Differential distribution of mannose-6-phosphate receptors and furin in immature secretory granules

Transcription

1 Journal of Cell Science 112, (1999) Printed in Great Britain The Company of Biologists Limited 1999 JCS Differential distribution of mannose-6-phosphate receptors and furin in immature secretory granules Andrea S. Dittié 1, *, Judith Klumperman 2 and Sharon A. Tooze 1, 1 Secretory Pathways Laboratory, Imperial Cancer Research Fund, 44 Lincoln s Inn Fields, London, WC2A 3PX, UK 2 University of Utrecht Medical School, Institute for Biomembranes, Heidelberglaan 100, 3584CX Utrecht, The Netherlands *Present address: Bayer AG, Apratherweg 18a, D Wuppertal, Germany Author for correspondence ( tooze@icrf.icnet.uk) Accepted 1 September; published on WWW 3 November 1999 SUMMARY In neuroendocrine cells sorting of proteins from immature secretory granules (ISGs) occurs during maturation and is achieved by clathrin-coated vesicles containing the adaptor protein (AP)-1. We have investigated the role of the mannose-6-phosphate receptors (M6PRs) in the recruitment of AP-1 to ISGs. M6PRs were detected in ISGs isolated from PC12 cells by subcellular fractionation, and by immuno-em labelling on cryosections. In light of our previous results, where greater than 80% of the ISGs were found to contain furin, we examined the relationship between furin and M6PR on ISGs. By immunoisolation techniques we find that 50% at most of the ISGs contain the cation-independent (CI)-M6PR. Using sequential immunoisolation we could demonstrate that there are two populations of ISGs: those that have both M6PR and furin, and those which contain only furin. Furthermore, using immobilized GST-fusion proteins containing the cytoplasmic domain of the CI-M6PR we have shown binding of AP-1 requires casein kinase II phosphorylation of the CI-M6PR fusion protein, and in particular phosphorylation of Ser Addition of these phosphorylated GST-CI-M6PR fusion proteins to a cellfree assay reconstituting AP-1 binding to ISGs inhibits AP- 1 recruitment to ISGs. Key words: Regulated secretion, AP-1, Clathrin-coated vesicle, Furin, Casein kinase II INTRODUCTION The molecular details of secretory granule formation from trans-golgi network (TGN) membranes is the subject of controversy and the precise mechanisms involved in secretory granule formation are not yet known. Regarding the selection of secretory granule content proteins experimental data has been obtained, using endocrine and neuroendocrine cell systems, which supports two disparate, but not mutually exclusive, models: (i) sorting is required for inclusion into ISGs and occurs via an active process utilizing membrane associated components, or (ii) sorting is not required for inclusion into ISGs and does not occur prior to or during budding from the TGN (for recent reviews see Tooze, 1998 and Arvan and Castle, 1998). Reconciliation of the experimental data is possible if one considers cell-type specific variations in the amount of newly synthesized regulated secretory proteins. The data to support the latter model is from cell types where greater than 90% of all newly synthesised proteins are delivered to the secretory granule. This large efflux into the regulated pathway, which could be analogous to bulk flow (Pfeffer and Rothman, 1987), may alleviate the need to selectively sort in the TGN the regulated secretory proteins from other soluble proteins; the latter may comprise only 10% of the protein in the regulated secretory pathway. In contrast, and consistent with the former model, in other cell types the regulated secretory proteins represent a smaller percentage of newly synthesised proteins and may have to be actively selected by sorting receptors to be concentrated in nascent secretory granules. Even less is known about the mechanisms involved in the biogenesis of the future secretory granule membrane surrounding the core proteins. The mature secretory granule (MSG) membrane contains a wide variety of proteins, for example the vacuolar H + -ATPase, cytochrome b 561 and synaptotagmin (for a recent comprehensive list see Apps, 1997), whose function, in many cases, is known but whose trafficking into the MSG has not been extensively studied. We have been characterizing the ISG in the neuroendocrine cell line PC12 in an attempt to address some of the current questions about the mechanisms of sorting of both the secretory granule content and membrane during secretory granule formation. ISGs are the intermediate vesicular compartment between the TGN and MSGs (Tooze et al., 1991), and were first identified in pituitary mammotroph cells by EM autoradiography as being the initial site of hormone concentration after the Golgi complex (Farquhar and Palade, 1981). Subsequent studies have added to the morphological definition of this compartment in a variety of cell types (for review see Tooze, 1991; Arvan and Castle, 1998). The consensus from morphological and more recent biochemical

2 3956 A. S. Dittié, J. Klumperman and S. A. Tooze studies is that the ISG contains prohormones, has a mildly acidic ph, and bears patches of a clathrin coat which is comprised of adaptor protein (AP)-1 and clathrin unlike MSGs which contain hormones, are acidic, and have no clathrin patches. The protein composition of the ISG membrane is not well characterized, however the cation-independent (CI)- mannose-6-phosphate receptor (M6PR), the cation-dependent (CD)-M6PR (Klumperman et al., 1998), and furin (Dittié et al., 1997) occur in ISGs but not in MSGs. The properties of, and differences between, ISGs and MSGs have contibuted to a model in which maturation from an ISG to an MSG requires removal of both soluble and membrane proteins. Secretion of soluble C-peptide (derived from proinsulin through endopeptidase cleavage) from ISGs occurs via a pathway referred to as the constitutive-like pathway (Kuliawat and Arvan, 1992), while vesicles budding from the ISG have been identified which are clathrin-coated and contain M6PRs (Klumperman et al., 1998). It is not yet known whether these pathways overlap, ie if the clathrin-coated vesicles also contain C-peptide. Finally, it has not yet been determined if the M6PR positive, clathrin coated membranes and vesicles also contain furin. The M6PRs and furin are found in TGN membranes, endosomes and the plasma membrane, and appear to cycle through these compartments (for review see Le Borgne and Hoflack, 1998 and Molloy et al., 1999). The cytoplasmic domains of M6PR and furin interact with either AP-1 in the TGN, or AP-2 at the plasma membrane, to promote inclusion of these proteins in clathrin-coated vesicles (CCVs) budding from these compartments. Sequences such as the tyrosine motif, the di-leucine motif and the acidic cluster (for review see Kirchhausen et al., 1997) in the cytoplasmic domain of the M6PRs are required for the TGN localization of the M6PRs. Two of these sequence motifs (the tyrosine motif and the acidic cluster) have also been shown to be required for TGN localization of furin. Furthermore, experiments performed with fibroblasts lacking M6PRs have confirmed that the acidic cluster of amino acids, when phosphorylated by casein kinase II (CKII), is important for the recruitment of AP-1 to TGN membranes and the formation of CCVs (Mauxion et al., 1996). Our previous experiments have demonstrated that the formation of a clathrin-coat on ISGs can be attributed to the recruitment of AP-1 by the cytoplasmic domain of furin. Importantly, furin present on the ISGs must be phosphorylated by casein kinase II (CKII) in order to bind AP-1 (Dittié et al., 1997). Since both the tyrosine motif and the acidic cluster are also present in M6PRs we have investigated whether or not AP- 1 recruitment onto ISGs in PC12 cells involves interaction with M6PRs. We have studied the distribution of the M6PRs in ISGs using subcellular fractionation and immunogold labelling of cryosections. We show here that the cytoplasmic domain of the CI-M6PRs can interact with AP-1 in a CKII phosphorylationdependent fashion and that M6PRs are also responsible for recruiting AP-1 to ISGs. In addition, using sequential immunoisolation techniques we have demonstrated that at most 50% of ISGs contain the CI-M6PR, while furin in contrast is present in greater than 80% of ISGs. Together these data establish that ISGs from PC12 cells exhibit a differential distribution of the M6PRs and furin, and both M6PRs as well as furin can recruit AP-1 to ISGs. MATERIALS AND METHODS Reagents Carrier-free [ 35 S]sulfate and 125 I-Protein A were from Amersham- Pharmacia Biotech, UK. Nucleotides, creatine phosphate, and creatine phosphokinase were from Boehringer Mannheim, Germany. Casein kinase II was prepared according to the method of Litchfield et al. (1990). Heparin and horseradish peroxidase (HRP type II) were from Sigma, UK. Tautomycin and microcystin LR were obtained from Calbiochem, UK. Fine chemicals were from Merck, UK, Life Technologies, UK and Sigma, UK. Cells and antibodies PC12 cells, a rat pheochromocytoma cell line (clone 251; Heumann et al., 1983), originally obtained from Dr H. Thoenen (Martinsried, Germany), were maintained as described (Tooze and Huttner, 1992). The mab against bovine γ-adaptin (100/3) (Sigma) was used at a dilution of 1:200. Polyclonal antibodies against furin (van Duijnhoven et al., 1992 and Dittié et al., 1997), and the cytoplasmic domain of the CI-M6PR (STO-49 and 52, prepared using the GST-fusion proteins, see below) were used at dilutions of 1:500-1:1000. Anti-SgII antibody (175) was as previously described (Dittié and Tooze, 1995). The antication dependent mannose-6-phosphate receptor (CD-M6PR) antibody, used at 1:1000 or as previously described (Klumperman et al., 1993), was a kind gift from Dr A. Hille-Rehfeld, Göttingen, Germany. Secondary antibodies were HRP-conjugated sheep antimouse or sheep anti-rabbit (Amersham-Pharmacia Biotech, UK). GST-fusion proteins and expression constructs Different regions of the cytoplasmic domain of the mouse CI-M6PR (Dittié et al., 1997, see Fig. 5A) were cloned via RT-PCR from NIH- 3T3 cell mrna and ligated in frame to the COOH terminus of GST using the pgex-3x system (Amersham-Pharmacia Biotech, UK). The fusion proteins were expressed in BL21 cells and purified using glutathione Sepharose 4B (Amersham-Pharmacia Biotech, UK). The fusion proteins were also used to generate polyclonal antibodies (STO-49 and STO-52). The full-length sequence of human furin preceded by the flag tag (Kodak, UK) was subcloned from hfur/flag pgem7zft (provided by Dr Gary Thomas, Oregon, USA) into pbk/cmv (Stratagene). Transient transfection was carried out using electroporation (Bio-Rad, UK) and cells. Six hours after transfection 5 mm Na butyrate was added to the cultures and they were incubated for an additional 18 hours before being processed for cryomicroscopy. Preparation of PC12 TGN, ISGs, MSGs and bovine brain cytosol PC12 cells were pulse-labelled with [ 35 S]sulfate and chased at 37 C (Tooze and Huttner, 1990; see figure legends for details). ISGs, constitutive secretory vesicle (CSVs), MSGs, and TGN were prepared from either labelled or unlabelled PC12 cells after preparation of a post nuclear supernatant (PNS) by velocity and equilibrium sucrose gradient centrifugation (Dittié et al., 1996). TGN was enriched in velocity gradient fraction 9, CSVs were found in equilibrium gradient (EG) fractions 5 and 6, ISGs in EG fractions 7-9, and MSGs in EG fractions The ISGs used for the binding assay contained ~0.5 mg/ml protein. Equal volumes of each fraction (100 µl from the velocity gradient fractions and 200 µl from the equilibrium gradient fractions) were used for western-blot analysis and the bound antibodies were visualized by ECL (Amersham-Pharmacia Biotech, UK). Bovine brain cytosol was prepared as described (Dittié et al., 1996). Protein concentrations were determined by protein assay (Bio-Rad, UK) using IgG as a standard. Horseradish peroxidase uptake To determine the position of early and late endosomes after sucrose

3 M6PRs are found in a subset of PC12 cell ISGs 3957 gradient centrifugation, PC12 cells were incubated with the endocytic fluid-phase marker HRP. Two 15 cm dishes of subconfluent PC12 cells each were washed twice in DMEM/0.1% HS/0.05% FCS (low serum DMEM) and incubated with 2 mg/ml HRP in low serum DMEM for 7 minutes at 37 C either without chase to label early endosomes or followed by a 30 minute chase with normal growth medium to label late endosomes. A PNS was prepared from each and subjected to velocity and equilibrium centrifugation. The HRP activity in each fraction was determined using o-dianisidine (Marsh et al., 1987). Immunoisolation Immunoisolation of [ 35 S]sulfate labelled ISGs using polyclonal antibodies against CI-M6PR and furin was carried out as described (Dittié et al., 1996), using 100 µl of the ISG containing equilibrium gradient fractions 8 and 9. When two rounds of isolation were performed, the supernatant of the incubation with the first round antibody was transferred to a fresh tube and incubated for the same time with the second round antibody. For competition experiments 0.1 mg/ml GST-CI-M6PR or GST-furin fusion proteins were added before the immunoisolation and were also included at the same concentrations throughout the immunoisolation. The final supernatant was TCA precipitated and resuspended in SDS sample buffer for PAGE analysis. RESULTS Localization of the CI- and CD-M6PRs by subcellular fractionation of PC12 cells Using immunogold labelling techniques M6PRs have been shown to be present in ISGs and to co-localize with AP-1 in both rat endocrine pancreatic β-cells and exocrine parotid and pancreatic cells (Klumperman et al., 1998). To demonstrate that the M6PRs are also present in ISGs of PC12 cells we fractionated PC12 cells (Dittié et al., 1996) and probed each fraction with antibodies specific for either the CI- or CD- M6PR. The CI-M6PR antibody was raised against a GSTfusion protein encoding most of the cytoplasmic domain of the mouse CI-M6PR and was characterized by comparison with a previously published antibody raised against the CI-M6PR (Brown et al., 1995; data not shown). Using these antibodies we found that the distribution of both the CI-M6PR and CD- M6PR in PC12 cell membranes after fractionation of a postnuclear supernatant using a velocity controlled sucrose gradient was complex (Fig. 1A). The CI-M6PR was found in Cell-free assay to reconstitute γ-adaptin binding to PC12 ISGs As previously described in detail (Dittié et al., 1996), ISGs were incubated with varying amounts of bovine brain cytosol in 25 mm Hepes, ph 7.2, 25 mm KCl, 2.5 mm MgOAc (binding buffer) and 100 µm GTPγS. After incubation for 30 minutes at 37 C the reaction was stopped, the ISGs were sedimented and the γ-adaptin bound was quantitated after western blotting using 125 I-Protein A. Competition experiments using GST-fusion proteins were performed as follows: 0.5 mg/ml bovine brain cytosol was preincubated for 15 minutes at 4 C in presence of 0.1 µm tautomycin and 1 µm microcystin LR with GST-fusion proteins (5 µm) either non-phosphorylated or phosphorylated by purified CKII as described by Dittié et al. (1997). This reaction mixture was then used for a standard binding assay. Binding of adaptor components to immobilized furin tails Fusion proteins were purified from BL21 cells expressing the different GST-CI-M6PR constructs (see Fig. 5A). The purified fusion proteins were dialyzed against 50 mm Tris-HCl, ph 7.4, 140 mm KCl, 10 mm MgCl 2 and incubated with or without GST-CKIIα (Dittié et al., 1997) in the presence of ATP and regenerating system for 1 hour at 37 C, then stopped by the addition of heparin. 10 µg of phosphorylated fusion protein were coupled to 30 µl of glutathione Sepharose 4B (50% w/v) for 4 hours at 4 C. Binding of AP-1 to the immobilized fusion proteins was performed as described before (Dittié et al., 1997). Electron microscopy of ultrathin cryosections PC12 cells were prepared for ultrathin cryosectioning and doubleimmunogold labeling according to the Protein A-gold method as previously described (Slot et al., 1991). In short, cells were fixed in a mixture of 2% paraformaldehyde and 0.2% glutaraldehyde in 0.1 M phosphate buffer for 2 hours at room temperature and stored overnight at 4 C in 2% paraformaldehyde. The cells were then washed in PBS/0.2 M glycine, scraped off the dish and pelleted in 10% gelatin in PBS, which was solidified on ice and cut into small blocks. After overnight infiltration with 2.3 M sucrose at 4 C, blocks were picked up in a 1:1 mixture of sucrose and methyl cellulose (Liou et al., 1996). PC12 cells overexpressing furin were incubated overnight with 5 mm butyrate prior to fixation. Sections were viewed in a JEOL 1010 transmission electron microscope operating at 80.0 kv. Fig. 1. Subcellular fractionation reveals that both the CI- and the CD- M6PR are found in fractions which contain ISGs but not MSGs. (A) Fractions from a velocity sucrose gradient of a post-nuclear supernatant obtained from PC12 cells were immunoblotted with antibodies to the CI- and CD-M6PRs. The location of the light and heavy vesicle population, and the TGN is shown. (B) The light or (C) heavy vesicle pool was subjected to equilibrium sucrose centrifugation. The location of the CSV (constitutive secretory vesicle), ISGs and MSGs are shown. Fractions from each gradient were analysed by immunoblotting with antibodies to the CI- and CD- M6PRs. We have reproducibly detected a strong signal for CD- M6PR in the fractions analysed in B although there is apparently very little immunoreactivity in the starting pool (fractions 2-4 in A). The location of the fractions enriched in ISGs and MSGs on the light and heavy vesicle pool gradients is illustrated. The position of the enriched fractions containing ISGs and MSGs was confirmed by analysis of the marker protein SgII (not shown). For all gradients fractions were collected from the top of the gradient (fraction 1 top).

4 3958 A. S. Dittié, J. Klumperman and S. A. Tooze most fractions, with a peak of immunoreactivity in fractions 5-9, while the bulk of the CD-M6PR immunoreactivity was found in fractions 5-9. To increase the resolution of the different compartments present in the velocity gradient fractions, a light and a heavy vesicle pool was prepared containing ISGs and MSGs, respectively, and subjected to equilibrium sucrose gradient centrifugation. ISGs which are present in fractions 2-4 of the velocity gradient, sediment to fractions 7-9 on the equilibrium gradient. MSGs, which are present in fractions 4-7 of the velocity gradient, sediment to a denser sucrose concentration and are found in fractions 9-12 on the equilibrium gradient, with the majority in fractions 11 and 12 (Tooze et al., 1991). After fractionation of the light vesicle pool (fractions 2-4) on equilibrium gradients immunoblotting revealed that both the CI-and CD-M6PR were found in fractions 6-9 (Fig. 1B). Likewise, after fractionation of the heavy vesicle pool (fractions 4-7) on equilibrium gradients the immunoreactive peaks of both the CI- and the CD-M6PR were also found in fractions 6-9 (Fig. 1C). These results demonstrate three points: first, and as expected, the CI-M6PR and the CD-M6PR have similiar fractionation profiles on the gradients described above; secondly, some of both the CI- and the CD-M6PR is present in the fractions containing the ISGs; lastly, neither the CI- nor CD-M6PR co-localized on equilibrium gradients with MSGs despite co-localization on the velocity gradient. Location of endosomal compartments in PC12 cells by subcellular fractionation As both M6PRs are present in early and late endosomes (for review see Hille-Rehfeld, 1995) it was important to determine what precentage of the immunoreactivity specific for the CIand CD-M6PRs on the velocity and equilibrium sucrose gradients is present in endosomal compartments. To identify endosomes we incubated PC12 cells with horseradish peroxidase (HRP) for different periods of time. The cells were then homogenized and fractionated as in Fig. 1, and the HRP activity was assayed for across the gradients. After HRP is internalized for 7 minutes to label early endosomes, the bulk of the early endosomes were present in fraction 5-7 of the velocity controlled sucrose gradient (Fig. 2A). These fractions (5-7) from the velocity gradients, which contain the bulk of the early endosomes were pooled and subjected to equilibrium sucrose gradient centrifugation. The distribution of the HRP activity across the equilibrium gradient demonstrated that the early endosomes were found in fractions 5-8, with a peak in fraction 7 (Fig. 2C). The light vesicle pool (fractions 2-4) of the velocity gradient, which is equivalent to the light vesicle pool in Fig. 1, contained a low level of HRP activity (Fig. 2B). We estimate that only about 20% of all early endosomes are found in the light vesicle fraction from the velocity sucrose gradient. Interestingly, after equilibrium sucrose gradient centrifugation of the light pool the HRP activity was in fractions 5-8 of the equilibrium gradient and exhibited a very similar distribution to the HRP activity found in the pool of velocity gradient fractions 5-7 after equilibrium sucrose gradient centrifugation (compare Fig. 2B and C). This result suggests that the early endosomal compartment is heterogeneous with respect to shape and size as it is widely spread across the velocity gradient but has a uniform density as it sediments to a defined position on the equilibrium A) VG: PNS from PC12 cells HRP activity (absorbance at OD44 5 mn) C) EG: pool of fractions B) EG: pool of fractions 2-4 HRP activity (absorbance at OD44 5 mn) HRP activity (absorbance at OD44 5 mn) min HRP pulse/no chase 7min HRP pulse/30min chase fraction No. Fig. 2. Analysis of the position of endosomal fractions, using HRP activity after internalization for 7 minutes or 7 minutes followed by a 30 minute chase. (A) HRP activity across the velocity sucrose gradient for a PNS from PC12 cells loaded with HRP for 7 minutes or 7 minutes followed by a 30 minute chase. The position of the fractions loaded onto the subsequent equilibrium gradients are shown. (B) Equilibrium sucrose gradient centrifugation of the pool of fractions 2-4 from the velocity sucrose gradient loaded with a PNS of PC12 cells after a 7 minute internalization of HRP. (C) Equilibrium sucrose gradient centrifugation of the pool of fractions 5-7 from the velocity sucrose gradient loaded with a PNS of PC12 cells after a 7 minute internalization of HRP. gradients. From these results it is possible that some of the CIand CD-M6PR immunoreactivity in the fractions enriched for ISGs is present in contaminating early endosomes. However, some CI- and CD-M6PRs are probably also present in ISGs because the peaks of immunoreactivity (Fig. 1B) and HRP activity (Fig. 2B) are not absolutely coincident. A 7 minute internalization of HRP followed by a 30 minute chase revealed

5 M6PRs are found in a subset of PC12 cell ISGs 3959 that the late endosomes and lysosomes rapidly sedimented into the bottom half of the velocity gradient (fractions 6-12 with a peak in fraction 9) well separated from the post-golgi vesicle pool in fractions 2-4 (data not shown). Antibodies to the CI-M6PR can immunoisolate immature secretory granules To determine more precisely what percentage of ISGs have M6PRs, we performed an immunoisolation experiment using fractions 8 and 9 from the equilibrium gradient which are enriched in ISGs. The ISG, isolated from cells labelled with [ 35 S]sulfate for 5 minutes followed by a 15 minute chase, contained [ 35 S]sulfate labelled SgII. After incubation with an anti-ci-m6pr antibody approximately 30% of the total ISGs containing [ 35 S]sulfate labelled SgII could be immunoisolated (Fig. 3A, lane 1, and Table 1). As Fig. 3A, lane 3, and Table 1 show this signal could be competed by pre-incubation of the antibody with the GST-fusion protein containing the cytoplasmic domain of the CI-M6PR (GST-CT87, see Fig. 5). Although sulfate-labelled ISGs were immunoisolated with the anti-ci-m6pr antibody the efficiency of immunoisolation was not very high: over the course of several experiments we observed that a maximum of 45% of the total sulfate-labelled ISGs was immunoisolated with the anti-ci-m6pr antibody (Table 1). The results obtained with the anti-ci-m6pr antibody were in contrast with those previously obtained using an anti-furin antibody with which, in similar immunoisolation experiments, greater than 85% of ISGs could be immunoisolated (Table 1 and Dittié et al., 1997). It was possible that the anti-ci-m6pr antibody did not recognize the cytoplasmic domain of the CI- M6PR as efficiently as the cytoplasmic domain of furin was recognized by the anti-furin antibody. To determine if the anti- M6PR antibody was efficiently binding the cytoplasmic domain of the M6PR in membranes we performed an immunoisolation experiment with an enriched fraction of TGN membranes, obtained from cells labelled with a 5 minutes pulse of sulfate. We found that the anti-m6pr receptor antibody was able to immunoisolate 84% of the total [ 35 S]sulfate present in the TGN fraction. In addition, the GST- CT87 fusion protein was able to compete the antibody binding to background levels (data not shown). These results suggest that the anti-m6pr antibody is able to recognize the cytoplasmic domain of the M6PR under the conditions used Fig. 3. Immunoisolation of ISGs containing [ 35 S]sulfate labelled SgII with antibodies directed against the cytoplasmic domain of CI-M6PR and furin. (A) ISGs were obtained from PC12 cells pulse-labelled for 5 minutes and chased for 15 minutes following velocity and equilibrium sucrose gradient centrifugation. ISGs (total sample lane 5) were incubated with an antibody to the CI-M6PR (lanes 1 and 2) or anti-ci-m6pr antibody preincubated with the GST-fusion protein used to immunize rabbits (lane 3 and 4). The material immunoisolated bound to the Staph A (P) was solubilized in SDS-sample buffer while the unbound (S) was TCA precipitated and solubilized in SDS-sample buffer. (B) Sequential immunoisolation was performed on the ISG fraction by incubation first (P1) with anti-ci-m6pr antibody (lanes 1, 4 and 7) or anti-furin antibody (lane 10 and 12). The unbound supernatant was then collected and subjected to a second round of immunoisolation (P2) using preimmune sera (lane 2), anti-ci-m6pr antibody (lanes 5 and 13), or anti-furin antibody (lane 8) or analysed directly as above. Positions of SgII and chromogranin B (CgB) are indicated.

6 3960 A. S. Dittié, J. Klumperman and S. A. Tooze Table 1. Quantitation of immunoisolation of ISGs with anti-m6pr and anti-furin antibodies Single round of immunoisolation: No. % total % total in Antisera of exp. in pellet supernatant Pre-immune 4 8.3±6.7 92±6.7 Anti-M6PR 5 29±4.4 72±4.4 Anti-M6PR ±2.4 98±2.4 GST-M6PR Anti-furin 3 85±1.0 15±1.0 Sequential round of immunoisolation: No. of % total in % total in % total in Antisera Exp. 1st pellet 2nd pellet supernatant 1 rd. anti-m6pr rd. pre-immune 1 rd. anti-m6pr rd. anti-m6pr 1 rd. anti-m6pr rd. anti-furin 1 rd. anti-furin 3 99± ± rd. anti-m6pr Either a single or sequential round of immunoisolation was performed with the indicated antibodies. For single immunoisolation experiments the antigenantibody complex was recovered with Staph A and is referred to as the pellet, while the unbound material was found in the supernatant. For sequential immunoisolation experiments the unbound supernatant from the first round was collected and used as the starting material for the 2nd round of immunoisolation. The percentages shown are derived from a total obtained by addition of the 35 S radioactivity present in SgII found in the pellets and supernatants and the errors are expressed as standard deviation. for immunoisolation, and does so as efficiently as the anti-furin antibody. An alternative explanation for the low immunoisolation efficiency of the ISGs with the anti-m6pr antibody is that the ISGs are a heterogeneous population with a variable amount of the CI-M6PR and furin in their membrane. ISGs which do not have sufficient amounts of the CI-M6PR would not be immunoisolated with the anti-ci MPR antibody, but would be efficiently immunoisolated with the anti-furin antibody if sufficient amounts of furin was present. To address this we performed a sequential immunoisolation experiment where the ISGs were incubated with the anti-ci-m6pr antibody, the antibody-antigen complexes were immobilized on Staph A and the supernatant was supplemented with the anti-furin antibody, or, in reverse order, the anti-furin antibody was followed by the anti-ci-m6pr antibody (Fig. 3B). As in Fig. 3A, incubation with the anti-ci-m6pr antibody resulted in approximately 30-45% of the labelled ISGs (Fig. 3B, lanes 1, 4 and 7, and Table 1) binding to the Staph A. A subsequent readdition of the anti- CI-M6PR antibody to the unbound fraction resulted in no further recovery of labelled ISGs (Fig. 3B, lane 5, and Table 1) compared to pre-immune serum (Fig. 3B, lane 2, and Table 1). However, addition of the anti-furin antibody to the unbound supernatant fraction resulted in the recovery of virtually all the remaining sulfate labelled ISGs (Fig. 3B, lane 8, and Table 1). To control for the efficiency of the sequential immunoisolation experiment we performed additional experiments with the anti-furin antibody. After immunoisolation with the anti-furin antibody in which greater than 85% of the ISGs (Fig. 3B, lane 10, and Table 1) were Fig. 4. Ultrathin cryosections showing Golgi (G) regions of wildtype (A,B) and furin overexpressing (C-E) PC12 cells, illustrating the presence of CD-M6PR (B) and furin (C-E) in secretory granules (S). (A) Immunogold labelling of SgII identified ~150 nm diameter organelles with a partially condensed content as secretory granules. Note that a dense staining on the cytoplasmic side of the membrane (arrows in A, B, C, D, and E), indicative for the presence of clathrin, covered large areas of some secretory granules, thereby defining them as ISGs. (B) SgII: 10 nm gold; CD-M6PR: 15 nm gold. The closed arrowheads point to CD-M6PR staining in SgII positive secretory granules (S). Also CD-M6PR negative secretory granules (open arrowhead) were regularly observed in the Golgi area. (C and D) SgII: 10 nm gold; furin 15 nm gold. Furin is localized in a SgIIpositive structures similar to a condensing vacuole (arrows) and secretory granules (S), some of which are clathrin coated (arrows). E) Furin (10 nm gold) and CD-M6PR (15 nm gold) sometimes colocalized to the same TGN vesicle (arrowheads), some of which were clathrin coated (upper vesicle). L, lysosome. Bars, 200 nm. immunoisolated, addition of the anti-ci-m6pr antibody to the supernatant failed to recover any additional labelled ISGs (Fig. 3B, lane 13, and Table 1). The signal obtained with the antifurin and anti-ci-m6pr antibodies used either in the first and second rounds could be competed by the addition of the respective GST-fusion proteins used to raise the antisera (see Fig. 3A and Dittié et al., 1997). These results demonstrate that while only 45% at most of the sulfate labelled ISGs could be immunoisolated with anti-ci-m6pr antibodies, most (greater than 85%) could be immunoisolated with the anti-furin antibody. Furthermore, the ISGs which were not bound to the Staph A via the anti-ci-m6pr antibodies could be immunoisolated with the anti-furin antibody. This suggests that while furin is present on virtually all ISGs, at least half of the ISGs do not have sufficient amounts of CI-M6PR to allow them to be immunoisolated. We have not investigated the distribution of the CD-M6PR because the anti-cd-m6pr antibody did not efficiently immunoisolate any membranes (even control TGN membranes) under the conditions used for immunoisolation. Finally, as an additional control for the efficiency of the immunoisolation experiments we used the same anti-ci-m6pr antisera to immunoisolate early endosomes, labelled after a 5 minute HRP internalisation, prepared from velocity gradients (pool of fractions 6 and 7). Early endosomes could be immunoisolated with both the anti-furin antibodies and anti- CI-M6PR antibodies (data not shown). Under conditions where quantative immunoisolation is obtained, comparison of the recovery of early endosomes revealed that with the anti-furin antibody approximately 2-fold more early endosomes were isolated than with the anti-ci-m6pr antibody (data not shown). Likewise with the anti-furin antibodies we were able to immunoisolate approximately 2-fold more ISGs than with the anti-ci-m6pr antibodies (see Fig. 3). These results imply that in PC12 cells furin is more abundant than the CI-M6PR in both early endosomes and ISGs. Localization of the CD-M6PR to immature secretory granules To confirm the co-localization of the M6PR and furin on ISGs we performed immunogold labelling of cryosections of PC12 cells (Fig. 4). Furin is present in very low amounts in cells (Creemers et al., 1993) and not readily detectable by

7 M6PRs are found in a subset of PC12 cell ISGs 3961 Fig. 4

8 3962 A. S. Dittié, J. Klumperman and S. A. Tooze immunogold labelling techniques (J. Klumperman, unpublished results) so to perform this analysis it was necessary to transiently transfect PC12 cells with furin. To unequivocally define ISGs in the Golgi area, we first localized SgII, a secretory granule marker protein, to organelles with a partially condensed dense core content (Fig. 4A). Large areas of some of these ISGs were coated with clathrin, identified by its characteristic appearance, consistent with our previous localization of γ-adaptin on isolated ISGs. Previous immunogold labelling studies (Klumperman et al., 1993) have shown that CI-M6PR and CD-M6PR co-localize to the same TGN membranes in HepG2 cells. We used anti-cd-m6pr antibodies to study the subcellular distribution of M6PRs in situ in PC12 cells. Double immunogold labelling of SgII and CD-M6PR (Fig. 4B) or SgII and furin (Fig. 4C and D) showed the presence of both CD-M6PR and furin in ISGs, some of which clearly exhibited a clathrin coat. Additional labelling in the Golgi area was found on TGN membranes; the CD-M6PR and furin were co-localized in clathrin-coated and in nonclathrin coated vesicles (Fig. 4E) that are likely to mediate transport between TGN and endosomes. These results demonstrate that the CI-M6PR and CD- M6PR are found in ISGs along with furin, and some of these ISG structures are coated with clathrin. Interestingly, our data indicate that not all ISGs contain the CI-M6PR. As we have previously demonstrated that furin is involved in AP-1 recruitment to ISGs (Dittié et al., 1997) and others have shown that M6PRs are key components for recruitment of AP-1 to the TGN (for review see Le Borgne and Hoflack, 1998) we were interested in role the M6PR plays in coat recruitment on ISGs. Preparation and phosphorylation of GST-CI-M6PR fusion proteins To determine if M6PRs are involved in clathrin coat recruitment to ISGs we used the established in vitro assay (Dittié et al., 1996) and fusion proteins containing the cytoplasmic domain of the CI-M6PR to compete for AP-1 binding. To develop a reagent containing the cytoplasmic domain of the CI-M6PR which could bind to AP-1 in cytosol we constructed a series of GST-fusion proteins containing different lengths of the cytoplasmic domain of the mouse CI-M6PR (Fig. 5A). The CI- M6PR cytoplasmic sequence has two stretches of acidic amino acids which are sites for CKII phosphorylation in vivo and in vitro, as well as a tyrosine-based sorting motif and a di-leucine sorting motif. We chose to express fusion proteins containing only the sites which interact with AP-1, that is the two CKII sites and the di-leucine motif. Fusion proteins were prepared which contained all three sites (GST- CT87), just the two CKII sites (GST- 4), only the N-terminal CKII site (GST- 9), or the C-terminal CKII site and the di-leucine site (GST-CT13). Bacteria expressing these constructs produced fusion proteins of the correct molecular mass which were purified on glutathione-sepharose (Fig. 5B). Expression of the GST-CT87 construct in bacteria gave rise to three bands, the full length fusion protein having the lowest mobility by SDS-PAGE (Fig. 5B, lane 3) while the other constructs produced a single band of the correct molecular mass (Fig. 5B). A) CI-M6PR (mouse): SP Y S S LL N luminal TM cytoplasmic C GST CT13 (aa ): N GST SLL C GST CT87 (aa ): N GST S S LL C GST 4 (aa ): N GST S S C Fig. 5. Construction and expression of truncated cytoplasmic domains of mouse CI-M6PR as GST fusion proteins. (A) Scheme of constructs. Various sections of the C-terminal region of the cytoplasmic domain of the mouse CI-M6PR were fused to GST to create the GST-fusion proteins illustrated. (B) These GST-fusion proteins were expressed and purified on glutathione Sepharose as detailed in Materials and Methods and analysed by Coomassie blue staining. Except for CT87, the major band of all constructs expressed was the correct size fusion protein. The full length fusion protein for CT87 was the slowest migrating band at 44 kda which corresponds to the predicted molecular mass. (C) The purified GST fusion proteins were incubated with recombinant CKIIα to test their ability to be phosphorylated. The full length fusion proteins, except for GST, were in each case phosphorylated by CKIIα. The arrowhead indicates the CKIIα fusion protein which is autophosphorylated. GST 9 (aa ): N GST S 2473 C

9 M6PRs are found in a subset of PC12 cell ISGs 3963 As AP-1 binding to the cytoplasmic domain of the CI-M6PR depends on CKII phosphorylation of the receptor (Le Borgne et al., 1993), we next tested if the GST-CI-M6PR fusion proteins could be phosphorylated by CKII by incubating the fusion proteins with recombinant CKIIα and [ 32 P]ATP (Fig. 5C). The full length form of all four GST-fusion proteins but not GST itself could be phosphorylated by recombinant CKIIα. The increased sensitivity obtained using 32 P revealed that the all the GST-fusion protein preparations contained low amounts of truncated fusion proteins but the full-length protein was the major form. Fig. 6. Optimal AP-1 binding to CI-M6PR in vitro requires CKII phosphorylation of Ser (A) GST or GST-fusion proteins bound to glutathione Sepharose were incubated with or without recombinant CKIIα at 37 C for 60 minutes, then incubated in bovine brain cytosol for an additional 30 minutes at 37 C. AP-1 bound to beads was assayed by immunoblotting with mab100/3. (B) Phosphorylation dependent stimulation of AP-1 binding to GSTfusion proteins is expressed relative to the unphosphorylated GSTfusion proteins. Values are obtained from 7 independent experiments for GST and GST-CT87 and 4 independent experiments for GST- CT13, 4 and 9. (C) GST and GST-fusion proteins furin and GST- CT87 were preincubated with or without purified CKIIα as above. Bovine brain cytosol was preincubated with the GST and the GSTfusion proteins for 15 minutes before incubation with ISGs in the binding assay. The reduction in the amount of bovine AP-1 recruited to the ISGs is expressed relative to the control, unphosphorylated GST. The error bars represent standard deviation of the mean. CKII phosphorylated CI-M6PR cytoplasmic domains can bind AP-1 and compete AP-1 recruitment to ISGs The GST-M6PR fusion proteins containing variable lengths of the cytoplasmic tail of CI-M6PR (see Fig. 5A) were phosphorylated with CKII, immobilized to glutathione- Sepharose and then incubated with bovine brain cytosol. AP- 1 binding was then assayed by immunoblotting with the bovine specific γ-adaptin mab 100/3. As shown in Fig. 6A, the immobilized fusion proteins GST-CT13, GST-CT87, and GST- 4 could efficiently bind AP-1 after phosphorylation by CKII. These constructs all have the C-terminal CKII site consisting of Ser Interestingly, neither the unphosphorylated fusion proteins, nor the phosphorylated GST- 9, which contains only the N-terminal CKII site close to the transmembrane region, bound AP-1. Although there was some binding of AP-1 to GST with and without incubation with CKII, there was a significant phosphorylation-dependent stimulation of AP-1 binding for GST-CT87, as well as slight stimulation for GST-CT13 and GST- 4 (Fig. 6B). These results suggest that the CKII site at the C-terminal end of the cytoplasmic domain of CI-M6PR, and in particular Ser 2474 is required for the efficient binding of AP-1. To determine if the CI-M6PR was involved in recruitment of AP-1 to ISGs from PC12 cells we performed a competition experiment using the cell-free binding assay (Dittié et al., 1996). This assay measures ARF1-dependent recruitment of exogenous bovine AP-1, present in bovine brain cytosol, onto ISG membranes. Bovine brain cytosol was preincubated with the phosphorylated or non-phosphorylated GST-CT87 fusion protein, and then the entire mixture was added to the assay containing ISGs in the presence of GTPγS (Fig. 6C). Recruitment of exogenous AP-1 was then assayed with mab 100/3 (Ahle et al., 1988). As controls we used GST or a GSTfurin fusion protein, previously characterized (Dittié et al., 1997), either with or without prior incubation with CKII. GST and GST after incubation with CKII were unable to compete for AP-1 binding to ISGs. The phosphorylated GST-furin fusion protein which has been previously shown to compete for binding of AP-1 to ISGs (Dittié et al., 1997) inhibited binding about 50%, while the phosphorylated cytoplasmic domain of the CI-M6PR (GST-CT87) was effective in reducing binding by about 70%. Both the unphosphorylated GST-furin and GST- M6PR fusion proteins did not significantly inhibit binding. These results demonstrate that after CKII phosphorylation the CI-M6PR cytoplasmic domain can effectively compete for AP- 1 recruitment to ISGs and reduces AP-1 binding below that observed with phosphorylated furin fusion protein. DISCUSSION There is a growing body of evidence that in both endocrine and exocrine cells, proteins are removed from the regulated secretory pathway by vesicles which originate from maturing secretory granules. Some soluble proteins, for example C- peptide which is derived from proinsulin in endocrine pancreatic cells (Arvan et al., 1991), are removed from the ISG and secreted by constitutive-like secretion (for review see Arvan and Castle, 1998) into the extracellular space, while other soluble molecules, such as lysosomal enzymes, are

10 3964 A. S. Dittié, J. Klumperman and S. A. Tooze transported from the ISG to the endosome. While the transport of lysosomal enzymes from the ISG to the endosome depends on M6PRs (Kuliawat et al., 1997), it is not clear how C-peptide is removed from the ISG, and if it occupies the same vesicles as the M6PR and lysosomal enzymes. Furthermore, the vesicular carrier for other molecules such as furin, which is found in ISGs but not MSGs, is not known although furin in the ISG can bind AP-1 (Dittié et al., 1997). These results suggest either that vesicles budding from the ISG carry a variety of cargo molecules or that there is more than one exit route from the ISG. The existence of multiple exit routes from the ISG is supported by data suggesting that there is a recycling pathway from the ISG back to the TGN, distinct from the constitutive-like secretory vesicles, to retrieve molecules such as the membrane form of peptidylglycine α-amidating monooxygenase (Milgram et al., 1994). To gain further insight into the process of vesicular transport from the ISG during secretory granule maturation, we have focused on trying to understand the role of AP-1 containing clathrin-coats in this process. Previous results using immunogold labelling suggest that the M6PR (Klumperman et al., 1998) uses the AP-1 coat-formation machinery to be removed from ISGS by CCVs. Sequences in the cytoplasmic domain of M6PRs that are implicated in AP-1 binding to TGN membranes, as well as the TGN localization of M6PRs, are the di-leucine motif, the tyrosine motif, and an acidic cluster of amino acids constituting a CKII phosphorylation site, which when phosphorylated may be the most important for high affinity binding of AP-1 (for review see Le Borgne and Hoflack, 1998). Similar motifs in furin are likewise thought to bind AP-1 in TGN membranes (Jones et al., 1995; Schäfer et al., 1995; Takahashi et al., 1995). As phosphorylation of the acidic cluster of amino acids in furin is an important signal for recruitment of AP-1 to ISGs (Dittié et al., 1997), and since furin is absent from MSGs, one can assume that furin is removed by AP-1 containing CCVs from the ISGs. While the CCVs containing the M6PR most probably go to the endosome, it is more difficult to make predictions about the final destination of the furin containing CCVs. One recent model suggests that the phosphorylated CKII site modulates retrieval of furin from the ISG to the TGN (Molloy et al., 1999) which raises several additional questions about the mechanisms of transport of other proteins out of the ISG, and in particular the M6PR and lysosomal enzymes. Does the M6PR recently shown to be present on ISGs and co-localized with AP-1 (Klumperman et al., 1998) also use the CKII phosphorylated acidic clusters in the cytoplasmic domain to bind AP-1 in the ISGs? Are furin and the M6PR found in the same AP-1 containing clathrin coated vesicles? and finally does the M6PR go back to the TGN from the ISG or does it go directly to an endosomal compartment? The recycling of M6PRs to the TGN seems most unlikely in view of recent data concerning the co-localization of syntaxin 6 and the CD-M6PR in cells with a regulated secretory pathway (Klumperman et al., 1998) and raises the possibilities that either furin and the M6PR are in different vesicles or that furin is also targeted to endosomes from the ISG and not to the TGN. As reported here the CI-M6PR cytoplasmic domain can recruit AP-1 from cytosol and this recruitment requires prior phosphorylation by CKII. AP-1 binding to ISGs can be inhibited by preincubation of the cytosol with the phosphorylated GST-CI-M6PR fusion protein containing both acidic clusters, the sites for CKII phosphorylation, and the dileucine motif. The GST-fusion proteins GST-CT13 and GST- 4 bind AP-1 better than GST- 9 (Fig. 6A), which does not contain the C-terminal CKII site, implying that the phosphorylation of the acidic clusters, and in particular the C- terminal phosphorylated Ser 2474 residue, is the most important signal for AP-1 binding. Furthermore, deletion of the dileucine motif resulted in a decrease in the amount of AP-1 bound in a manner which was stimulated by phosphorylation (see Fig. 6B) but did not affect the total amount bound (see Fig. 6A). Others (Johnson and Kornfeld, 1992; Mauxion et al., 1996) have reported that AP-1 binding requires the presence of di-leucine motifs and Chen et al. (1997) have demonstrated that di-leucine motifs are required for sorting of cathespin D by the CI-M6PR. Both in vivo, using furin as the substrate (Dittié et al., 1997) and in vitro, with both furin and the CI- M6PR as substrate, we have found that phosphorylation of the acidic cluster is the most important signal for recruitment of AP-1 to ISGs. Phosphorylation of the acidic cluster may regulate AP-1 binding to ISGs, and the removal of proteins from ISGs in CCVs. For example, if during maturation the dense core underwent a physical rearrangement, as documented for insulin (Michael et al., 1987), it might be advantageous only to remove proteins after this rearrangement had occurred thus ensuring that all soluble non-granule proteins were excluded from the dense-core. AP-1 recruitment to ISGs was inhibited in competition experiments using the phosphorylated cytoplasmic domains of the CI-M6PR. These results extend our previous results with furin (Dittié et al., 1997) and the results of Hoflack and colleagues (Le Borgne et al., 1993) which showed that CKII phosphorylation of the recombinant bovine CI-M6PR cytoplasmic domain GST-fusion protein was required to competitively inhibit AP-1 binding to TGN membranes. In addition, our results demonstrate that the phosphorylated CI- M6PR cytoplasmic tail can compete AP-1 recruitment to ISGs more effectively than furin. This probably reflects the higher affinity of AP-1 for the phosphorylated CI-M6PR cytoplasmic domain rather than the amount of furin and M6PR in the ISGs. Although the amount of M6PR present in the ISGs is difficult to estimate we can detect both the CI- and CD-M6PR in the ISG fraction by subcellular fractionation and the CD-M6PR by immunogold labelling on cryosections. Interestingly, using immunoisolation we found that at most only 50% of the ISGs had significant enough amounts of CI-M6PR to be immunoisolated, whereas most of the ISGs contained furin. Our results raise the possibility that there is a heterogenous populations of clathrin-coated vesicles forming from the ISGs in PC12 cells. Clathrin-coated vesicles could form from ISGs containing furin alone, or containing both furin and the M6PR. However, in PC12 cells ISGs can undergo homotypic fusion as part of their maturation to MSGs (Tooze et al., 1991; Urbé et al., 1998) and this homotypic fusion is thought to be a prerequisite for clathrin-coated vesicle formation (Tooze, 1991). If CCV formation is restricted to a late stage of granule maturation, homotypic fusion may serve to ensure that all ISGs have a uniform composition and that the CCVs emanating from the ISGs will be a homogenous population. Irregardless of the composition of the ISG derived CCVs it is most likely that these CCVs will go to the endosome. The alternative possibilities that