Address correspondence and reprint requests to Marco A. M. Prado, Laboratório de Neurofarmacologia, Departamento de Farmacologia, Instituto de Ciências Biológicas Universidade Federal de Minas Gerais, Avenue. Antônio Carlos, 6627 Belo Horizonte, MG, Brazil. E-mail: email@example.com
The pathways by which synaptic vesicle proteins reach their destination are not completely defined. Here we investigated the traffic of a green fluorescent protein (GFP)-tagged version of the vesicular acetylcholine transporter (VAChT) in cholinergic SN56 cells, a model system for neuronal processing of this cargo. GFP-VAChT accumulates in small vesicular compartments in varicosities, but perturbation of endocytosis with a dominant negative mutant of dynamin I-K44A impaired GFP-VAChT trafficking to these processes. The protein in this condition accumulated in the cell body plasma membrane and in large vesicular patches therein. A VAChT endocytic mutant (L485A/L486A) was also located at the plasma membrane, however, the protein was not sorted to dynamin I-K44A generated vesicles. A fusion protein containing the VAChT C-terminal tail precipitated the AP-2 adaptor protein complex from rat brain, suggesting that VAChT directly interacts with the endocytic complex. In addition, yeast two hybrid experiments indicated that the C-terminal tail of VAChT interacts with the µ subunit of AP-2 in a di-leucine (L485A/L486A) dependent fashion. These observations suggest that the di-leucine motif regulates sorting of VAChT from the soma plasma membrane through a clathrin dependent mechanism prior to the targeting of the transporter to varicosities.
A key process involved in synaptic vesicle biogenesis is the trafficking of synaptic vesicle proteins to axonal terminals. However, the mechanisms and pathways involved in this process have not been clearly elucidated. Classically, it is thought that the formation of synaptic vesicles involves the intracellular transport of synaptic proteins to nerve endings by tubulo-vesicular organelles (Mundigl et al. 1993; Nakata et al. 1998; Ahmari et al. 2000). These tubulo-vesicular organelles are proposed to bud from the trans-Golgi network (TGN) and traffic directly to the active zones of synapses (Nakata et al. 1998). Alternatively, it is possible that the formation of presynaptic active zones involves axonal transport of packages of heterogeneous vesicular organelles that become stabilized at axon initiated synaptic contacts (Ahmari et al. 2000). The membrane organelles transporting synaptic vesicle proteins are then remodeled into homogeneous vesicles, typical of the active zone (Ahmari et al. 2000).
Recent studies examining the intracellular trafficking of the homologous vesicular acetylcholine transporter (VAChT) and the vesicular monoamine transporter 2 (VMAT2) have provided interesting data regarding the targeting of these proteins to secretory vesicles. VAChT is predominantly targeted to synaptic-like microvesicles (SLMVs), whereas the VMAT2 is found mainly in large-dense core vesicles (LDCVs) in the PC12 cell model system (Liu et al. 1994; Weihe et al. 1996; Liu and Edwards 1997). Differences in the intracellular trafficking of VAChT and VMAT2 to SMLV and LDCVs, respectively, appear to be regulated by amino acid motifs in the carboxyl-terminal tails (C-tail) of the transporters (Varoqui and Erickson 1998; Krantz et al. 2000). A di-leucine motif (L485/L486) found in the C-tail participates in the internalization of VAChT, and may be necessary for recycling of the protein after synaptic vesicle exocytosis (Tan et al. 1998; Santos et al. 2001) or other trafficking events (Cho et al. 2000; Krantz et al. 2000).
Di-leucine signals are well-recognized motifs required for protein interactions with heterotetrameric clathrin adaptor protein (AP) complexes. Several distinct heterotetrameric AP complexes have been identified that exhibit specific protein trafficking functions. For example, the AP-1 complex is involved with the trafficking of proteins from the TGN, whereas the AP-2 complex mediates the internalization of proteins from the plasma membrane through clathrin-coated vesicles (Robinson and Bonifacino 2001). Di-leucine motifs interact with the β1 subunit of AP-1, as well as the µ1 and µ2subunits of AP-1 and AP-2 complexes, respectively (Rapoport et al. 1998; Rodionov and Bakke 1998; Kongsvik et al. 2002).
Mutation of the VAChT di-leucine motif to alanine residues relocates the transporter to the soma plasma membrane (Tan et al. 1998; Santos et al. 2001). Inhibition of endocytosis of VAChT has been suggested to be the cause of this change in location (Tan et al. 1998). However, the di-leucine motif might be primarily involved with trafficking of VAChT from the TGN (for example by interaction with AP-1 or another complex) and mutations could cause mis-sorting of the protein to the plasma membrane. Thus, it would be of interest to use conditions where clathrin-mediated endocytosis at the plasma membrane are affected to test whether VAChT endocytosis at the cell body is relevant for trafficking.
In the present work, we tested whether Dynamin I is necessary for VAChT endocytosis from the cell body plasma membrane of SN56 cells. Dynamin is a GTPase involved in some fission events and the dominant negative mutant dynamin I-K44A impairs clathrin mediated endocytosis (Damke et al. 1994; Damke et al. 2001). In cells overexpressing dynamin I-K44A endocytic intermediates are formed but fail to detach from the plasma membrane (Damke et al. 2001). We show here that dynamin I-dependent endocytosis is an essential step in the trafficking of VAChT. Moreover, we show that the C-terminal di-leucine motif (L485/L486) is used to recruit VAChT to vesicles and that the C-terminal region of VAChT associates with the µ-adaptin subunit of the heterotetrameric AP-2 adaptor complex. These data provide novel information on the pathways and mechanisms used by VAChT to reach varicosities and indicate the need for clathrin-mediated endocytosis in the biogenesis of cholinergic vesicles.
SN56 cells were a generous gift of Prof. Bruce Wainer, (Department of Pathology, Emory University School of Medicine, Atlanta, GA) and were maintained as previously described (Santos et al. 2001). For differentiation cells were incubated in medium lacking fetal bovine serum and supplemented with 1 mm dibutyril-cyclic-AMP (Sigma, St Louis, MO, USA) for at least 5 days. Medium was changed every two days. The SN56 cells were derived from septum neurons (Hammond et al. 1990) and present a number of cholinergic features including expression of synaptic vesicle proteins (Barbosa et al. 1999) and neuronal type calcium channels (Kushmerick et al. 2001). Such features are increased by differentiation (Blusztajn et al. 1992; Kushmerick et al. 2001).
Yeast two-hybrid interaction and cDNA constructs
We used a pMAL-C-terminal clone (Barbosa et al. 1999) containing the region encoding the carboxyl-terminal domain of mouse VAChT (amino acids 471–530) to prepare the yeast vector. The clone was digested with EcoRI and NotI and the fragment was fused in frame to the Gal4 DNA binding domain (BD) of pAS2-1 (Clontech Laboratories Inc., Palo Alto, CA, USA). A mutant of VAChT carboxyl-termini (L485A/L486A) was generated by PCR and cloned into pAS2-1 using EcoRI and BamHI restriction sites. All constructs were checked by automated DNA sequencing. Adaptin cDNAs, cloned into the Gal4 activation domain (AD) of pACT2 (pACT2-α adaptin, pACT2-β2, adaptin pACT2-µ2 adaptin, or pACT2-σ2 adaptin) have been previously described (Laporte et al. 1999) and were a kind gift from Dr S. Laporte, Department of Medicine, McGill University. Pairs of VAChT-BD and adaptin-AD fusion genes were cotransformed into PJ69–4a yeast strain and grown on leucine and tryptophan-free (–L–W) media. Specific interactions were assayed by growth of yeast clones on leucine, tryptophan, and adenine-free (–L–W–A) media. The pAS2-1 C-terminal construction did not allow the growth of transformant yeast in restrictive medium (–L–W–A) when coexpressed with empty or non-relevant pACT2 vectors.
The constructs containing dynamin I or the dominant negative mutant dynamin I-K44A have been described elsewhere (Zhang et al. 1997). The effect of dynamin I-K44A overexpression on clathrin-mediated endocytosis (estimated by transferrin uptake) was dose dependent with the maximum effect achieved with 1 µg of plasmid tranfected.
The complete coding region of the mouse VAChT gene was amplified from genomic DNA by PCR using Pfu polymerase (Barbosa et al. 1999; Santos et al. 2001). Detailed description of GFP-VAChT and GFP-VAChT L485A/L486A construction and characterization is given elsewhere (Santos et al. 2001). GST-VAChT C-tail (amino acids 471–530) was generated by PCR with restriction sites for BamHI and Not1, cloned into the pGEX 4T-3 vector (Amersham Pharmacia Biotech, Uppsala, Sweden) and sequenced. Expression of the fusion protein and purification with glutathione sepharose 4B was performed according to the manufacturer instructions. The fusion protein was recognized in immunoblots (not shown) by an antibody raised against the C-terminal region of VAChT (Weihe et al. 1996, a kind gift of Dr L. E. Eiden, NIH).
Western blotting and precipitation experiments
Animal use was approved by a local animal care committee and followed international standards. For preparation of rat brain extracts, the cortex was homogenized in 100 mm Tris-Cl pH 7.4, 150 mm NaCl, 1 mm EDTA 1% Triton X-100, 0.1% sodium dodecyl sulfate (SDS), 1% deoxycholic acid and protease inhibitor mixture (complete protease inhibitor tablets, double of the concentration suggested by the manufacturer) (Boehringer Mannheim,Indianapolis, IN, USA] and cleared by centrifugation (12 100 g × 10 min). Supernatants (500 µg of protein) were incubated for 60 min at 4°C with glutathione sepharose 4B beads previously bound to GST-VAChT C-tail region or GST. Afterwards, beads were precipitated and washed 5 times with homogenization buffer and after the last wash beads were resuspended in SDS-sample buffer. Proteins were resolved by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and transferred to nitrocellulose membrane as described (Barbosa et al. 1997). A monoclonal antibody antibovine α-adaptin (Sigma) was used to detect the AP-2 complex using enhanced chemiluminescence (ECL from Amersham Pharmacia Biotech, Piscataway, NJ, USA) and T-Mat X-ray film (Kodak, São José dos Campos, SP, Brazil).
Quantification of precipitated proteins were performed in a densitometer (Bio-Rad Model GS-700 imaging densitometer) controlled by the Multi-analyst software (Bio-Rad, Hercules, CA, USA). Films with distinct exposure times were used for quantification to insure linearity of signal. The amounts of GST and GST-C-terminal fusion protein in each experiment were estimated with an anti-GST antibody (Pharmacia) and used to normalize the amount of α-adaptin precipitated in the two conditions.
Cell transfection, confocal imaging and endosomal labeling
Cell transfection was performed with Lipofectamine 2000 (Life Technologies, Gaithesburg, MD, USA) according to the manufacturer's instructions. Confocal imaging has been previously described (Santos et al. 2001). Inhibition of clathrin-mediated endocytosis was estimated by transferrin labeling. Endosomes were visualized by incubating cells with 40 µg/mL of Alexa 568 labeled transferrin (Tfn-568 Molecular Probes, Eugene, OR, USA) at 37°C in 5% CO2 for 30 min. After incubation, cells were washed three times with ice cold phosphate-buffered saline and then fixed with 3% paraformaldehyde in phosphate-buffered saline for 20 min.
Analysis of transferrin uptake and neurite outgrowth
Maximum Z projections 8-bit images of cells previously labeled with Tfn-568 were analyzed by comparing the intensity of Tfn-568 fluorescence in transfected cells with that of neighboring non-transfected cells. Results are shown in gray levels. Cells with less than 50% uptake of Tfn were considered cotransfected with GFP and K44A.
We used Metamorph (Universal Imaging, West Chester, PA, USA) to measure (in µm) defined neurites budding from cells transfected with K44A or dynamin and grown in low-density cultures. The value obtained by measuring the lengths of all the neurites in a given transfected cell was averaged and the results for a large population of K44A transfected cells were compared with the results obtained with dynamin I transfected cells.
Overexpression of dynamin I-K44A blocks clathrin-mediated endocytosis in SN56 cells
In previous work we have characterized the distribution of GFP-VAChT in SN56 cells and showed that the introduction of the GFP to the amino-terminus of VAChT does not alter its trafficking to varicosity sites (Santos et al. 2001). We have also observed that the localization of GFP-VAChT is identical to that of an epitope-tagged VAChT construct (not shown), which has been previously shown to be targeted to SLMV (Krantz et al. 2000). To test whether GFP-VAChT reaches the plasma membrane of the soma or is directly targeted to varicosities from the TGN, we used a dominant-negative dynamin I mutant (K44A) and disrupted clathrin-mediated endocytosis (Damke et al. 1994; Damke et al. 2001). Internalization of fluorescent transferrin is blocked by 70–80% in response to overexpression of dynamin I-K44A, whereas overexpression of dynamin I had no effect on internalization, when compared with non-transfected cells (Fig. 1a). Dynamin I-K44A overexpression also decreased neuronal differentiation of SN56 cells, specifically inhibiting the number and length of neurites (70% of overall decrease). This decrease is significant and resembled the lack of neurite outgrowth and collapse of neurites observed in neurons where dynamin I activity was blocked by alternative methods (Kim and Wu 1987; Masur et al. 1990; Torre et al. 1994). Nevertheless, differentiation for 5 days prior to transfection allowed us to visualize cells that presented short neurites.
Co-transfection of wild-type dynamin does not alter the pattern of GFP-VAChT expression compared with GFP-VAChT expressed alone (Fig. 1b, Santos et al. 2001). The GFP-VAChT is present in puncta in both the cell body and neurites. We have demonstrated previously that the GFP-VAChT positive vesicles colocalize with Tfn, Rab5 and FM4-64, markers of endosomes and recycling vesicles, respectively (Santos et al. 2001). In contrast, the expression of dynamin I-K44A alters the subcellular distribution of GFP-VAChT, which is now localized to the plasma membrane and in large vesicular patches (Fig. 1c, arrows). Moreover, the presence of GFP-VAChT in recycling endosomes in the cell body (Fig. 1b and Santos et al. 2001) is completely suppressed, suggesting that dynamin I-K44A blocked endocytosis of VAChT. The patches were found, usually very close to the plasma membrane, resembling invaginating membrane that fail to pinch off in the presence of dynamin I-K44A as reported previously (Damke et al. 2001). In cells presenting neurites, GFP-VAChT is largely excluded from the neuritic processes (Fig. 1e, arrowheads). This is clearly different from the pattern of GFP-VAChT localization to neurites in cells coexpressing wild-type dynamin (Fig. 1b). The individual cells show the pattern of internalization of Tfn-568 expected for each condition (Figs 1c and f).
The di-leucine motif is necessary for VAChT recruitment to dynamin I-dependent patches
In previous experiments, the mutation of VAChT di-leucine motif (L485A/L486A) impaired internalization of theprotein, and VAChT was then located at the plasma membrane, thereby preventing the distribution of the transporter to small vesicles in varicosities (Santos et al. 2001). However, the pattern of GFP-VAChT L485A/L486A expression is distinct from that observed for wild-type GFP-VAChT coexpressed with dynamin I-K44A. GFP-VAChT L485A/L486A is diffusely localized in the somal membrane (Fig. 2a) and does not exhibit the patched appearance observed when dynamin I activity is blocked (Fig. 1e). As expected, the internalization of Tfn is unchanged in cells expressing GFP-VAChT L485A/L486A (Fig. 2b). We tested the assumption that the di-leucine motif in VAChT is used to recruit the protein to the patched vesicles observed in dynamin I-K44A expressing cells in cotransfection experiments. The expression of dynamin I-K44A does not alter the membrane distribution of GFP-VAChT L485A/L486A in SN56 cells indicating that the formation of GFP-VAChT patches is an active process that requires an intact di-leucine motif (Fig. 2d). Tfn-568 uptake is blocked in this representative cell indicating the cotransfection with dynamin I-K44A (Fig. 2e).
VAChT interacts with the AP-2 complex
Although synaptic vesicle proteins are constantly endocytosed at the nerve-terminal, only a few classical endocytic motifs have been identified in synaptic vesicle proteins. Our data suggest that the VAChT di-leucine motif is essential for the endocytosis of the transporter. However, the mechanisms involved with the internalization of the transporter are unknown. One possibility is that the VAChT di-leucine motif interacts with components of the heterotetrameric AP-2 adaptor complex at the plasma membrane.
To investigate the possibility that the di-leucine motif localized to the VAChT C-tail interacts with the AP-2 complex, we used a GST-VAChT C-tail fusion protein to test whether it could precipitate the AP-2 complex from rat brain. We find that the GST-VAChT C-tail fusion protein effectively coprecipitates the AP-2 complex from rat brain homogenates as demonstrated using an α-adaptin monoclonal antibody to identify the complex (Fig. 3a and inset). Glutathione beads bound to GST only were much less effective to precipitate the samples (Fig. 3a and inset). Similar results were obtained in experiments using SN56 cell extracts (not shown).
Previous experiments have shown that the AP-2 complex is not dissociated in conditions that resemble those used here to precipitate α-adaptin with the GST-VAChT C-tail fusion protein (Tebar et al. 1996). Thus, although we used an anti α-adaptin antibody to detect AP-2, the C-tail does not necessarily binds to this subunit of the complex. In order to identify which adaptin subunit interacts with the VAChT we used the yeast two-hybrid system (Fields and Song 1989). The VAChT C- tail was subcloned in the Gal4 DNA binding domain of pAS2-1 and yeast cells were cotransformed with this vector and complementary Gal4 activation domain vectors containing each of the four AP-2 subunits. As a positive control, we used the specific interaction of β-arrestin with β-adaptin (Fig. 3b), that was previously identified by yeast two-hybrid (Laporte et al. 1999). We found that the VAChT C-tail did not interact with either α or β2-adaptin subunits (Fig. 3b), but instead interacted with the µ2 subunit and to a lesser extent the σ2-adaptin subunit (Fig. 3b). In order to test whether the di-leucine motif (L485/L486) mediated the interaction of the C-terminal region of VAChT with µ2-adaptin, we repeated the transformations with the VAChT tail mutant (pAS2-1-C-tail L485A/L486A). This mutant failed to interact with any of the adaptin subunits including µ2-adaptin (Fig. 3b).
In the present study, we demonstrate that in cholinergic SN56 cells, VAChT trafficking is impaired by the dominant negative mutant of dynamin I-K44A. Our previous experiments using the GFP-VAChT L485A/L486A mutant indicated that the di-leucine motif is required for targeting VAChT to varicosities (Santos et al. 2001), but it was not clear whether this was the consequence of altered interactions with the endocytic machinery or mis-sorting of VAChT at the TGN. Dynamin I-K44A impaired clathrin- mediated endocytosis as assessed by Tfn-568 internalization experiments. Based on the determined role of dynamin I in the fission of endocytic vesicles from the plasma membrane (Damke et al. 2001), we expected that dynamin I-K44A should interfere with VAChT trafficking only if the protein reached the plasma membrane. GFP-VAChT was found in the plasma membrane and in large vesicles close to the plasma membrane when coexpressed with dynamin I-K44A suggesting that indeed, during GFP-VAChT traffic, one of the steps depends on the incorporation of the protein in the plasma membrane. These results agree with earlier observations by Tan et al. (1998) that indicated that VAChT is directed to the plasma membrane and undergoes endocytosis in non-neuronal cells.
Interestingly, we find here that the di-leucine motif is required for recruitment of the transporter into the vesicle intermediates generated by inhibition of dynamin I. We also provide evidence that the C-terminal tail of VAChT can interact with AP-2, likely due to an association with the µ subunit of the heterotetrameric AP-2 adaptor complex in a di-leucine dependent fashion. Although we have not shown that this interaction occurs ‘in vivo’ it is tempting to speculate that the di-leucine motif recruitment of VAChT to dynamin I generated vesicles occurs via AP-2. Therefore, interaction of the C-terminal VAChT tail with AP-2 defines an important role for clathrin-mediated endocytosis in the internalization of the protein and is perhaps critical for the generation of functional cholinergic synaptic vesicles.
The AP-2 adaptor complex is formed by α adaptin, β2-adaptin, µ2 adaptin and σ2 adaptin subunits. The complex recognizes both tyrosine (NPXY or YXXØ) and leucine-based motifs present in proteins to be internalized (Robinson and Bonifacino 2001). Di-leucine motifs have been suggested to interact with the β1-adaptin subunit (Rapoport et al. 1998). However, more recently the µ2 adaptin, which primarily recognizes tyrosine-based motifs, was alsodemonstrated to recognize leucine internalization signals (Rodionov and Bakke 1998; Kongsvik et al. 2002). Our results support previous suggestion of a region on the µ2-subunit that canrecognize internalization motifs based on leucine and di-leucine signals. These results however, do not exclude the possibility that other heterotetrameric APs may contribute to VAChT trafficking from the TGN or during exit from endosomes.
The extent by which synaptic proteins need to use the endocytic machinery at the cell body or dendrites before being directed to sites of synaptic contact is unknown. Identification of tubulo-vesicular organelles that carry synaptic vesicle proteins, and are present in the cell body of cultured neurons, led to the suggestion that these organelles bud from the TGN and are then directed to synapses (Nakata et al. 1998). Alternatively, it has been demonstrated that synaptophysin cycles through the plasma membrane and endocytic compartments prior to reaching SLMVs in PC12 cells (Régnier-Vigouroux et al. 1991). Synaptic proteins including synaptobrevin, synaptotagmin, VAChT and VMATs have amino acid motifs that may contribute to the regulation of their trafficking between different membrane compartments of neurons (Grote and Kelly 1996; West et al. 1997; Tan et al. 1998; Blagoveshchenskaya et al. 1999; Cho et al. 2000; Krantz et al. 2000; Krasnov and Enikolopov 2000). Moreover, in non-neuronal cells VAChT reaches the plasma membrane and is internalized to endosomes (Tan et al. 1998). The extensive decrease of GFP-VAChT trafficking to tips of SN56 processes in dynamin I-K44A transfected cells, with accumulation at the plasma membrane and in vesicles close to the plasma membrane, suggests that VAChT, similar to synaptophysin cycles at the cell body and that dynamin I participates in this process. Although there is an extensive change in GFP-VAChT distribution in cells overexpressing dynamin I-K44A, it should be noted thatwe did not test directly whether the patches are in contact with the extracellular milieu.
Why would synaptic vesicle proteins follow a more complicated trafficking pathway through the plasma membrane instead of being directed to processes and varicosities directly? Recent work by Smith and colleagues has indicated that synaptic proteins may travel to synapses as a preformed conglomerate of membrane organelles, suggesting that all the machinery necessary to form a synaptic contact are carried together (Ahmari et al. 2000). In this case, cell body and dendritic endocytosis may be necessary to recruit the correct proteins to these ‘proto-terminals’, as well as to maintain the localization of synaptic vesicle proteins in the mature terminal. Our observations are consistent with the notion that VAChT, and perhaps other synaptic proteins, use cell body endocytic machinery to generate vesicular precursors prior to their targeting to nerve-endings.
We thank Dr S. Laporte for the adaptin constructs and suggestions on the experiments, Dr Bruce Wainer for the gift of SN56 cells, Lee E. Eiden for the VAChT antiserum and G. S. Veloso for help with some experiments. We also thank D. Abraão, A. Pereira, E. E. P. Silva and C. A. Costa for technical assistance. This work was supported by PRONEX, PADCT, CNPq and FAPEMIG.