S-acylation regulates the trafficking and stability of the unconventional Q-SNARE STX19 Free

Over the past 30 years, the majority of the molecular machinery required for intracellular protein transport has been identified and functionally characterised. However, our understanding of how this machinery is regulated by post-translational modifications is less clear. This is especially true for lipid-based modifications, which are generally more difficult to study. It is estimated that up to 10% of the human proteome may be S-acylated (Blanc et al., 2015), and the number of confirmed S-acylated proteins is increasing rapidly due to proteomic-based approaches (Collins et al., 2017; Martin and Cravatt, 2009; Martin et al., 2011). However, for the majority of these proteins, it is unclear what role S-acylation plays in regulating their function. S-acylation is a reversible post-translational modification of free cysteine residues with fatty acids, predominantly palmitic acid (Smotrys and Linder, 2004); hence the process is also referred to as palmitoylation. However, other fatty acids with different chain lengths and unsaturation can also be added to S-acylated proteins (Muszbek et al., 1999). S-acylation not only targets peripheral proteins to membranes but also regulates protein trafficking, turnover and cell signalling (Akimzhanov and Boehning, 2015; Linder and Deschenes, 2007; Misaki et al., 2010; Schmick et al., 2014).

S-acylation, in humans, is catalysed by a family of 23 S-acyltransferases (Fukata et al., 2004). These enzymes are multi-spanning transmembrane proteins that are localised to a broad range of intracellular membranes including the ER, Golgi and plasma membrane (Greaves and Chamberlain, 2011a; Ohno et al., 2006). All S-acyltransferases (also known as zDHHC enzymes) have a highly conserved zinc finger-like domain that contains the amino acid sequence DHHC (Asp-His-His-Cys) which is essential for their activity. On the whole, it is not well understood how these enzymes recognise their substrates, so the extent of substrate specificity remains unclear. S-acylation is the only lipid-based modification that is reversible, and there are several examples where a cycle of S-acylation and de-S-acylation is critical for regulating trafficking and signalling (Akimzhanov and Boehning, 2015; Schmick et al., 2014). Two serine hydrolases, APT1 and APT2 (also known as LYPLA1 and LYPLA2) were initially thought to be the only enzymes that catalyse de-S-acylation in the cytoplasm (Duncan and Gilman, 1998; Tomatis et al., 2010). However, there is increasing evidence that there may be many more serine hydrolases involved in this process (Lin and Conibear, 2015; Martin et al., 2011; Yokoi et al., 2016).

Intracellular membrane fusion is driven by a family of evolutionarily conserved proteins known as SNAREs (Jahn and Scheller, 2006). The majority of SNAREs (32 out of 38) have a transmembrane domain at their C-termini, which targets them to membranes (Bock et al., 2001; Kloepper et al., 2007). However, there are several SNAREs that lack a membrane-spanning sequence. These include the R-SNARE Ykt6, the Q_a-SNAREs syntaxin 11 (STX11) and syntaxin 19 (STX19), and the Q_bc-SNAREs SNAP23, SNAP25, SNAP29 and SNAP47. Most of these SNAREs are targeted to membranes via the post-translational lipidation of conserved cysteine residues. For example, Ykt6 is both farnesylated and S-acylated (Fukasawa et al., 2004). SNAP23 and SNAP25 are S-acylated at a cysteine-rich domain located just downstream of the first SNARE motif (Greaves et al., 2010), and STX11 is S-acylated at its C-terminus (Hellewell et al., 2014). There is significant evidence that the lipidation of these SNAREs not only regulates their membrane targeting but also regulates their trafficking and function. For example, S-acylation of Ykt6 at C194 leads to a conformational switch that exposes the SNARE motif and makes it accessible for interaction with other SNAREs (Fukasawa et al., 2004; Gordon et al., 2017; Wen et al., 2010). Altering the levels of SNAP25 S-acylation by mutating key cysteine residues changes its localisation, suggesting that S-acylation can regulate SNAP25 trafficking (Greaves and Chamberlain, 2011b). S-acylation is also required for the correct targeting of STX11 to the immune synapse (Hellewell et al., 2014), where STX11 is thought to play a role in the regulation of lytic granule fusion with the plasma membrane (D'Orlando et al., 2013). Loss of STX11 function in humans causes the immunoproliferative disorder FHL4 (zur Stadt et al., 2005). Several mutations that cause this disorder result in the truncation of the cysteine-rich domain of STX11 (Hellewell et al., 2014), highlighting the importance of S-acylation for STX11 function.

STX19 is a close homologue of STX11 (38% amino acid identity; Wang et al., 2006) and has been identified as an S-acylated protein in a global proteomics screen (Kang et al., 2008). However, it is unclear which residues in STX19 are S-acylated and what role S-acylation has in regulating STX19 trafficking and function. Like STX11, STX19 has a restricted tissue distribution and its mRNA is predominantly detected in mucosal epithelium of the gut and the skin (Wang et al., 2006). Work from our group and others has shown that depletion of STX19 by means of siRNA perturbs the constitutive secretion of both soluble and transmembrane-anchored cargo (Gordon et al., 2010; Simpson et al., 2012). Immunoprecipitation studies have found that STX19 interacts with post-Golgi SNAREs thought to be involved in fusion with the cell surface, suggesting that STX19 may be a novel plasma membrane-localised Q-SNARE involved in biosynthetic transport (Gordon et al., 2010).

In this study, we sought to gain an insight into the function of STX19 by elucidating where STX19 is localised within the cell and determining how S-acylation regulates its trafficking. We have found that the highly conserved cysteine-rich domain of STX19 (KKRNPCRVLCCWCCPCCSSK) is S-acylated at several cysteine residues and is necessary and sufficient for targeting STX19 to tubular recycling endosomes and the plasma membrane. Blocking S-acylation by either treating cells with the 2-bromopalmitate or mutating cysteine residues in the cysteine-rich domain alters the steady-state localisation of STX19. In addition, these perturbations reduce expression levels of STX19 suggesting that S-acylation regulates STX19 stability. The S-acylation of STX19 can be catalysed by several Golgi-localised zDHHCs suggesting that STX19 is first S-acylated at the Golgi and subsequently traffics to the cell surface and tubular recycling endosomes. Overexpression of GFP-tagged STX19 perturbs the trafficking of Rab8 (which has Rab8a and Rab8b forms) suggesting that STX19 may be involved in the fusion of Rab8-positive vesicles with the plasma membrane.

To gain insight into the localisation of STX19 and its potential function, we transiently transfected HA-tagged STX19 into HeLaM cells (Fig. 1A). As previously reported, we observed recombinant STX19 at the plasma membrane and on intracellular membranes (Gordon et al., 2010; Wang et al., 2006). The intracellular labelling was very striking, with STX19 present on branched tubular structures that were between 2 and 20 μm in length and which emanated from the perinuclear region of the cell. A similar localisation pattern was also observed with a C-terminally tagged construct, indicating that the position of the tag did not interfere with STX19 targeting (Fig. 1A). To determine whether endogenous STX19 is also localised to branched tubular membranes, we generated a rabbit polyclonal antibody to STX19 and performed immunostaining. Like recombinant STX19, we observed a pool of endogenous STX19 localised to branched tubular structures. The tubular staining is specific, as the signal could be depleted using a previously validated siRNA targeting STX19 (Fig. 1B) (Gordon et al., 2010).

Endocytic recycling machinery proteins, such as Rab8 and MICAL-L1, are localised to tubular recycling endosomes (Rahajeng et al., 2012; Sharma et al., 2009). These endosomes contain endocytosed proteins, such as major histocompatibility complex class I (MHC-I) molecules and the glycophosphatidylinositol (GPI)-anchored proteins (CD55 and CD59). Thus, we hypothesised that STX19 may also be localised to this compartment. To determine whether this is the case, we performed colocalisation experiments with endogenous STX19, MICAL-L1, CD55 and recombinant Rab8 (Fig. 1C). As previously reported, MICAL-L1 and Rab8 were found on branched tubular membranes in the majority of cells. As predicted, these structures colocalise with endogenous STX19. CD55 is predominantly found at the cell surface in HelaM cells. However, a small intracellular pool that colocalises with STX19 can be observed. Taken together, our results indicate that a pool of endogenous STX19 is found on tubular recycling endosomes.

STX19 is a Q_a-SNARE that is predicted to have a folded H_abc domain and a single SNARE motif. Unlike the majority of Q_a-SNAREs, STX19 lacks the standard hydrophobic stretch of amino acids at its C-terminus that act as a membrane anchor. However, STX19 has a highly conserved cysteine-rich region at its C-terminus that has previously been predicted to be lipid modified (Fig. 2A, Table S1) (Wang et al., 2006). The conserved region consists of an upstream basic patch (K/R) followed by a series of cysteine residues separated by a bulky hydrophobic residue (W/F) and a proline residue. To investigate the importance of the cysteine-rich region in STX19 trafficking, we deleted this region and determined the localisation of the mutant protein. In addition, we also investigated the role of the SNARE motif in STX19 trafficking by deleting this region. Wild-type STX19 (STX19^1-294) was localised to the plasma membrane and Rab8-positive tubules, whereas the mutant construct lacking the cysteine-rich domain (STX19^1-277) was found in the cytoplasm (Fig. 2B,C; see Fig. S1 for representative images showing Rab8 co-staining). In addition to the observed mislocalisation, we also observed a significant reduction in the levels of the STX19^1-277 construct, so longer exposure times were required to obtain comparable images for this mutant construct (this phenotype is explored further in Fig. 6). These results suggest that the cysteine-rich domain of STX19 is required for targeting the protein to the plasma membrane and Rab8-positive tubular recycling endosomes. Deletion of the coiled-coil domain from STX19 (STX19^Δ218-271) did not significantly alter the localisation of STX19 suggesting that this domain is dispensable for its correct intracellular localisation (Fig. 2B,C).

To determine whether the cysteine-rich region of STX19 is preferentially targeting STX19 to tubular recycling endosomes, we replaced it with the transmembrane domain of the endosomal Q_a-SNARE STX13 (Prekeris et al., 2000). HeLaM cells were transfected with the hybrid construct (STX19^1-277/STX13^251-276) and the localisation of the protein determined. The hybrid construct was effectively targeted to the plasma membrane but its association with the Rab8-positive tubules was significantly reduced suggesting that the cysteine-rich region provides trafficking information that allows STX19 to be targeted to tubular recycling endosomes (Fig. 2B,C). To explore this observation further, we fused the cysteine-rich region onto GFP. HeLaM cells were transfected with either full-length STX19 (GFP–STX19^1-294) or the cysteine-rich region alone (GFP–STX19^275-294). Both constructs were found at the plasma membrane and on tubular recycling endosomes as defined by colocalisation with Rab8 (Fig. 2D). Taken together, these results indicate that the cysteine-rich region of STX19 is capable of targeting proteins to tubular recycling endosomes without the need for any additional sorting information. Interestingly, other S-acylated proteins, such as HRAS, are also trafficked to tubular recycling endosomes in HeLaM cells, suggesting that S-acylation may have a more general role in targeting proteins to this compartment (Fig. S2).

Our transfection studies indicated that the cysteine-rich domain of STX19 is able to target to membranes independently from the rest of the protein, suggesting that this is the domain most likely to be S-acylated. To test this hypothesis, we inhibited S-acylation in vivo by using 2-bromopalmitate (2-BP) (Resh, 2006). HeLaM cells were transfected with GFP–STX19^1-294 or GFP–STX19^275-294 and treated with 100 µM 2-BP overnight, and the steady-state distribution of the expression constructs were determined using fluorescence microscopy (Fig. 3A,C). In the control cells, GFP–STX19^1-294 and GFP–STX19^275-294 were localised to tubular recycling endosomes and the plasma membrane. However, in the cells treated with 100 µM 2-BP the constructs became predominantly cytosolic (Fig. 3A–D). We also observed that the levels of the constructs in the 2-BP-treated samples were markedly reduced compared to what was seen in the control samples, and the full-length STX19 construct (GFP–STX19^1-294) was most affected. Thus, we used MG132 to inhibit the degradation of the full-length STX19 construct (GFP–STX19^1-294). This degradation phenotype is explored further in Fig. 6. To determine whether S-acylation was also required for targeting endogenous STX19 to tubular recycling endosomes, we treated non-transfected cells with 100 µM 2-BP overnight and performed STX19 immunostaining (Fig. 3E,F). In control cells, STX19 is localised to tubular recycling endosomes where it colocalises with MICAL-L1. In the 2-BP-treated cells STX19 staining is lost from tubular recycling endosomes. It is worth noting that this treatment also perturbs the localisation of MICAL-L1. However, a significant pool of the protein still remains associated with the tubular recycling endosomes although they appear shorter in nature. Our results suggest that S-acylation plays an important role in targeting both recombinant and endogenous STX19 to tubular recycling endosomes.

STX19 has eight cysteine residues, with seven of them found in the conserved cysteine-rich domain at its C-terminus. To identify which of the seven cysteine residues are S-acylated, we developed an immunoprecipitation acyl-release assay. GFP–STX19^1-294 was transfected into HEK-293T cells and then immuno-isolated using GFP-Trap beads. The samples were reduced with TCEP (which does not affect acyl thioester bonds; Yang et al., 2010) and then alkylated to block all free (unmodified) cysteine residues. Samples were then treated with hydroxylamine at neutral pH to selectively cleave lipid–protein thioester bonds. The protein was then eluted from the resin and digested into peptides using trypsin. Digests were analysed by liquid chromatography tandem mass spectrometry (LC-MS/MS) and the positions of S-acylated residues inferred by quantitative analysis of the relative levels of non-alkylated cysteine residues (i.e. protected from alkylation by S-acylation) with and without hydroxylamine treatment. Analysis of the mass spectrometry data for GFP–STX19^1-294 identified C-terminal tryptic peptides (VLCCWCCPCCSSK) that showed evidence of S-acylation at several sites including a peptide where C284, C285, C287, C288, C290 and C291 were modified (Table S2). However, owing to technical problems, we were not able to generate enough material to produce reproducible data for the different acylated forms of full-length STX19. Thus, we repeated the experiments with GFP–STX19^275-294. As above, we identified C-terminal tryptic peptides (VLCCWCCPCCSSK) that showed evidence of S-acylation (Fig. 3G,H; Table S2). Several forms of the peptide were identified including unmodified (six alkylated cysteine residues), two S-acylation sites (four alkylated cysteine residues), three S-acylation sites (three alkylated cysteine residues), four S-acylation sites (two alkylated cysteine residues), five S-acylation sites (1 alkylated cysteine residues), six S-acylation sites (no alkylated cysteine residues) at a false discovery rate of 1% and a posterior error probability <0.01. Peptides containing residues predicted to be S-acylated were either exclusively identified in plus hydroxylamine samples or present in both plus and minus hydroxylamine samples but with at least a 5-fold increase in abundance after hydroxylamine treatment (Fig. 3G). The presence of a signal in the control (minus hydroxylamine samples) is likely arising from low-level loss of S-acylation during incubation steps (Jones et al., 2012). C284, C285, C287, C288, C290 and C291 all showed evidence of S-acylation (false localisation rate of 1%) (Fig. 3G). However, no evidence of S-acylation was observed for C280. The distribution of S-acylated forms of STX19 is striking in that the majority of STX19 is S-acylated at five or six cysteine residues and that the fractional occupancy of S-acylation (ratio of S-acylated to the total intensity of the peptide VLCCWCCPCCSSK) is high; 87% of STX19 was modified with at least four acyl groups (Fig. 3G). These results indicate that the six most C-terminal cysteine residues of STX19 can be S-acylated.

S-acylation is the only lipid modification that is reversible, and proteomic-based approaches have measured that up to 10% of the S-acyl-proteome is actively undergoing a cycle of S-acylation and de-S-acylation (Martin et al., 2011). Thus, we wondered whether STX19 could also be cycling on and off membranes in a similar way to what is seen for RAS and Ykt6 (Fukasawa et al., 2004; Lin and Conibear, 2015). To test this idea, we made use of a knock sideways approach, which is a protein re-targeting method (Robinson and Hirst, 2013). In this system, FKBP-tagged peripheral proteins are re-routed on to mitochondria that are expressing an FRB domain construct in the presence of rapamycin. For example, FKBP-tagged AP2 adaptor complex is completely rerouted from the plasma membrane on to mitochondria within 10 min of treatment with rapamycin (Robinson et al., 2010). Thus, if STX19 is cycling on and off membranes, we should be able to use this approach to re-direct it to mitochondria. HeLaM cells were transfected with FKBP–GFP (positive control), FKBP–Myc–STX19^1-294 and two constructs that cannot be S-acylated, FKBP–Myc–STX19^1-277 and FKBP–Myc–STX19^P1+P2+P3 (see Fig. 5 for fuller description of this construct). The cells were then treated with or without rapamycin for 10 min at 37°C, and fixed and stained for the indicated constructs. As expected, FKBP–GFP was efficiently retargeted to mitochondria (Fig. 4A). However, S-acylated STX19 (FKBP–Myc–STX19^1-294) remained associated with the plasma membrane (Fig. 4B), even after incubations of up to 1 h (data not shown). We also observed that, in some cells, it was the mitochondria that were rerouted to the plasma membrane, suggesting that STX19 cannot be easily removed from membranes (Fig. 4E). When we repeated these experiments with mutant forms of STX19 that cannot be S-acylated (FKBP–Myc–STX19^1-277 and FKBP–Myc–STX19^P1+P2+P3), we observed that these constructs were efficiently re-targeted to mitochondria (Fig. 4C,D). Taken together, our results indicate that S-acylated STX19 is stably associated with membranes and is not rapidly cycling on and off membranes.

The mass spectrometry experiments identified that the six most-C-terminal cysteine residues of STX19 can be S-acylated. These cysteine residues are organised into three pairs separated by a tryptophan and a proline residue, respectively (Fig. 5A). To gain an insight into which of the cysteine residues pairs are most important for STX19 targeting, we replaced them with leucine residues, which cannot be S-acylated. We have generated a series of constructs where each cysteine pair (P), was mutated separately (P1, P2 and P3) or in combination (P1+2, P2+3, P1+3 and P1+2+3) (Fig. 5A). The cysteine pairs are numbered from the C-terminus so P1 is a construct where the most-C-terminal cysteine pair was mutated. HeLaM cells were transfected with either wild-type or mutant HA-tagged STX19 constructs and then analysed by immunofluorescence microscopy (Fig. 5B,C). When we mutated each cysteine pair separately, no significant difference was observed between the localisation of the wild-type (HA-STX19^1-294) and mutant constructs (P1, P2 and P3), consistent with there being multiple sites of S-acylation, as determined by mass spectrometry. However, when we mutated several cysteine pairs simultaneously (P1+2 and P1+3) we observed a significant effect on STX19 localisation. In these mutants, STX19 is still targeted to the plasma membrane but there was a significant reduction in the number of cells showing tubular staining (Fig. 5C). In contrast, the construct where the last cysteine pair was intact (P2+P3) showed significant tubular staining, suggesting that this cysteine pair is most important for targeting STX19 to tubular recycling endosomes. As would be expected, the construct where all of the cysteine residues had been mutated (P1+2+3) behaved in a similar way to a construct where the cysteine-rich domain was deleted (HA–STX19^1-277). In both cases, the constructs were completely cytosolic and were not targeted to tubular recycling endosomes. As observed above, perturbing S-acylation of STX19 caused the levels of some of the mutant proteins (P1+2, P2+3, P1+3 and P1+2+3) to be reduced (see Fig. 6A) so longer exposure times were needed to obtain some of the images. In summary, our data suggest that the last cysteine pair (P1) is the most critical for targeting STX19 to tubular recycling endosomes. However, other cysteine pairs can compensate for its loss.

From the immunolocalisation studies, it became apparent that interfering with the S-acylation of STX19 perturbed its stability. To investigate this further, we transfected HeLaM cells with either wild-type (HA–STX19^1-294) or the mutant STX19 constructs (HA–STX19 P1, P2, P3, P1+2, P2+3, P1+3, P1+2+3, and HA–STX19^1-277) and performed immunoblotting using antibodies against HA and γ-adaptin as a loading control (Fig. 6A,B). As observed by microscopy, the levels of the mutant constructs, P1+2, P1+3, P1+2+3 and HA–STX19^1-277 were significantly reduced compared to wild-type STX19. We hypothesised that the mutant constructs may be degraded by ubiquitylation, so we repeated the experiment in the presence of the proteasomal inhibitor MG132 (Lee and Goldberg, 1998). Addition of MG132 to the cells significantly increased the levels of the mutant constructs indicating that they are being ubiquitylated and targeted for proteasomal degradation. In support of this hypothesis, we and others have identified that STX19 is ubiquitylated on several lysine residues by mass spectrometry (Fig. S3) (Boeing et al., 2016). Increasing the levels of the mutant constructs using MG132 does not rescue their ability to target to the plasma membrane and/or tubular recycling endosomes indicating that S-acylation is essential for this process (Fig. S4).

Our results suggest that when the cysteine-rich domain of STX19 is not S-acylated it becomes a novel degron. To test this hypothesis, we mutated all the cysteine residue pairs in a construct where the cysteine-rich domain is fused to the C-terminus of GFP (GFP–STX19^{275-294/P1+2+3}). HeLaM cells were transfected with GFP, GFP–STX19^275-294 or GFP–STX19^{275-294/P1+2+3} and their levels determined by immunoblotting (Fig. 6C). The levels of GFP–STX19^{275-294/P1+2+3} were significantly reduced compared to that of the wild-type STX19 reporter. As predicted, the mutant protein is not targeted to the plasma membrane or tubular recycling endosomes (Fig. 6D). It is possible that the degradation phenotype is simply caused by the insertion of multiple leucine residues within the cysteine-rich domain and does not reflect a true physiological process. To exclude this possibility, we used the drug 2-BP to block S-acylation of the wild-type construct. HeLaM cells were transfected with GFP–STX19^275-294 and treated with increasing concentrations of 2-BP overnight. The cells were then harvested and immunoblotting performed using antibodies against GFP and γ-adaptin (Fig. 6E,F). The levels of the reporter construct were dramatically reduced in a dose-dependent manner, suggesting that the cysteine-rich domain of STX19 is a degron that senses whether STX19 is S-acylated. Our immunoblotting results, from the truncated STX19 construct (HA–STX19^1-277), also suggest that there must be an additional degron located upstream of the cysteine-rich domain, as the truncated protein is still degraded even when the cysteine-rich domain is removed. Our results indicate that S-acylation regulates both the stability and membrane targeting of STX19.

S-acylation in mammals is catalysed by 23 S-acyltransferases (Fukata et al., 2004) (Fig. 7G). These enzymes are polytopic membrane proteins that have a conserved zDHHC motif and are localised to various membranes, including the endoplasmic reticulum (ER), Golgi, endosomes and plasma membrane (Ohno et al., 2006). To identify which zDHHC enzymes catalyse STX19 S-acylation, we have made use of an in vitro assay where S-acylation is measured based on the incorporation of 17-octadecynoic acid (17-ODYA) (Martin and Cravatt, 2009; Yap et al., 2010). This label is reacted with the azide dye, IR-800 and the level of labelling quantified using an infrared imaging system. HEK-293T cells were co-transfected with GFP–STX19^1-294 and 23 different S-acyltransferases and the amount of 17-ODYA labelling quantified by measuring the ratio of IR-800 signal to GFP signal (Fig. 7A,C). The samples were also probed for HA to verify the expression of the HA-tagged zDHHCs (Fig. 7B). In the control transfection (vector alone), there is a faint IR-800 signal at the predicted molecular weight for GFP–STX19^1-294, indicating that the protein is being S-acylated by the endogenous S-acyltransferases expressed in HEK-293T cells. This signal was significantly increased when zDHHCs 2, 3, 7, 11, 12 were co-transfected with STX19 suggesting that STX19 is a substrate for these enzymes (Fig. 7C). To determine whether the cysteine-rich domain construct is also S-acylated by the same enzymes, HEK-293T cells were co-transfected with GFP–STX19^275-294 and a subset of the zDHHC library (Fig. 7D–F). As with full-length STX19, the cysteine-rich domain construct is also S-acylated by zDHHCs 2, 3, 7, 11, 12. In addition, zDHHC14 and 15 also significantly increased the IR-800 signal for this construct. These enzymes also increased the level of S-acylation of full-length STX19, but the trend was not significant. To determine which of these enzymes most likely S-acylated STX19 in HeLaM cells, we examined their expression levels by using previously obtained microarray data (Fig. 7H) (Gordon et al., 2010; Kozik et al., 2013). zDHHCs 3, 7, 11 and 12 are all expressed in HeLaM cells. However, zDHHC2 is expressed at very low levels, suggesting that it is unlikely to play a major role in S-acylation in HeLaM cells. zDHHCs 3, 7, 11 and 12 are localised to the Golgi in mammalian cells (Ohno et al., 2006) suggesting that STX19 may be S-acylated at the Golgi and then traffic to the cell surface or tubular recycling endosomes. In support of this hypothesis, we find that overexpression of zDHHCs 3 or 7 significantly increase the Golgi-localised pool of GFP–STX19^275-294 (Fig. S5).

The precise function of STX19 is still unclear. However, endogenous STX19 shows a high degree of colocalisation with Rab8. Thus, we hypothesised that STX19 might play a role in Rab8 trafficking. To test this hypothesis, we overexpressed GFP (control), GFP–STX19^1-294 or GFP–STX19^1-277 (which cannot be S-acylated) and examined the steady-state localisation of Rab8 (Fig. 8A,B). In cells expressing GFP, Rab8 was localised to a range of membranous compartments, including tubular recycling endosomes, the cleavage furrow and the plasma membrane (Peränen, 2011). In cells with low to moderate levels of GFP–STX19^1-294, we found that Rab8 localisation was unperturbed and that in many cells there was good colocalisation with GFP–STX19^1-294 (Fig. 2D). However, in cells with moderate to high expression levels of GFP–STX19^1-294 (1.5-fold mean intensity and above) we found Rab8 to be targeted to membrane protrusions just at or under the plasma membrane (Fig. 8A). In cells overexpressing GFP–STX19^1-277 (which cannot be S-acylated), we did not observe any change in Rab8 localisation, even in cells that had very high expression levels (5-fold mean intensity and above). Our results suggest that STX19 may have a role in Rab8 trafficking and that STX19 must be S-acylated to perturb this pathway.

STX19 is one of the least-characterised Q-SNAREs in the human genome, and there are only a handful of publications that have investigated its biology. Work from several laboratories, including our own has identified that depleting STX19 using siRNA leads to defects in constitutive secretion (Gordon et al., 2010; Simpson et al., 2012) and endosome to trans-Golgi network transport (Breusegem and Seaman, 2014). Biochemical studies from our group have identified that STX19 interacts with the R-SNAREs VAMP3 and VAMP8 and the Q_bc SNAREs SNAP23, SNAP25 and SNAP29 supporting the idea that it has a role in membrane fusion at the plasma membrane. In this study, we have set out to elucidate where STX19 is localised and determine how S-acylation regulates its trafficking and function.

Our localisation data indicate that a pool of endogenous STX19 is localised to tubular recycling endosomes where it colocalises with Rab8 and MICAL-L1, proteins that have been shown to be involved in the endocytic recycling of proteins internalised by non-clathrin-mediated endocytosis (Rahajeng et al., 2012; Sharma et al., 2009). In addition, Rab8 has also been shown to be a key player in determining cell polarity (Vidal-Quadras et al., 2017) and have an involvement in exocytosis (Grigoriev et al., 2011). This is the first time that a SNARE has been localised to this compartment. Currently, it is unclear whether STX19 is: (1) simply recycling through this compartment back to the cell surface, (2) involved in the fusion of endocytic material with the plasma membrane, or (3) required for the fusion of material with tubular recycling endosomes. Overexpression of GFP–STX19 causes Rab8-positive vesicles to accumulate in membrane protrusions just at or under the plasma membrane. Our interpretation of this result is that GFP–STX19 is acting in a dominant-negative manner and is blocking the fusion of Rab8-positive vesicles with the plasma membrane. In the future, it will be interesting to more precisely define the role of STX19 in Rab8 trafficking and cell polarity.

Inhibiting STX19 S-acylation either by genetic or chemical means causes the protein to be degraded via the proteasome. Our experimental data suggest that STX19 contains at least two degrons. The first degron is formed when the cysteine-rich domain is not S-acylated, and the second is located upstream of this region. The degron present in the de-S-acylated cysteine-rich domain can efficiently target stable proteins, such as GFP, for degradation. At present, it is unclear what motifs or residues in STX19 are being recognised by the ubiquitylation machinery. However, the cysteine-rich domain of STX19 contains several bulky hydrophobic residues (tryptophan and phenylalanine) that most likely would not normally be exposed to the cytoplasm when this region is S-acylated. Many S-acylated proteins also contain bulky hydrophobic residues in close proximity to the lipidated cysteine residues (Collins et al., 2017), suggesting that this could be a general mechanism by which de-S-acylated proteins are recognised for degradation. In the future, it will be of interest to determine whether S-acylation is used as a mechanism to post-translationally regulate the levels of STX19.

To identify the enzymes involved in S-acylating STX19, we performed a screen using overexpressed zDHHCs. We found that full-length STX19 and the C-terminal cysteine-rich domain of STX19 are good substrates for several zDHHCs (2, 3, 7, 11 and 12) and both proteins were predominantly S-acylated by the same enzymes. These data indicate that the protein sequence of the cysteine-rich domain contains all the necessary information required for substrate recognition and S-acylation. zDHHCs 3 and 7 have been shown to have a broad substrate specificity and to S-acylate proteins including HRAS, SNAP25, SNAP23 and cysteine string protein (Greaves and Chamberlain, 2011a; Greaves et al., 2010, 2008; Lu et al., 2012). However, very few substrates have been identified for zDHHC11. Based on our expression analysis, zDHHCs 3, 7, 11 and 12 are the enzymes most likely to be involved in S-acylating STX19 in HeLaM cells. Owing to the significant level of redundancy observed between these enzymes (Ohno et al., 2012), we have not attempted to further define which enzymes S-acylate STX19. All of these zDHHCs are localised to the Golgi, suggesting that STX19 is first S-acylated at the Golgi and then traffics to the plasma membrane and/or tubular recycling endosomes.

One of the most striking observations in this study is that we have identified that S-acylation is a key determinant in targeting STX19 to tubular recycling endosomes, and that replacing the lipid anchor with a conventional transmembrane domain, from another Q-SNARE, significantly reduces the targeting of STX19 to this compartment. This data is consistent with S-acylation acting not merely as a membrane anchor but providing information that is critical for STX19 sorting. At present, it is unclear how S-acylation targets STX19 to tubular recycling endosomes. There are several possible mechanisms through which this could be achieved: (1) that S-acylation targets STX19 to lipid microdomains at the cell surface where it is subsequently internalised via non-clathrin-mediated endocytosis; and (2) that STX19 is directly recruited to this compartment from the cytosol through a cycle of S-acylation and de-S-acylation. At present, we favour mechanism 1 because: (a) there is a significant body of literature indicating that S-acylated proteins are associated with lipid microdomains at the cell surface (Chamberlain et al., 2001; Melkonian et al., 1999; Salaün et al., 2004), (b) STX19 is stably associated with membranes, and (c) STX19 is degraded when not S-acylated. We have observed that other S-acylated proteins (HRAS and S-acylated GFP) are also localised to tubular recycling endosomes, suggesting that S-acylation may be a key factor in targeting proteins to this compartment. In the future, it will be interesting to determine how general this mechanism is and what cellular machinery is involved in sorting S-acylated proteins to this compartment.

In conclusion, we have elucidated that S-acylation is critical for targeting STX19 to tubular recycling endosomes and that this modification may have a broader role in sorting proteins to this compartment. In addition, STX19 may also prove useful for those elucidating the mechanism of zDHHC substrate specificity, protein turnover and Rab8-dependent trafficking.

The polyclonal antibody against STX19 was generated by immunising rabbits with GST–STX19^1-100. The antiserum was affinity purified as previously described (Page and Robinson, 1995) and diluted 1:200 for microscopy. An anti-HA mouse monoclonal antibody (16B12) was purchased from BioLegend (cat. no. 901503; 1:200 for microscopy and 1:1000 for immunoblotting). A rabbit polyclonal and monoclonal antibody against Rab8A were purchased from ProteinTech (cat. no. 55296-1-AP; 1:200 for microscopy) and Cell Signalling Technology, respectively (cat. no. rabbit mAb #6975; 1:500 for microscopy). The rabbit polyclonal antibody against GFP (1:200 for microscopy and 1:1000 for immunoblotting) was previously described (Gordon et al., 2017). The Alexa-Fluor-488- and 594-conjugated secondary antibodies for fluorescence microscopy were purchased from ThermoFisher and used at 1:1000. The horseradish peroxidase (HRP)-conjugated secondary antibodies for immunoblotting were purchased from Jackson ImmunoResearch Laboratories and used at 1:5000. For click chemistry analysis, mouse anti-GFP antibody (JL8) was obtained from TaKaRa Clontech (cat. no. 632381, 1:1000 for immunoblotting); rat anti-HA antibody (3F10) was purchased from Sigma (cat. no. 000000011867423001, 1/1000 for immunoblotting) and the IR dye-conjugated secondary antibodies were purchased from LI-COR. All restriction enzymes were purchased from New England Biolabs. All primers were purchased from Sigma-Aldrich. GFP-Trap beads were purchased from Chromotek (Cat. no. gta-20).

Human STX19 (Gene ID 415117) was used as the template for all the constructs used in this study. Wild-type and mutant HA–STX19 expression constructs were generated by PCR and cloned into pIRES-NEO (GenBank accession number U89673). GFP-STX19 expression constructs (GFP–STX19^1-294 and GFP–STX19^1-277) were generated by PCR and cloned into pEGFP-C1 (GenBank accession number U55763). The GFP-tagged cysteine-rich domain construct (GFP–STX19^275-294) was generated by annealing oligonucleotide and ligating them into pEGFP-C3 (GenBank accession number U57607). FKBP–STX19^1-294 and FKBP–STX19^1-277 were generated by PCR and cloned into pIRES-NEO. The HA-tagged zDHHC expression library was a kind gift from Masaki Fukata (Division of Membrane Physiology, National Institute for Physiological Sciences, Japan). The pCAG-mGFP (GFP fused with GAP43 S-acylation sequence) was Addgene #14757 (deposited by Connie Cepko; Matsuda and Cepko, 2007). The GFP–HRAS construct was a kind gift from Ian Prior (Institute of Translational Medicine, University of Liverpool, UK). The MitoTRAP (pMito-mCherry-FRB) and GFP–FKBP constructs were a generous gift from Steve Royle (Cell and Developmental Biology, University of Warwick, UK; Cheeseman et al., 2013). The Strawberry–Rab8 construct was a kind gift from Anna Akhmanova (Cell Biology, University of Utrecht, Netherlands) (Meijering et al., 2011). All constructs were sequenced using the University of Sheffield's Core Genomic Facility.

HeLaM cells were originally obtained from the laboratory of Margaret S. Robinson (Cambridge Institute of Medical Research, University of Cambridge, UK) and HEK-293T (CRL-3216) cells were obtained from ATCC. Both cell lines were routinely tested for mycoplasma contamination. HeLaM and HEK-293T cells were grown in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS, Life Technologies), 100 IU/ml penicillin, 100 μg/ml streptomycin and 2 mM glutamine (Sigma-Aldrich) at 37°C in a 5% CO₂ humidified incubator. In some instances, HeLaM cells were treated with either 5 µM MG132 (Cayman Chemical) or 100 µM 2-bromopalmitate (Sigma-Aldrich). HeLaM and HEK-293 cells were transfected using polyethylenimine (PEI; ratio 1 part DNA to 5 parts PEI). Cells were analysed 24–48 h post transfection. siRNA transfections were performed as in Gordon et al. (2010).

HeLaM cells (∼2×10⁵) were seeded onto glass coverslips and grown overnight. The cells were rapidly fixed, by adding an equal volume of room temperature 4% paraformaldehyde in PBS to the culture media, as soon as the cells were removed from the incubator. After 1 min, the fixative was removed and replaced with fresh fixative. The cells were incubated for a further 15 min at room temperature. The fixative was removed and the cells washed three times with PBS. The cells were incubated for 5 min with 0.1 M glycine in PBS. The cells were permeabilised using 0.1% saponin in PBS containing 5% FBS. The cells were stained with the appropriate primary and secondary antibodies for 1 h at room temperature. The coverslips were mounted onto microscope slides using ProLong Gold antifade reagent (ThermoFisher) and observed with a 60× oil immersion objective (Olympus BX61 motorised wide-field epifluorescence microscope). Images were captured using a Hamamatsu Orca monochrome camera and processed using ImageJ.

Cells expressing GFP–STX19^275-294 were harvested 48 h post transfection by scraping and centrifugation (200 g for 5 min at 4°C). The cell pellets were resuspended in 1 ml of RIPA lysis buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 1 mM EDTA, 1% NP-40, 0.5% sodium deoxycholate, 0.1% SDS and protease inhibitors) and passed through a QIAShredder (Qiagen). The samples were then incubated for 1 hour at 4°C to ensure complete solubilisation. Insoluble material was removed by centrifugation at 17,000 g for 15 min at 4°C. The supernatant was then incubated with GFP-Trap beads for 3 h at 4°C (Rothbauer et al., 2008). The beads were washed three times with RIPA buffer and two times with Tris-buffered saline. The immuno-isolated protein was then reduced and alkylated using 5 mM TCEP and 50 mM iodoacetamide in 50 mM Tris-HCl pH 8. The beads were then washed five times with 50 mM Tris-HCl pH 7.5. The beads were then divided into two samples. Sample (a) was incubated with 1 M hydroxylamine in 50 mM Tris-HCl pH 7.5, to cleave thioester bonds, and sample (b) was incubated with 50 mM Tris-HCl pH 7.5 as a control. The beads were washed three times with 50 mM Tris-HCl pH 7.5. The protein was eluted from the beads using 8 M guanidine-HCl at 95°C for 10 min. The eluted protein was digested for 4 h at 37°C with 0.5 µg of trypsin prepared in 100 mM ammonium bicarbonate solution (pH 8). The peptides were dried overnight using a vacuum concentrator. Peptides were resuspended in 0.5% formic acid for 10 min at room temperature with agitation.

Peptides were analysed by nanoLC-MS/MS on an Orbitrap Elite (ThermoFisher) hybrid mass spectrometer equipped with a nanospray source, coupled with an Ultimate RSLCnano LC System (Dionex). The system was controlled by Xcalibur 2.1 (ThermoFisher) and DCMSLink 2.08 (Dionex) software. Peptides were desalted on-line using a micro-Precolumn cartridge (C18 Pepmap 100, LC Packings), and then separated using a 60 min RP gradient (4–32% acetonitrile and 0.1% formic acid) on an EASY-Spray column of 15 cm×50 µm internal diameter, PepMap C18, 2 µm particles and 100 Å pore size (ThermoFisher). The mass spectrometer was operated with a cycle of one MS (in the Orbitrap) acquired at a resolution of 60,000 at m/z 400, with the top 20 most-abundant multiply charged (2+ and higher) ions in a given chromatographic window subjected to collision-induced dissociation (CID) fragmentation in the linear ion trap. Fourier-transform MS (FTMS) target values of 10⁶ and an ion trap MSn target value of 10⁴ was used, and the lock mass (445.120025) enabled. Maximum FTMS scan accumulation time of 500 ms and maximum ion trap MSn scan accumulation time of 100 ms were used. Dynamic exclusion was enabled with a repeat duration of 45 s with an exclusion list of 500 and exclusion duration of 30 s. MS data was analysed using MaxQuant (Cox and Mann, 2008) version 1.5.5.1. Data were searched against human UniProt sequence databases (downloaded June 2015) using the following search parameters: trypsin with a maximum of two missed cleavages, 7 ppm for MS mass tolerance, 0.5 Da for MS/MS mass tolerance, with acetyl (protein N-terminus) and oxidation (M) set as variable modifications and carbamidomethyl (C) as a fixed modification. Previously S-acylated sites were identified by the lack of carbamidomethyl (C) (−57.0214), which was set as a variable modification. A maximum of six modifications per peptide was allowed. A protein false discovery rate (FDR) of 0.01 and a peptide FDR of 0.01 was used for identification level cut-offs. Label-free quantification of peptides with or without hydroxylamine treatment was performed using MaxQuant-calculated peptide intensities. S-acylation site occupancies were calculated by expressing peptide intensities of modified peptides against those of their unmodified counterparts.

Cells were washed three times with ice-cold PBS and lysed with SDS sample buffer containing 5% β-mercaptoethanol. The cell lysates were passed through a QIAShredder (Qiagen) three times to sheer chromosomal DNA. The lysates were boiled at 100°C for 10 min. The samples were then separated using SDS-PAGE and transferred overnight onto a 0.45 µm PVDF membrane (GE Healthcare Life Sciences). The membranes were blocked in PBS containing 5% milk powder and 1% Tween 20 at room temperature for 30 min. The membranes were probed with specific antibodies diluted in the blocking buffer. The probed membranes were washed thoroughly with PBS and the damp membranes incubated with Clarity ECL Western blot substrate for 5 min at room temperature. The excess substrate was removed and the signal detected using either X-ray film (Fujifilm) or a LiCOR c-DiGiT. Quantification of the immunoblots was performed using Image Studio Lite Ver 4.0 (Li-COR).

The 17-octadecynoic acid labelling coupled to click chemistry was performed as described previously (Martin and Cravatt, 2009; Yap et al., 2010) with some minor modifications. HEK-293T cells were seeded onto 24-well plates, grown overnight and co-transfected with 23 HA-tagged S-acyltransferases and GFP–STX19^1-294 or GFP–STX19^275-294 (0.8 μg of STX19 construct to 1.6 μg of HA–zDHHC construct). At 24 h post transfection, the cells were serum-starved in DMEM containing 1 mg/ml fatty-acid-free BSA for 30 min at 37°C. The cells were then labelled for 3 h at 37°C with 100 µM 17-octadecynoic acid. The cells were washed two times with ice-cold PBS and then lysed on ice in 50 mM Tris-HCl pH 8.0 containing 0.5% SDS and protease inhibitors. The cell lysates were then incubated with the click reaction buffer (final concentration of: 2.5 µM IR-800 dye, 2 mM CuSO₄, 0.2 mM TBTA and 4 mM ascorbic acid). The click reaction buffer and ascorbic acid were freshly prepared on the day of the experiment. The samples were vortexed and incubated for 1 h at room temperature with end-over-end rotation. After the click reaction, the proteins were precipitated using acetone and the pellets allowed to air-dry for 5 min. The pellets were resuspended in 1× SDS sample buffer containing 25 mM DTT and heated at 95°C for 5 min. The samples were separated by SDS-PAGE (12%) and transferred onto nitrocellulose membranes and probed with antibodies against GFP and HA.

All bar graphs presented in this manuscript were prepared using GraphPad Prism (GraphPad Software). For each experiment, a minimum of three biological repeats were performed. In all graphs the results are mean±s.e.m. An unpaired t-test was performed using GraphPad Prism to calculate P values.

We would like to thank Darren Robinson from Wolfson Light Microscopy Facility for providing technical assistance with the imaging. We would also like to acknowledge the Faculty of Science Mass Spectrometry Centre at the University of Sheffield.