Mutational analysis of the vacuolar sorting signal of procarboxypeptidase Y in yeast shows a low requirement for sequence conservation.

The core of the vacuolar targeting signal of yeast carboxypeptidase Y (CPY) is recognized by the receptor Vps10p and consists of four contiguous amino acid residues, Gln-Arg-Pro-Leu, near the amino terminus of the propeptide (Valls, L. A., Winther, J. R., and Stevens, T. H.(1990) J. Cell Biol. 111, 361-368; Marcusson, E. G., Horazdovsky, B. F., Cereghino, J. L, Gharakhanian, E., and Emr, S. D.(1994) Cell 77, 579-586). In order to determine the sequence specificity of the interaction with the sorting receptor, substitutions were introduced into this part of the propeptide by semirandom site-directed mutagenesis. The efficiency of vacuolar sorting by the mutants was determined by immunoprecipitation of CPY from pulse-labeled cells. It was found that amino acid residues Gln and Leu were the most important ones. While it appears that Gln is essential for proper function, Leu can be exchanged with the other hydrophobic amino acid residues, isoleucine, valine, and phenylalanine. Tolerance toward various substitutions for Arg is fairly high, while substitution of Pro for uncharged amino acid residues also resulted in only weak missorting. In addition to the low requirement for sequence conservation, the position of the targeting element relative to the amino terminus of the propeptide was analyzed and found not to be critical.

Eukaryotic cells consist of a number of compartments, each with a unique set of proteins, allowing the cell to regulate spatially its catalytic activities. This requires an efficient system for directing each newly synthesized protein to its correct functional location within the cell.
The targeting signals are ultimately encoded in the protein sequence, and these signals take many forms. Signals for protein import into mitochondria, for example, lack any obvious sequence homology but tend to form amphipatic ␣-helices or ␤-sheets, which allow them to bind to the mitochondrial membrane, whereafter specific recognition occurs (Hartl andNeupert, 1990, Baker andSchatz, 1991). Similarly, NH 2 -terminal leader peptides, which direct proteins into the ER 1 of eukaryotic cells and function in translocation through the plasma membrane in prokaryotes, also lack highly specific primary sequence determinants. Here more general physical properties such as hydrophobicity and charge are important for function (von Heijne, 1985). Conversely, targeting of some proteins to peroxisomes occurs via a fairly well defined signal consisting of three amino acid residues at the COOH terminus. The tripeptide SKL at the COOH terminus of firefly luciferase was shown to direct this protein into peroxisomes (Gould et al., 1987). Various substitutions in this tripeptide signal have been made, some of which are functional in peroxisomal import (Subramani, 1992).
The default route for soluble proteins in the secretory pathway is to the cell surface, i.e. they will be secreted if they lack any specific signals. Soluble ER proteins in Saccharomyces cerevisiae contain the COOH-terminal tetrapeptide HDEL, which allows them to be retained in the ER (Pelham, 1989). This is an example of a very specific signal, and recycling of HDEL-tagged proteins to the ER in yeast is mediated by a receptor, Erd2p (Lewis et al., 1990).
In mammalian cells, lysosomal targeting is mediated by the mannose 6-phosphate receptor and, ultimately, by the recognition of the structural elements of the lysosomal proteins by the phosphotransferase (Baranski et al., 1990). Although the yeast vacuole is equivalent to the lysosome of higher eukaryotes in many respects, no mannose 6-phosphate-mediated sorting mechanism has been found in yeast. The soluble vacuolar hydrolases proteinase A and carboxypeptidase Y (CPY) of S. cerevisiae contain vacuolar targeting information in their NH 2terminal propeptides (Johnson et al., 1987, Valls et al., 1987. The sorting of proteinase A and CPY was suggested to be receptor-mediated since overexpression resulted in mislocalization . Indeed, the sorting receptor for CPY was recently identified as the VPS10 gene product (Marcusson et al., 1994). The information necessary and sufficient for vacuolar localization of CPY is encoded by a sequence containing the tetrapeptide Gln 24 -Arg-Pro-Leu 27 near the NH 2 terminus of the precursor (Valls et al., 1990). The information for targeting of proteins to the vacuole of plant cells may also be found in short peptide domains (Chrispeels and Raikhel, 1992), but QRPL is the best characterized vacuolar targeting signal so far and is therefore often used as a basis for defining and comparing with other potential vacuolar targeting signals. None of the other known vacuolar proteins in yeast contain a QRPL sequence. However, since only mutations in the QRPL signal abolishing its function have been defined, it is difficult to deduce a consensus sequence for comparison with other systems.
Using degenerate oligonucleotides we have performed extensive mutagenesis on the part of the PRC1 gene encoding the QRPL signal and determined the efficiencies of sorting by immunoprecipitation of newly synthesized CPY in pulse-chase experiments. We found that many mutant forms of the QRPL signal are indeed able to sort pro-CPY correctly to the vacuole. Furthermore, we analyzed the importance of the distance of the QRPL signal from the NH 2 terminus of the propeptide and found that insertion of up to four amino acid residues did not affect the sorting efficiency seriously. These data together suggest a surprisingly low level of requirement of sequence conservation for function of the recognition of the sorting receptor. EXPERIMENTAL PROCEDURES Strains, Media, and Materials-Yeast strain JHRY20 -2C⌬3 (MATa ura3-52 leu2-3,112 his3-⌬200 ⌬prc1::HIS3;Valls et al. (1987)) was used for expression of wild-type and mutant PRC1 alleles. Strain SEY6210 (MAT␣ ura3-52 leu2-3,112 his3-⌬200 trp1-⌬901 lys2-108 suc2-⌬9;Robinson et al. (1988)) was used for expressing the CPYinvertase fusion. Escherichia coli cultures were grown in LB media (Sambrook et al., 1989). Yeast cultures were grown in standard YPD and SC media (Sherman, 1991). Restriction endonucleases, DNA polymerase I (Klenow fragment), and T4 DNA ligase were from Promega. Pfu DNA polymerase was from Stratagene. Deoxyribonucleotides were from Boehringer Mannheim. Zymolyase was from Seikagaku (Japan). A mixture containing 35 S-labeled methionine and cysteine (NEG-072 EXPRE 35 S 35 S from DuPont) was used for in vivo protein labeling. Oligonucleotides were synthesized on an Applied Biosystems 380A DNA synthesizer according to instructions of the manufacturer. Oligonucleotides were desalted on Sephadex G-25 columns equilibrated with 10 mM Tris-HCl, 1 mM EDTA, pH 7.5. Polymerase chain reaction was performed using a DNA thermal cycler from Perkin-Elmer.
Plasmid Constructions-All subcloning steps and transformations of E. coli and yeast were carried out using standard procedures (Sambrook et al., 1989;Ito et al., 1983). Plasmid pJW1445 was made from pLV9 (containing PRC1 on a URA3 CEN vector; Valls et al., 1987) by destroying the XhoI site with DNA polymerase I, Klenow fragment, and dNTPs. A polymerase chain reaction fragment was generated extending from the ClaI site upstream of PRC1 to the XbaI site in the PRC1 proregion. The downstream primer contained three mismatches resulting in a mutation in codon 19 from AAG to CGA, creating a XhoI site. This polymerase chain reaction fragment was cloned as an EcoRI XbaI fragment into pJW1445, resulting in the deletion of the sequence between ClaI and EcoRI. This plasmid was called pFV1. DNA sequencing showed that there were no errors in the part of the PRC1 open reading frame amplified by polymerase chain reaction. A 2.5-kilobase XhoI-XbaI fragment from the PEP4 gene on pPA2 (Rothman et al., 1987) was cloned as a stuffer fragment into pFV1, resulting in pFV2. pFV2 was used for further mutagenesis. Insertion of the stuffer fragment eliminated the possibility of subsequent inadvertent recovery and expression of wild-type PRC1. pFV2 was digested with XbaI and XhoI and purified from agarose gel. Subsequently, mutagenic oligonucleotides for both strands were synthesized, annealed, and ligated without phosphorylation of the oligonucleotide insert. After ligation, products were purified from the unligated oligonucleotides using Wizard TM polymerase chain reaction preps from Promega. The purified ligation products were heated to 65°C and slowly cooled down to room temperature before transformation of E. coli. This procedure ensures that the number of plasmids having multiple inserts is extremely low. The site-directed mutagenesis was done in different ways. For introduction of random sequence at any single codon in the five positions 23-27, a mixture of oligonucleotides was synthesized using all four nucleotides at the three positions in each of the codons. The five mixtures were ligated individually and introduced into E. coli. At least 500 individual transformants were obtained from each ligation, and the plasmid DNA isolated from these transformants resulted in five libraries, each theoretically containing PRC1 alleles with all possible mutations for each position. Plasmid DNA was reintroduced into E. coli, either after isolation from yeast or directly from the original pool. The latter step was carried out to ensure homogeneity of the propagated mutant plasmid propagated for sequencing.
Other libraries were created by insertion of oligonucleotide mixtures containing four extra codons inserted between positions 22 and 23. In order to eliminate stop codons but still obtain a fairly random collection of mutants, each of the additional new codons started with an adenine followed by two random nucleotides. This oligonucleotide mixture was ligated, and total plasmid DNA was isolated from a pool of ϳ150 individual E. coli transformants. Plasmid DNA was reintroduced into E. coli, and plasmid DNA from five of the transformants was used for sequencing. Oligonucleotides with several desired mutations or insertions were synthesized individually and cloned as described above. Mutations were verified by DNA sequencing.
Plasmid pJW1588 contains an allele of PRC1, PRC1-2112, with the mutations T15A and N368Q on a vector similar to pLV9 . These mutations remove the first and the last glycosylation site, and the resulting mutant protein is called aBCd-CPY. A SalI-PvuII fragment containing this mutant allele was subcloned into pRS315 (a LEU2 CEN vector; Sikorski and Hieter (1989)), resulting in pFV410.
The plasmid pFV127, containing the fusion between PRC1 and SUC2, was constructed as follows: a BamHI-SmaI fragment containing the first 156 codons of the PRC1 gene was subcloned in pSEYC306 (Johnson et al., 1987), resulting in pFV50. A SalI fragment of pSEYC306 was subcloned in pRS315 opened with SalI and XhoI, so that the 5Ј end of the SUC2 gene was nearest to the remaining SalI site, resulting in pFV99. Subsequently, a SmaI-SalI fragment from pFV50 was subcloned in pFV99 opened with SmaI and SalI.
DNA Sequencing-Template DNA was purified using the QIAGEN Tip20 plasmid kit. DNA sequencing was performed using a Taq DyeDeoxy TM terminator cycle sequencing kit and an Applied Biosystems 373A DNA sequencer according to the instructions of the manufacturer.
Plate Overlay Assays-CPY activity plate overlay assays were done using N-acetyl-DL-phenylalanyl-␤-naphtyl ester (Jones, 1991). For detection of extracellular invertase activity yeast transformants were replica-plated onto SC containing 2% fructose instead of glucose and incubated overnight at 30°C. Subsequently a plate overlay was used as described by Horazdovsky et al. (1994).
Immunoblotting-Colony immunoblotting was carried out as described by Roberts et al. (1991). However, the antibodies against CPY had been raised against natively glycosylated CPY. Therefore, to prevent carbohydrate-directed binding to other yeast proteins, the immunoblots were incubated with 1% periodate before incubation with the primary antibody as described earlier (Sørensen et al., 1994).
Pulse Labeling and Immunoprecipitation-Pulse labeling and immunoprecipitations were performed as described earlier  except for the following modifications. Labeling was carried out in SC media without sulfate and methionine, and cultures were labeled by the addition of 50 Ci of EXPRE 35 S 35 S. Spheroplasts were lysed in 1 ml of 10 mM Tris-HCl, 0.1% Triton X-100, and 2 mM EDTA, pH 8.0, to give the intracellular material. After SDS-polyacrylamide gel electrophoresis, gels were exposed to Molecular Dynamics PhosphorImager screens, and screens were scanned with a Molecular Dynamics PhosphorImager. Radioactivity in CPY-specific bands was detected and quantitated with ImageQuant 3.2 software.

Mutations at
Positions 23-28 -In the initial screen we wished to determine the degree of sequence conservation required for efficient recognition of the vacuolar sorting determinant of pro-CPY by its receptor, Vps10p. Thus, various mutant plasmid libraries were constructed covering the region containing the QRPL sequence using a cassette mutagenesis procedure. PRC1 mutant alleles were expressed from centromere plasmids in a ⌬prc1 yeast strain. To facilitate the mutagenesis, a novel XhoI site was introduced into the gene (Fig. 1). This XhoI site could be used in combination with the naturally occurring XbaI site just downstream of QRPL to replace the wild-type sequence with synthetic oligonucleotides. The mutation that created the XhoI site led to a K19R substitution in the presequence of CPY. As expected, this conservative change did not affect translocation of the precursor or presequence cleavage. Furthermore, since the mutant residue was removed by the latter process, there were no effects on further transport or processing. As an initial approach we wished to isolate mutants in the QRPL signal that were proficient in sorting. Thus, libraries of mutant plasmids were constructed at each of the positions Leu 23 through Leu 27 (pools Xaa 23 through Xaa 27 ) and more than 500 individual E. coli transformants were obtained from each oligonucleotide mixture. Cells from all transformant colonies of each ligation were pooled, and plasmid DNA was isolated from the pools and used for transformation of a ⌬prc1 yeast strain. Mutants were characterized according to two criteria: 1) the presence of intracellular CPY activity, determined by a plate overlay assay, and 2) secretion of CPY, detected by colony immunoblots. Mutants strongly deficient in sorting, as well as non-sense mutants, were not expected to give intracellular activity, while missorting mutants stain positive in the immunoblot. From each collection, 100 transformants were replica-plated and tested according to both criteria. This gave an initial indication as to the specificity of the sorting signal at each position, as well as the frequency of nonsense and frameshift mutations that would lead to absence of both activity and antigen. The number of colonies without CPY production varied between 5 and 7% in all collections, which corresponds to the expected frequency of stop codons. The most important result of this initial screen was that only two of the 100 yeast Xaa 24 transformants showed a wild-type phenotype. Screening of the Xaa 25 , Xaa 26 , and Xaa 27 pools showed a higher frequency (5-15%) of plasmids that gave rise to a wild-type or quasi-wildtype phenotype, while all plasmids from the Xaa 23 pool appeared to give a wild-type phenotype. Sequencing of the plasmid DNA from the two positive clones of the Xaa 24 pool showed that the wild-type phenotype was in both due to a Gln codon.
All plasmids giving what appeared to be a wild-type phenotype in the plate assays were sequenced, and to get a broader view of the specificity, several randomly chosen plasmids were sequenced from each library. The sorting efficiency of each mutant was determined quantitatively by 35 S pulse-chase labeling and immunoprecipitation. Thus, exponentially growing cultures were labeled with [ 35 S]methionine/cysteine, and CPY antigens were immunoprecipitated from intracellular and extracellular fractions. Samples were subsequently subjected to SDS-polyacrylamide gel electrophoresis, and autoradiograms were prepared using a PhosphorImager. Fig. 2 shows representative immunoprecipitations of mutants with different phenotypes. CPY that is correctly localized to the vacuole is processed to the mature form of 61 kDa. However, mislocalized pro-CPY is not matured and is found extracellularly as the precursor form of 69 kDa. Wild-type pro-CPY is sorted to the vacuole and converted to its mature form. However, a small fraction, around 5%, is mislocalized and is seen extracellularly as the pro-CPY precursor form (Fig. 2, lane 2). This level of missorting of the wild type is normal Valls et al., 1987Valls et al., , 1990. Appearance of CPY antigen extracellularly is not due to cell lysis, since no mature CPY is found extracellularly. Radioactivity in the CPY-specific bands was detected and quantified using storage phosphor technology (Johnston et al., 1990). The ratios of intra-and extracellular CPY were determined by quantification of at least two independent immunoprecipitations. The amounts of intracellular CPY relative to the total amounts synthesized are depicted in Fig. 3 together with data from an earlier study (Valls et al., 1990). Our determinations of the missorting phenotype of mutations Q24S and R25G are in good agreement with those obtained previously by Valls et al. (1990). The differences, 7% and 5% for Q24S and R25G, respectively, reflect the internal variation found in our own measurements. In addition to those clones selected for analysis on the basis of wild-type sorting in the plate screen and randomly selected clones, we constructed some directed mutants that were judged to supplement the collection well.
All the mutations in Gln 24 lead to more than 50% missorting, even the conservative Q24N mutation, which was introduced by directed mutagenesis. The combined results from the immunoprecipitation and the screening of colonies on plates strongly suggest that only glutamine will function well in sorting at this position.
In general, mutations in Arg 25 do not appear to have a strong effect on sorting. Most severe is the R25D mutation, which results in secretion of 48% of the newly synthesized pro-CPY. All the other mutations have a sorting efficiency between 74 and 95%. This is in good agreement with the large number of mutations (Asn, Asp, Glu, Gly, Lys, and Val) scored as wildtype or quasi-wild-type in the initial plate screen.
Mutations in Pro 26 give rise to a wide spectrum of phenotypes. P26R leads to missorting of almost all of the newly synthesized pro-CPY, whereas the P26S mutation still localizes 73% of the pro-CPY to the vacuole. Mutants with hydrophobic amino acid residues (Phe, Val, Leu, Ala) at this position show better than 80% sorting. Only proline (3 clones) and serine (one clone) were selected as correctly sorting in the plate assays of the random mutants at this position.
All mutations in Leu 27 except those with bulky hydrophobic residues give very severe missorting phenotypes. More than 95% of the L27A mutant pro-CPY is secreted, while the hydrophobic residues (Phe, Ile, Val) give rise to mutant forms that are sorted with wild-type efficiency.
Since the initial screen had shown that mutations at position 23 had very little effect on sorting, only three randomly selected mutants were analyzed quantitatively. As expected, the missorting phenotypes of these were not strong (Fig. 3).
Replacing the codon for Gly 28 by random DNA sequence was not possible using random substitutions since this codon forms part of the XbaI site used for ligating the mutagenic oligonucleotides (Fig. 1). Consequently, directed mutations were introduced to exchange Gly 28 with arginine, phenylalanine, or aspartic acid residues. Pulse-chase experiments showed that even these radical changes had limited effect, the strongest being Asp 28 , which gives missorting of 21% of the newly synthesized pro-CPY (Fig. 3).
Relevance of the Position of the Sorting Determinant Relative to the NH 2 Terminus of the Propeptide-The Candida albicans CPY homologue contains a QRPL-like sequence (QNPF) three amino acid residues from the most likely amino terminus of its precursor (Mukhtar et al., 1992). This spacing is thus identical to that of the QRPL element to the amino terminus of pro-CPY while there is no sequence conservation in the rest of the propeptide. To test whether the conservation of the spacing to the amino terminus of the propeptide is important for sorting we duplicated the leucine codon at position 23 of the S. cerevisiae PRC1. This mutation does not change QRPL or its direct FIG. 2. Immunoprecipitation of CPY antigen from selected mutants. Cells expressing three mutant PRC1 alleles and the wild type (WT) were labeled for 20 min with 35 S-labeled amino acids and chased with nonlabeled amino acids and sulfate for 60 min. Intracellular (I) and extracellular (E) material were separated, and CPY antigen was immunoprecipitated. The labeled precipitates were subsequently subjected to 8% SDS-polyacrylamide gel electrophoresis. Within the duration of the chase period all CPY reaches its final destination. environment but increases the distance to the amino terminus of the propeptide. Quantitative immunoprecipitation showed that there was no significant effect of this mutation on pro-CPY sorting (Table I).
In a more radical approach, four extra codons were introduced between codons 22 and 23. Each extra codon started with adenine followed by two random nucleotides to avoid stop codons. Five plasmids resulting from this mutagenesis were sequenced (Table I), and the sorting efficiencies of the respective CPY mutant proteins were determined in pulse-chase experiments. One of the mutants secretes 25% of the total amount synthesized, which is significantly more than the wild type. The secretion of the other of four mutant CPY proteins, however, is only slightly higher than that of the wild type.
CPY Mutant Proteins Do Not Poison the Receptor-Missorting of pro-CPY with altered targeting sequence is in most cases likely to be due to inability to interact functionally with the receptor. This has been directly shown by failure of the Q24K to chemically cross-link to the receptor (Marcusson et al., 1994). Mutations that result in pro-CPY molecules with a too high affinity for the receptor would potentially also lead to missorting. Inability to be released from the receptor in the acceptor compartment would thus lead to poisoning of the receptor. It has been shown that mutations Q24K and L27S do not poison the receptor (Valls et al., 1990). To identify putative receptorpoisoning mutants we used a fusion protein consisting of the first 156 amino acid residues of prepro-CPY and the SUC2encoded enzyme invertase. Invertase is normally secreted while the CPY-invertase hybrid protein is efficiently targeted to the vacuole (Johnson et al., 1987). Plasmid DNA from the four libraries containing QRPL mutations was introduced into ⌬suc2 cells containing the plasmid pFV127 directing the production of the CPY-invertase hybrid protein. Cells expressing a PRC1 allele that would lead to poisoning of the receptor were expected to secrete the pro-CPY-invertase fusion protein and show extracellular invertase activity. 500 yeast transformants from each of the Xaa 24 -Xaa 27 libraries were tested in a plate overlay assay for extracellular invertase activity. All were negative.
In parallel to this approach we tested mutations P26D, L27A, L27G, and L27R, which exhibited the strongest missorting phenotypes, using an approach very similar to that described by Valls et al. (1990), analyzing the competition between aberrant glycoforms of CPY and the QRPL mutant forms relative to the total amount of labeled CPY, intracellularly and extracellularly, after a 20-min 35 S pulse followed by a 60-min chase. Shaded bars indicate mutants that have been characterized in this study, while those indicated by hatched bars are from Valls et al. (1990). wt, wild type.  I Effect of the position of QRPL Sorting efficiencies of mutants having an altered distance to the amino terminus of the propeptide are shown. The extra amino acid residues listed in the first column are all inserted between residues 22 and 23 of the original sequence. The relative amount of intracellular and extracellular CPY was determined by pulse labeling and immunoprecipitation as described in Fig. 2. Immunoprecipitates were quantitated using a PhosphorImager. in immunoprecipitation experiments (not shown). This approach also suggested that receptor poisoning is a phenomenon that cannot be brought about by mutation of a single residue in the QRPL signal. Thus, we find it unlikely that single amino acid substitutions can lead to poisoning of the receptor.

DISCUSSION
The ligand-receptor pair, which is central to the present discussion, represents the first and best characterized member of a new class of intracellular sorting mechanisms. It has long been known that the signal for pro-CPY sorting to the vacuole resides in the proregion and that a sequence containing Gln 24 -Arg-Pro-Leu 27 is both necessary and sufficient for this function (Valls et al., 1987, Johnson et al., 1987. Through the use of genetic screens, this observation eventually led to the identification of the VPS10 gene, which encodes the receptor directly involved in QRPL recognition. The VPS10 gene product (Vps10p) is a very large type I transmembrane protein (1577 amino acid residues) with a short COOH-terminal cytoplasmic tail (164 amino acid residues). In cell fractionation as well as functional studies, it localizes to the distal Golgi apparatus together with the Kex2 protease (Marcusson et al., 1994, Graham et al., 1991. Chemical cross-linking experiments have shown that Vps10p interacts specifically with the QRPL signal, both in a pro-CPY context and in an invertase fusion context (Marcusson et al., 1994). Mutations that resulted in missorting of pro-CPY had been found in each of the QRPL residues (Valls et al., 1990), and it was directly shown that the Q24K mutant protein did not chemically cross-link to Vps10p (Marcusson et al., 1994). However, the level of sequence conservation required for productive pro-CPY-Vps10p interaction has not been investigated. Since the QRPL system has been described in such detail it has to some extent been regarded as a paradigm for comparison to other non-carbohydrate-dependent lysosomal/ vacuolar sorting systems. It has been suggested that the QRPL signal might be a part of a larger "consensus sequence" for lysosomal sorting both in mammals and yeast (McIntyre et al., 1994). The original screen would not address the validity of such a consensus since the mutants were identified by their inability to confer vacuolar sorting to pro-CPY, i.e. no functional mutants were isolated (Valls et al., 1990). This has also hampered the identification of other putative QRPL sequences by homology searches, as the significance of individual residues was not known. In the present work we have attempted to approach a consensus sequence for the specificity of the receptor-ligand interaction.
One of the results of the present study is the identification of the pivotal role of Gln 24 in the sorting process. All mutations at this position showed severe missorting phenotypes, and we failed in our attempts to isolate, by activity stain and immunoblotting, mutants that were able to functionally substitute for Gln 24 . We therefore conclude that this amino acid residue is essential for the proper recognition of the targeting element. There seems to be a similar importance for the structural conservation of Leu 27 , although several hydrophobic residues (Phe, Val, and Ile) were functional. Because of the mutational approach taken, we cannot exclude the possibility that tryptophan or methionine residues, encoded by rare codons, might also function.
Although missorting mutations can be found at positions Arg 25 , Pro 26 , and Gly 28 , the nature of the residues at these positions is clearly of lesser importance. The discussion of the phenotypes of these mutants to some extent depends on the definition of missorting. One should realize that pro-CPY sorting is never 100% efficient; we typically find around 5% of the newly synthesized wild-type pro-CPY to be missorted. This might be due to problems of stoichiometry at the site of inter-action, but it might also be due to an inherent lack of affinity. It is conceivable that CPY might also, under some conditions, be beneficial for extracellular peptide hydrolysis. Thus, there may not have been strong evolutionary pressure for 100% sorting efficiency.
In any event, the substitutions for Arg 25 in several cases lead to a clearly wild-type phenotype. However, it is not easy to rationalize the phenotypes from structural considerations since not only the conservative change to lysine but also radical changes to the hydrophobic amino acids leucine and valine and the polar glutamine fail to affect sorting. In addition, most other mutations obtained have only fairly weak missorting phenotypes, less than 30% being secreted in most strains. Only a substitution by aspartate appears to affect sorting strongly. Most of the mutants characterized at this position were selected on the basis of their proficiency in sorting. Thus, the relatively small number of strong missorting mutants reflects the selection procedure employed.
Mutations leading to amino acid substitutions for Leu 23 had a similar effect (Fig. 3). On immunoblots, no mutants could be identified that showed severe missorting. Thus, randomly chosen mutations were tested in immunoprecipitations, and only weak phenotypes were found. These findings support the notion that Leu 23 is not very important.
Proline is important for specific tertiary structures in many protein contexts. It is therefore surprising that many of the randomly selected mutations at position 26 only had limited effect. There appears to be a correlation between charge and missorting phenotype and a clear preference for hydrophobic residues. The phenotype of the Asp 28 mutant (21% missorting) could suggest that Gly 28 is just as much a part of the sorting signal as Phe 26 . Considering that glycine also has unique structural features, one might have expected more severe missorting phenotypes. As these two examples show, there is a remarkable insensitivity toward the tertiary structural environment of the QRPL signal. This notion is supported by the observation that fairly short fusions (30 propeptide residues) of the QRPL signal to invertase are able to direct this secreted protein to the vacuole (Johnson et al., 1987). Furthermore, the deletion of large fractions of the propeptide downstream of Asp 30 had no effect on sorting (Ramos et al., 1994). Indeed, the combined structural and biochemical analysis of the propeptide suggests that it may have a highly flexible and dynamic structure (Sørensen et al., 1993).
Considering the essential role of the propeptide in folding (Winther and Sørensen, 1991;Ramos et al., 1994), mutagenesis in the propeptide could result in aberrantly folded pro-CPY molecules, and these misfolded molecules might therefore not be recognized by the sorting receptor. It has been shown that mutants having amino acids 27-31 deleted are folded correctly and exit from the ER with normal efficiency (Valls et al. 1987). This suggested that changes in this part of the propeptide would not seriously affect folding. We examined one representative mutant for each position with shorter pulse and chase times and observed maturation half-lives shorter than 15 min (data not shown). The insertion mutants containing four extra residues, however, did show a minor folding defect suggested by a half-time of ER exit of about 30 min. This somewhat extended half-time of maturation also resulted in a small amount, less than 10%, of pro-CPY in the intracellular fraction after a 1-h chase. In all the other experiments there was no detectable pro-CPY left in the intracellular fractions after a 1-h chase. After the 20-min labeling the relative amount of CPY in the proform is typically 20 -40%. The minimal detection limit is 2.5%. This also shows that the half-time of maturation for all the single-amino acid mutants is less than 15 min. One should also bear in mind that it is unlikely that misfolded pro-CPY would escape the ER. Indeed, deletions in the propeptide that result in reduced efficiency of folding do not result in missorting (Ramos et al. 1994). Finally, the folded mature domain of CPY is not necessary for recognition by the receptor since CPY-invertase hybrid proteins containing as little as 30 amino acid residues of CPY are sorted efficiently (Johnson et al. 1987).
The QRPL signal does not conform to a SXX؉XL consensus (where ؉ is a positively charged residue and X is any amino acid residue) as suggested previously (McIntyre et al., 1994). The initial S is supposed to correspond to Ser 22 in pro-CPY, but the present study shows that this residue is not important, while Gln 24 is. Apart from the apparently essential function of Gln 24 , one should probably be cautious with defining a consensus. Although there are clear tendencies to suggest specificity for hydrophobic residues at position 27, it is difficult to rationalize the missorting phenotypes for the other residues. It is surprising that the specificity is not more stringent, especially since there does not appear to be any strict requirement for conservation of distance to the NH 2 terminus. The specificity may to some extent be determined by the exposure of the signal, possibly in a random coil. Such structures are probably not common among proteins en route to the harsh extracellular environment. Along these lines we suggest that the somewhat reduced sorting efficiency seen in some of the insertion mutants (Table I) is more likely to be due to steric interference than to disruption of a recognition element. The important and somewhat surprising conclusion of the present work is that there does not appear to be a conventional "consensus sequence" for ligands of Vps10p. It will thus be difficult to identify ligands by sequence alignment and, in a comparison across species, this problem would be further enhanced by slight changes in specificity likely to occur even between related ligand-receptor pairs.