A haploid genetic screen identifies the major facilitator domain containing 2A (MFSD2A) transporter as a key mediator in the response to tunicamycin

We used global gene disruption in a cell line haploid for all human chromosomes except for chromosome 8 to identify genes whose loss of function would render cells resistant to toxic doses of the N-linked glycosylation inhibitor TM (Fig. 1B). This insertional mutagenesis screening approach makes use of gene trap (GT) retroviruses that carry a strong splice acceptor site upstream of a promoterless and selectable marker gene. These haploid genetic screens have been recently used with success to identify cellular components required for influenza infection or toxicity to several bacterial toxins. GT insertions into genomic DNA of genes in KBM7 cells can lead to disruption of endogenous protein function and result in the generation of complete KOs (32).

KBM7 cells infected with GT retroviruses were grown in the presence of 500 ng/mL TM for ∼2 wk. Approximately 40 × 10⁶ GT cells were screened. Several TM-resistant clones were obtained, and GT mapping revealed seven independent insertions in the MFSD2A locus (Fig. 1C). MFSD2A constitutes the sole bona fide hit in our TM resistance screen, because it was the only locus for which multiple independent GT insertions were recovered. MFSD2A has been previously identified in humans as a putative receptor for Syncytin-2, an ancient retrovirus-derived envelope protein that endows placental trophoblast cells with the capability to fuse into syncytiotrophoblasts (33). In mice, MFSD2A expression is induced during fasting periods in the liver and also during exposure to low temperatures in brown adipose tissue (34). In another study, tumor-suppressive functions were assigned to MFSD2A owing to its down-regulation in non-small cell lung cancer samples, decreased colony formation in vitro, and reduced tumor growth of MFSD2A-overexpressing A549 cells in mouse xenograft studies (35). A potential transporter function for MFSD2A has not been explored so far.

MFSD2A belongs to the large major facilitator superfamily (MFS) of transporters of which the lactose permease LacY of Escherichia coli is a well-studied example. Members of this family are found in all kingdoms of life, usually have 12–14 transmembrane segments, and transport a plethora of molecules including ions, sugars, nucleotides, amino acids, and drugs (36–38). Approximately 25% of all known membrane transporters in prokaryotes are represented by the MFS group (39). They are polytopic membrane proteins that serve as secondary carriers acting as uni-, anti-, or symporters. We used several secondary structure/hydrophathy programs and found that the algorithms mostly predicted 10 or 12 transmembrane domains for human MFSD2A with both the N and C termini oriented to the cytoplasm. In light of the homology of MFSD2A to other members of the glycoside-pentoside-hexuronide (GPH) family of MFS transporters (40), it is likely that MFSD2A contains 12 TM helices (e.g., http://minnou.cchmc.org/). A BLASTP search reveals the presence of MFSD2A homologs in vertebrates, but interestingly, no obvious ortholog is present in yeast, Caenorhabditis elegans, or Drosophila. A second MFSD2A-related protein (42% identity and 59% similarity in humans) with unknown function, MFSD2B, is present in higher eukaryotes as well (34). MFSD2A also shows homology to diverse bacterial sugar transporters including the melibiose permease (MelB), a cation/melibiose symporter (Fig. S1). Interestingly, TM is a natural product that includes a nucleotide sugar analog as a building block (1, 41).

Both MFSD2A GT cell lines used throughout this study have a similar cell size distribution as WT cells (Fig. S2A). Semiquantitative RT-PCR analysis of these two MFSD2A GT lines (henceforth named MFSD2A GT1 and GT2) revealed no remaining MFSD2A transcripts using two different primer pairs (Fig. 1D). Importantly, reconstitution of the MFSD2A_GT2 line with an epitope-tagged MFSD2A cDNA construct fully reverted the TM resistance phenotype to levels comparable with WT KBM7 cells (Fig. 1E). Thus, loss of MFSD2A causes cellular TM resistance. Viability of the MFSD2A GT cells after 5 d of treatment with 500 ng/mL TM was ∼70% compared with only 6% of WT cells (Fig. 1F). In agreement with these results, TM-treated WT KBM7 but not MFSD2A GT cells showed up-regulation of UPR markers such as the ER chaperone GRP78, the proapoptotic CHOP transcription factor, and phospho-eIF2α, a readout for general inhibition of translation (Fig. 1G). Notably, MFSD2A GT cells were as susceptible as WT cells to other ER stress-inducing drugs such as brefeldin A (BFA), thapsigargin, or DTT and showed similar UPR marker induction (Fig. 1G and Fig. S2B). Moreover, pulse-chase experiments with ³⁵S-labeled methionine showed that glycosylation and trafficking of newly synthesized class I MHC heavy chains using two different antibodies proceed unperturbed in the presence of TM in MFSD2A GT but not in WT KBM7 cells (Fig. 1H). This shows that, in cells without MFSD2A, the N-glycosylation pathway is not blocked despite the presence of TM in the culture media, suggesting that the compound does not reach its intracellular target DPAGT1/GPT at the ER. The slight decrease in survival of TM-treated MFSD2A GT cells is suggestive of parallel mechanisms or alternative routes of TM entry into cells that impact cell viability on exposure to high concentrations of this compound. For instance, TM may be able to penetrate nonspecifically into the cell membrane by virtue of its lipophilic fatty acid moiety (41–43). Incomplete TM resistance may also be caused by other unknown proteins (for instance, MFSD2B) that act in a manner partially redundant to that of MFSD2A. Taken together, these results indicate that MFSD2A GT cells have no intrinsic ER stress-sensing defect and possess a functional UPR, and they suggest that TM-treated MFSD2A GT but not WT KBM7 cells maintain an unperturbed N-glycosylation pathway in the presence of the antibiotic. Therefore, TM resistance in MFSD2A mutant cells might be conferred upstream of known UPR components, presumably at the step of TM uptake into cells.

To establish where MFSD2A acts relative to the ER-luminal chaperone GRP78 and the TM target DPAGT1/GPT, which mediates the first step in the N-linked glycosylation pathway at the ER, shRNA-mediated double knockdown (KD) A549 cells were created (Figs. S2 C and D and S3 C and E show validation of hairpins; see Table S1 for shRNA sequences). The single GRP78 or DPAGT1 A549 KD cells are hypersensitive to TM and show a marked decrease in survival when exposed to different doses of TM, whereas MFSD2A-depleted cells display the opposite phenotype. Strikingly, simultaneous KD of MFSD2A/GRP78 or MFSD2A/DPAGT1 led to a TM-resistant phenotype very similar to the single MFSD2A KD phenotype (Fig. S3 A and B). Immunoblot analysis revealed a reduction of UPR marker expression in the shMFSD2A/shGRP78 or shMFSD2A/shDPAGT1 KD cells compared with control double KD cells (shLUC#221/shGRP78 or shLUC#221/shDPAGT1) or the single KD lines as expected (Fig. S3C). These findings were also recapitulated in KBM7 cells using four different DPAGT1 hairpins; single DPAGT1 KD KBM7 cells grown in TM-containing media quickly underwent apoptosis, whereas MFSD2A GT cells transduced with the same DPAGT1 hairpins were largely protected from cell death (Fig. S3D). In other words, to reveal TM hypersensitivity in cells with reduced GRP78 or DPAGT1 expression, MFSD2A function is required. The epistatic nature of the MFSD2A TM-resistant phenotype in the double KD cells is consistent with the hypothesis that MFSD2A might be involved in TM transport into cells upstream of GRP78 and DPAGT1/GPT.

To ask whether the function of MFSD2A in determining TM sensitivity is conserved across cells from different tissues, we infected several cancer cell lines with lentiviruses bearing shRNAs against MFSD2A for KD studies. Indeed, when using the most potent MFSD2A hairpin #1965, which leads to more than 70% or 80% reduction of MFSD2A mRNA levels in A549 and 786–0 cells, respectively (Fig. S2 C and D), significantly enhanced TM resistance was observed in A549, HT29, Jurkat T, HeLa, PC3, and 786–0 cells, which was evidenced by a greatly improved cell survival (Fig. S2 E–J). A similar, albeit weaker, TM resistance phenotype was noted when another MFSD2A hairpin (#1474) was used for KD studies; this finding is consistent with its less potent KD effect compared with #1965 in 786–0 cells (Fig. S2 D and J). Biochemical analysis of MFSD2A KD cells corroborated the finding that these cells experienced only little ER stress in response to TM as suggested by reduced GRP78 or CHOP expression (Fig. S2 K and L).

Having established the widespread TM resistance phenotype upon loss of MFSD2A, we next investigated its gain of function effects. For this purpose, stable cell lines overexpressing MFSD2A or control proteins were generated. Using various lysis conditions, we were not able to detect tagged MFSD2A by immunoblotting from protein extracts of stable expressing cell lines. We therefore immunoprecipitated MFSD2A before Western analysis using antibodies against the epitope tag. The molecular weight of the cloned MFSD2A isoform predicted a protein of ∼58.6 kDa; however, MFSD2A appeared as a smear of multiple bands, with the major band running around 70 kDa. TM treatment of Flag-MFSD2A–expressing cells before immunopurification (IP) or PNGase F digest of immunopurified MFSD2A led to a shift in band pattern compared with WT conditions, indicating that MFSD2A is a glycoprotein (Fig. 2 A and B). Interestingly, incubation of cells before IP with 10 ng/mL TM led to an identical band migration pattern as incubation with 1 μg/mL TM, suggesting that MFSD2A is already extensively deglycosylated at this low drug concentration (Fig. 2A). MFSD2A contains two predicted N-glycolysation sites (asparagines 217 and 227) in the extracelllular loop between transmembrane domains 5 and 6. We modified the WT MFSD2A cDNA by site-directed mutagenesis to abolish glycosylation of asparagines 217 and 227 (asparagine residues were converted into glutamine-encoding codons) in the protein (MFSD2AΔN-glyco). Flag-MFSD2AΔN-glyco immunopurified from stable A549 cells grown under normal conditions migrated similarly in a protein gel as Flag-MFSD2A treated with TM, and no additional band shift was apparent when Flag-MFSD2AΔN-glyco cells were treated with TM before IP (Fig. 2C). This supports the notion that at least one, if not both, of the two MFSD2A asparagine residues in this context are normally glycosylated.

No obvious phenotypic alteration could be detected in cells overexpressing MFSD2A under normal culture conditions, with the exception of a slightly increased cell size in A549 cells (Fig. S4A), but these cells showed a proliferation rate indistinguishable from controls. An increase in the G1 and decrease in S phase of the cell cycle upon MFSD2A overexpression in A549 cells was recently reported (35). We next assessed the consequences of varying TM concentrations on A549 or PANC1 cells that stably overexpress MFSD2A. We observed a pronounced decrease in cell survival and a corresponding increase of UPR parameters in MFSD2A-overproducing cells using low TM concentrations that neither altered ER stress marker expression like GRP78, phospho-eIF2α, or CHOP in control cells nor affected their viability (Fig. 2 D–F and Fig. S4B). Overexpression of MFSD2AΔN-glyco in conjunction with 50 ng/mL TM treatment led to a survival ratio comparable with WT MFSD2A expression, but the extent of apoptosis was somewhat reduced when a lower TM concentration (10 ng/mL) was used (Fig. S4C). The latter effect could be because of the decreased levels of the MFSD2A glycosylation mutant protein (Fig. 2C). Glycosylation of MFSD2A therefore is not a prerequisite for conferring TM sensitivity. A TM concentration as low as 1 ng/mL already had a strong effect on cell survival in MFSD2A-overexpressing A549 cells, whereas control cells showed no discernible effects at these concentrations (Fig. 2E). Additional control proteins containing transmembrane domains like MFSD2A such as Glut1, MFSD1, or lysosomal-associated membrane protein 1 (LAMP1), whose maturation involves processing through the secretory pathway, behaved indistinguishable from γ-Tubulin when overexpressed and treated with various TM doses (Fig. S4D), indicating that TM sensitivity in MFSD2A overexpressing cells is not caused nonspecifically through exhaustion of the endogenous UPR machinery (Fig. S4 D and E). Cell viability of γ-Tubulin– or Rap2a-expressing cells started to slightly decrease at TM concentrations ∼40 ng/mL TM, and their survival ratio at 100 ng/mL TM was comparable with MFSD2A treated with 1 ng/mL TM (Fig. 2E). By monitoring dose-dependent effects of TM on UPR markers in MFSD2A-overproducing A549 cells, we found measurable increases in GRP78 protein levels at TM concentrations as low as 100 pg/mL (∼120 pM) after 29 h of treatment. GRP78 levels were saturated at ∼20 ng/mL TM, and no additional increases with even 50-fold higher TM doses were observed. CHOP expression was apparent with 10 ng/mL TM treatment, consistent with the strong decrease in cell survival. However, GRP78 levels in Flag-γ-Tubulin–expressing cells started to rise at ∼40 ng/mL TM but did not reach a comparable GRP78 induction even with 1 μg/mL TM, the highest TM concentration used (Fig. 2G). This indicates that A549 MFSD2A overexpressors are at least ∼40- to 100-fold more sensitive to TM.

In accordance with the specific effects of TM on cells depleted of MFSD2A, no increased sensitivity to a variety of different ER stressors such as BFA, thaspigargin, DTT, and 2-DG as well as the topoisomerase II inhibitor etoposide was observed when MFSD2A was overexpressed (Fig. S4 E and F).

To gain insight into subcellular localization of MFSD2A, we performed immunofluorescence (IF) staining and confocal microscopy on cells transiently transfected or stably expressing tagged MFSD2A. MFSD2A expression showed some variability between cells but presented as plasma membrane (PM) staining and cytoplasmic punctae (Fig. 3A). The punctae might represent MFSD2A-containing vesicles that shuttle between the endomembrane compartment and the PM. We obtained similar results using transiently transfected human embryonic kidney 293T (HEK293T) or stable EGFP-MFSD2A fusion protein-expressing HeLa and PANC1 cells (Fig. 3A). To confirm the specificity of the observed MFSD2A expression pattern, we repeated the experiment using Flag-tagged MFSD2A that was cotransfected with a GFP-Pleckstrin homology (GFP-PH) PM reporter into PANC1 cells (44). In agreement with the previous staining pattern, PM localization of Flag-MFSD2A was evident in addition to accumulation at some discrete cytoplasmic spots (Fig. S5A). Consistent with our results, Esnault et al. (33) recently showed cell membrane staining of HA-tagged MFSD2A expressed in HeLa cells. Mouse MFSD2A was reported to localize mainly to the ER, but PM staining was acknowledged as well (34).

C-terminal amino acid residues of xenobiotic transporters are important to ensure correct targeting, function, and expression at the PM by interaction with adaptor proteins. We scanned the amino acid sequence of MFSD2A and found that the last three C-terminal residues (Ser-Ile-Leu) fit a class I PDZ binding motif (45). To test the importance of the C terminus for MFSD2A function regarding TM sensitivity, a series of C-terminal MFSD2A deletion constructs (Flag-MFSD2AΔC) lacking the last 3, 10, 23, or 40 amino acids was engineered. Expression of these stable A549 MFSD2A deletion mutant cell lines was verified by IP followed by Western blot analysis. Removal of C-terminal amino acids caused electrophoretic mobility shifts likely attributable to the reduced molecular weight of the mutant proteins and changes in glycosylation pattern. Expression levels of the MFSD2AΔC forms generally seemed to be diminished compared with MFSD2A full-length protein (Fig. 3B). The importance of certain C-terminal amino acid residues for correct trafficking and expression of several transporters has been appreciated before (46–48).

To assess functionality of the MFSD2A deletion constructs, stable MFSD2AΔC lines were treated with two different doses of TM for 3 d, after which time the surviving cells were counted. A549 cells that express WT MFSD2A showed the least numbers of viable cells at the end of the experiment (12% survival compared with 100% survival of control cell lines when treated with 10 ng/mL TM). Cells that expressed MFSD2A lacking the three C-terminal residues (MFSD2AΔC3) showed a significantly improved survival ratio (24% survival; P < 0.03 using Student two-tailed t test) compared with full-length MFSD2A. The viability ratio increased more in the MFSD2AΔC10 and MFSD2AΔC23 lines (59% and 58%, respectively). Notably, removal of the distal MFSD2A C-terminal 40 amino acids starting after the last transmembrane helix (MFSD2AΔC40) completely abrogated sensitivity to TM, and these cells showed the same viability as cell lines overexpressing control proteins without changes in GRP78 and CHOP levels (Fig. 3 C and D). IF staining of Flag-MFSD2AΔC40–expressing PANC1 cells showed that deletion of the last 40 amino acids prevented MFSD2A PM localization and instead, caused a reticular-like distribution that presumably reflects retention of the protein in the ER (Fig. S5B). Thus, we speculate that the C terminus of MFSD2A provides a binding platform for potential interaction partners required for correct MFSD2A trafficking and expression.

As mentioned earlier, TM interferes with glycoprotein biosynthesis by inhibiting the ER-localized glycosyltransferase DPAGT1/GPT. Its catalytic center has been proposed to act on the cytoplasmic side of the ER (49, 50). Cytoplasmic segments are also important for the catalytic activity of MraY, the bacterial homolog of DPAGT1/GPT (51). Thus, it is likely that, for TM to inhibit N-glycosylation, it does not need to permeate into the ER lumen. The TM resistance phenotype of cells without MFSD2A function could be explained by at least two possibilities. First, DPAGT1 levels might be increased in these cells. This seems unlikely, because quantitative real-time PCR for DPAGT1 in MFSD2A KD A549 and 786–0 KD cells did not show elevated DPAGT1 mRNA levels (Fig. S6 A and B). Second, a failure of TM to reach ER-localized DPAGT1 (for instance, through prevention of TM uptake at the plasma membrane as suggested by the pulse-chase experiment (Fig. 1H) and the MFSD2AΔC40 mutant that is absent from the PM and unable to confer TM sensitivity) (Fig. S5B) should also cause TM resistance. Hence, we were interested in examining the consequences of mistargeting MFSD2A to the ER. Specific ER targeting was achieved by adding a dilysine motif to the C terminus of MFSD2A (MFSD2AKKRQ). Staining for the ER marker Calnexin and Flag-tagged MFSD2AKKRQ indeed showed MFSD2A/ER colocalization and very little MFSD2A staining at the cell membrane (Fig. 3E). Treating stable ER-targeted MFSD2A A549 cells with TM showed no increase in sensitivity, and their survival was indistinguishable from control cells using 10 ng/mL TM (Fig. 3C). A slight sensitization was observed using 50 ng/mL TM, probably caused by residual PM expression of a small MFSD2A pool that escaped ER retention. However, the small effect on survival was minor compared with WT MFSD2A overexpression. Thus, MFSD2A PM localization is crucial for conferring cellular TM sensitization.

To learn more about a possible mechanism for how TM sensitivity is mediated by MFSD2A and to explore its potential TM transporter function, we made use of two small molecules, Chlorpromazine (CPZ) and Phloretin, that both affect ion transport across the PM. CPZ inhibits various ATPases, including Na⁺-K⁺-ATPase, thereby reducing both the electrochemical sodium gradient and sodium-dependent solute transport (52, 53). Phloretin impinges on nonelectrolyte and ion membrane transport and more specifically, inhibits sugar, urea, and chloride transport (54). Interestingly, simultaneous treatment of MFSD2A-overexpressing cells with TM and CPZ or Phloretin largely negated GRP78 and CHOP induction observed with TM treatment alone (Fig. S7A). This raised the possibility that MFSD2A-mediated transport processes are ion-dependent (e.g., sodium-dependent).

MelB, a bacterial sugar symporter of the MFS with homology to MFSD2A, catalyzes the coupled symport of a cation (Na⁺, Li⁺, or H⁺) with the energetically downhill movement of a galactoside such as melibiose, raffinose, or lactose (55). Through mutational analysis, several amino acids in MelB, including Arg-52 (R52), Asp-55 (D55), Asp-59 (D59), and Lys-377 (K377), have been identified that impair ion or sugar binding/translocation (56–62). The amino acid sequence alignment of MFSD2A and MelB shows that aforementioned residues are evolutionarily conserved (Fig. S1). We replaced selected amino acids of the WT sequence with alanines to create the corresponding MFSD2A mutants and asked whether these amino acid substitutions would alter MFSD2A activity regarding cellular TM susceptibility/transport. The following amino acids in MFSD2A were changed to alanine (A) residues: Arg-90 (R90A; corresponding to R52 in E. coli MelB), Asp-93 (D93A; D55 MelB), Asp-97 (D97A; D59 in MelB), and Lys-436 (K436A; K377 in MelB). A549 cells were transduced with the different MFSD2A mutant constructs by lentiviral infection and stable cell lines established before treatment with varying TM doses (ranging from 1 to 200 ng/mL TM) as described previously. We noted that, with the exception of the D93A and D97A mutants, all other MFSD2A mutant constructs showed reduced expression levels when Flag-immunopurified from stable A549-expressing cells, indicating that the amino acid substitutions render MFSD2A unstable (Fig. S7B). Of the MFSD2A constructs tested, only the D93A mutant showed similar survival ratios to the WT protein upon TM treatment, which is also reflected by their similar EC₅₀ values [9.5 ± 0.02 and 14.1 ± 0.09 ng/mL TM for Flag-MFSD2A (WT) and Flag-MFSD2A(D93A), respectively]. However, overexpression of MFSD2A(D97A) displayed a dramatically weakened TM sensitivity phenotype, leading to an at least 9.4-fold EC₅₀ shift [EC₅₀ for Flag-MFSD2A(D97A): ≥89.5 ± 0.09 ng/mL TM] (Fig. S7C). Importantly, IF localization studies show that transfected Flag-MFSD2A(D97A) has an expression pattern indistinguishable from WT (Fig. S5 A and C). The findings suggest that D93 is largely dispensable for MFSD2A's ability to hypersensitize to TM, whereas D97 is critical. These results are in agreement with the behavior of previously described MelB mutants: Asp to Cys substitutions at positions 55 and 59 (D55C and D59C) in WT MelB have been shown to cause loss of sodium-coupled transport of melibiose; however, the D55C mutant (corresponding to D93A in MFSD2A) still retained H⁺-coupled melibiose transport in contrast to D59C (D97A in MFSD2A), whose sugar transport activity is abolished because of elimination of cation (Na⁺ and H⁺) binding (56, 57, 59, 60). We surmise that the MFSD2A(D93A) mutant, which is equally potent in imparting TM sensitivity to cells as the WT protein, might still allow proton-coupled TM transport, whereas MFSD2A(D97A) does not.

The other two MFSD2A mutants tested in this assay, R90A and K436, displayed a similar dose-response curve as D97A or the negative controls (Flag-Glut1 and Flag-MFSD2AΔC40), but their reduced levels might have contributed to this phenotype (Fig. S7 B and C). For this reason, we transiently transfected HEK293T cells with the individual constructs to achieve comparable MFSD2A levels before treating the cells with TM. Both the transfected D97A and K436A mutants did not up-regulate the ER stress markers GRP78 and CHOP, whereas D93A and R90A did (Fig. S7D). Compared with WT Flag-MFSD2A, the K436A mutant displayed no appreciable PM localization but showed increased perinuclear staining suggestive of ER accumulation (Fig. S5D). Substitution of Ala, Ser, Gln, or Val for Arg-52 (R52) of MelB (R90 in MFSD2A) is critical for H⁺-coupled melibiose transport but not essential for Na⁺-coupled transport (61). Thus, by analogy, MFSD2A(R90A), despite its reduced expression levels in stably transduced A549 cells, might likewise still be able to couple sodium and TM cotransport, which is inferred from its almost WT-like behavior in our HEK293T transfection assay (Fig. S7D). Last, a mutant MelB carrier, where Lys-377 (Lys-436 in MFSD2A) was substituted for Val (K377V), had significantly reduced sugar transport activity (62), and we find that the corresponding MFSD2A(K436A) mutant no longer evokes TM susceptibility and UPR marker induction upon overexpression, which is suggestive of impaired TM transport into cells (Fig. S7 B and D). Together, our IF stainings and gain of function results as well as the functional analysis of MFSD2A missense mutants hinted to a role for MFSD2A as PM TM transporter.

Based on (i) sequence similarity of MFSD2A to bacterial sugar permeases of the MFS, (ii) mutational analysis of conserved residues that have been shown to affect cation/sugar transport in MelB and likewise, critically influence MFSD2A's ability to induce ER stress upon TM treatment, (iii) its presence at the PM, (iv) normal Class I MHC heavy-chain trafficking and glycosylation in MFSD2A GT cells despite the presence of TM, (v) TM hypersensitivity of MFSD2A-overexpressing cells, and (vi) the fact that exclusive ER localization of MFSD2A or cytoplasmic accumulation of C-terminal MFSD2A deletion mutants did not elicit TM hypersensitivity, we surmised that TM might be a substrate for MFSD2A transported across the PM. We analyzed TM-treated MFSD2A KO KBM7 as well as stable A549 or PANC1 overexpressor cell lines for their TM content by MS. Because of its lipophilic fatty acid chain, TM is expected to nonspecifically partition into the lipid membrane bilayer independent of the genotype of the cells (43). Indeed, no differences in the intensity of the TM signal between the mutant and WT cells were detected (Fig. S8A). It is probable that the MS signal derived from nonspecific membrane binding of TM prevented us from detecting intracellular differences in TM accumulation, especially because intracellular TM accumulation might be much lower than the TM on the PM (43). In contrast, we readily detected significant differences in TM content in a time-dependent manner when cells were treated for different time periods with TM using the overexpression lines (Fig. 4 A–C). For instance, MFSD2A-overexpressing A549 cells displayed an approximately fivefold increase (depending on the TM species) after 3 h compared with cells overexpressing Flag-γ-Tubulin (Fig. 4C), but significant differences could already be detected after 30 min of TM treatment (Fig. 4A). In PANC1 cells, an ∼2- to 32-fold higher TM amount was measured in MFSD2A-overexpressing cells relative to controls after a longer exposure to the compound (Fig. S8B). Reassuringly, Flag-MFSD2AKKRQ A549 cells did not show elevated TM uptake (Fig. S8C) even after 24 h TM treatment. The ratios of the individual TM species accumulated in cells reflect the original TM stock composition used for this experiment, indicating that MFSD2A does not discriminate between the different TM homologs. For the same reasons alluded to earlier (i.e., the lipophilic nature of TM), the magnitude of the result might underestimate TM accumulation in MFSD2A-overexpression cells. In summary, WT but not ER-targeted MFSD2A expression causes increased cellular TM accumulation.

Viability of cells in 96-well assay plates with clear flat bottoms (Costar) was determined using the CellTiter Glo assay (Promega), which measures cellular ATP content. For suspension cells (KBM7 and Jurkat T cells), typically 20,000 cells were seeded per well. Culture volume per well was 100 μL. The cells were grown in the absence or presence of the drug for several days depending on the experiment; 50 μL CellTiter Glo reagent was added at the end of the experiment to lyse cells according to the manufacturer's protocol. ATP-based bioluminescence levels were measured using an EnVision plate reader (Perkin-Elmer). For low-throughput viability assays, 50,000–60,000 adherent cells were seeded in six-well clusters 24 h before drug addition. The next day, old media was aspirated and replaced with 2 mL fresh DMEM premixed with TM (or other drugs) before addition to the cells. Treatment duration usually was between 3 and 6 d depending on the cell line and experiment. For determination of cell number, cells were trypsinized and resuspended in media in a total volume of 3 mL. A 0.5-mL aliquot diluted 1:20 in Isoton II Diluent solution (Beckman Coulter) was used for cell number measurement using a Z2 Coulter Counter (Beckman Coulter). Cells with diameters between 10 and 30 μm were counted. Survival ratio was calculated by dividing cell numbers (or the luminescence signal) of the treated samples by cell numbers (or luminescence signal) of the untreated samples.

Cells were harvested in lysis buffer containing 1% Nonidet P-40, 120 mM NaCl, 50 mM Tris⋅HCl (pH 7), 1 mM EDTA, 6 mM EGTA, 20 mM NaF, 1 mM benzamidine, 15 mM Na₄P₂O₇ × 10 H₂O and freshly added complete Mini protease inhibitor mixture (Roche), 30 mM β-glycerophosphate, 30 mM para-nitrophenyl phosphate (pNPP), 4 μM Pepstatin, 400 μM PMSF, 2 μM aprotinin, and 4 μM leupeptin. Cell extracts were briefly vortexed and rotated for 10 min at 4 °C before centrifugation at 16.2 × g for an additional 10 min. For immunoprecipitations, a 50% slurry of anti-Flag M2 Affinity beads (Sigma) was washed three times in lysis buffer before IP. Between 500 μg and 1 mg protein and 40–50 μL slurry were used for IP by rotating samples for 2 h at 4 °C. Immunoprecipitates were washed three times with lysis buffer before elution, denaturation in 50 μL sample buffer, and boiling for 3 min. Bio-Rad protein assay was used to quantify protein amounts. Proteins were resolved on 4–12% NuPAGE Novex gradient Bis⋅Tris gels (Invitrogen) and transferred onto PVDF membranes (Immobilon). IgG-HRP–coupled secondary antibodies (Santa Cruz) for normal immunoblotting or fluorescently coupled IRDye 680 and 800 secondary antibodies (LI-COR Biosciences) for LI-COR analysis were used at 1:5,000 in 5% milk/PBS-T for 1 h incubation at room temperature. Proteins bands were visualized with Western Lightning chemiluminescence (ECL)/plus-ECL reagent (Perkin-Elmer) or LI-COR infrared scanner. Metabolic labeling and pulse-chase analysis were performed as described (73).

HEK293T, A549, or PANC1 cells (60,000) were plated on coverslips (precoated with Fibronectin) in 12-well cluster plates; 24 h later, the cells were fixed for 15 min at room temperature in 4% paraformaldehyde solution (PFA) in PBS that was prewarmed to 37 °C, rinsed three times in PBS, and permeabilized for 5 min in 0.05% Triton-X in PBS. After an additional washing step with PBS, primary antibodies were diluted in 5% normal donkey serum, added to the cells, and incubated overnight at 4 °C. The slides were then rinsed three times with PBS and incubated for 1 h protected from light at room temperature with secondary antibodies (in 5% normal donkey serum) that were produced in donkey. After three more washes with PBS, the coverslips were mounted on glass slides using Vectashield (Vector Laboratories). Images were acquired either with a spinning disk confocal microscope (Perkin-Elmer) equipped with a Hamamatsu 1 × 1 k EM-CCD camera or a Zeiss Axiovert 200 M microscope equipped with a Zeiss AxioCam camera. Secondary Alexa Fluor antibodies were used at 1:2,000 dilution. Alexa Fluor 568 Phalloidin or Alexa Fluor 633 Phalloidin (Invitrogen) were used at 1:2,000, and Hoechst (Molecular Probes) was used at 1:2,000.

Stable A549 cells (75,000) expressing the protein of interest were seeded 24 h before the addition of TM (Fig. 4). Cells were washed three times with ice-cold PBS before lysis in 250 μL 80% methanol. Samples were pipetted up and down several times, briefly vortexed, and stored at −80 °C before MS analysis. Measurement of TM was carried out using a 4000QTRAP triple quardupole mass spectrometer (AB SCIEX) that was coupled to an Agilent 1100 HPLC pump (Agilent Technologies) and an HTS PAL autosampler (Leap Technologies). Chromatographic separations were achieved using a 150 × 2.1 mm Atlantis T3 column (Waters) that was initially eluted isocratically at 250 μL/min with 50% mobile phase A (0.1% formic acid in aqueous 10 mM ammonium formate) and 50% mobile phase B (0.1% formic acid in acetonitrile) for 2 min followed by an 8-min linear gradient to 100% mobile phase B. Mass spectrometer settings used for multiple reaction monitoring (MRM) analyses of TM species were determined by infusion of a references solution. MRM transitions monitored using a 75-ms dwell time and collision energy setting of 25 were m/z 817.6/596.6 (Tun 14:1 A and B), 833.6/612.6 (Tun 15:0), 831.6/624.7 (Tun 15:1 A and B), 845.6/624.7 (Tun 16:1 A and B), and 859.7/638.7 (Tun 17:1 A and B). Other instrument settings were curtain gas = 20, electrospray source potential = 5 kV, source temperature = 400 °C, gas1 = 35, gas2 = 40, and declustering potential = 60. Calibration curves were generated by analysis of a serially diluted TM reference solution. Cell extracts were analyzed directly with 10 μL injected from each sample. MultiQuant software (version 1.1; AB SCIEX) was used for peak integration, and data were manually reviewed for quality of integration. TM signal intensities were normalized to cell number and averaged from three biological replicates.