Brefeldin A promotes the appearance of oligosaccharyl phosphates derived from Glc3Man9GlcNAc2-PP-dolichol within the endomembrane system of HepG2 cells

We reported an oligosaccharide diphosphodolichol (DLO) diphosphatase (DLODP) that generates dolichyl-phosphate and oligosaccharyl phosphates (OSPs) from DLO in vitro. This enzyme could underlie cytoplasmic OSP generation and promote dolichyl-phosphate recycling from truncated endoplasmic reticulum (ER)-generated DLO intermediates. However, during subcellular fractionation, DLODP distribution is closer to that of a Golgi apparatus (GA) marker than those of ER markers. Here, we examined the effect of brefeldin A (BFA), which fuses the GA with the ER on OSP metabolism. In order to increase the steady state level of truncated DLO while allowing formation of mature DLO (Glc3Man9GlcNAc2-PP-dolichol), dolichyl-P-mannose Man7GlcNAc2-PP-dolichol mannosyltransferase was partially downregulated in HepG2 cells. We show that BFA provokes GA endomannosidase trimming of Glc3Man9GlcNAc2-PP-dolichol to yield a Man8GlcNAc2-PP-dolichol structure that does not give rise to cytoplasmic Man8GlcNAc2-P. BFA also strikingly increased OSP derived from mature DLO within the endomembrane system without affecting levels of Man7GlcNAc2-PP-dolichol or cytoplasmic Man7GlcNAc2-P. The BFA-provoked increase in endomembrane-situated OSP is sensitive to nocodazole, and BFA causes partial redistribution of DLODP activity from GA- to ER-containing regions of density gradients. These findings are consistent with BFA-provoked microtubule-dependent GA-to-ER transport of a previously reported DLODP that acts to generate a novel endomembrane-situated OSP population.

Presently, the machinery underlying OSP production has not been characterized at the molecular level, and how an OSP generation mechanism might distinguish between normal biosynthetic DLO intermediates, on the one hand, and, on the other, those that are surplus to requirements or possess aberrant structures is not understood. Recently, a Co 2+ -activated liver microsomal DLO diphosphatase (DLODP) activity that splits both mature and truncated DLO to yield OSP and DolP has been reported (18). This reaction might enable DolP recycling from unproductive or potentially dangerous DLO intermediates that occur in CDG. However, the potential physiological role of this enzyme in DLO regulation remains uncertain because, during density gradient centrifugation, this activity distributes similarly to a marker of the Golgi apparatus (GA), and differently to those of ER-situated DolP-dependent DLO biosynthetic enzymes (18). The complexity of this data prompted us to evaluate the subcellular compartmentalization and membrane topology of OSP production in human hepatocellular HepG2 carcinoma cells treated with drugs that perturb the endomembrane system. Here we show that brefeldin A (BFA), which causes fusion of the GA with the ER, provokes both the appearance of a truncated DLO intermediate that does not give rise to a cytoplasmic OSP, and complex GA-modified OSP within the endomembrane system. By contrast, in ALG12-deficient HepG2 cells, BFA did not affect the levels of either Man 7 GlcNAc 2 -PPdolichol or cytoplasmic Man 7 GlcNAc 2 -P that accumulated in this cell model of ALG12-CDG. These data HEPES/NaOH (pH 7.4) containing 250 mM sucrose and mammalian protease inhibitors]. To estimate the cell volumes, the scraped cells were centrifuged at 200 g at 4°C for 5 min, and then the total volume was adjusted to give 4 vol of homogenization buffer for each volume of cells. Cells were homogenized by 20 strokes in a loose fitting Dounce homogenizer. After centrifugation at 750 g for 5 min, the supernatant was collected and the pellet was again homogenized and centrifuged, as above, to give a pellet (P1). The two supernatants were combined and centrifuged at 6,000 g for 10 min and 100,000 g for 45 min to give P2 and P3, respectively, and a final supernatant (S). OptiPrep solutions were diluted into 10 mM HEPES/NaOH (pH 7.4) containing 250 mM sucrose. P3 was made 20% with respect to OptiPrep and layered onto a 30% OptiPrep cushion before overlaying with 15% OptiPrep (24). The tubes were centrifuged at 350,000 g for 90 min using a Beckman VTi 65.2 rotor.

Analytical procedures
Charged oligosaccharides were desalted on Biogel P2 columns prior to fractionation on quaternary aminoethyl (QAE)-Sephadex columns (25), as previously described (15). Neutral oligosaccharides were separated on silica-coated plastic sheets (0.2 mm thickness) developed in n-propanol/acetic acid/water (3:3:2) for 16-24 h (26). Radioactive components were detected on X-OMAT AR film by fluorography after spraying the dried TLC plates with En 3 hance and were quantitated by scintillation counting after their elution with water from the silica. Where indicated, fluorographs were scanned and images quantified using the BIO-1D software (Vilber Lourmat, Germany). Where appropriate, data were analyzed and plotted using GraphPad Prism version 5.00 for Windows, (GraphPad Software, San Diego, CA; http://www.graphpad.com). Oligosaccharides were derivatized with 2-aminopyridine (AP), separated by normal-phase and reversed-phase HPLC, and detected by on-line fluorometry, as previously described (27). OSPs were analyzed by concanavalin A (Con A)-Sepaharose chromatography exactly as previously described (28). The characterization of oligosaccharide structures using alkaline phosphatase, jack bean -mannosidase, jack bean -hexosaminidase, and endo--N-acetylglucosaminidase was carried out according to the protocols supplied by Sigma-Aldrich. Protein was detected using the BCA reagent (29). UDP-galactose glycoprotein galactosyltransferase (UGT) was assayed as previously described (30). For the DLODP assay: Glc 3-0 -[ 3 H]Man 5 GlcNAc 2 -PP-dolichol (16 × 10 3 cpm) was dried down into 1.5 ml centrifuge tubes before being resuspended in 5 l 1% NP-40. Then further components were added to give a final reaction mixture of 100 mM MES (pH 5.5), 1 mM CoCl 2 , 0.1% NP-40, 100 M SW, 100 M KIF, 1 mM CST, and 4.2-8.4 g protein in a volume of 50 l. Incubations were carried out at 37°C for 20 min and processed as described in (18).

Western-blot analysis
Samples prepared from density gradient fractions were diluted to the same protein concentration, heated with NuPAGE LDS sample buffer under reducing conditions according to the manufacturer's instructions, and subjected to SDS-PAGE; the primary antibodies were detected using horseradish peroxidasecoupled secondary antibodies.

RESULTS
Whereas some studies support the idea that OSPs are hydrolyzed from truncated DLO intermediates on the cytoplasmic face of the ER, others suggest that OSPs can be

Metabolic radiolabeling of cells
Cells were incubated in glucose-free RPMI 1640 medium containing 0.5 mM glucose, 2.0 mM fucose, and 2% dialyzed fetal calf serum and different reagents (detailed in each figure legend) for 1 h. Then 100 Ci/ml [2][3]

Permeabilization of HepG2 cells
Specific permeabilization of plasma membranes with SLO was performed as described previously (21): cells were incubated at 4°C with 0.5 ml precooled permeabilization buffer [5 mM HEPES (pH 7.0) containing 250 mM mannitol and 2 mM CaCl 2 ] containing 2 g/ml SLO. After 1 h, the SLO-containing medium was removed from the cells and combined with a subsequent 1 ml permeabilization buffer wash.

Recovery of DLO, OSP, neutral fOS, and glycoproteins from cells and medium
These methods are based on previously described procedures (6,22). Washed cells were suspended in 4 ml of methanol/100 mM Tris-HCl (pH 7.4) containing 4 mM MgCl 2 , 2:1. Four milliliters of CHCl 3 were added and the mixture shaken. After centrifugation, three phases were obtained: a water/methanol upper phase, a lower CHCl 3 phase, and an interphase containing precipitated protein and certain DLO. DLO was recovered from the lower CHCl 3 phase and from the CHCl 3 /methanol/water 10:10:3 extracts of the interphase material. The two organic phases were dried under vacuum and oligosaccharides were released from DLO after acid hydrolysis with 0.02 N HCl for 45 min at 100°C. For all experiments, except those reported in Figs. 2A, 7C, oligosaccharides derived from the two organic phases were pooled prior to further analysis. For the recovery of cellular OSP and neutral (n)fOS, the water/methanol upper phase was dried under vacuum and loaded onto 2 ml Dowex 50WX2 resin (H + form) and 2 ml Dowex 1X2 resin (acetate form) columns, the water wash containing nfOS was loaded onto charcoal columns and, after extensive washing with water, oligosaccharides were eluted with 30% ethanol. Cellular OSPs were eluted from the Dowex 1X2 columns using 20 ml 3 M formic acid. After drying, this material was subjected to acid hydrolysis with 0.02 N HCl for 45 min at 100°C then desalted using combined Dowex 50WX2 resin (H + form) and Dowex 1X2 resin (acetate form) columns. To recover secreted components, the labeling medium was collected and 2 vol of ice-cold acetonitrile were added. After vigorous shaking, the mixture was kept at 20°C for at least 24 h. After thawing and centrifugation at 3,200 g for 15 min at 4°C, the supernatant was collected and the protein precipitate was washed once with 3 ml ice-cold 66.7% acetonitrile. The two supernatants were pooled and dried under vacuum and processed for OSP and nfOS, as for the water/methanol upper phase material. Precipitated proteins recovered from the 10/10/3 extracts of the cellular interphase material and the acetonitrile-treated radiolabeling medium were dried and then solubilized with pronase.

Density gradient fractionation of HepG2 cell homogenates
HepG2 cell monolayers (80% confluent) were homogenized and fractionated according to a protocol devised for rat liver (23) with some modifications. Cells were washed twice with ice-cold PBS and scraped into ice-cold homogenization buffer [10 mM derived from mature DLO within the ER. Recently we reported on a Co 2+ -activated DLODP activity that liberates OSPs from both truncated and mature DLO, but that does not codistribute with DolP-dependent enzymes of the dolichol cycle during density gradient centrifugation of microsomal membranes (18). In fact, the bulk of this DLODP activity distributed similarly to GA-situated galactosyltransferase (UGT) (18). These results prompted us to examine the subcellular sites of OSP production in living cells using drugs that perturb the endomembrane system (Fig. 1A). The modes of action of several commonly used drugs to dissect out ER and GA function are shown in Fig. 1A. Both NZ and IMQ cause fragmentation and dispersal of the GA throughout the cytoplasm without affecting GA-dependent protein secretion (31,32). Whereas it is known that NZ primarily disrupts microtubules required for GA stabilization (33), the molecular target for IMQ has not yet been determined, but its effects on GA structure seem to require phospholipase D (34). BFA promotes fusion of elements of the GA with the ER to form a composite structure (35,36). Within this structure, glycoprotein processing normally associated with the cis, medial, and, to a lesser extent, trans GA can continue (37,38), but glycoprotein exit does not occur (39). This ER/GA composite structure ( Fig. 1A) has been termed the BFA compartment (38). Recently GCA has been shown to have similar effects on cultured cells to those of BFA (40). Thapsigargin inhibits the sarco/ER Ca 2+ -ATPase that is involved in maintenance of the ER Ca 2+ pool (41) and is known to inhibit ER-associated glycoprotein processing and also glycoprotein secretion (42). In order to examine the effect of perturbing the endomembrane system on DLO and OSP generation, HepG2 cells were preincubated with either carrier or the different drugs for 1 h prior to metabolic radiolabeling with [2-3 H]mannose for 2 h in the continued presence of drugs. TLC examination of oligosaccharides derived from DLO and OSP generated under these conditions is shown in Fig. 1B. In untreated cells, DLO occurs mainly as Glc-3 Man 9 GlcNAc 2 -PP-dolichol. Small amounts of OSP whose glycan structures comigrate with Man 7-5 GlcNAc 2 and Glc 3 Man 9 GlcNAc 2 are detected. The smaller OSP structures are observed in different cell types grown under normal conditions and in increased amounts in cells with blocks in the dolichol cycle where truncated Man 7-5 GlcNAc 2 -PPdolichol intermediates accumulate (10,14,15). Much less is known about the origins of OSPs that are derived from larger DLO (16,17). As shown in Fig. 1B, IMQ and NZ did not have a striking effect on the thin-layer chromatographic profiles of the oligosaccharides derived from either the DLO or OSP. By contrast, both BFA and GCA provoke the appearance of a truncated DLO, whose oligosaccharide moiety (Fig. 1B, a) comigrates with Man 8 GlcNAc 2 , and a family of OSP structures (Fig. 1B, b-f), whose glycan moieties do not comigrate with standard polymannose-type structures. TG caused an increase in radioactivity associated with DLO along with small changes in the profiles of oligosaccharides derived from both DLO and OSP, but did not provoke the striking changes seen with either BFA or GCA. The effects of the latter agents Subsequent to transport to the medial GA, N-glycans are modified by N-acetylglucosaminyltransferase I (GNT1), then MAN2A1, and finally by other N-acetylglucosaminyltransferases (such as GNT2). N-glycosylproteins are then transported to the trans GA where Nglycans are galactosylated by UGT. Finally, after vesicular transport to the trans Golgi network (TGN), N-glycans are sialylated by sialyltransferases (ST) before being packaged into vesicles for transport to their final destinations. BFA and GCA cause the cis, medial, and trans GA to fuse with the ER to form the BFA or GCA compartment (red dashed line). NZ and IMQ cause the GA to fragment and disperse throughout the cytoplasm. Thapsigargin (TG) blocks Ca 2+ transport into the ER and perturbs many aspects of ER function as well as protein secretion. The ER glucosidase I and II inhibitor, CST, the MAN1B1 and MAN1A1 inhibitor, KIF, and the MAN2A1 inhibitor SW are used to understand N-glycan processing reactions in the endomembrane system. B: HepG2 cells were treated with the carrier, DMSO, 20 M IMQ, 10 g/ml NZ, 4 g/ml BFA, 10 M GCA, or 100 nM TG as indicated, before metabolic radiolabeling with were chosen for further study because, although a truncated DLO became apparent, similar to the situation seen in cells from certain patients with CDG, the corresponding OSP was not observed, and instead, a population of novel OSP structures (Fig. 1B, b-f) was detected.

Golgi endomannosidase processing of DLO in BFA-treated cells
In order to further investigate the glycan structure of the BFA-induced DLO (Fig. 1B, a), HepG2 cells were treated with BFA for 3 h prior to harvesting and extracting DLO with organic solvents. Oligosaccharides were released from DLO, derivatized with the fluorophore, 2-AP, and resolved by HPLC as shown in Fig. 2A. As expected from data shown in Fig. 1B, the HPLC profiles reveal that, in the presence of BFA, the predominant DLO possesses an oligosaccharide moiety (labeled a) with an elution time corresponding to that of Man 8 GlcNAc 2 -AP. Material eluting under peak a in Fig. 2A could correspond to Man 8 GlcNAc 2 -AP or a glucosylated structure containing the same number of hexoses, such as, for example, Glc 3 Man 5 GlcNAc 2 -AP or Glc 1 Man 7 GlcNAc 2 -AP. To resolve this issue, peak a ( Fig.  2A) material was recovered and digested with jack bean -mannosidase in order to determine whether any of its -linked mannose residues were shielded by glucose residues. Greater than 95% of the starting material was converted into Man1,4GlcNAc 2 -AP, demonstrating that the predominant species eluting under peak a ( Fig. 2A) is not glucosylated and therefore contains eight residues of mannose (A. migrating as standard Man 9-8 GlcNAc 2 are observed, despite the occurrence of glucosylated DLO in this condition (compare right panel of Fig. 3A with Fig. 2C). In the presence of CST, KIF, and SW, the OSP glycan structures are similar to those of DLO observed under the same conditions (compare right panel of Fig. 3A with Fig. 2C), comprising predominantly Glc 3 Man 9 GlcNAc 2 and Man 8 GlcNAc 2 species. The changes in distributions of the novel OSP species that are noted to occur with the different glycosidase inhibitors suggest that BFA-provoked OSPs originate from DLO in the lumen of the BFA compartment and then undergo processing by SW- and KIF-sensitive enzymes that are normally present in the GA. The unexpected appearance of this novel population of OSPs in BFA-treated cells prompted us to confirm that these structures do, in fact, possess a phosphate moiety linked at the C1 position of their reducing N-acetylglucosamine (GlcNAc) residues. To do this, cells were treated with the three glycosidase inhibitors and BFA, metabolically radiolabeled, and, after solvent extraction, total MeOH/H 2 O upper phases were dried down and desalted on a Biogel P2 column. Ion-exchange fractionation of the desalted material on QAE-Sephadex (Fig. 3B, left panel) demonstrates that although the bulk of radioactivity is associated with previously described nfOS and the tetrasaccharide (Glc 3 Man) produced by endomannosidase ( Fig. 2A,  MANEA1), a peak of radioactivity is eluted with 75 mM NaCl. TLC examination of this purified charged material after further desalting revealed a mixture of slow migrating components comprising structures that give rise to predominantly Glc 3 Man 9 GlcNAc 2 and Man 8 GlcNAc 2 , and Glc 3 Man 9 GlcNAc and Man 8 GlcNAc , after alkaline phosphatase and endoH treatments, respectively (Fig. 3B, middle panel). These data confirm that the BFA-provoked structures contain a phosphate residue linked to the reducing GlcNAc moiety of the oligosaccharide (Fig. 3B, right panel).

Identification of -hexosaminidase-sensitive complex-type OSPs in BFA-treated cells
Next the glycan structure of the OSPs recovered from BFA-treated cells was examined using Con A-Sepharose chromatography. Data shown in Fig. 3C demonstrate that GA-modified complex-and hybrid-type structures constitute 80-90% of total OSP when cells are metabolically radiolabeled in the absence of glycosidase inhibitors. Upon addition of SW, the distribution of OSP changes, with hybrid-type structures becoming predominant (Fig. 3C). As expected, addition of KIF resulted in the predominance of polymannose-type OSPs. It was noted that the quantity of material associated with the 10 mM -methyl glucose eluates from the Con A-Sepharose columns does not change strikingly when cells are treated with the glycosidase inhibitors. Because it is known that certain truncated polymannose-type oligosaccharides, such as those occurring in truncated DLO intermediates (Man 7-5 GlcNAc 2 -PP-dolichol), elute with complex bi-antennary oligosaccharides under these conditions (47,48), it is likely that the small Man 7-5 GlcNAc 2 -P species detected in cells radiolabeled in experiments were conducted in the presence of the mannosidase and glucosidase inhibitors described in Fig. 1A. Neither control nor BFA-induced DLO profiles were affected by the presence of the MAN2A1 (Fig. 1A) inhibitor SW (Fig.  2C). By contrast, whereas addition of the ER mannosidase I (MAN1B1; Fig. 1A) and MAN1A1 (Fig. 1A) inhibitor, KIF, did not affect the DLO profile seen in the absence of BFA, it blocked the appearance of the BFA-induced species containing seven hexose units without affecting the quantity of component a (Fig. 2C). Finally, the addition of the glucosidase I and II inhibitor, CST (Fig. 1A), as well as both KIF and SW, to the cells caused an increase in the proportion of triglucosylated DLO in both control and BFAtreated cells, but did not completely block the appearance of the BFA-provoked DLO bearing eight hexose residues (Fig. 2C). These data suggest that MANEA1, which is insensitive to CST, KIF, and SW (43) and is known to hydrolyze glucosylated DLO in vitro (44), might underlie the appearance of Man 8 GlcNAc 2 -PP-dolichol in BFA-treated cells. Next, the Man 8 GlcNAc 2 -AP structure ( Fig. 2A, HPLC peak a) was digested with S. plicatus endo--N-acetylglucosaminidase H (endoH) to yield Man 8 GlcNAc, which was then rederivatized with AP and subjected to reversed-phase HPLC with a standard mixture of Man 8 GlcNAc-AP isomers, as shown in Fig. 2D. The chromatographs demonstrate that the (d2,d3)Man 8 GlcNAc 2 isomer that is the hallmark of endomannosidase action is the major oligosaccharide structure associated with Man 8 GlcNAc 2 -PP-dolichol derived from BFA-treated cells. Finally, Man 8 GlcNAc 2 -PP-dolichol is not observed in KIF/CST/BFA-treated Chinese hamster ovary (CHO) cells (Fig. 2E) that are known to be sensitive to BFA (45), but express an inactive MANEA1 (46). The ensemble of these data demonstrates that glucosylated DLO species are trimmed by MANEA1 in BFA-treated HepG2 cells.

BFA induces the appearance of a novel family of OSPs in HepG2 cells
In order to understand why the appearance of Man-8 GlcNAc 2 -PP-dolichol is not accompanied by the appearance of Man 8 GlcNAc 2 -P in BFA-treated cells, OSPs isolated from the glycosidase-treated cells described in Fig. 2C were examined by TLC. In the absence of BFA, as shown in Fig. 3A (left panel), irrespective of the presence of glycosidase inhibitors, Glc 3 Man 9 GlcNAc 2 is detected along with the smaller components migrating as standard Man 7-5 GlcNAc 2 s that were described in Fig. 1B. When SW and KIF are added to the cell cultures, the OSP profiles are modified to a small extent. Upon addition of CST to the inhibitor mix, a population of larger structures migrating between Glc 3 Man 9 GlcNAc 2 and Man 9 GlcNAc 2 is noted in addition to the smaller structures. In the absence of glycosidase inhibitors, BFA provokes the appearance of a complex mixture of OSPs (  the absence of BFA (Fig. 1B and left panel of Fig. 3A) also occur in BFA-treated cells and that these components contribute significantly to the radioactivity associated with the 10 mM -methyl glucose eluates. Next, in order to confirm the presence of complex and hybrid type OSPs in BFA and BFA-and SW-treated cells, respectively, OSP obtained under these two conditions were treated with alkaline phosphatase and then subjected to either jack bean -mannosidase or -hexosaminidase, as shown in Fig. 3D, E. OSPs generated by cells in the absence of SW largely resisted the mannosidase and were more susceptible to hexosaminidase, yielding predominantly a digest product behaving as Man 3 GlcNAc 2 . These data are consistent with the presence of OSP containing the trimannosyl core substituted with one or more residues of GlcNAc (Fig. 3D).
OSPs recovered from the SW-treated cells are moderately sensitive to -mannosidase and yield an oligosaccharide that comigrates with Man 5 GlcNAc 2 after digestion with hexosaminidase. These data are in accordance with the presence of hybrid-type OSP in which the Man 5 GlcNAc 2 core structure is substituted with one or more residues of GlcNAc (Fig. 3E). Previous studies with this type of structure have shown that the exposed mannose residues are somewhat resistant to jack bean -mannosidase cleavage (49). Therefore BFA provokes the appearance of GA-modified complex OSP and suggests that Man 8 GlcNAc 2 -PPdolichol could give rise to Man 8 GlcNAc 2 -P that is then further processed by GA enzymes. Fig. 4A), the DLO pool is small (about 2% of total radioactivity incorporated into cells) and turns over rapidly, and under the radiolabeling conditions employed, is probably close to steady state. During this radiolabeling period, the DLO pool gives rise to radioactive N-glycosylproteins (60-70% of all radioactive components recovered from the cells and medium), nfOS (20-40% of all radioactive components recovered from the cells and medium), and recovered from the cells and medium. The effects of BFA on the DLO and OSP pools are not accompanied by striking effects on the amounts of radioactive N-glycosylproteins and nfOS generated during the radiolabeling period. These data suggest that if OSP catabolism is occurring, it does not contribute significantly to nfOS production.

BFA-provoked cellular OSPs do not arise due to a block in their secretion
Why does BFA provoke the appearance of OSP from mature DLO in the lumen of the endomembrane system? There are several explanations. First, OSP could be generated constitutively in the ER and then be secreted into the extracellular medium after passage through the GA. Under these circumstances, the BFA-provoked appearance of OSP would be due to a block in OSP secretion. To test this hypothesis, OSPs were recovered from both the cells and medium of KIF + SW-treated cells radiolabeled in either the presence or absence of BFA. As shown in Fig. 4E, the amount of radioactivity recovered as OSP in the medium of cells radiolabeled in the absence of BFA cannot account for the BFA-provoked increase in cellular OSP in BFAtreated cells. Because OSPs could be dephosphorylated in the medium to yield nfOS, these neutral structures were also quantitated in the medium; but again, the difference in the quantities of nfOS found in the medium of cells radiolabeled in either the absence or presence of BFA is not sufficient to explain the BFA-provoked increase in intracellular OSPs (Fig. 4B).

BFA-provoked OSP accumulation is partially blocked by NZ
Next, we wanted to determine whether the BFA-promoted appearance of OSP is the result of, first, BFA-mediated inhibition of a normal degradative process that clears potential constitutively generated OSP from the lumen of the ER, or of, second, either the BFA-provoked creation of a nonphysiological OSP generation process or a BFA-promoted increase in a normally occurring phenomenon. For example, BFA-provoked OSP might be caused by mixing of a GA-situated DLODP activity with ER-situated DLO in a manner similar to that known to occur for the BFA-provoked increase in sphingomyelin biosynthesis, where BFA is thought to cause mixing of GA-situated sphingomyelin synthase with ER-situated ceramide (31). Alternatively, a block in ER export from the BFA compartment could trap a putative secretory OSP-generating enzyme in the ER, where it could hydrolyze DLO. In order to answer these questions, we sought to determine whether BFA-provoked OSP generation was due to a simple block of vesicular transport out of the ER, or whether retrograde movement of elements of the GA into the ER is required. To do this, we examined the effect of NZ on the ability of BFA to provoke the appearance of OSP. As mentioned earlier, the organization of the GA requires microtubules and, as shown in Fig. 5A, NZ disrupts GA morphology, leading to the production of functional GA-derived vesicles that distribute throughout the cell cytoplasm. Although BFA-provoked redistribution of the GA to the ER requires intact OSP (0.2-0.4% of all radioactive components recovered from the cells and medium). Misfolded glycoproteins also give rise to nfOS through ER-associated degradation processes (50). The changes in nfOS levels seen with the different glycosidase inhibitors are expected because, as previously reported (21), inhibition of the cytoplasmic and lysosomal mannosidases by SW leads to the accumulation of nfOSs that are produced during glycoprotein biosynthesis. In addition, the Golgi/MAN1B1 inhibitor, KIF (51), and the ER glucosidase inhibitor, CST (52), are known to be able to perturb normal ER-associated degradation processes and are expected to alter the levels of nfOS that are generated by these mechanisms. In BFAtreated cells (hatched bars in Fig. 4A), the DLO pool sizes increase by 1.3-to 2-fold over control values, whereas there are 5- to 10-fold increases in OSP. Although the BFA-provoked increase in OSP is striking, the radioactivity associated with OSP under these conditions does not exceed 2% of the total radioactivity incorporated into components microtubules, the ability of BFA to block post ER vesicular transport is not microtubule dependent. Accordingly, it has been demonstrated that pretreatment of cells with NZ reduces BFA-provoked fusion of the GA with the ER (33), while not affecting the ability of BFA to block transport out of the ER or pre Golgi compartments. As expected, therefore, under our experimental conditions, pretreatment of cells with NZ for up to 4 h before the radiolabeling period does not inhibit glycoprotein secretion (compare NZ hatched bars with DMSO hatched bars in Fig. 5B). In addition, NZ pretreatment has only a minor effect on the capacity of BFA to block glycoprotein secretion (compare DMSO solid bars with NZ solid bars in Fig. 5B). Data shown in Fig.  5C demonstrate that NZ pretreatment, before the addition of BFA, does not reduce the incorporation of radioactivity into either DLO or N-glycans. TLC of the BFA-provoked OSP, after 0 or 4 h NZ pretreatment, is shown when cells are radiolabeled in the presence of the glycosidase inhibitors (Fig. 5D, left panel). Quantitation of OSP recovered from two experiments is shown in Fig. 5D (right panel), and reveals that when BFA and NZ are added to the cells at the same time (0 h NZ), there is no inhibition of BFAprovoked OSP when compared with OSP quantitated in the absence of NZ (BFA alone), but after 2 and 4 h of NZ pretreatment there is a 40-50% reduction in BFAprovoked OSP. It was also noted that pretreatment of cells with NZ caused a change in the distribution of the BFAprovoked OSP. After 4 h NZ pretreatment, the ratio of Man 8 GlcNAc 2 -P to Glc 3 Man 9 GlcNAc 2 -P is reduced when cells are pretreated with NZ (left panel of Fig. 5D). A similar experiment was performed in the absence of glycosidase inhibitors and, as shown in Fig. 5E, preincubation of cells for both 0 and 4 h with NZ reduces the appearance of BFA-provoked complex-type OSP. Accordingly, the BFA-provoked appearance of OSP is partially inhibited by pretreating the cells with NZ for 2-4 h before addition of BFA, and suggests that microtubules are required for this process.

Examination of the effects of BFA on the subcellular localization of Co 2+ -activated DLODP activity
One possible explanation for the NZ-sensitivity of BFAprovoked appearance of OSP is that the retrograde transport of a normally GA-resident DLODP activity into the ER is required. Previous studies employing density gradient fractionation of mouse liver microsomes have indicated the presence of a Co 2+ -activated DLODP activity that codistributes with GA-situated UGT (see Fig. 1A). Whereas UGT-containing membranes derived from homogenates of normally cultivated cells are separated from heavier membranes of the ER during density gradient centrifugation, those derived from homogenates of BFA-treated cells partially shift to denser regions of density gradients that  (pretreatment time). Subsequently, the cells were treated with SW + KIF + CST and BFA (2 g/ml) or (DMSO) in radiolabeling medium, containing NZ or DMSO where appropriate, then 1 h later, radiolabeled as described in Fig. 2C. Glycoproteins recovered from the cells and medium were quantitated by scintillation counting after pronase digestion. The quantity of radioactivity associated with medium glycoproteins is expressed as a percentage of the total radioactivity recovered from cell and medium glycoproteins. C: The radioactivity incorporated into cellular and medium N-glycans and DLO derived from BFA-treated cells of the experiment described in (B). D (left panel): OSPs recovered from cells pretreated with NZ for 0 and 4 h before radiolabeling conducted in the presence of the glycosidase inhibitors and BFA were examined by TLC. The abbreviations used are as described for Fig. 3A. D (right panel): In two independent experiments conducted in the presence of SW + KIF + CST and BFA, where cells underwent 0, 2, and 4 h NZ pretreatment, OSPs were resolved by TLC. After fluorography, radioactive components were quantitated by densitometry. In these experiments control incubations, conducted in the presence of SW + KIF + CST, in which cells were not pretreated with NZ, were radiolabeled in the presence of BFA, but not NZ (conditions described in Fig. 1B). The quantity of OSP determined in this condition is set at 100%. The mean and standard error of the mean are shown. E (left panel): OSPs recovered from cells pretreated with NZ for 0 and 4 h before radiolabeling, conducted in the absence of the glycosidase inhibitors, but in the presence of BFA, were examined by TLC. The abbreviations used are as described for Fig. 3A. E (right panel): After fluorography, radioactive components were quantitated by densitometry, and the ratio of the quantity of OSPs in NZ pretreatment conditions to that of BFA alone (set at 100%) is presented. BFA has little effect on cytoplasmic OSP production from truncated DLO Next, we wanted to explore the relationship between the BFA-provoked OSPs and those generated from truncated DLO species. To this end, a HepG2 cell model of ALG12-CDG was generated by downregulating ALG12 expression in order to provoke the accumulation of Man-7 GlcNAc 2 -PP-dolichol. The time courses of inhibition of ALG12 mRNA expression and cell growth, after transduction of cells with three ALG12-targeting and one nontargeting siRNAs, are shown in Fig. 7A and Fig. 7B, respectively. Whereas all ALG12-targeting siRNAs inhibit ALG12 mRNA expression maximally between 2 and 4 days post transfection, cell growth is maximally affected between 4 and 8 days post transfection. Further experiments were performed with ALG12-targeting siRNA (2), whose ability to downregulate ALG12 mRNA expression was found to be the least transient of the three siR-NAs. Metabolic radiolabeling experiments revealed that contain ER markers (53). Accordingly, we asked whether the Co 2+ -activated DLODP activity could be redistributed to ER-containing fractions during density gradient fractionation of BFA-treated HepG2 cells. Data shown in Fig.  6A demonstrate that UGT and DLODP cofractionate in the absence of BFA, and after BFA treatment both activities partially redistribute to heavier membranes containing the ER markers, CNX and RPN1, as shown in Fig. 6B. As has been reported previously, BFA can cause a shift of ER marker proteins to lighter regions of density gradients (54). The BFA-provoked shift in density of DLODP-and UGT-containing membranes is less pronounced than that observed for either medial GA-situated MAN2A1 (Fig. 6C) or formimidoyltransferase cyclodeaminase (Fig. 6D), whose precise location within the GA is less well-characterized. Accordingly, taken along with the metabolic radiolabeling experiments described above, these subcellular fractionation studies demonstrate that a mechanism for the BFA-provoked increase of OSP involving microtubuledependent redistribution of a GA DLODP activity to the ER cannot be excluded. Fig. 6. Density gradient fractionation of control (Ctrl) and BFAtreated HepG2 cell homogenates. HepG2 cells were treated with 10 g/ml BFA for 1 h prior to harvesting and preparation of microsomes, as described in the Materials and Methods. The microsomes were submitted to OptiPrep density gradient fractionation. A: UGT and DLODP activities recovered from the gradient fractions were summed and the percentages of the total recovered in each fraction are shown. B-D: The ER markers, RPN1 and CNX, and the GA markers, MAN2A1 and formimidoyltransferase cyclodeaminase (FTCD), were quantitated by densitometry after SDS-PAGE and Western blot. The results are presented as described above. E: An OptiPrep gradient was generated in the absence of microsomal protein and the UV absorbance of each fraction was measured. The density of each fraction was calculated by comparing these absorbancies to those of OptiPrep solutions of known concentration and density. that endomannosidase could potentially also regulate DLO. Immuno-gold electron microscopy has shown that although 85% of endomannosidase is found associated with the cis GA, 15% is found to be associated with the cis GA network or ER/GA intermediate compartment (ERGIC) (55). Many proteins recycle between the ER and ERGIC the accumulation of Man 7 GlcNAc 2 -PP-dolichol in ALG12downregulated cells coincided with the inhibition in cell growth noted between 4 and 8 days post transfection (Fig.  7C). Whereas large OSPs were below the limit of detection in control-transfected cells, the appearance of the OSP, Man 7 GlcNAc 2 -P, coincided with the accumulation of Man 7 GlcNAc 2 -PP-dolichol in cells transfected with ALG12targeting siRNA. Next, glycosidase inhibitor-treated, ALG12downregulated cells were metabolically radiolabeled in either the presence or absence of BFA, as described in Fig. 1B. As seen in Fig. 8A, BFA provokes the appearance of Man 8 GlcNAc 2 -PP-dolichol and Man 8 GlcNAc 2 -P as well as Glc 3 Man 9 GlcNAc 2 -P, without having a major effect on the quantity of radioactivity associated with either Man 7 GlcNAc 2 -PP-dolichol or Man 7 GlcNAc 2 -P. We showed that the BFA-promoted appearance of OSP is sensitive to NZ (Fig. 5B). By contrast, data in Fig. 8B show that in ALG12downregulated cells, neither DLO nor OSP is affected by treating the cells with NZ before radiolabeling. In a similar experiment to that described in Fig. 8A, cell plasma membranes were selectively permeabilized with the bacterial pore-forming protein, SLO, at the end of the radiolabeling period, to yield an SLO perfusate (cytosol fraction: Cyt) and a residual cellular fraction comprising intracellular membrane bound compartments (MBCs). As expected from our previously reported findings using cells from an ALG12-CDG patient, Man 7 GlcNAc 2 -P is recovered mainly in the cytosolic compartment when cells are radiolabeled in the absence of BFA (left panel of Fig. 8B) and this situation is largely unaffected when cells are radiolabeled in the presence of BFA (right panel of Fig. 8B). By contrast, the BFA-provoked OSPs are predominantly recovered from the MBC. The ensemble of these results demonstrates that BFA provokes the appearance of a population of OSPs within the endomembrane system that is distinct from that generated constitutively when truncated DLO accumulates (15).

DISCUSSION
Here we examined DLO and OSP metabolism in cells treated with drugs that are known to perturb membrane flux along the endomembrane system. BFA and GCA, two agents known to fuse elements of the GA with the ER, were shown to have similar effects on DLO and OSP metabolism. For the first time, we show that BFA promotes MA-NEA1 processing of DLO. The function of MANEA1 is not clear, although it is thought that this enzyme deglucosylates glucosylated glycoproteins that leave the ER after escaping ER glucosidases. The endomannosidase-mediated removal of the d1 mannose residue (see Fig. 2B) means that reglucosylation of these N-glycans by ER-situated UDP-Glc glycoprotein glucosyltransferase cannot occur. Accordingly, it has been speculated that MANEA1 could play a role in the glycoprotein quality control system, during which UDP-Glc glycoprotein glucosyltransferasetagged misfolded glycoproteins are bound to either CNX or calreticulin to promote folding (55). Our results suggest Fig. 8. Effect of BFA and NZ on OSP production in a HepG2 cell model for ALG12-CDG. A: Cells were transfected with ALG12-targeting siRNA 2, and 6 days later were pretreated with 2.5 g/ml BFA and the glycosidase inhibitors (SW + KIF + CST), or the glycosidase inhibitors alone, for 1 h prior to addition of [2-3 H]mannose and further incubation for 2 h. DLO and OSP were extracted from cells and, after acid hydrolysis, the resulting neutral oligosaccharides were resolved by TLC. B: Cells were transfected as described in (A) and 6 days later were pretreated with DMSO or 20 g/ml NZ for 4 h. Prior to [2][3] H]mannose addition, the medium was replaced with radiolabeling medium supplemented (left panels) or not (right panels) with the glycosidase inhibitors (SW + KIF + CST) and containing DMSO or NZ. Cells were then incubated for 1 h before addition of [2-3 H]mannose and further incubation for 2 h. DLO and OSP were extracted from cells as described in the Materials and Methods, resolved by TLC, and then eluted and quantified by scintillation counting. Radioactivity associated with DLO or OSP recovered from cells pretreated with NZ was expressed as a percentage of controls without NZ. C: Cells were transfected with ALG12-targeting siRNA 2 and radiolabeled in the absence or presence of BFA exactly as described in (A), rinsed with ice-cold PBS, and permeabilized with SLO as described in the Materials and Methods. OSPs were recovered from the SLO perfusate (Cyt) and residual cellular material (MBCs). In control experiments, SLO was not added to the permeabilization buffer, and in this case the OSP recovered from the MBC represents total OSP (T). After acid hydrolysis of these components, released oligosaccharides were re- synthesized on the cytoplasmic face of the ER, any block in their synthesis might be expected to give rise to cytoplasmic OSPs (7,14). By contrast, Man 7 GlcNAc 2 -PP-dolichol is generated in the lumen of the ER and how this structure gives rise to cytoplasmic OSP is not clear. It is possible that, after generation within the lumen of the ER, Man 7 GlcNAc 2 -P could be rapidly and efficiently transported into the cytosolic compartment. Nevertheless, in vitro experiments using semi-intact cells derived from a ALG12-CDG patient demonstrated that while Man 7 GlcNAc 2 -P was recovered predominantly in the cytosol, nfOS and glycosylated acceptor peptide generated in the same experiments remained predominantly within the lumen of the ER (15). In the face of this data, it was proposed that if Man 7 GlcNAc 2 -PP-dolichol accumulates in the lumen of the ER, it could potentially be flipped back onto the cytoplasmic face of the ER, where it could be hydrolyzed to yield cytoplasmic Man 7 GlcNAc 2 -P (15). Indeed, protein-mediated Man-7 GlcNAc 2 -PP-dolichol flipping across artificial membranes has been demonstrated, but was shown to be much slower than the flipping of Man 5 GlcNAc 2 -PP-dolichol, and about as efficient as the flipping of Man 9 GlcNAc 2 -PP-dolichol (60). Thus, data from these in vitro flippase assays do not support the idea that the selectivity of the DLO flippase could regulate access of luminally generated DLO intermediates to a cytoplasmically oriented OSP-generating mechanism. Nevertheless, it is quite possible that, in vivo, the selectivity of lumen-to-cytoplasm DLO flipping is somewhat different to that observed in artificial membranes in vitro. In the present study, we show that BFA promotes the accumulation of Man 8 GlcNAc 2 -PP-dolichol and that, in this situation, cytoplasmic Man 8 GlcNAc 2 -P is not detected. In the presence of BFA and glycosidase inhibitors, Man 8 GlcNAc 2 -P is detected within the BFA compartment, but at present we do not know whether Man 8 GlcNAc 2 -P is derived directly from Man 8 GlcNAc 2 -PP-dolichol hydrolysis or whether Glc 3 Man 9 GlcNAc 2 -P is cleaved from Glc 3 Man 9 GlcNAc 2 -PP-dolichol and subsequently hydrolyzed by MANEA1 to yield Man 8 GlcNAc 2 -P. The latter hypothesis would predict that the OSP-generating mechanism within the BFA compartment might be selective for mature triglucosylated DLO and that BFA might amplify the OSP generation from mature DLO that has been previously noted to occur at low levels in various cell types (17). However, at present, it is not possible to say whether the BFApromoted OSP population described here has the same origin as that of the previously described OSP population that appears to be derived from mature DLO.
Whatever the origins of the larger OSP structures seen in control or BFA-treated cells, their relationship to the cytosolic structures generated from truncated DLO intermediates containing seven or fewer residues of mannose is clearer. Here, we show that despite striking increases in OSPs derived from mature DLO within the BFA compartment, the generation of Man 7 GlcNAc 2 -P from Man 7 GlcNAc 2 -PP-dolichol is largely unaffected by BFA. Furthermore, we show that, in the presence of BFA, Man 7 GlcNAc 2 -P is still recovered from the cytoplasmic compartment of cells, even though the BFA-provoked OSPs are recovered from and it is not inconceivable that, in the absence of BFA, ERGIC-situated endomannosidase transits through the ER where it could potentially act on glucosylated DLO and/or glycoproteins. Alternatively, glucosylated DLO intermediates that escape the ER could be rapidly deglucosylated in the ERGIC or cis GA as part of a catabolic pathway. Nevertheless, we cannot rule out the possibility that the endomannosidase-generated (d2,d3) Man 8 GlcNAc 2 -PP-dolichol intermediate could be slowly remannosylated by the ALG9 gene product that adds the d2 and d3 terminal nonreducing mannose residues (see Fig.  2B) to the growing DLO in the lumen of the ER. In the light of our results, it will be interesting to evaluate the role of endomannosidase in DLO regulation in normally cultivated cells.
We also show for the first time that BFA promotes the appearance of complex type OSPs within the endomembrane system. A high proportion of these OSPs are shown to possess terminal nonreducing GlcNAc residues consistent with their processing by enzymes that are normally found in the cis and medial GA. Such structures have also been shown to predominate on glycoproteins trapped within the BFA compartment (37,38). These observations suggest a generalized deficiency in galactosylation by normally trans GA-situated UGT in the BFA compartment (see Fig. 1A). Using fluorescent microscopy, it has been shown, in many cell types, that BFA causes fusion of cis, medial, and trans elements of the GA with the ER to form the BFA compartment (Fig. 1A). By contrast, the trans Golgi network was shown to fuse with elements of the endosomal system. Why is glycoprotein and OSP galactosylation reduced in BFA-treated cells? There is a progressive acidification of organelles of the endomembrane system from pH 7.2 in the ER to pH 6.0 in the trans Golgi network (56), and studies show that GA-situated glycosyltransferases form biologically important homo-and heterodimers at lower pH (57,58). Therefore enzymes normally resident in distal GA cisternae are perhaps less active in the BFA compartment, whose pH is likely to be similar to that of the ER (59).

What is the relationship between the BFA-promoted OSPs and other OSP populations?
The oligosaccharide structures found in the different OSP pools merit discussion because they throw light on the compartmentalization and membrane topology of the OSP-generating processes. Studies using cells harboring mutations in genes encoding enzymes of the dolichol cycle demonstrate that only when DLO structures containing seven or fewer residues of mannose accumulate are there concomitant increases in the corresponding OSP structures (7,14,15). Either OSPs are not produced from larger DLO structures in increased amounts or, if they are, they are rapidly metabolized. Previously, we proposed that the selectivity of this OSP generating process is not related to the selectivity of the OSP generating enzyme, but is due to the capacity of the accumulated DLO to be exposed to the cytoplasmic face of the ER (15). Because DLO structures containing less than five residues of mannose are from the ER of a newly synthesized DLODP whose normal subcellular localization is in a post ER compartment. In this situation DLO would be cleaved in a nonphysiological process. We can rule out this explanation because pretreating cells with NZ before BFA treatment essentially causes an ER export block and, under these conditions, we show that there is actually a reduction in BFA-provoked OSPs. The recently described Co 2+ -activated DLODP activity codistributes with the trans GA marker, UGT (see Fig.  1A), during density gradient fractionation of mouse liver microsomes (18). Here we show that these DLODP and UGT activities are associated with membranes of similar density in both the absence and presence of BFA, suggesting that these two activities may colocalize in the GA. Interestingly, MAN2A1, which is considered to be a marker for the medial GA (see Fig. 1A), shows a greater BFA-induced shift on the density gradient than UGT or DLODP, showing that Co 2+ -activated DLODP and MAN2A1 are differently distributed in HepG2 membranes. When HepG2 cells are treated with BFA, there is an increase in codistribution of membranes containing ER markers with those containing Co 2+ -activated DLODP, suggesting that fusion of DLO-containing ER membranes with DLODP-containing GA membranes could underlie the increased OSPs seen in BFA-treated cells. To summarize, presently we do not know why BFA provokes an increase in OSPs derived from mature DLO in the lumen of the endomembrane system, but taking into account that this process is blocked by NZ and that subcellular fractionation studies demonstrate a BFA-provoked shift in the density of DLODPcontaining membranes toward that of ER membranes, we cannot exclude the hypothesis that BFA causes the mixing of a GA-situated DLODP activity with ER-situated DLO.
What is the physiological significance of OSP generation from mature DLO within the endomembrane system?
If it is hypothesized that BFA merely blocks the catabolism of OSPs that are constitutively generated from mature DLO within the ER, then OSP generation in cells derived from ALG6-CDG and ALG8-CDG patients, which accumulate Glc 1-0 Man 9 GlcNAc 2 -PP-Dol, may actually be much higher than previously noted (14,15). In this situation OSP release may reflect a mechanism involved in reducing accumulation of unwanted DLO and recycling DolP. On the other hand, the physiological significance of the alternative mechanism, where translocation of a GA-situated DLODP into the ER underlies BFA-provoked OSP generation, remains obscure. The key to understanding this phenomenon may lie in finding out whether Golgi DLODP comes into contact with DLO under physiological conditions. For example, during mitosis, when GA enzymes associate with the ER (62), could Golgi DLODP play a role in DLO control? Alternatively, under other circumstances, DLO could flow out of the ER into the GA, and be hydrolyzed by DLODP.
To conclude, BFA causes MANEA1 processing of DLO so that the unusual (d2,d3)Man 8 GlcNAc 2 -PP-dolichol structure accumulates. This truncated DLO does not give rise to cytoplasmic Man 8 GlcNAc 2 -P, as has been shown for within the endomembrane system. Finally, we report that whereas the BFA-promoted appearance of luminal OSPs is partially blocked by NZ, the generation of Man 7 GlcNAc 2 -P from Man 7 GlcNAc 2 -PP-dolichol is not. These data point to independent processes for the generation of these two OSP pools.

Why does BFA promote the appearance of complex type OSPs within the endomembrane system?
There are two general explanations: BFA either blocks the catabolism or secretion of constitutively generated OSPs, or it increases OSP generation. The first explanation is that OSPs are normally generated from mature DLO within the membrane system, but in the absence of BFA they are either secreted into the extracellular medium or are rapidly further metabolized and difficult to detect. We show that in the absence of BFA, small amounts of radioactivity are associated with OSP-like material in the extracellular medium. However, due to the small amounts recovered, this material could not be characterized and it remains to be determined whether or not it actually corresponds to OSPs. Furthermore, the summed radioactivity associated with OSP-like material and nfOS recovered from the medium of control cells was not sufficient to account for the amount of radioactivity associated with cellular OSPs recovered from BFA-treated cells. So, does BFA block further metabolism of normally generated OSPs? We cannot rule out this possibility; but in experiments with concanamycin A, the vacuolar proton pumping ATPase inhibitor, or the ionophore, monensin, which increase the pH of acidic organelles such as lysosomes, endosomes, and GA (61), we were unable to reproduce the effects of BFA on OSP metabolism (A. Massarweh et al., unpublished observations). Furthermore, NZ inhibits the BFA-promoted appearance of OSPs. If the only effect of BFA is to block the degradation of constitutively generated OSPs in the endomembrane system, then it has to be hypothesized that NZ relieves this blockade. Although this cannot be ruled out, it seems unlikely, as we demonstrate that NZ does not affect the capacity of BFA to block the secretory pathway.
The second explanation for the effect of BFA on OSPs is that this drug promotes an increase in OSP generation from mature DLO. For example, BFA could increase the expression of an OSP generating enzyme, or increase the access of an OSP generating enzyme to DLO within the endomembrane system. In this regard, although the recently described Co 2+ -activated DLODP activity is probably not the only OSP-generating enzyme, its activity was found to be similar in both control and BFA-treated HepG2 cells (A. Massarweh et al., unpublished observations). BFA is known to provoke the unfolded protein response and it is possible that this leads to changes in expression of enzymes involved in OSP metabolism. However, for the 3 h BFA incubation times reported here, BFA did not activate the unfolded protein response, as judged by changes in either CNX or protein disulphide isomerase expression measured by Western blot (A. Massarweh et al., unpublished observations). BFA could also block the export smaller DLO species that accumulate in certain types of CDG. Nevertheless, BFA provoked a striking increase in OSPs recovered from within the endomembrane system without affecting the quantity of a cytoplasmic OSP derived from truncated DLO. Therefore the use of BFA has delineated at least two OSP-generating mechanisms, but further understanding of the compartmentalization and membrane topology of both DLO and DLODP activities will be required before insights into the function of OSPgenerating processes can be achieved.