Exogenous cholesterol promoted de novo biosynthesis and cancer cell migration in lung adenocarcinoma cells

Lung tissue is a cholesterol-low tissue compared to breast, brain, and liver tissues (S1A Fig). However, cholesterol levels of lung adenocarcinoma were significantly higher than that of adjacent lung tissues (Fig 1A). We observed similar features in in vitro lung adenocarcinoma cellular models, which showed significantly higher cholesterol levels compared to normal lung epithelial cells (Fig 1B), suggesting the involvement of cholesterol metabolism in lung cancer.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Lung adenocarcinoma cells sense exogenous cholesterol as signal and acquire cellular cholesterol via de novo biosynthesis, instead of uptake, for cell migration.

(A) The cholesterol levels of human lung adenocarcinoma and adjacent lung tissues (n = 20). (B) The cholesterol levels of lung cancer and normal epithelial cells (n = 3). (C, D) The cholesterol (C) and qPCR of Ldlr mRNA (D) levels of hepatoma and lung cancer cells treated with 5 μg/ml cholesterol for 24 h in 5% LPDS medium (RMPI1640 containing 5% LPDS, n = 3). (E) Cholesterol biotin (CB) uptake in A549 or HepG2 cells examined by immunofluorescent staining after cells were treated with 5 μg/ml CB cultured in 5% LPDS medium for 24 h. (F, G) Dil-LDL uptake of A549 (F) or HepG2 (G) cells treated with 10 μg/ml Dil-LDL cultured in 5% LPDS medium for 24 h examined by immunofluorescent staining. (H) CB uptake in A549 or HepG2 cells with or without MβCD pretreatment compared to untreated cells examined by immunofluorescent staining. (I) The cholesterol levels of A549 cells treated with 5 μg/ml cholesterol with or without MβCD pretreatment cultured in LPDS medium for 24 h (n = 3). (J) LC-MS analysis of ¹³C-acetate-derived ¹³C fractional contribution in cholesterol and dehydrocholesterol extracted from A549 cells treated with 5 μg/ml cholesterol cultured in LPDS medium (n = 3). (K) qRT-PCR analyses of Dhcr7 and Dhcr24 of lung cancer cells (n = 3). (L) The 7-dehydrocholesterol (7-DHC) and 24-dehydrocholesterol (24-DHC) levels of lung cancer cells were analyzed by ELISA kits (n = 3). (M) The 7-DHC levels of lung cancers cultured in LPDS medium treated with or without 5 μg/ml Cholesterol for 24 h detected by ELISA kit (n = 3). (N) The 7-DHC, cholesterol levels, and qRT-PCR analyses of metastatic-related genes of A549 cells cultured in LPDS medium treated with 5 μg/ml cholesterol with or without 10 μm fluvastatin for 24 h (n = 4). Scale bars, 20 μm (E–H). Data were presented as mean ± SEM. *, P < 0.05; **, P < 0.01. LPDS, lipoprotein deficient serum; CB, Cholesterol biotin; DiI-LDL, Human DiI-Low Density Lipoprotein; TC, total cholesterol; 7-DHC, 7-dehydrocholesterol; 24-DHC, 24-dehydrocholesterol; MβCD, methyl-β-cyclodextrin. The data underlying this figure can be found in the Supporting information file S1 Data.

https://doi.org/10.1371/journal.pbio.3002621.g001

To understand the role of cholesterol in lung cancer development, we treated lung cancer cell lines with cholesterol and compared their responses to the typical high cholesterol tumor hepatoma cells. We found that cholesterol addition did not increase, even slightly suppressed the intracellular cholesterol in hepatoma cells (Fig 1C), which is in consistent with previous reports of an inhibitory role of extracellular cholesterol uptake on cholesterol synthesis in liver cancers [28]. In contrast, cholesterol treatment significantly enhanced intracellular cholesterol levels in lung cancer cells (Fig 1C), accompanied with elevated metastatic gene programs at both mRNA and protein levels, as well as higher migration capacities (S1B–S1D Fig), without obvious influence on cell proliferation (S1E Fig).

To understand why exogenous cholesterol treatment increased intracellular cholesterol synthesis in lung adenocarcinoma cells but not in hepatoma cells, we further treated lung adenocarcinoma cells with biotin-labeled cholesterol (CB) to trace the fate of exogenous cholesterol and comparing it with hepatoma cells [6]. We observed that CB were predominantly retained on cell membranes of lung cancer cells, in sharp contrast to hepatoma cells wherein CB were actively internalized into cell plasma (Figs 1E and S1F). Consistent with this, we found significantly lower expression of cholesterol receptor LDL receptor (LDLR) in lung adenocarcinoma cells compared to hepatoma cells (Fig 1D). Cholesterol is mainly transported and taken up by cells in LDL particles via LDLR. Similarly, we found that florescent LDL (Dil-LDL) were still retained on membranes in lung cancer cells 4 h after treatment (Figs 1F and S1G), while Dil-LDL were notably taken up into cytoplasm in hepatoma cells in a time-dependent manner (Figs 1G and S1G).

Cholesterol has been shown to function as a signal molecule by binding to and modifying factors in Hedgehog signaling pathway enriched on cell membrane [29]. Interfering membrane signal transduction in lung cancer cells with Methyl-β-Cyclodextrin (MβCD), a substance depleting membrane cholesterol [30] (S1H Fig), we found that the subsequent administration of biotin-labeled cholesterol no longer bound to membrane and failed to increase cellular cholesterol levels in lung adenocarcinoma cells (Fig 1H and 1I). These results suggested that exogenous cholesterol may induce different signals in lung adenocarcinoma cells or in hepatoma cells, one involving membrane signaling while one engaging cytosolic cascade. Critically, we treated lung adenocarcinoma cells with ¹³C-acetate tracer and performed metabolic flux analysis. Subsequent ¹³C tracing indicated that exogenous cholesterol potently activated de novo cholesterol synthesis in lung cancer, as evident by the significant increases in ¹³C-dehydrocholesterol (DHC) and ¹³C-cholesterol in treated group compared to control (Fig 1J).

The last step of de novo cholesterol synthesis is the conversion of 24-dehydrocholesterol (24-DHC) or 7-dehydrocholesterol (7-DHC) into cholesterol, catalyzed by 24-Dehydrocholesterol Reductase (DHCR24) or 7-Dehydrocholesterol Reductase (DHCR7), respectively [31]. Interestingly, we found significantly higher mRNA level of DHCR7 compared to DHCR24, while consistently, 7-DHC is the dominant form of cholesterol precursor in lung adenocarcinoma cells (Fig 1K and 1L). Cholesterol treatment significantly increased 7-DHC levels in lung cancer cells (Fig 1M). Fluvastatin inhibits cholesterol biosynthesis [32]. We found that exogenous cholesterol-induced increase in intracellular 7-DHC and cholesterol levels, as well as metastatic gene programs expression (Fig 1N) in lung cancer cells were blunted by Fluvastatin treatment.

Overall, these data indicated that exogenous cholesterol promoted cell migration by potently activating de novo cholesterol synthesis and inducing 7-DHC and cholesterol levels in lung adenocarcinoma, possibly as a way to circumvent the typically low-cholesterol levels in lung tissues.

FOXA3 is a key transcriptional factor to mediate extracellular cholesterol-induced de novo cholesterol biosynthesis in lung adenocarcinoma cells

We next set out to study the mechanism that exogenous cholesterol activated de novo cholesterol synthesis. Dose and time-dependent transcriptional analysis revealed that cholesterol treatment significantly increased Hmgcs1 expression, without impacting critical factors in cholesterol uptake (Ldlr), classic cholesterol biosynthetic regulators (Srebp2, Insig, and Scap), rate-limiting enzyme in biosynthesis cascade (Hmgcr), or cholesterol efflux regulators (Lxr, Apoe, Abca1, Abcg1) in lung adenocarcinoma cells (S2A and S2B Fig). Besides, we did not detect USP20, a deubiquitinase recently reported to stabilize HMGCR in liver [33], in lung adenocarcinoma cells (S2A and S2B Fig). Interestingly, exogenous cholesterol treatment did not increase cholesterol synthesis in normal primary hepatocytes (S2C Fig), indicating these effects may be unique to lung cancer cells. Since extracellular signals often function by inducing changes in intracellular transcription factors, we then performed RNA-seq analysis (S2D Fig) on cholesterol-treated cells and control cells. GO and KEGG analysis revealed a significant enrichment of biological processes related to metastasis, including cell migration, cell junction assembly, and wound healing (S2E and S2F Fig). We also investigated key transcriptional factors that featured significant modulations during this process. Subsequent analysis rendered 34 differentially expressed transcriptional factors, among which FOXA3, a key transcription factor in lipid metabolism [21–25], was top ranked and piqued our interest (Figs 2A and S2G). Indeed, qPCR analysis confirmed that Foxa3 mRNA levels were significantly up-regulated upon cholesterol treatment in lung adenocarcinoma cells, but not in normal lung epithelial cells (Fig 2B). Exogenous cholesterol covalently modified and activated key factors in hedgehog signaling pathway to induce intracellular signaling cascade [6]. Consistently, we found increased level of Gli2, a downstream mediator of hedgehog signaling pathway, in our RNA-seq data (S2G Fig). This was further confirmed by a screen of GLIs family members in lung adenocarcinoma cells, which showed that cholesterol treatment specifically induced GLI2 expression (Fig 2C). Importantly, in silico analysis predicted a putative GLI2 binding site on Foxa3 promoter (S2H Fig), which was confirmed by ChIP assay (Fig 2D). Furthermore, GLI2, but not GLI1 or GLI3, enhanced luciferase activity of FOXA-luc construct (Fig 2E), while GLI2 overexpression dose-dependently promoted Foxa3 transcription in luciferase assay and in qPCR analysis (Fig 2F and 2G). These results suggested that exogenous cholesterol induced Foxa3 transcription in lung adenocarcinoma cells via GLI2.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. FOXA3 is a key transcriptional factor to mediate cholesterol-induced de novo cholesterol biosynthesis in lung adenocarcinoma cells.

(A) Volcano analysis of differentially expressed genes of RNA-seq data of A549 cells treated with or without 5 μg/ml cholesterol (p < 0.05). Red, blue, and gray dots suggest up-regulation, down-regulation, and not changed, respectively. (B) Foxa3 mRNA levels of normal lung epithelia cells and lung adenocarcinoma cells treated with or without 5 μg/ml cholesterol for 24 h in LPDS medium (n = 4). (C) qPCR analysis of GLIs in A549 cells treated with or without 5 μg/ml cholesterol for 24 h in LPDS medium (n = 3). (D) Enrichment of GLI2 occupancy on Foxa3 promoter in A549 cells analyzed by ChIP assay (n = 3). (E) Effects of GLIs on Foxa3 transcriptional activities analyzed by luciferase assay (n = 3). (F, G) Dose-dependent effect of GLI2 on Foxa3 transcriptional activity analyzed by luciferase assay in 293T (F) and by qPCR analysis in A549 cells (G) (n = 3). (H) ChIP-seq visualization of FOXA3 occupancy on the proximal region (top) and in silico analysis of putative FRE (down) on Hmgcs1 promoter in H1299 cells. (I) The binding of FOXA3 on predicted FRE of Hmgcs1 promoter in A549 cells was analyzed by ChIP (n = 3). (J) Dose-dependent Hmgcs1 transcriptional activities in HEK293T cells transiently expressing either vector (PCDH) or FOXAs. (K) Luciferase activities of WT, FRE mutation (Mut), or FRE deletion (Del) Hmgcs1 promoter reporters in HEK293T cells transiently expressing either vector (NC) or FOXA3 (n = 3). (L, M) qRT-PCR (L) and immunoblotting (M) of FOXA3 and HMGCS1 levels in NC or FOXA3 overexpressed A549 cells (n = 3). (N, O) The cholesterol (N) and 7-DHC (O) levels in NC or FOXA3 overexpressed A549 cells (n = 3). (P, Q) qRT-PCR (P) and immunoblotting (Q) of FOXA3 and HMGCS1 levels in control (shNC) or FOXA3 knockdown (shFOXA3) A549 cells (n = 3). (R, S) The cholesterol (R) and 7-DHC (S) levels in control (shNC) or FOXA3 knockdown (shFOXA3) A549 cells (n = 3). (T, U) qRT-PCR analysis of Hmgcs1 and metastatic gene program (T), cholesterol and 7-DHC (U) levels in in control (shNC) or FOXA3 knockdown (shFOXA3) A549 cells treated with or without cholesterol (n = 3). Data were presented as mean ± SEM. *, P < 0.05; **, P < 0.01. FC, fold change; RLU, relative luciferase unit; con, control; chol, cholesterol; IP, FOXA3 immunoprecipitation; NC, target gene overexpression control; FOXA3, forkhead box protein A3; FRE, FOXA3 response element; Del, FRE deletion; Mut, FRE mutation; shNC, target gene knockdown control; shFOXA3, FOXA3 knockdown; ChIP, chromatin immunoprecipitation. The data underlying this figure can be found in the Supporting information file S1 Data.

https://doi.org/10.1371/journal.pbio.3002621.g002

Then, to elucidate the function of FOXA3 in de novo cholesterol synthesis, we performed ChIP-seq to examine its direct target genes, which revealed that HMGCS1 was top ranked (Fig 2H and S1 Table). Besides, overlapping RNA-seq dataset with ChIP-seq dataset revealed that Hmgcs1 is one of the top candidates that both served as direct target genes of FOXA3 and featured significant increase in expression levels upon cholesterol treatment (S2I Fig). HMGCS1 is the critical enzyme that governing the initial step in cholesterol biosynthesis, catalyzing energy substrates into HMG-CoA that is essential for mevalonate formation [34]. Consequently, FOXA3 overexpression or exogenous cholesterol treatment both increased the levels of HMG-CoA, the direct enzymatic product of HMGCS1 (S2J and S2K Fig). ChIP assay further confirmed FOXA3 binding at a canonical FOXA binding site from −1,445 bp to −1,439 bp on Hmgcs1 promoter (Fig 2H and 2I) [35]. In addition, FOXA3, but not its family member FOXA1 or FOXA2, dose-dependently promoted Hmgcs1 transcription in luciferase assay (Fig 2J), while deletion or mutation of the putative FOXA3 binding site on Hmgcs1 promoter abolished transcriptional activation of FOXA3 on Hmgcs1 (Fig 2K). Consistently, FOXA3 overexpression increased, while FOXA3 knockdown decreased Hmgcs1 expression, cellular cholesterol, and 7-DHC levels in lung cancer cells (Fig 2L–2S). Importantly, in accordance with our previous data, exogenous cholesterol treatment increased cellular cholesterol and 7-DHC levels, as well as enhanced the expressions of Foxa3, Hmgcs1, and metastatic gene program in lung adenocarcinoma cells, while these effects were abolished in the scenario of FOXA3 knockdown (Fig 2T and 2U). Overall, these data suggested that FOXA3 is a key mediator of exogenous cholesterol-induced de novo cholesterol biosynthesis and progression, possible via the GLI2-FOXA3-HMGCS1 regulatory axis, in lung adenocarcinoma.

FOXA3 promotes lung cancer cell migration

Having shown that FOXA3 is involved in cholesterol synthesis, we next examined the contribution of FOXA3 to lung adenocarcinoma metastasis. We found in Oncomine database that lung adenocarcinoma tissues from patients characterized significantly higher FOXA3 expression compared to adjacent tissues (S3A Fig). We further confirmed enhanced FOXA3 levels in tumor tissues versus adjacent tissues using samples from lung adenocarcinoma patients by qPCR (S3B Fig), western blot (S3C Fig), and immunohistological analyses (Fig 3A). Moreover, Kaplan–Meier plotter for the correlation between FOXA3 levels and the disease-free survival of lung cancer patients from The Cancer Genome Atlas (TCGA) database revealed that high FOXA3 levels were associated with poor prognosis (Fig 3B).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. FOXA3 deletion alleviates lung cancer cell metastasis in genetic mouse model.

(A) Representative images and quantification for FOXA3 nuclear staining of FOXA3 immunostaining in human lung adenocarcinoma samples and adjacent tissues (n = 6). (B) TCGA database of disease-free survival rate of lung adenocarcinoma patients with different Foxa3 expression levels (Foxa3 low, n = 231; Foxa3 high, n = 289). (C–L) Urethane-induced lung adenocarcinoma mice models (C). Foxa3^+/+ (n = 6) and Foxa3^-/- (n = 6) mice were intraperitoneally injected with 1 mg/g body weight urethane once a week for 10 weeks, and 15 weeks after last injection, mice were sacrificed and lung weights (D), lung tumor numbers (E), liver metastatic tumors (F), and HE staining of liver tissues (G) were analyzed. (H, I) The expression of Hmgcs1 and metastatic gene program in lung (H) and liver (I) tissues of urethane-induced Foxa3^+/+ and Foxa3^-/- mice. (J) The protein levels of Fibronectin, E-cadherin, N-cadherin, and HMGCS1 in lung (Left) and liver (right) tissues of urethane-induced Foxa3^+/+ and Foxa3^-/- mice. (K, L) The cholesterol (K) and 7-DHC (L) levels in lung and liver tissues of urethane-induced Foxa3^+/+ and Foxa3^-/- mice. Scale bars, 50 μm (G). Data were presented as mean ± SEM. *, P < 0.05; **, P < 0.01. IHC, immunohistochemistry; IP, intraperitoneal injection; Foxa3^+/+, wild-type mice; Foxa3^-/-, Foxa3 Knockout mice; TC, total cholesterol; 7-DHC, 7-dehydrocholesterol. The data underlying this figure can be found in the Supporting information file S1 Data.

https://doi.org/10.1371/journal.pbio.3002621.g003

To demonstrate the role of FOXA3 in lung adenocarcinoma metastasis, we investigated the influence of genetic FOXA3 deficiency in in vivo lung adenocarcinoma spontaneous metastasis models by treating WT (FOXA3^+/+) and FOXA3 knockout (FOXA3^-/-) mice with urethane, a chemical lung carcinogen (Fig 3C). We found that compared to WT mice, FOXA3^-/- mice had similar lung weight and significantly lower numbers of lung tumor nodes (Fig 3D and 3E). Of note, the migration of lung cancer to the liver was largely reduced in FOXA3^-/- mice (Fig 3F and 3G), accompanied with significantly reduced HMGCS1 and metastatic markers in both lung and liver, as well as suppressed cholesterol and 7-DHC levels in lung (Fig 3H–3L).

Additionally, to eliminate the possibility that the suppressed tumor colonization to liver in FOXA3 ^-/- mice may be due to the reduced tumor incidence in their lung, we manipulated FOXA3 levels in lung adenocarcinoma cells and assessed its impacts on cancer migration ability both in vitro and in vivo. FOXA3 overexpression (FOXA3-OE) showed no effects on cell proliferation as shown by CCK8 analysis (S4A and S4B Fig), while significantly enhanced cell migration and invasion as shown by transwell assays (Fig 4A), expressions of metastatic gene program (Fig 4B and 4C), and wound-healing assays (S4C Fig). The impacts of FOXA3 on lung adenocarcinoma cell migration capability in vivo were further analyzed using xenograft mouse model. Mice were subjected to tail vein injection using luciferase-labeled FOXA3-OE lung cancer cells. Detailed analysis showed that compared to control group, FOXA3-OE cells resulted in stronger luciferase signal (Fig 4D), increased circulating tumor cells (CTC) in mice circulation (Fig 4E), reduced survival rate (Fig 4F), as well as increased lung colonization areas (Fig 4G), cholesterol (Fig 4H) and 7-DHC (Fig 4I) levels. Moreover, HMGCS1 levels and EMT markers (S4D Fig) in lung tissues were also elevated in FOXA3-OE xenograft mouse models. To further strengthen the role of FOXA3 in tumor metastasis, we have also established subcutaneous implantation models of murine lung cancer cells LLC with FOXA3 overexpression and examine the metastatic capability of these subcutaneous tumors compare to their controls. Consistently, FOXA3 overexpression promotes colonization of lung tissue and subsequent growth of subcutaneous LLC tumors (S4E and S4F Fig). It has been suggested that FOXA3 promoted metastasis of esophageal cancer cells through regulation of other FOXA members [36]. We thus examined FOXA1 and FOXA2 levels in A549 lung adenocarcinoma cells upon FOXA3 knockdown and found that the levels of FOXA1 and FOXA2 were not altered (S4G Fig), suggesting the regulatory role among FOXA family members might be cancer cell type specific. Conversely, FOXA3 knockdown (FOXA3-KD) cells exhibited opposite performance versus FOXA3-OE both in vitro and in vivo, as shown by similar proliferation rate (S4H Fig), decreased metastatic makers (Fig 4J and 4K), subdued in vitro cell migration/invasion (Figs 4L and S4I), as well as weaker luciferase signal (Fig 4M), less CTC in mice circulation (Fig 4N), increased survival rate (Fig 4O), decreased lung colonization areas (Fig 4P), decreased cholesterol (Fig 4Q) and 7-DHC levels (Fig 4R), as well as decreased HMGCS1 levels, along with the EMT markers (S4J Fig) in lung tissues of FOXA3-KD xenograft mouse models. We have also measured Srebp2 mRNA levels upon FOXA3 regulation in A549 and H441 lung cancer cells. The results showed that FOXA3 has no obvious effect on Srebp2 levels (S4K and S4L Fig). Overall, these data suggested that FOXA3 promotes lung cancer metastasis.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. FOXA3 promotes lung cancer cell metastasis both in vitro and in vivo.

(A) Representative images (left) and quantification (right) of transwell assay examining the effect of FOXA3 on A549 cell migration and invasion (n = 4). (B, C) The mRNA (n = 3) (B) and protein levels (C) of metastatic gene program of control (NC) and FOXA3 overexpressed (FOXA3) A549 cells. (D–I) Mice were injected intravenously with control (NC) or FOXA3 overexpressed (FOXA3) A549 cells (n = 8 per group). After 6 weeks, mice were analyzed for IVIS imaging (D), qRT-PCR analysis of hGapdh/mGapdh in circulating tumor cells (E), mice survival rate (F), HE staining (G), cholesterol levels (H) and 7-DHC levels (I) of lung tissues. (J, K) The mRNA (n = 3) (J) and protein levels (K) of metastatic genes in control (shNC) or Foxa3 knockdown (shFOXA3) A549 cells. (L) Representative images (left) and quantification (right) of transwell assay examining the effect of FOXA3 knockdown on A549 migration and invasion (n = 3). (M–R) Mice were injected intravenously with control (shNC) or FOXA3 knockdown (shFOXA3) A549 cells (n = 8 per group). After 8 weeks, mice were analyzed for IVIS imaging (M), qRT-PCR analysis of hGapdh/mGapdh in CTC (N), mice survival rate (O), representative images of HE staining (P), the cholesterol (Q) and 7-DHC levels in lung tissues (R). Scale bars, 50 μm (C, G, L, P). Data were presented as mean ± SEM. *, P < 0.05; **, P < 0.01. IVIS, In Vivo Imaging System; HE, hematoxylin and eosin staining; CTC, circulating tumor cells. The data underlying this figure can be found in the Supporting information file S1 Data.

https://doi.org/10.1371/journal.pbio.3002621.g004

HMGCS1 promotes lung cancer cell migration and is the downstream effector of FOXA3

In addition to FOXA3, Kaplan–Meier plotter from TCGA database also revealed that high GLI2 and HMGCS1 levels were both positively correlated with poor prognosis and decreased overall survival in lung cancer patients (Figs 5A and S5A), suggesting the vital role of GLI2-FOXA3-HMGCS1 axis in lung cancer metastasis. Indeed, we found that HMGCS1 overexpression (HMGCS1-OE) in lung adenocarcinoma cells promoted lung cancer cell migration both in vitro (S5B and S5C Fig) and in vivo (Fig 5B–5E). Conversely, HMGCS1 knockdown (HMGCS1-KD) showed opposite effects (Figs 5F–5I, S5D and S5E). We have also established the subcutaneous implantation models of murine lung cancer cells LLC with HMGCS1 overexpression or knockdown. The analysis of metastatic capability of these subcutaneous tumors suggested that HMGCS1 overexpression promoted (S5F and S5G Fig), while HMGCS1 knockdown suppressed colonization of lung tissue and subsequent growth of subcutaneous tumors (S5H and S5I Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. HMGCS1 promotes lung cancer cell metastasis both in vitro and in vivo and is the downstream effector of FOXA3.

(A) Kaplan–Meier plotter showing the correlation between Hmgcs1 levels and the overall survival of lung adenocarcinoma cancer patients from TCGA database (Hmgcs1 low, n = 966; hmgcs1 high, n = 960). (B–E) Mice were injected intravenously with control (NC) or HMGCS1 overexpression (HMGCS1) A549 cells (n = 8 per group). After 6 weeks, mice were analyzed for IVIS imaging (B), qRT-PCR analysis of hGapdh/mGapdh in circulating tumor cells (C), mice survival rate (D), and representative images of HE staining of lung metastatic nodules (E). (F–I) Mice were injected intravenously with control (shNC) or HMGCS1 knockdown (shHMGCS1) A549 cells (n = 8 per group). After 8 weeks, mice were analyzed for IVIS imaging (F), qRT-PCR analysis of hGapdh/mGapdh in circulating tumor cells (G), representative images of HE staining of lung metastatic nodules (H), and mice survival rate (I). (J–M) Mice were injected intravenously with control (shNC+NC) or FOXA3 knockdown with (shFOXA3+HMGCS1) or without (shFOXA3+NC) HMGCS1 overexpression A549 cells (n = 6 per group). After 6–8 weeks, mice were analyzed for IVIS imaging (J), qRT-PCR analysis of hGapdh/mGapdh in circulating tumor cells (K), representative images of HE staining of lung metastatic nodules (L), and the mRNA levels of metastatic gene program in lung tissues (M). (N–Q) Mice were injected intravenously with control (NC+shNC) or FOXA3 overexpression with (FOXA3+shHMGCS1) or without (FOXA3+shNC) HMGCS1 knockdown A549 cells (n = 6 per group). After 6–8 weeks, mice were analyzed for IVIS imaging (N), qRT-PCR analysis of hGapdh/mGapdh in circulating tumor cells (O), representative images of HE staining of lung metastatic nodules (P), and the mRNA levels of metastatic gene program in lung tissues (Q). Scale bar, 50 μm (E, H, L, P). Data were presented as mean ± SEM. *, P < 0.05; **, P < 0.01. IVIS, In Vivo Imaging System; HE, hematoxylin and eosin staining; CTC, circulating tumor cells. The data underlying this figure can be found in the Supporting information file S1 Data.

https://doi.org/10.1371/journal.pbio.3002621.g005

Furthermore, we found that FOXA3 deletion inhibited lung colonization, while HMGCS1 overexpression rescued this phenomenon in mice (Fig 5J–5M). Moreover, FOXA3 overexpression promoted cancer cell migration both in vivo and in vitro, while HMGCS1 knockdown significantly abolished these effects (Figs 5N–5Q and S6A–S6D). These results suggested that HMGCS1 is the downstream effector of FOXA3 in cholesterol biosynthesis and lung adenocarcinoma progression.

Cholesterol induces cell membrane composition change for lung adenocarcinoma progression

We have shown that FOXA3 promoted de novo cholesterol synthesis and cell migration in lung adenocarcinoma via GLI2-FOXA3-HMGCS1 axis. In order to understand the influences of elevated intracellular cholesterol levels in lung adenocarcinoma cells, we analyzed RNA-seq data in Fig 2 with Gene Ontology (GO) analysis, which demonstrated that cellular membrane is top ranked and the most affected event after exogenous cholesterol treatment (Fig 6A), in accordance with previous report showing cholesterol as a critical contributor in membrane fluidity and membrane remodeling [37]. Indeed, we found that exogenous cholesterol treatment significantly increased cholesterol and 7-DHC incorporation into membrane compared to cytosol in lung adenocarcinoma cells (Fig 6B and 6C). Furthermore, exogenous cholesterol administration in lung adenocarcinoma cells resulted in enhanced cancer cell migration and invasion (Fig 6D), increased expression of metastatic markers, as well as Caveolin-1 (Fig 6E) and downstream signaling (S6E Fig). Notably, these effects were abolished by MβCD treatment, which disrupts membrane signal transduction (Figs 6D, 6E, and S6E).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Cholesterol induces cell membrane composition change for lung adenocarcinoma cell migration.

(A) GO enrichment analysis on differentially expressed genes (p < 0.05) from RNA-seq in A549 cells treated with or without 5 μg/ml cholesterol in LPSD medium for 24 h. (B, C) The ratio of cholesterol (B) and 7-DHC (C) levels in membrane or cytosol of A549 cells treated with or without 5 μg/ml cholesterol in LPSD medium for 24 h. (D) Representative images (left) and quantification (right) of transwell assay examining the migration or invasion of A549 cells treated with cholesterol (5 μg/ml) with or without MβCD (2.5 mM) pretreatment. (E) The effect of cholesterol with or without MβCD pretreatment on metastatic gene program and caveolin1 levels detected by immunoblotting. (F, G) The effects of FOXA3 knockdown (F), FOXA3 overexpression with or without HMGCS1 knockdown (G) on the protein levels of caveolin-1 examined by immunoblotting. (H, I) The effect of FOXA3 on the ratio of cholesterol (H) and 7-DHC (I) levels in membrane or cytosol of A549 cells. (J) Representative images (left) and quantification (right) of transwell assay examining the migration or invasion of FOXA3 overexpressed A549 cells pretreated with or without MβCD for 24 h. (K) The effect of MβCD pretreatment on metastatic gene levels in FOXA3 overexpression A549 cells examined by immunoblotting. Scale bar, 50 μm (D, J). Data were presented as mean ± SEM. *, P < 0.05; **, P < 0.01. n = 3 per group for cell experiments. GO, Gene Ontology; Con, Control; Chol, Cholestrol. The data underlying this figure can be found in the Supporting information file S1 Data.

https://doi.org/10.1371/journal.pbio.3002621.g006

We also found that FOXA3 levels in lung adenocarcinoma cells mimics the effects of exogenous cholesterol, as FOXA3-KD decreased caveolin-1 levels, while FOXA3-OE increased caveolin-1 levels and could be blocked by HMGCS1 knockdown (Fig 6F and 6G). Moreover, FOXA3-OE in lung cancer cells significantly increased cholesterol and 7-DHC incorporation into membrane versus cytosol (Fig 6H and 6I), as well as enhanced cell migration and metastatic marker expression, which were largely abolished by MβCD treatment (Fig 6J and 6K). Overall, these data suggested that membrane composition change at least partially contributed to FOXA3-HMGCS1 axis-induced cholesterol biosynthesis and metastasis in lung adenocarcinoma cell.

Magnolol inhibits Foxa3 transcription and blocks metastatic program of patient-derived lung adenocarcinoma organoids

Having established the role of FOXA3 in lung adenocarcinoma metastasis, we next examined whether pharmacological blockade of FOXA3 would lead to inhibition of metastasis. Via a Foxa3 transcriptional reporter screening (S2 Table), we identified magnolol, an active substance extracted from the bark of Magnolia officinalis and widely used as Chinese traditional medicine for its anti-inflammatory and antispastic effects, potently inhibits Foxa3 expression (Fig 7A). Of note, we demonstrated that magnolol treatment in lung adenocarcinoma cells effectively blocked FOXA3-OE-induced Hmgcs1 expression and metastatic gene program (Fig 7B), cell migration/invasion (Fig 7C), and de novo cholesterol synthesis (Fig 7D). These effects of magnolol are not due to toxicity towards lung adenocarcinoma cells, as the treating concentration does not affect cell viability (S7A–S7C Fig). In order to assess whether magnolol is a selective FOXA3 inhibitor, we investigated the effects of magnolol treatment on the expression of FOXA family members and the results suggested that magnolol suppresses FOXA3 expression specifically in a dose-dependent manner (S7D and S7E Fig). In addition, we analyzed the effects of magnolol on lung adenocarcinoma cells with or without FOXA3. Our results showed that magnolol treatment did not exert additional inhibitory effects on the expression of Hmgcs1 and EMT in the absence of FOXA3 (S7F and S7G Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. Magnolol as a lead compound against lung adenocarcinoma progression in PDOs.

(A) Dose-dependent effect of magnolol on Foxa3 transcriptional activity. (B) qPCR analysis of the mRNA levels of Foxa3, Hmgcs1, and metastatic genesin FOXA3 overexpressed cells treated with or without magnolol at 10 μm (n = 3). (C) Representative images (left) and quantitation (right) of cell migration and invasion in FOXA3 overexpression A549 cells treated with or without magnolol (n = 4). (D) The cholesterol and 7-DHC levels of FOXA3 overexpressed A549 cells treated with or without magnolol (n = 3). (E, F) Relative mRNA levels of Foxa3 and Hmgcs1 (E) and correlation analysis (F) in lung adenocarcinoma PDOs (n = 6) and adjacent tissue PDOs (n = 6). (G, H) mRNA (G) and protein levels (H) of FOXA3, HMGCS1, and metastatic genes of lung adenocarcinoma PDO (n = 4) treated with or without 10 μm magnolol for 24 h. (I) The cholesterol (left) and 7-DHC (right) levels of lung adenocarcinoma PDO (n = 4) treated with or without magnolol. (J) Schematic illustration of exogenous cholesterol promotes endogenous de novo cholesterol synthesis via GLI2-FOXA3-HMGCS1 transcriptional axis, membrane composition change, and cancer metastasis in lung adenocarcinoma. Scale bar, 50 μm (C). Data were presented as mean ± SEM. *, P < 0.05; **, P < 0.01. Mag, Magnolol; PDO, patient-derived organoid. The data underlying this figure can be found in the Supporting information file S1 Data.

https://doi.org/10.1371/journal.pbio.3002621.g007

For clinical relevance, we obtained lung adenocarcinoma and adjacent tissue from patients to establish PDOs, a promising model for personalized cancer treatment. Consistent with our previous observation, Foxa3 and Hmgcs1 levels were higher in cancer PDO compared to adjacent PDO in lung adenocarcinoma patients (Fig 7E). Moreover, Hmgcs1 levels were positively correlated with Foxa3 levels in PDO samples (Fig 7F). In these PDO samples, we chose 4 samples featured highest Foxa3 and Hmgcs1 expression and subjected them for magnolol treatment. Importantly, magnolol effectively inhibited both mRNA (Fig 7G) and protein (Fig 7H) levels of metastatic gene program, as well as cholesterol and 7-DHC levels (Fig 7I) in these PDO samples, indicating magnolol as a promising lead compound in preventing lung adenocarcinoma development, especially in personalized treatment for cancers characterized high-Foxa3 and high-Hmgcs1 expression. In addition, we compared side-by-side the effects of magnolol (Mag) with other extensively studied anti-cancer therapies, i.e., fluvastatin (Fluv) and cisplatin (Cisp) in PDOs by examining cell viability (S7H Fig), proliferation marker Ki67 (S7I Fig), Hmgcs1 (S7J Fig), and EMT gene expressions (S7K–S7O Fig) to compare their anti-neoplastic and anti-metastatic effects on cancer cells. The results of in vitro PDO experiments showed that compared to fluvastatin and cisplatin, magnolol exhibited relatively mild effects on cell viability and proliferative genes, while strongly inhibited Hmgcs1 expression and cell metastatic gene programs. Furthermore, we established xenograft models using nude mice implanted with human A549 non-small-cell lung cancer (NSCLC) cells and treated them with magnolol or control to evaluate the effects of magnolol. The results demonstrated that magnolol exhibited significant inhibitory effects on lung adenocarcinoma colonization, suggesting its clinical translation potential (S8A–S8D Fig).

Taken together, our study suggested that lung adenocarcinoma cells responded to exogenous cholesterol with a unique positive feedback mechanism, which activated GLI2-FOXA3-HMGCS1 transcriptional axis to enhance de novo cholesterol biosynthesis and cellular cholesterol accretion, eventually promoted lung adenocarcinoma progression at least partially by promoting membrane composition change, while the FOXA3 inhibitor magnolol blocked this signaling cascade and prevented cancer metastasis (Fig 7J).

Ethics statement

All animal studies were carried out following the guidelines approved by the Ethics Committee of Animal Experiments of East China Normal University (m20210254). Tumor and adjacent normal lung tissue were obtained and used for research purposes with approval from the Medical Ethical Council of Shanghai Chest Hospital. The written informed consent was obtained from all patients. The study abides by the Declaration of Helsinki principles.

Cell culture and reagents

Beas-2B, HPAEpiC, A549, H441, H1299, HepG2, Huh7, and HEK293T cells were original purchased from the American Type Culture Collection (Manassas, Virginia, United States of America). LLC, HEK293T, HepG2, and Huh7 cells were cultured in DMEM, and Beas-2B, HPAEpiC, A549, H441, H1299 were cultured in RPMI 1640, supplemented with 10% FBS, 100 U/ml penicillin and 100 μg/ml streptomycin at 37°C in a humidified atmosphere with 5% CO2 except specially indicated. All cells were tested for mycoplasma contamination periodically. DMEM, RPMI1640, FBS, 100 U/ml penicillin and 100 μg/ml streptomycin, and 0.25% trypsin were purchased from Thermo Scientific (Gibco, Grand Island, New York, USA). Lentivirus system plasmids pCDH-CMV-MCS-EF1-Puro, pLKO.1-TRC, psPAX2, and pMD2.G were purchased from Addgene (Cambridge, Massachusetts, USA). Cisplatin (HY-17394), Fluvastatin (HY-14664), and Magnolol (HY-N0163) were purchased from MCE (Shanghai, China). Cholesterol (C3045), MβCD (C4555), and Filipin III (SAE0087) were purchased from Sigma (St. Louis, Missouri, USA). Human DiI-Low Density Lipoprotein (DiI-LDL) (20614ES76) was purchased from Yeason (Shanghai, China) and Cholesterol-biotin was purchased from Biobond (Biobond, Shanghai, China).

Plasmids and shRNA lentiviral constructions

FLAG-tagged hFoxa3 and hHmgcs1 were generated by the ligation of the full-length open-reading frame into the pCDH-CMV-puro vector (Addgene, 167463). Myc-tagged GLI2, myc-tagged GLI1, and Myc-tagged GLI3 were generated by the ligation of the full-length open-reading frame into the pcDNA3.1 vector (Addgene, 182494). Plasmids for knockdown of Foxa3 or hmgcs1 were constructed into pLKO.1 vectors (Addgene, 10879), with a scrambled small hairpin RNA sequence (Addgene, 1864) serving as the control. The sequences for Foxa3 or Hmgcs1 knockdown are listed in S3 Table. psPAX2 (12260, Addgene) and pMD2.G (12259, Addgene) were used to prepared lentivirus. High-titer lentivirus was packaged in HEK293T cells using EZ Trans (Life iLab, Shanghai, China). Viral supernatants were collected 72 h after transfection, and the lentivirus was concentrated by ultracentrifugation, 0.45 μm filtered and applied to cancer cells in the presence of 1 μg/ml polybrene for 24 h. Cells were selected with 1 μg/ml puromycin for 10 days. Gene expression levels were confirmed by qPCR and immunoblotting.

Mice and tumor models

All mice were maintained under a standard humidity- and temperature-controlled environment on a 12-h light/dark cycle, with free access to food and water. All mice in our experiments were gender and age matched. For lung metastasis models, 6-week-old male BalB/C nude mice were purchased from Shanghai Research Center for Model organisms, and 5 × 10⁶ luciferase-labeled A549 cells were injected into mice via tail vein. The tumors were detected by luciferase-based noninvasive bioluminescence imaging using the platform of IVIS Lumina II (PerkinElmer, Massachusetts, USA), and 6 to 8 weeks later, mice were euthanized and lung metastasis were measured by counting lung surface tumor number and by HE staining of lung paraffin sections. For magnolol treatment, magnolol (15 mg/kg) was administrated subcutaneously every other day for about 6 weeks. Wild-type (WT, Foxa3^+/+) and FOXA3-null (Foxa3^-/-) mice have been previously described [23]. For urethane-induced lung cancer, 8-week-old male Foxa3^+/+ and Foxa3^-/- mice were injected intraperitoneally with 1 g/kg urethane (ethyl carbamate; Sigma, Missouri, USA) 1 injection per week for 10 weeks. Mice were monitored once per week, and the lung and liver tissues of mice were collected 15 weeks after urethane treatment. The number of lung tumor bearing mice in each group was observed and counted and the tumor size was recorded by the diameter.

For LLC spontaneous lung metastatic models, to initiate the study, 1 × 106 LLC cells with FOXA3 overexpression, HMGCS1 overexpression, or HMGCS1 knockdown were suspended in 200 μl of phosphate-buffered saline (PBS) and subcutaneously injected into the right flank of mice. Upon reaching a size of 100 mm² (about 3 weeks), the tumors were surgically excised. Subsequently, 20 to 26 days post-tumor removal, lung metastasis was evaluated through HE staining on paraffin sections of lung tissue.

RNA extraction and real-time PCR

Total RNA was extracted using RNAiso Plus (Takara, Beijing, China) according to the manufacturer’s instructions. The mRNA was reverse transcribed using PrimeScript RT Master Mix (Takara, Beijing, China). Real-time PCR was performed using SYBR Green qPCR Master Mix with Roche 480 system. The relative expression levels were measured using the relative quantitation ΔΔCt method and normalized to gapdh expression. The primers used for real-time PCR are listed in S3 Table.

Immunoblotting

Cell lysates were prepared and immunoblotings were performed as previously described [66]. Briefly, tissues were homogenated and cells were lysed in RIPA lysis buffer. Protein samples were electrophoresed on SDS-PAGE and transferred onto nitrocellulose membranes. The membranes were incubated with appropriate primary antibodies overnight at 4°C, washed with TBST and incubated with IRDye secondary antibodies (926–68071 and 926–322210) from LI-COR Biosciences (LI-COR biosciences, Nebraska, USA). The protein signals were detected by Odyssey CLX imaging system (LI-COR biosciences, Nebraska, USA). α-tubulin was used as a loading control. Antibodies used for immunoblotting are listed in S4 Table.

Immunohistochemistry

Dehydrated paraffin sections were pretreated with 3% H2O2 for 30 min to quench endogenous peroxidase activity and then blocked in PBS containing 2.5% horse serum. FOXA3 staining were performed by incubating with mouse anti-FOXA3 (sc-74424, Santa Cruz) in a humidified chamber at 4°C overnight. Sections were washed with PBS and incubated with IgG-HRP (Huabio, Hangzhou, China) secondary antibody at room temperature. The signal was detected by DAB kit (Servicebio, Wuhan, China). Images were acquired using an Olympus microscope and quantitative analysis was conducted with ImageJ.

Immunofluorescence

A549 or HepG2 were cultured on coverslips in 24-well plates for 24 h, fasted for 12 h, and treated with or without 2.5 mM MβCD for 30 min, cells were washed thoroughly with RPMI1640 medium twice, then switched to fresh RPMI1640 medium supplemented with 5% lipoprotein-deficient serum (LPDS medium) [66] and treated with 5 μg/ml Cholesterol-Biotin for 24 h. Cells were then fixed, permeabilized, and incubated with primary antibodies against biotin (Abclonal, A18848) overnight at 4°C. Cells were then washed and incubated with Cy3 Goat Anti-Rabbit (Abclonal, AB_2769089). Dil-LDL uptake was performed as previously described [67]. Briefly, the cells were put in LPDS medium at 4°C for 30 min, then switched to fresh LPDS medium supplemented with 20 μg/ml Dil-LDL at 4°C for 1 h. Cells were then incubated at 37°C with fresh LPDS medium and fixed with 4% paraformaldehyde for visualization. Filipin III (Sigma, SAE0087) staining was performed according to manufacturer’s protocol. Briefly, cells were fixed with 4% paraformaldehyde (PFA) and stained with 50 μg/ml filipin III for 30 min at 4°C for visualization. DAPI was used to visualize the nuclei. Images were collected using a Leica SP8 confocal microscope and analyzed using a Leica LAS AF software. Quantitative analysis was conducted with ImageJ.

Cell migration and invasion

The assay was performed using Transwell migration chambers (8 mm pore size; Corning Costar, New York, USA) according to the instructions as previously described [67]. For invasion assays, the chambers were precoated with matrigel (356231, Corning, New York, USA) according to the instructions. For both assays, 1 × 10⁵ cells were suspended with serum-free culture medium and seeded into the upper well of the chamber, the lower well was filled with normal culture medium containing 10% FBS functions as a chemoattractant. Cells were cultured for 16 h in migration assay and 24 h in invasion assay, the non-migrated or noninvasive cells were scratched from the upper surface of the transwell. Cells on the lower surface of the Insert chambers were fixed with neutral formalin, stained with hemotoxylin/eosin, and counted under Nikon digital camera (Tokyo, Japan) at ×200 magnification.

Detection of circulating tumor cells

The relative circulating tumor cells (CTCs) was examined by detecting the ratio of hGapdh expression to mGapdh expression by real-time PCR. The primers were listed in S3 Table.

Total cholesterol, dehydrocholesterol, and HMG-CoA detection

For cellular assays evaluating intracellular cholesterol, dehydrocholesterol, and HMG-CoA levels, cells were cultured in LPDS medium (culture medium plus 5% LPDS) before indicated manipulation. Cholesterol levels were examined with a total cholesterol assay kit (Sigma), and 7-dehydrocholesterol, 24-dehydrocholesterol, and HMG-CoA levels were detected with ELISA kits (Zeye Bio and Milbio, Shanghai, China) following the manufacturer’s instructions.

Drug screening and luciferase reporter assays

A collection of 850 Natural compounds was obtained from targetmol (Wellesley Hills, Massachusetts, USA). Foxa3 or control reporter plasmids with pRL-SV40 (Renilla luciferase control reporter vector) were co-transfected into HEK293T cells by lipofectamine 2000 (Invitrogen, New York, USA). Cells were plated into 96-well plates and were added each of 850 Natural compounds at 10 μm. Luciferase activity was measured using the Dual-Luciferase Reporter Assay System (Promega).

Lung cancer clinical specimens and PDO

PDOs were established by BioGenous as previously reported [68]. Briefly, tissue derived from surgical resections was cut into small pieces and enzymatically digested using 800 U/ml collagenase I (GIBCO) and 10 mM Y-27632 (Sigma-Aldrich) before embedding in Matrigel (Corning). After Matrigel solidification for 15 min at 37°C, cells were overlaid with human normal airway or NSCLC organoid medium. Human normal airway and NSCLC organoid medium is composed of 1× Advanced DMEM/F12 (GIBCO), 1× Glutamax (GIBCO), 100/100 U/ml Penicillin/Streptomycin (GIBCO), 10 mM HEPES (GIBCO), 1× B27 supplement (GIBCO), 1.25 mM N-Acetylcystein (Sigma-Aldrich), 10 mM nicotinamide (Sigma-Aldrich), 500 ng/ml human recombinant R-spondin1 (OrganRegen), 100 ng/ml human recombinant Noggin (OrganRegen), 25 ng/ml human recombinant FGF-7 (OrganRegen), 100 ng/ml human recombinant FGF-10 (OrganRegen), 500 nM A83-01 (Selleck), 500 nM SB202190 (Selleck), 5 mM Y-27632 (Selleck), and 1× Puromycin (Invitrogen) was added to prevent microbial contamination. Organoids were passaged approximately every week by incubating in TrypLE Express (GIBCO) for 2 to 5 min at 37°C to dissociate organoids and replating in fresh Matrigel. Organoids were cryopreserved in Organoid Cryopreservation Medium (bioGenous) as master and working biobanks. Organoids < passage 20 were used in experiments.

RNA-seq preparation and RNA-seq data analysis

RNA-seq preparation and RNA-seq data analysis were performed as previously reported [69]. Briefly, total RNA was isolated using the standard RNAiso Plus Protocol and RNA quality was examined. RNA samples were purified with magnetic oligo(dT) beads after denaturation and were reverse transcribed into first-strand cDNA. Fragmented DNA samples were end blunted and adenylated at the 3′-end. Adaptors were ligated to construct a library. After DNA quantified by Qubit (Invitrogen, New York, USA) and cBot cluster generation, DNA samples were then sequenced by an Illumina HiSeqX Ten SBS instrument (Genergy Bio, Shanghai, China). Raw data were converted into Fastq format. The number of transcripts of each sample was calculated based on the number of fragments per kilobase of transcript per million fragments mapped (FPKM). The FPKM value was calculated by Cuffnorm software for each sample, and the values were log2 transformed. The differential gene transcripts between different samples were calculated using DESeq software. The thresholds for determining differentially expression transcripts are P < 0.05 and fold change≥2. Processed RNA-seq data were filtered by removing genes with low read counts. Read counts were normalized using TMM normalization and CPM (counts per million) were calculated to create a matrix of normalized expression values. Selected gene sets were annotated by GO analysis by DAVID database (DAVID 6.8) [66]. Heatmap was generated using heatmapper website (http://www.heatmapper.ca/expression/). The raw data of RNA-seq were uploaded into the NCBI Sequencing Read Archive (SRA) under accession number PRJNA845911.

Membrane and cytosol components extraction

A549 cells were treated with 5 μg/ml cholesterol in RPMI1640 containing 5% LPDS medium for 24 h, the cytosol and membrane components were obtained using Membrane and Cytosol Protein Extraction Kit (Beyotime, Shanghai, China) according to the manufacturer’s protocol and stored at −80°C for further analysis.

LC-MS analyses for cholesterol and its precursors

¹³C-acetate was used as a tracer to evaluate the impact of exogeneous cholesterol on de novo cholesterol synthesis. A549 cells were maintained in RPMI1640 containing 5% LPDS (Invitrogen). Prior to incubation with cholesterol or ¹³C tracers, A549 cells were fasted for 16 h, then A549 cells were washed with PBS and incubated with RPMI1640 containing 5% LPDS (Invitrogen) and 5 mM ¹³C-acetate (Sigma), and cells were treated with or without cholesterol for 24 h. Then, the cells were collected and stored at −80°C for further analysis.

Cholesterol and its precursors, dehydrocholesterol in cells or cellular components were extracted with chloroform/methanol mix solution and analyzed using an I-Class UHPLC coupled to a Q-TOF mass spectrometer (Waters Corp., Milford, Massachusetts, USA) under positive ESI mode. A Q-TOF mass analyzer was operated at 22,000 mass resolution with the scan range from 50 to 1,500 m/z. The data were analyzed by MassLynx v4.2 software (Waters Corp., Milford, Massachusetts, USA) and normalized with internal standard and protein concentration measured by the BCA protein assay kit.

ChIP and ChIP-seq

ChIP was performed using a SimpleChIP Plus Enzymatic Chromatin IP Kit (9003, Cell signaling) according to manufacturer’s instructions. Briefly, Flag-tagged FOXA3 overexpressed H1299 lung cancer cells were cross-linked by 1% formaldehyde for 10 min and stopped by 0.125 M glycine solution. Cells were then washed twice with ice-cold PBS and scraped into PBS, the extracted chromatin was digested and fragmented into 200 to 1,000 bp by the Bioruptor (Scientz–IID; Scientz) in SDS lysis buffer, and then immunoprecipitated by ChIP grade antibodies against Flag (F1804, Sigma) or a normal IgG (12–371, Sigma) using Magnetic Beads. After that, the DNA was uncross-linked with proteins. PCR was then performed using the primers of the putative FOXA3 binding site. Primers were listed in S3 Table.

For ChIP-seq assay, a total amount of 50 ng DNA per sample was used to generate libraries using the NEBNext Ultra II Library Prep Kit for Illumina (NEB), reactions were prepared according to the manufacturer’s manual. Libraries were sequenced 50 bp single-end on an Illumina HiSeq2500 platform (Illumina). Alignments were filtered by Skewer to exclude reads with low quality, adapter and short (less than 50 bp) sequence fragments. For data analysis, Fastqc software was used to carry out the quality of the processed sequences. Then, the filtered sequencing reads were mapped to the human reference sequence for the human genome using Bowtie2 (v.2.2.5), the original sequence of the sample was compared with the reference genome for unique mapping analysis and filtered the sequences that can align to multiple (>1) genome locations. After unique alignment, duplicated reads were removed.

For peak calling, MACS2 call peak (v2.1.1) was used for the enrichment regions of reads on the genome for each ChIP replicates and Input as background separately, using p value <0.01. The raw data of ChIP-seq were uploaded into the NCBI SRA under accession number PRJNA845827. The regarding binding genes were listed in S1 Table.

Cell viability

For cell lines, the assay was conducted using an Enhanced Cell Counting Kit-8 (CCK8, C0042, Beyotime) following the manufacturer’s instructions. Briefly, 5,000 cells per well in 100 μl were seeded in 96-well plates. After 24 h, the cells were treated with the magnolol at the indicated concentrations for an additional 24 h. Subsequently, CCK8 were added to each well and incubated for 0.5 to 2 h for further analysis.

For PDO cell viability assessment, we employed the ATP cell viability assay (Biogenous E238003) following the manuscript instructions. Briefly, PDO was added with an equivalent volume of the assay reagent to the culture plate, following with vigorously shake for 5 min and incubation at room temperature for 20 min. Analysis of the relative vitality of the organoids was based on the obtained chemiluminescence reading at a wavelength of 560 nm.

Online data mining

Clinical data of Hmgcs1 and Gli2 expression analysis for lung patients were downloaded from the Kaplan–Meier Plotter (https://kmplot.com/analysis). Clinical data of Foxa3 expression analysis for lung patients were downloaded from GEPIA (http://gepia.cancer-pku.cn/) [70] or Oncomine (https://www.oncomine.org/). In the analysis process, gene expression data and relapse free and overall survival information are downloaded from TCGA and the GTEx projects, using a standard processing pipeline. The database is handled by a PostgreSQL server, and gene expression and clinical data were integrated simultaneously. The survival plot and logrank P value are calculated and plotted.

Statistical analysis

Data analyses were performed by Graph-Pad Prism 8 (GraphPad Software). Unpaired Student’s t test was used for comparison between 2 groups. Log-rank (Mantel–Cox) test was used to analyze survival curve. The data are shown as means ± SEM. p < 0.05 was considered as statistically significant, *p < 0.05, **p < 0.01.