Maturation of Lipoprotein Lipase in the Endoplasmic Reticulum

The maturation of lipoprotein lipase (LPL) into a catalytically active enzyme was believed to occur only after its transport from the endoplasmic reticulum (ER) to the Golgi apparatus. To test this hypothesis, LPL located in these two subcellular compartments was separated and compared. Heparin affinity chromatography resolved low affinity, inactive LPL displaying ER characteristics from a high affinity, active fraction exhibiting both ER and Golgi forms. The latter forms were further separated by β-ricin chromatography and were found to have comparable activities per unit of LPL mass. Thus, LPL must reach a functional conformation in the ER. Active LPL, regardless of its cellular location, exhibited the expected dimer conformation. However, inactive LPL, found only in the ER, was highly aggregated. Kinetic analysis indicated a concurrent formation of LPL dimer and aggregate and indicated that the two forms have dissimilar fates. Whereas the dimer remained stable even when confined to the ER, the aggregate was degraded. Degradation rates were not affected by proteasomal or lysosomal inhibitors but were markedly reduced by ATP depletion. Lowering the redox potential in the ER by dithiothreitol caused the dimer to associate with calnexin, BiP, and protein-disulfide isomerase to form large, inactive complexes; dithiothreitol removal induced complex dissociation with restoration of the functional LPL dimer. In contrast, the LPL aggregate was only poorly associated with ER chaperones, appearing to be trapped in an irreversible, inactive conformation destined for ER degradation.

Most circulating dietary-and liver-derived fatty acids arrive at the heart, adipose tissue, and skeletal muscle esterified as triglycerides located within the hydrophobic core of chylomicrons and very low density lipoproteins. At the luminal surface of tissue capillaries, the esterified fatty acids are then released following the hydrolysis of core triglycerides by lipoprotein lipase (LPL). 1 LPL is synthesized by the tissue parenchymal cells and is secreted to the surface of the capillary endothelium, where it binds to heparan sulfate proteoglycans (1,2). The free fatty acids released by the action of LPL are subsequently taken up by the subjacent tissue for storage or oxidation.
The acquisition of enzymatic activity by newly synthesized LPL, a process known as maturation, has remained an unresolved question. Mature, fully functional LPL is a homodimer that binds heparin with high affinity (2,3). As a secreted protein, nascent LPL is vectorially transported from its site of synthesis at the cytoplasmic face of the ER into the lumen, where asparagine-linked glycosylation takes place. It has been well documented that initial glucose trimming from these nascent glycan chains is a prerequisite for acquisition of LPL catalytic activity (4 -6). This requirement suggests that, like most glycoproteins, newly synthesized LPL interacts with the folding chaperones calnexin or calreticulin. Binding to these chaperones occurs through the glucose residue remaining on the core glycan chain following the action of glucosidases I and II (7,8). However, whether LPL acquires its fully active form early in the ER or requires transport to the Golgi complex for maturation has been the subject of contradictory reports. Thus, Vannier and Ailhaud (9) postulated that LPL existed in the ER as an inactive 55-kDa monomer precursor; assembly into dimers and acquisition of catalytic activity appeared to occur only after transport to the Golgi apparatus. In a later study, following de novo synthesis of LPL in the presence of carbonyl cyanide m-chlorophenylhydrazone (CCCP), Park et al. (10) established that LPL dimerized and attained high affinity for heparin in the ER, although it exhibited no detectable enzyme activity. In adipocytes from cld/cld mice, as well as in normal adipocytes treated with an inhibitor of glucosidases I and II, LPL was found inactive and, based on its glycan structure, appeared to be present in the ER. The lack of activity in both cases was attributed to a defect in transport to the Golgi (6). Likewise, long term treatment of adipocytes with chlorate resulted in the accumulation of inactive LPL in the ER, suggesting that chlorate treatment impeded translocation to the Golgi, where final LPL maturation presumably occurred (11).
In contrast to these studies, other reports suggested that translocation to the Golgi was not necessary for attainment of LPL catalytic activity. For example, LPL analyzed in perfused hearts or in adipocytes from guinea pig revealed the existence of a significant pool of active LPL that was endo H-sensitive, suggesting its localization in the ER (3,12). Similar results were attained in CHO cells, following arrest of ER-Golgi transport at 16°C, or by affixing the ER retention signal KDEL to the LPL carboxyl terminus (4). The conditions used in the latter studies did not affect the normal protein folding mechanisms in the ER, unlike the use of chaperone binding inhibitors or energy-depleting agents such as CCCP. In fact, in a recent study of LPL in the Cld cell line (derived from cld/cld mouse fibroblasts) or in Lec23 (a glucosidase I-deficient CHO cell line), it was shown that LPL was retained in the ER because the mutations rendered it misfolded and transport-incompetent rather than a defect in transport impeding LPL maturation (13).
In the present study, a prime objective was to verify whether LPL maturation occurs in the ER by attempting to isolate a fully functional LPL fraction that exhibited only ER processing. Our characterization of such a form corroborates the fact that LPL becomes fully active in the ER, and only this transportcompetent form enters the Golgi en route to secretion. Our findings also establish the existence of an inactive LPL pool in the ER. However, this pool is not a precursor to the active form and does not acquire a native conformation. Rather, it contains "off-pathway," misfolded LPL molecules that form large aggregates. These aggregates are degraded in the ER, conforming to the stringent quality control mechanism of the cell. However, when the cells were stressed by reducing disulfide bonds, a large, inactive LPL complex exhibiting different properties from the aggregate was formed. Under these conditions, the functional dimer itself coalesced, losing enzyme activity in the process. Unlike the normal aggregates occurring in control cells, the stress-generated complex was found to be LPL arrayed with several folding chaperones; upon return of the cells to normal conditions, the complex dispersed to reform fully functional, transport-competent enzyme.

EXPERIMENTAL PROCEDURES
Cell Lines and Media-All cell cultures were maintained in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with either 10% fetal bovine serum (for CHO and EcR-CHO cells) or 10% calf serum (for 3T3-L1 adipocytes), antibiotics (50 units/ml penicillin and 50 g/ml streptomycin), 0.1 mM nonessential amino acids, and 1.0 mM sodium pyruvate. CHO cells were the proline auxotroph derivatives (Pro5) obtained from the ATCC (Manassas, VA). The ecdysone-inducible EcR-CHO cell line (Invitrogen) was maintained in selection medium containing 0.25 g/ml Zeocin TM . The 3T3-L1 preadipocyte cell line was a generous gift from Dr. Miklos Peterfy (UCLA). These cells were differentiated into adipocytes by the addition of 5 M dexamethasone, 0.5 mM methylisobutylxantine, and 10 g/ml insulin.
Expression Constructs-The termination codon of human LPL cDNA (nucleotides 426 -428 of coding) was replaced with a BamHI site and cloned into the pcDNA6/V5-His B expression vector (Invitrogen) as a HindIII/BamHI fragment. This placed the 3Ј-end of human LPL in frame with the V5 epitope tag of pcDNA6/V5-His B, which was subsequently used for antibody detection of the expressed LPL protein. The construct was driven by a cytomegalovirus promoter and contained a bovine growth hormone polyadenylation site.
Regulated expression of the epitope-tagged LPL construct was achieved using the pIND(SP1) ecdysone-inducible expression vector (Invitrogen). The entire coding region of human LPL (including the epitope tag) was isolated as a PmeI fragment from pcDNA6/V5-His B/LPL, and this fragment cloned into the PmeI site of the pIND(SP1). The construct was driven by a ponasterone A-inducible promoter containing three cis-acting SP1 elements, the latter elements incorporated to achieve enhanced expression levels. The pIND(SP1) was used in combination with the EcR-CHO cell line, which stably expresses a functional ecdysone receptor comprising the human retinoid X receptor and a modified version of the Drosophila ecdysone receptor (VgEcR).
The coding region of the asialoglycoprotein H2b receptor cDNA was amplified by RT-PCR from HepG2 RNA, and the natural termination codon was mutagenized and subcloned in pcDNA6/V5-His (Invitrogen) in frame to add the C-terminal V5 epitope tag. The coding region of the construct was sequenced in its entirety and used to transiently transfect CHO cells.
Transfection, Selection, and Harvesting of Cells-CHO and EcR-CHO cells were stably transfected with pcDNA6/V5-His LPL and pIND-LPL, respectively, using the calcium phosphate method (14). The selection medium contained 10 g/ml blasticidin S HCl (CHO) or 600 g/ml hygromycin (EcR-CHO). LPL-expressing colonies were identified by assaying medium for LPL activity following incubation for 16 h with 10 units/ml heparin (CHO), or 10 units/ml heparin and 10 g/ml ponasterone A (EcR-CHO). LPL-expressing EcR-CHO colonies were further screened for their potential to undergo inducible expression in response to increasing concentrations of ponasterone A (1-10 M; 16-h incuba-tion). Only colonies that responded with at least a 4-fold increase in LPL activity were selected. pcDNA6/V5-His H2b was transiently transfected into CHO cells using the reagent FuGENE 6 (Roche Molecular Biochemicals), according to information supplied by the manufacturer. Based on a ␤-galactosidase control expression vector, about 20 -30% of CHO cells were transfected using this reagent. The CHO cells were analyzed 24 h post-transfection.
Cells were subcultured onto 100-mm plates and propagated in 8 ml of medium; however, experiments were conducted using 3 ml of medium to increase the concentration of secreted LPL. Unless otherwise indicated, fresh medium containing heparin (10 units/ml; Sigma) was added to the culture medium 4 h prior to experiments and was also included at the same concentration during experimental procedures. Incubations of LPL-transfected cells with various inhibitors were carried out for up to 3 h, as the secretion of the dimer and turnover of the aggregate occur with a half-life of about 30 min. Experiments with the H2b subunit of the asialoglycoprotein receptor were carried out for 6 h, due to the slower turnover rate of this protein.
Cells were harvested after washing plates twice with phosphatebuffered saline (PBS), scraping contents in 1 ml of PBS and spinning the slurries at 500 ϫ g for 5 min at 4°C. The resulting cell pellets were stored at Ϫ80°C. For analysis, cell pellets from individual plates were suspended in 0.6 ml of lysis buffer (0.2% sodium deoxycholate, 10 units/ml heparin in 10 mM Tris-HCl, pH 7.5) and sonicated for 6. Lysates were centrifuged at 1000 ϫ g for 5 min at 4°C to remove any remaining cellular debris.
Chromatography and Sucrose Gradient Centrifugation-Heparin Sepharose chromatography was carried out as previously described (13). For ␤-ricin chromatography of the active, high affinity LPL fraction obtained after heparin-Sepharose chromatography, Ricinus communis agglutinin (RCA 120 ) linked to agarose (Sigma) was utilized. The column buffer was 10 mM Tris-HCl, pH 7.5, containing 0.1% Triton X-100, and 1.0 M NaCl, used to mimic the elution buffer of the preceding heparin-Sepharose step. The sample was pretreated with neuraminidase (see below), and the unbound LPL was collected as the column flow-through; the bound LPL was eluted with 0.2 M galactose (Sigma) in column buffer.
Rate zonal centrifugation of LPL was carried out using a 12-ml, 5-20% linear sucrose gradient as described previously (15). Generally, the sample applied to each gradient consisted of a cell lysate derived from one plate; centrifugation was carried out at 200,000 ϫ g for 22 h at 4°C. Following centrifugation, fractions of 0.48 ml were manually collected from the top. To ensure maximal recovery of the large LPL aggregate that sedimented to the bottom of the tube during centrifugation, the last fraction was vigorously shaken prior to removal from the gradient tube.
Cell Labeling-LPL-expressing CHO cells were grown in 175-cm 2 flasks to confluence. After washing cells in PBS, prewarmed Dulbecco's modified Eagle's medium containing 0.1 mM glutamate and 1.0 mM sodium pyruvate, but no methionine, cysteine, or serum, was added, and cells were allowed to equilibrate for 20 min. Tran 35 S-label (Amersham Pharmacia Biotech) was added to a final concentration of 130 Ci/ml. The labeling reaction was stopped after 5 or 20 min of pulse time by quickly removing the labeling medium and replacing it with 25 ml of ice-cold stopping solution (PBS containing 15 g/ml cycloheximide). The stopping solution was gently swirled over the cell monolayer for 30 s and removed, and an additional wash with stopping solution was repeated. The cells were harvested by scraping into 10 ml of fresh stopping solution. The cells were centrifuged as described above, and the resultant cell pellets were stored at Ϫ80°C.
Glycosidase Digestions-Samples selected for endoglycosidase H (endo H) digestion were adjusted to pH 5.9 by the addition of 0.1 M sodium phosphate buffer, pH 5.5. If samples contained NaCl concentrations of Ͼ0.75 M, they were diluted at least 2-fold with 50 mM sodium phosphate buffer, pH 5.9. To facilitate endo H digestion, proteins were denatured by adding SDS to 0.5% and heating the sample in a boiling water bath for 2 min. After chilling on ice, 10 milliunits of endo H (Roche Molecular Biochemicals) were added, and the samples were incubated at 37°C for 18 -20 h.
Samples selected for endo-␤-N-acetylglucosaminidase D (endo D) digestion were adjusted to pH 6.5 and denatured with SDS as described above for the endo H digestions. Triton X-100 was then added to a final concentration of 2.5%, followed by the addition of 5 milliunits of endo D (Calbiochem) and incubation at 37°C for 18 -20 h. A positive control (LPL containing only endo D-sensitive, Man 5 oligosaccharides) was always included to ensure that endo D digestion occurred.
Neuraminidase digestion was used for samples subjected to ␤-ricin chromatography (see above). Since these samples contained 1.5 M NaCl, the salt was diluted to 0.75 M by the addition of 10 mM Tris-HCl, pH 7.5, containing 0.1% Triton X-100; 50 milliunits of neuraminidase (Roche Molecular Biochemicals) were added, and the reaction was carried out for 1 h on ice. No loss of LPL activity occurred as a result of neuraminidase digestion.
Co-immunoprecipitation-IgG was isolated from rabbit antisera against calnexin, calreticulin, PDI, BiP, and normal rabbit serum (Stressgen) by protein A-agarose chromatography. Chicken anti-ERp57 IgG was a kind gift from M. Kito (Kyoto University). Samples representing 3% of the LPL aggregate obtained after sucrose gradient centrifugation of untreated or DTT-treated cells were mixed with 40 g of IgG and incubated overnight at 4°C. For the ERp57 sample, 50 g of rabbit anti-chicken IgG (Pierce) was added, and the incubation continued for 2 h. Purified Staph A (a slurry of denatured Staphylococcus aureus membranes containing protein A (see Ref. 16) was then added, and the samples were incubated for 30 min at 4°C with continual mixing using a rotary shaker. Following centrifugation, Staph A pellets were washed twice with lysis buffer (10 mM Tris-HCl, pH 7.5, 0.2% sodium deoxycholate, and 10 units/ml heparin), and bound proteins were dissociated by boiling for 2 min in 10 mM Tris-HCl, pH 7.5, containing 1% SDS. Co-precipitated LPL was detected by Western blotting (see below).
Immunoprecipitation and Western Blotting-LPL immunoprecipitation was carried out using the Fab fragments of chicken anti-bovine LPL IgG as described (17). For Western blot detection of LPL and H2b expressing the C-terminal V5 epitope tag, horseradish peroxidase-conjugated anti-V5 antibody was used (Invitrogen; 1:5000). For detection of endogenous LPL in 3T3-L1 adipocytes, affinity-purified chicken antibovine LPL IgG was used (final concentration, 0.75 g/ml), followed by biotinylated rabbit anti-chicken IgG and horseradish peroxidase-conjugated streptavidin (17). Visualization of detected proteins was carried out using a horseradish peroxidase-sensitive chemiluminescent substrate (SuperSignal®; Pierce). Quantitation of lipase bands was carried out by densitometric scanning of Western blots and autoradiographs using 1D Image Analysis software (Eastman Kodak Co.).
Other Methods-LPL activity in lysates and medium was assayed using a triolein substrate prepared by sonication as described (14). When required, LPL-specific activity (i.e. activity per unit of LPL mass) was calculated by dividing the activity loaded onto electrophoretic gels by the densitometric value of the LPL band determined after Western blot analysis. Since densitometry does not provide an absolute protein concentration, we reported relative specific activity, arbitrarily setting a value of 1.0 for a chosen sample from the blot. Cellular protein was assayed using the BCA Protein Assay kit (Pierce).

Intracellular LPL Is Located
Primarily in the ER-In the absence of heparin in the culture medium, although LPL is secreted from the cells, a significant amount remains attached to the cell surface through interaction with heparan sulfate proteoglycans. Only a fraction of LPL is present in the medium, where it accumulates with time. Thus, after 4 h of incubation in the absence of heparin, about 60% of the total LPL activity expressed in CHO cells was cell-associated, and the remainder was found in the medium (Fig. 1, histograms). The addition of heparin, an analog of heparan sulfate proteoglycans, causes the release of LPL into the medium. Under these conditions, after 4 h of incubation, only 20% of the activity was shown to be cell-associated. In the absence of heparin, endo H digestion of the cellular LPL revealed a mixture of forms that were either resistant (57 kDa), partially resistant (55 kDa), or fully sensitive (52 kDa) to glycan removal ( Fig. 1, subjacent Western blots). The 57-and 55-kDa forms represent LPL that is either present in or has traversed the Golgi apparatus, where one or both of its glycan chains have undergone processing to the endo H-resistant form. The 52-kDa form represents LPL residing in the ER, bearing only fully sensitive, high mannose glycan chains. In contrast to the abundant level of LPL bearing complex glycan structures observed in the absence of heparin, cells incubated in the presence of heparin displayed a substantial reduction in the levels of these Golgi-processed forms. Apparently, the majority of the Golgi-processed LPL formerly at-tached to the cell surface was released by heparin into the medium. The strictly intracellular Golgi-processed LPL pool must be relatively small; thus, almost all of the intracellular mass is composed of LPL in the ER.
The ER Contains both Inactive and Active Forms of LPL-As lysates of heparin-treated cells contained predominantly high mannose LPL, it was reasonable to attribute the catalytic activity expressed in those cells to enzyme residing in the ER (see Fig. 1). However, since published evidence suggested that LPL in the ER was inactive (6,10,11), it was possible that the expressed activity was solely derived from the small Golgi pool present in these cells.
As a first step in resolving this uncertainty, LPL from CHO cells incubated for 4 h in heparin was subjected to heparin-Sepharose chromatography. We have shown previously that this procedure can separate inactive from active LPL by elution at increasing ionic strengths (13). Indeed, as shown in Fig. 2 (top), an inactive LPL fraction eluted at a lower salt concentration (0.75 M NaCl) than active LPL, which bound with higher affinity and eluted at 1.5 M NaCl (see activity profile and the subjacent Western blot of eluted LPL protein). Endo H digestion of the peak fractions revealed that low affinity, inactive LPL was exclusively located in the ER, whereas high affinity, active enzyme was located in both ER and Golgi compartments.
To determine whether LPL in both ER and Golgi contributed to the catalytic activity of the high affinity fraction, the two LPL forms were separated based on their organelle-specific glycan structures. This was accomplished by ␤-ricin chromatography. As the ␤-ricin toxin specifically binds galactose, it interacts only with Golgi-processed (complex) glycoproteins, provided that their penultimate galactose residues have been exposed. Accordingly, the terminal sialic acid of LPL from the active sample was removed by neuraminidase digestion prior to its application onto a ␤-ricin-agarose column. It was expected that the unbound fraction would contain LPL exhibiting high mannose glycans, while the bound sample, eluting at 0.2 M galactose, would contain LPL bearing complex sugars.

FIG. 1. Heparin induces redistribution of LPL activity and mass. LPL-transfected CHO cells were incubated for 4 h in medium
containing no heparin or supplemented with heparin at 10 milliunits (mu)/ml. LPL activity was determined in cell lysates (black bars) and media (gray bars) and reported as milliunits of activity/mg of total cellular protein, where 1 milliunit is equivalent to 1 nmol of fatty acid released per min. Samples of lysates containing 2 milliunits of LPL activity were incubated in the absence (Ϫ) or presence (ϩ) of endo H, followed by Western blot analysis for detection of LPL mass and state of glycan processing, respectively.
As shown in Fig. 2 (bottom), ␤-ricin chromatography separated the high affinity fraction into two peaks of activity: one that emerged in the flow-through and a second peak that bound to the column and eluted at 0.2 M galactose. The equivalent of 2 milliunits of LPL activity from each peak was subjected to Western blotting following incubation with or without endo H; the flow-through peak was also subjected to endo D digestion (see below). Endo H digestion confirmed that the flow-through contained only LPL with high mannose glycans, whereas the galactose eluate contained only LPL with complex glycan chains. Scanning densitometry of the undigested bands was carried out in order to compare relative specific activities (LPL activity per unit of enzyme mass), and these were found to be similar: 1.0 for the flow-through and 0.85 for the galactose eluate. Thus, high mannose LPL was at least as active as LPL bearing complex glycan chains. Finally, to ascertain that the high mannose form indeed represented LPL residing in the ER, this fraction was subjected to endo D (endo-␤-N-acetylglucosaminidase D) digestion. This enzyme cleaves only the Man 5 form of glycan chains, which, although it is endo H-sensitive, is characteristic of glycoproteins that have reached the cis-Golgi compartment and have been trimmed by the Golgi mannosidase I (18). As evident from the resistance of high mannose LPL to endo D digestion, this pool of lipase had not yet reached the cis-Golgi and must represent LPL located in the ER. Consequently, the ER must possess the machinery necessary for maturation of LPL to its fully functional form. Strikingly, however, the ER also contained a substantial pool of inactive LPL that eluted from heparin-Sepharose at 0.75 M NaCl (Fig. 2, top). Thus, our next objective was to characterize the inactive form of LPL residing in this compartment.
Active LPL Is a Dimer, whereas Inactive LPL Is Highly Aggregated-It has been widely accepted that LPL with low affinity to heparin is a monomer (19 -23). Therefore, it was possible that inactive LPL eluting from heparin-Sepharose at 0.75 M NaCl (see Fig. 2, top) was a monomer. Moreover, since we found the inactive form of LPL to be exclusively in the ER, a putative monomer in this compartment might represent the precursor of the dimeric, active form as suggested by others (9). To verify whether inactive LPL was indeed a monomer, the low affinity (0.75 M NaCl) fraction was subjected to density gradient centrifugation. Previous studies (15) have shown that when subjected to centrifugation at 200,000 ϫ g for 22 h through a 5-20% sucrose gradient, active LPL sediments to around 11.5% sucrose (corresponding to fractions 10 -12), consistent with its dimeric molecular mass of 110 kDa. Under these conditions, a monomer would be expected to sediment slightly earlier, at around 9% sucrose (fractions 7 and 8). However, as shown in Fig. 3A, all LPL eluting from heparin-Sepharose at 0.75 M NaCl sedimented to the bottom of the gradient during centrifugation, indicating a composition of high order aggregates (Ͼ500 kDa).
To rule out the possibility that LPL aggregation resulted from prolonged centrifugation, the procedure was repeated and used to compare the sedimentation profile of LPL from CHO cell lysates with that of LPL secreted into the medium. Previous experiments have indicated that unlike cellular LPL, all secreted lipase protein was active and eluted from heparin-Sepharose with the characteristics of the high affinity form. 2 Thus, medium should contain only dimeric LPL, without the presence of an aggregate. Indeed, while active LPL in both cells and medium migrated as a dimer, only the cellular sample revealed an aggregate that was devoid of activity (Fig. 3B). Thus, LPL aggregation was not an artifact of centrifugation. Moreover, while endo H digestion of the dimer revealed that it contained both high mannose and complex forms, the aggregate was exclusively high mannose (see the endo H analysis above the peak fractions in Fig. 3B). Thus, the inactive LPL aggregate appeared to be recognized as unfit for transport, causing its retention in the ER. Lysates of cells preincubated with heparin for 4 h were subjected to heparin-Sepharose chromatography. The column was washed with 0.28 M NaCl, and bound LPL was sequentially eluted with 0.75 M NaCl and 1.5 M NaCl. Fractions were analyzed for LPL activity and mass (graph and subjacent Western blots). Note that elution of low affinity LPL protein emerging at 0.75 M NaCl does not exhibit activity; LPL activity is associated solely with the high affinity form eluting at 1.5 M NaCl. Aliquots of the peak LPL fractions were subjected to endo H analysis for assessment of subcellular location. Low affinity LPL displayed only high mannnose glycans, consistent with an exclusive ER location. While a high mannose form was also present in the 1.5 M NaCl fraction, the additional presence of a complex form in this fraction indicates that high affinity LPL is located in both ER and Golgi compartments. B, separation of high affinity LPL into ER and Golgi forms. ␤-Ricin chromatography was used to isolate LPL into ER and Golgi forms based on organelle-specific glycan structures. The LPL complex form, containing a penultimate galactose residue due to Golgi processing, binds ␤-ricin with high affinity after the removal of terminal sialic acid residues by neuraminidase (see "Experimental Procedures"). It remains bound to the column until elution with 0.2 M galactose (arrow). In contrast, the LPL high mannose form characteristic of ER glycoproteins emerges as unbound protein in the column flow-through. Note that both the high mannose and complex high affinity forms of LPL exhibit activity, suggesting that both possess a native conformation. The high mannose LPL fraction was also subjected to endo D digestion, to verify that the glycan chains had not yet been trimmed to Man 5 by cis-Golgi mannosidase. Indeed, the resistance of the high mannose form to endo D digestion confirmed its exclusive location in the ER.
The possibility was also considered that aggregation resulted from LPL overexpression since the cell line employed in this study was clonally selected to express high LPL activity. Accordingly, LPL was stably transfected into an ecdysone-inducible CHO cell line (EcR-CHO) and expressed at increasing levels by induction with ponasterone A (24). When LPL in these cells was subjected to sucrose gradient sedimentation, an aggregate was present at all levels of expression, and its amount increased proportionally with increasing levels of the LPL dimer (Fig. 3C). Moreover, a significant pool of inactive aggregate was also detected in differentiated 3T3-L1 adipocytes, cells that are untransfected but endogenously express LPL. Taken together, these experiments rule out the possibility that the aggregate was an artifact resulting from transfection and/or overexpression.
Aggregated LPL Is Not a Precursor of the Native Enzyme-The co-localization in the ER of inactive, aggregated LPL together with its active, native form suggested that they might share a precursor-product relationship. Although aggregation is often the result of irreversible misfolding, it has also been shown to occur as an intermediary step between the nascent polypeptide chain and the properly assembled, native protein (25,26).
To test for a precursor-product relationship, cells were pulselabeled for 5 or 20 min with [ 35 S]methionine. At each time point, the cell lysates were subjected to sucrose gradient cen- A, LPL with low affinity to heparin sediments as an aggregate. Inactive LPL with low affinity to heparin (see Fig. 2 top) was subjected to sucrose gradient centrifugation. Above the indicated gradient fractions (1-26) is a gray box denoting the region where an LPL dimer of 110 kDa is expected to sediment, based on internal molecular weight standards (glucose-6-phosphate dehydrogenase (114 kDa) and malic dehydrogenase (74 kDa)). A putative monomer should sediment slightly ahead of this region, around fractions 7 and 8. Western blots of the resulting fractions demonstrate that, contrary to its expected sedimentation as a monomer, low affinity LPL is highly aggregated. B, only intracellular LPL contains an aggregate that is located exclusively in the ER. Lysate obtained from one plate of LPL-expressing CHO cells and medium representing 20% of the total from one plate were subjected to sucrose gradient centrifugation. Analysis of LPL activity and mass (graphs and subjacent Western blots) revealed a nearly identical profile of active, dimeric LPL. However, only cellular extracts contained an inactive aggregate. Endo H analysis of the intracellular LPL dimer and aggregate demonstrates the presence of ER and Golgi forms in the former but an exclusive high mannose (i.e. ER) structure of the latter (see endo H analysis above peak LPL fractions). C, LPL aggregation is not the result of overexpression and is present in cells endogenously expressing LPL. LPL was stably transfected into the ecdysone-inducible EcR-CHO cell line and expressed at increasing levels by induction with ponasterone A. Cell lysates containing LPL activity over an 8-fold range were subjected to sucrose gradient centrifugation, and the resulting fractions were analyzed for LPL activity and mass. Lanes 1-3 in the top and bottom panels represent the equivalent of 5% of the peak dimer fraction and aggregate fraction, respectively; LPL activity (milliunits (mu)) per lane is also indicated for each dimer fraction. As can be seen at all levels of expression, active LPL sedimenting as a dimer was accompanied by a proportional amount of inactive LPL sedimenting as an aggregate. LPL was similarly analyzed in differentiated 3T3-L1 cells that endogenously express LPL activity (lane 4). Note the substantial amount of LPL aggregate in these cells. trifugation to separate dimeric, functional LPL from the inactive aggregate. Aliquots from the appropriate fractions containing these two separated forms were immunoprecipitated and subjected to endo H digestion. Western blotting was used to estimate LPL mass, and the same blots were then subjected to autoradiography to assess incorporated label.
As shown in Fig. 4 (Western blot panel), steady state levels of the LPL dimer consisted of the expected mixture of Golgi and high mannose forms, whereas the aggregate existed only as its anticipated high mannose form. Analysis of radioactivity indicated that even after a short pulse time of 5 min, the extent of [ 35 S]methionine incorporation was equivalent in the dimer and aggregated forms (see autoradiograph panel in Fig. 4). Both newly synthesized LPL forms were still in the ER, evident from their complete endo H susceptibility. At the longer pulse time (20 min), the dimer contained both endo H-sensitive and -resistant forms, indicating the normal progression of functional LPL to the Golgi; in contrast, the aggregate remained at all times in the ER.
The incorporation of label per unit of LPL mass (i.e. radiospecific activity) at the two pulse times was calculated for LPL in the ER, represented by the high mannose forms. If the aggregate was a precursor of the dimer, it should incorporate label prior to the appearance of label in the dimer. As shown in the graph of Fig. 4, no such delay was found, although, within the time frame of the experiment, an expected lag between a potential precursor and its product could have been detected. 3 This finding kinetically ruled out the possibility that the active dimer in the ER was formed from the inactive aggregate.
Degradation of the Aggregate and Secretion of the Dimer-Since the inactive LPL aggregate was not converted into a dimer and, based on endo H susceptibility, was restricted from advancing into the secretory pathway, it followed that it must be degraded within the ER. To test this hypothesis, the fate of the dimer and aggregate was followed after arrest of de novo protein synthesis. CHO cells pretreated with heparin for 4 h were incubated for increasing time periods with heparin and cycloheximide (Cx). The expected secretion of active LPL was shown by the progressive appearance of increasing amounts of LPL activity in the medium balanced by a proportional decrease of activity in the cells (Fig. 5A). After 1 h, most of the activity could be accounted for in the medium, whereas slightly FIG. 4. Pulse labeling of LPL indicates concurrent formation of LPL dimer and aggregate. LPL-expressing CHO cells were pulselabeled with [ 35 S]methionine/cysteine for 5 or 20 min. Lysates of the labeled cells were subjected to sucrose gradient centrifugation, and the fractions containing dimer and aggregated LPL were identified by Western blotting (data not shown). Aliquots of dimer and aggregated LPL were immunoprecipitated followed by endo H digestion. The resulting samples were subjected to SDS-PAGE followed by immunoblotting with horseradish peroxidase-conjugated anti-V5 antibody. The same blots were subsequently exposed to film. Since the cells employed in this experiment were not preincubated with heparin, note at all times the substantial amount of LPL bearing complex glycan chains in the dimer fraction in contrast to the exclusive high mannose composition of the aggregate (see Western blot, upper panel, indicating LPL levels at steady state). While LPL synthesized during the 5-min pulse period is all in the ER, some of the dimer has reached the Golgi after 20 min (lower panel, autoradiograph). The ratio of radiolabeled LPL to total LPL mass in the ER (i.e. radiospecific activity) was calculated and plotted as a function of pulse time (graph below). Both dimer and aggregate pools in the ER appear to incorporate the 35 S label simultaneously, without any indication of a precursor-product relationship.
FIG. 5. The fate of intracellular LPL: Secretion of the dimer and degradation of the aggregate. Cells pretreated with heparin were incubated in the presence of Cx (15 g/ml) for up to 60 min. A, active intracellular LPL is stable and is eventually secreted. LPL activity was assayed in both cells and medium (see legend to Fig. 1) and normalized to total cell protein (milliunits (mu)/mg). Note that intracellular activity is secreted to the medium without a loss in combined activity, indicating that once formed, active LPL remains stable and does not undergo degradation. B, the inactive aggregate is degraded intracellularly. Cells and medium treated with Cx were subjected to sucrose gradient centrifugation, and LPL in the peak dimer and aggregate fractions was detected by Western blot analysis as described under "Experimental Procedures." Comparison of LPL in cells and medium following inhibition of de novo protein synthesis indicates secretion of the dimer but intracellular degradation of the aggregate. more than 10% remained in the cells. As the combined activity (cells plus medium) remained constant, activity was not lost within the time period analyzed. Indeed, commensurate with activity, the LPL dimer was secreted to the medium without any apparent loss of protein during 60 min (Fig. 5B). However, unlike the dimer, the LPL aggregate disappeared from the cells and did not appear in the medium (Fig. 5B), indicating its degradation in the ER.
We raised the question of whether the selective degradation of the LPL aggregate was a consequence of its retention in the ER, where it was exposed to proteolytic factors, while the dimer "escaped" degradation by being secreted. To test this possibility, the cycloheximide experiment was repeated in the presence of brefeldin A (BFA), an agent that abolishes secretion and causes resorption of Golgi components into the ER (27). As shown in Fig. 6A, while secretion of activity was completely abolished in cells treated with both Cx and BFA, LPL activity remained constant even after 3 h. In accordance with the activity, the amount of dimer protein did not change in the cells, indicating its resistance to degradation (Fig. 6B). In contrast to the dimer, the aggregate underwent degradation, albeit at a rate slower than in the absence of BFA (compare Fig. 6B with Fig. 5B). Since both dimer and aggregate displayed complex glycan forms (Fig. 6C), it was evident that the two pools were exposed to a similar ER environment, containing resorbed Golgi-processing enzymes. Nevertheless, even in the presence of a BFA-altered ER, the cells recognized the nonnative conformation of the aggregate and selectively channeled it toward degradation.
Characteristics of the LPL Aggregate-The degradation of aberrantly folded proteins is most commonly carried out by the cytosolic ubiquitin-proteasomal pathway, following retrotranslocation from the ER (28). To test whether the LPL aggregate was disposed of in this manner, we tested the effect of proteasomal inhibitors on the rate of LPL degradation. As a known model for proteasomal degradation, we used the H2b subunit of the human asialoglycoprotein receptor (29). Although a substantial amount of newly synthesized H2b reaches the cell surface following processing in the Golgi (the mature form), some H2b is retained in the ER (precursor) and proteasomally degraded (30). CHO cells transfected with H2b or LPL were incubated in the absence or presence of 10 M MG-132. As shown in Fig. 7A (upper panel), untreated CHO cells exhibited primarily the mature, cell surface form of the receptor (lane 1). When the cells were treated with MG-132 (lane 2), both precursor and mature forms accumulated to high levels, since the precursor was protected from proteasomal degradation. The effect of MG-132 on H2b turnover was also examined in the presence of cycloheximide. In contrast to untreated cells, in the absence of protein synthesis, the mature form was barely detectable as both mature and precursor forms were turned over (lane 3). However, when MG-132 was added, levels of the mature form increased because of its replenishment by that portion of precursor normally undergoing proteasomal degradation (lane 4). Similar results were obtained in the presence of a 100 M concentration of the inhibitor N-acetyl-leucyl-leucylnorleucinal (data not shown).
Unlike H2b, the levels of the LPL aggregate were not affected by the proteasomal inhibitor. Comparable amounts were detected in the absence or presence of MG-132 (Fig. 7A, lower  panel, lanes 1 and 2), and the aggregate disappeared when protein synthesis was abolished, regardless of the proteasomal inhibitor (lanes 3 and 4). Similar results were obtained with an array of other proteasomal inhibitors: lactacystin (50 M), clasto-lactacystin ␤-lactone (10 M), N-acetyl-leucyl-leucyl-methional (100 M), and N-acetyl-leucyl-leucyl-norleucinal (100 or 200 M). Thus, the LPL aggregate must be degraded in the ER via a nonproteasomal pathway. Lysosomal degradation of the aggregate was also considered; however, no effect was seen when the cells were incubated with leupeptin (LP; 10 M) or chloroquine (100 M) (Fig. 7B, see LP as representative). Since no accumulation of LPL aggregate was observed when either proteasomal or lysosomal inhibitors were used in the absence of cycloheximide (data not shown), this finding ruled out the possibility that the degradation of the aggregate required synthesis of a short lived protein. However, when cells were depleted of ATP by incubation in 2-deoxy-D-glucose, the rate of degradation was drastically reduced (Fig. 7B, 2DG). Thus, the LPL aggregate is turned over by an energy-dependent pathway that is other than proteasomal or lysosomal.
Since interchain disulfide bonds often occur within aggregates of misfolded proteins, their presence in the LPL aggregate was evaluated by comparing its migration in SDS-PAGE under reducing and nonreducing conditions (with and without ␤-mercaptoethanol). The LPL dimer, whose migration is unaffected by the presence or absence of ␤-mercaptoethanol was included as a control. Indeed, even in the absence of reducer, the dimer fraction migrated during SDS-PAGE to the expected FIG. 6. The fate of the dimer and aggregate remains unchanged even when confined to a common intracellular compartment. CHO cells preincubated with heparin were incubated with Cx (15 g/ml) and BFA (5 g/ml) for up to 3 h. A, LPL activity remains stable even when retained in the cells. LPL activity in the cells and medium was assayed as described in Fig. 5 and showed the expected abolishment of secretion by BFA. Nevertheless, retained intracellular activity remained relatively constant. B, the dimer remains stable while the aggregate is degraded. The lysates described for A were subjected to sucrose gradient centrifugation and LPL in the peak dimer, and aggregate fractions were detected (see legend to Fig. 5). Note the stability of the dimer and the degradation of the aggregate. C, glycan processing of dimer and aggregate in BFA-treated cells. Dimer and aggregated LPL from cells incubated for 90 min with Cx and BFA were subjected to endo H digestion. Since BFA causes resorption of Golgi components into the ER, both dimer and aggregated LPL contain complex glycan chains. 57-kDa region of the gel. However, under these nonreducing conditions, the bulk of the aggregate fraction was not detected in this region of the blot (Fig. 7C) and in fact was even unable to penetrate the stacking gel. Thus, it appeared that misfolded LPL in the ER associates into large aggregates that feature interchain disulfide bonds.
Disruption of Disulfide Bond Formation and Its Effect on the LPL Dimer and Aggregate-The presence of interchain disulfide bonds within the aggregate suggested that reduced conditions in the environment of the ER might cause dissociation of the aggregate into LPL monomers. To generate reducing conditions, we employed DTT, a membrane-permeable reductant that abolishes formation of disulfide bonds without affecting most other cellular functions, including secretion (31).
When CHO cells were incubated with Cx and 4 mM DTT, LPL activity declined drastically. Thus, after 1 h, 10% of the original activity was recovered in the cells, and 4% was recovered in the medium; after 2 h, only 3% activity remained in the cells, and no activity could be detected in the medium. Concurrent with the activity, the LPL dimer disappeared, but levels of the FIG. 8. Effect of disruption of disulfide bond formation on LPL dimer and aggregate. A, DTT abolishes LPL activity and induces aggregation of the LPL dimer. Heparin-treated cells were incubated with Cx (15 g/ml) and 4 mM DTT for up to 2 h (left panel) or were preincubated with BFA (5 g/ml) for 3 h to allow for complete degradation of the LPL aggregate prior to a similar treatment with DTT (right panel). After analyzing LPL activity, dimer and aggregate fractions obtained from sucrose gradient centrifugation were subjected to Western blotting. Note in the left panel the disappearance of the dimer with a concomitant increase of an LPL aggregate. The right panel clearly shows that the source of the DTT-induced aggregate is indeed the dimer, since no aggregate was present at the start of the DTT treatment. B, DTT-induced aggregation is reversible. Cells were pretreated for 2 h with Cx (15 g/ml) and DTT (4 mM) to allow for complete disappearance of the dimer and formation of the aggregate (0 h). The medium was then removed and replaced with fresh medium containing only Cx. Regeneration of LPL activity was followed in the cells and medium for an additional 2 h (histogram), and the fate of dimer and aggregate were assessed by Western blotting. Note the gradual disappearance of the aggregate and the formation of dimer that was eventually secreted, reflecting the pattern of regenerated functional LPL. C, only dimer present in the ER is retained in the cells as an aggregate. LPL dimer present in cells prior to incubation in Cx and DTT (0 h) and the dimer-generated aggregate induced by DTT treatment for 1 and 2 h were subjected to endo H digestion and Western blot analysis. LPL in medium of cells treated for 2 h in Cx and DTT was also analyzed. Note that the aggregate was formed exclusively from the high mannose dimer located in the ER, whereas LPL bearing Golgi-processed complex chains was secreted even in the presence of DTT.

FIG. 7. Characteristics of the LPL aggregate.
A, degradation of the aggregate is not impeded by proteasomal inhibitors. CHO cells transfected with the H2b subunit of the human asialoglycoprotein receptor or with human LPL were pretreated with heparin and incubated with or without the proteasomal inhibitor MG-132 (10 M) in the presence or absence of Cx (15 g/ml) for 6 h (H2b) or 3 h (LPL). The cells were harvested, and the LPL-containing lysates were subjected to density gradient centrifugation to isolate the aggregate. H2b lysate samples containing 5 g of protein or equivalent amounts of the LPL aggregate fractions were analyzed by Western blotting. B, degradation of the LPL aggregate is not impeded by lysosomal inhibitors, but is energy-dependent. LPL-transfected CHO cells were pretreated with heparin for 4 h. For the energy-depletion experiments, the cells were washed twice with glucose-free Dulbecco's modified Eagle's medium and the incubations carried out in the same medium. The cells were incubated either with Cx alone (15 g/ml) or with Cx in the presence of 10 M leupeptin (LP) or 50 mM 2-deoxyglucose (2DG). Following incubations, the cells were washed thoroughly with PBS, lysed, and subjected to sucrose gradient centrifugation. Aggregated LPL was detected by Western blot analysis. C, evidence of interchain disulfide bonds in the LPL aggregate. Samples of dimer and aggregate obtained from untreated cells as described for B were subjected to SDS-PAGE in the absence or presence of 0.25 M ␤-mercaptoethanol (␤-ME), followed by Western blot analysis. Although only the 55-kDa region of the gel is shown, little aggregate could be detected even in the stacking gel when ␤-mercaptoethanol was absent. aggregate increased (Fig. 8A, left panel). The simultaneous increase in the amount of aggregate with decreased dimer levels suggested that the aggregate was generated from the dimer, implying that the origins of the aggregate after 2 h in DTT were different than in untreated cells (i.e. at time 0). To ascertain that the source of the DTT-generated aggregate was indeed the dimer, cells were preincubated with Cx and BFA for 3 h, conditions that allow the original aggregate to degrade while retaining the LPL dimer in the ER (see Fig. 6B). At this point (t ϭ 0), DTT was added to the culture, and the incubation continued for an additional 2 h. As shown in Fig. 8A (right  panel), in the absence of de novo protein synthesis, DTT induced the gradual disappearance of the dimer with the simultaneous appearance of an LPL aggregate. This finding confirmed that the aggregate formed under perturbed redox conditions was generated from the dimer, in striking contrast to the de novo origins of the aggregate in untreated cells.
The aggregating effect of DTT on native LPL was reversed when DTT was removed from the culture. Thus, when cells incubated for 2 h with Cx and DTT were washed and the medium was replaced with Cx alone (t ϭ 0), there was a gradual decrease in the levels of aggregate, with a concomitant appearance of active dimer that was rapidly secreted (Fig. 8B, histogram). This finding indicated that DTT treatment did not cause irreversible misfolding of the dimer to create the aggregate; rather, DTT arrested it in a condition that permitted proper reassembly upon withdrawal of the reducer.
Since the LPL dimer was present as both ER and Golgi forms, the question arose as to the source of the dimer creating the DTT-induced aggregate in the cells. Susceptibility to endo H clearly demonstrated that the aggregate was not indiscriminately created from both forms. Instead, as shown in Fig. 8C, the aggregate was generated solely from LPL dimer residing in the ER (left panel, Cells), whereas dimer in the Golgi was secreted unhindered even in the presence of DTT (right panel, Medium). Thus, creating a reducing environment in the ER invoked conditions causing the reversible aggregation of LPL dimers, since subsequent restoration to its normal oxidative state permitted disaggregation and a return of LPL to its native conformation.
Association with Chaperones-The reversible nature of the DTT-generated LPL aggregate suggested that it was associated in the ER with resident folding chaperones. Since DTT treatment induced a prolonged association of a number of misfolded proteins with calnexin and BiP (32)(33)(34), we tested the association of the LPL aggregate with these chaperones. We also included in the analysis ERp57 and calreticulin, because ERp57 is directly involved in disulfide bond oxidation and isomerization and works in concert with both calnexin and calreticulin (35). Last, PDI was included, since it may promote "anti-chaperone" activity under reducing conditions by forming large, insoluble aggregates from misfolded proteins (36).
The aggregate fraction from untreated and DTT-treated cells was immunoprecipitated with antibodies to the various chaperones, and the presence of LPL in these precipitates was determined by immunoblotting (Fig. 9). Included on the blots was an aliquot of the aggregate fraction before immunoprecipitation (Fig. 9, control lane) to confirm that equivalent amounts of aggregate were sampled from DTT-treated versus untreated cells. As shown in the upper panel, only a minimal association of the LPL aggregate was detected among the chaperones tested. In stark contrast, a substantial amount of LPL aggregate from DTT-treated cells was associated with calnexin and BiP and, to a lesser extent, with PDI. Clearly, the nature of the two aggregates is different, as evidenced both by their different origins and by dissimilar associations with ER chaperones. The association of chaperones with the DTT-induced aggregate most likely explains its reversible nature upon removal of the reductant. This feature clearly distinguishes it from the aggregate originating from untreated cells, which appears to be trapped in an irreversible conformation destined for degradation.

DISCUSSION
The concept of Golgi activation of an inactive LPL precursor is historically rooted in experiments where drug-induced perturbation of intracellular transport and energy balance was invoked. Thus, when LPL-depleted Ob17 cells were treated with CCCP, inactive LPL accumulated in the ER. On the other hand, monensin, a blocker of medial-to trans-Golgi traffic, had no effect on enzyme activity. Hence, it was concluded that "activation" of LPL occurred after exit of the enzyme from the ER but before reaching the trans-Golgi compartment (37). At the time, it was not taken into account that CCCP (an uncoupler of oxidative phosphorylation) depletes ATP required for regulated binding and release from folding chaperones (38,39). Moreover, when cells treated with the glycosylation inhibitor tunicamycin displayed only inactive LPL in the ER, activation in the cis/medial-Golgi was specifically linked to the processes of addition and trimming of mannose residues, followed by the addition of N-acetylglucosamine residues (40,41). Again, at that time, the crucial importance of N-linked glycosylation for interaction with ER folding chaperones calnexin and calreticulin (7,8,42,43) was not considered. Even after establishing that formation of monoglucosylated LPL in the ER was an essential step for acquisition of activity (4,5,44), it was concluded by some that lack of glucose trimming simply prevented transport of an inactive precursor from the ER to the Golgi (6). Yet folding and assembly of proteins is greatly impaired in the absence of glucose trimming, as calnexin and calreticulin specifically associate with monoglucosylated intermediates of glycoproteins in the ER (7). Since even minor folding defects can lead to ER retention by virtue of quality control (45), the data in the studies above can all be interpreted as failure of LPL to acquire a functional conformation by virtue of its inability to properly interact with crucial ER folding factors. The end result would be retention in the ER of an improperly folded, inactive form of LPL.
Taking into account this reevaluation of the published data, we hypothesized that the ER is the exclusive site of LPL maturation and that acquisition of lipase activity does not hinge on an undefined Golgi-based modification, as has been proposed (6). Thus, a prime objective in the present study was to confirm FIG. 9. Differential association of the original LPL aggregate and DTT-generated aggregate with ER chaperones. Similar amounts of LPL aggregate from untreated cells and from cells treated with DTT for 2 h were immunoprecipitated with antibodies against an array of ER chaperones. Normal rabbit IgG (rabb. IgG) was used as a negative control. Co-immunoprecipitated LPL was detected by Western blot analysis using an epitope-specific antibody (see "Experimental Procedures"). Proportional aliquots of samples used for these experiments were included on the blot ("control") to ensure that similar amounts of aggregate from untreated and DTT-treated cells were compared. that LPL acquired full catalytic activity in the ER. The approach of drug-or stress-induced conditions was avoided, and we attempted instead to isolate an active form from the ER of normal, untreated cells. This was accomplished by a combination of heparin-Sepharose and ␤-ricin chromatography (Fig. 2). LPL bearing both ER-and Golgi-type glycans was enzymatically active, and, most important, both forms had similar specific activity (milliunits/LPL mass). Thus, we were able to demonstrate for the first time in a direct manner that LPL becomes fully functional in the ER.
While we found that LPL maturation occurs in the ER, we also found that this compartment contained a substantial amount of inactive LPL with low affinity to heparin (Fig. 2). We questioned whether this form could be the proposed monomeric precursor to active LPL (41,46), since it has become common belief that LPL with low affinity to heparin is monomeric (19 -23). However, we found that low affinity LPL, rather than being a monomer, migrated on sucrose gradients as a high molecular weight aggregate. We further showed that the LPL aggregate was confined to the ER, that it was neither an artifact of centrifugation nor overexpression in CHO cells, and that it was also present in 3T3-L1 cells that endogenously express LPL (Fig. 3).
Aggregation is not an unusual feature of proteins in the ER. Rather, there is increasing evidence that aggregation is a common by-product of protein folding in the ER, resulting from poor solubility of folding intermediates (28,47). But a tendency of newly synthesized proteins to aggregate has also been observed during the normal folding pathway, where transient aggregation proceeds to dissociation and productive folding into a native configuration (26, 48 -50). Because of these precedents and because it has been historically postulated that LPL maturation occurs through an inactive intermediate, we entertained the possibility that the LPL aggregate represented a precursor of the mature enzyme. However, several experiments in the present study show that this is not the case. First, no kinetic evidence was observed to link the LPL aggregate to formation of dimeric enzyme in the ER (Fig. 4). Second, after inhibiting de novo protein synthesis, no gain in activity was detected upon disappearance of the aggregate; instead, active LPL was secreted from the cells concurrent with degradation of the aggregate (Fig. 5). This observation was corroborated when the experiment was conducted in the presence of BFA, where secretion was abolished. In this case, as the aggregate disappeared, neither the amount of enzymatic activity nor dimeric LPL mass changed significantly over a period of 3 h (Fig. 6). Third, DTT treatment clearly displayed the nature of an inactive LPL complex possessing the ability to generate active LPL dimer (Fig. 8B), contrasting sharply with the characteristics of the original aggregate that was eventually degraded. Thus, the LPL aggregate is not a precursor form but presumably originates from nascent molecules that have reached a "dead end" conformation, off the normal folding pathway (28). As opposed to the correctly folded and properly assembled active LPL, which was stable regardless of whether it was secreted or retained in the cell, the aggregate was retained in the ER and eventually degraded. The fact that the aggregate did not accumulate, despite its continued production, must be due to quality control mechanisms that recognize and target improperly folded polypeptides to degradation (28,45,51). The disposal of aberrant proteins, known as ER-associated protein degradation (ERAD) (52), efficiently removes these forms from the secretory pathway.
The precise ERAD pathway of the LPL aggregate remains to be determined. Unlike a majority of misfolded and/or unassembled proteins, such as H2b, that are degraded via ubiquitin-proteasome proteolysis (53,54), the disappearance of the LPL aggregate was not affected by the presence of proteasome inhibitors (Fig. 7A). Thus, the LPL aggregate belongs to a class of misfolded proteins that undergo ER-associated degradation by mechanisms distinct from the proteasomal pathway. Other examples of proteins in this class include the human thyroperoxidase (55) and misfolded forms of ␣ 1 -antitrypsin (56,57).
Lysosomal degradation of the LPL aggregate was also examined, although misfolded aggregates in the ER are usually retained by quality control mechanisms and thus become transport-incompetent (45). Nevertheless, the lysosomal pathway was considered, since a mutant form of human LPL lacking activity (G142E) was reported to be degraded in this manner (58). However, in contrast to this mutant form of LPL, lysosomal degradation of the aggregate was ruled out (Fig. 7B).
The depletion of ATP greatly reduced the rate of degradation of the LPL aggregate (Fig. 7B), implying that the process was energy-dependent. Several possibilities linking energy depletion with inefficient ERAD of LPL may be considered. For instance, some ER chaperones possessing ATPase activity, notably BiP and PDI (59,60), have been shown to associate with abnormal or excess proteins prior to their degradation in a pre-Golgi compartment (61,62). However, we were unable to detect a robust association of BiP and PDI with the LPL aggregate at steady state (Fig. 9). A second possibility is that energy depletion affects retrotranslocation of the LPL aggregate from the ER to the cytoplasm, since this is also an ATPdependent process (63). Although the degradation site of the LPL aggregate is not known, retrotranslocation to the ER remains a valid possibility, since the cytosol contains proteases such as the multicatalytic modular system with the evolutionary conserved tricorn protease at the core (64, 65) and tripeptidyl peptidase II (66).
While we have not detected in our studies aglycosylated or ubiquitinylated LPL 4 (features indicating retrotranslocation and targeting toward cytosolic degradation), such intermediates are often difficult to detect even when examining proteins that are known to undergo proteasomal degradation (30). Presumably, this is because such intermediates are rapidly turned over and are present in the cytoplasm in very small amounts. Thus, it is still a possibility that the LPL aggregate might be targeted to the cytoplasm for degradation.
The presence of interchain disulfide bonds in the LPL aggregate (Fig. 7C) suggested that the LPL aggregate resulted from poorly soluble nascent polypeptides or partially folded intermediates coming into close contact (67). However, the LPL complex induced by DTT treatment was very different from the original aggregate, both in its origin and its fate (Fig. 8). The DTT-induced complex was generated from dimeric LPL and was readily dissociated back into functional dimers when DTT was removed. In contrast, the original aggregate, which was not a precursor of the dimer, was recognized by the ER as aberrant and consequently targeted for degradation. We assume that this degradation occurred even in the presence of DTT, since the latter blocks disulfide bond formation without adverse effects on most cellular functions, including ATP synthesis (68). Thus, the "aggregate" present after 2 h in Cx plus DTT (or in Cx plus DTT and BFA) was presumably all generated from the preexisting dimer.
We investigated the association of LPL in the DTT-generated complex with chaperones, since DTT often evokes interactions with ER folding factors (69). Moreover, it was clear that formation of the complex occurred exclusively in the ER, since LPL located in the Golgi was readily secreted even in the presence of DTT (Fig. 8C). We found that this inactive complex was largely associated with BiP and calnexin and, to a lesser degree, with PDI. This was in contrast to the original LPL aggregate, where only weak associations were found (Fig. 9). The robust association in DTT-treated cells most likely resulted from the unfolded protein response that became activated by the accumulation of reduced proteins (including LPL) in response to the altered redox potential in the ER. The unfolded protein response elicits transcriptional up-regulation of many genes encoding folding factors in the ER and the cytoplasm (52, 70 -72).
The combination BiP/calnexin has been implicated in the folding of a number of proteins, both in the normal folding pathway (73) and in the unfolded protein response, where interactions are often enhanced (32,33). Interestingly, DTT promotes association of proteins with calnexin regardless of their glycosylated state (32,33), although normally glucose trimming to form monoglucosylated proteins is essential before an interaction with calnexin can occur. This could explain the DTT-induced calnexin association with LPL, although the latter had already acquired a functional conformation and was presumably no longer glucosylated. Moreover, a DTT-induced complex between LPL and BiP/calnexin presumably occurred even in BFA-treated cells (Fig. 8A), when LPL had already acquired a complex glycan structure (see Fig. 6C).
PDI was another chaperone detected at low levels in the original LPL aggregate but showing increased associations with LPL in the DTT-generated complex (Fig. 9). Like BiP, PDI is an important component of the quality control machinery in the ER that assists proteins both in proper folding (74) and in degradation (75). However, at low concentrations, PDI may express anti-chaperone activity, promoting protein aggregation and degradation (76). Indeed, it has been found that PDI and BiP are specifically incorporated into large aggregates with unfolded lysozyme, with a stoichiometry suggesting one antichaperone molecule with multiple molecules of unfolded proteins (77). A similar stoichiometry in our case could explain the low levels of PDI and BiP detected in the original LPL aggregate. In contrast, the presence of DTT resulted in the accumulation of an LPL complex that contained considerably higher levels of PDI (Fig. 9). It has been shown that binding of PDI to unfolded proteins is independent of lectin chaperones (35); calnexin may even impede PDI-assisted refolding of proteins as has been shown for RNase B (78). Thus, it may be that the complex in DTT-treated cells is a mixture of PDI-associated or calnexin/BiP-associated LPL molecules. Future experiments will be required to address this issue.
In conclusion, our findings suggest LPL aggregation as a normal yet dead end pathway that can occur when the nascent enzyme attempts to undergo folding in the ER. Perhaps aggregation provides a simple, efficient means to segregate misfolded LPL from the folded, transport-competent dimer solely on the basis of different solubility. Whether this process plays a regulatory function in the availability of LPL for lipid metabolism remains to be determined.