The role of palmitoylation and transmembrane domain in sorting of transmembrane adaptor proteins

Integral membrane proteins of eukaryotic cells comprise almost 30% of all proteins encoded by the human genome (Almén et al., 2009). These undergo sorting into target compartments such as the endoplasmic reticulum (ER), Golgi complex, mitochondria or the plasma membrane to perform their function. Mis-localisation can lead to a loss-of-function of these proteins, resulting in cell malfunction and even development of diseases (Howell et al., 2006). Structural and sequence motifs responsible for protein sorting events have been studied for more than 30 years. During that time a number of determinants of specific membrane protein localisation have been identified. These include sequence motifs in cytosolic and lumenal domains (e.g. KDEL for the ER or YxxØ for clathrin-dependent processes; Traub, 2009), N- and O-glycosylation of the extracellular domains (Fiedler and Simons, 1995; Potter et al., 2006; Proszynski et al., 2004), and physical properties of the transmembrane domain (TMD) such as length and hydrophobicity (Cosson et al., 2013; Sharpe et al., 2010). This wide variety of sorting signals is further extended by the fact that more than one determinant can define the localisation of proteins in the cell (Alonso et al., 1997; Duffield et al., 2008). No universal signal or motif has been demonstrated for the sorting of proteins to the plasma membrane. Therefore, all parts of the protein need to be tested for their impact on the proper localisation when investigating protein function using various mutants.

Transmembrane adaptor proteins (TRAPs) are topologically related type I integral membrane proteins that enable the association of signalling effectors and other enzymes with the plasma membrane of eukaryotic cells. They contain only a short extracellular part, single TMD and the cytosolic domain with multiple protein-protein interaction motifs mediating their functions (Fig. 1A). TRAPs do not have any enzymatic activity. In mammalian white blood cells, they function as crucial check-points or regulators of the main signalling events defining the basic function of these cells by facilitating the assembly of signalling complexes at the plasma membrane (Hořejší et al., 2004). Three TRAP subfamilies were defined in these cells: (i) monomeric, palmitoylated TRAPs (pTRAPs), (ii) monomeric, non-palmitoylated TRAPs and (iii) dimeric TRAPs frequently associated with large multisubunit receptor complexes (Stepanek et al., 2014). The latter group represents a challenge when studying the impact of various motifs on protein localisation because of their dimeric character and strong association with other subunits of receptor complexes. For example, CD247 dimer (ζ chain) facilitates the surface expression of the multi-subunit TCR/CD3 complex in T lymphocytes (Weissman et al., 1989). Non-palmitoylated, monomeric TRAPs (LAX1 and SIT1) have an extracellular part of intermediate size (38–40 residues) with potential glycosylation sites. Their glycosylation status can influence the plasma membrane localisation as shown for a number of surface glycoproteins (Fiedler and Simons, 1995). On the contrary, none of the known pTRAPs [LAT, PAG1, NTAL (also known as LAT2), LIME1, PRR7, SCIMP] are likely to be glycosylated because of their very short extracellular part (3–20 amino acids) and lack of glycosylation sites. In addition, pTRAPs are missing a signal peptide and use their unique TMD for membrane insertion, similar to type II membrane proteins. It can be hypothesised that it is the TMD length or composition that defines the localisation of such proteins to the plasma membrane, as observed for type II proteins in the past (Munro, 1991).

Palmitoylation, a reversible post-translational modification of eukaryotic proteins (Bijlmakers, 2009), is another factor that can contribute to plasma membrane targeting of pTRAPs. In the process of palmitoylation, palmitoyl-CoA produced by fatty acid metabolism at the ER membrane is enzymatically reacted with cysteine residues in the cytoplasmic parts of target proteins by the DHHC family of proteins (Zeidman et al., 2009). No consensus palmitoylation motif has been found to date (Zeidman et al., 2009). In all pTRAPs, the palmitoylation sites are thought to be the cytosolic cysteine residues within CxxC or CxC motifs adjacent to the transmembrane domain, although for some family members it has not yet been proven experimentally. Palmitoylation stabilises the plasma membrane localisation of a number of myristoylated or farnesylated peripheral membrane proteins such as Ras or Src family proteins (Rocks et al., 2010). The importance of palmitoylation for plasma membrane localisation of LAT, a member of the pTRAP subfamily, has also been reported (Hundt et al., 2009). Palmitoylation is also believed to target some pTRAPs to putative sphingolipid- and cholesterol-enriched membrane microdomains called lipid rafts (Hořejší et al., 2010; Levental et al., 2010). Levental and co-workers recently combined these two concepts and suggested that the capacity to associate with lipid rafts is a determinant of protein plasma membrane localisation (Diaz-Rohrer et al., 2014). In this study, we were interested in the impact of palmitoylation and TMD sequence (including adjacent sequences) on the plasma membrane localisation of pTRAP family proteins. Using a panel of mutant proteins and live cell imaging we reveal a complexity of determinants for plasma membrane targeting, specific for each individual membrane protein. Our data confirm the dominant impact of TMD length and/or hydrophobicity on plasma membrane localisation of proteins, but we also demonstrate that secondary sorting determinants such as palmitoylation, flanking sequences and the presence of the extracellular domain also contribute to plasma membrane localisation of proteins with suboptimal TMD length. We provide evidence that the intracellular domain (ER exit motifs) facilitates the flow of proteins towards the plasma membrane. The data indicate that more than one sorting determinant defines the dynamic localisation of proteins in cells.

The absence of a glycosylated extracellular domain suggests that TMD length and composition and/or other factors such as palmitoylation determine localisation of pTRAPs. The effect of TMD length and composition has recently been analysed by Munro and colleagues (Sharpe et al., 2010). They used a bioinformatics approach and found a strong correlation between the length and hydrophobicity of the TMD and the protein localisation. According to their results, TMDs of 22 residues and longer preferentially localise to the plasma membrane. In addition, specific distribution of side chain size and hydrophobicity throughout the TMD also correlated with protein subcellular distribution. Based on these data they created an algorithm (available online at www.tmdsonline.org as ‘TMD organelle predictor’) which predicts the localisation of membrane proteins in fungi and vertebrates to various cellular membranes (ER, Golgi complex and plasma membrane) based purely on the amino acid sequence. The algorithm has an overall success rate of 82% for prediction of plasma membrane localisation. However, analysis of all six human monomeric pTRAPs using this algorithm predicted the localisation of only three proteins to the plasma membrane, and localisation of the remaining three to the ER (Table 1). Five members of the pTRAP family have TMD lengths of 22–23 residues (Table 1) indicating that these proteins should be efficiently sorted to the plasma membrane. NTAL possesses a shorter TMD of 19 residues which falls outside of this range and was predicted to reside in the ER. LIME and PRR7, both with long TMDs, are also predicted to localise to the ER. As available data show that all these transmembrane adaptors are localised at the surface of human cells (Stepanek et al., 2014), we concluded that other determinants, not considered by the prediction algorithm, might contribute to the plasma membrane localisation of pTRAPs. It is worth noting that the bioinformatics tool ‘TMD organelle predictor’ was not designed for palmitoylated proteins.

In order to better characterise what determines plasma membrane localisation of these adaptors we selected three representative members of the pTRAP protein family: LAT, PAG and NTAL (Stepanek et al., 2014). To visualise sorting of these pTRAPs, green fluorescent protein (GFP) fusion variants (Fig 1A,B; Fig 2A; Fig. S1A) were expressed in a human T cell line (Jurkat, native environment) and cells of epithelial origin (HeLa, non-native environment) and imaged using live cell confocal microscopy. First, we verified the expression of wild-type variants of these fusion proteins and found that LAT, PAG and NTAL all almost exclusively localised in the plasma membrane of Jurkat (Fig 1C, Fig 2B; Fig. S1B) and HeLa cells (data not shown). Palmitoylation of pTRAPs is essential for their function (Stepanek et al., 2014) and is mainly thought to be a targeting signal for lipid rafts (Levental et al., 2010). Immunofluorescence of a non-palmitoylatable mutant of LAT revealed its mislocalisation to the intracellular membranes of human T cells (Hundt et al., 2009). We were, therefore, interested whether palmitoylation controls plasma membrane localisation of other pTRAPs, PAG and NTAL. pTRAPs are palmitoylated at the membrane proximal cysteines of the cytosolic tail (Fig. 1A). Transient expression of GFP fusion proteins with these cysteines mutated to serines (CS variants) demonstrated hindered plasma membrane localisation for non-palmitoylated LAT-CS (Fig. 1C,D, Table 3). Conversely, only a weak effect of Cys→Ser mutation was observed for PAG and NTAL proteins (Fig. 2B,C; Fig. S1B). Retention of LAT-CS in the Golgi complex was shown not to result from reduced kinetics of sorting; no plasma membrane localisation was observed in cells with blocked protein synthesis for 4–6 h (25 µg/ml cycloheximide; data not shown). Rather, LAT-CS mislocalisation led to its degradation. The surface expression of non-mutated LAT was unaffected in cyclohexamide-treated cells.

Changes in structure can cause re-orientation of proteins in membranes. Indeed, bioinformatics analysis (TMHMM 2.0; Krogh et al., 2001) suggests that non-palmitoylated LAT-CS is a type II membrane protein (cytosolic N-terminus). Native LAT protein is a type I protein (extracellular N-terminus). We, therefore, tested the orientation of LAT and LAT-CS proteins using a glycosylation assay (see Fig. S1C for more details) (van Geest and Lolkema, 2000). No difference in orientation between LAT and LAT-CS was observed (Fig. S1C).

We were further interested in whether TMD, together with palmitoylation, are sufficient for plasma membrane localisation and generated LAT and PAG mutants missing the intracellular domain (LAT: Δ34−262 residues; PAG: Δ44–432 residues). Both palmitoylated and non-palmitoylated versions of short variants were tested. Fig. 1C,D shows poor plasma membrane localisation of palmitoylated short LAT variant and retention of its non-palmitoylated counterpart in the ER. Similarly to LAT, removal of the intracellular domain of PAG reduced sorting to the plasma membrane. By contrast, its palmitoylated and non-palmitoylated variants exhibited comparable distribution between the ER and plasma membrane, confirming that palmitoylation does not influence sorting of PAG to the plasma membrane (Fig. 2B). Co-localisation analysis with markers of the Golgi complex (Golgi7) and ER (Sec61) demonstrated that non-palmitoylated LAT remained localised to the Golgi complex (Fig. 1E) whereas non-palmitoylated LAT missing the intracellular domain was retained in the ER (Fig. 1F). This was supported by quantitative analysis of plasma membrane localisation efficiency. Plasma membrane localisation efficiency of non-palmitoylated LAT lacking the intracellular domain was comparable to the negative control represented by ER marker Sec61 (Fig. 1D). In summary, we demonstrate here that palmitoylation is essential for plasma membrane localisation of LAT but not for PAG and NTAL and that cytoplasmic domain of LAT and PAG also plays an important role in this process.

The glycosylated extracellular domain of some proteins has been demonstrated to provide sufficient plasma membrane localisation signal (Potter et al., 2006; Proszynski et al., 2004). We were, therefore, interested in whether the lack of a large glycosylated extracellular domain in pTRAPs could be the reason for their increased dependence on palmitoylation and cytoplasmic sequences for plasma membrane localisation. To test the role of the extracellular domain on pTRAP localisation, we selected the extracellular region of T cell surface molecule CD4, which is relatively large, N-glycosylated and contains four immunoglobulin-like domains. In the native CD4 molecule it is followed by a transmembrane domain and a short cytoplasmic domain. The transmembrane domain is directly followed by a palmitoylation site (composed of two cysteine residues), which is very similar to the palmitoylation sites in pTRAPs. As shown in Fig. 3A, mutations of these cysteines did not have any effect on CD4 surface expression in human T cells. We fused the extracellular domain of CD4 (CD4ex) to the full coding sequence of LAT and its variants lacking the palmitoylation motif and/or cytoplasmic domain (Fig. 3B,C). The addition of the CD4 extracellular domain to the native LAT sequence (CD4ex-LAT) did not affect its surface expression in human T cells (Fig. 3D, upper left panel). Conversely, when this domain was attached to the non-palmitoylated LAT-CS mutant (CD4ex-LAT-CS), clear plasma membrane localisation was observed (Fig. 3D, lower left panel). This is in contrast to LAT-CS lacking the extracellular domain which exhibited no plasma membrane localisation (Fig. 1C). Similarly, the presence of the CD4 extracellular domain in the non-palmitoylatable CD4ex-LAT-ΔCT-CS variant missing the intracellular domain of LAT caused its weak but reproducible plasma membrane localisation (Fig. 3D, lower right panel). These data demonstrate the role of a glycosylated extracellular domain in plasma membrane localisation of proteins lacking other sorting determinants. Moreover, the additional effect of protein palmitoylation on the surface expression was reproduced by chimeric LAT proteins containing the extracellular domain of CD4.

Characterisation of chimeric and mutant LAT and PAG proteins suggests important roles for palmitoylation and both extra- and intracellular domains in protein sorting (Figs 1–3). To further address this effect we designed model TRAP-like proteins containing artificial TMDs followed by the intracellular domain of CD247, and transiently expressed these proteins in Jurkat cells. ‘TRAP-like’ here refers to an overall topological similarity to TRAP family proteins: monomers with a very short extracellular domain, a single transmembrane domain and an intracellular domain lacking enzymatic function. Our work was inspired by earlier findings that highly hydrophobic artificial sequences can provide a signal for membrane anchorage of model proteins (Munro, 1991, 1995). In these studies, the model TMDs were fused to the extracellular and intracellular domain of CD8α (UniProt ID P01732) and alpha-2,6-sialyltransferase (ST6GALNAC2), non-palmitoylated membrane proteins of type I and type II, respectively. The lengths of these TMDs were found to be important for protein sorting, with longer TMDs preferentially localising to the plasma membrane and shorter TMDs to the Golgi complex. Other parts of these proteins outside TMDs were also shown to play a role.

For our experiments we selected variants of peptide LW21 as an artificial transmembrane domain. LW21 has 21 hydrophobic residues and has been used in vitro as a model transmembrane peptide in a number of biophysical studies (Fastenberg et al., 2003; Kaiser et al., 2011; Machánˇ et al., 2014). The length of this peptide is close to the optimal TMD length for plasma membrane localisation of vertebrate membrane proteins (Sharpe et al., 2010). We found TRAP-like proteins with LW21 to be expressed at the surface of transiently transfected Jurkat cells (data not shown), and subsequently designed a TMD of 19 hydrophobic residues to represent a suboptimal (short) plasma membrane localisation signal. LW19 TMD was obtained by modifying the LW21 sequence by removing two C-terminal tryptophan residues (Table 2). The remaining tryptophan residues at the extracellular end of the hydrophobic stretch mimic the TMDs of PAG and NTAL. We generated TRAP-like protein variants composed of LW19 TMD, the intracellular domain of CD247 and C-terminal GFP for visualisation. CD247 was selected to limit the impact of ‘backbone’ on plasma membrane localisation. A GFP fusion protein of CD247 is retained in the ER in the absence of remaining subunits of the TCR/CD3 complex (Fig. S2), suggesting that CD247 does not contain any dominant sorting sequence mediating its transport further down the exocytic pathway. In addition, CD247 is not palmitoylated. However, it shares a similar overall structure with other TRAP proteins such as LAT. Because the sequence of LW19 TMD is symmetric but the plasma membrane is asymmetric, we therefore generated two versions of LW19 TMD, one of which was made symmetric by adding 2 lysine residues at both ends of the hydrophobic core, and one of which was made asymmetric by adding 2 glutamic acid residues at the N-terminus (extracellular end) and 2 lysine residues at the C-terminus (intracellular end) of the hydrophobic core (Table 2; Fig. 4A,B). Designed model proteins [LW19(Sym)] maintained type I protein orientation in the membrane (Fig. S1C). Asymmetric TRAP-like protein LW19(Asym) localised at the plasma membrane and Golgi complex but not in the ER in Jurkat cells (Fig. 4C,D). By contrast, symmetric TRAP-like protein LW19(Sym) was largely maintained at the level of Golgi complex and ER (Fig. 4C,D). Weak surface expression was detected for this protein, but on very few cells (<10%). A less prominent difference in membrane distribution was observed for LW19 TRAP-like proteins lacking the intracellular domain (Fig. 4C–E). A small proportion (∼20%) of cells expressing LW19-ΔCT protein with the asymmetric TMD [LW19(Asym)-ΔCT] exhibited detectable sorting to the plasma membrane (Fig. 4E), but GA and ER localisation was observed in the remaining cells (Fig. 4C). However, no surface expression was observed for symmetric variant LW19(Sym)-ΔCT even though >200 cells were inspected. These data experimentally demonstrate the importance of asymmetry for the plasma membrane protein sorting of proteins with suboptimal TMDs (Fig. 4C,D). However, a longer and highly hydrophobic asymmetric TMD was sufficient for plasma membrane localisation of TRAP-like proteins regardless of whether symmetric and asymmetric versions of LW25 TRAP-like proteins were tested (Fig. 4D,F,G).

Figs 1, 2 and Fig. S1 demonstrate the importance of palmitoylation for the plasma membrane sorting of LAT, but not of PAG and NTAL proteins. Because asymmetric LW19 TMD was able to mediate efficient sorting of our CD247-based artificial TRAP-like protein to the plasma membrane, we were interested in whether this TMD could rescue plasma membrane targeting of non-palmitoylated LAT. For this purpose, we replaced the TMD and palmitoylation motif in LAT with asymmetric LW19 TMD (Fig. 5A). Interestingly, the exchange of the TMD sequence of LAT protein for the highly hydrophobic, asymmetric LW19 sequence led to it sorting to the plasma membrane in the absence of palmitoylation (Fig. 5B). We did not observe any difference between asymmetric LW19 variants with CD247 or LAT ‘backbones’ (compare Fig. 4C with Fig. 5B). However, the absence of the intracellular domain again caused a reduction of LAT-LW19(Asym)-ΔCT plasma membrane localisation (Fig. 5B,D). Thus, we next tested whether palmitoylation can improve plasma membrane localisation of this variant. For this experiment we inserted a LAT palmitoylation site (CVHC sequence) into LAT-LW19(Asym)-ΔCT TRAP-like protein. Palmitoylation increased its plasma membrane localisation to almost complete surface expression level, comparable to native LAT (compare Fig. 1C with Fig. 5C). This effect was caused by the palmitoylation since the insertion of a non-palmitoylatable SVHS sequence had no impact (Fig. 5C,D). Our observations support the view that plasma membrane localisation of proteins is primarily determined by their TMD but supplementary signals, such as palmitoylation, can increase sorting efficiency.

In this study we have investigated the impact of the TMD, its flanking sequences and palmitoylation on the plasma membrane localisation of pTRAP family proteins. These proteins are essential for the proper function of various cells, facilitating signalling processes taking place at the plasma membrane. Their plasma membrane localisation is, therefore, a prerequisite for their function.

We studied the localisation of three pTRAP proteins: LAT, PAG and NTAL. Based on published data describing the correlation between TMD length and protein sorting to various cellular compartments (Sharpe et al., 2010), one would expect that in the absence of palmitoylation LAT and PAG, with TMDs of 23 residues, would localise to the plasma membrane, whereas NTAL, with a 19 amino acid TMD, would localise to the ER and Golgi complex. Intriguingly, our live cell imaging data showed that these predictions were valid only for PAG, whereas non-palmitoylated LAT was retained mainly in the Golgi complex, in agreement with data published by Hundt et al. (2009). Non-palmitoylated NTAL was sorted to the plasma membrane. Moreover, mislocalisation of LAT-CS to the Golgi complex caused its degradation. These experiments demonstrated that, in addition to TMD length, other sorting determinants such as palmitoylation also define the behaviour of pTRAPs in cell membranes.

Palmitoylation is essential for the proper function of LAT, PAG and NTAL (Brdicka et al., 2002, 2000; Posevitz-Fejfár et al., 2008; Zhang et al., 1998). Here we demonstrated that non-palmitoylated PAG and NTAL showed unaltered sorting. These observations indicate that plasma membrane localisation is important but not sufficient for the function of pTRAPs in lymphocyte signalling. Super-resolution microscopy recently identified non-homogenous distribution of palmitoylated proteins in the plasma membrane (Fukata et al., 2013; Owen et al., 2010; Saka et al., 2014). This suggests that palmitoylation of PAG and NTAL, which is unresolvable using confocal microscopy, could determine their precise localisation at the plasma membrane. A loss of palmitoylation would then result in incorrect distribution of PAG and NTAL at the plasma membrane of Jurkat cells and consequently disrupt the spatio-temporal control required for their proper involvement in cell signalling (Cebecauer et al., 2010).

All LAT and PAG variants lacking the intracellular domain showed inefficient release from the ER. This was especially the case for non-palmitoylated LAT-CS-ΔCT, which was almost exclusively resident in the ER (Fig. 1F). The data suggest that the intracellular domain encodes signals for ER exit whereas TMD and proximal sequences define the fate of membrane proteins further down the exocytic pathway. Both, LAT and PAG proteins contain DxE and YxxΦ motifs reported to provide signals for efficient exit of some proteins from the ER (Sevier et al., 2000). It was shown that mutation or deletion of a larger sequence containing these motifs, enables recycling of membrane proteins between the ER and Golgi complex and limits their sorting to the surface (Fossati et al., 2014). We speculate that by deletion of the entire intracellular domain, including combined DxE and YxxΦ motifs, we have reduced the ability of these variants to enter the later sorting machinery of the trans-Golgi/endosomal network.

We also designed artificial TRAP-like proteins containing a short model LW19 TMD and showed that, similarly to a previously used TMD composed of a stretch of 17 leucine residues (Munro, 1995), it was insufficient for sorting to the plasma membrane. Because the addition of two more hydrophobic residues in LW21 led to weak but detectable expression in the plasma membrane (data not shown), we concluded that LW19 is a suboptimal TMD for plasma membrane localisation, positioned at the borderline of the sorting length scale. This unique feature allowed us to investigate the impact of palmitoylation and proximal sequence(s) on sorting of TRAP-like proteins in more detail. For TMD proximal sequences, basic residues are enriched near the cytosolic end of the TMD of integral membrane proteins. These residues play a role during the insertion of the hydrophobic stretch of a nascent protein into the membrane (Andersson et al., 1992; Nilsson et al., 2005). By contrast, a slight increase in the presence of acidic residues near the extracellular end of the TMD was found in a comprehensive study with >700 vertebrate plasma membrane proteins analysed (Sharpe et al., 2010). Indeed, acidic amino acids are found close to the extracellular end of the TMD in some single-spanning membrane proteins of human lymphocytes (e.g. LAT, NTAL, CD8α, DAP-12). Interestingly, replacement of lysine residues with glutamic acid at the extracellular end of suboptimal LW19 TMD, and thereby generation of asymmetric LW19(Asym), led to an increase in plasma membrane localisation. Together with the preference of less voluminous hydrophobic amino acids for the exoplasmic half of the plasma membrane (Quiroga et al., 2013; Sharpe et al., 2010), these data support the importance of asymmetry in proteins with a suboptimal TMD. In addition, the increased surface localisation of palmitoylated LAT-LW19(Asym_CVHC)-ΔCT compared with its non-palmitoylatable variant LAT-LW19(Asym_SVHS)-ΔCT again supports the view that palmitoylation can facilitate protein sorting.

In this study we show the importance of the unique primary sequence of TMDs for the localisation of integral proteins in cell membranes. As such, we provide a novel standpoint for discussion on why so many TMDs have evolved when a handful would be sufficient for targeting membrane proteins to their proper cellular compartments (Spira et al., 2012). By contrast, no conclusive data were obtained to uncover the difference between LAT and PAG TMDs, which are formed of the same number of amino acids (23 residues). TMD sequence analysis shows that all tested TRAP proteins contain a similar number of hydrophobic (∼ 80%) and no charged residues. No clear difference could be found by calculating mean hydrophobicity of their TMDs (Table 2) (Tossi et al., 2002). Further experiments are required to uncover the exclusive character of these TMDs, especially whether the limited presence of bulky amino acids in the hydrophobic core or the asymmetric distribution of charged residues flanking the TMD provide sufficient signal, comparable to asymmetric LW19 TRAP-like proteins tested herein. It would also be interesting to investigate how the exceptionally short 19 residue TMD of NTAL drives plasma membrane localisation in the absence of palmitoylation.

To summarise, we have confirmed the dominant impact of TMD length and hydrophobicity on plasma membrane localisation of proteins, but also provided evidence that secondary sorting determinants such as palmitoylation or flanking sequences have evolved for proteins with suboptimal TMD length. We additionally demonstrate that the presence of ER exit motifs in the intracellular domain further influences the transport of proteins towards the plasma membrane, supporting the view that more than one determinant plays in concert to define the precise localisation of proteins in cells.

Jurkat T and HeLa (or HEK293) cell lines were grown in RPMI-1640 and DMEM media, respectively (Sigma-Aldrich), supplemented with glutamine and 10% foetal calf serum (Life Technologies) at 37°C under 5% CO₂ in a humidified incubator. These cells originate from the cell bank of the Institute of Molecular Genetics in Prague, Czech Republic, are regularly tested for surface markers, morphology and contamination.

Jurkat cells were transiently transfected using Neon^® transfection system (Life Technologies). According the manufacturer's instructions, 1 µg of vector DNA per shot per 50,000 cells was used. HeLa and HEK293 cells were grown at >50% confluency on clean coverslips for minimum of 18 h before transfection. For transfection, 500 ng of DNA was mixed with 1.5 µl of Fugene HD transfection reagent (Promega) in 25 µl OptiMEM, incubated for 20 min at ambient temperature and dispersed onto cultured cells in 24-well plates. Live cell imaging was performed 16–24 h after transfection.

To immunoprecipitate GFP-tagged proteins from the transfected cells, transfectants were lysed in lysis buffer [50 mM Tris pH 7.5, 150 mM NaCl, 1 mM AEBSF (Sigma-Aldrich)] containing 1% n-dodecyl-β-D-maltoside followed by immunoprecipitation with GFP-specific rabbit antisera (kind gift from Llewelyn Roderick, University of Wales). The immunoprecipitated proteins were detected by immunoblotting with the same anti-GFP antibody.

Plasmid pXJ41-EGFP was prepared by PCR amplification of the EGFP sequence of vector pEGFP-N1 with a spacer (GSGGGS) attached to the N-terminus (primers T198 and T199; see list of all primers in Table S1) and sub-cloning into the BamHI and XhoI restriction sites of vector pXJ41 (Xiao et al., 1991). For further cloning, restriction sites KpnI and BglII in the pXJ41-EGFP construct were eliminated by mutation. We designed a modular system for cloning of pTRAP variants with or without the intracellular domain and with the possibility to exchange the sequence encoding the TMD and flanking 4–5 residues (Fig. S3A). A DNA fragment encoding the LAT sequence was synthesised by GeneArt (Invitrogen) and includes the 5′ UTR and leader sequence of human CD148, followed by a Myc-tag and the entire coding sequence of LAT without the ATG start codon. The Myc-tag sequence and the coding sequence of LAT were separated by a KpnI restriction site. A silent mutation was inserted into the LAT sequence (residues GGC→GGA) to introduce a BamHI restriction site between the TMD and the intracellular domain. Restriction sites NotI and BamHI (partial digestion) were used for sub-cloning and generation of a pXJ41-LAT-EGFP vector. A pXJ41-LAT-ΔCT-EGFP variant was generated by deletion of the sequence encoding the intracellular domain using BamHI restriction sites and self-ligation. CS mutation was introduced into the pXJ41-LAT-ΔCT-EGFP vector by amplification of its TMD sequence (using primers T255 and T256) and sub-cloning of the LAT-CS TMD sequence instead of the LAT TMD using KpnI and BamHI cloning sites. A full length version was generated by sub-cloning of the LAT intracellular sequence into pXJ41-LAT-ΔCT-CS-EGFP using BamHI restriction sites.

A pXJ41-PAG-ΔCT-EGFP vector was prepared by amplification of the PAG TMD sequence (using primers T257 and T258) from plasmid PAG/pEFIRES-N1 and sub-cloning it into vector pXJ41-LAT-ΔCT-EGFP instead of the LAT TMD sequence using cloning sites KpnI and BamHI. The PAG intracellular region was amplified from PAG/pEFIRES-N1 (using primers T259 and T260) and sub-cloned into vector pXJ41-PAG-ΔCT-EGFP using BamHI restriction sites to generate pXJ41-PAG-EGFP. A CS version of PAG-ΔCT was prepared the same way except for the use of primer T263 instead of T258 to introduce the CS mutation. The PAG intracellular region amplified from plasmid PAG/pEFIRES-N1 was than sub-cloned into vector pXJ41-PAG-ΔCT-CS-EGFP to generate pXJ41-PAG-CS-EGFP. Full-length LAT and PAG sequences were also cloned as fusions with mCherry fluorescent protein by replacing the EGFP sequence in pXJ41-LAT-EGFP and pPAG-EGFP with a sequence encoding mCherry from plasmid pcDNA3.1-mCherry (kind gift from Marco Purbhoo, Imperial College London).

The DNA sequence of the CD4 extracellular region was amplified (using primers T246 and T297) from plasmid CD4/pEGFP-N1 and sub-cloned into vectors pXJ41-LAT-EGFP, pXJ41-LAT-CS-EGFP, pXJ41-LAT-ΔCT-EGFP and pXJ41-LAT-ΔCT-CS-EGFP through EcoRI and KpnI restriction sites, to generate pXJ41-CD4ex-LAT-EGFP, pXJ41-CD4ex-LAT-CS-EGFP, pXJ41-CD4ex-LAT-ΔCT-EGFP and pXJ41-CD4ex-LAT-ΔCT-CS-EGFP constructs.

To generate pXJ41-LW19(Sym)-EGFP and pXJ41-LW19(Asym)-EGFP, DNA fragments encoding symmetric and asymmetric variants of LW19 fused to the CD247 intracellular domain were synthesised by GeneArt (Invitrogen) and sub-cloned into plasmid pXJ41-LAT-EGFP using restriction sites KpnI and BamHI, replacing the coding sequence of LAT. During the fragment synthesis, the native BamHI site in the CD247 sequence was removed and a BglII restriction site newly inserted into the CD247 sequence (AGGAGC→AGATCT) to separate the TMD and the intracellular domain. pXJ41-LW25(Sym)-EGFP and pXJ41-LW25(Asym)-EGFP constructs were generated from respective LW19 constructs by site-directed mutagenesis. To achieve this, from each template [pXJ41-LW19(Sym)-EGFP or pXJ41-LW19(Asym)-EGFP], two separate overlapping cDNA fragments were synthesised by PCR with primer pairs T295/T292, and T291/T296. These fragments were then fused in a subsequent PCR reaction using flanking primers T295 and T296 and subcloned into the KpnI and BamHI sites of pXJ41-LAT-EGFP, replacing the coding sequence of LAT. ΔCT variants of all these LW constructs were generated by deletion of the sequence encoding the intracellular domain using BglII and BamHI restriction and self-ligation for the generation of constructs pXJ41-LW19(Sym)-ΔCT-EGFP, pXJ41-LW19(Asym)-ΔCT-EGFP, pXJ41-LW25(Sym)-ΔCT-EGFP and pXJ41-LW25(Asym)-ΔCT-EGFP.

The asymmetric variant of LW19 in a LAT ‘backbone’ was amplified from the pXJ41-LW19(Asym)-EGFP construct (using primers T289 and T290) and sub-cloned into vectors pXJ41-LAT-EGFP and pXJ41-LAT-ΔCT-EGFP using restriction sites EcoRI and BamHI to generate vectors LAT-LW19(Asym) and LW19(Asym)LAT-ΔCT, respectively. To introduce the CVHC motif into the sequence of LAT-LW19(Asym)-ΔCT construct we amplified its TMD sequence using primers T290 and T294, and then sub-cloned the mutated sequence into the pXJ41-LAT-ΔCT-EGFP vector using EcoRI and BamHI restriction sites to generate the LAT-LW19(Asym_CVHC)LAT-ΔCT construct. The cloning strategy for LAT-LW19(Asym_SVHS)LAT-ΔCT construct was the same with the exception of using a different set of primers (T290, T298).

The pXJ41-NTAL-EGFP vector was generated by amplification of the NTAL coding sequence (using primers NTAL wt fwd EcoRI, NTAL rev BamHI) from the NTAL-pFLAG-CMV plasmid and sub-cloning into vector pXJ41-LAT-EGFP through EcoRI and BamHI restriction sites. We inserted the CS mutation in the NTAL sequence by linking PCR (using primers NTAL wt fwd EcoRI, NTAL rev BamHI, NTAL CS fwd, NTAL CS rev) using NTAL-pFLAG-CMV plasmid as a template. The PCR product was sub-cloned into vector pXJ41-LAT-EGFP through EcoRI and BamHI restriction sites to generate pXJ41-NTAL-CS-EGFP.

A version of pXJ41-LAT-ΔCT-EGFP without the CD148 leader, c-Myc and 5′ UTR was generated by amplification of the LAT-EGFP sequence (using primers T287 and LAT BamHI rev) and sub-cloning into the former pXJ41-LAT-ΔCT-EGFP vector using EcoRI and BamHI restriction sites. pXJ41-LAT-ΔCT-CS-EGFP* without the CD148 leader, c-Myc and 5′ UTR was generated using the same cloning strategy (with use of primers T287 and T255). No difference between plasma membrane localisation of LAT variants with and without the CD148 leader, c-Myc and 5′ UTR was detected (Fig. S3B,C).

Cells were imaged on poly-L-lysine-coated (to immobilise cells) glass-bottom 8-well chamber slides (Lab-Tek^®, Thermo Scientific) supplemented with pre-heated, colour-free RPMI-1640 medium. Images were taken using a Leica SP5 TCS AOBS Tandem laser scanning confocal microscope equipped with Leica HyD hybrid detector, 63×1.3 NA glycerine immersion objective (Leica PLAN APO) and live cell support chamber. LAS AF image software (Leica Microsystems) was used for acquisition. 3–4 sections were taken per cell, focused on the Golgi complex, ER and the plasma membrane. Minor contrast and/or level adjustments were applied and images were processed for publishing by Fiji/ImageJ (Schindelin et al., 2012).

For colocalisation studies, markers for the ER (Sec61-mCherry) and Golgi complex (Golgi 7-mApple) were co-transfected into Jurkat cells and imaged in parallel with GFP constructs using the TRIC channel setup (Leica Microsystems). mApple-Golgi-7 was a gift from Michael Davidson (Addgene plasmid #54907) and mCherry-Sec61 beta from Gia Voeltz (Addgene plasmid #49155).

Kinetic studies were performed using transiently transfected Jurkat T cells. After 16 h of culture, 25 µg/ml cyclohexamide was added to media and live cell imaging was performed after 0, 2, 4, 6 and 20 h.

The hydrophobic core or TMD of all tested proteins (Table 2) were determined using the bioinformatics tool TMHMM 2.0 prediction of transmembrane helices in proteins (http://www.cbs.dtu.dk/services/TMHMM-2.0/) (Krogh et al., 2001). For in silico prediction of protein membrane compartmentalisation we employed http://www.tmdsonline.org/predict.html#0 (Sharpe et al., 2010). Mean hydrophobicity of protein TMDs was calculated using HydroMCalc algorithm provided by A. Tossi (University of Trieste). No difference in mean hydrophobicity of TMDs was found when an alternative algorithm was used (Fauchere and Pliska, 1983).

We would like to thank Anthony I. Magee (Imperial College London) and Kvido Stříšovský (IOCB, Prague) for critical reading of the manuscript. We would also like to thank Radek Šachl for help with MATLAB script development.