Inhibition of transferrin recycling and endosome tubulation by phospholipase A2 antagonists.

We report here that a broad spectrum of phospholipase A(2) (PLA(2)) antagonists produce a concentration-dependent, differential block in the endocytic recycling pathway of transferrin (Tf) and Tf receptors (TfRs) but have no acute affect on Tf uptake from the cell surface. At low concentrations of antagonists (approximately 1 microm), Tf and TfR accumulated in centrally located recycling endosomes, whereas at higher concentrations (approximately 10 microm), Tf-TfR accumulated in peripheral sorting endosomes. Several independent lines of evidence suggest that this inhibition of recycling may result from the inhibition of tubule formation. First, BFA-stimulated endosome tubule formation was similarly inhibited by PLA(2) antagonists. Second, endocytosed tracers were found in larger spherical endosomes in the presence of PLA(2) antagonists. And third, endosome tubule formation in a cell-free, cytosol-dependent reconstitution system was equally sensitive PLA(2) antagonists. These results are consistent with the conclusion that endosome membrane tubules are formed by the action of a cytoplasmic PLA(2) and that PLA(2)-dependent tubules are involved in intracellular recycling of Tf and TfR. When taken together with previous studies on the Golgi complex, these results also indicate that an intracellular PLA(2) activity provides a novel molecular mechanism for inducing tubule formation from multiple organelles.


From the Department of Molecular Biology and Genetics, Cornell University, Ithaca, New York 14853
We report here that a broad spectrum of phospholipase A 2 (PLA 2 ) antagonists produce a concentrationdependent, differential block in the endocytic recycling pathway of transferrin (Tf) and Tf receptors (TfRs) but have no acute affect on Tf uptake from the cell surface. At low concentrations of antagonists (ϳ1 M), Tf and TfR accumulated in centrally located recycling endosomes, whereas at higher concentrations (ϳ10 M), Tf-TfR accumulated in peripheral sorting endosomes. Several independent lines of evidence suggest that this inhibition of recycling may result from the inhibition of tubule formation. First, BFA-stimulated endosome tubule formation was similarly inhibited by PLA 2 antagonists. Second, endocytosed tracers were found in larger spherical endosomes in the presence of PLA 2 antagonists. And third, endosome tubule formation in a cell-free, cytosoldependent reconstitution system was equally sensitive PLA 2 antagonists. These results are consistent with the conclusion that endosome membrane tubules are formed by the action of a cytoplasmic PLA 2 and that PLA 2 -dependent tubules are involved in intracellular recycling of Tf and TfR. When taken together with previous studies on the Golgi complex, these results also indicate that an intracellular PLA 2 activity provides a novel molecular mechanism for inducing tubule formation from multiple organelles.
Endocytic compartments serve important sorting functions for the segregation and delivery of membrane-bound receptors, their cognate ligands, and soluble cargo to specific intracellular destinations (1,2). Early light and electron microscopic studies revealed that some membrane receptors become concentrated into endosomal tubules, 50 -70 nm in diameter and extending from the main vesicular/vacuolar body of endosomes, before recycling back to the plasma membrane (3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16). These results suggested that tubules may serve to segregate membrane and receptors away from the bulk fluid volume of the vacuolar part of the endosome. From other studies, Maxfield and colleagues (2,17,18) proposed that recycling of endocytosed TfRs 1 occurs in part by the iterative formation and budding of narrow 50 -70-nm tubules from endocytic compartments. These studies supported the idea that tubule formation provides a mechanism for the separation of membrane from soluble luminal contents based simply on differences in membrane surface-tovolume ratios (19,20).
In addition to membrane tubules, evidence has shown that cytoplasmic coat proteins, such as clathrin, COPI coatomers, and the Vps5-Vps17 complex in yeast cells, are also involved in mediating recycling from endosomal and lysosomal compartments by forming coated vesicles (15,16,(21)(22)(23)(24)(25)(26)(27)(28). Moreover, members of the ADP-ribosylation factor family of GTP-binding proteins appear to regulate endocytic trafficking, perhaps by mediating coat protein binding (29 -34). Thus, in a manner analogous to trafficking in the secretory pathway, various types of ADP-ribosylation factor-regulated coated vesicles may also package receptors and membrane for recycling in the endocytic pathway. As with COPI and clathrin-coated vesicles of the Golgi complex and trans-Golgi network (TGN), respectively, binding of coat proteins to endosomal membranes is sensitive to brefeldin A (BFA) (15,23,25), a potent inhibitor of trafficking through the secretory pathway (35). As a result of BFA's action, some endosomes are stimulated to form extensive membrane tubules, as do Golgi and TGN membranes (36 -41). However, unlike its effects on the secretory pathway, BFA does not significantly inhibit endocytic uptake or trafficking through endocytic compartments (37,41,42).
Our understanding of the exact roles that tubules and vesicles play in endocytic recycling is incomplete because, as yet, no trafficking intermediates have been definitively characterized. One possibility is that endosome tubules provide a mechanism for bulk segregation of membrane (including receptors) from fluid content, whereas coated vesicles provide additional specificity for receptor packaging and membrane fission (1,2). Tubules and vesicles could work in concert with coated vesicles budding from the tips of tubules, or they could work independently (3,15,37,39).
The mechanisms of endosome tubule formation are not known; however, they are probably different from that of coated vesicle formation because tubulation is enhanced in the presence of BFA, i.e. when coat proteins are prevented from binding. By way of comparison, recent studies on Golgi membranes have strongly suggested that tubulation, both in vivo and in a cell-free reconstitution system, requires the action of a cytosolic Ca 2ϩ -independent PLA 2 activity (43)(44)(45). Moreover, we found that constitutive and BFA-stimulated retrograde trafficking from the Golgi complex to the endoplasmic reticulum was strongly inhibited by a variety of compounds that antagonize cytoplasmic Ca 2ϩ -independent PLA 2 enzymes (43,46). In vitro studies suggest that the PLA 2 acts directly on the membranes to induce tubule formation (44,45). These results raise the possibility that endosomes and Golgi membranes utilize the same mechanisms to induce tubules, which are involved in different membrane trafficking events. Here we provide evidence that tubulation of endosome membranes, both in vivo and in an in vitro cell-free reconstitution system, was potently inhibited by a variety of PLA 2 antagonists, which also significantly inhibited recycling of Tf-TfR complexes from various endosomal compartments in vivo.
Rat clone 9 hepatocytes and HeLa cells were grown in modified Eagle's medium (MEM) with 10% Nu-Serum or Calf Serum-Supplemented and 1% penicillin/streptomycin. Mouse RAW 264.7 macrophages were grown in Dulbecco's modified Eagle's medium with 10% fetal bovine serum and 1% penicillin/streptomycin. All cells were maintained at 37°C in a humidified atmosphere of 95% air, 5% CO 2 .

Binding, Endocytosis, and Trafficking of 125 I-Tf, Fluorescent Tf, and Horseradish Peroxidase (HRP)
HeLa cells were grown for 2 days to ϳ70 -80% confluence in 12-well plates for 125 I-Tf binding, uptake, and recycling experiments or on glass coverslips for fluorescent Tf uptake and recycling experiments. For all binding, uptake, and recycling experiments, cells were first rinsed three times in MEM (without serum) and incubated in the same for 30 min at 37°C. Also, unless otherwise indicated, MEM did not contain serum, which inhibits the effects of PLA 2 inhibitors. The subsequent procedures for assaying binding, uptake, and recycling described below were modified from previous studies (48).
Biochemical Determination of 125 Tf Trafficking-To measure the effects of PLA 2 inhibitors on the relative amounts of TfRs found on the cell surface, cells were incubated in the presence or absence of 10 M BEL in MEM for 45 min at 37°C in a CO 2 incubator. Plates were then transferred to a cold room and placed on ice, and cells were rinsed three times with ice-cold MEM containing a 1:10 dilution of the normal sodium bicarbonate (MEM1/10). All subsequent surface binding procedures were done at 4°C. Cells were briefly rinsed with an acid wash (500 mM NaCl, 0.2 N acetic acid, pH 2) to remove any surface-bound Tf and then rinsed three times with MEM1/10, which restored the pH to 7.4. Medium was removed and then replaced with 0.5 ml of MEM1/10 containing a saturating amount of 125 I-Tf (467 ng of Tf/well ϭ 0.208 Ci/well). Cells were incubated for 2 h with gentle shaking. The medium was removed, and cells were washed six times with MEM1/10 (by which time the last wash had only background levels of 125 I). Cells were extracted by the addition of 1 N NaOH and gentle shaking for 30 min. The extract was collected and counted in a ␥ counter.
To measure the effects of PLA 2 inhibitors on the endocytic uptake of Tf from the cell surface, 12-well plates were placed on ice, rinsed with ice-cold acid wash and then washed three times with ice-cold MEM. The medium was removed, and cells were incubated in the presence or absence of 10 M BEL in ice-cold MEM for 10 min on ice. The medium was replaced with 0.5 ml of MEM plus 10 M BEL (37°C), as appropriate, which also contained 125 I-Tf (amounts described above for sur-face binding). Cells were allowed to internalize the bound 125 I-Tf for 10 min at 37°C. Following removal of the medium, cells were rinsed with ice-cold acid wash, washed six times with ice-cold MEM, and then extracted with 1 N NaOH as above.
To measure the effect of PLA 2 inhibitors on the recycling of Tf from internal compartments back to the cell surface, cells were incubated with MEM containing 125 I-Tf (0.17 Ci/ml) for 45 min at 37°C in a CO 2 incubator. The medium was removed, and cells were washed six times in ice-cold MEM (while on ice) and then incubated in the presence or absence of 10 M BEL in MEM for 5 min at 4°C. The medium was removed; 0.5 ml of 37°C efflux medium (MEM containing 0.1 mM desferrioxamine, 3 g/ml Tf) with or without 10 M BEL as appropriate was added; and cells were further incubated for various periods of time at 37°C. At the end of each time period, medium was collected, and intracellular 125 I-Tf was harvested by extracting cells in 1 N NaOH. The amounts of 125 I-Tf in both the medium and intracellular samples were determined.
Morphological Analysis of Fluorescent Tf Trafficking-To visualize the endocytosis and trafficking of fluorescent Tf, cells were treated essentially as above for biochemical labeling experiments. In some cases, various specific pulse-chase protocols were conducted, which will be described below. Cells were incubated in FITC-Tf (20 g/ml) and/or Rhod-Tf (30 g/ml) to label various endocytic compartments. In many cases after labeling with fluorescent Tf, cells were fixed and processed for immunofluorescence to localize TfR as described (47) or the trans-Golgi/TGN membrane protein, galactosyltransferase (using polyclonal antibodies at a 1:600 dilution). In most cases, fluorescently labeled cells were viewed with standard epifluorescence optics; however, in a few specific instances, where noted, cells were viewed by scanning confocal microscopy with a Bio-Rad 600 confocal microscope.
Endocytosis of soluble HRP and its visualization by diaminobenzidine cytochemistry for electron microscopy were performed as previously described (47).

Preparation of Endosomes
A suspension of 12-nm colloidal gold particles (Au 12 ) was prepared as described (49), stabilized with bovine serum albumin (BSA), resuspended in MEM (for clone 9 cells) or Dulbecco's modified Eagle's medium (for RAW 264.7 cells), and stored at 4°C until used. Monolayers of clone 9 hepatocytes or RAW 264.7 macrophages were grown to confluence in 4 -7 220-mm 2 plastic tissue culture plates. To label endosomes with Au 12 -BSA, cells were briefly rinsed once with MEM at 37°C and incubated with prewarmed Au 12 -BSA suspension (in MEM; 75 ml/dish) for 15 min (clone 9) or 10 min (RAW 264.7) at 37°C. Labeling was terminated by rapidly pouring off the Au 12 -BSA suspension and washing cells three times with ice-cold MEM; plates were kept on ice as the cells were harvested with a rubber policeman. All subsequent steps in the preparation procedure were carried out at 4°C. The cells were harvested in a table top centrifuge (IEC) for 5 min, and the supernatant was discarded. The pellet of packed cells was thoroughly resuspended in homogenization buffer (250 mM sucrose, 10 mM Tris-HCl, 1 mM EDTA, pH 7.4) and centrifuged as above. This step was repeated once, and the supernatant was discarded. The cells were resuspended in homogenization buffer to yield a 20% (v/v) cell suspension and homogenized by passage (2-4 times) through a Balch/Rothman ball bearing homogenizer (50). The homogenate (ϳ8 ml) was centrifuged as above to yield a postnuclear supernatant. This centrifugation step was repeated if the postnuclear supernatant contained any nuclei as detected by phasecontrast microscopy. The postnuclear supernatant (4 ml) was layered on top of a sucrose step gradient in Beckman SW28.1 centrifuge tubes containing the following concentrations of sucrose at each step (3.4 ml/step): 0.5, 1.0, 1.5, and 2 M sucrose. Each sucrose solution was buffered with 1 mM Tris, 1 mM EDTA, pH 7.4. Additionally, the 0.5 M sucrose step contained 500 mM KCl, which served to wash endosomes as they moved through the gradient and lowered the background level of in vitro tubulation. The gradients were centrifuged at 113,000 ϫ g for 2 h in a Beckman SW28.1 rotor. Material at the 1.0/1.5 M sucrose interface contained the highest concentration of Au 12 -BSA-loaded endosomes, was collected with a Pasteur pipette, and was either used immediately or slowly frozen and stored at Ϫ80°C. Tubulation-competent endosomes had a useful shelf life of ϳ2 days at Ϫ80°C.

Preparation of Bovine Brain Cytosol
Bovine brain cytosol used in these in vitro studies was prepared as previously described (51). Two fractions of brain cytosol from this preparation were used in these studies. The first, called "crude cytosol" was the supernatant obtained immediately after centrifuging the brain ho-mogenate at 95,500 ϫ g for 2 h. The second fraction, called "bovine brain cytosol" or BBC, was prepared from the above supernatant by precipitation with 60% ammonium sulfate, reconstitution in buffer, and dialysis as described (51). Unless otherwise indicated, all experiments used this BBC preparation. Typically, 1.5 kg of brains yielded over 600 ml of cytosol with a protein concentration of ϳ25 mg/ml.

In Vitro Tubulation Assay
The tubulation of isolated endosomes was visualized by a whole mount EM-negative stain assay as used to document the tubulation of Golgi membranes (52). In a typical assay, 12.5 l of bovine brain cytosol was mixed with 87.5 l of tubulation buffer (final concentrations in the complete assay mixture: 50 mM KCl, 1 mM MgCl 2 , 25 mM Tris-HCl, 10 mM HEPES, pH 7.4) containing various combinations of other reagents such as ATP and/or GTP. This buffer mixture and either fresh or thawed endosome-enriched fractions were equilibrated to 37°C for 15 min. Tubulation was initiated by slowly adding an equal volume (20 l) of the cytosol/buffer mixture to the endosome-enriched fractions, followed by incubation at 37°C for various periods of time. Drops of the endosome suspension were placed on Formvar-and carbon-coated EM grids for 15 min at room temperature, and grids were negatively stained with 2.0% phosphotungstic acid, pH 7.2, and viewed by electron microscopy (52). In an alternate method, drops of the endosome suspension were placed on Formvar-and carbon-coated EM grids for 5 min at room temperature, and glutaraldehyde (from 25% stock solution) was added directly to the drops (to a final concentration of 0.1%) and allowed to fix the organelles for 5 min. The fixed organelles were then stained with 2% phosphotungstic acid as described above.
Endosome tubulation was quantified by determining the percentage of gold-loaded organelles that displayed at least one membranous tubule or that were completely tubular. For each experimental condition, at least 150 endosomes were counted from triplicate samples. A tubule was defined as a membranous extension, 50 -70 nm in diameter and at least twice as long, that was continuous with the lipid bilayer of the endosome. For practical reasons, however, this assay can only provide a semiquantitative estimation of the tubulation activity. Ideally, we would like to determine the number of tubules/endosome. Unfortunately, determining this number slowed analyses of the experiments to unacceptable levels. Therefore, an endosome with one tubule was equal to an endosome with multiple tubules. From our qualitative observations, however, we noticed a clear positive correlation between the number of endosomes with tubules and the number of tubules/endosome.

PLA 2 Antagonists Inhibit Tf-TfR Recycling at Multiple
Steps-Based on our previous studies, which revealed that PLA 2 antagonists were potent inhibitors of BFA-stimulated, tubule-mediated retrograde trafficking from the Golgi complex (43), we asked here if these same antagonists affected putative tubule-mediated trafficking steps in the endocytic recycling pathway. To investigate this question, the intracellular trafficking of TfR and its ligand, Tf, were followed using various pulse-chase experiments and biochemical and morphological assays. We found that incubation of cells with the irreversible inhibitor of Ca 2ϩ -independent PLA 2 enzymes, BEL, had little effect on a brief pulse (10 min) of 125 I-Tf uptake (Fig. 1A). However, when cells were first pretreated with BEL for 45 min at 37°C and then assayed for the presence of 125 I-Tf binding sites, we found a significant decrease on BEL-treated cells (Fig.  1A). These results suggest that although BEL does not prevent uptake from the cell surface, it does prevent recycling of TfR from internal compartments. To directly test this idea, cells were incubated with 125 I-Tf for 45 min to allow its internalization and delivery to endosomal compartments, washed free of uninternalized ligand, and then chased in 125 I-Tf-free medium in the presence or absence of BEL for various periods of time. The results showed that the addition of 10 M BEL substantially inhibited the recycling of 125 I-Tf to the extracellular medium (Fig. 1B). Using a fixed chase time point (60 min), we found that increasing concentrations of BEL produced a linear and saturable inhibition of 125 I-Tf recycling with an IC 50 of ϳ4 M (data not shown).
To determine where in the recycling pathway Tf trafficking was inhibited, cells were pulse-labeled with FITC-Tf for 45 min and then chased (in medium without FITC-Tf) in the presence or absence of various PLA 2 inhibitors. Cells were then prepared for immunofluorescence labeling of the TfR. In control cells with no chase, Tf and the TfR were found to almost entirely colocalize in numerous punctate vesicles located throughout the cytoplasm (Fig. 2, A and B). When cells were pulse-labeled and chased in medium without antagonists for 60 min, FITC-Tf staining was significantly reduced, consistent with its recycling and loss into the extracellular medium (Fig. 2, C and D). However, quite different results were obtained when BEL was included in the chase medium.
In the presence of 10 M BEL, FITC-Tf was not lost from cells but was instead found to colocalize with TfRs in punctate vesicles located throughout the cytoplasm (Fig. 2, E and F), a pattern similar to that of untreated cells. Interestingly, in 1 M BEL, both FITC-Tf and TfR were found to colocalize in a more compact, juxtanuclear staining pattern (Fig. 2, G and  H). Similar dose-dependent results were obtained with a structurally different PLA 2 inhibitors, ONO-RS-082, N-(pamylcinnamoyl)anthranilic acid, and aristolochic acid (data not shown). These results confirm the biochemical studies above showing that PLA 2 antagonists inhibit Tf-TfR recy- cling and further suggest that BEL produced a concentrationdependent block in recycling at two different steps.
Previous studies have established that following endocytosis, Tf-TfR complexes move sequentially through peripherally located early sorting endosomes, then to a centrally located tubular recycling endocytic compartment, and finally back to the cell surface for release into the medium (for a review see Ref. 2). Thus, our above results suggest that 10 M BEL may block trafficking of Tf-TfR out of peripheral early endosomes and that 1 M BEL allows transport to, but not out of, centrally located recycling endosomes. To examine this issue further, we took advantage of the fact that ONO-RS-082 is a reversible inhibitor of Ca 2ϩ -independent PLA 2 enzymes (53) and displayed the same concentration-dependent inhibition of Tf recycling as BEL (data not shown). As with BEL, when cells were pulselabeled with FITC-Tf and then chased in medium without FITC-Tf but with high (10 M) concentrations of ONO-RS-082, FITC-Tf remained located along with TfRs in vesicles throughout the cytoplasm (Fig. 3, A and B). When similarly treated cells were then washed free of ONO-RS-082 and incubated for an additional 15 min, both FITC-Tf and TfR moved to the centrally located endocytic compartment (Fig. 3, C and D). This central compartment was indeed the recycling compartment, because incubation in ONO-RS-082-free medium for an additional 45 min (60 min total) resulted in a significant loss of FITC-Tf from the cells (Fig. 3, E and F). Finally, the TfRs that recycled to the cell surface following recovery from ONO-RS-082 were competent to re-endocytose another round of FITC-Tf added to the medium (Fig. 3, G and H).
Although our results are consistent with a concentration-dependent block at two stages of the Tf/TfR recycling pathway, the PLA 2 antagonists might also be causing a mistargeting of receptor-ligand complexes to or from other compartments. For example, the compact juxtanuclear compartment in which FITC-Tf and TfRs accumulate in cells treated with 1 M ONO-RS-082, might instead be elements of the TGN, given the close association between central recycling endosomes and the TGN (2). To investigate this further, cells were incubated with FTIC-Tf for 45 min in the presence of 1 M ONO-RS-082 and then chased in ligand-free medium for an additional 30 min in the presence of ONO-RS-082. Cells were fixed; processed for immunofluorescence to localize ␤1,4-galactosyltransferase, an enzyme found in both trans-Golgi cisternae and the TGN of HeLa cells (54,55); and then visualized by confocal microscopy. The results from both individual and merged images show that FITC-Tf and galactosyltransferase labeled separate compartments (Fig. 4).
Another possible explanation for the accumulation of FITC-Tf and TfR within the central juxtanuclear compartment (when treated with 1 M BEL or ONO-RS-082) is that peripheral sorting endosomes redistribute to the center of the cell, rather than receptor-ligand complexes exiting the peripheral sorting endosomes. To distinguish between these possibilities, cells were pulse-labeled with Rhod-Tf for 60 min, chased in ligand-free medium for 15 min, and then incubated with a second pulse-chase of FITC-Tf (as for Rhod-Tf), all in the presence of 1 M ONO-RS-082. The results showed that there was significant overlap in staining of the two endocytic markers within vesicles of the central compartment (Fig. 5). Because FITC-Tf was ultimately delivered to Rhod-Tf-containing endosomes, we conclude that 1 M ONO-RS-082 did not affect the function or distribution of early sorting endosomes, through which FITC-Tf must have entered and exited on its way to the central sorting endosomes.
One possible mechanism by which PLA 2 antagonists might alter the trafficking of Tf-TfR is by changing the pH of acidified endosomal compartments. However, we found no difference in the ability of control or treated (10 M ONO-RS-082 for 30 min at 37°C) cells to take up the pH-sensitive dye acridine orange into endosomal and lysosomal compartment (data not shown).
PLA 2 Antagonists Inhibit BFA-stimulated Endosome Tubule Formation-Trafficking from the central recycling compartment has been proposed to involve the formation of thin membrane tubules, whose presence is enhanced by BFA. Because we have previously shown that BFA-induced Golgi tubulation was inhibited by PLA 2 antagonists, we asked if BFA-stimulated tubulation of endosomes was similarly affected. Cells were first incubated with FITC-Tf for 45 min to label all endocytic compartments and then treated with BFA in the presence or absence of BEL. In both untreated cells and cells treated with BEL alone (10 M for 20 min), FITC-Tf and TfR were primarily located in punctate vesicles throughout the cytoplasm (Fig. 6, A-D). Treatment of cells with BFA alone for 15 min resulted in the extensive formation of membrane tubules that were labeled with both FITC-Tf and TfR (Fig. 6, E and F). However, pretreatment of cells for just 5 min with BEL (10 M) before a 15-min incubation in BFA significantly inhibited endosome tubule formation (Fig. 6, G and H). This inhibition was confirmed at the EM level in cells that had been incubated with soluble HRP to identify endosomes. In control cells, HRP-labeled endosomes were generally spherical, ϳ200 -400 nm in diameter (Fig. 7A). Incubation of cells with BFA for 10 min resulted in the formation of numerous tubular endosomes, ϳ50 -70 nm in diameter and up to several m in length (Fig.  7B). Pretreatment of cells with ONO-RS-082 for 5 min inhibited this BFA-stimulated tubulation (Fig. 7C).
A broad spectrum of PLA 2 antagonists were tested for their ability to inhibit BFA-stimulated tubulation in cells whose endosomes were labeled by prior uptake with FITC-Tf and subsequently by immunofluorescence localization of the TfR. We found that all PLA 2 antagonists tested were highly active inhibitors of BFA-stimulated tubulation, with IC 50 values in the low micromolar range (Table I). To more directly examine the ability of low concentrations of PLA 2 antagonists to inhibit tubulation of the recycling compartment, cells were incubated with FITC-Tf in the continuous presence of 1 M ONO-RS-082 for 45 min. In these cells, FITC-Tf was primarily located in the central recycling compartment (Fig. 2, G and H). Similarly treated cells were then either kept in 1 M ONO-RS-082 or washed free of the antagonist, BFA was immediately added, and the cells were further incubated for various periods of time (Fig. 8). In cells washed free of ONO-RS-082, BFA stimulated tubule formation within 5 min until, by 15 min, ϳ57% of the cells exhibited tubules emanating from the central recycling compartment. In contrast, cells kept in 1 M ONO-RS-082 did not exhibit BFAstimulated tubulation.
Formation of Endosome Tubules in Vitro-To complement these in vivo studies, we have developed a cell-free system that reconstitutes tubule formation from isolated endosomes. Endosomes were unambiguously identified in subcellular fractions by first incubating cultured clone 9 hepatocytes with Au 12 -BSA gold particles for 15 min at 37°C, to allow fluid phase particle uptake into early and late endosomes prior to cell homogenization (39). Cells were then homogenized, an endosome-enriched fraction was obtained, and endosomes containing Au 12 -BSA were visualized by EM-negative staining (arrowheads in Fig. 9).
Endosomes incubated with an organelle-free extract of bovine brain cytosol (BBC) alone (Fig. 9D), or BSA plus ATP (0.1 mM) (Fig. 9E), at 37°C for 30 min were primarily round vesicles and vacuoles, 0.25-1.5 m in diameter (average of 0.45 m; 50 random measurements), and they rarely exhibited tubular extensions. In contrast, incubation with BBC (1.5 mg/ml) and ATP (0.1 mM) induced the dramatic formation of extensive membrane tubules (Fig. 9, F-H). These tubules were generally uniform in diameter (average of 53 nm; 32 random measure-  (Fig. 9, F and H), which were morphologically similar to those seen extending from endosomes in vivo (Fig. 9, A-C) (in the latter case, endosomes that had been labeled by endocytic uptake of soluble HRP). Also, endosomes fixed with glutaraldehyde after incubation but before negative staining also exhibited a strict dependence on cytosol for tubulation (data not shown). Quantitation of these results showed that incubation with BBC or ATP (0.1 mM) alone had little effect but that incubation with both resulted in a 5-10-fold increase in the percentage of endosomes with at least one tubule (Fig. 10A). In these experiments, ϳ50 -75% of the endosomes were induced to form at least one tubule. Heating cytosol to 100°C for 30 min before assaying greatly reduced the level of tubulation activity (Fig.  10A), and incubation with BBC and GTP or GTP␥S (0.1 mM) did not support tubulation (data not shown). Also, because the tubulation mixture contained EDTA (0.5 mM), and the addition of excess Ca 2ϩ had no effect on tubulation (data not shown), we conclude that Ca 2ϩ is not required for this activity. A time course of tubulation using the complete mixture of BBC (1.5 mg/ml) and ATP (0.1 mM) at 37°C showed that tubulation was fairly rapid, with a t ϭ 4 min (Fig. 10B). In addition, a cytosol concentration curve determined that tubulation was a saturable process at ϳ1.5 mg/ml, with half-maximal activity at 0.15 mg/ml (Fig. 10C).
The tubulation activity could be followed through the steps used to prepare the BBC fraction from crude cytosol and characterized following various treatments (Fig. 10D). The tubulation activity in BBC was found to be insensitive to a number of agents including N-ethylmaleimide, N-ethylmaleimide in the presence of dithiothreitol, or nocodazole, a compound the depolymerizes microtubules (data not shown). Also, treating endosomes with BFA (10 g/ml) alone or with BBC seemed to have little effect on the tubulation activity. Finally, this tubulation activity appeared not be a general perturbant of membrane structure because the active BBC preparation used in these experiments had no effect on isolated lysosomes. 2 To investigate the origin of the membrane that forms tubules, we determined the average diameter of the main bodies of randomly selected nontubulated (50 selected) and tubulated (25 selected) endosomes. These measurements revealed that tubulated endosomes have, on average, a smaller diameter across their vacuolar body (0.31 m) than their nontubulated counterparts (0.45 m). The decrease in body diameter is consistent with the idea that the membrane for tubules originates from the main body of the endosomes. In tubulated endosomes, Au 12 -BSA particles were most commonly found in the tubular extensions (Fig. 9, F-H), suggesting some weak interaction with the membrane. However, with regard to tubulation, this interaction is irrelevant, since endosome tubules formed without any gold particles being present within the lumen of the tubule (Fig. 9F).
PLA 2 Antagonists Inhibit in Vitro Endosome Tubulation-Cytosol-dependent endosome tubulation was also examined for 2 A. van Vante and W. J. Brown, unpublished data.

FIG. 9. Morphological observations of endosome tubules in vivo and in vitro.
A-C, rat clone 9 hepatocytes were incubated with HRP (15 min at 37°C) and processed for DAB cytochemistry and thin section EM. HRP-labeled endosomes display membrane tubules, 50 -60 nm in diameter and varying lengths. D-H, endosomes isolated from clone 9 cells that had been allowed to endocytose Au 12 -BSA particles were prepared and visualized by electron microscopy negative staining. Endosomes were identified by the presence of endocytosed Au 12 -BSA particles (arrowheads in all panels). Due to the sparse surface density of endosomes on the electron microscopy grids, representative endosomes from various experimental conditions are shown. Endosomes incubated in the presence of bovine brain cytosol alone (D) or BSA plus ATP and GTP (E) were primarily vesicular or vacuolar with few tubular extensions. Incubation of endosomes with bovine brain cytosol plus ATP, however, resulted in the formation of multiple, short tubules (F) or few long tubules (G and H). In F, an endosome with tubules of varying lengths is shown (arrows), and all have the same morphology; as tubules protrude from the vacuolar membrane, they taper into a narrow shaft, 50 -70 nm in diameter, and terminate in a bulbous end. The tubules maintain this morphology after reaching lengths of Ͼ0.5 m, which is similar to that of tubules seen in vivo (A-C). Standard incubation conditions were 30 min at 37°C. Bars, 0.25 m. sensitivity to PLA 2 antagonists and to see if it was active on endosomes from another source. For these experiments, endosomes were prepared from RAW 264.7 macrophages, and BBC was incubated with a PLA 2 antagonist before the addition to isolated endosomes. The results showed that the structurally distinct compounds BEL and ONO-RS-082 significantly inhibited cytosol-dependent endosome tubulation in vitro even at 1 M (Fig. 11), concentrations similar to those that inhibited endosome tubulation and receptor recycling in vivo. In other studies, we have found that this BBC preparation contains both BEL-and ONO-RS-082-sensitive and -insensitive PLA 2 activities (43,44). As with the Golgi complex (45), we found that a nonspecific PLA 2 , snake venom PLA 2 , did not increase endosome tubulation above background levels. Quantitatively similar inhibition by PLA 2 antagonists was obtained using endosomes from clone 9 hepatocytes (data not shown).

DISCUSSION
This paper provides evidence for the role of a cytoplasmic Ca 2ϩ -independent PLA 2 in the recycling of Tf-TfR from endosomal compartments to the cell surface and the formation of endosome tubules both in vivo and in vitro. These results are similar to those of recent studies that demonstrated a nearly identical sensitivity of Golgi membrane tubulation and retrograde trafficking to PLA 2 antagonists (43,45,46). The fact that a broad spectrum of PLA 2 antagonists were found to inhibit trafficking events, suspected or known to involve membrane tubules, from both endosomes and the Golgi complex, suggests that a cytoplasmic PLA 2 activity plays a general role in tubule formation from organellar membranes.
We found that higher concentrations of PLA 2 antagonists (ϳ10 M) inhibited trafficking from peripheral sorting endosomes to central recycling endosomes, and lower concentrations (1 M) inhibited trafficking from central recycling endosomes to the cell surface. At either concentration, it appeared that transport out of an endocytic compartment rather than fusion with a target was the inhibited step. Although not shown directly, several lines of evidence support this idea. First, uptake and trafficking of endocytic tracers in the presence of low or high concentrations of BEL did not appear to result in the accumulation of small transport vesicles or tubules. Second, at the concentrations of inhibitors used here or in previous studies on trafficking through the Golgi complex, no acute effect on any known coated vesicle-mediated transport step was found, including endocytosis from the cell surface and anterograde trafficking through the secretory pathway (43). Third, BFA-stimulated tubulation of Tf-TfR-enriched endosomes was also potently inhibited within the same concentration range for all PLA 2 antagonists examined. And, fourth, cytosol-dependent endosome tubulation in the cell-free system was also potently inhibited by the same PLA 2 antagonists.
The conclusion that PLA 2 antagonists inhibit the tubulemediated exit of Tf-TfRs from both early sorting and late recycling endosomes is also consistent with our current understanding of the Tf-TfR recycling pathway (1, 2). For example, morphological studies have shown that after endocytosis from the cell surface, Tf-TfR complexes enter an early sorting endosome compartment located in the periphery of the cell. The complexes segregate into tubular extensions and are then packaged into vesicles and/or tubules that relocate to the central recycling compartment (9,17,56). Based on various pulsechase experiments, Maxfield and colleagues (17,18,57) have suggested that sorting within this compartment occurs by iterative fractionation, yielding multiple rounds of tubule and vesicle formation and fusion. By mechanisms that are not fully understood, these tubulo-vesicular membranes, enriched in Tf-TfR, form an extensively interconnected tubular recycling compartment (58 -62), from which Tf-TfRs recycle back to the plasma membrane. Interestingly, BFA, which enhances endosome tubule formation (36,37,39,40), has been reported to increase the number of various receptors, including TfRs, on the cell surface (39,42,63), perhaps as a consequence of stimulating tubule-mediated recycling. Thus, the observation that FIG. 10. Quantitation of in vitro endosome tubule formation. A, isolated endosomes, loaded by prior uptake with Au 12 -BSA, were incubated under various conditions as indicated for 30 min at 37°C and then visualized by electron microscopy negative staining. The concentrations of the various reagents present in the incubations were as follows: 1.5 mg/ml bovine serum albumin (BSA), 1.5 mg/ml bovine brain cytosol (BBC), 0.1 mM ATP. Heated refers to BBC that was heated to 100°C for 30 min before assaying. B, time course of endosome tubulation in vitro. Endosomes were incubated in the presence of bovine brain cytosol (1.5 mg/ml) and ATP (0.1 mM) for various times at 37°C. Endosome tubulation was fairly rapid, with t ϭ 4 min under these conditions. C, cytosol concentration dependence of endosome tubulation in vitro. Endosomes were incubated with varying concentrations of bovine brain cytosol in the presence of ATP (0.1 mM) for 30 min at 37°C. D, characterization of tubulation activity in bovine brain cytosol. Isolated endosomes were incubated under various conditions as indicated, in the presence of ATP (0.1 mM) for 30 min at 37°C. Crude refers to the organelle free extract resulting after bovine brains were homogenized and centrifuged at high speed. BBC, our standard cytosol preparation of material present after precipitation with 60% ammonium sulfate followed by extensive dialysis in dialysis buffer. The concentrations of various reagents present in the incubations were as follows: 1.5 mg/ml crude cytosol or BBC, 0.1 mM ATP, 10 g/ml BFA, 1 mM N-ethylmaleimide (NEM). The percentage of tubulation on the y axis is the percentage of gold-loaded endosomes with tubular extensions (determined as described under "Experimental Procedures"). Error bars, 1 S.D. determined from at least three independent experiments. PLA 2 antagonists inhibit Tf-TfR trafficking from sorting and recycling endosomes is consistent with the idea that membrane tubules play an important role in mediating these steps.
A role for PLA 2 in endosome-to-endosome fusion has also been suggested based on a similar pharmacological approach (64). However, the concentration of PLA 2 antagonists required to inhibit endosome fusion either in vivo or in vitro was ϳ10 -50-fold higher than that for inhibition of Tf-TfR recycling or endosome tubulation. Thus, at the concentrations used here, membrane fusion would not have been severely affected, a conclusion that is supported by finding that delivery of Tf to endosomal vesicles/vacuoles was not inhibited. These studies do, however, raise the interesting possibility that PLA 2 -dependent tubulation could be linked to a PLA 2 -dependent fusion event.
The cell-free system that reconstitutes endosome tubulation in vitro is an important adjunct to in vivo tubulation assays. The endosome tubules produced in vitro were uniform in diameter, 50 -70 nm, matching their in vivo counterparts (3, 4, 6 -8, 12, 18, 65, 66), and they often possessed flared or bulbous tips (see Fig. 7), which were also observed on endosome tubules in vivo (11). Interestingly, the requirements for endosome tubulation in vitro were nearly identical to those for Golgi membrane tubulation (51,52), including 1) dependence on a heatlabile, nondialyzable protein factor found in cytosol preparations; 2) the lack of requirement for Ca 2ϩ ; and 3) sensitivity to the same spectrum of PLA 2 inhibitors (44).
Although our in vivo and in vitro results suggest that cytoplasmic PLA 2 activity functions as a general tubulation factor, this activity was not found to induce tubulation of lysosomes, despite the well documented ability of lysosomes to become tubular in certain cells (67). However, tubular lysosomes are larger in diameter than endosome or Golgi tubules and may rely solely on microtubules for their formation (68,69). Thus, tubulation by a cytoplasmic PLA 2 may involve specific interactions with a subset of organelles on the cytoplasmic sides of their membranes.
In addition to PLA 2 -mediated mechanisms, intracellular membrane tubules could also form in other ways. For example, several recent studies have shown that the clathrin-coated vesicle-associated GTPase, dynamin, can be induced to assemble onto membranes and form tubules under conditions where the GTPase activity is inhibited (70 -73). It is important to note, however, that the tubules in our cytosol-dependent, PLA 2 inhibitor-sensitive in vitro assay were not affected by either GTP or GTP␥S and did not appear to be coated by spirals of dynamin polymers. Therefore, the two mechanisms appear to be quite different.
Taken together, these results suggest that a cytoplasmic PLA 2 activity is required for tubulation and membrane trafficking from multiple organelles, including endosome tubule formation and Tf-TfR recycling. One caveat is that we cannot definitively rule out that a phospholipase A 1 , lysophospholipase, or other lipase activity is responsible, because some of the tested antagonists have been shown to inhibit these activities as well (74,75). However, our conclusion is strengthened by finding that structurally unrelated compounds (BEL, ONO-RS-082, quinacrine, etc.) all have in common the ability to inhibit PLA 2 activity. BEL is particularly interesting, because it displays ϳ1000-fold more specificity for a cytoplasmic Ca 2ϩindependent versus Ca 2ϩ -dependent PLA 2 (74,75), and because our in vitro tubulation activity does not require Ca 2ϩ , a Ca 2ϩ -independent PLA 2 may be the relevant enzyme. A variety of Ca 2ϩ -independent, cytoplasmic PLA 2 forms, with no definitive known function, have been characterized to varying degrees (76 -79), and others are likely to be discovered. One of the most striking findings is that in tubulation, both endosomes and Golgi membranes display virtually identical sensitivities to PLA 2 antagonists. Thus, these results suggest similar mechanisms of tubule formation on these organelles and a novel role for a cytoplasmic Ca 2ϩ -independent PLA 2 in tubule-mediated intracellular trafficking.