Regulation of Endosome Sorting by a Specific PP2A Isoform

Antibodies against PP2A regulatory subunits have been described previously (Kamibayashi et al., 1994; Tehrani et al., 1996). PP2A catalytic subunit antibody was obtained from B. Wadzinsky (Vanderbilt University, Nashville, TN). mAb M1 was from Eastman Kodak Co. (Rochester, NY). PP1 and PP2A catalytic subunits were provided by A.A. DePaoli-Roach (Indiana University, Indianapolis, IN). All chemical reagents were obtained from Sigma Chemical Co. (St. Louis, MO) unless otherwise noted.

BSC-40 and PACS-1 control (C1) and antisense (AS19) cells were cultured as previously described (Thorne et al., 1989; Wan et al., 1998). HeLa cells expressing TS-Dyn I under control of Tet-suppressing elements were obtained from S. Schmid (The Scripps Research Institute, La Jolla, CA), and maintained as described (Damke et al., 1995). For immunofluorescence analyses, cells were cultured directly on glass coverslips to a density of 50–80% confluence before experimental manipulation as described in figure legends. Cells were fixed in 4% paraformaldehyde and processed for immunofluorescence as previously described (Molloy et al., 1994). mAb M1 was detected using either FITC- or TXR-conjugated goat anti– mouse IgG2b-specific secondary antibodies (Fisher Scientific Co., Pittsburgh, PA).

Epitope-tagged furin (fur/f), the CKII phosphorylation site point mutants (fur/f-DDD and fur/f-ADA), and the glutathione-S-transferase (GST)– Fur_cd fusion protein were generated previously (Molloy et al., 1994; Jones et al., 1995). Dynamin I WT and K44E cDNAs were each excised from pSVL (Herskovits et al., 1993) using EcoRI. The inserts were then blunted with Klenow and cloned into the vaccinia recombination vector pZVneo cut with StuI. SV-40 small t WT and the truncated small t mut 3 (Sontag et al., 1993) were excised from pCMV5 using EcoRI and BamHI, and then blunted with Klenow and inserted into pZVneo cut with StuI. Recombinant vaccinia expressing the various constructs were generated by standard methods (VanSlyke et al., 1995).

HeLa cells expressing the temperature-sensitive dominant-negative dynamin I were cultured in 35-mm plates for 3 d at permissive temperature in the absence of tetracycline. The cells were then infected with vaccinia virus expressing fur/f with either the serine to alanine (fur/f-ADA) or serine to aspartic acid (fur/f-DDD) substitutions within the cd CKII site (multiplicity of infection [m.o.i.] = 10) and allowed to express at nonpermissive temperature, 37°C. At 6 h postinfection the cells were placed on ice, rinsed with cold PBS and surface proteins were labeled for 1 h using 0.5 mg/ml EZ-link NHS-SS-biotin (Pierce Chemical Co., Rockford, IL) in PBS. After labeling, the cells were rinsed three times with PBS containing 50 mM glycine to quench unreacted biotin, and then refed with prewarmed medium (except for 0 time points) and placed at permissive temperature (31°C) for 10, 20, 30, or 40 min. At each time point, cells were transferred to ice, rinsed with cold PBS, and then either harvested directly in modified RIPA buffer (mRIPA, 50 mM Tris-HCl, pH 8, 150 mM NaCl, 1% NP-40 [Calbiochem-Novabiochem Corp., La Jolla, CA] and 1% sodium deoxycholate) for immunoprecipitation of total labeled furin using mAb M1 as previously described (Molloy et al., 1994), or were stripped of remaining surface biotin by washing three times for 10 min in Tris-buffered saline (TBS; 10 mM Tris-HCl, pH 8, 150 mM NaCl) with MesNa (50 mM). The surface-stripped samples were rinsed with PBS supplemented with Hepes (10 mM, pH 7.5) and harvested in mRIPA for immunoprecipitation of internalized labeled furin with mAb M1. The immunoprecipitates were resolved by SDS-PAGE on 8% gels, transferred to nitrocellulose membranes and probed with avidin HRP (1:2,000 dilution of 1 mg/ml in TBS + 0.05% Triton X-100) and developed by chemiluminescence (Renaissance; NEN Life Science Products, Boston, MA) to detect biotinylated furin. In double-strip analyses, cells processed as described above after the initial 20-min uptake period were then incubated for a second 20-min period at permissive temperature to allow for recycling of the internal pool to the cell surface. After the second chase period, cells were placed on ice and either harvested directly (control) or subjected to a second MesNa strip as described above before harvesting and analysis by Western blot.

Baculoviruses expressing PP2A catalytic, A, Bα, and Bβ subunits have been described previously (Kamibayashi et al., 1994). The recombinant baculovirus encoding the B′α subunit was generated as described (Tehrani et al., 1996). For expression experiments, Sf9 cells growing in log phase were seeded in 25-cm flasks (3 × 10⁶ cells/flask) and infected with baculovirus recombinants (m.o.i. = 2). The cells were harvested at 64–72 h postinfection by trituration and then pelleted and washed with PBS. The washed cell pellet was then resuspended in 0.5 ml of ice-cold harvest buffer (50 mM Tris, pH 7.0, 1 mM EDTA, and 2 mM DTT with pepstatin, leupeptin, PMSF, aprotinin, and E64) and the cells broken open by passage (10×) through a 25-gauge needle. The lysates were clarified by low speed centrifugation at 500 g and brought to 10% glycerol before performing phosphatase assays (see below).

Substrates for phosphatase assays were prepared by phosphorylating GST–Fur_cd fusion protein and phosphorylase b in vitro with CKII or phosphorylase kinase, respectively. The phosphorylation reactions (100 μl) contained 50 μg of substrate, and 100 μM of ³²P-labeled ATP (3,000 cpm/ pmol). After phosphorylation for 1 h at 30°C, the reactions were run over two G25 spin columns to remove free ATP. Dephosphorylation assays were conducted in a 50-μl reaction containing lysates, buffer (25 mM Tris, pH 7.0, 0.2 mM MnCl₂, 1 mM DTT, and 0.2 mg/ml BSA) and ∼10 μM phosphorylated substrate (either phosphorylase a or GST–Fur_cd). Aliquots (5 μl in triplicate) of each reaction were removed at 15 and 30 min of incubation and applied to P81 filter paper (Whatman Inc., Clifton, NJ) squares that were washed three times with 75 mM phosphoric acid to remove free phosphate and then dried and subjected to scintillation counting. In some cases, the aliquots were mixed with SDS sample buffer, resolved by SDS-PAGE, and then analyzed using a PhosphoImager (model 445SI; Molecular Dynamics Inc., Sunnyvale, CA). Both measurement techniques gave identical results.

Affinity chromatography for purification of endogenous furin-directed phosphatase from BSC-40 cell extract was performed essentially as described (Moorhead et al., 1994). In brief, microcystin (MC)-LR (LC Laboratories, Woburn, MA) was derivatized with aminoethanethiol to generate a primary amine group for coupling to NHS-activated sepharose (Pharmacia Biotech Inc., Piscataway, NJ). Approximately 0.25 mg of derivatized MC was linked to 1 ml of Sepharose resin. For the preparation of extracts, three 15-cm plates of confluent BSC-40 cells were rinsed three times with cold PBS and one time with harvest buffer (see above). The cells were then scraped from the plates and lysed by passage (10×) through a 25-gauge needle. The extract was then clarified by low-speed centrifugation (10 min at 500 g) and brought to 10% glycerol. The clarified extract was then cycled over the MC column (0.5 ml/min) for 1 h at 4°C to allow binding of the phosphatase. The column was then washed with 10 vol of harvest buffer and eluted with 3M NaSCN. Each eluted fraction (10 ml) was dialyzed over night against harvest buffer before assaying for phosphorylase a and GST–Fur_cd-directed phosphatase activities as described above.

Phosphorylation of the furin-cd by CKII is necessary for maintaining the steady-state distribution of the endoprotease in the TGN and also for the localization of cell surface-internalized furin to early endosomes. By contrast, sorting of furin from early endosomes to the TGN requires dephosphorylation of the furin-cd (Jones et al., 1995). These findings were confirmed by antibody uptake studies (Fig. 1). Incubation of cells expressing epitope (FLAG)- tagged furin (fur/f) with mAb M1 demonstrated the retrieval of cell surface fur/f to the TGN (Fig. 1,A). However, treatment of cells with tautomycin, a potent inhibitor of PP1 and PP2A (Takai et al., 1995), resulted in the accumulation of internalized fur/f in peripheral punctate structures characteristic of early endosomes (Fig. 1,B). The effect of tautomycin is specific for a furin-directed phosphatase since (a) this drug has no effect on the internalization of a nonphosphorylatable furin construct, fur/ fS_773,775A (ADA) (Jones et al., 1995), and (b) internalization of a phosphorylation mimic furin construct, fur/ fS_773,775D (DDD), also resulted in accumulation in the peripheral punctate compartments (Fig. 1,C). Double-labeling experiments showed that internalized fur/f-DDD and transferrin colocalized within these peripheral vesicles (Fig. 1, D and E). Together these results demonstrate that after internalization, phosphorylated furin localizes to early endosomes and sorting of the endoprotease from this compartment to the TGN requires the activity of a tautomycin-sensitive furin phosphatase.

The dynamic nature of the early endosomal sorting system suggests that the phosphorylation state-dependent accumulation of furin in early endosomes is not likely due to static retention in these structures. Therefore, we tested the possibility that this accumulation reflected a directed cycling of phosphorylated furin between the cell surface and early endosomes. To facilitate quantitative analyses of internalization and recycling, we first investigated the mechanism of furin endocytosis as a means by which to increase the surface pool of the protease. The presence of functional tyrosine- and dileucine-based internalization signals within the furin-cd suggested that endocytosis occurs via a clathrin-dependent pathway (Schäfer et al., 1995; Ohno et al., 1996). Therefore, we expressed a dominant-negative form of the GTPase dynamin I, K44E, to effectively block clathrin-dependent endocytosis (Herskovits et al., 1993; van der Bliek et al., 1993) and examined its effect on furin trafficking. Immunofluorescence analysis of furin internalization from the cell surface (Fig. 2) showed that although overexpression of native dynamin I had no effect on uptake (compare Fig. 2,A with Fig. 1,A), the K44E dominant-negative mutant blocked uptake and resulted in accumulation of the enzyme at the cell surface (Fig. 2,B). This effect on furin internalization mimicked the block in TfR endocytosis as observed with TRITC-transferrin uptake (Fig. 2, C [Dyn I] and D [K44E]).

Based on this result, quantitative analyses of furin trafficking were performed in stably transfected HeLa cells expressing a temperature-sensitive and reversible dominant-negative form of dynamin I (TS-Dyn I). Incubation of cells expressing fur/f with TS-Dyn I at nonpermissive temperature resulted in a dramatic increase in the amount of furin detected by surface biotinylation when compared with control cells at permissive temperature (data not shown). Therefore, we used this system to compare the trafficking of fur/f-DDD, to fur/f-ADA (Fig. 3). After surface labeling, cells expressing either fur/f-DDD or fur/f-ADA were chased for the increasing times to allow endocytosis/recycling. At each time point duplicate samples were either processed to detect internalized (I) or total (T) surface labeled furin (Fig. 3,A). The ratio of protected/total furin was plotted quantitatively as a percentage of surface furin internalized versus time at the permissive temperature (Fig. 3,B). At time points up to 20 min the internalization of fur/f-DDD and fur/f-ADA were similar. However, at longer incubation times the percentage of fur/ f-DDD in internal compartments decreased while the percentage of internalized fur/f-ADA continued to increase. Additional double-strip experiments showed that a significant portion (∼50%) of the fur/f-DDD found in internal compartments at the 20-min time point was reexpressed at the cell surface during a subsequent 20-min chase (Fig. 3 B, inset). Reexpression of substantial amounts of fur/f-DDD at the cell surface within 20–30 min is consistent with the recycling time of transferrin receptor in HeLa cells (Bleil and Bretscher, 1982), and suggests that the phosphorylated form of furin undergoes a similar local cycling between early endosomes and the cell surface.

The cell surface/early endosome recycling of furin requires multiple sorting motifs within the enzyme's cd. Internalization of furin from the cell surface is directed by the canonical tyrosine and di-leucine-based endocytosis signals (Schäfer et al., 1995). Recycling from the early endosome to the cell surface, however, requires the CKII-catalyzed phosphorylation of the furin-cd AC (Fig. 3). Whereas the hydrophobic endocytosis signals bind directly to the clathrin adaptor AP-2 (Ohno et al., 1996), association of the phosphorylated furin-cd AC with the clathrin sorting machinery requires a connector protein, PACS-1 (Wan et al., 1998). A requirement for PACS-1 for the early endosomal localization of phosphorylated furin was demonstrated by monitoring furin internalization in PACS-1–deficient cells (Fig. 4). Parallel plates of control or PACS-1–deficient cells expressing fur/f were incubated with mAb M1 for 1 h in the presence of tautomycin. In agreement with Fig. 1, tautomycin treatment caused the accumulation of internalized fur/f in transferrin-containing early endosomes (Fig. 4,A). By contrast, in the absence of PACS-1, internalized fur/f failed to localize to early endosomes and instead showed a faint and dispersed staining pattern throughout the cytosol (Fig. 4,C). Early endosomal localization of internalized transferrin, however, was unaffected by PACS-1 depletion (Fig. 4, compare B with D). The inability of internalized fur/f to localize to early endosomes in the PACS-1–deficient cells was not the result of defective furin endocytosis since after short times of antibody uptake both the control and PACS-1–deficient cells showed similar punctate staining patterns (Fig. 4, E and G). Rather, the faint staining pattern in panel B reflects a shared requirement for PACS-1 in the localization of phosphorylated furin both to early endosomes and to the TGN (Wan et al., 1998). Indeed, mAb M1 uptake in the absence of tautomycin showed that in control cells fur/f is retrieved to the TGN and to a peripheral endosome/lysosome population (Fig. 4,F). In PACS-1 antisense cells, however, internalized fur/f fails to localize to either the TGN or to the peripheral endosome/lysosome population (Fig. 4 H). Together, these observations show that the sorting, but not the internalization, of furin requires PACS-1.

After internalization into early endosomes, phosphorylated furin may either (a) undergo a PACS-1–dependent recycling to the cell surface (Figs. 3 and 4) or (b) be directed to the TGN by a tautomycin-sensitive furin phosphatase (refer to Fig. 1) (Jones et al., 1995). The sensitivity to tautomycin suggested the furin phosphatase is either PP1 or PP2A. Therefore, the ability of both PP1 and PP2A catalytic (C) subunits to dephosphorylate the phosphorylated furin-cd in vitro was determined (Fig. 5). Surprisingly, neither phosphatase was able to dephosphorylate the furin-cd, whereas both enzymes efficiently dephosphorylated a control substrate, phosphorylase a. Similar results were obtained for PP2B (data not shown).

Whereas neither the PP1 nor the PP2A catalytic subunits were active against the phosphorylated furin-cd, both cell extracts and bovine brain cytosol contained a tautomycin- and okadaic acid-sensitive furin phosphatase activity (Fig. 6,A). Because of the high total protein concentrations required to detect furin phosphatase activity, we were unable to identify the phosphatase by differential sensitivity to these inhibitors. Therefore, an affinity chromatography procedure was used to isolate the furin phosphatase. Cell extract containing furin phosphatase activity was applied to a microcystin (a potent PP1/PP2A inhibitor) affinity column (Moorhead et al., 1994, 1995). Although microcystin normally covalently binds to both PP1 and PP2A (MacKintosh et al., 1995), derivatization for coupling to the column matrix blocks the reactive group and changes the reagent to a tight-binding reversible inhibitor. Bound material was then eluted with NaSCN, dialyzed and assayed for phosphatase activity using either the phosphorylated furin-cd or phosphorylase a as substrate (Fig. 6 B). The microcystin affinity column efficiently bound both the phosphorylase a and furin-cd phosphatase activities from the load sample. However, only the phosphorylase a activity could be recovered from the eluate. The elution conditions were not responsible for the loss of furin-directed activity since treatment of extract with NaSCN followed by dialysis did not inactivate the furin phosphatase. Furthermore, neither the pooling of all eluted fractions from the column before dialysis nor the addition of the column flow-through to the assay was able to restore furin-directed phosphatase activity (data not shown). These results suggested that the depletion of furin phosphatase activity was not simply the result of separating the catalytic moiety from a required cofactor during purification.

The enhanced total phosphorylase a–directed activity recovered in the eluate fractions is consistent with previous reports (Moorhead et al., 1994, 1995), and reflects the purification of PP1 away from endogenous inhibitors. Interestingly, despite the ability of microcystin to bind and inhibit both PP1 and PP2A, only PP1 isoforms from muscle myofibrils and glycogen particles have been successfully isolated using this affinity technique (Moorhead et al., 1994, 1995). It is possible, therefore, that exposure of bound PP2A to the high concentrations of chaotropic agent required to dissociate it from the resin leads to an irreversible inhibition of the enzyme. These results, together with those in Fig. 5, indicate that the furin phosphatase is indeed a PP1/PP2A-like enzyme (tautomycin-sensitive, binds to the microcystin column). Furthermore, the inability to recover the furin phosphatase activity from the affinity column implicates PP2A.

Although the PP2A C subunit is active towards numerous substrates in vitro, the predominant form of the enzyme in vivo is a trimeric complex with additional A and B subunits (for reviews see Barford, 1996; Hubbard and Cohen, 1993; Mayer-Jaekel and Hemmings, 1994). The A subunit promotes association of the C subunit with one of a variety of regulatory B subunits. Recent cloning studies reveal several unique families of regulatory subunits with multiple members derived from both separate genes and alternative splicing. Although the role of complex formation was originally envisioned as restricting the otherwise broad substrate specificity of the isolated C subunit, recent studies show that the B subunits can act as positive regulators to enhance catalytic activity toward particular substrates both in vitro and in vivo (Sontag et al., 1996). Moreover, the distinct subcellular localization of the different B subunits could act to target isoforms of PP2A to particular locations within the cell (Sontag et al., 1995; McCright et al., 1996; Okamoto et al., 1996). Together, these findings suggest a potentially high degree of isoform-specific PP2A characteristics.

To test the possibility that such isoform-specific composition could modulate the activity of PP2A toward furin, we adopted a strategy of phosphatase isoform reconstitution by baculovirus expression (Kamibayashi et al., 1994). Sf9 cells were coinfected with baculovirus recombinants expressing combinations of the C, A, and one of the various B subunits (Fig. 7,A). Furin phosphatase activity was strictly dependent on B subunit composition. Whereas expression of C subunit alone or coexpression of C and A subunits failed to generate increased furin phosphatase activity, coexpression of the C and A subunits with B family regulatory subunits (Bα and Bβ) resulted in a selective increase in furin phosphatase activity (two- to fourfold). Importantly, the furin phosphatase activity was B family– specific since coexpression of C and A subunits with a member of the B′ family, B′α, failed to stimulate furin phosphatase activity. Indeed, B′α selectively inhibited the basal furin-directed activity. This effect is most likely due to displacement of endogenous Sf9 cell B family regulatory subunits from holoenzyme complexes by the overexpressed B′α, which has a higher affinity for the AC complex in vitro (Tehrani et al., 1996). As expected, both the phosphorylase a and furin phosphatase activities were inhibited completely by submicromolar concentrations of okadaic acid and tautomycin. Furthermore, Western blot analyses of the Sf9 extracts using subunit-specific antisera show that (a) the appropriate regulatory subunits were indeed expressed by the recombinant baculoviruses and (b) similar levels of catalytic subunit were expressed in each combinatorial infection (Fig. 7 B). This latter observation indicates that the dramatic changes in furin dephosphorylation are not the result of differential stabilization of the catalytic subunit, but rather reflect the inherent ability of the regulatory subunits to facilitate the recognition of furin by PP2A holoenzyme.

The in vitro data described above clearly implicate PP2A isoforms containing B family subunits in the dephosphorylation of furin. The characterization of SV-40 virus-transforming factors has also provided a mechanism through which to probe the importance of these particular PP2A isoforms in vivo. The small t antigen of SV-40 contributes to the transforming capacity of the large T antigen, at least in part, by disrupting the mitogen-activated protein (MAP) kinase signaling pathway. Recent studies have shown that this effect of small t derives from the ability of the SV-40 protein to displace B family regulatory subunits from PP2A, resulting in a change in PP2A substrate specificity and hyperphosphorylation of several members of the MAP kinase cascade (Sontag et al., 1993). We used this characteristic of small t antigen to examine directly the importance of PP2A for the correct routing of furin in vivo.

To test this possibility, replicate plates of BSC-40 cells were coinfected with vaccinia virus recombinants expressing fur/f and either small t or an inactive small t mutant (mut3) lacking the PP2A binding site (Sontag et al., 1993). The effect of small t expression on the recycling of furin from the cell surface was then assessed in immunofluorescence studies as described above (Fig. 8). As seen previously, antibody uptake showed that recycling furin is primarily localized to the TGN/late endosome in control cells (Fig. 8,A). Coexpression of small t, however, resulted in an accumulation of the internalized furin in peripheral endosome-like structures (Fig. 8,C) as seen in cells treated with tautomycin (Fig 8,B). This effect required the PP2A-binding function of small t since the inactive mut3 construct had no discernible effect on furin trafficking (Fig. 8,D). These results complement the in vitro data described in Fig. 7 and show that PP2A isoforms containing B family regulatory subunits control the dephosphorylation dependent transfer of furin from early endosomes to the TGN.

The mechanisms by which the endosomal system achieves and controls its complex sorting functions are largely unknown. Studies of furin recycling, however, demonstrate several factors that affect sorting within the early endosomal system. Here we report the identification of cellular machinery that directs the phosphorylation state-dependent trafficking of furin through endosomal compartments. Cell surface furin is endocytosed in a clathrin-/dynamin I–dependent step by direct interaction of the tyrosine- and dileucine-like internalization motifs with the AP-2 adaptor (refer to Fig. 2) (Schäfer et al., 1995; Ohno et al., 1996). After localization to transferrin containing early endosomes (refer to Fig. 1), furin undergoes a phosphorylation state-dependent sorting step that is controlled by the activities of CKII, and PP2A isoforms containing B family regulatory subunits (Jones et al., 1995) (refer to Fig. 7). Phosphorylated furin molecules bind the connector protein PACS-1 and are placed in a local cycling loop between the early endosome and the cell surface (refer to Figs. 3 and 4). By contrast, furin molecules dephosphorylated by PP2A are sorted to the TGN (refer to Figs. 5–8).

Several findings indicate that phosphorylation/dephosphorylation of the furin-cd by CKII and PP2A directly affect furin trafficking in the TGN/endosomal system (Jones et al., 1995; Dittié et al., 1997; Wan et al., 1998). For example, the accumulation of furin in early endosomes observed upon treatment of cells with tautomycin (refer to Fig. 1) (Jones et al., 1995) is replicated by the fur/f-DDD point mutant designed to mimic phosphorylation at the CKII site. Furthermore, this effect of tautomycim requires an intact, phosphorylatable CKII site in the furin-cd (Jones et al., 1995). Similarly, the effect of depletion of the phosphorylation state-dependent furin-binding protein PACS-1 upon trafficking is selective for furin and proteins containing related AC motifs (Wan et al., 1998). Although these results suggest direct effects on furin sorting, it seems likely that CKII and PP2A also have additional global roles in controlling membrane traffic within the TGN/endosomal system.

The regulation of furin trafficking by the combined activities of CKII and PP2A indicates a link between signaling pathways and control of protein localization within the TGN/endosomal system. Together with their known function in cell cycle progression, our studies support a broad role for CKII and PP2A in regulating diverse cellular processes. Although modulation of CKII activity in vivo has not been demonstrated (Allende and Allende, 1995), PP2A can be regulated in several ways. The importance of holoenzyme composition in generating a furin-directed phosphatase (refer to Fig. 7) illustrates the high level of substrate specificity determined by subunit composition. Similarly, the B family is required for the PP2A-catalyzed dephosphorylation of tau both in vitro and in vivo. Although a selective role for B′ family subunits has not been demonstrated in mammalian cells, deletion of the B′ homologue in yeast, Rts1, results in a temperature-sensitive growth defect that can be rescued by rabbit B′ family subunits (Zhao et al., 1997). Together, these studies show that the PP2A regulatory subunits can act as positive effectors for select substrates.

Regulatory subunit composition may also modulate PP2A by targeting the phosphatase to select compartments. For example, B′ subunits differentially partition between the nucleus and cytoplasm (McCright et al., 1996), whereas Bα subunits target a population of PP2A to microtubules where they are positioned for efficient dephosphorylation of tau (Sontag et al., 1995). The microtubule-associated PP2A also undergoes cell cycle-dependent modulation of its activity, indicating an additional level of regulation. The observation that both the catalytic and regulatory subunits of PP2A are phosphoproteins (Mayer-Jaekel and Hemmings, 1994; McCright et al., 1996) introduces the possibility that this second tier of regulation represents rapid, reversible changes in PP2A function linked to second messenger signaling pathways.

The effect of SV-40 small t expression on furin sorting (Fig. 8) provides an unequivocal demonstration of PP2A's role in directing trafficking in the endocytic pathway. The selective displacement of B family regulatory subunits by small t offers a sensitive and direct diagnostic tool for delineating the role of PP2A isoforms in vivo. Our data do not, however, exclude a potential contribution by PP1 to the regulation of furin sorting. At limiting concentrations in vitro, PP1 is at least an order of magnitude more sensitive to tautomycin whereas PP2A is more sensitive to okadaic acid (Takai et al., 1995). Initial studies of the phosphorylation state-dependent sorting of furin, however, showed that, at low concentrations (100 nM), tautomycin affected sorting while similar concentrations of okadaic acid had no effect. Okadaic acid at high concentrations (e.g., 1 μM) not only affected furin trafficking, but also caused a dramatic dispersal of the paranuclear staining, consistent with disruption of the microtubule network and fragmentation of the TGN and Golgi (Lucocq, 1992; Reaven et al., 1993; Horn and Banting, 1994), precluding the use of this inhibitor in evaluating protein sorting. This preferential inhibition of furin trafficking by tautomycin could reflect cell type differences in the permeability of the inhibitors and/or their relative ability to penetrate the cellular compartments associated with PP2A-dependent sorting. Although the empirical determination of effective concentrations by examination of residual PP1 and PP2A activities in treated cells can facilitate the application of inhibitors as diagnostic tools (Favre et al., 1997), the possibility of resistant or sensitive pools of the enzymes remains a complicating factor.

The requirement of PACS-1 for the sorting of phosphorylated furin in early endosomes points to several commonalties between the early endosome/cell surface furin cycling loop and the phosphorylation- and PACS-1–dependent localization of the endoprotease to the TGN (Wan et al., 1998). In the model shown in Fig. 9, the cell surface/early endosome local cycling loop represents a mirror image of a TGN/endosome cycling pathway. As for endocytosis of cell surface furin to early endosomes (Schäfer et al., 1995), efficient export of TGN-localized furin to a post-TGN endosomal compartment requires the canonical tyrosine- and/or dileucine-based internalization signals (Wan et al., 1998). These hydrophobic sorting signals bind directly to the clathrin adaptor AP-2 at the cell surface, whereas budding from the TGN uses AP-1 (Alconada et al., 1996; Honing et al., 1996; Ohno et al., 1996; Wan et al., 1998). Although furin in the peripheral recycling pathway colocalizes with TfR, demonstrating its presence in early endosomes, the identity of the post-TGN recycling compartment is not established. The recovery of phosphorylated furin from immature secretory granules in neuroendocrine cells (Dittié et al., 1997), however, provides an analogy for the TGN recycling loop. PACS-1 directs the retrieval step of both cycling loops by linking phosphorylated furin to the clathrin sorting machinery. Binding assays in vitro show PACS-1 connects the phosphorylated furin-cd to AP-1, which is consistent with a role for this adaptor in the TGN localization of membrane proteins. This finding, however, does not exclude the possibility that other adaptors (e.g., AP-3) may also mediate retrieval. The composition of the adaptor species used in the PACS-1–directed recycling between early endosomes and the plasma membrane has yet to be established. Interestingly, the prevalence on endosomes of clathrin-coated buds which contain neither α or γ adaptin (Stoorvogel et al., 1996) indicates that unique adaptor species may mediate the retrieval step within the peripheral cycling loop.

As reflected in Fig. 9, depletion of PACS-1 leads to sorting defects at both the level of endosomal trafficking and TGN localization. This model, however, does not address why nonphosphorylatable furin, fur/f-ADA (which does not bind PACS-1), displays some trafficking characteristics unique from fur/f in PACS-1 antisense cells (compare Fig. 4 and Wan et al., 1998 with Takahashi et al., 1995 and Jones et al., 1995). Most notable is the finding that fur/f-ADA mislocalizes from either the TGN or early endosomes to larger punctate membrane structures whereas fur/f in PACS-1 antisense cells shows very fine, dispersed vesicular staining. There are several possible explanations for these results, including (a) the presence of additional phosphorylated-furin binding proteins, (b) multiple roles for PACS-1 in organizing endosomal sorting compartments, and (c) the presence of binding proteins which interact selectively with the nonphosphorylated furin cd but not with fur/f-ADA. This latter possibility implies that nonphosphorylated AC motifs constitute positive sorting signals for trafficking within the TGN/endosomal system.

While PACS-1 directs the retrieval of phosphorylated furin in both the TGN and peripheral cycling loops, the transport of furin between these loops requires dephosphorylation by PP2A. This dephosphorylation-dependent trafficking step provides a mechanism by which cells may control the distribution of furin between the biosynthetic and plasma membrane/endosomal processing compartments. Furin at the cell surface is tethered by interaction with the actin binding protein ABP-280 (Liu et al., 1997). This interaction modulates the rate of furin internalization and may also act to generate processing sites at the cell surface (e.g., bacterial toxin activation) (Gordon et al., 1995). The cycling of furin between the cell surface and endosomes may reflect a requirement for the formation of peripheral processing compartments in which the enzyme can be concentrated with its substrates. In addition, the internal milieu of the early endosomes (e.g., acidic pH) facilitates the processing of some substrates such as pseudomonas and diphtheria toxins (Gordon et al., 1995). Thus, due to the highly dynamic and transitory nature of the endosomal system, continuous cycling of phosphorylated furin between the plasma membrane and early endosomes could be the best mechanism for maintaining functional concentrations of the protease within the peripheral processing compartments. Similarly, the TGN cycling loop might function to optimize processing of substrates in the biosynthetic pathway. This concept is consistent with the observation that, in some cells, endogenous furin can be predominantly localized to a post-TGN processing compartment (Sariola et al., 1995).

The PACS-1–directed trafficking of furin suggests that the localization of additional proteins containing acidic cluster motifs may be regulated via the phosphorylation state-dependent sorting machinery. Indeed, a number of itinerant secretory pathway membrane proteins including the cation-independent (Meresse et al., 1990) and cation-dependent (Mauxion et al., 1996) mannose 6 phosphate receptors, carboxypeptidase D (Xin et al., 1997), and sortilin (Petersen et al., 1997) have CKII phosphorylatable ACs within their cytosolic domains. In the case of the MPRs this phosphorylation has been linked to the efficient function of the receptor, however, it is not clear how this modification effects trafficking. Interestingly, many herpes virus envelope glycoproteins (e.g., VZV-gE, HSV-1-gE, and HCMV-gB) have phosphorylatable AC sorting motifs on their cds (Edson et al., 1987; Norais et al., 1996; Yao et al., 1993). Recent data indicate that the phosphorylation of these sites can influence trafficking of the envelope proteins (Alconada et al., 1996; Fish et al., 1998) and that PACS-1 binds VZV-gE cytosolic domain (Wan et al., 1998), raising the possibility that the CKII-mediated sorting system is used in viral biogenesis and/or spread.

Our studies of the regulation of furin sorting between the early endosome and plasma membrane also reveal several parallels with the cellular machinery that controls the phosphorylation-dependent resensitization of G protein– coupled receptors. The trafficking of both furin and βAR depend upon (a) the phosphorylation state of the cytosolic domain (Ferguson et al., 1995, 1996), (b) the presence of a phosphorylation state-dependent connector protein (either PACS-1 or β arrestin) (Goodman et al., 1996) that provides a link to the clathrin-sorting machinery, and (c) dephosphorylation by specific endosome-associated PP2A isoforms (Pitcher et al., 1995).

In summary, our results reveal the integrated roles of identified components of the endosomal trafficking machinery that direct the phosphorylation state-dependent sorting of furin. The importance of kinase (CKII) and isoform-specific phosphatase (PP2A) activities in regulating protein routing also indicates that these sorting events are likely under the control of second messenger systems. Future studies, therefore, will focus on how this dynamic endosomal sorting system responds to intracellular signaling pathways in order to control the distribution and activity of membrane proteins in vivo.

We thank B.G. Jones (University of Sheffield, Sheffield, UK) for contributions to the early part of this study. We also wish thank S.L. Schmid for the TS-Dyn I cell line, A.A. DePaoli-Roach for purified PP1 and PP2A, Wadzinsky for PP2A catalytic subunit antibody, R. Valle (Worcester Foundation, Shrewsbury, MA) for dynamin I and K44E plasmids, and P. Cohen (University of Dundee, Dundee, UK) for helpful discussions.

This work was supported by National Institutes of Health grants to G. Thomas (DK-44629 and DK-37274).

AC

acidic cluster

cd

cytosolic domain

GST

glutathione-S-transferase

MC

microcystin

m.o.i.

multiplicity of infection

PACS-1

phosphofurin acidic cluster-sorting protein

PP2A

protein phosphatase 2A