| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Francis D. Gibbons,
Matthew Copeland,
Frederick P. Roth, and
Kevin Struhl*
Department of Biological Chemistry and Molecular Pharmacology, Harvard Medical School, Boston, Massachusetts 02115
Received 25 April 2005/ Accepted 3 June 2005
 |
ABSTRACT |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
INTRODUCTION |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The molecular functions of activated Hog1 kinase in transcriptional stimulation are surprisingly complex. Activated Hog1 is recruited to the promoters of osmoinducible genes by specific transcription factors (1, 35). In the case of Sko1, which acts as a repressor of osmoinducible genes via the general corepressor complex Cyc8-Tup1 (34), the association with Hog1 leads to its phosphorylation and the conversion to an activator (33, 35). This repressor/activator switch involves the additional recruitment of the chromatin-modifying complexes SAGA and Swi/Snf, but it does not result in the dissociation of Cyc8-Tup1 (35). Furthermore, the nuclear localization and repressor functions of Sko1 are regulated by protein kinase A (31), which phosphorylates Sko1 at multiple sites in vitro (33).
The redundant operation of various transcription factors in the transcriptional osmostress program makes this adaptive response difficult to describe by genomic profiling experiments. Subsets of Hot1- and Msn2/4-dependent genes have been identified by expression-based microarrays (38), but these represent a small proportion of the genes induced by hyperosmotic shock. It is clear that understanding the yeast osmostress response will require knowledge of the physical interactions of the transcriptional regulators with their genomic targets during stress.
Here, we used genomewide location analysis to identify the in vivo target promoters of Sko1. Our results differ substantially from an initial genome-wide location analysis (22), but are in good accord with a report that appeared after the work described here was completed (11). We validate 20 Sko1 target sites by direct chromatin immunoprecipitation experiments and demonstrate that Sko1 is important to various extents for osmotic induction of genes encoding transcription factors involved in other stress responses and Ptp3, which encodes a phosphatase that regulates Hog1 kinase. Our results reveal important information about the biological function of Sko1 and about transcription factor networks that operate after osmotic stress.
|
MATERIALS AND METHODS |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Chromatin immunoprecipitation. Cells were grown in rich medium (YPD) to an optical density at 600 nm of 0.8. Chromatin immunoprecipitation was performed essentially as described previously (20), except that insoluble material was removed from the broken cells by 2 min centrifugation in a minicentrifuge. After sonication, soluble chromatin fragments were obtained by spinning for 30 min in a minicentrifuge. (HA)3-Sko1 was immunoprecipitated with antibodies against the HA epitope (12CA5 ascites). Input and immunoprecipitated samples were assayed by quantitative PCR in real time using the Applied Biosystems 7700 sequence detector. An internal fragment of the POL1 gene (nucleotides 2499 to 2717) was used as a negative control.
For targeted chromatin immunoprecipitation experiments, immunoprecipitations were performed on three independent chromatin preparations. Immunoprecipitation efficiencies were calculated in triplicate by dividing the amount of PCR product in the immunoprecipitated sample by the amount of PCR product in the input sample. All data (Fig. 1 and Table 1) are presented as fold immunoprecipitation over the POL1 coding sequence control.
|
|
1 (–318/–127, STR9987/9988), HIR2 (–274/–38, STR9983/9984), MLH3 (–886/–669, STR9965/9966), FOL1 (–338/–159, STR9961/9962), and open reading frame free region on chromosome V (STR5332/5333). Primer sequences are available upon request. Microarray hybridization. Microarray preparation, amplification of total and immunoprecipitated DNA, fluorescent labeling and microarray hybridization were performed as described previously (29). For chromatin immunoprecipitation analysis of Sko1, three independent chromatin samples were immunoprecipitated and the resulting DNA fragments amplified and labeled with Cy5 fluorescent dye (Amersham Biosciences). Three independent total chromatin samples were similarly amplified and labeled with Cy3 fluorescent dye. Microarrays were hybridized with a mixture of labeled fragments from immunoprecipitated and total chromatin samples. Slides were scanned on an Axon scanner, and data were analyzed with Axon GenePix 4.0 software. Raw data were then further analyzed in Microsoft Excel. Poor-quality or undetectable DNA spots were removed and the data were normalized for equal background-subtracted median fluorescence of Cy3 and Cy5 over the entire arrays. Spots were ranked in descending order by their median Cy5/Cy3 ratios (see the supplemental material).
Feature names were obtained from the Saccharomyces Genome Database (14). Statistical significance was computed for each spot using the Chipper software (9) for each spot to describe their likelihood of being bound by Sko1. Individual P values were then combined by Stouffer's method and are given in Fig. 1 and Table 1.
We tested some of the intergenic regions that were identified over different P value ranges by standard chromatin immunoprecipitation analysis. We regarded a DNA sequence to be bound by Sko1-HA when the respective fragment was at least 2.5-fold enriched over the unbound POL1 control and when there was no enrichment in the absence of the HA epitope. Some intergenic regions showed slight enrichment (about twofold) independently on the presence of the HA antibody and therefore represent unspecific antibody recognition or just "sticky" DNA fragments.
Sequence analysis. The AlignAce program (13, 41) was used to identify common motifs across multiple promoter regions. CRE sites were searched using the DNA pattern feature of the RSA-tools webpage (http://rsat.ulb.ac.be/rsat/dna-pattern.cgi).
Analysis of mRNA levels. Yeast strains were grown in YPD to an optical density at 600 nm of 0.8 and were treated or not with 0.4 M NaCl for 10 min. Total RNA was extracted from 50 ml of culture by acid phenol treatment (15) and DNase digested (RQ1 DNase, Promega). Total RNA was reverse transcribed using Superscript II reverse transcriptase (Invitrogen) following the manufacturer's instructions and RNase treated (Roche). As a control, each sample was additionally mock treated (without reverse transcriptase). Appropriately diluted samples were analyzed by quantitative PCR in real time using primers amplifying TBP1 (nucleotides 521 to 677), MSN2 (nucleotides 1671 to 1886), MSN4 (nucleotides 1505 to 1658), PTP3 (nucleotides 2513 to 2680), PTP2 (nucleotides 1997 to 2201), MOT3 (nucleotides 1127 to 1432), MGA1 (nucleotides 1143 to 1280), and ROX1 (nucleotides 903 to 1088).
|
RESULTS |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
We created a ranked list of putative Sko1-promoter interactions, which contains 33 intergenic regions with a P value of <10–4 (group I) and 34 intergenic regions with a P value of between 10–4 and 0.004 (group II) based on a minimum of two independent replicate experiments.
Direct analysis by standard chromatin immunoprecipitation (Fig. 1) confirmed Sko1 binding to 16 out of 18 group I targets (89%) and 4 out of 10 group II (40%) as defined by >2.5-fold enrichment of the respective locus comparing HA-Sko1-tagged and untagged control strains. The two group I regions that did not pass the test (TYE1 and GAC1) were weakly enriched by HA-Sko1, but at a level just below the 2.5-fold cutoff. HA-Sko1 binding was not observed to five regions characterized as being in group I or II based on just one experiment, indicating that results based on one experiment are not reliable. Three group II regions (ICY1, HOR7, and RSN1) were enriched independently of Sko1-HA (about twofold), probably due to cross-reactivity of the HA antibody with other cross-linked proteins in the chromatin preparations or to the general stickiness of these genomic regions during the immunoprecipitation procedure. The existence of these nonspecifically enriched regions emphasizes the importance of performing comparisons between tagged and untagged strains.
Based on the percentage of confirmed targets of those tested from groups I and II, we estimate that there are approximately 40 Sko1 target loci under the conditions tested in vivo. In addition, the experimentally determined false discovery rates of specifically or nonspecifically enriched regions for groups I and II are in good accord with false discovery rates predicted by the P values. As such, this analysis indicates that Chipper provides relatively accurate P values for identifying enriched genomic regions. Our results differ significantly from an initial genomewide location analysis (22), but are in good accord with a later report that appeared after the work described here was completed (11). The existence of this later report permits us to combine P values from experiments performed in two different laboratories, thereby providing a more reliable assessment of Sko1 target sites. With respect to the data generated here, this combined analysis casts doubt on a few group I regions (PRR2 and MNN9) and it provides strong evidence that some untested group II (FSH1, YAP6, RPS4A, and PET9) or even lower-ranked (CIN5 and ALD4) regions are bona fide Sko1 targets.
Sko1 targets include genes encoding stress-defense proteins, putative regulators, and transcription factors. Sko1 has been characterized as a repressor/activator protein with an important function in the transcriptional response to hyperosmotic stress. Among the 39 group I and II intergenic regions bound by Sko1 with the most significant P values, 20 (50%) were associated with a gene induced >3-fold by osmotic stress in at least one transcriptional profiling experiment (32, 38). This relationship between Sko1 binding in vivo and transcriptional induction in response to osmotic stress is highly significant (P = 10–7).
Table 1 summarizes the 20 intergenic regions associated with Sko1 that were confirmed by targeted chromatin immunoprecipitation experiments. Although there is a correlation between Sko1 binding and transcriptional response to osmotic shock (represented by FAA1/COT1, GAT2, FMP43/YAP1802, CWP1, ROX1/YPR063c, CIN1/PUT4, HXT5, MOT3, MSN2, ALD6, STL1, and SED1), we identified a number of promoters that are unknown to respond to osmolarity (represented by RHO3/RPI1, SRL1/YOR246c, MGA1, UTH1, SPO20/SOK2, SOR1, PTP3, and CPA1). These promoters may respond weakly to osmotic stress (i.e., less than threefold) or a Sko1-dependent effect might be masked by other factors that regulate the expression of these genes. Conversely, we and others (11) did not detect Sko1 binding to three genes, SFA1, GLR1, and YML131W, shown to have a modest Sko1-dependent effect on transcription (39). This observation suggests the possibility that Sko1 indirectly affects transcription of SFA1, GLR1, and YML131W, although the lack of Sko1 binding might be an artifact of the microarray experiments. We observed no detectable binding to the GRE2 promoter and only weak binding to the AHP1 promoter, both of which have been previously shown to be direct Sko1 targets in vivo (35). In general, the failure to detect Sko1 binding at a given region in these microarray experiments is less significant than the ability to detect Sko1 binding.
A number of the experimentally confirmed or group I Sko1 targets have known or likely functions in the direct relief from hyperosmotic stress. In particular, Sko1 target genes encode cell wall proteins (Sed1 and Cwp1), vacuolar or cytoplasmic transporters (Stl1, Hxt5, Put4, and Cot1), and confirmed or likely enzymatic activities (Faa1, Sor1, Ald6, and Yor246c). Most of them were previously identified by genomic profiling as being strongly induced upon salt stress. More interestingly, Sko1 targets also include known or putative regulators (Ptp3, Cin1, Rho3, Rpi1, Uth1, Prr2, and Srl1) or transcription factors (Msn2, Mot3, Mga1, Rox1, and Gat2). With the exception of Ptp3 and Msn2 (discussed later), none of these molecules has been functionally related to osmotic or salt stress adaptation, although it should be noted that subtle effects on the osmotic stress response might not have been detected by the phenotypic assays employed to date. It will be therefore of special interest to unravel their contribution to the osmotic stress response.
Sequence comparison of target promoters suggests that Sko1 binding in vivo requires additional factors beyond recognition of ATF/CREB motifs. Sko1 belongs to the family of ATF/CREB DNA binding proteins, and it associates with the T(G/T)ACGT(C/A)A consensus sequence via its bZIP domain. We therefore searched for ATF/CREB motifs, allowing one mismatch to the consensus, in the 20 confirmed Sko1 target regions, and 19 of the 20 intergenic regions contain at least one ATF/CREB motif within the upstream 1,000 nucleotides for a total of 47 sites, but only one of these motifs is identical to the consensus. The number of ATF/CREB motifs in the confirmed Sko1 target regions is expected by chance, because the yeast genome contains over 15,000 such motifs among the 6,000 intergenic regions. The average distance of these ATF/CREB motifs from the ATG initiation codon is 549 bp, a position expected by chance and considerably further upstream than the 200 to 300 bp that is typical for bona fide binding sites in Saccharomyces cerevisiae. Furthermore, an attempt to find common motifs by AlignAce resulted in several potential candidates of modest statistical significance, but the ATF/CREB motif was not among them.
Although many of the ATF/CREB motifs in the confirmed Sko1 target regions are of dubious functional significance, 15 of these motifs are conserved among the four Saccharomyces sensu stricto yeasts, S. cerevisiae, S. mikatae, S. bayanus, and S. paradoxus (Table 2). In addition, sequence comparison of these evolutionarily conserved motifs reveals additional sequence preferences in the ±4 position that immediately flanks the motif (Table 3). The preferences for A at the –4 position and T at the +4 position are remarkably similar to those of Gcn4 for the related AP-1 site (30). These observations suggest that at least some of the evolutionarily conserved ATF/CREB motifs may be relevant for Sko1 binding in vivo (4, 18). It should be noted, however, that some of the conserved ATF/CREB motifs are located in the coding region of the upstream gene, casting doubt on their functional significance.
|
|
In addition to Sko1, Saccharomyces cerevisiae has two other ATF/CREB factors, Aca1 and Aca2 (8). These three ATF/CREB proteins bind the consensus and near-consensus ATF/CREB sites with comparable affinity, and they modulate the expression of artificial ATF/CREB-driven promoters. However, the biological functions of these proteins are different, as judged by their mutant phenotypes (8). As a consequence, it is very likely that these three ATF/CREB proteins associate with distinct subsets of promoters in vivo, although this has yet to be determined experimentally. Distinct in vivo binding patterns of the ATF/CREB proteins might reflect differential preferences for sequence variations of the CRE motif, as observed for the glucose-regulated Cat8 and Sip4 transcription factors (42). Alternatively, the ATF/CREB proteins might interact with different factors, thereby resulting in distinct promoter selectivity.
Sko1 regulates the expression of Msn2, a transcriptional activator protein that mediates the general stress response. Msn2 and Msn4 are homologous proteins that activate transcription of many defense genes in response to a wide variety of different stress conditions (25, 43). Msn2 and Msn4 accumulate in the nucleus upon these various stress conditions, whereupon they bind to promoters containing stress response elements (10), Genomic profiling experiments have revealed a subset of stress genes that depend on Msn2/4 for the response to hyperosmotic stress (38). As Sko1 binds the MSN2 promoter region (Fig. 1) and expression of Msn2 is induced in response to osmotic stress (25), we analyzed MSN2 RNA levels in wild-type and sko1 and hog1 mutant cells under normal growth conditions and after osmotic shock (Fig. 2).
|
Sko1 regulates the osmoinducible expression of MOT3 and MGA1, genes that encode transcription factors. In addition to Msn2, Sko1 target genes also include those encoding the DNA-binding transcription factors Mot3, Rox1, and Mga1. MOT3 and MGA1 transcription is highly inducible by salt stress (6-fold and 15-fold, respectively), and this induction is significantly reduced in a sko1 strain and even more strongly affected in a hog1 strain. In contrast, expression of ROX1 is only slightly induced upon osmotic stress, and this induction is independent of Sko1 or Hog1. Thus, Sko1 directly regulates the transcription of at least three genes encoding transcription factors, MSN2, MOT3, and MGA1. In these three cases, Sko1 significantly contributes to but does not fully account for osmotic induction in wild-type cells.
Mot3 and Rox1 synergistically repress the transcription of hypoxic genes (17, 24, 44). Both genes are inducible by hyperosmotic stress, which points to a physiological function of both repressors upon salt stress. This function is not known to date, although a physiological connection between anaerobicity and osmotic stress tolerance has been reported recently (19). Mot3 is also involved in the regulation of ergosterol synthesis and vacuolar function (12), which might be critical under water stress and ion imbalance during osmotic shock. Sko1 binds to the MOT3 and ROX1 promoters, but it appears to play a more important role in osmotically inducible expression of MOT3. We suggest that Sko1-dependent induction of Mot3 repressor might further repress hypoxic genes, allowing the cell to devote more of its energy to deal with hyperosmotic stress.
The biological function of Mga1 is poorly understood, although Mga1 overexpression causes the induction of pseudohyphal growth (23). Although pseudohyphal growth is a long-term phenomenon associated with nutrient deprivation, the process may also be initiated in a transient manner upon other environmental stresses. We speculate that, in response to osmotic stress, Sko1 induces Mga1 to initiate pseudohyphal formation, and that progression towards this distinct cellular state is stopped upon successful adaptation to hyperosmotic conditions.
Sko1-regulated expression of Ptp3, a tyrosine phosphatase that acts on Hog1, provides evidence for a transcriptional feedback mechanism. Sko1 binds in vivo to the PTP3 promoter (Table 1). PTP3 and PTP2 encode two homologous protein tyrosine phosphatases that negatively regulate the HOG pathway by dephosphorylation and sequestration of Hog1 MAP kinase (16, 28, 48). We tested whether Sko1 was responsible for the stress-regulated expression of PTP3 and PTP2 (Fig. 2). PTP3 expression is induced 3.5-fold upon osmotic shock, whereas PTP2 expression is only mildly increased upon stress. Induction of PTP3 transcription is completely dependent on Sko1, while the mild activation of PTP2 seems to be independent.
Our results are strongly suggestive of a transcriptional feedback mechanism in which stress-activated Hog1 kinase directly activates Sko1 by multiple phosphorylations (33, 35) at the PTP3 promoter, thereby inducing PTP3 expression and hence Ptp3 phosphatase levels in response to osmotic stress. As a consequence, increased Ptp3 phosphatase levels favor dephosphorylation of tyrosine-phosphorylated Hog1 kinase, leading to inactivation of the HOG pathway at the later stages of osmotic stress adaptation. As PTP2 expression is much less inducible by osmotic stress and is independent on Sko1, we speculate that Ptp2 might be responsible for HOG inactivation under other environmental stresses recently reported to activate Hog1, such as heat, oxidative or citric acid stress (2, 21, 47).
|
DISCUSSION |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
Transcription factor networks have been identified as an important component of complex transcriptional programs such as the cell cycle (45), but also at a general level (11, 22). Our results indicate that Sko1 activates a regulatory network in addition to its induction of genes that encode proteins that directly relieve osmotic stress. First, Sko1 activates transcription of MSN2, permitting the initial response to a specific stress to reinforce the more general stress response. Second, Sko1-mediated activation of MOT3 permits the response to a specific stress to down-regulate genes that mediate a response to a different stress, allowing the cell to focus its energy on responding to the actual environmental insult. Third, Sko1-mediated activation of MGA1 might contribute to sending cells down a developmental program associated with more stressful conditions.
In a different vein, Sko1 activates PTP3, which encodes a tyrosine phosphatase that dephosphorylates Hog1, the ultimate signaling component that mediates the response to osmotic stress. In this regard, the response to osmotic stress is transient, and increased expression of Ptp3 phosphatase should be appropriately timed to contribute to the down-regulation of the response. Taken together, the transcriptional regulatory network activated by Sko1 is likely to be important for fine-tuning the response to osmotic stress (Fig. 3). Comparable regulatory networks are likely to be mediated by the other activator proteins that participate in the osmotic stress response.
Our results contribute to the emerging view that the relationships between protein binding in vivo, DNA sequence motifs, and transcriptional function are complex. Although many Sko1 target promoters confer induced transcription upon osmotic stress, others show minimal or no regulation. Although ATF/CREB motifs are likely to be important for Sko1 binding, these motifs are not overrepresented among Sko1 targets, indicating that the motifs per se are not sufficient to account for Sko1 binding in vivo. The complex relationships between protein binding in vivo, DNA sequence motifs, and transcriptional activity are not specific to Sko1, and indeed have been observed in many experiments in yeast and human cells (3, 11, 26). In fact, it has been difficult to identify DNA sequence motifs for in vivo binding by many of the yeast DNA-binding proteins that have been analyzed by chromatin immunoprecipitation on a genomewide basis (11). Taken together, these observations indicate that biological specificity and regulatory complexity depend on many parameters and are not easily reduced to simple relationships.
|
ACKNOWLEDGMENTS |
|---|
This work was supported by an EMBO Long Term Fellowship and the "Ramón y Cajal" program of the Spanish Ministry of Science and Technology to M.P. and a grant to K.S. from the National Institutes of Health (GM30186).
|
FOOTNOTES |
|---|
Supplemental material for this article may be found at http://ec.asm.org/.
Present address: Instituto de Biología Molecular y Celular de Plantas (IBMCP), Universidad Politécnica de Valencia, Camino de Vera s/n, 46022 Valencia, Spain.
|
REFERENCES |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Appl. Environ. Microbiol. | Infect. Immun. | J. Bacteriol. |
|---|---|---|
| Mol. Cell Biol. | Microbiol. Mol. Biol. Rev. | ALL ASM JOURNALS |
