Lysosomal autophagy inhibition (LAI) with hydroxychloroquine or DC661 can enhance cancer therapy, but tumor regrowth is common. To elucidate LAI resistance, proteomics and immunoblotting demonstrated that LAI induced lipid metabolism enzymes in multiple cancer cell lines. Lipidomics showed that LAI increased cholesterol, sphingolipids, and glycosphingolipids. These changes were associated with striking levels of GM1+ membrane microdomains (GMM) in plasma membranes and lysosomes. Inhibition of cholesterol/sphingolipid metabolism proteins enhanced LAI cytotoxicity. Targeting UDP-glucose ceramide glucosyltransferase (UGCG) synergistically augmented LAI cytotoxicity. Although UGCG inhibition decreased LAI-induced GMM and augmented cell death, UGCG overexpression led to LAI resistance. Melanoma patients with high UGCG expression had significantly shorter disease-specific survival. The FDA-approved UGCG inhibitor eliglustat combined with LAI significantly inhibited tumor growth and improved survival in syngeneic tumors and a therapy-resistant patient-derived xenograft. These findings nominate UGCG as a new cancer target, and clinical trials testing UGCG inhibition in combination with LAI are warranted.

Significance:

We discovered UGCG-dependent lipid remodeling drives resistance to LAI. Targeting UGCG with a drug approved for a lysosomal storage disorder enhanced LAI antitumor activity without toxicity. LAI and UGCG inhibition could be tested clinically in multiple cancers.

This article is highlighted in the In This Issue feature, p. 247

Autophagy inhibition impairs tumor growth in preclinical cancer models when combined with chemotherapy, MAPK targeted therapy, or immunotherapy (1, 2). Targeting autophagy at the level of the lysosome is an attractive approach in multiple disorders, including cancer (3, 4), but the clinically used lysosomal autophagy inhibitor (LAI) hydroxychloroquine (HCQ) rarely has single-agent antitumor activity in preclinical models. More potent LAIs like GNS561 and dimeric chloroquines (e.g., DC661) have single-agent antitumor activity in preclinical models (5, 6). However, either in preclinical studies of the novel LAIs or in the clinical trials conducted with HCQ combinations in which patients responded, tumor regrowth on continued therapy is common, indicating the development of resistance (5, 7–10). We and others have shown that HCQ, GNS561, and DC661 bind to and inhibit the lysosomal enzyme palmitoyl-protein thioesterase 1 (PPT1), and PPT1 knockout recapitulates the phenotype of these inhibitors (5, 6, 11). Therefore, HCQ and its derivatives are targeted therapies and not simply weak bases that affect lysosomal pH. Importantly, DC661 and related optimized derivatives are more cell-penetrant within an acidic tumor microenvironment than HCQ (5). Novel autophagy inhibitors, including LAIs, are entering clinical trials, and new strategies to address LAI resistance will significantly enhance their clinical utility.

Using melanoma as a model, we have found that LAIs induced a small change in the proteome but a large change in the lipidome. Sphingolipid and cholesterol pathways were upregulated and contributed to cell survival in the face of LAI. The increased activity of these lipid metabolism pathways was associated with a striking accumulation of GM1+ membrane microdomains (GMM). From these pathways, UGCG emerged as a druggable target for LAI resistance and cancer cell survival.

LAI Induces Proteome and Lipidome Changes Leading to the Accumulation of Sphingolipids and Cholesterol

LAI, by either HCQ or DC661, leads to lysosomal dysfunction, but other pathways could be perturbed that contribute to cell death or survival in the face of LAI. To investigate this in an unbiased manner, we performed proteome analysis of A375P melanoma cells treated with either DC661 or HCQ. Out of 4,264 cellular proteins identified, DC661 or HCQ treatment significantly increased only 157 and 88 proteins and decreased 34 and 43 proteins, respectively (Fig. 1A). Ingenuity pathway analysis (IPA) demonstrated that the most significantly activated pathways (positive z-score) in response to LAI were related to lipid metabolism (Fig. 1B). We found that 22 lipid metabolism proteins were significantly upregulated in DC661-treated compared with vehicle-treated cells (Fig. 1C). There was almost complete overlap in the patterns of proteins that were increased by DC661 (3 μmol/L) and by higher concentrations of the less potent lysosomal inhibitor HCQ (30 μmol/L), with 12 proteins significantly increased in both groups (Fig. 1C). Importantly, these 12 proteins, e.g., apolipoprotein B-100 (APOB), UDP-glucose ceramide glucosyltransferase (UGCG), hydroxymethylglutaryl-CoA synthase (HMGCS1), low-density lipoprotein receptor (LDLR), and scavenger receptor class B type 1 (SRB1), are all involved in sphingolipid and cholesterol metabolism. Considering lipid-related proteins (excluding APOB, which is likely from bovine source), DC661 or HCQ treatment produced the largest fold change in UGCG (∼7-fold increase; Fig. 1D). Based on these findings, we characterized the global lipidome of A375P cells and found extensive remodeling following DC661 [1,380/1,917 (72%) of lipids species changed significantly] or HCQ [1,410/1,917 (73%) of lipids species changed significantly] treatment (Supplementary Fig. S1A and S1B). In the cells treated with low concentrations of DC661 or high concentrations of HCQ, 19% of the lipid species were changed at least 2-fold (Fig. 1E). The most striking treatment-related increases occurred in glycosphingolipids (hexosylceramides), followed by ceramides and sphingomyelins (Fig. 1F). Filipin staining demonstrated that DC661 or HCQ treatments significantly increased cholesterol in the plasma membrane and lysosomal-associated membrane protein 1-positive (LAMP1+) lysosomes (Fig. 1G). To further confirm the lipid changes specifically in lysosomes, we purified lysosomes using lysosomal immunoprecipitation (Lyso-IP; ref. 12) from DC661- and vehicle-treated A375P cells (Fig. 1H) and characterized the lipidome. LAI induced a significant increase in sphingomyelins, ceramides, and cholesterol esters in the lysosomal lipidome (Fig. 1I). This increase in cholesterol esters was expected because DC661 raises lysosomal pH (5), and decreased lysosomal acid lipase activity was observed following DC661 treatment (Supplementary Fig. S1C), Altogether, these results demonstrated that LAI causes upregulation of lipid metabolism enzymes and accumulation of sphingolipid and cholesterol species.

Figure 1.

LAI leads to the accumulation of sphingolipids and cholesterol. A–D, LC/MS-MS–based proteome analysis of A375P cells treated with DC661 (3 μmol/L) or HCQ (30 μmol/L) for 24 hours. All experiments were done in triplicate. A, Volcano plots with significant changes relative to control (FDR < 5% and |FC| ≥1.5) highlighted in red (increase) and blue (decrease). B, IPA of elevated proteins (FDR < 5% and |FC| ≥1.5) in DC661- or HCQ-treated cells compared with control samples. The black line is set at threshold 3.0 (P < 0.001). Chol, cholesterol; Zym, zymosterol; Mva, mevalonate; GGPP, geranylgeranyldiphosphate; LXR/RXR/FXR, liver X receptor/retinoid X receptor/farnesoid X receptor. C, Heat map of lipid metabolism proteins significantly elevated (FDR < 5% and |FC| ≥1.5); names in bold indicate proteins which were significantly increased by both DC661 and HCQ. D, Fold change in the lipid metabolism proteins significantly elevated by both DC661 (3 μmol/L) and HCQ (30 μmol/L; FDR < 5% and |FC| ≥1.5) in A375P cells. E and F, LC/MS-MS–based lipidome analysis of A375P cells treated with DC661 (3 μmol/L) or HCQ (30 μmol/L) for 24 hours. E, Volcano plots of A375P lipidome showing significantly changed lipid species (FDR < 5% and |FC| ≥2) highlighted in red (increase) and blue (decrease) induced by DC661 (3 μmol/L) or HCQ (30 μmol/L) treatments for 24 hours. F, Mean peak area ± standard deviation of lipid classes; SM, sphingomyelin, Cer, ceramide, Hex1Cer, hexosylceramide. The number of species detected in each class is indicated below the label. G, Representative images of filipin staining and quantification. A total of at least 50 single cells were counted from multiple images/experimental group (each dot represents a single cell). Merged images show colocalization (yellow arrows) of filipin and lysosomes (LAMP1) in A375P cells treated with DC661 (1 μmol/L) or HCQ (30 μmol/L) for 24 hours. H, Schema for lysosomal purification by immunoprecipitation (Lyso-IP). I, LC/MS-MS lipidome analysis of lysosomes purified from A375P cells treated with vehicle control or DC661 (3 μmol/L) for 24 hours. Mean peak area ± standard deviation of significantly elevated lipid classes. ChE, cholesterol esters. Mean ± SEM. Scale: 10 μm. *, P ≤ 0.05; ****, P ≤ 0.0001. One-way ANOVA followed by Dunnett multiple comparisons procedure (F); Welch t test (G); Student t test (I).

Figure 1.

LAI leads to the accumulation of sphingolipids and cholesterol. A–D, LC/MS-MS–based proteome analysis of A375P cells treated with DC661 (3 μmol/L) or HCQ (30 μmol/L) for 24 hours. All experiments were done in triplicate. A, Volcano plots with significant changes relative to control (FDR < 5% and |FC| ≥1.5) highlighted in red (increase) and blue (decrease). B, IPA of elevated proteins (FDR < 5% and |FC| ≥1.5) in DC661- or HCQ-treated cells compared with control samples. The black line is set at threshold 3.0 (P < 0.001). Chol, cholesterol; Zym, zymosterol; Mva, mevalonate; GGPP, geranylgeranyldiphosphate; LXR/RXR/FXR, liver X receptor/retinoid X receptor/farnesoid X receptor. C, Heat map of lipid metabolism proteins significantly elevated (FDR < 5% and |FC| ≥1.5); names in bold indicate proteins which were significantly increased by both DC661 and HCQ. D, Fold change in the lipid metabolism proteins significantly elevated by both DC661 (3 μmol/L) and HCQ (30 μmol/L; FDR < 5% and |FC| ≥1.5) in A375P cells. E and F, LC/MS-MS–based lipidome analysis of A375P cells treated with DC661 (3 μmol/L) or HCQ (30 μmol/L) for 24 hours. E, Volcano plots of A375P lipidome showing significantly changed lipid species (FDR < 5% and |FC| ≥2) highlighted in red (increase) and blue (decrease) induced by DC661 (3 μmol/L) or HCQ (30 μmol/L) treatments for 24 hours. F, Mean peak area ± standard deviation of lipid classes; SM, sphingomyelin, Cer, ceramide, Hex1Cer, hexosylceramide. The number of species detected in each class is indicated below the label. G, Representative images of filipin staining and quantification. A total of at least 50 single cells were counted from multiple images/experimental group (each dot represents a single cell). Merged images show colocalization (yellow arrows) of filipin and lysosomes (LAMP1) in A375P cells treated with DC661 (1 μmol/L) or HCQ (30 μmol/L) for 24 hours. H, Schema for lysosomal purification by immunoprecipitation (Lyso-IP). I, LC/MS-MS lipidome analysis of lysosomes purified from A375P cells treated with vehicle control or DC661 (3 μmol/L) for 24 hours. Mean peak area ± standard deviation of significantly elevated lipid classes. ChE, cholesterol esters. Mean ± SEM. Scale: 10 μm. *, P ≤ 0.05; ****, P ≤ 0.0001. One-way ANOVA followed by Dunnett multiple comparisons procedure (F); Welch t test (G); Student t test (I).

Close modal

Cholesterol and Sphingolipid Salvage Pathways Confer Cytoprotection from LAI

We sought to determine if the induction of the cholesterol or sphingolipid metabolism pathways was a part of the cytotoxic response to LAI or a compensatory cytoprotective response to LAI-mediated cytotoxicity by investigating the effects of targeting key enzymes in these pathways in combination with DC661 treatment. We first depleted membrane cholesterol with methyl-β-cyclodextrin (MBCD; ref. 13). Cotreatment with DC661 and MBCD significantly decreased the viability of A375P cells compared with DC661 alone (Fig. 2A). The addition of water-soluble cholesterol to the medium significantly mitigated the enhanced cytotoxicity elicited by the MBCD and DC661 combination (Fig. 2A). We next targeted 3-hydroxy-3-methylglutaryl-coenzyme A (HMGCoA) reductase, which catalyzes the rate-limiting step in de novo cholesterol synthesis, using three clinically relevant statins: simvastatin, atorvastatin, and lovastatin (Fig. 2B; Supplementary Fig. S2A and S2B). No combination of statin and DC661 enhanced cytotoxicity compared with DC661 alone. Because de novo cholesterol synthesis did not play a role in DC661 cytotoxicity, we next targeted extracellular cholesterol uptake, carried out by LDLR. When LDL particles bind to the LDLR, the complex is internalized into endocytic vesicles and trafficked to the lysosome, where cholesterol is released and made available for use within the cell (14). We used a specific anti-LDLR antibody to prevent the binding of LDL to the LDLR (15). Anti-LDLR antibody + DC661 showed significantly enhanced cytotoxicity compared with isotype control + DC661 (Fig. 2C). Because LDLR requires lysosomal accumulation to deliver scavenged cholesterol to the cell, the enhanced cytotoxicity elicited with anti-LDLR and DC661 was initially surprising. We have previously shown that DC661 significantly impairs but does not completely abrogate lysosomal acidification (5). To further investigate this finding, we performed proteomics on isolated lysosomes and found a significantly increased level of LDLR and APOB in the DC661-treated samples (Supplementary Fig. S2C). This suggests that despite lysosomal impairment induced by DC661 (Supplementary Fig. S1C; ref. 5), LDL uptake and trafficking to the lysosome by LDLR was intact, but receptor recycling or degradation was blocked. A lysosome-independent means of scavenging cholesterol occurs through SRB1, which binds to high-density lipoprotein (HDL; ref. 16). Inhibition of SRB1 using the selective chemical inhibitor BLT-1 (17) resulted in a modestly enhanced DC661 cytotoxicity (Fig. 2D). Next, we focused on enzymes that regulate sphingolipid metabolism, specifically those responsible for generating ceramides and hexosylceramides, sphingolipid classes that were increased after LAI. Mammalian ceramide pools are maintained by de novo synthesis, which starts with the condensation of palmitoyl-CoA and serine, leading to the formation of the intermediate sphingoid base sphinganine, enabling ceramide synthesis. Alternatively, salvage pathways that include sphingomyelin hydrolysis and glycosphingolipid recycling can lead to ceramide formation (18). Once ceramides are generated, ceramide-modifying enzymes can add complexity to form hexosylceramides and gangliosides. Inhibition of de novo ceramide synthesis using the selective serine palmitoyltransferase inhibitor myriocin did not enhance the cytotoxicity of DC661 (Fig. 2E). We next targeted ceramide synthase isoforms 1 to 6 using a pan-ceramide synthase inhibitor, fumonisin-B1, and did not observe a difference in viability between DC661 plus fumonisin-B1 and DC661 alone (Fig. 2F). Each ceramide synthase isoform, from ceramide synthase 1 to ceramide synthase 6, synthesizes ceramides with varying fatty acid chains (19) and is involved in both de novo and salvage pathways. Lipidome analysis following DC661 treatment demonstrated a predominant increase in the levels of individual ceramide species with long (C20–26) fatty acid chains (Fig. 2G), which are synthesized predominantly by ceramide synthase 2. Therefore, we targeted ceramide synthase 2 using a more potent and specific inhibitor, ST-1074 (20, 21). ST-1074 and DC661 enhanced cytotoxicity in A375P cells compared with DC661 alone (Fig. 2H). We then targeted sphingomyelin hydrolysis using the acid sphingomyelinase inhibitor siramesine and observed no effect on DC661 toxicity (Fig. 2I). In our lipidome analysis, hexosylceramides with longer fatty acid chains were the most significantly increased hexosylceramide species (Fig. 2J). Notably, hexosylceramides are glycosphigolipids synthesized by the rate-limiting enzyme UGCG. Combining the UGCG inhibitors Genz-123346 (free base) or its FDA-approved analogue, eliglustat (hemitartarate salt), with DC661 strikingly enhanced the cytotoxicity of DC661 in A375P cells (Fig. 2K; Supplementary Fig. S2D). The inhibitor studies reported above were repeated at least three times to compare and confirm the relative augmentation of DC661 cytotoxicity imparted by each inhibitor (Fig. 2L). We concluded that inhibition of either cholesterol uptake receptors (LDLR, HDLR) or sphingolipid salvage pathway enzymes (ceramide synthase 2, UGCG) significantly augmented DC661 cytotoxicity (Fig. 2L), suggesting these proteins could be part of a resistance mechanism. The cholesterol uptake receptors and sphingolipid metabolism enzymes whose inhibition enhanced DC661-mediated cytotoxicity (LDLR, SRB1, UGCG, and ceramide synthase 2) increased in a concentration- and time-dependent manner following DC661 treatment (Fig. 2M and N).

Figure 2.

Inhibition of either cholesterol uptake receptors or sphingolipid salvage pathway enzymes augments DC661 cytotoxicity to A375P cells. A, MTT assay graph of 3-hour pretreatment of 2 mmol/L MBCD followed by the addition of DC661 (0.3 μmol/L) or DC661 + water-soluble cholesterol (Chol; 8 μmol/L). B, MTT assay plot with increasing concentrations of DC661 (0.01–10 μmol/L), with and without simvastatin. C, MTT assay graph of 2-hour pretreatment with anti-LDLR (5 μg/mL) or its isotype control IgG (5 μg/mL) followed by the addition of DC661 (0.3 μmol/L). D–F, Representative MTT assay plots with increasing concentrations of DC661 (0.01–10 μmol/L), with and without (D) BLT-1, (E) Myriocin, and (F) Fumonisin B1 (FB1). G, Fold change increase in the average peak area values (3 replicates) of sphingosine (d18:1) ceramide species increased by DC661 and HCQ treatments and synthesized by different isoforms of ceramide synthase (CerS), CerS1-6. H–I, MTT assay plots with increasing concentrations of DC661 (0.01–10 μmol/L) with and without (H) specific CerS2 inhibitor ST1074 and (I) acid sphingomyelinase inhibitor siramesine (Sira). J, Fold-change increase in the average total peak area (3 replicates) values of different species of Hex1Cer induced by DC661 and HCQ. K, MTT assay of DC661 (0.01–10 μmol/L) with and without UGCG inhibitor Genz-123346 (Genz). L, Bar graph showing average IC50 values ± SEM of MTT assays from three independent experiments. Dashed line partitions sphingolipid and cholesterol pathways. M and N, Representative Western blots of whole-cell lysates, probed for sphingolipid metabolism proteins and cholesterol uptake receptors whose inhibition significantly augmented DC661 cytotoxicity in MTT assays, after treatment of A375P cells with (M) different concentrations of DC661 and (N) different durations with DC661 (3 μmol/L). *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001; ****, P ≤ 0.0001; ns, nonsignificant. One-way ANOVA followed by Tukey (A, C) or Dunnett (L) multiple comparisons procedures.

Figure 2.

Inhibition of either cholesterol uptake receptors or sphingolipid salvage pathway enzymes augments DC661 cytotoxicity to A375P cells. A, MTT assay graph of 3-hour pretreatment of 2 mmol/L MBCD followed by the addition of DC661 (0.3 μmol/L) or DC661 + water-soluble cholesterol (Chol; 8 μmol/L). B, MTT assay plot with increasing concentrations of DC661 (0.01–10 μmol/L), with and without simvastatin. C, MTT assay graph of 2-hour pretreatment with anti-LDLR (5 μg/mL) or its isotype control IgG (5 μg/mL) followed by the addition of DC661 (0.3 μmol/L). D–F, Representative MTT assay plots with increasing concentrations of DC661 (0.01–10 μmol/L), with and without (D) BLT-1, (E) Myriocin, and (F) Fumonisin B1 (FB1). G, Fold change increase in the average peak area values (3 replicates) of sphingosine (d18:1) ceramide species increased by DC661 and HCQ treatments and synthesized by different isoforms of ceramide synthase (CerS), CerS1-6. H–I, MTT assay plots with increasing concentrations of DC661 (0.01–10 μmol/L) with and without (H) specific CerS2 inhibitor ST1074 and (I) acid sphingomyelinase inhibitor siramesine (Sira). J, Fold-change increase in the average total peak area (3 replicates) values of different species of Hex1Cer induced by DC661 and HCQ. K, MTT assay of DC661 (0.01–10 μmol/L) with and without UGCG inhibitor Genz-123346 (Genz). L, Bar graph showing average IC50 values ± SEM of MTT assays from three independent experiments. Dashed line partitions sphingolipid and cholesterol pathways. M and N, Representative Western blots of whole-cell lysates, probed for sphingolipid metabolism proteins and cholesterol uptake receptors whose inhibition significantly augmented DC661 cytotoxicity in MTT assays, after treatment of A375P cells with (M) different concentrations of DC661 and (N) different durations with DC661 (3 μmol/L). *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001; ****, P ≤ 0.0001; ns, nonsignificant. One-way ANOVA followed by Tukey (A, C) or Dunnett (L) multiple comparisons procedures.

Close modal

Of the proteins in the cholesterol and sphingolipid salvage pathways whose inhibition augmented DC661 cytotoxicity, we focused on UGCG because UGCG and hexosylceramide levels showed the largest and most significant treatment-related fold changes in the proteomics and lipidomics data, respectively (Fig. 1D and F). To determine if UGCG also increases with LAI in other cancer cells, we treated DLD-1, a colorectal cancer cell line; MIA PaCa-2, a pancreatic cancer cell line; A549, a lung cancer cell line; and WM4380, a melanoma cell line generated from a BRAFV600E patient-derived xenograft (PDX), with DC661 and observed a concentration-dependent increase in UGCG (Supplementary Fig. S2E). High concentrations of HCQ, a weaker lysosomal inhibitor, also increased UGCG in A375P cells (Supplementary Fig. S2F). To determine if UGCG levels also increase in response to nonlysosomal autophagy inhibition, we treated A375P cells with the autophagy inhibitors SBI-0206965 [which targets Unc-51 like autophagy activating kinase 1 (ULK1)] or SAR405 [which targets vacuolar sorting protein-34 (VPS34)] and found that concentrations which effectively blocked autophagy induction at the level of autophagic vesicles formation did not increase the UGCG levels (Supplementary Fig. S2G and S2H). Further, unlike DC661 + Genz-123346 or DC661 + eliglustat combination treatments which produced augmented cytotoxicity compared with single-agent cytotoxicity in long-term colony formation assays, the addition of Genz-123346 or eliglustat to either SBI-0206965 or SAR405 did not produce augmented cytotoxicity to A375P cells (Supplementary Fig. S2I).

LAI-Induced Lipid Remodeling Is Associated with the Formation of GMM

Because our proteomics, lipidomics, imaging, and chemical inhibition studies pointed to DC661-induced increases in UGCG activity, sphingolipid accumulation, and cholesterol accumulation on plasma membrane and intracellular compartments, we investigated the possibility of these changes reflecting the accumulation of lipid microdomains. Previous work has demonstrated that detergent-resistant, cholesterol (filipin+) and complex glycosphingolipid monosialotetrahexosylganglioside (GM1)–containing lipid microdomains can be identified on lysosomal (22–24) and plasma membranes (25). The detection of GM1 using cholera toxin B-subunit (CTxB) provides a means to visualize the presence of these microdomains (22, 26–28). Immunofluorescence microscopy of a labeled CTxB demonstrated increased GM1+ signals in both plasma membrane and lysosomes following DC661 and HCQ treatments (Fig. 3A; Supplementary Fig. S3A). DC661 or HCQ treatment also increased levels of the prototypical membrane lipid microdomain marker proteins flotillin-1 and flotillin-2 (29) in a concentration-dependent manner in the whole-cell lysates of A375P cells (Fig. 3B; Supplementary Fig. S3B). Flotillin-1 and flotillin-2 were also found to be significantly increased in the proteome of detergent-resistant membrane fractions of DC661-treated A375P cells (Supplementary Fig. S3C). We next used high-resolution confocal microscopy with deconvolution and the CTxB probe to demonstrate that increased CTxB signal following DC661 treatment was partitioned as microdomains in the plasma membrane and was not diffusely distributed across the membrane (Fig. 3C). CTxB+ GMM on the plasma membrane also contained flotillin-2 (Fig. 3D). Focusing on the lysosomes, both high-resolution confocal microscopy and super-resolution stimulated emission depletion (STED) microscopy with deconvolution demonstrated that following DC661 treatment, increased CTxB signal was arrayed on ∼0.5 to 1 μm (average diameter of a lysosome) circular structures alternating with LAMP1. These structures suggest that GM1 was not accumulating within the lumen of the lysosome but rather increasing on the surface after LAI (Fig. 3E and F). Notably, LAMP1 has been reported to be a lysosomal membrane protein which is excluded from flotillin+ lipid microdomains (28, 30, 31); therefore, the lack of colocalization at the super-resolution level provides further evidence that the CTxB signal reflects distinct membrane lipid microdomains. Even though CTxB and LAMP1 do not colocalize at the super-resolution level, their appearance of colocalization (Fig. 3A) in standard confocal microscopy reflects the proximity of GMM and LAMP1 on lysosomes. As super-resolution was not necessary for the conclusions drawn in subsequent experiments, we used standard confocal microscopy to detect GMM.

Figure 3.

LAI increases GMM. A, Representative confocal images of A375P cells treated with DMSO (vehicle control) or DC661 (1 μmol/L) for 24 hours and probed for GM1 with Alexa Fluor 488–labeled cholera toxin-subunit B (CTxB) and lysosome (LAMP1 antibody); yellow arrows: colocalization. Mean ± SEM of Mander's correlation coefficients for colocalization of LAMP1 over CTxB from three separate image fields with a total of at least 20 cells/group. B, Immunoblot of lysates from A375P cells treated with different concentrations of DC661 for 24 hours and probed for flotillin-1 and flotillin-2. In C–F, A375P cells were treated with DC661 (1 μmol/L) for 24 hours. C–E, Representative high-resolution confocal images (350–500× magnification and deconvolved) of GM1 (CTxB) with (C) membrane (Membrite stain), (D) flotillin-2 (maximum intensity projection image of 3 planes), and (E) LAMP1. Microdomain localization of CTxb (green arrow; C); colocalization of Flotillin-2 and CTxB (yellow arrow; D); and lack of colocalization of LAMP1 (red arrow) and CTxB (green arrow) in super-resolution confocal microscopy (E). F, Super-resolution STED images (400× magnification and deconvolved) of GM1 (CTxB) and LAMP1. Lack of colocalization of LAMP1 (purple arrow) and CtxB (yellow arrow) in STED microscopy. G, A375P cells were given PPT1 siRNA or control scrambled siRNA (Scr siRNA) for 72 hours and labeled for GM1 (CTxB) and lysosomes (LAMP1); yellow arrow: colocalization. H–J, Representative confocal images of (H) DLD-1, (I) MIA PaCa-2, and (J) A549 cancer cells treated with DC661 (0.6 μmol/L for DLD-1, 1 μmol/L for MIA PaCa-2, and A549 cells) for 24 hours and were labeled for GM1 (CTxB) and lysosomes (LAMP1); yellow arrow: colocalization. All images are representative of at least two independent experiments. Scale: 10 μm (A, GJ); 5 μm, inset 0.5 μm (CF). **, P ≤ 0.01; Student t test (A).

Figure 3.

LAI increases GMM. A, Representative confocal images of A375P cells treated with DMSO (vehicle control) or DC661 (1 μmol/L) for 24 hours and probed for GM1 with Alexa Fluor 488–labeled cholera toxin-subunit B (CTxB) and lysosome (LAMP1 antibody); yellow arrows: colocalization. Mean ± SEM of Mander's correlation coefficients for colocalization of LAMP1 over CTxB from three separate image fields with a total of at least 20 cells/group. B, Immunoblot of lysates from A375P cells treated with different concentrations of DC661 for 24 hours and probed for flotillin-1 and flotillin-2. In C–F, A375P cells were treated with DC661 (1 μmol/L) for 24 hours. C–E, Representative high-resolution confocal images (350–500× magnification and deconvolved) of GM1 (CTxB) with (C) membrane (Membrite stain), (D) flotillin-2 (maximum intensity projection image of 3 planes), and (E) LAMP1. Microdomain localization of CTxb (green arrow; C); colocalization of Flotillin-2 and CTxB (yellow arrow; D); and lack of colocalization of LAMP1 (red arrow) and CTxB (green arrow) in super-resolution confocal microscopy (E). F, Super-resolution STED images (400× magnification and deconvolved) of GM1 (CTxB) and LAMP1. Lack of colocalization of LAMP1 (purple arrow) and CtxB (yellow arrow) in STED microscopy. G, A375P cells were given PPT1 siRNA or control scrambled siRNA (Scr siRNA) for 72 hours and labeled for GM1 (CTxB) and lysosomes (LAMP1); yellow arrow: colocalization. H–J, Representative confocal images of (H) DLD-1, (I) MIA PaCa-2, and (J) A549 cancer cells treated with DC661 (0.6 μmol/L for DLD-1, 1 μmol/L for MIA PaCa-2, and A549 cells) for 24 hours and were labeled for GM1 (CTxB) and lysosomes (LAMP1); yellow arrow: colocalization. All images are representative of at least two independent experiments. Scale: 10 μm (A, GJ); 5 μm, inset 0.5 μm (CF). **, P ≤ 0.01; Student t test (A).

Close modal

PPT1 is the molecular target of chloroquine derivatives (5). Like DC661, knockdown of PPT1 in A375P cells increased GMM formation on plasma membrane and lysosomes (Fig. 3G; Supplementary Fig. S3D and S3E). DC661 treatment increased GMM formation on plasma membranes and lysosomes in multiple human cancer cell lines, including PDX-WM4380 (Supplementary Fig. S3F), DLD-1, MIA PaCa-2, and A549 (Fig. 3HJ; Supplementary Fig. S3G). Altogether, these results demonstrated that LAI upregulates UGCG and is associated with assembly of GMM in plasma membrane and lysosomes of multiple cancer lines.

UGCG Inhibition Augments Lysosomal Membrane Permeabilization and Abrogates DC661-Induced GMM Formation

A long-term colony formation assay showed that the combination of DC661 and Genz-123346 significantly reduced colony formation compared with either DC661 or Genz-123346 alone in melanoma, lung cancer, pancreatic cancer, and colon cancer cell lines (Fig. 4A). The cytotoxicity of DC661 + Genz-123346 was found to be synergistic in both human (A375P) and mouse (B16-F10) melanoma cell lines, respectively (Fig. 4B; Supplementary Fig. S4A). To check the effects of UGCG inhibition on GMMs, A375P cells were treated with vehicle control, DC661, Genz-123346, eliglustat, or the combinations and were stained with CTxB. Quantitative analysis of images showed that DC661-induced GMM formation in both the plasma membrane and lysosomes was significantly reduced by the addition of Genz-123346 (Fig. 4C), and a similar trend was observed with the DC661 and eliglustat combination (Supplementary Fig. S4B). Further, the addition of Genz-123346 to DC661 abrogated the increased cholesterol level induced by DC661 (Supplementary Fig. S4C). DC661 is known to induce lysosomal membrane permeabilization (LMP), which can be imaged by measuring galectin-3+ puncta in cells (refs. 5, 32; Supplementary Fig. S4D). In A375P cells, the combinations of DC661 + Genz-123346 or DC661 + eliglustat significantly increased the percentage of galectin-3 puncta-positive cells compared with DC661 alone (Fig. 4D; Supplementary Fig. S4E), demonstrating that augmented cytotoxicity in DC661 + UGCG inhibition groups was associated with increase in LMP and reduction of GMM. Furthermore, this augmented LMP in the DC661 and Genz-123346 combined treatment group was associated with increased lysosomal pH (Supplementary Fig. S4F) and cleaved caspase-3 levels relative to the single treatment groups (Fig. 4E; Supplementary Fig. S4G). We investigated whether this synergistic toxicity of DC661 and Genz-123346 combination was due to a compensatory rise in ceramides resulting from the inhibition of UGCG. Our lipidomics results showed that although DC661 + Genz-123346 significantly reduced hexosylceramide levels, as expected, there was no significant increase in ceramide levels compared with either DC661 or Genz-123346 (Supplementary Fig. S4H and S4I), demonstrating that increased ceramide levels due to the combination treatment could not explain the induction of apoptosis, LMP, or cytotoxicity. To complement the chemical inhibition of UGCG, we used a siRNA-based genetic approach. Immunoblotting demonstrated an efficient knockdown of UGCG by pooled siRNA (Supplementary Fig. S4J). Like chemical inhibition with Genz-123346, in a long-term colony formation assay of A375P cells, DC661 + UGCG knockdown completely abolished colony formation (Fig. 4F). siUGCG reduced GMM (Fig. 4G) and significantly augmented DC661-mediated LMP (Fig. 4H) compared with DC661 or control scramble siRNA treatment in A375P cells. Also, increased cleaved caspase-3 was observed in A375P cells treated with DC661 and siUGCG (Fig. 4I). Taken together, these results demonstrate that UGCG inhibition augments DC661-induced cytotoxicity, LMP, and activation of apoptosis, and is associated with a reduction in GMM formation.

Figure 4.

Chemical or genetic inhibition of UGCG synergistically augments DC661-mediated cytotoxicity and abrogates DC661-induced GMM formation. A, Seven-day colony formation assays in different cancer lines treated with DC661 (0.3 μmol/L for A375P and 0.1 μmol/L for other lines), Genz-123346 (Genz; 5 μmol/L for B16-F10 and A549; 10 μmol/L for A375P, MIA PaCa2, and DLD-1) or their combination. B, BLISS synergy and antagonism 3D plots of human (A375P) and mouse (B16-F10) melanoma cells treated with indicated combinations of Genz and DC661. C and D, Representative confocal images of A375P cells or A375P-galectin-3-GFP cells treated with DC661 (1 μmol/L), Genz (20 μmol/L), or combination for 24 hours. C, GM1 (CTxB) and lysosomes (LAMP1); yellow arrow: colocalization. Quantification of CTxB intensity from at least 50 single cells/experimental group (each dot represents a single cell). D, A375P-galectin-3-GFP cells with galectin-3 (Gal3) puncta (white arrows) and quantification. E, Immunoblotting for cleaved caspase-3 (c-csp3) in the lysates of A375P cells. F–I, For UGCG genetic inhibition, A375P cells or A375P-galectin-3-GFP cells were given UGCG siRNA or control scrambled siRNA (Scr siRNA) for 48 hours, followed by treatment with either DMSO or DC661 (1 μmol/L) for 24 hours. F, Colony formation assay in A375P cells treated with DC661 (0.6 μmol/L), Genz (10 μmol/L), or combination following the UGCG knockdown as shown in the panel. G, Representative confocal images of cells labeled for GM1 (CTxB) and lysosomes (LAMP1); yellow arrow: colocalization. H, A375P-galectin-3-GFP cells with Gal3 puncta (white arrows) and quantification. I, Immunoblot of c-csp3 in the whole-cell lysates of A375P cells. All experiments were repeated at least twice. Scale: 10 μm (C, G), 20 μm (D, H). ****, P ≤ 0.0001; ns, nonsignificant. Welch ANOVA followed by Dunnett T3 multiple comparisons procedure (C). In D and H, each dot represents the percentage of puncta+ cells in an image field, a total of 100 cells were counted/group. We did not observe any significant number of puncta+ cells in Genz, Scr, and UGCG siRNA alone control groups. We excluded control groups and performed a two-tailed t test between two groups (D and H).

Figure 4.

Chemical or genetic inhibition of UGCG synergistically augments DC661-mediated cytotoxicity and abrogates DC661-induced GMM formation. A, Seven-day colony formation assays in different cancer lines treated with DC661 (0.3 μmol/L for A375P and 0.1 μmol/L for other lines), Genz-123346 (Genz; 5 μmol/L for B16-F10 and A549; 10 μmol/L for A375P, MIA PaCa2, and DLD-1) or their combination. B, BLISS synergy and antagonism 3D plots of human (A375P) and mouse (B16-F10) melanoma cells treated with indicated combinations of Genz and DC661. C and D, Representative confocal images of A375P cells or A375P-galectin-3-GFP cells treated with DC661 (1 μmol/L), Genz (20 μmol/L), or combination for 24 hours. C, GM1 (CTxB) and lysosomes (LAMP1); yellow arrow: colocalization. Quantification of CTxB intensity from at least 50 single cells/experimental group (each dot represents a single cell). D, A375P-galectin-3-GFP cells with galectin-3 (Gal3) puncta (white arrows) and quantification. E, Immunoblotting for cleaved caspase-3 (c-csp3) in the lysates of A375P cells. F–I, For UGCG genetic inhibition, A375P cells or A375P-galectin-3-GFP cells were given UGCG siRNA or control scrambled siRNA (Scr siRNA) for 48 hours, followed by treatment with either DMSO or DC661 (1 μmol/L) for 24 hours. F, Colony formation assay in A375P cells treated with DC661 (0.6 μmol/L), Genz (10 μmol/L), or combination following the UGCG knockdown as shown in the panel. G, Representative confocal images of cells labeled for GM1 (CTxB) and lysosomes (LAMP1); yellow arrow: colocalization. H, A375P-galectin-3-GFP cells with Gal3 puncta (white arrows) and quantification. I, Immunoblot of c-csp3 in the whole-cell lysates of A375P cells. All experiments were repeated at least twice. Scale: 10 μm (C, G), 20 μm (D, H). ****, P ≤ 0.0001; ns, nonsignificant. Welch ANOVA followed by Dunnett T3 multiple comparisons procedure (C). In D and H, each dot represents the percentage of puncta+ cells in an image field, a total of 100 cells were counted/group. We did not observe any significant number of puncta+ cells in Genz, Scr, and UGCG siRNA alone control groups. We excluded control groups and performed a two-tailed t test between two groups (D and H).

Close modal

UGCG Is a Driver of Resistance to LAI

We next determined the effects of UGCG overexpression in A375P cells. Overexpression of UGCG in A375P cells was confirmed using immunoblotting (Fig. 5A). UGCG overexpression led to a significant resistance to LAI cytotoxicity in 72 hours (Fig. 5B) as well as in 7-day cell culture assays (Fig. 5C). Further, UGCG overexpression increased the levels of GMM in the plasma membrane and lysosomes (Fig. 5D) and decreased DC661-mediated LMP (Fig. 5E) and cleaved caspase-3 levels (Fig. 5F). To determine if UGCG is a cancer-dependency gene, we interrogated the DepMap database (33). UGCG is a dependency gene in multiple cancer cell lines of different lineages, as characterized by CRISPR-mediated knockout of UGCG affecting the survival of these cancer lines (Fig. 5G). Across all cancer types, 42 of 1,054 cell lines (3.9%) had a probability of dependency based on a chronos score ≤ −0.5, and 207 of 1,054 (19.6%) cell lines had a probability of dependency based on a chronos score ≤ −0.25. Skin cancer cell lines demonstrated the highest incidence of UGCG dependency. In patients with melanoma in The Cancer Genome Atlas (TCGA; n = 211), we observed that UGCG expression in lymph node tissue was related to disease-specific survival (Fig. 5H). Patients with melanoma in the high UGCG expression group had a lower five-year survival rate (55.8%, n = 47) compared with patients with low UGCG expression (68.4%, n = 164). There were no significant differences in clinicopathologic variables among patients with low and high UGCG expression (Supplementary Table S1). In the validation cohort of similar patients [n = 130; Gene Expression Omnibus (GEO) database accession GSE65904; ref. 34], the two Kaplan–Meier disease-specific survival curves were significantly different (P = 0.03; Fig. 5I; Supplementary Table S2).

Figure 5.

UGCG expression in therapy resistance, melanoma cell survival, and disease-specific survival of patients with melanoma. A–F, A375P cells expressing vector (Vector) or a UGCG-DDK plasmid (UGCG) were used. A, Immunoblot in the whole-cell lysates confirming overexpression of UGCG. B, Three-day MTT assay with DC661. C, Seven-day colony formation assay with DC661 (0.6 μmol/L). In the experiments from D–F, DC661 (3 μmol/L) was used to treat the cells for 24 hours. D, GM1 (CTxB) and lysosomes (LAMP1), and colocalization (yellow). E, Representative images of A375P cells overexpressing UGCG and galectin-3-GFP; white arrows indicate galectin-3 (Gal3) puncta+ cells. F, Immunoblot of cleaved caspase-3 (c-csp3). G, Box plots show the chronos score (cancer dependency) for UGCG knockout in multiple cancer lines (numbers of cell lines shown in the parenthesis) of different lineages (Y-axis) in The Cancer Dependency Map (DepMap). The cell lines (shown by dots) with chronos score <0 (red line) show UGCG dependency. H, Kaplan–Meier survival curves showing the disease-specific survival (years) of patients with melanoma in the Training Data Set (n = 211) with low and high UGCG levels. I, Kaplan–Meier survival curves showing the disease-specific survival (years) of patients with melanoma (n = 130) in the validation data set with low and high UGCG levels. Scale, 10 μm (D), 20 μm (E). ***, P ≤ 0.001; ****, P ≤ 0.0001. Two-way ANOVA followed by Tukey multiple comparisons procedure (B). Log-rank (Mantel–Cox) test (B).

Figure 5.

UGCG expression in therapy resistance, melanoma cell survival, and disease-specific survival of patients with melanoma. A–F, A375P cells expressing vector (Vector) or a UGCG-DDK plasmid (UGCG) were used. A, Immunoblot in the whole-cell lysates confirming overexpression of UGCG. B, Three-day MTT assay with DC661. C, Seven-day colony formation assay with DC661 (0.6 μmol/L). In the experiments from D–F, DC661 (3 μmol/L) was used to treat the cells for 24 hours. D, GM1 (CTxB) and lysosomes (LAMP1), and colocalization (yellow). E, Representative images of A375P cells overexpressing UGCG and galectin-3-GFP; white arrows indicate galectin-3 (Gal3) puncta+ cells. F, Immunoblot of cleaved caspase-3 (c-csp3). G, Box plots show the chronos score (cancer dependency) for UGCG knockout in multiple cancer lines (numbers of cell lines shown in the parenthesis) of different lineages (Y-axis) in The Cancer Dependency Map (DepMap). The cell lines (shown by dots) with chronos score <0 (red line) show UGCG dependency. H, Kaplan–Meier survival curves showing the disease-specific survival (years) of patients with melanoma in the Training Data Set (n = 211) with low and high UGCG levels. I, Kaplan–Meier survival curves showing the disease-specific survival (years) of patients with melanoma (n = 130) in the validation data set with low and high UGCG levels. Scale, 10 μm (D), 20 μm (E). ***, P ≤ 0.001; ****, P ≤ 0.0001. Two-way ANOVA followed by Tukey multiple comparisons procedure (B). Log-rank (Mantel–Cox) test (B).

Close modal

UGCG Inhibition Combined with LAI Impairs Tumor Growth in a Syngeneic Tumor Model

To determine the effects of targeting UGCG in vivo, we first determined the effects of UGCG inhibition on DC661 cytotoxicity in B16-F10 melanoma mouse cells, which we used to generate tumors in immunocompetent C57BL/6 mice. Like A375P cells, DC661 treatment increased UGCG levels (Fig. 6A), and B16-F10 cells treated with DC661 and Genz-123346 showed significantly increased cytotoxicity compared with DC661 (Fig. 6B). DC661-induced changes in UGCG levels were also correlated with an increase in membrane lipid microdomain marker proteins flotillin-1 and flotillin-2 (Fig. 6C). To model therapy resistance in vivo, we grew B16-F10 melanoma tumors and did not start treatment until tumors were large (150 mm3). Clinically, large tumors are less likely to respond to targeted therapy or immunotherapy, and our previous preclinical studies demonstrated that this was also true for autophagy inhibition (35, 36). We used an FDA-approved highly selective UGCG inhibitor, eliglustat. Though high-dose HCQ (60 mg/kg) given daily, low-dose DC661 (0.5 mg/kg) given once every two days, or low-dose daily eliglustat (30 mg/kg) demonstrated little single-agent efficacy, low-dose intermittent DC661 + eliglustat and high-dose continuous HCQ + eliglustat demonstrated significant tumor growth inhibition (Fig. 6D). At euthanasia, the combination treatment significantly reduced tumor weight and size (Fig. 6E and F). Importantly, these combinations were well tolerated with no significant change in body weight after the treatments were started (Fig. 6G), demonstrating that the reliance on UGCG and presumably on sphingolipid metabolism following lysosomal inhibition was more prominent in tumor cells compared with cells of other vital organs. We next measured GMM in the harvested tumor tissue. Concordant with our in vitro findings, whereas DC661 and HCQ treatment induced GMM formation in the tumor tissues, the combination of eliglustat with either DC661 or HCQ significantly abrogated increased GMM formation (Fig. 6H).

Figure 6.

Antitumor activity of UGCG inhibition and LAI in immunocompetent mice bearing melanoma tumors. A, Immunoblots of UGCG in the whole-cell lysates from B16-F10 cells treated with different concentrations of DC661 for 24 hours. B, MTT assay plot of B16-F10 cells treated with DC661 (0.01–10 μmol/L) ± Genz-123346 (Genz). C, Immunoblots of flotillin 1 and 2 in the lysates from B16-F10 cells treated with different concentrations of DC661 for 24 hours. D, B16-F10 average tumor volumes ± SEM in different cohorts (n = 8) of syngeneic C57BL/6J mice, measured every day following the treatment plan. Adjusted P values are shown for measurements on day 9. E, B16-F10 tumor weights at the end of the experiment. F, Representative pictures of animals from each group with tumors shown in red circles, and isolated tumor pictures from each animal with a scale bar of 3 cm. G, Percent change in the average body weights in each cohort of tumor-bearing animals with respect to day 1 of treatment; body weights were measured every 2 days as shown. H, Representative immunofluorescence images and quantification (mean ± SEM) in the tumor tissue sections labeled for GM1 (CTxB) and lysosome (LAMP1) with their merged pixels shown in yellow. Each circle in the bar represents an animal and the average intensity of fluorescent-labeled CTxB; at least 80 cells were counted/animal. Scale: 50 μm. *, P ≤ 0.05; **, P ≤ 0.01; ****, P ≤ 0.0001; ns, not significant. Two-way ANOVA followed by Tukey multiple comparisons procedure (D); Welch ANOVA test followed by Dunnett T3 multiple comparisons procedure (E); one-way ANOVA followed by Tukey multiple comparisons procedure (H).

Figure 6.

Antitumor activity of UGCG inhibition and LAI in immunocompetent mice bearing melanoma tumors. A, Immunoblots of UGCG in the whole-cell lysates from B16-F10 cells treated with different concentrations of DC661 for 24 hours. B, MTT assay plot of B16-F10 cells treated with DC661 (0.01–10 μmol/L) ± Genz-123346 (Genz). C, Immunoblots of flotillin 1 and 2 in the lysates from B16-F10 cells treated with different concentrations of DC661 for 24 hours. D, B16-F10 average tumor volumes ± SEM in different cohorts (n = 8) of syngeneic C57BL/6J mice, measured every day following the treatment plan. Adjusted P values are shown for measurements on day 9. E, B16-F10 tumor weights at the end of the experiment. F, Representative pictures of animals from each group with tumors shown in red circles, and isolated tumor pictures from each animal with a scale bar of 3 cm. G, Percent change in the average body weights in each cohort of tumor-bearing animals with respect to day 1 of treatment; body weights were measured every 2 days as shown. H, Representative immunofluorescence images and quantification (mean ± SEM) in the tumor tissue sections labeled for GM1 (CTxB) and lysosome (LAMP1) with their merged pixels shown in yellow. Each circle in the bar represents an animal and the average intensity of fluorescent-labeled CTxB; at least 80 cells were counted/animal. Scale: 50 μm. *, P ≤ 0.05; **, P ≤ 0.01; ****, P ≤ 0.0001; ns, not significant. Two-way ANOVA followed by Tukey multiple comparisons procedure (D); Welch ANOVA test followed by Dunnett T3 multiple comparisons procedure (E); one-way ANOVA followed by Tukey multiple comparisons procedure (H).

Close modal

UGCG Inhibition Combined with LAI Impairs Tumor Growth and Increases Survival in Xenograft and PDX Tumor Models

A375P cells were implanted into the flanks of NOD/SCID animals, and mice were randomized to treatment cohorts when tumors reached an average of 100 mm3. Similar to B16-F10 tumors, the eliglustat and DC661 combination produced significant tumor growth inhibition (∼80%) in mice bearing A375P tumors compared with either treatment alone (Fig. 7A). The eliglustat and DC661 combination significantly prolonged the survival of tumor-bearing mice compared with DC661 alone (median 34 days vs. 17 days: log-rank P < 0.0001; Fig. 7B). The combination of eliglustat (60 mg/kg) and DC661 (1 mg/kg) was also tolerable in animals for long-term treatment with no significant body weight change after the treatments were started (Fig. 7C). To test this combination in a more clinically relevant tumor model, we utilized the PDX-WM4552 cell line, which was derived from a patient who never responded and progressed on the combination of HCQ with the BRAF inhibitor dabrafenib and the MEK inhibitor trametinib in the BAMM trial (NCT02257424; ref. 37). The combination of Genz-123346 and DC661 significantly enhanced the cytotoxicity of DC661 in PDX-WM4552 cells (Fig. 7D and E). In PDX-WM4552 cells, DC661 treatment increased UGCG levels (Supplementary Fig. S5A), and Genz-123346 treatment completely abrogated DC661-induced GMM formation on plasma membranes and lysosomes (Supplementary Fig. S5B). We previously showed that PDX-WM4552 grown on the flanks of NOD/SCID mice was highly resistant to dabrafenib + trametinib treatment (38). To determine if combined UGCG inhibitor and LAI had activity in this highly therapy-refractory tumor model, we generated PDX-WM4552 tumors on the flanks of NOD/SCID mice and randomized mice into treatment cohorts when tumors reached an average size of 135 mm3. Treatment with either DC661 (3 mg/kg, every two days) or eliglustat (30 mg/kg, every day) alone produced minimal tumor growth inhibition, whereas the combination of DC661 and eliglustat produced significant tumor growth inhibition (86%; Fig. 7F). When control tumors reached an average volume of 2500 mm3, tumor weight and size demonstrated the activity of this combination (Supplementary Fig. S5C–S5E). DC661 and eliglustat treatment significantly improved survival of mice compared with DC661 alone (median 20 days vs. 9 days: log-rank P < 0.0011; Fig. 7G). The DC661 and eliglustat combination was tolerable to animals with no significant body weight changes (Fig. 7H). Taken together, these findings demonstrate that LAI induces UGCG-dependent lipid remodeling that resists LMP leading to cancer cell survival. These LAI-induced changes were also correlated with increased GMM formation on plasma membrane and lysosomes. The UGCG inhibitor eliglustat reverses the LAI-induced changes in glycosphingolipids, abrogates GMM formation, and augments LAI-induced LMP, resulting in cancer cell death (Fig. 7I).

Figure 7.

UGCG inhibition combined with LAI impairs tumor growth in xenograft and therapy-resistant PDX tumor models. A–C, A375P tumors were generated in the flanks of NOD/SCID mice, and animals were treated with either vehicle, DC661 (1 mg/kg, every 2 days), eliglustat (60 mg/kg, every day), or the combination as shown in A. A, Tumor volumes (mean ± SEM) in different cohorts (n = 10–15) measured every day. Adjusted P values are shown for the measurements on day 18. B, Kaplan–Meier survival curves showing the percentage of tumor-bearing animals that survived in different cohorts (n = 10–11). Animals were removed from the study when tumor size reached ∼2,000 mm3. C, Percent change in the average body weights in each cohort of tumor-bearing animals with respect to day 1 of treatment. Body weights were measured every three days, as shown. D, MTT assay graph for PDX-WM4552 cells treated with DC661 (0–0.3 μmol/L) alone or in combination with Genz-123346 (Genz, 20 μmol/L). E, Seven-day colony formation assay with DC661 (0.1 μmol/L) and Genz (5 μmol/L) in PDX-WM4552 cells. F–H, PDX-WM4552 tumors were generated in the flanks of NOD/SCID mice. After tumors reached 135 mm3, mice were treated with either vehicle, DC661 (3 mg/kg), eliglustat (30 mg/kg), or the combination as shown in F. F, Tumor volumes (mean ± SEM) in different cohorts (n = 8–10) measured every day. Adjusted P values are shown for measurements on day 9. G, Kaplan–Meier survival curves showing the percentage of animals that survived in different cohorts (n = 6–9). H, Percent change in the average body weights in each cohort of tumor-bearing animals with respect to day 1 of treatment. Body weights were measured every 3 days, as shown. I, Graphical sketch illustrating that UGCG-dependent lipid remodeling predominantly increases glycosphingolipid (GSL) level and is associated with the increased formation of GMM on the plasma membrane and lysosomes. UGCG upregulation is a druggable resistance mechanism to LAI-induced LMP-associated cancer cell death. *, P ≤ 0.05; ***, P ≤ 0.001; ****, P ≤ 0.0001; ns, not significant. Two-way ANOVA followed by Tukey multiple comparisons procedure (A, D, and F); log-rank (Mantel–Cox) test (B and G).

Figure 7.

UGCG inhibition combined with LAI impairs tumor growth in xenograft and therapy-resistant PDX tumor models. A–C, A375P tumors were generated in the flanks of NOD/SCID mice, and animals were treated with either vehicle, DC661 (1 mg/kg, every 2 days), eliglustat (60 mg/kg, every day), or the combination as shown in A. A, Tumor volumes (mean ± SEM) in different cohorts (n = 10–15) measured every day. Adjusted P values are shown for the measurements on day 18. B, Kaplan–Meier survival curves showing the percentage of tumor-bearing animals that survived in different cohorts (n = 10–11). Animals were removed from the study when tumor size reached ∼2,000 mm3. C, Percent change in the average body weights in each cohort of tumor-bearing animals with respect to day 1 of treatment. Body weights were measured every three days, as shown. D, MTT assay graph for PDX-WM4552 cells treated with DC661 (0–0.3 μmol/L) alone or in combination with Genz-123346 (Genz, 20 μmol/L). E, Seven-day colony formation assay with DC661 (0.1 μmol/L) and Genz (5 μmol/L) in PDX-WM4552 cells. F–H, PDX-WM4552 tumors were generated in the flanks of NOD/SCID mice. After tumors reached 135 mm3, mice were treated with either vehicle, DC661 (3 mg/kg), eliglustat (30 mg/kg), or the combination as shown in F. F, Tumor volumes (mean ± SEM) in different cohorts (n = 8–10) measured every day. Adjusted P values are shown for measurements on day 9. G, Kaplan–Meier survival curves showing the percentage of animals that survived in different cohorts (n = 6–9). H, Percent change in the average body weights in each cohort of tumor-bearing animals with respect to day 1 of treatment. Body weights were measured every 3 days, as shown. I, Graphical sketch illustrating that UGCG-dependent lipid remodeling predominantly increases glycosphingolipid (GSL) level and is associated with the increased formation of GMM on the plasma membrane and lysosomes. UGCG upregulation is a druggable resistance mechanism to LAI-induced LMP-associated cancer cell death. *, P ≤ 0.05; ***, P ≤ 0.001; ****, P ≤ 0.0001; ns, not significant. Two-way ANOVA followed by Tukey multiple comparisons procedure (A, D, and F); log-rank (Mantel–Cox) test (B and G).

Close modal

HCQ was the first autophagy inhibitor tested as a cancer therapeutic in clinical trials. Single-agent autophagy inhibition with HCQ produced a 0% response rate in a pancreatic cancer clinical trial (39). Even in combination with other standard-of-care anticancer agents, HCQ showed variable results. In some cases, there were no signs of clinical activity (8, 9). There were clear signs of increased response rates in other clinical trials that did not translate into long-term progression-free survival (37, 40, 41). The finding that PPT1 is the target of HCQ paved the way for the development of novel lysosomal autophagy inhibitors such as DC661 (5) and GNS561 (6). PPT1 inhibition by GNS561 has also been shown to induce LMP-mediated cell death in vitro (6). GNS561 has modest single-agent efficacy in tumor models, and a phase I clinical trial of GNS521 showed no responses in patients with treatment-refractory hepatocellular carcinoma (42). The findings here suggest that UGCG-mediated lipid remodeling plays a role in resistance to PPT1 inhibitors. Of note, our limited studies here demonstrated that inhibition of upstream autophagy targets ULK1 or VPS34 did not increase the levels of UGCG, suggesting the compensatory changes in lipid metabolism may be specific to LAI.

Although there are few reports suggesting that cancer cells can adapt to genetic autophagy inhibition (e.g., ATG7 knockout) by upregulating an NRF2-dependent antioxidant response (7) or by altering their mitochondrial dynamics and using mitochondrial-derived vesicles to maintain their functional mitochondria (43), the molecular basis of resistance related to LAI remains underexplored. We demonstrated that LAI induced significant changes in a small percentage of the proteome of melanoma cells with increases in proteins involved in autophagy and lipid metabolism. Further, we found that LAI causes a striking remodeling of the cellular lipidome with enrichment in cholesterol and sphingolipids, especially glycosphingolipids, including GM1. These changes are likely due to tumor cells scavenging cholesterol and sphingolipids from the extracellular matrix to increase GMM formation. Alternatively, GMM may not play a causative role in LAI resistance, and changes in lipid metabolism enzymes and lipid species directed by enzymes such as UGCG may simply facilitate a metabolic adaptation to lysosomal impairment. Regardless of whether or not GMM are critical for lysosomal repair following injury, the ability to measure them on cell surfaces allows the possibility of GMM serving pharmacodynamic biomarkers of LAI and/or UGCG inhibitors. We found GMM formation increased in multiple human cancer cell lines following LAI and was abrogated following UGCG inhibition, suggesting the findings in melanoma models are likely to be applicable broadly across other cancers.

GMM are glycosphingolipid and cholesterol-rich microdomains within lipid membranes. There is a significant controversy over whether GMM can be referred to as lipid rafts, implying a somewhat rigid structure, or are simply transient associations of freely mobile lipid species and proteins within membranes (44). Our super-resolution and standard microscopy experiments together have provided further evidence that the accumulation of GMM observed with DC661 occurs in a domain-like structure and can be reversed with UGCG inhibition. However, we acknowledge that this pattern is merely an association and further work is needed to demonstrate whether GMM are mechanistically critical for lysosomal membrane repair. Regardless, the association between GMM and UGCG activity could be translated into bioassay to measure pharmacologic UGCG inhibition in clinical trials. Targeting UGCG with eliglustat improves symptoms in Gaucher disease (45), a lysosomal storage disorder. Eliglustat is an oral therapy with favorable pharmacokinetics and pharmacodynamics in humans for long-term daily use (45, 46). Interestingly, lysosomal accumulation of glucosylceramides in Gaucher disease, due to a defect in lysosomal glucocerebrosidase, has been hypothesized to promote cancer growth (47). In a mouse model of Gaucher disease that is driven by a mutation in the glucocerebrosidase gene, engraftment of melanoma cells produced larger tumors in comparison with the wild-type animals with melanoma tumor (47). We found that targeting UGCG augmented DC661 cytotoxicity in vitro and in vivo in 3 different melanoma tumor models.

Our findings suggest that pursuing other druggable targets besides UGCG involved in cholesterol and sphingolipid metabolism could augment LAI treatment. In a recent study, SRB1-dependent HDL uptake was reported as a metabolic vulnerability in renal cancer (48). Chemical inhibition of CerS2, which is involved in both de novo and salvage ceramide synthesis, but not serine palmitoyltransferase, the rate-limiting enzyme of de novo ceramide synthesis, augmented DC661 cytotoxicity in MTT assays, confirming the involvement of salvage but not de novo ceramide synthesis after LAI. Notably, ceramides with longer-chain fatty acids synthesized by CerS2 have also been reported to have antiapoptotic functions (49), which could contribute to LAI resistance observed in our study. Importantly, ceramides have been found in the LDL or HDL particles that can be imported into the cell by LDLR or SRB1, respectively (50, 51). Further work is needed to nominate other cholesterol and sphingolipid metabolism enzymes as putative therapeutic targets in melanoma and other cancers. Because UGCG was found to be a cancer-dependency gene, a UGCG inhibitor could have single-agent activity in multiple cancers, as recently shown in prostate cancer (52). The striking in vivo antitumor activity and tolerability of combined LAI and UGCG inhibition suggests that a clinical trial combining eliglustat with HCQ or a more potent PPT1 inhibitor could be launched.

Cell Lines and Cell Culture

The human A375P (CRL-3224), DLD-1 (CCL-221), MIA PaCa-2 (CRL-1420), and A549 (CCL-185) and mouse B16-F10 (CRL-6475) cell lines were purchased from the ATCC. The human melanoma PDX lines WM4380 and WM4552 were obtained from Meenhard Herlyn, Wistar Institute. All cell lines were tested for Mycoplasma biannually by the MycoAlert Mycoplasma Detection Kit (Cambrex Bio Science Rockland, Inc.) and authenticated using short-tandem repeat fingerprinting by University of Pennsylvania Core facilities. The A375P cell line was cultured in RPMI 1640 (Invitrogen, 11875), and other cell lines were cultured in DMEM, supplemented with 10% fetal bovine serum (F6178, Sigma-Aldrich) and 1× antibiotic antimycotic solution (A5955, Sigma-Aldrich). Cells were grown at 37°C in the presence of 5% CO2.

Chemicals and Reagents

Chemicals purchased included HCQ sulfate (Spectrum Chemicals, 747-36-4); DC661 (Selleckchem, S8808); Methyl-Beta-Cyclodextrin (332615, Sigma); Cholesterol-Water Soluble (C4951, Sigma); Simvastatin (10010344, Cayman Chemical); BLT-1 (HY-116767, MedChemExpress); Myriocin (63150-5, Cayman Chemical); Fumonisin B1 (62580-1, Cayman Chemical); Siramesine hydrochloride (6740, Tocris Bioscience); Genz-123346 (BioVision, B2286); Eliglustat hemitartarate (HY-14885A, MedChemExpress LLC); Eliglustat hemitartarate (21487, Cayman Chemical); SBI-0206965 (18477, Cayman Chemical); SAR-405 (16979, Cayman Chemical); and LysoSensor Green DND-189 (Thermo Fisher; L7535). ST1074 was synthesized by Holger Stark and Aleksandra Zivkovic (21).

Proteomics

Melanoma A375P cells were seeded in 60-mm dishes at 0.7 × 106 cells per dish. Frozen cell pellets were lysed with 50 mmol/L Tris pH 7.4, 1% SDS, 150 mmol/L NaCl, 1 mmol/L EDTA, 0.15 mmol/L PMSF, 1 μg/mL pepstatin, and 1 μg/mL leupeptin. Samples were processed using previously published methods (53, 54). Raw MS data were analyzed using MaxQuant 1.6.5.0 (http://www.maxquant.org) with a Uniprot human sequence database (October 10, 2019) and a common contaminants database including trypsin, keratins, bovine proteins, and Mycoplasma (55). Tryptic peptide specificity with a maximum of 2 missed cleavages, fixed modification on cysteine (carbamidomethylation), and variable methionine oxidation or N-terminal acetylation were used in the search (56). A cutoff of 1% false discovery rate (FDR) was used for peptides and proteins. Match between runs was enabled; proteins were quantified using label-free quantitation (57). Statistical analysis was performed using Perseus 1.6.2.3 (http://www-perseus-framework.org; refs. 58, 59). Proteins were required to be identified by at least 3 unique peptides and have 3 valid values (nonzero quantitation) within a sample group, and contaminants and reverse proteins were filtered from the data set. Missing values were imputed from a normal distribution. Pairwise comparisons between conditions were performed at the protein level using Student t test with permutation-based FDR with s0 = 0.1 and 250 randomizations. Significant changes were defined as FDR < 5% and an absolute fold change greater than 1.5 (or 2.0). IPA with default parameters (http://www.ingenuity.com/pa/) was conducted as previously reported (60). Significantly enriched canonical pathways were reported using a threshold of −log (P) of 3 (P < 0.001).

Lipidomics

Melanoma A375P cells were seeded in 60-mm dishes at 0.7 × 106 cells per dish. When cells were at ∼50% confluence, they were treated with DMSO (control), DC661 (3 μmol/L), or HCQ (30 μmol/L) for 24 hours. Samples were processed according to previously published methods (60). Lipids were identified and quantitated using LipidSearch 4.2 (Thermo Fisher Scientific). The identified lipid species were filtered by expected adducts and identification grade based on class. Peak areas were normalized to the EquiSPLASH standards for supported classes and further normalized based on the total area for each sample to correct for variation in cell number. Log-transformed data were analyzed in Perseus 1.6.2.3 to determine lipid fold changes between conditions (58, 59).

Fluorescence Microscopy and Quantification Methods

For cholesterol detection, A375P cells were grown on a polylysine coated 8-well μSlide with #1.5H glass coverslip bottom for high-resolution microscopy (80807, ibidi USA). After the treatments, cells were washed with physiologic buffered solution (A14291DJ, Thermo Fisher Scientific) and fixed with 4% paraformaldehyde (19943, Affymetrix) for 15 minutes. Unused paraformaldehyde was quenched by the 1.5 mg glycine/mL PBS solution for 10 minutes at room temperature. After washing, cells were stained with 0.05 mg/mL filipin complex (F-9765, Sigma) solution in PBS for 2 hours at room temperature. Cells were washed 3× with PBS, and anti-LAMP1 antibody was added without permeabilization and incubated overnight at 4°C to probe for lysosomes. After the primary antibody was washed off, species-specific Alexa-561–labeled secondary antibody was applied and incubated for 1 hour at room temperature. Cells were imaged using Olympus IX71 inverted confocal microscope equipped with Visitech super-resolution VT iSIM scan head (Visitech International Ltd.) and Hamamatsu ORCA Flash 4.0 sCMOS camera with the excitation UV laser (405 nm) for filipin and red laser (561 nm) for Alexa-561, with a 60× water-immersion objective.

For GMM and lysosome detection, cells were grown in a polylysine coated 8-well μSlide with #1.5H glass coverslip bottom for high-resolution microscopy (80807, ibidi USA). A Vybrant Alexa Fluor-488 Lipid Raft Labeling Kit (V34403, Thermo Fisher Scientific) that uses Alexa Fluor-488 labeled CTxB (AF-488-CTxB) was used to detect GM1, as per the manufacturer's protocol. For intracellular GM1 staining, cells were permeabilized after the fixation step of the standard protocol using the 0.1% Triton X-100 solution in PBS (A14291DJ, Thermo Fisher Scientific) for 10 minutes and were labeled again with AF-488-CTxB as per the manufacturer's protocol. For lysosomes, monoclonal LAMP1 primary antibody (sc-28320, Santa Cruz Biotechnology, Inc.) at 1:50 dilution was added overnight at 4°C after the intracellular AF-488-CTxB staining, followed by the standard protocol of immunofluorescence. For GM1 and lysosome detection in animal tumors, a part of tumor tissue was flash-frozen in OCT compound (Sakura Finetek USA, Inc.) immediately after euthanasia. Vybrant Alexa Fluor 594 Lipid Raft Labeling Kit (V34405, Thermo Fisher Scientific) was used to stain GM1 in the frozen tumor tissue sections following the manufacturer's protocol using AF-594-CTxB at 1:100 dilution. Nuclei were stained using NucBlue reagent (Hoechst 33342; R37605, Thermo Scientific) as per the manufacturer's protocol. Secondary antibody was AlexaFluor-594 (A-11032, Thermo Fisher Scientific). Cells were imaged using an Olympus IX71-inverted confocal microscope equipped with Visitech super-resolution VT iSIM scan head (Visitech International Ltd.) and Hamamatsu ORCA Flash 4.0 sCMOS camera with a 100×, 1.40 NA oil-immersion objective.

For image quantifications, mean fluorescence background intensity was subtracted from all images before quantifying raw integrated density. Labeled CTxB and filipin intensity quantifications were done on a single-cell level using ImageJ/FIJI software (61); raw integrated density of fluorescence was measured separately on plasma membrane and intracellular lysosomes by creating separate regions of interest in a cell. A total of at least 50 cells were counted in multiple image fields per condition. Mander's correlation coefficient for colocalization of LAMP1-positive lysosomes over AF-488-CTxB was calculated using JACoP (Just Another Colocalization Plugin; ref. 62) plugin of ImageJ/FIJI software in multiple image fields with at least 25 cells in each experimental group.

High-Resolution Confocal Microscopy with Deconvolution

A375P cells were grown in an 8-well μslide for high-resolution confocal microscopy (80807, ibidi USA). GM1 and lysosomes (LAMP1) were labeled similarly as mentioned in the standard confocal microscopy method above. For labeling membrane lipid microdomain markers GM1 and flotillin-2 together, AF-488-CTxB surface labeling was followed by the addition of a monoclonal flotillin-2 primary antibody (610384, BD Transduction Laboratories) at 1:10 dilution for 2 hours at 4°C followed by fixation and permeabilization (0.01% Triton X-100) to label intracellular GM1 and flotillin-2. Secondary antibody was AlexaFluor-555 (A-21422, Thermo Fisher Scientific). For cell membrane detection with AF-488-CTxB labeling, MemBrite Fix 640/660 cell-surface staining kit (30097-T, Biotium) was used as per the manufacturer's instructions. Nuclei were stained using NucBlue (R37605, Thermo Scientific). Images were acquired with a 100× (1.46 NA) oil-immersion objective lens on a laser scanning Leica SP5 confocal microscope. An additional optical zoom factor of 3.5–5.0× was applied and Z-stacks were acquired at 0.17–0.21 μm intervals. The resulting image stacks were processed in Huygens deconvolution software to yield high-resolution images.

Super-Resolution STED Microscopy with Deconvolution

A375P cells were grown on STED-compatible poly-D-lysine coated precision coverslips (GG-12-15H-PDL, Neuvitro Corporation). GM1 and lysosomes (LAMP1) were labeled similarly as mentioned in the high-resolution confocal microscopy method above. The nuclei staining was omitted. Coverslips were mounted on glass slides using ProLong Gold Antifade Mountant (P36930, Thermo Scientific). Images were acquired with 100×/1.4 NA HC PL APO CS2 oil-immersion objective lens on a Leica TCS SP8-gated STED microscope using an optical zoom factor: 4×, Z-step size: 120 nm, pixel size: 25 nm and scan speed: 600 Hz with unidirectional scanning. All images were deconvolved using Huygens Professional v. 20.10.0p2. A 4-line averaging was done for both STED channels with the following settings for green channel excitation: 490 nm, 8%–10% laser power, green channel depletion laser 592 nm, total power 40% with 60% toward z-depletion, HyD detector using time-gating window 1.2 to 6.5 ns, pinhole set to 0.77 A.U. = 116.38 μm, and for red channel excitation: 555 nm, 8% laser power, red channel depletion laser 660 nm, total power 50 with 60% toward z-depletion, HyD detector using time-gating window 1.5 to 6.1 ns, pinhole set to 0.7 A.U. = 103.1 μm.

For A375P-Galectin-3 cells and A375P-mcherry-GFP-LC3 reporter cells, pEGFP-hGal3 was a gift from Tamotsu Yoshimori (Addgene plasmid #73080; ref. 63) and was used for creating stable A375P-Galectin-3 line for LMP detection. For LMP, 100 cells in multiple image fields were assessed for the percentage of puncta-positive cells. pDEST-CMV mCherry-GFP-LC3B WT plasmid was a gift from Robin Ketteler (Addgene plasmid #123230; ref. 64) and was used for generating stable A375P-mcherry-GFP-LC3 reporter cells. Images were acquired using Olympus IX71 inverted confocal microscope equipped with Visitech super-resolution VT iSIM scan head (Visitech International Ltd.) and Hamamatsu ORCA Flash 4.0 sCMOS camera with a 60×/water objective lens.

For lysosomal lipase assay, the LysoLive Lysosomal Acid Lipase Assay Kit (ab253380) was used following the manufacturer's instructions. Briefly, A375P cells were grown in an 8-well μSlide (80807, ibidi USA) and were treated with vehicle control (DMSO), DC661 (1 μmol/L), or Bafilomycin (20 nmol/L) for 24 hours followed by 2 hours of incubation of LipaGreen substrate. Cells were imaged using a 20× air objective lens on a camera-based confocal microscope. LipaGreen mean fluorescence raw integrated density was measured from ∼400 cells from two independent experiments/group using ImageJ/FIJI software.

Immunoblotting

Treatments were given and cells were collected by scraping. Whole-cell lysates were prepared using 1% SDS lysis buffer (1% SDS; 50 mmol/L Tris; 150 mmol/L NaCl; 1 mmol/L EDTA; 150 mmol/L PMSF; 1 μg/mL Leupeptin; 1 μg/mL Pepstatin; pH 7.5). Protein concentration was measured using Pierce BCA Protein Assay Kit (23225, Thermo Scientific). Protein (15–30 μg) was used for SDS-PAGE and was transferred onto a PVDF membrane (1620177, Bio-Rad). The membrane was blocked using 5% BSA or 5% nonfat dry milk as per the optimized conditions and incubated with primary antibody overnight at 4°C. Primary antibodies included β-Actin (3700, Cell Signaling Technology), UGCG (H00007357-M03, Abnova), Flotillin-1 (610820, BD Transduction Laboratories), Flotillin-2 (sc-28320, Santa Cruz Biotechnology), LDLR (AF2148, R&D Systems), CerSyn2 (SAB2101321, Sigma-Aldrich), Cleaved Caspase-3 (9661, Cell Signaling Technology), SR-B1 (SAB3500048, Sigma-Aldrich), and PPT1 (TA800501, Origene). PVDF membranes were washed with 1× TBS-T (CS-9997, Cell Signaling Technology, Inc.) and incubated for 1 hour at room temperature with species-specific horseradish peroxidase–conjugated secondary antibody (CS-7076S, CS-7074S, Cell Signaling Technology Inc.; or SC-2304, Santa Cruz Biotechnology, Inc.). Membranes were subsequently washed and developed using Pierce ECL Western Blotting substrate (32106, Thermo Scientific) and autoradiography films (XAR ALF 1318, Lab Scientific, Inc.).

Synergy Testing

DC661 and Genz cross-titrations were added to two white 384-well plates (Greiner 781098) using Beckman Coulter ECHO650 instrument (triplicates per plate). B16F10 (DMEM, 10% FBS, P/S) and A375P (RPMI, 10% FBS, P/S) were trypsinized, washed, counted, and added to each 384-well plate at a density of 500 cells/25 μL per well. Assay plates were incubated at 37°C, 5% CO2 for 72 hours. After 72 hours, CellTiter-Glo solution was prepared (Promega #G7573), and 12 μL was added to each well. The assay plates were placed in an orbital shaker for 15 minutes, and luminescence signal measured with BMG Labtech ClarioStarPlus plate reader. For normalized % inhibition calculation, the observed lowest viability value (L) on each plate and the average viability value at dose 0 of two drugs (H) on the same plate were used as controls. Use of O stands for the observed viability at a given combined dose level, the normalized % inhibition was calculated as 100 − (O − L)/(H − L) × 100. Data analysis for drug combination effect was done using Combenefit software with Bliss surface model (65).

Overexpression and Knockdown

Human UGCG pcDNA3.1+/C-(k) DYK (OHu61224D) plasmid and negative control vector backbone pcDNA3.1+/C-(K)-DYK were purchased from GenScript USA Inc. Plasmids were transfected (1 μg/mL) into A375P cells using jetPRIME reagent (101000046, Polyplus-transfection) as per the manufacturer's protocol; transfected cells were stably selected using G418 sulfate antibiotic (10131035, Gibco). UGCG overexpression was confirmed by immunblotting using anti-DYKDDDDK tag antibody (2368S, Cell Signaling Technology). For knockdown experiments, human UGCG siRNA (sc-45404) or human PPT1 siRNA (sc-105216) along with negative control scrambled siRNA (sc-44236) were purchased from Santa Cruz Biotechnology and were transfected (40 nmol/L) into A375P cells using jetPRIME reagent (101000046, Polyplus-transfection) as per the manufacturer's protocol.

Lysosomal Immunoprecipitation

Lysosomes were purified using lysosomal immunoprecipitation (Lyso-IP) method (12). A375P cells were infected with pLJC5-Tmem192-3xHA lentivirus and selected using 1 mg/mL puromycin. Approximately 3 × 106 A375-P-Tmem192-HA cells were treated with either 3 μmol/L DC661 or DMSO vehicle control for 24 hours in culture conditions. After treatment, cells were washed in PBS, harvested in 0.5 mL cold KPBS (136 mmol/L KCl, 10 mmol/L KH2PO4, pH 7.25), and gently homogenized using 20 strokes in a 2 mL Dounce homogenizer. About 2.5% of the homogenate was reserved for whole-cell lysate analysis and the remainder was centrifuged at 3,000 × g for 2 minutes at 4°C to remove cell membrane debris. Homogenate supernatant was then transferred into a clean 1.5 mL tube and was incubated with 50 μL anti-HA beads for 15 minutes at 4°C. Samples were then precipitated by placing tubes on DynaMag (12321D, Thermo Fisher Scientific), and rocked gently for 2 minutes at room temperature. The supernatant was reserved for unbound fraction analysis. The IP fraction was washed 3 times with KPBS containing 8 mmol/L CaCl2 and suspended in PBS for lipidomics analysis.

MTT (3-[4,5-Dimethylthiazol-2-yl]-2, 5 Diphenyl Tetrazolium Bromide) Cell Viability Assay

Two thousand cells were plated in triplicates in each well of a 96-well plate and 3-day MTT assays were performed using the Cell Viability Kit (11465007001, Roche) following the manufacturer's protocol. Cells were cotreated with DC661 ± chemical inhibitors of sphingolipid/cholesterol pathway enzymes. MBCD pretreatment was given for 3 hours followed by cotreatment of the cells with MBCD ± DC661 ± water-soluble cholesterol. LDLR was neutralized by the 2 hours pretreatment with anti-LDLR antibody followed by LDLR ± DC661 treatment. An isotype antibody (AB-108-C, R&D Systems) was used as the control.

Colony Formation Assay

One thousand cells per well were seeded into 12-well plates and were treated the next day; cells were kept under the treatment for a total of 7 days. Colonies formed in each well were washed with PBS, fixed with cold methanol at −20°C for 10 minutes and stained with crystal violet 1% aqueous solution (V5265; Sigma).

Analysis of UGCG Expression and Disease-Specific Survival among Melanoma Patients

The analysis of UGCG expression was done using the z-score normalized expression of UGCG downloaded from the Memorial Sloan Kettering Cancer Center's cBioPortal website based on RNAseqV2 expression data from TCGA patients included in the Pan-Cancer Atlas study (the Learning Data Set). These data were generated by the TCGA Research Network (https://www.cancer.gov/tcga). A cutoff point for UGCG expression was identified using the Contal and O'Quigley method (66) using the maximum Q statistic based on the log-rank test to find a cutoff point for a biomarker such that the survival curves of the low and high groups are most different. The algorithm has been implemented in a SAS macro % FINDCUT by Mandrekar and Meyers (available from: https://www.lexjansen.com/mwsug/2015/DV/MWSUG-2015-DV-07.pdf; ref. 67). UGCG expression for the validation data set is accessible at the NCBI GEO database (accession GSE65904; ref. 34). This analysis used the percentile of the cutoff point from the Learning Data Set to define two groups; the log-rank test was then used to evaluate whether the two disease-specific survival curves were significantly different. Cutoff point identification and survival analyses were done with SAS software v9.4.

In Vivo Mouse Studies

C57BL/6 mice were used for B16-F10 syngeneic, and NOD/SCID gamma mice (The Jackson Laboratory) were used for A375P and PDX-WM4552 xenograft studies. Tumor generation, tumor measurement, and tumor harvesting were performed as previously described (38). Briefly, 0.5 × 106 B16-F10, 5 × 106 A375P, or 1 × 106 PDX-WM4552 cells were subcutaneously injected with an equal volume of Matrigel (354248, Corning) in the right flank of animals. Tumors were measured using electronic calipers, and tumor volume was calculated as V = L × W2 × 0.5. For the survival experiment, mice were sacrificed when signs of morbidity that met ethical criteria for sacrifice were observed or when the tumor reached to an average size of 2,000 to 2,500 mm3. All animal experiments were performed in accordance with the protocols approved by the Institutional Animal Care and Use Committee of the University of Pennsylvania. All compounds were administered via the intraperitoneal route (100 μL injection volume). Eliglustat (HY-14885A, MedChem Express) and HCQ injections were prepared in PBS. DC661 injections were prepared in a formulation of DMSO: PEG 400: Tween 80: Water (10:10:10:70; V/V).

Statistical Analysis

If not described elsewhere, experimental data were analyzed with GraphPad Prism 9 software. Student t test was used to evaluate whether the difference between two group means was statistically significant. The null hypothesis was rejected if the P value was less than 0.05. The ANOVA was used to compare group means when more than two groups were compared. In cases where the group variances were not equal (assessed using the Bartlett test or Brown–Forsythe test), Welch t test or Welch ANOVA test was used.

Data Availability

The proteomics data have been deposited into the MassIVE public repository (https://massive.ucsd.edu/ProteoSAFe/dataset.jsp?task=f98972f2b1c74e6e8e988a5329b5729e) with accession MSV000089020 and the ProteomeXchange repository (http://proteomecentral.proteomexchange.org) with accession PXD032120.

S.L. Harper reports grants from The Wistar Institute during the conduct of the study. A.R. Goldman reports grants from the NIH during the conduct of the study. P.A. Gimotty reports grants from the NIH during the conduct of the study. D.W. Speicher reports grants from the NIH during the conduct of the study. R.K. Amaravadi reports grants and other support from Pinpoint Therapeutics, personal fees from Deciphera and Tasca, and other support from Deciphera, Bristol Myers Squibb, and Novartis outside the submitted work. No disclosures were reported by the other authors.

V. Jain: Conceptualization, data curation, formal analysis, supervision, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. S.L. Harper: Conceptualization, data curation, formal analysis, validation, investigation, methodology, writing–review and editing. A.M. Versace: Data curation, formal analysis, validation, investigation, methodology, writing–review and editing. D. Fingerman: Resources, investigation, methodology. G.S. Brown: Resources, investigation, methodology. M. Bhardwaj: Resources, investigation. M.S. Crissey: Resources. A.R. Goldman: Validation, writing–review and editing. G. Ruthel: Formal analysis, investigation. Q. Liu: Formal analysis. A. Zivkovic: Resources. H. Stark: Resources. M. Herlyn: Resources. P.A. Gimotty: Formal analysis, writing–review and editing. D.W. Speicher: Conceptualization, resources, supervision, investigation, visualization, methodology, project administration, writing–review and editing. R.K. Amaravadi: Conceptualization, resources, supervision, funding acquisition, validation, investigation, visualization, project administration, writing–review and editing.

This work was supported by HHS/NIH/NCI grants to R.K. Amaravadi (P30 CA016520-45 and P01 CA114046) and to R.K. Amaravadi and D.W. Speicher (R01 CA266404 and P50 CA174523). We acknowledge the grants U5CA4224070, P01CA114046, P50CA261608, and RO1CA238237, and the Dr. Miriam and Sheldon Aldelson Foundation. H. Stark received financial support for the design and synthesis of ST1074 from the German Research Foundation (DFG) on Research Training Group No 2158 (GRK). The Wistar Proteomics and Metabolomics Core Facility was supported by Cancer Center Support Grant CA010815, and the Thermo Q-Exactive HF-X mass spectrometer was purchased with NIH grant S10 OD023586. The Wistar Molecular Screening Facility and Genomics Facility in part is supported by NIH grant S10 OD023586. The authors thank Joel Cassel and Isabela Batista Oliva at The Wistar Molecular Screening Facility and Genomics Facility for their assistance with synergy experiment testing. The authors thank the PennVet imaging core for high-resolution confocal imaging and Dr. Andrea Stout (UPenn CDB Microscopy Core) for training and assistance with STED and confocal imaging and image analysis.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.

Note: Supplementary data for this article are available at Cancer Discovery Online (http://cancerdiscovery.aacrjournals.org/).

1.
Amaravadi
RK
,
Kimmelman
AC
,
Debnath
J
.
Targeting autophagy in cancer: recent advances and future directions
.
Cancer Discov
2019
;
9
:
1167
81
.
2.
Towers
CG
,
Thorburn
A
.
Therapeutic targeting of autophagy
.
EBioMedicine
2016
;
14
:
15
23
.
3.
Zoncu
R
,
Perera
RM
.
Built to last: lysosome remodeling and repair in health and disease
.
Trends Cell Biol
2022
;
32
:
597
610
.
4.
Cao
M
,
Luo
X
,
Wu
K
,
He
X
.
Targeting lysosomes in human disease: from basic research to clinical applications
.
Signal Transduct Target Ther
2021
;
6
:
1
28
.
5.
Rebecca
VW
,
Nicastri
MC
,
Fennelly
C
,
Chude
CI
,
Barber-Rotenberg
JS
,
Ronghe
A
, et al
.
PPT1 promotes tumor growth and is the molecular target of chloroquine derivatives in cancer
.
Cancer Discov
2019
;
9
:
220
29
.
6.
Brun
S
,
Bestion
E
,
Raymond
E
,
Bassissi
F
,
Jilkova
ZM
,
Mezouar
S
, et al
.
GNS561, a clinical-stage PPT1 inhibitor, is efficient against hepatocellular carcinoma via modulation of lysosomal functions
.
Autophagy
2021
;
18
:
678
94
.
7.
Towers
CG
,
Fitzwalter
BE
,
Regan
D
,
Goodspeed
A
,
Morgan
MJ
,
Liu
CW
, et al
.
Cancer cells upregulate NRF2 signaling to adapt to autophagy inhibition
.
Dev Cell
2019
;
50
:
690
703
.
8.
Rangwala
R
,
Leone
R
,
Chang
YC
,
Fecher
LA
,
Schuchter
LM
,
Kramer
A
, et al
.
Phase I trial of hydroxychloroquine with dose-intense temozolomide in patients with advanced solid tumors and melanoma
.
Autophagy
2014
;
10
:
1369
79
.
9.
Vogl
DT
,
Stadtmauer
EA
,
Tan
KS
,
Heitjan
DF
,
Davis
LE
,
Pontiggia
L
, et al
.
Combined autophagy and proteasome inhibition: a phase 1 trial of hydroxychloroquine and bortezomib in patients with relapsed/refractory myeloma
.
Autophagy
2014
;
10
:
1380
90
.
10.
Kinsey
CG
,
Camolotto
SA
,
Boespflug
AM
,
Guillen
KP
,
Foth
M
,
Truong
A
, et al
.
Protective autophagy elicited by RAF→MEK→ERK inhibition suggests a treatment strategy for RAS-driven cancers
.
Nat Med
2019
;
25
:
620
7
.
11.
Vergoten
G
,
Bailly
C
.
Binding of hydroxychloroquine and chloroquine dimers to palmitoyl-protein thioesterase 1 (PPT1) and its glycosylated forms: a computational approach
.
J Biomol Struct Dyn
2021
;
40
:
8197
205
.
12.
Abu-Remaileh
M
,
Wyant
GA
,
Kim
C
,
Laqtom
NN
,
Abbasi
M
,
Chan
SH
, et al
.
Lysosomal metabolomics reveals V-ATPase- and mTOR-dependent regulation of amino acid efflux from lysosomes
.
Science
2017
;
358
:
807
13
.
13.
Mahammad
S
,
Parmryd
I
.
Cholesterol depletion using methyl-β-cyclodextrin
.
Methods Mol Biol
2015
;
1232
:
91
102
.
14.
Brown
MS
,
Goldstein
JL
.
A receptor-mediated pathway for cholesterol homeostasis
.
Science
1986
;
232
:
34
47
.
15.
Grin
PM
,
Dwivedi
DJ
,
Chathely
KM
,
Trigatti
BL
,
Prat
A
,
Seidah
NG
, et al
.
Low-density lipoprotein (LDL)-dependent uptake of Gram-positive lipoteichoic acid and Gram-negative lipopolysaccharide occurs through LDL receptor
.
Sci Rep
2018
;
8
:
1
1
.
16.
Shen
WJ
,
Azhar
S
,
Kraemer
FB
.
SR-B1: a unique multifunctional receptor for cholesterol influx and efflux
.
Annu Rev Physiol
2018
;
80
:
95
.
17.
Nieland
TJF
,
Shaw
JT
,
Jaipuri
FA
,
Duffner
JL
,
Koehler
AN
,
Banakos
S
, et al
.
Identification of the molecular target of small molecule inhibitors of HDL receptor SR-BI activity
.
Biochemistry
2008
;
47
:
460
72
.
18.
Quinville
BM
,
Deschenes
NM
,
Ryckman
AE
,
Walia
JS
.
A comprehensive review: sphingolipid metabolism and implications of disruption in sphingolipid homeostasis
.
Int J Mol Sci
2021
;
22
:
5793
.
19.
Levy
M
,
Futerman
AH
.
Mammalian ceramide synthases
.
IUBMB Life
2010
;
62
:
347
56
.
20.
Morris
AP
,
Le
TH
,
Wu
H
,
Akbarov
A
,
van der Most
PJ
,
Hemani
G
, et al
.
Trans-ethnic kidney function association study reveals putative causal genes and effects on kidney-specific disease aetiologies
.
Nat Commun
2019
;
10
:
1
4
.
21.
Schiffmann
S
,
Hartmann
D
,
Fuchs
S
,
Birod
K
,
Ferreirs
N
,
Schreiber
Y
, et al
.
Inhibitors of specific ceramide synthases
.
Biochimie
2012
;
94
:
558
65
.
22.
Kaushik
S
,
Massey
AC
,
Cuervo
AM
.
Lysosome membrane lipid microdomains: novel regulators of chaperone-mediated autophagy
.
EMBO J
2006
;
25
:
3921
33
.
23.
Bagshaw
RD
,
Mahuran
DJ
,
Callahan
JW
.
A proteomic analysis of lysosomal integral membrane proteins reveals the diverse composition of the organelle
.
Mol Cell Proteomics
2005
;
4
:
133
43
.
24.
Schmidt
F
,
Thywißen
A
,
Goldmann
M
,
Cunha
C
,
Cseresnyés
Z
,
Schmidt
H
, et al
.
Flotillin-dependent membrane microdomains are required for functional phagolysosomes against fungal infections
.
Cell Rep
2020
;
32
:
108017
.
25.
Hanafusa
K
,
Hotta
T
,
Iwabuchi
K
.
Glycolipids: linchpins in the organization and function of membrane microdomains
.
Front Cell Dev Biol
2020
;
8
:
589799
.
26.
de Chatterjee
A
,
Mendez
TL
,
Roychowdhury
S
,
Das
S
.
The assembly of GM1 glycolipid- and cholesterol-enriched raft-like membrane microdomains is important for giardial encystation
.
Infect Immun
83
:
2030
42
.
27.
Kenworthy
AK
,
Petranova
N
,
Edidin
M
.
High-resolution FRET microscopy of cholera toxin B-subunit and GPI-anchored proteins in cell plasma membranes
.
Mol Biol Cell
2000
;
11
:
1645
55
.
28.
Han
WQ
,
Chen
WD
,
Zhang
K
,
Liu
JJ
,
Wu
YJ
,
Gao
PJ
.
Ca2+-regulated lysosome fusion mediates angiotensin II-induced lipid raft clustering in mesenteric endothelial cells
.
Hypertens Res
2016
;
39
:
227
36
.
29.
Banning
A
,
Tomasovic
A
,
Tikkanen
R
.
Functional aspects of membrane association of reggie/flotillin proteins
.
Curr Protein Pept Sci
2011
;
12
:
725
35
.
30.
Garver
WS
,
Heidenreich
RA
,
Erickson
RP
,
Thomas
MA
,
Wilson
JM
.
Localization of the murine Niemann-Pick C1 protein to two distinct intracellular compartments
.
J Lipid Res
2000
;
41
:
673
87
.
31.
Dermine
JF
,
Duclos
S
,
Garin
J
,
St.-Louis
F
,
Rea
S
,
Parton
RG
, et al
.
Flotillin-1-enriched lipid raft domains accumulate on maturing phagosomes
.
J Biol Chem
2001
;
276
:
18507
12
.
32.
Aits
S
,
Kricker
J
,
Liu
B
,
Ellegaard
AM
,
Hämälistö
S
,
Tvingsholm
S
, et al
.
Sensitive detection of lysosomal membrane permeabilization by lysosomal galectin puncta assay
.
Autophagy
2015
;
11
:
1408
24
.
33.
Tsherniak
A
,
Vazquez
F
,
Montgomery
PG
,
Weir
BA
,
Kryukov
G
,
Cowley
GS
, et al
.
Defining a cancer dependency map
.
Cell
2017
;
170
:
564
76
.
34.
Cirenajwis
H
,
Ekedahl
H
,
Lauss
M
,
Harbst
K
,
Carneiro
A
,
Enoksson
J
, et al
.
Molecular stratification of metastatic melanoma using gene expression profiling: prediction of survival outcome and benefit from molecular targeted therapy
.
Oncotarget
2015
;
6
:
12297
.
35.
Joseph
RW
,
Elassaiss-Schaap
J
,
Kefford
R
,
Hwu
WJ
,
Wolchok
JD
,
Joshua
AM
, et al
.
Baseline tumor size is an independent prognostic factor for overall survival in patients with melanoma treated with pembrolizumab
.
Clin Cancer Res
2018
;
24
:
4960
7
.
36.
Puzanov
I
,
Amaravadi
RK
,
McArthur
GA
,
Flaherty
KT
,
Chapman
PB
,
Sosman
JA
, et al
.
Long-term outcome in BRAF(V600E) melanoma patients treated with vemurafenib: patterns of disease progression and clinical management of limited progression
.
Eur J Cancer
2015
;
51
:
1435
43
.
37.
Mehnert
JM
,
Mitchell
T
,
Huang
AC
,
Aleman
TS
,
Kim
BJ
,
Schuchter
LM
, et al
.
BAMM (BRAF Autophagy and MEK inhibition in Melanoma): a phase I/II trial of dabrafenib, trametinib and hydroxychloroquine in advanced BRAFV600-mutant melanoma
.
Clin Cancer Res
;
2022;
,
28
:
1098
106
.
38.
Ojha
R
,
Leli
NM
,
Onorati
A
,
Piao
S
,
Verginadis
II
,
Tameire
F
, et al
.
ER translocation of the MAPK pathway drives therapy resistance in BRAF-mutant melanoma
.
Cancer Discov
2019
;
9
:
396
415
.
39.
Wolpin
BM
,
Rubinson
DA
,
Wang
X
,
Chan
JA
,
Cleary
JM
,
Enzinger
PC
, et al
.
Phase II and pharmacodynamic study of autophagy inhibition using hydroxychloroquine in patients with metastatic pancreatic adenocarcinoma
.
Oncologist
2014
;
19
:
637
8
.
40.
Karasic
TB
,
O'Hara
MH
,
Loaiza-Bonilla
A
,
Reiss
KA
,
Teitelbaum
UR
,
Borazanci
E
, et al
.
Effect of gemcitabine and nab-paclitaxel with or without hydroxychloroquine on patients with advanced pancreatic cancer: a phase 2 randomized clinical trial
.
JAMA Oncol
2019
;
5
:
993
8
.
41.
Zeh
HJ
,
Bahary
N
,
Boone
BA
,
Singhi
AD
,
Miller-Ocuin
JL
,
Normolle
DP
, et al
.
A randomized phase II preoperative study of autophagy inhibition with high-dose hydroxychloroquine and gemcitabine/nab-paclitaxel in pancreatic cancer patients
.
Clin Cancer Res
2020
;
26
:
3126
34
.
42.
Harding
JJ
,
Awada
A
,
Roth
G
,
Decaens
T
,
Merle
P
,
Kotecki
N
, et al
.
First-in-human effects of PPT1 inhibition using the oral treatment with GNS561/ezurpimtrostat in patients with primary and secondary liver cancers
.
Liver Cancer
2022
;
11
:
268
77
.
43.
Towers
CG
,
Wodetzki
DK
,
Thorburn
J
,
Smith
KR
,
Caino
MC
,
Thorburn
A
.
Mitochondrial-derived vesicles compensate for loss of LC3-mediated mitophagy
.
Dev Cell
2021
;
56
:
2029
42
.
44.
Levental
I
,
Levental
KR
,
Heberle
FA
.
Lipid rafts: controversies resolved, mysteries remain
.
Trends Cell Biol
2020
;
30
:
341
53
.
45.
Bennett
LL
,
Turcotte
K
.
Eliglustat tartrate for the treatment of adults with type 1 gaucher disease
.
Drug Des Devel Ther
2015
;
9
:
4639
47
.
46.
Yew
NS
,
Zhao
H
,
Hong
EG
,
Wu
IH
,
Przybylska
M
,
Siegel
C
, et al
.
Increased hepatic insulin action in diet-induced obese mice following inhibition of glucosylceramide synthase
.
PLoS One
2010
;
5
:
e11239
.
47.
Dubot
P
,
Astudillo
L
,
Therville
N
,
Sabourdy
F
,
Stirnemann
J
,
Levade
T
, et al
.
Are glucosylceramide-related sphingolipids involved in the increased risk for cancer in gaucher disease patients? review and hypotheses
.
Cancers
;
2020;
,
12
:
475
.
48.
Riscal
R
,
Bull
CJ
,
Mesaros
C
,
Finan
JM
,
Carens
M
,
Ho
ES
, et al
.
Cholesterol auxotrophy as a targetable vulnerability in clear cell renal cell carcinoma
.
Cancer Discov
2021
;
11
:
3106
25
.
49.
Sassa
T
,
Suto
S
,
Okayasu
Y
,
Kihara
A
.
A shift in sphingolipid composition from C24 to C16 increases susceptibility to apoptosis in HeLa cells
.
Biochim Biophys Acta
2012
;
1821
:
1031
7
.
50.
Iqbal
J
,
Walsh
MT
,
Hammad
SM
,
Hussain
MM
.
Sphingolipids and lipoproteins in health and metabolic disorders
.
Trends Endocrinol Metab
2017
;
28
:
506
18
.
51.
Carlier
A
,
Phan
F
,
Szpigel
A
,
Hajduch
E
,
Salem
JE
,
Gautheron
J
, et al
.
Dihydroceramides in triglyceride-enriched VLDL are associated with nonalcoholic fatty liver disease severity in type 2 diabetes
.
Cell Rep Med
2020
;
1
:
100154
.
52.
Vykoukal
J
,
Fahrmann
JF
,
Gregg
JR
,
Tang
Z
,
Basourakos
S
,
Irajizad
E
, et al
.
Caveolin-1-mediated sphingolipid oncometabolism underlies a metabolic vulnerability of prostate cancer
.
Nat Commun
2020
;
11
:
1
16
.
53.
Beer
LA
,
Tang
HY
,
Sriswasdi
S
,
Barnhart
KT
,
Speicher
DW
.
Systematic discovery of ectopic pregnancy serum biomarkers using 3-D protein profiling coupled with label-free quantitation
.
J Proteome Res
2011
;
10
:
1126
38
.
54.
Goldman
AR
,
Bitler
BG
,
Schug
Z
,
Conejo-Garcia
JR
,
Zhang
R
,
Speicher
DW
.
The primary effect on the proteome of ARID1A-mutated ovarian clear cell carcinoma is downregulation of the mevalonate pathway at the post-transcriptional level
.
Mol Cell Proteomics
2016
;
15
:
3348
60
.
55.
Cox
J
,
Mann
M
.
MaxQuant enables high peptide identification rates, individualized p.p.b.-range mass accuracies and proteome-wide protein quantification
.
Nat Biotechnol
2008
;
26
:
1367
72
.
56.
Cox
J
,
Neuhauser
N
,
Michalski
A
,
Scheltema
RA
,
Olsen
JV
,
Mann
M
.
Andromeda: a peptide search engine integrated into the MaxQuant environment
.
J Proteome Res
2011
;
10
:
1794
805
.
57.
Cox
J
,
Hein
MY
,
Luber
CA
,
Paron
I
,
Nagaraj
N
,
Mann
M
.
Accurate proteome-wide label-free quantification by delayed normalization and maximal peptide ratio extraction, termed MaxLFQ
.
Mol Cell Proteomics
2014
;
13
:
2513
26
.
58.
Tyanova
S
,
Temu
T
,
Sinitcyn
P
,
Carlson
A
,
Hein
MY
,
Geiger
T
, et al
.
The Perseus computational platform for comprehensive analysis of (prote)omics data
.
Nat Methods
2016
;
13
:
731
40
.
59.
Tyanova
S
,
Cox
J
.
Perseus: a bioinformatics platform for integrative analysis of proteomics data in cancer research
.
Methods Mol Biol
2018
;
1711
:
133
48
.
60.
Alicea
GM
,
Rebecca
VW
,
Goldman
AR
,
Fane
ME
,
Douglass
SM
,
Behera
R
, et al
.
Changes in aged fibroblast lipid metabolism induce age-dependent melanoma cell resistance to targeted therapy via the fatty acid transporter FATP2
.
Cancer Discov
2020
;
10
:
1282
5
.
61.
Schindelin
J
,
Arganda-Carreras
I
,
Frise
E
,
Kaynig
V
,
Longair
M
,
Pietzsch
T
, et al
.
Fiji: an open-source platform for biological-image analysis
.
Nat Methods
2012
;
9
:
676
82
.
62.
Bolte
S
,
Cordelières
FP
.
A guided tour into subcellular colocalization analysis in light microscopy
.
J Microsc
2006
;
224
:
213
32
.
63.
Maejima
I
,
Takahashi
A
,
Omori
H
,
Kimura
T
,
Takabatake
Y
,
Saitoh
T
, et al
.
Autophagy sequesters damaged lysosomes to control lysosomal biogenesis and kidney injury
.
EMBO J
2013
;
32
:
2336
47
.
64.
Agrotis
A
,
Pengo
N
,
Burden
JJ
,
Ketteler
R
.
Redundancy of human ATG4 protease isoforms in autophagy and LC3/GABARAP processing revealed in cells
.
Autophagy
2019
;
15
:
976
97
.
65.
di Veroli
GY
,
Fornari
C
,
Wang
D
,
Mollard
S
,
Bramhall
JL
,
Richards
FM
, et al
.
Combenefit: an interactive platform for the analysis and visualization of drug combinations
.
Bioinformatics
2016
;
32
:
2866
8
.
66.
Contal
C
,
O'Quigley
J
.
An application of changepoint methods in studying the effect of age on survival in breast cancer
.
Comput Stat Data Anal
1999
;
30
:
253
70
.
67.
Meyers
JP
,
Mandrekar
JN
.
Cutpoint determination methods in survival analysis using SAS®: updated %FINDCUT macro
.
Proc SAS Glob Forum
2015;3249
.
Available from
: https://support.sas.com/resources/papers/proceedings15/3249-2015.pdf.