Lysosomal targeting of the ABC transporter TAPL is determined by membrane-localized charged residues

The human lysosomal polypeptide ABC transporter TAPL (ABC subfamily B member 9, ABCB9) transports 6–59-amino-acid-long polypeptides from the cytosol into lysosomes. The subcellular localization of TAPL depends solely on its N-terminal transmembrane domain, TMD0, which lacks conventional targeting sequences. However, the intracellular route and the molecular mechanisms that control TAPL localization remain unclear. Here, we delineated the route of TAPL to lysosomes and investigated the determinants of single trafficking steps. By synchronizing trafficking events by a retention using selective hooks (RUSH) assay and visualizing individual intermediate steps through immunostaining and confocal microscopy, we demonstrate that TAPL takes the direct route to lysosomes. We further identified conserved charged residues within TMD0 transmembrane helices that are essential for individual steps of lysosomal targeting. Substitutions of these residues retained TAPL in the endoplasmic reticulum (ER) or Golgi. We also observed that for release from the ER, a salt bridge between Asp-17 and Arg-57 is essential. An interactome analysis revealed that Yip1-interacting factor homolog B membrane-trafficking protein (YIF1B) interacts with TAPL. We also found that YIF1B is involved in ER-to-Golgi trafficking and interacts with TMD0 of TAPL via its transmembrane domain and that this interaction strongly depends on the newly identified salt bridge within TMD0. These results expand our knowledge about lysosomal trafficking of TAPL and the general function of extra transmembrane domains of ABC transporters.

membrane (PM), whereas the TMD0 itself is trafficked to lysosomes. If both parts are co-expressed, coreTAPL interacts noncovalently with TMD0, which leads to lysosomal localization (11).
For lysosomal membrane proteins, two intracellular biosynthetic routes are described (13,14). The direct route leads from the trans-Golgi network (TGN) via early endosomes and late endosomes to lysosomes. In contrast, the indirect route includes an intermediate trafficking step at the PM, subsequent endocytosis, and transport to early endosomes. Short linear targeting motifs determine individual trafficking steps. Most common are tyrosine (YXX⌽)-or dileucine ((D/E)XXXL(L/I))based motifs in a cytosolic region of the lysosomal membrane protein. These sequences are recognized by cytosolic adaptor proteins such as the five adaptor protein complexes (AP-1 to -5) or the monomeric GGAs (Golgi-localizing, ␥-adaptin ear domain homology, ARF-binding proteins). None of these consensus sequences associated with protein trafficking can be found in the TMD0 of TAPL. However, an increasing body of evidence clearly demonstrates that lysosomal localization can also be mediated by atypical targeting determinants ranging from anomalous leucine or tyrosine motifs to post-translational modifications (15).
In this study, we investigated the trafficking of TAPL to elucidate the intracellular route to lysosomes, especially with regard to a possible PM intermediate step. In mutational studies, we were able to determine conserved, charged residues within TMD0 that are essential for individual trafficking steps and thus form the targeting determinants of TAPL. Immunoprecipitation-MS (IP-MS) allowed us to identify YIF1B, a factor in ER-to-Golgi trafficking, as an interaction partner of TAPL involved in the targeting process. The interaction between the two proteins is mediated by the TMD of YIF1B and TMD0 of TAPL, with significantly reduced interaction if the salt bridge between two conserved, charged residues in TMD0 is disrupted.

Intracellular route of TAPL
To decipher the intracellular route of TAPL to lysosomes, we applied the retention using selective hooks (RUSH) assay (16), which enables synchronization of protein trafficking and thus visualization of intermediate steps. HeLa Kyoto cells were transiently transfected with a bicistronic plasmid. One gene coded for the invariant chain (Ii, CD74) containing an N-terminal, cytosolic ER retention signal and core streptavidin as hook, whereas the other coded for TAPL, fused C-terminally with a streptavidin-binding peptide (SBP) followed by eGFP (Fig. 1A). In the absence of biotin, TAPL colocalized with the hook protein, demonstrating efficient ER retention, and did not overlap with GM130 (cis-Golgi), EEA1 (early endosomes), and LAMP-1 (lysosomes) (Fig. S1A). By adding biotin, the interaction between TAPL and the hook protein was outcompeted, and TAPL was released and trafficked to lysosomes in a time-dependent manner. Within 15 min after the addition of biotin, TAPL accumulated in a perinuclear area partially overlapping with the cis-Golgi marker GM130, which represents the first trafficking step from ER to the Golgi (Fig. 1B and Fig. S1B).
After 45-60 min, TAPL colocalization with GM130 was reduced, and a fraction of TAPL was found in distinct vesicular structures stained with the early endosomal marker EEA1. Finally, a first overlap with the late endosomal and lysosomal marker LAMP-1 was observed after 60 -90 min. In this context, it should be noted that trafficking periods vary between cells dependent on cellular TAPL levels, with faster kinetics for low expressing cells. According to the observed localizations, TAPL takes an intracellular route from the ER, via Golgi and early endosomes to late endosomal and lysosomal compartments.
To exclude the possibility that we missed a short-lived PMtrafficking step that reflects the indirect route, we inhibited endocytosis by the dynamin inhibitor Dyngo-4a (17). HeLa Flp-In T-REx cells were used to allow stable, inducible expression of tapl under the control of the tetracycline-regulated promotor. To inhibit endocytosis during TAPL synthesis and trafficking, Dyngo-4a was added together with the inducer doxycycline. 2 h after induction, TAPL was detected in lysosomes, but no PM localization was observed (Fig. 1C). To verify the inhibitory function of Dyngo-4a, we analyzed the clathrinand dynamin-dependent endocytosis of the transferrin receptor, which can be triggered by the addition of transferrin (18,19). In the presence of Dyngo-4a, transferrin was not taken up by endocytosis, in contrast to the inactive control compound Dyngo-Ø, which did not interfere with endocytosis, as proven by immunofluorescence microscopy as well as by flow cytometry (Fig. S2). Thus, Dyngo-4a is capable of fully blocking dynamin-dependent endocytosis. Collectively, these data demonstrate that TAPL solely takes the direct route from the ER via Golgi and early endosomes to lysosomes.

Residues involved in intracellular trafficking
Intracellular trafficking steps are mediated by targeting motifs. However, none of the tyrosine-or dileucine-based motifs classically associated with lysosomal targeting are found in the cytosolic loops of TMD0, the targeting domain of TAPL. To elucidate how TMD0 mediates individual targeting steps along the intracellular route, we performed multiple sequence alignments of TAPL orthologues of phylogenetically distant species to identify conserved residues (Fig. S3).
As derived from a recently published secondary structure model of TMD0 (Fig. 2) (20), all eight conserved residues, three leucines and five charged residues found in species from humans to C. elegans, are localized in the transmembrane helices 1, 2, and 3. Because charged residues in the hydrophobic core of the membrane often accompany special functions, we focused our targeting studies on the charged residues Asp-17 in TMH1, Asp-49 and Arg-57 in TMH2, and Lys-100 in TMH3. The following studies were performed with HeLa and HEK cells, in which endogenous TAPL expression was detected at the mRNA level (21). However, at the protein level, endogenous TAPL could not be detected by either immunoblot or immunofluorescence microscopy, most likely due to low abundance of TAPL in these cells (7). Therefore, analysis of intracellular localization of TAPL mutants was carried out using stably transfected HeLa Flp-In T-REx cells. This system enabled mild overexpression of TAPL by means of low inducer concentrations and thus led to a more robust localization analysis across Lysosomal trafficking of TAPL the cell population than would be possible in transiently transfected cells.
First, we replaced the conserved, positively charged residues with alanine and the conserved, negatively charged residues with asparagine. Asp-49 was always substituted in combination with Asp-45, which is one helical turn upstream of Asp-49 in TMH2 and could compensate for charge deletion of Asp-49. Subcellular localization was examined by confocal immunofluorescence microscopy (Fig. 3A), and lysosomal localization was quantified by Pearson correlation of 50 single cells (Fig. 3B). The substitution of any of the four conserved, charged residues interfered with trafficking of TAPL. TAPL K100A still displayed a substantial overlap with LAMP-1 (Pearson correlation coefficient (PCC) ϭ 0.54 Ϯ 0.02) but was significantly reduced compared with WT TAPL (PCC ϭ 0.8 Ϯ 0.1). Further reduction in lysosomal localization was detected for TAPL D45N,D49N (PCC ϭ 0.24 Ϯ 0.03) and TAPL R57A (PCC ϭ 0.20 Ϯ 0.02). Most strikingly, TAPL D17N was not detected in LAMP-1positive compartments at all (PCC ϭ Ϫ0.23 Ϯ 0.03). TAPL D17N also showed no overlap with the cis-Golgi marker GM130 but strong colocalization with the ER marker protein disulfide-isomerase (Fig. S4). Because of this result, we replaced both conserved positive charged residues and observed that TAPL R57A,K100A (PCC ϭ Ϫ0.19 Ϯ 0.02) behaved like TAPL D17N , showing no lysosomal localization. TAPL mutants, which are unable to reach lysosomes due to these substitutions, are localized in the ER. A pronounced ER localization could mask a weak lysosomal population in case of TAPL D17N . Therefore, we treated the cells with cycloheximide (CHX) and chased them for 5 h to inhibit protein synthesis and allow trafficking (Fig. S5). Lysosomal localization of TAPL D45N,D49N (PCC ϭ 0.27 Ϯ 0.02) and TAPL K100A (PCC ϭ Figure 1. Intracellular route of TAPL. A, constructs for the RUSH assay. The invariant chain of MHC class II (luminal domain is not depicted), containing a cytosolic HA tag, streptavidin, and an ER retention signal, served as ER "hook" protein. TAPL was C-terminally fused to SBP and eGFP. To guarantee simultaneous expression, both genes were cloned in a bicistronic internal ribosome entry site (IRES) vector under the control of the cytomegalovirus promotor (pCMV) and separated by a synthetic intervening sequence (IVS) and an IRES. B, time-dependent trafficking of TAPL. HeLa Kyoto cells were transiently transfected with the bicistronic IRES vector coding for both RUSH constructs and incubated with biotin for the indicated times. Subsequently, cells were fixed and immunostained by ␣-HA (hook), ␣-GM130 (cis-Golgi), ␣-EEA1 (early endosomes), and ␣-LAMP-1 (late endosomes/lysosomes). White arrowheads at the zoom-in point to an overlap of TAPL and subcellular marker. Scale bar, 10 m; inset, 5 m. C, TAPL is not found on the plasma membrane. TAPL-FLAG expression was induced by doxycycline in stably transfected HeLa Flp-In T-REx cells in the presence of 30 M Dyngo-4a for 2 h. Cells were fixed and immunostained by ␣-FLAG (TAPL) and ␣-LAMP-1 (lysosomes) or ␣-MHC I (plasma membrane). Scale bars, 10 m and 5 m (inset).

Lysosomal trafficking of TAPL
0.59 Ϯ 0.02) was unaltered (Fig. S6). In contrast, lysosomal localization of TAPL R57A was significantly increased (PCC ϭ 0.45 Ϯ 0.02) by CHX treatment, implying a more challenging folding of TAPL R57A . Importantly, TAPL D17N and TAPL R57A,K100A did not show colocalization with LAMP-1, confirming that these TAPL variants were unable to reach lysosomes and were stuck in the ER. In summary, substitution of the conserved, charged residues within TMD0 impacts lysosomal targeting of TAPL, and therefore these residues determine its subcellular localization. Moreover, TAPL was trapped in the ER if the positive charges at Arg-57 and Lys-100 or the negative charge at Asp-17 were deleted. Because the deletion of these charged residues had the same effect, we hypothesize that Asp-17 forms an intramolecular salt bridge with one of these two positively charged residues to yield a trafficking-competent conformation. It should be noted that the deletion of the positive charge at position 57 had a stronger effect than the charge removal at position 100.
If the ER localization of TAPL D17N was caused by disruption of a salt bridge, the simultaneous inversion of the charges of Asp-17 and Arg-57 or Lys-100 should yield a trafficking competent TAPL variant. To prove any lysosomal localization, we again applied CHX for 5 h (Fig. 4). TAPL D17R (PCC ϭ Ϫ0.29 Ϯ 0.02), like TAPL D17N , showed no overlap with LAMP-1. TAPL R57D (PCC ϭ Ϫ0.32 Ϯ 0.02), in stark contrast to its neutral substitution TAPL R57A , also fully lacked colocalization with LAMP-1, whereas TAPL K100D (PCC ϭ 0.51 Ϯ 0.02) was trafficked as well as TAPL K100A to lysosomes. These findings indicated that Arg-57 is more relevant for yielding a traffickingcompetent conformation than Lys-100 and also possibly in closer proximity to Asp-17 for forming a salt bridge. This hypothesis was supported by the fact that TAPL D17R,R57D was indeed trafficking-competent and showed colocalization with lysosomes (PCC ϭ 0.49 Ϯ 0.02), strongly indicating a salt bridge between Asp-17 and Arg-57. In contrast, TAPL D17R,K100D (PCC ϭ Ϫ0.27 Ϯ 0.02) was not found in lysosomes, pointing to a minor role of Lys-100 in the context of Asp-17.

TAPL D17N is correctly folded
Defects in trafficking and especially ER retention can be caused not only by missing trafficking determinants but also by misfolding. Therefore, we tested correct folding of TMD0 wt and its mutants in vivo by the interaction with coreTAPL. We transiently co-expressed TMD0 variants containing a C-terminal FLAG tag and coreTAPL in HEK293T cells and performed co-immunoprecipitation with an ␣-FLAG antibody. Core-TAPL was precipitated together with all TMD0 variants but not in the absence of TMD0, demonstrating correct folding of all TMD0 mutants (Fig. 6A). Because substituting Asp-17 showed the strongest effects in subcellular trafficking and ER retention, we addressed the folding state of TMD0 D17N and TAPL D17N in more detail. First, we evaluated folding of TMD0 D17N by solution NMR. 15 N-Labeled variants of TMD0 wt (cf-TMD0) and TMD0 D17N (cf-TMD0 D17N ) were synthesized by cell-free (cf) expression, and 15 N, 1 H BEST-TROSY spectra were recorded (Fig. 6B). The well-resolved peak distribution and peak overlaps between both spectra indicated that cf-TMD0 D17N is a folded protein with a conformation similar to cf-TMD0. Next, the localization of TMD0 in stably transfected HeLa Flp-In T-REx cells was analyzed (Fig. 6C). TMD0 wt significantly overlapped with the lysosomal marker LAMP-1, whereas TMD0 D17N was not localized in lysosomes. Adding the cytosolic, tyrosinebased lysosomal targeting sequence of LAMP-2C at the C terminus of TMD0 D17N partially restored lysosomal localization (Fig. 6D). Therefore, ER retention must be due to the absence of a lysosomal targeting determinant in TMD0 D17N , whereas misfolding can be excluded. Lysosomal localization of full-length TAPL D17N was also recovered if co-expressed with TAPL wt in HeLa Kyoto cells, demonstrating that TAPL D17N dimerizes with TAPL wt and is not retained in the ER (Fig. 7, A and B). Finally, correct folding and dimerization of TAPL D17N was demonstrated by the comparable ATP-dependent peptide transport activity of TAPL wt and TAPL D17N in crude membranes derived from HeLa Flp-In T-REx cells (Fig. 7C). Taking these results together, we conclude that altered lysosomal trafficking and ER or Golgi retention of TAPL mutants are not caused by misfolding but are due to the absence of a targeting determinant. Figure 2. Conserved, charged residues within TMD0. Conserved residues of human, mouse, chicken, zebrafish, and sea lamprey TAPL, as well as C. elegans HAF-4 and HAF-9 (see Fig. S3), are depicted in a TMD0 secondary structure model (20). Charged residues within TMH1-3 investigated in this study are highlighted by a boldface border.

Lysosomal trafficking of TAPL Trafficking chaperones
Because the impaired ER to Golgi trafficking of TAPL D17N and TAPL D17R is not due to misfolding, we hypothesized that the identified salt bridge stabilizes a specific conformation of TMD0 that allows transient interactions with trafficking chaperones. The only known interaction partners for TAPL are LAMP-1 and -2B, identified by tandem affinity purification of digitonin-solubilized membranes. However, these proteins are negligible for TAPL trafficking, because in double LAMP-1/2 knockout cells, TAPL is correctly targeted to lysosomes (12). Therefore, we aimed to identify components of the trafficking interactome that are assumed to interact with TAPL only transiently and are consequently associated with TAPL in low abundance. Thus, we performed a fast, single-step purification followed by MS and applied Comparative Proteomic Analysis Suite (CompPASS) to enable identification of high-confidence candidate interacting proteins (HCIPs) (22). In short, Comp-PASS ranks identified proteins based on peptide abundance and occurrence in predetermined proteomic data sets. A protein frequently found across several samples in the data sets is ranked lower than a unique one in a given sample. HeLa Flp-In T-REx cells nontransfected or stably expressing TAPL-HA or coreTAPL-HA were solubilized by the detergent Nonidet P-40 (NP40), purified in a single step with an ␣-HA antibody, and analyzed by tandem MS. CompPASS analysis revealed 14 HCIPs of TAPL that were absent in coreTAPL samples or samples derived from untransfected cells. Among these HCIPs (Fig.  8A), the Yip1-interacting factor homologue B (YIF1B) is the only one that is a transmembrane protein and also involved in trafficking of a transmembrane protein. YIF1B is a member of the FinGER protein family and, together with YIF1A, the human homologues of yeast YPT-interacting protein 1 (YIP1). In prior studies, both proteins were implicated in ER-to-Golgi protein trafficking (23,24). YIF1B is composed of an N-terminal cytosolic domain, followed by a C-terminal TMD composed of five predicted TMHs (Fig. S11A). First, we verified the interaction of TAPL with YIF1B by co-immunoprecipitation. Endogenous YIF1B was co-immunoprecipitated with C-terminally FLAG-tagged TAPL and TMD0 stably expressed in HeLa Flp-In T-REx cells, but not with coreTAPL ( Fig. 8B). Importantly, TMD0 was pulled down in a reverse immunoprecipitation via YIF1B if transiently co-expressed in HEK293T cells  Table S1.

Lysosomal trafficking of TAPL
( Fig. 8C), demonstrating that YIF1B interacts with the TMD0 of TAPL. The TMD of YIF1B, devoid of its cytosolic tail, was still able to interact with TAPL in transiently co-transfected HEK293T cells (Fig. 8D), validating that the interaction occurs via the TMDs of both proteins.
Next, we investigated whether YIF1B is essential for TAPL targeting. HeLa Flp-In T-REx YIF1B knockout clones were created by CRISPR/Cas9 using two separate sgRNAs targeting exon 2 or exon 3. The knockout was confirmed by immunoblot and sequencing of the targeted exons (Fig. S8A). In the YIF1B knockout cell lines, TAPL showed no alteration in localization between 2 and 24 h after induction of expression compared with the original cell line (Figs. S8B and S9A). Next, YIF1B knockout cells were transfected with the aforementioned vector for the RUSH assay, yielding no differences in the early trafficking process of TAPL (Fig. S9B). This demonstrates that YIF1B is not essential for TAPL targeting and its absence does not impact trafficking kinetics. Redundancy in subcellular trafficking is common; therefore, we investigated whether YIF1A, with a sequence identity of 53% to YIF1B, can compensate for YIF1B deficiency (Fig. S10). Nevertheless, YIF1B, but not YIF1A, was co-immunoprecipitated with TMD0 if transiently co-expressed in HEK293T cells, consistent with the MS results.
To assess whether YIF1B plays a role in TAPL targeting, even if it is not an essential factor, we aimed to outcompete alternative binding partners as well as endogenous YIF1B by overexpression of truncated YIF1B (TMD YIF1B ). If we assume that YIF1B is involved in ER to Golgi targeting of TAPL and harbors its targeting determinant in its cytosolic domain, the overexpression of TMD YIF1B should influence TAPL targeting. In contrast to YIF1B, which is found predominantly in the ER, TMD YIF1B is strongly enriched in the cis-Golgi (Fig. S11, B and C), implying that a targeting or retrieval signal is indeed localized in the cytosolic domain of YIF1B. Upon transient co-expression in HeLa Kyoto cells, TAPL showed a reduction in lysosomal localization dependent on the TMD YIF1B level (Fig. 9, A, B, and E), with more pronounced effects in cells with higher TMD YIF1B levels. TAPL was strongly enriched in lysosomes of cells not transfected by YIF1B or TMD YIF1B . Overexpression of TMD YIF1B significantly decreased colocalization of TAPL with LAMP-1, compared with cells overexpressing full-length YIF1B (Fig. 9, D and E). Interestingly, TAPL colocalized strongly with TMD YIF1B in cis-Golgi (Fig. 9C).
Finally, we tested our initial hypothesis that D17N substitution in TMD0 affects the interaction with YIF1B. HEK293T cells were transiently transfected with TMD0-FLAG or TMD0 D17N -FLAG, and immunoprecipitation was carried out using ␣-FLAG antibody. The amount of endogenous YIF1B pulled down by TMD0 D17N was reduced by a factor of 4 compared with TMD0 (Fig. 9, F and G), based on the four individual immunoprecipitations we performed (Fig. S12). This indicates a conformational change of the TMD0 interaction interface by disruption of the salt bridge between Asp-17 and Arg-57.  Table S1.

Lysosomal trafficking of TAPL
In summary, the newly identified transient interaction partner YIF1B interacts via its TMD YIF1B directly or indirectly with TMD0. Although YIF1B is not essential for lysosomal targeting of TAPL, overexpression of TMD YIF1B strongly interferes with the correct localization of TAPL. Interaction of both proteins is strongly dependent on the conserved Asp-17 within TMD0.

Discussion
In this study, we demonstrated that TAPL takes the direct route from the Golgi to lysosomes. Further, we found four conserved, charged amino acids in the transmembrane helices of TMD0, which affect the subcellular localization of TAPL and, therefore, represent atypical targeting determinants. Additionally, we identified YIF1B as a new interaction partner of TAPL, being the first piece of the, presumably more complex, trafficking interactome.
By synchronizing intracellular trafficking using the RUSH assay, we proved that TAPL chooses the direct route to lysosomes via the Golgi and early endosomes but not the indirect route via the PM. The lack of PM localization is supported by the use of the dynamin inhibitor Dyngo-4a and missing surface biotinylation (11). Furthermore, TAPL was not detected on the PM in bone marrow-derived dendritic cells (8). This is in contrast to lysosome-localized ABCB6, which takes the indirect route and can be accumulated at the PM by adding Dyngo-4a (25). Although the trafficking kinetics should be analyzed with caution due to their dependence on cells with high TAPL-SBP-eGFP levels for detection, rough estimates can be made based on our results. ER-to-Golgi trafficking occurs within 15-30 min after the addition of biotin. After 30 -60 min, Golgi localization decreases, and early endosome localization is detectable. To reach lysosomes, TAPL needs up to 90 min. This time frame is

Lysosomal trafficking of TAPL
in line with data from inducible HeLa Flp-In T-REx cells used in this study, where TAPL co-localized with LAMP-1 2 h after induction of expression. Moreover, the trafficking time of TAPL is in good agreement with that of other proteins in the secretory pathway and lysosomal membrane proteins (26 -28).
Atypical cytosolic targeting determinants, which differ from the consensus tyrosine or dileucine motifs, can be found, for example, in the Longin domain of VAMP7 (29) as PY motifs ((L/P)PXY) in LAPTM5 (30) or as "extended acidic dileucine signal" in TMEM106B (31). For some proteins, these motifs are, by themselves, not sufficient to mediate lysosomal localization but are dependent on additional motifs. One example is the lysosomal targeting of TMEM106B, which additionally depends on N-glycosylation of its luminal loops (32). Another type of atypical sorting determinants is TMH parameters like their length, amount, and positioning of hydrophobic or charged amino acids or interplay with lipids (33). In the case of TAPL, charged amino acids within the TMHs of TMD0 determine its subcellular localization. Charged residues within TMHs are often essential for protein function and complex assembly. For instance, correct complex assembly and therefore PM localization of the T-cell receptor depends on charged residues in the TMHs (34,35). In the case of TAPL D17N , full ER retention is not due to an incomplete assembly of TAPL homodimers because TAPL D17N is fully active in peptide transport. By means of immunoprecipitations, NMR analysis of  Table S1.

Lysosomal trafficking of TAPL
cf-TMD0 D17N , and restoration of lysosomal localization of TMD0 D17N by supplying the targeting motif of LAMP-2C, we demonstrated that ER retention was also not caused by misfolding. In conclusion, whereas Asp-17 is indispensable for targeting, it is not essential for the structural integrity of TAPL or TMD0.
YIF1B shuttles between ER, ERGIC, and Golgi (23). Furthermore, it is involved in lysosomal trafficking of TAPL, because overexpression of TMD YIF1B , which accumulates in the Golgi, has a strong impact on TAPL targeting. Moreover, TMD0 of TAPL interacts transiently with TMD YIF1B , and this interaction is strongly weakened by mutating Asp-17 of TAPL. Therefore, we postulate that subcellular targeting of TAPL depends on its interactions with other transmembrane proteins that link it to components of the trafficking machinery. Because YIF1B is not implicated in trafficking steps beyond the Golgi, other transmembrane proteins probably interact with TAPL at the TGN and endosomes to mediate further trafficking steps. This would coincide with the observed strong Golgi retention of TAPL upon Asp-45, Asp-49, and Lys-100 substitution. Such a piggyback mechanism is described for the ABC transporter ABCD4, which interacts with the classical tyrosine-based motif exhibiting LMBD1 (36). Similarly, the endosomal and lysosomal localization of MHC class II is the result of its interaction with the invariant chain (37,38), which blocks premature peptide binding (39) and also exhibits a dileucine motif (40).
YIF1B, implicated in anterograde transport (23), plays a role in 5-HT 1A receptor trafficking (41). However, this cell surface-localized protein interacts with YIF1B via its cytosolic tail and not its TMD. In contrast, the closely related YIF1A, which shares high sequence identity in its TMD with YIF1B, interacts with VAPB via its transmembrane domain (24), indicating that these proteins may have multiple bind-ing sites for different proteins. Despite their similarities in the TMD, YIF1A is unable to interact with TMD0, as demonstrated by immunoprecipitation.
CRISPR/Cas9-mediated knockout of YIF1B in our stable cell line system did not alter TAPL localization and is therefore apparently not an essential factor for targeting. Thus, other transmembrane proteins also have to be considered for TAPL ER to Golgi trafficking. Redundant trafficking pathways and mechanisms are observed for several proteins. For instance, three different trafficking routes were described for LAMP-1/2. Usually, these proteins are trafficked in a clathrin-dependent manner. They can either take the direct route (42), mediated by AP-1 (43) and AP-3 (44), or the indirect route, mediated by AP-2 (45). Additionally, trafficking from the TGN to late endosomes by non-clathrin-coated "LAMP carriers" was described (46,47).
Of the four conserved charged residues, Asp-17 substitutions revealed the strongest impact on TAPL localization. Mutation of Asp-17 retained TAPL completely in the ER, whereas single substitutions of the other conserved amino acids allowed partial lysosomal localization. Intriguingly, the other two members of the TAP family, TAP1 and TAP2, which form heterodimers, also have a conserved aspartate in their first and a conserved arginine in their second TMH (48). It was shown that this aspartate is essential for an intermolecular salt bridge with tapasin. This interaction is placing TAP in the peptide loading complex, which is responsible for MHC class I loading. Based on molecular dynamics simulations, this aspartate forms an intramolecular salt bridge with the arginine, which is replaced by an intermolecular salt bridge with Lys-428 in the transmembrane helix of tapasin, whereas the arginine side chain snorkels to the polar headgroups of the lipid bilayer to avoid an unfavorable uncompensated charge in the TMD. Based on the strong

Lysosomal trafficking of TAPL
impact of charge inversion substitutions of Asp-17 and Arg-57 on TAPL trafficking, as well as the partial rescue of lysosomal localization by simultaneous charge reversal of both, we demonstrated that Asp-17 in TAPL does not form an intermolecular salt bridge but an intramolecular one with Arg-57. However, deletion of the charge at Arg-57 was partially compensated by Lys-100 because R57A substitution, unlike D17N, did not abolish lysosomal localization. This indicates that Asp-17 and Lys-100 are able to form a salt bridge but yield a conformation that is less trafficking-competent than the WT conformation. Charge inversion of Asp-17 and Lys-100 (TAPL D17R,K100D ) does not produce a trafficking-competent conformation, possibly due to the positive charge of Arg-57, which would be in close proximity to D17R, causing destabilizing effects. In line with the proposed salt bridges, deletion of both conserved positive charges again completely prohibited lysosomal targeting. Summing up our data, we propose that disruption of the salt bridge between Asp-17 and Arg-57 induces subtle conformational changes in the binding interface. These changes weaken the interaction with YIF1B and possibly other trafficking chaperones, resulting in increased ER localization of TAPL.
Interestingly, substitution and charge reversals of Asp-45, Asp-49, and Lys-100 strongly increased Golgi localization and also abolished lysosomal localization, further highlighting the importance of the conserved, charged residues. Whether a salt bridge is formed between Lys-100 and the negative cluster of Asp-45/Asp-49 cannot be derived from these experiments. These substitutions seem to impact the second trafficking step of TAPL from the Golgi to early endosomes, indicating an involvement in a second protein interaction interface, separate from that affected by Asp-17 and Arg-57 substitutions.
In summary, TAPL targeting is determined by conserved, charged residues in its TMD0, forming intramolecular salt bridges. Disruption of these salt bridges arrests TAPL either in . HeLa Flp-In T-REx cells stably expressing FLAG-tagged variants of TAPL wt , coreTAPL or TMD0 wt were solubilized by NP40, immunoprecipitated by ␣-FLAG antibody (FLAG), and immunoblotted. Specificity of immunoprecipitation was verified by an IgG isotype control (IC) and using nontransfected cells. The contrast in the right part of ␣-YIF1B immunoblot was enhanced due to low signal intensity. C, reverse co-immunoprecipitation of YIF1B and TMD0. HEK293T cells transiently transfected with TMD0 wt -FLAG, HA-YIF1B, or a combination of both were lysed by NP40. YIF1B was immunoprecipitated by ␣-HA antibody (HA) and immunoblotted. Specificity of immunoprecipitation was verified by an IgG isotype control and using cells transfected with a plasmid devoid of a gene of interest (Ϫ/Ϫ). D, TMD YIF1B interacts with TAPL. HEK293T cells transiently transfected with TAPL wt -FLAG alone or in combination with HA-YIF1B (fl.), HA-cyt YIF1B (cyt.), or HA-TMD YIF1B (TMD) were solubilized by NP40, and immunoprecipitation was performed by ␣-FLAG antibody. Specificity of immunoprecipitation was verified by an IgG isotype control and using cells transfected with a plasmid devoid of a gene of interest (Ϫ/Ϫ).

Lysosomal trafficking of TAPL
the ER or Golgi by rendering TMD0 trafficking-incompetent. This inability to travel along the direct route to lysosomes is due to a diminished capability to interact with interaction partners involved in trafficking as exemplified with YIF1B.

Peptides and antibodies
Primary and secondary antibodies used in this study are listed in Table S2. ␣-YIF1B was a kind gift from Michèle Darmon (Centre de Psychiatrie et Neuroscience, Paris, France) (41). ␣-C8 was obtained through the AIDS Research and Reference Reagent Program, Division of AIDS, NIAID, National Institutes of Health: Chessie 8 from Dr. George Lewis (49). HA peptide was purchased from Sigma-Aldrich/Merck and Charité (Berlin, Germany). RRYQNSTCL peptide (Charité) was labeled with 5-iodoactamidofluorescein (I9271; Sigma-Aldrich/Merck) as described elsewhere (50).

Cloning
Cloning primers, corresponding templates, plasmids, and cloning techniques are listed in Table S3. The Q5 site-directed mutagenesis kit (New England Biolabs) was used according to the manufacturer's instructions. Restriction enzyme cloning was performed by digesting backbone and PCR product with the listed restriction enzymes (Thermo Fisher Scientific or New England Biolabs). Str-Ii_VSVGwt-SBP-EGFP (16) was a gift from Franck Perez (Institute Curie, Paris, France), and pSpCas9(BB)-2A-Puro (PX459) version 2.0 was a gift from Feng Zhang (Addgene plasmid 62988) (51). HA-YIF1A and HA-YIF1B genes were synthesized by Eurofins Genomics. All constructs were verified by DNA sequencing.

DNA extraction
To extract DNA from mammalian cells for PCR, 0.1 ϫ 10 6 cells were transferred in a tube, washed three times with PBS  Table S1. F, Asp-17 of TAPL is important for YIF1B interaction. HEK293T cells transiently transfected with TMD0 wt -FLAG or TMD0 D17N -FLAG were lysed by NP40. TMD0 variants were immunoprecipitated by ␣-FLAG antibody (FLAG) and immunoblotted using ␣-FLAG and ␣-YIF1B. Specificity of immunoprecipitation was verified by an IgG isotype control (IC) and using cells transfected with a plasmid devoid of a gene of interest (Vector). G, YIF1B and TMD0-FLAG immunoblot signals from four independent experiments were quantified, and the TMD0 wt -FLAG-normalized ratio of YIF1B to TMD0 is plotted. Individual immunoblots used for quantification are shown in Fig. S12. Individual data points are shown as squares, and mean value and corresponding S.D. (error bars) are depicted in red. (4.3 mM Na 2 HPO 4 , 1.47 mM KH 2 PO 4 , 137 mM NaCl, 2.7 mM KCl, pH 7.4), and frozen at Ϫ20°C. Pellet was resuspended in TE buffer (10 mM Tris-HCl, 1 mM EDTA, pH 8.0), incubated at 80°C for 10 min, and then incubated at Ϫ20°C for 5 min. Proteinase K (0.5 g/l, Thermo Fisher Scientific) was added and incubated at 50°C for 30 min and at 80°C for 10 min. Samples were centrifuged at 10,000 ϫ g for 2 min, and 1 l of supernatant was used as template for PCR.

Cell culture
HeLa Kyoto, HEK293T, and HeLa Flp-In T-REx cells were cultured at 37°C, 5% CO 2 , and 95% humidity. HeLa Kyoto and HEK293T were cultured in Dulbecco's modified Eagle's medium (DMEM) (Gibco/Thermo Fisher Scientific) with 10% fetal calf serum (FCS; Capricorn Scientific). For culturing stable cell lines of the HeLa Flp-In T-REx system, DMEM with 10% tetracycline-free FCS (Bio&Sell) was used. Selection of stable HeLa Flp-In T-REx cells was performed with 200 g/ml hygromycin B (Thermo Fisher Scientific) in combination with 2 g/ml blasticidin S HCl (Thermo Fisher Scientific). Selection of transiently transfected HeLa Flp-In T-REx cells was performed with 1 g/ml puromycin (Thermo Fisher Scientific). Induction of expression in stable HeLa Flp-In T-REx cells was performed with 1 ng/ml to 5 g/ml doxycycline (D9891; Sigma-Aldrich/ Merck), depending on the gene of interest and application. For CHX (2112, Cell Signaling Technology) treatment, cells were induced for 19 h and then treated with 25 g/ml CHX for an additional 5 h. All cells were tested regularly for mycoplasma contamination.

Transfection
Transfections of HeLa Kyoto, HeLa Flp-In T-REx, and HEK293T Flp-In T-REx cells were performed with Lipofectamine 2000 (Thermo Fisher Scientific) in a 1:2.5 ratio (g of DNA/l of transfection reagent). HEK293T cells were transfected using 18 mM polyethyleneimine (PEI) stock solution in a 1:5 ratio (g of DNA/l of transfection reagent). We used 0.8 g of DNA/well in a 24-well plate, 2.5 g DNA/well in a 6-well plate, and 15 g for 10-cm dishes. DNA and Lipofectamine 2000 or PEI were diluted in Opti-MEM I medium (Thermo Fisher Scientific), incubated for 5 min, mixed, and incubated for 15 min prior to transfection. Cells were seeded 6 -20 h prior to transfection to ensure complete adhesion. Lipofectamine 2000 and PEI transfections were performed at 80 -95 and 40% confluence, respectively.

CRISPR/Cas9
Exon and intron sequences were obtained from the Ensembl Genome Browser (http://www.ensembl.org). 3 Exon 2 or 3 was used as an input sequence for sgRNA design. sgRNAs were picked according to their score and potential off-target effects (Table S3). Corresponding DNA was ordered from Eurofins Genomics. SpCas9 -2A-Puro version 2.0 was used for transfection of HeLa-Flp-In T-REx cells. Selection was performed using puromycin. Post-selection limited dilution cloning was per-formed to obtain monoclonal cells. DNA was extracted and used for PCR. PCR products were separated by 2% agarose gel, extracted, purified, and sequenced. Knockout was further confirmed by immunoblot (52,53).

Confocal imaging
The intracellular route of TAPL was imaged using a confocal laser-scanning TCS SP5 microscope (Leica) with a plan-apochromat ϫ63/1.4 oil differential interference-contrast objective. Sequential settings for scans were used. For excitation, the following laser lines were used: 405 nm (diode laser) for 4Ј,6diamidino-2-phenylindole, 488 nm (argon laser) for eGFP, 561 nm (diode-pumped solid-state laser) for Cy3, and 633 nm (helium-neon laser) for Alexa Fluor 647. Intensities of channels were adjusted over the whole image for better visualization of overlap and exported by Leica Application Suite X.
All other experiments were performed using a confocal laserscanning LSM880 microscope (Zeiss) with a plan-apochromat ϫ63/1.4 oil differential interference-contrast M27 objective. Sequential settings for scans were used. For excitation, the following laser lines were used: 488 nm (argon laser) for Alexa Fluor 488 and eGFP, 543 nm (helium-neon laser) for Cy3, 594 nm (helium-neon laser) for Alexa Fluor 568, and 633 nm (helium-neon laser) for Alexa Fluor 647. Intensities of channels were adjusted over the whole image for better visualization of overlap and exported by Zen blue (version 2.3 lite, Zeiss).

Quantification of colocalization
Unprocessed micrographs were exported by ZEN blue as .tiff files without compression. Images were loaded into Fiji/ImageJ (57,58). Then 50 -90 individual cells were outlined as regions of interest, and the PCC was determined by Coloc 2. Automated threshold selection (Costes) was performed. PCC values above threshold were plotted by GraphPad Prism version 5 (GraphPad Software) as a vertical scatter plot with means and S.E. p Ͻ 0.05 was considered significant. For intensity plots, intensity along the line was measured by Zen blue, normalized, and plotted by GraphPad Prism 5.

Statistical analysis
Statistical significance analyses were performed using GraphPad Prism 5 (GraphPad Software). For pairwise comparison statistics, unpaired two-tailed Student's t tests were applied. For multiple comparison of flow cytometry data, oneway analysis of variance with post hoc Tukey's test was used. For multiple comparisons of PCC values obtained from colocalization analysis, the nonparametric Kruskal-Wallis test with post hoc Dunn's test was used. Test results are listed in Table S1.

Deconvolution of images
All microscopy images, except maximum intensity z-projections, used for the figures were deconvoluted. Point-spread functions (PSFs) were generated by PSF Generator (59) for Fiji/ ImageJ, using the Born-Wolf algorithm. Images were deconvoluted using DeconvolutionLab2 (60) for Fiji/Image2, using the Richardson-Lucy algorithm with 50 iterations.

Synthesis of Dyngo compounds
Dyngo-4a and Dyngo-Ø were synthesized as described recently (61) using the reflux method. Products were dissolved as 30 mM stock solutions in 100% DMSO (D2650, Sigma-Aldrich/Merck) and stored at Ϫ20°C. Compounds were diluted in DMEM without FCS to 30 M directly before usage.

Transferrin uptake assay
Transferrin conjugated with Alexa Fluor 647 (T23366, Thermo Fisher Scientific) was reconstituted in H 2 O at 5 mg/ml. Cells were washed with Opti-MEM I (without FCS) twice and pre-incubated with 30 M Dyngo-4a, Dyngo-Ø, or DMSO in Opti-MEM I for 30 min at 37°C. Then cells were incubated with 25 g/ml transferrin AF647 and 30 M Dyngo-4a or Dyngo-Ø in Opti-MEM I at 37°C for 30 min. Afterward, cells were trypsinized, mixed with DMEM plus 10% FCS, centrifuged at 600 ϫ g at 4°C, and washed once with cold PBS and twice with cold FACS buffer (1% FCS in PBS). To assess background signal, a sample was trypsinized after the pre-incubation, mixed with DMEM plus 10% FCS, and centrifuged at 600 ϫ g at 4°C. Subsequently, cells were washed twice with cold PBS, incubated with transferrin AF647 and Dyngo-4a in Opti-MEM I for 30 min on ice, and then washed once with cold PBS and twice with cold FACS buffer. Cells were fixed in 1% paraformaldehyde in FACS buffer prior to analysis. All samples were analyzed by FACS Celesta (BD Biosciences) with ex/em 633/670 nm. Data analysis was performed with FlowJo version 10 to determine the mean fluorescence intensity. The statistics are based on three separately prepared samples. Immunoprecipitation 50 l of Dynabeads M-280 sheep anti-mouse or sheep antirabbit (Life Technologies) were washed with 3 ml of IP buffer (20 mM Tris, 150 mM NaCl, 5 mM MgCl 2 , pH 7.4) supplemented with 0.1% BSA. Washed beads were incubated in 800 l of IP buffer with either 2.5 l of anti-FLAG M2, 2.5 l of anti-HA, or 5 l of mouse monoclonal IgG1-as isotype control overnight at 4°C and washed with 3 ml of IP buffer. 8.8 ϫ 10 6 cells were used for each individual IP. Harvested cells were stored at Ϫ80°C, thawed on ice, and solubilized using IP buffer containing 0.5% NP40 (Sigma-Aldrich/Merck) or 1% digitonin (Millipore/Merck) and 1ϫ HP protease inhibitor mix (Serva) for 1 h at 4°C. Lysate was centrifuged at 20,000 ϫ g for 20 min at 4°C. Supernatant was incubated with antibody-coated beads for 2 h at 4°C and washed with 3 ml of IP buffer containing 0.05% NP40 or 0.1% digitonin. Proteins were eluted by incubation at 90°C for 20 min in elution buffer containing 2ϫ SDS sample buffer (125 mM Tris, 4% (w/v) SDS, 4 mM EDTA, 0.02% (w/v) bromphenol blue, 20% (v/v) glycerol, pH 6.8) without reducing agents in 25 mM NaOAc, pH 5.0. Samples were analyzed by 10% SDS-PAGE or 10% Tricine-SDS-PAGE followed by immunoblotting.

Peptide transport
40 ϫ 10 6 HeLa Flp-In T-Rex cells were washed with 60 ml of PBS, harvested by centrifugation for 10 min at 1000 ϫ g and 4°C, and stored at Ϫ80°C. For membrane preparation, cells were thawed on ice, resuspended in Tris buffer (20 mM Tris, 1 mM DTT, 2.5 mM benzamidine, 1 mM PMSF, pH 7.4) and disrupted by Dounce homogenization (40 times) with a tissue grinder (Wheaton). Sucrose was added to a final concentration of 250 mM, and cells were Dounce-homogenized again (10 times). Cell lysate was sequentially centrifuged for 4 min at 200 ϫ g and 8 min at 700 ϫ g at 4°C. Subsequently, supernatant was pelleted for 45 min at 100,000 ϫ g and 4°C. Crude membranes were resuspended in PBS, and aliquots were snap-frozen Lysosomal trafficking of TAPL in liquid nitrogen and stored at Ϫ80°C. Protein concentration in membranes was determined by a Bradford assay (Thermo Fisher Scientific). For peptide transport, crude membranes (120 g of protein) were incubated in 50 l of PBS supplemented with 3 mM MgCl 2 and 3 M NST-F (RRYQNSTC Fluorescein L) peptide. Peptide transport was started by the addition of 3 mM ATP for 12 min at 37°C. Transport was stopped with 1 ml of ice-cold stop buffer (PBS supplemented with 10 mM EDTA). Membranes were collected on microfilter plates (MultiScreen plates, Durapore Membrane, 1.2-m pore size; Merck) preincubated with 0.3% (w/v) polyethyleneimine. Filters were washed three times with 250 l of ice-cold stop buffer and incubated at room temperature for 10 min in lysis buffer (PBS, 1% SDS, pH 7.5). The amount of transported peptides was analyzed on a fluorescence plate reader (CLARIOstar, BMG LABTECH) at ex/em ϭ 485/520 nm. ATP-dependent peptide accumulation in crude membranes from nontransfected cells was subtracted, and transport was normalized to TAPL amount determined by immunoblot.

MS-based proteomics
Anti-HA-immunoprecipitation was performed as described previously (22,(63)(64)(65). Briefly, expression of TAPL-HA and coreTAPL-HA was induced by the addition of 4 g/ml doxycycline for 24 h in HeLa Flp-In T-REx cells. Parental nontransfected HeLa Flp-In T-REx cells were used as negative control. For each sample, 6.4 ϫ 10 7 cells were harvested, frozen in liquid nitrogen, and stored at Ϫ80°C. Cells were lysed in 3 ml of MCLB buffer (50 mM Tris, 150 mM NaCl, 0.5% NP40, pH 7.4) supplemented with cOmplete EDTA-free protease inhibitor tablets (Roche Applied Science, 2 tablets for 50 ml of MCLB buffer) on ice for 30 min. Lysate was cleared by centrifugation (18,000 ϫ g, 10 min, 4°C) followed by filtering through a 0.45-m spin filter (Millipore). For isolation of HA-tagged proteins, cell lysate was incubated overnight at 4°C with 60 l of equilibrated ␣-HA-agarose beads (Sigma-Aldrich/Merck). Subsequently, beads were washed four times with 1 ml of MCLB and 1 ml of PBS, respectively. 50 l of 250 g/ml HA peptide was added to dry beads and incubated for 30 min at room temperature. Elution was repeated twice, obtaining a final volume of 150 l. Proteins were precipitated with 20% TCA, resuspended in 20 l of 50 mM ammonium bicarbonate, pH 8.0, containing 10% acetonitrile and 750 ng trypsin (Promega), and incubated for 4 h at 37°C. Desalting was performed using stage tips. Samples were analyzed as technical duplicates on a LTQ Velos mass spectrometer (Thermo Fisher Scientific), and spectra were identified by Sequest searches as described previously (66). For CompPASS analysis, the identified peptides were compared with IP-MS data of 99 unrelated bait proteins, which were previously processed using the same experimental conditions (22), to obtain weighted and normalized D scores (WD N score). Proteins with WD N Ն 1.0 and APSM (average peptide spectral matches) Ն 4 were considered as high-confidence candidate interacting proteins. To account for co-purifying (background) proteins in HeLa cells and coreTAPL-binding proteins, proteins found in these two IP conditions were subtracted from the list of TAPL HCIPs. The MS proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE (67) partner repository with the data set identifier PXD010989.