A Phosphorylation Site Regulates Sorting of the Vesicular Acetylcholine Transporter to Dense Core Vesicles

At least two distinct populations of secretory vesicles mediate the regulated exocytotic release of neurotransmitters. Synaptic vesicles (SVs) appear almost exclusively in nerve terminals where they cluster at the synaptic cleft (Calakos and Scheller 1996). In contrast, large dense core vesicles (LDCVs) appear throughout the cell, and release transmitter more slowly and in response to different stimuli than SVs (De Camilli and Jahn 1990; Martin 1994). Although SVs are generally considered to contain classical transmitters, whereas LDCVs store neural peptides, both types of regulated secretory vesicles contain monoamines, indicating the potential for release of a single transmitter through two distinct mechanisms (Thureson-Klein 1983; Bruns and Jahn 1995). Since classical transmitters, including monoamines, are synthesized in the cytoplasm, the subcellular location of the transporters required for packaging them into secretory vesicles determines the site of vesicular storage, and hence, the mode of exocytotic release.

Secretory vesicles exhibit four distinct neurotransmitter transport activities, including one for monoamines, a second for acetylcholine (ACh), a third for γ-aminobutyric acid (GABA), and a fourth for glutamate (Liu and Edwards 1997b). All of these activities depend on the H⁺ electrochemical gradient across the vesicle membrane, and exchange lumenal protons for cytoplasmic transmitter (Schuldiner et al. 1995). Pharmacologic manipulation of vesicular transport indicates the importance of these activities for behavior. The antihypertensive drug reserpine inhibits vesicular monoamine transport and induces a syndrome resembling depression (Frize 1954). In addition, amphetamines induce efflux from vesicular monoamine stores (Sulzer et al. 1995) and can cause psychosis. Thus, changes in vesicular transport activity have the potential to influence behavior, but the extent to which they undergo regulation has remained unknown.

Molecular cloning has begun to identify the proteins responsible for neurotransmitter transport into secretory vesicles. One family of proteins includes two vesicular monoamine transporters (VMATs) and the vesicular acetylcholine transporter (VAChT). VMAT1 is expressed principally in peripheral, nonneural tissues and VMAT2 is expressed by central monoamine neurons (Varoqui and Erickson 1997; Liu and Edwards 1997b). Consistent with the observed release of monoamines from both SVs and LDCVs, the VMATs occur on multiple populations of secretory vesicles. In brain, VMAT2 localizes preferentially to LDCVs, but also resides on SVs, as well as tubulovesicular structures in the cell body and dendrites of central dopamine neurons (Nirenberg et al. 1995, Nirenberg et al. 1996). Similarly, VMAT2 resides predominantly on LDCVs rather than the synaptic-like microvesicles (SLMVs) present in PC12 cells (Liu et al. 1994; Erickson et al. 1996). In contrast, VAChT localizes predominantly to SVs in brain (Gilmor et al. 1996; Weihe et al. 1996) and to lighter membranes, including the SLMVs in PC12 cells (Liu and Edwards 1997a; Varoqui and Erickson 1998). However, VAChT also occurs at lower levels on LDCVs in PC12 cells (Liu and Edwards 1997a) and in neurons (Lundberg et al. 1981; Agoston and Whittaker 1989). Thus, two closely related vesicular neurotransmitter transporters localize to distinct populations of secretory vesicles that differ in their mode of release.

To understand how VMAT2 and VAChT localize to different secretory vesicles, we have sought to identify their sorting signals. As integral membrane proteins, the transporters presumably depend on signals in their cytoplasmic domains that interact with a cytosolic sorting machinery. We have found previously that dileucine motifs located COOH-terminal to transmembrane domain (TMD) 12 of both VMAT2 and VAChT (see Fig. 1 A) mediate internalization of the transporters from the plasma membrane (Tan et al. 1998). Dileucine motifs implicated in the endocytosis of certain other proteins require the presence of acidic residues at positions −4 and −5 relative to the two leucines (Pond et al. 1995; Dietrich et al. 1997) and the VMATs contain highly conserved glutamates at both of these positions. However, replacement of these glutamates with alanine does not impair the internalization of VMAT2 (Tan et al. 1998). Since dileucine motifs mediate trafficking at multiple sites in the secretory pathway, these residues may influence other events, distinct from endocytosis. Although VAChT, like the VMATs, contains a glutamate at −4 relative to the dileucine, it contains a serine at −5, raising the possibility that the charge of this residue accounts for the localization of the transporters to distinct populations of secretory vesicles. In addition, the presence of a serine at −5 relative to the dileucine in VAChT suggests that phosphorylation of this residue may influence the sorting of the transporter. Indeed, phosphorylation of serine residues at −5 relative to the dileucine motifs in CD4 (Shin et al. 1991) and CD3γ (Dietrich et al. 1994) dramatically influences the trafficking of these proteins.

We now report that the serine at −5 relative to the dileucine motif in VAChT (Ser-480) undergoes phosphorylation by a calcium-dependent isoform of protein kinase C (PKC). Replacement of Ser-480 by alanine, which prevents phosphorylation, reduces the expression of VAChT on LDCVs relative to the wild-type protein. In contrast, substitution of Ser-480 by glutamate, to mimic the sequence of VMAT2 at this position and the phosphorylation event, increases the expression of VAChT on LDCVs. Consistent with the importance of acidic residues upstream of a dileucine motif in sorting to LDCVs, replacement of Glu-478 and -479 upstream of the dileucine-like motif in VMAT2 reduces localization to LDCVs. The results provide some of the first information about signals involved in sorting to different classes of secretory vesicles. In addition, they suggest that phosphorylation regulates the targeting of VAChT to LDCVs, providing a potential mechanism to regulate transmitter release.

All cultures were grown in media containing penicillin and streptomycin at 37°C in 5% CO₂. Monkey kidney COS cells were maintained in DME containing 10% calf serum. PC12 cells were maintained in DME containing 5% calf serum and 10% equine serum (Greene and Tischler 1976). For transient transfection, COS cells were electroporated with 15–20 μg DNA using a BioRad Gene Pulser (Peter et al. 1994). Stable PC12 cell transfectants expressing hemagglutinin (HA)-tagged transporters were prepared by electroporation of wild-type and mutant VAChT cDNAs in a modified pCDNA3 expression vector (see below). After selection in G418, isolated colonies (∼80 per construct) were picked and screened for expression by immunofluorescence. At least two independently isolated cell lines were analyzed for each VAChT construct.

We used the method of Kunkel et al. 1991 to introduce epitope tags and point mutations into the VAChT cDNA. In brief, we prepared single-stranded uracil-containing template in the dut⁻, ung⁻ CJ236 strain of E. coli, annealed the mutagenic primer, extended with T7 DNA polymerase, and transformed the reaction product into XL1-Blue. Oligonucleotides containing silent restriction enzyme sites to facilitate mapping were synthesized using an Applied Biosystems PCR-Mate. The mutations were verified by sequence analysis and a small fragment containing the desired mutation transferred into the wild-type VAChT cDNA in the expression vector pCDNA3 (Invitrogen) with an RSV promoter replacing the original CMV promoter. For expression of bacterial glutathione S-transferase (GST) fusion proteins (Smith and Johnson 1988), cDNA fragments of wild-type and mutant VAChT were cut at VAChT nucleotide 1,473 using StuI and the multiple cloning site of the parent construct, and the 0.4-kb fragment inserted in frame into the SmaI site of the pGEX-5X-1 bacterial expression vector (Pharmacia Biotech).

For metabolic labeling with ³²P_i, cells were washed three times in media lacking phosphate, and incubated for 4 h at 37°C in the presence of 0.5–1.0 mCi/ml ³²P_i (ICN). For kinase stimulation experiments, pharmacological agents at the concentrations indicated in the text were added from stock solutions by rapid mixing during the last 15 min of labeling. After labeling, the cells were washed in ice-cold PBS and harvested by scraping into 1 ml 50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 50 mM NaF, 0.2 mM Na₃VO₄, 10 mM EDTA, 5 mM EGTA, 10 μg/ml PMSF, 2 μg/ml leupeptin, 2 μg/ml pepstatin, 1% vol/vol Triton X-100 detergent (homogenization buffer, HB). After removal of the nuclei and cell debris by centrifugation at 14,000 g for 5 min at 4°C, SDS was added to the supernatant to a final concentration of 0.2%. For immunoprecipitation, the mixture was incubated overnight at 4°C with either the polyclonal antiserum to VAChT prebound to protein A–Sepharose (Sigma Chemical Co.) or with an mAb to HA (Berkeley Antibody Co.) prebound to protein G–Sepharose. Immune complexes were washed 4 times in HB containing 0.2% SDS, resuspended in 2× Laemmli sample buffer and the proteins separated by electrophoresis through 10% polyacrylamide. The gels were then fixed in 10% acetic acid, 50% methanol, dried, and submitted to autoradiography.

Proteins were separated by electrophoresis through polyacrylamide containing SDS and transferred to nitrocellulose or PVDF using either a semidry or liquid transfer apparatus. The filters were then incubated in PBS containing 0.1% Tween 20 and 5% nonfat dry milk, and stained in the same buffer with either a primary rabbit polyclonal antibody to VAChT at 1:1,000 (Liu and Edwards 1997a) or an mAb to HA (Berkeley Antibody Co.) at 1:1,000, followed by the appropriate secondary antibody conjugated to HRP (1:1,000). The complex was then visualized by chemiluminescence (Pierce Chemical Co.), and the exposed films scanned using a UMAX flatbed scanner and Adobe Photoshop for Macintosh. The digitized images were quantitated using NIH Image 1.61 software.

Phosphoamino acid analysis was performed as previously described (Krantz et al. 1997). In brief, extracts prepared from cells metabolically labeled with ³²P_i were immunoprecipitated with the polyclonal antiserum to VAChT as described above, and the immunoprecipitates separated by electrophoresis through polyacrylamide. After autoradiography, the radiolabeled band was excised from the gel, rehydrated in 50 mM ammonium bicarbonate, and the protein was eluted overnight in 0.2% SDS, 2% β-mercaptoethanol. The eluate was precipitated with 20% TCA and partially hydrolyzed by boiling in 5.7 M HCl for 60 min. The hydrolysate was washed first with distilled water, then with 7.8% acetic acid, 2.2% formic acid (pH 1.9 buffer), resuspended in 10 μl pH 1.9 buffer containing phosphoamino acid standards, and spotted onto thin layer cellulose plates. Electrophoresis was performed at 4°C using pH 1.9 buffer for the first dimension and 5% acetic acid, 0.5% pyridine (pH 3.5 buffer) for the second dimension. The standards were then stained with ninhydrin and the plates submitted to autoradiography.

To express GST fusion proteins, Escherichia coli were grown overnight in 1.6% tryptone, 1% yeast extract, 0.5% NaCl (2× YTA media) at 37°C, and induced in 0.1 mM isopropyl β-d-thiogalactoside (IPTG) for an additional 3–6 h at room temperature. Bacteria were then pelleted, resuspended in PBS, and disrupted by vigorous sonication for 1–2 min at 0°C. Cell debris was removed by centrifugation at 14,000 g and the resulting supernatant was either used immediately or stored at −70°C. To partially purify the fusion protein, the cleared extract was bound to glutathione-Sepharose beads for 20 min at room temperature in PBS, washed twice in PBS, and once in either 20 mM Tris, pH 7.5, 0.5 mM DTT, 10 mM MgCl₂ (lysate kinase buffer, LKB), or 50 mM MES, pH 6.0, 1.25 mM EGTA, 12.5 mM MgCl₂ (PKC buffer, PKCB). Aliquots of fusion protein (∼1 μg) bound to glutathione-sepharose (10–20 μl bed vol) were then incubated with either 1 μl postnuclear supernatant (PNS; ∼10 μg total protein) from COS or PC12 cells (see below) for 20 min at 30°C in LKB containing 2 mM CaCl_2, unless otherwise indicated, and 200 μM ATP and γ[³²P]ATP, to a final specific activity of 500 μCi/μmol; or with 20 ng (0.02 units) of the catalytic fragment of PKC from rat brain (PKC-M; Calbiochem) for 20 min at 30°C in PKCB containing 125 μM ATP and γ[³²P]ATP, to a final specific activity of 5,000 μCi/μmol. For experiments using kinase inhibitors, the ATP concentration was reduced to 1 μM. The reactions were stopped by washing with cold PBS containing 15 mM EDTA, the phosphorylated proteins were eluted with 20 μl of 10 mM glutathione in 50 mM Tris-HCl, pH 8.0, and the eluates were added to an equal volume of 2× Laemmli sample buffer before separation by electrophoresis through 12.5% polyacrylamide. The gels were fixed and stained with Coomassie blue, and then dried and submitted to autoradiography.

Preparation of a PNS was performed as previously described (Krantz et al. 1997). In brief, cells were harvested using either trypsin or a cell scraper after one freeze–thaw cycle. The cells were resuspended in 10 mM Hepes-KOH, pH 7.4, 320 mM sucrose, 10 mM EDTA, 5 mM EGTA, 20 μg/ml PMSF, 2 μg/ml leupeptin, 2 μg/ml pepstatin, 1 mM DTT (resuspension buffer), and disrupted in a bath sonicator (Branson) using 15 pulses at an intermediate setting. Nuclei and cell debris were removed by sedimentation at 1,300 g for 5 min at 4°C, and the supernatant stored at −70°C before use. To reduce competition with kinase inhibitors, the endogenous ATP in the added cell extract was reduced tenfold by dilution in resuspension buffer, followed by reconcentration to the original volume by filtration through a Centricon 10 filtration device (Amicon).

For immunostaining, PC12 cells were plated onto glass coverslips coated with poly-d-lysine and Matrigel (Collaborative Research), and treated with 50 ng/ml NGF for 2–5 d before fixation in 100 mM NaPO₄, pH 7.3 containing 4% paraformaldehyde. The cells were then permeabilized in PBS, pH 7.4, containing 5% bovine serum and 0.1% TX-100 detergent, incubated with primary antibody in the same buffer (1:100–1:1,000), washed three times for 5 min, and incubated in secondary antibody (1:100) for 1 h. After washing, the coverslips were mounted onto glass slides using Antifade (Molecular Probes), and visualized by indirect immunofluorescence with a BioRad 1024 confocal microscope. Digital images were processed using NIH Image and Adobe Photoshop.

To label the dense core granules with [³H]serotonin, PC12 cells were incubated in standard medium containing 5 μCi [³H]serotonin (Amersham)/15 ml for at least 4 h before cell fractionation. Cell membranes were prepared by homogenization at a clearance of 10 μm in cold Hepes-KOH-buffered saline (HBS) containing 2 mM Mg EGTA, 1 μg/ml leupeptin, 1 μg/ml pepstatin, and 10 μg/ml PMSF. The nuclei and other cell debris were pelleted by centrifugation at 1,300 g for 5 min. For sucrose gradient equilibrium centrifugation, the PNS was layered onto a linear 0.65–1.55 M (or 0.6–1.6 M, as indicated) sucrose gradient in 10 mM Hepes-KOH and sedimented to equilibrium at 30,000 rpm for 8–12 h in an SW41 rotor at 4°C. 17 fractions were collected from the bottom of the tube, stored at −70°C, and the radioactivity of fractions from cells preloaded with [³H] serotonin measured in Ecolume (ICN) using a Beckman 3800 scintillation counter. For two-step separations, the PNS from cell lysates was first layered onto a linear 0.3–1.2 M sucrose velocity gradient in 10 mM Hepes and sedimented at 25,400 rpm for 30 min in an SW41 rotor at 4°C. The radioactivity in each fraction of the velocity gradient was determined by scintillation counting, the two peak fractions were pooled, adjusted to a final sucrose concentration of 0.6 M, layered onto a linear 0.6–1.6 M sucrose gradient, and sedimented to equilibrium at 30,000 rpm for 8–12 h in an SW41 rotor at 4°C, with 12 fractions collected from the bottom of the tube. Immunoblots were probed using an antibody to the HA epitope tag, and were then optically scanned and the digitized images quantitated using NIH Image software. The total amount of HA immunoreactivity in each fraction of the second gradient was expressed as a percentage of the total HA immunoreactivity in the PNS loaded onto the first gradient, or

as shown in Fig. 9 as

Using secretogranin II (SgII) and ³H-serotonin as markers, fractions 7–10 of the second gradient include the peak of LDCVs. The sum of HA immunoreactivity for fractions 7–10 of the second gradient was then normalized for the recovery of LDCVs as follows:

This measure is similar to a recently described granule targeting index (Blagoveshchenskaya et al. 1999). Note that the number of fractions collected from one-step and two-step procedures were not identical, and that LDCVs therefore localize to different fractions.

Magnetic beads conjugated to secondary mouse antibody (Dynal) were coated with antisynaptophysin mouse mAb by incubation overnight in PBS, pH 7.4, containing 2% FCS (immunoisolation buffer), followed by 4 washes in this buffer. For each reaction, 10 μl bead slurry was added to 30 μl PNS and mixed for 1 h at 4°C. After washing in immunoisolation buffer, the bound material was eluted in 2× Laemmli sample buffer, subjected to Western analysis in parallel with samples of starting material diluted to yield images of comparable intensity, and quantitated using Adobe Photoshop and NIH image. We then expressed the amount of VAChT in the immunoisolated material as a fraction of the amount of VAChT in the starting material.

Undifferentiated PC12 cells were fixed for 2 h at room temperature with a mixture of 2% formaldehyde and 0.2% glutaraldehyde in 0.1 M phosphate buffer, pH 7.4, and stored in 4% formaldehyde until further processing. Ultrathin cryosectioning was performed according to the method described by Liou et al. 1996 and double immunogold labeling as described by Slot et al. 1991. Labeling was carried out with the polyclonal antibody to VAChT, which labeled only PC12 cells overexpressing wild-type or mutant VAChT. For double-labeling with synaptophysin, we used a polyclonal rabbit antiserum (Synaptic Systems). Semiquantitative analysis of the relative subcellular distributions of the S480A and S480E mutants was performed in double-blind experiments. To establish the percentage of total label that was present on LDCVs, sections were scanned along a fixed track and the gold particles associated with LDCVs determined as a proportion of all gold particles no more than 25 nm from a membrane. In a separate measurement, we used a similar approach to determine the association of gold particles with LDCVs relative to the association with ∼50-nm vesicles. The latter category includes SLMVs, but may also include Golgi-derived vesicles and constitutively recycling vesicles of a similar size. For all quantitations, 1,000 gold particles were counted in ten separate counting sessions using at least two independently labeled grids.

We first used heterologous expression in COS cells to determine whether VAChT undergoes regulation by phosphorylation. Like the VMATs, VAChT sorts to endocytic vesicles in these cells (Liu et al. 1994; Liu and Edwards 1997a) and the acidic nature of this compartment supports the transport activity observed in biochemical studies (Liu et al. 1992). To facilitate the analysis, we previously engineered an epitope from HA into the lumenal loop between TMDs 1 and 2 of VAChT, as well as VMAT2. These epitope tags do not interfere with the transport activity or subcellular localization of either transporter (Tan et al. 1998).

To assess the phosphorylation of VAChT, we used metabolic labeling with ³²P_i. Immunoprecipitation with an HA antibody of COS cells labeled continuously with ³²P_i for 4 h shows a prominent labeled protein of ∼70 kD in cells transiently transfected with the VAChT cDNA (Fig. 1 B, right). The strongly labeled protein corresponds to the largest immunoreactive species of VAChT observed in transfected cells by Western analysis with an antibody to VAChT (Fig. 1 B, left). An ∼50-kD immature form of VAChT (Liu and Edwards 1997a) and an ∼35-kD immunoreactive species undergo phosphorylation to a lesser extent (Fig. 1B and Fig. D). The amount of protease inhibitors used to prepare the cell extracts influences the amount of the ∼35-kD form, suggesting that it results from proteolytic degradation (data not shown), but we have used the optimal cocktail of protease inhibitors for these experiments, and not observed substantial variation in the recovery of full length VAChT. Recognition by the COOH-terminal antibody to VAChT, but not the more NH₂-terminal antibody to the HA tag suggests that the ∼35-kD species represents a phosphorylated COOH-terminal fragment (compare Fig. 1 B and 2 A). The analysis of VMAT2–VAChT chimeras further supports localization of a phosphorylation site(s) to the COOH terminus (data not shown). Since the COOH terminus of VAChT contains serine, threonine, and tyrosine residues that might serve as phosphorylation sites, we performed phosphoamino acid analysis of the VAChT protein expressed in COS cells and metabolically labeled with ³²P_i (Boyle et al. 1991). Hydrolysates of immunoprecipitated VAChT contain phosphoserine, but not phosphothreonine or phosphotyrosine residues, indicating that phosphorylation of VAChT occurs exclusively on serine (Fig. 1 C).

Although the COOH terminus of VAChT contains several serine residues, we have focused on Ser-478 and -480 (Fig. 1 A). Both of these residues show strong conservation in vertebrates (Erickson et al. 1994; Roghani et al. 1994; Naciff et al. 1997) and reside within a consensus sequence for phosphorylation by multiple kinases including PKC (Pearson and Kemp 1991). In addition, Ser-480 occurs five residues upstream from the dileucine motif required for efficient internalization of VAChT from the plasma membrane (Tan et al. 1998) and the VMATs contain a highly conserved glutamate at the equivalent position (Liu et al. 1992; Erickson and Eiden 1993; Krejci et al. 1993). Further, an acidic residue at the −5 position relative to a dileucine motif influences the sorting of other membrane proteins (Pond et al. 1995) and phosphorylation at this position regulates the trafficking of CD3γ and CD4 (Shin et al. 1991; Dietrich et al. 1994). To determine whether VAChT undergoes phosphorylation on Ser-480 or Ser-478, we used site-directed mutagenesis (Kunkel et al. 1991) to replace these residues individually with alanine. Metabolic labeling and immunoprecipitation of the mutants expressed in COS cells shows reduced phosphorylation of the S480A, but not the S478A mutant (Fig. 1 D). Although the S480A mutant shows low levels of residual phosphorylation on the ∼70-kD species, the mutation eliminates labeling of the COOH-terminal ∼35-kD fragment (Fig. 1 D, also see 2 A). Simultaneous mutagenesis of both Ser-478 and Ser-480 did not further reduce the phosphorylation (data not shown). Thus, the COOH terminus of VAChT appears to undergo phosphorylation predominantly, if not solely, on Ser-480.

To extend the analysis of VAChT phosphorylation to neuroendocrine cells, we used heterologous expression in rat pheochromocytoma PC12 cells. Metabolic labeling with ³²P_i of PC12 transformants stably expressing VAChT shows prominent labeling of an ∼70-kD band not observed in untransfected cells (Fig. 2 A, right). As in COS cells, the S480A mutation greatly reduces the labeling of both this band and the ∼35-kD species, and Western analysis of the cell extracts confirms the equivalent expression of both wild-type and mutant VAChT (Fig. 2 A, left). Thus, phosphorylation of the VAChT COOH terminus occurs on Ser-480 in both COS and PC12 cells.

To determine whether VAChT phosphorylation undergoes regulation, we have pharmacologically stimulated several protein kinases in the stably transfected PC12 cells. Using the calcium ionophore A23187 in the presence of external calcium to activate such calcium-dependent kinases as CaMKII and facilitate phosphorylation by calcium-dependent (conventional) isoforms of PKC (cPKC), we found an increase in phosphorylation of VAChT (Fig. 2 B, lane 2) relative to untreated cells (Fig. 2 B, lane 1). Stimulation of cPKC and other PKC isoforms with the phorbol ester phorbol 12,13-diacetate (PDA) activates VAChT phosphorylation even more strongly (Fig. 2 B, lane 3). In contrast, stimulation of cAMP-dependent protein kinase (PKA) with the cell-permeable cAMP analogue 8-bromo-adenosine 3′,5′-cyclic monophosphate (8-Br-cAMP) modestly inhibits VAChT phosphorylation. These results suggest phosphorylation of VAChT by PKC, with the potential for modulation by other signaling pathways. Mutagenesis of Ser-480 dramatically reduces PDA-induced changes in VAChT phosphorylation (Fig. 2 C), indicating that Ser-480 undergoes PKC-dependent phosphorylation.

To identify the kinase(s) responsible for phosphorylating VAChT, we have extended the analysis in vitro. As a substrate for phosphorylation, we produced a bacterial protein containing GST fused to the COOH terminus of VAChT (GST-VAChT). Since phosphorylation in vitro may show less specificity than in intact cells, we first tested the specificity of GST-VAChT phosphorylation. In the presence of γ[³²P]ATP and COS or PC12 cell lysate, GST-VAChT undergoes robust phosphorylation (Fig. 3A and Fig. E). The replacement of Ser-480, but not Ser-478 with alanine abolishes this modification, consistent with the results of metabolic labeling in intact cells and indicating the specificity of in vitro phosphorylation for the appropriate residues (Fig. 3A and Fig. E). Similar to the changes observed in cells with the calcium ionophore A23187, addition of Ca²⁺ to the in vitro reaction increases phosphorylation of the fusion protein (Fig. 3B and Fig. E). Calmodulin does not further alter the response to increased Ca²⁺, consistent with data from intact cells implicating Ca²⁺-dependent isoforms of PKC, which do not require calmodulin for activation.

We used the in vitro phosphorylation reaction and specific inhibitors to identify the kinase(s) responsible for phosphorylating VAChT. The PKC inhibitor H7 partially blocks VAChT phosphorylation by COS (Fig. 3C and Fig. D) or PC12 (Fig. 3 E) cell extracts, but H7 also inhibits such other kinases as PKA (Hidaka et al. 1984). We tested the effect of bisindolylmaleimide I (BIS), a highly specific PKC inhibitor (K_i ∼0.01 μM) that inhibits other kinases only at much higher concentrations (K_i for PKA ∼2 μM; Toullec et al. 1991). BIS inhibits phosphorylation of VAChT by ∼50% at 0.1 μM, implicating PKC in VAChT phosphorylation. Increasing concentrations of BIS do not further reduce phosphorylation, suggesting that distinct kinase(s) may also contribute to VAChT phosphorylation. However, both peptide and nonpeptide inhibitors of PKA (H89; Chijiwa et al. 1990) and CamKII (KN93; Sumi et al. 1991) do not inhibit VAChT phosphorylation at concentrations that selectively inhibit these enzymes (data not shown). Thus, although the identity of the additional kinases remains unclear, the effects of calcium, PDA, and BIS on VAChT phosphorylation suggest that a calcium- and diacylglycerol-activated form of PKC mediates phosphorylation of the transporter at Ser-480. Indeed, purified PKC robustly phosphorylates Ser-480 in vitro (Fig. 3 F). These results are consistent with a previous report showing that VAChT undergoes PKC-dependent phosphorylation in synaptosomes prepared from hippocampus (Barbosa et al. 1997).

Previous structural analysis of the VMATs and VAChT has implicated the central region containing 12 TMDs in the translocation of substrate (Merickel et al. 1995; Finn and Edwards 1998). In contrast, the COOH termini appear to be required for localization rather than transport (Tan et al. 1998). Nonetheless, we compared the ACh transport activity of PC12 cells expressing the wild-type protein with that of cells expressing the S480A mutant and a mutant in which Ser-480 was replaced by glutamate to mimic the phosphorylation event (S480E). Using a standard transport assay, the activities of the S480A and S480E mutants were similar to wild-type (data not shown), suggesting that phosphorylation of Ser-480 does not affect catalytic function.

Since COOH-terminal signals appear to govern the membrane trafficking of VMATs and VAChT (Tan et al. 1998; Varoqui and Erickson 1998), we have focused on the possible role of phosphorylation at Ser-480 on the localization of VAChT using transfected PC12 cells stably expressing wild-type and mutant proteins. After treatment with nerve growth factor to induce neurite formation, double-staining shows extensive colocalization of wild-type VAChT with the endosome/SLMV marker synaptophysin in both the cell body and at the tips of processes (Liu and Edwards 1997a; Fig. 4). To determine the role of phosphorylation in the localization of the wild-type protein, we used PDA to stimulate PKC, and tautomycin (100 nM) to inhibit protein phosphatases 1 and 2A, and observed a decrease in the perinuclear localization of wild-type VAChT (data not shown). However, the S480A mutant showed similar changes in localization after treatment with PDA and tautomycin, excluding a role for phosphorylation at this site in the observed changes. Indeed, PKC and protein phosphatases contribute to many membrane trafficking events (Davidson et al. 1992; De Mattheis et al. 1993). To assess the specific role of Ser-480 phosphorylation on VAChT localization, we therefore used mutants that prevent (S480A) or mimic (S480E) phosphorylation specifically at this site.

The S480A mutant shows similar, extensive colocalization with synaptophysin, similar to wild-type VAChT. However, S480E localizes less with synaptophysin in the cell body. Rather, the distribution of S480E more closely resembles that of the LDCV marker SgII than wild-type VAChT or S480A (Fig. 5). The results suggest that an acidic residue −5 relative to the dileucine motif influences the steady-state distribution of VAChT.

To characterize further the effect of Ser-480 mutations on the membrane trafficking of VAChT, we have used density gradient fractionation (Liu et al. 1994). As noted previously, wild-type VAChT partially colocalizes with the SLMV/endosome marker synaptophysin in light fractions on equilibrium sedimentation gradients. However, VAChT localizes to additional compartments including LDCVs, and appears as a relatively broad peak extending into heavier fractions (Liu and Edwards 1997a; Varoqui and Erickson 1998). HA-tagged VAChT shows a similar distribution (Fig. 6 A). The S480A mutant also appears preferentially in light fractions (Fig. 6 B), and exhibits a single peak that coincides more precisely with the peak of synaptophysin than does wild-type VAChT. The S480E mutant, which mimics phosphorylation at Ser-480, also comigrates with synatophysin. However, S480E shows increased cofractionation with heavier membranes relative to wild-type VAChT and the S480A mutant (Fig. 6C and Fig. D), confirming that an acidic residue at position −5 relative to the dileucine motif influences subcellular localization.

The immunofluorescence and fractionation experiments suggest that the Ser-480 mutations also change the expression of VAChT on SLMVs and/or endosomes. To further quantitate these differences, we immunoisolated vesicles expressing synaptophysin from cells expressing wild-type and mutant VAChT, and determined the amount of immunoreactive VAChT in synaptophysin-positive (syn⁺) vesicles relative to the total amount in cell lysates (Fig. 6 E). The S480A mutant sorts to syn⁺ vesicles in proportions similar to wild-type VAChT (∼15%). The low proportion presumably reflects the localization of VAChT to multiple membrane compartments, suggested by the distribution on sucrose gradients, as well as, perhaps, the inefficiency of immunoisolation. In contrast to the wild-type protein and S480A, the S480E mutant localizes to syn⁺ vesicles in lower proportions (∼7%). Since synaptophysin resides on both endosomes and SLMVs in PC12 cells, the results indicate reduced expression of S480E on one or both of these membranes. To determine whether the Ser-480 mutations specifically affect sorting to SLMVs, we separated endosomes from SLMVs using velocity gradient sedimentation. However, we did not detect consistent differences in the proportion of wild-type and mutant VAChT expressed on SLMVs (data not shown), suggesting that the increase in targeting of S480E to heavy fractions on sucrose density gradients correlates primarily with a decrease in localization to syn⁺ endosomes.

Unlike VAChT, wild-type VMAT2 contains acidic residues at both positions −4 and −5 relative to the dileucine-like motif (Glu-478 and Glu-479; see Fig. 1). The importance of acidic charge upstream of the dileucine motif in VAChT suggests that Glu-478 and Glu-479 may also affect VMAT2 trafficking. We therefore replaced Glu-478 and -479 with alanine (E478A/E479A). Immunofluorescence of stably transfected PC12 cells shows that wild-type VMAT2 colocalizes with the LDCV marker SGII at the tips of processes (Fig. 7, a–c). However, the E478A/E479A mutant differs from SGII and shows consistently strong labeling in cell bodies (Fig. 7, d–f). Indeed, the E478A/E479A mutant colocalizes with synaptophysin to a greater extent than wild-type VMAT2 (Fig. 7, g–l). Thus, acidic residues upstream of the dileucine motif appear to be important for the localization of VMAT2, as well as VAChT.

VMATs localize almost exclusively to LDCVs in PC12 cells, and therefore cofractionate to a high degree with SGII in the heavy fractions of sucrose equilibrium sucrose gradients (Liu et al. 1994). Increased localization of the VAChT mutant S480E to heavy fractions suggests that neutralization of Glu-478 and Glu-479 might conversely reduce the localization of VMAT2 to LDCVs. HA-tagged VMAT2 cofractionates nearly identically with SGII, consistent with its localization to LDCVs (Fig. 8 A). In contrast, the E478A/E479A mutant cofractionates with SgII to a lesser extent, and the profile of VMAT2 immunoreactivity across the gradient differs greatly from that of the LDCV marker (Fig. 8 B). Since wild-type VMAT2 resides almost entirely on LDCVs, these results indicate that neutralization of Glu-478 and -479 reduces the proportion of VMAT2 expressed on LDCVs. Acidic residues upstream of the dileucine motif in VMAT2 thus influence sorting of the transporter to this population of regulated secretory vesicles.

Replacement of Ser-480 to mimic phosphorylation redistributes VAChT to the tips of processes in PC12 cells, and to heavier fractions in density gradients. To determine whether the S480E mutation specifically increases localization of VAChT to LDCVs, we have used additional membrane fractionation and immunoelectron microscopy to assess the localization of the S480E mutant to LDCVs. A two-step fractionation procedure involving velocity sedimentation, followed by equilibrium sedimentation (Stinchcombe and Huttner 1994), separates LDCVs from other membranes of similar size and density. A small percentage of wild-type (Liu et al. 1994) and HA-tagged VAChT (∼4.5%) resides on LDCVs (Fig. 9). However, mutations at Ser-480 change the amount of HA-tagged VAChT in LDCV fractions relative to the amount of transporter in starting material loaded onto the first velocity gradient (Fig. 9). The S480A mutant localizes to LDCVs to a slightly lesser extent (∼3.4%), and the S480E mutation increases expression on LDCVs to ∼10.6%, 2.3-fold more than wild-type, and about threefold more than the S480A mutant. The low proportion of VAChT on LDCVs, even in the case of the S480E mutant may reflect the inefficiency of the two-step procedure, which results in more highly purified LDCVs at the expense of yield. Nonetheless, the S480A mutation modestly reduces, and the S480E mutation clearly increases, the expression of VAChT on LDCVs, indicating that an acidic residue at position 480 increases localization of VAChT to LDCVs, and suggesting that a negatively charged phosphoserine at this position in wild-type VAChT may similarly direct the transporter to LDCVs.

The subcellular location of wild-type VAChT, as well as the S480A and S480E mutants, was analyzed in more detail by immunogold labeling of ultrathin cryosections (Fig. 10). LDCVs in PC12 cells appear in cryosections as 75–125-nm diam compartments with an electrodense core (Fig. 10 E) that labels for SgII (not shown). In agreement with previous observations, wild-type VAChT localizes to LDCVs, as well as numerous small vesicular profiles found in the area near the Golgi complex (Fig. 10 A), near endosomes (Fig. 10 B), or on membranes dispersed throughout the cytoplasm (Weihe et al. 1996; Liu and Edwards 1997a; Fig. 10 C). Label was also found directly over the Golgi complex, plasma membrane, and endosomes. Double-labeling shows that VAChT and synaptophysin colocalize to a high degree in the small vesicular profiles (Fig. 10 C) which largely represent the SLMVs of PC12 cells (de Wit et al. 1999). VAChT appears in early endosomes as well, characterized by a limited number of internal membranes (Fig. 10 B), but is absent from densely filled late endosomes and lysosomes (Fig. 10 A). Notably, the limiting membrane of the vacuolar part of early endosomes (also known as the sorting endosome) contains relatively little VAChT label, whereas dense labeling appears over surrounding vesicles. A similar steady-state distribution over early endosomes has previously been shown for synaptophysin (Clift-O'Grady et al. 1990; Cameron et al. 1991; Regnier-Vigouroux et al. 1991) and implies rapid sorting of proteins into recycling vesicles.

The S480A and S480E mutants label the same compartments as wild-type VAChT (Fig. 10D and Fig. E), but with different relative distributions. Of all the gold particles present in a section, 17.5% S480A and 29.4% S480E appear over the membrane of LDCVs (Table). In addition, when the distribution of gold particles between only LDCVs and SLMVs was quantitated, 40% S480A and 56% S480E were found over LDCVs (Table). Consistent with the immunofluorescence and biochemical fractionation, immunoelectron microscopy thus provides quantitative evidence for the increased localization of S480E to LDCVs.

In addition to differences in their contents, subcellular location and mode of release, LDCVs and SVs differ in their biogenesis (Bauerfeind and Huttner 1993; Kelly 1993; Martin 1994). Electron microscopic observations have indicated that LDCVs bud directly from the trans-Golgi network (TGN; Farquhar et al. 1978; Tooze and Tooze 1986; Orci et al. 1987). After a maturation process that involves the removal of proteins destined for other intracellular membranes (Tooze et al. 1991; Klumperman et al. 1998), LDCVs undergo regulated exocytosis. In contrast, SVs do not form at the TGN. Rather, proteins destined for SVs sort to constitutive secretory vesicles at the TGN, arrive at the plasma membrane in the absence of stimulation, and enter SVs only after recycling from the cell surface (Clift-O'Grady et al. 1990; Cameron et al. 1991; Regnier-Vigouroux et al. 1991). However, relatively little is known about the signals that sort proteins to either LDCVs or SVs (Grote et al. 1995; Haucke and De Camilli 1999).

Previous studies of protein sorting to LDCVs have focused on their soluble contents. Although recent work has implicated the processing enzyme carboxypeptidase E in sorting to LDCVs (Cool et al. 1997), most neural peptides sort into LDCVs as a result of aggregation under the particular redox and pH conditions of nascent LDCVs (Chanat and Huttner 1991; Rindler 1998). Thus, the soluble cargo of LDCVs are not thought to contain a specific sorting sequence, but rather to enter the regulated secretory pathway as a result of their general biophysical properties. In contrast, integral membrane proteins of the LDCV have the potential to interact directly with sorting machinery through their cytoplasmic domains which presumably contain specific sorting signals.

Most of the integral membrane proteins present on LDCVs also occur at substantial levels on SVs. These include the vSNARE proteins involved in vesicle docking and fusion (Schmidle et al. 1991; Glenn and Burgoyne 1996), as well as certain isoforms of the calcium sensor synaptotagmin (Schmidle et al. 1991; Walch-Solimena et al. 1993). Heterologous expression in neuroendocrine cells of the endothelial protein P-selectin also confers expression on both LDCVs and SVs (Norcott et al. 1996). In contrast, VMATs localize preferentially to LDCVs in PC12 cells, indicating that they sort more specifically to the regulated secretory pathway (Liu et al. 1994). The closely related VAChT also occurs on LDCVs, but at a much lower level than the VMATs (Liu and Edwards 1997a; Varoqui and Erickson 1998). VAChT preferentially localizes to lighter membranes, including endosomes and SLMVs in PC12 cells, indicating substantial differences in trafficking from the VMATs. Recent observations also indicate that the differential trafficking of the transporters depends on signals contained in their cytoplasmic, COOH-terminal domains (Tan et al. 1998; Varoqui and Erickson 1998).

Using a combination of immunofluorescence, biochemical fractionation, and immunoelectron microscopy, we now show that mutation of Ser-480 regulates membrane trafficking of VAChT to LDCVs. By immunofluorescence, the distribution of wild-type and the S480A mutant protein resembles that of synaptophysin. In contrast, the localization of S480E differs from wild-type VAChT and S480A, and more closely resembles that of the LDCV marker SGII. Equilibrium sedimentation through sucrose also reveals a larger proportion of S480E than wild-type VAChT in heavy fractions, comigrating with LDCVs.

Mutation of acidic residues upstream of the dileucine motif in VMAT2 (Glu-478 and -479) also alters transporter localization. However, unlike VAChT, wild-type VMATs localize almost exclusively to LDCVs (Liu et al. 1994). The effect of VAChT mutations at Ser-480 suggest that neutralization of glutamate residues upstream of the dileucine motif in VMAT2 might reduce localization to LDCVs. Indeed, immunofluorescence and sucrose equilibrium gradients show that replacement of Glu-478 and -479 by alanine partially redistributes VMAT2 to lighter membranes and away from LDCVs. The E478/479A mutation does not eliminate localization of VMAT2 to LDCVs, and additional signals may contribute to the sorting of VMAT2. The results nonetheless suggest that acidic residues upstream of a dileucine motif are involved in the trafficking of VMAT2 to regulated secretory vesicles.

In addition, two-step fractionation shows that the S480E mutation increases localization of VAChT to LDCVs relative to wild-type protein and the S480A mutant. Quantitative immunoelectron microscopy further confirms the increased expression of S480E on LDCVs, although the difference from S480A appears less marked than the threefold increase observed by the two-step fractionation procedure. Losses of starting material during fractionation may have amplified slightly the difference in localization between S480A and S480E. Nonetheless, analysis of the mutants strongly suggests that a negative charge five residues upstream of the dileucine motif increases the sorting of VAChT to LDCVs.

Both gradient fractionation and immunoisolation indicate a corresponding decrease in the sorting of S480E to synaptophysin containing vesicles. However, synaptophysin localizes to both endosomes and SLMVs in PC12 cells (Clift-O'Grady et al. 1990; Cameron et al. 1991). We therefore performed velocity gradient fractionation to specifically determine the effect of S480E on SLMVs. Since the S480E mutant sorts to SLMVs in proportions similar to wild-type and S480A, the negative charge upstream of the dileucine motif may not influence sorting to SLMVs, at least in PC12 cells. In neurons, however, synaptophysin localizes almost exclusively to SVs, and the dominant pathway of SV biogenesis is likely to differ from that used in PC12 cells (Shi et al. 1998), leaving open the possibility that phosphorylation of Ser-480 influences the localization to SVs in neurons.

Very few, if any, other signals responsible for localization to LDCVs have previously been identified. The protein P-selectin, which localizes to secretory granules in endothelial cells and to LDCVs after heterologous expression in PC12 cells (Norcott et al. 1996), has a sequence at its cytoplasmic COOH terminus (KDDG) that resembles the sequence of basic and acidic residues in the VMATs and VAChT. However, unlike the VMATs and VAChT, selectin does not contain a dileucine-like motif flanking the charged residues. Indeed, replacement of the KDDG sequence in P-selectin by four alanines does not apparently reduce targeting to LDCVs, and other signals in P-selectin mediate localization to secretory vesicles (Blagoveshchenskaya et al. 1999).

The signals that direct neurotransmitter transporters to secretory vesicles presumably interact with a cytoplasmic sorting machinery that may include clathrin adaptor proteins, such as AP-1 and -2, and the related AP-3 and AP-4 (Bonifacino and Dell'Angelica 1999). Dileucine motifs are generally considered to interact directly with the β subunit of adaptors (Rapoport et al. 1998). However, acidic residues and phosphorylation sites upstream of dileucine signals have also been implicated in the internalization of plasma membrane proteins. These residues may bind directly to the adaptors, or they may modulate adaptor binding indirectly (Dietrich et al. 1997). Regardless of the precise mechanism, the spacing between these upstream residues and the leucines appears important and in some cases, the upstream acidic residues and the two leucines function as a single motif (Dietrich et al. 1997). These observations suggest that the acidic residues upstream of the dileucine motif in VAChT and the VMATs act in concert with the two leucines, or modulate their ability to interact with the cytoplasmic sorting machinery.

The signal we have identified could act at several sites in the cell to affect localization to LDCVs. First, it could act in the TGN to direct the vesicular transporters toward the regulated secretory pathway (LDCVs) and away from the constitutive secretory pathway. Second, it could influence retrieval from immature LDCVs. Immature LDCVs contain resident TGN proteins, such as furin (Dittie et al. 1997), and lysosomal proteins, such as cathepsin B (Kuliawat et al. 1997). Maturation of the LDCVs involves the removal of these proteins by an AP-1– and clathrin-dependent mechanism (Dittie et al. 1996; Klumperman et al. 1998), and phosphorylation of an acidic patch by casein kinase II promotes retrieval by recruiting a novel cytosolic protein (Wan et al. 1998). Acidic residues upstream of the dileucine motifs present in the neurotransmitter transporters might thus act by suppressing retrieval from immature LDCVs, thereby increasing expression on mature LDCVs. Third, the dileucine-based signals may influence the recycling of VAChT and the VMATs from the cell surface after exocytosis. Upstream acidic residues may suppress local recycling to endosomes or synaptic vesicles and so direct the proteins to the TGN and then to LDCVs. The precise membrane trafficking step(s) affected by acidic residues upstream of the dileucine motif in vesicular transporters remain to be determined. However, the role of the dileucine motif in internalization from the plasma membrane (Tan et al. 1998) indicates that the dileucine motif itself functions at multiple trafficking steps, with the upstream acidic residues perhaps required for a subset of these events.

Phosphorylation of Ser-480 in VAChT provides a negative charge similar to the glutamate present at this position in the VMATs. Metabolic labeling in unstimulated PC12 cells indicates that constitutive phosphorylation may contribute to the localization of wild-type VAChT. However, since wild-type VAChT localizes to LDCVs in amounts similar to the S480A mutant, constitutive phosphorylation presumably occurs at low levels. Activation of PKC increases VAChT phosphorylation, and may increase the localization of the transporter to LDCVs.

The localization of VAChT to LDCVs in PC12 cells has relevance for the site of ACh storage in vivo. Although SVs are generally considered to store and release from ACh at the neuromuscular junction and in brain (Gilmor et al. 1996; Weihe et al. 1996), ACh also appears in vesicles similar to, or identical to LDCVs by biochemical fractionation (Lundberg et al. 1981; Agoston and Whittaker 1989). In addition, light microscopic analysis of brain sections shows a distribution for VAChT very similar to that of VMAT2, with strong staining of cell bodies and dendrites not observed with other synaptic vesicle proteins, including the vesicular GABA transporter (Schafer et al. 1995; Gilmor et al. 1996; Chaudhry et al. 1998). Thus, the regulation of VAChT trafficking by phosphorylation of Ser-480 may also influence the distribution of the protein in vivo.

A change in the trafficking of VAChT has the potential to affect transmitter release. A number of observations indicate that vesicular transport activity influences the amount of transmitter released per vesicle, or quantal size. Mice heterozygous for a disruption of the VMAT2 gene show dramatically reduced storage and release of dopamine and serotonin (Fon et al. 1997). VMAT2 heterozygotes (±) also show behavioral differences from wild-type animals (Takahashi et al. 1997; Wang et al. 1997). In addition, overexpression of rat VAChT in Xenopus laevis embryos generates larger postsynaptic quantal events in cocultured myocytes (Song et al. 1997). Alterations in the localization of VAChT to LDCVs (or SVs) may therefore affect the amount of transmitter stored and released by these vesicle populations.

The observations have relevance for certain forms of neural plasticity. The neuromuscular junction exhibit considerable variation in quantal size and the mechanism(s) appears to be presynaptic (Van der Kloot 1990; Williams 1997). Drugs that inhibit vesicular ACh transport block the regulation, implicating VAChT in the mechanism, and protein phosphorylation also appears to be involved (Van Der Kloot and Molgo 1994). In the central nervous system, PKC also modulates ACh release (Tanaka et al. 1986; Allgaier et al. 1988). Activation of PKC increases VAChT phosphorylation, as well as ACh release in hippocampal synaptosomes, and the changes in transmitter release require VAChT function (Barbosa et al. 1997). The phosphorylation of VAChT by PKC at Ser-480 may thus contribute to multiple forms of neural plasticity by influencing the distribution of the transport protein on neurosecretory vesicles. The difference that we have observed between trafficking of S480A and S480E mutants indicates the potential for phosphorylation at Ser-480 to influence quantal size. The magnitude of the differences in localization resembles the magnitude changes in quantal size observed for other forms of synaptic plasticity (Malenka and Nicoll 1993). Further, the localization of VAChT to LDCVs and its regulation by phosphorylation of Ser-480 may specifically contribute to the giant endplate potentials observed at the neuromuscular junction (Van der Kloot 1990), and help to account for this poorly understood mode of ACh release.

We thank the members of the Edwards lab for thoughtful discussion, and Victor Faundez, Jack Roos, and Ted Fon for helpful comments on the text. Rene Scriwanek and Tom van Rijn are acknowledged for the preparation of electron micrographs.

This work was supported by a Postdoctoral Fellowship for Physicians from the Howard Hughes Medical Institute (to D.E. Krantz), a National Science Foundation Graduate Research Fellowship (to R.I. Wilson), a Young Investigator Award from the National Alliance for Research on Schizophrenia and Affective Disorders (NARSAD; to Y. Liu), the National Parkinson Foundation (Y. Liu), and the National Institutes of Mental Health (P.K. Tan and R.H. Edwards) and Drug Abuse (R.H. Edwards).