Proteins are covalently modified with a variety of lipids, including fatty acids, isoprenoids, and cholesterol. Lipid modifications play important roles in the localization and function of proteins. The focus of this review is S-palmitoylation, the reversible addition of palmitate and other long-chain fatty acids to proteins at cysteine residues in a variety of sequence contexts. The functional consequences of palmitoylation are diverse. Palmitoylation facilitates the association of proteins with membranes, mediates protein trafficking, and more recently has been appreciated as a regulator of protein stability. Members of a family of integral membrane proteins that harbor a DHHC cysteine-rich domain mediate most cellular palmitoylation events. Here we focus on DHHC proteins that modify Ras proteins in yeast and mammalian cells.
Proteins are covalently modified with lipids through multiple mechanisms. Lipid modifications can be broadly divided into two categories: those that occur in the cytoplasm or on the cytoplasmic face of membranes, and those that occur in the lumen of the secretory pathway. Three common lipid modifications that occur in the cytoplasm are N-myristoylation, S-palmitoylation, and prenylation (Fig. 1). N-myristoylation is the covalent addition of the fatty acid myristate to an N-terminal glycine residue via an amide linkage . Prenylation is the addition of an isoprenoid, either a C15 farnesyl or a C20 geranylgeranyl group, to a C-terminal cysteine residue via a thioether linkage . Lastly, S-palmitoylation is the covalent addition of a long-chain fatty acid to a cysteine residue via a thioester linkage [3–5].
The best-characterized lipid modification that occurs in the lumen of the secretory pathway is the attachment of glycosylphosphatidylinositol (GPI) anchors . This lipid modification is composed of a phosphatidylinositol connected through a carbohydrate linker to the protein. Following addition of the GPI moiety in the endoplasmic reticulum (ER), the protein traffics through the secretory pathway to the cell surface, where the GPI anchor tethers the protein to the extracellular face of the plasma membrane.
The secreted morphogens Hedgehog and Wnt have recently gained attention as targets for lipid modification . Like GPI-anchored proteins, these proteins receive their lipid modifications in the lumen of the secretory pathway. Hedgehog is modified with both cholesterol and palmitate. As the protein undergoes an autoprocessing event that results in cleavage between the glycine and cysteine residues of a GCF motif, a cholesterol moiety is added to the now C-terminal glycine residue of the N-terminal cleavage product (Fig. 1). This same cleavage product is then N-palmitoylated on the cysteine at its extreme N-terminus (Fig. 1). The process of N-palmitoylation is postulated to occur via a thioester intermediate utilizing the thiol of the cysteine residue, followed by a spontaneous rearrangement to form the amide linkage. Wnt, on the other hand, is modified twice with acyl groups. The first modification of Wnt to be discovered was S-palmitoylation. Murine Wnt-3a is S-palmitoylated at the conserved cysteine 77. S-palmitoylation is not required for secretion of the protein but is important for Wnt's ability to signal . The second modification recently discovered on Wnt-3a was O-acylation at serine 209. Interestingly the acyl group identified by MS is palmitoleic acid (16:1), a monounsaturated fatty acid (Fig. 1). As opposed to the S-palmitoylation, the O-acylation of Wnt is associated with exit from the ER and subsequent secretion .
The focus of this review will be S-palmitoylation, hereafter referred to as palmitoylation, which occurs on the cytoplasmic surface of membranes. Palmitoylation is unique among lipid modifications in that it is reversible [3,4]. The regulation of palmitoylation status is mediated by two classes of enzymes. Protein acyltransferases (PATs) are responsible for catalyzing the addition of palmitate to the substrate, and protein acylthioesterases are responsible for the removal of the palmitate. A protein could therefore undergo several rounds of palmitoylation and depalmitoylation during its lifetime, either constitutively or in response to signals. In the remainder of this review, we will briefly discuss the sequence context of the sites of palmitoylation and the role of palmitoylation in the function of proteins. We will then discuss the enzymes responsible for palmitoylation, including our own work on a PAT that modifies mammalian Ras proteins.
Myristoylation and prenylation have well-defined consensus sequences. Myristoylation absolutely requires an N-terminal glycine residue that is exposed upon cleavage of the initiator methionine or generated by a proteolytic cleavage event . Prenylation occurs at a C-terminal CaaX motif, where C represents a cysteine, a is typically an aliphatic amino acid, and the identity of amino acid X determines whether the protein will be modified with a farnesyl or geranylgeranyl group. Geranylgeranylation of proteins can also occur at C-terminal CC or CXC motifs . Palmitoylation, on the other hand, has no single sequence requirement outside of the presence of a cysteine residue. Despite the absence of a universal palmitoylation motif, many palmitoylated proteins can be categorized according to the sequence context of their palmitoylation. To begin, palmitoylated proteins can be subdivided into two categories: proteins synthesized on free ribosomes that associate peripherally with membranes, and proteins containing transmembrane domains (TMDs).
Palmitoylated proteins peripherally associated with membranes can be further subdivided into dually lipidated proteins and solely palmitoylated proteins . Palmitoylation is often found adjacent to other lipid modifications. This is a feature of numerous signal transducers. G-protein α-subunits of the Gi family and many nonreceptor tyrosine kinases are palmitoylated at one or more cysteines adjacent to the myristoylated glycine, whereas N-Ras and H-Ras are palmitoylated at one or two cysteines, respectively, adjacent to the farnesylation site. The sites of palmitoylation of proteins exclusively modified with palmitate are found throughout proteins, frequently at pairs or longer stretches of cysteine residues in close proximity. Proteins with TMDs are often modified at cysteine residues at the interface of the cytoplasm and membrane or in cytoplasmic C-terminal tails.
Functions of palmitoylation
As the field of palmitoylation has advanced, it has become clear that the roles of palmitoylation are diverse. The most commonly described function of palmitoylation is to increase the affinity of a soluble protein for membranes, which can thereby affect the protein's localization and function. This is the case for both solely palmitoylated proteins and dually lipidated proteins. This raises the question of why a protein with one lipid modification needs a second lipid modification to associate with membranes. Biophysical studies of lipidated peptides and model membranes, as well as more recent work using fluorescence bleaching techniques and live cell imaging, are consistent with the kinetic membrane trapping model proposed by Silvius to explain the behavior of lipid-modified proteins in cells (Fig. 2) [11–13]. This hypothesis states that myristoylation and prenylation promote transient interactions with membranes. Accordingly, they are able to sample different membranes in the cell. The addition of the palmitate by a PAT yields a dually lipidated protein that has a long-lived association with the membrane. The subcellular localization of the PAT then determines where its substrate becomes stably anchored to membranes. The substrate can then move to different compartments but will do so by vesicle-mediated transfer. Depalmitoylation of a substrate will return the protein to a state where it is rapidly cycling on and off membranes and trafficking through a vesicle-independent mechanism.
Regulation of protein trafficking by palmitoylation has been best illustrated for palmitoylated forms of the small GTPase Ras [4,5,14]. All Ras isoforms undergo a complex series of post-translational modifications at their C-terminal CaaX motif: farnesylation, proteolytic removal of the aaX motif and carboxylmethylation of the farnesylated cysteine. N-Ras and H-Ras are further modified with palmitate at one or two cysteines, respectively, upstream of the farnesylated cysteine. Farnesylation occurs in the cytoplasm, whereas proteolysis and carboxylmethylation occur on the cytoplasmic face of the ER. H/N-Ras is then palmitoylated early in the secretory pathway (at either the ER or Golgi) and moves to the plasma membrane by vesicular transport. Here it is available to participate in Ras signaling events at the cell surface. At the plasma membrane, H/N-Ras is depalmitoylated and traffics by a diffusion-limited process. At the Golgi, Ras can be repalmitoylated and again be stably anchored to membranes. It appears that the more stable the membrane attachment, the more prominent the plasma membrane localization, as the dually palmitoylated H-Ras accumulates to a greater extent at the plasma membrane than the monopalmitoylated N-Ras, which is more prominently distributed to the Golgi. This has important functional consequences, as Ras proteins signal at the Golgi in response to a different repertoire of activators than they see at the cell surface . Accordingly, the signaling output from the cell will depend on the localization of H/N-Ras, which in turn is regulated in part by its palmitoylation status.
Another function for palmitoylation is to modulate protein stability. One way in which palmitoylation performs this function is as a quality control checkpoint. This is the case for the yeast chitin synthase Chs3, a protein with six to eight TMDs. Palmitoylation of Chs3 appears to be an indicator of proper folding. When palmitoylation is blocked, Chs3 aggregates and is retained in the ER. Although Chs3 is not unstable or degraded, chitin deposition on the cell surface is markedly decreased. Thus, palmitoylation is playing an important role in a quality control mechanism for a multimembrane-spanning protein .
A second example in which palmitoylation plays a role in quality control is the yeast SNARE protein Tlg1 , which is localized in the Golgi apparatus and mediates the fusion of vesicles with the late Golgi compartment. Tlg1 is palmitoylated on two cysteine residues adjacent to its TMD. Palmitoylated yeast SNARE proteins are modified shortly after synthesis and insertion into the ER membrane, and maintain the acylation for the duration of the protein's life. Thus, the function of the palmitoylation must be constitutive rather than regulated. Palmitoylation of Tlg1 prevents interaction of the SNARE with the E3 ubiquitin ligase Tul1. This ligase specifically recognizes TMD polar residues . Integral membrane proteins with polar TMDs are likely to be misfolded, due to unfavorable interactions with the lipid bilayer. Palmitoylation near its TMD may orient Tlg1 in a more favorable conformation with respect to the membrane, allowing it to escape ubiquitination. Mutation of either Tlg1 cysteine residue results in a protein that interacts with Tul1, is ubiquitinated, and is targeted to the vacuole interior for degradation . Again, palmitoylation is playing a role in quality control.
A second potential role for palmitoylation in protein stability is in regulating protein degradation. This appears to be the role of acylation for the yeast sphingoid long-chain base kinase Lcb4. Lcb4 is a soluble protein palmitoylated at two internal cysteine residues, which anchors it to the membrane. Under normal conditions, Lcb4 is downregulated upon entry of the cells into stationary phase. This does not occur when palmitoylation of Lcb4 is blocked by mutation. Thus, in contrast to the example presented above, the half-life of the Lcb4 protein is negatively regulated by palmitoylation .
The examples described above illustrate that palmitate exerts its effects in diverse ways. It will be interesting to determine how widely palmitoylation is used as a mechanism to regulate protein stability and to define the mechanism or mechanisms by which palmitate promotes stability in some contexts but destabilizes proteins in others.
Mechanism of palmitoylation
Elucidating the mechanism by which proteins are palmitoylated has been a matter of intense investigation for some time. Two hypotheses have been proposed for how proteins are modified with palmitate, and in recent years evidence has been found for both . One hypothesis is spontaneous acylation of the palmitoylated protein. In this model, palmitate is transferred from acyl-CoA to a reactive cysteine residue in the protein in the absence of any other protein factors. In mitochondria, the high local acyl-CoA concentrations are thought to support spontaneous acylation of several mitochondrial enzymes [20,21]. More recently, several lines of evidence suggest that spontaneous acylation accounts for palmitoylation of the yeast transport protein Bet3. First, Bet3 rapidly and stoichiometrically incorporates palmitate in vitro when incubated with palmitoyl-CoA at physiological concentrations and pH . Second, when produced in bacteria, which lack the capacity to palmitoylate proteins, Bet3 is stoichiometrically modified with palmitate . Third, Bet3 remains palmitoylated in yeast strains deleted for genes that encode palmitoylating enzymes (see below) . Interestingly, palmitoylation of Bet3 appears to play a structural role in stabilizing the protein .
The second hypothesis is that palmitate transfer from palmitoyl-CoA to a protein substrate is mediated by an enzyme. Traditional biochemical approaches demonstrated the existence of PAT activities in membranes, but the instability of detergent-solubilized enzyme prevented molecular identification of the species responsible for PAT activity . As described below, genetic screens in yeast uncovered the first PAT candidates.
Lipid modification of Ras proteins is conserved from yeast to humans. Thus, Ras proteins in Saccharomyces cerevisiae are farnesylated and palmitoylated, similar to mammalian H-Ras and N-Ras. Our collaborator R. Deschenes and his coworkers (University of Iowa) isolated a nonfarnesylated Ras mutant that was dependent upon palmitoylation for its ability to support the viability of yeast . The protein contains a C-terminal extension of basic amino acid residues that prevents farnesylation, but palmitoylation is preserved at the cysteine palmitoylated in the wild-type protein. The combination of the basic amino acids and the palmitoylation of the protein substitutes for the prenylation and palmitoylation found on the wild-type protein. This led to the idea that Ras2 PAT candidates could be isolated by screening for mutations that were lethal in the presence of the palmitoylation-dependent Ras2 protein. Using the palmitoylation-dependent Ras2 yeast strain, mutants were isolated that significantly reduced or blocked Ras2 palmitoylation and that resulted in lethality. Two genes of interest came out of this screen, designated ERF2 and ERF4 (effect on Ras2 function) . Erf4 had previously been identified as a partial suppressor of an activated form of Ras that confers heat shock resistance and was designated SHR5. Erf2 and Erf4 form a protein complex and localize to the ER, where Ras2 undergoes postfarnesylation processing before trafficking to the plasma membrane [27,29–31]. Deletion of either ERF2 or ERF4 caused both a reduction in Ras2 palmitoylation and a relocalization of Ras2 to internal membranes [27,31].
We collaborated with the Deschenes group to test whether Erf2 and Erf4 had PAT activity for yeast Ras in vitro. Sandra Lobo (University of Iowa) showed that incubation of the purified Erf2–Erf4 complex with Ras2 and palmitoyl-CoA promoted incorporation of palmitate into Ras2, establishing the Erf2–Erf4 complex as the Ras PAT. A notable feature of Erf2 is the presence of an Asp-His-His-Cys (DHHC) motif embedded in a cysteine-rich domain (CRD). This domain was critical for PAT activity for Ras in vitro and for Ras function in vivo. Concurrently, N. Davis (Wayne State University) identified the yeast ankyrin-repeat-protein Akr1 as a PAT for yeast casein kinase 2 (Yck2) . Hence, Erf2–Erf4 and Akr1 represent the first PATs validated by both genetic and biochemical criteria.
Akr1, like Erf2, is an integral membrane protein containing a DHHC-CRD that is critical for its PAT activity (Fig. 3). Unlike Erf2, Akr1 has an additional two TMDs and a series of N-terminal ankyrin repeats. Interestingly, both Erf2 and Akr1 become palmitoylated (autoacylated) when incubated with palmitoyl-CoA. Autoacylation and transfer of palmitate to the substrate are abolished by mutation of the cysteine within the DHHC motif [32,33]. This is consistent with the formation of an acyl-enzyme intermediate, but additional studies are required to establish this. An important distinction between these first two PATs is the requirement for accessory proteins. Akr1 is active as a PAT in the absence of a stoichiometric accessory protein, whereas the DHHC protein Erf2 requires Erf4 for PAT activity.
It was of obvious interest to identify a mammalian ortholog of the Ras PAT. The gene product of ZDHHC9 was chosen as a candidate, due to its homology to Erf2. DHHC9 shares 70% sequence identity with Erf2 in the DHHC-CRD and 31% identity overall. Initial work with DHHC9 demonstrated that the protein alone could not act as a Ras PAT, suggesting the need for a mammalian ortholog of Erf4. No obvious orthologs for Erf4 were evident in mammalian genomes. However, through iterative blast searches, we eventually identified a candidate, Golgi complex protein (GCP) of 16 kDa (GCP16) . GCP16 was initially characterized as an interacting protein of the golgin family member GCP170 . J. Swarthout in our laboratory, in collaboration with Deschenes and coworkers, demonstrated that DHHC9 and GCP16 are functional orthologs of Erf2 and Erf4 . We showed that GCP16 and DHHC9 coimmunoprecipitated from tissue culture cells and displayed a similar subcellular distribution. Both proteins localized to the Golgi, with a portion of DHHC9 also found at the ER (Fig. 4A). This distribution is consistent with the predicted localization of a Ras PAT in the Golgi based on studies of Ras trafficking in mammalian cells [12,13]. DHHC9 and GCP16 were purified as a stoichiometric complex from insect cells infected with recombinant DHHC9 and GCP16 baculoviruses. Purified DHHC9/GCP16 palmitoylated both H-Ras and N-Ras in vitro. Like the Erf2–Erf4 complex, DHHC9 required GCP16 to function as the PAT for H-Ras.
Prior to this study, it was not clear whether palmitate transfer to substrates by DHHC PATs was catalytic or whether they acted stoichiometrically as palmitate transfer proteins . To address this, Swarthout performed a kinetic analysis of N-Ras palmitoylation by DHHC9/GCP16. Examination of the time course reveals that the reaction is rapid, linear to 8 min, and reaches a maximum by 10 min (Fig. 4B). The N-Ras substrate curve produced a turnover number of 3 min−1 (Fig. 4C), suggesting that the reaction is indeed catalytic .
The DHHC family of proteins
The DHHC-CRD defines a family of PATs with seven members in yeast and 23 in humans . Yeast has continued to serve as an excellent model with which to elucidate the biological roles of this family of enzymes. An important prediction of the kinetic trapping model described above is that the localization of PATs will dictate in which compartment soluble substrates become associated with membranes. The yeast DHHC proteins are widely distributed on yeast membranes. Erf2 , Swf1  and Pfa4  reside on the ER. Akr1 and Akr2 are localized at the Golgi [24,38]. Pfa5 is found at the plasma membrane  and Pfa3 is localized on the vacuole . Genetic and/or biochemical analyses have established additional enzyme–substrate pairs. We demonstrated that the DHHC protein Pfa3 palmitoylates the vacuolar protein Vac8 . Lcb4 is palmitoylated by Akr1 . Swf1 palmitoylates yeast SNARE proteins such as Tlg1 , and Pfa4 palmitoylates Chs3 .
Recently, a global analysis of palmitoylation was performed in yeast that has considerably expanded the palmitoyl proteome . Palmitoylated proteins were tagged by exchanging the palmitate on modified cysteine residues with biotin using a thiol-reactive crosslinker (acylbiotin exchange method ). Palmitoylated proteins were isolated and trypsinized, and the tryptic peptides were analyzed by MS. The analysis identified 35 new palmitoyl proteins, including a family of amino acid permeases and numerous signaling and trafficking proteins. The same type of analysis was performed in yeast strains, where the genes that encode DHHC proteins were deleted singly or in combination. This analysis revealed putative enzyme–substrate pairs and demonstrated that DHHC proteins account for most cellular palmitoylation events .
There are 23 human DHHC genes, denoted ZDHHC1–ZDHHC24 (ZDHHC10 is not annotated as a gene). Significant progress has been made in characterizing the large family of enzymes encoded by these genes with respect to substrate identification, localization, and expression patterns . One of the most fruitful approaches was pioneered by Fukata et al. . They cloned 23 murine DHHC genes, expressed them individually in tissue culture cells, and screened for the ability to increase radiolabeled palmitate incorporation into a substrate of interest. The neuronal scaffold protein PSD-95 was the first protein analyzed. Four DHHC proteins were identified as candidate PSD-95 PATS: DHHC2, DHHC3, DHHC7, and DHHC15. DHHC15 was further characterized biochemically and shown to palmitoylate PSD-95 in vitro, and a dominant negative form of DHHC15 reduced palmitoylation in cells. A similar approach was used with the substrate endothelial nitric oxide synthase (eNOS) . DHHC2, DHHC3, DHHC7, DHHC8 and DHHC21 were identified as candidate PATs by the coexpression assay. Attenuation of DHHC21 expression in endothelial cells decreased eNOS palmitoylation, further supporting its identity as an eNOS PAT. In both studies, there were functional consequences of inhibiting the DHHC proteins for the localization or activity of the substrates. Given the physiological and pathophysiological importance of many palmitoylated proteins, the development of DHHC protein inhibitors would provide important tools with which to explore the biology of palmitoylation and could potentially have therapeutic potential.
The medical importance of DHHC proteins is emerging as mutations in ZDHHC genes have been associated with human diseases. A female patient with X-linked mental retardation was determined to have a balanced reciprocal translocation t(X;15) that disrupts transcription of DHHC15. Analysis of additional patients revealed five nucleotide exchanges and two 4 bp deletions in the 3′-UTR region of DHHC15, further supporting linkage of the disease to DHHC15 . A second study of X-linked metal retardation revealed mutations in the gene ZDHHC9. Truncating and missense mutations were found, with the amino acid changes being in the DHHC-CRD, although the consequences of these mutations have yet to be discerned . These findings and others demonstrate the importance of investigating the biology and enzymology of DHHC proteins and expanding our understanding of palmitoylation as a regulatory post-translational modification.