PDMP Blocks Brefeldin A–induced Retrograde Membrane Transport from Golgi to ER: Evidence for Involvement of Calcium Homeostasis and Dissociation from Sphingolipid Metabolism

d-sphingosine, BFA, cyclosporin A, thapsigargin, and E3A5 mouse monoclonal anti–β-COP antibody were obtained from Sigma Chemical Co. (St. Louis, MO). The succinimidyl ester of C₆-NBD. Bodipy™-503/512-C₅, and indo-1-HS/AM were from Molecular Probes (Eugene, OR). PDMP, d,l-threo-lyso-PDMP, C₆-Cer, and 1-β-d-glucosylsphingosine were purchased from Matreya, Inc. (Pleasant Gap, PA). TRITC/FITC anti–rabbit/ mouse IgG F(ab′)₂ fragments and ATP were from Boehringer Mannheim GmbH (Mannheim, Germany). MK571 and PSC833 were kind gifts from Dr. E. Vellenga (University of Groningen, the Netherlands). Anti-ManII polyclonal antibodies were kindly provided by Dr. A. Velasco (University of Sevilla, Spain), and anti-GMP_t-1 antibodies were provided by Dr. I. Sandoval (Centro de Biologia Molecular Severo Ochoa, Consejo Superior de Investigaciones Cientificas, Universidad Autonoma de Madrid, Spain).

Monocultures of HT29 G+ or normal rat kidney (NRK) cells were grown in Dulbecco's minimal essential medium (DME, containing 25 mM glucose), supplemented with 10% (vol/vol) decomplemented (56°C, 30 min) FCS, in a water-saturated atmosphere of 5% CO₂/95% air. During the exponential phase of growth, the culture medium was changed every 48 h. For maintenance purposes, cells were trypsinized once (HT29) or twice (NRK) a week and plated at appropriate densities to obtain confluent layers after 1 wk of culture.

(Fluorescent) short-chain sphingolipids/PDMP were synthesized according to Kok and Hoekstra (1993). C₆-NBD-Cer and C₆-NBD-PDMP were synthesized from d-erythro-sphingosine and d,l-threo-lyso-PDMP, respectively, and the succinimidyl ester of C₆-NBD. C₅-Bodipy-Cer was synthesized from d-erythro-sphingosine and Bodipy™-503/512-C₅. C₆-GlcCer was synthesized from 1-β-d-glucosylsphingosine and hexanoic acid. For cell incubations, C₆-(NBD-)sphingolipid/PDMP solutions were made by ethanol injection in serum-free culture medium (0.5% [vol/vol] final ethanol concentration; see also Kok et al., 1992).

Approximately 10⁷ HT29 G+ or NRK cells were incubated with 10 μM C₆-NBD-Cer for 60 min at 37°C. After these incubations, the cells together with the incubation medium were subjected to lipid extraction by the procedure of Bligh and Dyer (1959). C₆-NBD-sphingolipids were separated by high performance thin layer chromatography (HPTLC) using CH₃Cl/CH₃OH/20% (wt/vol) NH₄OH (14:6:1, vol/vol/vol) as the running solvent system. Individual C₆-NBD-sphingolipid species (Cer, GlcCer, and SM) were scraped from the HPTLC plates and eluted from the silica with 1% (vol/vol) Triton X-100 by vigorous shaking at 37°C. Individual C₆-NBD-sphingolipid species were quantified by measuring the NBD fluorescence in a fluorimeter at an excitation wavelength of 465 nm and an emission wavelength of 530 nm.

Approximately 10⁷ HT29 G+ cells were washed with Hanks' solution and incubated with 10 μM C₆-NBD-PDMP for 30 min at 37°C. The incubation buffer was aspirated and replaced by fresh Hanks' solution, followed by incubations at 37°C for various time intervals. At each time point, both the cell fraction and the incubation medium were subjected to lipid extraction (see above). The amount of C₆-NBD-PDMP in each fraction was measured in a fluorimeter as described above. In all experiments, mock-treated cells were subjected to lipid extraction and fluorescence measurement. These background values were subtracted from the values of cells, which had been incubated with C₆-NBD-PDMP.

For sodium dithionite quenching studies, the initial 37°C incubation with 10 μM C₆-NBD-PDMP was followed by a 30-min incubation in Hanks' solution containing 30 mM Na₂S₂O₄ (diluted from a stock solution of 1 M Na₂S₂O₄ in 1 M Tris, pH 10; c.f. McIntyre and Sleight, 1991). Control cells were incubated in Hanks' solution without sodium dithionite. To study the effect of sodium dithionite on C₆-NBD-Cer, cells were incubated with 10 μM C₆-NBD-Cer for 30 min at 37°C, followed by a “back- exchange” at 2°C. To this end, cells were incubated with 5% (wt/vol) BSA in Hanks' solution for 30 min at 2°C, followed by extensive washing. Thereafter, cells were incubated in Hanks' solution with or without 30 mM Na₂S₂O₄ for 30 min at 37°C, followed by a back-exchange at 2°C. Back-exchanges were performed to ensure that plasma membrane pools of C₆-NBD-lipid did not contribute to the measured values in control cells. These pools consist of C₆-NBD-Cer, which was initially inserted in the plasma membrane, and/or C₆-NBD-sphingolipid products, which were synthesized in the Golgi apparatus and transported to the plasma membrane. In cells subjected to dithionite, quenching of this pool of C₆-NBD-lipid will occur. In contrast to the experiments with C₆-NBD-Cer, in the case of C₆-NBD-PDMP, similar results were obtained, regardless of whether the back-exchange had been carried out.

In all experiments, total lipid phosphorus was determined in each fraction after perchloric acid destruction according to Böttcher et al. (1961). In addition, in cell fractions the number of cells was counted. All C₆-NBD-PDMP/Cer determinations were normalized to total lipid phosphorus.

Cells were grown on glass coverslips contained in 35-mm-diam Petri dishes. Experiments were carried out 72 h after passage. Before experiments, the cells were cooled (30 min) and washed several times with ice-cold Hanks' solution. Cells were incubated with C₆-NBD-Cer (5 μM) in Hanks' solution at 2°C for 1 h, followed by a wash with Hanks' buffer. Subsequently, the cells were subjected to 30-min incubations with Hanks' solution, containing 5 μg/ml BFA, 100 μM (lyso-)PDMP, or both, at 37°C. Occasionally, either C₆-Cer or C₆-GlcCer (both 100 μM), or one of the MDR inhibitors was included. Also during these incubations, C₆-NBD-Cer (2 μM) was present. Finally, a back-exchange was performed by incubating the cells with 5% BSA in cold (2°C) Hanks' solution for 30 min, followed by extensive washing. In the case of intracellular PDMP localization, cells were incubated with C₆-NBD-PDMP (100 μM) in Hanks' solution for 30 min at 37°C. The effect on Golgi morphology was in this case monitored by using C₅-Bodipy-Cer. For this purpose, cells were preincubated with 5 μM C₅-Bodipy-Cer for 60 min at 37°C to ensure that high enough concentrations of Bodipy-sphingolipids at the Golgi apparatus were reached for a spectral shift in fluorescence to occur (Pagano et al., 1991). For dithionite quenching, cells were incubated with Hanks' solution containing 30 mM Na₂S₂O₄ (diluted from a stock solution of 1M Na₂S₂O₄ in 1 M Tris, pH 10; c.f. McIntyre and Sleight, 1991) for 30 min at 37°C, allowing fluid-phase uptake of this NBD quencher in endocytic compartments to occur. Hence, this procedure will result in loading of the endo/lysosomal system with the NBD quencher (Kok et al., 1995).

For immunocytochemical studies, following various incubation schemes with BFA and/or PDMP, cells were fixed in −20°C methanol for 30 min. Thereafter, cells were washed four times with PBS, followed by a 30-min incubation with PBS/1% (wt/vol) BSA. Fixed cells were incubated overnight at 4°C with the primary antibody diluted in PBS/1% (wt/vol) BSA (anti-ManII: diluted 1:4,000 for NRK cells and 1:80 for HT29 G+ cells; 18B11: diluted 1:100), followed by two washes with PBS/1% (wt/vol) BSA. Thereafter, fixed cells were incubated for 60 min at room temperature with the secondary antibody TRITC anti–rabbit IgG F(ab′)₂ fragment (dilution 1:30), followed by two washes with PBS/1% (wt/vol) BSA. In case of double immunofluorescence staining, the cells were incubated with two primary antibodies (anti-ManII and anti–β-COP: diluted 1:20) overnight and subsequently with FITC anti–rabbit IgG F(ab′)₂ fragment + TRITC anti–mouse IgG F(ab′)₂ fragment (dilutions 1:30) for 60 min. Finally, fixed cells were mounted and examined by fluorescence microscopy. Fluorescence microscopy was performed with an Olympus (Melville, NY) AX-70 research microscope, equipped with a PM20 photomicrography system. Photomicrographs were taken with 10-s (NBD fluorescence) or 1-min (FITC/TRITC fluorescence) exposure times using Ilford (Cheshire, UK) HP5 film that was processed at 1,600 ASA. All images from HT29 cells were taken with a confocal scanning laser microscope (True Confocal Scanner 4D; Leica, Heidelberg, Germany) equipped with an argon-krypton laser and coupled to a Leitz DM IRB inverted microscope (Leica). Images were taken at 488 nm for NBD fluorescence and 562 nm for Bodipy/TRITC fluorescence.

Intracellular calcium concentrations were measured as described previously (Sipma et al., 1995; Filipeanu et al., 1997). After trypsinization, the cells were suspended in Hanks' solution at a concentration of 10⁷ cells/ml. The cells were incubated with 2 μM indo-1/AM at room temperature for 30 min in the dark. The cells were collected by centrifugation and washed twice with Hanks' solution before fluorescence measurements. Intracellular indo-1 fluorescence was measured in a spectrophotometer (model Aminco^®-Bowman; Spectronic Instruments, Inc., Rochester, NY), using 10⁶ cells/ml for each measurement. Measurement was performed at 23°C to prevent dye compartmentalization and leakage. The excitation wavelength was 349 nm, and emission at 410 and 490 nm were acquired with a frequency of 1 Hz. Drugs were added with a Hamilton syringe to a magnetically stirred cell suspension in a volume <1% of the total cell suspension volume (3 ml). Actual [Ca²⁺]_i values were calculated by a classical equation (Grynkiewicz et al., 1985) using R_max and R_min values obtained at the end of each experiment using 60 mg/ml saponin and 20 mM EGTA, pH 8.0, respectively. Autofluorescence of the cells, i.e., the signal remaining after 10 mM MnCl₂ addition, was subtracted before calculation. Calcium-free medium is defined as Hanks' buffer to which 1.5 mM EGTA is added (final [Ca²⁺]_i < 1 μM).

As shown in Fig. 1,B, incubation of HT29 G+ cells with BFA resulted in loss of discrete Golgi structures that had been labeled with C₆-NBD-Cer (compare to control cells in Fig. 1,A). However, when BFA incubation had been preceded by PDMP treatment (Fig. 1,E), discrete Golgi structures were still observed. Apparently, PDMP blocked the BFA-induced retrograde C₆-NBD-lipid flow from Golgi to ER. The observed effects were critically dependent on the order of addition of both drugs. Thus, when BFA was allowed to act before PDMP was added to the cells (Fig. 1,F), discrete Golgi structures were no longer detectable, i.e., all Golgi membranes had already merged with the ER compartment by BFA-induced retrograde flow. In the case when both drugs were added simultaneously (Fig. 1 D), Golgi compartments did not disappear.

To investigate whether the effect of PDMP was specific for (C₆-NBD-)lipid flow or was due to a general block of membrane flow, the following experiments were performed. Cells were treated with BFA and/or PDMP, as in Fig. 1, A–F, but instead of monitoring the fate of C₆-NBD-Cer in living cells, the localization of a Golgi resident protein was assessed after fixation of the cells. As shown in Fig. 1, G–N, the fate of a cis/medial-Golgi resident protein, ManII, was very similar to that observed for C₆-NBD-Cer in living cells. Indeed, BFA treatment (Fig. 1,H) caused a loss of discrete Golgi structures, while PDMP resulted in maintenance of the Golgi compartment, provided that it had been administered before (Fig. 1,M) or together with (Fig. 1 L) BFA. In conclusion, not only is BFA-induced retrograde lipid flow inhibited by PDMP, but the entire membrane flow is blocked.

The effect of PDMP on BFA-induced retrograde Golgi to ER membrane flow was not a feature that occurred only in the human HT29 G+ colon tumor cell line. In an entirely different cell type, the NRK fibroblast, PDMP affected BFA-induced retrograde membrane flow as well, as indicated by either C₆-NBD-Cer labeling (Fig. 2, A–F) or ManII staining (Fig. 2, G–N). Note that when PDMP was administered simultaneously with BFA in NRK cells (Fig. 2, D and L), the Golgi complex appeared swollen, while C₆-NBD-lipid staining suggested the presence of tubules (Fig. 2,D). These structural changes were not observed when PDMP was allowed to exert its effect before BFA addition. It thus appeared that PDMP interfered with the ongoing redistribution of membranes into the ER, just after the onset of this process induced by BFA. A trans-Golgi resident protein, GMP_t-1 (Alcade et al., 1992, 1994), was also affected by PDMP treatment (Fig. 2, P–V), although in this case some residual Golgi structure was observed in BFA-treated cells.

Although most experiments were performed at the time scale of 1 h, the effect of PDMP on BFA-induced retrograde Golgi to ER membrane flow was also observed after extended incubation time intervals. BFA + PDMP treatment during 3 h did not result in disappearance of discrete Golgi structures (Fig. 3,A; c.f. Fig. 2,L). When PDMP was removed from the medium, its inhibiting effect on BFA-induced retrograde membrane flow turned out to be reversible. After an initial incubation with PDMP + BFA for 30 min (Fig. 3,B), a subsequent incubation with BFA alone resulted in a gradual loss of discrete Golgi structures (Fig. 3, C and D). Note that only 3 h after the withdrawal of PDMP, the effect of BFA was completed.

The effect of PDMP on BFA-induced retrograde membrane flow was concentration dependent. This was observed when cells were incubated with PDMP at various concentrations, ranging from 10 to 100 μM (in steps of 10 μM), followed by incubation with both PDMP (at the respective concentrations) and BFA. In HT29 cells, the inhibition of BFA-induced retrograde membrane flow, as monitored by C₆-NBD-Cer staining, became apparent at a PDMP concentration of 50 μM, although not all cells were affected at this concentration. NRK cells appeared to be slightly less sensitive, as inhibition became apparent at 80 μM. Furthermore, the inhibitory effect of PDMP required the intact molecule because lyso-PDMP, which lacks the fatty-acyl moiety, was not able to block BFA-induced retrograde membrane flow in both HT29 G+ and NRK cells (Fig. 4).

BFA-induced dissociation of coatomer from membranes precedes and is required for redistribution of Golgi membranes into the ER compartment (Scheel et al., 1997). When the distribution of the coatomer protein subunit β-COP (Aridor and Balch, 1996; Bednarek et al., 1996) was analyzed in PDMP- and BFA-treated NRK cells, it turned out that its release from the Golgi complex, as induced by BFA, was not prevented by PDMP (Fig. 5). Regardless of whether BFA-treated cells were initially (Fig. 5,L) or simultaneously (Fig. 5,H) incubated with PDMP, β-COP was released, indistinguishably from its release in cells treated with BFA alone (Fig. 5,D). Note that release of β-COP occurred under conditions where the Golgi complex did not disappear (Fig. 5, G and K), while PDMP by itself did not affect the distribution of β-COP (Fig. 5 F, compare with B and D).

A correlation between anterograde membrane transport in the biosynthesis pathway and SM biosynthesis has been suggested by Rosenwald et al. (1992) for CHO cells. Since PDMP is known to affect sphingolipid metabolism in other cell types as well (Radin et al., 1993), our results regarding inhibition of BFA-induced retrograde membrane transport might be explained similarly. However, while GlcCer biosynthesis is very sensitive to PDMP in all cell types tested (Table I; c.f. Rosenwald et al., 1992), the effect of PDMP on SM biosynthesis is less drastic and, furthermore, cell type dependent. In CHO cells, SM biosynthesis is inhibited at PDMP concentrations above 40 μM (Rosenwald et al., 1992), coinciding with concentrations that impair anterograde VSV-G protein transport. In the present work, using NRK cells, inhibition of SM biosynthesis occurred with 100 μM PDMP (Table I). Yet in HT29 G+ cells, no inhibition of SM biosynthesis was found (Table I). In conclusion, PDMP had different effects on C₆-NBD-SM biosynthesis in NRK and HT29 cells, while in both cell types BFA-induced retrograde membrane flow was inhibited. Therefore, these results argue against a correlation between inhibition of BFA-induced retrograde membrane flow and the biosynthesis of SM.

Apart from an effect on the levels of GlcCer and SM, PDMP treatment could result in an increase in Cer. The latter has been correlated with an inhibition of protein trafficking through the secretory pathway (Rosenwald and Pagano, 1993). To establish whether the block in BFA- induced retrograde membrane flow from Golgi to ER could be attributed to an increase in Cer or a decrease in GlcCer, the following experiments were performed. C₆-NBD-Cer–labeled HT29 G+ cells were initially incubated with either C₆-Cer or a combination of PDMP and C₆-GlcCer. This was done to mimic a possible increase in Cer by PDMP or to compensate for the loss of GlcCer due to treatment with the drug, respectively. If the action of PDMP proceeded through changes in sphingolipid levels, one could expect that C₆-Cer mimicked the action of PDMP or that the effect of the drug was counteracted by supplying the cells with GlcCer. As shown in Fig. 6, this was not the case, even at the high concentrations (100 μM) of (nonfluorescent) short-chain sphingolipids that were used. Incubation with 100 μM C₆-Cer did not result in preservation of discrete Golgi structures (Fig. 6, A and C) during subsequent incubation with BFA, while the addition of C₆-GlcCer did not result in loss of discrete Golgi structures in PDMP-treated cells (Fig. 6, B and D). In conclusion, inhibition of BFA-induced retrograde membrane flow does not appear to be correlated to changes in either GlcCer or Cer levels.

To further investigate the mechanism of action of PDMP on membrane flow, we synthesized a fluorescent analogue of the drug, C₆-NBD-PDMP, enabling the study of uptake, efflux, and intracellular localization of the drug. As shown in Fig. 7,A, C₆-NBD-PDMP was rapidly taken up by HT29 G+ cells. Incubation with 10 μM C₆-NBD-PDMP for 30 min at 37°C resulted in uptake of 1.58 pmol C₆-NBD-PDMP/nmol cellular lipid phosphorus (n = 2), or 3.74 × 10⁷ molecules/cell. This was 3.8% of the total amount of C₆-NBD-PDMP administered to the cells. Half of this uptake was already achieved after 1.8 min (Fig. 7,A). Interestingly, when the C₆-NBD-PDMP–containing incubation solution was removed and the cells were subsequently incubated in fresh Hanks' solution, a rapid efflux of C₆-NBD-PDMP occurred (Fig. 7,B). Half of the initial cell- associated pool was released into the incubation solution after only 40 s, while within 30 min 89% was released. The residual intracellular pool remained constant during prolonged incubation up to at least 3 h (Fig. 7 B). TLC analysis revealed that C₆-NBD-PDMP was not metabolized to lyso-PDMP and NBD-C₆–fatty acyl chain during these incubations (data not shown).

When HT29 G+ cells were incubated with 100 μM C₆-NBD-PDMP for 30 min at 37°C, a staining pattern was observed (Fig. 8,A), which is indicative of a primary localization of the drug in endocytic organelles. This was further examined by loading the cells with sodium dithionite. This NBD fluorescence quencher is taken up into the endocytic pathway by fluid-phase endocytosis and transported to endocytic compartments, including lysosomes. As a result, the fluorescence of all NBD-labeled molecules residing in endocytic compartments will be quenched (Kok et al., 1995). When such an incubation was performed on cells that had previously taken up C₆-NBD-PDMP, intracellular fluorescence was quenched, as judged by confocal scanning laser microscopy (Fig. 8,B). Indeed, when the extent of NBD quenching was measured in a parallel experiment with the same incubation protocol (see Materials and Methods), 91.3% (±5.6; n = 3) quenching of C₆-NBD-PDMP was found. It should be noted that, given the experimental setup (see Materials and Methods), the pool of C₆-NBD-PDMP, which is quenched under these conditions, is the residual intracellular pool after large scale efflux (see above). When HT29 G+ cells were labeled with C₆-NBD-Cer, followed by uptake of sodium dithionite, Golgi labeling was readily observed (Fig. 8 C), indicating that sodium dithionite did not acquire access to the Golgi apparatus. Determination showed that in this case only 16.7% (±6.9; n = 3) quenching of NBD fluorescence occurred. In conclusion, although the cell-biological effect of PDMP is evidently associated with Golgi functioning, the drug itself did not reach this organelle since C₆-NBD-PDMP could not be detected in the Golgi apparatus (c.f. Rosenwald and Pagano, 1994).

An important question to be addressed was whether the fluorescent derivative of PDMP inhibited BFA-induced retrograde Golgi to ER membrane flow, similarly to native drug. HT29 G+ cells were incubated with C₆-NBD-PDMP (100 μM), followed by BFA treatment (Fig. 8,D, c.f. to Fig. 1,E) or pretreated with BFA, followed by incubation with C₆-NBD-PDMP (Fig. 8,E, compare to Fig. 1,F). To identify the Golgi in this case, the cells were prelabeled with C₅-Bodipy-Cer instead of C₆-NBD-Cer. Concentration of C₅-Bodipy-sphingolipids at the Golgi apparatus results in red fluorescence (Pagano et al., 1991), which can be discriminated from the green NBD fluorescence of the C₆-NBD-PDMP. As shown in Fig. 8, D and E, C₆-NBD-PDMP was able to prevent BFA-induced Golgi–ER merging, provided that the fluorescent PDMP analogue was added before BFA. Thus, C₆-NBD-PDMP behaved similarly to PDMP, in terms of inhibition of BFA-induced retrograde membrane flow.

Recently, a correlation was established between the MDR status of cells and the accumulation of glucosylceramides (Lavie et al., 1996). Furthermore, agents that reverse multidrug resistance were shown to inhibit GlcCer synthase, the target of PDMP (Lavie et al., 1997). In view of this similarity in target, we examined whether such MDR- reversing agents could also inhibit BFA-induced retrograde membrane transport. When HT29 G+ or NRK cells were incubated with 50 μM MK571, a specific inhibitor of multidrug resistance protein (MRP) function (Gekeler et al., 1995), followed by incubation with BFA, discrete Golgi structures were observed (Fig. 9, A and C). Thus, similar to PDMP, this MRP-specific MDR-reversing agent was able to block BFA-induced retrograde membrane flow from the Golgi to the ER. MK571 also caused an inhibition of GlcCer biosynthesis from C₆-NBD-Cer, be it much less efficient than PDMP. In comparison, in HT29 G+ cells 100 μM PDMP inhibited C₆-NBD-GlcCer biosynthesis 34.7-fold (±12.3; n = 3), while 50 μM MK571 caused an inhibition of only twofold (±0.4; n = 3). MDR-reversing agents acting primarily on P-glycoprotein (P-gp), such as cyclosporin A and PSC833 (Friche et al., 1991; Watanabe et al., 1995), had no effect on C₆-NBD-GlcCer biosynthesis (data not shown). In NRK cells, cyclosporin A and PSC833 did not affect BFA-induced retrograde membrane flow from Golgi to ER (Fig. 9,D). In HT29 cells, both P-gp drugs showed some effect, although not as prominent as MK571 (Fig. 9 B). Interestingly, when C₆-NBD-PDMP uptake was measured (as shown above) in HT29 G+ cells, its cellular accumulation during a 10-min incubation was increased to 140% (n = 2) in the presence of MK571, while cyclosporin A and PSC833 did not modulate its accumulation. The absence of effects of P-gp antagonists is compatible with results from Rosenwald and Pagano (1994), showing that in CHO cells that were selected for increased resistance to PDMP toxicity, this resistance was not mediated by P-gp.

To explore the molecular mechanism behind the inhibiting effect of PDMP on BFA-induced retrograde membrane flow, we measured the intracellular calcium concentration in HT29 cells and its potential modulation by PDMP. The rationale for such experiments is rooted in the observations made by Ivessa et al. (1995), who reported that BFA-induced retrograde membrane transport from Golgi to ER depends on calcium homeostasis. When HT29 G+ cells were incubated with PDMP, modulation of [Ca²⁺]_i was indeed observed (Fig. 10,A). The average basal [Ca²⁺]_i was 144 ± 11 nM (n = 78). Addition of 100 μM PDMP resulted in a transient increase of 154 ± 24 nM (n = 8) above basal [Ca²⁺]_i. The rise in [Ca²⁺]_i depended on the concentration of PDMP (Fig. 11,A), with a modest increase observed at 50 μM (22 ± 12 nM; n = 7) and a large enhancement at 150 μM (445 ± 70 nM; n = 6). Below 50 μM, responses were not observed, in accordance with the absence of effect on BFA-induced retrograde membrane transport at these concentrations. Furthermore, in the very same sample of cells, lyso-PDMP did not elicit any Ca²⁺ response, in contrast to PDMP (Fig. 10,B). This observation agrees well with the ineffectiveness of lyso-PDMP regarding modulation of BFA-induced retrograde membrane transport (Fig. 4).

A possible interaction between PDMP and receptor-mediated [Ca²⁺]_i increases was subsequently studied using extracellular ATP to stimulate P₂ purinoceptors on the plasma membrane of HT29 cells (Nitschke et al., 1993; Zhang and Roomans, 1997). An ATP response could be evoked after an initial PDMP response (Fig. 10,A). However, occasionally a very large increase in [Ca²⁺]_i was observed after addition of 150 μM PDMP. Under these circumstances, the ATP response was completely abolished (Fig. 10,D). The extent of inhibition of the ATP response was dependent on the concentration of PDMP (Fig. 11,B). Interestingly, with the reversed order of addition, PDMP did still evoke a calcium response even after a very high ATP response (Fig. 10,E). Next we analyzed the PDMP response in terms of its components. In calcium-free medium, 100 μM PDMP elicited a calcium response (Fig. 10,C), which was very similar to that in calcium-containing medium (155 ± 21 nM; n = 7). In fact, at all concentrations of PDMP, the two responses overlapped (Fig. 11,A), indicating that the PDMP response can be attributed almost entirely to calcium release from intracellular stores, with a minor contribution of influx from the medium. A large part of the calcium release appeared to be associated with the ER compartment since (a) the response is effectively reduced after stimulation of the P₂ purinoceptor by ATP (Fig. 10,E), which releases calcium from the ER store (c.f. Salter and Hicks, 1995; Liu et al., 1996), and (b) a similar inhibition occurs after addition of thapsigargin (Fig. 10,G), a known blocker of the Ca²⁺-ATPase present on the ER (Lytton et al., 1991), which is followed by a PDMP (100 μM) response of 23 ± 4 nM (n = 11) in calcium-free medium. However, the observation that the PDMP response could not be fully blocked by either ATP or thapsigargin strongly suggests that PDMP released calcium from another intracellular pool in addition to the ER. Finally, a third component of the PDMP response was a striking inhibition of calcium influx, as induced by P₂ purinoceptor stimulation or thapsigargin treatment. As shown in Figs. 10,E and 11,B, PDMP blocked the sustained [Ca²⁺]_i increase (influx) component of the ATP response. Furthermore, a drastic inhibition of the influx component of the thapsigargin response occurred, as shown in Figs. 10,F and 11,C. In contrast to PDMP, none of the other calcium mobilizing compounds ATP or thapsigargin or the Ca²⁺ ionophore ionomycin were able to inhibit BFA-induced retrograde membrane transport from Golgi to ER in our studies in HT29 and NRK cells (Fig. 12).

Our results show that PDMP inhibits BFA-induced retrograde membrane flow from the Golgi apparatus to the ER, since the drug impaired the transfer of both lipid and protein components of Golgi membranes to the ER in cells treated with BFA. This reveals a novel action of PDMP, in addition to its known effect on membrane transport through the secretory pathway (Rosenwald et al., 1992). Interestingly, with respect to its mechanism(s) of action, PDMP neither directly counteracted the effect of BFA by interference with the drug's ability to induce release of β-COP from Golgi membranes (c.f. Scheel et al., 1997), nor exerted its effect by modulation of sphingolipid metabolism. Rather, the data strongly support the view that the PDMP-mediated impediment of membrane flow is accomplished via a specific modulation of calcium homeostasis.

Given the well-known effect of PDMP on sphingolipid metabolism, we investigated a potential relation with membrane transport. Rosenwald et al. (1992) have suggested that the inhibition by PDMP of anterograde membrane flow in the Golgi and from Golgi to the plasma membrane is correlated to down-regulation of SM biosynthesis. We do not favor this hypothesis as an explanation for the effect of PDMP on BFA-induced retrograde membrane flow because PDMP did not affect C₆-NBD-SM biosynthesis in HT29 G+ cells while still inhibiting BFA- induced retrograde membrane flow. PDMP is not likely to exert its effect through changes in the levels of GlcCer or Cer either. A rise in the cellular Cer level by exogenous addition of C₆-Cer did not mimic the effect of PDMP on BFA-induced retrograde membrane flow. Exogenous addition of C₆-GlcCer, which would compensate for the loss of cellular GlcCer, did not counteract the effect of PDMP. Furthermore, at concentrations below 50 μM, PDMP did not affect BFA-induced retrograde membrane flow but still inhibited GlcCer biosynthesis efficiently (data not shown). In conclusion, the inhibition of membrane flow does not appear to be correlated to fluctuations in sphingolipid levels.

MDR inhibitor studies were prompted by recent observations on the correlation between GlcCer biosynthesis and MDR (Lavie et al., 1996, 1997). Our own observations regarding the fast kinetics of C₆-NBD-PDMP efflux from HT29 cells and its specific modulation by the MRP inhibitor MK571 suggested involvement of MRP pumps in the action of PDMP on membrane transport. Although a specific effect of MK571 (in contrast to the P-gp inhibitors cyclosporin A and PSC833) on BFA-induced retrograde membrane transport was revealed, this experimental approach did not provide insight into the mechanism of action of PDMP. Nevertheless, a correlation with sphingolipid metabolism was again refuted, given the efficient inhibition of BFA-induced retrograde membrane flow by MK571 at a concentration of 50 μM, whereas the effect on GlcCer biosynthesis was marginal compared with PDMP.

More insight into the underlying mechanism of action of PDMP was obtained when modulation of calcium homeostasis by PDMP was investigated, inspired by results from Ivessa et al. (1995), who showed that redistribution of the Golgi complex into the ER is linked to calcium homeostasis in NRK and HeLa cells. An intracellular calcium response was indeed triggered by PDMP, in a similar concentration range as observed for the occurrence of inhibition of BFA-induced retrograde transport (both starting at 50 μM PDMP), while lyso-PDMP served as a negative control regarding both intracellular calcium responses and modulation of BFA-induced retrograde membrane transport. With respect to the low intracellular calcium response at 50 μM PDMP, it should be noted that measurements were performed at room temperature (to avoid compartmentalization of indo-1), which may give rise to an underestimation of the responsiveness at 37°C. Furthermore, as will be pointed out, the absolute size of the calcium response does not appear to be critical. Ivessa et al. (1995) reported that drugs affecting calcium equilibrium, such as thapsigargin and ionomycin, inhibited retrograde membrane transport in the ER/Golgi system in BFA-treated cells. In the present work, we have obtained evidence indicating that a more specific modulation of calcium homeostasis is involved. In our studies in HT29 and NRK cells, ATP, thapsigargin, and ionomycin did not inhibit BFA-induced retrograde transport, suggesting that membrane transport is not associated with calcium release from the ER pool but rather is linked to a specific calcium event. Our detailed analysis of the components of the calcium response to PDMP indicates that in addition to the ER, another PDMP-sensitive calcium pool exists. We have not yet characterized this pool, but it appears to be different from the recently described thapsigargin- and inositol-1,4,5-trisphosphate–insensitive calcium pool (Pizzo et al., 1997), since the latter is ionomycin sensitive. Thus, the size of the calcium response to PDMP, which is mostly determined by the ER release component, is not critical for membrane transport inhibition. Rather, the intracellular localization of the PDMP-sensitive, thapsigargin- and ATP-insensitive calcium pool is likely to be relevant. This pool may function in concert with the shutdown of calcium influx from the medium, as induced by PDMP, constituting the basis of the observed inhibition of BFA-induced membrane transport. A contribution of calcium influx inhibition to the effect of PDMP on membrane transport appears likely given the slow kinetics of reversibility of membrane flow inhibition (Fig. 3). In this respect, it was found that after a single administration of PDMP to cells calcium influx inhibition persisted for at least 2 h (data not shown). The persistent presence of a (C₆-NBD-)PDMP pool of about 13% (Fig. 7 B) correlates well with the persistent effect of PDMP on membrane flow and calcium influx inhibition.

Supporting evidence for an indirect mode of action of PDMP, i.e., not related to its interference with sphingolipid metabolism in the Golgi apparatus, includes: (a) PDMP acts very quickly, as observed when PDMP, added to cells 3 min after the addition of BFA, still inhibited BFA-induced retrograde membrane transport; only when the preincubation with BFA was extended to 10 min could PDMP not prevent merging of Golgi with ER (data not shown). These results are compatible with modulation of a fast acting Ca²⁺ signaling system within the cell. A prerequisite is its rapid uptake into the cell, which was shown to occur using a fluorescent derivative of PDMP (C₆-NBD-PDMP; Fig. 7,A). (b) C₆-NBD-PDMP was largely found in endocytic compartments (c.f. Rosenwald and Pagano, 1994), while the fraction residing in the Golgi was too small to be discerned by confocal scanning laser microscopy. It should be noted that the fluorescent analogue of PDMP was as effective as the native drug in causing membrane flow inhibition, ruling out possible artifacts of fluorescent derivatization. Chase experiments revealed that two pools of C₆-NBD-PDMP could be discerned after 30 min incubation with the drug, one of which was rapidly excreted, the other being very stable (Fig. 7 B). Also, this stable pool was residing primarily in endocytic compartments since it was almost completely quenched by sodium dithionite. These results show that C₆-NBD-PDMP could not be detected in the Golgi apparatus, and thus indicate the unlikeliness of a direct effect of PDMP on Golgi membrane structure or Golgi membrane proteins involved in transport. A general disrupting effect of PDMP on membrane structure can also be excluded given the transient nature of the intracellular calcium response to PDMP, since a sustained influx would be expected in case of membrane disruption.

All our experiments concerning the effects of PDMP on retrograde membrane flow from Golgi to ER were performed in the context of BFA treatment. Therefore, no conclusions can be drawn with respect to the effect of PDMP on retrograde membrane flow under physiological circumstances, i.e., without BFA, a pathway that functions in the delivery of Golgi- and plasma membrane–localized proteins to the ER (Johannes et al., 1997; Cole et al., 1998). In this respect, PDMP has recently been shown to shift the steady-state distribution of infectious bronchitis virus M protein and ERGIC-53 from the Golgi to the ER in BHK cells under conditions not involving BFA (Maceyka and Machamer, 1997). However, the Golgi stack markers ManII and giantin were not redistributed, indicating that this effect of PDMP was related to cycling of specific proteins rather than to overall membrane flow. This effect of PDMP was attributed to Cer accumulation, although it was neither completely reversed by metabolic inhibitors of Cer formation nor mimicked by exogenous C₆-Cer addition.

We propose two models for the mechanism by which PDMP and PDMP-induced [Ca²⁺]_i changes could inhibit BFA-induced retrograde membrane flow. (a) PDMP acts independently of BFA in a process (e.g., the actual fusion process) involved in the merging of ER and Golgi membranes during the direct, vesicle budding-uncoupled fusion reaction. (b) PDMP specifically antagonizes one of the actions of BFA. Clearly, PDMP does not act by interference with BFA-induced release of β-COP (Fig. 5). This is consistent with observations from Ivessa et al. (1995), who showed that calcium mobilizing agents that prevent BFA-induced retrograde membrane flow did not interfere with β-COP release. This shows that although BFA-induced dissociation of coatomer from membranes is required for redistribution of Golgi membranes into the ER (Scheel et al., 1997), it is not sufficient. In this respect, it has recently been shown in permeabilized cells that in the absence of NAD⁺, BFA can dissociate coatomer from the Golgi complex without affecting the structure of the organelle (Mironov et al., 1997). NAD⁺- and BFA-dependent ADP-ribosylation of BARS-50 was shown to play a role in the Golgi disassembling activity of BFA. On the other hand, it cannot be excluded that PDMP (partly) exerts its effect by detoxification of BFA, as has been shown to occur for forskolin (Nickel et al., 1996), an activator of adenylyl cyclase, which also inhibits BFA-induced retrograde membrane flow (Lippincott et al., 1991). However, we consider this possibility unlikely, given the fast kinetics of PDMP action (no preincubation with PDMP is required to obtain inhibition of BFA-induced retrograde membrane flow; even addition of PDMP 3 min after BFA is effective), compared with the kinetics of secretion of BFA conjugates.

In conclusion, PDMP inhibition of BFA-induced retrograde membrane flow in the ER/Golgi system is dissociated from its well-known effect on glycosphingolipid metabolism and does not involve direct interference with the integrity of Golgi membranes. Rather, it appears to be related to specific modulation of calcium homeostasis, a novel action of the drug, and mimicked by a specific antagonist of the MRP pump. Thus, PDMP is a novel tool to interfere with BFA-induced retrograde Golgi to ER membrane transport and provides new insights into the physiological mechanisms underlying this process. Future work will aim at further identification of PDMP-sensitive calcium pools and the assessment of the role of these pools and calcium influx inhibition in membrane transport between Golgi and ER. Furthermore, we will investigate whether the effects of PDMP on BFA-induced retrograde membrane transport are a reflection of inhibition of retrograde transport under physiological circumstances (e.g., without BFA).

The authors thank Jacqueline Plass and Coen Wiegman for their contribution to the initial observations on PDMP inhibition of BFA-induced retrograde membrane transport, Han Roelofsen for his help with confocal scanning laser microscopy, and Peter van der Syde for the skillful technical assistance in photo-editing/printing.

BFA

brefeldin A

Cer

ceramide

C₆-NBD

6-(N-[7-nitro-2,1,3-benzoxadiazol-4-yl]amino)hexanoyl or hexanoic acid

GlcCer

glucosylceramide

ManII

mannosidase II

MDR

multidrug resistance

MRP

multidrug resistance protein

NRK

normal rat kidney

PDMP

d,l-threo-1-phenyl-2-decanoylamino-3-morpholino-1-propanol

P-gp

P-glycoprotein

SM

sphingomyelin