Glycerophospholipid-driven lipid metabolic reprogramming as a common key mechanism in the progression of human primary hepatocellular carcinoma and cholangiocarcinoma

Abstract

Metabolic reprogramming, a key mechanism regulating the growth and recurrence of hepatocellular carcinoma (HCC) and cholangiocarcinoma (CCA), still lacks effective clinical strategies for its integration into the precise screening of primary liver cancer. This study utilized ultra-high-performance liquid chromatography with quadrupole time-of-flight mass spectrometry to conduct a comprehensive, non-targeted metabolomics analysis, revealing significant upregulation of lipid metabolites such as phosphatidylcholine and lysophosphatidylcholine in patients with HCC and CCA, particularly within the glycerophospholipid metabolic pathway. Hematoxylin and eosin and immunohistochemical staining demonstrated marked upregulation of phospholipase A₂ in tumor tissues, further emphasizing the potential of lipid metabolism as a therapeutic target and its important part in the course of cancer. This work provides a new viewpoint for addressing the clinical challenges associated with HCC and CCA, laying the groundwork for the broad application of early diagnosis and personalized treatment strategies, and ultimately aiming to provide tailored and precise therapeutic options for patients.

Introduction

The incidence and fatality rates of primary liver cancer make it a prominent malignancy worldwide. It includes HCC and CCA, which make up approximately 80% and 15% of all cases, respectively [1, 2]. Patients with these tumors still have a poor prognosis despite advances in detection and therapy because of late-stage diagnosis and high recurrence rates. Metabolic reprogramming is a key area of cancer research, and numerous studies have showed that HCC and CCA exhibit significant metabolic alterations, including lipid metabolic reprogramming, glucose metabolic reprogramming, and amino acid metabolic reprogramming [3,4,5]. The development, high recurrence rates, and terrible prognosis of HCC and CAA are intimately linked to these metabolic alterations, highlighting the critical role of metabolic reprogramming in the development of these cancers [5, 6].

The metabolic reprogramming is a hallmark of malignant tumors, driving changes in cancer cell metabolism to support growth and maintain the tumor microenvironment [3, 7]. In HCC and CCA, enhanced anabolic processes and reduced catabolic processes lead to a metabolic imbalance [3, 8]. Conventional diagnostic methods often miss these early changes, resulting in late-stage diagnoses and poor prognoses [9]. In chemotherapy-resistant tumors alter pathways such as glycolysis and glutamine metabolism are altered [10]. Immunotherapy efficacy is limited by insufficient reprogramming of the immunosuppressive tumor microenvironment [11]. Targeting tumor metabolism by reprogramming the tumor microenvironment and enhancing antitumor immunity can improve immunotherapy outcomes and the patient prognosis, and thus identifying the specific metabolic reprogramming patterns of each cancer type is essential.

This study employed the LCMS-9030 UPLC‒Q-TOF/MS platform for comprehensive, nontargeted metabolomics analysis to investigate the metabolic characteristics of HCC and CCA. By identifying DEMs and their enrichment in various metabolism pathways, this study aimed to elucidate cancer-associated metabolic changes. The analysis revealed the significant upregulation of key metabolites involved in glycerophospholipid metabolism in HCC and CCA patients, indicating common metabolic reprogramming in these cancers. H&E and immunohistochemical staining demonstrated a marked upregulation of phospholipase A2 (PLA2) in tumor tissues, further emphasizing the importance of lipid metabolism in cancer progression (Scheme 1). This study systematically reveals, for the first time, the expression patterns of 24 metabolites that were significantly altered in expression in HCC patients and CAA patients, with a particular emphasis on their enrichment in lipid metabolism pathways. These findings offer new perspectives on the metabolic reprogramming that occurs in HCC and CCA. The study revealed the significant upregulation of the key enzyme, PLA2, in the glycerophospholipid metabolism pathway in HCC and CCA tumor tissues. These findings further emphasize the vitality of lipid metabolism reprogramming in the growth and advancement of HCC and CCA, further highlighting the importance of lipid metabolism in cancer.

Scheme 1

Diagnosis strategies for metabolic reprogramming in human primary HCC and CCA. This study utilized UPLC‒Q-TOF/MS to conduct a thorough, non-targeted metabolomics analysis, identifying a notable upregulation of lipid metabolites, including PC and LysoPC, in patients with HCC and CCA, which were particularly enriched in the GPL metabolic pathway. H&E and IHC staining revealed a significant increase in PLA2 expression in tumor tissues, underscoring the crucial role of lipid metabolism in cancer progression and its potential as a therapeutic target. These findings offer fresh insights into the clinical challenges of HCC and CCA, supporting the advancement of early diagnostic and personalized treatment strategies to deliver customized and precise therapeutic options for patients

These results offer fresh perspectives on the metabolic reprogramming of HCC and CCA and underline the potential of lipid metabolites as early detection and diagnostic biomarkers. This study lays the groundwork for future clinical applications, including the significance of cultivating targeted and personalized treatment strategies to enhance patient care.

Materials and methods

Materials

Reference materials

Reference materials: The isotopic internal standards LysoPC (19:0) and SM (12:0) were acquired from Avanti Polar Lipid, Inc. (Birmingham, Alabama, USA). We purchased Phe-d5, Trp-d5, CA-d4, and CDCA-d4 from Bepure (Manhage Shanghai Biotechnology Co., Ltd., Shanghai, China) in order to get Carnitine C8:0-d3. The suppliers of FFA C16:0-d3, FFA C16:0-d3, and Carnitine C16:0-d3 were Isoreag (ZZBIO Co., LTD, Shanghai, China). We purchased Carnitine C2:0-d3 from Cmass Company (Shanghai, China).

Experimental reagents

Experimental reagents included acetonitrile, formic acid, and methanol graded for LC/MS obtained from ANPEL Scientific Instrument Co., Ltd. (Shanghai, China). Fluka provided LC‒MS-grade ammonium bicarbonate (Honeywell, Seelze, Germany). Wahaha Group Co., Ltd. provided the purified water (Hangzhou, China).

Methods

Enrollment criteria for patients with primary HCC or CCA

The inclusion criteria for patients with primary HCC or CCA were as follows: (i) histologically or cytologically confirmed diagnosis of primary HCC or CCA or imaging findings consistent with HCC or CCA with underlying cirrhosis; (ii) aged ≥18 years; (iii) an Eastern Cooperative Oncology Group (ECOG) performance status of 0–2; (iv) a Child–Pugh score of A or B (≤7); (v) those with HCC should be at Barcelona Clinic Liver Cancer (BCLC) stage A-C; (vi) those with CCA should have locally advanced or metastatic disease not amenable to surgical resection or local ablative therapies; (vii) no prior systemic therapy for advanced HCC or CCA, and those who received local therapies such as transcatheter arterial chemoembolization or radiofrequency ablation must have recovered from treatment-related toxicities; (viii) at least one measurable lesion according to Response Evaluation Criteria in Solid Tumours (RECIST; version 1.1); and (iv) laboratory tests meeting the following requirements: absolute neutrophil count (ANC) ≥1.5 × 10⁹/L, platelet count ≥100 × 10⁹/L, hemoglobin level ≥9 g/dL, serum creatinine level ≤1.5× upper limit of normal (ULN) or creatinine clearance ≥60 mL/min, total bilirubin level ≤2× ULN, and aspartate transaminase (AST) and alanine transaminase (ALT) levels ≤5× ULN.

This study retrospectively collected data from HCC and CCA patients treated at the Fifth Affiliated Hospital of Wenzhou Medical University (CTR20210090) between January 2020 and December 2023, including tumor and adjacent nontumor tissue samples. All tissues were obtained through surgical resection using standards to ensure sufficient tissue for metabolomic sequencing analysis. The Declaration of Helsinki was followed by the ethics committees at each participating site when approving the research protocol, adhering to the guidelines set forth by Nature journal. Every patient provided written, informed consent for the utilization of their data in this research endeavor.

The exclusion criteria were as follows: (i) other active malignancies (except adequately treated basal cell carcinoma, squamous cell carcinoma of the skin, or carcinoma in situ of the cervix); (ii) severe or uncontrolled systemic diseases (e.g., uncontrolled diabetes, active infections, or cardiac diseases); (iii) pregnancy or lactation; (iv) history of substance abuse or any condition that might interfere with compliance; (v) previous liver transplantation; (vi) prior systemic anti-cancer therapies for HCC or CCA; and (vii) immunosuppressive therapy within 14 days before enrollment.

Metabolomic analysis of tumor tissues

Sample pre-treatment

Serum samples were thawed in a refrigerator at 4 °C, and all experimental procedures were conducted on ice. After vortexing, the samples were moved to an Eppendorf (EP) tube, with 50 μL for each sample, and 200 μL of precooled methanol containing mixed internal standards was added. After five minutes of vortexing, the samples were subjected to centrifugation at 13,000 g for a duration of ten minutes at a temperature of 4 °C. For every sample, the supernatant (150 μL) was moved to an additional EP tube. The extract from all the samples remaining after the supernatants were transferred was mixed to create samples for quality control (QC).

Both the test and QC samples were generated concurrently as follows: the extracts were lyophilized (Labconco Corporation, Kansas City, Missouri, USA) to obtain the polar components.The extracts were reconstituted in 80 μL of an 80% acetonitrile solution, centrifuged, and the supernatant was moved to a lined vial for examination prior to the mass spectrometry analysis. For the purpose of extracting metabolites, each sample was examined in the ion modes of both favorable and unfavorable. The samples were stored at a temperature of 4 °C in an autosampler throughout the evaluation. A random sampling method was employed to ensure data reliability and to avoid the impact of instrument signal fluctuations. Additionally, the QC samples were uniformly integrated into each batch of data collection to monitor instrument stability.

Chromatographic conditions

The positive ion mode analyses used the following conditions. The chromatographic column was a UPLC BEH C8 (1.7 μm, 50.0 × 2.1 mm; Waters, Milford, Massachusetts, USA). The mobile phases consisted of water containing 0.1% formic acid for phase A, and acetonitrile containing 0.1% formic acid for phase B. The flow rate was set to 0.4 mL/min with a column temperature of 60 °C, and the injection volume was 5 μL. The elution gradient initiated at 5% B and remained constant for half a minute before increasing to 40% B after two minutes. It then reached full saturation at 100% B after an additional duration of eight minutes before being maintained for another two minutes. Finally, the mixture returned to its original gradient composition of 5% B after10 minutes and was allowed to equilibrate for two minutes.

The chromatographic column utilized for negative ion mode was a UPLC HSS T3 column (1.8 μm, 50 mm × 2.1 mm; Waters, Milford, Massachusetts, USA). The subsequent mobile phases were as follows: the composition of Phase B consisted of 95% methanol and water with the addition of ammonium bicarbonate at a concentration of 6.5 mM, while Phase A contained ammonium bicarbonate at the same concentration. The injection dosage was set to 5 μL, with a flow speed of 0.4 mL/min and a column temperature kept at 60 °C. The elution gradient was initially set at 2% B and stored for thirty seconds. It was then raised to 40% B and maintained for 2 minutes. Finally, it was elevated to 100% B and held for 8 minutes. It was equilibrated for two minutes and restored to the original gradient of 2% B at ten minutes.

Mass spectrometry data acquisition parameters

The desolvation tube temperature is adjusted to 250 °C, while the heater module temperature is maintained at 400 °C. Additionally, the interface temperature is regulated to be 300 °C. The ion source temperatures are configured as ESI (+) and ESI (-). Moreover, the ion source interface voltage is adjusted to +4.50 kV and -3.5 kV for positive as well as negative polarities correspondingly. Nitrogen gas at a stream rate of 3.0 L/min serves as the nebulizer gas, whereas air at a stream rate of 10 L/min functions as the heater gas. Lastly, argon gas acts as the collision gas. MS1 scan (m/z 70–1050); MS2 DDA (m/z 70–1050); ID: OFF is the scan mode.

Data preprocessing and statistical analysis

The LC/MS metabolomics data was preprocessed using LabSolutions Insight LCMS 4.0 software (Shimadzu, Kyoto, Japan) for automated peak detection, alignment, and integration. The compounds were recognized using the company’s internal database, which includes standards for more than 700 metabolites, retaining characteristic peaks detected in more than 80% of the samples. After applying the 80% rule for feature selection, the missing values were substituted with the minimum nonmissing value of the corresponding peak. The raw data were normalized using internal standards, and the results of the metabolite analysis were further analyzed. Simca-P software (version 14.1, Sartorius AG Umetrics, Goettingen, Germany) was used to perform multivariate studies, such as orthogonal partial least squares discriminant analysis (OPLS-DA), principal component analysis (PCA) and orthogonal partial least squares discriminant analysis (OPLS-DA). TBtools software (version 2.042) was used to generate clustering heatmaps of differentially abundant metabolites. Hierarchical clustering analysis (HCA), t-tests, and false discovery rate (FDR) adjustments utilized the Benjamini–Hochberg method were conducted using R software (version 4.2.3). The QC sample analysis of correlation was conducted using GraphPad Prism software (version 8.0). The assessment of fold changes and visualization in volcano plots, The MetaboAnalyst 6.0 website was utilized to do the correlation analysis and metabolic pathway analysis of metabolites that varied in abundance. (https://www.metaboanalyst.ca/).

Clinical metabolomics correlation analysis

In this study, a thorough examination was done to determine and identify metabolic pathways in HCC and CCA patients. Using a Venn diagram and KEGG pathway enrichment analysis, differently abundant metabolites across the HCC and CCA groups were examined in order to identify and screen metabolic signaling pathways unique to tissue samples from HCC and CCA patients.

The “VennDiagram” R package (https://cran.r-project.org/web/packages/VennDiagram/index.html) was used for Venn diagram analysis, while the “clusterProfiler” R package (https://bioconductor.org/packages/release/bioc/html/clusterProfiler.html) was utilized for KEGG pathway enrichment analysis.

Histopathological analysis

Hematoxylin and eosin (H&E) staining was used to typical tissue slices from patients with CCA and HCC in order to assess the histological architecture. A Hematoxylin and Eosin Staining Kit (Beyotime, C0105M, Shanghai, China) was used to accomplish the H&E staining. the stained slices were observed using a light microscope to capture images for further analysis.

IHC analysis

Tissue slices were deparaffinized, rehydrated, and antigen-retrieved for the immunohistochemistry (IHC) examination. Hydrogen peroxide was used to stop endogenous peroxidase activity, and a protein blocking solution was used to stop nonspecific binding. After that, the sections were treated with primary antibodies (anti-phospholipase A2; Abcam, ab23705; Cambridge, United Kingdom) against certain markers were stored at 4 °C for the entire night. The slices were cleaned and then treated with secondary antibodies (Abcam, ab6715, Cambridge, United Kingdom) and visualized using a DAB substrate kit (Abcam, ab64238, Cambridge, United Kingdom). Counterstaining was performed with hematoxylin.

The semiquantitative analysis of the immunohistochemical (IHC) staining employed a scoring system dependent on the number and proportion of positive cells. The proportion of positively stained cells was categorized into four groups: 0% (0), 1-25% (1), 26-50% (2), 51-75% (3), and 76-100% (4). Staining strength was assessed on a scale ranging from absence of staining (0) to faint (1), moderate (2), and strong (3) staining. Multiplication of the intensity score by the percentage score yielded the final result. A statistical analysis was performed to compare marker expression levels between HCC and CCA tissues.

Results

Quality control and data analysis of metabolomics samples

This study utilized the UPLC-Q-TOF/MS platform for a comprehensive, untargeted metabolomics analysis of HCC and CCA samples, and identified 467 metabolites and 11 internal standards in both positive and negative ion modes. PCA showed tight clustering of the quality control (QC) samples, indicating stable performance and reliable data quality (Supplementary Figure 1A). The first main component scores from the PCA, remained within 2 standard deviations, confirming that instrument’s variability was within the normal range and suitable for further analysis (Supplementary Figure 1B). Pearson’s correlation analysis between the first and last QC data matrices revealed a very strong correlation, further verifying the excellent data quality (Supplementary Figure 1C).

Metabolomic characteristics of patients with primary HCC

In this study, the univariate differential examination of the identified metabolites detected in both positive and negative ionization modes from the primary liver cancer (HCC) and control (Ctrl) groups identified 122 significantly differentially expressed metabolites (DEMs) with fold changes (FCs) > 1.5 or < 0.67 and P-values < 0.05. Among these genes, 86 were significantly upregulated and 36 were downregulated (Fig. 1A). The hierarchical clustering analysis was performed based on the Bray-Curtis distance showed high similarity within each group (Fig. 1B). The PCA results revealed clear metabolic phenotype differences between the HCC and Ctrl groups (Fig. 1C). OPLS-DA demonstrated strong classification performance (R²Y = 0.832, Q² = 0.771; Fig. 1D). The permutations test was conducted with 200 iterations validated the model’s robustness, showing no overfitting (Q² intercept = -0.371) and confirming an acceptable predictive accuracy (Fig. 1E). A ROC curve analysis of the OPLS-DA model indicated excellent separation between the HCC and Ctrl groups, with an AUC of 0.998 (Fig. 1F).

Fig. 1

Untargeted LC‒MS metabolomics analysis of primary liver cancer. A Volcano plot of the Ctrl and HCC groups. The horizontal axis of the graph represents the log2 value of the fold change (FC), and the vertical axis represents the -log10 value of the significance P value. Significantly different metabolites are presented in red if they satisfy FC > 1.5 and P value < 0.05; metabolites are presented in blue if they satisfy FC< 0.67 and P value < 0.05. Nonsignificantly different metabolites are presented in gray. B Hierarchical clustering analysis of Ctrl and HCC samples. C Unsupervised PCA score plots of metabolomic phenotypes between the Ctrl and HCC groups. Model parameter: R²X = 0.81 (cumulative variance proportion of 2 principal components). D Score plot of the OPLS-DA model to maximize the intergroup differentiation of the metabolomic data between the Ctrl and HCC groups. Model parameters: 1 predictive component + 1 orthogonal component, R²Y = 0.832, and Q² = 0.771. E OPLS‒DA permutation test diagram of the Ctrl and HCC groups. F ROC plot of the OPLS-DA model differentiating the Ctrl and HCC groups. G Hierarchical clustering heatmap of differentially abundant metabolites between the Ctrl and HCC groups. H Correlation heatmaps of differentially abundant metabolites between the Ctrl and HCC groups. I KEGG pathway enrichment diagram (bar chart)

The variable importance in projection (VIP) scores obtained from the OPLS-DA model identified biologically significant DEMs, which were then used to construct clustering heatmaps and for a correlation analysis, and pathway analysis. The hierarchical clustering heatmap revealed 46 significant DEMs (OPLS-DA VIP > 1 and P-value < 0.05), including phosphatidylcholine (PC), lysophosphatidylcholine (LysoPC), phosphatidylethanolamine (PE), and bile acids (Fig. 1G). Acylcarnitines, such as oleoylcarnitine and hexadecanoyl CAR, were present at relatively higher expression levels in the HCC group (FC = 2.396736 [P < 0.00001] and 2.058736 [P < 0.0001], respectively), suggesting their role in fatty acid β-oxidation. Bile acids, such as glycochenodeoxycholic acid (GCDCA), exhibited significantly higher levels in the HCC group (FC = 7.005840, P < 0.0001), indicating their involvement in lipid digestion, absorption, and liver dysfunction. Additionally, the levels of amino acids such as valine and phenylalanine were significantly increased in the HCC group, reflecting potential metabolic reprogramming.

The correlation analysis and significance testing of the DEMs shown in the correlation heatmap revealed several significant relationships among lipid metabolites between the primary HCC group and the control group (Fig. 1H). Different PCs exhibited strong positive correlations, such as between PC 36:5 and PC 40:8 (R= 0.50422, P < 0.00001). Lysophosphatidylethanolamine (LysoPE) also showed significantly positively correlated with PCs, such as between LysoPE 22:6 and PC 38:6 (R= 0.70638, P < 0.00001), indicating potential synergistic roles in lipid metabolism. Some PCs and bile acids also displayed significant negative correlations, such as PC 38:6 and GCDCA (R = −0.58467, P < 0.00001), suggesting complex inverse interactions that possibly reflecting regulatory relationships between lipid and bile acid metabolism. Dehydroepiandrosterone sulfate (DHEA-S) was positively correlated with various metabolites, such as with PC 40:8 (R = 0.50379, P < 0.00001), while 3-dehydroteasterone was weakly correlated with specific metabolites, including PC 38:6 (R = −0.26617, P < 0.05).

The metabolic pathway enrichment study of DEMs in primary HCC revealed significant alterations in lipid metabolism pathways (Table 1). Key metabolites included various PCs, such as 36:0, 36:3, and 38:6, as well as LysoPCs, such as 22:4 and 20:1, indicating disruptions in cell membrane metabolism and signaling pathways. Additionally, the levels of phosphatidylethanolamines (PEs) and LysoPEs, such as PE 34:2, PE 34:1, LysoPE 22:6, and LysoPE 16:0, were significantly changed, reflecting altered membrane fluidity and flexibility. Relevant alterations in the concentrations of bile acids have been observed, including glycochenodeoxycholic acid (GCDCA) and taurochenodesoxycholic acid (TCDCA), suggest liver dysfunction and bile acid metabolism dysregulation. Elevated levels of acyl carnitines, such as oleoylCAR and hexadecanoylCAR, which are involved in fatty acid β-oxidation, indicate enhanced or abnormal fatty acid metabolism in liver cancer. The maps of enriched KEGG pathways (bubble chart and bar chart) clearly show that these differentially expressed metabolites are significantly enriched in lipid metabolism-associated pathways, the processes encompassed include phospholipid metabolism, bile acid synthesis, and fatty acid metabolism (Fig. 1I and Supplementary Figure 1D). These findings reveal the important part of lipid metabolism in the process and progression of HCC, providing new research directions and potential therapeutic targets.

Table 1 Enrichment analysis of metabolic pathways of differential metabolites in primary liver cancer

Metabolomic characteristics of patients with primary CCA

Univariate differential analysis of metabolites detected in both positive and negative ion modes identified 130 significant DEMs between the primary CCA and Ctrl groups with FCs > 1.5 or < 0.67 and P-values < 0.05. Of these, 67 were upregulated, and 63 were downregulated (Fig. 2A).The hierarchical clustering analysis, based on the Bray-Curtis distance, revealed a high degree of similarity among the CCA samples (Fig. 2B). PCA revealed distinct metabolic phenotypes between the CCA and control groups, highlighting significant metabolic differences (Fig. 2C). OPLS-DA yielded R²Y = 0.787 and Q² = 0.671, suggesting good model fit and predictability (Fig. 2D). Model robustness was confirmed through 200 permutation tests, with a Q² intercept of −0.389 and a positive slope, indicating no overfitting (Fig. 2E). ROC curve analysis of the OPLS-DA model indicated excellent discriminatory power, with an AUC of 0.998 (Fig. 2F).

Fig. 2

Untargeted LC‒MS metabolomics analysis of primary cholangiocarcinoma. A Volcano plot of the Ctrl and CCA groups. B Hierarchical clustering analysis of the Ctrl and CCA groups. C Unsupervised PCA score plots of metabolomic phenotypes between the CCA and Ctrl groups. Model parameter: R²X = 0.801 (cumulative variance proportion of 2 principal components). D Score plot of the OPLS-DA model to maximize intergroup differentiation of the metabolomic data between the Ctrl and CCA groups. Model parameter: 1 predictive component + 1 orthogonal component, R²Y = 0.787, Q² = 0.671. E OPLS-DA permutation test diagram of the Ctrl and CCA groups. F ROC plot of the OPLS-DA model differentiating the Ctrl and CCA groups. G Hierarchical clustering heatmap of differentially abundant metabolites between the Ctrl and CCA groups. H Correlation heatmaps of differentially abundant metabolites between the Ctrl and CCA groups. I KEGG pathway enrichment diagram for the Ctrl and CCA groups

Through the analysis of the heatmap of DEMs between the CCA group and the control group, this study identified significant alterations in the lipid metabolism pathway of CCA patients (Fig. 2G). There were 49 DEMs between the CCA and Ctrl groups, primarily comprising PCs, LysoPCs, and amino acids. Specifically, several PCs (e.g., 40:8 and 36:1) and LysoPCs (e.g., 18:0 and 16:0) were highly raised in the CCA group. These metabolites exhibited high VIP scores in the PLS-DA model, indicating their substantial contribution to the group differences. Additionally, sphingolipids (e.g., SM 38:3) were significantly upregulated (P < 0.05 after multiple testing correction). Reprogramming of lipid metabolism is believed to perform a pivotal role in the formation and progress of CAA, as indicated by these findings. Additionally, phospholipids and LysoPCs have the possibility to serve as valuable biomarkers for the early detection and diagnosis of CCA.

Analysis of the correlation heatmap of significant DEMs between CCA and Ctrl groups revealed that LysoPC and lysophosphatidic acid (LPA) were highly correlated, suggesting their involvement in similar metabolic pathways or regulatory mechanisms (Fig. 2H). Strong correlations were also observed among LysoPCs of varying chain lengths and saturation levels, such as 16:0, 18:0, and 20:0, indicating that they might be produced or regulated by similar mechanisms. Additionally, phospholipids such as PE 34:2, PE 38:6, and PC 36:3 displayed strong intercorrelations, reflecting the coordinated regulation of cell membrane metabolism. Choline, an essential precursor in phospholipid metabolism, was positively correlated with multiple PCs, further confirming its crucial role in phospholipid biosynthesis. In contrast, polyunsaturated phospholipids such as PC 40:8 were negatively correlated with various LysoPCs, potentially indicating complementary or competitive relationships among them. The high correlations of Pro-Leu with multiple phospholipids suggest its potential role in lipid metabolic regulation. Additionally, DHEA-S and tryptophan were significantly correlated with several lipids, suggesting their involvement in lipid metabolic regulation or cross-pathway interactions.

The KEGG enrichment pathway analysis of lipid metabolites in patients with CCA revealed significant alterations across multiple lipid metabolic pathways (Table 2, Fig. 2I, and Supplementary Figure 1E). Notably, several PCs and PEs showed significant changes, indicating an enrichment in phospholipid metabolism pathways and suggesting they may play crucial roles in cell membrane remodeling and signal transduction. Additionally, metabolites involved in glycerolipid metabolism, such as triglycerides and diacylglycerols, exhibited significant changes, highlighting the importance of glycerolipid metabolism in energy storage and lipid signaling in CCA. The enrichment of bile acid metabolism pathways, including metabolites such as GCDCA, suggests a potential role for bile acids in cholestasis and liver function impairment. Furthermore, the significant alteration in DHEA-S within the steroid hormone metabolism pathway implies that steroid hormones might play a regulatory role in CCA pathogenesis. The marked changes in arachidonic acid and its metabolites underscore the importance of the arachidonic acid metabolism pathway in inflammatory responses and cell signaling in patients with CCA. Changes in acylcarnitines, such as CAR 18:1 and CAR 16:1, indicate alterations in fatty acid oxidation and energy metabolism, further supporting the enrichment of fatty acid metabolism in CCA.

Table 2 Enrichment analysis of metabolic pathways associated with differential metabolites in primary cholangiocarcinoma

Common metabolomic features of patients with primary HCC and CCA

Univariate statistical analysis identified 37 significant DEMs between the HCC and CCA groups with FCs > 1.5 or < 0.67 and P-values < 0.05. Among these, 32 were significantly upregulated, and five were significantly downregulated (Fig. 3A). HCA based on Bray–Curtis distance demonstrated a high overall similarity between the samples (Fig. 3B). PCA of the metabolomics data for the HCC versus Ctrl group revealed a clear distinction in metabolic phenotypes (Fig. 3C). The OPLS-DA classification yielded R²Y = 0.697 and Q² = 0.38 (Fig. 3D). The robustness of the OPLS-DA model was confirmed using 200 permutation tests, with a Q² intercept of −0.532 and a positive slope, indicating no overfitting and confirming the model’s reliability and predictive accuracy (Fig. 3E). Additionally, the ROC curve based on the OPLS-DA model for the HCC and CCA groups had an AUC of 0.988 (Fig. 3F).

Fig. 3

Interaction analysis of untargeted LC‒MS metabolomics in primary liver cancer and cholangiocarcinoma. A Volcano plot of the HCC and CCA groups. B Hierarchical clustering analysis of HCC and CCA samples. C Unsupervised PCA score plots of metabolomic phenotypes between the HCC and CCA groups. Model parameter: R²X = 0.853 (cumulative variance proportion of 2 principal components). D Score plot of the OPLS-DA model to maximize the intergroup differentiation of the metabolomic data between the HCC and CCA groups. E Model parameters: 1 predictive component + 1 orthogonal component, R²Y = 0.697, and Q² = 0.38. OPLS-DA permutation test diagram of the HCC and CCA groups. F ROC plot of the OPLS-DA model differentiating the HCC and CCA groups. G Hierarchical clustering heatmap of differentially abundant metabolites between the HCC and CCA groups. H Correlation heatmaps of differentially abundant metabolites between the HCC and CCA groups. I KEGG pathway enrichment diagram for the HCC and CCA groups

Hierarchical clustering heatmap (Fig. 3G) analysis identified 33 significant DEMs between the HCC and CCA groups (OPLS-DA VIP > 1, P < 0.05), comprised primarily of PCs, LysoPCs, and LysoPEs and mainly involved in lipid metabolism pathways. Additionally, metabolites such as lysine, LPA, LysoPC, PC, SM, and some fatty acids such as palmitoleic acid were upregulated in the HCC group, suggesting that they play crucial roles in membrane structure, signal transduction, and energy metabolism in liver cancer cells. These alterations in metabolite levels reflect significant metabolic reprogramming in HCC, particularly in membrane synthesis, lipid signaling, and apoptosis pathways, thereby highlighting potential therapeutic targets.

Correlation heatmap (Fig. 3H) analysis indicated that the significant DEMS were primarily involved in lipid metabolism pathways, including LysoPC, LPA, PC, and SM. LysoPC was strongly positively correlated with other lipid metabolites, such as LPA, suggesting their crucial roles in membrane lipid synthesis and signal transduction in liver cancer and CCA cells. The strong correlation between PC and SM metabolites further corroborates their importance in cell membrane structure and apoptosis signaling.

KEGG enrichment pathway analysis (Table 3, Fig. 3I, and Supplementary Figure 1E) indicated that the significant DEMs between the primary HCC and CCA groups were predominantly involved in lipid metabolism pathways, including various LysoPCs, PCs, LPAs, and SMs. Since these metabolites play crucial roles in membrane lipid metabolism, structural integrity, signaling, and energy storage, the observed changes suggest a reprogramming of lipid metabolism, affecting cell membrane composition and signaling pathways, which are vital for cancer cell behavior. This lipid metabolism reprogramming in primary HCC and CCA highlights potential metabolic vulnerabilities and therapeutic targets for these cancers. Overall, these changes in metabolite levels reflect significant lipid metabolism reprogramming in primary HCC and CCA, highlighting potential therapeutic targets.

Table 3 Enrichment analysis of metabolic pathways of interrelated differential metabolites in primary liver cancer and cholangiocarcinoma

Identification of DEMs in clinical samples from patients with primary HCC and CCA

Twenty-four significant DEMs were commonly identified in patients with HCC and CCA, with significant enrichment in lipid metabolism pathways such as PE biosynthesis, PC biosynthesis, phospholipid biosynthesis, and bile acid biosynthesis (Fig. 4A-C, Supplementary Figure 2). These elevated DEMs included CAR 18:1, LysoPC 20:3, PC O-34:1, PC 36:3, PC 36:0, PC 36:1, PC 34:1, PC 32:0, PC 30:0, PC 34:0, PC 32:1, PC 35:1, and PC 34:3 (Fig. 4D). The DEMs significantly upregulated in patients with HCC and CCA were primarily enriched in the GPL metabolism pathway within lipid metabolism (Fig. 4E). H&E and IHC staining of select tissue samples from patients with HCC and CCA revealed that the expression of PLA2, a key enzyme in the GPL metabolism pathway, was significantly higher in HCC and CCA tumor tissues than in adjacent non-tumor tissues (Fig. 4F-I). The expression level of PLA2 was approximately 267.6% ± 2.505% higher in the HCC group than in the Ctrl group (P < 0.05) and approximately 99.0% ± 8.963% higher in the CCA group than in the Ctrl group (P < 0.05).

Fig. 4

Identification of differentially abundant metabolites in clinical samples from patients with primary liver cancer and cholangiocarcinoma. A Venn diagram of differentially expressed metabolites in HCC and CCA patients. B-E KEGG pathway enrichment analysis results for commonly differentially expressed metabolites in HCC and CCA patients: bubble plot (B), pathway impact plot (C), metabolite expression level bar plot (D) and metabolic pathway map of glycerophospholipid metabolism (E). F Representative images of H&E staining and IHC staining for PLA2 in human HCC and adjacent nontumor tissue sections. G Representative images of H&E staining and immunohistochemical staining for PLA2 in human CCA and adjacent nontumor tissue sections. H Semiquantitative analysis of (F). I Semiquantitative analysis of (G). HCC, hepatocellular carcinoma. CCA, cholangiocarcinoma. Ctrl, adjacent nontumor tissue sections. The data are presented as the means ± SEMs; n = 3 biologically independent samples. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001

Discussion

Metabolic remodeling is a key characteristic of cancer, and the use of advanced metabolomics techniques to elucidate the metabolic characteristics of cancer is a crucial avenue for cancer treatment [5, 12, 13]. HCC and CCA, both of which are high-incidence and high-mortality malignancies, often rely on complex metabolic reprogramming mechanisms, including enhanced glycolysis, glutamine metabolism, and lipid metabolism, to support cancer cell proliferation and survival [14,15,16]. These metabolic changes not only affect the efficacy of chemotherapy and immunotherapy but also suggest that targeting specific metabolic pathways could aid in early detection and the development of personalized treatment strategies [16,17,18].

This comprehensive, nontargeted metabolomics analysis conducted using the LCMS-9030 UPLC-Q-TOF/MS platform identified the significant upregulation of key metabolites involved in glycerophospholipid metabolism in HCC and CCA patients (Figs. 1, 2 and 3). Lipid metabolism takes a critical part in the proliferation, survival, and metastasis of HCC and CCA cancer cells [19, 20]. AAlterations in lipid metabolic pathways can influence membrane fluidity, signal transduction, and energy storage, thereby promoting cancer progression [21, 22]. Specifically, GPLs such as PC and LysoPC are pivotal in various cellular processes and are often dysregulated in cancer [23, 24]. In HCC and CCA, reprogramming of GPL metabolism is considered a key mechanism driving cancer progression [6, 21]. Studies have shown significant alterations in the lipid profiles of patients with HCC and CCA, particularly a decrease in polyunsaturated fatty acids (PUFAs) and an increase in monounsaturated fatty acids (MUFAs) [21, 25, 26], which are associated with cancer cell proliferation, migration, and immune suppression [21].

The marked upregulation of PLA2 in tumor tissues, as demonstrated by H&E and IHC staining, further highlights the significance of lipid metabolism in cancer progression and its potential as a viable therapeutic target (Fig. 4). Importantly, GPLs and their metabolites, such as LPA and prostaglandin E₂ (PGE2), mediated by PLA2, are believed to play critical roles in HCC and CCA [21, 27, 28]. The cancer-related signaling pathways Rho, Ras/MAPK, PI3K/AKT, and phospholipase C are activated by LPA through its receptors (LPAR1-6). These pathways control cell proliferation, autophagy, migration, and the epithelial-mesenchymal transition (EMT) [21, 29, 30]. PGE2 promotes cell proliferation, invasion, and migration in liver cancer cells through the prostaglandin E receptor 2 (PTGER2/EP2) while inhibiting anti-tumor immune responses [21, 31, 32].

The study’s findings shed light on the metabolic changes linked to HCC and CCA, emphasizing the pivotal role of lipid metabolites in the formation and progress of cancer. The significant rise in the levels of glycerophospholipid-related metabolites in cancer patients suggests their potential as early detection and diagnostic biomarkers. This finding is especially important because early detection of HCC and CCA can significantly improve patient. The upregulation of key enzymes such as PLA2 suggests that inhibiting these enzymes could disrupt the metabolic reprogramming that supports tumor growth and survival. This study opens new avenues for targeted and personalized treatment strategies aimed at improving patient outcomes.

Advantages and limitations

In this study of two major malignancies, the metabolic profiles of HCC and CCA patients were analyzed for the first time, revealing the important role of the phospholipid metabolome in these two major tumors. The lipid metabolism process was found to be an important factor leading to the crosstalk of these two tumors, and the corresponding metabolic targets were identified. This study systematically analyzed the metabolic profiles of patients with primary HCC and CCA, providing detailed insights into the lipid metabolic reprogramming associated with these cancers. By identifying key metabolites and enzymes involved in lipid metabolism, particularly PLA2, this study suggests potential biomarkers for early diagnosis and targeted therapy with direct clinical significance. However, the present results depend on a limited number of samples and may not be completely typical of the broader group of patients with HCC and CCA. In addition, although the study identified key metabolic pathways and enzymes, it did not provide functional validation or mechanistic insights into how these metabolic changes contribute to cancer development. Notably, this study is the first to systematically reveal the expression patterns of significantly differentially expressed metabolites in HCC and CCA, contributing to a deeper understanding of cancer metabolism.

Conclusions

This study revealed the expression patterns of 24 significantly DEMs in primary HCC and CCA patients, highlighting their enrichment in lipid metabolism pathways. The levels of these metabolites, especially those included in the glycerophospholipid metabolism pathway, were significantly elevated in HCC and CCA patients. H&E and immunohistochemical staining revealed that the key enzyme PLA2 in glycerophospholipid metabolism was highly upregulated in HCC and CCA tumor tissues compared with adjacent nontumor tissues.

The clinical significance of this project lies in its systematic revelation of significantly distinctively expressed metabolites in patients with HCC and CAA , as well as their enrichment of these distinctively expressed metabolites in lipid metabolism pathways. These finding provides new targets for early diagnosis and individualized treatment. The study highlights the critical role of lipid metabolism reprogramming in the development of HCC and CCA, particularly the significant upregulation of the key enzyme PLA2 in glycerophospholipid metabolism. The present findings suggest that these metabolites might serve as possible indicators, providing new diagnostic tools for clinical practice. This discovery not only contributes to improving the early diagnosis rate of HCC and CCA, thereby enhancing patient prognosis, but also provides a theoretical foundation for developing more precise and personalized treatment strategies, which are ultimately promising for significantly improving overall patient outcomes.