Analysis of DHHC Acyltransferases Implies Overlapping Substrate Specificity and a Two-Step Reaction Mechanism

Authors

  • Haitong Hou,

    1. University of Osnabrück, Department of Biology, Biochemistry section, Barbarastrasse 13, 49076 Osnabrück, Germany
    Search for more papers by this author
  • Arun T. John Peter,

    1. University of Osnabrück, Department of Biology, Biochemistry section, Barbarastrasse 13, 49076 Osnabrück, Germany
    Search for more papers by this author
  • Christoph Meiringer,

    1. University of Osnabrück, Department of Biology, Biochemistry section, Barbarastrasse 13, 49076 Osnabrück, Germany
    Search for more papers by this author
    • 2

      Present address: Roche Diagnostics GmbH, 82377 Penzberg, Germany

  • Kanagaraj Subramanian,

    1. University of Osnabrück, Department of Biology, Biochemistry section, Barbarastrasse 13, 49076 Osnabrück, Germany
    Search for more papers by this author
    • 3

      Present address: The Scripps Institute, 10550 North Torrey Pines Road, La Jolla, CA 92037, USA

  • Christian Ungermann

    Corresponding author
    1. University of Osnabrück, Department of Biology, Biochemistry section, Barbarastrasse 13, 49076 Osnabrück, Germany
      Christian Ungermann, Christian.Ungermann@biologie.uni-osnabrueck.de
    Search for more papers by this author

Christian Ungermann, Christian.Ungermann@biologie.uni-osnabrueck.de

Abstract

Asp-His-His-Cys (DHHC) cysteine-rich domain (CRD) acyltransferases are polytopic transmembrane proteins that are found along the endomembrane system of eukaryotic cells and mediate palmitoylation of peripheral and integral membrane proteins. Here, we address the in vivo substrate specificity of five of the seven DHHC acyltransferases for peripheral membrane proteins by an overexpression approach. For all analysed DHHC proteins we detect strongly overlapping substrate specificity. In addition, we now show acyltransferase activity for Pfa5. More importantly, the DHHC protein Pfa3 is able to trap several substrates at the vacuole. For Pfa3 and its substrate Vac8, we can distinguish two consecutive steps in the acylation reaction: an initial binding that occurs independently of its central cysteine in the DHHC box, but requires myristoylation of its substrate Vac8, and a DHHC-motif dependent acylation. Our data also suggest that proteins can be palmitoylated on several organelles. Thus, the intracellular distribution of DHHC proteins provides an acyltransferase network, which may promote dynamic membrane association of substrate proteins.

Function of proteins in the endomembrane system is largely determined by their subcellular localization. Membrane association of peripheral membrane proteins may be facilitated by positively charged amino acid stretches or the covalent addition of lipids (like prenylation or myristoylation), although the binding affinity to membranes mediated by these interactions is weak (1,2). An alternative means of membrane association is protein palmitoylation (3). The thioester bond between the palmitate and the cysteine of the target protein is generated with the help of acyltransferases (4,5). A family of polytopic transmembrane proteins, named DHHC cysteine-rich domain (CRD) proteins, according to a consensus sequence in their cytosolic domain, has been identified to function as acyltransferases in yeast and other eukaryotes (3,6). Whereas 23 DHHC isoforms have been found in humans, only 7 are present in yeast. They are located on distinct membranes of the endomembrane system; three proteins (Erf2, Swf1 and Pfa4) localize to the endoplasmic reticulum (ER), two on the Golgi (Akr1 and Akr2), one on the vacuole (Pfa3) and Pfa5 is present on the plasma membrane (7). On the basis of deletion analyses, substrate-specific activities have been demonstrated for Erf2 towards Ras2 (8), for Swf1 towards the SNARE Tlg1 (9), for Akr1 towards Yck1/2 and Lcb4 (10,11), for Pfa4 towards the chitin synthase Chs3 (12) and for Pfa3 towards the vacuolar fusion factor Vac8 (13,14). So far, neither activity nor substrates have been described for Pfa5.

Using knock-out strains of up to six DHHC proteins, Davis and co-workers conducted a proteomic survey to identify palmitoylated proteins and their connection to DHHC proteins (15). A large portion of palmitoylated peripheral membrane proteins can be subdivided into three groups according to the positions of palmitoylated cysteines; those that have cysteines proximal to the N-terminal myristolyation motif (e.g. Vac8; I), adjacent to the C-terminal prenylation sequence (e.g. Ras2; II) and near the C-terminus of the protein (e.g. Yck3, Yck2; III). Most of these proteins are either found at the plasma membrane or the vacuole. Group I proteins and also some members of group II are mostly dependent on Erf2. Some depend on Akr1, whereas others could not be assigned to any specific DHHC. This observation suggests that multiple DHHCs could have overlapping specificities, although this has not been shown so far. In fact, it has not been possible to identify a clear consensus sequence for palmitoylation (15).

Here, we present evidence that five of the DHHC proteins when overexpressed mediate palmitoylation and localization of several peripheral membrane protein substrates such as the vacuolar fusion factor Vac8, the casein kinases Yck1-3 or Ras2. We also find that Pfa5 has acyltransferase activity. For Pfa3, we provide evidence for a two-step reaction mechanism, which includes substrate binding. Our data suggest that the endomembrane distribution of multiple DHHCs with a broad substrate specificity provides a framework for the localization and modification of palmitoylated proteins.

Results

Expression and localization of overexpressed DHHCs

Upon deletion of Pfa3, the vacuole localization of Vac8, a palmitoylated peripheral membrane protein, is reduced, suggesting that it acts as the main acyltransferase for Vac8 (13,14). Interestingly, the mislocalization of Vac8 to the cytosol is enhanced upon a combined deletion of AKR1, AKR2, PFA3, PFA4 and PFA5 (akr1Δ akr2Δ pfa3Δ pfa4Δ pfa5Δ; termed 5 ×Δ in the remaining text; Figure 1A,B), and Vac8 is poorly acylated under these conditions (15). This suggests that several DHHC proteins contribute to Vac8 localization, although this was actually not shown previously (15). We therefore decided to test this hypothesis directly and placed selected DHHC proteins, which localize to ER, Golgi, plasma membrane and endosome/lysosome, under the control of the galactose promotor and tagged them C-terminally with Protein A or green fluorescent protein (GFP) (Figure 2). Akr2 was not tested because of its lethality upon overproduction (16).

Figure 1.

Vac8 localization in cells lacking DHHC proteins. A) Cells with the indicated DHHC deletions expressing genomically integrated Vac8-GFP were analysed by fluorescence microscopy. 4 ×Δ = akr 2Δ pfa3Δ pfa4Δ pfa5Δ; 5 ×Δ = akr 1Δ akr2Δ pfa3Δ pfa4Δ pfa5Δ. Size bar: 10 μm. B) Subcellular fractionation. Cells containing untagged Vac8 were lysed and separated as described in Methods. The total cell lysate (T) corresponds to the same amount used for the separation of membrane pellet (P) and supernatant (S). Proteins were analysed by SDS-PAGE and western blot using antibodies against the vacuolar V-ATPase subunit Vma6, Vac8 and Arc1, a cytosolic marker protein.

Figure 2.

Overexpression and localization of DHHC proteins. A) DHHC overexpression level. Total cell lysates were prepared from BJ VAC8-TAP and the 5 ×Δ strains expressing the Protein A-tagged DHHC proteins. Proteins were analysed by SDS-PAGE and western blotting using antibodies against the Protein A tag or Vti1. B) Determination of the Vac8 and Pfa3 expression level in wild-type cells. Indicated amounts of lysate of BJ3505 cells expressing chromosomally TAP-tagged Vac8 and Pfa3 were analysed by SDS-PAGE and western blot with antibodies against the Protein A tag and Vti1. C) Intracellular localization of GFP-tagged DHHC proteins. Cells were grown to mid-log phase and analysed by fluorescence microscopy as described in Methods. Size bar: 10 μm.

Upon induction in galactose, overexpression of all DHHCs was detected (Figure 2A). Pfa3 was then expressed at a comparable level as its substrate Vac8. Our quantification indicates that wild-type cells contain 10-fold more Vac8 than Pfa3 (Figure 2B), indicating a 10-fold overproduction of Pfa3 in our overexpression strain. Akr1 overexpression levels were similar to Pfa3, whereas Erf2, Pfa4 and Pfa5 expression was significantly lower (Figure 2A). We analysed the subcellular localization of the overexpressed DHHCs by fluorescence microscopy. Despite their correct localization to the ER (Pfa4, Erf2), plasma membrane (Pfa5) and vacuole (Pfa3), a significant portion of GFP-tagged DHHCs was localized to the vacuole lumen, which might be a result of the overproduction (Figure 2C). Akr1 was found in multiple dots that could correspond to the Golgi.

Several DHHC proteins can palmitoylate Vac8

We then tested whether overproduction of these DHHCs would be sufficient to localize selected substrates to their target organelle. Initially, we focused on the vacuolar Vac8 fusion protein. Vac8 is myristoylated at the N-terminal glycine and contains three cysteines at position 4, 5 and 7 that can be palmitoylated and thus support membrane localization (17–19). Previous studies suggested that Pfa3 (13,14), and Akr1 are involved in Vac8 palmitoylation (15), even though a deletion of Akr1 alone shows no effect on Vac8 membrane localization (Figure 1A). To our surprise, any of our overexpressed DHHC was sufficient to confer vacuole membrane localization (Figure 3A,D) and palmitoylation on Vac8 (Figure 3E). In addition, this indicates that DHHC proteins are functional despite their overexpression and C-terminal tagging. Interestingly, even the overexpression of Erf2, which is already present at endogenous levels in the 5xΔ strain, was sufficient to rescue Vac8 localization and palmitoylation. Our data also show that Pfa5 has acyltransferase activity (Figures 3–5). Thus, Vac8 vacuole localization and palmitoylation appears to occur independently of the localization of the DHHC protein and its identity.

Figure 3.

Localization of Vac8 upon overexpression of DHHC proteins. (A–C) Localization of Vac8-GFP (A), Vac8 (C4,5,7A)-GFP (B) and Vac8 (G2A)-GFP (C) in the respective DHHC overexpression strains. Scale bar 10 μm. D) Subcellular fractionation was conducted as in Figure 1B. DHHC proteins were detected via the C-terminal Protein A tag. Pfa5 could not be detected in this experiment because of its low expression (see Figure 2A). E) Palmitoylation of wild-type Vac8 was determined by biotin switch as described in Methods. In brief, protein samples were pretreated with NEM to quench free cysteines, then reisolated and incubated with hydroxylamine to cleave the palmitate thioester. Free sulfhydryl groups were then cross-linked to BMCC biotin, and biotinylated proteins were enriched on streptavidin beads. Load (4%) corresponds to the sample prior to the pull down.

Figure 4.

Targeting of SH4-GFP by DHHCs. Cells expressing genomically integrated SH4-GFP were analysed by fluorescence microscopy. Localization of SH4-GFP is shown in the wild-type (A) and 5 ×Δ (B) background with different DHHCs overproduced. C) SH4-GFP localization in the indicated DHHC deletion strains. Size bar: 10 μm.

Figure 5.

Figure 5.

Overlapping substrate specificity of DHHC proteins. A) Overview of DHHCs and corresponding substrates. Substrates and their acyltransferases as described in the literature are listed. The palmitoylation sites of each substrate are shown in bold. B) and C) Localization of GFP-Yck3 in the indicated strains was analysed by fluorescence microscopy as before. Size bar = 10μm. D) Rescue of cell morphology by DHHC overexpression in akr1Δ cells. The indicated cells were grown in glucose (YPD) and galactose (YPG) and analysed by DIC optics. E) Localization of GFP-tagged Yck2 by fluorescence microscopy in cells overexpressing the indicated DHHC proteins in the akr1Δ background. F) Localization of GFP-Ras2 in the presence or absence of the indicated overexpressed DHHC proteins in the erf2Δ background.

Figure 5.

Figure 5.

Overlapping substrate specificity of DHHC proteins. A) Overview of DHHCs and corresponding substrates. Substrates and their acyltransferases as described in the literature are listed. The palmitoylation sites of each substrate are shown in bold. B) and C) Localization of GFP-Yck3 in the indicated strains was analysed by fluorescence microscopy as before. Size bar = 10μm. D) Rescue of cell morphology by DHHC overexpression in akr1Δ cells. The indicated cells were grown in glucose (YPD) and galactose (YPG) and analysed by DIC optics. E) Localization of GFP-tagged Yck2 by fluorescence microscopy in cells overexpressing the indicated DHHC proteins in the akr1Δ background. F) Localization of GFP-Ras2 in the presence or absence of the indicated overexpressed DHHC proteins in the erf2Δ background.

We then tested for the specificity of this reaction. Localization of Vac8 required its N-terminal cysteines. A mutant lacking the cysteines (Vac8 C4,5,7A) was primarily cytosolic (Figure 3B). The Vac8-GFP membrane staining observed in some overexpression backgrounds corresponds most likely to the ER, and Vac8 is known to bind the ER protein Nvj1 (20). Indeed, a similar localization of Vac8 is already observed in the 5 ×Δ mutant (Figure 3A). Moreover, loss of Vac8 myristoylation (Vac8 G2A) led to a complete mislocalization of Vac8 to the cytosol (Figure 3C). This indicates that myristoylation alone might allow for some association of Vac8 with membranes as previously observed (17). Interestingly, only Pfa3 overproduction rescued the localization of either Vac8 mutant, which might be because of its direct interaction and subsequent efficient palmitoylation (see below).

Palmitoylation of multiple substrates by DHHC proteins

We then asked how DHHC overproduction would affect substrates that are not exclusively localized to the vacuole. In previous studies, we used a fusion protein consisting of the N-terminal 18 amino acid sequence, termed Src homology 4 (SH4) domain, of Vac8 and GFP (13,18). This SH4-GFP fusion protein is found on the vacuole and on the plasma membrane in wild-type cells (Figure 4A), but looses vacuole localization in the absence of Pfa3 (Figure 4C) as previously shown (13). A deletion in akr1 did not affect the localization of SH4-GFP, but diminished plasma membrane localization (Figure 4C). Moreover, the fusion protein was poorly localized to membranes if several DHHCs were lacking (Figure 4B, right panel), indicating that it could be a substrate of multiple DHHCs. When we overproduced Erf2, Pfa4 and Pfa5 in wild-type or 5 ×Δ cells, SH4-GFP accumulated at the plasma membrane (Figure 4), and was palmitoylated. In contrast, overexpression of the vacuolar Pfa3 efficiently trapped SH4-GFP at the vacuole (Figure 4A,B, bottom panel). On the basis of these results, we speculate that palmitoylated SH4-GFP traffics by default along the secretory pathway to the plasma membrane. At the vacuole, Pfa3 can palmitoylate and trap SH4-GFP due to the terminal character of this organelle in the endomembrane system.

To further investigate the redundancy of DHHC proteins, we extended this analysis to other palmitoylated peripheral membrane proteins, which—in part—had been linked to certain DHHC proteins (Figure 5A). The first substrate, Yck3, contains several C-terminal cysteines, which have been shown to be modified by Pfa3 (13) and Akr1 (15,21). In fact, Yck3 is poorly localized (Figure 5B) and palmitoylated in the 5 ×Δ mutant (15). When we tested Yck3 in our DHHC overexpression strains, we observed clear vacuole localization in all cases (Figure 5B), which was not observed if the C-terminal cysteines were lacking (Figure S1A). This indicates that any DHHC can efficiently palmitoylate Yck3.

We then asked, how this DHHC-mediated palmitoylation connects to the transport of Yck3, which occurs via the AP-3 pathway that connects Golgi and vacuole directly (21). In the absence of the AP-3 subunit Apl5, Yck3 is found at the plasma membrane and—because of endocytosis—in the lumen of the vacuole (Figure 5C). Overexpression of selected DHHCs like the ER-localized Pfa4 or the plasma membrane DHHC Pfa5 did not restore vacuolar rim staining (Figure 5C). In contrast, overproduction of Pfa3 redirected Yck3 to the vacuole rim (Figure 5C, bottom). As a control, we show that Vac8 trafficking to the vacuole is independent of an apl5 deletion (Figure S1B). Thus, overproduced Pfa3 can bypass the sorting information of Yck3, presumably by recruiting it directly from the cytosol to the vacuole surface.

As a second example, we analysed the casein kinase Yck2. Yck2 and its homolog Yck1 are dually palmitoylated at their C-termini in an Akr1-dependent manner (10). These kinases are responsible for multiple phosphorylation events at the plasma membrane and are mislocalized if Akr1 is absent (22). As a consequence, akr1Δ cells have an abnormal shape (15). We considered wild-type cell morphology as an indication for functional rescue, when we screened our overexpressed DHHC proteins (Figure 5D). Whereas wild-type cells exhibited a comparable shape in glucose and galactose, akr1Δ cells were elongated in either medium. Overexpression of all DHHCs, but the vacuole-localized Pfa3 rescued the morphology defect of akr1Δ cells (Figure 5D). This suggested that Yck1/2 palmitoylation can only occur if the DHHC proteins are localized along the secretory pathway. Alternatively, Pfa3 might be able to recognize Yck1/2, but would trap them at the wrong location. To distinguish between these possibilities, we tagged Yck2 with an N-terminal GFP tag to determine its localization. As previously described (10), Yck2 was mislocalized to the cytosol in akr1Δ cells (Figure 5E). Akr1, Pfa4 or Pfa5 overproduction directed Yck2 to the plasma membrane (Figure 5E), consistent with the rescue in morphology observed before (Figure 5D). In the Pfa4 overexpression strain, some Yck2 was concentrated in one to two punctate structures, which interfered with visualization of plasma membrane localization. Only for Erf2 overexpression we observed a mixed result: cell morphology was only rescued if Yck2 was not GFP tagged (Figure 5D,E). This could be because of the poor recognition of Yck2 in the presence of the GFP tag or the slightly higher expression of Yck2. More importantly, overexpression of Pfa3 led to a striking vacuolar concentration of Yck2, indicating that Pfa3 is able to recognize Yck2, but traps it at the wrong location, where it cannot conduct its function (Figure 5D,E).

We selected the plasma membrane-localized Ras2-GTPase as a third example to analyse the response to overexpressed DHHCs. Ras2 has a dual palmitoyl and prenyl anchor at its C-terminus. The ER-localized DHHC Erf2 has been identified as an acyltransferase for Ras2. In the absence of Erf2, Ras2 is poorly palmitoylated, but still found primarily at the plasma membrane (8,23,24). In addition, GFP-Ras2 staining in the cytosol is slightly increased, and a weak vacuolar ring is visible (Figure 5F). Overexpression of Erf2, Akr1, Pfa4 and Pfa5 in the erf2 deletion background rescued the plasma membrane localization of Ras2. Only Pfa3 overproduction redirected a significant portion of Ras2 to the vacuole membrane in the erf2Δ background (Figure 5F). Interestingly, in the 5 ×Δ cells (which still have Erf2) such a Pfa3-mediated redirection does not occur (Figure S2). Potentially, Erf2 palmitoylation of Ras2 in wild-type cells occurs preferentially early in the secretory pathway, and the potential low turn over of Ras2 may not expose it to Pfa3, even if it is overexpressed.

We conclude that DHHC proteins have a broad specificity, exemplified by the ability of Pfa3 to trap proteins at the vacuole.

Recognition of Vac8 by Pfa3

During our analysis of the substrate specificity of the DHHC proteins, we noticed that the dedicated enzyme-substrate pair Pfa3 and Vac8 behaved differently. Pfa3 was able to trap even non-myristoylated or non-palmitoylated Vac8 at vacuoles (Figure 3B,C). We therefore decided to analyse this interaction in more detail. First, we generated a mutant form of Pfa3, in which the central cysteine (C134) of the DHHC motif was mutated to its non-functional DHHA form (14,25). This mutation in Pfa3 did not affect its trafficking to the vacuole (Figure 2C,6A). Unexpectedly, the Pfa3 C134A rescued vacuolar localization of Vac8-GFP to a similar extent as wild-type Pfa3 (Figure 6B). We performed subcellular fractionation to confirm the localization of wild-type Vac8 (Figure 6C), and used the established biotin switch protocol (Figure 6D) to assess the palmitoylation status of Vac8. Unlike wild-type Pfa3 (Figure 6C, lanes 2 and 3), the C134A mutant is not able to efficiently retain Vac8 on vacuoles during subcellular fractionation (Figure 6C, lanes 11 and 12) or confer palmitoylation (Figure 6D, lanes 7 and 8). This suggests that the myristoylated Vac8 is able to bind at least transiently to membranes and is retained at the vacuole by directly binding to Pfa3 C134A in vivo. During the subcellular fractionation procedure, this binding is probably reduced because of the dramatic dilution. However, some wild-type Vac8 is still found on membranes and could be co-isolated with Pfa3 C134A, but not the wild-type Pfa3 (Figure 6E, lanes 5 and 6), suggesting that Pfa3 C134A is able to interact with non-palmitoylated Vac8 at the vacuole.

Figure 6.

Interaction of Pfa3 and Vac8. A) Localization of GFP-tagged Pfa3 (C134A) by fluorescence microscopy. Size bar = 10μm. B) Localization of Vac8-GFP, Vac8 (C4,5,7A)-GFP and Vac8 (G2A)-GFP in cells overexpressing Pfa3 (C134A) was determined by fluorescence microscopy as previously described. C) Subcellular fractionation. The indicated cells carrying the Vac8-GFP wild-type or Vac8 with the C4,5,7A or G2A mutations in addition to endogenous Vac8 were grown in YPG to overexpress Pfa3 and subjected to subcellular fractionation (see Figure 1B and Methods). Proteins were analysed by SDS-PAGE and western blot using antibodies against the vacuolar V-ATPase subunit Vma6, Vac8 and Arc1. Pfa3 was detected via the C-terminal Protein A tag. D) Palmitoylation of wild-type Vac8 in the presence of overexpressed Pfa3 and Pfa3-C134A was determined by biotin switch as described in Methods. E) Immunoprecipitation of Pfa3 and Pfa3 (C134A). C-terminal Protein A-tagged Pfa3 or Pfa3 (C134A) were immunoprecipitated as described in Methods, and their interaction with wild-type Vac8 was determined by SDS-PAGE and western blot using an antibody against Vac8. F) Palmitoylation of Vac8 (C4,5,7A)-GFP and Vac8 (G2A)-GFP in the presence of overexpressed Pfa3 and Pfa3-C134A was determined by biotin switch as described in Methods. G) Interaction of Pfa3 (C134A) with Vac8 mutant proteins. Cells with or without overexpressed Pfa3 (C134A), which expressed the indicated Vac8-GFP fusion protein, were subjected to immunoprecipitation and analysed as in (E).

We analysed whether the Pfa3 C134A mutant could interact equally well with Vac8 lacking the N-terminal cysteines (Vac8 C4,5,7A) or the myristoylation site (G2A). As mentioned earlier, both mutant proteins localize to vacuoles in the presence of overexpressed wild-type Pfa3 (Figure 3A–C). Interestingly, overexpression of Pfa3 C134A also conferred vacuole localization on Vac8 C4,5,7A, but not the G2A mutant (Figure 6B), suggesting that myristoylation is a prerequisite for the association of Vac8 with the enzymatically inactive Pfa3 C134A mutant. Indeed, neither mutant is palmitoylated (Figure 6F, lanes 7 and 8). As observed for the Vac8 wild-type protein, the interaction between the Pfa3 mutant and the Vac8 C4,5,7A was not strong enough to completely withstand the subcellular fractionation protocol (Figure 6C, lanes 14 and 15), albeit some membrane-bound Vac8 C4,5,7A was also co-purified with Pfa3 C134A (Figure 6G, lanes 10 and 11).

We noticed that the non-myristoylated Vac8 G2A was completely cytosolic upon overexpression of enzymatically inactive Pfa3 (Figure 6B), but localized if wild-type Pfa3 was overproduced (Figure 3C). Indeed, Vac8 G2A was efficiently palmitoylated under these conditions (Figure 6F, lane 6). This is not observed, if Pfa3 is expressed at endogenous levels (18), suggesting that increasing the amount of Pfa3 and thus the acyltransferase activity on the vacuole can compensate for the inefficient membrane binding of a non-myristoylated substrate protein. One issue is, however, puzzling. Although the G2A mutant is efficiently palmitoylated upon Pfa3 overproduction, a large fraction is found in the supernatant during subcellular fractionation (Figure 6C, lanes 8 and 9). It is possible that either palmitoylation itself is not sufficient to support Vac8 G2A membrane localization or a portion of Vac8 G2A is released from membranes during the lysis by increased depalmitoylation. When we analysed the pellet fraction, we could detect Vac8 G2A palmitoylation in the membrane fraction, which was comparable to the amount observed in the total fraction (Figure S3). Thus, recruitment of Vac8 to Pfa3 is facilitated by its myristoyl anchor, but can be bypassed by higher expression levels of Pfa3. In wild-type cells, this is not possible because of the 10-fold lower amount of Pfa3 compared to Vac8 (Figure 2B).

To obtain additional insight into the acyltransferase enzymology, we mutagenized Pfa3 and followed Vac8 palmitoylation and localization. Initially, we made truncation mutants within the cytosolic DHHC domain. Any deletion within the DHHC domain inactivated Pfa3 (not shown). We therefore generated point mutants and followed their consequences on Vac8-GFP localization and palmitoylation. To this end, we focused on amino acids (primarily cysteines) within the DHHC cytosolic domain, which were changed individually or in pairs to alanine. We then followed Vac8 localization as a read-out. Mutagenesis of conserved or non-conserved cysteines in combination led to missorting of Vac8 to the cytosol (Figure 7A). We then focused on individual amino acids. Mutants in C106, C109 or C215 localized Vac8 to the vacuolar rim and promoted palmitoylation (Figure 7A). However, cysteines immediately preceding the DHHC box (C120, C126) or mutating histidine residues (H118, H119) rendered Pfa3 inactive for Vac8 palmitoylation (Figure 7A). Furthermore, we observed that mutations in the conserved DHHC box had distinct effects on Vac8 localization and palmitoylation. As shown above (Figure 6), the DHHA (C134A) mutant could localize Vac8 without palmitoylating the protein, whereas a DHAC (H133A) mutant was inactive (Figure 7A). Surprisingly, a AHHC (D131A) mutant, which was expressed at a similar level in comparison to all other mutants analysed (not shown), was functional in both assays, suggesting that the Asp residue is not essential for palmitoylation. Our data therefore suggest that conserved residues preceding the DHHC box co-operate in the palmitoylation reaction.

Figure 7.

Mutagenesis of the DHHC cysteine-rich domain sequence. A) Overview of the Pfa3 mutants generated and the consequences on localization. The marked cysteine (indicated by numbers), histidine and aspartate residues were changed individually or in pairs to alanine. For localization, Pfa3 was overexpressed as before in the Vac8-GFP background. Palmitoylation was determined by the biotin switch protocol. B) Model of the interaction between Vac8 and Pfa3. Palmitate is indicated by the light gray color.

In summary, our data suggest a two-step acylation reaction , in which Pfa3 first recruits Vac8 by direct binding, followed by the transfer of palmitate and subsequent release of the protein laterally to the membrane. Overexpression of the active acyltransferase can compensate for poor initial membrane binding if the substrate lacks a myristoylation site. Owing to the loss of acyltransferase activity of Pfa3-DHHA, the enzyme-substrate complex is stabilized, leading to a prolonged residence time of Vac8 on overproduced Pfa3.

Discussion

Broad specificity of DHHCs

Previous palmitoylation studies revealed the specificity of enzymes for substrate proteins: the yeast Pfa3 protein promotes efficient Vac8 palmitoylation, and a subset of human DHHC proteins including Hip14, mediate PSD-95, SNAP-25 and synaptotagmin palmitoylation (14,26,27). However, the redundancy of DHHC palmitoyl acyltransferases has been implicated by several studies. Some proteins, such as yeast Vac8 or Ras2 were still partially localized even if the identified DHHC protein was deleted (13,24). Only a combination of several DHHC deletions led to a loss of palmitoylation of several proteins, including Meh1, Yck3 and Vac8 (15).

In this study, we show that five yeast DHHCs can palmitoylate multiple substrates if overexpressed. This observation may explain the compensation in palmitoylation upon loss of one DHHC protein, even if the remaining DHHC proteins are present at their endogenous expression level (15). In addition, we show that Pfa5 has acyltransferase activity (Figures 3–5). The redundancy in substrate recognition might be an important mechanism as proteins with a high turn over at membranes like the mammalian Ras GTPase (28) could be regulated at the level of DHHC activity. Our observation that Akr1 can localize Ras2 is therefore consistent with a Ras repalmitoylation at the mammalian Golgi (28). The depalmitoylation of Ras in mammalian cells could be because of a ubiquitous thioesterase or a relatively low half-life of the thioester bond. So far, the Apt1 protein has been identified in yeast and mammalian cells as a Gpa1 (Gα)-specific thioesterase (29,30), but additional studies will be necessary to address its general function in depalmitoylation.

Overexpression may bypass DHHC regulators, which otherwise ensure palmitoylation by a selected DHHC protein. In this respect, we were surprised to see rescue of activity if just Erf2, but not its co-factor Erf4 (8), was overexpressed. Potentially, a single Erf4 protein can support multiple Erf2 proteins.

The vacuolar DHHC Pfa3 protein is unusual in that it can mislocalize several proteins to the vacuole, including Ras2, Yck3, SH4-GFP and Yck2. We suggest that this recruitment occurs directly from the cytosol. In support of this notion, we observed that Yck3, which is sorted via the adapter protein complex (AP)-3 pathway to the vacuole (21) and can be modified by the ER-localized Pfa4 (Figures 3,4), can reach the vacuole surface even if the AP-3 pathway has been disrupted (Figure 5).

This retention at the vacuole might be because of the slow turn over of palmitoylated proteins at the vacuole or the fast repalmitoylation after hydrolysis of the thioester bond. Why did we then not observe an accumulation of substrates at the nuclear ER in cells overexpressing Pfa4 or at the Golgi in Akr1 overexpressing cells? Even before the DHHC family had been identified it was postulated that the distribution of acyltransferases to different compartments is sufficient to kinetically trap substrates (31,32). Our data are consistent with a model that proteins, which become modified at the ER, Golgi or plasma membrane, enter secretory or endocytic vesicles, leave the organelles soon after their palmitoylation and get sorted based on their sorting information. Proteins lacking sorting information (SH4-GFP) or the sorting machinery (Yck3 in apl5Δ) seem to get palmitoylated at the ER and Golgi and will accumulate at the plasma membrane. Transport of some palmitoylated proteins to the plasma membrane may therefore not require a specific sorting signal. Future studies using mutants that block secretion will be necessary to test this model.

In our experiments, we use overexpression as a tool to show the versatility of DHHC proteins for their substrates in yeast. However, in mammalian cells co-overexpression has been widely used to identify the DHHC and substrate pairs (26,27,33,34). Since we can show that upon overproduction almost any DHHC tested can recognize each analysed substrate, it should be taken into consideration that data from overexpression approaches that reveal specific enzyme-substrate pairs require a very careful interpretation.

Pfa3 provides insight into the palmitoylation mechanism

Our analysis suggests at least two stages in the Pfa3-mediated acylation reaction: (i) binding of the substrate to the DHHC protein, followed by (ii) palmitoylation and lateral release into the membrane. In support of this, we find that wild-type Pfa3 can recruit Vac8 lacking the N-terminal cysteines, and an enzymatically inactive DHHA variant of Pfa3 mediates vacuole localization of wild-type Vac8 and a mutant lacking the N-terminal cysteines. The latter observation is consistent with our previous observation, where we showed that a GFP-fusion protein containing the N-terminal 18 amino acids of Src, which is positively charged but lacks cysteines, is localized to vacuoles in a Pfa3-dependent manner (13). In addition, the palmitoylation-independent recruitment of Pfa3 is dosage-dependent; Vac8 C4,5,7A membrane localization is observed when Pfa3 is overproduced but not in wild-type cells (18), indicating that additional Pfa3 can recruit Vac8 C4,5,7A to the vacuole even though it is only loosely bound.

Recently, a palmitoylation-independent function has been suggested for the DHHC protein Swf1 based on the analysis of the corresponding DHHA mutant (35). As Pfa3-DHHA is able to localize Vac8 to vacuoles, we suggest that Swf1 DHHA functions in a similar manner in its regulation of the actin cytoskeleton.

Several palmitoylated peripheral membrane proteins require a second membrane-binding domain, such as myristoylation (Vac8) and prenylation (Ras2). For Vac8, myristoylation is a prerequisite to promote an efficient interaction with Pfa3. The poor recognition of non-myristolated Vac8 G2A can be bypassed by overproducing Pfa3, which can bind and then palmitoylate Vac8 G2A (Figure 6F). Surprisingly, Vac8 G2A does not remain membrane associated during subcellular fractionation. Our data suggest that the retention of the G2A mutant in vivo is primarily because of the interaction with and palmitoylation by Pfa3.

Palmitoylation via Pfa3 requires the CRD proximal to the DHHC box. Beside wild-type Pfa3, only Pfa3 with a mutation of the active site cysteine is able to promote the localization of Vac8 to the vacuole membrane, whereas mutants within the CRD or a DHAC mutant, which maintain the central cysteine, inactivate Pfa3. Surprisingly, the conserved Asp residue of the DHHC box is not essential for palmitoylation activity of Pfa3, at least under the conditions analysed. Similar mutations in Erf2 (D to A) and Akr1 (DH to AA) caused inactivation of the protein (10,23). Future experiments using purified proteins will be necessary to analyse the catalytic mechanism in detail.

In summary, our analysis of Pfa3 and other DHHC proteins provides novel insight into the enzymatic mechanism by DHHC acyltransferase and their unexpected broad substrate specificity. We suggest that the broad distribution of DHHC proteins along the endomembrane system could provide an acyltransferase network to ensure the protein palmitoylation or repalmitoylation. Future studies will need to address the acyltransferase mechanism and the potential regulation of DHHC activity.

Methods

For yeast strains and molecular biology, see Supporting Information.

Microscopy

Cells were centrifuged at 3000 ×g and washed with phosphate buffered saline (PBS) buffer. Images of yeast cells were acquired with digital camera (SPOT pursuit XS monochrome; Diagnostic Instruments, Inc.) on a LEICA DM5500B microscope equipped with differential interference contrast (DIC) optics and bandpass filter for GFP (37).

Biochemical procedures

Cell lysates were prepared as described (18,36). Subcellular fractionation was done as before (36,37). Pellet and supernatant were separated by centrifugation at 20000 ×g for 10 min at 4°C. For immunoprecipitations, cells were disrupted by glass beads in the PBS buffer (9 mm Na2HPO4, 1.8 mm KH2PO4, 136 mm NaCl, 2.7 mm KCl, pH 6.8) with 0.5% Triton-X-100, 1× protease inhibitor cocktail (PIC; 0.1 μg/mL leupeptin, 1 mmo-phenanthroline, 0.5 μg/mL pepstatin A, 0.1 mm pefablock) at 4°C. The lysate was clarified by centrifugation (20000 ×g, 10 min, 4°C) and the supernatant containing 2 mg protein in 400 μL was incubated with IgG sepharose beads (Sigma) for 2 h at 4°C on a nutator. Beads were washed twice and proteins were eluted by boiling in SDS sample buffer. They were then analysed by SDS-PAGE and western blotting.

Biotin switch assay

The biotin switch assay was done as before (13,37) with slight modification. After quenching of free cysteines by 25 mm n-ethylmaleimide (NEM) and following the chloroform/methanol precipitation, each sample containing 1 mg protein was resuspended in 150 μL resuspension buffer (2% SDS, 8 m urea, 100 mm NaCl, 50 mm Tris–HCl, pH 7.4) and incubated with 300 μm 1-Biotinamido-4-(4′[maleimidoethyl-cyclohexan]carboxamido)butane (BMCC-biotin) biotin BMCC in the presence of 900 μL 1 m hydroxylamine, pH 7.4 on a turning wheel for 2 h at 4°C. Proteins were precipitated by chloroform/methanol and resuspended in 100 μL resuspension buffer. The samples were diluted with 1 mL PBS buffer containing 0.1% Triton-X-100, 5 mm EDTA, 0.1 PIC, and incubated on a turning wheel for 15 min at 4°C. Two or 4% of sample was saved as a loading control. The remaining sample was incubated with 30 μL neutravidin-agarose beads for 1 h at 4°C on a turning wheel, washed twice with PBS buffer containing 0.5 m NaCl, 0.1% Triton-X-100 and once with PBS. Proteins were eluted by boiling in SDS sample buffer and analysed by SDS-PAGE and western blotting.

Acknowledgments

We thank Nicolas Davis for yeast strains, Angela Perz for experimental support, Clemens Ostrowicz for comments and all members of the Ungermann lab for fruitful discussions. This work was supported by the SFB431. C.U. is supported by the Hans-Mühlenhoff foundation. A.T. JP was supported by a mobility fellowship of the Boehringer Ingelheims Fonds.

Ancillary