Binding of BiP to the Processing Enzyme Lymphoma Proprotein Convertase Prevents Aggregation, but Slows Down Maturation*

Lymphoma proprotein convertase (LPC) is a subtili-sin-like serine protease of the mammalian proprotein convertase family. It is synthesized as an inactive precursor protein, and propeptide cleavage occurs via intramolecular cleavage in the endoplasmic reticulum. In contrast to other convertases like furin and proprotein convertase-1, propeptide cleavage occurs slowly. Also, both a glycosylated and an unglycosylated precursor are detected. Here we demonstrate that the unglycosylated precursor form of LPC is localized in the cytosol due to the absence of a signal peptide. Using a reducible cross-linker, we found that glycosylated pro-LPC is associated with the molecular chaperone BiP. In addition, we show that pro-LPC is prone to aggregation and forms large complexes linked via interchain disulfide bonds. BiP is associated mainly with non-aggregated pro-LPC and pro-LPC dimers and trimers, suggesting that BiP prevents aggregation. Overexpression of wild-type BiP or a dominant-negative BiP ATPase mutant resulted in reduced processing of pro-LPC. Taken together, these results suggest that binding of BiP to pro-LPC prevents aggregation, but results in slower

Many secretory proteins are synthesized as inactive precursor proteins. Activation requires the endoproteolytic cleavage of their propeptides, usually carboxyl-terminal of a basic motif composed of two or more basic residues. In 1990, the prototype of the mammalian enzyme family responsible for this activation step, furin, was discovered (1,2). Since then, six additional members of this proprotein convertase (PC) 1 family have been isolated: PC1 (also called PC3), PC2, PC4, PACE4, PC5 (also called PC6), and lymphoma proprotein convertase (LPC; also called PC7 or PC8) (reviewed in Refs. [3][4][5]. Some members, like furin, PACE4, PC5, and LPC, exhibit a widespread tissue distribution, whereas the others have a more restricted expression pattern (6). PC1 and PC2 are expressed in neuroendocrine cells, whereas expression of PC4 is restricted to germ cells.
They all are multidomain enzymes composed of a conserved amino-terminal propeptide followed by a subtilisin-like catalytic and a middle domain. They diverge at the carboxyl terminus, although common elements like a cysteine-rich domain (present in furin, PACE4, and PC5) are found. Furin, LPC, and the PC6B isoform are type I transmembrane proteins, whereas the other convertases are soluble.
The latest member of the family, LPC, was first discovered in the chromosome breakpoint region of a high-grade lymphoma carrying a t(11;14)(q23;q32) translocation (7). It is widely expressed, both during embryogenesis and adult life (8,9), and has a substrate specificity similar but not identical to that of furin (10,11). LPC is the least conserved family member, and phylogenetic analysis indicates that LPC is more related to kexin in yeast than to any other mammalian PC (12,13). The mature protein is concentrated in the trans-Golgi network (TGN), although under certain conditions, it can be resolved from TGN markers like furin and TGN38 (10,14).
A key event in the maturation of PCs is the cleavage of the amino-terminal propeptide. This process has been found to be autocatalytic and intramolecular for furin (15,16), PC1 (17)(18)(19), PACE4 (20), and LPC (10). Propeptide cleavage is a prerequisite for exit out of the endoplasmic reticulum (ER) and requires the presence of both the middle and catalytic domains (21)(22)(23)(24). The requirement for sequences in the middle domain was illustrated recently in a patient with a multihormonal syndrome including severe childhood obesity. Two compound heterozygous mutations were found in the PC1 gene (25). In addition to a splice site mutation leading to a frameshift, a missense mutation (Gly 3 Arg) at the carboxyl terminus of the middle domain was found that prevented processing of pro-PC1 and led to retention in the ER.
The only exception to this type of processing is PC2, which requires the presence of the molecular chaperone 7B2 for transport and activation, a process that takes place in the TGN (26,27). The necessity of 7B2 for activation of pro-PC2 in vivo has unequivocally been established using a knockout mouse model (28). The mechanism of propeptide cleavage of PC2 is not yet fully understood, and both intermolecular (29) and intramolecular (30) autocatalytic mechanisms have been suggested.
Cleavage of the propeptides of furin (31) and PC1 (17) occurs rapidly after biosynthesis, with a half-life of the pro form of only a few minutes. The conversion of pro-PACE4 (20, 32) and pro-LPC (10), on the other hand, takes hours rather than minutes. The prosegments of both furin and LPC can act in trans as potent inhibitors of the enzymes themselves (33). In the case of furin, the cleaved propeptide remains associated with the enzyme during transport to the TGN, where a second internal cleavage of the propeptide is required for release and full activation of furin (34). Although this mechanism might apply to other PCs as well, LPC and PC4 lack such a second cleavage site marked by a cluster of basic residues. Activation of LPC must therefore follow an alternative pathway.
This study was performed to obtain insight in the slow, autoproteolytic maturation of LPC in both transiently and stably transfected Chinese hamster ovary (CHO) cells. We have analyzed the origin of the unglycosylated precursor form observed previously (10). Furthermore, we have used a crosslinking approach to identify proteins interacting with pro-LPC. Finally, we characterized the large aggregates of pro-LPC that are formed upon transient, but not stable, overexpression.

MATERIALS AND METHODS
Plasmids and Mutagenesis-The cloning of the 2.6-kilobase pair human LPC cDNA in the mammalian expression vector pcDNA3 (Invitrogen) and the construction of the active-site serine mutant Lser (containing a S265A mutation) and the mutants L⌬c and L⌬tc have been described before (10). Mutant L⌬c lacks the cytoplasmic tail of LPC, whereas mutant L⌬tc also lacks the transmembrane anchor. Mutant L⌬stc contains a stop codon six amino acids carboxyl-terminal of the end of the middle domain (VWSAVD-stop) and was constructed by cloning the following primer set in the AccI restriction site in LPC and the XbaI site in pcDNA3: 5Ј-AGACTAGCTCGAGT-3Ј (sense) and 5Ј-CTAGACTCGAGCTAGT-3Ј (antisense). Mutation of the second inframe ATG codon to ATA (mutant ⌬atg2) was achieved using the QuickChange site-directed mutagenesis kit (Stratagene), according to the guidelines of the supplier, and the following mutagenic primers: 5Ј-GTTCCCTGGGTCATAGGCCTGGCAGGGAC-3Ј (sense) and 5Ј-GT-CCCTGCCAGGCCTATGACCCAGGGAAC-3Ј (antisense). Mutations were confirmed by nucleotide sequence analysis. The generation of the hamster BiP ATPase mutant (G37) has been described previously (35).
Cell Lines and DNA Transfer-The medium, serum, and supplements used for maintenance of cells were obtained from Life Technologies, Inc. Generation of the CHO-DHFR Ϫ cell line stably overexpressing LPC (CHO-LPC) has been described before (10). The cells were cultured in ␣-minimal essential medium containing ribonucleosides and deoxyribonucleosides supplemented with 10% fetal calf serum and 250 g/ml G418. CHO and COS-1 cells were grown in Dulbecco's modified Eagle's/Ham's F-12 medium (1:1) supplemented with 10% fetal calf serum.
CHO and COS-1 cells were plated 1 day before transfection. Cells (8 -10 ϫ 10 5 /10-cm 2 culture plate) were transfected with 2 g of DNA and 6 l of Fugene (Roche Molecular Biochemicals) and used for experiments the next day. CHO-LPC cells were plated 1 day before the experiments. All experiments were performed at least twice.
Radiolabeling, Immunoprecipitation, and Western Blotting-Cells (8 -10 ϫ 10 5 /10-cm 2 culture plate) were starved for 1 h in methioninefree RPMI 1640 medium and then labeled in the same medium containing 100 Ci/ml [ 35 S]methionine and chased with Dulbecco's modified Eagle's/Ham's F-12 medium (1:1) for the times indicated in the figure legends. In the case of overnight labeling, 5% dialyzed fetal calf serum was added to the labeling medium, and starvation was omitted. Before lysis, the cells were washed with phosphate-buffered saline (PBS) and incubated for 10 min in 20 M N-ethylmaleimide, a sulfhydryl-alkylating agent that blocks all free thiol groups. Lysis and immunoprecipitation were performed as described previously (10) using 5-10 l of anti-LPC polyclonal antibodies directed against the catalytic (MP1) or cytoplasmic (KP1) domain or 2-5 l of an anti-BiP polyclonal antibody (described in Ref. 35 or obtained from Affinity BioReagents). For the experiments in Fig. 3A, cells were lysed in 100 l of 50 mM Tris-HCl (pH 7.8), 150 mM NaCl, and 1% SDS and boiled for 5 min. Subsequently, the samples were diluted with 900 l of 50 mM Tris-HCl (pH 7.8), 150 mM NaCl, 1% Triton X-100, and 1% sodium deoxycholate and used for immunoprecipitation. In some cases, as indicated below, 10 mM DTT was added to the lysis buffer before boiling of the samples. Immunoprecipitates and cell lysates were analyzed by SDS-polyacrylamide gel electrophoresis. Western blots were incubated with 1:2000 dilutions of MP1 or anti-BiP polyclonal antibody, followed by incubation with a peroxidase-conjugated secondary antibody (Dako) and chemiluminescence detection (Renaissance, PerkinElmer Life Sciences).
Separation of Cytosol and Membranes-Cells (8 -10 ϫ 10 5 ) were pulse-labeled for 20 min and subsequently snap-frozen in liquid nitrogen. After quick thawing, the cytosol was gently washed out of the cells with 500 l of PBS; and after one additional wash (discarded), the cells were lysed in 1 ml of the lysis buffer. The cell lysate contained all transmembrane proteins and about half of the soluble ER proteins (using glycosylated pro-L⌬tc as marker protein for soluble ER proteins). The cytosol-containing extract was centrifuged at 13,000 rpm for 10 min at 4°C, and the supernatant was diluted 1:1 with the lysis buffer. Both fractions were used for immunoprecipitation as described above.
Cross-linking-Cross-linking with the reducible cross-linker dithiobis(succinimidyl propionate) (Pierce) was performed essentially as described (36). Cells (8 -10 ϫ 10 5 /10-cm 2 culture plate) were pulse-labeled for 2 h (transient transfections) or overnight (stable cell line) with [ 35 S]methionine, unless indicated otherwise in the figure legends. Cells were placed on ice, washed with PBS, and incubated with N-ethylmaleimide as described above. Subsequently, cells were cross-linked in PBS containing 2 mM dithiobis(succinimidyl propionate) (from a freshly made 20 mM stock in dimethyl sulfoxide) and 0.5% Nonidet P-40 for 30 min on ice. The cross-linker was quenched with 20 mM glycine (final concentration), and the lysate was immunoprecipitated as described above. For the sequential immunoprecipitations carried out for Fig. 2, the Sepharose beads were boiled in 100 l of 50 mM Tris-HCl (pH 7.8), 150 mM NaCl, and 1% SDS containing 10 mM DTT. The eluate was subsequently transferred to a clean tube; diluted in 900 l of 50 mM Tris-HCl (pH 7.8), 150 mM NaCl, 1% Triton X-100, and 1% sodium deoxycholate; and used for the second immunoprecipitation.
Sucrose Gradients-Velocity sedimentation was performed based on a previously described protocol (37). Cells (8 -10 ϫ 10 5 ) were lysed in 400 l of PBS containing 1% Triton X-100 and protease inhibitors (Complete medium, Roche Molecular Biochemicals) and loaded on top of a 10 -40% (w/v) linear sucrose gradient in PBS containing 0.1% Triton X-100. Gradients were centrifuged for 12 h at 45,000 rpm in an SW 55 rotor at 4°C. 500-l fractions were collected from the top of the gradient and either diluted 3-fold in PBS and used for immunoprecipitation or precipitated with 4 volumes of methanol and used for Western blotting. Pellets used for immunoprecipitation were resuspended in 100 l of 50 mM Tris-HCl (pH 7.8), 150 mM NaCl, and 1% SDS; boiled for 5 min; and diluted with 900 l of 50 mM Tris-HCl (pH 7.8), 150 mM NaCl, 1% Triton X-100, and 1% sodium deoxycholate. Pellets used for Western blotting were resuspended in sample buffer and boiled for 5 min. Parallel gradients containing 100 g of bovine serum albumin or thyroglobulin were run as molecular markers. Aliquots of the fractions were run on SDS-polyacrylamide gel and stained with Coomassie Blue. (10), we noticed that biosynthesis of LPC occurred via an unusual pattern. At early time points, two bands of 92 and 102 kDa were observed, whereas at later time points, the 102-kDa protein disappeared with the concomitant increase of a 90 -92-kDa band. The latter protein was shown to represent mature LPC, which lacks the propeptide of ϳ100 amino acids. Deglycosylation experiments performed on samples from early time points revealed that both the 92-and 102-kDa bands represented pro-LPC, with the former one being unglycosylated. To elucidate the molecular basis of the unglycosylated form, we decided to analyze its subcellular localization. As shown in Fig. 1 (left  panel), glycosylated pro-LPC was found exclusively in the crude membrane fraction, whereas the vast majority of unglycosylated LPC was found in the cytosol. This raised the question of whether the unglycosylated form had never been translocated into the lumen of the ER or if it had been retrotranslocated from the ER for ER-associated degradation in the cytosol (38,39). We noticed that the ATG start codon does not contain a Kozak consensus sequence (CTGATGC in the LPC gene instead of G/AXXATGG) (38). In contrast, the second methionine contains a favorable Kozak consensus sequence (GTCATGG). To determine if the second ATG codon might be utilized as a translation initiation codon, we substituted it for an isoleucine codon (ATA). This resulted in the total disappearance of the unglycosylated form ( Fig. 1, right panel), showing that the unglycosylated pro-LPC detected in these transfections is the result of translation initiation at the second ATG codon. This methionine is located 35 amino acids carboxyl-terminal of the first methionine, two amino acids aminoterminal of the major signal peptide cleavage site, and five amino acids amino-terminal of the minor signal peptide cleavage site (11). Translation initiation at this ATG codon will result in the biosynthesis of a protein without a functional signal peptide. Therefore, it will not be translocated into the lumen of the ER or glycosylated. It will comigrate, however, with deglycosylated pro-LPC that has been translocated into the ER and whose signal peptide has been cleaved.

Biosynthesis of Cytosolic Pro-LPC-In a previous study
Interactions of Pro-LPC with BiP-Autoproteolytic cleavage of the propeptide of LPC is a slow process, with a half-life of Ͼ1 h. To obtain insight into the molecular basis for this slow maturation, we investigated potential interactions with molecular chaperones using chemical cross-linking. As shown in Fig.  2 (first panel), transiently transfected wild-type LPC and the active-site serine mutant Lser were cross-linked with a 78-kDa protein. The same result was achieved using a cell line stably expressing moderate amounts of LPC (Fig. 2, third panel), indicating that this interaction was not due to a high-level expression in transient transfections. The cross-linking with Lser showed that this interaction occurred in the ER since this mutant is autoprocessing-incompetent and therefore retained in the ER (10). Based on its molecular mass, we speculated the identity of the protein to be BiP. BiP is a molecular chaperone that binds transiently to unfolded proteins and more stably to misfolded proteins (39). Therefore, the experiment was repeated using an anti-BiP antibody (Fig. 2, second panel), which resulted in co-immunoprecipitation of a protein with the same electrophoretic mobility as pro-LPC, but not LPC. Final proof was obtained by sequential immunoprecipitations (Fig. 2,  fourth panel). After the first immunoprecipitation with either anti-BiP or anti-LPC antibody, the samples were boiled under reducing conditions and immunoprecipitated with anti-LPC or anti-BiP antibody, respectively. Taken together, these results unequivocally demonstrate that BiP is noncovalently bound to pro-LPC.
Aggregation of Pro-LPC-Subsequently, transient cotransfections of full-length LPC with the truncated mutant L⌬c, which lacks the cytoplasmic domain of LPC, were performed (Fig. 3A). The antibody KP1 is directed against Cys 585 -Ser 604 in the cytoplasmic domain (10) and therefore does not recognize L⌬c. L⌬c is, however, recognized by MP1, which is directed against Lys Ϫ2 -Arg 18 in the catalytic domain of LPC (10). Immunoprecipitation of cotransfected Lser and L⌬c using MP1 (Fig. 3A, fifth lane) showed the expected bands: pro-Lser (glycosylated and unglycosylated), pro-L⌬c (glycosylated and unglycosylated), and processed L⌬c. Lser was used in this experiment rather than LPC because processed LPC has only a slightly higher molecular mass than pro-L⌬c, which might obfuscate the results. The use of Lser did not have any qualitative effect (compare with Fig. 3B). When the immunoprecipitation was performed with KP1 (Fig. 3A, sixth lane), which recognizes only pro-Lser, glycosylated pro-L⌬c was found to coprecipitate. The observation that unglycosylated pro-L⌬C and processed L⌬c do not coprecipitate provides evidence that the coprecipitation of glycosylated pro-L⌬c is not due to a post-lysis effect or direct binding of the antibody to pro-L⌬c. This unexpected result suggests that pro-Lser and pro-L⌬c form multimers or aggregates. To characterize the nature of the interactions, the cell extracts were boiled for 5 min in the presence or absence of 10 mM DTT before immunoprecipitation. As shown in Fig. 3A (last four lanes), heat denaturation of the sample was insufficient for dissociation of the complexes. The coimmunoprecipitated proteins were dissociated only when the samples were reduced. These data demonstrate that pro-LPC forms complexes linked via interchain disulfide bonds. Next, we investigated if coimmunoprecipitation was dependent on transmembrane anchorage of LPC or on the presence of the stalk domain, i.e. the region between the middle and transmembrane domains (Fig. 3B). We also tried to obtain a first indication about the size of the complexes by transfecting three times less wild-type LPC than mutant cDNA. The fifth and sixth lanes confirm that the results obtained with Lser in Fig.  3A are similar to those obtained with wild-type LPC. In addition, the sixth, eighth, and tenth lanes show more radiolabeled pro-L⌬c, pro-L⌬tc, and pro-L⌬stc than pro-LPC. In other words, on average, Ͼ1 molecule is coimmunoprecipitated per molecule of pro-LPC, indicating the formation of complexes containing multiple LPC molecules. Finally, the last four lanes demonstrate that mutants missing both the stalk and trans- membrane regions of LPC still form mixed-disulfide complexes with full-length LPC, suggesting that associations were mediated through the luminal domains of LPC. It should be noted that L⌬stc displayed very limited processing even though it contained six amino acids in addition to the topologically equivalent minimal sequence required for maturation of furin (40). Cotransfection of this truncated mutant seemed to affect maturation of LPC as well since less of the 90 -92-kDa form of LPC was observed.
Analysis of LPC by Velocity Sedimentation-The coimmunoprecipitation experiments described above suggest the formation of aggregates. To obtain insight in the size of these aggregates, we analyzed cell extracts of both transiently and stably transfected CHO cells by velocity centrifugation on 10 -40% sucrose gradients. Under steady-state conditions, the vast majority of LPC in the stably transfected cells (Fig. 4A) was in the processed form, concentrated in the third fraction and, to a lesser extent, in fractions 2 and 4. Small amounts of pro-LPC were detected in fractions 3 and 4, with little or none in the bottom fractions. In parallel gradients, bovine serum albumin (66 kDa) was concentrated in fractions 2 and 3. This suggests that under these conditions, most of the pro-LPC and mature LPC (90 -92 kDa) is in the monomeric form, although we cannot exclude that the extraction of the proteins from the cells has disrupted di-or multimeric forms, held together by weak interactions. In contrast, when the transiently transfected cells were examined in the same way (Fig. 4B), only a small fraction of pro-LPC was detected in fractions 3 and 4. The majority of pro-LPC was detected in the pellet, with a trace in fraction 10. Apparently, under these conditions, most of the pro-LPC is present in large aggregates over 660 kDa. However, most of the LPC in these cells is soluble, suggesting that a portion of pro-LPC must remain soluble for processing. Furthermore, these data suggest that aggregation occurs in the ER since most of the aggregated protein is in the pro-LPC form.
Since little non-aggregated pro-LPC was detected under steady-state conditions, we used radiolabeled CHO cells to further investigate the maturation of LPC (Fig. 5). After overnight labeling (Fig. 5A), the distribution was similar to that shown in Fig. 4B, with most of the pro-LPC in the pellet and LPC in fractions 3 and 4. However, after 1 h of labeling (Fig.  5B), we found, in addition to the aggregated pro-LPC, a substantial amount in fractions 3-6 and smaller amounts in fractions 7-10. This indicates that dimers, trimers, and, to a lesser extent, small oligomers have formed.
Under these conditions, cross-linked cell extracts were separated on a sucrose gradient (Fig. 5C) and immunoprecipitated with anti-BiP antibody. BiP was found in all fractions except fractions 1 and 2, with the highest levels occurring in fractions 6 -8, which coincides with the highest levels of coimmunoprecipitated pro-LPC. A small amount of BiP was found in the pellet together with LPC. Most of the pro-LPC protein associated with BiP was found in the fractions with molecular masses between ϳ200 and 600 kDa. This suggests that several BiP molecules are bound to the pro-LPC dimers and trimers. It is of interest that small amounts of LPC were detected in fractions 6 and 7. Combined with the results in Fig. 5B, where some LPC was found in higher molecular mass fractions (fractions 5 and 6), this suggests that processing of pro-LPC to LPC might occur while still in dimeric or trimeric form and associated with BiP. It should be mentioned, however, that this represents only minor amounts (see also Fig. 2). In conclusion, these results show that BiP binds primarily to non-aggregated pro-LPC.

Inhibition of Processing of Pro-LPC by BiP and a BiP ATPase Mutant-
The results obtained with the velocity sedimentation experiments suggested that BiP assists in the folding and maturation of pro-LPC and that failure to bind BiP results in pro-LPC aggregation. In stably transfected cells with moderate expression levels of pro-LPC, sufficient BiP is present to assure complete conversion of pro-LPC, whereas in transiently transfected cells, the high levels of expression saturate the folding machinery, leading to aggregation of interdisulfide chainlinked complexes. To study the role of BiP in the folding and processing of pro-LPC in more detail, we performed experiments using wild-type BiP and the BiP ATPase mutant G37 (Fig. 6). This mutant has retained the capacity for high-affinity binding of polypeptides, but does not undergo a conformational change upon ATP binding, resulting in a failure to release substrate proteins (41). Overexpression of this mutant in African green monkey COS-1 cells, in which endogenous BiP is not recognized by the hamster-specific antibody, resulted in the detection of two immunoreactive bands (Fig. 6A). Wild-type BiP was expressed as one band corresponding to the lower band seen after expression of the BiP mutant. The upper band probably represents a difference in glycosylation. 2 This mutantspecific band allowed us to estimate the level of overexpression compared with endogenous BiP in CHO cells (Fig. 6B). Based on immunofluorescence results, we usually observed transfection efficiencies of Ͼ50% in CHO cells (data not shown). Therefore, we estimated that the ratio of endogenous BiP versus mutant BiP in transfected cells was 2:1 to 3:1. This moderate overexpression of the BiP mutant appeared to be quite potent in the inhibition of pro-LPC processing (Fig. 6C). In the absence of recombinant BiP, ϳ70% of the immunoprecipitated LPC was in the processed form. After overexpression of wild-type BiP, maturation was reduced to 50%; and in the presence of mutant BiP, maturation was further reduced to ϳ30%. These results show that maturation of LPC is reduced when binding to BiP is enhanced. DISCUSSION Previously, we have shown that maturation of LPC is slow (10), in contrast to most other convertases. In addition, we noticed the presence of an unglycosylated precursor. In this study, we have investigated early stages of biosynthesis to unravel the molecular mechanism underlying this slow maturation. We have found that the unglycosylated precursor is not translocated. The most likely explanation for this observation is the lack of a strong Kozak consensus sequence for initiation of translation at the ATG start codon, thus leading to an alternatively initiated LPC isoform, starting 36 codons downstream and encoding an LPC protein without a signal sequence. Alternatively, it has been shown that a number of misfolded or unassembled proteins are recognized by ER chaperones including BiP and are subsequently retrotranslocated, deglycosylated, and degraded via the proteasomal pathway (42,43). Given the binding of pro-LPC to BiP (described below) and the tendency of LPC to aggregate, this possibility was examined. However, after mutation of the second ATG codon, the unglycosylated precursor could not be detected anymore, showing that aberrant translation initiation rather than retrotranslocation formed the basis of this phenomenon.
Newly synthesized LPC was found to form noncovalent interactions with BiP. BiP is an ER chaperone with multiple functions (for a recent review, see Ref. 44). BiP was initially identified as an immunoglobulin heavy chain-binding protein (45) and as a glucose-regulated protein, GRP78 (46). It binds transiently to newly synthesized proteins in the ER to keep them in a folding-competent state. It stably binds misfolded or unassembled proteins, preventing their transport to a post-ER compartment. BiP has also been shown to play a role in translocation, retrotranslocation, and sealing of the luminal end of the translocon pore before and early during translocation (47)(48)(49). BiP recognizes heptapeptide motifs and prefers those with aliphatic residues, although binding is dependent on the rate and stability of folding and not simply on the presence of such consensus sequences (50,51).
Mainly pro-LPC was found to bind to BiP, consistent with the fact that BiP does not bind to folded proteins. Since propeptide cleavage of LPC is an intramolecular process occurring in the ER (10), it requires an active, correctly folded protein. BiP was found to bind mainly the non-aggregated form, suggesting that it plays a role in preventing pro-LPC aggregation. In transiently transfected cells, the high level of expression probably saturates the folding machinery, resulting in aggregation of part of the newly synthesized protein. Shortly after biosynthesis, a substantial amount of pro-LPC was detectable in the fractions of the sucrose gradient containing 100 -300-kDa proteins, probably as dimers and trimers. Oligomeric complexes were also formed, albeit at lower amounts. Small amounts of processed LPC appeared to coimmunoprecipitate in the fractions containing dimers and trimers. This might reflect that sufficient folding can be achieved to allow autoproteolytic processing while still part of a dimer or trimer. At later stages, LPC is found only as monomer. The large amount of cross-linked BiP in Fig. 2 and the size of the cross-linked pro-LPC⅐BiP complexes (200 -600 kDa) in Fig. 5 indicate the simultaneous interaction of several BiP molecules with pro-LPC and pro-LPC dimers or trimers. Binding of multiple BiP molecules has been reported for thyroglobulin, which folds slowly and can bind up to 10 BiP molecules (52). It has also been reported that pro-PACE4 forms dimers (20). In that study, transfected PACE4 was cross-linked with a non-reducible cross-linker, and both monomeric and dimeric pro-PACE4 proteins were detected. However, whether BiP-containing higher molecular mass complexes were formed was not investigated.
Like other HSP70 proteins, BiP has an ATPase-and substrate-binding domain. BiP can undergo cycles of binding and release from substrates, which is dependent on adenine nucleotide exchange. Here we have used a BiP mutant (53) that has lost ATPase activity and binds substrates with high affinity, but slow dissociation. Binding of pro-LPC to this mutant (and to a lesser extent, to overexpressed wild-type BiP) reduced propeptide processing. The relatively low expression of this mutant compared with endogenous BiP explains the incomplete inhibition of maturation. The reduced processing upon overexpression of wild-type BiP might seem to contradict the role of BiP as a molecular chaperone. However, it should be kept in mind that folding is a dynamic process, often requiring several on/off cycles of more than one chaperone (54). For instance, the thiol oxidoreductases protein-disulfide isomerase and ERp72 might play a role in isomerization of nonnative disulfide bonds. Overexpressed BiP might compete with other chaperones for unfolded pro-LPC, hence reducing folding.
The results of this study suggest that pro-LPC is prone to aggregation and that binding of BiP prevents it from doing so. On the other hand, it is likely that BiP binding causes slow maturation and export from the ER, as was also observed for thyroglobulin (55).
The high-mass pro-LPC⅐BiP aggregates observed in transiently transfected cells are covalently linked by disulfide bonds and are present in relatively large amounts compared with cells that stably overexpress recombinant LPC. Combined with pulse-chase experiments (10), where only part of the newly synthesized pro-LPC was processed in transient transfections, whereas most of the pro-LPC was cleaved in stable cell lines, we conclude that formation of these aggregates is a dead-end pathway.
The results obtained in this study might also apply to other PCs. It has been shown for most convertases that propeptide cleavage is a prerequisite for exit from the ER. Binding to BiP would be sufficient to prevent forward transport of the unprocessed precursor. The rapid folding and propeptide cleavage of some convertases like furin and PC1 suggest only very transient interactions, which might be difficult to detect. Active-site mutants, on the other hand, are unable to exit the ER and FIG. 6. Binding of BiP to LPC reduces maturation. Cells were transiently transfected with wild-type or ATPase mutant hamster (Bi-Pmut) BiP cDNA as indicated. Ϫ, transfected with empty vector. A, shown is a Western blot of COS-1 cell lysates. Note that the hamsterspecific anti-BiP antibody did not recognize endogenous BiP. B, shown is a Western blot of CHO cell lysates, in which endogenous BiP was also recognized by the antibody. C, CHO cells were cotransfected with LPC, pulsed for 1 h, and chased for 1 h, and LPC was immunoprecipitated. might be instrumental to study BiP interactions. In contrast, pro-PC2 is able to leave the ER, and no interaction with BiP was detected (27), although it should be mentioned that in this study, no cross-linker was used. Pro-PC2 was found to bind 7B2, however, which acts as a helper protein involved in pro-PC2 transport and activation in the TGN. Expression of endogenous LPC is below detection levels under the conditions used in this study. Therefore, all experiments were performed using recombinant LPC. It might be argued that the slow maturation of LPC and the binding of BiP are due to the absence of sufficient amounts of an LPC-specific chaperone. Although we cannot exclude this possibility, it seems unlikely for several reasons. First of all, maturation of recombinant furin is fast, whereas endogenous expression is very low as well (31). Furthermore, the overexpression of PC1 in cell lines that do not express endogenous PC1 does not seem to affect its fast propeptide cleavage (17). Finally, the slow maturation of recombinant PACE4 was similar to its endogenous counterpart (32). It therefore seems that overexpression of proprotein convertases does not have a pronounced effect on their maturation kinetics. The binding of BiP might reflect a somewhat unstable folding of newly synthesized LPC, inhibiting aggregation by keeping it in a folding-competent state, resulting in efficient but slow maturation.