Myc linked to dysregulation of cholesterol transport and storage in nonsmall cell lung cancer

Abstract

Nonsmall cell lung cancer (NSCLC) is a leading cause of cancer-related deaths. While mutations in Kras and overexpression of Myc are commonly found in patients, the role of altered lipid metabolism in lung cancer and its interplay with oncogenic Myc is poorly understood. Here we use a transgenic mouse model of Kras-driven lung adenocarcinoma with reversible activation of Myc combined with surface analysis lipid profiling of lung tumors and transcriptomics to study the effect of Myc activity on cholesterol homeostasis. Our findings reveal that the activation of Myc leads to the accumulation of cholesteryl esters (CEs) stored in lipid droplets. Subsequent Myc deactivation leads to further increases in CEs, in contrast to tumors in which Myc was never activated. Gene expression analysis linked cholesterol transport and storage pathways to Myc activity. Our results suggest that increased Myc activity is associated with increased cholesterol influx, reduced efflux, and accumulation of CE-rich lipid droplets in lung tumors. Targeting cholesterol homeostasis is proposed as a promising avenue to explore for novel treatments of lung cancer, with diagnostic and stratification potential in human NSCLC.

Lung cancer is the leading cause of cancer-related mortality, with nonsmall cell lung cancer (NSCLC) the most common subtype (1). Mutations in Kras are found in more than 30% of NSCLC cases, while the RTK/RAS/RAF pathway is activated in 76% of cases (2). In addition, Myc is frequently overexpressed, focal amplifications occurring in more than 30% of lung adenocarcinomas (3). Myc is a transcription factor with numerous functions in healthy cellular processes, including regulation of cell cycle and cell growth. Deregulation of Myc leads to uncontrolled cell proliferation in many tissues and is implicated in tumorigenesis of some, perhaps all, tumors (4, 5). Inhibition of Myc is therefore of interest as a cancer treatment (6); however, its complex regulation of transcription factors and genes are not fully understood.

Enhanced lipid synthesis is recognized as a signature of cancer (7–9). Excess lipids are stored in lipid droplets, providing a source of energy for rapidly dividing cancer cells and structural components for building new membranes. Furthermore, lipids are increasingly being recognized as critical to signaling pathways in cancer. Cholesterol is a particularly important lipid messenger for signal transduction, control of membrane fluidity, and regulation of the innate immune response (10–12), while cholesteryl ester (CE) accumulation has been implicated in prostate cancer aggressiveness (13). The link between oncogenic Myc and lipid metabolism is relatively underexplored. Previously, we showed that dysregulated Myc modulated the production of eicosanoids, critical for proliferation and cell survival, in lung adenocarcinoma (14). Other recent studies link Myc to increased lipogenesis in tumors (15, 16), while switching from a high- to low-fat diet attenuates the Myc transcriptional program in prostate cancer (17). It is evident that lipids and their interaction with oncogenes, such as Myc, play a complex and elegant role in tumorigenesis and offer an underexploited therapeutic avenue.

Here we use a transgenic mouse model of Kras-driven lung adenocarcinoma with reversible activation of Myc. We explore the effect of Myc activity on cholesterol homeostasis in lung tumors by integrating surface analysis MS-based lipidomics, transcriptomics, and quantitative gene expression analysis. Our results reveal that increased Myc activity favors cholesterol influx over efflux in tumors and leads to the accumulation of CEs stored in lipid droplets. Deactivation of Myc triggers the clearance of cholesterol through increasing efflux, decreasing influx and further increasing cholesterol esterification. No accumulation of cholesterol occurred in tumors in which Myc was never activated. These findings provide new insights into the role of oncogenic Myc and dysregulation of cholesterol homeostasis in lung cancer.

MATERIALS AND METHODS

Tumor models

Mice were maintained on a regular diet in a pathogen-free facility on a 12 h light/dark cycle with continuous access to food and water. Lung tumors were generated in adult Kras^tm4Tyj/+ (LSL-Kras^G12D) or LSL-Kras^G12D; R26^{LSL-CAG-c-MycER/LSL-CAG-c-MycER} (R26^LSL-CMER) mice as previously described (5). Briefly, mice were anesthetized (isoflurane), and the intranasal instillation of adeno-Cre virus (7 × 10⁸ – 3.5 × 10⁹ plaque-forming units; University of Iowa Viral Vector Core) was performed by placing virus droplets on the nose of the mouse. This resulted in the expression of Cre-recombinase removing the stop element sporadically in the lung epithelium of both the Kras and Rosa26 alleles. Consequently, these lung epithelial cells express oncogenic Kras^G12D, driving the formation of lung adenocarcinomas. Tamoxifen added to the diet of LSL-Kras^G12D; R26^LSL-CMER mice results in the activation of the c-MycER^T2 protein, specifically within the Cre-deleted tumor tissues (18). The subsequent removal of tamoxifen results in the rapid deactivation of Myc^ER.

Twelve adeno-Cre-infected LSL-Kras^G12D; R26^LSL-CMER mice were maintained on a tamoxifen-containing diet for 1 month. Tamoxifen was removed from the diet for 24 h (n = 3) or 72 h (n = 3) to deactivate Myc before culling (“Myc inactive”). The remaining mice were continuously fed a tamoxifen-containing diet (“Myc activated”). LSL-Kras^G12D were also fed a tamoxifen-containing diet; however, Myc expression was unaffected (“Kras only”; n = 4). Lung samples were collected, snap-frozen in liquid nitrogen, and stored at –80°C. All animal procedures were approved by the UK Home Office and the University of Cambridge.

Oil red staining for neutral lipids was performed according to an established protocol (19). Fresh-frozen unfixed tissue sections were incubated with working oil red O solution (Sigma-Aldrich) for up to 10 min and quickly rinsed in tap water. This was followed by counterstaining with Mayer’s Haematoxylin (VWR) and a prolonged wash in tap water. Slides were then mounted with glycerol and imaged immediately.

Microarray

Lung cells from adeno-Cre-infected R26^LSL-CMER and LSL-Kras^G12D; R26^LSL-CMER mice were isolated as previously described (20) 14 days after treatment with (LSL-Kras^G12D; R26^LSL-CMER) or without (R26^LSL-CMER) tamoxifen. Single cells were flow-sorted for positivity of GFP directly into Trizol (Thermo Fisher Scientific). To increase RNA yield, cells from two mice of the same genotype and treatment were combined, resulting in two biological replicates for each condition. RNA was extracted with Trizol following the manufacturer’s protocol. RNA was amplified using the Ovation Pico WTA v2 kit (NuGEN Technologies) and subsequently labeled using the BiotinIL kit (NuGEN). RNA was assessed for concentration and quality using a SpectroStar (BMG Labtech) and Bioanalyser (Agilent Technologies). The concentration, purity, and integrity of the resulting cDNA were measured using the Nanodrop ND-1000 (Thermo Fisher Scientific) and by Bioanalyser. cDNA was hybridized to the MouseWG-6 v2 BeadChip overnight followed by washing, staining, and scanning using the Bead Array Reader (Illumina).

Raw data were loaded into R using the lumi package (21) and divided into subsets according to the groups being compared. Subsets were then filtered to remove any nonexpressed probes using the detection P value from Illumina. Across all samples, probes for which the intensity values were not statistically significantly different (P > 0.01) from the negative controls were removed (∼25,000 probes remained). Following filtering, the data were transformed using the variance stabilization transformation (22) from lumi and then normalized to remove technical variation between arrays using quantile normalization. Comparisons were performed using the limma package (23), with results corrected for multiple testing using false discovery rate correction.

Pathway enrichment analysis was performed on differentially expressed genes (P < 0.05; −1.5 > fold change > 1.5) using gene set enrichment analysis (24) to identify overrepresented gene ontology terms for biological process. Pathways were ranked according to their normalized enrichment scores (NESs) and false discovery rates (q). Network analysis was performed using Cytoscape (25). Transcription factor analysis was carried out with TRRUST database in the EnrichR platform (26, 27) to identify potential upstream regulators that influence the expression of dysregulated genes. These were ranked according to P value and predicted activation status.

Lipid profiling

Samples were embedded in Tissue-Tek OCT, and 12 μm tissue sections were prepared on glass microscope slides using a cryostat. Sections were dried in a vacuum desiccator for 30 min prior to analysis. Adjacent sections were stained with H&E. Lipids for MS were extracted using liquid extraction surface analysis (LESA) at user-defined points (1 mm²) across the tissue surface by dispensing 0.8 μl 1:2:4 chloroform-methanol-isopropanol with 10 mM ammonium formate and incubating for 2.5 s. Analytes extracted from the surface were directly infused by a Triversa Nanomate (Advion BioSciences) with a capillary voltage of 1.2 kV, capillary temperature of 200°C, and 0.3 gas flow for 1 min into an LTQ Orbitrap Elite mass spectrometer (Thermo Fisher Scientific). Mass spectra were acquired in the positive ion mode from 200 to 1,000 m/z at 60,000 mass resolution.

Data were converted to mzML format, and features were extracted using an in-house R script. Lipid annotation was performed by accurate mass using the LipidMaps database (<5 ppm) (28). Metaboanalyst software was used to further analyze the data (29). Features with >50% missing values were removed, and the remaining missing values were imputed using the k-nearest neighbors method (30). The features with the lowest 10% mean signal intensity were removed; data were then normalized to the total ion count, mean-centered, and scaled by dividing by the standard deviation of each variable. Principal components analysis identified outliers that were subsequently excluded. A volcano plot represents a combination of fold change and unpaired t-test P values for individual variables. Heatmap was generated in Metaboanalyst using the Ward clustering method with Euclidean distance measure. Features were centered and scaled using the “Autoscale” function, and the top 25 hits based on their significance (ANOVA) were calculated.

Transcript quantification

Total RNA was purified from tumor-containing lung tissue using RNeasy Mini Kit (Qiagen). Approximately 20 mg of each tissue sample was lysed and homogenized in Trizol (1 ml) using a TissueLyzer (Qiagen). The samples were centrifuged at 12,000 g for 15 min after the addition of chloroform (200 μl) and the RNA-containing aqueous phase combined with 1 vol 70% ethanol. Samples were loaded on spin columns, and the procedure was performed according to the manufacturer’s guidelines. Purified RNA concentration was quantified (260 nm) using a NanoDrop 100 (Thermo Fisher Scientific). Genomic DNA contamination was eliminated using RT² First Strand Kit (Qiagen), and cDNA was produced using an RT² First Strand Kit (Qiagen). The relative abundance of transcripts of interest was assessed using quantitative polymerase chain reaction in RT² SYBRgreen Mastermix (Qiagen) with a StepOnePlus detection system (Applied Biosciences). RT² primer assays for mouse Rn18s (endogenous control), Ldlr, Olr1, Scara1, Hmgcr, Soat1, Apoc1, Abca1, Abca3, Abcg1, Srebf1, Srebf2, Ppara, Pparg, and Nr1h3 were obtained from Qiagen. Thermocycler (PTC-200; MJ Research) parameters were as follows: incubation, 95°C for 10 min; elongation, 95°C for 15 s; and cooling, 60°C for 1 min. Elongation and cooling were performed in 40 cycles. Expression levels were normalized to endogenous controls using the ΔΔC_T method, and fold changes were reported relative to the “Kras only” group. Statistical significance was determined using one-way ANOVA (α = 0.05) and adjusted for multiple comparisons using the Holm-Sidak method.

RESULTS

Enrichment of lipid-related pathways in lung cells expressing high Myc and oncogenic Kras

First, we determined pathways modulated by Myc and Kras by studying the gene expression of lung cells from tamoxifen-treated LSL-Kras^G12D; R26^LSL-CMER mice compared with untreated R26^LSL-CMER mice. Using fluorescence-activated cell sorting, we sorted lung cells based on GFP positivity from the R26^LSL-CMER allele. We compared the gene expression profiles for lung cells expressing both active (tamoxifen-treated) Myc^ER and oncogenic Kras^G12D with those that had inactive (untreated) Myc^ER. Pathway analysis of differentially expressed genes (P < 0.05; −1.5 > fold change > 1.5) revealed that biological processes relating to cell cycle, DNA replication and repair, ribosome biogenesis, and RNA processing were significantly overrepresented (P < 0.05; NES >1) in the Myc^ER/Kras^G12D-expressing cells as expected (supplemental Table S1, Fig. 1A). Interestingly, pathways pertaining to immune signaling, inflammation, lipid metabolism, and transport were also significantly enriched (P < 0.05; NES < 1). Of the lipid-related pathways, key terms included lipid metabolic process, lipid localization, cellular response to lipid, steroid metabolic process, lipid catabolic process, and fatty acid metabolic process (supplemental Table S1, Fig. 1B).

Fig. 1.

Pathway analysis for differentially expressed genes in Kras^G12D/Myc^ER-positive lung cells. Gene expression was compared (n = 2) between GFP-positive cells sorted by fluorescence-activated cell sorting from LSL-Kras^G12D; R26^LSL-CMER mice treated with tamoxifen and untreated R26^LSL-CMER mice (each biological replicate represents combined cells from two mice). Gene set enrichment based on gene ontology (biological process), with subsequent network analysis (P < 0.05; q < 0.1; A). The size of the circles (nodes) reflect the P value. The thickness of the lines (edges) reflect the degree of overlap. Gene sets are color-coded according to their broad function. Selected enriched gene ontologies are shown (q < 0.1; B), with the full pathway list in supplemental Table S1. Activation status of upstream regulators (full list in supplemental Table S2). The intensity of the color scale reflects a magnitude of −log10(P); red and blue colors show the predicted activated and deactivated transcription factors, respectively (C).

Transcription factor analysis identified several upstream regulators significantly (P < 0.05) associated with the input gene expression data sets (Fig. 1C; full list in supplemental Table S2). As expected, there was predicted activation of upstream regulators associated with cancer (MYC, TP53), regulation of cell cycle and division (MYC, MYCN, E2F1, and E2F4), and response to hypoxia (HIF1A). Predicted inhibition of key transcriptional regulators of lipid metabolism was found. These included the peroxisome proliferator-activated receptors (PPARA, PPARD, and PPARG), retinoic acid receptor, and sterol regulatory element-binding proteins (SREBPF1 and SREBPF2; regulators of cholesterol biosynthesis and uptake).

In situ extraction of lipids from lung tumors

Given the changes to lipid metabolic pathways, we carried out lipid profiling on LSL-Kras^G12D; R26^LSL-CMER and LSL-Kras^G12D mice. Both sets of mice express oncogenic Kras^G12D, which drives the formation of lung adenocarcinomas. Using tamoxifen-inducible activation of Myc^ER, lung tumors in LSL-Kras^G12D; R26^LSL-CMER mice expressed high levels of Myc^ER in its active or inactive form. Myc expression in lung tumors from LSL-Kras^G12D mice was unmodified (“Kras only”). LESA-MS was used to extract lipids from the surfaces of tissue sections. Sampling locations could be precisely defined and were guided by adjacent H&E-stained tissue sections. LESA-MS involves a liquid microjunction being formed when a droplet of solvent makes contact with the tissue surface (31, 32). Lipids were dissolved in the solvent droplet and directly infused into a high-resolution mass spectrometer. More than 500 features were detected and annotated by their accurate mass and database searching. These covered multiple lipid classes, including phosphatidylcholines, phosphatidylethanolamines, triacylglycerides (TAGs), plasmalogens, diacylglycerides, lysophosphatidylcholines, lysophosphatidylethanolamines, sphingomyelins, ceramides, and CEs.

A comparison of global lipid profiles for Kras and Myc tumors suggested a relative increase in CEs and plasmalogens when Myc was activated (Fig. 2A, supplemental Figure S1) compared with the Kras control. In addition, there was a relative decrease in DAGs and TAGs (particularly those containing shorter fatty acyl chains; supplemental Figure S1). Examining the LESA-MS spectra, a clear increase in the peak at m/z 668.6348 was observed when Myc was activated compared with “Kras only” tumors (Fig. 2B). This peak was assigned as the ammonium adduct of CE(18:1). We next carried out an in-depth analysis of all the lipid features across Myc activated/deactivated tumors and their paired nontumor tissue and Kras control tumors. Of the top 25 features, almost half corresponded to CEs (Fig. 2C, supplemental Table S3). These were increased in tumors in which Myc was activated compared with Kras control and further increased following Myc deactivation. The observed changes were specific to tumors.

Fig. 2.

LESA-MS analysis of lung tumors. LESA was used to extract lipids directly from tissue surfaces at user-defined points. Direct comparison of lipid profiles for “Kras only” and “Myc activated” tumors (four biological replicates per group; A). Representative spectra from “Kras only” and “Myc activated” lung tumors (B). Highlighted peak is the feature corresponding to cholesteryl oleate, CE(18:1), at m/z 668.6348. Heatmap analysis of lipid features across the tumor groups (“Kras only” = 4, “Myc activated” = 4, “Myc deactivated” = 6) and paired nontumor (NT) tissue from the Myc activated/deactivated groups. The top 25 features (based on ANOVA) are shown (C).

Cholesteryl esters accumulate in lung tumors with high Myc

Intratumor CE accumulation has been implicated in prostate, breast, and pancreatic cancers (13, 33, 34). However, it has not been reported in NSCLC, and neither has it been linked to Myc. Our LESA-MS results point to an increase in CE upon Myc activation and a further increase following Myc deactivation (Fig. 3A). This was confirmed by LC-MS/MS analysis of bulk tissue extracts (supplemental Figure S2). An analysis of the CE fatty acyl chain compositions showed that CEs containing a MUFA were disproportionately increased compared with their saturated and polyunsaturated counterparts (Fig. 3A).

Fig. 3.

CEs differentiate lung tumors with varying Myc activity. Comparison of individual CE species across the tumor groups revealed that activation of Myc resulted in a large global increase in the abundance of CEs and free cholesterol, with further increases upon Myc deactivation. The inset shows changes in CEs across different fatty acid compositions (A). The bound to free cholesterol ratio (CE to CB) was calculated in tumors and in the nontumor tissue from the “Myc activated” group (B). Oil red O staining of neutral lipids was performed. Representative images (scale bar = 50 μm) are shown for each group (C). The CE to TAG ratio was calculated in tumors and in the “Myc activated” nontumor tissue (D). Data are means ± SEMs; *P < 0.05, **P < 0.01, and ***P < 0.001.

The accumulation of free cholesterol can induce an apoptotic response, while the biologically inert esterified form is preferred for storage. The measurement of free cholesterol with LESA-MS has limitations: because CEs fragment readily during the ionization process, it is difficult to reliably distinguish free cholesterol from CE fragments. We mitigated this limitation by using a low capillary temperature, tuning the source parameters to minimize CE fragmentation. Furthermore, isobaric species of cholesterol cannot be ruled out, and thus we refer to the species measured at m/z 369.3519 as a “cholesterol-like backbone” (CB). We calculated the ratio of bound (CE) to free (CB) cholesterol in tumors and found an increase on Myc activation compared with Kras, with a further increase over time with Myc deactivation (Fig. 3B). These changes were not recapitulated in the corresponding nontumor tissue. This highlights that the dysregulation of cholesterol storage is limited to the tumor environment and demonstrates the importance of spatially resolved metabolic analysis over bulk tissue measurements.

Cholesterol and CEs may be incorporated into lipid membranes or stored in lipid droplets. In order to establish whether there was any change in lipid droplet content with Myc, we carried out oil red O staining for neutral lipids. There was significant lipid droplet staining in the adjacent nontumor tissue, which did not vary noticeably between the different groups (Fig. 3C). These lipid droplets may arise from the presence of pulmonary surfactant, of which cholesterol is a major component (35). In contrast, lipid droplet staining in the tumor region was low in the “Kras only” group and substantially greater in tumors with high Myc activity (Fig. 3C). Consistent with the MS results, lipid droplet staining was further increased following the deactivation of Myc. Because lipid droplets are primarily composed of TAGs and CEs, we calculated the ratio of CE to TAG to test whether the increased CE content was the result of increased storage of neutral lipids more generally (Fig. 3D). Interestingly, this ratio was substantially increased in tumors with Myc activation compared with Kras and to paired nontumor tissue, suggesting that changes to cholesterol/CE are not purely the result of increased lipid storage per se.

Cholesterol transport and storage linked to Myc activity

Transcriptomics and lipidomics experiments point to alterations in cholesterol homeostasis in the lung with Myc activity. To investigate this further, we extracted lung RNA from the different groups and carried out transcript quantification. We specifically targeted genes involved in cholesterol influx and efflux, as well as those involved in cholesterol synthesis and esterification. No significant differences were noted in expression for Hmgcr, the rate-determining step in cholesterol synthesis (Fig. 4). On the other hand, expression for genes governing cholesterol influx (Ldlr, Olr1, and Scara1) increased with Myc activation and decreased with time after Myc deactivation (Fig. 4). The reverse trend was established for those genes whose main role is cholesterol efflux out of the cell. These transcripts, which included Abca1, Abca3, and Abcg1, decreased with Myc activity (Fig. 4). Expression of Soat1, responsible for esterification of cholesterol, increased with time following deactivation, in line with the increased CE species noted by lipid profiling (Fig. 4). APOC1, which plays a major role in lipoprotein metabolism and cholesterol transport, has been proposed as a diagnostic/prognostic marker in lung cancer (36). Here, gene expression for Apoc1 was markedly increased by Myc activation (Fig. 4).

Fig. 4.

Cholesterol transport and storage linked to Myc activity. Transcript quantification in lung tissue for genes related to cholesterol synthesis (Hmgcr), influx (Ldlr, Olr1, Scara1), efflux (Abca1, Abca3, Abcg1), esterification (Soat1), transport (Apoc1), and regulation (Nr1h3, Srebpf1). Data are means ± SEMs. P values were calculated using ANOVA (*P < 0.05, **P < 0.01, and ***P < 0.001).

Finally, we examined gene expression for nuclear receptors and transcription factors considered important for regulating cholesterol homeostasis. Srebf1 expression decreased when Myc was activated (Fig. 4), while no link to Myc was found for Srebf2 or PPAR (Ppara, Pparg) gene expression (supplemental Figure S3). Nr1h3, which codes for the α subunit of LXR, was also decreased when Myc was activated, increasing following Myc deactivation (Fig. 4). LXRs form heterodimers with the retinoid X receptor and PPARs to regulate lipid homeostasis, in particular acting as cholesterol sensors, promoting the transcription of genes that protect from cholesterol overload (37). Its decrease with Myc activation is consistent with the downregulated transcription of cholesterol efflux transporters and the predicted suppression of upstream regulator PPAR (Fig. 1C).

To assess whether changes to cholesterol transport and storage were recapitulated in human NSCLC, we mined two data sets from the GEO repository (Fig. 5A). Tumor and nontumor samples from a population of nonsmoking women with NSCLC were compared in GSE19804 (38, 39). We found significantly increased expression for SOAT1 and significantly decreased expression of the ABC transporters ABCA1, ABCA3, and ABCG1. Interestingly, in a separate study (40) comparing the two main subtypes of NSCLC, the expression of SOAT1 was increased in adenocarcinoma compared with squamous cell lung cancer (GDS3627; Fig. 5B).

Fig. 5.

Dysregulation of cholesterol efflux and storage in human NSCLC. Gene expression for ABC transporters was decreased and SOAT1 increased in lung tumors from a population (n = 60) of nonsmokers with NSCLC (compared with paired nontumor tissue; GSE19804; A). SOAT1 expression was also increased in lung adenocarcinoma (n = 40) compared with squamous cell carcinoma (n = 18; GDS3627; B). Kaplan-Meier survival estimates for two expression groups of ABCA3 mRNA in lung adenocarcinoma patients (N = 514; TCGA; C).

Last, survival data representing months of disease-specific survival were downloaded from the CBioPortal (41) and applied to the lung adenocarcinoma data set (N = 514) from the TCGA PanCancer atlas (42). Based on the mRNA expression z-score threshold relative to all samples, patients were classified into two expression groups per gene of interest. The correlation between expression level and patient survival were examined using Kaplan-Meier survival curves. Interestingly, we found that low ABCA3 expression was associated with a significantly lower 5-year survival rate (log-rank P value = 0.0008; Fig. 5C).

Overall, we have shown that pathways regulating cholesterol transport and storage are modulated at the gene expression level in lung tumors, are linked to oncogenic Myc, and have potential as diagnostic markers or patient stratification in human NSCLC.

DISCUSSION

Through a combination of transcriptomics, lipidomics, and quantitative gene expression analysis, we show that Myc activity in lung tumors is linked to a disruption in cholesterol homeostasis. Transcriptomics and pathway analysis revealed the dysregulation of lipid metabolism in Myc^ER/KRas^G12D-positive cells compared with the controls. By extracting lipids directly from tumor surfaces, we found a dramatic increase in CEs when Myc was activated compared with when Myc was unmodified. Finally, we showed that the expression of genes relating to cholesterol influx, efflux, and esterification were linked to Myc activity.

There was a striking correlation between the expression of ABC transporters and Myc activity. These transporters are expressed in both alveolar epithelial cells and macrophages and regulate cholesterol clearance out the cell to circulation via HDLs. A decrease in their expression leads to increased cholesterol loading and inflammation in cells (43–45). It is therefore plausible that the accumulation of cholesterol in tumors with high Myc activity is a result of the predicted deactivation of PPAR/LXR that promotes cholesterol efflux through the induction of transporter expression.

The link between Myc and cholesterol uptake is less clear. Increased expression for several genes pertaining to cholesterol uptake was found in tumors with high Myc activity. On the other hand, SREBP, a key regulator of cholesterol uptake and biosynthesis, was predicted to be deactivated in the upstream regulator analysis of oncogenic MYC/KRAS lung cells. One possible explanation is that SREBP is suppressed by the high levels of cholesterol in tumor cells, which could also account for the corresponding decrease of short-chain TAGs in MYC-activated tumors.

Taken together our data show that upon activation of Myc there is a net accumulation of cholesterol that results from the dysregulation of influx/efflux pathways. Esterification of free cholesterol and storage in lipid droplets avoids cholesterol-induced apoptotic death of cancer cells. When Myc is subsequently deactivated, there is a return toward homeostasis. The influx of cholesterol decreases and efflux increases, and to process accumulated cholesterol, further increases to cholesterol esterification occur (Fig. 6).

Fig. 6.

Proposed mechanism for control of cholesterol transport by Myc. The activation of Myc in Kras-driven lung tumors results in increased cholesterol influx and reduced cholesterol efflux, while the synthesis of cholesterol appears largely unaffected at the gene expression level. Subsequent deactivation of Myc results in decreased cholesterol influx and increased efflux. Esterification prevents the accumulation of excess free cholesterol in the cell and is increased with Myc deactivation.

The accumulation of nonadipocyte lipid droplets more generally has been associated with aggressive cancers and linked to inflammation, mitochondrial dysfunction and oxidative stress (12, 46, 47). These lipid droplets maintain a store of excess fatty acids packaged into TAGs and CEs. Hydrolysis of these neutral lipid-containing droplets releases fatty acids and free cholesterol that may be then used by cancer cells for altering membrane fluidity and fueling cellular proliferation and downstream cellular signaling (48, 49). A recent study has shown that Myc and Ras cooperate in NSCLC to reprogram inflammation and immune response. Deregulation of Myc in lung epithelia triggers the release of specific cytokines, which recruit macrophages. These macrophages in turn stimulate angiogenesis, inflammation, and clearance of specific immune cells (50). Interestingly, cholesterol imbalance has been implicated as a contributor to immune dysfunction; for instance, cholesterol loading in macrophages promotes toll-like receptor signaling and activation of the inflammasome (11). Undoubtedly, dysregulation of cholesterol homeostasis will have an important effect on many cellular functions, including inflammatory and immune responses. The measurement of oxysterols may provide a missing link to further understand these complex mechanisms in future studies. These understudied bioactive metabolites of cholesterol act as ligands for LXR, suppress SREBP (51), and have been linked to the modulation of immune response and cancer (52).

This study furthermore suggests new potential drug targets for lung cancer therapy. Treatment with statins, which reduce serum cholesterol levels and proinflammatory signaling, are already being explored in preclinical trials for a variety of cancers. Statins have also been demonstrated to benefit a variety of pulmonary diseases in which inflammation plays a role, as well as improving survival rates for patients with stage IV NSCLC (10, 53, 54). Another intriguing prospect for NSCLC treatment is by the inhibition of SOAT1. This is currently being explored to treat aggressive cases of prostate cancer, while studies have shown SOAT1 inhibition to prolong survival and suppress tumor growth and metastasis in mouse models of different cancers (33, 55, 56). Last, because Western diets are typically high in cholesterol, with obesity a risk factor for developing a variety of cancers, dietary intervention in combination with drug therapy may be a fruitful area for future clinical trials (57).

To conclude, a murine model of lung adenocarcinoma was used to link changes in cholesterol transport and storage to Myc activity, pointing to dysregulation of cholesterol homeostasis in NSCLC. The accumulation of cholesterol may play a role in downstream signaling or be used to fuel cellular proliferation. Targeting cholesterol metabolism is therefore proposed as a promising avenue for exploring novel treatments of lung cancer.

Data availability

All data can be found in the manuscript and supplementary material. Raw microarray data have been deposited in the ArrayExpress database at EMBL-EBI (www.ebi.ac.uk/arrayexpress) under accession number E-MTAB-9131.