In this study, 33 patients were involved who had pathologically confirmed prostate adenocarcinoma staged as T2-3N0M0-1b based on the 2008 AJCC/UICC staging classification. Additionally, 18 healthy volunteers, who were within a similar age range as the patients, participated. Consequently, we successfully collected a total of 72 urinary samples. These samples were categorized into four groups:

Synchronous OM-HSPC was defined as the number of bone metastasis lesions detected on both PSMA PET/CT and MRI imaging ≤ 5, without visceral/other metastasis. The criteria patients’ selection was performed according to the following criteria.

CIRT was administered using the Siemens IONTRIS particle therapy system. The delineation of the clinical target volume (CTV) was determined based on the ESTRO ACROP consensus (11). Depending on the different risk groups, the CTV for CIRT comprised the entire prostate gland and part of the seminal vesicles. For intermediate-risk patients, the CTV included the complete prostate gland and the lower 1-1.5 cm of the seminal vesicles, and for high-risk patients, the CTV encompassed the entire prostate gland and the lower 2-2.5 cm of the seminal vesicles. For patients with synchronous OM-HSPC, the CTV included the entire prostate gland and entire seminal vesicle and the positive bone lesion(s) identified on both PSMA PET/CT and MRI imaging.

CIRT was carried out using the Siemens IONTRIS particle therapy system. The prescribed dose was determined using the Syngo planning system (V13B, Siemens, Germany), taking into account the local effect model (LEM) 1. For intermediate- and high-risk patients, the CIRT dose was 65.6 Gy (RBE) at 4.1Gy(RBE) per daily fraction. For OM-HSPC patients, the CIRT doses was 65.6 Gy (RBE) at 4.1Gy(RBE) per daily fraction to the whole prostate and the positive bone lesion(s) was treated using metastasis-directed therapy (MDT) with 49.2 Gy (RBE) at 4.1Gy(RBE) per daily fraction. 100% of the CTV was prescribed to be covered by 95% isodose line, while 100% of the planning target volume (PTV) was prescribed to be covered by 90% isodose line. In order to ensure accurate treatment delivery, both bladder and rectum preparation were performed before each treatment.

Urine samples were collected from the patients within a time frame of 4 hours before the administration of the initial fraction and within 4 hours after the completion of the final fraction. Following collection, the samples were promptly stored at a temperature of 4°C. To eliminate potential bacterial contamination, a 0.22 µm membrane filter was utilized. In each thawed sample, 80 μL of urine was combined with 160 μL of acetonitrile/methanol (1:1, v/v) in a 1.5 mL tube and placed in an ice bath. After vortexing for a duration of 30 seconds, the mixture underwent further treatment through ultrasonication for 10 minutes while placed in an ice/water bath. Subsequently, the tube was stored at a temperature of -20°C for at least 1 hour. After centrifugation (12000 rpm, 15 minutes) at 4°C, 200 μL of supernatant was transferred to a sample vial insert (250 μL) at 4°C, awaiting injection into LC-MS. Simultaneously, a pooled quality control (QC) sample was prepared by combining equal volumes (5 µL) of the supernatant from each sample.

Ultrapure water was acquired from the Millipore Milli-Q Integral 5 Ultrapure Water Systems (Billerica, USA). LC-MS grade acetonitrile and methanol were procured from SCRC (Shanghai, China). Ammonium hydroxide and ammonium formate were purchased from Merck (Darmstadt, Germany). All standard compounds were obtained from Tansoole (Shanghai, China), ThermoFisher (Waltham, USA), Merck, Aladdin (Shanghai, China), and Macklin (Shanghai, China).

The UPLC system (ExionLC™ 2.0, AB SCIEX, USA) coupled with a quadruple time-of-flight mass spectrometer (X500B, AB SCIEX, USA) was used to collect the metabolomics data of PBMC samples. Metabolite separation in both positive and negative modes was performed using an ACQUITY UPLC BEH Amide Column, 130 Å, 1.7 µm, 2.1 mm × 100 mm (Waters, USA). The water phase (A) was prepared by mixing 25 mmol of ammonium hydroxide and 25 mmol of ammonium formate with 1 L of water, while the organic phase (B) consisted of acetonitrile. The total liquid flow rate was set at 0.3 mL/min, and the gradient elution program was as follows: 0–1 min: 95% B, 1–14 min: 95% B to 65% B, 14–16 min: 65% B to 40% B, 16–18 min: 40% B, 18–18.1 min: 40% B to 95% B, 18.1–23 min: 95% B. The injection volume was 2 μL, and the column oven temperature was maintained at 40 °C.

During the analysis queue, all samples were randomly injected. Mass axis calibration, blank sample injection, pooled QC injection, and laboratory mixed standard compound solution injection were performed sequentially every eight sample injections. For electrospray ionization source parameters, ion source gas 1 and ion source gas 2 were set at 55 psi, curtain gas at 35 psi, temperature at 600 °C, and spray voltage at 5500 V or -4500 V in positive or negative modes, respectively. The TOF MS scan parameters included a mass scan range of 50–1000 Da, an accumulation time of 0.2 s, a declustering potential of 60 V or -60 V in positive or negative modes, respectively, and a collision energy of 10 V or -10 V in positive or negative modes, respectively. The acquisition of MS/MS spectra was triggered using the information-dependent acquisition mode with a maximum of ten candidate ions, and dynamic background subtraction was enabled. For TOF MS/MS scan parameters, the mass scan range was set as 25–1000 Da, the accumulation time as 0.05 s, the collision energy as 35 V or -35 V in positive or negative modes, respectively, and the collision energy spread as 15. The default settings were retained for other items.

The raw data files (.wiff2 and.wiff.scan) were converted to mzXML format and mgf format using ProteoWizard with the Peak Picking filter to centroid the peaks. The relative quantitative matrixes of all samples were generated using our large-scale metabolomics data processing software based on the R language (Software copyright registration Number: 2023SR0256527). Initially, all metabolite peaks were annotated using MetDNA2 (http://metdna.zhulab.cn/). Subsequently, the statistical significance of unknown peaks was further annotated with available reference standards in our laboratory and web-based resources, including the Human Metabolome Database (HMDB) (http://www.hmdb.ca/), Massbank (http://www.massbank.jp), and PubChem (pubchem.ncbi.nlm.nih.gov).

14 gene sets of glutamine metabolism were obtained from MSigDB (https://www.gsea-msigdb.org/gsea/msigdb) (12, 13) via inputting “glutamine” to search blank. After the removal of duplicate genes, a gene set consisting of 128 genes was established.

The RNA-seq data and corresponding clinical information of the Cancer Genome Atlas (TCGA) (14) prostate cancer cohorts were acquired from the UCSC (http://xena.ucsc.edu/) (15) public database.

PC3 cells were obtained from the Chinese Academy of Sciences Cell Bank and cultured in 1640 medium (Gibco Thermo Fisher Scientific, Inc. USA) supplemented with 10% fetal calf serum (FBS) (Gibco Thermo Fisher Scientific, Inc. USA), 100 U/mL penicillin, and 100 μg/mL streptomycin (Solarbio, Beijing, China) and cultured at 37°C in a 5% CO₂ humid atmosphere. Four groups of cells were designed in each in vitro experiment: negative control (NC) with complete medium [NC gln(+)], NC with glutamine-free medium [NC gln(-)], CIRT with completed medium [CIRT gln(+)] and CIRT with glutamine-free medium [CIRT gln(-)].

Mice were obtained from Hangzhou Ziyuan Laboratory Animal Technology Co., Ltd. The animal ethics committee of Shanghai Proton and Heavy Ion Center approved the use of mice for experiments.

Male BALB/c-nude mice aged 6 weeks were obtained, and PC3 cells were suspended in the culture medium at a concentration of 5 × 10⁶ cells/mL. A 200 μL cell suspension was then subcutaneously injected into the right flank of 10 nude mice. The volume of tumors was measured using the following formula: length × width² × 0.5 (mm³). Irradiation commenced on day 10 after tumor inoculation when the mean tumor volume reached approximately 100 mm³. Euthanasia was carried out by exposing the mice to carbon dioxide. The percentage of chamber volume replaced by carbon dioxide flow was approximately 50% vol/min.

Harvested tumor samples were utilized for further metabolomic analysis, and the steps were as follows: approximately 12 mg (record accurate weight) solid tissue was added to methanol/water (1:1, 480 μL solution per 12 mg tissue) in an ice bath. After homogenization with an automatic tissue separator or other methods, 480 μL homogenate was added to 720 μL acetonitrile/methanol (2:1). After vortexing for 30 s, the mixture was further treated with ultrasonication for 10 min in an ice/water bath. Then, this tube was refrigerated at −20 °C for at least 1 h. After centrifugation (12000 rpm, 15 min) at 4 °C, 1 mL of supernatant was transferred to a new tube. Then, all solvent was removed by a vacuum concentrator at room temperature for 3 h or longer. The obtained metabolite solid could be stored at −80 °C until testing. Before testing with LC-MS, the metabolite solid was redissolved in 100 μL acetonitrile/water (1:1). After vortexing for 30 s, the mixture was further treated with ultrasonication for 5 min in an ice/water bath. Then, the mixture was centrifuged (12000 rpm, 15 min) at 4 °C. The obtained solution was stored in sample vial inserts (250 μL) at 4 °C until LC-MS injection.

Transwell assay using 24-well plates with an 8-µm pore chamber (Corning, Inc.), and Matrigel matrix was diluted 1:8 to coat the upper side of the membrane at the bottom of the Transwell chamber. Cells were added to the upper chamber (3 × 10^4 cells/well). Meanwhile, 1640 medium with 10% FBS was added to the lower compartment. Then, the plates were incubated for 48 h in the incubator. After incubation, cells that migrated to the lower surface of the filter membrane were fixed with 4% paraformaldehyde and stained with 0.5% crystal violet. Cells remaining on the upper surface of the filter membrane were gently scraped with a cotton swab. The lower surfaces were captured by an inverted microscope, and the counting was repeated three times.

PC3 cells were seeded into a 6-well plate. When the cell confluence reached 80%, we scratched the monolayer cells with a sterile micropipette tip. The floating cells were then washed with PBS. Wound healing within the scrape line was observed at 0 h, 16 h and 24 h. Triplicate wells for each condition were examined.

The data analysis was conducted using MetaboAnalyst (https://metaboanalyst.ca/). A volcano plot was utilized to compare group differences, employing fold change (FC) analysis and the Wilcoxon signed-rank test to identify statistically significant metabolites (p<0.05, |FC|>1.5). Heatmaps were generated to visualize variations in metabolite profiles. Pathway analysis was conducted based on significantly altered metabolites. Box plots were employed to illustrate mean values with standard deviations. Statistical analysis involved the use of the Mann-Whitney test and Pearson’s chi-square test, with a significance level set at p<0.05.

Compared to healthy humans, the level of metabolites in treatment-naïve PCa patients was generally elevated ( Figure 1A ). Enrichment analysis based on differential metabolites revealed the typical metabolic reprogramming pathway in patients with prostate cancer ( Figure 1B ). The examination of a volcano plot revealed that 223 metabolites were dramatically elevated and 10 were markedly downregulated ( Figure 1C ). Using an UpSet plot, the hit number (the number of metabolites belonging to enriched particular pathways) and the overlapping relationship of the top 5 pathways were explored, aspartate and glutamate metabolism ranked first (10 hits), followed by arginine biosynthesis, glycine, serine, and threonine metabolism, cysteine and methionine metabolism, and galactose metabolism ( Figure 1D ). Five overlapping metabolites (glutamine, 2-oxoglutaramate, argininosuccinic acid, aspartate, and glutamate) were identified between aspartate and glutamate metabolism and arginine synthesis, the two most crucial processes ( Figures 1E–I ). To satisfy the increasing energy requirement, the elevated pathways and five metabolites might contribute to the initiation and progression of PCa. Determining the response of these metabolites to therapy is essential for clinical guidance.

The heatmap based on the top 50 differential metabolites between treatment-naïve and post-naADT groups is shown in Figure 2A . Most of the metabolic expression was suppressed after naADT. The enriched pathways of these differential metabolites are shown in Figure 2B . The volcano plot demonstrated 164 significantly downregulated metabolites and 2 significantly elevated metabolites ( Figure 2C ). The UpSet plot was used to examine the hit count and overlapping relationships of the top 5 metabolic pathways, including arginine and proline metabolism (7 hits), purine metabolism (10 hits), arginine biosynthesis (5 hits), glycine, serine, and threonine metabolism (7 hits), and alanine, aspartate, and glutamate metabolism (7 hits) ( Figure 2D ). Four metabolites, glutamine, 2-oxoglutaramate, argininosuccinic acid, and glutamate, were significantly suppressed (p<0.05) by naADT in both the alanine, aspartate, and glutamate metabolism pathways and the arginine biosynthesis metabolism pathway ( Figures 2E–H ). Despite the fact that naADT considerably inhibited PCa metabolism, the alanine, aspartate, and glutamate metabolism pathways, which were the most active in PCa, were not the most inhibited by naADT. In addition, the post-naADT levels of glutamine, 2-oxoglutaramate, argininosuccinic acid, and glutamate remained considerably higher than those in healthy samples ( Figures 2E–H ).

The patients were separated into three groups, intermediate-risk and high-risk, and OM-HSPC, to examine the response to ADT at various stages. The heatmap ( Figure 3A ) with the top 50 metabolites revealed distinct differences between the three groups. Notably, the OM-HSPC group had significantly higher metabolic activity than the localized stages. Volcano plots ( Figure 3B ) identified 99 metabolites that were significantly elevated in the comparison of the high-risk and intermediate-risk groups, Similarly, 112 upregulated and 2 downregulated metabolites were also significantly altered in OM-HSPC compared to localized stages (intermediate- and high-risk) PCa ( Figure 3C ). From the two groups of upregulated metabolites, 60 overlapping metabolites ( Figure 3D ) were identified ( Supplementary Table 1 ), which could be considered key risk metabolites associated with staging under ADT. Pathway enrichment analysis and UpSet plots revealed that the 60 identified metabolites were associated with multiple metabolic pathways, including alanine, aspartate, and glutamine metabolism (6 hits); Glycine, serine, and threonine metabolism (5 hits); Glyoxylate and dicarboxylate metabolism (4 hits); Arginine biosynthesis (3 hits); Amino sugar and nucleotide sugar metabolism (3 hits) ( Figures 3E, F ). Glutamine was discovered to be enriched in aspartate and glutamine metabolism as well as arginine biosynthesis. This finding suggested that glutamine levels differed significantly between post-naADT stages. Moreover, the glutamine levels of the intermediate- and high-risk groups treated with naADT were not substantially greater than those of the healthy samples, while glutamine levels were still considerably greater in the OM-HSPC group than in healthy samples or those with localized PCa ( Figure 3G ).

These findings highlight that glutamine plays a pivotal role in the initiation and progression of PCa. Furthermore, naADT appears to be less effective in reducing glutamine levels in OM-HSPC, compared to its impact on localized PCa, to the levels observed in healthy individuals. Consequently, the elevated glutamine levels in OM-HSPC suggest that relying solely on systemic naADT treatment may be inadequate in managing PCa, which could lead to suboptimal patient outcomes.

The genes associated with glutamine metabolism from MSigDB and the BCR-related genes identified by univariate Cox regression (p<0.05, HR>1) from TCGA shared 13 genes, indicating that glutamine metabolism may play a risk factor in PCa BCR ( Figure 4A ). Based on TCGA data, a LASSO analysis with the glmnet package revealed 9 of the 13 genes ( Figures 4B, C ). Samples were categorized into low- and high-GBS groups based on the glutamine and BCR-related gene score (GBS) generated from the levels of expression of the 9 genes. Figure 4D shows the gene expression pattern for these groups. According to the findings, samples with relatively high GBS had considerably shorter BCR-free times (P< 0.001) ( Figure 4E ).

For each gene in PCa, a comparison of paracancerous and tumor tissue was carried out to learn more about the oncogenic role of these genes. TARS2, SLC9A, SLC7A9, NOS3, NDUFB10, FPGS, and CAD all showed significant changes (P< 0.05) ( Figure 4F ), showing that glutamine metabolism may play a role in the development and recurrence of PCa. Using TCGA PCa baseline data, the relationship between GBS and clinical PCa characteristics was also explored ( Table 3 ). The findings showed that patients in the high-GBS group had significantly higher T stages and Gleason scores (P < 0.001).

The results indicate a notable correlation between heightened glutamine-related metabolism and advanced stages of PCa, as well as a reduced BCR-free survival duration. As such, reducing glutamine levels in OM-HSPC patients might enhance their prospects for improved local control.

The metabolic profiles of OM-HSPC patients before and after CIRT are depicted in Figure 5A . The enrichment analysis uncovered significantly downregulated metabolic pathways ( Figure 5B ). The volcano plot revealed that 3 metabolites were considerably elevated and 120 were significantly downregulated ( Figure 5C ). The UpSet analysis found the top 5 altered metabolic pathways to be alanine, aspartate, and glutamate metabolism (6 hits); Glycine, serine, and threonine metabolism (5 hits); Arginine biosynthesis (5 hits); Glyoxylate and dicarboxylate metabolism (4 hits) ( Figure 5D ). The Venn diagram represented the association between the 123 differential metabolites and the 60 risk metabolites identified in prior investigations ( Figure 5E ). After CIRT, glutamine levels in all three groups were reduced to healthy levels (p>0.05) ( Figure 5F ).

85% (51 out of 60) risk metabolites were considerably decreased in patients treated with naADT followed by CIRT. In addition, the glutamine levels of OM-HSPC patients were reduced to a level comparable to that of healthy individuals. Based on present conclusions, we depicted the changes among core risk metabolites influenced by treatments in Figure 6 .

Due to the impossibility of obtaining tissue samples from PCa patients treated with CIRT, an experiment on mouth was conducted to confirm the inhibitory effects of CIRT on glutamine levels in tumor tissue. The mice were divided into two independent groups: normal control (NC) and CIRT. The CIRT group was irradiated with a single fraction of 2.5 Gy carbon ion. NC was the nontreatment control group. Tumor size was measured every three days during this time period, and tumor samples were collected 14 days after irradiation, ( Figure 7A ). It was evident that CIRT dramatically inhibited tumor growth compared to the NC group ( Figure 7B ). Metabolomic analysis of tumor tissue indicated distinct metabolites in the two groups ( Figure 7C ), with seven downregulated and seven upregulated ( Figure 7E ). Enrichment analysis found the differential pathways ( Figure 7D ), which revealed a substantial decrease in glutamine levels in the CIRT group compared to the NC group, with a p value less than 0.001 ( Figure 7F ). This result showed the ability of CIRT to inhibit glutamine metabolism in PCa tissue.

Given that glutamine plays a crucial role in the progression of PCa, we employed PC3 cells as a prostate model to determine whether glutamine deprivation could enhance the inhibitory effect of CIRT on PCa invasion and migration. In Transwell tests, 48 hours after 2 Gy carbon ion irradiation, less cells retained the capacity to infiltrate Matrigel to the bottom of the insert ( Figures 8A, B ). When glutamine was removed from the cell media, cell invasion was drastically reduced. In addition, the cell migratory ability was also decreased by the single dose of 2 Gy carbon ion irradiation in wound healing assay. When glutamine was eliminated from the cell media, cell migration was further remarkably impaired. These results preliminarily indicated that glutamine deprivation could significantly enhance the capacity of CIRT to inhibit tumor metastasis.