The Stress-induced Tfs1p Requires NatB-mediated Acetylation to Inhibit Carboxypeptidase Y and to Regulate the Protein Kinase A Pathway*

The Saccharomyces cerevisiae N-terminal acetyltransferase NatB consists of the subunits Nat3p and Mdm20p. We found by two-dimensional PAGE analysis that nat3 (cid:1) exhibited protein expression during growth in basal medium resembling protein expression in salt-adapted wild-type cells. The stress-induced carboxypeptidase Y (CPY) inhibitor and phosphatidylethanolamine-binding protein family member Tfs1p was identified as a novel NatB substrate. The N-terminal acetylation status of Tfs1p, Act1p, and Rnr4p in both wild type and nat3 (cid:1) was confirmed by tandem mass spectrometry. Furthermore it was found that unacetylated Tfs1p expressed in nat3 (cid:1) showed an (cid:1) 100-fold decrease in CPY inhibition compared with the acetylated form, indicating that the N-terminal acetyl group is essential for CPY inhibition by Tfs1p. Phosphatidylethanolamine-binding proteins in other organisms have been reported to be involved in the regulation of cell signaling. Here we report that a number of proteins, whose expression has been shown previously to be dependent on the activity in the protein kinase A (PKA) signaling pathway, was found to be regulated in line with low PKA activity in the nat3 (cid:1) strain. The involvement of Nat3p and Tfs1p in PKA signaling was supported by caffeine growth inhibition studies. First, growth inhibition by caffeine addition (resulting

The Saccharomyces cerevisiae N-terminal acetyltransferase NatB consists of the subunits Nat3p and Mdm20p. We found by two-dimensional PAGE analysis that nat3⌬ exhibited protein expression during growth in basal medium resembling protein expression in salt-adapted wild-type cells. The stress-induced carboxypeptidase Y (CPY) inhibitor and phosphatidylethanolamine-binding protein family member Tfs1p was identified as a novel NatB substrate. The N-terminal acetylation status of Tfs1p, Act1p, and Rnr4p in both wild type and nat3⌬ was confirmed by tandem mass spectrometry. Furthermore it was found that unacetylated Tfs1p expressed in nat3⌬ showed an ϳ100-fold decrease in CPY inhibition compared with the acetylated form, indicating that the Nterminal acetyl group is essential for CPY inhibition by Tfs1p. Phosphatidylethanolamine-binding proteins in other organisms have been reported to be involved in the regulation of cell signaling. Here we report that a number of proteins, whose expression has been shown previously to be dependent on the activity in the protein kinase A (PKA) signaling pathway, was found to be regulated in line with low PKA activity in the nat3⌬ strain. The involvement of Nat3p and Tfs1p in PKA signaling was supported by caffeine growth inhibition studies. First, growth inhibition by caffeine addition (resulting in enhanced cAMP levels) was suppressed in tfs1⌬. Second, this suppression by tfs1⌬ was abolished in the nat3⌬ background, indicating that Tfs1p was not functional in the nat3⌬ strain possibly because of a lack of N-terminal acetylation. We conclude that the NatB-dependent acetylation of Tfs1p appears to be essential for its inhibitory activity on CPY as well its role in regulating the PKA pathway.
N-terminal acetylation, together with N-terminal methionine cleavage, is the most common protein modification in eukaryotic cells. Over 40% of all yeast proteins and almost 90% of mammalian proteins are estimated to be N-terminally acetylated (1). N-terminal acetylation occurs co-translationally when nascent peptides are between 20 and 50 amino acids long (2,3). The addition of acetyl groups is catalyzed by N-terminal acetyl transferases (NATs). 1 In Saccharomyces cerevisiae there are three known NATs: NatA, NatB, and NatC (1). NatA consists of the subunits Ard1p and Nat1p (4). In addition, Nat5p was recently shown to be associated with Ard1p and Nat1p but does not seem to be important for NatA function (5). NatC consists of the subunits Mak3p, Mak10p, and Mak31p (6). NatB consists of the subunits Nat3p (7) and Mdm20p (8). Ard1p, Nat3p, and Mak3p are related by sequence homology and are the catalytic subunits of NatA, NatB, and NatC, respectively, and all exhibit acetyl-CoA binding sites (1).
NATs act on substrates with a specific but degenerated Nterminal amino acid sequence where certain amino acids in the N-terminal region are required for the activity of each NAT and where suboptimal amino acid residues can diminish the activity (9). In the case of NatB an N-terminal sequence of MD-, ME-, MN-, or MM-is required for acetylation. All proteins with an N-terminal sequence of MD-or ME-that have been characterized so far have been acetylated, while only a subset of the proteins with MN-or MM-at their N terminus are acetylated (10).
Nat1p and Ard1p orthologs in mouse were shown recently to form a complex with acetyltransferase activity (11), and the presence of orthologs to both the catalytic subunits Nat3p, Ard1p, and Mak3p and the auxiliary subunits Mdm20p, Nat1p, and Mak10p in numerous higher eukaryotic model organisms and in humans indicates that the NATs may be found in all eukaryotes. Similarities between the N-terminal amino acid sequences of acetylated proteins in higher eukaryotes and in yeast indicate highly conserved molecular mechanisms of recognition and/or acetyl group addition over large evolutionary distances (10).
Despite the fact that so many proteins are N-terminally acetylated few cases where the N-terminal group is of biological importance for protein function have been reported. The acetylation of the N terminus of the viral coat protein Gag catalyzed by NatC is essential for assembly or maintenance of the viral coat particle in yeast (12). The unacetylated form of fetal hemoglobin, HbF, has been shown to stabilize the hemoglobin tetramer compared with the acetylated form (13). Tropomyosin isolated from striated muscles is dependent on N-terminal acetylation to polymerize and to bind to F-actin in the correct way (14), and acetylation of actin in Dictyostelium has been shown to affect the weak interactions between actin and myo-sin (15). From a medical perspective it is also interesting to note that the human NAT1 ortholog NATH has been shown to be strongly overexpressed in papillary thyroid carcinomas (16); however, the mechanistic implications of this for the disease are not known at present.
S. cerevisiae strains lacking NatA, NatB, or NatC are viable but exhibit various phenotypes. The phenotypes are most severe in strains lacking NatB, and the effects are believed to be related to the lack of acetylation of the two NatB substrates actin and tropomyosin 1 (8). Both these protein need to be acetylated to interact and to form stable actin filaments (17). nat3⌬ cells lacking functional NatB have been reported to exhibit many phenotypes including slow growth, sensitivity to various growth inhibitors when grown on agar plates (e.g. deficiency in utilization of nonfermentable carbon sources), reduced mating of MATa cells, inability to grow at 37°C, abnormal mitochondrial and vacuolar inheritance (8,17), and random budding polarity in diploid cells (18).
In this work we investigated the protein expression pattern of the nat3⌬ strain during growth in high salt. We showed that the protein expression for this mutant in basal medium is very similar to the protein expression found in salt-adapted wildtype cells. Furthermore we identified Tfs1p as a novel NatB substrate. Tfs1p is known to be an inhibitor of the protease carboxypeptidase Y (CPY) (19), and we showed that N-terminal acetylation is important for the inhibitory activity of Tfs1p. Furthermore we found experimental evidence supporting that Tfs1p negatively regulates signaling in the protein kinase A (PKA) pathway and that this inhibitory effect is acetylation-dependent.

EXPERIMENTAL PROCEDURES
Strains, Media, and Growth Conditions-The S. cerevisiae strains used in this study are listed in Table I. Strains in the FY1679 background were kindly provided by Dr. Bogdan Polevoda (7). The multicopy plasmid pKT1067 with the TFS1 gene and a ura marker was kindly provided by Dr. J. Winter with the kind permission of Dr. K. Tatchell. Double deletion mutants were obtained by mating of single mutants, sporulation of diploids, and dissection of tetrads followed by subsequent selection on plates containing 200 g/ml kanamycin. Deletions were confirmed by PCR. Transformations were performed using the lithium acetate/polyethylene glycol method. Cells were grown at 30°C in SD medium (0.14% yeast nitrogen base without amino acids (Difco), 2% (w/v) glucose, 0.5% ammonium sulfate, 1% succinic acid) supplemented with the appropriate amino acids. The medium was buffered to pH 5.8. Cells from stationary phase cultures were inoculated to OD 610 ϭ 0.07. The cell cultures were grown with continues shaking at 180 rpm. Cells grown for protein extraction were harvested by centrifugation at 3000 rpm for 4 min at 4°C. All samples prepared were grown, harvested, and analyzed in triplicates. Microcultivation was performed in 350-l cultures for 48 h in a Bioscreen analyzer C (Labsystems Oy, Helsinki, Finland) according to procedures in an earlier report (20). Duplicates of all samples were run on each plate. Two identical plates with cells from separate overnight cultures were run in parallel. The caffeine concentrations used were 0.5, 1, and 1.5 mg/ml. Strains from the FY background were used for twodimensional PAGE analysis, preparation of protein extracts for MS/MS analysis, and for phenotype analysis in flask culture, while strains from the BY background were used for CPY activity measurements and phenotype analysis in microtiter culture.
Two-dimensional Electrophoresis-For the preparation of radiolabeled proteins 10 Ci of [ 35 S]methionine (Amersham Biosciences catalog no. SJ-235) was mixed with 10 ml of cell culture for one-fourth of a generation time under continuous shaking at 30°C. The isotope was added at A 610 ϭ 0.35. Protein extracts of radiolabeled proteins were prepared by glass bead disruption, and determination of incorporated [ 35 S]methionine was performed as described previously (21). Protein concentrations were determined with trichloroacetic acid precipitation using a Lowry-based commercial kit (Sigma catalog no. P-5656).
Protein extract with a total radioactivity of 2 ϫ 10 6 dpm was loaded on each analytic gel, and 1 mg of protein was loaded on each preparative gel. The immobilized pH gradient strips used were 18 cm long and covered pH 3-10 (non-linear gradient) (Amersham Biosciences). Two-dimensional electrophoresis was performed as described previously (21) using a Multiphor II (Amersham Biosciences) for running the first dimension and an Ettan DALT II (Amersham Biosciences) for running the second dimension.
Gels with radiolabeled proteins were dried and exposed to image plates for roughly 72 h. The plates were scanned using a phosphorimaging system (Bio-Rad Molecular Imager FX) with a resolution of 200 ϫ 200 m. The raw files were processed and put together in a match set in the two-dimensional gel analysis software PDQuest version 6.2.0 (Bio-Rad). The spot detection was checked manually, and matching of all spots was manually performed. Signal quantities in the individual spots on the gels were normalized to the total signal from all spots in each gel, and comparative quantification between corresponding spots on the different gels in the match set was performed. The proteins on the preparative gels were visualized with Coomassie Blue staining according to a published protocol (21).
Protein Trypsinization, Mass Spectrometry, and Analysis of MS/MS Data-Gel pieces were cut out and in-gel trypsinized as described elsewhere (21). Peptides were eluted in 8 l of elution buffer containing 2% acetonitrile and 0.05% formic acid. Peptide separation was performed on a Finnigan Surveyor chromatography work station (Finnigan Corp., San Jose, CA) using a 150 ϫ 0.18-mm C 18 HyPurity column (Thermo Hypersil) or a 150 ϫ 0.18-mm C 18 column produced in-house. Mobile phase 1 Tfs1p Is a NatB Substrate consisted of 95% water, 4.95% acetonitrile, and 0.05% formic acid, and mobile phase 2 consisted of 99.95% acetonitrile and 0.05% formic acid. Liquid chromatography was performed using a linear gradient starting at 0% and reaching 50% phase B after 50 min. The flow rate was 2 l/min. The liquid chromatography device was directly coupled to a Finnigan LCQ DECA ion trap mass spectrometer (Thermofinnigan Corp., San Jose, CA). The mass spectrometer was run in data-dependent scan mode where the three most dominant ions in each MS scan were selected for MS/MS analysis. Dynamic exclusion was set to a maximum of three MS/MS scans in a row, and an exclusion time of 1 min was applied. Where indicated mass spectrometry was performed using MALDI-TOF as described elsewhere (22). The SEQUEST search algorithm was used to correlate experimental mass data to theoretical mass data derived from the sequence in the yeast FASTA data base from the National Center for Biotechnology Information (NCBI). Protein identifications based on tandem mass spectra correlating to one tryptic peptide were considered valid for identification. For singly charged peptides, only spectra with a cross-correlation to a tryptic peptide of 1.5 or more were accepted. The corresponding value for multiple charged peptides was 2.0. Only peptides with a ⌬Cn score larger than 0.1 were accepted. In cases where the identity of the protein had already been established additional peptides with lower scores were taken into account after manual control of the MS/MS spectrum. MALDI-TOF spectra were interpreted using Mascot software. The MALDI-TOF analysis was performed at the SWEGENE proteomics center in Göteborg on a Micromass mass spectrometer.
CPY Activity Measurement-Protein extracts for CPY activity measurement were prepared by glass bead disruption in MES buffer containing 50 mM MES, 1 mM EDTA, and 2.5% methanol with a pH of 6.5. Protein concentrations were determined with a Bradford assay with bovine serum albumin as standard.
CPY activity was determined by measuring hydrolysis of N-(3-[2-furyl]acryloyl)-Phe-Phe (FA-Phe-Phe) (Sigma) over time in MES buffer. Protein samples were mixed with substrate in a total volume of 1 ml and a FA-Phe-Phe concentration of 0.3 mM. Absorbance at 337 nm was measured with a Beckman DU7400 spectrophotometer. Endogenous CPY activity was measured by mixing 150 g of whole cell protein extract with substrate. To measure CPY activity in protein mixtures with a known CPY/Tfs1p ratio, commercially available CPY (Sigma catalog no. 21943) was mixed with protein extracts from prc1⌬ strains lacking endogenous CPY. The purity of the CPY was confirmed on one-dimensional minigels using a standard protocol (data not shown). The protein mixture was incubated at room temperature for 10 min and then mixed with FA-Phe-Phe. The CPY concentration used was 24 nM. The amount of Tfs1p in the protein extracts was quantified by two-dimensional PAGE analysis. Substrate hydrolysis was measured at fixed CPY concentration and various Tfs1p concentrations. The following equation was used to calculate K i(app) .
V i represents the CPY activity in the presence and V 0 represents the CPY activity in the absence of inhibitor. (23). Bright Field and Fluorescence Microscopy-Bright field microscopy was performed on cells harvested from flask cultures. All samples were diluted to OD 610 ϭ 0.1 before cells were observed. Approximately 200 cells were examined for each sample. Microscopy pictures were taken using a Leica DM R microscope. Actin staining was performed using phalloidin-TRITC (Sigma catalog P-1951). Staining was preceded by fixation in 3.7% formaldehyde for 40 min.
Calculation of Growth Variables and Growth Ratios-Growth rate was calculated as described elsewhere (24). To standardize the growth behavior of each strain, a wild type was included as a reference in each run, and a logarithmic strain coefficient (LSC) was calculated according to Equation 2, where ref kj is the growth variable of the kth measurement of the reference strain in environment j, x kj is the growth variable of strain i in environment j, and r indicates the run. From the LSC values a logarithmic phenotypic index (LPI), which describes the sensitivity of the strain for a specific inhibitor, was calculated according to Equation 3.

RESULTS
nat3⌬ Exhibited Biphasic Growth-To study the salt-imposed stress response in the nat3⌬ strain, wild-type FY1679 and nat3⌬ cells were grown and labeled in SD medium and SD medium supplemented with 1 M NaCl in flask cultures. In both media the nat3⌬ strain grew slower than the wild type. However, an interesting feature of nat3⌬ growth was that the growth rate in early exponential phase was faster then in late exponential phase independent of the salinity of the medium. The breakpoint between the two phases was recorded at around OD 610 ϭ 0.5 (Fig. 1A). The generation time of the nat3⌬ in basal medium was 3.2 Ϯ 0.1 h in the first exponential phase (wild type grow in basal medium at 1.7 Ϯ 0.1 h) and 5.8 Ϯ 0.05 h in the second exponential phase, and a similarly prolonged doubling time was also observed in the second exponential phase for the saline cultures. In a previous study nat3⌬ was reported to be salt-sensitive when grown on agar plates (7). We found that the generation time for the nat3⌬ strain was indeed longer in saline medium compared with the wild type, 7.8 Ϯ 0.9 h (first exponential phase) compared with 4.0 Ϯ 0.1 h for the wild type. However, the ratio between growth rates in the presence and absence of salt was roughly 2 for both strains, indicating that the nat3⌬ strain was not specifically slow growing in salt but that it had a general growth defect. In contrast, the stationary phase yield of nat3⌬ grown in saline medium was ϳ50% lower than for salt-free nat3⌬ culture, while the corresponding diminished yield value for the salt-grown wild type was less than 10% (Fig. 1A). We conclude that the reported agar plate salt defect (7, 8) is a combined effect from a generally slower rate of growth for the nat3⌬ strain but most significantly from a decreased yield in salt.
In addition, bright field microscopy revealed that nat3⌬ cells exhibited an increased number of multiple buds. The highest number of cells with two buds or more was observed for exponentially growing cultures where almost half of the nat3⌬ cells had multiple buds (Fig. 1, B and C) independent of which of the two exponential phases were analyzed. On the contrary, less then 10% of wild-type cells had two buds, and no cells with three buds or more were observed. In addition, as reported earlier (17), the mutant displayed defective actin cable formation, which was observed in both the first and second exponential phases (data not shown).
Mass Spectrometric Confirmation of Act1p and Rnr4p as NatB Substrates-Protein expression was analyzed by twodimensional PAGE with protein extracts from wild type and Tfs1p Is a NatB Substrate nat3⌬ exponentially growing in SD medium in the absence and presence of salt. The lack of N-terminal acetylation on substrate proteins in NAT mutant strains is characterized by a shift in the isoelectric point of the protein. The horizontal shift in two-dimensional gels can thus be used to identify NAT substrates. Initially two earlier identified NatB substrate candidates (7), Act1p and Rnr4p, were isolated from two-dimensional gels and characterized from both wild-type and nat3⌬ extracts. Their N-terminal acetylation status was confirmed using electrospray ionization tandem mass spectrometry (ESI-MS/MS). Proteins of wild-type origin were found to be acetylated, while proteins in nat3⌬ lacked the acetyl group in both Act1p (Fig. 2, A and B) and Rnr4p ( Fig. 2G and data not shown). The wild-type Act1p N-terminal peptide ion was found at m/z 993.4 (charge state, ϩ2), while the corresponding Act1p ion in the nat3⌬ was found at m/z 972.4 (charge state, ϩ2). The difference of 42 Da indicated an N-terminal acetylation differ-ence for the two Act1p forms in its N-terminal peptide. The unacetylated status of the N terminus of the Act1p produced in the nat3⌬ strains was confirmed in the MS/MS mode where all detected b-series ions, which include the N-terminal end of the peptide, were shifted by 42 Da and thus indicated an acetylation difference. In this fragmentation series it was evident that the acetyl modification was present at one of the first two amino acids; the b1 fragments in this series could not be detected. However, in the case of Rnr4p we could identify the whole series of fragments and even score the acetylated methionine at its modified mass of 190.1 Da (Fig. 2, G and H; acetylated and oxidized methionine) confirming that this NatBdependent modification occurs at the N-terminal methionine.
Tfs1 Is a Novel Salt-induced NatB Substrate-Besides the earlier identified NatB substrates Act1p and Rnr4p we found a salt-induced (Ͼ4-fold) protein to be a NatB substrate (Fig. 3). In the wild type this protein was positioned on the two-dimensional gel corresponding to a mass of ϳ24 kDa and a pI value of 6.5. In the nat3⌬ this protein apparently shifted its pI to a slightly more basic value (Fig. 3). The protein spots were cut out of the two-dimensional gels from both the wild-type and the nat3⌬ strains, in-gel trypsinized, and identified as YLR178C/ Tfs1p by ESI-MS/MS analysis (Table II). The theoretical mass and pI value of Tfs1p is 24 kDa and 6.5, respectively, further supporting the identification. The N-terminal peptide from both the wild-type and the nat3⌬ forms of Tfs1p were identified and characterized by MS/MS analysis. Similar to Act1p and Rnr4p, the wild-type form of the N-terminal peptide of Tfs1p was found to be 42 Da heavier than the nat3⌬ form. In addition, the b-series fragment ions exhibited a 42-Da shift between the wild-type sample and the nat3⌬ sample, supporting its Nterminal acetylation status (Fig. 2, D, E, and F). The y-series ions, which include the C terminus, were of the same m/z values in both samples. The reason why the pI shift of Tfs1p in nat3⌬ was not recorded earlier is probably that previous studies of protein expression in nat3⌬ (7) have not included samples grown in salt; Tfs1p is poorly expressed in wild-type cells grown in the absence of salt and could therefore easily be missed.
N-terminal Acetylation of Tfs1p Appears Essential for Its Inhibitory Activity on CPY-The NatB substrate Tfs1p is reported to be an inhibitor of the vacuolar protease CPY (19). It has been shown previously that Tfs1p expressed in Escherichia coli inhibits CPY with lower efficiency than Tfs1p expressed in yeast (25). Since prokaryotes do not express NATs it has been claimed that the lack of N-terminal acetylation is the reason for the difference in inhibitory activity. In addition, our two-dimensional PAGE analysis revealed that many proteins in the high molecular weight region of the gels, e.g. Hsc82p, Hsp82p, Eft1p, Met6p, and Aco1p, were down-regulated in nat3⌬ (Fig.  4). This may indicate an elevated protease activity potentially caused by increased CPY activity in nat3⌬ compared with wild type because of the non-functional unacetylated CPY inhibitor Tfs1p in this strain.
To substantiate that the N-terminal acetylation of Tfs1p influences the inhibition of CPY even for natively expressed Tfs1p (to circumvent any secondary effects from heterologous expression in another host like E. coli resulting in inclusion bodies) a CPY inhibition test of endogenous CPY activity was first performed. Breakdown activity on the CPY substrate FA-Phe-Phe was initially measured in whole cell extracts. The strains included in the study were wild type, nat3⌬, tfs1⌬, prc1⌬ (PRC1 coding for CPY), mdm20⌬ (the second subunit of the NatB system), and nat3⌬tfs1⌬. CPY inhibition was measured in strains grown in basal medium. The prc1⌬ strain, lacking CPY, did not exhibit any CPY activity indicating that the assay is specific for this single protease (Fig. 5). Most Tfs1p Is a NatB Substrate importantly, the nat3⌬ and the mdm20⌬ strains, both lacking NatB, showed a 50% increased CPY activity (loss of CPY inhibition) compared with wild type. A similar increase in CPY activity was recorded for the tfs1⌬ strain. In addition, the double mutant nat3⌬tfs1⌬ exhibited a CPY activity similar to that of tfs1⌬ and nat3⌬, indicating an epistatic relationship and that the N-terminal acetylation of Tfs1p is of vital importance for its inhibition of CPY.
To ensure that the difference in CPY activity between nat3⌬ and wild type is not caused by differences in expression or modification of the endogenous CPY in the mutant background, a prc1⌬ strain and a prc1⌬nat3⌬ were transformed with a multicopy plasmid expressing TFS1. Varying amounts of protein extracts were mixed with a fixed amount of commercially available CPY, and the CPY activity/inhibition was determined as described above. The amount of Tfs1p in the protein extract was quantified by two-dimensional PAGE analysis. In Fig. 6 the residual CPY activity of the two different protein mixtures is plotted against the Tfs1p/CPY ratio. The K i(app) values for the acetylated and the unacetylated forms of Tfs1p were estimated to 3 ϫ 10 Ϫ9 and 2 ϫ 10 Ϫ7 M, respectively. These results clearly show that Tfs1p with an N-terminal acetyl group inhibited CPY ϳ100-fold more efficiently then the unacetylated form of Tfs1p.
Global Protein Expression in nat3⌬ Indicates a Connection to the General Stress Response via PKA-In total 582 spots were matched between two-dimensional gels of whole cell protein extracts from wild-type and nat3⌬ cells grown in basal and saline medium. The proteins were labeled at OD 610 ϭ 0.35 when nat3⌬ cells were in the first phase of exponential growth (Fig. 1). 64 proteins were found to be down-regulated 3-fold or more and 51 were up-regulated 3-fold or more in nat3⌬ compared with wild type when grown in basal medium (only including spots where the regulation was statistically significant by log Student's t test criteria). Regulated spots were, if enough material was available for analysis, identified by mass spectrometry (Table III).
Surprisingly we found that many of the expression changes scored in the nat3⌬ strain in basal medium mimicked changes observed in the wild type during stress (Fig. 7, A-D). A hierarchical clustering of wild type and nat3⌬ grown in the presence and absence of salt revealed that nat3⌬ grown in basal medium clustered close to wild type grown in salt and that wild type grown in basal medium clustered far from the other samples (Fig. 7E). The most strongly altered regulation in nat3⌬ compared with the wild type was Zps1p, which was found to be up-regulated 38-fold in basal medium. Zps1p, a homolog to the Candida albicans protein Pra1p with similarity to zinc metalloproteases (26), has been reported previously to be up-regulated in alkaline medium (27) and to be located in the vacuole in S. cerevisiae (28). The DNA damage-induced (29) and flocculent specific (30) protein Ddr48p was up-regulated 5.4-fold in nat3⌬. Ddr48p has been reported previously to be induced when exposed to various stresses like ethylmethane, sulfonate, heat shock, and osmotic stress (31,32). The glycolytic enzymes enolase I and glyceraldehyde 3-phosphate dehydrogenase I (Eno1p and Tdh1p, respectively) were both up-regulated in nat3⌬. Tdh1p has been reported to be induced in response to osmotic stress (33) as well as changes in the intracellular redox balance (34). The unacetylated form of Tfs1p was up-regulated 4-fold in nat3⌬ compared with the acetylated form in wild type (Fig. 3). It is noteworthy that many of the down-regulated proteins during salt stress in the wild type, e.g. Adh1p, Gdh1p, and Met6p, also displayed repression in nat3⌬ grown in the absence of salt.
Tfs1p, identified as a NatB substrate in the present work, has been reported previously to be a multicopy suppressor of the temperature-sensitive cdc25-1 mutant (35). Cdc25p is an activator of the PKA pathway. The expression of many of the salt stress-induced proteins has been shown previously to be partly or fully dependent on the level of PKA activity (33). Protein expression in nat3⌬ was compared with the expression of proteins previously shown to be regulated by the PKA pathway. Among five proteins reported to be up-regulated and eight proteins reported to be down-regulated in a fully or partially PKA-dependent manner, all but one exhibited the same type of regulation in nat3⌬ during growth in basal medium (Fig. 8) indicating low PKA activity in this strain background.
However, the salt regulation of protein expression is complex, and various pathways besides PKA, such as the high osmolarity glycerol mitogen-activated protein kinase pathway, the cell integrity pathway, and the calcineurin/calciumdependent pathway, are involved. An intricate interplay between these different stress-activated pathways sets the final stress output (36). In particular, several of these pathways exhibit partial requirement for PKA or even an antagonistic role to PKA. The latter can be exemplified by the opposing roles played by the calcineurin pathway and PKA (37) and the PKA repression of the nuclear localization of the important transcription factors Msn2/4p and Sko1p during osmotic stress (36). Thus, we do not find it surprising that the salt regulation observed in the nat3⌬ strain differs markedly from the wildtype response. In fact, eight of the 46 proteins induced by stress in the wild type were down-regulated in nat3⌬ during stress (e.g. Fig. 7, A, C, and D), exemplifying the antagonistic roles played by PKA to certain other stress pathways.
Phenotypic Analysis of Caffeine Sensitivity Supported That NatB and Tfs1p Are Linked to the PKA Pathway-The putative link between NatB and the PKA pathway from the expression data presented here is interesting given that some mammalian homologs to the NatB substrate Tfs1p have been shown to affect signaling via interaction with certain kinases (see "Discussion"). Even more importantly, overexpression of TFS1 has been shown to suppress the temperature sensitivity of a point mutation in CDC25 (Cdc25p is believed to act as a positive factor for Ras (35)). The link to the PKA pathway can be experimentally studied by the use of the growth inhibitor caffeine since caffeine inhibits cAMP phosphodiesterase and Tfs1p Is a NatB Substrate thereby increases the intracellular concentration of cAMP. High concentrations of cAMP result in a high PKA activity. Thus, to further investigate the connection among NatB, Tfs1p, and the PKA pathway, we compared the phenotype of various single and double deletion mutants in the presence of growthinhibiting concentrations of caffeine.
The concentrations of caffeine applied resulted in growth rate retardation in the wild type in the range of 30 -300% ( Fig.  9 and data not shown). The link between Tfs1p and PKA was apparent in the strong suppression of the caffeine-imposed growth defect by a TFS1 deletion (Fig. 9A). The suppressive effect was most apparent at the higher caffeine concentrations, clearly seen in the stepwise increased LPI values for tfs1⌬ with increased concentration of the inhibitor (Fig. 9C). A positive LPI indicates resistance to that environment compared with the response in a reference strain (24). However, the suppressive effect was totally absent if the TFS1 gene was deleted in the nat3⌬ background, supporting the notion that the Tfs1p link to PKA is fully dependent on its N-terminal acetylation (Fig. 9C). The suppressive effect of tfs1⌬ (possibly from lowering the intracellular cAMP level) on caffeine growth inhibition was in line with the observed protein expression change that indicated a lower cAMP level in the nat3⌬ strain (with an unacetylated and non-functional Tfs1p).
Is the tfs1⌬ suppression of growth inhibition by caffeine related to its role as a CPY inhibitor? To address this question we deleted the TFS1 gene in the prc1⌬ background. Most importantly and similar to the case where TFS1 was deleted in the nat3⌬ background, the double deletion did not show any improvement in growth compared with the single prc1⌬ strain indicating an epistatic relation. In addition, the single deletion prc1⌬ displayed resistance to caffeine in the same range as for tfs1⌬ (Fig. 9C). These results indicate that the tfs1⌬ suppressive effect is epistatic and "upstream" of CPY.

DISCUSSION
Sequence Requirements for NatB Substrates-N-terminal acetylation is believed to occur on almost 50% of all yeast The raw correlation score. c The difference in correlation score of the top two candidate peptides (55). d The number of peptide ion fragments in the spectra compared to the total number of fragment ions expected.

FIG. 4. Down-regulated spots in nat3⌬.
Spots significantly down-regulated in nat3⌬ during growth in basal medium compared with wild type are indicated with circles on the two-dimensional image. Displayed is a synthetic image of the two-dimensional gel from the wild type generated using PDQuest software.
Tfs1p Is a NatB Substrate proteins (1). However, the loss of the NAT3 gene, which encoded the catalytic subunit for NatB, is reported to influence the N-terminal acetylation on a limited number of yeast proteins; 10 NatB substrates have been identified previously, including Act1p, Tpm1p, Rnr4p, four ribosomal proteins (Rps21ap, Rps21bp, Rps28ap, and Rps28bp), and three subunits of the 26 S proteasome (Pre1p, Rpt3p, and Rpn11p) (Table IV) (7, 17, 38 -40). The sequence requirement for NatB substrates is proposed to be rather strict, and so far all native MD-and ME-termini studied have been found to be acetylated (10). However, the number of proteins in the whole yeast proteome with these termini is almost 600, indicating that many more proteins are potentially modified by the NatB system in yeast.
By extending our analyses of the nat3⌬ strain to salt stress conditions, we here report on the identification of yet another NatB substrate, Tfs1p. Tfs1p is the first native yeast protein with an MN-N terminus that has proven to be a NatB substrate; this N-terminal sequence specificity for NatB was pro-posed previously in studies on mutated variants of iso-1-cytochrome c (Table IV). NatA-dependent N-terminal acetylation in yeast requires the removal of the initial methionine before completion of the nascent protein chain by the action of either of the two methionine aminopeptidases, Map1p or Map2p (9). Experimental evidence has been presented that removal of methionine is dependent on whether the penultimate residue has a radii of gyration that is less than 1.29 Å. Contrary to NatA, NatB acetylates on the first methionine of nascent chains not subjected to methionine cleavage. Since asparagine, similar to aspartate and glutamate, has a radii of gyration greater than 1.29 Å, proteins with an asparagine as the second residue will retain their N-terminal methionine and are thus potentially good NatB substrates.
Tfs1p has the N-terminal amino acid sequence MNQAI-. Some mutant proteins with an N-terminal MN-remain unacetylated; i.e. iso-1-cytochrome c, which has an N-terminal MN-QFL-or MNEKL- (41). To our knowledge Tfs1p is the first protein (native or mutant form) with an MNQ-N terminus found to be acetylated. This indicates that the amino acid sequence determining whether a protein with an MNQ-N terminus will be acetylated by NatB stretches beyond the third amino acid. This is not surprising given that the sequence requirements for acetylation in general have been reported to extend even beyond the fifth amino acid residue for MC-and MT-termini (7). Thus, data indicates that upstream inhibitory residues are present in some putative NatB substrates with MN-terminal sequences, providing less precise theoretical predictions about the acetylation status of these proteins.
Tfs1p Belongs to the PEBP Family of Inhibitors-Tfs1p is a member of the widespread PEBP family. Crystal structure determination of both mammalian and prokaryotic members of the family has revealed that they display almost identical ␤-fold topology despite rather low sequence similarity, suggesting conserved function throughout the family (42). PEBPs have been associated with different forms of protein inhibition activity. In rat and human cells the Tfs1p sequence homolog Raf-1 kinase inhibitor protein (RKIP; formally known as PEBP) has been shown to suppress Raf-1 kinase activity. By binding to either Raf-1 or its substrate MEK-1, RKIP prevents phosphorylation of MEK-1 and thereby regulates the mitogen-  Ϫ3.6 Hsp82p Ϫ3.2 Eft1p Ϫ3.8

Met6p
Ϫ5 Aco1p Ϫ4 Gln1p Ϫ4.1 Gdh1p Ϫ3. Tfs1p Is a NatB Substrate activated protein kinase pathway that activates the extracellular signal-regulated kinases (43,44). In addition to inhibiting Raf-1, RKIP was recently shown to inhibit transcription factor nuclear factor B (NF-B) activation by physically interacting with NF-B-inducing kinases (45). The multiple kinase inhibitory functions of RKIP together with genetic studies that indicate that PEBPs in plants function in signal transduction pathways (46 -49) have led to the suggestion that PEBPs represent a widely conserved family of protein kinase regulators. In this study we also showed that Tfs1p acts as an inhibitor of the protein kinase A signaling pathway in yeast in line with its reported role as a multicopy suppressor of the cdc25-1 mutation (35).
Tfs1p has been characterized previously as a high affinity inhibitor of the vacuolar enzyme CPY (19,25), which is a serine-type protease involved in the C-terminal processing of peptides and proteins. We have confirmed this CPY inhibitory function in vitro, and in addition we identified numerous high molecular weight proteins that were down-regulated in vivo in nat3⌬, indicating a generally increased protease activity. Furthermore PEBP from mouse brain, to which Tfs1p shows 31% sequence identity, has been shown to be an inhibitor of several different serine proteases including thrombin, neuropsin, and chymotrypsin (50), supporting a more general role of PEBP members as protease inhibitors in diverse organisms. The fact that Tfs1p acts both on the protein kinase A signaling pathway and as a potent inhibitor of the vacuolar protease CPY makes Tfs1p the first member of the PEBP/RKIP family with manifested dual inhibitory roles.
Functional Significance of the N-terminal Acetylation of Tfs1p-Since the N-terminal portion of Tfs1p has been shown to be important for the inhibitory action on CPY (19) and the fact that an unacetylated form of Tfs1p, expressed as inclusion bodies in E. coli, displayed impaired inhibitory activity on CPY (25), we performed a CPY activity/inhibition assay to study the impact of the N-terminal acetylation of endogenously expressed Tfs1p. The assay clearly showed that the acetylated form of Tfs1p inhibited CPY much more efficiently than the unacetylated form; K i(app) of Tfs1p expressed in nat3⌬ was increased 100-fold compared with Tfs1p expressed in wild type. Tfs1p expressed in E. coli has been shown to have K i values for anilidase and peptidase activity of CPY increased by 700-and 60-fold, respectively, compared with Tfs1p expressed in yeast (25). We thus believe that our data together with earlier data on heterogenously expressed Tfs1p firmly establish the inhibitory activity of Tfs1p to be dependent on the existence of the N-terminal acetyl group.
What are the molecular mechanisms for this major effect on inhibitory activity of Tfs1p from only this minor protein modification? First of all, it would be expected that the acetylated N terminus would be exposed on the surface of the Tfs1 protein.
The structure of the yeast Tfs1p has not been determined; however, the structural conservation between the bacterial and mammalian homologs indicates that the yeast form would be structurally rather similar. Importantly the determined crystal structures of both bacterial forms expose their N termini (42). Currently no structure for the complex between a PEBP inhibitor and a carboxypeptidase has been determined. However, the structure of the human carboxypeptidase A2 in complex with the leech inhibitor has been determined by x-ray crystallography (51). Even if the leech carboxypeptidase inhibitor does not display any sequence homology to the yeast Tfs1p, it is interesting to note that it is the terminus of the leech inhibitor (however, in this case the C terminus) that extends into the active site of the carboxypeptidase and displaces the reactive water molecule. It can be hypothesized that Tfs1p mechanistically acts in a similar way but instead protrudes its acetylated N terminus into the CPY active site. Consistent with this view is the experimental finding that the specific CPY reagents phenylmethylsulfonyl fluoride and p-chloromercuribenzoate reacted more rapidly against the catalytic site residues Ser 146 and Cys 341 of CPY when in complex with the unacetylated Tfs1p compared with Tfs1p that is properly acetylated (25).
General Functional Significance of NatB-dependent Acetylation-The overall importance of the NatB modification system in yeast is revealed during growth since nat3⌬ cells grow at approximately half the rate of wild-type cells. Although NatA appears to have many more substrates than NatB, strains deleted in NatA subunits grow better in normal growth media than strains lacking a functional NatB (7). 2 In this study we show that Tfs1p is N-terminally acetylated by NatB and demonstrate that the acetyl group is essential for the inhibition of CPY and for regulating the PKA pathway. Tfs1p is the third NatB substrate that depends on its acetyl group to be fully functional. Considering that none of the numerous known NatA substrates have been reported to exhibit altered function 2 R. Caesar and A. Blomberg, unpublished results.

FIG. 7. Comparison of protein expression between samples.
A-D, zoom-in portions of two-dimensional gels. Arrows indicate protein with clearly altered expression in the four samples: A, Met6p indicated; B, Tdh1p indicated; C and D, unidentified protein indicated. E, protein quantities of 582 spots were obtained by quantitative two-dimensional PAGE and used for global quantitative analysis of similarity in expression. Spot quantities close to background level (Ͻ50 ppm) were set to 50 ppm before log 2 transformation and clustering using average linkage hierarchical clustering. The uncentered Pearson correlation coefficient was used as a measure of similarity.
Tfs1p Is a NatB Substrate FIG. 9. tfs1⌬ suppression of caffeine sensitivity. Shown are the results of microcultivation of different gene deletion strains using growth-inhibitory concentrations of caffeine. A, growth of wild type (filled symbols) and tfs1⌬ (open symbols) in low (squares; 0.5 mg/ml) and medium (circles; 1 mg/ml) caffeine. B, growth of nat3⌬ (filled symbols) and nat3⌬tfs1⌬ (open symbols) in low (squares; 0.5 mg/ml) and medium (circles; 1 mg/ml) caffeine. C, display of LPI values for the indicated different strains. White bars represent low concentration (0.5 mg/ml), striped bars represent medium concentration (1 mg/ml), and black bars represent high concentration (1.5 mg/ml) of caffeine. Note that for the three single deletions the response is compared with the response in the wild type, while for the double deletion the nat3⌬ and prc1⌬ single deletions have been used as the reference strain for nat3⌬tfs1⌬ and prc1⌬tfs1⌬, respectively. ref., reference.

FIG. 8. Expression changes for PKA-regulated proteins in nat3⌬.
Shown is the -fold expression change of 13 reporter proteins for PKA-dependent expression (33). -Fold change indicates the relative difference in expression between nat3⌬ compared with wild type, both grown in basal medium.
Tfs1p Is a NatB Substrate 38541 in its unacetylated form, the frequent functional importance of NatB-mediated acetylation is striking. The functional importance of NatB-mediated acetylation of actin and tropomyosin has been reported. When expressed in strains defective in NatB these proteins partially lose their activity. N-terminal acetylation of both proteins is important for F-actin-tropomyosin interaction, which is required to stabilize actin filaments (17). The loss of actin-tropomyosin interaction in nat3⌬ is believed to cause dramatic changes in the cell physiology, and the co-occurrence of this defect with a nonfunctional Tfs1p, and possibly other proteins with altered function, makes the analysis of the nat3⌬ phenotype complex. However, in an elegant series of experiments Polevoda et al. (8) established the great importance of acetylation of actin and tropomyosin for the recorded nat3⌬ phenotypes on a great variety of growth conditions. It was shown that many of these nat3⌬ phenotypes, like NaCl or thermal (37°C) sensitivity, could be suppressed by expressing alleles of actin and tropomyosin that were altered in their N-terminal sequences and that had earlier been shown to be dominant suppressors of mdm20⌬ (17) and thus believed to be functional even without the corresponding N-terminal acetylation. Full to partial suppression of the phenotypes was found on most of the environ-ments tested, highlighting the importance of Act1p and Tpm1p acetylation in establishing the nat3⌬ phenotype. Similar dominant acetylation-independent variants of Tfs1p are not presently available, making corresponding analysis of the phenotypic consequences of non-acetylation of Tfs1p in isolation unfeasible.
The Acetylation of Tfs1p Appears to Be Essential for Its Functional Link to the PKA Pathway-As has been described above, Tfs1p orthologs exhibit kinase inhibitory activity. The observation that overexpression of Tfs1p suppresses the phenotype of cdc25-1 (35) suggests a connection to the control of cell signaling in S. cerevisiae as well. Cdc25p functions as an activator of Ras, which in turn activates the downstream portion of the PKA pathway (Fig. 10). By comparing protein expression in nat3⌬ with expression of proteins reported to be regulated by the PKA pathway we found indications of a low PKA activity in nat3⌬. Transcriptional regulation by the promotor element stress-responsive element (CCCCT) has been demonstrated to be highly dependent on the level of PKA activity (52). All of the proteins listed as up-regulated in nat3⌬ in Table III, except Eno1p and Zps1p, have two or more stressresponsive element in their promoters strengthening a possible PKA-mediated regulation. TFS1 has two stress-responsive elements and has been reported to be strongly up-regulated during steady-state growth in 1 M NaCl 3 and during various stress conditions including oxidative stress, osmotic stress, and heat shock (53). In this work the up-regulation during osmotic stress was confirmed at the protein level, and we also found that Tfs1p is strongly up-regulated in nat3⌬ both in basal medium and in salt-supplemented medium. Interestingly the TFS1 regulation during salt stress and osmotic stress is completely abolished in a MSN2/4 deletion strain indicating a PKA-dependent regulation (53). Since the TFS1 promoter contains stress-responsive elements this provides a setup for a negative feedback loop of the general stress response; as de-scribed below, more Tfs1p leads to greater inhibition of Ira2p, which leads to increased PKA activity resulting in lower expression of TFS1 (Fig. 10).
Caffeine inhibition of cAMP phosphodiesterase leads to increased levels of cAMP that results in high PKA activity. We predict from our phenotypic growth data that inactivation of Tfs1p, via gene deletion or loss of the N-terminal acetyl group, decreases the cAMP level and thus suppresses the caffeine mediated growth inhibition. If the target for inhibition by Tfs1p is upstream of the action of caffeine, a deletion of TFS1 should counteract the caffeine-induced phenotype in wild-type background (Fig. 10). Our data, showing that tfs1⌬ grew better than wild type during caffeine inhibition, supports this assumption.
In line with our model, it was reported during the latter phase of this work that Tfs1p physically interacts strongly with Ira2p and that Tfs1p in that way inhibited the GTPase promoting activity of Ira2p on Ras (54).
In addition, a deletion of TFS1 in nat3⌬ background should not show the same suppressive effect on caffeine growth inhibition if the N-terminal acetylation of Tfs1p is essential for its activity. Interestingly we found that nat3⌬ and nat3⌬tfs1⌬ are equally sensitive against caffeine indicating the caffeine suppressive effect to be acetylation-dependent. In line with the presented physical interaction between Tfs1p and Ira2p by Benedetti and co-workers (54), one can predict that N-terminal acetylation of Tfs1p is important for this interaction. Since Tfs1p is dependent on its acetylation status it might at first seem strange that the nat3⌬ strain is more caffeine-sensitive than the wild type (Fig. 9). However, the earlier mentioned strong influence on the nat3⌬ phenotypes from unacetylated and not fully functional actin and tropomyosin could mask the suppressive effect of the non-functional Tfs1p on caffeine growth inhibition. In fact, at higher caffeine concentrations the relative resistance of nat3⌬ was increased similarly to the tfs1⌬ strain; if the caffeine data is displayed as a relative response (setting the LPI at the lowest caffeine concentration to 1) the caffeine dose-response pattern is almost identical between nat3⌬ and tfs1⌬. We interpret this relative dose-response behavior as a positive caffeine suppressive effect from a nonfunctional Tfs1p also in the nat3⌬ strain.
Interestingly we also found the suppression of growth inhibition by caffeine by TFS1 deletion to be dependent on an intact CPY. This would place CPY mechanistically downstream of Tfs1p. Bruun et al. (19) predicted previously that the suppressive effect on the cdc25-1 mutation from TFS1 overexpression was not caused by complete inhibition of CPY (which in principle could be the result of TFS1 overexpression) since prc1⌬ did not suppress the cdc25-1 mutation. However, TFS1 suppression of cdc25-1 could still be dependent on CPY activity but be operational via some mechanisms other than CPY inhibition, and we here propose a model in which Ira2p is selectively degraded in the vacuole by CPY and in which this delocalization is dependent on Tfs1p physically interacting with both Ira2p and CPY (Fig. 10). This interesting putative interconnection between the dual functionalities of the Tfs1 protein certainly deserves further in depth analysis.