Yeast as a Tractable Genetic System for Functional Studies of the Insulin-degrading Enzyme*

We have developed yeast as an expression and genetic system for functional studies of the insulin-degrading enzyme (IDE), which cleaves and inactivates certain small peptide molecules, including insulin and the neurotoxic Aβ peptide. We show that heterologously expressed rat IDE is enzymatically active, as judged by the ability of IDE-containing yeast extracts to cleave insulin in vitro. We also show that IDE can promote the in vivo production of the yeast a-factor mating pheromone, a function normally attributed to the yeast enzymes Axl1p and Ste23p. However, IDE cannot substitute for the function of Axl1p in promoting haploid axial budding and repressing haploid invasive growth, activities that require an uncharacterized activity of Axl1p. Particulate fractions enriched for Axl1p or Ste23p are incapable of cleaving insulin, suggesting that the functional conservation of these enzymes may not be bidirectionally conserved. We have made practical use of our genetic system to confirm that residues composing the extended zinc metalloprotease motif of M16A family enzymes are required for the enzymatic activity of IDE, Ste23p, and Axl1p. We have determined that IDE and Axl1p both require an intact C terminus for optimal activity. We expect that the tractable genetic system that we have developed will be useful for investigating the enzymatic and structure/function properties of IDE and possibly for the identification of novel IDE alleles having altered substrate specificity.

The insulin-degrading enzyme (IDE 1 ; EC 3.4.24.56) has broad substrate specificity, being able to cleave and inactivate a number of small molecules, including the A␤ peptides and insulin (1)(2)(3). In animal models, IDE deficiency correlates with increased levels of insulin and A␤ and increased risks of type 2 diabetes and Alzheimer disease (4 -6). Despite these and other findings, the biological role of IDE and its importance in the clearance of insulin and A␤ remains to be fully clarified (7).
The Ste23p and Axl1p proteins from Saccharomyces cerevisiae have significant sequence homology to IDE (36 and 19% identity, respectively) (8). Genetic and mutational studies indicate that Ste23p and Axl1p are required for the proteolytic maturation of the yeast a-factor mating pheromone, a lipidmodified peptide that is produced by a multistep process (9,10). This pheromone is produced by MATa haploid cells and is required for yeast mating, the fusion of MATa and MAT␣ haploid cells to form a diploid cell. Although Ste23p and Axl1p can independently promote a-factor maturation, the enzymes are not fully redundant. Ste23p is significantly less efficient than Axl1p at producing biologically active pheromone (9).
Axl1p has several cellular roles. In addition to its role in pheromone production, Axl1p is required for the efficient postconjugation fusion of haploid mating partners (11). This role may be indirectly related to the ability of Axl1p to produce a-factor, because limiting levels of pheromone similarly lead to fusion defects (12). Axl1p is also required for maintenance of the axial budding pattern that is characteristic of haploid yeast (8,9). In the absence of Axl1p, haploid yeasts exhibit a bipolar budding pattern that is typical of diploid cells. The function of Axl1p in this process does not require its proteolytic activity, suggesting that Axl1p is a bi-functional enzyme (9). Axl1p also represses the invasive growth of haploid yeast (13,14). Whether the proteolytic, bud site selection, or another activity of Axl1p is required for this process has not been reported. By contrast to Axl1p, the only reported cellular role for Ste23p is in a-factor production. Additional functions for this enzyme are likely because Ste23p is expressed in both the MAT␣ and diploid cell types that do not produce a-factor. 2 IDE, Ste23p, and Axl1p are members of the M16A protease subfamily. M16A proteases are characterized by a core inverted zinc metalloprotease motif that is typically located within the first 200 residues of these enzymes and a pair of glutamic acid residues that are 70 and 77 amino acids distal to the core motif (i.e. HXXEHX 69 EX 6 E). The spacing between the core motif and the distal glutamates is invariant for all 46 M16A enzymes identified so far, except in the case of Axl1p where the spacing is 76 residues. The histidine residues and most distal glutamate residue are putative zinc ligands, and mutational alteration of these residues inactivates IDE and Axl1p (9,15,16). Some members of the M16A subfamily, including IDE, are sensitive to thiol modifiers, suggesting that a cysteine residue(s) is critical for proper structure of the active site or the overall tertiary or quaternary structure of these proteins (1,15). Sequence alignment of IDE and its yeast orthologs reveals two conserved cysteine residues. One is within the core metalloprotease motif, and this residue can reportedly be altered without affecting enzymatic activity (15). The other conserved * This work was supported in part by funds from the University of Georgia. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
§ Supported by a University of Georgia graduate school assistantship.
cysteine is exactly 67 amino acids distal to the extended motif (HXXEHX 69 EX 6 EX 67 C). This distal cysteine is not conserved in Escherichia coli Protease III, a bacterial M16A enzyme that is insensitive to thiol modifiers (17).
The sequence similarity between IDE, Ste23p, and Axl1p has prompted us to investigate the hypothesis that these proteins have conserved enzymatic properties and substrate specificity. To this end, we present evidence indicating that rat IDE can substitute for Axl1p and Ste23p in a-factor production but not for the other known functions of Axl1p. More importantly, these findings establish yeast as a genetically tractable system for future studies of IDE function. We have made practical use of this system to create and characterize novel mutations that alter the enzymatic properties of IDE.

EXPERIMENTAL PROCEDURES
Strains and Media-The yeast strains used in this study are listed in Table I. Plasmid-bearing versions of these strains were generated by transformation with the indicated plasmids according to published methods (18). Strains were routinely grown at 30°C on synthetic complete dropout (SCϪ) media, as described previously (19).
Plasmids-The plasmids used in this study are indicated in Table II. Plasmids p80 (CEN URA3 AXL1) and p137 (CEN URA3 STE23) have been described previously and were kindly provided by Dr. C. Boone (University of Toronto) (9). The other plasmids used in this study were created by PCR-directed recombination-mediated plasmid construction and/or standard molecular methods (20).
The general strategy for the epitope tagging of AXL1, STE23, and IDE encoding plasmids involved the creation of a novel restriction site at the end of the respective open reading frame and subcloning of a triply iterated HA tag. pWS371 (CEN URA3 AXL1-2HA) was created by modifying p80 as follows. An NcoI site was inserted immediately prior to the stop codon of AXL1 into which a DNA fragment encoding a triply iterated HA tag was inserted; in actuality, two tandem copies of the tag were inserted. The plasmid was further modified to delete a portion of the 5Ј-untranslated region corresponding to an XhoI fragment in order to reduce the overall size of this plasmid from 12.1 to 10.6 kb; this deletion had no effect on protein expression. pWS482 (CEN URA3 STE23-2HA) was created by modifying p137 in a similar manner to that described for p80. In this instance, a BamHI site was introduced immediately prior to the stop codon of STE23 into which a triply iterated HA tag was inserted; as with pWS371, two tandem copies of the tag were actually inserted. pWS511 (2 URA3 P PGK -IDE-HA) was created by PCR amplification of IDE from pSR␣-rat IDE such that the PCR product had ends homologous to the parent yeast vector (pSM703) and a BglII restriction site immediately before the stop codon of IDE (21). The BglII site was used for insertion of the triply iterated HA epitope tag. pWS496 is identical to pWS511 except that it contains two copies of the triply iterated HA tag. Restriction digest analysis and DNA sequencing were used to confirm the presence of the tag in each of the plasmids described above.
The general strategy for the creation of site-directed mutations in the above plasmids also involved PCR-directed recombination-mediated plasmid construction. In brief, a DNA fragment encoding an appropriate segment of the target open reading frame was amplified by PCR where one of the oligonucleotides contained the mutation of interest and typically a silent restriction site. The DNA fragments were cotransformed with a target plasmid that had been linearized or gapped with restriction enzymes near the intended site of mutation. Recombinant plasmids were recovered from yeast, amplified in E. coli, and screened by restriction analysis and/or sequencing to verify the presence of the mutation. All mutant alleles were confirmed by immunoblot to encode the full-length protein.
The epitope tagging of Ste23p was not straightforward. DNA sequencing of an intermediate plasmid in the construction pWS482 revealed an extra nucleotide near the 3Ј end of the open reading frame that was not present in the public sequence of STE23. The presence of the extra nucleotide was confirmed in the STE23 sequence derived from three different sources: p137 and PCR products derived from chromosomal amplification of the STE23 gene from two distinct strain backgrounds (IH1783 and YPH499). An independent group (22) has proposed recently the identical sequence annotation as part of an effort to identify mistakes in published genome sequences. The extra nucleotide alters and extends the translation of the STE23 open reading frame such that amino acids 971-988 now code for 18 distinct amino acids and an additional 39 amino acids are gained as an extension to the original proposed translation.
Protein Extract Preparations-Whole cell protein extracts for Western analysis were prepared by the NaOH/trichloroacetic acid method (23). In brief, mid-log cells (an amount equivalent to 2 ml of a 1.0 A 600 culture) were harvested by centrifugation, washed once with cold water, resuspended in 1 ml of cold water, and treated with a solution of 2 N NaOH, 1 M ␤-mercaptoethanol. Proteins released were precipitated with trichloroacetic acid (11.5% final), recovered by centrifugation, resuspended in Urea Sample Buffer (USB; 250 mM Tris, pH 8.0, 6 M urea, 4% SDS, and 0.01% bromphenol blue), heated, and cleared of insoluble material before analysis by SDS-PAGE and immunoblotting. HAtagged proteins were detected by chemiluminescence (ECL kit, Roche Applied Science) after immunodecorating proteins with mouse anti-HA and horseradish peroxidase-conjugated rabbit anti-mouse antibodies.
Particulate fractions used for in vitro assays were prepared by mechanical breakage of yeast cells expressing either IDE, Ste23p, or Axl1p (24,25). In brief, mid-log cells were harvested, washed with cold 10 mM NaN 3 , and treated with Tris/dithiothreitol (100 mM Tris, pH 9.4, 10 mM dithiothreitol) for 10 min on ice. Cells were resuspended in Oxalyticase Buffer (50 mM KP i , pH 7.5, 1.4 M sorbitol, and 10 mM NaN 3 ) containing Oxalyticase (1 g/A 600 ; Enzogenetics, Corvallis, OR) and incubated for 30 min at 30°C with gentle mixing. The spheroplasts were harvested by centrifugation, resuspended in cold Lysis Buffer (50 mM Tris, pH 7.5, 0.2 M sorbitol, and 1 mM EDTA) containing protease inhibitors (1 g/ml each leupeptin, chymostatin, pepstatin, aprotinin, and 1 mM phenylmethylsulfonyl fluoride), and lysed using a glass Dounce homogenizer. The primary lysate was cleared twice of cell debris by centrifugation (500 ϫ g, 10 min) and fractionated into supernatant and particulate fractions by centrifugation (16,000 ϫ g, 10 min). The fractions were adjusted to 1 mg/ml with Lysis Buffer and stored at Ϫ80°C as aliquots.
To assess the effect of chaotropic agents on the association of IDE, Ste23p, and Axl1p with the yeast particulate fraction, freshly prepared samples of the particulate fraction were exposed to either 1 M NaCl, 0.1 M Na 2 CO 3 (pH 11.5), 1% SDS, or buffer alone for 10 min on ice, and the samples were subjected to centrifugation (16,000 ϫ g, 10 min). Equivalent portions of each supernatant and particulate fraction were analyzed by SDS-PAGE and immunoblot as described above.
Insulin Degradation Assay-Insulin degradation assays were carried out essentially as described previously but using yeast-derived lysates as the source of enzyme activity (4). In brief, reactions were assembled to contain 0.5 mg/ml yeast lysate and 60 pM (ϳ10,000 cpm) of 125   and centrifugation (16,000 ϫ g, 10 min) at 4°C. The supernatant containing insulin fragments was transferred to a new tube, and the radioactivity associated with the supernatant and particulate fractions was determined using a Wallac ␥-counter (PerkinElmer Life Sciences). Yeast Mating Assay-To evaluate the ability of plasmid-transformed MATa axl1⌬ ste23⌬ strains to promote a-factor production, patch mating tests were performed using IH1793 (MAT␣ lys1) and established methods (26). This test provides an indirect assessment of M16A enzyme function because a-factor production in the MATa cells is entirely dependent on the plasmid-encoded copy of the M16A enzyme. In brief, master plates were prepared by patching MATa yeast strains expressing Axl1p, Ste23p, or IDE onto YEPD agar plates. After 2-3 days of growth, the patches were replica-printed onto lawns of MAT␣ cells (IH1793) spread on minimal media agar plates, and the replica-printed plates were incubated for 2 days at 30°C. The MAT␣ cell suspensions were prepared to approximately the same cell density in solutions of 1, 10, or 100% YEPD, which were prepared as mixtures of YEPD and the appropriate amount of sterile H 2 O. The trace amount of YEPD added to the minimal media allows for limited survival of the auxotrophic haploid cells, which is necessary for the mating process. Decreasing the YEPD amount shortens the survival window of the haploid cells and thus results in decreased mating efficiency. The diploid cells that result from mating events are prototrophic, and thus the growth of diploids in this test is indicative of mating and a functional M16A enzyme.
Mass Spectroscopy-The a-factor mating pheromone produced by yeast expressing IDE, Ste23p, or Axl1p was purified according to published methods and analyzed by mass spectroscopy (26). In brief, yeast were cultured in polypropylene culture tubes, and the a-factor secreted from the yeast cultures was recovered by washing the polypropylene culture tubes with methanol; secreted a-factor adsorbs to polypropylene and can be removed with organic solvents. The enriched a-factor samples were concentrated by speed-vac, desalted using Zip-tip C18 beads (Millipore), washed three times with 0.1% trifluoroacetic acid, eluted with 70% acetonitrile, and subjected to MALDI-TOF/TOF mass spectroscopy using ␣-cyano-4-hydroxycinnamic acid matrix and a 4700 Proteomics Analyzer spectrometer (Applied Biosystems, Foster City, CA). a-Factor Halo Assay-The total secreted a-factor produced by the saturated cultures of indicated strains was recovered as described above. The enriched samples were dried by a speed-vac and resuspended in 50 l of MeOH. 2-Fold serial dilutions of the samples prepared in YEPD were spotted onto a lawn of RC757 (MAT␣ sst2-1 his6 met1 can1 cyh2) or RC631 (MATa sst2-1 his6 met1 can1 cyh2 rme ade2-1 ura1) cells; the latter served as a control for the unlikely possibility that a toxic product was being produced by the IDE-expressing strain (27). The formation of a spot in the lawn is indicative of the presence of pheromone in the sample, and the relative potency of pheromone can be determined from the serial dilutions of the sample. The highest dilution having biological activity is referred to as the end point and is equivalent to a concentration of 12 pg/l of a-factor (28).
Invasive Growth Assay-The ability of haploid yeast cells to invade yeast agar was determined using a plate-washing assay (14). In brief, cell suspensions were spotted onto SC-ura plates, and the cells were grown for 4 days at 30°C. The plates were washed under running H 2 O while gently rubbing the surface of the agar plate with a gloved finger.  Agar plates were scanned prior to and immediately after washing.
Budding Assay-Mid-log cells grown in selective liquid media were harvested to concentrate the cell suspension ϳ2 times, and the cells were treated with 10 g/ml Calcofluor (Sigma) for 5 min. The cells were washed twice with H 2 O, absorbed onto polylysine-coated glass slides, and viewed at ϫ100 with a Zeiss Axioplan microscope equipped with fluorescence optics. Fluorescent images were captured using a digital camera (Optronics, DEI-750). Bud scars on 100 cells were evaluated and categorized into three patterns: axial, bipolar, and random. The axial pattern was defined as bud scars located solely at one pole of the cell; the bipolar pattern had scars at both poles, and the random pattern displayed at least one bud scar in the region between both poles (8,9).

IDE Can Be Heterologously Expressed in
Yeast-Many of the yeast enzymes required for a-factor production can be functionally replaced with orthologs from other eukaryotic species (29 -34). Thus, we hypothesized that IDE could rescue the mating defect of a strain lacking Axl1p and Ste23p. Prior to evaluating the ability of IDE to substitute for the functions of the yeast M16A enzymes, we first determined whether yeast could be used for the heterologous expression of IDE. We created plasmids encoding epitope-tagged versions of the genes encoding IDE and its yeast orthologs. For all the proteases, a DNA fragment encoding the HA epitope tag was placed at the 3Ј end of each gene. The addition of this tag did not alter the activities of these enzymes (see Fig. 3B).
Evaluation of protein extracts from strains expressing the tagged proteases revealed that each tagged protein could be detected as a protein of the expected size by immunoblot (Fig.  1A). We also observed that IDE and Ste23p were expressed at ϳ10 times higher levels than Axl1p as determined by comparison of immunoblot signals. The constitutive phosphoglycerate kinase (PGK) promoter was used to drive IDE expression, so the abundant expression of IDE was not unexpected. Native promoters were used to drive Ste23p and Axl1p expression, and the relative abundance of Ste23p over Axl1p was somewhat unexpected, especially because Ste23p does not promote a-factor production as efficiently as Axl1p. Thus, the reason for the decreased ability of Ste23p to promote mating cannot be simply attributed to reduced protein expression by comparison to Axl1p.
IDE Is Associated with a Particulate Yeast Fraction-The subcellular distribution of IDE in mammalian systems is reported to be primarily cytosolic, with extracellular and peroxisomal localizations also being described (35,36). To ascertain the effect of heterologous expression on the subcellular distribution of IDE, we subjected a total yeast lysate to differential fractionation. By comparison of equal amounts of loaded protein by immunoblot, we observed that IDE and the yeast M16A enzymes were highly enriched in the particulate fraction associated with 16,000 ϫ g centrifugation of the lysate (P16), although a significant amount of Ste23p was found in the supernatant fraction (Fig. 1B). In this particular experiment, Axl1p was encoded on a multicopy plasmid to facilitate its detection, hence its stronger signal relative to that observed in total extracts (see Fig. 1A).
Although the localization of Ste23p has not been reported previously, Axl1p is known to transiently associate with components that form a sub-plasma membrane complex that is required for establishing bud sites. Whether IDE or Ste23p can assemble into this complex is unknown. To better understand the nature of association that IDE, Axl1p, and Ste23p have with the P16 yeast fraction, we performed extractions of the particulate fraction with various chaotropic agents (Fig. 1C). Our analysis revealed that high pH and detergent treatments significantly disrupted the association of these proteins with the P16 fraction, whereas salt and buffer alone had a minor impact. This profile is consistent with IDE, Ste23p, and Ax1p being peripheral membrane proteins and/or components of a large macromolecular complex that sediments under our experimental conditions.
Yeast-expressed IDE Retains Proteolytic Activity-We next determined whether yeast-expressed IDE could cleave insulin, a well characterized substrate of IDE, using a previously described insulin degradation assay and yeast-derived particulate fractions (4,25). Because of the high degree of enrichment for M16A enzymes in yeast particulate fractions, these samples were used exclusively as the source of activity for these in vitro assays. Our analysis of enzymatic activity indicated that the IDE-containing samples had insulin cleaving activity, whereas samples containing Axl1p or Ste23p had no more activity than a sample prepared from yeast lacking these enzymes (Fig. 2A). The source of the residual activity in these preparations is unknown but is likely due to a nonspecific enzymatic activity rather than that of an additional IDE homolog because 1,10-phenanthroline, a well documented IDE inhibitor, does not inhibit the residual activity (Fig. 2B). Whether the Axl1p and Ste23p proteases in these preparations are incapable of cleaving insulin because of altered substrate specificity, sub-optimal reaction conditions, or other reasons has not yet been determined.
IDE is described as a thiol-and zinc-dependent metalloprotease (35). Having determined that IDE can be heterologously expressed in a functional form, we next wanted to confirm that the observed insulin degrading activity detected had the hallmarks of IDE-mediated insulin degradation. We examined the effect of alkylating (i.e. iodoacetamide and N-ethylmaleimide) and metal ion-chelating agents (i.e. 1,10-phenanthroline and EDTA) on IDE activity (Fig. 2B). Alkylating agents inhibited IDE-dependent insulin degrading activity. An inhibitory effect was also observed for 1,10-phenanthroline but not with the nonchelating agent 4,7-phenanthroline that is structurally similar. EDTA did not measurably inhibit enzymatic activity; extensive pretreatment with EDTA is reportedly required for inhibition of IDE (15). The inhibitor profile observed for insulin degrading activity in yeast was identical to that reported for IDE, which further establishes that yeast can synthesize IDE possessing all the enzymatic properties described for that of IDE found in metazoans (35).
IDE Can Promote Yeast Mating-Many of the yeast enzymes required for a-factor production can be functionally replaced by orthologous enzymes from other species (29 -34). In similar fashion, we have now determined that IDE expression can rescue the mating defect of a yeast strain lacking both AXL1 and STE23 (MATa axl1⌬ ste23⌬) (Fig. 3A, top row). The mating defect in this strain is due to an inability to produce yeast a-factor, but whether Axl1p and Ste23p, and by extension IDE, participate directly or indirectly in pheromone production has not been rigorously established (9). Nevertheless, this result establishes that IDE and the yeast enzymes Axl1p and Ste23p can similarly promote yeast a-factor production and thus have an evolutionarily conserved activity. The simplest explanation for our findings is that these enzymes directly participate in pheromone production and that these enzymes have shared substrate specificity with respect to cleavage of the a-factor precursor.
Yeast mating is sensitive to a number of variables, including available nutrients. Under depleted nutrient conditions, mating is less efficient. We have taken advantage of this property of yeast mating to evaluate the effectiveness of IDE in rescuing the mating defect of yeast lacking both Ste23p and Axl1p. As expected, we observed that decreasing nutrient conditions correlated with decreased mating for all strains. Ste23p and IDEexpressing strains showed significant reductions in mating relative to the Axl1p-expressing strain (Fig. 3A). Epitopetagged versions of these enzymes had a similar activity profile in this assay (Fig. 3B). The relatively poor ability of Ste23p to promote mating, with respect to Axl1p, is consistent with pre-vious reports on the properties of this enzyme and may be attributable to a number of factors, including altered substrate specificity and/or an altered subcellular localization pattern that prevents interaction with substrates (9). Likewise, the reduced mating observed for IDE may be attributable to similar factors. The reduced mating observed for IDE and Ste23pexpressing cells is not simply due to low expression of these enzymes because both are expressed at significantly higher levels than Axl1p (see Fig. 1).
The most straightforward explanation for the reduced mat-

FIG. 2. IDE can be functionally expressed in yeast.
A, particulate fractions were isolated from yeast expressing IDE, Axl1p, or Ste23p. An equivalent amount of each fraction (5 g) was assayed for insulin degradation activity during a time course according to established methods (see "Experimental Procedures"). The degradation of insulin is indicated by the presence of recovered insulin fragments that cannot be precipitated by trichloroacetic acid. B, the effect of various agents on the activity of yeast-expressed IDE was evaluated. Activities were determined as in A, but only for the t ϭ 0-min and t ϭ 120-min time points. The abbreviations used are as follows: NEM, 1 mM Nethylmaleimide; IAA, 1 mM iodoacetamide; 1,10, 1 mM 1,10-phenanthroline; 4,7, 1 mM 4,7-phenanthroline; and EDTA, 10 mM EDTA.

FIG. 3. Yeast-expressed IDE promotes yeast mating and a-factor pheromone production.
A, the mating competence of MATa ste23⌬ axl1⌬ cells transformed with the indicated M16A enzyme encoded on a plasmid was evaluated using a patch mating test. The selective growth of diploid cells on the minimal media is indicative of mating. Patch mating tests were conducted using MAT␣ lys1 lawns containing various amounts of nutrients (100, 10, and 1% YEPD; top to bottom). Lowering the amount of YEPD in the lawn increases the stringency of the mating test and allows for discrimination of differences in mating efficiency not otherwise observable under permissive mating conditions (100% YEPD). B, patch mating tests were conducted as in A but using MATa strains expressing HA-epitope tagged M16A proteases under mildly stringent mating conditions (10% YEPD). C, the a-factor produced by each of the strains described in B was recovered from the walls of culture tubes using an organic solvent, and concentrated samples were analyzed by a spot halo test. Each strain was grown to saturation in the same volume of media. The formation of a spot in the lawn of MAT␣ (RC757) cells is indicative of the presence of pheromone in the sample, and the relative pheromone potency can be determined by serial dilution of the sample. The highest dilution having activity is referred to as the end point and is equivalent to a concentration of 12 pg/l of a-factor (28). D, 2-fold serial dilutions of the IDE sample prepared in YEPD were spotted onto a lawn of RC631 (MATa sst2-1 his6 met1 can1 cyh2 rme ade2-1 ura1) serving as a control for the possibility that a toxic product was being produced by the IDE-expressing strain (27). The strains used were IH1783 transformed with pRS316 (wild type (WT)), and Y272 transformed with pRS316 (vector), p80 (AXL1), p137 (STE23), pWS491 (IDE), pWS371 (AXL1-2HA), pWS482 (STE23-2HA), or pWS496 (IDE-2HA).
ing observed with Ste23p and IDE-expressing strains is that these strains have reduced a-factor production by comparison with the Axl1p-expressing strains. Thus, we compared the relative amounts of a-factor produced by these strains over the lifetime of a culture. We used a bioassay that relies on the natural growth-arrest response of MAT␣ cells to the a-factor pheromone to detect and quantifiably measure a-factor activity in our samples. Our analysis revealed that both IDE and Ste23p-expressing strains produce significantly less a-factor than an Axl1p-expressing strain, despite the fact that IDE and Ste23p are significantly overexpressed relative to Axl1p (Fig.  3C). The IDE-derived sample was also bioassayed using MATa cells. These cells fail to undergo growth arrest in response to the a-factor pheromone and thus served as a control for the unlikely scenario that IDE-expressing strains were producing a toxic secreted product. No growth inhibition of MATa cells was observed with the IDE-derived sample (Fig. 3D).
The a-factor mating pheromone is an isoprenylated and carboxymethylated dodecapeptide (YIIKGVFWDPAC[farnesyl]methyl). Alterations to the chemical form of a-factor (e.g. lipid removal or primary sequence alterations) can impact the function of this pheromone (37). For example, a-factor lacking a tyrosine residue has 25% of the biological activity of full-length pheromone as judged by yeast mating tests (37). Thus, the reduced production of a-factor pheromone by IDE and Ste23pexpressing yeast could be attributable to cleavage of the a-factor precursor at a site other than the Asn-Tyr cleavage site recognized by Axl1p. To determine whether the pheromone products produced by IDE, Ste23p, and Axl1p were identical, we determined the mass of the a-factor species generated by these enzymes using MALDI-TOF/TOF mass spectroscopy. Several major species were detected in these samples, including a 1629-Da peak that corresponds exactly to the mass of bona fide a-factor (Fig. 4, A-C; IDE, Ste23p, and Axl1p samples, respectively). This species was not in the negative control (Fig. 4D). The 1629-Da species was still observed in the samples after the data were de-isotoped (Fig. 4, E-G) and remained absent in the negative control (Fig.  4H). Based on these observations, we reasoned that the 1629-Da species is indeed a-factor, which implies that IDE, Ste23p and Axl1p have similar cleavage specificities.
Although no other species were apparent in the de-isotoped data for the IDE and Ste23p-derived samples in the indicated mass range, additional species were evident in the other samples. In the Axl1p-derived sample, three additional species were observed (1480, 1526, and 1718 Da) (Fig. 4G). One distinct species was observed in the negative control (1587 Da). These additional species were not reproducibly detected between experiments. 2 The mass range shown in all panels is inclusive of the masses of the theoretical MFA1-derived a-factor species either lacking an N-terminal tyrosine (1467 Da) or extended by an N-terminal asparagine (1744 Da), but none of the four unidentified species can be matched to these alternative a-factor cleavage products. The nature of these other species is therefore unknown.
Genetic Evaluation of IDE Mutants-The ability of yeastexpressed IDE to promote a-factor production suggests that yeast mating can be used as a phenotype to evaluate the function of IDE mutant alleles. To test this hypothesis, we created site-directed and deletion mutations of IDE and tested the ability of these mutants to promote mating. Using our genetic system, we confirmed that residues comprising the core metalloprotease motif (HXXEH) of IDE were essential for promoting mating (Fig. 5A). These residues were required for the activity of Ste23p and Axl1p as well (Fig. 5, B and C, respectively). The most distal glutamate of the extended motif (HXXEHX 69 EX 6 E) was also determined to be essential for the activity of all three enzymes, whereas the penultimate glutamate was dispensable for Axl1p activity. The penultimate glutamate is required for the activity of mitochondrial processing peptidase, an M16B protease, and has been proposed to aid in metal coordination of this enzyme (38).
In addition to addressing the importance of established active site residues, we investigated the functional importance of cysteine residues in the function of IDE. We initially investigated the role of two cysteine residues that are invariably conserved between IDE, Ste23p, and Axl1p (HXCEHX 69 EX 6 EX 67 C), suspecting that one of these residues is the likely target of sulfhydrylmodifying agents that inactivate certain M16A enzymes. Independent mutations at these residues did not alter the abilities of IDE, Ste23p, or Axl1p to promote mating (Fig. 5). Mutation of the proximal cysteine of IDE (C110A) reportedly does not alter the thiol sensitivity profile of IDE (15), and similarly, we found that mutation of the distal cysteine of IDE (C257A) had no effect on the thiol sensitivity of IDE. 2 We also determined that combining the mutations in one molecule (C110A/C257A) did not alter mating function (Fig. 5A). Moreover, the sensitivity of IDE to thiol modifiers was unaffected. 2 Similar results were observed when other cysteine residues that are conserved between IDE and Ste23p or Axl1p (C819A and C414A, respectively) were mutated (Fig. 5A).
M16A enzymes are large proteins, typically having a molecular mass in excess of 100,000 Da. The metalloprotease motif of M16A enzymes is localized near the N-terminal end of these enzymes. In the absence of structural data for these enzymes, we sought to determine whether the catalytic domain of IDE was self-contained within the N-terminal portion of the enzyme. C-terminally truncated forms of IDE were created and evaluated for the ability to promote pheromone production. These truncations were created as fusions to GFP, which in and of itself did not alter the ability of IDE to function ( Fig. 6A; IDE 1-1019). Systematic deletions of the C terminus of the IDE revealed that relatively short truncations inactivated IDE as judged by yeast mating tests. Deletion analysis of Axl1p revealed a similar requirement for an intact C terminus, although the deletion required for Axl1p inactivation was considerably larger (Fig. 6A, Axl1p 1-1076). Site-directed mutation of a residue conserved between IDE and Axl1p that was at the functional/nonfunctional truncation boundary of Axl1p (i.e. Ser-965 in IDE and Ser-1081 in Axl1p) did not reveal an essential requirement for this residue in either enzyme (Fig. 5).
For the site-directed and truncation mutants created for this study, we evaluated protein expression. By using immunoblots, the site-directed mutants were judged to be expressed as well as the relevant wild type control, except for Ste23p E192A that was expressed at a level approximately one-quarter that of wild type. 2 With the exception of this mutant, our results rule out the trivial possibility that inactive IDE, Axl1p, and Ste23p mutants are nonfunctional because of poor expression. We cannot exclude the possibility that these mutants are nonfunctional because of malfolding or other unknown reasons. Evaluation of the Axl1p truncation mutants revealed that these were expressed at levels similar to that of the full-length Axl1p-GFP fusion protein. By contrast, the IDE truncation mutants were not similarly expressed (Fig. 6B). We observed decreased expression for certain truncations, suggesting that the C terminus of IDE, for reasons unknown, is required for normal steady state expression. Decreased expression alone cannot account for the loss of activity for IDE, because one nonfunctional truncation (IDE-(1-973)) was expressed comparably to the full-length IDE-GFP fusion. More importantly, the expression levels observed for all the nonfunctional truncations of Axl1p and IDE, which were encoded on multicopy plasmids, were significantly more than that of the respective full-length fusion encoded on a low copy plasmid, as judged by immunoblot analysis, suggesting that lowered expression alone cannot account for the absence of function in these mutants.
Yeast-expressed IDE Cannot Substitute for Other Known Functions of Axl1p-Axl1p is involved in regulating several cellular processes besides pheromone production. For example, Axl1p is required for repressing haploid invasive growth (14). Haploid invasion occurs more readily upon the deletion of AXL1, and this phenotype can be fully reversed by the introduction of a plasmid-borne copy of AXL1. In order to address the ability of IDE to substitute for this other function of Axl1p, the invasive phenotype of an Axl1p-deficient strain (axl1) that expressed IDE was examined. We also investigated the ability of a protease active site mutant of Axl1p to suppress invasive growth. We found that IDE expression could not repress invasive growth, whereas the Axl1p active site mutant suppressed invasive growth (Fig. 7).
Axl1p also has a well described role in the maintenance of the axial budding pattern of haploid yeast (9). Ste23p does not reportedly have a role in this process. To determine whether IDE could substitute for the function of Axl1p in axial bud site selection, we expressed IDE in a haploid strain defective for axial budding, and we evaluated the budding pattern by Calcofluor staining of bud scars. Consistent with previous findings, we observed a bipolar budding phenotype in the absence FIG. 4. Mass spectroscopic analysis of a-factor produced by IDE, Ste23p, and Axl1p-expressing yeast strains. The a-factor species secreted by the indicated MATa strains were enriched from conditioned media as described in Fig. 3C, and samples were subjected to MALDI-TOF/TOF mass spectroscopy using ␣-cyano-4-hydroxycinnamic acid as the matrix. The strains used were Y272 transformed with pWS496 (IDE-2HA), pWS482 (STE23-2HA), pWS372 (AXL1-2HA), or pRS316 (vector) (A-D, respectively). De-isotoping of the data shown in A-D resulted in the data presented in E-H, respectively. All strains also contained pSM463 (2 TRP1 MFA1).
of Axl1p expression (Table III; vector). An axial budding phenotype was observed upon the introduction of a plasmid encoding Axl1p into the axl1 strain (AXL1), whereas bipolar budding was the predominant pattern when either STE23 or IDE was introduced.

DISCUSSION
The majority of studies on IDE has relied on in vitro biochemical assays (1). More recently, knock-out and overexpression mouse models have been described for IDE (4 -6). Both in vitro and in vivo model systems have led to a better understanding of the biochemical and physiological properties of IDE, which has proposed roles in type 2 diabetes and Alzheimer disease (4, 39 -41).
In this study, we have developed yeast as a tractable genetic model system for studying the functional properties of IDE. We have determined that yeast can be used to express a functional form of IDE and, more importantly, that IDE can promote a-factor production, an activity normally associated with the yeast enzymes Axl1p and Ste23p that have homology to IDE. These results imply that members of the M16A metalloprotease family, to which these enzymes belong, may have shared substrate specificity. The fact that all three enzymes can promote a-factor production supports this hypothesis. However, our inability to demonstrate insulin cleavage by the yeast enzymes implies that the substrate specificity of Ste23p and Axl1p may be more restricted. Whether the yeast enzymes cleave other established IDE substrates is currently under investigation. We are also investigating the ability of other M16A enzymes to cleave a-factor, because the studies of these enzymes may benefit from the development of a tractable genetic system.
An additional finding from our study is that IDE is enriched FIG. 6. Evaluation of GFP-labeled IDE and Axl1p truncation mutants. A, full-length and C-terminally truncated versions of IDE and Axl1p were created as fusions to GFP, and the activities of these fusions were evaluated by patch mating tests under permissive conditions (100% YEPD). The truncations exhibiting robust mating were as active as the full-length fusion as judged by mating tests under stringent conditions, except for Axl1-(1-1081) that had a reduced mating phenotype. The amino acids to which GFP was fused are indicated by the arrows and are shaded in gray. The residues within the C-terminal regions of IDE and Axl1p that are conserved are shown in the protein alignment below the mating tests (arrowheads). The residue in this region that was targeted for site-directed mutational analysis (see but does not exclusively partition into a yeast particulate fraction. We suspect that heterologously expressed IDE is either being incorporated into membranes as a loosely associated peripheral membrane protein or it is assembling into a macromolecular complex that partially sediments under the conditions used for isolation of the particulate fraction. Although IDE reportedly associates with peroxisomes, this localization is unlikely under our experimental conditions because we did not impose an induction that is required for the formation of yeast peroxisomes (42). Moreover, the putative C-terminal peroxisomal targeting signal found in IDE, although potentially functional in yeast, is blocked by an epitope tag in our yeastexpressed enzyme. Alternatively, IDE may be associating with other undefined membrane sites. Axl1p has been reported to associate transiently with components that form a sub-plasma membrane complex that is required for budding (43). Thus, it is also conceivable that IDE is assembling into this complex, but as a nonfunctional component. Ultimately, defining the subcellular localization Ste23p may provide insight into the subcellular tar-geting of IDE because these two enzymes appear more conserved in sequence and enzymatic properties than IDE and Axl1p. The genetic model system that we have developed should have far reaching utility for the characterization of mutant IDE alleles. We have provided a practical example of this utility by detailing the functional importance of residues that compose the extended metalloprotease motif (HXXEHX 69 EX 6 E). In addition, our mutational study allows us to make certain conclusions about the cysteine residue(s) that imparts the sensitivity of IDE to thiol modifiers. Provided that a single cysteine residue is the target of thiol modification in IDE, our study excludes as targets the two cysteine residues that are invariably conserved between all three enzymes (Cys-110 and Cys-257 in IDE) and two cysteines that are semi-conserved between IDE and the yeast M16A enzymes (Cys-414 and Cys-819 in IDE). The double C110A/C257A mutant also remains sensitive to thiol modifiers. We are currently developing an in vitro assay for monitoring the activities of Ste23p and Axl1p. We expect a determination of whether these enzymes are sensitive to thiol  bottom, respectively). The schematic has been drawn to scale. Each representation has been divided into several domains. The borders of domains I-VI are operationally defined as sites where insertions (Ͼ10 residues) are found in the Axl1p sequence as determined by multiple sequence alignment (ClustalW) of all four proteins. Domains at the ends of the molecules (dark gray) represent sequences that are of variable length and are not conserved between the enzymes. The N-terminal sequences range between 19 and 69 residues, and the C-terminal sequences range between 35 and 127 residues. The relative position of the shortest C-terminal truncation yielding a nonfunctional mutant is shown (filled arrowhead). The values below the schematic represent the percent identity between the indicated IDE domain and the corresponding domains of either Ste23p, E. coli Protease III, or Axl1p as determined by using DNA Strider 1.3 at default settings. B-D, several conserved motifs can be identified within domains I (B), II (C), and V (D). For the purposes of this figure, a conserved motif was defined as a block of amino acids (Ն10) having Ն50% identical or highly conserved residues (i.e. S/T, K/R, or A/I/V). The second motif listed for domain I is found within the X 69 sequence of the extended metalloprotease motif, whereas the third is found within the X 57 sequence. The residues in boldface were mutated in this study. Neither Cys-257, which is found near the end of domain I, nor Cys-414, which is found at the beginning of domain IV, are within conserved motifs. modifiers to aid in the final identification of cysteine residues that are targeted by thiol modifiers in this enzyme family. As a second practical example of the utility of our genetic system, we demonstrate for the first time that a C-terminal region is required for the activity of IDE. In the absence of structural information for IDE, we suspect that its C-terminal region may be required for stabilizing the overall tertiary or possibly quaternary structure of IDE or its active site.
Our genetic system will also be useful for investigating the role of other residues in the function of IDE. For example, several conserved sequence motifs can be identified by multiple sequence alignment of M16A enzymes (Fig. 8). Our future studies will be aimed at determining whether these motifs are important for the function of M16A enzymes. In addition, we will investigate whether E. coli Protease III can support a-factor production in yeast. If so, our system could conceivably be used to investigate the structure/function relationships for this enigmatic protease, which also lacks a defined cellular role (44).
Another potential utility of our genetic model system is the theoretical ability to rapidly identify IDE mutants having altered substrate specificity. We have already demonstrated that our system is amenable to screening specific mutant IDE alleles for the ability to promote mating. The same approach could be used to identify IDE mutants having altered specificity toward the a-factor. We envision a positive genetic selection that takes advantage of the observations that IDE-dependent mating is essentially nonexistent at reduced nutrient levels (Fig. 3A). Under these highly stringent mating conditions, yeast harboring IDE mutants could be screened with the expectation that those having enhanced a-factor production would now be matingcompetent. These IDE mutants would represent candidates having improved a-factor recognition (i.e. altered substrate specificity). Conceivably, these mutants might also have altered specificity for insulin and/or A␤. IDE mutants having enhanced activity toward these substrates could potentially be used as therapeutic agents for diabetes or Alzheimer disease.
Our data clearly demonstrate that the Axl1p/Ste23p-dependent step in a-factor production can be supported by IDE. This result, in and of itself, is not surprising given the high degree of similarity between these enzymes. Of greater curiosity is the observation that all the yeast enzymes required for a-factor production can be functionally replaced by their mammalian counterparts. The first three steps associated with a-factor production (i.e. isoprenylation, CaaX proteolysis, and carboxymethylation) are part of the biosynthetic pathway of isoprenylated proteins, such as Ras and RhoB. The subsequent two proteolytic steps associated with a-factor production have no analogous counterparts in mammals. Thus, the ability of mammalian enzymes to promote these latter steps is entirely serendipitous, or it reflects the existence of an orphan biosynthetic pathway that integrates all the steps associated with a-factor biosynthesis. If the latter, by analogy, we suspect that the molecule produced by this pathway would serve as a secreted signaling molecule, possibly functioning to regulate cellcell fusion events such as those that occur during the development of certain tissues. Curiously, the morphological and biochemical differentiation of L 6 myoblasts into myotubes has been reported to involve IDE (45,46). Thus, by analogy to a-factor production, we hypothesize that a role of IDE in these cells might be to produce a signaling molecule derived from a precursor having an aliphatic amino acid motif and an N-terminal extension that is readily cleaved by IDE. Whether such a molecule exists in mammalian systems remains to be determined.