Portal de Periódicos da CAPES

Swf1-dependent palmitoylation of the SNARE Tlg1 prevents its ubiquitination and degradation

2005; Springer Nature; Volume: 24; Issue: 14 Linguagem: Inglês

10.1038/sj.emboj.7600724

1460-2075

Javier Valdez Taubas, Hugh R.B. Pelham,

Ubiquitin and proteasome pathways

Article23 June 2005free access Swf1-dependent palmitoylation of the SNARE Tlg1 prevents its ubiquitination and degradation Javier Valdez-Taubas Javier Valdez-Taubas MRC Laboratory of Molecular Biology, Cambridge, UK Search for more papers by this author Hugh Pelham Corresponding Author Hugh Pelham MRC Laboratory of Molecular Biology, Cambridge, UK Search for more papers by this author Javier Valdez-Taubas Javier Valdez-Taubas MRC Laboratory of Molecular Biology, Cambridge, UK Search for more papers by this author Hugh Pelham Corresponding Author Hugh Pelham MRC Laboratory of Molecular Biology, Cambridge, UK Search for more papers by this author Author Information Javier Valdez-Taubas1 and Hugh Pelham 1 1MRC Laboratory of Molecular Biology, Cambridge, UK *Corresponding author. MRC Laboratory of Molecular Biology, Hills Road, Cambridge CB2 2QH, UK. Tel.: +44 1223 402290; Fax: +44 1223 412142; E-mail: [email protected] The EMBO Journal (2005)24:2524-2532https://doi.org/10.1038/sj.emboj.7600724 PDFDownload PDF of article text and main figures. ToolsAdd to favoritesDownload CitationsTrack CitationsPermissions ShareFacebookTwitterLinked InMendeleyWechatReddit Figures & Info Protein palmitoylation is a post-translational modification that affects a great number of proteins. In most cases, the enzymes responsible for this modification have not been identified. Some proteins use palmitoylation to attach themselves to membranes; however, palmitoylation also occurs in transmembrane proteins, and the function of this palmitoylation is less clear. Here we identify Swf1, a member of the DHHC-CDR family of palmitoyltransferases, as the protein responsible for modifying the yeast SNAREs Snc1, Syn8 and Tlg1, at cysteine residues close to the cytoplasmic end of their single transmembrane domains (TMDs). In an swf1Δ mutant, Tlg1 is mis-sorted to the vacuole. This occurs because unpalmitoylated Tlg1 is recognised by the ubiquitin ligase Tul1, resulting in its targeting to the multivesicular body pathway. Our results suggest that one role of palmitoylation is to protect TMDs from the cellular quality control machinery, and that Swf1 may be the enzyme responsible for most, if not all, TMD-associated palmitoylation in yeast. Introduction Many proteins are post-translationally modified by the addition of a palmitate molecule to a cysteine by thioesterification. This modification is better referred to as S-acylation because the cysteines can be modified by lipids of different chain lengths (Berthiaume and Resh, 1995; DeMar and Anderson, 1997). Several kinds of proteins have been found to be palmitoylated (Linder and Deschenes, 2004). Some hydrophilic proteins like Ras and G proteins make use of this modification to attach themselves to membranes, often in combination with prenyl groups. However, many transmembrane proteins are also modified by palmitate. Among these are the transferrin receptor (Omary and Trowbridge, 1981; Alvarez et al, 1990), the mannose 6-phosphate receptor (Schweizer et al, 1996), the human asialoglycoprotein receptor (Yik et al, 2002) and the calcium sensor synaptotagmin (Veit et al, 1996). The function of this kind of palmitoylation is less clear, but it has been shown to affect trafficking and function of several proteins (Bijlmakers and Marsh, 2003). Modification of transmembrane proteins occurs in the cytoplasmic domain, often near or within a transmembrane domain (TMD); there are no clear sequence requirements and neither the cytoplasmic nor the TMD sequences appear to have an influence (Yik and Weigel, 2002). However, adding a cysteine to a nonpalmitoylated protein does not necessarily result in palmitoylation (Ponimaskin and Schmidt, 1998). Although palmitoylation of proteins has been known for many years, only a few palmitoyltransferases have been described and an enzyme responsible for modifying TMDs has not previously been identified. In the yeast Saccharomyces cerevisiae, Akr1 has been reported to modify Yck2 type I casein kinase, a hydrophilic protein (Roth et al, 2002), and Erf2, in combination with an additional subunit, Erf4, modifies Ras2 (Jung et al, 1995; Bartels et al, 1999; Lobo et al, 2002). In mammalian cells, the γ2 subunit of the GABAA receptor was found to be palmitoylated by GODZ (Keller et al, 2004). Very recently, Huntingtin interacting protein 14 (HIP14) was shown to have palmitoyltransferase activity and, upon overexpression, to transform cells to form aggressive tumours (Ducker et al, 2004). HIP14 and other related proteins modify the cytosolic neuronal protein PSD-95 (Fukata et al, 2004; Huang et al, 2004). All these proteins belong to a family of proteins that share a 50-residue-long DHCC cysteine-rich domain (CRD), which is a variant of the zinc-finger domain C2H2 (Putilina et al, 1999). Members of this family have been suggested to have a general role in palmitoylation (Roth et al, 2002). There are at least 23 distinct family members in the human genome and seven in yeast (Linder and Deschenes, 2004), which presumably differ in their substrate specificity, intracellular location or expression patterns. In addition, a protein lacking the DHCC-CRD, Ykt6, has recently been reported to possess palmitoyltransferase activity towards both Vac8 and itself (Dietrich et al, 2004; Fukasawa et al, 2004). Members of the SNARE family of membrane fusion proteins (Sollner et al, 1993; Rothman, 1994) are known to be substrates for palmitoylation. Some of these lack a TMD and are anchored either by multiple palmitate residues, as in the case of SNAP-25 (Veit et al, 1996), or, for Ykt6, by a combination of prenylation and palmitoylation (Fukasawa et al, 2004). However, others contain TMDs with adjacent cysteine residues that could, in principle, be palmitoylated. In one case, that of the yeast Snc proteins, palmitoylation has been demonstrated. This modification occurs soon after synthesis, while the protein is still in the ER, and is of unknown function (Couve et al, 1995). Interestingly, viral glycoproteins that mediate membrane fusion are also often palmitoylated, and although this is not always necessary for function, in at least one case, that of Rous sarcoma virus, loss of palmitoylation causes the protein to be rapidly degraded in lysosomes (Ochsenbauer-Jambor et al, 2001). Swf1 is a member of the S. cerevisiae DHCC-CRD family, and an swf1 null mutant shows synthetic lethal interactions with many mutants that influence membrane recycling through endosomes and the Golgi apparatus, including Tlg1, Snc2, Ric1, Ypt6 and many others (Tong et al, 2004). Here we show that Swf1 is required for the modification not only of Snc1, but also of the endosomal SNAREs Tlg1 and Syn8. The most notable effects of Swf1 are observed for Tlg1, a SNARE involved in the recycling of proteins from endosomes to Golgi (Holthuis et al, 1998a, 1998b; Lewis et al, 2000). Lack of Tlg1 palmitoylation results in its entry into multivesicular bodies (MVBs) and ultimately to its degradation in the vacuolar compartment. This occurs because nonacylated Tlg1 is recognised by the ubiquitin ligase Tul1 (Reggiori and Pelham, 2002) and becomes ubiquitinated, a signal for entry into MVBs (Katzmann et al, 2001; Reggiori and Pelham, 2001). Results SWF1 is required for the palmitoylation of Snc1 Snc1, the yeast homologue of synaptobrevin, is involved in vesicle docking and fusion with the plasma membrane. Intracellular traffic of this protein has been extensively studied and it is known to cycle between the plasma membrane, endosomes and the Golgi apparatus (Lewis et al, 2000). This protein and its very close homologue Snc2 are known to be post-translationally modified by the addition of a palmitate (Couve et al, 1995). In order to identify the enzyme(s) responsible, we examined the modification of Snc1 in wild-type (WT) cells and in mutants lacking each of the seven yeast members of the DHCC-CDR family of putative palmitoyltransferases. Under carefully controlled conditions (see Materials and methods section), palmitoylation of Snc1 can be detected by a reduction in electrophoretic mobility of the protein on polyacrylamide gels. Figure 1A shows that Myc-tagged Snc1 migrated slightly faster when expressed in the swf1Δ strain, compared to WT cells and all the other mutants. Identical results were obtained with myc-tagged Snc2 (data not shown). Figure 1.Palmitoylation of Snc1 requires Swf1. (A) Immunoblot of myc-tagged Snc1 expressed in the indicated strains. (B) Extracts of cells expressing myc-tagged Snc1 or the Cys–Ala mutant (Snc1ΔC) were treated with hydoxylamine (H) or, as a control, Tris (T) before gel electrophoresis and immunoblotting. Equal amounts of total protein were loaded in each lane in both (A) and (B), as confirmed by staining of the blot. Download figure Download PowerPoint To demonstrate that the shift was due to a change in the palmitoylation status of Snc1, we treated the samples with either 1 M hydroxylamine, pH 7.4, which selectively removes thioester-linked palmitates (Chamberlain and Burgoyne, 1998) or 1 M Tris–HCl, pH 7.4, as a control. After this treatment, the difference in mobility was abolished, and hydroxylamine-treated Snc1 expressed in a WT strain comigrated with Snc1 expressed in an swf1Δ mutant (Figure 1B). Snc1 has only one cysteine in its sequence, in position 95, and therefore that is the only site susceptible to S-acylation. A mutant version of Snc1 in which cysteine 95 was mutated to alanine (Snc1ΔC) shows the same shift towards lower molecular weight that is observed for the nonmutated Snc1 expressed in an swf1Δ mutant, confirming that the shift reflects modification of the cysteine (Figure 1B). It has been suggested that nonpalmitoylated Snc1 is less stable than the modified form (Couve et al, 1995). We did observe a slight reduction in the amount of Myc-tagged Snc1 in an swf1Δ mutant compared to WT cells. However, this does not seem to be due specifically to the lack of modification of Snc1, since we did not observe any reproducible reduction in the levels of Snc1ΔC in a WT strain. The effect may be an indirect one caused by the loss of Swf1-dependent modification of some other substrate(s). This is not inconsistent with the observations of Couve et al, since the unmodified Snc1 that they observed resulted from massive overexpression of the protein, which presumably saturates the transferase and would be expected to affect other substrates by competition. Swf1 modifies other SNARES Having established an assay capable of detecting palmitoylation, we searched for additional substrates of Swf1. We focused on other SNAREs with an endosome/Golgi location, several of which have cysteines near or within their TMDs (Figure 2). Figure 2.TMD sequences of yeast and human SNAREs. Cysteine residues are highlighted, and the hydrophobic TMD sequences underlined. The major locations of the proteins are indicated: PM, plasma membrane; endo, endosomes. Download figure Download PowerPoint Syn8 is an endosomal SNARE, homologous to mammalian syntaxin 8 (Lewis and Pelham, 2002). Western blot analysis of WT or swf1Δ cells probed with anti-Syn8 antibody showed increased mobility of the protein in the swf1Δ mutant. This difference was abolished by treatment of the protein extract from the WT strain with 1 M hydroxylamine (Figure 3A). Analysis of all seven members of the DHCC-CDR family showed that only Swf1 contributes to the modification of Syn8 (Figure 3B). Figure 3.Swf1-dependent modification of Syn8 and Tlg1. (A) Extracts were treated with hydroxylamine (H) or Tris (T) as in Figure 1B, and immunoblotted using anti-Syn8 antiserum. (B) Syn8 mobility was compared in the indicated deletion mutants by immunoblotting. (C) Tlg1 was analysed as in panel A, using anti-Tlg1 antiserum. (D) The indicated strains were transformed with an empty vector (−), or a plasmid expressing either C-terminally GFP-tagged WT Swf1 or the equivalent DHHC mutant (Swf1*). In the left panel, extracts were treated with the cysteine-specific biotinylation reagents, and biotinylated proteins recovered with streptavidin beads before immunoblotting. In the right panel, samples were treated sequentially with N-ethylmaleimide to block free cysteines, then with hydroxylamine to remove palmitate, then with the biotinylation reagent. The loading control represents an example of a cysteine-containing protein detected by probing the same blot with fluorescent streptavidin, to demonstrate the reproducible recovery of biotinylated protein. Controls in which biotinylation was omitted resulted in no detectable Tlg1 being recovered from the streptavidin beads. Download figure Download PowerPoint Next, we analysed Tlg1 (Holthuis et al, 1998a), a SNARE whose function overlaps that of Syn8. Again, there was a small but reproducible increase in the mobility of the protein from swf1Δ cells compared to WT cells, which was abolished by hydroxylamine treatment (Figure 3C). As the mobility difference was small, we sought other ways to test the modification of Tlg1. Attempts to label the protein in vivo with [3H]palmitate were unsuccessful, possibly because the palmitate does not turn over and the rate of synthesis of Tlg1 is relatively low. As an alternative, we used a maleimide derivative (Biotin-BMCC) that biotinylates free cysteine residues, allowing unpalmitoylated Tlg1 to be purified on streptavidin beads and detected by immunoblotting. This is possible because Tlg1 contains only two cysteines, which are the potential sites of palmitoylation. Figure 3D shows that, in swf1 mutant cells, but not in WT cells, Tlg1 reacted well with this reagent, indicating free cysteines. These became inaccessible when the swf1 cells were transformed with a plasmid expressing Swf1, but not with a version of Swf1 in which the cysteine residue of the DHHC sequence, known to be crucial for palmitoyltransferase activity (Lobo et al, 2002; Roth et al, 2002), was mutated to alanine. The faint bands detected at 38 and 45 kDa are discussed below. In a complementary experiment, we treated cell extracts first with N-ethylmaleimide to block free cysteines, then used hydroxylamine treatment to remove palmitate, and finally the biotinylation reagent to detect the exposed cysteines. This is an established assay for palmitoylation (Drisdel and Green, 2004; Politis et al, 2005), and, as expected, gave a strong signal only when WT Swf1 was present (Figure 3D). We conclude that both Syn8 and Tlg1 have the hallmarks of palmitoylation, and depend on Swf1 for their modification. GFP-Tlg1 is mislocalised and degraded in swf1Δ cells Despite their lack of palmitoylation, we found no evidence for substantial mis-sorting of either Snc1 or Syn8 in swf1Δ cells, as judged by examination of GFP-tagged proteins (not shown). However, the effects on Tlg1 were more dramatic. This SNARE is normally localised to the TGN/endosomal compartment, and, when expressed as a GFP-fusion protein, it showed the dotty distribution characteristic of this compartment in yeast. A fraction of the GFP-tagged protein reached the vacuole, where it could be seen on the outer vacuolar membrane (Figure 4A). In contrast, in swf1Δ cells, the fluorescence was mostly inside the vacuole (Figure 4A). Since Tlg1 was tagged with GFP at its amino-terminus, which is normally cytoplasmic, the appearance of GFP inside the vacuole can only occur if the protein enters the internal vesicles of MVBs, which are formed by invagination of the endosomal membrane. Following fusion of the MVB with the vacuole, these vesicles are delivered into the vacuole for degradation. Since GFP itself is resistant to proteolysis, whereas the attached Tlg1 protein is not, free GFP accumulates in the vacuole. In confirmation of this, Western blotting revealed an increase in the amount of free GFP in the swf1Δ background (Figure 4B). Strains lacking other members of the DHHC-CRD family showed much less vacuolar proteolysis of GFP-Tlg1 (Figure 4C). Figure 4.Vacuolar targeting of Tlg1 in swf1Δ cells. (A) GFP-Tlg1 visualised in WT and swf1Δ cells, and GFP-Tlg1M2, in which the two cysteine residues are replaced with serines, together with the single cysteine mutants M3 (CC-CS) and M4 (CC-SC), in WT cells. With the mutant proteins, or in swf1 mutant cells, GFP is visible inside vacuoles. (B) Immunoblotting with anti-GFP shows the generation of free (vacuolar) GFP from GFP-Tlg1 in swf1Δ cells, and from GFP-Tlg1M2 in WT cells. The lower panel shows the same samples after immunoprecipitation with anti-Tlg1 antibodies. Note that full-length GFP-Tlg1M2 is not abundant enough to be clearly visible unless concentrated by immunoprecipitation. Numbers indicate apparent sizes in kilodaltons. (C) Immunoblotting shows that degradation of GFP-Tlg1 occurs preferentially in swf1Δ cells. (D) Immunoblotting with anti-Tlg1 shows reduced levels of endogenous protein in swf1Δ cells, and the appearance of bands with reduced mobility corresponding approximately in size to the addition of one and two ubiquitins. Equivalent loading was demonstrated by probing the same blot with antibodies to Pgk1. Tlg1 was also immunoprecipitated from the cells, and the precipitates probed with anti-Tlg1 (centre panel) and anti-ubiquitin (right panel). (E) WT and swf1Δ cells were treated with cycloheximide and samples taken at intervals for the assay of Tlg1 by immunoblotting. Note the logarithmic scale of the vertical axis. Download figure Download PowerPoint To test whether the proteolysis of Tlg1 was caused by its lack of palmitoylation, or by some other effect of the swf1Δ mutation, we mutated the two adjacent cysteine residues located at the cytosolic end of the Tlg1 TMD. Mutation of both these residues to serine (construct Tlg1M2) resulted in delivery of Tlg1 to the vacuole interior in WT cells (Figure 4A and B). A single mutation of either cysteine was sufficient to trigger entry into the vacuole (Figure 4A, Tlg1M3 and Tlg1M4). Palmitoylation of Tlg1 prevents recognition by Tul1 Degradation in the vacuole is normally associated with ubiquitination, which signals entry into MVBs. Indeed, when GFP-Tlg1 was concentrated by immunoprecipitation, we could detect faint bands in swf1Δ cells that were larger than the fusion protein; similar bands were detectable with the Tlg1M2 mutant in WT cells, despite the low levels of intact protein in this case (Figure 4B, lower panel). Immunoblotting of endogenous Tlg1 also revealed not only a reduction in the amount of protein in swf1Δ cells but also proteins migrating at 38 and 45 kDa (Figure 4D), as previously detected in Figure 3D. Immunoprecipitation of Tlg1 followed by immunoblotting with anti-ubiquitin antibody showed that the 45 kDa band, at least, is a ubiquitinated form of Tlg1 (Figure 4D). The 38 kDa band did not react with our monoclonal anti-ubiquitin, though it is not clear whether this is due to a genuine lack of ubiquitin or due to a particular property of the antibody. Some ubiquitination of endogenous Tlg1 could also be detected in WT cells. This may represent molecules that fail to be modified with palmitate, because we were unable to detect ubiquitinated bands when palmitoylated Tlg1 was selectively purified using the hydroxylamine-dependent biotinylation procedure (right-hand panel of Figure 3D; data not shown). Since ubiquitination results in degradation, the clear prediction is that Tlg1 should turn over more rapidly in swf1Δ cells than in WT cells. To test this, we added cycloheximide to block protein synthesis, and followed the fate of Tlg1 by immunoblotting. Figure 4E shows that, under these conditions, the SNARE was very stable in WT cells, but in the swf1Δ strain it was degraded with a half-life of about 6 h. Our results indicate that Tlg1 is ubiquitinated in swf1Δ cells. We therefore sought a ubiquitin ligase capable of recognising unpalmitoylated Tlg1. Tul1 is a ring domain ubiquitin ligase that recognises polar TMDs and targets them for degradation (Reggiori and Pelham, 2002). Another protein, Bsd2, also recognises TMDs and mediates protein ubiquitination by recruiting the HECT domain ubiquitin ligase Rsp5 (Hettema et al, 2004). To see whether these proteins contribute to the degradation of unpalmitoylated Tlg1, the Tlg1 cysteine to serine mutant Tlg1M2 was transformed into bsd2Δ, tul1Δ and bsd2Δ tul1Δ strains and the transformants analysed by fluorescence microscopy and Western blotting. Figure 5A shows that Tlg1M2 still enters the vacuolar lumen in the absence of either Tul1 or Bsd2, but not in the double mutant. In agreement with this, Western blot analysis showed that, although degradation of GFP-Tlg1 was reduced in the single mutants, only in the double mutant was it blocked completely (Figure 5B). Thus, Tlg1M2 is a substrate for both Tul1 and the Bsd2 system. Figure 5.Unpalmitoylated Tlg1 is recognised by Tul1 and Bsd2. (A) GFP-tagged Tlg1 mutants with the cysteines replaced by serines (Tlg1M2) or leucines (Tlg1M7) were visualised in the indicated strains. (B) Immunoblotting with anti-GFP antibodies of the strains shown in (A) expressing GFP-Tlg1M2. (C) GFP-Tlg1M7 analysed as in (B) in the indicated strains. (D) GFP-Tlg1 and GFP-Tlg1M2 analysed in an swf1 tul1 double mutant. Sequences of the C termini of the various forms of Tlg1 are shown below. Download figure Download PowerPoint Since Tul1 and Bsd2 recognise polar TMDs, it could be that their effect on Tlg1M2 is a consequence of the presence of the serines, rather than the absence of palmitoylation. Indeed, the M2 mutant, with serines, was less stable than WT Tlg1 in an swf1 mutant, a situation in which the protein should have free cysteines. We therefore generated a new construct (Tlg1M7) in which the cysteines were replaced by hydrophobic leucines. This mutant also entered the vacuole in WT cells, though to a lesser extent than the serine-containing Tlg1M2. Interestingly, the leucine mutant remained a substrate for Tul1, but was completely unaffected by Bsd2 (Figure 5A and C). It was also unaffected by Swf1, confirming that Swf1 stabilises Tlg1 by acting directly on the cysteine residues, rather than indirectly through some other change to the protein-sorting machinery. Finally, we constructed an swf1Δ tul1Δ double mutant to test whether unpalmitoylated WT protein, with free cysteines, is also recognised by Tul1. In this strain, GFP-Tlg1 was completely stable, as shown by the absence of free GFP. However, the serine-containing M2 mutant was still degraded, due to the presence of Bsd2 (Figure 5). We conclude that it is the absence of palmitoylation that renders Tlg1 sensitive to Tul1. In contrast, recognition by Bsd2 occurs only when the polarity of the TMD is increased by the introduction of serine residues. Although removal of Tul1 prevented degradation of Tlg1 in swf1Δ cells, it did not suppress swf1 phenotypes that we have observed, such as the inability of the cells to grow on lactate or the reduced level of Snc1 (data not shown). Thus, these other phenotypes must reflect additional functions of Swf1, or of unpalmitoylated SNAREs, and are not simply a consequence of low Tlg1 levels. Swf1 is localised to the ER Palmitoylation of Snc1 occurs rapidly after synthesis (Couve et al, 1995). To locate Swf1, we tagged it at the N and C termini with a triple HA epitope tag or GFP, and confirmed that the tagged proteins retained the ability to prevent degradation of GFP-Tlg1 (Figure 6A). Direct observation of both N- and C-terminal GFP tagged versions under the fluorescence microscope showed a weak signal, localised to the ER and most prominently the nuclear envelope (Figure 6B; data not shown). Immunofluorescence of HA-tagged Swf1 yielded similar results, although in this case some additional intracellular dots were observed (Figure 6B). Thus, Swf1 is in a position to modify newly synthesised SNAREs before they exit the ER. This is consistent with the observation that blocking ER exit with a temperature-sensitive sec18 mutation does not prevent palmitoylation of Snc1 (Couve et al, 1995). It is likely that the modification persists for the lifetime of the protein, but we cannot rule out the possibility that some Swf1 is present in the Golgi or endosomes, and can act there to repalmitoylate proteins that lose their modification. Figure 6.Localisation of Swf1. (A) Tagged Swf1 is functional. Immunoblotting of GFP-Tlg1 in swf1Δ cells transformed also with an empty plasmid (PRS316), or plasmids expressing Swf1 tagged with HA at the N-terminus (HA-Swf1) or C-terminus (Swf1-HA), or GFP at the N-terminus (GFP-Swf1). (B) Fluorescence and immunofluorescence of the N-terminally GFP-tagged Swf1 in swf1Δ cells, and HA-tagged Swf1 in WT cells. Two different fields are shown for Swf1-HA. Identical results were obtained with GFP and HA tags at either the N- or C-terminus, in either swf1Δ or WT cells. Download figure Download PowerPoint Discussion In this paper, we have presented evidence that at least three different yeast SNAREs, Snc1, Tlg1 and Syn8, are palmitoylated on cysteine residues close to the cytoplasmic end of the TMD. Since several other SNAREs, both in yeast and in vertebrates, contain cysteine residues in a similar position (Figure 2), it is likely that they are all similarly modified. We have also shown that modification of all three of the SNAREs that we have studied requires the ER-localised DHHC-CRD protein Swf1. By analogy with other DHHC-CRD proteins, Swf1 is likely to be the palmitoyltransferase that acts directly on newly synthesised SNAREs, and presumably also on other proteins with appropriately positioned cysteine residues. In support of this, we have shown that the cysteine residue in the DHHC motif of Swf1 is essential for modification of Tlg1 in vivo. The equivalent residues in Erf2 and Akr1 have been shown to be required for palmitoyltransferase activity both in vivo and in vitro (Lobo et al, 2002; Roth et al, 2002). The specificity of Swf1 function is striking. Six other DHHC-CRD palmitoyltransferases are present in yeast. Of these, Erf2 is localised to the ER (Bartels et al, 1999), and we have found that Yol003c is also in the ER, yet neither can substitute for Swf1 even when overexpressed (our unpublished observations). Since the actual transfer of palmitate from palmitoyl-CoA to cysteine residues can occur spontaneously, it has been suggested that the role of the transferases is simply to present a palmitoyl-thioester to the substrates, either as palmitoyl-CoA or as an enzyme-linked intermediate (Roth et al, 2002; Dietrich and Ungermann, 2004), in much the same way that E3 ubiquitin ligases present ubiquitin thioesters to their substrates. The conserved DHHC-CRD is likely to be involved in the palmitoylation reaction itself (Roth et al, 2002), while other features of the individual transferases recognise specific substrates. For example, prenylated Ras proteins may be recruited by Erf4, a cytoplasmic protein that associates tightly and specifically with Erf2, while the soluble casein kinase Yck2 may be recruited via the ankyrin repeats that are a feature of its specific transferase, Akr1. In the case of Swf1, an unusual feature is the presence of five predicted TMDs, the first close to the amino-terminus; most DHHC-CRD proteins have four, or in some cases two, three or six. It may be that the 5-TMD structure includes a pocket for interaction with the single TMDs of SNARE proteins, or potentially with any free TMD. This feature may be conserved: humans and mice have only one DHHC protein (DHHC-4) that fits the Swf1 topology, and it is more closely related to Swf1 than to any other yeast protein. However, not all species have an Swf1-like protein with five TMDs. Thus, there may be some structural diversity in the palmitoyltransferases that recognise TMDs. Once palmitoylated, the SNARES leave the ER and thus are largely separated from the transferase. The modification is therefore likely to be permanent, rather than a transient change associated with each membrane fusion event. Its contribution, if any, to the membrane fusion activity of the SNAREs is evidently subtle, since unpalmitoylated mutants of both Snc1 and Tlg1 can support growth under conditions in which the SNAREs themselves are essential (our unpublished observations; Lewis et al, 2000). Furthermore, even the multiple palmitoylation defects of swf1Δ cells cause only a mild defect in the sorting of vacuolar proteases, a sensitive assay for membrane traffic between Golgi and endosomes (Bonangelino et al, 2002). Nevertheless, the synthetic lethality of an swf1 mutation with others that affect such traffic does indicate an important role for palmitoylation of one or more substrates. Since TMDs are already firmly embedded in the membrane, it seems unlikely that palmitoylation is involved in anchoring them. However, it might help to increase the solubility of the TMD in a sterol-rich bilayer. TMD-associated cysteines are present almost exclusively in SNAREs associated with the plasma membrane, endosomes and TGN, membranes rich in sterols (Figure 2). Increasing bilayer solubility may help SNAREs to partition into, and be sorted with, sterol-rich regions of membrane. It may also help the TMDs to dissociate from each other when the SNARE complex is disrupted following a fusion event. Indeed, we have observed that, upon lysis of swf1Δ mutant cells, Tlg1 readily forms disulphide linkages with other SNAREs to which it is bound, implying that the cysteines, and hence normally the palmitoyl groups, are located close to the interfaces between TMDs in the SNARE complex. In the case of Tlg1, one clear effect of palmitoylation is to prevent interaction with the ubiquitin ligase Tul1, w