Melatonin inhibits lipid accumulation to repress prostate cancer progression by mediating the epigenetic modification of CES1

2.1 MLT reduced lipid accumulation in PCa

To investigate whether melatonin is associated with the risk of PCa, a systematic review and meta-analysis were performed according to a flowchart (Figure S1A). Men with lower MLT levels had increased risk of PCa (Table S1), and low melatonin levels increased the incidence of advanced PCa (Figure 1A).

To investigate the influence of circadian rhythm disorder on PCa, we used mouse-derived prostate Rm-1 cells to generate a PCa model with C57BL/6 mice. The levels of MLT in mice housed and exposed to LAN were significantly lower than those in the normal light rhythm group, and supplementation with MLT in drinking water increased the level of melatonin in the blood (Figure 1B), indicating that LAN-induced circadian rhythm disorder decreased the MLT levels in these mice. Rm-1 PCa tumors in the mice exposed to LAN showed significantly increased size, growth rate, and weight, compared to those of the tumors from the normal light rhythm group and from the LAN group receiving supplemental MLT, indicating that the LAN-induced decrease in the level of MLT stimulated PCa growth in the mice. Comparison of the control group with the ENZ treatment group exposed to normal light or LAN indicated that the Rm-1 xenograft tumors were not sensitive to ENZ, suggesting that Rm-1 is a mouse-derived CRPC cell line (Figures 1C and 1D). The result of the real-time PCR analysis indicated positive expression of AR-V7 in the Rm-1 tumors (Figure S1C). The results of IHC and Tunel assays indicated that circadian rhythm disorder promoted cell proliferation and inhibited the apoptosis rate, while MLT supplementation had the opposite effect (Figure S1D). Moreover, comparison of the ENZ-alone treatment group with ENZ + MLT treatment group indicated that MLT treatment increases the sensitivity of Rm-1 tumors to ENZ regardless of exposure to normal light rhythm or LAN (Figures 1C and 1D).

MLT is a potent antioxidant and a key regulator of energy metabolism, especially lipid metabolism.^21-23 Several studies have demonstrated that the occurrence and progression of PCa, especially CRPC, are significantly related to the lipid metabolism.¹⁰ The meta-analysis of the association of melatonin with lipid levels demonstrated that MLT is negatively correlated with TG, T-CHO, and LDL and positively correlated with HDL in the blood (Figures 1E, S1B, and Table S2). To investigate the influence on lipids, we assayed the TG and T-CHO contents in the blood and tumor tissues of the groups. The results indicated that the TG and T-CHO levels were higher in the LAN-exposed group compared to those in the normal light rhythm group and lower than those in the MLT treatment group (Figure 1F). Similar results were shown by BODIPY staining (Figure 1G). Overall, these results indicate that MLT reduced lipid accumulation in PCa.

2.2 MLT upregulated the expression of the lipid metabolism-related gene CES1 in PCa cells

To determine the regulatory mechanisms of MLT in PCa, we performed RNA sequencing with C4-2 cells treated with MLT (1 mM). A comparison of treated C4-2 cells with control cells led to the identification of 1333 differently expressed genes (DEG), including 614 upregulated and 719 downregulated genes (|logFC| > 1, p < 0.05) in the C4-2 cells treated with MLT (Figures 2A and S2A). Subsequently, GO analysis showed that MLT treatment mediated most of changes in lipid metabolism pathways (Figure 2B). In addition, the result of GO analysis in Figure 2B also suggested that MLT could have an influence on cytokine-cytokine receptor interaction and transcriptional misregulation pathways in PCa.

Lipids are mainly stored in the cells as lipid droplets. A BODIPY assay was carried out to detect the lipid droplets, and the results demonstrated that treatment with MLT decreased the fluorescence in the C4-2 and 22RV1 cells compared to that in the corresponding control cells, indicating that MLT inhibits lipid deposition in PCa cells (Figure 2C). The melatonin receptor has two subtypes, MT1 and MT2; MT1 is predominantly expressed in prostate tissue, and MT2 is negligibly expressed. To investigate whether the effect of MLT on lipid accumulation in PCa cells depends on MT1, we generated an siRNA for MT1 (Figure S2B); the knockdown of MT1 eliminated the effect of melatonin on the depletion of lipid droplets in C4-2 and 22RV1 cells (Figure 2C). As quantitative indicators of lipid accumulation, the cholesterol and TG contents were detected, and the results demonstrated melatonin inhibits lipid accumulation, and the knockdown of MT1 eliminated the effect of melatonin on PCa cells (Figure 2D). Then, the results of the LC/MS assay demonstrated that treatment with MLT downregulated the components in lipid droplets, TG, PE, SM, and PC (Figures 2E and S2C).

To investigate the mechanism of lipid metabolism changes induced by melatonin in PCa, lipid-related genes were identified in the top 30 DEGs based on the RNA sequencing results; three genes (DHRSX,²⁷ HNF4A,²⁸ and CES1²⁹) were upregulated, and six genes (SLC7A10,³⁰ IGFBP1,³¹ ALDH1A2,³² ACSBG2,³³ ZIC1,³⁴ and SLC38A4³⁵) were downregulated in the C4-2 cells treated with MLT compared with the control cells (Figure 2A). Besides, the top 30 DEGs included other functional genes, for example, CCL20³⁶ and MILR1, ³⁷ are related to tumor microenvironment and inflammation. The results of the Western blot assay of the levels of proteins encoded by these lipid-related genes indicated that compared with the control cells, CES1 was the most significantly and consistently changed in the C4-2 and 22RV1 cells treated with MLT (Figure S2D). Finally, we demonstrated that MLT promoted the CES1 expression in a dose-dependent and an MT1-dependent manner (Figures 2F and S2E).

2.3 CES1 was downregulated and predicted poor prognosis in PCa

To investigate the effect of CES1 on PCa, we first analyzed CES1 expression in the independent PCa datasets (The Cancer Genome Atlas [TCGA], Taylor). The expression of CES1 was downregulated in PCa tissues compared to normal prostate tissues, and the high expression of CES1 is negatively correlated with tumor stage, metastasis, and Gleason score (Figure 3A). Receiver operating characteristic (ROC) curve showed that CES1 could be used as a diagnostic biomarker in PCa (Figure S3A). Multivariate analyses were carried out and demonstrating that CES1 was an independent prognostic marker in PCa (Figure S3B). Immunohistochemical (IHC) staining was carried out to determine the CES1 protein expression in the PCa specimens, and we found that the protein expression of CES1 was significantly lower in the PCa tissues compared with that in the normal prostate tissues, and high expression was negatively related to Gleason score and tumor stage (Figure 3B), indicating that CES1 is a key gene in PCa. Rabbit IgG (R-IgG) as negative control (Figure S1F). The protein levels of CES1 in the PCa cell lines were lower than the level in the normal RWPE-1 prostate epithelial cell line (Figure 3C). Furthermore, Kaplan-Meier analysis of the data obtained from Taylor and TCGA databases indicated that patients with PCa who express low levels of CES1 (above the 50th percentile) had shorter BRFS (p = 0.0068) and DFS (p = 0.0282) than patients with PCa who express high levels of CES1 (below the 50th percentile) (Figures 3D and S3C). Therefore, we hypothesized that CES1 may be a tumor suppressor gene in PCa.

To prove our hypothesis that CES1 is a key gene in PCa, gene set enrichment analysis (GSEA) was executed to look for relative gene sets with CES1 expression in the PCa TCGA dataset. The results showed that pathways related to cell cycle and metastasis and pathways related to lipid accumulation and metabolism are associated with CES1 expression (Figures 3E and S3D). Then, using sgRNA (CRISPR-Cas9) or lentivirus, we acquired 22RV1 and C4-2 cell lines with knocked out or overexpressed CES1 (Figure S3E). Transwell and CCK-8 assays were executed to examine whether CES1 is involved in the regulation of tumor growth and metastasis of PCa cells. C4-2 and 22RV1 cells with stable overexpression of CES1 had significantly inhibited proliferation, migration, and invasion (Figures 3F-3H).

2.4 MLT decreased lipid accumulation and cell activity in PCa cells by upregulating CES1 expression

To investigate whether MLT affects PCa progression by CES1, 22RV1 and C4-2 cells with CES1 knockout expression were treated with MLT. The results of Western blot indicated that MLT promotes the expression of CES1, and MLT negligibly promotes CES1 expression after inhibition of endogenous CES1 (Figure 4A). The results of CCK-8 and Transwell assays suggested that CES1 knockout can almost completely abolish the repressions of MLT on the proliferation, invasion, and migration of C4-2 and 22RV1 cells (Figures 4B, 4C, and S3F). Compared with the control cells, the cells treated with MLT showed lower fluorescence intensity, while the fluorescence intensity in the CES1-knockout PCa cells with/without MLT treatment was similar, suggesting that CES1 knockout significantly abrogated the effects of MLT on the elimination of lipid deposits in PCa cells (Figure 4D). A similar result was found by TG and T-CHO content assays (Figure 4E). These results indicate that MLT inhibits lipid accumulation and cell activity by upregulating CES1.

To investigate the mechanism of the effect of CES1 on lipid metabolism, a GSEA was performed; the results indicated that CES1 is positively related to the PPAR and ER lumen signaling pathway in the PCa TCGA database samples (Figure S3G). Previous studies have demonstrated that activation of the PPAR pathway leads to lipid degradation,³⁸ and lipid droplet depletion increases ER stress and inhibits cell activity by mediating apoptosis.³⁹ Combined RNA sequencing data indicated that treatment with MLT increases the expression of the PPARa pathway-related genes CPT1C⁴⁰ and ME1⁴¹ (Figure S3I). We hypothesized that MLT/CES1 mediates lipid depletion and apoptosis through the PPARa-ER stress pathway. Our previous studies showed that melatonin reduced lipid drops accumulation in PCa cells by increasing the expression of the lipid metabolism-related gene CES1. And a study by Bo Qiu et al from the University of Pennsylvania demonstrated that reduction of lipid drops induced ER stress through activation of UPR sensors PERK, IRE-1α, and ATF6 signaling pathways.³⁹ Therefore, we performed western blot assay to detect these signaling pathways. The results of Western blot demonstrated that MLT promotes the expression of PPARa, and the levels of phosphorylation of PERK, phosphorylation of eIF2a, and phosphorylation of IRE1a in MLT-treatment cells are higher than the control cells, and ATF6 was significantly elevated in the nucleus after MLT treatment, while knocking out CES1 eliminated these effects (Figure 4F). These results indicated that MLT treatment activated the three pathways (PERK, IRE1a, and ATF6), which are involved in the UPR response.

Overall, these results indicate that MLT inhibits lipid accumulation and cell activity by regulating CES1 expression in PCa.

2.5 MLT inhibited CRPC progression and reversed enzalutamide resistance

Several studies have proved that the occurrence and progression of PCa, especially CRPC, are significantly related to the lipid metabolism.¹⁰ Higher levels of T-CHO, TG, and LDL increase the risk for PCa, and lower HDL level is associated with a lower risk for PCa. The systematic review and meta-analysis were carried out to demonstrate that high cholesterol levels are related to high incidence of advanced PCa (Figure 5A). These results indicated that cholesterol has a closely relationship with PCa progression. Cholesterol, which is stored in LDs in the form of cholesterol esters, is the raw material for androgen synthesis. Moreover, increased androgen synthesis in tumor cells is an important factor in the progression of PCa to CRPC and enzalutamide resistance. Since MLT-mediated CES1 expression inhibits lipid accumulation in PCa cells and decreases intracellular cholesterol content (Figures 2D and 4E), we hypothesized that MLT/CES1 reduces cholesterol, leading to the interference of PCa progression.

To test this hypothesis, we initially performed a GSEA with TCGA PCa datasets and found that CES1 is associated with a set of positively regulated cholesterol efflux genes and a set of metabolism-related steroid hormones genes (Figure S3J). Then, CCK-8 assays were performed to assess the effects of cholesterol and/or melatonin on PCa cells at various stages. LNCaP cells are sensitive to androgen; C4-2 cells constitute a CRPC cell line derived from LNCaP cells and express functional endogenous AR⁴²; 22RV1 cells constitute a CRPC and ENZ-resistant cell line that expresses AR-V7.⁴³ Cholesterol treatment promoted the proliferation of PCa cells, and the LNCaP cells cultured without androgen were more sensitive to cholesterol treatment than the CRPC cell lines (C4-2 and 22RV1 cells) suggesting that the effect of cholesterol on PCa progression is associated with androgen; MLT inhibited the proliferation of these cells, and cholesterol supplementation partially reversed MLT-mediated inhibition (Figure 5B).

Moreover, we generated a C4-2-ENZR cell model (C4-2-enzalutamide resistance cell) to test whether MLT/CES1 influences enzalutamide resistance. We found that the expression of CES1 in the C4-2-ENZR cells is lower than that in the C4-2 cells (Figure 5C); MLT treatment had the opposite effect (Figure S3K). MLT or overexpressed CES1 reduced the cholesterol and TG contents in C4-2-ENZR cells (Figures 5E and S3H). Considering the RNA sequencing data, we determined that treatment with MLT significantly decreased the expression levels of androgen synthesis-related genes CYP11A1⁴⁴ and STARD4⁴⁵ (Figure S3I). The Western blot data demonstrated that the levels of the proteins encoded by the androgen synthesis-related genes CYP11A1 and STARD4 were significantly lower in the MLT-treated C4-2-ENZR and 22RV1 cells than those in the control cells (Figure 5E), indicating that MLT influences intratumoral androgen synthesis in enzalutamide-resistant CRPC. Thus, we performed an LC/MS (liquid chromatography-mass spectrometry) analysis to evaluate the levels of testosterone (T) and dihydrotestosterone (DHT) in tumor cells. The results showed that MLT seriously decreased the contents of T and DHT in the C4-2-ENZR cells and had a negligible effect in the absence of endogenous CES1, demonstrating that MLT inhibited intratumoral androgen synthesis though CES1 (Figure 5F).

Finally, we conducted CCK-8 assays and found that the C4-2-ENZR and 22RV1 cells were resistant to ENZ treatment and that treated-MLT restored the sensitivity of ENZ treatment in 22RV1 and C4-2-ENZR cells (Figure 5G).

2.6 MLT-mediated epigenetic modification upregulated CES1 expression

Abnormal DNA methylation is an important epigenetic alteration in cancer. In tumors, 5′- UTRs and CpG islands (CGIs) are hypermethylated relative to prostate tissue.⁴⁶ The results of the analysis for the CES1 genomic DNA sequence in the 2-kilobase promoter regions indicated that there are CGIs in these regions (Figure 6A). Besides, we used the Cancer Cell Line Encyclopedia database to evaluate the methylation status of the CES1 promoter in several PCa cell lines and found that the methylated CES1 promoter is relatively enriched in CRPC cells (Figure 6B). The results of bisulfite sequencing PCR (BSP) indicated there were noticeable hypermethylation of the CES1 promoter in PCa cells compared with that in normal prostate epithelial cells, while similar result in PCa tissues compared with that in paired normal prostate tissues (Figures 6C and S4A).

Considering the downregulation of CES1 in PCa cells and tissues, the results indicate that DNA methylation may be the key regulatory factor of CES1 expression. To determine the mechanism by which CES1 expression is upregulated in PCa by MLT treatment, BSP was performed to assess the changes in the methylation level of the CES1 promoter after MLT treatment. As shown in Figure 6D, MLT treatment reduced CGIs methylation of the CES1 promoter. The results of the luciferase reporter assays showed that the luciferase activity of the cells transfected with a vector containing the CES1 promoter was promoted by the MLT treatment of the C4-2 and 22RV1 cells in a dose-dependent manner (Figure 6E). Overall, these results indicate that MLT-mediated demethylation upregulates CES1 expression.

2.7 MLT regulated methylation of CES1 promoter though SIRT1-mediated DNMT1 deacetylation

In mammals, genomic methylation status is established by DNA methyltransferases (DNMTs), including DNMT1, DNMT2, DNMT3a, DNMT3b, and DNMT3L.⁴⁷ Western blot was carried out to determine whether MLT influences DNA methylation of the CES1 promoter via DNMTs, and the results indicated that the level of DNMT1 was significantly and consistently decreased in a dose-dependent manner, in the 22RV1 and C4-2 cells treated with MLT, compared to the control cells (Figure 7A). Luciferase reporter assays demonstrated that the luciferase activity of the cells transfected with a vector containing the CES1 promoter was inhibited by DNMT1 overexpression in 293T and 22RV1 cells (Figure 7B). Note that 5-azacytidine (5-AzaC), an inhibitor targeting DNMT1, upregulated CES1 protein expression in a dose-dependent manner (Figure 7C). Moreover, MLT decreased the activity of DNMT1 in a dose-dependent manner (Figure 7D). Then, chromatin immunoprecipitation (ChIP) assays were executed. A strong enrichment between the CGIs of CES1 promoter and anti-DNMT1 antibody was observed in the 22RV1 and C4-2 cells, while MLT decreased this enrichment in dose-dependent (Figures 7E and S4B). These results indicate that MLT suppresses DNA methylation of CES1 by inhibiting DNMT1 binding to CGIs.

Several studies have demonstrated that SIRT1 deacetylates the DNMT1 protein and alters its activity.⁴⁸ Numerous findings have reported an increase in the activity of SIRT1 after MLT treatment.⁴⁹ We hypothesized that MLT regulates DNMT1 by SIRT1-mediated deacetylation. The results of the Western blot showed that MLT promoted SIRT expression and decreased the DNMT1 protein level in an MT1-dependent manner (Figure 7F). Moreover, LAN-mediated circadian rhythm disorder downregulated the expression of Ces1 and Sirt1, upregulated the expression of Dnmt1, while MLT supplementation eliminated the effects in the mouse tumor tissues (Figure 7G). We constructed siRNA for SIRT1 (Figure S4C). The results of the coimmunoprecipitation (co-IP) assay showed that MLT treatment increased SIRT1 binding to DNMT1 and induced DNMT1 deacetylation in dose-dependent manner, and SIRT1 knockout eliminated the effect of melatonin on the DNMT1 expression (Figures 7H, S4D, and S4E).

Overall, these results indicate that MLT reduces the methylation level of the CES1 promoter by upregulating SIRT-mediated DNMT1 deacetylation.

2.8 MLT inhibited tumor growth and reversed enzalutamide resistance of CPRC by mediating CES1 expression in vivo

To study the role of MLT/CES1 in vivo, C4-2 cells were lentivirally transduced to knockout CES1 expression, implanted subcutaneously in NCG mice. C4-2-cell xenograft tumor treated with MLT (200 mg/kg, intragastric administration) was significantly smaller with a reduced growth rate, and weight compared with tumors treated with a vehicle (EtOH), and C4-2 tumor with stably knocked out CES1 expression showed an opposite pattern of tumor growth compared with that of the vector-transduced tumor. Knockout of CES1 reversed the effect of MLT on tumor growth (Figures 8A and 8B). Notably, BODIPY staining revealed sparse lipid deposition in the MLT-treated tumor compared to that in the control tumor; knockout of CES1 expression increased lipid accumulation and reversed the effect of melatonin (Figure 8C). Moreover, the results of the IHC and IF assays indicated that melatonin upregulated the levels of CES1, PPARa, ER stress markers, and caspase-3-mediated apoptosis, and CES1 knockout abolished these MLT-induced effects, R-IgG as negative control (Figures 8D and S5A). Overall, these results indicate that MLT inhibits tumor growth by upregulating the CES1 expression in vivo.

Considering these results, we studied the MLT-mediated reversal of enzalutamide resistance in vivo. C4-2-ENZR cells were implanted subcutaneously in NCG mice. The results indicate that tumors treated with MLT alone reduced tumor size, growth rate, and weight compared with those in the control group; however, the size, growth rate, and weight of the tumors treated with ENZ alone have no significant differences from those in the control group; those of the tumors in mice fed MLT and ENZ were the most obviously decreased (Figures 8E and 8F), indicating that MLT increased the sensitivity of enzalutamide-resistant CRPC to ENZ. BODIPY staining and IF and IHC assays were performed with the xenograft tumors. MLT treatment decreased lipid accumulation, upregulated the levels of CES1, inhibited cell proliferation, decreased androgen synthesis, and increased the apoptosis rate (Figures 8G, 8H, and S5B). Moreover, the results of IHC demonstrated that LAN-induced circadian rhythm disorder downregulated the expression of Ces1, Ppara, Perk, caspase-3, Cyp11a1, and Psa, R-IgG as negative control. MLT treatment abolished these effects in the PCa tissues of the C57BL/6 mice (Figures S6A and S6B). Thus, the data indicate that MLT reverses enzalutamide resistance of CPRC.

2.9 Ces1-knockout mice confirmed the role of endogenous Ces1

Ces1-knockout mice (Ces1^−/−) were generated to investigate the endogenous CES1's role in vivo. IHC and HE staining assays were performed and demonstrated that loss of CES1 induced the higher ratio of prostatic intraepithelial neoplasia (PIN) in the anterior prostates (Figure 9A), suggesting that the absence of CES1 mediated the transformation of normal prostate tissue into tumorous tissues. R-IgG as negative control (Figure S6C). As quantitative indicators of lipid accumulation, the contents of TG and cholesterol measurement and BODIPY staining assays were carried out in the prostate tissues. As shown in Figures 9B and 9C, lipid accumulation was increased in the Ces1^−/− prostate tissues compared to the control mouse prostate tissues. Moreover, as shown in Figure 9D, the extent of Ki67 staining was markedly increased, while the extent of Perk and caspase-3 staining was significantly reduced in the Ces1^−/− prostate tissues, indicating that Ces1 repressed cell proliferation and induced ER stress-mediated apoptosis. The percentage of apoptotic cells was decreased in the Ces1^−/− prostate tissues, suggesting that the lack of endogenous CES1 facilitated cell survival; the Ces1^−/− prostate tissues had lower Ppara staining intensity comparing to the control group and higher intensity of Cyp11a1 and Psa staining, indicating that Ces1 mediated LD elimination and reduced androgen synthesis in the prostate cells.

Figure 9E shows the schematic diagram of this process. In an MT1-dependent manner, MLT regulated epigenetic modification of CES1 through SIRT1-mediated DNMT1 deacetylation. MLT-mediated CES1 expression induced lipid depletion, leading to ER stress-related apoptosis and blocking the intratumoral androgen synthesis that results in the inhibition of PCa progression and reversal of the enzalutamide resistance.

4.1 Cell culture and reagents

Human PCa cell lines LNCaP, C4-2, 22RV1, PC3, and DU145 and mouse-derived PCa cell line Rm-1 were cultured in RPMI 1640 medium supplemented with 10% fetal bovine serum (FBS) (Gibco, USA) at 37°C and 5% CO₂. RWPE-1 cells were incubated in K-SMF medium (Gibco) at 37°C and 5% CO₂. Note that 293T cells were incubated in DMEM supplemented with 10% FBS at 37°C and 5% CO₂. All cell lines were obtained from the American Type Culture Collection (ATCC).

Lentiviruses for the expression or knockout of CES1 and the corresponding control vectors were purchased from GeneChem. siRNAs for SIRT or MTNR1A (MT1) and control siRNA were obtained from RiboBio. All transfection steps were performed according to the manufacturers’ instructions. PCa cells were treated with various concentrations (0, 0.5, and 1 mM) of melatonin (Sigma) for 48 h according to previous studies.³⁷

4.2 Human samples

Paired adjacent noncancer and cancer paraffin-embedded tissue arrays for IHC staining were bought from Shanghai Outdo Biotech (HProA150CS03), Xi'an Alena Bio (DC-Pro11027 and DC-Pro11036), Xi'an ZK Bioaitech (M821601), and Suzhou CCELL (20190522R80). Detailed clinical information (pathological and clinical stage diagnoses which were given by pathologist) was directly provided by the companies. The fresh PCA tissues and adjacent normal tissues used for BSP (bisulfite genomic sequence polymerase chain reaction [PCR]) were obtained from the Department of Urology, Union Hospital, Tongji Medical College, Wuhan, China; the samples were frozen and stored in liquid nitrogen. Prior informed consent was obtained from each patient, and the study was approved by the Institutional Research Ethics Committee.

4.3 Animals and melatonin administration

All animal procedures have been approved by the Animal Care and Use Committee of Tongji Medical College of Huazhong University of Science and Technology.

Male C57BL/6 mice, 4- to 5-week-old, which had been surgically castrated, were bought from the Vital River Laboratory Animal Technology Co. Ltd. (Beijing, China). The animals were individually maintained in an isolated room at 25°C, with humidity of 60 ± 10% and a diurnal lighting schedule controlled by an electronic timer according to the experimental design (12 h light:12 h dark; 12L:12D, 200–300 lux; lights on 06:00 h, off at 18:00 h). One week before the tumor was implanted, one-half of the animals were changed to continuous light (12L:12LAN, 12 h light:12 h light at night), as described previously. Rm-1 cells in PBS (2 × 10⁶) were subcutaneously inoculated into each mouse. When the tumor reached approximately 0.5 cm, MLT, ENZ, or Vehicle (ETOH) treatment was initiated. MLT (Solarbio, Beijing, China) was dissolved in 0.02% ethanol to a final concentration of 60 mg/ml. As described previously, the stock solution was diluted to 0.1 mg/ml by adding drinking water in lightproof bottles and was administered to the mice from 18:00 to 06:00 to simulate MLT with high normal nocturnal physiological levels. The animals treated with intragastrically were administered 10 mg/kg ENZ (MedChemExpress) daily for approximately 7 weeks as described in a previous study. The present study included eight groups of five mice each: group I (12L:12D, EtOH), group II (12L:12D, ENZ), group III (12L:12D, MLT), group IV (12L:12D, MLT + ENZ), group V (12L:12LAN, EtOH), group VI (12L:12LAN, ENZ), group VII (12L:12LAN, MLT), and group VIII (12L:12LAN, ENZ + MLT). Tumor size was measured every 5 days using a digital caliper. After 55 days, the mice were euthanized, and the tumors were removed, measured, stored, and fixed in phosphate-buffered formalin for future study.

Male NCG mice (4- to 5-week-old), purchased from GemPharmatech., Ltd. (Nanjing, China), were surgically castrated prior to the use in the experiments in a subcutaneous xenograft model investigating the role of MLT/CES1 or MLT-induced reversal of enzalutamide resistance in vivo. Cells (5 × 10⁶) in a 1:1 mixture of PBS:Matrigel (BD, USA) were subcutaneously inoculated into each mouse. The animals were treated with MLT (200 mg/kg by intragastric administration). Measure the tumor size every 5 days. The mice were sacrificed on day 55 after cell implantation, and the tumors were excised, measured, stored, and fixed in phosphate-buffered formalin for future study.

Ces1-knockout mice (Ces1^−/−) constructed by the CRISP/Cas9 technique were purchased from Cyagen Bio-Technique Co., Ltd. (Suzhou, Jiangsu, China).

4.4 Cell proliferation assays

Five thousand cells/well were plated in 96-well plates for cell proliferation assay and cultured overnight at 5% CO₂ and 37°C. According to the Cell Counting Kit-8 (CCK-8) manufacturer's instructions, cell proliferation rate was determined. All experiments were carried out in triplicate.

4.5 Transwell assays

To assess the migration and invasion, cells were cultured to 80% fusion in six-well plates, then incubated in serum-free medium for 24 h. The Transwell chamber used for invasion tests was pre-coated with Matrigel (BD Biosciences, San Jose, CA, USA). The cells (2 × 10⁴) were cultured in the top chamber of an insert, and medium containing 10% FBS was added to the lower chamber as a chemoattractant. The chambers were incubated at 37°C and 5% CO₂ for 48 h. Then, fixed the cells on the lower surface in 100% methanol, then stained in 0.05% crystal violet. Five random fields were selected to count the number of cells for analysis. All experiments were performed in triplicate.

4.6 RNA sequencing

C4-2 cells were treated with DMSO or 1 mM MLT for 48 h, and three duplicate samples were collected for RNA sequencing. The RNA sequencing was executed by OE Biotech. Using TRIzol reagent (Invitrogen), RNA was extracted. Using a TruSeq Stranded Total RNA Ribo-Zero Gold kit, removed rRNA. Then reverse transcription was carried out to obtain cDNA, which was quality controlled by optical density and electrophoresis through agarose gel. The libraries were sequenced on an Illumina HiSeq X Ten platform. Generated 150 bp paired-end reads. Using Trimmomatic, raw data of fastq format were firstly processed, and the low-quality reads were removed to obtain the clean reads.⁷³ About 80 M clean reads for each sample were retained for subsequent analyses. Mapped the clean reads to the human genome (GRCh38) using HISAT2.⁷⁴ Differential expression analysis was carried out by using the DESeq R package (version: 1.18.0).⁷⁵ DEGs were further analyzed based on a |LogFC| > 1 and a p value < 0.05.

4.7 LC-MS/MS analysis

C4-2 cells, treated with DMSO or 1 mM MLT for 48 h, were collected and stored at -80°C for the LC/MS analysis. The LC-MS/MS detection and analysis were executed by Shanghai Lu Ming Biological Technology Co. Ltd. Identified the metabolites by using Progenesis QI (Waters Corporation, Milford) data processing software and public databases (http://www.hmdb.ca/ and http://www.lipidmaps.org/), and in-house databases.

4.8 BODIPY staining

The cells or fresh frozen sections were fixed in 4% paraformaldehyde solution and incubated in the dark with BODIPY for 30 min and DAPI for 10 min. Five random fields were selected and analyzed using ImageJ software. All experiments were carried out implemented in triplicate.

4.9 Measurement of MLT, TGs, and T-CHO

According to the MLT ELISA kit's instructions (Abcam, ab213978, USA), the levels of MLT in the plasma were measured. TG and T-CHO contents in the cells and plasma were measured using a T-CHO assay kit (Nanjing Jiancheng Bioengineering, cat. A111-1-1) and a TG assay kit (Nanjing Jiancheng Bioengineering, cat. A110-1-1), respectively, according to the manufacturers’ instructions. Blood collections (100 μl) were performed by caudal vein puncture, and blood was added to a procoagulant tube and centrifuged at 3000 rpm for 10 min at 4°C to obtain serum. The cells or tissues were lysed in PBS by sonication and centrifuged at 12,000 rpm for 10 min at 4°C to obtain extracts. The results were determined spectrophotometrically at 510 nm. A BCA kit was used to detect the protein concentration as a control.

4.10 Bisulfite genomic sequence PCR (BSP)

BSP was performed by Servicebio (Wuhan) according to the methods described previously (81). The UCSC Genome Browser (Human assembly hg38 from December, 2013) was used to predict the human CES1 promoter-related CpG islands (CGIs); the primers for BSP were designed by using MethPrimer online software (http://www.urogene.org/methprimer/). Treated genomic DNA samples with sodium bisulfite, and PCR was performed. Bisulfite sequencing analysis was carried out using an ABI3730XL sequencing system.

4.11 Western blotting

Western blotting was performed according to the methods described previously (81). The primary antibodies included anti-CES1 (1:2000, Proteintech, 16912-1-AP), anti-MT1 (1:1000, Abcam, ab203038), anti-SIRT1 (1:5000, Proteintech, 13161-1-AP), anti-DNMT1 (1:1000, Abcam, ab188453), anti-DNMT2 (1:1000, Proteintech, 19221-1-AP), anti-DNMT3a (1:1000, Abcam, ab227823), anti-DNMT3b (1:1000, Abcam, ab79822), anti-DNMT3L (1:500, Proteintech, 14939-1-AP), anti-PPARa (1:1000, Proteintech, 15540-1-AP), anti-phospho-PERK (Thr980) (1:1000, CST, 3179), anti- pospho-eIF2α (Ser51) (1:1000, CST, 3398), anti-ATF6 (1:1000, CST, 65880), anti-Bip (1:1000, CST, 3177), anti-IRE1a (1:1000, CST, 3294), anti-p-IRE1a (1:1000, Abcam, ab124945), anti-Caspase 3 (1:1000, CST, 9662), anti-CYP11A1 (1:1000, Proteintech, 13363-1-AP), anti-STARD4 (1:1000, Abacam, ab202060), anti-DHRSX (1:2000, Abcam, ab171736), anti-HNF4A (1:2000, Abcam, ab92378), anti-MGAT4C (1:1000, Proteintech, 17841-1-AP), anti-SLC7A10 (1:1000, Novus, NBP2-85763), anti-IGFBP1 (1:1000, Proteintech, 13981-1-AP), anti-ALDH1A2 (1:1000, Proteintech, 13951-1-AP), anti-ACSBG2 (1:1000, Novus, NBP1-54963), anti-ZIC1 (1:3000, Abcam, ab134951), anti-SLC38A4 (1:1000, Proteintech, 20857-1-AP), anti-acetyl Lysine (1:300, abcam, ab190479), anti-HDAC1 (1:1000, Abcam, ab109411), anti-β-actin (1:10000, Proteintech, 20536-1-AP), and anti-GAPDH (1:10000, Proteintech, 10494-1-AP).

4.12 Immunohistochemical staining and TUNEL assays

The tissue array sections were processed as reported previously (81). The primary antibodies included anti-CES1 (1:200, Proteintech, 16912-1-AP), anti-Ki67 (1:20000, Proteintech, 27309-1-AP), anti-PERK (1:50, Abcam, ab192591), anti-caspase-3 (1:200, Proteintech, 19677-1-AP), anti-CYP11A1 (1:200, Proteintech, 13363-1-AP), and anti-PSA (1:200, Proteintech, 10679-1-AP). The protein expression levels in the tissue sections were assessed based on the immunoreactive score (IRS). IRS scores of 0–1 indicated negative staining; scores of 2–3 indicated mild staining; scores of 4–8 indicated moderate staining, and scores of 9–12 indicated strong positive staining.

The TUNEL assay was performed using a TMR (red) TUNEL cell apoptosis detection kit (Servicebio, China) based on the manufacturer's instructions. Five random fields were selected to count the number of TUNEL-positive cells, and the apoptosis index in each field was calculated based on the percentage of TUNEL-positive cells relative to the total number of cells.

4.13 co-IP assays

Using Triton lysis buffer supplemented with a freshly prepared protease inhibitor cocktail and PMSF, the cells were lysed for 30 min on ice. Centrifugated at 12,000 × g for 10 min at 4°C. Then 20 μl of protein A/G PLUS-agarose (Santa Cruz, CA, USA) was added into 100 μl of the supernatant and incubated overnight with a primary anti-DNMT1 antibody (1:250, Proteintech, China) at 4°C. Then, using cold Triton lysis buffer, the immunoprecipitate was washed three times and then boiled with 2 × SDS loading buffer for 10 min. Then, Western blot assay was carried out according to the protocol described above.

4.14 Quantitative real-time PCR assays

Quantitative real-time PCR was executed as previously described.⁷⁶ The primer sequences for AR-V7 were as follows: Forward, 5′-CCATCTTGTCGTCTTCGGAAATGTTA-3′, Reverse, 5′-TTTGAATGAGGCAAGTCAGCCTTTCT-3. The housekeeping gene was GAPDH: forward, 5′-TCAAGAAGGTGGTGAAGCAG-3′, reverse, 5′-CGTCAAAGGTGGAGGAGTG-3′. We evaluate relative expression based on ΔΔCT value and used the student's t-test as statistical test.

4.15 ChIP assays

ChIP assays were carried out as described previously.⁷⁷ The primer sequences for the CES1 promoter CGIs region were as follows: P1-For: TGCCTTTATCCTGGCCACTC, P1-Re: TGCCTTTATCCTGGCCACTC; P2-For: TGAACATCCTGAGATCGCCC, P2-Re: GATCCAGCCGTAAGGGGACT; P3-For: CGGGGCACGACTTCCATATTA, P3-Re: GGGCGATCTCAGGATGTTCA; P4-For, CCGCACACACGAGAAATTTAGG, P4-Re, CTCCGTGATCCAGCCGTAAG. Used anti-DNMT1 antibody for IP, and normal IgG as a negative control.

4.16 Luciferase reporter assays

Luciferase reporter assays were executed as described previously.⁷⁸ CES1 promoter (CES1-WT) and blank (CES1-MUT) vector plasmids and a DNMT1 overexpression vector and blank vector plasmid were purchased from GeneChem. According to the manufacturer's instruction, renilla luciferase was used for the normalization of the luciferase activity detected using the dual-luciferase reporter assay kit (Promega, Madison, WI, USA).

4.17 Bioinformatics analysis

The mRNA expression of the genes in PCa tissues, adjacent normal tissues and the clinical data of the patients were obtained from The Cancer Genome Atlas (TCGA) database (http://www.cbioportal.org/public-porta) and the GEO database (Gene Expression Omnibus, https://www.ncbi.nlm.nih.gov/geo/, namely, the Taylor GSE21034, Wallace Prostate GSE6956, and Yu Prostate GSE 6919 datasets). GSEA was performed to identify the pathways enriched with the gene set. We used C2: curated gene sets for GSEA. Based on p value (p < 0.05) and FDR (FDR < 0.25), we identify gene sets which are associated with CES1 in PCa TCGA dataset. Then based on CORE ENRICHMENT in the result of GSEA, we identify genes which are associated with CES1 in the gene set. These genes and the DEGs in RNA sequencing are intersected, then obtain the statistically significant genes which are associated with MLT/CES1 in PCa.

4.18 Systematic review and meta-analysis

The literature search was executed independently by two investigators. We searched literature in the database Medline, PubMed, EMBASE, the Cochrane library, and CNKI, from inception to November 2020; selected appropriate studies for subsequent analyses, according to the inclusion criteria in Figures S1A and S1B. Using a standardized data extraction form, two investigators extracted information from each study independently. After quality assessment, statistics was performed in random model or fix effect model, using the software STATA 14.0.

4.19 Statistical analysis

Statistical analyses were performed using SPSS 22.0 (IBM, NY, USA) or GraphPad Prism 7.0 (GraphPad Software, San Diego, California, USA). A Kaplan–Meier analysis and log-rank test were carried out to assess the survival of the patients. Statistical analyses were executed using Mann-Whitney U test, Student's t-test, paired Student's t-test, and receiver operator characteristic curve. p values < 0.05 were considered statistically significant.