Differential palmitoylation regulates intracellular patterning of SNAP25

Intracellular-membrane-fusion events in eukaryotes are regulated by members of the soluble N-ethylmaleimide-sensitive fusion protein-attachment protein receptor (SNARE) protein family (Jahn and Scheller, 2006; Sollner et al., 1993), and these proteins are the minimal membrane fusion machinery in vitro (Weber et al., 1998). One of the most extensively characterised membrane fusion pathways is regulated exocytosis, which involves the controlled fusion of intracellular secretory vesicles with the plasma membrane. Regulated exocytosis mediates the secretion of a wide variety of essential molecules, including neurotransmitters, peptides and hormones, in response to specific intracellular signals, most often a rise in cytosolic Ca²⁺ levels. In neurons and neuroendocrine cells, regulated exocytosis is dependent upon the interaction of the plasma membrane SNAREs syntaxin 1 and SNAP25 with the vesicle SNARE VAMP2 (Blasi et al., 1993; Gary et al., 1994; Sadoul et al., 1995; Schiavo et al., 1993; Sollner et al., 1993).

Although SNAP25 has been intensively studied as an exocytotic SNARE protein, more recent work has shown that SNAP25 depletion also inhibits the trafficking of cargo from sorting endosomes to recycling endosomes (REs) in PC12 cells (Aikawa et al., 2006a). Consistent with this, previous work has also highlighted an important role for SNAP23, a ubiquitous homologue of SNAP25, in endosomal recycling of internalised transferrin to the basolateral plasma membrane in Madin–Darby canine kidney (MDCK) cells (Leung et al., 1998). The endosomal function of SNAP25 is supported by an ARF6-dependent cycling pathway, operating between the plasma membrane and the RE and trans Golgi network (TGN) compartments, which ensures a sufficient pool of SNAP25 to support endosomal membrane fusion (Aikawa et al., 2006b); disruption of this pathway inhibits trafficking of cargo to REs (Aikawa et al., 2006a). Although this cycling pathway is central to the intracellular functions of SNAP25, it is not clear how entry into this pathway is achieved.

The majority of SNARE proteins are anchored to membranes by transmembrane domains. SNAP25, however, is an interesting exception and the membrane binding of this protein is regulated by palmitoylation, a common post-translational modification, most often involving the attachment of palmitic acid groups (C16:0) to cysteine residues (Resh, 2006). Previous work defined the minimal palmitoylation domain of SNAP25b, the major SNAP25 isoform expressed in adult brain (Bark et al., 1995); this region comprises residues 85–120 and includes four palmitoylated cysteine residues present at amino acid positions 85, 88, 90 and 92 (Gonzalo et al., 1999). It is important to note that palmitoylation is not simply a hydrophobic membrane anchor. Recent work, in different systems, has highlighted the diverse array of effects that palmitoylation can have on proteins, including regulating intracellular sorting, mediating association with membrane microdomains, regulating protein–protein interactions and modulating protein stability (Greaves and Chamberlain, 2007; Greaves et al., 2009b; Linder and Deschenes, 2007; Resh, 2006). Furthermore, the versatility of palmitoylation as a protein regulator is enhanced by its reversibility, with many proteins undergoing continuous cycles of depalmitoylation and repalmitoylation (Rocks et al., 2005).

Intracellular palmitoylation reactions are mediated by a large family of more than 23 DHHC (for aspartate-histidine-histidine-cysteine) palmitoyl transferases; these proteins are defined by the presence of an ~50-amino-acid cysteine-rich domain containing a DHHC motif (Fukata et al., 2004; Mitchell et al., 2006; Putilina et al., 1999). All DHHC proteins are predicted polytopic membrane proteins, associating with distinct intracellular compartments, including the endoplasmic reticulum (ER), Golgi and plasma membrane (Ohno et al., 2006). As DHHC proteins are membrane-associated, peripheral proteins that undergo palmitoylation require additional primary membrane-targeting signals that mediate membrane interactions before palmitoylation. Primary membrane-targeting signals include N-myristoylation (e.g. certain Gα subunits and SRC family kinases) and prenylation of C-terminal CAAX motifs (e.g. Hras and Nras). SNAP25 is not modified by either N-myristoylation or prenylation, and we have recently shown that the cysteine-rich domain of this protein is important for initial membrane interactions before palmitoylation. In particular, cysteine hydrophobicity and the hydrophobic character of the surrounding amino acids are required for efficient membrane association (Greaves et al., 2009a). Single cysteine-to-alanine mutations lead to a marked reduction in SNAP25 membrane association, whereas replacement of single cysteine residues with the more hydrophobic leucine residue preserves membrane binding (Greaves et al., 2009b). However, double cysteine replacement mutants dramatically inhibit the membrane association of SNAP25, even when leucine is the replacement amino acid. In addition to cysteine residues and the surrounding hydrophobic amino acids, membrane binding of SNAP25 also requires residues at the C-terminus of the minimal-membrane-targeting domain (in particular, Q116 and P117) (Gonzalo et al., 1999; Greaves et al., 2009a). The residue P117 appears to play an important role in determining the specificity of the interaction with DHHC proteins: in human embryonic kidney (HEK)-293 cells, mutation of this residue prevented palmitoylation by DHHC17 but not by DHHC3. There are several ways that P117 might regulate the interaction with DHHC17, including by directly participating in protein–protein interaction, by modulating the strength of initial membrane interaction before palmitoylation or by allowing SNAP25 to associate with specific membrane domains that facilitate an interaction with DHHC17.

The majority of evidence suggests that newly synthesised SNAP25 is palmitoylated at the Golgi (Gonzalo and Linder, 1998; Greaves et al., 2010). Interestingly, recent studies have defined an intracellular cycling pathway operating between the plasma membrane and the Golgi; this pathway is followed by a number of other peripheral palmitoylated proteins, including Hras and Nras (Goodwin et al., 2005; Rocks et al., 2010; Rocks et al., 2005). Following synthesis, Ras proteins are farnesylated on a C-terminal CAAX motif, which mediates transient interactions with intracellular membranes. When the farnesylated protein associates with the Golgi, it is palmitoylated by resident DHHC proteins, promoting stable membrane attachment and delivery to the plasma membrane through vesicular transport. Subsequent depalmitoylation releases Ras from membranes, and the process of palmitoylation at the Golgi is repeated. This cycling between the plasma membrane and the Golgi is thus mediated by a combination of vesicular transport and cytosolic diffusion, and is regulated by palmitoylation–depalmitoylation dynamics. It has been proposed that rapid depalmitoylation and membrane release prevents excessive accumulation of Ras proteins on endosomal membranes (Rocks et al., 2005).

By contrast, there is no evidence that SNAP25 cycles on and off membranes, or that cytosolic diffusion plays any role in trafficking or localisation of mature SNAP25. This difference between SNAP25 and Ras proteins might arise owing to a slower rate of SNAP25 depalmitoylation or might reflect the fact that SNAP25 has more palmitoylation sites and is therefore less likely to be in a fully depalmitoylated (and hence soluble) state at any given time. Studies analysing whether SNAP25 palmitoylation is constitutively dynamic have produced conflicting results, probably owing to the limitations of the techniques employed (Heindel et al., 2003; Kang et al., 2004; Lane and Liu, 1997). However, there is good evidence that SNAP25 palmitoylation is subject to dynamic regulation under some conditions, and palmitate turnover is notably enhanced following inhibition of synaptic activity in cortical neurons (Kang et al., 2004).

Although palmitoylation plays a dominant role in defining the intracellular localisation of proteins such as Ras, far less is understood about how multiple palmitoylation of cysteine-rich domains coordinates the trafficking of soluble proteins such as SNAP25. Similarly, the potential for dynamic interplay between palmitoylated cysteine residues in the regulation of protein sorting is poorly understood (Roy et al., 2005). As regulation of SNAP25 intracellular targeting is central to its dual function at the plasma membrane and endosomes, we have investigated how multiple palmitoylation impacts upon endosomal targeting and whether changes in SNAP25 palmitoylation might function as a switch to regulate this sorting pathway.

A previous study, in PC12 cells, described an intracellular pool of SNAP25 on REs and the TGN that cycles between these compartments and the plasma membrane (Aikawa et al., 2006b). This intracellular pool is readily observed for both endogenous SNAP25 and eGFP-tagged SNAP25b (Fig. 1A). As a first step towards defining the role of palmitoylation in regulating SNAP25 localisation, we sought to confirm the identity of the intracellular SNAP25 pool by quantitative colocalisation analysis. For this, PC12 cells were transfected with eGFP–SNAP25b and stained with antibodies against markers of specific intracellular compartments: GM130 (Golgi), TGN38 (TGN) and Rab11 (RE). Clear colocalisation of eGFP–SNAP25b with both Rab11 and TGN38 was observed, whereas we did not detect any overlap with GM130 (Fig. 1B), in good agreement with previous work (Aikawa et al., 2006b). Analysis of covariance of the fluorescence signals provided a quantitative measure of the colocalisation of SNAP25b with both Rab11 and TGN38 (Fig. 1B). Triple-labelling experiments further confirmed the overlap of eGFP–SNAP25b with the RE and TGN compartments and its absence from the Golgi (Fig. 1C). An intracellular pool that colocalised with Rab11 was also observed for both eGFP–SNAP25a- and eGFP–SNAP23-transfected cells (supplementary material Fig. S1), confirming that association with this intracellular compartment is a conserved feature of all SNAP25 protein isoforms.

As the intracellular functions of SNAP25 depend upon its accumulation on endosomal membranes (Aikawa et al., 2006a), we sought to determine the unique features of SNAP25 that allow it to traffic to RE and TGN membranes. Membrane binding of SNAP25 is mediated by multiple palmitoylation of a cysteine-rich cluster (Veit et al., 1996); the minimal palmitoylation domain for SNAP25b has been mapped to residues 85–120 (Gonzalo et al., 1999). The palmitoylation and membrane-binding domain of SNAP25b is sufficient for plasma membrane targeting; however, its role in RE and TGN localisation is not clear.

PC12 cells were co-transfected with eGFP–SNAP25(85–120) and full-length SNAP25b fused to the mCherry fluorescent protein. This co-transfection strategy allows the localisation of the minimal palmitoylated domain to be directly compared with that of full-length SNAP25 in the same cell. Fig. 2 shows that the intracellular localisation of the minimal palmitoylation domain was almost identical to that of full-length SNAP25, demonstrating that all the information required for RE and TGN targeting is contained within this region of SNAP25. By extension, we can conclude that SNAP25 localisation at the plasma membrane and RE and TGN is independent of its interaction with other SNARE proteins.

The observed localisation of the minimal palmitoylation domain is consistent with the idea that palmitoylation regulates RE and TGN targeting of SNAP25. However, another possibility is that palmitoylation only functions as a membrane anchor, and that amino acids downstream of the modified cysteine residues (residues 93–120) regulate targeting of SNAP25. To examine this possibility further, palmitoylation was blocked by mutating all four cysteine residues in full-length SNAP25 to leucine residues (termed 4CL). As this 4CL mutant is cytosolic (Greaves et al., 2009a), we ligated it to 15 amino acids from the C-terminal membrane-targeting domain of Kras (4CL-KrasMTD). This domain contains a farnesylation signal and a polybasic domain, to mediate membrane binding and plasma membrane delivery (Apolloni et al., 2000) (see Fig. 2A). This cysteine-less 4CL-KrasMTD SNAP25 mutant was targeted to the plasma membrane, in a manner similar to wild-type SNAP25; however, it displayed a clear loss of localisation to RE and TGN membranes (Fig. 2B,C). Taken together, these data support the notion that palmitoylation mediates targeting of SNAP25 to RE and TGN membranes.

To quantify the localisation of these SNAP25 mutants with respect to the wild-type protein, the intensity of the eGFP and mCherry intracellular (RE and TGN) fluorescence (Fi) and total cell fluorescence (Ft) was measured as described in the Materials and Methods section. The intracellular RE and TGN fluorescence was expressed relative to the whole-cell fluorescence for each fluorescent protein (Fi/Ft–Fi). The average cellular ratio of Fi/Ft–Fi(eGFP) to Fi/Ft–Fi(mCherry) was 1.046 for eGFP–SNAP25(85–120) relative to mCherry–SNAP25 (wild-type). This value suggests that the amount of targeting of the minimal palmitoylation domain to the RE and TGN is essentially the same as that of full-length SNAP25. By contrast, the eGFP–4CL-KrasMTD mutant had an average value of 0.581 relative to wild-type SNAP25 (Fig. 2B), demonstrating that this chimaeric construct has a quantifiable loss of RE and TGN association.

The results shown in Fig. 2 suggest that palmitoylation regulates the intracellular patterning of SNAP25. As palmitoylation is a reversible modification (Resh, 2006), it is possible that dynamic changes in SNAP25 palmitoylation might modulate the RE and TGN targeting without displacing the protein from membranes. Previous pulse–chase experiments, examining incorporation of radiolabelled palmitate, have produced conflicting results on whether palmitoylation of SNAP25 is dynamic under basal conditions (Heindel et al., 2003; Kang et al., 2004; Lane and Liu, 1997); these discrepancies are probably a consequence of the insensitivity of pulse–chase protocols to dynamic changes in multiply palmitoylated proteins. As a more sensitive test of whether SNAP25 palmitoylation is dynamic, PC12 cells were treated with the protein synthesis inhibitor cycloheximide for 2 hours before the addition of radiolabelled palmitate. Fig. 3A shows that incorporation of [³H]palmitate into SNAP25 was still readily detected in the presence of cycloheximide, consistent with the notion that mature SNAP25 is undergoing palmitoylation remodelling. The averaged results from three experiments revealed that [³H]palmitate incorporation in the presence of cycloheximide was ~50% that in the absence of the drug (Fig. 3A, right-hand panel). Fig. 3B confirms that cycloheximide effectively blocked protein synthesis, as judged by a complete loss of [³⁵S]methionine incorporation into cellular protein.

As palmitoylation is essential for targeting SNAP25 to RE and TGN membranes (Fig. 2), and because this modification is dynamic (Fig. 3A), we examined whether changes in SNAP25 palmitoylation, brought about by cysteine mutagenesis, affect its localisation. The precise stoichiometry of the palmitoylation of the cysteine-rich domain of SNAP25 is not known; however, [³H]palmitate incorporation was decreased when each of the four cysteine residues were mutated individually to leucine (Fig. 4A). This suggests that all cysteine residues in SNAP25 are modified to some degree by palmitoylation. To analyse whether the mutation of the cysteine residues affects SNAP25 localisation, PC12 cells were co-transfected with mCherry–SNAP25 wild-type and eGFP-tagged versions of each of the cysteine-to-leucine mutants. Interestingly, mutation of C88 and C90 promoted a marked increase in RE and TGN targeting (Fig. 4B; supplementary material Fig. S2), suggesting that dynamic palmitoylation remodelling might play a key role in regulating SNAP25 localisation. In support of this, quantitative analysis of multiple transfected cells revealed an increase in the RE and TGN targeting of all the SNAP25 cysteine mutants, which was substantially higher for the C88L and C90L proteins (Fig. 4B). Thus, decreasing the number of palmitoylation sites in SNAP25 enhances RE and TGN targeting, and specific cysteine residues (i.e. C88 and C90) appear to be more dominant in directing SNAP25 localisation. Importantly, the increased intracellular level of the cysteine mutants was not related to the fluorescent tags, as enhanced RE and TGN targeting of the C88L mutant was still observed when the tags were switched (Fig. 4C).

The results described in Fig. 4 clearly show that RE and TGN accumulation of eGFP–SNAP25b is enhanced when the number of palmitoylated cysteine residues is reduced from four to three. It is interesting to note that SNAP23 has an additional cysteine residue (five rather than four) in its cysteine-rich domain (Fig. 5). To determine whether functional differences between SNAP25b and SNAP23 might arise owing to the effects of the different number of cysteine residues on RE and TGN targeting, we examined a SNAP25b mutant with an extra cysteine (F84C) to mimic SNAP23 (Salaun et al., 2005a) (Fig. 5A). Introduction of this fifth cysteine into SNAP25 had no major effect on the intracellular localisation of SNAP25 (Fig. 5B), and quantitative analysis revealed an average value for Fi/Ft–Fi(eGFP):Fi/Ft–Fi(mCherry) of 1.0±0.022 (n=8).

To test whether the intracellular targeting of other SNAP25 proteins is also regulated by palmitoylation, we examined the localisation of eGFP–SNAP23 with a double cysteine mutation (C79F and C83F), mimicking the cysteine configuration of the SNAP25b (C88L) mutant. The intracellular localisation of this SNAP23 mutant was clearly enhanced compared with that of mCherry–SNAP23 (wild-type) (supplementary material Fig. S3). Thus, palmitoylation regulates the precise intracellular localisation of distinct SNAP25 proteins.

To ensure that the increased intracellular accumulation of the SNAP25 cysteine mutants did not reflect association with the Golgi, PC12 cells expressing eGFP–SNAP25(C88L) were stained with anti-GM130 antibodies. No colocalisation of the cysteine mutant with the Golgi marker was apparent (Fig. 6A). We also examined the possibility that the enhanced intracellular accumulation of the SNAP25(C88L) mutant reflected a lag in the rate of transport of newly synthesised protein to the plasma membrane, rather than a shift in the cycling of mature SNAP25 between the plasma membrane and the REs and TGN. For this, cells transfected with eGFP–SNAP25(C88L) and mCherry–SNAP25 (wild-type) were incubated in the absence or presence of cycloheximide for 3 hours before fixation. Cycloheximide treatment did not reduce the intracellular accumulation of SNAP25(C88L) compared with that of wild-type SNAP25 (Fig. 6B), suggesting that it is cycling of SNAP25 that is affected by this mutation.

Our analysis of the effects of cysteine mutagenesis on SNAP25 localisation is limited to single point mutations, as removal of more than one cysteine prevents membrane binding (Greaves et al., 2009a). It is important to note that this effect of double cysteine mutations is probably attributable to defects in the initial membrane interaction of SNAP25 before palmitoylation. Indeed, two palmitoylation sites should be sufficient for tight membrane binding (Shahinian and Silvius, 1995). To extend our investigation into how palmitoylation regulates SNAP25 localisation, we generated mutant proteins that would permit palmitoylation to be studied without confounding effects on primary membrane-targeting signals. For this, SNAP25 was truncated immediately following the cysteine-rich domain and attached to the farnesylated CAAX motif (CVLS) of Hras (Fig. 7A). The lysine preceding the CAAX motif of Hras was also added to the SNAP25(1–92) truncation mutant to maintain the spacing between the farnesylated cysteine residue and the most proximal palmitoylated cysteine residue (Fig. 7A).

Removal of amino acids downstream from the cysteine-rich region has previously been shown to prevent the stable membrane association of SNAP25 (Gonzalo et al., 1999; Greaves et al., 2009a), and membrane targeting of SNAP25 requires residues 85–120 (Gonzalo et al., 1999). Thus, SNAP25(1–92) displayed a dispersed localisation, consistent with a loss of membrane binding (Fig. 7B). There appeared to be a small amount of this construct associated with membranes, which probably reflects the weak membrane affinity of the intact cysteine-rich domain. By contrast, the SNAP25(1–92)-CAAX construct displayed an almost identical localisation to that of mCherry–SNAP25 (Fig. 7B). Thus, the dominant role of the SNAP25 cysteine-rich domain in directing intracellular targeting is apparent even when it is placed in a different sequence context. Furthermore, as observed with the full-length protein (Fig. 4), introduction of a C88L mutation into the chimaeric protein promoted a significant increase in the association with RE and TGN membranes, whereas a C85L mutation only had a minor effect (Fig. 7B).

The initial membrane interactions of the SNAP25(1–92)-CAAX mutants will be mediated by farnesylation, allowing greater scope to investigate whether SNAP25 localisation can be modified further by removing additional palmitoylation sites. Thus, we combined the C85L and C88L mutations, and observed a substantial increase in intracellular targeting compared with that for the single cysteine residue mutants (Fig. 7B), further highlighting the negative correlation between the number of palmitoylation sites and RE and TGN targeting. Fig. 7C shows that the localisation of the SNAP25(1–92,C85L,C88L)-CAAX mutant did not overlap with GM130, demonstrating that the intracellular accumulation of this mutant does not reflect Golgi targeting. The targeting of the SNAP25(1–92)-CAAX chimaeras to the plasma membrane and defined intracellular compartments was entirely dependent upon palmitoylation, as removal of all of the palmitoylation sites (4CL) led to a dispersed localisation (Fig. 7B). Some association of the SNAP25(4CL)-CAAX construct with intracellular membranes was visible, although this was largely distinct from the RE and TGN membranes that were labelled by wild-type SNAP25. Previous work has shown transient association of the CAAX motif of Ras with ER and Golgi membranes, and it is these membranes that are likely to be labelled by the 4CL-CAAX protein (Rocks et al., 2010).

The number and configuration of cysteine residues in the SNAP25(1–92,C85L,C88L)-CAAX mutant is similar to that of Hras (Fig. 7A). However, the localisation of the membrane-targeting domain of Hras (Fig. 7A, tH) was completely distinct from that of this SNAP25 mutant and only a very small fraction of Hras associated with the same intracellular membranes as wild-type SNAP25 (Fig. 7B). This result demonstrates that subtle differences in palmitoylation motifs can have dramatic effects on intracellular targeting.

This study highlights a hitherto unknown function for multiple palmitoylation in regulating the targeting of SNAP25 to RE and TGN membranes (Fig. 8). It is clear from this analysis that palmitoylation does not simply function as a membrane anchor for SNAP25 but also dictates the precise patterning of the protein across intracellular membranes. By employing cysteine residue mutagenesis, we found that decreasing the number of palmitoylation sites increased RE and TGN targeting. As palmitoylation is a reversible process, these observations highlight differential palmitoylation as a mechanism to regulate targeting of SNAP25 to specific intracellular membranes.

A recent study suggested that Golgi enzymes play a major role in regulating the palmitoylation of peripheral membrane proteins (Rocks et al., 2010). Consistent with this, Golgi integrity and Golgi DHHC proteins are important for palmitoylation of newly synthesised SNAP25 (Gonzalo and Linder, 1998; Greaves et al., 2010; Greaves et al., 2008). Following palmitoylation at the Golgi, SNAP25 is probably delivered to the plasma membrane by vesicular transport; however, it is not clear whether SNAP25 traffics directly to the plasma membrane from the Golgi or whether it transits via other compartments such as REs and the TGN. It is clear that SNAP25 levels at the RE and TGN compartments are not maintained by new protein synthesis, as cycloheximide had very little effect on this localisation (Aikawa et al., 2006b; Gonzalo and Linder, 1998). Instead, a constitutive cycling pathway replenishes RE and TGN compartments with SNAP25 derived from the plasma membrane (Aikawa et al., 2006b). The continued enrichment of the SNAP25(C88L) mutant at the RE and TGN compartments following cycloheximide treatment suggests that changes in plasma membrane to RE and TGN cycling contribute to the increased intracellular targeting of this mutant. Thus, it appears that palmitoylation facilitates entry of SNAP25 into the plasma membrane–RE-TGN cycling pathway but that a step of this cycling is refractive to the extent of SNAP25 palmitoylation. Interestingly, Martin and co-workers (Aikawa et al., 2006b) previously suggested that palmitoylation turnover on SNAP25 might represent a possible mechanism to regulate cycling between the plasma membrane and the REs and TGN.

The altered distribution of the SNAP25 cysteine mutants is consistent with an increased rate of internalisation or a decreased rate of exit from the RE and TGN. Many effects of palmitoylation on modified proteins are thought to reflect changes in association with cholesterol-rich membrane domains (Greaves and Chamberlain, 2007; Melkonian et al., 1999). These domains might also be intricately linked with the effects of differential palmitoylation on SNAP25 sorting, for example, by regulating SNAP25 exit from RE and TGN membranes. It should be noted, however, that we have previously shown that SNAP25(F84C) displays a higher in vitro association with cholesterol-rich membranes than that of wild-type SNAP25 (Salaun et al., 2005b). The observation that the SNAP25(F84C) mutant did not display a marked reduction in RE and TGN localisation might imply that an association with cholesterol-rich domains is not the sole factor accountable for the effects of palmitoylation on SNAP25 localisation. However, it is also possible that SNAP25 molecules with either four or five palmitoylated cysteine residues display the same affinity for cholesterol-rich domains, whereas three palmitoylated cysteine residues might support a weaker association with these domains. Thus, DHHC proteins at RE and TGN membranes might catalyse the complete palmitoylation of wild-type SNAP25 and the SNAP25(F84C) mutant, mediating a similar association with cholesterol-rich domains and facilitating RE and TGN exit. In this case the different association of SNAP25(F84C) and wild-type SNAP25 with cholesterol-rich domains in vitro could reflect heterogeneous palmitoylation of the proteins and the likelihood that more F84C molecules than wild-type SNAP25 will have at least four palmitate chains. In future studies, it will be interesting to determine the precise stage of trafficking that is perturbed by cysteine mutagenesis; this could be assessed using fluorescence photobleaching/photoactivation approaches.

Although it is clear that many proteins undergo rapid cycles of palmitoylation and depalmitoylation, there has been some debate about whether this is also true for SNAP25. In PC12 cells, the half-life of palmitoylation was reported to be ~3 hours compared with the half-life of the protein of ~8 hours (Lane and Liu, 1997). However, two subsequent reports did not detect constitutive palmitate turnover on SNAP25 (Heindel et al., 2003; Kang et al., 2004). It is important to emphasise that these studies all employed pulse–chase experiments to analyse the rate of palmitate turnover, examining the loss of the ³H radiolabel over time. A significant issue with pulse–chase analyses is the inherent lack of sensitivity to palmitoylation changes in multiply palmitoylated proteins. For example, if one palmitate group is removed from 40% of intracellular SNAP25, this might only amount to a 10% loss of radiolabel (Prescott et al., 2009). This level of change in palmitoylation status might have a major effect on SNAP25 trafficking or function but could not be reliably detected using pulse–chase experiments. Furthermore, recent analyses of palmitoylated proteins that undergo dynamic palmitoylation suggest that pulse–chase experiments also underestimate palmitate turnover (Rocks et al., 2005). As an alternative approach to test whether SNAP25 palmitoylation is dynamic, we examined [³H]palmitate incorporation in the presence or absence of cycloheximide to block protein synthesis. A previous study showed that palmitoylation of SNAP25 was largely complete following 1 hour of labelling and 2 hours of chase (Gonzalo and Linder, 1998). We therefore treated cells with cycloheximide for 2 hours before the addition of [³H]palmitate, in the continued presence of cycloheximide, for a further 4 hours. It is unlikely that the level of palmitoylation observed following cycloheximide treatment (50% of control values) could be due to incorporation into residual SNAP25 that had yet to be palmitoylated. Thus, we propose that SNAP25 does undergo constitutive palmitoylation remodelling and that this can be enhanced further, for example, in response to synaptic activity (Kang et al., 2004).

Previous studies have highlighted a dynamic Ras cycling pathway, operating between the Golgi and plasma membrane, which also appears to be followed by other peripherally localized palmitoylated proteins. The rapid palmitoylation–depalmitoylation dynamics of Hras and Nras, which occurs on the time-scale of a few minutes, is proposed to mediate the distribution of the proteins over the Golgi and plasma membrane and prevent excessive accumulation on endosomal membranes (Rocks et al., 2010; Rocks et al., 2005). In agreement with this idea, we observed that association of Hras with RE and TGN membranes was markedly lower than that of SNAP25. However, in PC12 cells, we did not detect an obvious Golgi pool of Hras, perhaps reflecting the high palmitoylation capacity of PC12 Golgi membranes and rapid Golgi exit dynamics.

The results obtained with the SNAP25 truncation mutants fused to the Hras CAAX motif are interesting for several reasons. The inverse correlation observed between cysteine number and targeting to RE and TGN membranes was in good agreement with our data for full-length SNAP25, and further suggests that the cysteine-rich domain of SNAP25 is an autonomous unit that is dominant and sufficient for determining SNAP25 localisation. By fusing the SNAP25 cysteine-rich domain to a CAAX motif, we were also able to investigate how combined cysteine mutations affect the localisation of SNAP25. The C85L,C88L-CAAX mutant displayed a substantial increase in intracellular localisation, far greater than that of either the single C85L-CAAX or C88L-CAAX mutants. This finding suggests that the effects of differential palmitoylation on RE and TGN targeting can be fine-tuned by the precise number and distribution of palmitoylated cysteine residues. The localisation of the C85L,C88L-CAAX mutant was intriguing as it was strikingly different to the localisation of Hras, even though the proteins had very similar palmitoylation motifs (Fig. 7A). On the basis of current models for the trafficking of peripheral palmitoylated proteins (Rocks et al., 2010), it might have been predicted that reducing the number of cysteine residues in the SNAP25-CAAX construct would change the distribution so that it was similar to Hras. Thus, this work clearly demonstrates that subtle differences in the palmitoylation domains of peripheral proteins can have major effects on intracellular trafficking.

Whereas Ras proteins undergo cytosolic diffusion and repalmitoylation at the Golgi, mature SNAP25 is not thought to undergo membrane–cytosol exchange. This might reflect a slower rate of depalmitoylation of SNAP25 or the fact that SNAP25 has more palmitoylation sites than Ras. Different rates of depalmitoylation might partly explain the distinct localisations observed for Hras protein and the SNAP25(C85L,C88L)-CAAX mutant. However, SNAP25 could also be subject to modification by post-Golgi DHHC proteins, preventing complete depalmitoylation and membrane release. Indeed, we recently reported that DHHC2, which localises to the plasma membrane in PC12 cells, is active against the SNAP25 protein family (Greaves et al., 2010); this DHHC protein is therefore a prime candidate to regulate SNAP25 palmitoylation dynamics and intracellular sorting. By contrast, the rapid depalmitoylation and cytosolic exchange of Ras proteins might indicate that these proteins are not regulated by post-Golgi DHHC proteins but this issue merits further investigation.

This work highlights the evolving view that palmitoylation should not be regarded simply as a mechanism to attach soluble proteins to membranes and emphasises how subtle differences in the number and configuration of palmitoylated cysteines can have major effects on protein trafficking. We predict that this post-translational modification will have many more intricate effects on SNAP25 targeting and function than are currently recognised.

MP Biomedicals supplied [³⁵S]L-methione and PerkinElmer provided [³H]palmitate. Cycloheximide was purchased from Sigma. Lipofectamine 2000, cell culture media and serum, rabbit anti-Rab11 antibody and Alexa-Fluor-conjugated secondary antibodies were from Invitrogen. Anti-SNAP25 monoclonal antibody was from Covance. Monoclonal antibodies against GM130 and TGN38 were supplied by BD Biosciences.

SNAP25b wild-type and cysteine mutants, cloned into pEGFPC2, were as previously described (Greaves et al., 2009a). The region encoding amino acids 85–120 of SNAP25b was amplified by PCR and cloned into pEGFPC2. To generate the 4CL-KrasMTD construct, a 4CL mutant was used as a template and an additional 45 nucleotides, encoding the amino acid sequence GKKKKKKSKTKCVIM from Kras, was inserted immediately upstream of the stop codon by site-directed mutagenesis. A plasmid encoding the membrane-targeting domain of Hras (tH) fused to the C-terminus of eGFP was kindly provided by Ian Prior (University of Liverpool, UK). SNAP25 mutants containing the KCVLS sequence from Hras were generated by site-directed mutagenesis of the pEGFPC2-SNAP25b construct. A GFP–Rab11 construct, provided by Giampietro Schiavo (London Research Institute, UK), was used to generate mCherry–Rab11. All constructs were verified by DNA sequencing (University of Dundee DNA sequencing service, Dundee, UK).

PC12 cells were grown in RPMI1640 with 10% horse serum and 5% fetal calf serum, in a humidified atmosphere containing 7.5% CO₂. Lipofectamine 2000 was used for transfections of 0.2 μg of each plasmid per coverslip.

For comparison of the intracellular localisations of SNAP25 wild-type and mutant proteins, PC12 cells were co-transfected with 0.2 μg each of the respective mCherry- and eGFP-tagged plasmids. Cells were fixed in 4% formaldehyde at 48 hours post-transfection. For antibody staining, fixed cells were permeabilised for 6 minutes in 0.25% Triton X-100 (in PBS with 0.3% BSA). Antibodies against GM130, TGN38, Rab11 and SNAP25 were used at a dilution of 1:50, and Alexa-Fluor-633-conjugated anti-(rabbit Ig) and anti-(mouse Ig) secondary antibodies were used at 1:400. Antibody incubations were for 60 minutes at room temperature.

Image data were acquired on Zeiss Axiovert and Leica SP5 confocal microscopes at Nyquist sampling rates and were deconvolved using Huygens software (Scientific Volume Imaging). Image J software was used to calculate Pearson's correlation coefficient (R) values for analysis of the co-variance of the fluorescence signals between the intracellular pool of eGFP–SNAP25 and endogenous antibody-labelled GM130, TGN38 and Rab11. To limit the contribution of SNAP25 plasma membrane fluorescence to the calculated R values, this analysis was performed on a region of interest that encompassed the fluorescent signals from the intracellular compartments. The ‘zoom’ images shown in Fig. 1 are representative of single slices from such regions of interest. Pearson's values were calculated from the region of interest applied to whole-cell image stacks. To quantify the changes to the localisation of defined SNAP25 mutants compared with that of wild-type SNAP25, summed z-projections of eGFP and mCherry fluorescence were acquired. The intracellular fluorescence intensity (Fi) and total cellular fluorescence intensity (Ft) were measured from identical regions of the eGFP and mCherry projection images using ImageJ software. The intracellular fluorescence was expressed as a fraction of the total fluorescence for each channel (Fi/Ft–Fi) and a ratio was calculated for this value for eGFP:mCherry (a ratio of 1 indicates identical levels of eGFP and mCherry constructs in the intracellular endosomal compartment, a ratio of greater than 1 indicates an increase in the eGFP fluorescence in the intracellular pool, and a value lower than 1 highlighted an increased level of mCherry fluorescence in the intracellular pool). Fluorescence images from cells coexpressing wild-type and mutant proteins were presented in pseudocolour using the Image J software.

Image analysis was performed in the CALM and IMPACT confocal imaging facilities at the University of Edinburgh. We thank Giampietro Schiavo for the kind gift of GFP–Rab11. This work was supported by a senior non-clinical fellowship award to L.H.C. from the Medical Research Council. Deposited in PMC for release after 6 months.