The C-terminal Region of Carboxypeptidase E Involved in Membrane Binding Is Distinct from the Region Involved with Intracellular Routing*

Carboxypeptidase E (CPE) is involved in the biosyn- thesis of numerous peptide hormones and neurotrans-mitters. Previously, the C-terminal region of CPE has been shown to participate in the binding of the protein to membranes and to also contribute to the sorting of CPE into the regulated pathway. In this study, the role of the C-terminal region of CPE was further examined using several approaches. A series of CPE mutants with C-terminal deletions was expressed in the baculovirus system; constructs with a deletion of 14 or 23 residues were expressed at levels comparable to wild-type CPE. In contrast, deletion of 33 or more residues eliminated CPE activity, and the resulting protein was not secreted from the cells. Even though CPE mutants with a deletion of 14 or 23 residues were expressed normally, the result- ing protein was mainly soluble, whereas approximately 55% of wild-type CPE was membrane associated. When expressed in AtT-20 cells, CPE with a deletion of 43 C-terminal amino acids was not secreted, whereas CPE with a deletion of 23 residues was secreted via the regulated pathway. Pulse-chase analysis revealed the pro- tein with a deletion of 43 residues to be degraded in a non-acidic intracellular compartment. To investigate whether the C-terminal region of CPE can confer membrane binding and regulated pathway sorting to an- other protein, portions of the CPE C-terminal region were attached to the C terminus of albumin and the fusion proteins expressed in AtT-20 cells. Of the constructs examined, only the protein containing 51 amino acids of CPE was sorted to the regulated pathway, although with reduced efficiency compared to endoge- nous CPE. Although the C-terminal 14 amino acids of CPE are sufficient determined

functions in the production of a large number of bioactive peptides (1). CPE is present in neuroendocrine tissues, and in several tissues this enzyme has been localized to the peptidecontaining secretory vesicles (2)(3)(4)(5). Purified CPE is able to cleave a wide variety of peptides with C-terminal basic residues, an important step in the production of biologically active peptides (2, 6 -8). Further evidence of the physiological role of CPE comes from the recent finding that a mouse strain with an inactive form of CPE (due to a point mutation) does not efficiently convert proinsulin into insulin (9).
CPE is present within secretory vesicles in both soluble and membrane-associated forms (2,10). However, the amino acid sequence (deduced from the DNA sequence) does not show any potential transmembrane-spanning helical regions, suggesting that CPE is membrane-bound through another mechanism (11,12). Also, the binding of CPE to membranes is pH-dependent, with maximal binding occurring at pH 5-6 and minimal binding at pH values greater than 8 (13,14). Although CPE has also been found to aggregate at acidic pH values, this process appears to be distinct from the membrane-binding process (15). Several lines of evidence suggest that the C-terminal region of CPE contributes to the membrane binding of this protein. First, the C-terminal region of CPE is predicted to form an amphiphilic ␣-helix (14). Second, synthetic peptides that correspond to the C-terminal region of CPE are able to bind to membranes in a pH-dependent manner (14). Third, the only difference found between the soluble and membrane-bound forms of CPE is within the C-terminal region; antisera directed against the C-terminal region of CPE show much stronger binding to the membrane forms of CPE compared to the soluble forms (14,16). However, this point is controversial as it has been reported that the soluble and membrane forms of CPE show comparable binding of C-terminally directed antisera (17). Thus, one focus of the present study was to further examine the role of the C-terminal region of CPE in membrane binding.
The pH-dependent membrane binding of CPE has been proposed to be a mechanism for the sorting of CPE (14,18). According to this hypothesis, the interaction of CPE with membranes in the trans Golgi network, with an internal pH in the 6 -6.5 range, would help drive the sorting of CPE into the regulated pathway. Consistent with this hypothesis was our recent finding that fusion proteins containing 51 amino acids of the C-terminal region of CPE attached to albumin are both membrane bound and partially sorted into the regulated pathway (18). Another focus of the present study was to investigate whether the sorting and membrane-binding regions within the C-terminal region of CPE were the same or distinct. Several approaches were used to study the C-terminal region of CPE. The first approach used deletion mutations within the C-terminal region to address the role of this domain in protein expression and membrane binding. In the second approach, fusion proteins consisting of albumin with small portions of CPE attached to the C terminus were expressed in AtT-20 cells and examined for sorting and membrane binding. The results of these analyses indicate that the membrane-binding and sorting domains within the C-terminal region are distinct. Our results support the hypothesis that the putative amphiphilic helical region of CPE contributes to the membrane binding but does not support the hypothesis that this region is also involved with intracellular routing of CPE.

MATERIALS AND METHODS
Generation of Constructs-CPE C-terminal deletions were generated using the Altered Sites II in vitro mutagenesis system (Promega). Rat CPE cDNA was subcloned into the BamHI and EcoRI sites of the vector pALTER-1, and mutagenic oligonucleotides with stop codons in different positions were used to generate C-terminally truncated forms of CPE. The hemagglutinin (HA) sequence recognized by the monoclonal antisera 12CA5 (HA tag, Fig. 1) was inserted into the N-terminal PstI site of wild-type CPE by ligation of the appropriate oligonucleotides. These oligonucleotides introduce an XbaI site on the N-terminal side of the HA sequence. The ⌬pro CPE construct ( Fig. 1) was created using polymerase chain reaction to introduce an XbaI site at the junction between the pre-domain and the pro-domain and then to ligate this small fragment encoding the pre-region into the expression vector, replacing the prepro-region of wild-type CPE. The construct with the N-terminal HA tag and the deletion of the 23 C-terminal residues was created by piecing together the cDNA encoding the N-terminal fragment (BamHI/SacI) of HA-tagged wild-type CPE and the cDNA encoding the C-terminal fragment (SacI/EcoRI) of CPE⌬23. The constructs were subcloned into the BamHI and EcoRI sites of the baculovirus transfer vector pVL1393 (PharMingen) to express it under polyhedrin promoter control in the baculovirus system. For expression in a mammalian cell line, constructs were subcloned into the eukaryotic expression vector pcDNA-3 (Invitrogen).
Human albumin cDNA in pGEM7zf (Promega) was used to generate albumin/CPE fusion proteins. Polymerase chain reaction was used to generate fragments of CPE encoding the various C-terminal regions, with a Bsu36I site on the 5Ј-end and a SacI site on the 3Ј-end. These fragments were subcloned into the appropriate sites within the albumin/pGEM plasmid; there is a Bsu36I site at the C terminus of albumin and a SacI site in the pGEM7zf vector. For the expression in AtT-20 cells, the constructs were subcloned into the BamHI and NsiI sites of the eukaryotic expression vector pcDNA/NEO (Invitrogen). Dideoxynucleotide sequencing was performed to confirm the sequence of the C-terminal regions of all constructs and the N-terminal region of the ⌬pro and HA-tagged constructs.
Expression of Proteins in AtT-20 Cells-The expression plasmids containing the sequences encoding albumin/CPE fusion proteins, CPE⌬43, and HA-tagged wild-type and CPE⌬23 were transfected into AtT-20 cells using the standard calcium phosphate procedure (19). Stable cell lines were selected using 0.6 mg/ml geneticin (G418). For the albumin/CPE fusion proteins and the CPE⌬43 constructs, 80 -100 colonies were isolated per construct and grown in 35-mm plates. Cells expressing the various constructs were identified by immunoprecipitation (for the albumin/CPE constructs) or by Western blot analysis (for CPE⌬43), as described (18). For each of these constructs, several positive clones were selected and analyzed as described below; the data shown are representative of two to five separate clones. HA-tagged wild-type and CPE⌬23 were similarly transfected into the AtT-20 cells, but individual cell lines were not isolated. Instead, G418 was used to select for stably transformed cells, and after 3 weeks in culture, the mixed population of cells was analyzed for regulated secretion as described below.
Expression of Mutants in Baculovirus-Recombinant baculovirus expressing the various constructs was generated using the Baculogold system (PharMingen), using the procedure recommended by the supplier. Sf9 cells (25 ml at 1.6 ϫ 10 6 /ml) in a 250-ml flask were infected with the recombinant baculovirus, and cells were incubated for 72 h at 27°C. Cells were recovered by centrifugation at 1000 ϫ g for 10 min. The cells were homogenized (Polytron, Brinkman) in 12.5 ml of 100 mM NaAc buffer, pH 5.5. Aliquots of cell homogenate and media were assayed for carboxypeptidase activity using 0.1 mM dansyl-Phe-Ala-Arg in 100 mM NaAc buffer, pH 5.5 (without CoCl 2 ), in a final volume of 250 l as described (20). Protein was determined using the Bradford reagent. Aliquots were also analyzed on a Western blot as described below.
To determine the amount of CPE that was membrane-associated, the cell homogenates were centrifuged at 50,000 ϫ g for 30 min. The supernatant was removed, and the pellet was homogenized (Polytron) in 0.1 M NaAc buffer, pH 5.5, and centrifuged as above. Then, the pellet was homogenized (Polytron) in 1 M NaCl in NaAc buffer and centrifuged as above. This step was repeated, and then the pellet was homogenized in 1% Triton X-100 and 1 M NaCl in NaAc buffer and centrifuged as above. This step was repeated, and the final pellet was resuspended in 0.1 M NaAc buffer. Aliquots of each fraction were assayed for CPE activity as described above. In a separate series of experiments, cells were extracted directly with 1% Triton X-100 and 1 M NaCl in 0.1 M NaAc buffer, pH 5.5, and the extract was applied to a p-aminobenzoyl Arg-Sepharose 6B affinity column as described (14). The ability of the various CPE constructs to bind to the column was assessed by Western blot analysis.
Fractionation of AtT-20 Cells-AtT-20 cells expressing the various fusion proteins were grown as described previously for wild-type AtT-20 cells (21). Cells from four 15-cm plates were scraped into 5 ml of phosphate buffered saline (PBS) and then pelleted at low speed. The pellet was gently resuspended in 3 ml of cold 0.25 M sucrose in 10 mM Tris, 1 mM MgAc 2 , pH 7.4, and passed through a cold steel ball-bearing homogenizer several times (22). The homogenate was diluted with 2.3 M sucrose in the same buffer to give a final concentration of 1.4 M sucrose. A step sucrose gradient was created using the following solutions: 5 ml of 2 M sucrose, 8 ml of 1.4 M sucrose containing the cell lysate, 18 ml of 1.2 M sucrose, and 6 ml of 0.8 M sucrose. The tubes were centrifuged in a SW-28 rotor for 12 h at 26,000 rpm at 4°C. Fractions (2-3 ml) were collected and diluted 10 times with 0.2 M sucrose and centrifuged using the Ti-42 rotor at 32,000 rpm for 6 h at 4°C. The pellets were resuspended and sonicated in 1 ml of 10 mM NaAc buffer, pH 5.5, and stored frozen at Ϫ20°C. CPE activity was determined as described (23). Acid phosphatase, a lysosomal marker, and ␣-mannosidase II, a Golgi marker, were assayed as described (24,25). Protein was determined by the Bradford assay. Western blot analysis was performed to identify fractions with CPE and albumin immunoreactivity. The fraction collected from the interface of the 0.8 and 1.2 M sucrose layers, which is enriched in Golgi and secretory vesicles (22) and contains a large amount of immunoreactive CPE and albumin (not shown), was used for further studies on the soluble/membrane distribution of the proteins.
A 100-l aliquot of the resuspended Golgi/vesicle fraction was centrifuged at 100,000 ϫ g for 30 min. The supernatants were removed ("soluble" fraction), and the pellets were resuspended and sonicated in 100 l of 1 M sodium chloride in 10 mM NaAc buffer following centrifugation again at 100,000 ϫ g for 30 min. The second supernatants ("membrane-1" extracts) were removed, the pellets were sonicated in 100 l of 1 M sodium chloride and 1% Triton X-100 in the same buffer and centrifuged again, and the supernatants were removed ("membrane-2" extracts). Aliquots of each of the three supernatants were analyzed by Western blotting as described below.
Western Blot Analysis-Protein samples were combined with SDSpolyacrylamide gel loading buffer, heated to 95°C for 5 min, cooled, separated on a 10% SDS-polyacrylamide gel, and then transferred to nitrocellulose. The CPE was detected using the standard enhanced chemiluminescence procedure (Amersham). The antiserum raised against an N-terminal peptide of bovine CPE has been previously described (14). Another antiserum was raised against the peptide SETLNF, which corresponds to the predicted C-terminal sequence of CPE, conjugated to keyhole limpet hemocyanin using the glutaraldehyde procedure, as described (26). A rabbit polyclonal antiserum was produced against this peptide (Hazleton Research Products, Denver, PA). This antiserum was additionally purified on an affinity column consisting of the same peptide conjugated to Sepharose 6B using the epichlorohydrin method (27). The affinity-purified antiserum was used at a dilution corresponding to 1:1000 of initial antiserum. An antiserum against human serum albumin (Cappel) was used at a dilution of 1:1000. The monoclonal antisera 12CA5, which recognizes the HA epitope, was a gift of Dr. John Backer (Albert Einstein College of Medicine) and was used at a dilution of 1:5000.
Examination of Regulated Secretion-To examine whether the albumin/CPE fusion proteins are secreted via the regulated pathway, the cells expressing these proteins were grown on 15-cm cell culture dishes to 90% confluency. The cells were washed twice with PBS, then treated with secretagogue (5 M forskolin) or control media for 30 min, and the secreted proteins were analyzed by Western blotting. AtT-20 cells expressing the HA-tagged wild type and CPE⌬23 were similarly treated, except that the cells were treated with control media, media containing 10 M forskolin, or media containing 0.1 g/ml tetradecanoylphorbol-13-acetate.
Pulse-Chase Analysis of AtT-20 Cells-Wild-type AtT-20 cells and AtT-20 cells expressing CPE⌬43 were labeled (pulse) with [ 35 S]Met (100 Ci/ml) for 30 min, washed twice with PBS, and then incubated in Dulbecco's modified Eagle's medium for 0, 30, or 180 min. Media were removed, and cells were washed with PBS and then frozen in 10 mM NaAc, pH 5.5, with 1 mM phenylmethylsulfonyl fluoride. The cells and the media were then subjected to immunoprecipitation as described previously (18). To examine the effect of different conditions and agents on the secretion of CPE, AtT-20 cells were labeled as described above and then during the 3-h chase subjected to different treatments including incubation at 15 and 20°C, brefeldin A (5 g/ml), chloroquine (100 M), and NH 4 Cl (10 mM). Media were collected, and the cells were washed twice with PBS and frozen in 10 mM NaAc, pH 5.5. Both cells and media were subjected to immunoprecipitation using the antisera raised against the N-terminal region of CPE as described (18).

RESULTS
Deletion Protein Analysis-To identify the portion of the C-terminal region of CPE that is involved with membrane binding and to test if this region is also involved with intracellular routing, a series of deletion mutants was created (Fig. 1). Using the baculovirus expression system, all of the constructs were expressed in the Sf9 insect cell line in comparable levels (Fig. 2, left). However, the two constructs with the largest C-terminal deletions (⌬33 and ⌬43) were not secreted from the Sf9 cells (Fig. 2, right). These two constructs also did not appear to be enzymatically active; the levels of CPE-like activity in either cells or media after infection with ⌬33and ⌬43expressing baculovirus were not substantially different from control virus-infected cells (Fig. 2, bottom). The other two Cterminal deletions (⌬14 and ⌬23) as well as a construct with the N-terminal pro-region deleted (⌬pro) showed enzyme activities comparable to that of wild-type CPE (Fig. 2, bottom). The ability of the various constructs to bind to a Sepharose benzoyl-Arg substrate affinity column correlated with the presence of enzyme activity; all of the constructs except CPE⌬33 and ⌬43 bound to this column (data not shown).
To examine whether the various deletions altered the soluble/membrane distribution of the active forms of CPE, cells expressing the constructs that showed enzyme activity were sequentially extracted with NaAc at pH 6.0 (soluble), with 1 M NaCl in NaAc buffer (membrane-1), and then with a combination of 1% Triton X-100 and 1 M NaCl in NaAc buffer (membrane-2). To ensure that each extraction was complete, two extractions with each buffer were performed and assayed separately. Typically, the second extraction with each buffer contained much less CPE activity than the first extraction (Fig. 3), indicating that the apparent membrane binding is not due to incomplete tissue disruption or to revesiculation. Approximately 40 -50% of the total wild-type CPE and ⌬proCPE activity was detected in the two soluble extracts, 30 -40% was detected in the membrane-1 extracts, and 15-20% was detected in the membrane-2 extracts (Fig. 3). Membranes after the extractions contained very little CPE activity (not shown). In contrast to the results with wild-type CPE and ⌬proCPE, the two C-terminal deletion mutants (CPE⌬14 and ⌬23) were much more abundant in the soluble fraction and virtually absent from the membrane-2 fraction (Fig. 3). This finding supports the hypothesis that the C-terminal region, which is predicted to form an amphiphilic helix, is important for the NaCl/ Triton X-100-dependent binding of CPE to membranes.
Removal of the C-terminal 14 amino acids has a minor effect on the net charge, with the elimination of two acidic groups and one basic group (Fig. 1). In contrast, removal of the C-terminal 23 residues has a larger effect on the net charge. To test whether the decrease in membrane binding of CPE⌬23 is due to an alteration in the net charge, which affects the ability of the protein to aggregate, we examined the pH-dependent aggregation of wild-type CPE and CPE⌬23. CPE was purified from baculovirus-infected Sf9 cells and tested for aggregation as described previously (15). Wild-type CPE and CPE⌬23 showed substantial aggregation at pH 5.0, less aggregation at pH 5.5, and virtually no aggregation at pH 6.0 or higher (data FIG. 1. Fusion proteins and deletion mutants. Top, diagram of prepro-CPE and amino acid sequences of wild-type (wt) CPE and various deletion mutants. The C-terminal sequence of wild-type CPE is indicated for the region beginning at Ser 381 and extending to the C terminus of wild-type CPE (Phe 434 ). The ⌬pro deletion construct contains the influenza hemagglutinin sequence (HA tag) in place of most of the pro-region as indicated. Bottom, diagram of pro-albumin, with portions of CPE attached to the C terminus, and amino acid sequences of the C-terminal portions of these constructs. Human albumin ends with the sequence LGL, and for the various fusion proteins the C-terminal Leu was replaced with the indicated sequence. Dashes indicate gaps in the sequence, and dots indicate continuing sequence not shown in the figure.

FIG. 2. Expression of CPE deletion mutants in baculovirus.
Top, Western blot analysis of cells (left) or media (right) after infection for 72 h with control baculovirus (BV), with baculovirus expressing wild-type CPE (wt), or with the various deletion mutants. The Western blot was performed as described under "Materials and Methods." Bottom, carboxypeptidase activity in the cell extracts and media was determined using dansyl-Phe-Ala-Arg as described previously (20), except that CoCl 2 (which activates CPE approximately 3-5-fold) was not included. Protein in the cell extracts was determined using the Bradford assay. Enzyme activity in both the cell and media samples was normalized to the amount of protein in the cell extracts, which varied less than 2-fold between samples. The enzyme determinations were performed with two to three separate infections, with less than 50% variation in the enzyme measurements between infections. For each infection, enzyme activity was measured in duplicate, with less than 10% variation between replicates. not shown). The similar pH-dependent aggregation of the two forms of CPE implies that this aggregation is not responsible for the membrane-binding properties of CPE, which show substantial differences between wild-type CPE and CPE⌬23 (Fig. 3).
To test whether the C-terminal 23 residues of CPE contributes to the efficient sorting of this protein into the regulated pathway, the HA-epitope tag sequence was introduced in the N-terminal region (to distinguish the mutant CPE from endogenous CPE) and the protein expressed in AtT-20 cells. The N-terminal HA sequence does not interfere with the enzymatic properties of CPE or the sorting of this protein into the regulated pathway (data not shown). AtT-20 cells expressing HAtagged CPE⌬23 were stimulated with secretagogues for 30 min, and the amount or protein in the media was analyzed by Western blot analysis. The monoclonal antisera that recognizes the HA sequence binds to a single protein in the media of transfected AtT-20 cells (Fig. 4); untransfected cells do not show any signal with this antisera (data not shown). Treatment of the cells with either forskolin or tetradecanoylphorbol-13-acetate stimulates the levels of HA-tagged CPE⌬23 in the media approximately 2-3-fold (Fig. 4). The same media samples were also analyzed for endogenous CPE, which was detected using an antisera directed against the C-terminal region (i.e. the region missing from the CPE⌬23 construct). The endogenous CPE that reacts with the C-terminally directed antisera is stimulated by the two secretagogues to the same extent as the CPE⌬23 (Fig. 4), indicating that the C-terminal region is not necessary for the efficient sorting of CPE into the regulated pathway.
CPE with a deletion of 43 C-terminal amino acids was expressed in AtT-20 cells to examine whether the C-terminal region is required for secretion in a mammalian cell line that normally expresses CPE. Immunoprecipitation of the [ 35 S]Metlabeled extract from cells expressing CPE⌬43 showed a major band of approximately 44 kDa and a minor band of 50 kDa, in addition to the endogenous proCPE (56 kDa) and various minor bands that were also detected with wild-type AtT-20 cells (Fig.  5). Analysis of 5 separate clones by [ 35 S]Met labeling and immunoprecipitation and 30 separate clones by Western blot analysis showed a similar presence of two additional proteins of approximately 44 and 50 kDa. Both forms of CPE⌬43 do not change during the 30-min chase but then largely disappear from the cells after 180 min of chase (Fig. 5). Neither form of CPE⌬43 appears in the media, indicating that the mutant is degraded during the 180-min chase. In contrast, the endoge-nous CPE becomes slightly smaller and shows a weaker signal after 30 min of chase. This is consistent with the removal of Met-containing N-terminal (propeptide) and C-terminal sequences from a portion of the molecules; altogether, 4 out of the 9 total Met residues in mouse proCPE are located within 13 residues of the N and C termini (GenBank accession no. X61232). After 180 min of chase, the endogenous CPE is detected in the media in amounts equal to those in the cells (Fig. 5).
To further investigate the fate of CPE⌬43 in the AtT-20 cells, the cells were labeled with [ 35 S]Met and then either analyzed directly or chased under a variety of conditions that block various cellular processes. After 3 h of chase, both the endogenous CPE and the two bands of CPE⌬43 were substantially decreased relative to the levels before the chase (Fig. 6). Analysis of the media showed the CPE, but not the CPE⌬43, to be secreted (not shown) as found for Fig. 5. The degradation of CPE⌬43 that occurs during the 3-h chase is blocked when the chase is performed at either 15 or 20°C (Fig. 6). Brefeldin A, which blocks transport between the endoplasmic reticulum and Golgi, does not block the degradation of CPE⌬43, although this does prevent the secretion of CPE (Fig. 6). Neither chloroquine nor ammonium chloride, which block the acidification of vesicles, prevents the degradation of CPE⌬43 (Fig. 6). Taken together, these results suggest that CPE⌬43 is degraded by a temperature-sensitive pre-Golgi enzyme.
Fusion Protein Analysis-Albumin fused with 51 residues of the C-terminal region of CPE has been previously shown to bind to membranes and to be sorted into the regulated pathway with low efficiency. Additional constructs (Fig. 1) were stably expressed in AtT-20 cells and compared to cells expressing either albumin with the 51 residues of CPE (Albϩ51) or albumin alone (Alb). Cells were fractionated by sucrose density centrifugation, and the enriched Golgi/secretory vesicle fraction was lysed and separated into soluble and membrane extracts. Western blot analysis showed a substantial amount of membrane-bound albumin in the cells expressing the Albϩ51, Albϩ23, and Albϩ14 constructs (Fig. 7). A similar result was obtained when the extraction was performed at pH 6.0 (data not shown). In contrast, the majority of immunoreactive albumin from the cells expressing either unmodified albumin or Albϩ51⌬14 was soluble (Fig. 7). The Albϩ51⌬14 construct showed slightly less membrane binding than the Alb control (Fig. 7), which could be due to the addition of numerous charged residues on the Albϩ51⌬14 construct (Fig. 1). For some of the cell lines, several bands of immunoreactive albumin were detected, which presumably represents differences in post-translational processing (Fig. 7). When cell extracts were probed with an antisera directed against the C-terminal 6 residues of CPE, bands were detected corresponding to the molecular masses of endogenous CPE (54 -56 kDa) and the albumin/CPE constructs (70 -75 kDa). With this antisera, immunoreactivity was detected only for the membrane-2 fraction (i.e. the extracts with both NaCl and Triton X-100) and not in the membrane-1 fraction (i.e. extracts with only NaCl) or soluble fraction (Fig. 7). This result, which is consistent with previous reports that soluble and membrane-bound forms of CPE differ in their C-terminal region (14), suggests that the soluble forms of the albumin/CPE constructs have undergone proteolysis within the C-terminal region of CPE.
Sorting of the albumin/CPE constructs into the regulated pathway was investigated by stimulating the cells with forskolin and then measuring either albumin or CPE immunoreactivity using Western blot analysis. For all cell lines, the secretion of endogenous CPE in the presence of forskolin was 200 -300% of the secretion of CPE from unstimulated cells (Fig. 8). This result indicates that the cell lines expressing the various albumin/CPE constructs possess a functional regulated secretory pathway. As previously reported, the forskolin-induced secretion of immunoreactive albumin from the Albϩ51-expressing cells is 135% of the control level of secretion (Fig. 8). In contrast, forskolin does not cause a significant change in the secretion of immunoreactive albumin from any of the other cell lines (Fig. 8). This result indicates that the membrane-binding domain (i.e. the C-terminal 14 residues) is not sufficient for sorting into the regulated pathway. Also, the Albϩ23 and the Albϩ51⌬14 constructs contained overlapping segments of the C-terminal region of CPE, and the lack of correct sorting of either construct suggests that the signal for sorting is not a short linear sequence.

DISCUSSION
The present study used two separate approaches: deletion mutation and fusion protein analysis. These different approaches provide complementary information. The fusion protein analysis reveals whether a region can confer a particular targeting pattern to another protein. A positive result is strong evidence that the particular region performs the targeting function, although it does not reveal whether additional regions participate. However, a negative result does not rule out the possibility that a region is involved with the function since the structure of this region within the fusion protein may not resemble the structure of this region in the native protein. The deletion approach provides information as to the consequences of the elimination of a particular region. A positive result is good evidence for a functional role, although it is hard to rule out a nonspecific change in structure, caused by the deletion, which causes the observed effect. Also, a negative result does not necessarily indicate the lack of a particular function since two or more regions could contribute, and deletion of a single region may not cause a dramatic effect. For these reasons, the combined approach of fusion proteins and deletion analysis provides stronger results than either approach alone.
A major finding of the present study is that the C-terminal 14 residues of CPE are both necessary (from the deletion mutant analysis) and sufficient (from the fusion protein analysis) for the NaCl-independent membrane binding at pH 5.5. This conclusion fits the hypothesis that the C-terminal 14 residues form an amphiphilic helix, which helps anchor the protein to membranes. Adjacent to this putative amphiphilic helix is a highly charged region that was previously proposed to contribute to the salt-dependent membrane binding (i.e. the membrane-1 form) (29). However, our finding that the mutant with a deletion of 23 C-terminal residues is still able to bind to membranes in a salt-dependent manner argues against this charged region being solely responsible for the ionic binding of CPE to membranes. In addition to CPE, many other proteins within secretory vesicles have been reported to exist in membrane-associated forms even though the protein does not contain a predicted transmembrane-spanning helix (30 -34). Some of these proteins (such as prohormone convertases 1 and 2) also contain a predicted C-terminal amphiphilic helix, although there is no direct sequence similarity between CPE and these other proteins.
The C-terminal region of CPE has been highly conserved among species. The last exon of CPE, which encodes the Cterminal 32 amino acids, is 100% identical in human, rat, mouse, and bovine CPE and contains only four conservative substitutions in Anglerfish CPE (Refs. 11, 12, 35, and 36 and GenBank accession no. X61232). Except for gp180 (37), a duck protein which may be the homologue of the recently discovered bovine carboxypeptidase D (38), none of the other members of the metallocarboxypeptidase gene family have significant homology to this region of CPE (39 -45). Furthermore, several family members (such as pancreatic carboxypeptidase A and B) are approximately 130 amino acids shorter than CPE, with most of the difference in size due to the length of the C-terminal region (39,40). Based on these sequence comparisons, it was expected that all of the C-terminal deletion mutants would be enzymatically active. Our finding that CPE lacking 33 or more C-terminal residues is inactive implies that the region between 23 and 33 residues from the C terminus of CPE is necessary for the production of enzymatically active CPE. Since CPE⌬33 and CPE⌬43 are unable to bind to a substrate affinity column, it is likely that they are not correctly folded, although a direct role for the C-terminal region in substrate binding is also possible. Our finding that CPE⌬23 is active and secreted from Sf9 cells is consistent with a study by Manser et al. (46), who found that CPE with a deletion of 26 C-terminal residues was both active and secreted from C6 cells.
The finding that CPE⌬43 is degraded and is not secreted from AtT-20 cells is consistent with the possibility that the C-terminal region is important for proper folding of the protein.
The effect of brefeldin A and other compounds on the degradation of CPE⌬43 is similar to the effect of these compounds on the degradation of other proteins that have been expressed in mammalian cell lines. Examples include T-cell receptor subunits expressed in fibroblasts (47), apolipoprotein B100 in HepG2 cells (48), the immunoglobulin chain in CH12 lymphoma-derived cells (49), and the heavy chain of class 1 major histocompatibility complex in human embryonic lung fibroblasts (50). As found for CPE⌬43, the degradation of these other proteins is insensitive to lysosomotrophic agents (chloroquine, weak bases) and to blockers of transport from endoplasmic reticulum to Golgi (such as brefeldin A), but their degradation is prevented by incubation at low temperatures.
Protein sorting into the regulated pathway is thought to involve one of two mechanisms. One mechanism requires that the regulated pathway proteins contain specific sequences that are recognized by a "sortase" protein that binds to and directs the sorting of these proteins (51). However, except for recent studies on the sorting of proopiomelanocortin (52), there is no evidence for a specific sorting signal on regulated pathway proteins, and a previous report describing a "sortase" (53) does not appear to be correct (54). The other potential mechanism involves the aggregation of regulated pathway proteins; these aggregates would then be sorted into the regulated pathway either by a size selection (i.e. if the aggregates are simply too large to fit into constitutive vesicles) or by some other mechanism (28,(55)(56)(57). The pH-dependent membrane binding of CPE was previously proposed to be a mechanism for aggregates of CPE (and possible co-aggregates of CPE and other proteins) to bind to secretory vesicle membranes, thus driving the sorting process (14). An important finding of the present study is that the membrane-binding domain of CPE is not sufficient for sorting of a fusion protein (Fig. 8) or necessary for the sorting of CPE (Fig. 4) into the regulated pathway. These results argue against our previous prediction that the membrane binding of CPE would contribute to the sorting of this protein (14,18).
In summary, there appear to be three separate functions within the C-terminal region of CPE. The 51 C-terminal amino acids direct the sorting of albumin into the regulated pathway, although with a lower efficiency compared to that of CPE. Since smaller portions of this C-terminal region are not effective in directing the sorting of albumin, this "sorting" domain is distinct from the other two regions identified in the present study. Another important region is located 23-33 amino acids from FIG. 7. Western blot analysis of AtT-20 cells expressing albumin/CPE fusion proteins. Either whole cell homogenates (center panel) or an enriched vesicle/microsomal fraction of the cells (left and right panels) were separated into soluble extracts (S), NaCl membrane extracts (M 1 ), or NaCl/Triton X-100 membrane extracts (M 2 ), as described under "Materials and Methods." The mobilities and apparent molecular weights of prestained standards (Bio-Rad) are indicated. Similar results were obtained in three separate experiments using two distinct clones for each construct.
FIG. 8. Relative amounts of albumin or CPE immunoreactivity secreted from fusion protein-expressing AtT-20 cells. Cells were stimulated with forskolin for 30 min, and the media were analyzed on Western blots using albumin-and CPE N-terminal-directed antisera as described under "Materials and Methods." For each antisera and for each construct, the data are shown relative to the level of immunoreactive protein in the control media. Error bars show standard error of the mean for the cell lines expressing Albϩ51 (n ϭ 7), Albϩ23 (n ϭ 6), Albϩ14 (n ϭ 6), Albϩ51⌬14 (n ϭ 16), and Alb (n ϭ 5). *, differs from control (p Ͻ 0.05) using Student's t test.
the C terminus; this region may be required for the proper folding of CPE since protein lacking this region is neither active nor secreted from cells. The third domain, which is located within the predicted amphiphilic helix of the C-terminal 14 residues, is involved with the salt-independent binding of CPE to membranes. The presence of multiple domains within the C-terminal region of CPE could account for the high degree of conservation of this region among CPE from different species.