Ketone drink enhances therapeutic efficacy in prostate cancer by targeting EZH2

Abstract

It is well established that EZH2, a lysine methyltransferase, is upregulated in most aggressive cancers, highlighting the importance of EZH2 in cancer progression. Recent research has shown that metabolic reprogramming is pivotal in various biological processes, including cancer. Despite this, evidence of EZH2’s role in regulating cancer metabolism remains limited. Our study reveals a negative correlation between EZH2 and HMGCS2, a gene belonging to the HMG-CoA synthase, in prostate and breast cancers. Interestingly, HMGCS2 is inversely related to cancer progression and prognosis in these cancers. Furthermore, HMGCS2 is epigenetically repressed by EZH2 both in vitro and in vivo. Notably, restored EZH2 reduces the elevated HMGCS2 levels observed upon EZH2 depletion. Overexpression of HMGCS2 decreases tumorigenesis in both prostate and breast cancers. Additionally, β-hydroxybutyrate (BHB), a downstream metabolite of HMGCS2, impedes prostate cancer progression by targeting EZH2 via direct protein-compound interaction-mediated protein degradation. More importantly, the ketone drink of BHB administration dramatically reduces tumor size and weight in a therapy-resistant, castration-resistant prostate cancer patient-derived xenograft model. Combining a ketone drink with FDA-approved drugs enzalutamide and Tazemetostat further suppresses tumor progression. Overall, the EZH2-HMGCS2-BHB regulatory network plays a critical role in the progression of prostate cancer, and a ketone drink is a novel therapeutic tool for patients with aggressive prostate cancer.

Introduction

Prostate cancer (PCa) is the second leading cause of cancer-related deaths in American men [1]. Despite advancements in treatment, some PCa cases progress to a hormone-refractory state known as castration-resistant prostate cancer (CRPC), which has a survival rate of only 1 to 2 years on average [2]. Enhancing survival rates necessitates a deeper understanding of the molecular mechanisms driving PCa development and its progression to CRPC. Both genetic and epigenetic factors play critical roles in PCa development. During tumor progression, epigenetic regulators reprogram normal and differentiated cells into undifferentiated cancer cells, promoting cell growth, invasion, and metastasis by upregulating oncogenes and downregulating tumor suppressors [3].

One of the critical epigenetic regulators is the Enhancer of zeste homolog 2 (EZH2), a histone methyltransferase that catalyzes the methylation of lysine 27 on histone H3 (H3K27) [4], leading to gene silencing and 3-D genome changes [5]. Overexpression of EZH2 is very common in most types of cancers, including prostate [6], breast [7], and melanoma [8], making it a potential target for cancer therapy [4]. Currently, Tazemetostat has been approved by FDA for some patients with follicular lymphoma or epithelioid sarcoma [9]. Several EZH2 inhibitors are undergoing clinical trials for solid tumors [10].

Cancer cells often undergo metabolic reprogramming, adapting their metabolism to support rapid growth and survival in a nutrient-poor and hypoxic tumor microenvironment [11]. This reprogramming allows cancer cells to thrive by altering their glucose and lipid metabolism among other pathways [11, 12]. As a result, targeting metabolic reprogramming has emerged as a promising strategy for cancer therapy. EZH2 has been shown to play a role in this process, influencing both glucose and lipid metabolism in cancer cells. By epigenetically silencing metabolic genes and directly regulating key metabolic enzymes, EZH2 helps rewire cancer cell metabolism to support tumor growth. For instance, in ovarian cancer, EZH2 upregulates isocitrate dehydrogenase 2, enhancing the tricarboxylic acid (TCA) cycle activity and promoting tumor growth independently of its methyltransferase activity [13]. Additionally, EZH2 modulates lipid metabolism by regulating genes involved in lipogenesis and lipolysis, as observed in adipocytes and glioblastoma [14, 15].

Ketogenesis is another metabolic process of interest, producing ketone bodies like β-hydroxybutyrate (BHB) as alternative fuel sources during low-carbohydrate availability states, such as fasting or ketogenic diets [16]. Cancer cells have altered metabolic pathways that allow them to use glucose and other nutrients more efficiently than normal cells [11]. While some studies suggest that ketogenic diets might limit cancer cell growth by restricting glucose availability [17], the relationship between ketogenesis and cancer remains complex and not fully understood.

HMGCS2, also known as 3-hydroxy-3-methylglutaryl-coenzyme A synthase 2, is a rate-limiting enzyme responsible for the synthesis of ketone bodies. Recent studies have reported that HMGCS2 is upregulated during ketogenesis, leading to BHB production [18]. While HMGCS2 expression increases in colorectal, oral, and rectal cancers [19, 20], it decreases in esophageal squamous cell carcinoma, liver, renal, and prostate cancers [21–25]. Understanding the dysregulation of HMGCS2 in various cancers could pave the way for novel therapeutic strategies targeting this enzyme. Further research is necessary to elucidate the mechanisms behind HMGCS2 dysregulation and to develop effective HMGCS2-targeted therapies.

Results

HMGCS2 is associated with PCa and breast cancer progression and prognosis

It is controversial whether HMGCS2 is a tumor suppressor or oncogene in various cancers. It has been reported that HMGCS2 was upregulated in colorectal, oral, and rectal cancers [19, 20]. However, it has also been reported that HMGCS2 was downregulated in esophageal squamous cell carcinoma, liver, renal, and prostate cancers [21–25]. When we analyzed the publicly available PCa transcriptome profiling datasets (TCGA-PRAD, SU2C [26], and Taylor dataset [27]), we observed that HMGCS2 expression levels were reversely associated with the progression of PCa (Fig. 1A, B). HMGCS2 transcript levels were highest in benign prostate tissues and lowest in metastatic castration-resistant prostate cancer (mCRPC) (Fig. 1A, B). In addition, HMGCS2 expression is low in progenitor-like tissue and neuroendocrine prostate cancer (NEPC) tissue [28] (Fig. 1C). More importantly, high expression of HMGCS2 shows significantly better prognosis with higher recurrence-free survivals in PCa patients compared to low expression of HMGCS2 (Fig. 1D, E). Consistently, the median disease-free survival time of breast cancer (BCa) patients group with a high HMGCS2 expression was significantly longer than that of the group with a low HMGCS2 expression (Fig. 1F) [29]. Additionally, we analyzed the expression of HMGCS2 in breast invasive carcinoma based on sample types. Primary breast invasive tumors showed lower expression of HMGCS2 compared to normal tissues (Fig. 1G) [30, 31]. Furthermore, normal breast tissues showed higher HMGCS2 expression compared to BCa tissues at stage 1, 2, 3, and 4, respectively. (Fig. 1H) [30, 31]. The basal subtype of BCa tissue showed the lowest expression of HMGCS2 (Supplementary Fig. 1A, B). These results indicate that HMGCS2 might be a tumor suppressor in PCa and BCa. Since the Androgen Receptor (AR) is the master regulator of PCa progression and influences numerous cellular mechanisms in PCa, we analyzed the correlation between AR and HMGCS2. However, no correlation was observed in two datasets (Supplementary Fig. 2A, B), suggesting that HMGCS2 functions independently of AR.

Fig. 1. HMGCS2 expression is lower in metastasis PCa and is negatively associated with EZH2 expression in PCa and BCa.

A, B HMGCS2 mRNA level in prostate tissues is shown based on sample types in normal prostate tissue (n = 52), primary (n = 500) and metastatic prostate tissues (n = 101) from TCGA-PRAD + SU2C dataset (A), and in normal prostate tissue (n = 29), primary (n = 131) and metastatic prostate tissues (n = 19) data from Taylor et al. (B). C, I HMGCS2 (C) and EZH2 (I) expression levels based on different groups of PCa by scRNA-seq. D, E Patients with PCa from Taylor et al. (total samples = 186) (D), TCGA-PRAD (n(high)=123, n(low)=123) (E) dataset were split into two groups based on mean expression of HMGCS2 and the resultant differences in disease free survival are shown by log-rank test. F Patients with BCa were split into two groups based on mean expression of HMGCS2 and the resultant differences in disease-free survival are shown by log-rank test (n(high) = 1014, n(low) = 1014). G, H HMGCS2 mRNA level in breast tissues is shown based on sample types (G) and individual cancer states (H) from TCGA data. J, K Inverse relationship between HMGCS2 and EZH2 in prostate adenocarcinoma (J) and BCa (K) using TCGA and METABRIC dataset. L Immunoblot of six different PCa cell lines for EZH2, HMGCS2, GAPDH, and H3 protein level. M Quantification of EZH2 and HMGCS2 protein levels in PCa cell lines. Protein levels were normalized to control values, with EZH2 in PC-3 cells set to 1.0 and HMGCS2 in LNCaP cells set to 1.0. Mean ± SE (n = 3). N Correlation of EZH2 and HMGCS2 Levels from PCa cell lines in Fig. 1M. PCa prostate cancer, BCa breast cancer.

HMGCS2 and EZH2 are negatively correlated in PCa and BCa

Using the TCGA-PRAD RNA-seq data from PCa patients, the expression level of HMGCS2 is negatively associated with the expression of EZH2 (Fig. 1J). While HMGCS2 level is high in non-malignant tissue and low in NEPC (Fig. 1C) [28], EZH2 level is low in non-malignant and high in NEPC in PCa tissues (Fig. 1I) [28]. Consistently, the expression level of HMGCS2 is negatively correlated with the expression of EZH2 in BCa as well (Fig. 1K). To confirm this relationship in PCa, we performed immunoblot analysis with six different PCa cell lines. EZH2 protein was highly expressed in DU-145 and PC-3 cell lines, but HMGCS2 protein was not detected (Fig. 1L, M). On the other hand, EZH2 protein was lightly detected and HMGCS2 protein was highly expressed in LNCaP cell lines (Fig. 1L, M), suggesting that EZH2 and HMGCS2 are negatively associated in PCa cell lines (Fig. 1N).

HMGCS2 is epigenetically repressed by EZH2

To investigate whether EZH2 regulates HMGCS2 expression, we performed RNA interference with two EZH2 siRNA duplexes in C4-2 PCa cell lines. Interestingly, EZH2 depletion markedly increased HMGCS2 mRNA level (Fig. 2A), and protein level (Fig. 2B), respectively. To examine whether EZH2’s enzymatic activity is essential for the increase of HMGCS2 expression, we treated PCa cells with EZH2 inhibitor GSK126, a S-Adenyl-l-methionine (SAM)-competitive EZH2 pharmacological enzymatic inhibitor [32]. GSK126 treatment increased HMGCS2 mRNA and protein levels in C4-2 prostate cell lines in a dose-dependent manner (Fig. 2C and D). Another EZH2 enzymatic inhibitor EPZ-6438 [33] also markedly increased HMGCS2 protein level in vitro (Fig. 2E), indicating EZH2’s enzymatic activity is required to increase HMGCS2 level. GSK126 treatment also increased HMGCS2 protein level in BCa cell lines T47D (Supplementary Fig. 3A), suggesting that EZH2-mediated HMGCS2 regulation is not limited to PCa. MS1943, an EZH2-selective degrader [34], remarkedly increased HMGCS2 mRNA and protein levels in C4-2 cell lines (Supplementary Fig. 3B and Fig. 2F). We further examined EZH2-mediated changes in HMGCS2 expression in DU-145 cell lines, which exhibit high EZH2 and low HMGCS2 levels (Fig. 1L). Although HMGCS2 protein levels were undetectable by western blot (data not shown), EZH2 depletion via siEZH2 significantly increased HMGCS2 mRNA levels in DU-145 cells (Supplementary Fig. 3C). Additionally, treatment with the EZH2 inhibitor EPZ-6438 and the EZH2 degrader MS1943 resulted in a dose-dependent increase in HMGCS2 mRNA levels (Supplementary Fig. 3D and E). To check EZH2-mediated H3K27me3 occupancy of HMGCS2 promoter regions, we scanned four different fragments upstream of 2 Kb from transcription start site by ChIP-qPCR. H3K27me3 was highly enriched at region 3 and 4 (−1000~−2000 bps) and this enrichment of H3K27me3 was dramatically decreased with 5 µM GSK126 treatment (Fig. 2G). This confirmed that the promoter of HMGCS2 is epigenetically repressed by EZH2 via H3K27 methylation.

Fig. 2. Depletion, inhibition, and degradation of EZH2 increased HMGCS2 level in PCa in vitro and in vivo.

A, C qPCR analysis showed HMGCS2 was increased in EZH2-depleted and inhibited C4-2 cells. B, D, E, F Immunoblot analysis confirmed increased-HMGCS2 protein level with depletion, inhibition, and degradation of EZH2. G ChIP-qPCR assay to monitor the enrichment of H3K27me3 at 2 kb upstream of the promoter regions of HMGCS2 in C4-2 cells with vehicle or 5uM of GSK126 treatment. F1(transcription start site (TSS)~−0.5 kb), F2 (−0.5~−1.0 kb), F3 (−1.0~−1.5 kb), F4 (−1.5~−2.0 kb). H, I Precastrated mice carrying LuCaP 35CR, an enzalutamide-resistant and abiraterone-resistant patient-derived xenograft model, received EPZ6438 (200 mg kg⁻¹ per day) for 28 days (five days per week). Tumor tissues were lysed and blotted for HMGCS2, EZH2, H3K27me3, H3, and β-actin. HMGCS2, EZH2, and H3K27me3 levels were normalized to the loading controls β-actin and H3, and the averages of the values normalized to β-actin and H3 were used. J, K C4-2 bearing NCG mice received MS8815 (25, 50, 100 mg kg⁻¹ per day) for 28 days (five days per week), and tumor tissues were used for immunoblotting analysis. Tumor tissues were lysed and blotted for HMGCS2, EZH2, H3K27me3, H3, and β-actin. HMGCS2, EZH2, and H3K27me3 levels were normalized to the loading controls β-actin and H3, and the averages of the values normalized to β-actin and H3 were used. MS: MS8815, Mean ± SE. Student t-test. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

HMGCS2 is increased in vivo by targeting EZH2

Furthermore, we were interested in whether EZH2 inhibitors and degraders regulate HMGCS2 level in vivo. We treated immunodeficient NCG mice harboring LuCaP 35CR [35], an enzalutamide-resistant and abiraterone-resistant CRPC patient-derived xenograft (PDX) model, with tazemetostat (EPZ-6438) [33], an FDA-approved EZH2 catalytic inhibitor in some patients with epithelioid sarcoma or follicular lymphoma. As shown in Fig. 2H and I, HMGCS2 protein levels were markedly increased in EPZ6438-treated PDX tumors when compared to those of the vehicle control group. We also utilized C4-2 bearing immunodeficient NCG mice model and treated them with a PROTAC-based EZH2 inhibitor MS8815 [36] by intraperitoneal injection for 28 days. As a result, MS8815 treatment also markedly increased HMGCS2 protein levels in mouse tumors in a dose-dependent manner (Fig. 2J and K). These results demonstrate that targeting EZH2 significantly increased HMGCS2 protein and transcript levels both in vitro and in vivo.

Re-expression of EZH2 represses increased-HMGCS2 upon EZH2 depletion

To study the role of EZH2 in regulation of HMGCS2, EZH2 was restored in EZH2-depleted PCa cells. While EZH2 depletion increased HMGCS2 mRNA level (Fig. 3A), re-expression of EZH2 rescued HMGCS2 back to low levels before depleting EZH2 (Fig. 3B). To further validate this result, doxycycline-inducible shEZH2 cell lines were generated. While EZH2 protein levels were decreased, HMGCS2 protein levels were markedly increased upon doxycycline induction (Fig. 3C and D). After changing into fresh culture media without doxycycline, EZH2 protein levels were restored to a high level at day 30, and the increased-HMGCS2 protein levels were decreased accordingly. These results further confirmed that EZH2 directly regulates HMGCS2 during cancer progression.

Fig. 3. Re-expression of EZH2 represses increased-HMGCS2 upon EZH2 depletion.

A, B qPCR analysis showed EZH2 and HMGCS2 mRNA level after depletion of EZH2 and reexpression of EZH2 in C4-2 cells. C, D Doxycycline-inducible EZH2-depletion cells were generated, treated with doxycycline for 4 days (1 ug/ml), and changed into fresh media. After 4 days of maintaining with fresh media (total 8 days from the start day of doxycycline treatment) and 26 days of maintaining of fresh media (total 30 days from the start day of doxycycline treatment), cell lysates were used for immunoblot analysis. The figure shows protein expression of EZH2 and HMGCS2 from immunoblot analysis with three different biological replicates. Mean ± SE. Student t-test. **p < 0.01, ***p < 0.001, ****p < 0.0001.

HMGCS2 is a tumor suppressor in PCa and BCa

Next, we were interested in the role of HMGCS2 in the progression of PCa. C4-2 cells were transiently transfected with empty vector (EV) control or FLAG-HMGCS2 plasmid. Overexpression was confirmed by western blot analysis (Fig. 4A). Overexpressing HMGCS2 significantly inhibited growth of C4-2 cells compared to EV control (Fig. 4B (left)). To investigate the role of HMGCS2 in PCa migratory capability, wound healing assay was performed in HMGCS2-overexpressed cells. HMGCS2 overexpression markedly decreased the migration of C4-2 cells (Fig. 4C (left)). Additionally, HMGCS2 overexpression decreased migrated cells by transwell chamber assay (Fig. 4D (left)). Furthermore, HMGCS2 overexpression also decreased cell proliferation (Fig. 4B (right), Supplementary Fig. 4B), migratory capability by wound healing assay (Fig. 4C (right), Supplementary Fig. 4C) and transwell chamber assay (Fig. 4D (right), Supplementary Fig. 4D) in PC-3 cells and BCa MDA-MB-231 cells. These results suggested that HMGCS2 may alter the major pathways involved in tumor proliferation and progression.

Fig. 4. HMGCS2 overexpression decreases PCa progression.

A Immunoblot analysis confirmed HMGCS2 overexpression in C4-2 and PC-3 cell lines. B HMGCS2 overexpression decreased cell viability in C4-2 and PC-3 cell lines by cell titer glo. C Wound healing assay was conducted to evaluate the migration potential of HMGCS2 overexpression in C4-2 and PC-3 cell lines. D Boyden chamber migration assay was performed in C4-2 and PC-3 cell lines. Mean ± SE. Student t-test. *p < 0.05, **P < 0.01, ***P < 0.001, ****p < 0.0001.

HMGCS2 downstream metabolite, BHB, inhibits PCa progression via targeting EZH2

Many studies have reported that when HMGCS2 is increased, the downstream metabolite, β-hydroxybutyrate (BHB), is increased as well [16]. BHB is one of the ketone bodies found in the circulation after fasting or ketogenic diets. Thus, we measured BHB level upon EZH2 depletion in C4-2 cells via targeted metabolomic analysis by LC-MS/MS and found out that EZH2 depletion increased cellular BHB level compared to siControl (Fig. 5A). Furthermore, BHB treatment inhibited the growth of PCa cells with IC₅₀ 93 mM (C4-2) and 99 mM (PC-3) respectively. (Supplementary Fig. 5A, B). We chose 50 mM of BHB for further assay since it is below IC₅₀ level and did not show cytotoxicity in C4-2 and PC-3 cell lines. BHB treatment decreased cell migratory capability by wound healing assay (Fig. 5B, Supplementary Fig. 5D), and transwell chamber assay (Fig. 5C, Supplementary Fig. 5E). Intriguingly, EZH2 protein levels were remarkably decreased upon BHB treatment in a dose-dependent manner in C4-2, PC-3, and 22Rv1 PCa cell lines (Fig. 5D). However, EZH2 mRNA levels were not changed by 50 mM of BHB in C4-2 cells (Supplementary Fig. 5C (left)), indicating EZH2 is not regulated by BHB at transcript level. To determine how BHB inhibits the cancerous properties of PCa cells, we performed RNA-sequencing (RNA-seq) transcriptome analysis in BHB-treated and control C4-2 cells and compared with our previously obtained RNA-seq data of GSK126-treated or EZH2-depleted C4-2 cells by shEZH2 and siEZH2. BHB-mediated altered gene expressions were similar to the altered genes with EZH2 depletion by shEZH2 and siEZH2 and enzymatic inhibitor GSK126 treatment (Supplementary Fig. 6). GSEA analysis of RNA-seq data revealed that BHB treatment decreased genes associated with E2F targets, DNA repair, G2M checkpoint, and MYC targets (Fig. 5E (left), Supplementary Fig. 7A–D). Similarly, EZH2 depletion also decreased these genes (Fig. 5E (right), Supplementary Fig. 8A–D). Additionally, BHB treatment and EZH2 depletion shared 24 positively enriched and 12 negatively enriched gene signatures (Fig. 5F). These findings suggest that BHB functions as an EZH2 degrader. Furthermore, we investigated whether HMGCS2 overexpression exhibits similar gene signature patterns. Our qPCR analysis revealed that HMGCS2 overexpression also downregulated genes associated with E2F targets, DNA repair, G2M checkpoint, and MYC targets (Supplementary Fig. 9C–F).

Fig. 5. BHB bound to EZH2 protein, and induced EZH2 degradation in a proteasome-dependent manner.

A After three days of transfection of siControl and siEZH2 in C4-2 cells, targeted metabolomics was conducted to measure BHB levels by HPLC-MS/MS. B Wound healing assay was performed after three days of treatment of BHB in C4-2. BHB was dissolved in RPMI medium, vehicle is RPMI medium. C Transwell migration assay was conducted with vehicle or BHB treatment for 3 days in C4-2. D EZH2 protein level was decreased with BHB treatment by western blot in C4-2, PC-3, and 22Rv1 cell lines. Vehicle (RPMI medium). E GSEA analysis of RNA-seq data in C4-2 cells after BHB treatment (left) and silencing EZH2 (right) revealed that genes of E2F targets, Myc targets, DNA repair, and G2M Checkpoint are enriched with both BHB treatment and EZH2 depletion. F Venn diagram with the shared signatures of BHB treatment and siEZH2. Number of positive enrichment (left) and negative enrichment (right) with BHB treatment and siEZH2. G C4-2 cells were treated with vehicle or BHB for three days. Cycloheximide was added to cells for the time indicated. The half-life of EZH2 was calculated. H MG132, a proteasome inhibitor, was used for 24 h. The cells were lysed and subjected to co-immunoprecipitation with anti-EZH2 antibody, followed by immunoblotting with anti-ubiquitin antibody. EZH2 levels were normalized to the loading controls β-actin. Ubiquitination levels were normalized to vehicle control. I Representative western blot obtained from pull-down assays using vehicle (RPMI medium) -conjugated or BHB-conjugated Sepharose beads. Binding proteins were eluted with BHB-containing buffer (0.1M Tris-HCl, BHB 0.5 mM). J GST and GST-EZH2 proteins were incubated with BHB-conjugated Sepharose beads for pull-down assays. Binding proteins were eluted with BHB-containing buffer (0.1 M Tris-HCl, BHB 0.5 mM). Student t-test. Mean ± SE. *p < 0.05, **p < 0.01, ***p < 0.001.

We then hypothesized that BHB might play a role in degrading EZH2 proteins. To test this hypothesis, we used cycloheximide (CHX) to block de novo protein synthesis and measured the half-life of EZH2 protein. As a result, BHB treatment shortened the half-life of EZH2 protein in C4-2 cells (Fig. 5G). We found that MG132, a proteasome inhibitor, could rescue BHB-mediated decrease of EZH2 protein levels (Fig. 5H), suggesting EZH2 undergoes degradation in a proteasome-dependent manner with BHB treatment. To further confirm that BHB induces EZH2 degradation, we employed MG132 to block EZH2 degradation, followed by co-IP with anti-EZH2 antibody to examine changes in EZH2 ubiquitination levels. As shown in Fig. 5H, BHB treatment dramatically increased ubiquitination of EZH2. These results demonstrate that BHB induced EZH2 degradation in a proteasome-dependent manner. Collectively, these data suggest that BHB could inhibit the progression of PCa via targeting EZH2.

BHB directly binds to EZH2 protein

Next, we were interested in investigating the mechanism by which BHB mediates EZH2 protein degradation. One possible hypothesis is that BHB binds to EZH2, further helps degrade EZH2 protein. To test this hypothesis, beads only or BHB-conjugated Sepharose beads were incubated with C4-2 cell lysates. After washing, the BHB-conjugated Sepharose beads binding proteins were eluted, followed by immunoblot analysis with anti-EZH2 antibody. As shown in Fig. 5I, EZH2 was bound to BHB. To examine if EZH2 directly binds to BHB, purified recombinant GST-only and GST-EZH2 protein were incubated with BHB-conjugated beads. As a result, BHB-conjugated beads did not bind to GST-only protein, but BHB-conjugated beads bound to GST-EZH2 protein, demonstrating that BHB binds to EZH2 directly (Fig. 5J). To further confirm this finding, conjugated beads were incubated with EZH2 protein at different temperatures, followed by western blot analysis. It was found that BHB-bound EZH2 was solubilized at high temperature (Supplementary Fig. 10), suggesting that ligand BHB interacts with target protein EZH2 and further degrades EZH2 protein level.

Ketone drink inhibits the progression of PCa in vivo

There are several forms of ketone supplements that could be used for our body including ketone salt, ketone esters, and ketone diol [37]. We chose BHB salt for our cell culture work since it is used by several researchers [38, 39], and it is easily dissolved in water. Ketone diol (1,3-butanediol), a precursor of BHB that is readily converted to BHB in our body through oxidation [40], was purchased from Amazon for in vivo experiment as it is already in the market and safe to use for in vivo study. Immunodeficient NCG mice were first castrated and tumor bits of LuCaP 35CR PDX were implanted subcutaneously. We utilized the LuCaP 35CR PDX model in vivo because it is enzalutamide-resistant, abiraterone-resistant CRPC model, and closely mimics the molecular and histological features of CRPC, providing a clinically relevant platform for evaluating therapeutic strategies in therapy-resistant advanced-stage disease. Castration was performed to simulate the androgen-depleted tumor environment characteristic of late-stage PCa. When the tumors were approximately 100 mm³ in size, mice were randomized and treated with either vehicle or two concentrations, 1:1 dilution with PBS and no-dilution, of ketone diol drink (100 μl x twice per day, 5 days per week). Ketone drink without dilution significantly reduced the tumor weight compared to vehicle group (Fig. 6A, B). Additionally, tumor volume was reduced in no-dilution group compared to vehicle (Fig. 6C). Ketone drink did not change the body weight (Fig. 6D). These results suggest that ketone drink could inhibit the progression of PCa in vivo without side effects. In addition, the undiluted ketone drink reduced EZH2 protein levels in tumor tissues, without affecting HMGCS2 protein levels (Fig. 6E, F).

Fig. 6. Ketone drink decreased the progression of PCa in vivo.

A–F Precastrated mice carrying LuCaP 35CR received vehicle or ketone drink (100 μl x twice per day) for 42 days (weekly schedule of five days on, two days off) (n = 5, each group). A The image of tumors was taken after sacrifice. B Tumor weight was measured after sacrifice. C Caliper measurements were taken twice a week to determine tumor volume. *P < 0.05, vehicle versus no dilution ketone drink. D Ketone drink does not change body wight in LuCap 35CR xenograft. E Tumor tissues were lysed and blotted for HMGCS2, EZH2, AR, H3 and β-actin. F HMGCS2, EZH2, and AR levels were normalized to the loading controls β-actin. Student t-test. Mean ± SE. *p < 0.05, **p < 0.01, ***p < 0.001.

Ketone drink increases the therapeutic efficacy of FDA-approved drugs in CRPC xenograft models

EPZ-6438 is an FDA-approved drug for patients with relapsed/refractory follicular lymphoma or metastatic epithelioid sarcoma. However, the therapeutic efficacy of EPZ6438 is limited because it only inhibits EZH2 enzymatic functions. We then wondered whether EPZ-6438 and BHB would have a synergistic effect on inhibiting PCa cell viability. As a result, EPZ-6438 and BHB synergistically reduced C4-2 PCa cell growth (Fig. 7A). Enzalutamide is an important hormonal therapy medication that blocks the androgen receptor (AR) and FDA-approved currently clinically well-used drug to treat advanced or metastatic PCa. BHB treatment significantly decreased AR protein level in a dose-dependent manner (Fig. 7B, C), but not mRNA level (Supplementary Fig. 5C (right)), consistent with previously reported EZH2/EED degrader astemizole [41]. Next, we tested the effect of the combination of enzalutamide and BHB on PCa growth in vitro. The combination of enzalutamide and BHB also had synergistic effects on the decrease of cell viability of C4-2 PCa cell lines (Fig. 7D). To assess the therapeutic effects of ketone drink and FDA-approved drugs (either EPZ-6438 or enzalutamide) in CRPC in vivo, we utilized a castration-resistant C4-2 xenograft mouse model. C4-2 cells responded to enzalutamide treatment, showing a decrease in AR protein levels and its downstream target, prostate-specific antigen (PSA), in vitro (Supplementary Fig. 11). We evaluated the effect of the combination of ketone drink and FDA-approved drugs (either EPZ-6438 or enzalutamide) on inhibiting the CRPC progression in vivo. Serum levels of BHB were successfully elevated in ketone drink alone as well as combination groups with ketone drink (Fig. 7E). Interestingly, ketone drink significantly enhanced the therapeutic efficacy of EZH2 inhibitor as well as enzalutamide in the xenograft model (Fig. 7F–H). Enzalutamide alone, EPZ-6438 alone, and ketone drink alone all reduced the tumor weight and tumor volume significantly compared to the vehicle group. Enzalutamide and ketone drink further significantly decreased the tumor weight compared to enzalutamide alone and ketone drink alone. Combination treatment of a ketone drink with EPZ-6438 achieved remarkable efficacy compared to the ketone drink alone and the EPZ-6438 alone group. Moreover, no additional adverse effects on the mice’s body weight were observed (Supplementary Fig. 12). Immunohistochemical (IHC) staining for Ki-67, a marker of cell proliferation, was conducted, and the combination of ketone drink and FDA-approved drugs had the greatest effects on decreasing tumor cell proliferation (Fig. 7I, J). These results demonstrated that ketone drink synergized the effects of the FDA-approved drugs to reduce the PCa tumor growth in vivo. Additionally, the ketone drink decreased EZH2 levels, and the combination of the ketone drink and FDA-approved drugs further reduced EZH2 levels in tumor tissues (Supplementary Fig. 13A, E). Meanwhile, the ketone drink decreased HMGCS2 levels regardless of the combination treatment (Supplementary Fig. 13C, G).

Fig. 7. Ketone drink increased therapeutic efficacy of EZH2 inhibitor and enzaluatamide in PCa xenograft model.

A, D Combination index plot of BHB and EPZ-6438 (A), BHB and Enzalutamide (D) in C4-2 cells. B, C Immunoblot analysis showed decreased-AR protein level with BHB treatment in C4-2 (B) and 22Rv1 (C). E-J Precastrated mice injected C4-2 cells received vehicle, enzalutamide(10 mg/kg/day), ketone drink (100 μl x twice per day), EPZ6438 (200 mg/kg/day), and combination by gavage for 4 weeks (five days per week) (n = 6/group). E Serum was collected when the mice were sacrificed, and the serum level of BHB was measured. F Caliper measurements were taken every week to determine tumor volume. G Tumor weight was measured after sacrifice. H The image of the tumors was taken after sacrifice. I Representative images of IHC staining for Ki-67. J Quantification of IHC images by ImageJ. Mean ± SE. Student t-test. *p < 0.05, **p < 0.01, ***p < 0.001.

Discussion

In this study, we discovered that EZH2 regulates HMGCS2, one of the rate-limiting steps of ketone metabolism enzymes, and HMGCS2 is negatively correlated with EZH2. Furthermore, we found that the downstream metabolite of HMGCS2, BHB, directly binds to EZH2 and decreases EZH2 protein level (Fig. 8). Moreover, a ketone drink as a monotherapy reduces the progression of PCa in vivo. Importantly, ketone drink increases the therapeutic efficacy of FDA-approved drugs, enzalutamide and EPZ-6438, in CRPC xenograft models.

Fig. 8. The proposed model of this study.

EZH2 represses HMGCS2, a rate-limiting enzyme of ketogenesis, epigenetically. Downstream metabolite, BHB, binds to EZH2 and induces degradation of EZH2 as a negative feedback loop, leading to decrease PCa progression. Created with BioRender.com.

It has been well-established in numerous previous studies that EZH2 is highly expressed in aggressive cancers and exerts a significant influence on cancer progression, including promoting cell proliferation, invasion, and metastasis [42, 43]. However, it is newly discovered that EZH2 regulates the ketogenesis metabolic process. This finding suggests the need for research into other aspects of EZH2 beyond its previously known effects. While the key mechanisms regulating EZH2 protein stability are post-translational modifications (PTMs), including phosphorylation, deacetylation, and methylation [44], it has not been reported that metabolites bind to EZH2 protein. Notably, it has not been reported that EZH2 regulates HMGCS2, nor that the downstream metabolite BHB binds to and degrades EZH2 in a negative feedback loop. To our knowledge, our research is the first to suggest that BHB could act as an EZH2 degrader, revealing a new dimension of BHB’s functionality. Our RNA-seq data revealed that BHB treatment showed regulatory gene-expression patterns similar to those seen with EZH2 inhibition and EZH2 depletion (Supplementary Fig. 6), indicating that BHB might serve as a natural EZH2 degrader. This mechanism could be more potent for advanced cancers than current EZH2 enzymatic inhibitors. While we found that the metabolite BHB interacts with EZH2 and induces EZH2 degradation in a proteasome-dependent manner, the specific mechanisms underlying EZH2 degradation remain unclear. It is possible that PTMs of proteins, such as ubiquitination, are involved in this process. Our mass spectrometry proteomic data showed that SUMO activating enzyme subunit 2 (UBA2) and Ubiquitin Conjugating Enzyme E2 D3 (UBE2D3) bind to BHB (data not shown), suggesting that these enzymes might play a role in the BHB-mediated degradation of EZH2. Future studies are needed to demonstrate the exact mechanisms of BHB-mediated EZH2 degradation, providing insight into a promising anti-cancer strategy.

Research related to ketogenesis is ongoing in various areas. The emerging evidence suggests that BHB could play a significant role in inhibiting colorectal cancer development and progression [38], highlighting its importance in cancer research. While preparing this manuscript, a very recent study reported that BHB might serve as an HDAC inhibitor and could enhance the therapeutic efficacy of immune checkpoint inhibitors in PCa. While Sean et al. demonstrated the effects of a ketogenic diet in enhancing PCa immunotherapies [45], our study focused on delaying PCa tumor progression using a ketone drink and FDA-approved drugs in a CRPC model, specifically via the EZH2-HMGCS2-BHB axis. We found outthat metabolite BHB binds to EZH2, which is newly discovered, and it is also possible that BHB might bind to other proteins, leading to delaying PCa tumor progression.

Importantly, we discovered that combining a ketone drink with enzalutamide, the current drug for metastatic CRPC, resulted in a significant decrease in PCa tumor growth. These findings propose a novel therapeutic approach for patients with advanced PCa. Using ketones in the form of beverages, rather than drugs, offers significant safety advantages. This approach bypasses the extensive regulatory and safety testing required for new drug development. In our study, serum BHB levels increased from 100 µM to 600 µM, which is lower than the 2-3 mM levels typically seen with a continuous ketogenic diet [45, 46]. This suggests that ketone drinks might reduce side effects associated with high BHB spikes. Our findings, combined with the Lu group’s discoveries, strongly suggest that BHB/ketone drinks are a novel therapeutic tool for patients with advanced PCa. Furthermore, evidence of their efficacy in reducing primary PCa tumors indicates their potential effectiveness against various types of cancer. Future studies are needed to determine the optimal concentration of ketone drinks and the resultant blood levels of BHB for maximum therapeutic effect without adverse side effects.

Herein, we observed that EZH2 epigenetically regulates HMGCS2, and downstream metabolite, BHB, induces the degradation of EZH2 in a negative feedback loop (Fig. 8). Furthermore, a ketone drink enhances the therapeutic efficacy of FDA-approved drugs in CRPC xenograft models. This finding underscores the significance of EZH2 and the ketone drink in PCa. However, there are several limitations to this study. For instance, there is no dataset demonstrating serum BHB levels in PCa patients. Additionally, the metabolic rewiring involving EZH2 alone, ketone body itself, and the combination of ketone bodies with current drugs remains to be fully elucidated. Despite these limitations, our study highlights the interplay between metabolism and epigenetic regulation in cancer progression, suggesting potential therapeutic strategies targeting metabolic and epigenetic pathways. While BHB (or ketone drinks) might be effective, clinical validation is needed to confirm their efficacy in PCa treatment.

Materials and Methods

Cell lines and reagents

The human PCa cell lines C4-2, PC-3, DU145, LNCaP, and 22Rv1 were purchased from the American Type Culture Collection (ATCC, VA, USA). The above cell lines were grown in RPMI 1640 medium (Gibco, NY, USA) supplemented with 10% fetal bovine serum (FBS). VCaP, HEK293T, and MDA-MB-231 cells were obtained from ATCC and maintained in DMEM medium (Gibco) with 10% FBS. All cell lines were authenticated by short tandem repeat (STR) genotyping and were used within 2 months of continuous culturing. Cells were mycoplasma-negative in routine tests. When indicated, cells were treated with EZP-6438 (Selleckchem, TX, USA), GSK126 (Cayman Chemical, MI, USA), enzalutamide (Selleckchem), and DL-β-Hydroxybutyric acid sodium salt (Sigma, MO, USA). MS1943 and MS8815 were synthesized by Dr. Jian Jin at the Icahn School of Medicine at Mount Sinai. For the xenograft study, a ketone drink (Ketone IQ, R-1,3-Butanediol) was purchased from H.V.M.N (CA, USA).

Antibodies

Primary antibodies used in this study are listed in Supplementary Table 1.

Primers

RT-qPCR and ChIP-qPCR are listed in Supplementary Table 2.

Transfection and transduction of cell lines

All the silencer siRNAs used in this study were purchased from Thermo Fisher (MA, USA) (siEZH2-1: s4916, siEZH2-2: s4918). Lipofectamine RNAiMAX (Invitrogen, MA, USA) was utilized for siRNA transfection according to the manufacturer’s protocol.

Lentiviral shRNA vector was purchased from Sigma (shEZH2: TRCN0000286227). EZH2 depletion utilized doxycycline inducible lentiviral human shRNA, which targeted EZH2 plasmids were generated (shEZH2-1: TRCN0000286227, shEZH2-2: TRCN0000293738). Lentivirus was packaged by co-transfecting the shRNA construct with the helper plasmids pVSVG and psPAX2 into HEK293T cells. The viruses were harvested at 48 h post-transfection. Transduction of C4-2 cells took place for 48 h with 10 ug/mL polybrene, followed by puromycin selection (3 mg/mL).

Cells were transiently transfected with either pcDNA3.1 empty vector (EV) or HMGCS2 ORF clone (GenScript, NJ, USA) using Lipofectamine 3000 following the manufacturer’s protocol (Invitrogen).

RNA isolation and RT-qPCR

RNA was extracted from cells using the RNeasy Plus Kit (QIAGEN, Germany). cDNA was synthesized from 2 μg of total RNA using the High-Capacity cDNA Reverse Transcription Kit with RNase Inhibitor (ThermoFisher Scientific). Each cDNA sample was amplified using PowerUP SYBR Green Master Mix (Applied Biosystems, CA, USA) using QuantStudio 6 Flex real-time PCR systems (Applied Biosystems). The 2−ΔΔCT method was used to calculate the relative gene expression levels, and β-actin was used as an endogenous control to normalize each sample. The primers are listed in Supplementary Table 1.

Western blot analysis

Protein samples were separated electrophoretically by SDS–PAGE, and semi-dry transferred to polyvinylidenedifluoride membranes (Biorad, CA, USA). The membranes were blocked for 60 min in Tris-buffered saline–Tween 20 with 5% nonfat milk. Thereafter, immunoblotting was performed with primary antibodies overnight. After washing for three times, the membranes were incubated with goat anti-mouse/rabbit IgG (H + L)–horseradish peroxidase secondary antibody (GenDEPOT (TX, USA), 1:2,000 dilution) for 1 h. The signals were developed using western ECL substrate (Bio-Rad), and images were scanned using a Bio-Rad imaging system. The relative protein level was evaluated using ImageJ software.

EZH2 rescue assay

The EZH2 plasmid was generously provided by Dr. Yu. For rescue assay, shEZH2 lentivirus (shEZH2: TRCN0000286227) was selected because it targets the 3’UTR region of endogenous EZH2, and ectopic EZH2 expression remains unaffected. Samples were collected 48 h after transfecting EZH2 plasmids into EZH2-deficient C4-2 cells, and RT-qPCR was performed to measure mRNA levels. Doxycycline-inducible EZH2-depletion cells were generated and treated with doxycycline (1 µg/mL) for 4 days, after which the media were replaced with fresh medium. Cell lysates were harvested for immunoblot analysis after 4 days of maintenance with fresh media (8 days from the start of doxycycline treatment) and again after 26 days of maintenance with fresh media (30 days from the start of doxycycline treatment).

Co-immunoprecipitation (co-IP)

C4-2 cells were treated with BHB or vehicle (RPMI medium) for three days. 10 μM of MG132 (Calbiochem, CA, USA) was used at two days post-treatment of BHB for 24 h. Cells were washed with cold PBS three times and lysed in NP-40 lysis buffer (Thermo Fisher Scientific) with protease and phosphatase inhibitor cocktails (Thermo Fisher Scientific). The lysate was kept on ice for 15 min followed by sonication. Insoluble material was removed by centrifugation. Lysates were pre-incubated with Dynabeads protein A/G (Invitrogen) to eliminate nonspecific binding. Then, anti-EZH2 antibodies were mixed into the lysates with newly added Dynabeads and incubated at 4 °C overnight. The immune complexes were collected using a magnetic separator and washed three times with lysis buffer. To denature proteins, beads were added to 2× reducing SDS sample buffer (Bio-Rad) and heated at 95 °C for 10 min. Protein samples were subjected to WB assay for further analysis.

Analysis of EZH2 protein stability by cycloheximide (CHX) treatment

C4-2 cells were treated with BHB or vehicle for three days. Cell lysates were collected at specific times after CHX addition (10 μg/ml). Total cell lysates were blotted for EZH2 and H3, while β-actin served as a loading control. The amount of the EZH2 protein at the zero time in each group was considered as 1.

Chromatin immunoprecipitation (ChIP)-qPCR analysis

The ChIP experiment was performed using the SimpleChIP Plus Sonication ChIP kit (Cell Signaling Technology) with the procedure provided by the manufacturer. Cells were cross-linked using paraformaldehyde solution (Invitrogen) and terminated with glycine solution. A chromatin fragment at an average size of 200 bp was obtained by cell lysis and sonication using a Diagenode bioruptor. DNA was isolated from samples by incubation with the anti-H3K27me3 (Cell Signaling Technology, MA, USA) antibody at 4 °C overnight, followed by washing and reversal of crosslinking. ChIP-qPCR assay was performed to monitor the enrichment of H3K27me3 marks at 2 kb upstream of the promoter regions of HMGCS2 in vehicle control and GSK126-treated C4-2 cells. Four pairs of primers (F1 (transcription start site~−0.5 kb), F2 (−0.5 ~ −1.0 kb), F3 (−1.0 ~ −1.5 kb), F4 (−1.5 ~ −2.0 kb)) at different regions were used to amplify fragments inside the HMGCS2 promoter region. Primers are listed in Supplementary Table 2. Immunoprecipitated DNA was calculated as a percentage of input DNA.

Pull-down assay

First, BHB-conjugated beads were synthesized according to a previous literature [47]. Briefly, (S)-(+)-4-Amino-3-hydroxybutyric acid (Sigma) was dissolved in coupling buffer (0.2 M NaHCO3, 0.5 M NaCl, pH 8.3), and prewashed NHS-activated Sepharose beads (Cytiva, MA, USA) were mixed and incubated overnight at 4 °C. After blocked and washed three times, BHB-conjugated beads were used for pull-down assay with either cells or a recombinant protein.

C4-2 cells were washed with PBS and homogenized with a syringe in binding buffer. After centrifugation, the supernatant was loaded into BHB or vehicle (RPMI medium) conjugated Sepharose beads. After three times of washing, the samples were eluted with BHB containing elution buffer (0.1 M Tris-HCl, BHB 0.5 mM), followed by western blot.

Recombinant GST protein (Abcam) and GST-EZH2 (BPS Bioscience, CA, USA) protein were incubated with BHB-conjugated Sepharose beads for two hours at room temperature. After three times of washing, the samples were eluted with BHB-containing elution buffer, followed by western blot.

Immunohistochemistry (IHC) staining

Tumor tissues were harvested after the mice were sacrificed. Part of the tumor tissues were fixed in 10% neutral-buffered formalin, processed, and embedded in paraffin. Ki67 (1:800) staining was conducted using the Immunoperoxidase Secondary Detection System (EMD Millipore), followed by Hematoxylin counterstaining. Detection was developed by secondary antibodies (EMD Millipore) and visualized with microscopes. The primary antibody used for IHC was listed in Supplementary Table 1.

RNA-Seq and data analysis

Total RNAs were extracted using the RNeasy Plus Kit (QIAGEN) and sequenced by the DNBseq-PE100. The RNA-seq reads were mapped to the human genome GRCh38 and assigned to the reference genes using HISAT2 v2.1.0. Read counts for each gene were calculated by featureCounts v1.6.1 [48]. Differential gene expression analysis was performed between the two groups of cells by R packages edgeR and limma to determine the log₂ fold change of each gene as the ranking metric [49, 50]. Adjusted P < = 0.05 was set up as a cutoff to define differential expressed genes (DEGs). GSEA (version 2.2.0) was used to analyze the BHB-altered gene signature from C4-2 cells treated with vehicle control and BHB or siControl and siEZH2. Hallmark gene signatures were used in the analysis, and the source data are provided with this paper. These results were used to determine overlapping hallmarks by creating Venn diagrams for both negative and positive enrichment scores [51]. For clustering analysis, we used a hierarchical clustering method with Spearman correlation distance to cluster samples based on the log-scaled FPKM. The data’s expression matrix was processed using the R software “Limma” for differential expression analysis. The significantly differentially expressed genes (DEGs) (|logFC| > 1, p-value < 0.05) were then visualized by a heatmap(“pheatmap”). GSK126, shEZH2_1/2, GSK126_con, shCTRL_1 data sets were used from GSE124268.

Bioinformatic analysis for public data

Normalized expression values of individual genes for PCa and BCa patients from the multiple cohorts were downloaded for analysis. TCGA-PRAD RNA-Seq data were downloaded from https://portal.gdc.cancer.gov. We obtained the RNA count from 101 metastatic PCa in the WC-SU2C cohort from the Genotypes and Phenotypes (dbGaP) database under accession code phs000915.v2.p2 [26]. RNA-seq data from Taylor et al. (GSE21032) [27] was downloaded from Gene Expression Omnibus (GEO). Data of Kaplan-Meier Survival Analysis were downloaded from https://betastasis.com, PCa, Taylor et al.) [27] and GEPIA (PCa, TCGA-PRAD) [52], and https://kmplot.com/analysis (BCa) [29]. EZH2 and HMGCS2 expression levels by scRNA-seq were obtained from https://pcatools.shinyapps.io/HuPSA-MoPSA/ [28]. Expression of HMGCS2 in BCa based on sample types and individual cancer stages was downloaded from https://ualcan.path.uab.edu/ [30, 31]. Co-expression plots of EZH2 and HMGCS2 in PCa (TCGA) [53] and BCa (METABRIC) [54, 55] were obtained from https://www.cbioportal.org.

Cell function assays

For the cell growth assay, cells were seeded in 96-well plates and treated with BHB and enzalutamide (or EPZ-6438), either alone or in combination, at concentration gradients for 72 h. Bioluminescence was measured to quantify cell viability using a CellTiter-Glo Luminescent Cell Viability Assay kit (Promega, WI, USA), and plates were read on a Tecan plate reader. Combination index (CI) values were calculated by Calcusyn (Biosoft, Ferguson, MO).

For the wound-healing assay, cells were cultured in a 35-mm dish with a 3-well culture insert (Ibidi) and grown to 90% confluency. After carefully removing the inserts, the dishes were refilled with serum-free RPMI 1640 medium and incubated in a humidified atmosphere at 37 °C. Images were captured under a microscope at the specified time points, and the percentage of wound closure was quantified using ImageJ software. Wound closure was calculated using the formula:

$$\mathrm{Wound}\mathrm{closure}\left(\%\right)=\left[\left(\mathrm{wound}\mathrm{area}\mathrm{at}\mathrm{time}0\right)-\left(\mathrm{wound}\mathrm{area}\mathrm{at}\mathrm{time}\mathrm{x}\right)\right]/\left(\mathrm{wound}\mathrm{area}\mathrm{at}\mathrm{time}0\right)\times 100.$$

For Boyden chamber migration assay, the migratory effect of PCa cells was assessed by their ability to pass through transwell inserts (Millipore). Cells diluted in 300 μl of serum-free RPMI medium were seeded into the upper compartments of the chambers. Meanwhile, the lower compartments of the chambers were filled with 900 μl of RPMI medium with 10% FBS. After 24 h, cells that had migrated to the lower surface of the chamber were fixed in methanol, stained with 0.1% crystal violet (Sigma), and subjected to microscopy inspection. The number of cells was expressed as the average number of cells counted in five random fields per filter.

Targeted metabolomics

After transfection of C4-2 cells with siEZH2 and siControl, the cells were harvested in 80% methanol. Cell lysates were incubated in -80 °C freezer for 5 mins and then vortexed 60 seconds at room temperature. Incubation and vortex steps were repeated two more times. Cell lysates were stored in -80 °C freezer for protein precipitation. After centrifugation, the supernatant was dried under nitrogen gas, and the pellets were used for protein analysis. Dried metabolites were reconstituted in 60% acetonitrile, and samples were analyzed by High-Performance Liquid Chromatography and Triple-quadruple Mass Spectrometry and Tandem Mass Spectrometry (HPLC-MS/MS). Specifically, the system consisted of a TSQ (Thermo) in line with an electrospray source (ESI) and a Vanquish (Thermo) UHPLC consisting of a binary pump, degasser, and auto-sampler outfitted with an XBridge C18 column (Waters, dimensions of 2.1 mm × 50 mm and a 3.5 μm particle size). Elution was at isocratic mode with mobile phase A (10 mM ammonium formate, 0.1% formic acid in water) and mobile phase B (acetonitrile) at A: B = 80: 20 (v/v) with 0.15 mL/min. In negative mode, the capillary of ESI source was set to 300 °C, with sheath gas at 35 arbitrary units, auxiliary gas at 3 arbitrary units and the spray voltage at 3 kV. A selective reaction monitoring (SRM) of the protonated precursor ion and the related product ion for BHB (m/z 103 → 59) was monitored. Peak area was integrated, and data acquisition and analysis were carried out by Xcalibur 4.1 software and TraceFinder 4.1 software, respectively (both from Thermo Fisher Scientific). The data were normalized to β-actin level of each sample.

Tumor xenograft and drug treatments

All procedures involving mice were performed in compliance with ethical regulations and with the approval of the Northwestern University Institutional Animal Care and Use Committee (IS00017714). Four-week-old NCG male mice were purchased from Charles River Laboratories and castrated. The LuCaP 35CR PDX was provided by E. Corey (University of Washington). One week later of castration, mice were implanted subcutaneously with LuCaP 35CR tumor bits. Mice were divided into treatment groups based on body weight when the tumor volume reached 100 mm³. The sample size for each group (n = 6) was chosen based on experimental feasibility and ethical considerations. Mice were given vehicle (PBS), 1:1 dilution of ketone drink with PBS, and no dilution of ketone drink (R-1,3-butanediol, H.V.M.N., 100 μl x twice by oral gavage daily). Blinding was not used in the animal studies. The tumor volume was measured with calipers using the formula (L*W*W/2). Tumor volume and body weight were measured twice weekly. After 42 days of treatments, mice were euthanized, tumors were excised, and weighed.

To evaluate the efficacy of the ketone drink and current drugs, pre-castrated male mice were anesthetized using 2% isoflurane (inhalation), and 3 × 10⁶ C4-2 cells were suspended in 100 μl of FBS with 50% Matrigel and implanted subcutaneously into the dorsal flank of the mice. When tumor volume reached 100 mm³, mice were randomly divided into six different groups and treated with 200 μl of vehicle control (0.5% Sodium carboxymethyl cellulose, 0.1% tween 80 in water), EPZ-6438 (200 mg/kg/day; 100 μl x twice by oral gavage daily), enzalutamide (10 mg/kg/day; 100 μl x twice by oral gavage daily), ketone drink (R-1,3-butanediol, H.V.M.N., 100 μl x twice by oral gavage daily) or in combination. Animals were treated by oral gavage on a weekly schedule of 5 days on, 2 days off. Tumor volume and body weight were measured weekly. After 28 days of treatments, mice were euthanized, tumors were excised, and weighed. Blood was obtained by a cardiac puncture when sacrificed. Serum was separated by centrifugation at 1,000 g for 10 min at 4 °C, and supernatant was transferred to −80 °C until BHB measurements. BHB was measured using a fluorescence assay kit (Cayman). The effects of drug treatment in suppressing target pathways were examined by IHC analysis.

Statistics and reproducibility

Statistical analysis was performed using GraphPad Prism (v.9.0) as described in the figure legends for each experiment. Data were presented as the mean ± s.e., and P values were determined by unpaired two-tailed t test. Significance was set at P < 0.05. No statistical methods were used to predetermine sample sizes, but our sample sizes are similar to those reported in previous studies. The results were reproducible and conducted with established internal controls. When feasible, experiments were repeated three or more times and yielded similar results. All samples that met proper experimental conditions were included in the analysis.

Author contributions

CY: Conceptualization, Writing – original draft, Methodology, Investigation, Formal analysis, Validation, Visualization, Writing-Review & Editing. RAS: Software, Investigation, Formal analysis, Data Curation, Visualization. RW: Investigation, TYW: Software, Investigation, Formal analysis, Data Curation, Visualization. XL: Software, Investigation, Formal analysis, Data Curation, Visualization. QL: Investigation. YR: Software, Investigation, Formal analysis, Data Curation, QM: Investigation. YYang: Investigation, Funding acquisition. XinZ: Investigation. YX: Investigation. XY: Investigation. XiaoyuZ: Methodology. JJ: Methodology. XD: Methodology. YYi: Investigation, Methodology. Funding acquisition. RY: Supervision, Project administration, Funding acquisition, Writing-Review & Editing. QC: Supervision, Resources, Project administration, Funding acquisition, Conceptualization, Writing-Review & Editing.

Funding

This work was supported in part by grants from Department of Defense W81XWH-20-1-0504 (Q.C.), HT9425-23-1-0661 (Y.Yi.), W81XWH-21-1-0146 (Y.Yang.), HT9425-23-1-0491 (Y.Yang.), NIH/NCI R01CA256741 (Q.C.), R01CA285684 (Q.C), R01CA259388 (R.Y.), R35GM142441 (R.Y.), Prostate SPORE P50CA180995 Development Research Program (Q. C., R.Y.), and the Polsky Urologic Cancer Institute of the Robert H. Lurie Comprehensive Cancer Center of Northwestern University at Northwestern Memorial Hospital (Q.C., R.Y.).

Data availability

The RNA-Seq data generated by this study have been deposited in the GEO under accession no. GSE269208 and GSE269259. All other data supporting the findings of this study are available from the corresponding authors upon reasonable request.

Competing interests

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: JJ is a cofounder and equity shareholder in Cullgen, Inc., a scientific cofounder and scientific advisory board member of Onsero Therapeutics, Inc., and a consultant for Cullgen, Inc., EpiCypher, Inc., Accent Therapeutics, Inc, and Tavotek Biotherapeutics, Inc. The Jin laboratory received research funds from Celgene Corporation, Levo Therapeutics, Inc., Cullgen, Inc. and Cullinan Oncology, Inc. Other authors declare no conflicts of interest.

Supplementary information

The online version contains supplementary material available at 10.1038/s41389-025-00567-0.