Protein acylation and localization in T cell signaling (Review)

Abstract

Many proteins with pivotal roles in T cell activation are modified by fatty acylation. Examples of these include transmembrane proteins such as the co-receptors CD4 and CD8, the adaptors LAT and Cbp/PAG, the pre-TCR as well as proteins synthesized on free cytosolic ribosomes, such as the Src-related tyrosine kinases Lck and Fyn. The two main types of fatty acylations in eukaryotic cells are N-myristoylation and S-acylation, the latter being more commonly referred to as palmitoylation. N-Myristoylation occurs exclusively on proteins synthesized on soluble ribosomes and provides substrates with an affinity for membranes. Palmitoylation modifies a wide range of substrates that includes both cytosolic and transmembrane proteins, its functions are diverse and in many cases not yet understood. Like myristoylation, palmitoylation promotes membrane-binding of cytosolic proteins, but it has also been implicated in protein targeting, trafficking, stability and activity. In addition, many palmitoylated proteins are insoluble in cold non-ionic detergent, and have therefore been proposed to localize to lipid rafts. The organization of receptors and signaling proteins into microdomains such as lipid rafts provides an attractive model for the initiation and propagation of T cell signaling, although many aspects of this are still poorly understood. This review will discuss the current evidence for the involvement of acylations in the localizations and functions of T cell signaling proteins.

• Myristoylation
• palmitoylation
• signaling
• T cells

N-Myristoylation and palmitoylation

N-Myristoylation and S-acylation both attach fatty acids to proteins, but in terms of chemistry, enzymology and substrates, these modifications could hardly be more divergent. N-myristoylation involves the cotranslational attachment of myristate, a 14-carbon saturated fatty acid, to an N-terminal glycine via an irreversible amide linkage [1]. This modification is catalysed by N-myristoyltransferases, two of which have been identified in humans (NMT1 and NMT2) [2]. Myristate is attached to glycines in the consensus sequence (M)GXXT/S, which requires the proteolytic removal of the N-terminal methionine. The crystal structure of NMT1 revealed that a myristoyl-CoA binding site restricts the length of the fatty acid that can be bound, and thus determines the preferential use of the short, relatively rare myristate [3].

Although myristoylation bestows an affinity for membranes upon substrates, the presence of myristate alone is not sufficient for stable membrane insertion. Instead, a second signal is required, which can either be a cluster of positively charged amino acids, palmitoylation, or protein-protein interactions [1], [4]. Thus, myristoylation is thought to target proteins to membranes where they are then trapped by a second signal. Myristoylation itself is not reversible, but dynamic regulation of membrane binding can be achieved through, for instance, phosphorylation, which prevents electrostatic interactions between positively charged amino acids and negatively charged phospholipids (myristoyl-electrostatic switch). In addition, for some proteins a myristoyl-conformation switch mechanism, which determines whether the myristate is exposed to the environment or buried within the protein, has been described [1], [4].

In contrast to myristoylation, S-acylation is a posttranslational and reversible modification that attaches fatty acids via a thioester bond to sulfhydryl groups of cysteines [1], [4], [5]. S-acylation was first demonstrated using tritiated palmitate [6], [7], a saturated 16-carbon fatty acid, and has therefore historically been referred to as palmitoylation. However, shorter (C14) and longer (up to C20) fatty acids can also be used, as well as mono- and poly-unsaturated ones. S-palmitoylation should be distinguished from N-palmitoylation, which attaches palmitate to the amide group of N-terminal cysteine residues. This modification has been described for secreted proteins such as Hedgehog and Wnt proteins, and is the topic of another review in this issue.

Two obstacles have stood in the way of rapid progress in the field S-acylation, which was first described almost twenty-five years ago: the absence of a well-defined consensus sequence and, until six years ago, the absence of known enzymes that catalyse this modification. Yet, novel tools have now been developed for large-scale identification of palmitoylated proteins, and a comprehensive picture of the palmitoyl-proteome appears within reach [8], [9]. Palmitoylated proteins known to date are a heterogeneous group that fall into one of four categories: cytosolic proteins with a myristate at the N-terminus; cytosolic proteins with a prenyl group at the C-terminus; cytosolic proteins without other lipid modifications; transmembrane proteins [1], [5], [10].

The purification of palmitoyltransferases (PATs) has proven to be extremely difficult, but genetic approaches in yeast have led to the cloning of two of these enzymes, Erf2p and Akr1p [11], [12]. These proteins have different substrate specificities, but share two features: they are multiple membrane spanning proteins and contain a DHHC motif in a cysteine-rich region (DHHC-CRD). The recognition of the common DHHC-CRD motif has allowed the identification of 23 PATs in mammalian genomes [13], [14]. The specificities of these enzymes, which are discussed in detail in another review in this issue, are still being worked out. The membrane localization of PATs suggests that substrates require some pre-existing membrane affinity. This fits well with the presence of additional lipid-modifications or transmembrane domains in most of these proteins.

Unique among lipid-modifications, is the reversibility of palmitoylation through the activity of palmitoyl-protein thioesterases. Evidence for palmitate cycling exists for numerous substrates. For Ras proteins, the constitutive cycling between depalmitoylation and repalmitoylation was shown to be important for localization and activity [15]. For other substrates such as Gα subunits and G protein coupled receptors, changes in the turnover rate of palmitate in response to specific signals are known to occur [5], [16]. This dynamic palmitoylation is discussed elsewhere in this issue.

Microdomains in T cell signaling

T cells are crucial for the adaptive immune response against pathogens. Activation of T cells is initiated when a pathogen-derived peptide associates with MHC proteins, and is recognized by the T cell receptor (TCR) [17]. This ligand recognition initiates a multitude of signaling pathways that eventually lead to the transcription of genes required for T cell proliferation and differentiation into effector cells (Figure 1).

Figure 1.  Early events in T cell activation. T cell activation is initiated by recognition of MHC/peptide complexes and leads to signaling as described in the main text. Light boxes in the CD3 and ζ chains are Immunoreceptor Tyrosine-based Activation Motifs where phosphorylation by Lck occurs. Thicker boxes in the plasma membrane represent lipid rafts.

The first detectable event following TCR engagement is the activation of the tyrosine kinase Lck [18]. Lck associates with CD4 and CD8, coreceptors that interact with non-polymorphic regions of MHC molecules in CD4+ and CD8+ T cells, respectively. The activation of Lck leads to the phosphorylation of tyrosines in the cytoplasmic domains of CD3 and TCRζ?chains. Phosphorylated TCRζ chains recruit the tyrosine kinase ZAP-70, which subsequently phosphorylates the transmembrane adaptor LAT. This leads to the assembly of a multiprotein complex that is essential for the propagation of signaling along several pathways, eventually resulting in the activation of transcription factors NFAT, AP-1 and NFκB.

T cell activation has been studied intensely, numerous signaling proteins have been identified and many pathways elucidated. Yet, challenges remain in understanding its initiation, extreme sensitivity and the ability to mount different responses against different ligands. One of the models that have been employed to explain these issues is the lipid raft hypothesis [19–23]. Lipid rafts are defined as membrane microdomains composed of highly-ordered saturated sphingolipids, intercalated with cholesterol, in an environment of less-ordered unsaturated glycerophospholipids [24], [25]. These domains can be isolated from artificial and biological membranes based on their relative insolubility in cold non-ionic detergents [26], [27]. In addition to specific lipids, some proteins are enriched in detergent insoluble fractions as well, including those with lipid anchors, such as GPI-linked proteins that insert into the extracellular leaflet of the plasma membrane and palmitoylated cytosolic proteins that insert into the inner leaflet (Figure 2). The attached saturated fatty acids of these proteins are thought to preferentially intercalate into the saturated fatty acids of lipid rafts. Transmembrane domains do not have a natural affinity for the highly ordered environment of lipid rafts, but some palmitoylated transmembrane proteins are detergent insoluble nonetheless. Palmitoylation is therefore frequently regarded as a raft-targeting signal. However, the use of detergent insolubility as a criterion for raft localization has come under attack because of its non-physiological nature, and this has opened the challenge to visualize plasma membrane microdomains in intact cells using microscopy-based methods. Results from several studies that apply FRET, single particle tracking, transmission electron microscopy or lipid-sensitive probes are consistent with the existence of microdomains in the plasma membrane, although conflicting reports exist as well. Thus, the debate on the nature and/or existence of these domains is still ongoing and a more extensive application of these imaging techniques will be required to settle this issue in the future. In the meantime, detergent insolubility should not be equated with lipid raft localization, although for historic as well as style reasons, the terms detergent resistant membranes (DRM) and rafts may occasionally be used interchangeably in this review.

Figure 2.  Lipid raft-associated proteins. Lipid rafts are characterized by the tightly packed saturated fatty acids with intercalated cholesterol. Three examples of raft-associated proteins are shown: GPI-linked proteins that associate via a lipid anchor with the extracellular membrane leaflet (A), acylated cytosolic proteins that associate via fatty acids with the intracellular leaflet (B), and transmembrane proteins that are acylated in their cytoplasmic domain (C).

The lipid raft model for T cell signaling proposes that the segregation of proteins into specific microdomains is crucial to prevent activation of signaling in resting cells, and to maximize propagation of signaling in activated cells [19–23]. According to this model, lipid rafts in resting T cells are homogeneously distributed and small, but will coalesce upon recognition of MHC/peptide antigens. This raft aggregation is thought to coincide with a re-shuffling of proteins into and out of rafts, recruiting those that promote signaling and excluding those that inhibit it. Two observations stand at the cradle of this model in T cells: (1) Crosslinking of GPI-linked proteins initiates T cell signaling by activating Lck; and (2) like GPI-linked proteins, Lck can be isolated from detergent insoluble membranes [28–30]. Further support for this model has come from the findings that T cell activation can be initiated by crosslinking of the raft lipid GM1 [31], and that the TCR and signaling molecules are recruited to DRM after T cell stimulation [32], [33]. Studies on palmitoylated proteins have further contributed to the shaping of this theory. For some of these, mutant forms deficient in palmitoylation are defective both in DRM localization and T cell signaling (see below). Therefore, a discussion on palmitoylated T cell proteins is inherently linked to the lipid raft theory.

Lck and Fyn

Lck is one of nine Src family proteins that are each composed of a unique domain, followed by SH3, SH2, tyrosine kinase domains and a regulatory tail. Src kinases do not share sequence similarities at the unique domain except for a short region at the very N-terminus, also referred to as the SH4 domain, where acylations occur. All Src kinases are myristoylated and, with the exception of Src and Blk, palmitoylated as well [34].

Lck is myristoylated at Gly2, and palmitoylated at Cys3 and 5 [35], [36]. Mutagenesis of either Gly2 or Ser6 in the consensus sequence (M)GXXT/S abrogates myristoylation of Lck, and as a consequence palmitoylation and membrane-association [35], [37], [38]. However, myristoylation alone is not sufficient for stable membrane-binding: Lck mutants with both palmitoylation sites deleted are myristoylated but do not associate with membranes [37], [39–41]. Hence, myristoylation most likely targets Lck to membranes where PATs are located, and subsequent palmitoylation leads to stable membrane-insertion. Myristoylation has also been shown to be important for the stability of Lck, since treatment of cells with the myristoylation inhibitor 2-hydroxymyristate resulted in a reduced half-life of the protein [42]. Substitution of either Cys3 or Cys5 by mutagenesis does not abolish palmitoylation or membrane-binding, indicating that both sites can be used [37], [38], [40], [41], [43–45]. However, in most studies the absence of Cys5 leads to only a minimal reduction in membrane binding, whereas deletion of Cys3 reduces this by approximately half. Thus, both single and dually palmitoylated Lck are likely to co-exist.

The essential role of Lck in the development and activation of T cells is well established [46]. An absence of Lck in knockout mice leads to a severe reduction in peripheral T cells, and mutant T cell lines that lack Lck are unable to respond to TCR stimulation. What are the roles of acylation in this function? To begin with, the activity of Lck is absolutely dependent on an ability to associate with membranes. Lck mutants with amino acid substitutions at either myristoylation or palmitoylation sites, are unable to reconstitute T cell activation in Lck-deficient cell lines [38], [40], [43], [44]. These soluble proteins do not interact with CD4, phosphorylate TCRζ, recruit ZAP-70 to membranes, or promote downstream signaling. Palmitoylation also appears to play important roles beyond membrane-binding. For instance, an Lck chimera that localizes to the plasma membrane through a transmembrane domain instead of palmitoylation is unable to efficiently reconstitute TCR signaling [44]. Similarly, Lck synthesized in the presence of the palmitate analogue 13-oxy-palmitic acid, cannot induce TCR signaling despite being associated with membranes and CD4 [47]. Oxy-palmitate is likely to be incorporated in place of palmitate since its presence decreases labeling of Lck with iodinated palmitate. This second function of palmitoylation has been postulated to represent a role in targeting Lck to lipid rafts. Indeed, a considerable fraction of Lck is normally insoluble in cold non-ionic detergent [41], [44], [47], but this is not the case for the abovementioned transmembrane Lck chimera or Lck modified with oxy-palmitate. The importance of palmitoylation for detergent insolubility of Lck has further been demonstrated using engineered proteins that contain a six-lysine motif instead of palmitoylation sites [37]. Such proteins can associate with membranes but do not interact with GPI-linked proteins in DRM. In addition, it has been shown that the first 10 amino acids of Lck can target otherwise cytosolic proteins to detergent insoluble membranes [41]. In the absence of detergents, a palmitoylation-dependent localization to specific microdomains was observed by transmission electron microscopy of plasma membrane sheets: a protein composed of the 10 N-terminal Lck residues linked to five lysines and a myc-tag localized to different areas of the membrane than a non-palmitoylated counterpart [48].

Fyn is another Src kinase with roles in T cell activation although its function is not yet completely understood. Experiments on mice and cell lines deficient for Lck have shown that Fyn can replace Lck activity to some extent but cannot sustain full T cell development and activation [49], [50]. Under wildtype conditions, Fyn activation appears to be dependent on and follow that of Lck. In contrast to Lck, Fyn may be required for dampening T cell responses after activation, and play a role in T cell anergy [51], [52]. Like Lck, Fyn is both myristoylated and palmitoylated. Fyn does not contain a myristoylation consensus sequence with a Ser or Thr at position 6, but instead Lys 7 and 9 are important for efficient myristoylation [53]. Myristoylation of Fyn is required for palmitoylation [54], but in contrast to Lck, non-acylated Fyn associates to some extent (∼ 25%) with membranes [55], [56], possibly because N-terminal positively charged amino acids contribute to membrane-binding. This may also explain the faster membrane-binding kinetics of newly synthesized Fyn compared to those of Lck [56], [57]. Although based on mutagenesis studies both Cys3 and Cys6 can be palmitoylated [58], mass spectrometry showed that only a minor fraction of Fyn is dually S-palmitoylated and that the majority is palmitoylated on Cys3 only [59]. This further revealed that in addition to palmitate, mono-unsaturated palmitoyl as well as stearate and oleate are attached by S-acylation. The finding that Cys3 is the main site of Fyn palmitoylation fits well with the importance of this site, in contrast to Cys6, for membrane binding, detergent insolubility, association with GPI-linked proteins and interaction with TCRζ? [45], [53], [60].

While it is obvious that palmitoylation of Lck and Fyn is important for detergent insolubility as well as for function, precisely how targeting to detergent insoluble membranes contributes to the activity of these proteins is not yet clear [19–23]. For instance, it has been reported that in resting murine T cells the majority of Lck is excluded from DRM, whereas Fyn is predominantly present in these domains, leading to the postulate that rafts function to segregate Lck and Fyn in the absence of a T cell stimulatory signal. Lck was demonstrated to have higher activity outside of DRM and to translocate to these domains upon T cell activation, where it then activates Fyn [61]. In this scenario, the activation of Lck may result from its dephosphorylation at the regulatory tail tyrosine by CD45, a phosphatase that is excluded from detergent insoluble domains. Conversely, the tyrosine kinase Csk is to some extent recruited to DRM following phosphorylation of the adaptor Cbp/PAG by Fyn, where it can inhibit Lck by phosphorylating the tail tyrosine [62], [63]. This negative feedback loop fits well with the transient activation of Lck detected with activation-specific antibodies [61], [64]. However, several issues require further clarification. For instance, the importance of Cbp for Csk activity has been disputed (see below), and the role of CD45 is ambiguous since it can also inactivate Lck by dephosphorylating a tyrosine in the kinase domain. Moreover, the very low amount of detergent insoluble Lck is not always observed in resting T cells, although the use of different types and percentages of detergents impedes a direct comparison between different studies.

Other studies have addressed the significance of an association with DRM by following the movements of Lck (and other T cell proteins) upon T cell activation. Several of these analyses used a fusion protein composed of the Lck acylation sites (aa 1–10) coupled to GFP (Lck10-GFP), alongside of Lck. In one report a CD28 costimulation-dependent accumulation of both Lck and Lck10-GFP at the T cell/APC interface was demonstrated by microscopy, which occurred concomitant with an increase in detergent insolubility of Lck [65]. The similar behaviour of Lck10-GFP and Lck points towards a lipid-based rather than protein-based recruitment of this palmitoylated protein to sites of T cell activation. Similarly, following the trajectory of single molecules by TIRF microscopy, the movements of both Lck-GFP and Lck10-GFP molecules were shown to slow down at sites of T cell stimulation [66]. This change in movement has been proposed to result from a trapping of rafts in the reorganized, denser actin cytoskeleton at TCR activation sites. A rearrangement of plasma membrane lipids at T cell activation sites has also been established with the fluorescence membrane probe Laurdan [67], the emission spectrum of which is dependent on the order and condensation of surrounding lipids. These studies provide important evidence for the occurrence of plasma membrane reorganizations at T cell activation sites, yet, what drives the accumulation of proteins and lipids at these sites is not completely clear. Whereas the data on Lck and Lck10-GFP above suggest that lipid-lipid interactions are important for this, other results indicate a predominant role for protein-protein interactions. For instance, Lck was shown to reside in TCR activation-induced CD2 signaling clusters in another TIRF study, whereas Lck10-GFP did not [68]. The formation of CD2 signaling clusters was found to be dependent on protein interactions via LAT tyrosines. Furthermore, in a biochemical analysis, plasma membrane isolates that represent TCR stimulation sites showed an enrichment of LAT, TCRζ and downstream signaling proteins, but not of Lck [69]. This again demonstrates a role for protein rather than lipid interactions in the stability of TCR assemblies. On the other hand, the activity of Lck was required for the formation of these TCR signaling complexes. A current model may therefore be as follows: Lck raft localization is important for the initial signaling that drives subsequent cytoskeletal changes, LAT activation, membrane condensation and the formation of stable signaling complexes at TCR activation sites. However, obtaining a clearer picture of this remains a challenge for the future.

CD4 and CD8

The TCR coreceptors CD4 and CD8 facilitate T cell activation in two ways: they bind to invariant regions on MHC proteins, which stabilizes interactions with antigen presenting cells, and they associate with Lck, which couples the TCR to downstream signaling pathways. Despite having similar functions, the composition and structure of CD4 and CD8 are quite distinct. CD4 is a monomer with an extracellular domain composed of four immunoglobulin-like domains, whereas CD8 is present on most T cells as a disulfide-linked heterodimer of CD8α and CD8β chains, each containing one extracellular immunoglobulin-like domain [70]. Nevertheless, at the cytoplasmic domains both proteins contain palmitoylation sites.

CD4 is palmitoylated at two cysteines located at the junction of the cytoplasmic and transmembrane domains, but this modification does not affect transport, cell surface expression or association with Lck [71]. Since CD4 was observed to be detergent-insoluble, further studies on palmitoylation have focused primarily on a role in lipid raft targeting, and the involvement of this in T cell signaling and HIV entry. An influence of CD4 palmitoylation on T cell signaling has been suggested in studies using Jurkat T cells transfected with either wildtype or palmitoylation-deficient CD4. Reduced levels of tyrosine phosphorylation induced by CD4 crosslinking were observed for the mutant protein, although differences with wildtype CD4 were not very pronounced [72]. In primary murine T cells, palmitoylation of CD4 was shown to be required for the recruitment of TCR and PKCθ to the immunological synapse [73], but the significance of this for T cell signaling was not further established. Thus, at present there is only scarce knowledge on the effect of CD4 palmitoylation on T cell signaling.

Moreover, the importance of palmitoylation for DRM localization of CD4 is not completely clear since contradictory reports on this exist. Two studies demonstrated that the absence of CD4 palmitoylation, either through deletion of palmitoylation sites or by growing cells in the presence of the inhibitor bromo-palmitate, led to a reduction in detergent insolubility [72], [74]. However, a complete absence of detergent insolubility was observed only when both palmitoylation and Lck association did not occur. In contrast, other studies failed to detect any role for palmitoylation in the detergent insolubility of CD4 [75], [76]. Instead, a novel raft-targeting signal that consists of the positively charged residues RHRRR was identified [75]. Mutations in this sequence located adjacent to the palmitoylation sites, abolished DRM localization without interfering with palmitoylation. This suggests that the RHRRR sequence is a dominant signal for raft targeting of CD4, and that although a contribution of palmitoylation cannot be excluded, it does not appear to be essential. In the absence of a role in raft localization, it is not clear how CD4 palmitoylation contributes to T cell signaling.

Conflicting reports also exist regarding the importance of lipid rafts in HIV entry, nonetheless, it has been consistently shown that CD4 palmitoylation is not required for this [74–76]. Raft localization of CD4 was also suggested to contribute to the downmodulation of CD4 by HIV Nef, but here again palmitoylation of CD4 does not appear to play a role [77]. Thus, clear consequences of CD4 palmitoylation have remained elusive so far.

In the case of CD8, palmitoylation was first demonstrated for the mouse CD8αβ heterodimer and occurs here on a single cysteine in the cytoplasmic domain of the CD8β chain [78]. The deletion of this palmitoylation site reduced detergent insolubility of CD8αβ. Absence of CD8β palmitoylation also resulted in a reduction of Lck activation, Ca2+ induction and recruitment of the TCR to DRM following T cell stimulation [79]. Thus, for mouse CD8αβ, palmitoylation appears to contribute to its function in T cell signaling. Interestingly, palmitoylation and detergent insolubility do not occur for the mouse CD8αα homodimer [78], which is expressed on a small subset of T cells as well as on NK and dendritic cells. Despite the fact that CD8α associates with Lck, the CD8αα homodimer is not an efficient TCR coreceptor [80], and this difference in function between CD8αα and CD8αβ has now been attributed to the absence of raft localization of CD8αα.

The role of CD8 palmitoylation is less straightforward when results on human CD8 are considered. Similar to the situation in the mouse, human CD8αβ localizes to DRM, whereas CD8αα does not [81]. Although this points towards a conserved function of these localizations, the involvement of palmitoylation is strikingly different between human and mouse CD8. Human CD8β contains not one, but two palmitoylation sites, but more surprisingly, CD8α is palmitoylated as well. Thus, the human CD8αα homodimer is not detergent insoluble despite being palmitoylated. Also unexpectedly, the CD8αβ heterodimer localizes to DRM in the absence of palmitoylation: deletion of palmitoylation sites in both CD8β and CD8α does not prevent detergent insolubility of CD8αβ. Instead, a cluster of positively charged amino acids functions as a raft-targeting signal for CD8β expressed on its own. However, this motif is not needed for CD8αβ heterodimers, which solely require the pairing of CD8α and CD8β extracellular domains for detergent insolubility. Hence, human CD8αβ has an intrinsic ability to localize to DRM, but palmitoylation does not appear to play a role in this. A role for palmitoylation in protein transport was also not detected, since non-palmitoylated CD8αα and CD8αβ are efficiently expressed at the cell surface [81]. Whether palmitoylation is important for the coreceptor function of human CD8αβ, has not yet been assessed.

Overall, it can be concluded that, whereas there is some evidence that palmitoylation contributes to CD4 and CD8 function, the mechanism for this is not understood. Furthermore, it is obvious that a simple relationship between palmitoylation and DRM localization does not exist. This is also clear from previous data on the Transferrin Receptor that similarly, is not detergent insoluble despite being palmitoylated [6].

The adaptors LAT and PAG/Cbp

The adaptor LAT is primarily expressed in T cells and is essential for the development and activation of these cells [82]. LAT is a transmembrane protein with an extracellular region of only a few amino acids and a large cytoplasmic domain. Tyrosines in the cytoplasmic domain provide docking sites, when phosphorylated, for Grb2, PLC-γ1, SLP-76 and PI3K. At the junction of the transmembrane and cytoplasmic domain, Cys26 and Cys29 are palmitoylated with Cys26 being the most prominent site [83]. The use of mutant proteins with deletions at these sites has shown that palmitoylation of LAT is essential for optimal recruitment and/or phosphorylation of Grb2, PLC-γ?, Vav and SLP-76 [84], [85]. In the absence of LAT palmitoylation, downstream events such as Ca2+ fluxes, Erk activation and the induction of transcription factors AP-1 and NFAT cannot occur.

The involvement of palmitoylation in LAT function has been attributed to a role in lipid raft targeting. In agreement with this, a large proportion of LAT localizes to detergent insoluble membranes in a palmitoylation-dependent manner [69], [83–85]. Furthermore, poly-unsaturated fatty acids (PUFAs) displace both Lck and LAT from DRM, but negative effects on T cell activation could be prevented by targeting LAT to these domains in a PUFA resistant manner [86]. Palmitoylation of LAT was further demonstrated to be important for the formation of TCR signaling assemblies: plasma membrane fractions isolated on anti-CD3 beads (also described with respect to Lck above) contained TCRζ??, Grb2 and PLCγ? only when LAT palmitoylation sites were intact [69]. On the other hand, the diffusion patterns of palmitoylation-deficient and wildtype LAT clusters were similar with respect to TCR activation-induced CD2 signaling clusters [48], arguing against a role for palmitoylation in the movement to and trapping at sites of T cell activation. The significance of LAT raft localization has also been challenged by the forced localization of LAT to non-raft domains, which does not affect its function [87]. A chimera composed of the cytoplasmic region of LAT and the transmembrane and extracellular domains of the adaptor LAX is excluded from DRM, but can reconstitute T cell activation in LAT-deficient T cells. Moreover, this LAX/LAT chimera can rescue T cell development in LAT-deficient mice, and permits the generation of fully functional peripheral T cells. Thus, in the context of LAX/LAT, palmitoylation is not essential for LAT function. Based on the difference in detergent insolubility between LAT and LAX/LAT, these data also argue against a role for rafts in LAT function, although it remains possible that these proteins target to the same membrane subdomains. LAT palmitoylation may serve functions distinct from raft targeting. For instance, it has been demonstrated that non-palmitoylated LAT mislocalizes to some extent to the Golgi region and has a shorter half-life than the wildtype protein, indicating that palmitoylation ensures proper plasma membrane localization and protein stability [88]. Other data are consistent with such a role as well. Peptides composed of the LAT extracellular and transmembrane regions only, do not associate with membranes in the absence of palmitoylation [89]. Lower protein levels compared to the wildtype protein were also shown for palmitoylation-deficient LAT in transfected Jurkat cells [69] and for LAT/GFP fusion proteins [48]. Therefore, defects of T cell activation in the absence of LAT palmitoylation may at least to some extent arise from inefficient expression of LAT at the plasma membrane.

Interestingly, differences in LAT palmitoylation between activated and anergic T cells have been reported. The induction of T cell anergy was shown to coincide with a reduction in LAT palmitoylation and a decrease of LAT levels at the immunological synapse [90]. This suggests that dynamic palmitoylation of LAT occurs and that the regulation of LAT palmitoylation contributes to T cell anergy. A pool of non-palmitoylated LAT exists in T cells, and the size of this may be controlled by the activation state of these cells. The palmitoylation of Fyn was not affected in anergic T cells, which indicates that perhaps the activity of specific palmitoyl transferases are regulated in T cells.

The adaptor Cbp/PAG shares with LAT some degree of sequence homology, has a comparable domain organization and contains two palmitoylation sites at similar positions as LAT [62], [63]. Cbp/PAG is ubiquitously expressed but given its almost exclusive detergent insolubility, an important role in T cell signaling has been postulated. The main function of Cbp/PAG was proposed to be the recruitment of Csk to membranes, resulting in the inactivation of Src kinases. Consistent with a role in T cells, Cbp/PAG is constitutively phosphorylated in resting T cells and dephosphorylated upon T cell activation. Furthermore, overexpression of Cbp/PAG inhibits T cell activation to some extent, which was not observed with Cbp/PAG mutants unable to associate with Csk [63], [91]. Palmitoylation of Cbp/PAG has not been studied extensively, but one recent study reported that palmitoylation-deficient Cbp/PAG is not detergent insoluble, and does not inhibit T cell activation when overexpressed [92]. Whereas these studies point towards a role for Cbp/PAG and its palmitoylation in controlling the activity of Csk in T cells, the generation of Cbp-/- knockout mice has raised strong doubts about this. In contrast to Csk-/- mice that are embryonically lethal, Cbp -/- mice are viable and healthy [93]. Moreover, T cell development and function was not compromised in Cbp-/- mice either [94]. In this light, the roles of Cbp palmitoylation and raft localization also remain unresolved.

Pre-TCR signaling

αβT cells undergo a highly regulated developmental program during maturation in the thymus, which ensures that only those cells that are able to recognize MHC proteins but do not respond to self-peptides, will leave the thymus and make up the T cell repertoire in the blood. One off the first hallmarks of T cell development is the completion of somatic gene recombination to generate a specific TCRβ chain. This TCRβ chain then pairs with a surrogate TCRα chain, pTα, and forms the pre-TCR receptor [95]. Signaling from the preTCR leads to several rounds of cell division, followed by the rearrangement of the TCRα gene, and the subsequent expression of the TCRαβ receptor. Lck and LAT, amongst other proteins, are essential for pre-TCR signaling, but intriguingly, a ligand does not appear to be involved [96]. Instead, it has been hypothesized that constitutive targeting of the pre-TCR to lipid rafts initiates signaling [97]. Raft localization of the pre-TCR has been suggested to depend on pTα palmitoylation. This hypothesis is based on the observations that in contrast to TCRαβ, the preTCR colocalizes with Lck in detergent insoluble fractions and in co-capping experiments. Moreover, the cytoplasmic domain of pTα, but not that of TCRα or TCRβ, is modified by palmitoylation. Furthermore, in an artificial model system using crosslinking of CD3 with cell surface calnexin, raft localization and palmitoylation of calnexin were shown to be required for T cell development [98]. Nevertheless, a convincing role for palmitoylation of pTα has not been demonstrated, as a mutation at the pTα palmitoylation site did not compromise T cell development [99]. In contrast, the deletion of a PPSRK sequence located close to the palmitoylation site did have some effect, but whether this motif is involved in raft localization has not been tested. Again, the role of pTα palmitoylation is not yet clear, but does not appear to correlate with lipid raft localization.

Table I.  Summary of the roles of palmitoylation related to raft localization for the proteins discussed in this review.

	Lck	Fyn	CD4	mCD8	hCD8	LAT	Cbp/ PAG	pTα
Acylation	myr/palm	myr/palm	Palm	palm	palm	palm	palm	palm
Palm. required for function?	+	+	+	+	+	+	+	-
Raft localization?	+	+	+	+	+	+	+	+
Palm. required for raft localization?	+	+	–	+	–	+	+	?
Raft local. required for function?	+	+	?	+	?	–	+	?
Alternative raft signals?	–	–	+	–	+	–	–	–

?: not yet studied.

Summary and conclusions

The role of acylation, in particular palmitoylation, in the function of T cell signaling proteins has been studied primarily in the context of lipid rafts over the past 10 years (summarized in Table I). Palmitoylation indeed has clear effects on detergent insolubility and function of Lck, Fyn and LAT, but the role of DRM localization is not yet understood. Palmitoylation also affects functions of CD4 and CD8, but here it does not appear to be essential for detergent insolubility. To obtain further insights into palmitoylation, it seems appropriate and timely to analyze this modification for T cell proteins from non-raft perspectives as well. Many areas have remained largely unexplored. For instance, it is not known whether palmitate cycling of T cell proteins is regulated, and/or whether this is influenced by T cell activation. Further, little is known about the effects of palmitoylation on subcellular localizations, in specific on the trafficking of proteins between different membrane compartments, or the endocytosis of transmembrane proteins. Additionally, effects of palmitoylation on lateral mobility of proteins in cell membranes can be further explored, as well as a possible interplay between palmitoylation and other protein modifications such as phosphorylation and ubiquitination. Lastly, the identification and regulation of DHHC proteins in T cells may yield important insights. Thus, a less raft-biased approach to palmitoylation may generate novel knowledge on this modification in general and its role in T cell signaling specifically. Declaration of interest: The author reports no conflicts of interest. The author alone is responsible for the content and writing of the paper.