LNCaP, 22Rv1, PC3, and DU145 cells were purchased from the American Type Culture Collection (ATCC). PC-3M cells were obtained from the Institute of Cell Research, Chinese Academy of Sciences (Shanghai, China). The cell lines were all authenticated by short tandem repeat (STR) test. LNCaP, 22Rv1, DU145, and PC-3M cells were maintained in RPMI-1640 medium (SH30809.01, Hyclone, Utah, United States) supplemented with 10% fetal bovine serum (FBS). PC3 cells were cultured in Ham’s F12-K medium (21127022, ThermoFisher, Waltham, MA, United States) with 10% FBS.

Cepharanthine hydrochloride (CH), a semi-synthetic derivative of cepharanthine (A1443), was purchased from MUST BioTechnology (Chengdu, China). Deferoxamine (DFO, S5742), ferrostatin-1 (Fer-1, S7243), RSL-3 (S8155), and wortmannin (S2758) were purchased from Selleck Chemicals (Houston, TX, United States). Matrigel (354234) was obtained from Corning Costar (Lowell, United States). Testosterone (T1500) and methyl-thiazolyldiphenyl-tetrazolium bromide (MTT, M2128) were purchased from Sigma-Aldrich (St. Louis, MO, United States). Cell Counting Kit-8 (CCK-8, K1018) was obtained from APExBIO (Houston, Texas, United States). TRIzol reagent (15596–026) and the BCA protein assay kit (23227) were purchased from Invitrogen (Waltham, MA, United States). PrimeScript RT reagent kit (RR047A) and SYBR Premix DimerEraser (RR091A) were purchased from TaKaRa (Kusatsu, Shiga, Japan). Enhanced chemiluminescence (ECL) blotting detection kit (34580) was purchased from ThermoFisher Scientific (Waltham, MA, United States). Bovine serum albumin (BSA, 1027–106) was from Gibco (Waltham, MA, United States). RIPA buffer (P0013C) was purchased from Beyotime (Shanghai, China). Crystal violet (DC079) was from Genview (Beijing, China). Paraformaldehyde (BL539A) was from Biosharp (Hefei, China).

All animal work was performed in accordance with the Institutional Animal Care and Use Committee of Wuyi University (No. N2022013) and AAALAC guidelines. Maximum tumor volume should not exceed 1000 mm³ for mice. 22Rv1 xenograft models were established by subcutaneous (s.c.) injection of 5 × 10⁶ cells in Matrigel into the left flank of male nude mice when the animals were 6–7 weeks of age. When tumors reached between 100–200 mm³, as measured by a caliper using the formula V = 1/2 ab² (a, long axis; b, short axis), the mice were randomized into four groups of five mice each. The grouped mice were then subjected to different treatments intraperitoneally (i.p.): saline (Model), 100 mg/kg DFO, 30 mg/kg CH, and a combination of 100 mg/kg DFO with 30 mg/kg CH. The sizes of the tumors were measured every other day. All the animals were sacrificed on the 13th day. Livers of the mice were collected for hepatotoxicity study.

Cells were plated in 96-well plates at a density of 4,000 cells per well and allowed to adhere overnight. Subsequently, the cells were subjected to incremental concentrations of CH treatment for either 24 or 48 h. Following the treatment, MTT solution (0.5 mg/mL) was added to the cell culture medium and incubated with the cells for an additional 4 h at 37°C. The supernatant was subsequently aspirated, and 100 μL of DMSO was added. The optical density of the dissolved formazan was quantified at 570 nm using a multifunctional microplate detector (Synergy Neo2, BioTek, Winooski, VT, United States). The optical density (OD570) of the control cells was designated as 100% viability.

The cells were seeded in 6-well plates at a density of 1 × 10³ cells per dish and allowed to grow for 7–9 days. When colonies containing 20–30 cells were formed, they were treated with different concentrations of CH for an additional 9 days. The colonies were finally fixed with 4% paraformaldehyde and stained with 0.1% crystal violet.

The transwell system (3422, Corning Costar, Lowell, United States) was used for invasion experiments. Matrigel (0.5 mg/mL) was added to the upper chamber, and a suspension of 2 × 10⁵ cells in 100 μL serum-free medium was inoculated after the matrigel solidified. In the bottom chamber, 500 μL of medium containing 20% FBS and CH was added. Following 24 h of culturing, the invaded cells were stained with 0.1% crystal violet, and images of the invaded cells were captured using an invert microscope (RVL-100-G, ECHO, San Diego, CA, United States).

The confluent cells were scratched in a line across the bottom of the culture dish with a pipet tip. The cells were then cultured with a medium containing 2% FBS. Cell migration was observed at 24 h and 48 h, respectively. The micrographs show the extent of scratch closure achieved under normal conditions in comparison to conditions inhibited by the addition of CH. Cell migration quantification was evaluated by measuring the average distance of the central gap.

Cells were treated with CH for 24 h. For apoptosis analysis, the cells were harvested and stained with AnnexinV Alexa Fluor 488/PI reagent (FXP022-050, 4A Biotech Co., Beijing, China). For total reactive oxygen species (ROS) analysis, the cells were stained using a Total ROS Assay Kit (88–5930, Invitrogen, CA, United States). For lipid peroxidation detection, we used BODIPY (581/591) C11 (D3861, Invitrogen, CA, United States) as the probe. This probe incorporates into membranes where it undergoes a fluorescence emission shift upon peroxidation. The stained cells were then detected by a flow cytometer (CytoFLEX LX, Beckman Coulter, Indianapolis, IN, United States).

Transfections were performed using polyethylenimine (PEI) for Western blotting analysis. For siRNA transfection, non-targeting control siRNA and siRNAs specific for human AR were constructed by Sangon Biotech (Shanghai, China). The sequences of the siRNAs (5′-3′) are: siRNA#1 (+), CCU​GCU​AAU​CAA​GUC​ACA​CAU​TT; siRNA#2 (+), CGC​GAC​UAC​UAC​AAC​UUU​CCA​TT; siRNA#3 (+), GAA​AUG​UUA​UGA​AGC​AGG​GAU​TT.

Cells were lysed in RIPA buffer containing Tris (pH 7.4, 50 mM), NaCl (150 mM), 1% NP40, 0.5% sodium deoxycholate, 0.1% SDS, and protease inhibitor. The concentrations of the extracted proteins were determined via BCA assay. The lysates were mixed with a loading buffer and denatured by heating at 95°C for 10 min. Protein lysates with an equal total protein content were loaded onto an SDS-PAGE (8%–12%) gel and run under constant voltage to separate proteins. The proteins on the gel were then transferred to PVDF membranes (ISEQ00010, Millipore, Billerica, MA, United States) and probed with the indicated primary antibodies overnight and subsequent species-specific HRP-conjugated secondary antibodies. The immunoreactive bands were detected using an ECL blotting detection kit. The band intensity was quantified by ImageJ software.

Primary antibodies of AR (5153), LC3B (2775), and GAPDH (2118) were purchased from Cell Signaling Technology (Bost on, MA, United States). PARP (sc-8007), cleaved-PARP (sc-56196), Bcl-2 (sc-509), Caspase-3 (sc-56053), P62 (sc-28359), ATG5 (sc-133158), FSP1 (sc-377120), GPX4 (sc-166570), and AR (441) (sc-7305) were obtained from Santa Cruz Biotechnology (Dallas, TX, United States). DHODH (14877-1-AP) was from Proteintech (Manchester, United Kingdom). ACSL4 (ab155282) was from Abcam (Cambridge, United Kingdom).

LNCaP cells at a confluence of 60%–70% were treated with CH for 12 h and fixed in 2.5% glutaric dialdehyde at 4°C overnight. After fixation, dehydration, infiltration, and embedding, the samples were sliced into ultra-thin sections and stained with uranyl/lead citrate. Images were captured under a transmission electron microscope (TEM for short, HT7700, Hitachi, Tokyo, Japan).

Total mRNA was extracted using the TRIzol reagent. The quantity and quality of RNA were determined using a spectrophotometer (DS-11 FX+, DeNovix, Wilmington, United States), and then the RNA was reversely transcribed into cDNA using a PrimeScript™ RT reagent kit with gDNA eraser (Perfect Real Time; RR047A, TAKARA, Kyoto, Japan) according to the manufacturer’s instructions. The q-PCR was performed using a 2 × SYBR Green qPCR Master Mix (Low ROX; B21703, Selleck, Texas, United States) in the real time fluorescence quantitative PCR instrument (LightCycler 96, Roche, Basel, Switzerland) in a 20 μL reaction. The primers are shown in Table 1.

High-throughput sequencing was performed using the TRIZOL method (T9424, Merck, Rahway, NJ, United States) to prepare of RNA samples. The quality and integrity of the processed samples were assessed by measuring nucleic acid concentration with a Nanodrop2000 (Thermo Fisher, Waltham, MA, United States) and evaluating integrity using an Agilent 2100 or LabChip GX (PerkinElmer, Waltham, MA, United States). Sequencing was conducted with the DNBSEQ-T7RS High-throughput Sequencing Kit (FCL PE150) V3.0 (940–000266–00, BGI Genomics, Shenzhen, China). The final library was loaded at a concentration of 1 pmol, as determined by a Qubit 3.0 fluorometer (Thermo Fisher, Waltham, MA, United States) with the Qubit dsDNA HS Assay Kit (1000017571, BGI Genomics, Shenzhen, China). Differentially expressed genes (DEGs) were identified with P values adjusted using the Benjamini–Hochberg method to control the false discovery rate. A corrected P value threshold of 0.05 and a minimum absolute fold change of 2 were applied to define significant differential expression. Additionally, KOBAS software was used to assess the statistical enrichment of DEGs in KEGG pathways, providing insights into their potential biological roles and interactions. The transcriptome data generated in the present study can be found in the NCBI BioProject under accession number PRJNA1157790.

All data are shown as mean ± SD from at least 3 independent experiments. The 2-tailed Student's t-test and one-way ANOVA were used for the statistical analysis. Statistical significance was defined as *, P < 0.05; **, P < 0.01; ***, P < 0.001.

The structure of CH used in this study is shown in Figure 1A. To evaluate the inhibitory effect of CH on PCa cells, we performed a cell viability assay using AR-independent cells DU145, PC-3M, PC3, and AR-dependent cells 22Rv1 and LNCaP. From the IC50 values shown in Figure 1B, we found that CH exhibits cytotoxicity across various PCa cell types, among which 22Rv1 and LNCaP cells showed the highest sensitivity. We primarily used these 2 cell lines for the following experiments. Since LNCaP and 22Rv1 are both androgen receptor (AR)-dependent prostate cancer cells, we also selected the AR-independent cell line PC3 for comparative studies in our subsequent experiments, as PC3 requires less drug concentration to achieve similar effects, making it more drug-sensitive. CH is shown here to be a promising drug to inhibit PCa in vitro, due to its inhibitory effects on colony formation (Figure 1C), transwell invasion (Figure 1D), and scratch wound healing (Figure 1E) in these three PCa cells.

To identify the nature of cell death induced by CH, we investigated the apoptotic pathway firstly. CH can induce apoptosis in LNCaP and 22Rv1 cells, but simultaneously, it can also trigger a certain degree of non-apoptotic cell death (Figures 2A, B). However, for PC3 cells, non-apoptotic cell death is the main form of cell death under the action of CH. Similarly, apoptotic marker analysis via Western blotting confirmed that LNCaP and 22Rv1 cells underwent a certain degree of apoptosis, while PC3 cells should have undergone a different form of cell death (Figure 2C). To determine the specifics of CH-induced non-apoptotic cell death, we further examined the ferroptosis pathway, as our previous studies found that CH can induce autophagy in various cells (Wang et al., 2019; Wang et al., 2020). As autophagy is also associated with ferroptosis, we speculate that the agent may induce ferroptosis in PCa cells. Iron chelator DFO and ferroptosis inhibitor Fer-1 were subsequently used in cell viability assays to validate the existence of CH-induced ferroptosis (Figure 2D). It is noteworthy that the concurrent administration of DFO or Fer-1 markedly diminished the susceptibility of LNCaP and 22Rv1 cell lines to CH, thereby highlighting the ferroptosis inducing potential of CH in these 2 cells. Under a TEM, the mitochondria of LNCaP cells treated by CH underwent shrinkage and increased membrane density compared to the mitochondrial features induced by RSL-3 (Kinowaki et al., 2021), a well-known ferroptosis inducer (Figure 2E). The total ROS levels of LNCaP cells (Figure 2F) and lipid peroxidation detection of LNCaP and 22Rv1 cells (Figure 2G) confirmed that CH induced ferroptosis in these AR-dependent cells. For PC3 cells, an AR-independent cell, the levels of CH-induced lipid peroxides did not significantly increase, and iron chelator also did not have a protective effect. These findings suggest that CH can trigger both apoptosis and ferroptosis in PCa cells, with a particularly strong effect observed in AR-dependent cells.

There is a complex relationship between ferroptosis and autophagy. To clarify the regulatory relationship between ferroptosis and autophagy under the treatment of CH, and to determine whether CH-induced ferroptosis is related to autophagy, we first investigated the autophagy flux in all 3 cell lines. As shown in Figure 3A, protein levels of LC3-II and the ratio of LC3-II/LC3-I were increased in a dose-dependent manner in CH-treated cells, indicating CH initiated autophagy in all 3 cell lines. The protein sequestosome 1 (p62/SQSTM1) is an autophagy receptor and a selective substrate for autophagy. P62 accumulation observed here indicates a reduced autophagic flux. Autophagic cell death may be the main reason for the toxicity of CH to PC3 cells. TEM analysis revealed that autophagosomes appeared frequently in RSL3 treatment, as indicated by red arrows in Figure 3B (middle). In the cells treated with CH, numerous autophagosomes can also be observed, but these autophagosomes may have engulfed more organelles, such as mitochondria (Figure 3B, right). We next employed the autophagy inhibitor wortmannin in flow cytometry assay (Figure 3C). Wortmannin can restore CH-induced lipid peroxidation elevation in LNCaP and 22Rv1 cells, but not in PC3 cells. overall, we illustrated that CH can induce incomplete autophagic flux in PCa cells, preventing autophagic substrates from being recycled by the cells, which may be the main reason for the toxicity of CH to PC3 cells. Blocking autophagy can reduce lipid peroxides caused by CH, indicating that CH-induced accumulation of autophagic cargos promotes ferroptosis.

To systemically study the mechanisms of CH-induced ferroptosis, we performed transcriptome analysis and pathway enrichment analysis of DEGs using the KEGG database (KEGG, 2022). From the results of 22Rv1 cells, the significance of DEGs in the ferroptosis pathway ranks 12th (Figure 4A), much higher than LNCaP and PC3 cells. Next, we performed quantitative real-time PCR validation experiments on these ferroptosis pathway-related genes that exhibited significant regulation in the 22Rv1 cell line, across the three different PCa cell lines (Figure 4B). In 22Rv1 cells (Figure 4B, middle), FTH1, SLC7A11, and HMOX1 were increased more than twofold. Among the three upregulated genes, FTH1 and SLC7A11 were both negative regulators in ferroptosis. Note that STEAP3 is downregulated by CH significantly, which results in ferroptosis induction. From the above findings, the mechanism of ferroptosis induced by CH in 22Rv1 cells is complex. The regulation of ferroptosis-related genes by CH does not solely target the induction of ferroptosis, but rather represents a comprehensive outcome of multi-gene regulation. In LNCaP cells (Figure 4B, top), there are only three genes—TFRC, SLC7A11, and SAT1—regulated more than twofold by CH treatment. However, in PC3 cells (Figure 4B, bottom), no gene was significantly regulated more than twofold. Overall, the mechanism of CH-induced ferroptosis in PCa varies among cells.

FSP1 is a glutathione-independent ferroptosis inhibitor molecule, which exerts a ferroptosis inhibition role parallel to GPX4. Therefore, we aim to determine the effect of CH on these two ferroptosis defense factors. As the above experiments have shown that under the treatment of CH, PC3 cells hardly exhibit ferroptosis, our subsequent mechanistic studies focused on LNCaP and 22Rv1 cells. Firstly, we found that CH downregulated the mRNA levels of GPX4 in LNCaP and 22Rv1 cells (Figure 5A). Meanwhile, CH reduces AR protein and its truncated variant AR-V7 (Figure 5B). To elucidate the intrinsic relationship between AR and GPX4, we browsed Oncomine (ONCOMINE, 2020) (the website has now ceased its operation) to determine AR and GPX4 expression, and found co-expression of these two genes within a research of PCa containing 102 cases (Figure 5C). As AR functions both as a nuclear receptor and a transcription factor, our initial validation focuses on the nuclear translocation of AR and its regulatory effects on GPX4 and its parallel ferroptosis suppressor FSP1. In AR-overexpressing cells treated with the AR ligand testosterone, the expression of protein AR leads to a modest increase in the protein levels of GPX4 and FSP1 (Figure 5D). However, in cells overexpressing protein AR and subsequently knocked down by siRNAs, in the presence of androgen, the expression of both GPX4 and FSP1 proteins decreases with the reduction of AR. However, in the absence of androgen, the levels of protein GPX4 and FSP1 are not influenced by AR levels (Figure 5E). This indicated that only nuclear localized AR exerts slight regulatory effects on both proteins. Furthermore, to verify whether GPX4 and FSP1 serve as target genes of AR, acting as a transcription factor, we used JASPAR (JASPAR, 2022) to predict whether AR could bind to the promoters of the two genes. The sequence logo representing conserved DNA binding sites where AR binds is shown in Figure 5F. Intriguingly, JASPAR predicts that AR is likely the promoter of both GPX4 and FSP1, because there are numerous potential binding domains in the promoter regions of these two genes (Figure 5G). However, regardless of AR overexpression or knockdown, the transcription levels of genes GPX4 and FSP1 show no significant and consistent accompanying changes (Figures 5H, I). Therefore, we infer that AR is not a direct transcription factor for genes GPX4 and FSP1. Overall, we conclude here that CH moderately regulates the expressions of GPX4 and FSP1 via nuclear AR, but does not involve its transcription factor function.

In our subsequent research, we conducted exploratory studies on other important regulatory factors of ferroptosis. Our findings revealed that the regulation of ACSL4 and DHODH by CH played a pivotal role in this process. CH up-regulated the transcriptional levels (Figure 6A, top) and protein levels of ACSL4 (Figure 6B, top), but down-regulated the transcriptional levels (Figure 6A, bottom) and protein levels (Figure 6B, bottom) of DHODH in LNCaP cells and 22Rv1 cells. The regulation of CH on ACSL4 and DHODH synergistically promotes the occurrence of ferroptosis.

To investigate the effect of CH on inducing ferroptosis in vivo, a xenograft mouse model was established. The 6–7-week-old nude mice were treated with CH (30 mg/kg) or CH combined with DFO (100 mg/kg) intraperitoneally for 12 days (Figure 7A). The animals were sacrificed on the 13th day. The body weights and tumor sizes were measured every other day. Figure 7B shows that during the administration period, no significant body weight change was observed within each group. Interestingly, a xenograft growth curve revealed a significant inhibition of tumor growth in the CH group. However, when combined with DFO, the xenografts exhibit a reduced sensitivity to CH (Figure 7C). To better illustrate the differences among the groups, we statistically analyzed tumor size and presented the results in the form of a bar chart (Figure 7D). From the chart, we found that DFO itself had some anti-tumor effects throughout the therapy period except the seventh day, while CH exhibited anti-tumor effect from the fifth day onward. Data from the seventh day to the 11th day showed significant differences in tumor size between the CH group and CH + DFO group, indicating the critical role of ferroptosis in the pharmacological mechanism of CH. HE stainings of the liver tissue showed no obvious hepatotoxicity in both groups during the 12-day experiment period (Figure 7E).