The Saccharomyces cerevisiae PCD1 Gene Encodes a Peroxisomal Nudix Hydrolase Active toward Coenzyme A and Its Derivatives*

The PCD1 nudix hydrolase gene ofSaccharomyces cerevisiae has been cloned and the Pcd1p protein characterized as a diphosphatase (pyrophosphatase) with specificity for coenzyme A and CoA derivatives. Oxidized CoA disulfide is preferred over CoA as a substrate with K m andk cat values of 24 μm and 5.0 s− 1, respectively, compared with values for CoA of 280 μm and 4.6 s− 1respectively. The products of CoA hydrolysis were 3′-phosphoadenosine 5′-monophosphate and 4′-phosphopantetheine. F− ions inhibited the activity with an IC50 of 22 μm. The sequence of Pcd1p contains a potential PTS2 peroxisomal targeting signal. When fused to the N terminus of yeast-enhanced green fluorescent protein, Pcd1p was shown to locate to peroxisomes by confocal microscopy. It was also shown to co-localize with peroxisomal thiolase by immunofluorescence microscopy. N-terminal sequence analysis of the expressed protein revealed the loss of 7 or 8 amino acids, suggesting processing of the proposed PTS2 signal after import. The function of Pcd1p may be to remove potentially toxic oxidized CoA disulfide from peroxisomes in order to maintain the capacity for β-oxidation of fatty acids.

digested with EcoRI and XhoI, and the gel-purified restriction fragment ligated between the EcoRI and XhoI sites of both pET17b(ϩ) and pPGY1. The resulting pET151C construct (from pET17b) yielded an N-terminal fusion of the T7 tag sequence and Pcd1p under the control of the T7 promoter while the pPGY151C construct (from pPGY1) generated the ATG initiator downstream of GAL1p, a galactose-inducible promoter, and yielded native Pcd1p when expressed in yeast. Both plasmids were used to transform E. coli XL1-Blue cells for propagation.
Expression of Pcd1p in E. coli and Preparation of Antibody-E. coli strain BL21(DE3) was transformed with pET151C. A single colony was picked from an LB agar plate containing 60 g/ml ampicillin and inoculated into 10 ml of LB medium containing 60 g/ml ampicillin. After overnight growth, the cells were transferred to 1 liter LB medium containing 60 g/ml ampicillin and grown to an A 600 of 0.6. Isopropyl-1-thio-␤-D-galactopyranoside was added to 0.4 mM and the cells induced for 4 h. The induced cells (4.2 g) were harvested, washed, and resuspended in 50 ml of sonication buffer (50 mM Tris-HCl, pH 8.0, 2 mM EDTA, 0.1 M NaCl). The cell suspension was sonicated and the inclusion bodies recovered by centrifugation at 10,000 ϫ g for 20 min. After washing by resuspension in sonication buffer containing 2.5 M urea, the inclusion bodies were dispersed in 27 ml of 25 mM Tris-HCl, pH 8.0, 8 M urea, 10 mM dithiothreitol (DTT) and the extract centrifuged at 100,000 ϫ g for 1 h. The supernatant was applied in 1-ml aliquots to a Mono Q HR 5/5 anion exchange column (Amersham Pharmacia Biotech) previously equilibrated in Buffer A (25 mM Tris-HCl, pH 8.0, 6 M urea) and the protein eluted with a linear gradient from 0 to 0.5 M NaCl in Buffer A. Homogeneous Pcd1p eluted at 0.27 M NaCl. Fractions containing the protein were dialyzed extensively against phosphate-buffered saline and used to generate a rabbit anti-Pcd1p polyclonal antiserum by standard procedures.
Expression of Pcd1p in S. cerevisiae and Purification-S. cerevisiae strain INVScI was transformed with pPGY151C. A single colony was picked from an SC-Ura (synthetic complete medium without uracil) agar plate and inoculated into 40 ml of SC-Ura medium supplemented with 5% glucose. After 36 h the cells were harvested by centrifugation and resuspended in 4 liters of SC-Ura ϩ 5% glucose and grown for another 24 h. The cells were again recovered by centrifugation, resuspended in 4 liters of SC-Ura ϩ 2% galactose, 1% raffinose, and grown for 24 h to fully induce expression of Pcd1p.
Solid (NH 4 ) 2 SO 4 was added to the pooled fraction to a final concentration of 1 M and the sample was loaded at 1.5 ml/min on to a 15 ϫ 50-mm column of Phenyl-Sepharose CL-4B (Amersham Pharmacia Biotech) previously equilibrated in Buffer C (50 mM Tris-HCl, pH 7.5, 5 mM 2-mercaptoethanol) containing 1 M (NH 4 ) 2 SO 4 . After removal of unbound protein, a 100-ml reverse linear gradient from 1 to 0 M (NH 4 ) 2 SO 4 in Buffer C was applied. Fractions containing Pcd1p were identified by immunoblotting and pooled (52 ml) before dialysis against 10 mM sodium phosphate, pH 6.8, 10 M CaCl 2 .
The dialysate was applied at 1 ml/min to a 100 ϫ 7.8-mm Bio-Gel HPHT column (Bio-Rad) and the protein eluted with a 30-ml linear gradient from 10 to 350 mM sodium phosphate, pH 6.8, containing 10 M CaCl 2 . Homogenous Pcd1p eluted at 300 mM sodium phosphate, and fractions containing the pure protein were dialyzed extensively against 25 mM Tris-HCl, pH 7.5, 50 mM NaCl prior to assay.
Pcd1p-GFP Fusion Constructs and Subcellular Localization-Expression plasmids encoding C-terminal and N-terminal fusions of Pcd1p to yEGFP (11) were constructed by amplification of the PCD1 coding region from genomic DNA using the polymerase chain reaction and either the 36-mer oligonucleotide primers described above to give PCR product C, or the 36-mer forward and reverse primers d(AGAAAAGA-ATTCATGATATTAAGTCAGAGGAGGATG) and d(CAGTTTCTCGAG-CCAAAGCGAGCGGCACTCCAGCAG) to give product N. After amplification with Pfu DNA polymerase, both DNA products were recovered by phenol/chloroform extraction and digested with EcoRI and XhoI. The digested and gel-purified PCR product C was ligated between the EcoRI and XhoI sites of pUG36 to give pyEGFP-PCD1 while PCR product N was ligated between the EcoRI and SalI sites of pUG35 to give pPCD1-yEGFP. Both plasmids were propagated by transformation of E. coli XL1-Blue cells.
For microscopy, S. cerevisiae strain INVScI was transformed with either pyEGFP-PCD1 or pPCD1-yEGFP and grown in liquid or solid SC-Ura medium containing 2% glucose. For staining of mitochondria in living cells, cultures of exponentially growing transformed INVScI were resuspended in 10 mM HEPES, pH 7.4, 5% (w/v) glucose, 100 nM rhodamine B hexyl ester and incubated at 20°C for 30 min. For immunofluorescence microscopy, INVScI cells transformed with pPCD1-yEGFP were first grown in SC-Ura ϩ 2% glucose to mid log phase followed by growth for 18 h in SC-Ura ϩ 0.1% oleic acid, 0.02% Tween 40. Fixation and immunocytochemical staining were as described (12) using anti-3-oxoacyl-CoA thiolase (dilution 1:3000) followed by rhodamine-conjugated goat anti-rabbit IgG (1:50). Cells were viewed by conventional and confocal fluorescence microscopy on a Zeiss LSM510 confocal microscope with a 100 ϫ 1.4 numeric aperture objective.
Enzyme Assays and Product Identification-Potential substrates were screened by measuring the P i released by co-incubation of nucleotides with Pcd1p and either inorganic pyrophosphatase or alkaline phosphatase as described previously (9,13). Reaction products generated from substrates were identified by high performance anion-exchange chromatography. Reaction mixtures (100 l) containing 50 mM 1, 3-bis[tris(hydroxymethyl)methylamino]propane-HCl, pH 7.0, 5 mM MgCl 2 , 0.5 mM substrate, and 0.125 g of Pcd1p were incubated for up to 10 min at 37°C and applied to a 1-ml Resource-Q column (Amersham Pharmacia Biotech) at 2 ml/min in 5% buffer E. The elution system comprised Buffer D (0.045 M CH 3 COONH 4 , adjusted to pH 4.6 with H 3 PO 4 ) and Buffer E (0.5 M NaH 2 PO 4 , adjusted to pH 2.7 with CH 3 COOH) (14), with a gradient of 5-70% Buffer E over 10 min. Substrates requiring reducing conditions (CoA and 3Ј-dephospho-CoA) were pre-incubated for 5 min at 37°C with DTT before addition to the assay. The final assay concentration of DTT was 1 mM. Peaks were identified with the aid of standards and quantified by area integration for kinetic analysis.
Other Methods-Immunoblotting was performed as described previously (9). Protein concentrations were estimated by the Coomassie Blue binding method (15).

Cloning, Expression, and Purification of Pcd1p
The intronless PCD1 gene contains an open reading frame, YLR151C, that potentially encodes a 39,758-Da protein containing a nudix box (Fig. 1A). Initial and various attempts to express soluble Pcd1p to a high level in E. coli were unsuccessful, with all the recombinant protein being found in inclusion bodies. Resolubilization of this material also failed to produce protein with discernible enzyme activity. It was therefore decided to use the protein isolated from inclusion bodies to raise a polyclonal rabbit antibody that could then be used in an immunoassay for the purification of the overexpressed protein from yeast. The YLR151C ORF was amplified by PCR from genomic DNA and the PCR product inserted into the yeast centromere vector, pPGY1. The resulting plasmid, pPGY151C, was transformed into S. cerevisiae strain INVScI and expression of the insert induced by growth on galactose. By following the major immunoreactive species on Western blots of chromatographic fractions, the protein product of PCD1 was purified to homogeneity from extracts of the induced cells. The purified protein had a molecular mass of 38 kDa according to SDS-PAGE (Fig. 1B). Throughout the purification, the immunoblots revealed more clearly than the gels that the overexpressed Pcd1p actually comprised two species of very similar size (Fig. 1C). N-terminal sequencing of the two bands excised from these blots showed the upper and lower species to have the sequences MLSSKQLI and LSSKQLI, respectively, suggesting that Pcd1p may have undergone some N-terminal processing with the loss of either 7 or 8 amino acids (Fig. 1A).

Properties of the Protein
Substrates-Nucleotides were tested as potential substrates for Pcd1p using a spectrophotometric assay. No activity was found with the following compounds: (deoxy)nucleoside 5Јtriphosphates, nucleoside 5Ј-di-or monophosphates, diadenosine polyphosphates, nucleoside 5Ј-diphosphosugars, cytidine 5Ј-diphosphoalcohols, NAD ϩ , NADH, or FAD. However, substantial activity was found with CoA and some CoA derivatives (Table I). HPLC analysis of the products of CoA hydrolysis showed that the enzyme was a CoA diphosphatase, cleaving the diphosphate linkage in CoA to give 3Ј,5Ј-ADP and 4Ј-phosphopantetheine ( Fig. 2A). Pcd1p is the first nudix hydrolase to be described with such a substrate specificity. The K m and k cat for CoA were 280 M and 4.6 s Ϫ1 , respectively, while the corresponding values for oxidized CoA disulfide (CoASSCoA) were 24 M and 5.0 s Ϫ1 , respectively. Thus, the enzyme has a 13-fold higher k cat /K m ratio for CoASSCoA compared with CoA. These kinetic parameters were calculated by non-linear regression analysis of the data in Fig. 3A. The reciprocal plots in Fig. 3B clearly show that the enzyme follows Michaelis-Menten kinetics with these two substrates. The initial products of CoASSCoA hydrolysis were 3Ј,5Ј-ADP and what is presumed to be CoASSP, the mixed disulfide of CoA and 4Ј-phosphopantetheine, i.e. CoASSCoA lacking a single 3Ј,5Ј-ADP moiety (Fig. 2B). Significant accumulation of this product was observed with time before it too was degraded, presumably to 3Ј,5Ј-ADP and the dimer of 4Ј-phosphopantetheine, suggesting that CoASSP is not as efficient a substrate as CoASSCoA. When measured at a single fixed substrate concentration, moderate activity was also obtained with several short chain acyl-CoA esters while 3Ј-dephospho-CoA was a very poor substrate (Table I). Thus, the 3Ј phosphate on the adenosine moiety appears to be important for substrate recognition. Attempts to demonstrate enzyme activity in crude yeast extracts proved impossible due to 3Ј dephosphorylation of the CoA substrates. The mixed CoA-glutathione disulfide, which may exist in vivo but which is more usually found as an extraction artifact (16,17), was also a relatively poor substrate. Therefore, recognition of CoASSCoA as a good substrate must involve more than just the disulfide bond.
Pcd1p was optimally active at pH 7.0 with 5 mM Mg 2ϩ ions. Mn 2ϩ at 300 M supported 83% of the activity observed with 5 mM Mg 2ϩ ions. Like all other nudix hydrolases tested, Pcd1p was very sensitive to inhibition by fluoride ions with an IC 50 of 22 M using CoASSCoA as substrate (data not shown).
Subcellular Localization-A likely subcellular location for an enzyme with the properties described would be the mitochondria or peroxisomes, as these contain the major cellular CoA pools. The latter organelle is the sole site of fatty acid ␤-oxidation in yeast. The N-terminal 30 -40 amino acids of Pcd1p are rich in Lys, Arg, Ser, and Thr, suggesting they may comprise a mitochondrial targeting signal (Fig. 1A) (18). Indeed, the PSORT algorithm suggests a possible mitochondrial location (19) while a hydrophobic transmembrane segment following the potential leader sequence that could anchor the imported protein in the inner mitochondrial membrane is predicted by the TMpred (20) and TMHMM (21) algorithms (Fig. 1A). The sequence of Pcd1p does not contain a typical C-terminal tripeptide peroxisomal targeting signal (PTS1) (22); however, the sequence RRMLSSKQL in the N-terminal region (Fig. 1A) is a close match to the PTS2 N-terminal peroxisomal matrix tar-  Thus, Pcd1p could be either mitochondrial or peroxisomal. The similarity in peroxisomal and mitochondrial N-terminal targeting signals has been noted before (24).
In order to determine the true subcellular location of Pcd1p, yeast cells were transformed with expression vectors encoding Pcd1p fused to either the C terminus (pyEGFP-PCD1) or the N terminus (pPCD1-yEGFP) of yeast-enhanced green fluorescent protein (yEGFP) (11) and then examined by confocal microscopy. Cells transformed with pyEGFP-PCD1 showed a diffuse cytoplasmic fluorescence with no clear subcellular localization (Fig. 4, A and B) while cells transformed with pPCD1-yEGFP showed a clear punctate fluorescence characteristic of peroxisomes (Fig. 4C). The same cells stained with the mitochondrialspecific dye rhodamine B hexyl ester revealed a quite distinct pattern of mitochondrial staining (Fig. 4D). Superimposition of the latter two images showed only limited coincidence of green and red emissions due to physical overlap of some organelles (Fig. 4E). The structural integrity of the cells was apparent under bright field conditions (Fig. 4F). The peroxisomal location of Pcd1p in cells transformed with pPCD1-yEGFP (Fig.  4G) was confirmed by indirect immunofluorescence microscopy using an antibody to 3-oxoacyl-CoA thiolase (Fox3p), a known peroxisomal enzyme (12,25), and a rhodamine-conjugated second antibody (Fig. 4H). Both signals were clearly coincident (Fig. 4I). These results show that the N-terminal sequence of Pcd1p directs the enzyme to peroxisomes, most probably via the PTS2-like sequence RRMLSSKQL, but not to the mitochondria. Thus, Pcd1p is only the second protein identified in S. cerevisiae to be imported into peroxisomes by virtue of a PTS2 signal, the first being Fox3p (26). Interestingly, purified Fox3p has been reported to lack its 5 N-terminal amino acids (25) while purified Pcd1p appears to have lost 7 or 8 amino acids from its predicted sequence. This would lend support to the suggestion that the PTS2 signal in yeast may undergo specific proteolytic processing after import into the peroxisomes (25). Structure 1 shows the possible sites of proteolytic processing (arrowed) within the putative PTS2 sequences of Fox3p and Pcd1p (boxed).

DISCUSSION
In addition to the nudix box, the sequence of Pcd1p contains a second, contiguous signature motif to the N-terminal side identified in the PROSITE data bank as UPF0035 (Fig. 5). This motif has the sequence LLTXR(SA)X 3 RX 3 GX 3 FPGG and is present in a variety of related bacterial, fungal, animal, and plant putative protein sequences in the GenBank/EMBL nonredundant and expressed sequence tag data bases, some examples of which are shown in Fig. 5. Animals (mouse, rat, human, and Caenorhabditis elegans) have pairs of related sequences. The mouse gene sequences, Nudt7 and Nudt8, encode two 26-kDa proteins that share 34% sequence identity with each other and 26% and 20% sequence identity with Pcd1p, respectively. The Nudt7 gene product is also a peroxisomal CoA diphosphatase with a C-terminal tripeptide targeting signal, SKL. 4 Therefore, the UPF0035 motif may be a determinant of CoA substrate specificity among the nudix hydrolases. Since it overlaps with the predicted transmembrane segment, the latter may not be genuine. An additional sequence feature is the substitution of the usual glutamate residue in the nudix box (marked with * in Fig. 5) by either aspartate or glutamine and the inclusion of an extra amino acid in several cases between this residue and the invariant alanine (marked with † in Fig.  5). Thus, the consensus nudix box in this family of potential CoA diphosphatases is GX 5 DX 6 AXREXXEEXGU.
Pcd1p represents a new class of nudix hydrolase and a new class of enzyme. The existence of such an activity has been inferred previously but it has not been isolated. In E. coli, and presumably in other cells, regulation of the concentration of CoA includes turnover to form 3Ј,5Ј-ADP and 4Ј-phosphopantetheine, the latter being formed directly or by transfer to and removal from acyl carrier protein (27). The former route would require a CoA diphosphatase. CoA diphosphatase has also been proposed as an activity associated with the 400-kDa CoA synthesizing protein complex from S. cerevisiae in which it forms part of an alternative pathway for CoA biosynthesis that differs from the principal route of 3Ј-dephospho-CoA and CoA synthesis by this complex (28). This CoA/4Ј-phosphopantetheine cycle also includes hydrolysis of CoA to 3Ј,5Ј-ADP and 4Ј-phosphopantetheine, which then reacts with ATP to give 3Ј-dephospho-CoA, and then CoA. At the moment we do not know if Pcd1p is responsible for this activity. A recent comprehensive two-hybrid analysis of protein-protein interactions in S. cerevisiae revealed no interacting partners for Pcd1p (29). However, stable interactions requiring three or more partners would not have been detected.
Alternatively, the high activity of Pcd1p toward oxidized CoA disulfide and its peroxisomal location suggest a function that may be more in keeping with the proposal that a major role of the nudix hydrolases is the elimination of toxic nucleotides. Oxidative stress generates increased levels of several of the substrates for nudix hydrolases, e.g. 8-oxo-dGTP for the MutT protein (30), diadenosine tetraphosphate (Ap 4 A) for Ap 4 A hydrolase (31,32), and ADP-ribose for ADP-sugar diphosphatases (13). Many of the oxidative reactions in peroxisomes generate hydrogen peroxide and the resultant oxidizing environment would be expected to increase the CoA disulfide/CoA ratio (cf. the oxidized glutathione/glutathione ratio). Indeed, some organisms such as Staphylococcus aureus use a thiol/disulfide redox system based on CoA, CoA disulfide, and a CoA disulfide reductase instead of the more common glutathione system to maintain a reducing environment (33). In the probable absence of a specific CoA disulfide reductase to regenerate CoA within the yeast peroxisomes (33), accumulation of CoA disulfide could lead to a reduction in the ability to oxidize fatty acids, hence the need for Pcd1p. Since the S. cerevisiae NPY1 NADH diphosphatase is also peroxisomal, 3 both enzymes may participate in the maintenance and protection of essential nucleotide pools for ␤-oxidation. Although preliminary experiments with a yeast strain disrupted for PCD1 have failed to show any substantial deficiency in growth on oleic acid, 3 more detailed investigations are under way to determine the precise function of Pcd1p and the consequences of PCD1 disruption.
In conclusion, a nudix hydrolase with a previously undescribed enzyme activity has been characterized in yeast. It is the first nudix hydrolase to be shown to be peroxisomal and only the second protein known in S. cerevisiae to be targeted by a PTS2 signal sequence. It will be of interest to determine if a deficiency in the human orthologue is associated with impaired peroxisomal function and disease. FIG. 5. Multiple sequence alignment of Pcd1p and related sequences. A 44amino acid section of the translated S. cerevisiae PCD1 sequence encompassing the UPF0035 signature motif and the nudix box was aligned using the ClustalW program with the following sequences (data base accession numbers in parentheses): Schizosaccharomyces pombe YDH5 (SwissProt Q92350), E. coli YEAB (SwissProt P43337), Mycobacterium tuberculosis Rv3672c (SPTREMBL O69640), mouse Nudt7 (GenBank™ AA674232), mouse Nudt8 (GenBank™ AI854862), C. elegans Y38A8.1 (SwissProt Q23236) and Y87G2A.14 (GenBank™ AL110500), and Arabidopsis thaliana (SPTREMBL O22951 and O23663). "Invariant" amino acids of the nudix box and the UPF0035 motif are highlighted as white on black. See "Discussion" for significance of residues marked * and †.