Role of the proregion in the production and secretion of the Yarrowia lipolytica alkaline extracellular protease.

The yeast Yarrowia lipolytica secretes an alkaline extracellular protease (AEP). It is first synthesized as a precursor comprising a putative signal peptide, a stretch of 10 X-Ala or X-Pro sequences that are substrates for a dipeptidyl aminopeptidase, a large pro-region that contains a glycosylation site and two Lys-Arg sites that can be cleaved by a KEX2-like endoprotease and finally the mature protease itself. A defect in the XPR6 (KEX2-like) gene results in the secretion of an inactive proenzyme (Matoba, S., and Ogrydziak, D. M. (1989) J. Biol. Chem. 264, 6037-6043), showing that the proregion inhibits protease activity. To determine whether the proregion plays an additional role in protease secretion, we have generated deletions and point mutations in the corresponding region of the structural gene. In this paper we examine the effects of these mutations on AEP secretion and maturation and show that the proregion is essential for its secretion. All deletions affecting the proregion resulted in the intracellular accumulation of unprocessed precursors. Deletion of the glycosylation site in the proregion resulted in the production of an unglycosylated precursor that was secreted and matured correctly at 18 degrees C but accumulated in the cells at 28 degrees C. From these results, we propose that the AEP prosequence plays an additional essential role in guiding the proper folding of the protein into a conformation compatible with secretion.

Although the roles of signal sequences and other signals involved in routing proteins to different cellular organelles are now well established (Pugsley, 1989), the functions of proregions such as those present in a-factor (Julius et al., 1983), carboxypeptidase Y (Valls et al., 1987), antibiotics such as nisin (Kaletta and Entian, 1989), and several proteases (alytic protease , subtilisin (Ikemura et al., 1987), alkaline and neutral protease from Bacillus amyloliquefaciens (Nakayama et al., 1987;Vasantha et al., 1984), alkaline * This work was supported by grants from the Institut National de la Recherche Agronomique, Centre National de la Recherche Scientifique, and from Pfizer Inc. (Groton, CT). 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.
$ Recipient of a fellowship from the Ministere de la Recherche et Technologie.
§ T o whom correspondence should be sent.  (Davidow et al., 1987), and from lactic bacteria (Kok et al., 1975) are less well understood. The length of the proregion may vary from a few amino acids (seven for a-lytic protease;  to several hundred (e.g. 221 amino acids for the neutral protease from B. amyloliquefaciens (Vasantha et al., 1984) and 143 amino acids for the AEP (Davidow et al., 1987)). In addition, proregions may carry sites for post-transcriptional modifications such as glycosylation (e.g. one in the AEP propeptide and three in the a-factor propeptide). It has been proposed that propeptides may prevent the formation of active structures (production of zymogen for proteases (Docherty and Steiner, 1982;Wandesman, 1989) or prehormones (Julius et al., 1983)) or may function as sorting signals (e.g. to the vacuoles in the case of carboxypeptidase Y) to increase the efficiency of secretion of small peptides such as a-factor (Bussey, 1988) or to promote the correct folding of the protein (Ikemura et al., 1987;Silen et al., 1988;Stoller and Shields, 1989).
Although several reports have dealt with the role of proregions in the secretion of heterologous proteins by yeast (Brake et al., 1984;Ernst, 1988;Franke et al., 1988;Heslot et al., 1990) no analysis of the role of these regions in the extracellular secretion of their cognate protein has been reported so far. To investigate the role of a specific proregion we have used AEP which is secreted by the yeast Y. lipolytica in quantities that can reach 1-2 g/liter under optimal conditions (Tobe et al., 1976;. The XPR2 gene coding for the AEP has been cloned in several laboratories (Davidow et al., 1987;Matoba et al., 1988;Nicaud et al., 1989b). The DNA sequence revealed that it may code for a preproenzyme precursor of 46,903 Da. Following the putative signal sequence there is a stretch of X-Ala or X-Pro sequences that are possible substrates for a dipeptidyl aminopeptidase. This is followed by a proregion that includes a glycosylation site (AsnlZ3) and two Lys-Arg sequences (positions 59-60 and 156-157) that are possible substrates for a KEXZ-like endoprotease. The mature protein starts at amino acid 158 and has a molecular size of 30,524 Da.
Ogrydziak and collaborators (Matoba and Ogrydziak, 1989;Matoba et al., 1988) have shown that the AEP precursor undergoes a complex series of maturation steps. In the following section precursor size will be named according to Ogrydziak and co-workers (Matoba and Ogrydziak, 1989;Matoba et al., 1988). A glycosylated (pre)proprecursor polypeptide of 55 kDa generates an inactive proprecursor of 52 kDa after cleavage of the signal sequence and of the X-Ala/X-Pro sequences.
This glycosylated precursor may be secreted in the culture medium by an xpr6 mutant defective in KEX2-like activity (Matoba and Ogrydziak, 1989). This secreted polypeptide consists of a mixture of three proprecursors starting at Ala3', Val3', and G~u~~, confirming that processing involves a dipep-tidy1 aminopeptidase. Experiments with tunicamycin and endoglycosidase H-treated cells showed that both the 55-and the 52-kDa precursors contain about 2 kDa of N-linked oligosaccharide. Further processing generates 44-and 36-kDa precursors and the mature 32-kDa protease. The proregion can thus be divided into three parts: proI, proII, and proIII, with putative cleavage sites at Lys-Ar$' (proI-proII), Lys12' or L y P (pro11-proIII), and L y s -A~g l~~ (proIII-mature), respectively. Finally, Ogrydziak and collaborators (Matoba et al., 1988) have proposed the following precursor-product relationship: 55 kDa -P 52 kDa + 44 kDa + 36 kDa + 32 kDa. It remains unclear whether the 44-and 36-kDa precursors are part of the main processing pathway and whether signal peptide cleavage occurs early in the ER or later in the secretory pathway (Matoba et al., 1988).
We have shown that the AEP proregion is not absolutely necessary for extracellular secretion of heterologous proteins. Indeed, strains carrying the interferon-a1 gene coding for porcine interferon-a fused to XPR2 after the signal sequence or the pro1 site of AEP are able to secrete antiviral activity into the growth medium (Heslot et al., 1990). Similar results were obtained with prochymosin fusions (Franke et al., 1988). In this article we report the effects of deletions and point mutations on the proregion and the mature part of the protease on the production and secretion of AEP. We show that in addition to a zymogen function, the complete proregion is required for AEP secretion and that secretion at 28 "C depends on the glycosylation of the precursor. We also confirm that AEP processing corresponds to the precursor-product relationship 55 kDa + 52 kDa + 32 kDa and that the 55-kDa precursor corresponds to the precursor from which the signal sequence has been cleaved.

MATERIALS AND METHODS
Enzymes-All restriction enzymes were purchased from Appligbe. T 4 polynucleotide kinase was purchased from Amersham Corp., and the Klenow fragment of the DNA polymerase I and T4 DNA ligase were purchased from Boehringer Mannheim.
Strains, Plasmids, and Media-Strains and plasmids used are listed in Table I. Escherichia coli strain TG1 was used for oligonucleotide mutagenesis and for plasmid DNA propagation. Y. lipolytica isogenic strains JM12, which carries the wild-type XPR2 gene, and JM23, in which XPR2 is disrupted, were used as recipients for the different mutated XPR2 constructs. Restriction maps of the chromosome around the XPR2 locus are shown in Fig. 1B. The E. coli vectors used for mutagenesis were constructed by inserting a 1.88-kilobase Chi/ XbaI fragment from pINA152 (Nicaud et al., 198913) spanning the promoter and the preprosequence of the XPR2 gene into the Bluescript vectors KS+ and KS-(Stratagene) resulting in plasmid pINA303 (see Fig. 1A). The Y. lipolytica vector pINA310 includes the following features: (i) a NheI (or MluI)/XbaI cassette containing the promoter and the preproregion of the XPR2 gene that can be exchanged with the E. coli plasmids used for mutagenesis; (ii) the region coding for the mature AEP which can also be excised as an XbaI/ ClaI fragment for mutagenesis; (iii) unique MluI or NheI restriction sites in the XPR2 promoter region used to target the plasmid to the XPR2 locus in the chromosome; and (iv) the marker gene LEU2 from Y. lipolytica which can be excised using EcoRI. Plasmid pINA392 (J. M. Beckerich, this laboratory), used for the in vitro experiments, contains the XPRB gene placed under the control of the phage SP6 promoter. In this plasmid, the SP6 promoter of pSP64 (Boehringer Mannheim) is fused through the (destroyed) PstI site to the 5'noncoding sequence of AEP 13 bp upstream from the ATG. The derivative plasmids pINA467, pINA468, pINA469, and pINA472 containing Dell, De12, De13, and DelG, respectively (see "Oligonucleotide Mutagenesis"), were constructed by three-way ligation using the EcoRIISspI and SspIIBgll fragments from pINA392 (containing the SP6 promoter) and the BglI/EcoRI fragment from pINA310 derivatives containing the mutations Dell, De12, De13, and DelG.
YNB minimal medium and the inducing medium YPDm were described previously (Nicaud et al., 1989b). Skim milk plates (SKM) were prepared according to Ogrydziak and Mortimer (1977). Media used for growing E. coli were prepared as described in Miller (1972).
Oligonucleotide Mutagenesis-The oligonucleotides used for mutagenesis ( Fig. 2 A ) were synthesized using the solid phase phosphotriester method (Efimov et al., 1983) with a Cyclone DNA synthesizer (Biosearch) and were purified by high pressure liquid chromatography.
Site-directed deletions or other mutations were constructed by the primer extension method of Carter et al. (1985) with minor modifications. Single-stranded matrix DNA was obtained by infection of E. coli strain TG1 containing plasmid pINA303 grown in 2TY with the M13 helper phage M13 KO7 as described by the supplier (Stratagene). The oligonucleotides were phosphorylated at 37 "C for 30 min using polynucleotide kinase, then heated to 100 "C in the presence of the template for 3 min, allowed to anneal at 65 "C for 1 h and then for 30 min a t 37 "C. Extension with the DNA polymerase I (Klenow fragment) was performed at 37 "C for 90 min in the presence of ligase. T h e hybrid molecules were used to transform TG1 cells. However, when oligonucleotides a or b were used, hybrid molecules were digested with BglII restriction enzyme prior to transformation. Screening of transformants was done either by gel analysis of plasmid DNA (for the creation of restriction sites), by determining the size of the EcoRV/XbaI fragment, or by colony hybridization with 32P-labeled oligonucleotide ( [Y-~'P]ATP, 5,000 mCi/mmol, Amersham Corp.) (Davis et al., 1980). The deletions or mutations were confirmed by sequencing the preproregion from the XbaI site to the ATG codon according to Sanger et al. (1977). The mutagenized cassette was then introduced into the Y. lipolytica vector pINA310 by replacing either the NheIIXbaI or MluIIXbaI fragment by the mutagenized fragment. The resulting plasmids containing deletions or mutations are listed in Table I. Their corresponding fusions sites or amino acid mutations are shown in Fig. 2B, and the encoded proteins are represented schematically in Fig. 3.
Oligomer a was used to delete the entire proregion and create the De12 deletion which fused Pro3' (the last amino acid of the X-Ala/X-Pro segment) to AlaI5* (first amino acid of the mature form). The resulting plasmid was called pINA310-2. During this mutagenesis, we fortuitously obtained a deletion of the pd-pro11 region and an insertion of 14 amino acids (Dell deletion) which resulted in plasmid pINA310-1.
The resulting plasmid was called pINA310-3.
Oligomer c was used to destroy the glycosylation site in the proregion by replacing the threonine with a valine (DelG mutation). The resulting plasmid was called pINA317.
Oligomer d was used to inactivate AEP by the addition of four amino acids in a region close to the active site. Far this purpose, plasmid pINA310 was linearized by cleavage with BamHI. Ligation was performed in the presence of the phosphorylated oligomer d.
Plasmids were cut with BamHI prior to transformation to avoid ligation without linker. The presence of the linker was checked by digesting with the restriction endonuclease PuuI. The resulting mutation (Ll) was sequenced using oligomer e, which corresponds to the sequence near the BglII site from bp 1045 to 1063 (17). The resulting plasmid was called pINA320.
DNA Preparation-Plasmid DNA from E. coli grown in LB medium was prepared by the Holmes and Quigley (1981) procedure. Chromosomal DNA from Y. lipolytica was prepared according to Hoffman and Winston (1987). DNA fragments were separated by electrophoresis in agarose gels and purified using a Geneclean kit (Ozyme).
Transformation-Transformation of E. coli was performed as described by Dagert and Ehrlich (1979). Y. lipolytica integrative transformation was performed as described previously (Gaillardin and Ribet, 1987). Plasmid DNA was cut with NheI restriction endonuclease before transformation to target integration at the XPRB locus. To check the chromosomal structure of the XPR2 locus, we prepared chromosomal DNA, digested it with EcoRV, separated DNA fragments by gel electrophoresis, transferred them to Hybond N nylon (Amersham Corp.), and hybridized them to the 1.1-kilobase EcoRV fragment from pINA310 according to Davis et al. (1980). Protein Analysis-Precultures grown in the noninducing, selective medium YNB were used to inoculate 20 ml of the inducing medium YPDm (Nicaud et al., 1989a) at an initial density of 0.1 (OD,,) in 250-ml flasks. Cells were grown under aeration a t 28 "C, and samples were taken in the late exponential phase between 24 and 32 h after inoculation at a density of 15-20 units (OD,,). Cells and supernatant fractions were separated by centrifugation a t 12,000 X g for 10 min at 4 "C. Cells were washed twice with cold homogenization buffer (50 mM Tris-HC1, pH 6.8, 1 mM MgC12, 1 mM CaCl,), resuspended at the original cell density in this buffer together with 0.3 g of acid-washed glass beads (0.45 mm), and disrupted using a Braun homogenizer. Samples were assayed for AEP activity or were analyzed for protein content after treatment with phenylmethylsulfonyl fluoride (2 mM) and aprotinin (0.1%). All protein samples were denatured in an equal volume of 2 X sample buffer (100 mM Tris-HC1, p H 6.8, 2% 2mercaptoethanol, 4% SDS, 20% glycerol, 0.01% bromphenol blue) for 5 min at 100 "C. Insoluble material was discarded by centrifugation for 5 min at 12,000 X g a t 4 "C. For partial fractionation studies, disrupted cells were separated into soluble and insoluble (membrane) fractions by centrifugation a t 14,000 X g for 30 min prior to heating at 100 "C with sample buffer.
Tunicamycin and Endoglycosidase H Treatment-To prevent protein glycosylation, protease synthesis was derepressed as described under "Protein Analysis," and tunicamycin (10 mg/ml, Sigma) was added 5 h before culture samples were withdrawn. Protein deglycosylation was carried out with 1 milliunit of endoglycosidase H (Genzyme) for material derived from cells corresponding to 1 ml of culture at OD,, = 2.0, in sodium citrate buffer (50 mM, pH 5.5) containing phenylmethylsulfonyl fluoride (1 mM) and sodium azide (5 mM) for 24 h a t 37 "C. I n Vitro Transcription and Translation-Plasmid pINA392 and its derivatives were linearized downstream from the XPR2 gene by digestion with the restriction endonuclease EcoRI. One microgram of linearized DNA was transcribed and capped for 2 h a t 40 "C using SP6 RNA polymerase as described by the suppliers (Promega Biotec). Rabbit reticulocyte lysate (Amersham Corp.) was used to translate 1 pg of capped transcripts in the presence of 20 mCi of ["Slmethionine (Amersham Corp.) for 1 h a t 30 "C as described by the suppliers (Amersham Corp.). Aliquots (10 pl) were removed and mixed with 40 ml of sample buffer 1 X. Samples were heated a t 95 "C for 3 min. Aliquots (10 pl) were applied to the SDS-polyacrylamide gels.
Membrane Preparation-The membrane preparation described in Feldman et al. (1987) was adapted for our yeast. Cells were grown to late exponential phase (OD,,,, of 15) a t 28 "C in 200 ml of complete medium (YPDm) buffered with 50 mM sodium phosphate buffer, pH 6.8. Cells were first converted to spheroplasts by treatment with cytohelicase as described in Gaillardin et al. (1985). Spheroplasts were then lyzed gently in 8 ml of cold lysis buffer (20 mM Hepes/KOH, pH 7.5, 100 mM potassium acetate, 2 mM magnesium acetate, 2 mM dithiothreitol). Lysate was homogenized with 10 strokes in a Potter homogenizer and centrifuged a t 2 "C a t 900 X g for 5 min.
The supernatant, containing the membranes, was recovered for protease treatment. To a 200-ml aliquot Triton X-100 (0.1% final concentration), proteinase K (0.5 mg/ml), or both were added, and samples were incubated on ice for 1 h. Protease treatment was stopped by adding trichloroacetic acid (20% final concentration). Pellets were washed once with acetone and solubilized in 100 ml of sample buffer X 1 neutralized with 125 mM Tris base, by extensive vortexing and by heating at 100 "C for 5 min. Proteins (15 ml) were applied to the SDS-polyacrylamide gels and analyzed by immunoblotting as described under "Protein Analysis."

RESULTS
Synthesis of Alkaline Extracellular Protease-We have reported previously on the cloning of the XPRB gene coding for AEP. To study the role of the proregion of AEP we developed plasmids pINA303 for mutagenesis and pINA310 to introduce the mutation into the Y. lipolytica chromosome (see "Materials and Methods").
Since we expected that some of the constructs would have a trans effect on the wild-type protease itself, these mutations were studied in both Xpr+(JM12) and Xpr-(JM23) backgrounds (see Table I). Strain JM23 did not produce AEP but did so once transformed with a plasmid carrying a functional XPRB gene such as pINA310. In order to compare AEP secretion and production, we have verified that the regulation and the level of AEP production were identical in strain JM12 and in JM23WT (JM23 transformed with pINA310) (data not shown).
Construction of Large Deletions Affecting the Proregion of AEP-For our studies on the role of the proregion, we constructed several deletions in the corresponding region of XPR2. In the derivative designated Dell, the region between Lys-Arg" and Ala'58, the first amino acid of mature AEP is deleted. In De12, the complete proregion is absent, but the X-Ala/X-Pro stretch that may separate the pre-and proregion is retained. De12 also retains 18 amino acids downstream of the putative signal peptide processing site Ala15 (Matoba et  Gly Gly Gln Gly Ser Arg Ser Arg Ser Asn   FIG. 2. A , DNA sequence of the synthetic oligonucleotides used for site-specific mutagenesis, linker insertion, and sequence analysis. B, amino acid sequence of the fusion sites in the deletions or mutation constructs. 1 , pINA310-1 (fusion pre-proIIIAEP); 2, pINA310-2 (fusion pre-AEP); 3, pINA310-3 (fusion preproI-AEP); 4, pINA317, mutagenesis of the glycosylation site in the pro11 region; 5, pINA 320, linker insertion in the region coding of the mature AEP. Numbers refer to amino acid positions in wild-type AEP (Davidow et al., 1987). al., 1988) such that cleavage should not be affected. During these experiments we obtained a third clone corresponding to a pre-proIIIAEP fusion (De13). These three deletions were introduced into pINA310, and the corresponding plasmids (see Table I and Fig. 3) were used to transform both JM12 and JM23 (both strains are Leu-). LEU+ transformants of both strains were obtained at high frequency.
The chromosomal organization of the transformants was confirmed by Southern blotting as described under "Materials and Methods." All transformants with plasmids pINA310, pINA310-1, pINA310-2, and pINA310-3 resulted from single insertion and presented the expected pattern, i.e. a 1,349-, 1,104-, 977-, or 1,057-bp EcoRV fragment (respectively) in addition to the 4,365-bp EcoRV fragment corresponding to the disrupted gene in strain JM23 (data not shown). Similar results were obtained after transformation of the strain JM12 (the EcoRV fragment corresponding to the XPRB gene in JM12 has 1,349 bp).
Analysis of A E P Produced by the Various Constructs-AEP production by the above strains was tested on SKM plates. We showed previously that the size of the hydrolysis zone is related directly to gene copy number (Nicaud et al., 1989b). All JM12 transformants produced hydrolysis zones similar to those produced by the nontransformed strain except when transformed with pINA310 (two complete copies of XPR2 wild-type gene), in which case a larger halo was observed. None of the JM23 transformants formed halos except in the control experiment in which JM23 was transformed with Plasmids carrying the wild-type XPR2 gene and the mutated XPR2 constructs and schematic representation of their coding region. The structure of the wild-type AEP precursor (Matoba et al., 1988;Matoba and Ogrydziak, 1989) is shown at the top (WT). The signal sequence is indicated as a black box ( S S ) followed by vertical lines corresponding to the X-Ala/X-Pro stretch. The proregion is symbolized by an open rectangle divided in three parts (proI, proII, and proIII) according to Matoba et al. (1988), and the mature AEP sequence, by diagonal lines. The glycosylation site is indicated by a rhombus with its sequence underneath in lightface capital letters. A circle is used for the mutated glycosylation site. The modified sequences for the glycosylation site or for the linker insertion are indicated in bold capital letters. The abbreviations used are: KR, Lys-Arg sequence; AA, amino acid; 8, BamHI restriction site. The exact amino acid sequences of the fusion sites are shown in Fig. 2. pINA310 carrying the wild-type XPRB gene.
To examine the nature of the AEP produced by the strains carrying the deletions, supernatant and cell extracts from induced cell cultures were analyzed by immunoblotting using anti-AEP antibodies (see Fig. 4). No immunoreactive material was detected either in the supernatants or in the cell extracts of strain JM23 (disrupted for the XPR2 gene), and only the 32-kDa mature form was detected in both fractions from the strain JM12 containing the wild-type XPR2 gene or in the strain JM23 transformed with plasmid pINA310. No immunoreactive material was detected in the supernatant from other JM23 derivatives (Fig. 4A), indicating that the absence of a clearing zone on SKM plates was not due to the secretion of inactive material but to the absence of secreted protein.
Immunoreactive material with apparent molecular masses of 40,34, and 38 kDa was detected in cell extracts of Dell, Del% and De13 transformants, respectively.
The effect of the accumulation of the AEP derivatives on the secretion of the wild-type AEP was determined in derivatives of strain JM12. Production of the Dell, De12, and De13 AEP derivatives had no effect on extracellular proteins (Fig.  4B) or on the amount of AEP detected extracellularly. However, the cell extracts of the transformants carrying the deletions contained the same immunoreactive materials as the derivatives of strain JM23 together with an additional 55-kDa polypeptide. This polypeptide was not observed in strains containing one copy of the XPRB gene (e.g. JM12 or Role of Prosequence in Secretion of Yeast Protease JM23WT) but appeared in strains such as JM12WT, which carries two copies of XPRS. JM12 also produced twice the normal level of AEP.
Mutation Affecting the Glycosylutwn Site in the ProII Region and Its Effect on AEP Secretion-Glycosylation is often important for protein secretion. For example, tunicamycintreated cells are impaired in the secretion of a factor (Julius et al., 1984), invertase (Ferro-Novick et al., 1984), and acid phosphatase (Mizunaga et al., 1988). In some cases, inhibition was reported to be temperature dependent, being more pronounced at 30 "C than at 20 or 25 "C (Mizunaga et al., 1988;Ferro-Novick et al., 1984). To assess if the glycosylation site located in the pro11 region is important for AEP secretion, we destroyed the Asn-X-Thr recognition site by oligomutagenesis which changed the ACT codon coding for threonine to GTC coding for valine (see Figs. 2B and 3). This glycosylationminus mutation (DelG) was introduced into Y. lipolytica strain JM23 using pINA317, and the transformants were tested on SKM plates at both 18 and 28 "C. This latter temperature is close to the upper limit for Y. lipolytica growth. All transformants produced clear hydrolysis halos on SKM at 18 "C, whereas no halos were observed at 28 "C. This may have been because strains producing DelG either synthesized and cell extracts ( B ) were isolated from cells grown at 28 or at 18 "C and were separated on a 12% gel. an inactive protease due to misfolding induced by the DelG mutation or were unable to secrete AEP. To distinguish between these two possibilities we examined the accumulation of AEP by the strain carrying the DelG mutation. A strain producing a thermosensitive AEP encoded by pINA320, which has a linker insertion (Ll) in XPR2 close to the region encoding the active site of the enzyme2 (see "Materials and Methods"), was used as a control in these studies. Precultures were grown at 18 "C and then induced at 18 or 28 "C. Culture samples were collected between 24 and 32 h when grown at 28 "C or between 30 and 38 h when grown at 18 "C. Supernatant and cell fractions were used to determine AEP activities and for protein analysis by immunoblotting.
As shown in Fig. 5A, identical levels of extracellular mature AEP were produced at 18 "C by the three strains (wild type, DelG, and Ll). AEP produced by strain L1 migrated slightly more slowly than normal AEP, possibly because two of the four amino acids added by the linker L1 (Ser-Arg-Ser-Arg) are positively charged. Strains carrying DelG and L1 produced lower levels of extracellular mature AEP at 28 "C than did the wild-type strain. For strain L1 the lower level of extracellular mature AEP was due to an instability of the protein. Indeed L1 and wild-type AEP present similar half-lives at 18 "C, but L1 is degraded rapidly at 28 "C (data not shown). Cells grown at 18 "C contained mainly the 32-kDa polypeptide and small amounts of a 55-kDa polypeptide for wild-type and L1 strains or a 53-kDa polypeptide for strain DelG. All three strains were observed occasionally to contain small amounts of two other proteins (44 and 36 kDa, not visible on Fig. 5B). The patterns of intracellular proteins produced by the wildtype strain and L1 were identical irrespective of the growth temperature. In contrast, the DelG mutant contained large amounts of a 53-kDa polypeptide precursor (which also fractionates with membranes; see "Materials and Methods") when grown at 28 "C. The appearance of this precursor correlates with the decrease in the amounts of intracellular and extracellular mature AEP. These cells also contained increased amounts of immunoreactive material that may have resulted from degradation of accumulated precursors.
Characterization of AEP Produced by the Various Mutants-The molecular masses of the precursors that result from the translation of Dell, De12, De13, and DelG XPR2 derivatives were calculated to be 38,099,33,620,36,541,and 46,901 Da,respectively. Signal sequence cleavage after Ala15 would lower these figures by 1.55 kDa; further diaminopeptidase processing would lower them by an additional 1.55 kDa.
The observed values for the mutated precursors, 40, 34, 38, and 53 kDa (see preceding sections), are all higher than predicted for precursors having undergone signal sequence and diaminopeptidase processing (see also Matoba et al., 1988). Although there is no potential glycosylation site remaining in the proregion of the various constructs (deleted for Dell, De12, De13, or mutated for DelG), there are two such sites (AsnZg2 and AS^^^') in the mature part of the protein.
Neither of these sites is normally modified in the wild-type protein. Since the precursors accumulated in the cells, they could misfold in such a way that these glycosylation sites would be glycosylated. We therefore compared the electrophoretic mobility of precursors accumulated in JM12 and JM12 transformants grown in the presence or absence of tunicamycin. The Xpr' precursors were used as internal controls. As shown in Fig. 6, the accumulated mutant precursors had the same electrophoretic mobility in both cases whereas the wild-type (glycosylated) 55-kDa precursor was converted to a 53-kDa protein. Similar results were obtained when cell extracts were treated with endoglycosidase H (data not shown), indicating that Dell, De12, De13, and DelG polypeptides were not glycosylated. Sometimes we observed a faint band above the accumulated precursors in the presence of tunicamycin (see Fig. 6) which is not observed after endoglycosidase H treatment (data not shown). This may result from tunica- mycin-dependent inhibition of signal peptidase. Because higher mobility could not be explained by glycosylation, we compared the migration on acrylamide gels of in vivo products and polypeptides produced in vitro. For the latter, we used pINA392 in which XPR2 is under the control of SP6 promoter and constructed derivatives containing the Dell, De12, De13, and DelG mutations. Fig. 7A shows that the precursor produced by strain JM12 containing De12 migrated faster than the in vitro product, suggesting that the signal  (+) or not (-) with tunicamycin, and JM12 transformant containing De12 (De12, vu) with their in vitro conterparts (ut). Proteins were separated on a 12% gel. B, JMl2WT in uiuo precursor and in vitro product from A were resolved on a 7.5% gel. C, comparison of electrophoretic mobility of precursors produced by strain JM12 containing wild-type and JM23 containing DelG, with precursors produced by an xpr6 mutant. Proteins were separated on a 7.5% gel. WT, proteins from JMl2WT grown without (-) or with (+) tunicamycin (exposed 1 day); DelG, proteins from JM23DelG in vivo (uv) and its in vitro counterpart ( u t ) ; xpr6, xpr6-secreted precursor before (-) or after (+) treatment with endoglycosidase H (exposed 7 days). Proteins were transferred to nitrocellulose, developed with rabbit antibodies, and detected with 35S-labeled donkey anti-rabbit antibodies. mutants. Immunoblot analysis of proteins produced by wild-type ( W T ) in strain JM12 and by the deletion mutants in strain JM23. Membranes were prepared as described under "Materials and Methods." They were treated with proteinase K (0.5 mg/ml) ( P k ) in the absence (-) or presence (+) of Triton X-100 (0.1%) (Tx). Proteins were separated by SDS-polyacrylamide gel electrophoresis on a 10% gel, transferred to nitrocellulose, and developed with anti-AEP polyclonal antibodies.

3789
sequence was indeed cleaved. Similar results were obtained with Dell and De13 (data not shown). The abnormally high apparent molecular masses of these precursors may thus result from the presence of the 10 X-Ala/X-Pro sequences (Matoba et al., 1988;Matoba and Ogrydziak, 1989).
To identify the precursor accumulated in the DelG construct at 28 "C we compared the electrophoretic mobilities of the in vivo DelG product, the in vitro translated DelG polypeptide, the wild-type 55-kDa precursor, and the product secreted by an xpr6 mutant. The in vitro product corresponds to the complete precursor, i.e. pre-(X-Ala/X-Pro)"-pro-AEP.
The 55-kDa precursor may be the glycosylated form of this precursor or of (X-Ala/X-Pro)"-pro-AEP (Matoba et al., 1988). Glycosylated pro-AEP is secreted by the xpr6 mutant (Matoba and Ogrydziak, 1989). The corresponding deglycosylated form was obtained by treatment of protein from the supernatant from a culture of this strain with endoglycosidase H.3 Fig. 7 B shows that the deglycosylated 55-kDa wild-type precursor was smaller than the wild-type precursor synthesized in vitro. Similar results were obtained with the DelG construct as shown in Fig. 7C; the DelG protein, which has the same electrophoretic mobility as the deglycosylated 55-kDa wild-type precursor, was smaller than the DelG product synthesized in vitro and about 3 kDa larger than the 52-kDa xpr6 deglycosylated precursor. This suggests strongly that DelG and the 55-kDa wild-type precursor have undergone signal peptide cleavage but not diaminopeptidase processing.

Localization of A E P Produced by the Various Mutants-We
have shown that signal sequence cleavage occurred in the AEP derivatives. However, we could not determine if the main part of the protein was still in the cytoplasm or was fully translocated into the ER lumen. As the deletions did not fortuitously generate a new segment of hydrophobic amino acids which could function as a stop transfer signal, we predicted that the mutated AEP precursors were fully translocated to the ER lumen. T o confirm this hypothesis, we prepared membrane fractions as described under "Materials and Methods" and tested the protease sensitivity of these precursors. As shown in Fig. 8, all Del precursors as well as the 55-kDa wild-type precursor were protease resistant in the absence of detergent and became sensitive in its presence. Proteinase K activity was checked by its action on the total protein pattern (Coomassie Blue staining; data not shown). We observed that mature AEP is resistant to proteinase K treatment, which may be due to the fact that AEP is a serine protease, as is proteinase K. These results confirm that Del precursors have entered the secretory pathway.

DISCUSSION
In this study, we analyzed the secretion of alkaline extracellular protease by the yeast Y. lipolytica. We have identified the precursors and have shown that mutations in the proregion affect secretion drastically.
Secretion of AEP from Y. lipolytica is a rapid process that involves the synthesis of a pre-(X-Ala/X-Pro)"-pro-AEP polypeptide. The first precursor seen by pulse labeling is a 55-kDa polypeptide that contains about 2 kilodaltons of Nlinked carbohydrate. According to Matoba et ul. (1988) and