Glycolysis reprogramming predicts poor prognosis and drives therapy resistance via CLN6 in lethal prostate cancer

Abstract

Lethal prostate cancer is marked by tumor heterogeneity and resistance to androgen receptor signaling inhibitors (ARSIs). In this study we identify glycolysis as a driver of disease progression and therapy resistance. Using single-sample gene set enrichment analysis (ssGSEA) on the SU2C cohort, we demonstrate that elevated glycolysis activity is associated with poor progression-free and overall survival. The glycolysis-based prognostic score (GLY score) is derived from the HALLMARK_GLYCOLYSIS gene set which includes CLN6, SDHC, B4GALT2, RPE, NANP, and KIF20A, via LASSO-Cox regression. The GLY score effectively stratifies risk in the SU2C and WDCT cohorts, with higher scores predicting worse outcomes and increased SYNE1 mutation frequency. Pan-cancer analysis across TCGA datasets confirm its prognostic value. In vitro, enzalutamide-resistant prostate cancer cell lines exhibit heightened glycolysis, and 2-DG inhibition reverses this effect, restoring drug sensitivity. CLN6 knockdown reduces glycolytic activity and cell proliferation. The GLY score offers robust prognostic value, and CLN6 represents a promising therapeutic target for precision medicine in lethal prostate cancer.

Introduction

Prostate cancer ranks as the second most prevalent malignancy among men globally, contributing to over 10% of cancer-related mortality and standing as the fifth leading cause of cancer deaths worldwide [1]. In its localized stage, prostate cancer typically exhibits indolent growth, with patients often achieving favorable outcomes through surgical resection or radiotherapy. Over the recent decades, advancements in diagnostic techniques and therapeutic strategies have significantly enhanced the management of lethal prostate cancer, particularly through androgen deprivation therapy (ADT) and AR signaling inhibitors (ARSIs), including abiraterone, enzalutamide, apalutamide, and darolutamide, which have conferred substantial survival benefits [ 2, 3] .

Nevertheless, to further optimize patient prognosis, a deeper understanding of the complex interplay between molecular drivers and dysregulatory mechanisms that propel lethal prostate cancer progression is essential. The marked heterogeneity of lethal prostate cancer, coupled with the diverse clinical presentations observed across patients, poses challenges to such investigations. However, by focusing on prevalent genetic mutations and regulatory pathways, research findings are more likely to translate effectively into clinical practice, offering broader applicability and impact [4].

Metabolic reprogramming is a hallmark of tumor heterogeneity and therapy resistance, with glycolysis playing a pivotal role [ 5, 6] . In contrast to normal cells, which primarily rely on oxidative phosphorylation (OXPHOS) for energy production, tumor cells exhibit a pronounced preference for glycolysis. Despite generating less ATP per glucose molecule, glycolysis offers a faster rate of energy production, enabling tumor cells to meet the heightened metabolic demands of proliferation and survival [ 7, 8] . For instance, in pancreatic cancer, activation of the GFRA2-RET signaling axis by neurturin enhances glycolysis through phosphorylation of hexokinase 2, thereby driving tumor progression [9]. Similarly, in colorectal cancer, ATF4 promotes tumor growth by upregulating glycolytic activity [10]. In lethal prostate cancer, emerging evidence highlights the role of glycolysis in mediating treatment resistance [11]. However, the intricate regulatory networks governing glycolysis necessitate further investigation to fully elucidate its mechanistic contributions to cancer progression and resistance.

In the present study, we initially identified glycolysis as a critical pathway associated with poor prognosis in lethal prostate cancer through pathway enrichment analysis and in vitro assays. Subsequently, we developed a scoring system based on characteristic genes of the glycolysis pathway to predict therapy response and prognosis in lethal prostate cancer, which was validated using multi-omics data. Furthermore, we identified CLN6 as a novel hub gene within the glycolytic pathway of lethal prostate cancer and investigated its role through in silico and in vitro assays. Our findings highlight CLN6 as a potential therapeutic target for intervening in glycolysis-driven lethal prostate cancer.

Materials and Methods

Data processing

RNA expression profiles, mutation data, and clinical information for the SU2C and WCDT cohorts were retrieved from cBioPortal and the PCaDB database ( http://bioinfo.jialab-ucr.org/PCaDB/) [12]. Pan-cancer data from The Cancer Genome Atlas (TCGA) were sourced from the University of California, Santa Cruz (UCSC) Xena platform ( https://xenabrowser.net/datapages/). Data processing was conducted in accordance with methodologies outlined in our previously published studies [ 12, 13] .

Pathway enrichment analysis

Hallmark gene sets were obtained from the Molecular Signatures Database (MSigDB, updated 2021). Single-sample gene set enrichment analysis (ssGSEA) was performed using the “GSVA” R package (version 1.46.0) to calculate enrichment scores for each gene set across all samples [ 14, 15] . Associations between ssGSEA scores and clinical follow-up data were evaluated using the “survminer” (version 0.4.9) and “survival” (version 3.5-7) R packages.

Construction of the GLY score

Utilizing the HALLMARK_GLYCOLYSIS gene set, a six-gene expression signature was developed from the SU2C dataset with progression-free survival (PFS) as the endpoint, employing the least absolute shrinkage and selection operator (LASSO) regression method via the “glmnet” package (version 4.1-8) [ 16, 17] . The resulting gene expression signature was used to formulate the GLY score ( Equation 1).

$$\begin{array}{cr}\text{GLY score =}{\sum }_{i=1}^N\text{Coefficient}\times G\text{ene expression}& \text{(1)}\end{array}$$

As outlined in the formula, N represents the total number of genes included in the model, Coefficient denotes the LASSO regression coefficient assigned to each gene, and Gene expression refers to the expression level of each respective gene.

Tumor mutation analysis

The somatic mutation profiles were generated using the “maftools” R package [18], and the top 20 mutated genes were visualized with the “ComplexHeatmap” R package [19]. The GLY score was used to stratify samples from both datasets into high- and low-score groups, and differences in mutation frequencies between these groups were evaluated. Additionally, GLY scores were compared between the mutated and non-mutated groups.

In vitro assays

The establishment of enzalutamide-resistant PCa cell lines and the associated cellular functional experiments have been detailed in our previous studies [ 20, 21] . In brief, LNCaP and C4-2B cell lines were purchased from BeiNa Biotechnology Company (Beijing, China) and cultured in RPMI-1640 medium (BC-M-017-500 mL; Bio-Channel, Bury St Edmunds, UK) supplemented with 10% fetal bovine serum (BC-SE-FBS07; Bio-Channel) and 1% penicillin/streptomycin (15140122; Gibco) in a humid incubator with 5% CO ₂ at 37°C. LNCaP and C4-2B cells were cultivated in RPMI-1640 medium with 10 μM and 25 μM enzalutamide (S1250; Selleck, Houston, USA) for 3 months to generate enzalutamide-resistant cell lines (LNCaP_ENZR and C4-2B_ENZR). A total of 5000 cells per well were seeded into a 96-well plate, and a CCK-8 assay kit (MA0218; MeilunBio, Dalian, China) was used to examine the sensitivity to enzalutamide in LNCaP, C4-2B, LNCaP_ENZR and C4-2B_ENZR cells after treatment with gradient concentrations of enzalutamide (0 μM, 2.5 μM, 5 μM, 12.5 μM, 25 μM, and 50 μM). A total of 5000 cells per well were seeded into a 6-well plate, and a plate colony formation assay was used to examine the sensitivity of LNCaP, C4-2B, LNCaP_ENZR, and C4-2B_ENZR cells to enzalutamide treatment at 10 μM and 25 μM for 12 days. The lentivirus expressing short hairpin RNA (shRNA) targeting CLN6 was constructed by GENECHEM Co., Ltd ( Genechem, Shanghai, China). Western blot analysis was used to evaluate CLN6 expression in BPH-1, PC-3, DU145, 22Rv1, and C4-2 cells, as well as in LNCaP and C4-2B WT cells and LNCaP_ENZR and C4-2B_ENZR. The CLN6-targeting shRNA sequence was 5′-GCTGCTTTACTGCCTCTAAAG-3′, and a non-targeting shRNA (NC) sequence was 5′-TTCTTCGAACGTGTCACGT-3′.

Glucose uptake assay

Glucose uptake was assessed using the Glucose Uptake-Glo™ Assay kit (J1341; Promega, Madison, USA). Briefly, 2000 cells were seeded per well in a 96-well plate and washed twice with phosphate-buffered saline (PBS). Then, 50 μL of 1 mM 2-deoxyglucose (2-DG) solution was added to each well and gently mixed. After incubation at room temperature for 10 min, 25 μL of stop buffer was added, followed by gentle mixing. Then 25 μL of neutralization buffer was added to each well with mixing and 100 μL of 2DG-6P detection reagent was added. After incubation at room temperature for 30 min to 1 h, luminescence signals were recorded using a luminometer with an integration time of 0.3 to 1 s.

Lactate production assay

Lactate production was measured using the Lactate Assay Kit (A019-2-1; Nanjing Jiancheng Bioengineering Institute, Nanjing, China). The procedure was as follows: (1) 100 μL of cell culture supernatant was collected. (2) The enzyme working solution was prepared by diluting the enzyme stock solution at a 1:100 ratio with enzyme diluent. (3) Chromogenic working solution was prepared by dissolving chromogenic powder in 6 mL of chromogenic diluent. (4) A 12-well plate was divided into three groups, each containing 20 μL of distilled water, 3 mM standard solution, or test sample, with three replicates per group. (5) Each well received 1 mL of enzyme working solution and 200 μL of chromogenic working solution, was thoroughly mixed, and incubated in a 37°C water bath for exactly 10 min. (6) The reaction was terminated by adding 2 mL of stop solution to each well, and the absorbance was measured at 530 nm using a full-wavelength microplate reader.

Seahorse glycolysis stress test

The Seahorse XF Glycolysis Stress Test Kit (103020-100; Agilent, Santa Clara, USA) was used to evaluate glycolytic activity. The protocol included the following: (1) Hydration of the sensor cartridge plate with assay medium and incubation overnight at 37°C. (2) Preparation of Seahorse XF cell culture microplates, with cell confluency verified under a microscope. (3) Test reagent supplemented with 2 mM glutamine was prepared, the culture medium in the microplate was replaced by this test reagent, and the plate was incubated in a 37°C non-CO ₂ incubator for 1 h. (4) Stock solutions of glucose (100 mM), oligomycin (100 μM), and 2-DG (500 mM) were prepared. (5) The hydrated sensor cartridge was loaded with 56 μL, 65 μL, and 69 μL of glucose, oligomycin, and 2-DG solutions, respectively, in ports A, B, and C. (6) Seahorse XF software was used, the glycolysis stress test protocol was selected, and the sensor cartridge with added compounds was calibrated on the instrument tray. (7) After calibration, the hydrated plate was replaced by a cell culture microplate, and the assay was initiated.

Statistical analysis

Bioinformatics analyses were conducted using R software (version 4.3.0; The R Project for Statistical Computing, Vienna, Austria). For progression-free survival (PFS) and overall survival (OS), survival analyses were performed using the Kaplan–Meier method with log-rank tests and Cox proportional hazards regression models. Univariate and multivariate Cox regression analyses were conducted to identify independent prognostic factors. All survival analyses were performed using the survival (version 3.7.0) and survminer (version 0.5.0) packages in R. Tumor mutation burden was evaluated with the “ComplexHeatmap” package. Cell functional assay data were analyzed with GraphPad Prism (version 9.0; GraphPad Software, La Jolla, USA). Data from two groups were expressed as the mean ± standard deviation (SD) and compared using a two-sided Student’s t-test. For comparisons involving three or more groups, one-way analysis of variance (ANOVA) was applied. Statistical significance was defined as P  < 0.05.

Results

Glycolysis is a critical pathway associated with poor prognosis in lethal prostate cancer

As illustrated in Figure 1, our study workflow commenced with the assessment of pathway activity in SU2C lethal prostate cancer samples using ssGSEA based on the Hallmark pathway collection, with progression-free survival (PFS) as the primary endpoint. Integrated Cox regression analysis revealed that glycolysis exhibited the most statistically significant value among the top five pathways with a hazard ratio (HR) greater than 1 ( Figure 2A and Supplementary Table S1). KM survival analysis demonstrated that higher glycolytic pathway activity was consistently associated with poorer outcomes for both PFS and OS in lethal prostate cancer from the SU2C dataset ( Figure 2B–E). These findings underscore the pivotal role of glycolysis as a key determinant of adverse prognosis in lethal prostate cancer.

Figure 1 .

Flowchart of this study

Glycolytic activity was assessed in public prostate cancer cohorts, followed by construction and validation of a glycolysis-based prognostic score. Associations with genomic alterations and pan-cancer prognosis were analyzed, and in vitro experiments were conducted to validate the functional role of glycolysis and key regulators in enzalutamide-resistant prostate cancer.

Figure 2 .

Glycolysis pathway activity is associated with poor prognosis in lethal prostate cancer

(A) Cox regression analysis of pathway enrichment scores for progression-free survival (PFS) in the SU2C cohort. Hazard ratios (HR) and P-values were calculated using the Wald test, the top five pathways with HR > 1 and statistically significant P-values are shown. (B,C) Risk plot analysis of the SU2C cohort based on pathway enrichment scores for progression-free survival (B) and overall survival (C). Samples were stratified into high- and low-risk groups. Heatmaps displaying the expression profiles of the top five pathways. (D,E) Kaplan-Meier survival analyses of the SU2C cohort stratified by glycolysis pathway scores. Progression-free survival (D) and overall survival (E) were compared between high and low glycolysis pathway scores using the log-rank test.

Inhibiting glycolysis enhances the therapeutic efficacy for lethal prostate cancer

Enzalutamide is one of the first agents used to treat lethal prostate cancer. To investigate the relationship between glycolysis and treatment response, we established enzalutamide-resistant prostate cancer cell line models ( Figure 3AD). Following successful induction of resistance, both the LNCaP_ENZR and C4-2B_ENZR cell lines exhibited reduced sensitivity to enzalutamide. Subsequently, we assessed glucose uptake and lactate production rates in these resistant cell lines, revealing significantly elevated levels of both parameters in the resistant strains compared to their WT controls ( Figure 3E,F). To further validate the role of glycolysis, we employed the glycolytic inhibitor 2-DG. Treatment with increasing concentrations of 2-DG significantly suppressed proliferation in LNCaP_ENZR and C4-2B_ENZR cells in a dose-dependent manner, with the inhibitory effect becoming more pronounced over time, peaking at 72 h ( Figure 3G,H). In CCK-8 assays, co-treatment with enzalutamide and 2-DG markedly reduced proliferative capacity in both resistant cell lines compared to enzalutamide alone ( Figure 3I,J). Consistent results were obtained in colony formation assays, where 2-DG co-treatment similarly impaired clonogenic potential ( Figure 3K,L). These findings suggest that enzalutamide-resistant prostate cancer cells exhibit heightened glycolytic activity and that targeting glycolysis with 2-DG can effectively reverse this resistance.

Figure 3 .

Inhibition of glycolysis reverses enzalutamide resistance in prostate cancer cells

(A,B) Cell viability curves of wild-type (WT) and enzalutamide-resistant (ENZR) (A) LNCaP and (B) C4-2B cells treated with increasing concentrations of enzalutamide. (C,D) Colony formation assays showing reduced sensitivity to enzalutamide in LNCaP_ENZR (C) and C4-2B_ENZR (D) cells compared to WT controls. (E,F) Quantification of glucose uptake (E) and lactic acid production (F) in WT and ENZR cell lines. ENZR cells exhibited significantly higher glucose uptake and lactic acid production, indicating enhanced glycolysis. (G,H) Cell viability of LNCaP_ENZR (G) and C4-2B_ENZR (H) cells treated with different concentrations of 2-deoxy-D-glucose (2-DG) over 24, 48, and 72 h. (I,J) Dose-response curves for enzalutamide in the presence or absence of 0.5 mM 2-DG in LNCaP_ENZR (I) and C4-2B_ENZR (J) cells. 2-DG co-treatment enhanced enzalutamide sensitivity. (K,L) Colony formation assays of LNCaP_ENZR (K) and C4-2B_ENZR (L) cells treated with enzalutamide alone or in combination with 0.5 mM 2-DG.

Identification of glycolysis pathway hub genes associated with lethal prostate cancer

Having established the role of glycolysis in lethal prostate cancer, we sought to identify hub genes within this pathway. Given the complexity of the glycolytic pathway, which involves numerous genes, we aimed to enhance its translational potential by developing a concise gene panel to assess pathway activity. Using the SU2C dataset with PFS as the training cohort, we constructed a scoring system, termed the GLY score, by applying LASSO-Cox regression to all glycolysis pathway genes. The GLY score, derived from a panel of six key genes, CLN6, SDHC, B4GALT2, RPE, NANP, and KIF20A ( Figure 4A,B) was developed to assess glycolytic pathway activity in lethal prostate cancer. In the SU2C training cohort, the GLY score emerged as an independent prognostic factor for PFS ( Figure 4C), with higher GLY scores correlating with shorter PFS ( Figure 4D). Validation using the SU2C dataset with OS as the endpoint confirmed the GLY score as an independent prognostic indicator for OS in lethal prostate cancer ( Figure 4E), where elevated GLY scores are associated with reduced OS ( Figure 4F). Further validation in an external WDCT dataset demonstrated the robust predictive performance of the GLY score ( Figure 4G,H).

Figure 4 .

GLY score development and validation

(A,B) Identification of six hub genes contributing to the GLY score was performed using LASSO-Cox regression analysis. (C) GLY score as an independent prognostic factor for PFS in the SU2C training cohort. (D) Kaplan-Meier curve showing shorter PFS with higher GLY scores using the log-rank test. (E) Validation of the GLY score as an independent OS prognostic indicator in the SU2C testing cohort using the Wald test. (F) Kaplan-Meier curve correlating higher GLY scores with reduced OS using the log-rank test. (G,H) Robust predictive performance of the GLY score in the external WDCT validation cohort using the Wald test and log-rank test.

In addition, we investigated the association between the GLY score and genomic mutations. In the SU2C dataset, the high GLY score group exhibited a significantly higher mutation frequency, particularly in SYNE1, with a mutation rate of 24.3% compared to only 2.7% in the low GLY score group ( P  < 0.05; Figure 5A). Notably, samples with SYNE1 mutations displayed significantly higher GLY scores than those with wild-type SYNE1 ( Figure 5B). In the WDCT dataset, the high GLY score group also showed a slightly elevated mutation frequency, although the trend was less pronounced and lacked statistical significance ( Figure 5C,D).

Figure 5 .

Association of GLY score with genomic mutations

(A) Mutation frequency heatmap for the SU2C cohort, stratified by low and high GLY score groups, with the top 20 mutated genes displayed. (B) Box plots comparing GLY scores between mutated (Mut) and non-mutated (Not Mut) groups for selected genes in the SU2C cohort, with no significant differences (ns) observed. (C) Mutation frequency heatmap for the WCDT cohort, stratified by low and high GLY score groups, showing the top 10 mutated genes. (D) Box plots comparing GLY scores between mutated and non-mutated groups for selected genes in the WCDT cohort.

These findings highlight the GLY score as a streamlined and effective tool for evaluating glycolytic pathway activity in lethal prostate cancer while also reflecting heterogeneity at the level of gene mutations.

Pan-cancer analysis of the GLY score

Following our investigation of the role of the GLY score in lethal prostate cancer, we sought to evaluate its generalizability across other cancer types using the TCGA pan-cancer dataset. GLY scores were calculated for various cancer types, revealing that skin cutaneous melanoma (SKCM), uveal melanoma (UVM), and testicular germ cell tumors (TGCT) exhibited the highest scores, while adrenocortical carcinoma (ACC), thymoma (THYM), and thyroid carcinoma (THCA) had the lowest scores ( Figure 6A). We then assessed the clinical relevance of the GLY score across four prognostic endpoints: disease-free interval (DFI), disease-specific survival (DSS), OS, and progression-free interval (PFI). The results demonstrated that in ACC, bladder urothelial carcinoma (BLCA), kidney chromophobe (KICH), kidney renal clear cell carcinoma (KIRC), low-grade glioma (LGG), liver hepatocellular carcinoma (LIHC), mesothelioma (MESO), SKCM, and UVM, the GLY score was significantly associated with at least three prognostic endpoints ( Figure 6B). Notably, the GLY score acted as a risk factor in all these cancer types except KIRC, where it was a protective factor ( Figure 6C). These findings suggest that the GLY score has generalizable prognostic utility in specific cancer types. Intriguingly, in primary prostate cancer datasets PRAD, the GLY score was significantly associated only with DFI, indicating that it may be a distinctive feature of lethal prostate cancer rather than primary prostate cancer.

Figure 6 .

Pan-cancer prognostic analysis of GLY score

(A) Box plots illustrating the distribution of GLY scores across various cancer types within the TCGA. (B) Heatmap depicting the prognostic significance of the GLY score across the disease-free interval (DFI), disease-specific survival (DSS), overall survival (OS), and progression-free interval (PFI) endpoints. (C) Forest plot of hazard ratios (HR) with 95% confidence intervals (CI) from Cox regression analysis, showing the association of GLY score with tumor HR across multiple cancer types. HR and P-values were calculated using the Wald test, statistically significant P-values (P < 0.05) are indicated.

Hub gene validation in the GLY score

As described in Figure 4B, the GLY score comprises six genes, with CLN6 exhibiting the highest coefficient, suggesting its role as a hub gene. To validate this, we first evaluated CLN6 expression across various benign prostatic hyperplasia (BPH) and PCa cell lines. The results revealed significantly elevated CLN6 expression in PCa compared to BPH, with the highest levels observed in DU145 cells ( Figure 7A). Notably, CLN6 expression was also markedly increased in enzalutamide-resistant PCa cell lines (LNCaP_ENZR and C4-2B_ENZR) ( Figure 7A). Subsequently, we established stable CLN6-knockdown cell lines in LNCaP_ENZR and C4-2B_ENZR ( Figure 7B). Knockdown of CLN6 significantly reduced glucose uptake and lactate production rates in these cells ( Figure 7C,D). Similarly, in glycolysis stress tests, glucose addition triggered a rapid increase in ECAR, but the sh-CLN6 group consistently exhibited lower ECAR values than the sh-NC control group. Following oligomycin treatment to inhibit ATP synthase, cells relied more heavily on glycolysis, resulting in an elevated ECAR; however, the sh-CLN6 group displayed a lower ECAR peak than the controls. The addition of 2-DG subsequently reduced the ECAR to baseline levels in both groups ( Figure 7E,F). Finally, CLN6 knockdown significantly diminished enzalutamide resistance ( Figure 7G,H), as evidenced by reduced cell proliferation rates ( Figure 7I,J). These findings indicate that CLN6 knockdown attenuates glycolytic activity and proliferation in enzalutamide-resistant PCa cell lines.

Figure 7 .

Validation of CLN6 as a hub gene in enzalutamide-resistant prostate cancer cells

(A) Western blot analysis comparing CLN6 protein expression in benign prostatic hyperplasia (BPH) cells, prostate cancer (PCa) cell lines, and enzalutamide-resistant (LNCaP_ENZR, C4-2B_ENZR) versus wild-type (WT) LNCaP and C4-2B cells. (B) Confirmation of CLN6 knockdown efficiency by western blot analysis. (C) Bar graph quantifying glucose uptake rates in LNCaP_ENZR and C4-2B_ENZR cells, showing a significant reduction (**P < 0.01, ***P < 0.001) following CLN6 knockdown. (D) Bar graph depicting lactic acid production rates in LNCaP_ENZR and C4-2B_ENZR cells, with significant decreases (**P < 0.01) upon CLN6 knockdown. (E) Extracellular acidification rate (ECAR) profile for LNCaP_ENZR cells. (F) ECAR profile for C4-2B_ENZR cells. (G) Dose-response curve of LNCaP_ENZR cell proliferation under increasing enzalutamide concentrations. (H) Dose-response curve of C4-2B_ENZR cell proliferation, showing increased enzalutamide sensitivity following CLN6 knockdown. (I) Colony formation assay images of LNCaP_ENZR cells. (J) Colony formation assay images of C4-2B_ENZR cells under similar conditions, confirming diminished clonogenic potential with CLN6 knockdown.

Discussion

Androgen receptor (AR) signaling is a critical driver of PCa initiation and progression. The combination of ADT and ARSIs represents a cornerstone of treatment for lethal prostate cancer [ 2, 22] . Although many patients initially respond favorably to ARSIs, tumor heterogeneity results in diverse therapeutic outcomes, posing challenges for clinical management [23]. Furthermore, patients who develop resistance to prolonged ARSI therapy frequently have limited or no viable alternative treatment options [22]. Consequently, identifying novel therapeutic targets to overcome ARSI resistance in lethal prostate cancer is a crucial strategy for improving patient prognosis [24].

In prostate cancer, glycolysis displays a dynamic, stage-specific pattern that critically influences disease progression and metastatic spread [25]. Distinct from many solid tumors, glycolysis is downregulated during the early phases of prostate cancer development [26]. Instead, malignant transformation is characterized by a metabolic shift toward OXPHOS and fatty acid oxidation (FAO), yielding a more energy-efficient cellular phenotype [27]. In lethal prostate cancer, however, glycolysis is reactivated, mirroring the Warburg effect. This metabolic reprogramming is orchestrated by interactions with the tumor microenvironment, notably bone marrow adipocytes, which trigger hypoxia-inducible factor-1α (HIF-1α) signaling and upregulate glucose transporters and key glycolytic enzymes, such as hexokinase-2 (HK2) [28]. Increased glycolytic flux supports tumor cell survival in hypoxic niches, accelerates ATP generation, promotes immune evasion through lactate accumulation, and facilitates extracellular matrix remodeling and angiogenesis, thereby driving aggressive disease behavior [ 29, 30] .

Our findings align with prior reports demonstrating that activation of the glycolytic pathway is a significant predictor of poor prognosis in lethal prostate cancer patients treated with ARSIs. The GLY score, developed based on the glycolytic pathway, exhibited robust risk stratification capabilities across training, testing, and external validation cohorts, with higher GLY scores correlating with adverse prognostic outcomes. In contrast, in the TCGA-PRAD dataset, which primarily comprises primary prostate cancer cases, the GLY score did not demonstrate significant risk stratification capacity. Consistent with this, enzalutamide-resistant prostate cancer cells display elevated glycolytic activity, and inhibition of glycolysis using 2-DG effectively reverses this resistance. Additionally, in the SU2C cohort, we observed a higher mutation frequency of SYNE1 in the GLY-high subgroup. As a key component of the nuclear envelope–cytoskeleton linkage complex, SYNE1 mutations are known to induce genomic instability, which may synergize with the cellular stress associated with a high glycolytic state [31].

Within the GLY score framework, we identified CLN6 as a hub gene. CLN6 is a member of the neuronal ceroid lipofuscinosis (NCL) gene family, which is associated with a group of inherited neurodegenerative lysosomal storage disorders. Collectively, NCLs constitute the leading cause of childhood dementia [32]. The CLN6 gene encodes a highly conserved 311-amino acid protein with seven transmembrane domains and a molecular weight of approximately 36 kDa that is prevalent across vertebrate species. Localized to the endoplasmic reticulum, this protein is integral to lysosomal function [33]. Deficiency in CLN6 impairs the transport of lysosomal enzymes from the endoplasmic reticulum, resulting in reduced enzyme levels within lysosomes, thereby underscoring CLN6’s critical role in lysosomal biogenesis [34].

Research on CLN6 in cancer remains limited. Preliminary studies indicate that CLN6 is associated with poor prognosis in uveal melanoma [35]. In vitro assays demonstrated that CLN6 expression is upregulated in enzalutamide-resistant prostate cancer cell lines. Silencing of CLN6 attenuated proliferation and inhibited glycolytic activity in these resistant cells. The mechanism by which CLN6 regulates glycolysis, despite its ER localization, has remained elusive. Our preliminary co-immunoprecipitation (co-IP) and mass spectrometry experiments identified the key glycolytic enzyme LDHA as a direct interacting partner of CLN6 (data not shown; presented to reviewers). Although this finding requires confirmation by orthogonal methods such as co-IP/western blot, it suggests an intriguing regulatory axis. Given that CLN6 is a type II ER transmembrane protein with an N-terminal cytosolic domain exposed to the cytoplasm, we propose that CLN6 serves as a membrane-anchored scaffold to recruit LDHA to the ER surface, thereby establishing a local glycolytic microdomain [36]. Such spatially restricted metabolic hubs—clusters of glycolytic enzymes tethered to organelle membranes—have been increasingly recognized to generate ATP and metabolic intermediates in close proximity to high-energy-demand processes, such as protein folding, ER-to-Golgi trafficking, and Ca ^{2 +} homeostasis [ 37, 38] . In this model, CLN6 supports localized glycolysis at the ER interface; its depletion disrupts this microdomain, leading to reduced glycolytic flux—a finding consistent with our observation that CLN6 silencing inhibits glycolysis in resistant cells. As a member of the HALLMARK_GLYCOLYSIS gene set, CLN6 thus emerges as a non-canonical regulator of glycolysis through its role in the ER-lysosomal metabolic axis. This working hypothesis integrates organelle-specific metabolism with cancer cell adaptation and warrants further investigation in future research.

In conclusion, our study emphasizes the critical role of the glycolytic pathway in lethal prostate cancer and identifies CLN6 as a hub gene in addressing treatment resistance.

Supplementary Data

Supplementary data are available at Acta Biochimica et Biophysica Sinica online.

COMPETING INTERESTS

The authors declare that they have no conflict of interest.

Funding Statement

This work was supported by the grants from the Science and Technology Development Fund, Macau SAR (Nos. 0116/2023/RIA2, 006/2023/SKL, and 0090/2022/A), the Regional Joint Fund—Regional Cultivation Project (No. 2023A1515140040), and the National Natural Science Foundation of China (No. 82373166).