The fat controller: roles of palmitoylation in intracellular protein trafficking and targeting to membrane microdomains (Review)

Abstract

The attachment of palmitic acid to the amino acid cysteine via thioester linkage (S-palmitoylation) is a common post-translational modification of eukaryotic proteins. In this review, we discuss the role of palmitoylation as a versatile protein sorting signal, regulating protein trafficking between distinct intracellular compartments and the micro-localization of proteins within membranes.

• Palmitoylation
• protein sorting
• membrane microdomains
• lipid rafts
• protein trafficking

Abbreviations
TMD	=	transmembrane domain
DHHC	=	aspartic acid-histidine-histidine-cysteine
CSP	=	cysteine-string protein
SNAP25	=	synaptosomal-associated protein of 25 kDa
GPCR	=	G protein-coupled receptor
DRM	=	detergent-resistant membrane.

Introduction

The covalent attachment of fatty acids is a common modification of eukaryotic proteins [1]. This process, termed ‘acylation’, can occur co-translationally when myristate (C14, saturated) is added to the free amino group of a glycine residue present at position 2 in the polypeptide chain, following cleavage of the initiating methionine residue (‘N-myristoylation’). In contrast, the modification of cysteine residues with palmitic acid (C16, saturated) via a thioester linkage (‘S-palmitoylation’) is a post-translational modification, which can occur shortly after protein synthesis or at later stages throughout the life-time of a protein. Although palmitate is thought to be the major fatty acid attached to cysteines, it should be noted that other fatty acids may also be attached to cysteine residues by a similar thioester linkage (this point will be discussed in more detail in the section ‘Palmitoylation as a Signal for Retention’). Both N-myristoylation and S-palmitoylation can regulate protein-membrane interactions due to an affinity of the hydrophobic lipid groups for membranes. Although N-myristoylation is irreversible, several N-myristoylated proteins exhibit a ‘myristoyl switch’, where the myristate group is either exposed on the surface of the protein or sequestered into a hydrophobic pocket [2]. These structural changes facilitate dynamic regulation of membrane association of N-myristoylated proteins without removal of the myristate group. In contrast, the thioester bond linking cysteines and palmitate groups is labile and many S-palmitoylated proteins undergo cycles of palmitoylation and depalmitoylation [3].

A major lag in the study of cellular S-palmitoylation arose due to repeated failures to identify enzymes that catalysed this lipid modification [4]. Indeed, there was a period of greater than 20 years between the first reports of protein palmitoylation [5] and the subsequent identification of bona fide palmitoyl transferases [6–10]. During this period there was controversy as to whether S-palmitoyl transferases were required for cellular protein palmitoylation, or whether palmitoylation in vivo occurred spontaneously, as had been observed in vitro, e.g. [11]. A family of 23 mammalian palmitoyl transferases has since been identified by the presence of a signature DHHC-cysteine-rich domain. The large number of putative S-palmitoyl transferases (by contrast only 2 N-myristoyl transferases are present in vertebrates [12]), is consistent with the lack of a general sequence-specific consensus site for S-palmitoylation.

The DHHC family of S-palmitoyl transferases are predicted to have overall similar topologies with 4 or more transmembrane domains (TMDs) and the DHHC domain cytoplasmically orientated [13]. DHHC proteins have been localized to ER, Golgi, plasma membrane, endosomes and the yeast vacuole [14–16]; as yet the determinants that underlie the differential localization of DHHC proteins are unknown. Studies in yeast have revealed that DHHC proteins are responsible for the large majority of palmitoylation events [17], although a small number of proteins may be palmitoylated in a DHHC-independent manner (see e.g. [18]).

Over recent years, S-palmitoylation has emerged as a common requirement for efficient protein trafficking [19], regulating protein movement between several intracellular compartments. However, no universal mechanism to account for the effects of palmitoylation on protein sorting is currently known, and in many cases palmitoylation is likely to direct trafficking in a protein-specific manner. Here, rather than provide a comprehensive literature review of proteins that display palmitoylation-dependent sorting, we will instead focus more on specific studies that offer insight into the mechanisms that regulate this process. For more general information on protein palmitoylation, readers are referred to other review articles [1], [20], [21].

Palmitoylation and stable membrane binding

For many proteins, a primary function of palmitoylation is to enhance membrane affinity, allowing the modified protein to accumulate on membranes [22]. In several cases, N-myristoylation or the attachment of isoprenyl modifications provide substrate proteins with a weak membrane affinity, allowing the proteins to interact transiently with membranes [3], [23], [24], and hence to localize in proximity to membrane-bound DHHC palmitoyl transferases. The subsequent palmitoylation of cysteine residues proximal to the myristoylated/isoprenylated residues creates a ‘dual lipid’ anchor, and conversion from a single lipid modification to a dual modification greatly increases membrane affinity [22]. In this way, palmitoylation inhibits rapid cycling of the protein between cytosol and membrane by ‘trapping’ the modified protein on the membrane. The identity of the partner DHHC protein(s) therefore determines which membrane compartment the substrate localizes to. Following palmitoylation, the substrate may remain on the same membrane or sorting signals within the palmitoylated domain or other regions of the protein may facilitate subsequent trafficking.

Several other proteins are dependent upon palmitoylation for stable membrane attachment but are not modified by myristoyl/prenyl groups. Work from our group has indicated that an intrinsic weak membrane affinity may also operate as a mechanism to allow such substrates to access DHHC proteins [25], [26]. Cysteine-string protein (CSP) is multiply palmitoylated by Golgi-localized enzymes [26] on a central cysteine-string domain, and mutational analyses suggest that the cysteine residues play a direct role in membrane interaction prior to palmitoylation. In particular, the replacement of specific cysteines with more hydrophilic amino acids (i.e. serines) inhibits membrane binding and palmitoylation, whereas substitution with alanines has little effect on these parameters [25], suggesting that cysteine hydrophobicity contributes to membrane interactions prior to palmitoylation. Intriguingly, the introduction of more hydrophobic residues (leucines) at specific sites in the cysteine-string domain resulted in an increased membrane affinity of the unpalmitoylated protein [26] and prevented palmitoylation of the remaining cysteines [25]. The enhanced membrane affinity of such CSP mutants was proposed to inhibit palmitoylation by promoting association with membranes that do not contain partner DHHC proteins (in particular, the ER). Thus, the initial membrane affinity of CSP has to be set within an optimal range: if membrane affinity is too low the substrate cannot associate with membranes to allow access to DHHC enzymes, whereas if membrane affinity is too high, cycling between the cytosol and membrane is inhibited and the substrate sticks to inappropriate membranes (ie membranes lacking the partner DHHC proteins). Thus, membrane ‘sampling’ via either single lipid moieties or protein domains with an intrinsic weak membrane affinity may be a widespread mechanism employed by substrate proteins to search for their partner DHHC proteins.

Exiting the ER

As discussed in the previous section, several proteins rely on palmitoylation for stable membrane attachment. However, palmitoylation also frequently occurs on transmembrane proteins. Membrane-spanning domains of integral membrane proteins are predominantly inserted co-translationally into the ER membrane. For non-ER-resident proteins, several studies have highlighted the importance of palmitoylation for regulating ER exit. There is currently no unifying theme describing how palmitoylation regulates movement from the ER but one mechanism likely involves the reorientation of the TMDs. Lipoprotein receptor-related protein 6 (LRP6) is a co-receptor for Wnt, and functions in the canonical Wnt signalling pathway that regulates gene transcription via β-catenin [27–29]. LRP6 was shown to be palmitoylated on juxtamembrane cysteine residue(s) [30], and palmitoylation-deficient mutants of LRP6 were retained at the ER, preventing plasma membrane delivery. The trafficking of transmembrane proteins from the ER to other intracellular membranes (in particular the plasma membrane) may require conformational changes in the membrane-spanning region(s). The reason for this is that the distinct lipid [31] and protein [32] composition of intracellular membranes results in marked differences in bilayer thickness. Thus, in polarized liver cells, the width of the ER membrane was estimated at 37.5 Å, the Golgi at 39.5 Å and the apical plasma membrane at 42.5 Å [32]. As intracellular membranes vary in thickness, the length of hydrophobic transmembrane domains has been proposed to play an important role in intracellular sorting, presumably via matching of the TMD length to the thickness of the hydrophobic region of the lipid bilayer [33], [34] (Figure 1); thus, longer TMDs target to the plasma membrane and shorter TMDs to intracellular membranes such as the Golgi or ER [35–37]. Nevertheless, integral membrane proteins that are associated with the plasma membrane at steady-state must still transit through the ER-Golgi following synthesis, and this poses a problem regarding matching of the long TMDs of plasma membrane proteins with the thin hydrophobic core of the ER bilayer. Based upon the values calculated for rat liver membranes, a protein trafficking from the ER to the apical plasma membrane would need to be accommodated in membranes that vary in thickness by around 13%. Thus, an important question regarding LRP6 trafficking is how the long TMD (23 amino acids) is accommodated in the ER membrane prior to delivery to the plasma membrane. If the TMD is ‘designed’ for plasma membrane residence, then its presence in the ER membrane would be predicted to result in a positive mismatch (hydrophobic protein segment longer than thickness of the bilayer). Although this mismatch may be overcome through rearrangements of the lipid groups in the membrane bilayer, the protein may also re-orientate to minimize mismatch.

Figure 1.  Hydrophobic matching and mismatch of transmembrane domains and membrane bilayer. (i) Hydrophobic matching occurs when the length of a proteins TMD (blue cylinder) is similar to the length of the hydrophobic region of the bilayer. (ii) Negative mismatch occurs when the length of the TMD (red cylinder) is significantly shorter than the length of the hydrophobic region of the bilayer. (iii) Positive mismatch occurs when the length of the TMD (green cylinder) is significantly longer than the length of the hydrophobic region of the bilayer. (iv) Positive mismatch can be relieved by TMD tilting, which may be facilitated by palmitoylation (solidbrown line). This figure is reproduced in colour in Molecular Membrane Biology online.

One possibility is that palmitoylation facilitates the re-orientation of TMDs, maximizing hydrophobic matching. In support of this idea, a study in model membranes suggested that palmitoylation was able to tilt a transmembrane peptide [38] (see Figure 1). Intriguingly, plasma membrane delivery of the palmitoylation-deficient LRP6 mutant could be rescued by shortening the length of the TMD by as little as 2 amino acids [30]. This result was interpreted as suggesting that the role of LRP6 palmitoylation at the ER is to decrease the relative length of the TMD (i.e. by tilting). Deletion of 2 amino acids would be predicted to decrease the length of the LRP6 TMD by around 10%, consistent with the calculated difference in thickness of ER and apical plasma membrane of ∼13% [32]. As a rough guide, an intramembrane tilt angle of 20–25° would be required in order to accommodate a 23-amino acid spanning segment into the same bilayer thickness that would be occupied by a 21-amino acid segment. It is not clear how palmitoylation would cause tilting of TMD regions, but this could involve a direct interaction of the palmitate group with the TMD. Alternatively, the preference of the TMD and the palmitate group for less ordered and more ordered regions of the membrane, respectively, may result in TMD reorientation to accommodate the microdomain preferences of both groups [19].

If palmitoylation promotes tilting of the TMD of LRP6 then it would be predicted that a depalmitoylation step is required when LRP6 reaches the thicker plasma membrane to prevent a negative hydrophobic mismatch. However, characterization of the half-life of ³H-palmitate incorporation into LRP6 compared with half-life estimates of protein life-time suggested that palmitoylation of LRP6 is not dynamic. Other possible mechanisms whereby tilting of the TMD of palmitoylated LRP6 at the plasma membrane could be prevented or accommodated include: (i) the movement of palmitoylated LRP6 into more disordered regions of the plasma membrane; (ii) association with partner protein(s) at the plasma membrane might reverse the effects of palmitoylation on LRP6 transmembrane configuration; (iii) palmitoylation-dependent tilting of the LRP6 TMD may be dependent upon the specific lipid composition of the ER membrane, and the more ordered plasma membrane might prevent tilting; and (iv) the intrinsic forces driving hydrophobic matching may be dominant and palmitoylation might only affect orientation of TMDs when proteins are destabilized due to mismatching. However, with regard to this last point, it is interesting that deletion of 2 amino acids from the (palmitoylated) wild-type LRP6 protein caused ER retention. This was suggested to be caused by palmitoylation-mediated tilting of the shorter TMD, leading to a negative mismatch with the ER membrane [30].

The requirement for palmitoylation for ER export of proteins is conserved in lower eukaryotes and the polytopic yeast chitin synthase, Chs3p, which localizes to the plasma membrane and endosomes, is palmitoylated at the ER and this is essential for ER release [39]. In this case, inhibiting palmitoylation promoted aggregation of Chs3p in the ER, similar to that observed when the interaction of Chs3p with specific ER chaperones is prevented. Thus, as with LRP6, the role of palmitoylation in regulating ER exit of Chs3p may be related to preventing the effects of positive hydrophobic mismatch. Despite the important role played by palmitoylation in regulating ER exit of transmembrane proteins, not all substrates modified by ER-localized DHHC proteins require palmitoylation to leave the ER. The yeast endosomal SNARE protein Tlg1p is palmitoylated by the ER-localized DHHC protein Swf1p, however deletion of Swf1p does not lead to Tlg1p accumulation at the ER [40].

Palmitoylation as a signal for retention

The role of palmitoylation in regulating protein trafficking is well-established [19]. In addition, several studies have also reported that over-expression of specific DHHC proteins promotes the retention of the substrate protein on the membrane housing the DHHC protein. One interesting example of this is the accumulation of AMPA (α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid) receptor subunits on Golgi membranes following over-expression of DHHC3/GODZ, which was suggested to be a regulatory mechanism controlling cell surface expression of the receptor subunits [41]. However, this type of observation may not reflect a direct role for palmitoylation in membrane retention, and it is worthwhile noting that several proteins, such as the membrane fusion protein SNAP25, accumulate on intracellular membranes following over-expression of specific DHHC proteins (see e.g. [7]). However, in the absence of DHHC over-expression, SNAP25 is efficiently palmitoylated and traffics to the plasma membrane without showing obvious accumulation on Golgi membranes in neuronal/neuroendocrine cells [42]. This suggests that intracellular accumulation of SNAP25 (and most likely some other substrates) following DHHC over-expression may be a direct consequence of over-expression rather than palmitoylation acting as an intracellular retention signal. This intracellular retention might occur following co-transfection with DHHC proteins due to a lack of available accessory proteins that regulate substrate binding and release. For example, palmitoylation of mammalian and yeast Ras proteins is mediated by the DHHC protein DHHC9/Erf2p working in conjunction with GCP16/Erf4p [9], [43].

Despite these caveats, the notion that the attachment of fatty acids could prevent forward traffic from the Golgi (and other intracellular membranes) is an interesting proposal. In this regard, it is important to reflect that the use of the term ‘palmitoylation’ arose from experiments detailing the incorporation of [³H]-labelled palmitate into cellular proteins. However, when less biased approaches are employed, such as mass spectrometry, it is clear that although palmitate is a major lipid modification of acylated proteins, significant amounts of stearate (C18, saturated), palmitoleate (C16, unsaturated) and oleate (C18, unsaturated) can also be attached via thioester linkage [44–46]. Thus, an interesting question regards the lipid-coA specificity of DHHC proteins: are these specificities similar or do different enzymes have distinct lipid-coA preferences? The incorporation of significant amounts of unsaturated fatty acids into substrate proteins may well prevent forward transport by weakening the affinity of modified proteins for cholesterol/sphingolipid-rich lipid raft export domains at the Golgi [47], and this is clearly an area that warrants further investigation. A detailed description of lipid raft domains is provided later in the review.

Palmitoylation and lysosomal targeting

Several reports have shown that mutation of palmitoylation sites promotes protein degradation at lysosomes/vacuoles [21]. This is the case for neutral sphingomyelinase 2 and the chemokine receptor CCR5, palmitoylation-deficient mutants of which show reduced cellular expression levels that are restored by lysosomal inhibitors, a treatment that also reveals co-localization of the proteins with lysosomal marker proteins [48], [49]. In many cases the mechanisms by which palmitoylation prevents lysosomal targeting of proteins is not clear. However, for the yeast SNARE protein Tlg1p, vacuole targeting and destruction appears to result from an increased susceptibility to ubiquitylation when palmitoylation is blocked [40]. Modification of proteins with the protein ubiquitin often serves as a tag to target proteins for degradation at the proteosome or lysosomal compartment [50]. Ubiquitylated proteins are internalized into endosomal multivesicular bodies, which deliver cargo to lysosomes for subsequent destruction [50]. Similarly, palmitoylation-deficient mutants of the anthrax receptor, tumour endothelial marker 8 (TEM8), show enhanced ubiquitylation and lysosomal destruction [51]. Thus, modification by ubiquitin ligases may be a widespread signal regulating turnover of proteins when palmitoylation is perturbed (Figure 2).

Figure 2.  Regulation of protein sorting and turnover by S-palmitoylation. The schematic highlights specific examples of how palmitoylation contributes to protein sorting and stability. Proteins such as LRP6 and Chs3p (dark blue cyclinder) require palmitoylation to exit the ER and target the plasma membrane. Palmitoylation regulates the stability of Tlg1p (light blue cyclinder), which localizes to endosomes and the trans Golgi network, and also cell surface proteins such as TEM8 (green cyclinder). For both Tlg1p and TEM8, decreased palmitoylation leads to an increase in ubiquitylation (Ub), causing the proteins to be routed to the lysosome/vacuole and subsequently degraded. The internalization and recycling of many proteins (e.g. GPCRs) is also regulated by palmitoylation (brown cylinders). The inset highlights the role of palmitoylation in regulating protein micro-localization between raft and non-raft domains. ER, endoplasmic reticulum; ERGIC, ER-Golgi intermediate compartment; MVB, multivesicular body.

In addition to preventing lysosomal destruction, the differential usage of palmitoylation sites may also regulate protein sorting to the limiting membrane of lysosomes. For example, palmitoylation appears to be important in determining the localization of certain Src family kinases (SFKs) between the plasma membrane and lysosomes. The myristoylated c-Src [52] and p61^Hck[53] proteins were shown to associate with lysosomal membranes. In contrast, Lyn [52] and p59^Hck[53], which are myristoylated and palmitoylated are predominantly associated with the plasma membrane, but display an increased lysosomal targeting when palmitoylation sites are mutated. Thus, as well as ensuring that proteins avoid lysosomal destruction, the differential use of palmitoylation sites in closely related proteins may serve as a physiological mechanism to regulate trafficking to the lysosomal membrane and hence facilitate functional expression of proteins at this compartment.

Palmitoylation and trafficking in the endosomal system

Several G protein-coupled receptors (GPCRs) have been reported to be palmitoylated just downstream of the seventh transmembrane domain [54]. Palmitoylation is often required for efficient plasma membrane delivery of GPCRs [55–57]. In addition, palmitoylation has been proposed to regulate internalization of specific GPCRs from the plasma membrane. For example, palmitoylation-deficient mutants of Luteinizing hormone receptor exhibited a faster rate of ligand-induced internalization and subsequent degradation of ligand [58], without changes in ligand affinity [59]. Conversely, several other GPCRs exhibit decreased internalization when palmitoylation is blocked and, for the thyrotropin-releasing hormone receptor-1 and the V2 vasopressin receptor, this decreased internalization was suggested to reflect an inhibition of β-arrestin recruitment to the activated receptors [60], [61]; β-arrestin functions in clathrin-dependent internalization of activated GPCRs [62]. The palmitoylation-dependent regulation of protein-protein interactions is also important for internalization of other proteins, such as AMPA receptors. Palmitoylation of GluR subunits of AMPA receptors was suggested to inhibit binding to the cytoskeletal 4.1N protein, facilitating internalization of the receptor [41]. Palmitoylation of the synaptic vesicle (SV) Ca²⁺-binding protein, synaptotagmin I regulates internalization and SV sorting in neurons and neuroendocrine cells [63]. The decreased internalization of palmitoylation-deficient synaptotagmin I correlated with a reduction of multimerization, suggesting that assembly of synaptotagmin I into homo-oligomers in a palmitoylation-dependent pathway may be important for internalization [63]. In addition, as synaptotagmin I interacts with endocytic adaptor proteins [64], it is also possible that palmitoylation is somehow regulating this interaction. Palmitoylation may modulate protein-protein interactions by physically masking a binding site or by forcing a binding site into close membrane proximity, hence reducing availability for protein interaction. However, the effects of palmitoylation on protein interactions may also be more indirect, for example, by regulating the distribution of a protein between distinct membrane microdomains. For example, the enhanced ubiquitylation and internalization of palmitoylation-deficient TEM8 is linked to increased association with lipid raft microdomains [51], [65], bringing TEM8 into contact with the raft-localized ubiquitin ligase, Cbl. Precise membrane micro-localization of receptors is likely to have additional relevance for internalization, as specific endocytic retrieval pathways originate from defined regions of the plasma membrane [66]

The role of palmitoylation in regulating endocytic processes is not restricted to internalization, and the mucin-like MUC1 protein requires palmitoylation following internalization to facilitate traffic from recycling endosomes back to the plasma membrane [67]. Palmitoylation-deficient mutants of MUC1 in this case were found to exhibit a modest decrease in co-immunoprecipitation with the clathrin adaptor AP-1, which has been implicated in the recycling pathway from endosomes to the plasma membrane [68]. Furthermore, the internalized palmitoylation-deficient Luteinizing hormone receptor, which displayed a reduced rate of internalization, also exhibited a slower rate of recycling back to the plasma membrane [59].

Palmitoylation and protein traffic to cell compartments distinct from the secretory pathway

Certain members of the Stathmin family of microtubule regulatory proteins are palmitoylated at two cysteines within an N-terminal domain. Fusion of Stathmin family N-terminal domains to GFP led to a Golgi/vesicular distribution similar to the full-length proteins [69]. Interestingly though, mutation of the palmitoylation sites or inhibition of palmitoylation using 2-bromopalmitate led to a marked redistribution of the fusion proteins to mitochondria [69]. The relevance of these findings to the function of Stathmin proteins is unclear at this stage as mitochondrial localization was not observed following cysteine mutation or 2-bromopalmitate treatment of full-length proteins. Thus, it may be that for full-length Stathmins (de)palmitoylation co-operates with the dynamics of other Stathmin domains to regulate subcellular localization. Importantly, Drosophila wild-type Stathmin (full-length) revealed a marked mitochondrial distribution in biochemical fractions [69], supporting the notion that mitochondrial targeting is relevant to Stathmin function. Stathmins can regulate microtubule dynamics and thus one possible role of palmitoylation-dependent regulation of membrane localization of these proteins could be to facilitate microtubule interactions of several subcellular organelles [69]. These findings were not the first to suggest coupling of palmitoylation (or lack of) to mitochondria, as previous work had suggested that palmitoylated Ras may traffic via mitochondria en route to the plasma membrane [70]. A genetic screen for defects in Ras trafficking in yeast identified the class C VPS complex, which has established functions in endosomal/vacuolar trafficking and fusion [71]. Disruption of class C VPS proteins appeared to lead to accumulation of Ras2p on mitochondrial membranes, and a small fraction of endogenous Ras2p could also be detected on mitochondrial membranes of wild-type strains [70]. These results are consistent with the notion that palmitoylated Ras2p traffics via a novel mitochondria-dependent pathway in yeast. Together, these studies offer an intriguing insight into potential interplay between targeting of palmitoylated proteins to either the secretory pathway or mitochondria, and these areas of study warrant further investigation.

Palmitoylation/depalmitoylation was also proposed as a mechanism to regulate the localization of R7BP (RGS7 family binding protein) between the plasma membrane and nucleus [72]. R7BP was shown to be palmitoylated on 2 C-terminal cysteine residues, mutation of which resulted in a marked shift from the plasma membrane to nucleus. Interestingly, studies using 2-bromopalmitate suggested that palmitoylation of R7BP was dynamic, and inhibition of palmitoylation led to a shift of R7BP from the plasma membrane to nucleus. Importantly, the intracellular localization of R7BP determined the localization of R7 proteins and Gβ5, which failed to target to the plasma membrane in the absence of R7BP expression [72]. A C-terminal polybasic domain was found to play an important role in nuclear localization, and also to cooperate with palmitoylation for plasma membrane targeting [73], [74]. Thus, putative palmitoylation/depalmitoylation cycling of R7BP offers a system to regulate GPCR signalling, possibly facilitating transduction of plasma membrane signals from the plasma membrane to the nucleus [72]. Nuclear-plasma membrane shuttling of palmitoylated proteins may be fairly common [75–77], with palmitoylation/depalmitoylation as a central mechanism to regulate this differential targeting. It will be of significant interest to determine whether the palmitoylation status of R7BP is regulated by specific signals relevant to G protein signalling. In this regard, very little information relevant to the regulation of palmitoyl transferase or thioesterase activity/function is currently available, and understanding how palmitate turnover in proteins is regulated remains a central goal.

Recent work has also implicated perturbation of huntingtin (htt) palmitoylation as a contributing factor to the formation of cytoplasmic and nuclear inclusions that are a feature of Huntington's disease (HD). HD mutations involve the expansion of an unstable CAG (codon for glutamine) in the coding region of the N-terminus of htt leading to an expanded polyglutamine tract [78]. Expansion of the polyglutamine tract of htt inhibited palmitoylation of a proximal cysteine residue and interaction with the DHHC17/HIP14 palmitoyl transferase. Furthermore, mutation of the palmitoylation site in htt or downregulation of DHHC17 enhanced cytoplasmic and nuclear inclusions of htt proteins with both an expanded and a normal polyglutamine tract [79]. Thus, a single palmitate modification may play an important role in regulating the correct sorting (and aggregation) of the massive 350kDa htt protein.

Coupling of palmitoylation and ubiquitylation

The well-established link between palmitoylation and protein stability [21] is consistent with an interplay between palmitoylation and ubiquitylation as a signal for destruction. This idea is exemplified by palmitoylation of the yeast SNARE protein, Tlg1p, which prevents ubiquitylation and degradation of the protein via modulation of Tlg1p membrane topology [40]. Specifically, palmitoylation may regulate the orientation of the TMD and membrane proximal region of Tlg1p to avoid close contact of negatively-charged amino acids with the membrane surface [40], which appears to promote recognition by the ubiquitin ligase Tul1p. Palmitoylation of the anthrax toxin receptor, TEM8, also regulates ubiquitylation and subsequent endocytosis [51]. In this case however this appears to result from palmitoylation sequestering the receptor in membrane regions that are de-enriched in the ubiquitin ligase Cbl [51], an interesting example of how membrane micro-organization can regulate the accessibility of enzyme-substrate pairs.

Ubiquitylation is not always a signal for protein degradation however, and indeed ER-retained palmitoylation-deficient mutants of LRP6 were shown to be ubiquitylated but with no change in protein half-life [30]. In this case, ubiquitylation appeared to function as a specific ER retention signal for non-palmitoylated LRP6, as mutation of the ubiquitylated lysine residue rescued plasma membrane delivery of palmitoylation-deficient LRP6 [30]. Thus, palmitoylation of LRP6 is important (presumably by re-orientating the TMD in the plasma membrane) for avoidance of an ER-based ubiquitylation-retention system.

Membrane micro-localization of palmitoylated proteins

When cells or membranes are solubilized in non-ionic detergents (typically Triton X-100), a detergent-resistant membrane (DRM) fraction can be isolated, which is enriched in saturated phospholipids (in particular sphingomyelin), glycosphingolipids, cholesterol and specific proteins [80], [81]. These observations, coupled with studies showing phase separation of saturated phospholipids and cholesterol from unsaturated glycerophospholipids in model membrane systems [82], are consistent with the proposal that ‘raft’, microdomains enriched in saturated phospholipids and cholesterol, are present in cellular membranes [83]. Although the raft model of membrane organization has been extensively refined and tweaked over the past 10 years [84–88], the central idea holds that cell membranes, rather than being homogenous mixtures of protein and lipids, are more likely composed of a patchwork array of heterogenous microdomains. Phase separation of cholesterol and saturated phospholipids from unsaturated glycerophospholipids provides a simple concept for lipid microdomain formation, which undoubtedly occurs in protein-free lipid mixtures in vitro. Whether these lipid interactions are sufficient to promote domain formation in biologically complex cell membranes or whether the role of lipid interactions in the organising principles of cell membranes is secondary to protein-protein interactions is not clear.

Around half of the DRM protein pool in Madin-Darby canine kidney cells was estimated to be composed of palmitoylated proteins [89]. This finding is consistent with the notion that domains rich in cholesterol and saturated phospholipids form tightly-packed ordered domains that should readily accommodate saturated lipid groups such as palmitate. There is an extensive literature detailing the association of palmitoylated proteins with purified raft fractions (reviewed in [90], [91]) or with specific plasma membrane domains (for example see [92]), and the overwhelming rule is that palmitoylation serves as a strong signal for targeting to membrane rafts [90]. In some cases, palmitoylation alone may be sufficient for raft targeting or it may act synergistically with protein-protein interactions to facilitate raft association [93]. Palmitoylation-dependent association of proteins with raft microdomains can have important functions in intracellular signalling and membrane traffic pathways [90], [94]. However, palmitoylation is not a universal raft-targeting signal, and indeed, palmitoylation-deficient mutants of the anthrax receptor, TEM8, showed a dramatic enrichment in DRMs compared to wild-type palmitoylated protein [51]. For TEM8, it was proposed that palmitoylation may modulate membrane micro-localization by regulating a protein-protein interaction [51]. Based on the previous discussions of LRP6 palmitoylation and hydrophobic matching, another intriguing possibility is that palmitoylation may also tilt the TMD of TEM8, leading to a preference for thinner membrane regions, such as non-raft membranes. Thus, although palmitate may have a high affinity for raft domains, the effects of palmitoylation on protein conformation may have a dominant role in determining membrane micro-localization.

For other palmitoylated proteins, raft targeting is regulated by the extent of palmitoylation or by the position of the palmitoylated cysteine. DRM association of the membrane fusion proteins SNAP25 and SNAP23 correlates with the number of palmitoylation sites, with a 5-cysteine motif exhibiting a 3-fold greater enrichment in DRMs than a corresponding 4-cysteine motif [95]. H-ras has two palmitoylated cysteines, and although cys-181 is more important than cys-184 for plasma membrane localization, cys-184 is essential to determine the correct micro-localization of H-ras between cholesterol-sensitive and -insensitive domains at plasma membrane [96].

In addition to modulating protein micro-localization at the plasma membrane, palmitoylation may also regulate the distribution of proteins within intracellular membranes. For example, the yeast vacuolar fusion factor Vac8 is palmitoylated on up to three cysteine residues in an N-terminal SH4 domain, which also contains a myristoylation site. Palmitoylation cooperates with downstream region(s) of the protein to localize Vac8 to vacuole membranes [97], [98]. However, palmitoylation is required over and above simple vacuole targeting, as vacuole fusion was not efficiently rescued in Vac8 mutants expressing Vac8 fused to a myristoylation/polybasic sequence from Src, which displays efficient vacuole targeting [97]. Indeed, palmitoylation was suggested to mediate the association of Vac8 with specific subdomains of the vacuole membrane [99], possibly targeting Vac8 to ergosterol (yeast cholesterol) rich vacuole vertices where vacuole-vacuole fusion occurs [100], [101].

Role of membrane microdomains in sorting of palmitoylated proteins in the early secretory pathway

The differential affinity of palmitoylated and non-palmitoylated forms of a protein for specific membrane lipid compositions may be a major factor determining subsequent intracellular sorting. In a recent revised view of Golgi dynamics it was proposed that movement of proteins through the Golgi is regulated by a rapid exchange between processing domains and export domains [47]. Cholesterol and sphingolipid-rich regions of the Golgi membrane were proposed to represent export domains, in line with the increasing gradient of cholesterol/sphingolipids that occurs from the cis to the trans Golgi cisternae. The high affinity of palmitoylated proteins for cholesterol/sphingolipid-rich domains [89] may therefore enrich palmitoylated proteins in export domains, facilitating traffic through the Golgi. There is also some evidence that palmitoylation may allow direct binding of certain proteins to cholesterol [96], [102], which would similarly be predicted to facilitate forward transport.

The notion that palmitoylation may serve to couple proteins to specific Golgi export domains is supported by the observation that an ER-localized CSP mutant, which was palmitoylated by Golgi DHHC proteins following brefeldin A-induced mixing of ER and Golgi membranes, failed to exit the ER after brefeldin A washout [26]. Thus, in this case, for forward transport to occur, there may be a specific requirement for CSP palmitoylation to occur within the protein/lipid environment of intact Golgi membranes. These findings might be explained if palmitoylation is important to allow coupling of CSP to specific export domains in the Golgi [47].

One final point worth noting is that estimates of the thickness of cell membranes provide only an average for the whole membrane and do not take into account the presence of lipid microdomains that may have varying thickness [32]. Indeed, cholesterol has been proposed to increase membrane thickness [31], and palmitoylation could therefore stabilize newly-synthesized proteins at the ER or Golgi membrane by increasing affinity for cholesterol-enriched microdomains. Indeed, it is possible that LRP6 palmitoylation at the ER is also important to move the protein into thicker membrane regions and thus minimize hydrophobic mismatching [30].

Outlook

The versatility of S-palmitoylation as a tag mediating intracellular protein sorting, protein movement between membrane microdomains and protein turnover has been clearly emphasized over recent years. However, the effects of S-palmitoylation on membrane dynamics and protein stability are not easy to predict, and thus considerable effort will be required to gain a complete picture of the multitude of functions of palmitoylation within the cell. Of particular interest are the mechanisms that underlie the regulation of palmitoyl transferases and thioesterases, allowing cells to respond to specific stimuli by regulating the dynamics of S-palmitoylation. The recent identification of 23 mammalian DHHC palmitoyl transferases has re-invigorated the study of S-palmitoylation, and should facilitate a more rapid evaluation of the diverse array of mechanisms whereby palmitoylation dictates the targeting and ultimate fate of cellular proteins.

Acknowledgements

Work in the authors’ laboratory is funded by the Medical Research Council and the Wellcome Trust.

Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.