Therapies for head and neck squamous cell carcinoma (HNSCC) are, at best, moderately effective, underscoring the need for new therapeutic strategies. Ceramide treatment leads to cell death as a consequence of mitochondrial damage by generating oxidative stress and causing mitochondrial permeability. However, HNSCC cells are able to resist cell death through mitochondria repair via mitophagy. Through the use of the C6-ceramide nanoliposome (CNL) to deliver therapeutic levels of bioactive ceramide, we demonstrate that the effects of CNL are mitigated in drug-resistant HNSCC via an autophagic/mitophagic response. We also demonstrate that inhibitors of lysosomal function, including chloroquine (CQ), significantly augment CNL-induced death in HNSCC cell lines. Mechanistically, the combination of CQ and CNL results in dysfunctional lysosomal processing of damaged mitochondria. We further demonstrate that exogenous addition of methyl pyruvate rescues cells from CNL + CQ–dependent cell death by restoring mitochondrial functionality via the reduction of CNL- and CQ-induced generation of reactive oxygen species and mitochondria permeability. Taken together, inhibition of late-stage protective autophagy/mitophagy augments the efficacy of CNL through preventing mitochondrial repair. Moreover, the combination of inhibitors of lysosomal function with CNL may provide an efficacious treatment modality for HNSCC.
Head and neck squamous cell carcinoma (HNSCC) is a cancer originating from squamous cells, mainly of the larynx, pharynx, and oral cavity. The worldwide incidence and mortality is over 830,000 and 430,000 per year, respectively (1). Although traditional surgical and chemoradiotherapy offer patients some benefit, these modalities cause serious adverse events and relapse still occurs (2). Until the very recent approval of immune-checkpoint inhibitors (3), cetuximab, an epidermal growth factor receptor (EGFR) inhibitor, was the only targeted therapy approved in combination with chemoradiotherapy for first-line treatment in HNSCC. Unfortunately, this regimen causes significant adverse events in over 80% of patients and provides minimal therapeutic benefit (2). With cancer recurrence, second-line therapies offer a dismal 6% response (4). Despite attempts to identify subtypes of HNSCC that would be amenable to targeted therapies, these endeavors have proven difficult due to the molecular heterogeneity, late-stage of detection, and lack of susceptibility to multiple specific EGFR tyrosine kinase inhibitors (2, 5). Thus, there is an urgent need for new therapeutic approaches.
Sphingolipid metabolites, including ceramide, and the enzymes that regulate their levels are dysregulated in a myriad of pathologies, including cancers (6–8). Ceramide is well established as a proapoptotic lipid. However, ceramide metabolites, including sphingosine-1-phosphate and glucosylceramide, have been shown to be antiapoptotic and promote drug resistance (7, 8). Previously, C6-ceramide, a synthetic short-chain form of ceramide, has been shown to induce cell-cycle arrest and apoptosis in HNSCC in vitro (9). In fact, many chemotherapeutic approaches increase levels of ceramide within tumors leading to cell death (7, 8, 10–12). Our laboratory has engineered a water-soluble, nontoxic delivery platform for C6-ceramide, the ceramide nanoliposome (CNL; ref. 13), that is currently being evaluated in a phase I human trial for solid tumors (14). Like many chemotherapeutics, the efficacy of ceramide (or CNL) may be limited by drug-resistance mechanisms, including autophagy (15).
Autophagy is a cellular recycling process that sequesters damaged proteins or organelles within an isolation membrane which matures to form an autophagosome. The autophagosome then fuses with an acidified lysosome, a necessary step for degradation and recycling of its cargo (16). Autophagy is frequently dysregulated in cancer and has previously been identified as a mechanism of resistance to cell death (15). However, in the context of ceramide signaling, it is still unclear if ceramide induces protective autophagy-mediated survival or lethal autophagy-mediated cell death (17). Furthermore, the mechanism and stage at which ceramide induces autophagy appears to be multifaceted and cell-type dependent (6).
Our laboratory has previously published that inhibitors of microtubules (and possibly autophagosomes), such as vinblastine, augment the efficacy of CNL in solid (18) and nonsolid cancers (19, 20). Both ceramide signaling and autophagic inhibition have been independently shown to alter mitochondrial function. Ceramides have previously been shown to induce oxidative stress which can alter mitochondrial function, directly induce pore formation in mitochondria, and drive mitophagy (20–22). Mitophagy, a subclass of autophagy, is responsible for recycling damaged parts of the mitochondria to promote mitochondrial health and functionality (23). However, how inhibition of autophagic/lysosomal processing affects ceramide-dependent mitochondrial damage and/or mitophagy is undefined.
In the present work, we examine if direct inhibitors of lysosomal function enhance the therapeutic effect of CNL in drug-resistant HNSCC by circumventing autophagic resistance. We further evaluated the mechanisms by which inhibitors of late-stage autophagy/mitophagy can augment mitochondrial-dependent CNL-induced cell death. Taken together, this work defines a novel combinatorial therapeutic strategy to enhance the efficacy of CNL in drug-resistant HNSCC.
Materials and Methods
The HNSCC cell lines Cal27, FaDu, UNC-7, UNC-10, SCC-25, SCC-61, and OSC-19 were obtained from Mark Jameson's Lab (University of Virginia) and grown in DMEM/F-12 media (Thermo Fisher Scientific) supplemented with 10% FBS (Gemini Bio Products) and 1% antibiotic–antimycotic (Thermo Fisher Scientific). SCC-61 media were additionally supplemented with 0.5 mg/mL hydrocortisone (Millipore Sigma). The Primary Gingival Fibroblasts (PCS-201-018 ATCC) were grown in fibroblast basal media supplemented with Fibroblast Growth Kit-Low serum (ATCC PCS-201-041) and 1% antibiotic–antimycotic. TrypLE (Thermo Fisher Scientific) was used to passage cells. All cell lines were authenticated via DNA fingerprinting (University of Arizona) of early passage, confirmed Mycoplasma via MycoAlert System (University of Virginia) after thawing, and did not exceed 25 passages.
Chloroquine diphosphate salt (CQ; Millipore Sigma), methyl pyruvate (MP; Alfa Aesar), and 3-methyladenine (Selleck Chemicals) were dissolved in water. Working stocks of staurosporine (ApexBio), apilimod mesylate (Millipore Sigma), bafilomycin A1, rapamycin (LC Laboratories) Torin-1 (Cayman Chemical Company), and wortmannin (Selleck Chemicals) were prepared in DMSO. BAM15 was a generous gift from Dave Kashatus (University of Virginia).
Ceramide nanoliposome (CNL) formulation
Ceramide nanoliposomes were a generous gift from KeystoneNano and manufactured according to published methods (14). Control (ghost) formulations included all liposomal ingredients except C6-ceramide.
HNSCC cells were seeded on 96-well plates to achieve a similar confluency. After 24 hours, cells were pretreated for 4 hours (rapamycin and Torin 1), 2 hours (MP), or 1 hour [CQ, Bafilomycin (Baf), apilimod mesylate (AMS)] and subsequently treated with CNLs or ghost liposomes at concentrations indicated in the text. MTS assays were performed according to the manufacturer's instructions (Promega). Absorbance at 490 nm was determined with a Cytation 3 plate reader (Bio Tek). After subtracting the background absorbance (no cells), all values were normalized to their intraplate controls.
Cells were treated as indicated in the text. RIPA buffer (Alfa Aesar) with protease inhibitor (Thermo Fisher Scientific) and phosphatase inhibitor (Roche) was used to lyse the cells. Protein concentrations were determined with a BCA protein assay (Pierce). Protein (20–30 μg) was added to a NuPAGE 4–12% Bis-Tris gel (Thermo Fisher Scientific) and ran at 120 V for 2 hours and 20 minutes. Transfer to a nitrocellulose membrane (Bio-Rad Laboratories) was performed using the Bio-Rad Turbo-Transfer apparatus. Blocking with 5% BSA (Thermo Fisher Scientific) in TBST was done for 1 hour at room temperature and blots were cut before an overnight primary antibody incubation at 4°C. Primary antibodies were LAMP1 [sc-20011], LAMP2 [sc-18822], SQSTM1 (p62) [sc-28359], GSK-3α/β [sc-7291] (Santa Cruz Biotechnologies), Beta Actin [A5441] (Millipore Sigma), LC3B , Beclin 1 , Caspase-3  (Cell Signaling Technologies) and BNIP3 [ab10433], Rab7 [ab137029], and PINK1 [ab23707] (Abcam). Blots were washed 3 × 5 minutes with TBST. Horseradish peroxidase–conjugated secondary antibody against rabbit or mouse (Thermo Fisher Scientific) was then added for 1 hour at room temperature. Three additional 5-minute washes were performed prior to detection with enhanced chemiluminescence (Prometheus) and imaged with a G:Box Chemi XX6 (Syngene). If blots were reprobed, a stripping step was performed using Restore Western Blot Stripping Buffer (Thermo Fisher Scientific) according to the manufacturer's instructions. Densitometry of each target was performed using ImageJ software and was normalized first to β-actin, then to the vehicle treatment at the respective timepoint from the same blot.
Transmission electron microscopy (TEM)
Cells were treated as indicated in the text. After treatment, samples were fixed in 4% paraformaldehyde + 2% glutaraldehyde, post-fixed in 4% osmium tetroxide, dehydrated through a gradient of ethanol into propylene oxide, then embedded in the epoxy resin Embed 812. Samples were sectioned at 70 nm with a UC7 Ultramicrotome (Leica), placed on 200 mesh copper grids, and contrast-stained with uranyl acetate and lead citrate. Microscopic analysis was performed using a JEM-1230 TEM microscope (JEOL) and images captured with a 4K × 4K CCD camera.
Quantitative reverse transcription polymerase chain reaction
Treated cells were washed and RNA extracted using the RNeasy Mini Kit (Qiagen). RNA content was quantified using a Cytation 3 (BioTek). RNA (800–1,000 ng) was used in the iScript cDNA Synthesis Kit (Bio-Rad Laboratories) to synthesize cDNA. FAM probes (BNIP3-[qHsaCIP0040441], MAP1LC3B-[qHsaCEP0041298], LAMP1-[qHsaCEP0055037], BECN1-[qHsaCIP0030326], PSMB6-[qHsaCEP0052321], B2M-[qHsaCIP0029872]), and iTaq Universal Probes Supermix (Bio-Rad Laboratories) were used and CT values were measured using CFX Connect Real-Time PCR Connection System (Bio-Rad Laboratories). CT values were normalized first to the housekeeping gene control (PSMB6 and/or B2M), and then to their respective timepoint/vehicle controls.
Flow cytometry assays
Cells were treated as indicated in the text. The media containing floating cells and the adherent cells, after trypsinization, were combined. Cells were centrifuged and the supernatant discarded. To assess cell viability, autophagic vacuole presence, and mitochondrial permeability on the resultant cell pellet, Fixable Viability Dye 780 (Thermo Fisher Scientific), CYTO-ID Autophagy detection kit 2.0 (Enzo Biochem), and BD MitoScreen (JC-1) (BD Biosciences) dyes were used, respectively, according to the manufacturers' instructions. Samples were measured by the Attune Nxt Flow Cytometer (Thermo Fisher Scientific). Forward and side-scatter measurements were used to gate for singlets and exclude debris. Single-stain compensation controls were collected and gates were drawn accordingly.
ROS/superoxide assay and ATP measurement assay
Cells were treated as described in the text, and the ROS/Superoxide Detection Assay Kit (Cell-based) (ab139476; Abcam) and the Luminescent ATP Detection Assay Kit (ab113849; Abcam) were utilized according to the manufacturer's instructions. A Cytation 3 plate reader was used to analyze general ROS (Ex/Em: 490/525 nm), Superoxides (Ex/Em: 550–620 nm), and ATP (luminescence). Background was subtracted, and all results were normalized to the respective vehicle controls.
All statistics were performed on GraphPad Prism 8 (GraphPad Software). All experiments were repeated at least three times unless otherwise stated. A one-way ANOVA was performed with Tukey multiple comparison post hoc test to determine significance between all groups. Asterisks indicate significance: *, P < 0.05; **, 0.01; ***, 0.001. All error bars are standard deviation of the mean. Bliss scores were calculated using SynergyFinder (24).
CNL induces a time- and concentration-dependent decrease in cell viability by MTS and flow cytometry in HNSCC
To determine the response of HNSCC cell lines to CNL treatment, seven HNSCC cell lines and nontransformed primary gingival fibroblasts (PGF) were treated with varying concentrations of CNL for 24 hours (Fig. 1A). CNL induced a concentration-dependent decrease in cell viability in all cell lines with varying sensitivities (IC50 calculations are shown in Supplementary Fig. S1). No significant cell death occurred in all cell lines treated with ghost liposomes (Fig. 1A). The most resistant cancer cell line, FaDu (81% viable with 25 μmol/L CNL), was similar to the nontransformed PGF. In contrast, after 48 hours, FaDu cells exhibited a larger reduction in viability compared with PGF (Fig. 1B). PGF and FaDu cells were further analyzed by flow cytometry for cell death (Fig. 1C). Confirming the MTS results, PGF cells were more resistant to CNL than FaDu, and only showed significant cell death with 10 and 25 μmol/L CNL after 48 hours. These data support that, though variable, HNSCC cell lines are more sensitive than nontransformed cells to CNL.
CNL induces autophagy but not apoptosis
Because FaDu cells require a longer duration of CNL treatment to observe a robust response compared with other HNSCC cell lines, experiments to investigate mechanisms of ceramide resistance are focused on FaDu cells. Caspase-3 and PARP cleavage are indicative of increased caspase-dependent apoptotic cell death and a decrease in commitment to DNA repair (25). Supporting results from Fig. 1A, western blotting confirmed neither caspase-3 nor PARP protein cleavage in response to CNL within 24 hours in FaDu cells (Fig. 2A). As a positive control, staurosporine treatment resulted in cleavage as early as 6 hours. To further examine the lack of apoptotic markers, TEM imaging was used to identify morphologic differences in cells treated with CNL or ghost liposomes after 12 hours (Fig. 2B). Further ruling out canonical apoptosis, increased membrane blebbing and apoptotic body formation were not observed. Interestingly, FaDu cells showed a marked increase in the number of autophagosomes after CNL, but not ghost liposome treatment. This suggests that CNL may be triggering an autophagic response.
To evaluate this autophagic response, markers of autophagy were assessed by western blotting (Fig. 2C/2D). The two forms of the protein encoded by microtubule-associated proteins 1A/1B light chain 3B (MAP1LC3B or LC3B) were measured. The cytosolic form of LC3B, LC3B-I (top band), can be conjugated with phosphatidylethanolamine generating LC3B-II (lower band), a marker of increased autophagosomes in cells (16). In cells treated with CNL, a 4.6-fold increase of LC3B-II was observed at 24 hours (Fig. 2C and D). Moreover, LC3B transcripts also increased (Fig. 2E). The adenovirus E1B 19-kDa-interacting protein 3 (BNIP3), which is associated primarily with the mitophagy subclass of autophagy (26), was next assessed. Similar to LC3B-II, a 3.1-fold increase in BNIP3 protein expression was observed at 24 hours (Fig. 2C and D), which was consistent with BNIP3 RNA expression (Fig. 2E). Although LC3B-II is a marker of autophagosomes, it is not a direct measurement of late-stage autophagy or lysosomal function. Lysosome-associated membrane glycoprotein 1 (LAMP1) is a multifunctional protein located within the lysosomal membrane (16, 27). In contrast to LC3B-II, a time-dependent 0.56-fold decrease in LAMP1 was seen in cells treated with CNL (Fig. 2C and D). LAMP-1 transcript levels were not changed with CNL treatment (Fig. 2E), demonstrating the effect of ceramide on LAMP-1 is posttranscriptional. These data suggest that CNL alters multiple autophagic/mitophagic targets (LC3B, BNIP3, and LAMP1) through transcriptional, translational, and posttranslational mechanisms.
Inhibition of late-stage autophagy synergizes with CNL to induce cell death in resistant HNSCC
To determine the functional relevance of the CNL-driven autophagic response, early-stage (autophagosome formation and cargo sequestration) and late-stage (autophagosomal maturation and lysosomal fusion) autophagy was investigated utilizing inhibitors described in Supplementary Table S1; ref. 28). Two inducers of early-stage autophagy, Torin1 and Rapamycin, did not alter CNL-mediated cell death at 24 or 48 hours (Supplementary Fig. S2A and S2B). Consistently, inhibitors of early-stage autophagy, the PI3K inhibitors 3-methyladenine or wortmannin, did not alter viability in the presence or absence of CNL (Supplementary Fig. S2C and S2D).
In contrast to modulating early-stage autophagy, inhibition of late-stage autophagy using CQ significantly augmented CNL-induced cell death in CNL-resistant HNSCC cells. Specifically, nontransformed (PGF), ceramide-resistant (FaDu), moderately ceramide-sensitive (UNC-10), and ceramide-sensitive (SCC-61) cells were pretreated with CQ before CNL treatment and the viability was assessed 24 hours later (Fig. 3A–D). In PGF and FaDu cells, CQ had minimal impact on cell viability as a single agent, only decreasing viability at 25 μmol/L concentrations (Fig. 3A and B). However, CQ dramatically augmented CNL-driven reduction in cell viability at both 24 and 48 hours in FaDu cells, with negligible effects seen in PGF cells (Fig. 3A and B). Specifically, at 48 hours, CNL treatment reduced viability of FaDu cells by 22%, whereas CNL with 5 μmol/L CQ pretreatment resulted in a 68% decrease in viability (Fig. 3A and B). In FaDu cells, the combinatorial effect of CQ and CNL was highly synergistic with a Bliss score of 26.9 at 48 hours (Fig. 3B). Synergism was also observed with UNC-10 cells, where CNL + CQ had a Bliss score of 16.4 (Fig. 3C). In contrast, Bliss score analysis revealed minimal synergy between CQ and CNL in PGF (Fig. 3A) and CNL-sensitive SCC-61 cells (Fig. 3D). The combinatorial decrease in viability by CNL and CQ treatment in FaDu cells was confirmed via flow cytometry (Fig. 3E). Moreover, TEM analysis of CQ and CNL dual-treated FaDu cells revealed a marked increase in the number of autophagic vacuoles compared with either treatment alone (Fig. 3F).
To confirm the specificity of synergy between CNL and CQ, FaDu cells were next treated with other inhibitors of late-stage autophagy/lysosomal function, Baf and AMS. Baf, which inhibits V-ATPase-dependent acidification and autophagosome-lysosome fusion (29), dramatically sensitized cells to CNL-induced cell death at 24 hours (Fig. 3G). Similarly, AMS, an inhibitor of the enzyme PIKFyve shown to prevent fusion of autophagosomes with lysosomes (30), also induced synergistic cell death (Fig. 3H). Taken together, these data suggest that late-stage, but not early-stage autophagy, is a resistance mechanism to CNL treatment.
Inhibiting late-stage autophagy alters lysosomal-mediated CNL signaling
Although late-stage autophagic inhibition sensitized FaDu cells to CNL-induced cell death, the underlying mechanism remained unknown. Thus, the effects of CNL and CQ on autophagy-related targets (LC3B-II, BNIP3, LAMP1, and Rab7) were analyzed to ascertain if these drugs elicit a response that corresponds with the combinatorial treatment (Fig. 4A and B). The combination of CQ and CNL significantly increased LC3B-II levels 5.1-fold, compared with 3.0-fold by CQ and 1.8-fold by CNL alone. Furthermore, BNIP3 levels increased 9.0-fold after CQ + CNL compared with smaller changes from either drug alone. The late-endosome marker Rab7 increased 1.4-fold after CQ + CNL treatment with no changes observed with either compound alone. The protein expression increases of BNIP3 and LC3B were consistent with mRNA measurements (Fig. 4C). In contrast, the CNL-induced reduction of LAMP1 and LAMP2 was unaffected by CQ.
To further confirm the combinatorial effect of CQ and CNL on autophagy, autophagic vacuole formation was measured using Cyto-ID dye incorporation. Autophagic vacuoles increased 2.3-fold by CQ, 1.7-fold by CNL, and 4.5-fold by the combination (Fig. 4D). The combination of rapamycin and CQ was used as a positive control for vacuole presence (Fig. 4D). These data are consistent with TEM images (Fig. 4F) demonstrating elevation of autophagic vacuoles. Taken together, these data support a buildup of CNL-driven autophagic vacuoles and autophagic/mitophagic proteins after inhibition of lysosomal function by CQ.
MP partially rescues from CNL + CQ–driven autophagic/mitophagic response
Because ceramide has been shown to stress cells (such as via ROS generation; ref. 31) and autophagic and mitophagic response can generate energy as well as repair mitochondria under stressed conditions (32), the synergism of CNL and CQ may be attributed to a compromised autophagic/mitophagic stress response. To assess this and circumvent the necessity of a cellular energy-generating response, MP at a concentration consistent with previous studies (31, 33) was used as an exogenous energy source. FaDu cells were pretreated with MP prior to vehicle, CQ, CNL, or CQ and CNL together and autophagic protein targets were measured by western blotting (Fig. 5A and B). MP significantly decreased CNL + CQ–induced LC3B-II protein expression by 47%. Similarly, MP decreased BNIP3 protein levels by 85% after CNL or CQ + CNL treatments. These protein changes were also observed at the RNA level (Fig. 5C). Other mitophagic proteins, PINK1 and p62, were also significantly increased after CQ + CNL, an effect that was decreased by MP (Fig. 5A and B). This effect was also seen with Rab7, a late-endosomal marker, but not the early endosomal marker Rab5. Beclin 1, a driver of autophagy previously shown to be increased by ceramide (34), remained unchanged (Fig. 5A and B). Validating the MP-driven decrease in LC3B-II and BNIP3 proteins, TEM reveals that MP significantly reduced autophagic vacuoles (Fig. 5D). In contrast to LC3B-II, BNIP3, p62, PINK1, and Rab7, CNL-induced reductions in LAMP1 and LAMP2 expression was not rescued by MP (Fig. 5A and B). A similar MP-mediated rescue from CQ + CNL protein changes was observed for LC3B, BNIP3, p62, Rab7, LAMP1, and LAMP2 in the moderately resistant UNC-10 cell line (Supplementary Fig. S3). Taken together, these data show that MP limits CQ + CNL–driven dysregulation of the autophagic and/or mitophagic pathways, but not lysosomal degradation.
MP rescues from CNL + CQ–driven oxidative stress and mitochondrial dysfunction
As MP reduced the CQ + CNL–driven elevations in mitophagic targets (BNIP3, p62, and PINK1), the effects of MP in the presence or absence of CNL and/or CQ on mitochondrial activity were next determined. Oxidative stress levels were determined through measuring total ROS species (hydrogen peroxide, peroxynitrite, hydroxyl radicals, nitric oxide, and peroxy radical) and superoxide levels. Although CQ alone did not alter ROS, CNL and the combination of CQ + CNL increased ROS generation 78% and 190%, respectively (Fig. 6A). Additionally, MP did not alter basal levels of ROS, but significantly reduced the CNL and CQ + CNL–induced increase in ROS to the same level as the vehicle group (Fig. 6A). Although neither CQ nor CNL changed superoxide levels as single agents, MP treatment decreased superoxide levels by 51% compared with the vehicle control (Fig. 6B). Strikingly, MP reduced the CQ + CNL superoxide levels from a 71% increase to a 33% decrease compared with basal levels (Fig. 6B). This MP-mediated protection from CQ + CNL–induced oxidative stress and superoxide levels was repeated in UNC-10 cells (Supplementary Fig. S4A and S4B). Mitochondrial permeability was next measured using a cationic dye that only aggregates in healthy mitochondria with an intact mitochondrial potential (Fig. 6C). MP significantly protected FaDu cells from mitochondrial permeability by reducing CNL-decreased mitochondrial potential from 70% to 31% and CQ + CNL from 92% to 31% over basal levels. Supporting the rescue of mitochondrial function by MP, MP elevated the CQ + CNL–induced decrease in ATP levels from an 87% decrease to 46% from basal (Fig. 6D). Finally, MP pretreatment reduced CQ + CNL–induced cell death from 83% to 13% as assessed by flow cytometry in FaDu cells (Fig. 6E). A similar rescue was seen in UNC-10 cells via flow cytometry (Supplementary Fig. S4C) and morphologically (Supplementary Fig. S4D). In summary, MP rescues CNL + CQ–dependent mitochondrial dysfunction, consistent with augmented cellular survival.
Targeted EGFR inhibition (cetuximab) and immune-checkpoint inhibitors (nivolumab and pembrolizumab) are now routinely utilized with conventional therapy for patients with HNSCC. Highlighting the urgent need for new therapies, these inhibitors only increase survival by months and response rates remain very poor (2–5). C6-ceramide, which has previously been shown to have a large therapeutic window for cancer and augments the efficacy of chemotherapeutic regimens (10, 14, 18, 35–37), is an attractive therapeutic option for HNSCC. C6-ceramide decreased cell viability in HNSCC cell lines in a concentration and time-dependent manner. Mechanistically, CNL induced mitochondrial damage through the production of oxidative stress and increased mitochondrial permeability. Inhibition of mitophagic repair with late-stage autophagic or lysosomal inhibitors (CQ, BAF, and AMS) synergized with CNL in ceramide-resistant HNSCC cancer cells. To circumvent mitochondria dysfunction, an exogenous energy source, MP, rescued CQ + CNL–induced mitochondrial damage and cell death. Taken together, these studies document that targeting the lysosomal portion of mitophagy-mediated repair mechanisms augments CNL-induced cell death.
Although the role of ceramide to regulate autophagy is well documented, the function of autophagy, whether lethal or protective, remains a point of contention (17). Mechanistically, ceramide can increase the expression of autophagic/mitophagic proteins LC3B-II (38), Beclin 1 (34), and BNIP3 (38), and dissociate autophagy-inhibiting Beclin 1:BCL-2 complexes (39), downregulate nutrient transporters (31, 40), and directly bind LC3B-II to traffic autophagosomes to the mitochondria (22). Consistent with these studies, increased LC3B-II and BNIP3 expression in response to CNL was observed. However, Beclin 1 was not altered, an observation reported by others (20). In a novel observation, ceramide decreased LAMP1 and LAMP2, two proteins which comprise 50% of the lysosomal membrane and promote lysosomal stability, fusion, autophagy, and metastatic potential (16, 27, 41, 42). Further work is needed to determine if ceramide-induced elevation of LC3B-II is a marker of reduced LAMP-dependent lysosomal breakdown of LC3B-containing autophagosomes.
We, and others, have previously demonstrated that inhibitors of microtubule assembly, autophagy, and/or lysosomal maturation, induce ceramide-dependent “autophagic cell death” in both solid and nonsolid tumor models (18–20, 43–45). However, we now identify that the synergy between CNL and multiple “autophagy inhibitors” (CQ, Baf, and AMS) occurs at the level of inhibiting lysosome function, explaining why inhibitors of early autophagy fail to synergize in our model and others (18). Of note, some inhibitors of acid ceramidase, a ceramide-metabolizing enzyme, enhance ceramide-induced cell death (46–50). Interestingly, some of these inhibitors also induce rapid destabilization of the lysosome (51) and, thus, may indirectly induce or augment ceramide-dependent cell death by blocking lysosome-mediated protective autophagy/mitophagy. Taken together, there is a great body of direct and indirect evidence that it is specifically lysosomal inhibition, not general autophagic inhibition, which synergizes with ceramide, likely by reducing protective autophagy or mitophagy.
Although ceramide promotion of mitochondrial dysfunction, autophagy, and synergism with lysosomal inhibitors have been described previously, the mechanistic links between these findings are still undefined (52, 53). This study may be the first to document that disruption of mitophagy with lysosomal inhibitors exacerbates ceramide-induced mitochondrial damage as evidenced by enhanced ROS production, diminished ATP levels, and increased membrane permeability. These findings are further supported by CNL + CQ–dependent increases of proteins linked to mitophagy (BNIP3, Rab7, PINK1, and p62). These data support previous findings that ceramide increased BNIP3 expression (38), with this effect exacerbated by lysosomal inhibitors. The role of BNIP3 is controversial in cancer; BNIP3 may be a prosurvival protein, possibly through reducing ROS and inducing autophagy/mitophagy (54, 55) or, in contrast, directly induce nonapoptotic cell death (38, 56). Regardless, BNIP3 remains a suitable marker for mitophagy and/or damaged mitochondria (26, 55). Like BNIP3, the canonical late-endosomal protein Rab7 is increased by ceramide and also regulates mitochondrial function, promoting mitophagy and mitochondria-lysosomal contact sites (57). Ceramide-induced accumulation of PINK1 also suggests mitochondrial damage, as PINK1 drives mitophagy through recruitment of Parkin and subsequent lysosome-driven mitochondrial degradation (58). PINK1 accumulation is further exacerbated by CQ + CNL, suggesting additional mitochondria dysfunction due to limited lysosomal functionality. In contrast, ceramide reduced p62 expression, a finding recently demonstrated by the Cabot lab (20). Although the necessity of p62 in autophagy, mitophagy, and NrF2-driven ROS clearance are debated, decreases are consistent with mitochondrial breakdown via autophagy/mitophagy (59). Of interest, CQ cotreatment with ceramide reversed and, in fact, increased protein levels of p62. Although these observed changes in p62 certainly warrant further causative investigation, we hypothesize that CQ inhibits lysosomal breakdown of damaged mitochondria, preventing ceramide-driven decreases of p62. As we demonstrate altered hallmarks of multiple markers of mitophagy (BNIP3, p62, PINK1, LC3B) after CQ + CNL, further studies are required to elucidate the full signaling cascade and which of these targets are effectors of this cell death phenotype. Together, these data indicate that lysosomal inhibitors augment mitochondrial-dependent, ceramide-induced, cell death, likely through inhibition of protective mitophagy.
MP has previously been used to decrease the cytotoxic effects of ceramide. This protection has been attributed to providing an energy source either after ceramide-driven nutrient transporter downregulation (4F2hc, mCAT-1, GLUT-1; ref. 31) and/or reduced glycolysis attributed to a ceramide-induced decrease in glyceraldehyde 3-phosphate dehydrogenase (33). Additionally, MP can rescue from mitochondrial damage induced by SIGMAR1 mutations (60). Consistent with these data, MP limited CQ + CNL–mediated increases in mitophagy-dependent LC3B, BNIP3, Rab7, PINK1, and p62 expression. Taken together, these data suggest that MP may prevent the need for an autophagic or mitophagic response after ceramide administration by providing an emergency alternative energy source to the ceramide-damaged mitochondria.
Although ceramides induce mitochondrial pore formation, membrane depolarization (21, 52), and mitophagy (22, 43, 53), the role of lysosomal inhibitors to augment or prevent mitochondrial-dependent, ceramide-induced cell death is still somewhat controversial. Although prior studies with the Cabot lab support CNL-driven mitophagy via colocalization of LC3B-II with LAMP1, the prosurvival/prodeath role of this mitophagy was not explored. This current body of work expands those findings to support a “protective” mitophagic phenotype. In contrast to our studies, the Ogretmen group demonstrated that blocking lysosomal acidification with Baf prevented FMS-like receptor tyrosine kinase-3 inhibition–induced cell death, believed to be through C18-ceramide generation in a model of acute myeloid leukemia (43). They demonstrated that a mitochondria-targeted ceramide, C18-Pyr-Cer, induced colocalization of lysosomes and mitochondria and interacted with LC3B-II to drive lethal autophagy/mitophagy. Moreover, they identified a protein, PERMIT, responsible for transporting C18-ceramide to the mitochondria in HNSCC (61). Our data suggest that mitophagy is a controlled compensatory repair mechanism of damaged mitochondria as opposed to a cell death mechanism, and that therapeutic inhibition of autophagic/lysosomal processes contributes to lethal mitophagy in cancer. Discrepancies between the present and previous studies may be due to differences in the therapeutic approach (exogenous CNL addition versus increasing endogenous C18-ceramide through ceramide synthase 1), differences in chain length, trafficking, and metabolism of these ceramides, and potentially type of mitophagy induced. Our studies extend or modify the work of the Ogretmen lab, suggesting a critical role of lysosomal processing to repair damaged mitochondria. Further studies are needed to elucidate key differences between protective and lethal mitophagy.
Therapeutically, these studies support that disruption of interconnected autophagic/mitophagic or lysosomal pathways augment the mitochondrial-dependent cell death mechanisms induced by ceramide. Of interest, the primary chemotherapeutics used to treat HNSCC, cisplatin and 5-fluorouracil (62), both increase ceramide levels in various cancer models, including HPV+ HNSCC (11, 12). Furthermore, inhibiting autophagy with CQ sensitizes HNSCC cells to both cisplatin and 5-fluorouracil (11, 63). CNL is currently being evaluated in a phase I monotherapy trial and is well tolerated. CQ is also well tolerated and while previously appreciated as an antimalarial drug, it is now in multiple single agent and combinatorial trials for cancer (64). Thus, agents like CQ or AMS may potentially be repurposed to improve the efficacy of ceramide-based therapeutics, including CNL, in drug-resistant tumors via blocking mitophagy and exacerbating ceramide-induced mitochondrial damage. Taken together, ceramide or ceramide-generating therapeutics in combination with inhibitors of autophagy/lysosome function may be a viable treatment strategy for HNSCC.
The synergistic effects of CNL + CQ upon mitochondrial dysfunction points to a novel therapeutic intersection between ceramide and lysosomal inhibitors. We further conclude that therapeutic synergy between CNL and CQ is attributed to ceramide-induced mitochondrial stress, which can no longer be mitigated by mitophagy due to inhibition of lysosome function on which mitophagy relies. The promise of enhanced ceramide efficacy by mitigating mitophagy-dependent resistance bodes well for the combination CNL + CQ as a treatment strategy for HNSCC.
Disclosure of Potential Conflicts of Interest
M. Kester reports being a co-inventor of the ceramide nanoliposome and co-founder of Keystone Nano, Inc., which has licensed the technology from Penn State; reports grants from NIH NCI (2P01 CA171983-06A1) during the conduct of the study; other from Keystone Nano, Inc. (Keystone Nano, Inc. has licensed ceramide nanoliposomes from Penn State Research Foundation; M. Kester is CTO and co-founder of Keystone Nano) outside the submitted work; and has a patent for Method and System for Systemic Delivery of Growth Arresting Bioactive Lipids issued and licensed to Keystone Nano, Inc. (US9028863) and a patent for ceramide anionic liposome composition issued and licensed to Keystone Nano, Inc. (US8747891). No potential conflicts of interest were disclosed by the other authors.
The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.
J.J.P. Shaw: Conceptualization, data curation, software, formal analysis, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. T.L. Boyer: Conceptualization, data curation, formal analysis, investigation, writing–original draft, writing–review and editing. E. Venner: Conceptualization, data curation, validation, investigation, writing–review and editing. P.J. Beck: Data curation, formal analysis, validation, investigation. T. Slamowitz: Data curation, formal analysis, validation, investigation. T. Caste: Data curation, formal analysis, validation, investigation. A. Hickman: Formal analysis, investigation. M.H. Raymond: Resources, data curation, validation, investigation, methodology. P. Costa-Pinheiro: Conceptualization, data curation, methodology, writing–review and editing. M.J. Jameson: Conceptualization, resources, supervision, project administration, writing–review and editing. T.E. Fox: Conceptualization, resources, data curation, formal analysis, supervision, methodology, project administration, writing–review and editing. M. Kester: Resources, supervision, funding acquisition, project administration, writing–review and editing.
This work is supported by the NIH/NCI 2 T32 CA009109 and 1 P01 CA171983. The authors would like to thank Stacey Criswell and Adrian Halme from the UVA Microscopy Core.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.