Mutagenesis of the Amino Targeting Signal of Saccharomyces cereuisiae 3-Ketoacyl-CoA Thiolase Reveals Conserved Amino Acids Required for Import into Peroxisomes in Viuo*

Saccharomyces cerevisiae peroxisomal 3-ketoacyl-CoA thiolase is a soluble matrix protein that does not end in a consensus peroxisomal targeting signal-1. The amino terminus of S. cerevisiae peroxisomal thiolase is conserved in 6 of 11 residues with the amino terminus of rat thiolase B, shown to act as a peroxisomal targeting sig-nal-2 (Swinkels, B. W., Gould, 5. J., Bodnar, A. G., Rachu- binski, R. A, and Subramani, S. (1991) EMBO J. 10,3255-3262). Unlike mammalian peroxisomal thiolases, there is no extensive cleavage of S. cerevisiae thiolase upon im- port into peroxisomes. We demonstrate by in vivo expression that the amino-terminal 16 amino acids of S. cerevisiae thiolase are necessary and sufficient for targeting to peroxisomes. This result implies that yeast, like mammalian cells, can target proteins to the peroxisomal matrix by at least two different routes. We also demonstrate by targeted mutagenesis and in vivo ex- pression of mutated thiolase genes that three amino acids conserved in the amino

The amino-terminal 11 amino acids of Saccharomyces cerevisiae thiolase are identical in 6 of 11 residues with the PTS-2 of rat peroxisomal thiolase. A comparison of the amino-terminal regions of mammalian and yeast peroxisomal thiolases reveals three amino acids identical in all thiolases (Fig. 1). Because of these similarities, we wished to determine whether the amino terminus of S. cereuisiae thiolase functions as a PTS-2. We report that the amino-terminal 16 amino acids of S. cereuisiae thiolase act as a PTS-2 and that the arginine at position 4 and the leucines at positions 5 and 12 are critical for efficient targeting of thiolase to peroxisomes.
Dansformation of East-Plasmid DNA was introduced into 20-1.11 aliquots of cells by electroporation with a BRL Cell-Porator equipped with a voltage booster set to 4-ka resistance and delivering a pulse with a field strength of approximately 7.5 kV.cm-l (Nuttley et aZ., 1993). Cells were diluted into 100 pl of 1 M sorbitol and spread onto selective YNBD agar plates.
Yeast Strains-The DL-1 strain of S. cerevisiae (MATa, leu2, ura3, his3) was used to express a proteinA(F'rA)-based passenger protein and to construct a strain (STUD; MATa, Zeu2, ura3, his3, THI::URAJ) having its nuclear thiolase gene (THI) disrupted by homologous recombination with the URA3 gene. A HinfI fragment containing the URA3 gene from the plasmid YCp50 was made blunt with the Klenow fragment of DNA polymerase and inserted by ligation into pGEM7Zff +) cut with SmaI to yield pGEMUIU3. The THI gene was excised from pSG524 as a BamHI-XbaI fragment and subcloned between the BamHI-XbaI sites of a derivative pGEM7Zf(+) with its SphI site destroyed. The THI coding sequence between codons Cys-107 andksp-359 was removed by digestion with SphI and ClaI. This portion was replaced with an SphI-ClaI fragment from pGEMURA3 containing the URA3 gene to yield pSTUD3. pSTUD3 was digested with BamHI and SspI, and the resulting fragment (Fig. 2, STUD) was introduced into DL-1 cells by electroporation. Colonies capable of growth on glucose in the absence of uracil but incapable of growth on oleic acid when replicaplated onto YNO were subjected to further testing by Southern and Western blot analysis to confirm the success of the disruption of the THI gene.
Thiolase Expression Constructs-The S. cerevisiae THI gene was expressed from plasmids pSG524 and pSG522 (Fig. 2) under the control of the oleic acid-responsive S. cerevisiae fatty acyl-CoA oxidase promoter (Wang et al., 1992) in the plasmid pRS315 (Sikorski and Hieter, 1989). To synthesize thiolase initiating at the AUG encoding Met-17, pSG522 Alignment of the amino termini of the known pemxisomal thiolases. The initiating methionines of the rat B, human, S. cerevisiae, Candida tropicalis, and Yarrowia lipolytica thiolases are aligned with the second methionine of rat thiolase A. The shaded regions highlight sequences necessary and sufficient for targeting rat and S. cereuisiae thiolases to peroxisomes. Italicized amino acids of S. cerevisiae thiolase were subjected to mutagenesis. Amino acids designated by asterisks are identical in all thiolases. Downward pointing arrows represent the known or putative cleavage positions of mammalian thiolases.
was digested with NcoI and religated unto itself to obtain pSG522ANcoI (Fig. 2).
Construction of a Passenger Protein-The construction of pgGPrAS-TOP from pSPUTKgGPrA and the structure of the passenger protein gGPrA have been described in detail elsewhere (Janiak et al., 1994). Complementary oligonucleotides containing an in-frame stop codon (5'-GATC'R'ATAGGCGGCGGCG and 5'-GATCCGCCGCCGCCTATAA; the stop codonFthe noncoding strand is underlined) were annealed and ligated into a BamHI site immediately downstream of DNA encoding four IgG-binding domains of Staphylococcus aureus protein A and a 27-amino acid segment of chimpanzee globin modified to contain a site for glycosylation (gGPrA).
The resultant plasmid was digested to completion with EcoRV and partially with NcoI to yield a fragment containing the entire sequence coding for gGPrA. This segment was subcloned for expression by digestion of pSG524 and pSG522 to completion with XbaI, blunting with Klenow polymerase, digestion with NcoI, and purification of the vector. Ligation of the insert to the pSG524derived vector resulted in the plasmid pgGPrASCTN (Fig. 2), which directs the expression of the amino-terminal 16 amino acids of thiolase fused to the passenger protein gGPrA. The pSG522-derived construct (pgGPrASTOP, Fig. 2) directs the expression of gGPrA alone. quence 5'-C'R'GGATCCGCTAGCCATGC'R'CMGACTACAAAGTA-Construction of PTS-2 Mutants-An oligonucleotide having the se-TCAAGGATCATITGGTGGAGAGCGCCATGGCG was synthesized in a manner such that the composition of the nucleotides at the positions indicated in bold was 90% of the indicated nucleotide and 3.3% of each of the three other nucleotides (Ausubel et al., 1987). The oligonucleotide was heated at 70 "C for 5 min and allowed to cool slowly to 16 "C to anneal the palindrome (underlined) at the 3'-end of the oligonucleotide. The annealed oligonucleotides were extended at 16 "C for 16 h with the Klenow fragment of DNA polymerase in the presence of trace amounts of &"PIdATP. Extension was completed by a further 2-h incubation at 37 "C with fresh Klenow fragment. The extended product was digested with excess BamHI. The cleavage products were subjected to electrophoresis on a 5% polyacrylamide gel, which was then exposed to x-ray film. The radiolabeled band migrating at -154 base pairs was excised, eluted, and digested with NcoI. The resulting double-stranded oligonucleotide was Ligated into BamHI-NcoI-digested pSG524. Plasmid DNA from randomly selected colonies was sequenced by the dideoxy method using a primer (5'-CAATAACTACATC'IT) that hybridized immediately downstream of the region targeted for mutagenesis. The STUD yeast strain was electroporated with selected mutant plasmids. Transformants were checked for growth on oleic acid-agar, and the intracellular location of thiolase was determined by subcellular fractionation.
Thiolase Assay-Yeast cells grown in SCIM were pelleted and disrupted with glass beads in buffer containing 50 nm Tris-HC1 (pH 7.5), 50 m~ NaC1, 0.1 nm EDTA, 0.1 m~ ZnCl,, and 1 nm phenylmethylsulfonyl fluoride (Needleman and Tzagoloff, 1975). Thiolase activity was assayed as described by Suebert et al. (1968) except that enoyl-CoA hydratase was not required for the enzymatic formation of the substrate. 3-Hydroxydecanoyl-CoA was added at a final concentration of 50 p to a solution of 50 m~ Tris-HC1 (pH 9.01, 50 nm KC], 25 m~ MgCI2, 50 pgml-' bovine serum albumin, 1 m~ NAD, 1 m~ sodium pyruvate, 25 milliunits of pig heart 3-hydroxyacyl-CoA dehydrogenase, 1.8 units of rabbit muscle lactate dehydrogenase. The formation of 3-ketodecanoyl-CoA was monitored as an increase in absorbance at 303 nm. Measurement of thiolase activity was initiated by the addition of cell lysate containing 1% (w/v) Triton X-100 and of CoASH to a final concentration of 150 p. Activity was calculated using the extinction coefficient E = 13.9 cm2ymol-'.
SDS-Polyacrylamide Gel Electrophoresis and Western Blot Analy-sisSDS-polyacrylamide gel electrophoresis was performed on 10% polyacrylamide gels (Laemmli, 1970). Western blotting was performed essentially as described (Burnette, 1981). Nitrocellulose was blocked with 1% skim milk in 20 m~ Tris-HC1 (pH 7.5), 150 nm NaCl, 0.05% (v/v) Tween 20. Thiolase was detected with rabbit anti-thiolase serum, followed by lZ5I-Protein A (Amersham). The gGPrA-based passenger protein was detected with 1251-labeled rabbit IgG. Blots were exposed to X-Omat AR film or to a storage phosphor screen, which was scanned on a PhosphorImager and quantified using software provided by the manufacturer (Molecular Dynamics). Nytran membrane (Schleicher and Schuell) was used to compare the electrophoretic mobilities of in uitro-translated thiolase and thiolase isolated from peroxisomes. The membrane was first probed with anti-thiolase serum followed by I T -Protein A. f i r washing, the blot was dried, soaked in 10% 2,5-diphenyloxazole in toluene, and exposed to X-Omat AR film at -70 "C.
In Vitro Panscription and Panslation-pGEM7Zf+) containing a BamHI-XbaI insert encoding thiolase was linearized with XbaI and transcribed using SP6 RNA polymerase. The RNA was translated in rabbit reticulocyte lysate in the presence of ~-[~~Slmethionine.
Miscellaneous-Total protein content was determined as described

RESULTS
Thiolase of S. cerevisiae Zs Not Cleaved-Forms A and B of rat peroxisomal thiolase and human peroxisomal thiolase are cleaved upon import into peroxisomes (Hijikata et al., 1987;Tager et al., 1985). Cleavage results in mature forms of thiolase that are substantially shorter than their newly synthesized counterparts. S. cerevisiae peroxisomal thiolase synthesized in vitro had the same electrophoretic mobility in SDS-polyacrylamide gel electrophoresis as thiolase from purified peroxisomes (Fig. 3). This result suggests that thiolase is not cleaved upon import into peroxisomes of S. cerevisiae. However, because of the limited resolution of this approach, removal of a few amino acids from the newly synthesized thiolase upon import into peroxisomes cannot be ruled out. Nevertheless, S. cerevisiae thiolase is imported into peroxisomes without the extensive cleavage observed for mammalian peroxisomal thiolases.

The Targeting Signal of S. cerevisiae Peroxisomal Thiolase
Resides at Its Amino Terminus-A mutated gene (Fig. 2,  pSG522ANcoI) coding for a truncated form of S. cerevisiae peroxisomal thiolase lacking the first 16 amino acids at its amino terminus was expressed in the THZ gene disruption strain, STUD. This truncated thiolase failed to support growth on oleic acid-agar (Fig. 4, A1-16) and was localized to the cytosol (Fig.  5,A1-16). Thiolase A1-16 retained enzymatic activity ( Table I), indicating that the inability to grow on oleic acid-agar was due not to a loss of enzymatic activity by the truncated thiolase but to its lack of import into peroxisomes. Moreover, since the truncated thiolase retained normal levels of enzymatic activity, its inability to be imported into peroxisomes was probably not the result of a gross alteration of structure but rather the result of the removal of the peroxisomal targeting signal of thiolase.
The amino-terminal 16 amino acids of thiolase were fused to the gGPrA passenger protein to show their sufficiency for import into peroxisomes. This passenger has been shown to be cytosolic in the absence of added targeting information (Janiak et al., 1994). However, it is passive to translocation and is correctly targeted by both carboxyl-terminal and amino-terminal signals to a variety of cellular compartments, including the yeast peroxisome (Janiak et al., 1994). The gGPrA passenger encoded by pgGPrASTOP (Fig. 2) was exclusively localized to the cytosol (Fig. 6, panel A, upward pointing arrowheads). The fusion protein (Fig. 2, gGPrASCTh9 composed of the aminoterminal 16 amino acids of thiolase attached to the amino terminus of gGPrA (Fig. 6, panels B and C, upward pointing ar-rowheads) co-fractionated with endogenous thiolase (Fig. 6, downward pointing arrowheads). The thiolase-gGPrA fusion construct was released from the organelle pellet by treatment with the detergent Triton X-100, indicating that the thiolase-gGPrA fusion protein is targeted and imported into peroxisomes and does not form an insoluble inclusion body (data not shown; McCollum et al. (1993)).

Mutations in Several Conserved Amino Acids within the Amino-terminal PTS of S. cerevisiae Thiolase Impede or Abolish
Import into Peroxisomes-A comparison of peroxisomal thiolases reveals a conservation among all thiolases of certain amino acids found within the 11-amino acid PTS of rat thiolase (Swinkels et al., 1991;Osumi et al., 1991). In S. cerevisiae peroxisomal thiolase, these conserved residues are Arg-4, Leu-5, and Leu-12 (Fig. 1). Gln-6 and His-11 are conserved among S. cerevisiae thiolase and the mammalian thiolases, but they are not conserved in the thiolases from other organisms. The importance of these amino acids in targeting S. cerevisiae thiolase to peroxisomes was determined by random mutagenesis of the corresponding codons followed by expression of the mutant thiolase genes in vivo. Subcellular localization of mu- Immunodrtrction of thiolasr was hy rahhit antl-thiolasr srrum followed hy l""I-protrin A. Strain drsignations arr as in Fig. 4. tant thiolases was determined using two assays. The first was a functional assay involving growth of S. cerevisiae on oleic acid-agar. The thiolase gene disruption strain STUD (Fig.

4,
NULL.) and the STUD strain carrying a plasmid expressing the gene coding for Al-16 thiolase f Fig. 4.11-16) could not grow on oleic acid-agar, while the STUD strain carrying a plasmid expressing the wild-type thiolase gene (Fig. 4, W T ) could grow on this medium. Therefore, growth on oleic acid-agar necessitates correct targeting of thiolase to peroxisomes. The second assay was biochemical, involving subcellular fractionation followed by immunodetection with anti-thiolase antibodies (Fig. 5). The inability of the STUD strain expressing certain mutant thiolase genes to grow on oleic acid-agar was not due to the synthesis of enzymatically inactive thiolase in these transformants. All mutant thiolases were enz-matically active ( Table I ) at levels comparable to that of wild-type thiolase. The specific activities of mutant thiolases vaned between 2.8 and 11.0 nmol.min".mg of protein". The specific activity of wild-type thiolase was 6.6 nmol.min".mg of protein". One mutant, L5R*, did not grow on oleic acid-agar (Fig. 4 ) and showed almost no thiolase activity (0.14 nmol.min".mg of protein"); however, this low activity was due not to the production of normal levels of a poorly active thiolase hut to the low levels of thiolase synthesized (Fig. 5). A second isolate of this mutant, L5R. still showed reduced growth on oleic acid-agar (Fig. 4). even though it made increased amounts of thiolase (Fig. 5 ) that
Only the mutant L5R and the double mutant L5VIQ6P showed retarded growth on oleic acid-agar at both 23 "C and 30 "C ( Fig. 4) and in the almost total mislocalization of thiolase to the cytosol (Fig. 5). The single mutation Q6P resulted in a moderate reduction in targeting efficiency (Fig. 5). The Leu to Val substitution at position 5 in L5VlQ6P is conservative and might therefore be expected to result in little or no further diminution of targeting over that observed in Q6P alone. However, it appears that these side-by-side substitutions act synergistically to abolish targeting and import of thiolase into peroxisomes (Fig. 51, although we cannot formally rule out the possibility that the single mutant L5V would show the same characteristics as the L5V/Q6P double mutant. Indeed, for all double mutants, it is difficult to assess whether a single mutation in one or the other of the two amino acids or whether the two mutations acting in combination is responsible for the observed behavior of a particular double mutant thiolase. The L5R mutant examined initially was very poorly expressed (designated as L5R*) and was replaced by an alternative isolate expressing the same mutation at an acceptable level (Fig. 5). Neither L5R* nor L5R was effectively targeted, and neither could support growth on oleic acid-agar (Fig. 4).
The mutants R4S and L5Q supported growth on oleic acidagar at 30 "C (Fig. 4). These two mutants cultured at 30 "C showed similar amounts of thiolase in their 20 kgP fractions as did mutants that showed poor growth on oleic acid-agar at 30 "C (Figs. 4 and 5, compare R4S and L5Q with R4G, L12S, and L5R). This observation suggests that only small amounts of thiolase need be correctly targeted to peroxisomes to support growth on oleic acid-agar.
Two mutants (R4G and L12S) targeted thiolase poorly (Fig.   5). These mutants grew poorly at 30 "C but were able to grow better at 23 "C on oleic acid-agar (Fig. 4). This temperature sensitivity may be due to thermal destabilization within the thiolase PTS, resulting in reduced interaction between it and factor(s) involved in the recognition of the PTS. Four mutations (S2A, R4A/L12M, L12V, and L12M) had a negligible effect on targeting (Fig. 5 ) and all showed reasonable growth on oleic acid-agar (Fig. 4). The conversion of Ser-2 to Ala (S2A) through the introduction of an NcoI site at the initiation codon of the thiolase gene in pSG522 (Fig. 2) was not expected to affect the targeting of thiolase, as this amino acid is not conserved in the different thiolases. The mutations L12V and L12M were tolerated, most likely because the substituted amino acids occupy the same approximate volume as Leu, and because they maintain hydrophobicity at this position. In contrast, mutants with bulky aromatic side chains at position 12 (L12F and L12W) were less efficiently targeted (Fig. 5). The larger aromatic side chains may cause steric interference between the thiolase PTS and presumed import receptorb).
The double mutant R4ALl2M was efficiently targeted (Fig.  5 ) and supported strong growth on oleic acid-agar (Fig. 4). This result is difficult to explain given that the more conservative substitution R4K was less efficiently targeted and had decreased growth on oleic acid-agar. If the reduced volume of the Ala side chain compensated for the loss of the positively charged Arg side chain, then the mutant R4S should be expected to be accommodated as well as R4A. However, this was not the case, as the mutant R4S was poorly targeted. The possibility remains that the second amino acid mutation (L12M) in the double mutant compensates in a n unknown way for the loss of Arg at position 4.
All thiolases characterized to date have a conserved Gln or Asn corresponding to position 6 in S. cerevisiae peroxisomal thiolase. Mutation of Gln at this position to Pro (Q6P) resulted in reduced targeting of the mutated thiolase vis-a-vis wild-type thiolase (Fig. 5). This reduced targeting might be the result of the loss of hydrogen bonding capacity at this position in the mutant. The mutant Q6H retains hydrogen bonding capacity and was more efficiently targeted than Q6P but still less well than wild-type thiolase. In contrast, thiolase in which the conserved His at position 11 is replaced by Gln (H11Q) was divided almost equally between the 20 kgS and 20 kgP. The mutant H l l L was targeted more efficiently than H11Q.
Of the three double mutants encompassing Q6 and H11 (QGR/HllN, QGFUHllT, and QGFUHllY), Q6R/H11Y is the most interesting in that while -40% of thiolase was localized to the 20 kgP in yeast grown a t 30 "C (Fig. 51, this mutant was incapable of growth on oleic acid-agar at 30 "C (Fig. 4). Immunolocalization of thiolase to the 20 kgP cannot distinguish between thiolase correctly localized to peroxisomes and thiolase that is mislocalized to one or more compartments that co-sediment with peroxisomes in the 20 kgP. Density gradient centrifugation showed a preferential immunolocalization of the Q6FUH11Y double mutant to fractions enriched for mitochondria (data not shown). Strong immunofluorescence from cytosolic thiolase did not permit an unequivocal localization of the QGR/HllY mutant by light microscopy (data not shown). The effect of mutating His-11 of S. cerevisiae thiolase may be similar to the effects seen by mutating the corresponding His residue (His-17) in the targeting signal of rat peroxisomal thiolase B. Mutation of His-17 of rat peroxisomal thiolase B to Arg, Lys, Leu, or Val resulted in the targeting of DHFR-PTS-2 fusion constructs to mitochondria and in mislocalization to the cytosol in CHO cells (Osumi et al., 1992). The change of Gln to Arg at position 6 in the double mutant Q6R/HllY may exert an influence on the mistargeting of this mutant to mitochondria, for while the single mutants HllL and HllQ had more thiolase and a similar amount of thiolase, respectively, in their 20 kgP fractions than did Q6FUH11Y (Fig. 51, they still allowed for growth at both 23 "C and 30 "C on oleic acid-agar (Fig. 4). Therefore, in these two single mutants, some portion of the thiolase found in the 20 kgP fractions must be correctly sorted to peroxisomes so as to permit growth. DISCUSSION We have shown that the amino-terminal 16 amino acids of S. cerevisiae peroxisomal thiolase are both necessary and sufficient for protein targeting to peroxisomes in uiuo. Therefore, S. cerevisiae peroxisomal thiolase has a PTS-2, akin to the PTS-2s of rat thiolases A and B. However, unlike the amino termini of peroxisomal thiolases from human and rat, the amino terminus of S. cerevisiae peroxisomal thiolase is not extensively cleaved upon import. In contrast, we have recently observed that peroxisomal thiolase is cleaved upon import into peroxisomes of the yeast E: lipolytica (Nuttley et al., 1994). Therefore, the phenomenon of thiolase cleavage is conserved in at least one yeast species.
Mutations of amino acids within the PTS-2 of S. cerevisiae thiolase affect the localization and import of thiolase as indicated by subcellular fractionation and the ability to restore growth on oleic acid-agar to a thiolase-deficient strain. Although there is not always a tight correlation between increased amounts of thiolase in peroxisomes and increased growth on oleic acid-agar in the "spot" growth assay presented in Fig. 4, the results are consistent with a few conclusions. Substitution of Leud by Arg completely abolishes import of thiolase by peroxisomes and retards growth on oleic acid-agar at both 23 "C and 30 "C. Substitution of Leu-5 by Gln compromises severely targeting to peroxisomes, causing most of the thiolase to be mislocalized to the cytosol. However, the small amount of thiolase imported into peroxisomes in the L5Q mutant permits reduced growth on oleic acid-agar. Similarly, when Leu-12 is replaced by hydrophobic amino acids with similar volumes (Met and Val), there is little effect on targeting of thiolase to peroxisomes. However, when Leu-12 is replaced by much larger hydrophobic amino acids (Phe and "p), thiolase targeting to peroxisomes is reduced. Substitution of a polar residue (Ser) at Leu-12 abolishes thiolase targeting to peroxisomes at 30 "C. The Leu residues at -23 and -16 of the PTS-2 of rat thiolase B correspond to Leu-5 and Leu-12 of S. cerevisiae thiolase. Helical wheel analysis showed that these Leu residues align in the two thiolases, suggesting that they form part of a hydrophobic surface required for interaction of the PTS-2s with their respective receptors. However, although our results suggest that small hydrophobic amino acids are required at positions 5 and 12, we have no direct evidence that the PTS-2 of S. cereuisiae thiolase forms a helical structure.
Arg-4 is a critical residue within the PTS-2 of S. cerevisiae thiolase. Substitution ofkg-4 by either Ser or Gly causes most of the thiolase to be mislocalized to the cytosol. Even the conservative substitution of Arg-4 by Lys results in significant mislocalization of thiolase to the cytosol, suggesting that the precise positioning of a positively charged group within the overall structure of the signal is required for maximum functioning of the targeting signal.
The existence of S. cerevisiae mutants that synthesize thiolases altered in their PTS and which are less efficiently targeted to peroxisomes than wild-type thiolase provides us with the means for isolating proteins that could interact with the thiolase PTS. We are currently exploring the possibility of suppressing the defect in peroxisomal import of certain mutant thiolases by overexpression of genes contained in a multicopy plasmid vector library. Such a strategy should allow us to identify genes whose products are involved in recognizing the thiolase PTS-2, thereby initiating import of thiolase into the peroxisomal matrix.