidine kinases) contain a proline residue five residues distal to the phosphorylatable histidine (Grebe and Stock, 1999; Tomomori et al., 1999). The placement of this proline inside Helix 1 coincides with a disordered region in the EnvZ helices and may create a kink in two of the four helices in the four-helix bundle, as shown in Fig. 10. In this case, changes in the orientation of the linker helix could allow rotation of the phosphorylatable histidine without disturbing the remainder of the four-helix bundle. We propose that increasing or decreasing pressure on the membrane as the result of changing osmotic conditions causes a twist that propagates through the linker and influences the alignment of the phospho-accepting histidine in the H box located in Helix 1 and the catalytic part of the kinase domain, and that the alignment is more favourable under high kinase activity conditions such as hypo-osmotic stress or the sln1* mutant than it is under low kinase activity conditions such as hyperosmotic stress.
Although kinase levels are reduced relative to wild type in the ΔCC mutant (Fig. 3), the sln1 and sln1-C/EBP mutants that improve the hydrophobic interface do not appear to elevate kinase levels, as measured by our in vitro kinase assay (Fig. 8). This suggests that the role of the coiled-coil domain in kinase activity may not be to increase or decrease autophosphorylation, but rather to regulate some other aspect of Sln1p activity such as phosphotransfer or dephosphorylation. Most histidine kinases are bifunctional, possessing phosphatase activity (typically directed against the aspartyl phosphate of the response regulator) in addition to kinase activity (Egger et al., 1997); the ratio of the two activities dictates flux through the pathway. Thus, pathway activity can be modulated by regulation of the kinase or the phosphatase. Interestingly, it has been proposed that the spatial arrangement of domain A relative to domain B of EnvZ may be altered under different osmotic conditions to favour one activity over the other (Zhu et al., 2000). Although no Sln1p-associated phosphatase activity has yet been identified using the Sln1p-associated receiver domain as substrate (A. D. Ault et al., in preparation), studies involving the Ssk1p and Skn7p response regulators have not yet been conducted. A second possible mechanism accounting for the activated phenotypes is a change in phosphorelay efficiency. Recent biochemical analysis of another sln1* mutant, sln1 (p114) mutant (A. D. Ault et al., in preparation) has revealed that its activated phenotype is at least partially attributable to changes in the efficiency of the Sln1p to Ypd1p phosphorelay step.
Many details of linker region regulation of kinase activity remain to be defined. Structural analysis of several histidine kinase domains is needed to test the structural predictions of the coiled-coil hypothesis fully. To date, no structures exist that include the region N-terminal to the kinase domain. In the absence of structural information, we have used functional analysis of coiled-coil mutants and chimeras to demonstrate the involvement of a coiled-coil domain in Sln1p signalling. Further analysis of the in vivo and in vitro characteristics of SLN1 mutants and chimeras is expected to provide additional mechanistic detail.
Media were prepared as described by Sherman et al. (1986) and included synthetic complete medium (SC) lacking one or more specific amino acids (e.g. SC–leucine), rich medium (YPD) and synthetic dextrose (SD). Plates for the detection of β-galactosidase activity contained 50 mg ml−1 Xgal unless noted otherwise and were prepared as described by Larson et al. (1983). Yeast strains were grown at 30°C unless noted otherwise. Yeast transformation was performed by a modified LiOAc method (Ito et al., 1983; Gietz et al., 1995)
All yeast strains (Table 1) used in this work are from our laboratory collections or were constructed for this study. Strains are isogenic or congenic with the wild-type strain S288C as noted (Winston et al., 1995).
All plasmids used in this work are from our collections or were constructed for this study (Table 2).
The two-hybrid vector plasmids pAS1-CYH2 and pACTII contain the GAL4 DNA-binding domain (amino acids 1–147) and GAL4 activation domain (amino acids 768–881) driven by the ADH1 promoter respectively (Durfee et al., 1993; Harper et al., 1993). Vector pAS1-CYH2 contains the yeast TRP1 marker, and pACTII contains the LEU2 marker. Both plasmids have one copy of the haemagglutinin (HA) epitope tag inserted behind the GAL4 sequence and preceding the poly cloning sites.
The cytoplasmic SLN1 kinase construct was amplified from SLN1 plasmid pGY111 (Yu et al., 1995) using synthetic primers SLN1060F (5′-CTA GCA CAC TGG ATC CTG CAA CCA ATT GTA CG-3′) and SLN3231R (5′-CTC AGT CTC GAG TTA TCA GCG GCC GCC TGT GGG GAT GTT TCT ACT TG-3′) and digested with BamHI and XhoI. The 2.2 kb BamHI–XhoI SLN1 kinase fragment was cloned in BamHI–SalI-digested pAS-CYH2 vector and BamHI–XhoI-digested pACTII vector respectively. The junctions of the insertion and the entire SLN1 sequence in both plasmids were verified by DNA sequence analysis.
pCLM994 was constructed by cloning the SacII to EcoRV fragment from pHL597 into the SacII–EcoRV-digested pRS313 (Sikorski and Hieter, 1989). pHL597 is a full-length SLN1 clone with six copies of the myc epitope added in frame by insertion into a newly created NotI site at the end of the ORF. The sln1ΔC1 plasmid, pCLM1001, was constructed by polymerase chain reaction (PCR) amplification of SLN1 sequences from pHL597 using synthetic oligonucleotide primers SLN718F (5′-GTA TAT AAT AGT CAA GGC-3′) and SLN1578R (5′-TCA TCG CTT CTA GAA TTC TGT TAA ATC GGA AAG TTC ATC-3′), followed by digestion with StuI and XbaI and cloning into pCLM994 also digested by StuI and XbaI. Amino acids 527–540 are deleted in this construct. The sln1ΔC2 plasmid, pCLM1002, was constructed in two steps. First, the product of PCR amplification of SLN1 sequences from pHL597 using primers SLN1705F (5′-TCA TCG CTT CTA GAA GTC TTT ATT GCC AAT ATT TCG CAC G-3′) and SLN2080R (5′-GGT GAA CTT TAG TGC ATT GGA-3′) was digested with XbaI and ClaI and ligated into pCLM994 (XbaI, ClaI) to generate the intermediate plasmid pCLM996. pCLM996 was then digested with ClaI and ligated to the ClaI fragment from pCLM994. Amino acids 544–568 are deleted in this plasmid. The sln1-C/EBP plasmid, pCLM1021, was constructed by digesting pCLM1002 with XbaI, treating with calf intestinal phosphatase (CIP) and co-transforming with the PCR product generated using primers SLN1574F (5′-CAG AAA CCT TCA ATA CTA TGA CAG ACG CAT TAG ACC AAC ATT ATG CTC TTC GCA ACG TGG AGA CGC-3′) and SLN1705R (5′-ATA CCA TTT AAA GGC GTT CTC AAT TCG TGC GAA ATA TTG GCA ATA AAG ACG CTC AGC TGT TCC ACC-3′) with pGB008 (Landschulz et al., 1988; Bustos and Schleif, 1993) as template. The sln1-C/EBP+X plasmids, pCLM1033 (+1), pCLM1034 (+2), pCLM1035 (+3) and pCLM1036 (+4), were constructed by digesting pCLM1002 with XbaI, treating with CIP and co-transforming with the PCR amplification product using pGB008 (Landschulz et al., 1988; Bustos and Schleif, 1993) as template, forward primer SLN1574F and one of the following reverse primers: (+1) SLN1704R (5′-ATA CCA TTT AAA GGC GTT CTC AAT TCG TGC GAA ATA TTG GCA ATA AAG ACG GTG CTC AGC TGT TCC ACC-3′); (+2) SLN1701R (5′-ATA CCA TTT AAA GGC GTT CTC AAT TCG TGC GAA ATA TTG GCA ATA AAG ACG GTT TTG CTC AGC TGT TCC ACC-3′); (+3) SLN1698R (5′-ATA CCA TTT AAA GGC GTT CTC AAT TCG TGC GAA ATA TTG GCA ATA AAG ACG GTT TTT GCG CTC AGC TGT TCC ACC-3′); and (+4) SLN1695R (5′-ATA CCA TTT AAA GGC GTT CTC AAT TCG TGC GAA ATA TTG GCA ATA AAG ACG GTT TTT GCT TCG CTC AGC TGT TCC ACC-3′).
pHL581 and pADA847 are Sln1K537–947 and Sln1K570–947 in pGEX-KG respectively. SLN1 DNA corresponding to amino acids 537–947 or 570–947 was amplified using primers SLN1609F (5′-ATG ACA GAC GCA TGA ATT CAA CAT TAT GCT CTT CTA G-3′) or SLN1711F (5′-TTT CTA GAC GGA TCC ATT GCC AAT ATT TCG CAC G-3′) and SLN2827R (5′-CTT TCT ACT CTC GAG GAT TAA ATT CGT C-3′). PCR products were digested with EcoRI or XbaI and XhoI for ligation into pGEX-KG (Guan and Dixon, 1991).
pCLM1032, pCLM1037 and pCLM1038 are Sln1-C/EBP+0, +3 and +1, respectively, in pGEX-KG. They were constructed using the pCLM1021 and pCLM1033 (+1) and pCLM1035 (+3) plasmids as templates in PCR reactions with primer SLN1612F (5′-ACG CAT GGA ATT CAA CAT TAT GCT CTT CGC-3′) and SLN2080R (above). PCR products were digested with EcoRI and ClaI (pCLM1021) or BplI and ClaI (pCLM1033 and pCLM1035) and the fragments cloned into pGEX-SlnK537–947.
To introduce amino acid substitutions at amino acid 550 of Sln1, the degenerate primer SLN1704R [5′-GGT TTT TGC TTC ATT TGC GGC CTC TGC CTC AAT CTT GGC TGC TTC CAG CTG TTT (T,G,A)(T,A,C)(C,G) CCT CGC CCT AAC TCT TTC-3′], corresponding to the reverse complement of Sln1 amino acids 544–568, was synthesized. This primer was designed to introduce 18 types of codons, encoding 14 different amino acids. SLN1704R and SLN1551F (5′-GCG CAA CTG GAT CCC TGA TGA ACT TTC CG-3′) were used to amplify degenerate PCR products from an SLN1 plasmid template under normal PCR conditions. These 154 bp degenerate PCR products containing substitutions at amino acid 550 were co-transformed with XbaI-linearized pWT1165 (SLN1, HIS3, CEN) into JF2057 (sln1Δ, ssk1Δ) carrying the P-lacZ reporter (Yu and Fassler, 1993). The XbaI site in SLN1 is situated close to T550, and recombination events between the corresponding sequences in the PCR product and the plasmid repair the gap in the plasmid. Transformants were selected on plates lacking histidine and screened for an increase in the normal pale blue colour of this strain on plates containing 50 μgml−1 Xgal. Six out of 110 transformants screened exhibited a blue colour on Xgal media and at least a twofold increase in β-galactosidase activity indicating reporter activation. SLN1 plasmids in these transformants were isolated and subjected to sequence analysis. Four were T550F substitutions, one was T550L, and one was T550I. The T550D mutant was isolated in this screen by analysis of transformants whose colour appeared to be whiter than the pale blue of the starting strain. Of four white transformants analysed, only one (T550D) had a mutation.
Two-hybrid constructs carrying the GAL4 DNA-binding domain and GAL4 activation domain fusions were co-transformed into the two-hybrid test strain pJ69-4A (James et al., 1996). Transformants were grown in selective media to log phase. Cells were washed twice with synthetic media before resuspension at a concentration of 2 × 108 cells ml−1. Tenfold dilutions (2 × 107, 2 × 106 and 2 × 105) were spotted (2 μl) on test and control plates incubated at 30°C. Complete plates were SD + uracil, methionine, adenine and histidine; 3 amino-triazole (3-AT) plates were SD + uracil, methionine, adenine plus 1.0 mM 3-AT.
Yeast strain FY251 carrying myc-tagged and HA-tagged fusion constructs were grown to OD600 = 0.5 in selective media with 2% raffinose. Galactose (4%) was added to half the culture to induce protein expression from the PGAL-driven myc-tagged construct. Cells were collected by centrifugation, resuspended in 200 μl of lysis buffer [25 mM Tris, pH 8.0, 10 mM MgCl2, 1 mM EDTA, 0.1% Triton X-100, 1 mM dithiothreitol (DTT), phenylmethylsulphonyl fluoride (PMSF) and protease inhibitors] and broken by glass bead lysis. Cell debris was removed by microcentrifugation. Anti-myc antibody (5 μl) was added to 100 μl of extract and incubated on a rocker at 4°C for 1–2 h. An aliquot of 25 μl (50% slurry) of washed (in lysis buffer) anti-IgG agarose (Sigma A-6531; goat anti-mouse) was added to the mixture and incubated at 4°C for 1–2 h. Agarose beads were washed three times with 1 ml of PBS buffer, collected by microcentrifugation and resuspended in 25 μl of 2.5× SDS loading buffer. Samples were boiled for 2 min, and proteins were separated by polyacrylamide (10%) gel electrophoresis. Proteins were transferred to nitrocellulose membrane, probed with anti-HA antibody (1:1000) and visualized using anti-mouse horseradish peroxidase (HRP) and ECL chemiluminescence (Amersham).
GST fusion proteins were expressed in DH5α cells. One litre cultures were grown to an OD600 of 0.6–1.0 in LB media and induced with 0.2 mM IPTG at 16°C for 20 h. After induction, cells were harvested by centrifugation (5000 g for 20 min), resuspended in lysis buffer [10% glycerol, 50 mM Tris, pH 7.6, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, 0.1% β-mercaptoethanol, 1× PIC (0.05 mg ml−1 leupeptin, 0.001 mg ml−1 pepstatin, 0.001 mg ml−1 chymostatin, 1× aprotinin (Sigma)] and stored frozen at −80°C.
Frozen bacterial cells were thawed and lysed by French press in the presence of 1 mM PMSF (added immediately before lysis). The lysate was cleared by centrifugation. GSH-agarose beads were added as a 50% slurry to the supernatant. (3 ml of beads per 5 g of cell paste or 1 l of culture), and the mixture was incubated at 4°C for 1 h. Beads were pelleted and washed four times in lysis buffer, followed by four washes with storage buffer (50 mM Tris, pH 7.6, 50 mM KCl, 5 mM MgCl2, 0.1% β-mercaptoethanol, 50% glycerol) and stored at −20°C.
To cleave Sln1 kinase from GSH-agarose beads (Novagen), a 50% slurry (1ml) of bead-bound protein was washed four times in thrombin cleavage buffer (50 mM Tris, 150 mM NaCl, 10 mM CaCl2) before the addition of 5 units of biotinylated thrombin. Cleavage reactions were incubated at 25°C for 4–6 h with streptavidin beads (25 μl) added to the reaction during the final 30 min. Proteins were quantified by Bio-Rad protein assay and Coomassie staining of SDS–polyacrylamide gels.
Sln1 kinase domains (380 nM) were incubated with 10 μM [γ-32P]-ATP (400 Ci mM−1) for 15 min in kinase reaction buffer (50 mM Tris, pH 7.6, 50 mM KCl, 5 mM MgCl2, 0.1% β-mercaptoethanol, 50% glycerol) at room temperature. Beads were washed four times with reaction buffer to remove unincorporated label. Aliquots were quenched in 5× SDS loading buffer. Samples were resolved by SDS–PAGE and the gels analysed using a Packard Instant Imager.
We thank S. Moye-Rowley, R. Malone and J. Lu for critical reading of this manuscript. This work was supported by the National Institutes of Health (GM56719) and grant T32AG (A.D.A.) from the NIH/National Institute of Aging.