Desumoylation of the Endoplasmic Reticulum Membrane VAP Family Protein Scs2 by Ulp1 and SUMO Regulation of the Inositol Synthesis Pathway

Abstract

Posttranslational protein modification by the ubiquitin-like SUMO protein is critical to eukaryotic cell regulation, but much remains unknown regarding its operation and substrates. Here we report that specific mutations in the Saccharomyces cerevisiae Ulp1 SUMO protease, including its coiled-coil (CC) domain, lead to the accumulation of distinct sumoylated proteins in vivo. A prominent ∼50-kDa sumoylated protein accumulates in a Ulp1 CC mutant. The protein was identified as Scs2, an endoplasmic reticulum (ER) membrane protein that regulates phosphatidylinositol synthesis and lipid trafficking. Mutation of lysine 180 of Scs2 abolishes its sumoylation. Notably, impairment of either cellular sumoylation or cellular desumoylation mechanisms inhibits cell growth in the absence of inositol and exacerbates the inositol auxotrophy caused by deletion of SCS2. Mutants lacking the Ulp2 SUMO protease are the most severely affected, and this defect was traced to the mutants' impaired ability to induce transcription of INO1, which encodes the rate-limiting enzyme of inositol biosynthesis. Conversely, inositol starvation induces a striking change in the profiles of total cellular SUMO conjugates. These results provide the first evidence of cross-regulation between the SUMO and inositol pathways, including the sumoylation of an ER membrane protein central to phospholipid synthesis and phosphoinositide signaling.

INTRODUCTION

The SUMO (small ubiquitin-related modifier) proteins are a family of protein modifiers conserved from the yeast Saccharomyces cerevisiae to humans. Modification of a protein by SUMO (sumoylation) has a variety of outcomes, including alterations of its localization, stability, or interaction with other molecules (18). Increasingly, SUMO is being identified as a key coordinator of protein interaction networks, and disruption to sumoylation has been linked to a variety of disorders, including many cancers and neurodegenerative diseases (25, 41).

The yeast S. cerevisiae expresses a single SUMO protein called Smt3. Smt3 is synthesized as an inactive precursor; the three C-terminal residues are removed to yield mature Smt3, which terminates with a pair of Gly residues. Following precursor cleavage, Smt3 enters an enzymatic cascade involving a series of enzymes designated E1, E2, and E3. The heterodimeric E1-activating enzyme Uba2-Aos1 first forms a high-energy thioester bond with the C terminus of Smt3 (22). Activated Smt3 is then transferred to the active-site cysteine of the E2 enzyme Ubc9 (20), from which it is then conjugated to a lysine side chain in a substrate (21). Smt3-substrate conjugation is usually assisted by one of several E3 protein ligases (21, 50). Additional Smt3 proteins can be appended to previously conjugated Smt3 molecules, generating a polySUMO chain.

Sumoylation is reversible, and substrates can be desumoylated by specific proteases, termed ubiquitin-like protein-specific proteases (ULPs) (10). S. cerevisiae has two SUMO proteases, Ulp1 and Ulp2, which cleave Smt3 from distinct sets of substrates (29, 31). Only Ulp1 efficiently processes the Smt3 precursor, whereas Ulp2 is largely responsible for the disassembly of polySUMO chains (4, 29, 31).

Ulp1 is essential for viability (29, 31). Despite its importance, much remains unknown about how it functions in vivo. The Ulp1 protein can be divided into three broad domains: an N-terminal domain (residues 1 to 340) involved in its tethering to the nuclear basket of the nuclear pore complex (NPC), a predicted coiled-coil (CC) domain (residues 340 to 403), and a C-terminal catalytic domain responsible for the SUMO-deconjugating and precursor-processing activities (30). Little is known about the function of the CC domain. Previously, we reported that the CC domain was required for Ulp1 localization to esc1Δ-induced nuclear basket clusters but not for its general localization to the nuclear periphery (28). Presumably, the CC domain contributes to the interaction of Ulp1 with nuclear basket proteins and disruption of these interactions alters the way in which Ulp1 contacts the NPC.

We had previously generated a conditional allele of ULP1 named ulp1-333 (ulp1ts) (29, 31). Our sequence analysis revealed multiple point mutations in ulp1ts, and in our current study, we have correlated the accumulation of distinct sumoylated proteins in vivo with specific mutations. Most strikingly, a single point mutation in the CC domain of Ulp1 (K352E) was found to have a significant impact on the desumoylation of several substrates. To help characterize substrates that accumulate in ulp1ts cells, we purified sumoylated proteins from the mutant and analyzed them by mass spectrometry. This investigation identified proteins spanning a broad range of biological functions and included Scs2, a type II integral endoplasmic reticulum (ER) membrane protein and member of the vesicle-associated, membrane protein-associated protein (VAP) family (23, 27). VAP members are conserved in all eukaryotic organisms and have been linked to the regulation of many cellular functions, including membrane trafficking, lipid transport and metabolism, the unfolded protein response, and microtubule organization (see reference 27 for a review). Disruption of the human VAP-B gene by a single missense mutation, P56S, has been identified in three forms of familial motor neuron disease (36, 37).

A major function of the Scs2 protein is the regulation of inositol synthesis (16, 23, 35). Inositol is an essential compound in S. cerevisiae and serves as the structural core for a number of second-messenger signaling molecules and major phospholipid bilayer components (1, 6). If sufficient quantities of inositol are not provided in the media, yeast must synthesize it de novo. To do this, glucose 6-phosphate is converted to inositol 3-phosphate, followed by dephosphorylation (6). Formation of inositol 3-phosphate is catalyzed by the rate-limiting enzyme inositol-3-phosphate synthase, which is encoded by INO1 (26). INO1 is part of a network of phospholipid synthesis genes regulated by the inositol-responsive element (UAS_INO) and the transcription factors Ino2, Ino4, and Opi1 (5) (see Fig. 7A).

Fig 7.

SUMO pathway mutants have defects in inositol regulation. (A) Schematic of Scs2 function in inositol synthesis. (B) ulp2Δ and ubc9ts mutant cells grow poorly on medium lacking inositol, and ulp1ts, ulp2Δ, and ubc9ts mutant cells interact synthetically with the scs2Δ mutant. Serial dilutions of yeast cultures were spotted onto plates and grown under the indicated conditions. d, day. (C) Supplementation with mature Smt3 does not suppress the ulp1ts scs2Δ double mutant defect. Yeast cells were transformed with pRS316-SMT3GG or empty vector and grown at 30°C. (D) Mature Smt3 suppresses the inositol sensitivity in ulp2Δ cells and very weakly suppresses the genetic interaction between ulp2Δ and scs2Δ cells. Yeast cells were transformed with pRS316-SMT3GG or empty vector and grown at 30°C. (Some variability in ulp2Δ suppression was observed; this is probably due to a threshold level of Smt3 being required, based on more uniform suppression with vectors expressing higher levels of Smt3.) (E) Scs2 sumoylation is not required for its role in inositol regulation. Cells deleted for SCS2 were transformed with pRS316-SCS2, pRS316-scs2-R180,R181, or empty vector and grown at 36.5°C for 3 days. Two transformants of each genotype are shown. (F) OPI1 deletion partially suppresses the inositol sensitivity of ulp2Δ cells but not that of ubc9ts cells.

Based on our results indicating that Scs2 is a Ulp1 substrate, we further investigated the links between Scs2, the SUMO pathway, and inositol synthesis. We found that Scs2 is monosumoylated on lysine 180 and is desumoylated solely by Ulp1. We also uncovered an unexpected connection between SUMO and inositol; specifically, several SUMO pathway enzymes are required for growth under conditions of low levels of inositol. For at least one of these mutants, the ulp2Δ strain, the poor growth is due at least in part to a defect in transcriptional derepression of INO1. Inositol starvation, in turn, alters SUMO conjugate accumulation. Together these results suggest an interdependent relationship between SUMO and inositol, in which protein sumoylation regulates inositol production and inositol modulates SUMO protein conjugation in the cell.

MATERIALS AND METHODS

Yeast strains.

Yeast strains used are listed in Table 1. Standard media and techniques were used for their growth and construction (13). The scs2Δ (MHY5891), scs2Δ ulp1ts (MHY5893), scs2Δ ubc9ts (MHY6259), opi1Δ (MHY6000), opi1Δ ulp1ts (MHY6006), opi1Δ ubc9ts (MHY6405), and opi1Δ ulp2Δ (MHY6624) strains were made by PCR amplifying the scs2Δ::kanMX4 and opi1Δ::kanMX4 cassettes from the gene deletion library (Open Biosystems) and by integrating the amplified alleles into the wild-type (WT) (MHY1395), ulp1ts (MHY4709), ubc9ts (MHY5809), and ulp2Δ (MHY6147) genomes. The strains used for the liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis of the sumoylated proteins, MHY5341 and MHY5343, were constructed by integrating the GAL1 promoter, either with or without a sequence encoding an in-frame His₆ tag, in front of the SMT3 chromosomal locus of strain MHY4806. The GAL1 promoter was generated by performing PCR using the pFA6a-kanMX6-PGAL1 plasmid (33) as a template. Amplified DNA was transformed into MHY4806. Colonies were selected on G418 plates containing galactose, and integration was confirmed by anti-Smt3 immunoblotting. These strains were maintained in media containing 2% galactose.

Table 1.

Yeast strains used in this study

Strain	Genotype	Source or reference
W303	MATα ade2-1 ura3-1 his3-11 trp1-1 leu2-3,112 can1-100	R. Rothstein
BY4741	MATa his3Δ1 leu2Δ0 met15Δ0 ura3Δ0	43
MHY4201	MATα ulp1Δ::kanMX4 his3Δ0 leu2Δ0 ura3Δ0 (YCplac22-ulp1ts-NAT)	28
MHY4479	MATα ulp1Δ::kanMX4 his3Δ0 leu2Δ0 ura3Δ0 (pRS415-ulp1-N*-GFP)	This study
MHY4480	MATα ulp1Δ::kanMX4 his3Δ0 leu2Δ0 ura3Δ0 (pRS415-ulp1-C*-GFP)	This study
MHY4566	MATa yra1-2::his5+ ade2-1 ura3-52 his3-11 trp1-11 leu2-3,112 can1-100 bar1Δ	24
MHY4709	MATα ulp1Δ::LEU2 ade2-1 ura3-1 his3-11,15 trp1-1 leu2-3,112 can1-100 (YCplac22-ulp1ts)	This study
MHY4806	MATα ulp1Δ::LEU2 yra1-2::his5+ ade2-1 ura3-1 his3-11,15 trp1-1 leu2-3,112 can1-100 (YCplac22-ulp1ts)	This study
MHY5341	MATα ulp1Δ::LEU2 yra1-2::his5+ kanMX6::GAL1-H6-SMT3 ade2-1 ura3-1 his3-11,15 trp1-1 leu2-3,112 can1-100 (YCplac22-ulp1ts)	This study
MHY5343	MATα ulp1Δ::LEU2 yra1-2::his5+ kanMX6::GAL1-SMT3 ade2-1 ura3-1 his3-11,15 trp1-1 leu2-3,112 can1-100 (YCplac22-ulp1ts)	This study
MHY5809	MATα ubc9Δ::TRP1 ubc9ts::LEU2 ade2-1 ura3-1 his3-11,15 trp1-1 leu2-3,112 can1-100	This study
MHY5816	MATα ulp2Δ::HIS3 ade2-1 ura3-1 his3-11,15 trp1-1 leu2-3,112 can1-100	This study
MHY5891	MATα scs2Δ::kanMX4 ade2-1 ura3-1 his3-11 trp1-1 leu2-3,112 can1-100	This study
MHY5893	MATα scs2Δ::kanMX4 ulp1Δ::LEU2 ade2-1 ura3-1 his3-11,15 trp1-1 leu2-3,112 can1-100 (YCplac22-ulp1ts)	This study
MHY5904	MATα SCS2-V5::kanMX6 ade2-1 ura3-1 his3-11 trp1-1 leu2-3,112 can1-100	This study
MHY5910	MATα SCS2-V5::kanMX6 ulp1Δ::LEU2 yra1-2::his5+ ade2-1 ura3-1 his3-11,15 trp1-1 leu2-3,112 can1-100 (YCplac22-ulp1ts)	This study
MHY5941	MATα SCS2-V5::kanMX6 ulp1Δ::LEU2 ade2-1 ura3-1 his3-11,15 trp1-1 leu2-3,112 can1-100 (YCplac22-ulp1ts)	This study
MHY6000	MATα opi1Δ::kanMX4 ade2-1 ura3-1 his3-11 trp1-1 leu2-3,112 can1-100	This study
MHY6006	MATα opi1Δ::kanMX4 ulp1Δ::LEU2 ade2-1 ura3-1 his3-11,15 trp1-1 leu2-3,112 can1-100 (YCplac22-ulp1ts)	This study
MHY6147	MATα ulp2Δ::HIS3 scs2Δ::kanMX4 ade2-1 ura3-1 his3-11 trp1-1 leu2-3,112	This study
MHY6259	MATα scs2Δ::kanMX4 ubc9Δ::TRP1 ubc9ts-LEU2::leu2-3,112 ade2-1 ura3-1 his3-11,15 trp1-1 can1-100	This study
MHY6405	MATα opi1Δ::kanMX4 ubc9Δ::TRP1 ubc9ts-LEU2::leu2-3,112 ade2-1 ura3-1 his3-11,15 trp1-1 can1-100	This study
MHY6595	MATα ulp1Δ::kanMX4 his3Δ0 leu2Δ0 ura3Δ0 (pRS415-ulp1-N*-3N-GFP)	This study
MHY6624	MATα opi1Δ::kanMX4 ulp2Δ::HIS3 ade2-1 ura3-1 his3-11,15 trp1-1 leu2-3,112 can1-100	This study
MHY6770	MATα ulp1Δ::kanMX4 his3Δ0 leu2Δ0 ura3Δ0 (pRS415-ULP1-GFP)	This study
MHY6772	MATα ulp1Δ::kanMX4 his3Δ0 leu2Δ0 ura3Δ0 (pRS415-ulp1-N*-3C-GFP)	This study
MHY6774	MATα ulp1Δ::kanMX4 his3Δ0 leu2Δ0 ura3Δ0 (pRS415-ulp1-Q272R,N332S-GFP)	This study
MHY6776	MATα ulp1Δ::kanMX4 his3Δ0 leu2Δ0 ura3Δ0 (pRS415-ulp1-K352E-GFP)	This study

To tag Scs2 with a C-terminal V5 epitope, a DNA sequence encoding V5 was introduced in frame at the 3′ end of the chromosomal SCS2 open reading frame (ORF) in the WT (MHY1395), ulp1ts (MHY4709), and ulp1ts yra1-2 (MHY4806) strains to generate strains MHY5904, MHY5941, and MHY5910, respectively. The V5 tag sequence (with a Gly₆ linker) was generated by PCR using the pFA6a-6×Gly-V5-kanMX6 plasmid (11) as a template. Yeast transformants were selected on G418 plates, and integration was confirmed by anti-V5 immunoblotting.

The ULP1-GFP strain MHY6770 and its corresponding mutants (MHY4479, MHY4480, MHY6595, MHY6772, MHY6774, and MHY6776) were made by transforming plasmids bearing various ulp1-GFP fusion alleles (Table 2) into a strain deleted for the chromosomal ULP1 locus but carrying either a URA3-marked ULP1 plasmid or a NAT-marked ulp1ts plasmid. Following transformation, colonies that had lost the original covering plasmid were identified.

Table 2.

Plasmids used in this study

Plasmid	Description	Source or reference
16-9-9	YCplac22-ulp1ts(333)	29
18-6-3	pFA6a-kanMX6-PGAL1	33
2-8-8	pRS316	42
4-1-4	YCplac33	12
43-1-2	pFA6a-6×Gly-V5-kanMX6	11
47-6-2	pRS316-SCS2	This study
47-7-1	pRS316-scs2-K180R,K181R	This study
48-5-3	YCplac33-HA-scs2-K181R	This study
48-5-7	YCplac33-HA-scs2-K180R,K181R	This study
48-6-6	YCplac33-HA-scs2-K180R	This study
49-4-1	pRS415-ulp1-N*-3N-GFP	This study
54-3-6	pRS316-SMT3(GG)	This study
56-2-1	pRS415-ulp1-N*-3C-GFP	This study
56-4-5	pRS415-ulp1-Q272R,N332S-GFP	This study
56-6-9	pRS415-ULP1-GFP	This study
56-7-5	pRS415-ulp1-K352E-GFP	This study
BOK32	pRS415-ulp1-N*-GFP	This study
BOK35	pRS415-ulp1-C*-GFP	This study
BOK58	YCplac22-ulp1ts-NAT (NAT marker inserted into AflII site of ulp1ts)	This study
pKY166	YCplac33-HA-SCS2	23
pMR1036	INO1-CYC-lacZ reporter	7
60-4-3	pRS426-INO1	46

Plasmids.

The plasmids used are listed in Table 2. Genomic SCS2 and SMT3 and their endogenous promoters and terminators were PCR amplified, cloned into different plasmids, and confirmed by DNA sequencing. QuikChange mutagenesis (Stratagene) was used to delete the DNA sequence encoding the last 3 amino acids of Smt3 and to mutate lysine codons 180 and/or 181 to arginine in SCS2. All mutations were confirmed by DNA sequencing.

Plasmid pRS415-ULP1-GFP bears the wild-type (WT) ULP1 genomic sequence and the endogenous ULP1 promoter fused to DNA sequence encoding a C-terminal green fluorescent protein (GFP) tag. Plasmid pRS415-ulp1-C*-GFP contains a WT ULP1 sequence at its 5′ end (corresponding to residues 1 to 417) and a mutant ulp1-333 sequence at its 3′ end (corresponding to residues 418 to 621) (O. Kerscher and M. Hochstrasser, unpublished data). Plasmid pRS415-ulp1-N*-GFP has the reverse configuration, with a mutant ulp1-333 sequence at its 5′ end (corresponding to residues 1 to 417) and a WT ULP1 sequence at its 3′ end (corresponding to residues 418 to 621). Plasmid pRS415-ulp1-N*-3N-GFP was made by ligating the ∼1.5-kb SalI-NotI fragment of pRS415-ULP1-GFP to the ∼9-kb SalI-NotI fragment of pRS415-ulp1-N*-GFP. Plasmid pRS415-ulp1-N*-3C-GFP carries the ∼1.5-kb SalI-NotI fragment from pRS415-ulp1-N*-GFP ligated to the ∼9-kb SalI-NotI fragment of pRS415-ULP1-GFP. QuikChange mutagenesis was used to correct the K352E mutation in pRS415-ulp1-N*-3C-GFP to generate pRS415-ulp1-Q272R,N332S-GFP. Plasmid pRS415-ulp1-K352E-GFP was made by mutating the lysine 352 codon in pRS415-ULP1-GFP. All constructs were confirmed by DNA sequencing.

Western immunoblot assays.

For all experiments except the inositol starvation experiments, cells were grown at 30°C overnight in 2 ml yeast extract-peptone-dextrose (YPD) or selective medium, diluted to an optical density at 600 nm (OD₆₀₀) of ∼0.25 in 5 ml fresh medium, and grown for an additional 6.5 h. For inositol starvation, cells were grown at 30°C overnight in 4 ml synthetic defined (SD) complete medium, diluted to an OD₆₀₀ of ∼0.45 in 5 ml fresh medium, and grown for 2 h. Cells were then washed once with water, resuspended in either 5 ml SD complete (with inositol) or SD−inositol (without inositol) medium, and grown for an additional 5 h.

A volume of culture corresponding to 3.5 to 5 OD₆₀₀ equivalents of cells was centrifuged, and cell pellets were resuspended in 750 μl 20% trichloroacetic acid (TCA) (Sigma) and kept at −80°C overnight. Samples were thawed, pelleted, and resuspended in 400 μl 20% TCA and an equal volume of glass beads, followed by disruption in a FastPrep-24 instrument (MP Biomedicals) at 4°C, using two 20-s pulses. The supernatant was removed from the beads and centrifuged at high speed. Protein pellets were washed once with 950 μl 2% TCA, resuspended in SDS gel-loading buffer plus 150 mM Tris (added to neutralize any residual TCA), and boiled for 5 min. After electrophoresis through a 6 to 15% gradient SDS-polyacrylamide gel, the proteins were transferred to a polyvinylidene difluoride (PVDF) membrane and immunoblotted as previously described (28). The primary antibodies used were anti-Smt3 (29), anti-V5 (Invitrogen), anti-HA.11 16B12 (Covance), anti-GFP (Roche Diagnostics), and anti-PGK 22C5 (Molecular Probes), and the secondary antibodies consisted of either donkey anti-rabbit IgG linked to horseradish peroxidase or sheep anti-mouse IgG linked to horseradish peroxidase (GE Healthcare). Enhanced chemiluminescence (ECL) reagents were used for detection as directed by the manufacturer (GE Healthcare).

Purification of SUMO conjugates for LC-MS/MS.

Cells were grown in 5 ml yeast extract-peptone-galactose medium (YPGal) at 30°C overnight. Cultures were diluted into 50 ml YPGal and incubated at 30°C for a second night. The following morning, cells were diluted to an OD₆₀₀ of 0.3 in 500 ml YPGal and incubated at 30°C for 7 h. A volume of culture corresponding to 295 OD₆₀₀ equivalents of cells was harvested and washed once with water, and the cell pellets were frozen at −80°C overnight. Thawed pellets were resuspended in 7 ml buffer L (8 M urea, 50 mM sodium phosphate [pH 7.0], 500 mM NaCl, 1% NP-40) and split evenly into 12 2-ml tubes. Six hundred microliters of glass beads was added to each tube. Samples were lysed in a FastPrep-24 instrument for two 20-s pulses at 4°C. Lysed samples were cleared by centrifugation at 15,000 rpm for 5 min at 4°C. Supernatants originating from the same cells were pooled and subsequently divided into 10 tubes of 700 μl each. One hundred fifty microliters of preequilibrated Talon Superflow resin (Clontech Laboratories) was added to each tube, and the tubes were rotated for 2 h at 4°C. Samples were centrifuged at 3,000 rpm for 1 min, and the supernatant was removed. The resin was washed twice by adding 1 ml buffer L and rotating the mixture for 5 min at 4°C. After the second wash, the resin was resuspended in 150 μl buffer L and two samples of the same type were pooled into 5-ml disposable columns (Pierce). Each column was washed once with 1.5 ml buffer L followed by 1.5 ml buffer W (8 M urea, 50 mM sodium phosphate [pH 7.0], 500 mM NaCl, 1% NP-40, 15 mM imidazole). Bound protein was eluted by the addition of 1 ml of 200 mM EDTA to each column and precipitated by the addition of 20% TCA, incubation on ice for 10 min, and centrifugation for 5 min at 13,200 rpm. Pellets were washed once with 2% TCA. One pellet of precipitated eluate was resuspended in 60 μl SDS gel-loading buffer containing 150 mM Tris. Resuspended samples of each type were pooled, boiled for 5 min, and resolved by SDS-PAGE. After the gel was stained with GelCode Blue (Thermo Scientific), matched gel slices from each sample were excised with clean razor blades and used for LC-MS/MS analysis.

Protein identification by mass spectrometry.

The mass spectrometry analysis was performed according to an optimized procedure of liquid chromatography coupled with tandem mass spectrometry (LC-MS/MS) (48). In brief, gel slices were subjected to in-gel trypsin digestion to generate peptides. The peptides were then analyzed by capillary reverse-phase chromatography and tandem mass spectrometry on a hybrid LTQ-Orbitrap instrument (Thermo Fisher Scientific, San Jose, CA). The acquired MS/MS spectra were matched against a database containing proteins that were potentially present in the samples. The matching peptides were further rigorously filtered by the target-decoy search strategy (39) to reduce the protein false-discovery rate to <0.1% in the final list. All accepted proteins sharing peptides were grouped together, and the top protein with the highest peptide-matching number was selected to represent the group.

Immunopurification of Scs2-V5.

Cells expressing Scs2-V5 (or control cells lacking the V5 epitope) were grown at 30°C overnight in 5 ml YPD, diluted to an OD₆₀₀ of ∼0.25 in 50 ml fresh medium, and grown for an additional 6.5 h. A volume of culture corresponding to 40 OD₆₀₀ equivalents of cells was harvested, and cell lysates were generated as described for Western immunoblot assays. Following a 2% TCA wash, protein pellets were resuspended in 100 μl buffer R (0.06 M Tris-HCl [pH 6.8], 200 mM Tris, 10% glycerol, 2% sodium dodecyl sulfate, 5% β-mercaptoethanol). Samples were incubated at 100°C for 5 min, followed by centrifugation at 13,200 rpm for 10 min. The supernatants (100 μl) were transferred to a plugged Handee spin column (Pierce; 10-μm-pore-size polyethylene filter) and diluted 6.6-fold in buffer NLB (10 mM potassium acetate, 20 mM HEPES [pH 7.5], 2 mM MgCl₂, 0.1% Tween 20, 1% Triton X-100, 5 mM N-ethylmaleimide [NEM], and a mini, complete, EDTA-free protease inhibitor cocktail tablet [Roche Diagnostics]). Following addition of 45 μl of anti-V5 agarose (Sigma), the columns were rotated for 2 h at 4°C. The resin was washed four times with 600 μl buffer NLB, resuspended in 50 μl SDS gel-loading buffer lacking β-mercaptoethanol, and incubated at 100°C for 10 min. Eluates were collected by centrifuging the column at 13,200 rpm for 1 min. Finally, 5% β-mercaptoethanol was added to each eluate, and the samples were boiled for 5 min.

Growth analysis by cell dilution series.

Cells were grown overnight in YPD or selective medium, pelleted and resuspended in water at 1 OD₆₀₀ equivalent per ml, and subsequently diluted in 5-fold steps. Cells were spotted on plates, and growth was observed over several days. For analysis of cells in the absence of inositol, synthetic defined media were prepared with a yeast nitrogen base lacking both amino acids and inositol (For Medium, Ltd.).

β-Galactosidase assays using a UAS_INO1-lacZ reporter.

WT (MHY1395), ulp2Δ (MHY5816), scs2Δ (MHY5891), and ulp2Δ scs2Δ (MHY6147) yeast strains were transformed with the reporter plasmid INO1-CYC1-lacZ (pMR1036) (7). Independent transformants were inoculated in liquid minimal medium lacking leucine and supplemented with 75 μM inositol (SD−Leu plus 75 μM inositol) and grown overnight at 30°C. Cultures were diluted in fresh SD−Leu plus 75 μM inositol and grown to mid-log phase (4 h at 30°C). To starve cells of inositol, the cultures were pelleted, washed once with water, and resuspended in selective medium lacking inositol (SD−Leu−Ino). Cells were grown in SD−Leu−Ino for 5 h at 30°C to an OD₆₀₀ of ∼1.0 to 1.3 and used for β-galactosidase assays. A two-tailed, paired t test was used for statistical comparison of activity values.

RESULTS

Specific sumoylated proteins accumulate in response to distinct mutations in ULP1.

We showed previously that the ulp1ts mutant, a temperature-sensitive yeast mutant which contains the ulp1-333 allele, accumulates multiple sumoylated proteins at permissive and semipermissive temperatures (29) (Fig. 1A). In some instances, combining ulp1ts with mutations in other genes that interact with ulp1ts led to additional alterations in SUMO conjugate profiles. Such was the case with the yra1-2 mRNA export mutant. Not only was the ulp1ts yra1-2 mutant synthetically lethal at intermediate temperatures (i.e., 33°C) (data not shown), it also displayed a novel sumoylated protein of ∼36 kDa (Fig. 1A, open arrowhead).

Fig 1.

Accumulation of sumoylated proteins in ulp1 mutants. Whole-cell lysates were analyzed by Western immunoblotting. “Light” and “dark” refer to short and long exposure times, respectively. (A) Sumoylated proteins (black arrowheads) accumulate in ulp1ts cells. An additional ∼36-kDa species (open arrowhead) builds up in ulp1ts yra1-2 cells. (B) Schematic of the nine point mutations that comprise the ulp1ts allele (ulp1-333) and the mutant constructs derived from ulp1ts. NPC, nuclear pore complex; CC, coiled-coil domain. (C) Mutations in the N- and C-terminal domains of Ulp1 differentially alter preSmt3 processing and Smt3 conjugates. A single dot indicates equivalent increases in the ulp1-N* and ulp1-C* mutants; double dots indicate greater increases in the ulp1-N* mutant; triple dots indicate greater increases in the ulp1-C* mutant; quadruple dots indicate decreases in the ulp1-C* mutant. (D) Expression of mature Smt3 in ulp1-C* cells grown at 37°C enhances accumulation of sumoylated proteins (arrowheads). (E) Different sumoylated proteins accumulate when Ulp1 contains mutations in either its CC domain (bracket) or N-terminal region (arrowheads).

As a first step to analyzing the basis for the accumulation of specific SUMO conjugates in ulp1ts cells, we analyzed the sequence of the ulp1-333 allele, which had been generated by low-fidelity PCR (29). The mutant gene was found to have nine missense mutations dispersed throughout its sequence (Fig. 1B). To determine which of these mutations were responsible for the mutant phenotype, we created multiple ulp1 plasmids containing various subsets of the nine ulp1-333 mutations and analyzed their effects on growth and SUMO conjugate accumulation by anti-Smt3 immunoblotting. Separation of the C-terminal catalytic domain mutations (ulp1-C*) from the N-terminal/CC domain mutations (ulp1-N*) demonstrated partially overlapping effects; multiple bands increased in both mutants (Fig. 1C, compare bands indicated by single and double dots). However, the ulp1-C* mutant also displayed some unique effects, including prevention of growth at 37°C (not shown). The precursor Smt3 accumulated only in the ulp1-C* mutant, and the levels of an ∼40-kDa sumoylated protein decreased in this strain (Fig. 1C, indicated by triple and quadruple dots, respectively).

Mutations in the C-terminal domain of Ulp1 hinder its catalytic activity, including its ability to convert preSmt3 into mature Smt3 (Fig. 1C) (30). Consequently, the full effect of the ulp1-C* mutations on SUMO protein deconjugation might be masked by the reduced sumoylation that occurs in these cells, particularly at higher temperatures. We therefore provided ulp1-C* cells with mature Smt3 expressed from a plasmid and grew the transformants at 37°C, which exacerbates the effect of the C-terminal point mutations. An anti-Smt3 immunoblot analysis of cell lysates (Fig. 1D) showed that all of the sumoylated proteins that accumulate in ulp1ts cells (Fig. 1C) also accumulated in ulp1-C* cells expressing mature Smt3. Thus, the C-terminal Ulp1 protease domain is required for the desumoylation of all sumoylated proteins that accumulate in the ulp1ts mutant.

Further resolution of the N-terminal and CC domain mutations from ulp1-333 showed that the accumulation of specific sumoylated proteins correlated with distinct mutations in Ulp1. The ulp1-N*-3N mutant, which has only the three N-terminal mutations in the Ulp1 NPC-targeting domain, accumulated a sumoylated species of ∼35 kDa (Fig. 1E). Most notable, however, was the effect of a single point mutation, K352E, in the predicted CC domain. This allele was associated with enhanced levels of multiple sumoylated proteins, including species of ∼50 kDa and ∼70 kDa (Fig. 1E, last lane). The K352E mutation did not detectably alter Ulp1 localization or its steady-state levels (Fig. 1E and 2). These results indicate that both the N-terminal and CC domains of Ulp1 contribute to substrate desumoylation, but the requirement for each is substrate specific.

Fig 2.

Localization and levels of the ulp1-K352E protein are similar to those of WT Ulp1 (ULP1-GFP). (A) Fluorescence (GFP) and differential interference contrast (DIC) images of live yeast strains. (B) Cell lysates from the cultures used in panel A were assayed by anti-GFP immunoblotting.

Identification of sumoylated proteins from a ulp1ts mutant.

The above-described data imply that Ulp1 interacts with or gains access to its various substrates differently. To help understand such differences, the identities of Ulp1 substrate proteins must be determined. In an earlier study (28), we had examined the function of Ulp1 in mRNA export. While extending these findings, we found that combining ulp1ts with yra1-2, a mutant allele of the mRNA export factor YRA1 gene, led to the appearance of a novel sumoylated species (Fig. 1A). We therefore used the ulp1ts yra1-2 mutant for the identification of SUMO conjugates by LC-MS/MS.

A sequence encoding a His₆ tag was integrated at the 5′ end of endogenous SMT3, and His₆-Smt3, along with its conjugates, was affinity purified under strongly denaturing conditions. Western blotting showed that the isolation of SUMO conjugates was specific to the His₆ tag-bearing strain, as no sumoylated proteins were purified from a nontagged control (Fig. 3A). The eluted proteins were resolved in a preparative SDS-polyacrylamide gel, and matching gel slices were extracted from the His₆-tagged and nontagged sample lanes (Fig. 3B, dashed boxes). A comparison of these samples by GelCode Blue staining showed that similar arrays of contaminating proteins were purified in both. Proteins in the excised slices were trypsinized and analyzed by LC-MS/MS in two groups: group 1 corresponded to proteins of 35 to 60 kDa, and group 2 corresponded to proteins of 60 to 150 kDa (Fig. 3B, solid boxes).

Fig 3.

Purification of Smt3 conjugates. (A) Proteins were purified under denaturing conditions on Talon Superflow resin from ulp1ts yra1-2 cells expressing either His₆-SMT3 (H) or untagged SMT3 (S). Lysates were analyzed by anti-Smt3 immunoblotting. Arrowheads indicate sumoylated proteins whose migration shifts due to the His₆ tag. (B) Eluates from the purification shown in panel A were resolved on a 10% SDS gel and stained with GelCode Blue. Matched gel slices (dashed boxes) were cut from lanes H and S and analyzed by LC-MS/MS. Samples were processed in two groups (solid boxes). M, molecular mass standards.

Table 3 summarizes the MS/MS results and lists proteins that had a number of sequenced peptides in the His₆-tagged sample that was at least three times that of the nontagged control. His₆-Smt3 has a molecular mass of ∼13 kDa. Most of the proteins purified in the sumoylated substrate group were confirmed to be modified based on a change in their apparent masses of 13 kDa or more. Two exceptions, Cdc48 and Hsp104, were noted; these anomalies suggest that only sumoylated fragments of these proteins were isolated or that their identification was a result of contamination. Of the remaining proteins identified, several are well-characterized sumoylated substrates, such as the septin Cdc3 and the DNA sliding clamp PCNA (15, 19), and others had been detected in previously conducted proteomic studies (14, 38, 47, 51). We were unable to identify the ∼36-kDa sumoylated conjugate that increased specifically in ulp1ts yra1-2 cells.

Table 3.

Proteins identified as sumoylated substrates by the presence of a ≥3-fold-higher number of peptides

Protein (predicted mass in kDa)	No. of peptides per sample expressing:		Group(s)a	Functionb
	Untagged Smt3 (S)	His6-Smt3 (H)
Cdc3 (60)	0	6	2	Component of the septin ring
Cdc48 (92)	0	6	1	ATPase with homology to mammalian p97
Gpd1 (43)	0	3	1	NAD-dependent glycerol-3-phosphate dehydrogenase; key enzyme of glycerol synthesis
Hsp104 (102)	0	6	1, 2	Heat shock protein that cooperates with Ydj1 and Ssa1 to refold and reactivate denatured, aggregated proteins
Lys1 (41)	0	3	1	Saccharopine dehydrogenase; catalyzes the conversion of saccharopine to l-lysine
Pol30 (29)	0	5	1	Proliferating cell nuclear antigen; functions as the sliding clamp for DNA polymerase delta
Pgk1 (45)	8	34	1, 2	3-Phosphoglycerate kinase; catalyzes the transfer of high-energy phosphoryl groups from the acyl phosphate of 1,3-bisphosphoglycerate to ADP to produce ATP
Rps0a (28)	0	3	1	Protein component of the small (40S) ribosomal subunit
Rps0b (28)	0	3	1	Protein component of the small (40S) ribosomal subunit
Rps3 (27)	0	3	1	Protein component of the small (40S) ribosomal subunit
Rsc8 (63)	0	4	2	Component of the RSC complex
Scs2 (27)	0	7	1, 2	Integral ER membrane protein that regulates phospholipid metabolism
Tfg1 (82)	0	4	2	TFIIF largest subunit; involved in both transcription initiation and elongation of RNA polymerase II
Tup1 (78)	0	8	2	General repressor of transcription
Yel007w (73)	0	3	2	Putative protein with sequence similarity to Schizosaccharomyces pombegti1+ (gluconate transport inducer 1)
Zuo1 (49)	0	3	1	Cytosolic ribosome-associated chaperone

^aGroup 1 consists of proteins of 35 to 60 kDa. Group 2 consists of proteins of 60 to 150 kDa.

^bRSC complex, chromatin structure remodeling complex; TFIIF, transcription factor II.

The proteins in Table 3 span a range of cellular functions, from chromatin remodeling to metabolism. Several components of the small (40S) ribosomal subunit were identified, as were mediators of various biosynthetic pathways. One of the identified proteins, Scs2, is a highly conserved regulator of lipid trafficking and inositol biosynthesis and is an integral ER/nuclear membrane protein (27). Ulp1 is largely tethered to the nuclear side of the NPC, which might regulate its access to Scs2. Intriguingly, a previous proteomic study had suggested that Scs2 sumoylation increases under certain stress conditions (38, 51), although its sumoylation had never been verified (38, 51). Based on these observations, we decided to examine the potential sumoylation of Scs2 in more detail.

Scs2 is covalently modified by Smt3.

To confirm that Scs2 is sumoylated, we integrated a sequence encoding a V5 epitope at the 3′ end of endogenous SCS2 to allow immunological detection of Scs2. If Scs2 is indeed sumoylated, the V5 tag should cause a shift in the molecular weight of sumoylated Scs2 that is observable by anti-Smt3 immunoblotting if this species is sufficiently abundant. In ulp1ts cells, a sumoylated protein of ∼50 kDa accumulated; this is close to the size predicted for monosumoylated Scs2 (Fig. 4A, lane 4). The migration of this species was slightly slower when Scs2 was tagged with V5 (Fig. 4A, lanes 2 and 3 compared to lane 4), suggesting that it corresponds to sumoylated Scs2. In support of this, the 50-kDa sumoylated protein disappeared in cells from which SCS2 had been deleted (Fig. 4B); this was observed in both WT and ulp1ts backgrounds. Finally, when Scs2-V5 was immunoprecipitated from ulp1ts cells under denaturing conditions and the precipitated proteins were evaluated by anti-Smt3 immunoblotting, a sumoylated ∼50-kDa species was detected (Fig. 4C). Together, these data demonstrate that Scs2 is a sumoylated protein whose sumoylation levels greatly increase in the ulp1ts mutant.

Fig 4.

Sumoylation of Scs2. (A) An ∼50-kDa sumoylated protein accumulates in ulp1ts cells and migrates more slowly when Scs2 is tagged with the V5 epitope. Cell lysates were analyzed by Western blotting. (B) Accumulation of the 50-kDa species (arrowhead) in ulp1ts cells is suppressed by deletion of SCS2. Cell lysates were subjected to Western blotting. A single dot indicates a second sumoylated protein whose presence depends on SCS2. (C) Smt3 is associated with Scs2 following lysis under denaturing conditions. Scs2-V5 was immunoprecipitated (IP), and eluates were analyzed by immunoblotting (IB). Double dots indicate an IgG heavy chain.

Scs2 is desumoylated by Ulp1 but not Ulp2.

Sumoylated Scs2 accumulated in the ulp1ts mutant, suggesting that it is a substrate of the Ulp1 protease. To determine whether sumoylated Scs2 is also desumoylated by the other yeast SUMO protease, Ulp2, the level of sumoylated Scs2 was compared in mutants of these two proteases. WT, ulp1ts, and ulp2Δ cells were transformed with a plasmid expressing hemagglutinin (HA)-tagged Scs2, and whole-cell lysates were subjected to Western blotting. Both sumoylated HA-Scs2 and sumoylated endogenous Scs2 accumulated to high levels in ulp1ts but not ulp2Δ cells (Fig. 5). Therefore, Smt3-Scs2 is deconjugated by Ulp1 but not Ulp2.

Fig 5.

Sumoylated Scs2 accumulates in ulp1ts but not ulp2Δ cells. Cell lysates from WT, ulp1ts, or ulp2Δ cells transformed with a YCplac33-HA-SCS2 plasmid were analyzed by anti-HA immunoblotting. (A) Anti-HA blot. “Light” and “dark” refer to short and long exposure times, respectively. (B) The same extracts as those in panel A were analyzed by anti-Smt3 immunoblotting. Sumoylation of both endogenous Scs2 and HA-Scs2 is observed in ulp1ts cells.

Scs2 is sumoylated on lysine 180.

In Smt3 conjugates, Smt3 is ligated to a lysine side chain(s) in the substrate, and the lysine is often part of a consensus motif, ψ-K-X-(E/D), where ψ is a hydrophobic residue, typically I, V, or L, and X represents any amino acid (44). Scs2 contains one consensus motif, VKKE, with K180 at the consensus lysine position, which falls between the major sperm protein (MSP) domain and the transmembrane domain (Fig. 6A). We mutated lysines 180 and 181 to arginines, which cannot be conjugated to Smt3. Anti-Smt3 immunoblotting of ulp1ts scs2-R180,R181 cell extracts revealed a complete loss of sumoylated Scs2 (Fig. 6B). Single lysine substitutions showed that only the K180R mutation prevented the accumulation of sumoylation (Fig. 6C). Neither mutation affected total Scs2 protein levels, although there was a notable migration difference dependent on each, suggesting that these residues may be important for other Scs2 modifications (Fig. 6C) (see Discussion). These data strongly suggest that Scs2 is sumoylated on lysine 180.

Fig 6.

Mutation of Scs2 lysine 180 abolishes Scs2 sumoylation. (A) Domain architecture of Scs2. MSP, major sperm protein domain; TM, transmembrane domain; VKKE, sumoylation consensus motif. (B) Mutating lysines 180 and 181 to arginines in the consensus motif blocks accumulation of sumoylated Scs2 in ulp1ts cells. ulp1ts scs2Δ mutant cells expressing pRS316-SCS2, pRS316-scs2-K180,181R, or an empty vector (-) were lysed, and extracts were analyzed by Western blotting. (C) The K180R single mutation, but not the K181R mutation, blocks Scs2 sumoylation. ulp1ts scs2Δ mutant cells bearing the YCplac33 vector, YCplac33-HA-SCS2, or the indicated lysine mutant alleles were lysed and assayed by immunoblotting.

Sumoylation regulates inositol synthesis.

One of the best-characterized functions of Scs2 is its role in the promotion of inositol synthesis. When inositol levels are low, Scs2 sequesters the transcriptional repressor Opi1 from UAS_INO promoter sites (Fig. 7A) (5). In the absence of Scs2, Opi1 remains constitutively bound to UAS_INO-containing promoters, thereby inhibiting INO1 transcription and ultimately blocking de novo production of inositol. Consequently, scs2Δ cells grow poorly on medium lacking inositol and die under these conditions at elevated temperatures (23).

As Scs2 is modified by Smt3, we wished to determine whether protein sumoylation also regulates inositol production. Yeast containing mutations in one of the SUMO proteases or in the E2 SUMO-conjugating enzyme Ubc9 were grown on medium lacking inositol. While the ulp1ts mutant consistently showed at most a mild growth defect under these conditions, ubc9ts and especially ulp2Δ mutants were more strongly impaired (Fig. 7B). All three mutations resulted in enhanced inositol auxotrophy when they were combined with a deletion of SCS2.

The Ulp1 and Ulp2 SUMO proteases cleave Smt3 from protein substrates (29, 31), but mutations in either enzyme also cause a general inhibition of sumoylation. For ulp1 mutants, this is due, at least in part, to an inability of the cell to process the Smt3 precursor. In contrast, cells lacking Ulp2 accumulate polySmt3 chains (4), and these chains appear to act as a sink for Smt3 and limit the amount of free Smt3 available for conjugation (D. Su and M. Hochstrasser, unpublished observations). When the growth assays using medium lacking inositol were repeated with cells expressing exogenous mature Smt3, the growth defect of ulp1ts scs2Δ cells was not rescued (Fig. 7C). However, addition of extra mature Smt3 partially suppressed the inositol auxotrophy of ulp2Δ cells; interestingly, this suppression appeared much weaker if cells also lacked SCS2 (Fig. 7D). Collectively, these results demonstrate that protein sumoylation is required for growth in the absence of inositol, and when Scs2 is absent, the protein desumoylation functions of Ulp1 and Ulp2 become critical as well.

Potentially, the sumoylation of Scs2 itself contributes to the regulation of inositol synthesis. To test this hypothesis, cells deleted for endogenous SCS2 and expressing plasmid-borne SCS2 or scs2-R180,R181 were tested for complementation on medium lacking inositol. Both plasmids fully complemented the lethality of scs2Δ (and scs2Δ ulp1ts) on SD-inositol medium at high temperatures (Fig. 7E). This finding suggests that despite the requirement for efficient cellular sumoylation under conditions of low levels of inositol, sumoylation of Scs2 does not seem to be essential.

Scs2 acts upstream of Opi1 in the transcriptional regulation of inositol synthesis, and deletion of OPI1 has been shown to suppress the inositol auxotrophy of scs2Δ cells (3). To determine whether the SUMO enzymes act in the same pathway, double mutants with mutations in SUMO pathway enzymes and opi1Δ were constructed. Interestingly, deletion of OPI1 partially suppressed the inositol auxotrophy of ulp2Δ cells but not that of ubc9ts cells (Fig. 7F). In fact, ubc9ts opi1Δ cells grew slightly worse than either single mutant, although this had already been seen under inositol-replete conditions. These data suggest that SUMO plays multiple roles in regulating inositol synthesis, both upstream of Opi1 (in the case of Ulp2) and either downstream or in a parallel pathway (in the case of Ubc9).

Impaired transcriptional induction of INO1 in ulp2Δ cells.

The ability of opi1Δ to suppress the inositol auxotrophy of the ulp2Δ cells raised the possibility that the absence of Ulp2 prevents normal transcriptional induction of INO1 on medium lacking inositol. As a first test of this idea, we determined whether an increased dosage of the INO1 gene could enhance growth of ulp2Δ cells grown in the absence of inositol (46). Indeed, substantial suppression of inositol auxotrophy in ulp2Δ cells was observed (Fig. 8A). To determine more directly whether Ulp2 was required for INO1 transcriptional derepression, we measured the UAS_INO-dependent induction of a lacZ reporter gene (7). In WT cells, a large increase in β-galactosidase (β-Gal) activity was observed after 5 h of inositol starvation at 30°C (Fig. 8B). Under these conditions, INO1 induction was partially blocked in scs2Δ cells, as expected. The induction defect was almost as great in the ulp2Δ mutant. Loss of both Scs2 and Ulp2 led to an even more striking defect in INO1 induction. Notably, the growth defect of ulp2Δ cells on plates lacking inositol was stronger than that of the scs2Δ single mutant (Fig. 7B), consistent with the inference that inositol auxotrophy in ulp2Δ cells is not due simply to a lack of Scs2 activity.

Fig 8.

Loss of Ulp2 impairs INO1 transcription. (A) Cells transformed with either empty high-copy-number (HC) vector (pRS426) or pRS426-INO1 were spotted in serial dilutions on the indicated media and grown for 3 days at 30°C. (B) An INO1 promoter-lacZ reporter plasmid (pMR1036) was transformed into the indicated strains, and after growth in high-inositol medium (75 μM), cells were harvested, washed, and grown in the indicated minimal medium (+INO, 75 μM inositol) at 30°C for 5 h, followed by activity assays. Each bar represents the average activity of five transformants. Error bars indicate the standard deviations. By a paired t test, the P values for the differences between mean values under inositol-depleted (−INO) conditions were as follows: WT versus ulp2Δ strains, 0.028; WT versus scs2Δ strains, 0.0046; ulp2Δ versus scs2Δ strains, 0.025; ulp2Δ versus ulp2Δ scs2Δ strains, 0.00027; and scs2Δ versus ulp2Δ scs2Δ strains, 0.00034. All differences were therefore considered statistically significant.

We conclude that a significant component of inositol auxotrophy in ulp2Δ cells results from a failure to fully induce INO1 transcription and that Scs2 and Ulp2 contribute in distinct ways to transcriptional derepression during inositol starvation.

Inositol starvation changes protein-SUMO conjugate patterns.

The impact of SUMO pathway disruption on inositol auxotrophy prompted us to investigate whether inositol starvation, in turn, affects sumoylation levels. WT and ulp1ts cells were grown in either the presence or absence of inositol, and SUMO conjugate accumulation was assessed by anti-Smt3 immunoblotting. Figure 9 shows that, in general, the levels of both free Smt3 and total Smt3 conjugation increased in response to inositol starvation, regardless of genotype. Similar effects were seen for ulp2Δ, ubc9ts, and scs2Δ cells (data not shown). Interestingly, an ∼40-kDa sumoylated protein in ulp1ts cells showed the opposite trend and decreased upon inositol starvation. The level of sumoylated Scs2 appeared unchanged. These results suggest that inositol starvation induces a cellular response that leads to enhanced global sumoylation but also stimulates the desumoylation of at least one specific protein.

Fig 9.

Effect of inositol starvation on sumoylation. Cells were grown either in the presence (+) or absence (−) of inositol for 5 h at 30°C. Cell lysates were analyzed by Western blotting. There was some experimental variation in the extent of Smt3 (free and conjugate) upregulation observed upon inositol starvation, but the trend remained consistent. An ∼40-kDa band (arrowhead) decreased upon inositol starvation. “Light” and “dark” refer to short and long exposure times, respectively. MW, molecular weight.

DISCUSSION

Over the past decade, SUMO has emerged as a key regulator of protein function. Here we have shown that distinct residues of the Ulp1 SUMO protease are required for the desumoylation of specific proteins and have found a contribution of the protease's CC domain to this specificity. We identified the integral membrane protein Scs2 as one of the major sumoylated proteins that accumulates in ulp1 mutant cells. Scs2 is modified by Smt3 on lysine 180 and is desumoylated solely by Ulp1. Genetic analysis revealed that sumoylation of Scs2 is not required for its role in promoting inositol synthesis. Importantly, however, the SUMO pathway in general is needed for survival under low-inositol conditions, and the SUMO pathway mutant with the most severe inositol auxotrophy, the ulp2Δ strain, is defective in transcriptional induction of INO1, which encodes the rate-limiting enzyme for inositol biosynthesis. Conversely, inositol starvation was found to induce a general upregulation of sumoylation in cells while leading to the desumoylation of at least one specific protein. These data provide the first evidence that protein sumoylation regulates inositol metabolism in the cell.

A new role for the Ulp1 coiled-coil domain.

Sequence analysis of the ulp1-333 (ulp1ts) allele revealed nine point mutations. The desumoylation of multiple proteins, including an ∼50-kDa protein which we subsequently identified as sumoylated Scs2, was dependent not only on the catalytic domain but also on the integrity of a putative CC domain preceding the catalytic domain. Little is known about this region of Ulp1, and this is the first evidence showing that it is required for the desumoylation of specific substrates. The data presented here suggest that the CC domain of Ulp1 contributes to its ability to interact productively with at least a subset of substrates. The K352E point mutation is predicted to interfere with the specificity of coiled-coil formation. Cells expressing ulp1-K352E-GFP do not have an obvious localization defect compared to those expressing Ulp1-GFP, nor does the mutation cause Ulp1 to become unstable (Fig. 1E and 2). It is possible that the CC domain subtly yet significantly affects the way in which Ulp1 interacts with the NPC. Mutating the CC domain might then change how Ulp1 is situated at the nuclear pore and consequently limit its access to certain substrates. Such an explanation may also account for why deletion of the CC domain was previously found to prevent Ulp1 from localizing to esc1Δ-induced nuclear basket clusters despite the concentration of the Ulp1-binding proteins Nup60 and Mlp1 at these sites (28). Alternatively, or in addition, the K352E mutation might directly inhibit the interaction of Ulp1 with certain substrates, including Scs2, without altering its NPC interactions more generally. The VAP family proteins are characterized by a variable CC domain (23, 27).

Function of Scs2 sumoylation.

Preventing the sumoylation of Scs2 did not affect its role in inositol regulation (Fig. 7E) (although subtle defects might have been missed), nor did it appear to alter other known functions of Scs2, such as cortical ER inheritance or telomeric silencing (data not shown) (8, 9, 32). Consequently, a specific function cannot yet be attributed to sumoylated Scs2. Determining the precise functional consequences of sumoylation can often be challenging. SUMO is frequently found to coordinate networks of protein interactions involving multiple sumoylated proteins and SUMO-interacting proteins, and inhibiting just one of these modifications is often insufficient to disrupt the network. Therefore, one possibility is that the impairment of Scs2 sumoylation is not enough to disrupt its interactions with other proteins.

It is also possible that mutation of the sumoylated lysine of Scs2 (lysine 180) to arginine blocked not only SUMO addition but also another modification on this lysine or a nearby site. Other known lysine modifications include methylation, acetylation, and ubiquitylation (49). Scs2 was reported to be ubiquitylated in one large-scale study (40). Some evidence suggests that Scs2 can have modifications other than sumoylation. Full-length Scs2 has been shown by SDS-PAGE, in both this study and others, to migrate significantly slower than expected from its predicted size of 27 kDa (23). Mutating lysines 180 and 181 to arginine either individually or together shifted the apparent size of Scs2 closer to its predicted size (Fig. 6C). In contrast, sumoylated HA-Scs2 and HA-scs2-K181R migrated similarly, suggesting that sumoylation may mask the effects of other modifications. Considering these observations, it is plausible that Scs2 sumoylation may work in concert with other Scs2 modifications to regulate its function.

SUMO regulates inositol synthesis in multiple ways.

Unexpectedly, genetic analysis revealed multiple roles for the SUMO pathway in regulating inositol production (Fig. 7). Both sumoylation and desumoylation activities are required for normal growth under conditions of inositol starvation. It is noteworthy that earlier genome-wide screens for genes required for inositol prototrophy (45) or for transcriptional changes in response to inositol starvation (17) failed to find any link to the SUMO system. The former screen used the yeast gene deletion collection. Because most SUMO pathway components are required for viability (and ulp2Δ is not in the collection), these factors were not identified. The results of the latter screen suggest that transcriptional changes in response to inositol do not generally include SUMO pathway genes. We note, however, that WSS1, a gene that modulates SUMO function, is moderately repressed in inositol-replete media (17).

Sumoylation is needed both upstream of the canonical Scs2-mediated pathway and either downstream or in a parallel pathway, as the inositol auxotrophy of the ulp2Δ mutant was (partially) suppressed by deletion of OPI1 but that of the ubc9ts mutant was not. Moreover, synthetic interactions were observed between the ulp1ts, ulp2Δ, and ubc9ts mutants and the scs2Δ mutant and between the ubc9ts and opi1Δ mutants. These interactions suggest requirements for sumoylation and desumoylation in the absence of SCS2 or OPI1. These requirements could include additional roles in regulating inositol levels directly or, alternatively, in regulating a pathway that becomes essential when inositol synthesis is impaired.

We looked at this more closely with the ulp2Δ mutant, the SUMO pathway mutant with the most severe growth defect on media lacking inositol. In addition to being suppressed by opi1Δ, the inositol auxotrophy of ulp2Δ cells was also suppressed by increased dosage of the INO1 gene (Fig. 8A). Most importantly, ulp2Δ cells fail to fully derepress UAS_INO-dependent transcription upon inositol starvation, similar to what was observed with an scs2Δ strain, and transcription was further impaired in a ulp2Δ scs2Δ double mutant (Fig. 8B). Therefore, Ulp2 is important for the transcriptional activation of INO1, perhaps by desumoylation of a specific transcription factor or histones (34). Further studies will be needed to identify the relevant sumoylated substrates regulating UAS_INO-dependent transcription. Candidate substrates include the transcriptional activators Ino2 and Ino4; however, we have been unable to detect sumoylation of these proteins (not shown). Known SUMO targets, such as histones or the general corepressor Tup1, are additional candidates.

We also found that inositol starvation led to a general increase in sumoylated proteins in the cell (Fig. 9). Increased sumoylation has previously been reported to occur in response to various stressors, such as hydrogen peroxide, ethanol, or aging (2, 51). Inositol starvation could be inducing a general stress response in which cells increase overall sumoylation. Interestingly, we found that levels of free Smt3 also increased following inositol starvation. Preliminary experiments measuring SMT3 transcript levels by real-time quantitative PCR (qPCR) did not indicate a significant difference in the amounts of transcript present under high- or low-inositol conditions (data not shown). Potentially, translation of SMT3 mRNA might be upregulated or degradation of Smt3 protein might be reduced under these conditions. Another interesting observation was that despite the overall increase in sumoylation, the levels of an ∼40-kDa sumoylated protein decreased in response to inositol starvation. This indicates that in addition to a general stress response, inositol starvation leads to specific changes in the modification of certain proteins. Together with the genetic data, these results suggest a close relationship between the SUMO conjugation and inositol synthesis pathways, wherein SUMO regulates the synthesis of inositol and, conversely, inositol levels impact protein sumoylation patterns in the cell.