S-palmitoylation determines TMEM55B-dependent positioning of lysosomes

Lysosomes and late endosomes (LEs) are highly dynamic organelles. Their spatiotemporal distribution depends on active transport mainly driven by microtubule motors such as kinesins and dynein (Bonifacino and Neefjes, 2017; Saftig and Puertollano, 2021). Continuous transport of lysosomes within the cell is essential for the maintenance of various cellular functions, such as the fusion of lysosomes with autophagosomes or the plasma membrane (lysosomal exocytosis), the tubulation of lysosomes, and the formation of contact sites with other organelles, including the endoplasmic reticulum (Ballabio and Bonifacino, 2020; Saftig and Puertollano, 2021). Various motor protein–cargo complexes mediating the microtubule-dependent positioning of lysosomes have been described, and kinesin 1 and kinesin 3 facilitate the anterograde transport of lysosomes. These protein complexes depend on the cytosolic BLOC-1-related complex (BORC; Guardia et al., 2016). The BORC interacts with the Ragulator complex, which is attached to the lysosomal membrane via N-myristoylation (Pu et al., 2017, 2015). ARL8B, a small GTPase, recruits the BORC and its effector SKIP (also known as PLEKHM2), which directly interacts with kinesin 1.

Dynein is the only retrograde microtubule motor. GTP-activated membrane-bound RAB7, which resides on the lysosomal membrane, recruits its effectors, Rab-interacting lysosomal protein (RILP) and PLEKHM1, to promote dynein-driven retrograde transport of LEs and lysosomes. Another protein complex mediating retrograde transport of LEs and lysosomes independent of RAB7 depends on the integral lysosomal membrane protein TMEM55B (also known as PIP4P1) and the cytosolic scaffold c-Jun N-terminal kinase (JNK)-interacting protein 4 (JIP4, also known as JLP and SPAG9). JIP4 binds to TMEM55B, thereby recruiting the dynein–dynactin complex (Willett et al., 2017). TMEM55B has two transmembrane segments, a large N-terminal cytosolic domain, a small luminal loop and a short cytosolic C terminus (Fig. 1A). Overexpression of TMEM55B induces perinuclear clustering of lysosomes at the microtubule-organizing center (MTOC), and TMEM55B knockdown leads to a more peripheral distribution of lysosomes due to increased dynein-dependent retrograde transport processes (Willett et al., 2017). Initially, TMEM55B was assigned as a phosphatidylinositol-4,5-bisphosphate 4-phosphatase (Ungewickell et al., 2005). However, this presumed catalytic function has been challenged and later refuted (Willett et al., 2017). TMEM55A (also known as PIP4P2) is a paralog of TMEM55B, which is also localized in lysosomes. The amino acid sequences of human TMEM55A and TMEM55B are 51% identical and 63% similar. However, TMEM55A overexpression does not affect lysosomal positioning upon overexpression, implicating independent functions (Willett et al., 2017). Both TMEM55A and TMEM55B contain conserved dileucine-based sorting motifs close to their N termini that mediate adaptor protein-dependent transport. Mutagenesis of this motif leads to partial missorting of both proteins to the plasma membrane (Willett et al., 2017).

In a previous proteome-wide screen for TMEM55B-interacting proteins, TMEM55B was co-immunoprecipitated with subunits of the vacuolar H⁺-ATPase (V-ATPase) and the Ragulator complex (Hashimoto et al., 2018), which are essential constituents of mTORC1 (mechanistic target of rapamycin complex 1) signaling, the nutrient-sensing and -signaling machinery of the cell.

TMEM55B is embedded in raft-like microdomains of membranes (Takemasu et al., 2019), a feature that is often dependent on the S-palmitoylation of transmembrane proteins (Ballabio and Bonifacino, 2020; Hashimoto et al., 2018; Takemasu et al., 2019; Zhang et al., 1998). S-palmitoylation is an important post-translational modification characterized by the reversible attachment of a C16 saturated fatty acyl chain to cytoplasmic cysteine residues by different palmitoyl acyltransferases (PATs) (Charollais and Van Der Goot, 2009). As a result of partitioning into membrane microdomains, S-palmitoylation often affects the intracellular trafficking and function of membrane proteins. In several proteomic studies, TMEM55A and TMEM55B have been described as candidate proteins that undergo S-palmitoylation (Dowal et al., 2011; Ivaldi et al., 2012; Martin et al., 2012; Merrick et al., 2011).

Here we reveal that TMEM55A and TMEM55B are bona fide S-palmitoylated proteins. Under steady-state conditions, TMEM55B is present in multiple S-palmitoylated forms in different cell lines and murine tissues. Up to five S-palmitoylated cysteine residues are used for the lipidation of TMEM55B. S-palmitoylation of TMEM55B is dispensable for partitioning into detergent-resistant membranes (DRMs). A mutant TMEM55B protein with all cytoplasmic cysteines mutated to serine is retained in the Golgi instead of being transported to lysosomes. Interestingly, additional mutation of the critical dileucine-based sorting motif of TMEM55B leads to a partial localization of TMEM55B at the plasma membrane, suggesting a critical role of S-palmitoylation in adaptor protein-dependent post-Golgi sorting of TMEM55B.

TMEM55A and TMEM55B have been identified in previous proteome-wide screening studies that aimed to identify S-palmitoylated proteins in different cell lines and tissues (Dowal et al., 2011; Ivaldi et al., 2012; Martin et al., 2012; Merrick et al., 2011). These findings prompted us to experimentally validate the putative S-palmitoylation of TMEM55A and TMEM55B and investigate the impact of this modification on the biology of the two proteins. To this end, we utilized a resin-assisted capture (acyl-RAC) assay that specifically enriches S-palmitoylated proteins (Fig. 1B) (Forrester et al., 2011). In this assay, free thiols are first blocked with methyl methanethiosulfonate (MMTS), which reversibly blocks free cysteines. Thioesters are then cleaved with neutral hydroxylamine (NH₂OH), and the newly liberated thiols are captured with a thiol-reactive Sepharose resin (Forrester et al., 2011). HeLa cells were transfected with hemagglutinin (HA)-tagged TMEM55B. Transfected cells were analyzed using the acyl-RAC assay followed by immunoblotting of the resulting fractions (Fig. 1C). We included control samples from cells that were treated with the S-palmitoylation inhibitor 2-bromopalmitate (2-BP) (Webb et al., 2000) or with NaCl instead of hydroxylamine in the thioester cleavage step as negative controls. Furthermore, the presence or absence of a known S-palmitoylated protein (LAMTOR1; Nada et al., 2009) and a non-palmitoylated lysosomal protein (LAMP2) in these fractions was analyzed. Ectopically expressed TMEM55B–HA was robustly identified in the hydroxylamine-treated fraction containing the S-palmitoylated proteins (as was LAMTOR1). LAMP2 (as a non-palmitoylated protein) was absent in the pull-down fractions, confirming that TMEM55B is S-palmitoylated. Treatment with 2-BP significantly reduced the S-palmitoylation of TMEM55B–HA, providing additional evidence for the specificity of the assay (Fig. 1C). Like TMEM55B, TMEM55A–HA was also detected in the hydroxylamine-treated fraction, indicating that both TMEM55 paralogs undergo S-palmitoylation (Fig. S1A).

In an independent click-chemistry-based experimental approach, we used metabolic labeling of TMEM55B–HA-expressing HeLa cells with the clickable palmitate analog 17-octadecynoic acid (17-ODYA; Martin, 2013) (Fig. S1B). TMEM55B–HA was precipitated from the cell lysates via the HA-tag, and the click reaction was performed with a ‘clickable’ fluorescent dye. The samples were then analyzed by SDS-PAGE and fluorescence scanning as well as immunoblot. Metabolic labeling with varying pulse times (2–24 h) revealed a band at the expected molecular weight of TMEM55B–HA (∼35 kDa) with increasing pulse-time-dependent intensity. This strongly suggests the incorporation of the clickable 17-ODYA and provides direct evidence for S-palmitoylation of TMEM55B (Fig. S1B).

Human and mouse TMEM55B are rich in cysteines. Both contain 19 cysteine residues in total (out of 277 and 284 amino acids for the human and mouse proteins, respectively). These cysteine residues are unevenly distributed over the full-length protein sequence. While the very N-terminal end of TMEM55B lacks any cysteine residue, cysteines are frequent in the middle part of the protein and the first N-terminal transmembrane segment (Fig. 1A). Notably, most cysteines are evolutionarily conserved among different species, implicating essential preserved functions (Fig. S1C). In order to investigate whether TMEM55B is S-palmitoylated at a single cysteine residue or at multiple cysteine residues, we employed a modified acyl-RAC assay based on the attachment of a polyethylene glycol (PEG) group on each previously S-palmitoylated cysteine after thioester cleavage (Percher et al., 2016). The attached PEG leads to a distinctive shift of the apparent molecular weight of the target protein after SDS-PAGE (Percher et al., 2016). SDS-PAGE of lysates from cells transfected with TMEM55B–HA and treated with the acyl-PEG exchange chemicals followed by immunoblotting with antibodies directed against HA revealed four to six distinct bands, each representing non-palmitoylated and differentially S-palmitoylated forms of TMEM55B with an increasing number of occupied S-palmitoylation sites, as depicted in Fig. 1D. Mono- and di-palmitoylated forms were the most abundant species detected. Non-palmitoylated TMEM55B was also detected to a substantial degree. Analysis of endogenous TMEM55B in lysates from mouse embryonic fibroblasts (MEFs) yielded similar results, excluding the possibility that the S-palmitoylation observed was an artifact resulting from TMEM55B overexpression (Fig. S2A). Analysis of the S-palmitoylation of different murine tissues (brain, testis and spleen) revealed a similar S-palmitoylation pattern of TMEM55B among all the tested tissues, with mono- and di-palmitoylated forms as the most abundant species (Fig. 1E). However, it should be noted that the quantification of the band intensities might be biased by the PEGylation status, due to differences in the blotting procedure, and might not quantitatively reflect the degree of S-palmitoylation. Analysis of the dynamics of TMEM55B S-palmitoylation in a pulse-chase assay using clickable palmitate and parallel labeling of the protein by the clickable methionine analog L-azidohomoalanine (AHA) revealed a similar half-life of the protein backbone and the S-palmitoylated TMEM55B (Fig. S2B), indicating an absence of dynamic de- or re-palmitoylation.

To more directly address the functional consequence of TMEM55B S-palmitoylation, we investigated the effect of abolishing S-palmitoylation of TMEM55B by mutating all cytosol-exposed and transmembrane-embedded cysteine residues to serine (TMEM55B^NoCys–HA). The absence of S-palmitoylation of the TMEM55B^NoCys–HA construct was confirmed using the acyl-RAC assay. In contrast to wild-type TMEM55B–HA, no binding of TMEM55B^NoCys–HA to the thiol-reactive Sepharose beads was observed (Fig. 2A). Immunoblot analysis of HeLa cells transfected with wild-type TMEM55B–HA or TMEM55B^NoCys–HA revealed reduced total levels of the protein lacking cysteines in the ‘input fractions’ used for the assay (Fig. 2A). Treating the cells with the proteasome-inhibitor MG132 rescued this effect partially and increased the levels of TMEM55B^NoCys-HA, indicating a fraction of non-palmitoylated TMEM55B is degraded via the proteasome (Fig. S3A).

We also applied the 17-ODYA labeling click-chemistry-based approach to analyze the S-palmitoylation status of the TMEM55B^NoCys–HA variant (Fig. 2B). Metabolic labeling of cells transfected with TMEM55B–HA or TMEM55B^NoCys–HA followed by immunoprecipitation, click-chemistry and fluorography revealed a band at the expected molecular weight of TMEM55B–HA for the wild-type protein, but not for the TMEM55B^NoCys–HA, validating the absence of S-palmitoylation. In comparison to wild-type TMEM55B–HA, the TMEM55B^NoCys–HA mutant was expressed at a lower level; however, similar levels of the tagged proteins were detected in the immunoprecipitated fractions, indicating that the immunoprecipitation step remedied most of the difference, allowing an appropriate comparison (Fig. 2B). Finally, we validated the absence of S-palmitoylation of the TMEM55B^NoCys–HA variant in the acyl-PEG assay. In contrast to the cells transfected with wild-type TMEM55B–HA, no molecular weight shift was observed for the TMEM55B^NoCys–HA in those fractions where PEG was added and disulfide linkages were cleaved by NH₂OH (Fig. S3B). As additional controls for this assay, N-Ras and LAMP1 immunoblotting were included for S-palmitoylated proteins and non-palmitoylated proteins, respectively, additionally validating the specificity of the assay.

While the S-palmitoylation of soluble cytosolic proteins serves as a hydrophobic membrane anchor, the S-palmitoylation of membrane proteins often affects their subcellular trafficking, localization and sorting between different organelles (Charollais and Van Der Goot, 2009; Flannery et al., 2010; Tanimura et al., 2006; Vergarajauregui and Puertollano, 2006). Therefore, we studied the trafficking of non-palmitoylated TMEM55B and analyzed whether the lack of S-palmitoylation affected its subcellular localization. Under steady-state conditions, both endogenous and ectopically expressed wild-type TMEM55B is found on LEs and lysosomes (Willett et al., 2017), where it extensively colocalizes with LAMP2 but not with other organelle marker proteins, such as the Golgi marker GM130 (also known as GOLGA2; Fig. 2C). Notably, TMEM55B overexpression induces a perinuclear clustering of lysosomes due to the recruitment of the retrograde microtubule motor complex dynein–dynactin (Willett et al., 2017). Remarkably, overexpressed TMEM55B^NoCys–HA showed, in contrast to the wild-type TMEM55B–HA, only little colocalization with LAMP2, but a high degree of colocalization with GM130, indicating extensive localization of non-palmitoylated TMEM55B to the Golgi (Fig. 2C). Quantification of the images and representation of the signals as the Pearson correlation coefficient (representing a measure of the covariance in the two signals) or the Manders overlap coefficient (as a measure of the fraction of overlapping pixels) revealed a decreased colocalization of TMEM55B^NoCys–HA with LAMP2 and an increased colocalization with GM130 (Fig. 2C,D). While the transient expression of wild-type TMEM55B–HA induced robust perinuclear clustering, heterologous expression of the TMEM55B^NoCys–HA mutant resulted in the dispersed distribution of lysosomes throughout the cells, which was comparable to the distribution in non-transfected cells (Fig. 2E). These results indicate that lysosomal membrane localization of TMEM55B is required to exert its effect on lysosomal positioning and that S-palmitoylation of TMEM55B modulates lysosomal positioning.

TMEM55B has previously been detected in lipid rafts (Hashimoto et al., 2018). In good agreement with these findings, in MEF cells, endogenous TMEM55B exhibited a patchy and clustered distribution in the LAMP1-positive lysosomal membrane (Fig. 3A). These findings indicate a concentration of TMEM55B in protein complexes, raft-like domains or organelle contact sites. In contrast, LAMP1 staining showed a rather smooth ring-like staining pattern. It should be noted that LAMTOR1, which was previously identified as a physical interactor of TMEM55B (Hashimoto et al., 2018), has been found in raft-like subdomains of LEs due to its myristoylation and S-palmitoylation (Nada et al., 2009).

Palmitoylation is a major determinant of protein targeting to raft-like subdomains of membranes (Levental et al., 2010). This prompted the question of whether S-palmitoylation of TMEM55B is required for its partitioning into raft-like membrane subdomains in a manner similar to the partitioning of LAMTOR1. Upon sucrose gradient fractionation of membranes from transfected HeLa cells, we confirmed the localization of TMEM55B in raft-like domains, because TMEM55B–HA was enriched in fractions containing the typical DRM marker flotillin1 (Fig. 3B). The non-raft protein transferrin receptor (TFR) was absent from these DRM fractions. Endogenous TMEM55B showed an even stronger enrichment in DRMs compared to the overexpressed protein and was virtually absent from the non-raft fractions upon sucrose gradient centrifugation and analysis of the fractions by immunoblotting (Fig. 3C). Surprisingly, incubation of cells with the palmitoylation inhibitor 2-BP did not affect the partitioning of TMEM55B–HA into DRMs (Fig. 3D), and similarly, the non-palmitoylated TMEM55B^NoCys–HA construct was still found in substantial amounts in the DRM fraction (Fig. 3E). These experiments indicate that S-palmitoylation is not essential for the partitioning of TMEM55B into raft-like domains or DRMs in lysosomes.

Our immunofluorescence studies revealed localization of TMEM55B^NoCys–HA within the Golgi (Fig. 2C), implicating either retention of the protein lacking cysteines and palmitoylation in the Golgi or rapid retrograde transport (‘retrieval’) from post-Golgi compartments (i.e. endosomes or LEs) back to the Golgi. To investigate these two possibilities experimentally, we blocked retrograde transport through early endosomes pharmacologically by applying chloroquine. Chloroquine treatment causes swelling of endocytic compartments and interferes with retrograde transport to the Golgi of Golgi-resident proteins such as TGN46 (also known as TGOLN2; Chapman and Munro, 1994; Kent et al., 2012; Reaves and Banting, 1994; Reaves et al., 1998; Rous et al., 2002) (Fig. 4). Chloroquine treatment of cells transfected with wild-type TMEM55B–HA (Fig. 4, upper panels) resulted in perinuclear clustering of enlarged lysosomes and extensive colocalization of TMEM55B–HA and LAMP2, but no colocalization of TMEM55B–HA with GM130 or TGN46 was observed. Endogenous TGN46 displayed, in contrast to untreated cells where it colocalized with GM130, vesicular localization in the chloroquine-treated cells. In contrast to cells transfected with TMEM55B–HA, cells transfected with TMEM55B^NoCys–HA showed little colocalization between HA and LAMP2 (Fig. 4, lower panels) but extensive colocalization between HA and GM130. Little colocalization between TMEM55B^NoCys–HA and TGN46 was observed in the chloroquine-treated cells. Also, strongly reduced colocalization of TGN46 and GM130 was observed, validating retention of TGN46 in post-Golgi vesicles upon chloroquine treatment. These experiments suggest that TMEM55B is retained in the Golgi rather than being transported back from post-Golgi compartments to the Golgi by retrograde transport mechanisms.

Having established that TMEM55B is modified at multiple cysteine residues, we wondered which of the cysteine residues are modified by the addition of palmitate. As described above, up to five cytosolic TMEM55B cysteines are palmitoylated under steady-state conditions (Fig. 1D). The high number of cysteines (19 in total) contained in the palmitoylation-accessible regions of the protein complicated the identification of individual palmitoylated residues. In a first attempt to address this issue, we generated constructs where all cysteine residues in the cytosolic N terminus were mutated to serine residues (TMEM55B^N-term_NoCys–HA). Another construct was generated for the C terminus containing the two transmembrane segments, the luminal loop and the cytosolic C-terminal tail (TMEM55B^C-term–^NoCys–HA) (Fig. 5A). Analysis of lysates from cells transfected with these constructs by immunoblotting (Fig. 5B, left panel) revealed that all constructs were expressed, though TMEM55B^NoCys–HA was again expressed at a significantly lower level, indicating that cysteines are critical for stability (see also Fig. S3).

Interestingly, we observed an altered migration behavior during SDS-PAGE with a minor shift towards a slightly lower apparent molecular weight for the TMEM55B^N-term_NoCys–HA variant compared with TMEM55B–HA that might be caused by altered migration behavior due to S-palmitoylation during SDS-PAGE, as observed previously for other S-palmitoylated proteins (Diez-Ardanuy et al., 2017). Analysis of lysates from cells transfected with these constructs using the acyl-RAC assay revealed specific binding of all variants to the thiol-reactive resin, qualitatively indicating S-palmitoylation of all constructs except TMEM55B^NoCys–HA (Fig. 5B, right panel). However, the TMEM55B^N-term_NoCys–HA construct bound with lower efficiency, which might be a hint that this variant is S-palmitoylated to a lower extent as compared to TMEM55B^C-term_NoCys–HA.

We next analyzed the cellular localization of the TMEM55B constructs with mutations in the N- or C-terminus using immunofluorescence. In contrast to the perinuclear clustering of lysosomes induced by overexpression of wild-type TMEM55B and colocalization with LAMP2, a lack of colocalization with LAMP2 was observed for TMEM55B^NoCys–HA, as described before (Fig. 5C). Interestingly, no colocalization between LAMP2 and the TMEM55B construct lacking cysteines in the N terminus (TMEM55B^N-term_NoCys–HA) was observed, similar to the construct lacking all cysteines. TMEM55B^N-term_NoCys–HA showed extensive colocalization with the Golgi marker GM130 and did not alter the distribution of lysosomes (Fig. 5C). In contrast, the construct lacking cysteines in the transmembrane domains and the C terminus (TMEM55B^C-term_NoCys–HA) showed extensive colocalization with LAMP2 comparable to that of the wild-type protein and induced the typical perinuclear accumulation of lysosomes. No colocalization with GM130 was observed (Fig. 5C).

Our data indicate that S-palmitoylation of the cytosolic N-terminal domain of TMEM55B (but not the cysteines within the transmembrane domains or cytosolic C terminus) is critical for Golgi export and sorting to lysosomes and is indirectly critical for inducing perinuclear clustering of lysosomes.

The Golgi localization of the TMEM55B^NoCys–HA mutant (Fig. 2C) suggests a direct impact of S-palmitoylation on intracellular trafficking and post-Golgi sorting. As described above, the sorting of TMEM55B depends on a dileucine-based motif (ERSPLL, in which the glutamic acid residue and the L11 and L12 leucine residues are critical) within the N terminus of TMEM55B (Willett et al., 2017) (Fig. 6A). Dileucine-based motifs are recognized by adaptor protein complexes, facilitating the concentration of cargo into clathrin-coated vesicles and subsequent intracellular sorting. They are typically found close to transmembrane segments of the cargo protein (Bonifacino and Traub, 2003). In TMEM55B, however, the dileucine-based motif is distal of the transmembrane domains, close to the N-terminus, with more than 200 amino acid residues between the sorting motif and the first transmembrane domain (Fig. 6A).

Interestingly, S-palmitoylation affects adaptor protein-dependent sorting of mucolipin-1, an ion channel and integral lysosomal membrane protein (Vergarajauregui and Puertollano, 2006). Palmitoylation mediates endocytosis from the plasma membrane by bringing the C-terminal endocytosis motif closer to the membrane, thus facilitating its interaction with the adaptor protein complex AP-2 (Vergarajauregui and Puertollano, 2006). To investigate whether S-palmitoylation of TMEM55B is required for its interaction with adaptor proteins, we analyzed the sorting of the TMEM55B^NoCys–HA mutant by additionally mutating the dileucine motif. The resulting TMEM55B^NoCys_LL/AA mutant was transfected into HeLa cells, and its localization was analyzed using confocal microscopy (Fig. 6B). In agreement with previously published results (Willett et al., 2017), TMEM55B^LL/AA was missorted to the plasma membrane, with only a minor fraction colocalizing with the lysosomal marker LAMP2 and very little colocalization with the early endosomal marker EEA1 (Fig. S4). TMEM55B^NoCys was again exclusively detected in the Golgi (based on GM130 co-staining). TMEM55B^NoCys_LL/AA lost this exclusive Golgi localization but was instead found to a significant degree at the plasma membrane. These findings suggest that upon mutagenesis of the sorting motif and in the absence of S-palmitoylation, a fraction of TMEM55B can exit the Golgi and is missorted to the plasma membrane. These results are in agreement with the assumption that S-palmitoylation affects the adaptor protein- and dileucine-dependent cellular sorting of TMEM55B.

Our experiments suggest that S-palmitoylation of the N terminus allows the dileucine-based motif to localize in closer proximity to the membrane. To support this hypothesis experimentally, we generated a construct in which the majority of the cytosolic N-terminal domain of TMEM55B was deleted so that the dileucine motif is forced into close proximity to the membrane (TMEM55B^C-term+LL–HA; Fig. 6C). In agreement with our hypothesis, and in contrast to our observations of an N-terminal deletion mutant also missing the dileucine motif (TMEM55B^C-term–HA), TMEM55B^C-term+LL–HA was found in post-Golgi compartments containing the lysosomal marker LAMP2 and at the plasma membrane, as determined by immunofluorescence, indicating partial transport to lysosomes (Fig. 6D). Notably, the expression of the TMEM55B^C-term+LL–HA construct did not result in lysosomal clustering. This suggests that other regions of the N-terminal domain are necessary for this ‘lysosomal positioning’ function.

Lysosomes are highly dynamic organelles, and their microtubule-dependent spatial distribution in the cell is orchestrated by a sophisticated system of motor proteins that needs to be tightly controlled (Ballabio and Bonifacino, 2020; Bonifacino and Neefjes, 2017). The transport of these organelles is regulated by motor protein attachment and protein complexes facilitating cargo engagement (Saftig and Puertollano, 2021). Our understanding of protein complexes mediating this sophisticated system has significantly improved in the past few years, and the complex of TMEM55B and JIP4 is among the most recently characterized motor–cargo adaptor machineries (Willett et al., 2017). How this complex is regulated, for example by post-translational modifications, remains to be determined. As we demonstrated in this study, TMEM55B is subject to S-palmitoylation, a reversible post-translational modification. TMEM55B can also be modified by phosphorylation. A reversible phosphorylation by Erk MAPKs at two distinct serine residues of TMEM55B may regulate TMEM55B-dependent lysosomal positioning (Takemasu et al., 2019). The precise mechanism of how phosphorylation alters this function, however, remains enigmatic. It will be interesting to analyze in future experiments whether phosphorylation and S-palmitoylation of TMEM55B reciprocally affect each other.

Though we cannot exclude a regulatory function of S-palmitoylation in general or of specific S-palmitoylation sites in regard to lysosomal positioning, for example by modulating the interaction of TMEM55B and JIP4, we propose that S-acylation regulates the post-Golgi trafficking of TMEM55B. The closely related protein TMEM55A does not interact with JIP4 (Willett et al., 2017), and it does not mediate lysosomal positioning. However, TMEM55A is also S-palmitoylated, indicating that S-palmitoylation is required for shared functions of the two TMEM55 proteins. Moreover, it is evident from our experiments that blocking S-palmitoylation of TMEM55B leads to complete retention of the protein in the Golgi and abrogates sorting of TMEM55B to lysosomes.

S-palmitoylation of transmembrane proteins is well known to affect intracellular sorting and, in particular, facilitates post-Golgi sorting events (Ernst et al., 2019; Guardia et al., 2016). Mutation of the S-palmitoylation sites or pharmacological inhibition of S-palmitoylation of synaptotagmin VII (SYT7) using 2-BP blocks its trafficking to lysosomes (Flannery et al., 2010). As a proposed mechanism, an S-palmitoylation-dependent interaction with the tetraspanin CD63 has been suggested, which is essential for lysosomal sorting and contains a critical lysosomal sorting motif (Flannery et al., 2010). TMEM55B might similarly require an interaction partner for post-Golgi trafficking, and this interaction might be regulated by S-palmitoylation.

Mutagenesis of the dileucine-based sorting motif in TMEM55B leads to missorting of the protein to the plasma membrane, implicating that the dileucine motif is critical for the adaptor protein-mediated intracellular sorting to lysosomes (Willett et al., 2017). Many lysosomal integral transmembrane proteins harbor the dileucine motif in the membrane-proximal cytosolic part of the protein (Bonifacino and Traub, 2003). It is interesting that the sorting motif in TMEM55B is located distal to the transmembrane segment. This supports the idea that S-palmitoylation brings the dileucine motif in close proximity to the membrane bilayer where it can bind to adaptor proteins (Fig. 6E), which also bind phosphatidylinositol membrane lipids and the membrane-bound ARF1 (Bonifacino and Traub, 2003; Ren et al., 2013).

The results of our experiments with a TMEM55B variant that only contains the C terminus and the two transmembrane domains with the dileucine motif in direct proximity to the membrane support this assumption. However, the lack of specific adaptor protein-mediated sorting in the Golgi (e.g. by deleting or mutating sorting motifs) typically leads to missorting of the cargo to the plasma membrane (Ballabio and Bonifacino, 2020), as observed for the analyzed dileucine-mutated TMEM55B. In contrast, retention in the Golgi was demonstrated for the palmitoylation-deficient TMEM55B mutant, arguing against this explanation. To our surprise, mutagenesis of the dileucine motif in the non-palmitoylated TMEM55B led to partial missorting to the plasma membrane, suggesting that S-palmitoylation plays a role in the adaptor protein-mediated sorting but that additional factors are involved. These factors might involve protein–protein interactions in the Golgi that depend on S-palmitoylation and might act independently or in concert with the adaptor-protein mediated sorting machinery. S-palmitoylation of the integral lysosomal membrane protein mucolipin-1 mediates its adaptor protein-dependent transport by facilitating binding of the canonical motif to the membrane-proximal AP-2 complex (Vergarajauregui and Puertollano, 2006). However, more direct effects of S-palmitoylation on protein sorting are known. Lack of S-palmitoylation of linker for activation of T cells (LAT), an adaptor molecule mediating T cell receptor signaling, which is palmitoylated on one or two juxtamembranous cysteines, leads to its retention in the Golgi (Tanimura et al., 2006). A lack of S-palmitoylation abrogates partitioning into lipid rafts (Tanimura et al., 2006), and the mutant protein is unable to traffic to the plasma membrane despite the presence of the transmembrane portion (Zhang et al., 1998), similar to TMEM55B. For non-palmitoylated TMEM55B, however, we did not observe significant alterations of partitioning into raft-like domains, suggesting a different mechanism. In this regard, it is interesting to note that the transfected TMEM55B construct lacking cysteines in the N-terminal domain was still S-palmitoylated. A contribution of the membrane-proximal and/or membrane-embedded cysteine residues cannot be ruled out.

Another possibility of how S-palmitoylation might regulate post-Golgi sorting has been proposed recently. Palmitoylation of juxtamembranous cysteines affects membrane tilting and sorts palmitoylated proteins into unique subdomains, from which they are sorted in cargo vesicles for anterograde transport (Ernst et al., 2018). The observation that the mutant TMEM55B protein lacking cysteines in the C terminus, including the two transmembrane domains, was still retained in the Golgi suggests that such a mechanism contributes to the observed trafficking defects.

The Golgi is a central hub for S-palmitoylation. Notably, 17 of 23 DHHC PATs exhibit significant overlap with endogenous Golgi markers, and nine localize to the Golgi exclusively (Ernst et al., 2019). Future experiments will address the identification of the responsible PAT(s) for TMEM55A and TMEM55B.

Our study highlights the importance of S-acylation as a critical determinant in regulating the subcellular trafficking of lysosomal transmembrane proteins. Moreover, our findings increase our knowledge of TMEM55B and its role as a central component of the TMEM55B–JIP4-dependent retrograde sorting machinery.

Analytical grade chemicals were purchased, if not stated otherwise, from Sigma-Aldrich (St Louis, MO, USA). The following antibodies were used in the study: anti-TMEM55B (immunoblot, 1:2000; immunofluorescence, 1:300; #23992-I-AP, Proteintech), anti-HA (immunoblot, 1:2000; immunofluorescence, 1:300; #3F10, Roche/Sigma-Aldrich), anti-LAMP2 (immunoblot, 1:2000; immunofluorescence, 1:300; clone Abl93, Developmental Studies Hybridoma Bank), anti-GAPDH (immunoblot, 1:2000; sc-365062, Santa Cruz Biotechnology), anti-N-Ras (immunoblot, 1:2000; sc-31, Santa Cruz Biotechnology), anti-flotillin1 (immunoblot, 1:1000; #610820; BD Bioscience), anti-TFR (immunoblot, 1:1000; #ab84036, Abcam), anti-NPC1 (immunoblot, 1:2000; #ab134113, Abcam), anti-GM130 (immunofluorescence, 1:300; #610822, BD), anti-LAMTOR1 (immunoblot, 1:500; #8975, Cell Signaling Technology), anti-EEA1 (immunofluorescence, 1:300; #3288, Cell Signaling Technology) and anti-TGN46 (immunofluorescence, 1:300; 13573-1-AP, Proteintech). Fluorophore-conjugated secondary antibodies against the corresponding primary antibody species (Alexa Fluor 488-, Alexa Fluor 594- and Alexa Fluor 647-conjugated) were purchased from Invitrogen/Molecular Probes and were diluted 1:500. 2-Bromopalmitate and chloroquine were purchased from Sigma-Aldrich. cDNA coding for mouse TMEM55B or TMEM55A were cloned into pcDNA3.1 Hygro(+) (Invitrogen).

Site-directed mutagenesis was performed according to standard protocols. A PCR mixture was prepared as follows: 1× Phusion GC buffer containing 1.5 mM MgCl₂ (Thermo Fisher Scientific), dNTP mix to a final concentration of 0.2 mM and 3% DMSO (Thermo Fisher Scientific). Overlapping forward and reverse primers (Sigma-Aldrich) containing the desired mutation were added to a final concentration of 0.2 μM, and finally, Phusion High-Fidelity DNA Polymerase (Thermo Fisher Scientific) was added. The PCR reaction was performed in a FlexCycler2 PCR machine (Analytik Jena) with an annealing temperature of 54°C and an elongation temperature of 72°C. The PCR product was treated with 10 units of DpnI (Thermo Fisher Scientific) for 1 h at 37°C to digest the template plasmid (methylated strand) before transformation.

For isolation of DRMs, cells were harvested in phosphate-buffered saline (PBS) containing cOmplete protease inhibitors (Roche/Sigma-Aldrich) and lysed for 1 h at 4°C in Solution A [50 mM Tris-HCl (pH 7.4), 150 mM NaCl, 1 mM EDTA, 1 mM Na orthovanadate, 1 mM NaF and 0.25% (w/v) Triton X-100]. The lysates were adjusted with 63.75% (w/v) sucrose in Solution A to a final concentration of 42.5% (w/v) sucrose. Sucrose was added at the following concentrations and volumes on top of the lysate: 1.5 ml 42.5% (w/v) sucrose, 7 ml 35% (w/v) sucrose and 2 ml 5% (w/v) sucrose. The DRMs were separated for 16 h at 4°C and 38,000 rpm in an SW41 Ti rotor (Beckman Coulter). Fractions of 1 ml volume were collected after centrifugation and prepared for further analysis.

Cell lysates were prepared as described previously (Kissing et al., 2015). Proteins were separated on 10–12.5% SDS-PAGE gels and transferred to nitrocellulose membranes (Whatman, GE Healthcare, 10426994) by semi-dry blotting. The Amersham ECL Advanced Western Blotting Detection kit (GE Healthcare, RPN2135) was used for the detection of the antibody complexes.

HeLa cells were purchased from CLS Cell Lines Service and used at low passage numbers. HeLa and mouse embryonic fibroblast (MEF) cells were cultured in Dulbecco's modified Eagle's medium (DMEM, Gibco) supplemented with 10% (v/v) fetal bovine serum (FBS; Sigma-Aldrich) and antibiotics (penicillin-streptomycin) at 37°C in a humidified atmosphere at 5% CO₂. Cells were transfected with plasmid DNA using the stable cationic polymer polyethylenimine (PEI). DNA was incubated with PEI in a ratio of 1:2.5 for 20 min at room temperature. The solution was dripped onto the adherent cells in DMEM containing 10% FCS. The medium was changed 4 h after the addition of the PEI–plasmid complex.

The acyl resin-assisted capture assay (acyl-RAC) was performed as described previously (Forrester et al., 2011) with minor modifications. Cells were harvested in PBS containing cOmplete protease inhibitors (Roche/Sigma-Aldrich) and lysed in blocking buffer (100 mM HEPES, pH 7.5, 1.0 mM EDTA and 2.5% SDS) by sonification two times for 30 s each with a 30 s break. Free thiol groups were blocked for 4 h at 40°C by adding methyl methanethiosulfonate (MMTS) with a final concentration of 0.1%. Afterwards, proteins were precipitated by adding three volumes of ice-cold acetone and incubating at −20°C for 20 min followed by centrifugation at 16,000 g for 2 min. The pellets were washed five times with ice-cold 70% acetone and air dried. Protein pellets were dissolved in 300 µl binding buffer (100 mM HEPES, pH 7.5, 1.0 mM EDTA and 1% SDS) for up to 1 h at 40°C. Undissolved particles were pelleted for 1 min at 16,000 g. 30 µl of the solution was directly prepared as the input control for SDS-PAGE. 120 µl of the solution was incubated with 19 μl 2 M NaCl as a control or with 19 μl 2 M NH₂OH, pH 7.0 for cleaving palmitate residues to generate new free thiol groups. Proteins containing free thiol groups were captured for 3 h at room temperature by the addition of Thiol Sepharose 4B bead slurry in binding buffer (55 mg beads and 275 μl binding buffer were mixed to give 550 μl of bead slurry) and precipitated by centrifugation. The pellets were washed five times each in 1 ml binding buffer followed by analysis by SDS-PAGE.

The acyl-polyethylene glycol exchange (Acyl-PEG) assay was performed as described previously (Percher et al., 2016). Cells were harvested in PBS containing cOmplete protease inhibitor cocktail and lysed for 10 min at 37°C in 125 μl Buffer A [4% (w/v) SDS and 5 mM EDTA in PBS containing cOmplete protease inhibitors] followed by addition of 375 µl Buffer B (5 mM EDTA in PBS containing cOmplete protease inhibitors). The solution was sonicated three times each for 20 s. The protein concentration was adjusted to 0.5 μg/μl by the addition of Buffer D [1% (w/v) SDS in PBS containing cOmplete protease inhibitors]. Parts of this lysate were prepared for SDS-PAGE as an input control. To reduce possible disulfide bonds, 2×500 μl lysates were incubated for 30 min at room temperature with a final concentration of 10 mM Tris(2-carboxyethyl)phosphine (TCEP). Free thiol groups were blocked for 3 h at room temperature by the addition of N-ethylmaleimide (NEM) with a final concentration of 40 nM. Proteins were precipitated using acetone. The pellets were dissolved in 125 μl Buffer A and either incubated for 1 h at 37°C with Buffer T (1.33 M Tris-HCl, pH 7.0, 5 mM EDTA and 13 mM TCEP) as a control or with Buffer H (1.33 M hydroxylamine, pH 7.0, 5 mM EDTA and 13 mM TCEP) for cleavage of S-palmitoylation. Proteins were precipitated using acetone, and the pellets were dissolved for 1 h at 37°C in 110 μl Buffer C [2% (w/v) SDS in PBS containing cOmplete protease inhibitors]. The concentration was adjusted to 0.5 μg/μl by adding Buffer C, and the lysate was split into two 100 μl fractions. Either 10 μl of 200 mM NEM was added as control, or 10 μl of 200 mM methoxy-PEG (mPEG) was added to bind to the free thiol groups. Samples were incubated for 1 h at room temperature. The proteins were precipitated using acetone. The pellets were dissolved in Laemmli buffer, heat denatured for 2 min at 95°C and used for SDS-PAGE.

Click chemistry was performed as described previously (Martin, 2013). Cells were treated with 50 µM 17-ODYA in DMEM containing 10% FCS for different time periods to label palmitoylated proteins, followed by immunoprecipitation of the protein of interest. For pulse-chase labeling of palmitic acid and methionine, cells were starved for 1 h in starvation medium (DMEM high glucose, 200 µM L-cysteine, 4 mM glutamine, 1 mM sodium pyruvate and 5% dialyzed FCS) followed by a 2 h pulse in starvation medium containing 50 µM 17-ODYA and 50 µM L-azidohomoalanine (AHA; Click Chemistry Tools). The chase was performed for different time periods in DMEM containing 10% FCS, 2 mM L-methionine and 500 µM palmitic acid, followed by immunoprecipitation of the TMEM55B–HA with an antibody against HA. Cells were lysed in immunoprecipitation buffer [IP buffer; 0.1 M phosphate buffer, pH 7.4, 1% (v/v) Triton X-100 and cOmplete protease inhibitors] followed by sonification and precipitation of cell debris. The lysates were incubated with the primary antibodies at 4°C overnight. Magnetic Protein-G-coupled Dynabeads were washed three times with IP buffer followed by blocking overnight at 4°C in IP buffer containing 5% BSA. The magnetic beads were washed three times with IP buffer and resuspended in a suitable volume of IP buffer. The lysate–antibody mix was combined with the bead slurry and incubated for 1 h at room temperature. The beads were washed three times each for 10 min at room temperature in IP buffer followed by three washing steps, each of 5 min, in Reaction buffer (0.1 M phosphate buffer, pH 7.0). The click reaction of 17-ODYA (alkyne) with Cy3–Azide or Alexa Fluor 647–azide and of AHA (azide) with Alexa Fluor 488–alkyne was performed according to the manual of the CuAAC Biomolecule Reaction Buffer Kit (THPTA based) from Jena Bioscience. In the case of double labeling, click reactions were performed sequentially.

A membrane preparation was performed by homogenization of cell pellets via sonification in homogenization buffer (10 mM Tris-HCl, pH 7.5, 140 mM NaCl, 250 mM sucrose and cOmplete protease inhibitors in H₂O). Parts of the lysate were directly adjusted to a final concentration of 1% Triton X-100 (v/v) and prepared for SDS-PAGE. The rest was separated into a soluble fraction and a membrane fraction by 100,000 g centrifugation for 1 h. As a washing step, the membrane fraction was dissolved in 400 µl homogenization buffer and centrifuged at 100,000 g for 1 h. The pellet was dissolved in homogenization buffer containing 1% Triton X-100 (v/v) and prepared for SDS-PAGE.

Indirect immunofluorescence was performed as described previously (Gonzalez et al., 2018). Semi-confluent cells were grown on coverslips and fixed with 4% (w/v) paraformaldehyde (PFA; Roth) in PBS for 20 min at room temperature. After permeabilization of the cells with 0.2% (w/v) saponin (Roth) in PBS and quenching of PFA-induced fluorescence by adding 0.12% (w/v) glycine (Roth) in PBS containing 0.2% saponin, cells were blocked for 1 h with 10% FCS in PBS containing 0.2% saponin. Coverslips were incubated overnight at 4°C with the indicated primary antibodies diluted in PBS containing 0.2% saponin. The fluorophore-coupled secondary antibodies (Alexa Fluor 488, 594 and 647; Thermo Fisher Scientific) were added for 1 h at room temperature, followed by the embedding of the coverslips in 17% (w/v) Mowiol 4-88 mounting solution (Calbiochem) containing 20 mg/ml 1,4-diazabicyclo[2.2.2]octane (DABCO; Sigma-Aldrich) and 5 μg/ml 4′,6-diamidino-2-phenylindole (DAPI; Sigma-Aldrich) for nuclear staining. Images were analyzed with a Zeiss Axio Observer.Z1/7 Airyscan microscope equipped with a 40× lens (C-Apochromat 40×/1.20 W Korr). Image acquisition and processing were performed with the Zen 3.1 (blue edition) software (Carl Zeiss Microscopy GmbH, Germany).

The colocalization of proteins in immunofluorescence images was quantified by measuring the Pearson correlation coefficient or Manders overlap coefficient. Single cells were framed and analyzed using the Zen 3.1 (blue edition) software ‘Colocalization’ module (Carl Zeiss Microscopy GmbH, Germany). The corresponding thresholds were obtained based on single-antibody staining.

The distribution of LAMP2-positive organelles was measured using ImageJ software (NIH, Bethesda, MD, USA). Cells were framed, and LAMP2 signals were measured three times in each cell using the ‘plot profile’ tool between the MTOC and the plasma membrane.

Unless stated otherwise, a two-tailed unpaired t-test was performed using GraphPad Prism Software version 5.03. Values of P<0.05 were considered significant. Values are expressed as mean±s.e.m., and significance is designated as *P<0.05, **P<0.01, ***P<0.001 and ****P<0.0001.

We thank Sebastian Held, Sophie Reiher and Annika Detje for their excellent technical assistance. Laurence Abrami and Gisou van der Goot are acknowledged for discussions and help in establishing the palmitoylation assays.

The peer review history is available online at https://journals.biologists.com/jcs/article-lookup/doi/10.1242/jcs.258566