Plasma untargeted metabolomics reveals the metabolic landscape of NKTCL and aids in liquid biopsy for disease detection Authors: Siyu Qian, Yue Zhang, Mingzhi Zhang, Qingjiang Chen, Xudong Zhang Published: 2025 ABSTRACT BackgroundMetabolic reprogramming is a significant hallmark of NK/T-cell lymphoma (NKTCL). A comprehensive analysis of the alterations in the metabolic landscape of NKTCL is crucial for facilitating early non-invasive screening and providing new therapeutic strategies.MethodsIn this study, we employed a non-targeted approach using liquid chromatography-high-resolution mass spectrometry (LC-HRMS) to analyze the plasma of 92 NKTCL patients and 28 healthy controls, comparing the metabolic profiles of both groups.ResultsWe found significant differences in the plasma metabolic landscape between NKTCL patients and healthy individuals. The differential metabolites were primarily enriched in purine metabolism, arginine and proline metabolism, pyrimidine metabolism, and arginine biosynthesis pathways. Three of these metabolites were used to construct a diagnostic model, demonstrating good diagnostic and therapeutic efficacy.ConclusionsUtilizing LC-HRMS for non-targeted metabolomics can aid in the early screening of NKTCL and provide new therapeutic insights and theoretical foundations.Supplementary InformationThe online version contains supplementary material available at 10.1186/s40001-025-03641-w. Introduction Natural killer/T-cell lymphoma (NKTCL) is a rare and aggressive subtype of non-Hodgkin lymphoma with a higher incidence in East Asia and Latin America [1]. Originating from NK cells or T cells, NKTCL is characterized by unique biological features and clinical presentations, often closely associated with Epstein-Barr virus (EBV) infection [2]. Despite recent progress in treating NKTCL with the DDGP regimen developed at our center, as well as significant improvements in patient outcomes through immunotherapy, many patients remain resistant to conventional therapies or experience high relapse rates [3]. Therefore, there is an urgent need to identify novel therapeutic targets and diagnostic and stratification biomarkers to optimize disease management and offer new clinical treatment options. Liquid biopsy has gained widespread attention in recent years due to its noninvasive nature, ease of sampling, and cost-effectiveness in disease diagnosis [4]. Various genomic and protein-based disease biomarkers have been established through this approach [5]. Metabolic reprogramming, recognized as a hallmark of cancer, plays a pivotal role in tumor metabolism by directly participating in metabolic processes under the influence of genetic modifications and environmental stressors [6]. This makes metabolic indicators a valuable tool for noninvasive diagnosis and novel therapeutic strategies. Advances in technology have enabled more precise detection of a broad range of small molecule metabolites, allowing for high-throughput screening of abnormal metabolites, which has already been validated in colorectal, liver, and lung cancers [7]. As a unique lymphoma subtype, NKTCL exhibits resistance to anthracycline-based chemotherapy, while asparaginase-based antimetabolic therapy has significantly improved patient survival outcomes, highlighting the critical role of metabolomics in NKTCL research [8, 9]. In this study, we employed ultra-performance liquid chromatography-tandem mass spectrometry (UPLC-MS/MS) to perform plasma metabolomic analysis on NKTCL patients and healthy controls (HC). Our aim was to identify differential metabolites between NKTCL patients and healthy individuals, characterize the associated regulatory pathways, and provide novel insights and theoretical foundations for NKTCL diagnosis and treatment. Materials and methods Patients, sample collection, and study design From April 2020 to October 2023, we enrolled a total of 92 patients diagnosed with NKTCL at the First Affiliated Hospital of Zhengzhou University. Patients eligible for this study were required to meet the following inclusion criteria: (1) a pathologically confirmed diagnosis of NKTCL; (2) age between 18 and 70 years; and (3) availability of complete clinical information. Exclusion criteria included: (1) a history of any prior malignancy; (2) severe chronic diseases affecting the heart, liver, kidneys, or other major organs that could interfere with the study outcomes; (3) incomplete clinical data or patients who are unable to provide full clinical information. In addition, 28 healthy controls were recruited for comparison. The inclusion criteria for healthy controls were as follows: (1) no history of malignancy; (2) no chronic diseases, such as hypertension, diabetes, or other metabolic or endocrine disorders; (3) age between 18 and 70 years; (4) the male-to-female ratio was between 2.0 and 2.5; (5) no symptoms and passing a routine medical examination; (6) no other disease history, such as metabolic syndrome or autoimmune disorders. This study is retrospective, and patient data were meticulously gathered from medical records and supplemented by telephone interviews. A summary of the detailed clinical data is presented in Table 1. Table 1Clinical characteristics of patients and controlsCharacteristicNKTCL Group (n = 92)HC Group (n = 28)Age, years46.6 ± 14.548.8 ± 14.0Sex Male6520 Female278 BMI22.94 ± 3.6622.62 ± 4.24CA stage Ⅰ-Ⅱ44– Ⅲ-Ⅳ48–PINK-E score Low(0–1)65– Intermediate(2)16– High(≥ 3)11–ECOGa score < 291– ≥ 21– Bone marrow involvement12–BMI Body Mass Index, CA stage Cancer Annals Staging, PINK(-E) Prognostic index for Natural Killer Cell Lymphoma(-Epstein-Barr virus), ECOG Eastern Cooperative Oncology Group Performance Status Table 1Clinical characteristics of patients and controlsCharacteristicNKTCL Group (n = 92)HC Group (n = 28)Age, years46.6 ± 14.548.8 ± 14.0Sex Male6520 Female278 BMI22.94 ± 3.6622.62 ± 4.24CA stage Ⅰ-Ⅱ44– Ⅲ-Ⅳ48–PINK-E score Low(0–1)65– Intermediate(2)16– High(≥ 3)11–ECOGa score < 291– ≥ 21– Bone marrow involvement12–BMI Body Mass Index, CA stage Cancer Annals Staging, PINK(-E) Prognostic index for Natural Killer Cell Lymphoma(-Epstein-Barr virus), ECOG Eastern Cooperative Oncology Group Performance Status Clinical characteristics of patients and controls BMI Body Mass Index, CA stage Cancer Annals Staging, PINK(-E) Prognostic index for Natural Killer Cell Lymphoma(-Epstein-Barr virus), ECOG Eastern Cooperative Oncology Group Performance Status This study received approval from the Ethics Committee of the First Affiliated Hospital of Zhengzhou University. All participants gave informed consent for the use of their plasma samples, adhering to the principles of the Helsinki Declaration. Plasma was obtained from peripheral blood using EDTA tubes, followed by centrifugation at 1000 g for 10 min. The resulting plasma samples were promptly frozen at − 80 °C until metabolomic analysis. All samples were collected from volunteers who had fasted overnight for at least 10 h. GC/MS analysis A total of 50 µL of plasma supernatant was combined with 200 µL of methanol containing the internal standard, [13C2]-myristic acid, at a concentration of 5 µg/mL. The mixture was vortexed for 5 min and subsequently centrifuged at 20,000 × g for 10 min to precipitate proteins. An aliquot of 100 µL of the supernatant was taken and evaporated to dryness. The resulting residue was derivatized at room temperature using 30 µL of pyridine oxime containing 10 mg/mL methoxyamine for 16 h, followed by silylation with 30 µL of MSTFA + 1% TMCS for 1 h. Next, 30 µL of heptane containing myristic acid methyl ester (15.0 µg/mL) was added as a quality control reference standard, and the mixture was thoroughly combined. Quality control (QC) samples were prepared by pooling aliquots from the study samples. A 0.5 μL portion of the prepared sample was injected into a Shimadzu GC/MS QP2010Ultra/SE, which was equipped with a 30 m × 0.25 mm ID fused silica capillary column chemically bonded with a 0.25 µm DB1-MS stationary phase. The column temperature program began with an initial hold at 80 °C for 3.0 min, followed by a ramp to 300 °C at a rate of 20 °C/minute, and a final hold at 300 °C for 5.0 min. The transfer line temperature was maintained at 220 °C, while the ion source temperature was set to 200 °C. Ions were generated using a 70 eV electron beam at a current of 3.2 mA. After a solvent delay of 160 s, mass spectra were recorded at a rate of 25 spectra per second across a mass range of 50–700 m/z. To minimize variability during instrumental analysis, all samples were analyzed in a randomized order, with QC samples evenly distributed among the injections. Metabolites were identified by comparing the acquired mass spectra and retention indices of the analytes with those in the Wiley 9, NIST (National Institute of Standards and Technology) library 2.0, and an in-house mass spectrometry library. A match factor greater than 80% was considered reliable. Statistical analysis Raw data files generated from UHPLC-MS/MS were processed using Compound Dis coverer 3.3 (CD3.3, Thermo Fisher) for peak alignment, peak picking, and quantification of each metabolite. Statistical analyses for metabolite identification and relative quantification were conducted using R (R version R-3.4.3), Python (Python version 2.7.6), and CentOS (CentOS version 6.6). Annotation was performed using the KEGG database (https://www.genome.jp/keggpathway.tm). Principal Component Analysis (PCA) and Partial Least Squares Discriminant Analysis (PLS-DA) were carried out using metaX. Statistical significance was determined using one-way analysis (t test), with metabolites considered different if VIP > 1, p < 0.05, and fold change > 2 or < 0.5. Volcano plots were generated using R's ggplot2 package, based on the metabolites' log2(Fold Change) and log10 (p values). The functions and metabolic pathways of these metabolites were explored using the KEGG database, with pathways showing a p value < 0.05 deemed statistically significant. Results Reliability of the analytical method The total ion chromatograms (TICs) of plasma samples from NKTCL and HC showed good reproducibility in terms of retention time and signal intensity (Figure S1). One QC sample was injected for every 20 experimental samples to ensure system stability. PCA confirmed high reproducibility, with tight clustering of QC samples in the score plot (Fig. 1A). Further supervised analysis using OPLS-DA revealed clear classification between the two groups, with both the positive and negative modes showing strong separation between NKTCL and HC, demonstrating robust metabolic differences (Fig. 1B, C). The untargeted metabolomics analysis conducted in this study identified a total of 1,228 metabolites, including alkaloids, benzenes, lipids, nucleotides and their derivatives, organic acids and their derivatives, organic nitrogen compounds, organic oxygen compounds, organic heterocycles, and other unknown compounds (Fig. 1D). Differential metabolites with P < 0.05, VIP > 1, and an absolute log2FC value > 1 were selected for further analysis.Fig. 1Global metabolome analysis. A 3D-PCA analysis showing the separation of NKTCL, healthy controls (HC), and quality control (QC) samples. Blue represents NKTCL, red represents HC, and yellow represents QC samples. B, C OPLS − DA score plots showing the metabolic separation between NKTCL and HC. B Positive pattern; C Negative pattern. D Volcano plot of metabolites showing the differential expression of metabolites. Each point represents a metabolite, with color indicating its significance and direction of change (upregulated or downregulated) Global metabolome analysis. A 3D-PCA analysis showing the separation of NKTCL, healthy controls (HC), and quality control (QC) samples. Blue represents NKTCL, red represents HC, and yellow represents QC samples. B, C OPLS − DA score plots showing the metabolic separation between NKTCL and HC. B Positive pattern; C Negative pattern. D Volcano plot of metabolites showing the differential expression of metabolites. Each point represents a metabolite, with color indicating its significance and direction of change (upregulated or downregulated) Identification of differentially expressed metabolites between NKTCL and HC samples A total of 72 differential metabolites were identified based on the criteria (P < 0.05, VIP > 1, and absolute log2FC value > 1, Table S1), including amino acid derivatives, lipid derivatives, nucleotides, vitamins, and carbohydrates (Fig. 2A). Upon applying more stringent criteria (Log2FC > 2, P < 0.05, VIP > 1), we identified 24 metabolites with significant differential expression, including 8 metabolites upregulated and 16 downregulated in the NKTCL group compared to the HC group (Fig. 2B). Figure 2C and D show heatmaps for amino acid and lipid metabolism, respectively, where distinct differences were observed in key metabolites. Notably, abnormalities were found in amino acid metabolism, including L-aspartic acid, arginine, and proline, which are involved in key cellular processes such as protein synthesis and energy metabolism. In addition, significant alterations were observed in lipid metabolism, with hydroxy fatty acids, such as 2-Hydroxycaproic acid, 20-Hydroxy-eicosatetraenoic acid, and 9,10-dihydroxy-octadecenoic acid, showing differential levels. These findings highlight the complex metabolic dysregulation in NKTCL, with implications for its aggressive behavior and potential therapeutic targeting.Fig. 2Differential plasma metabolites between the NKTCL and HC groups. A Volcano plot showing the changes of all metabolites. The x-axis represents the Log2 fold change (Log2FC), and the y-axis represents the -Log10 (adjusted P value). Blue and red dots represent downregulated and upregulated metabolites in NKTCL (P < 0.05, VIP > 1, and |Log2FC|> 1), respectively. B A table listing the differentially expressed metabolites, along with their fold changes (FC) and P values. C, D Heatmaps illustrating the differential levels of amino acids (C) and fatty acids (D) between NKTCL and HC, highlighting metabolic changes associated with NKTCL, including alterations in amino acid and fatty acid metabolism Differential plasma metabolites between the NKTCL and HC groups. A Volcano plot showing the changes of all metabolites. The x-axis represents the Log2 fold change (Log2FC), and the y-axis represents the -Log10 (adjusted P value). Blue and red dots represent downregulated and upregulated metabolites in NKTCL (P < 0.05, VIP > 1, and |Log2FC|> 1), respectively. B A table listing the differentially expressed metabolites, along with their fold changes (FC) and P values. C, D Heatmaps illustrating the differential levels of amino acids (C) and fatty acids (D) between NKTCL and HC, highlighting metabolic changes associated with NKTCL, including alterations in amino acid and fatty acid metabolism Metabolic pathway analysis For a better understanding of the metabolic dysregulation that may contribute to the development of NKTCL, we performed metabolite set enrichment analysis using all metabolites with P < 0.05, VIP > 1, and absolute log2FC value > 1 in the KEGG database. The results revealed significant alterations in several key metabolic pathways, with amino acid and lipid metabolism showing the most prominent changes (Fig. 3A). Figure 3B displays the top 20 significantly enriched pathways, with four pathways ─ purine metabolism, arginine and proline metabolism, pyrimidine metabolism, and arginine biosynthesis ─ demonstrating statistical significance, underscoring their potential role in the pathogenesis of NKTCL (Fig. 4.Fig. 3Differential metabolic pathways between the NKTCL and HC groups. A KEGG classification analysis of the overall metabolic differences between NKTCL and HC, highlighting enriched metabolic pathways. B Specific metabolic pathways enriched in KEGG for the differentially expressed metabolites, identifying key metabolic processes altered in NKTCL Differential metabolic pathways between the NKTCL and HC groups. A KEGG classification analysis of the overall metabolic differences between NKTCL and HC, highlighting enriched metabolic pathways. B Specific metabolic pathways enriched in KEGG for the differentially expressed metabolites, identifying key metabolic processes altered in NKTCL Diagnostic model To develop a simple and interpretable diagnostic model for NKTCL, we narrowed down the selection of metabolites from the top 10 metabolites in Fig. 2B. Considering the high correlation among metabolites, we ultimately selected three specific metabolites [2-Methylbutyl beta-D-glucopyranoside, 4-(2,3-dihydro-1,4-benzodioxin-6-yl)butanoic acid, and 4-(octyloxy)benzoic acid] to construct the diagnostic model, ensuring both clear recognizability and low correlation. The individual diagnostic performance of these three metabolites was evaluated by receiver operating characteristic (ROC) curve analysis (Figure 4A). The performance of the diagnostic model was evaluated using a five-fold repeated ten-fold cross-validation technique, with results provided in the supporting information. The sensitivity and specificity of the five computational algorithms included in the model are summarized in Figure 4B. Taken together, among the five computing algorithms evaluated, random forest demonstrated the best performance, with an average accuracy of 90% and an AUROC curve consistently exceeding 0.9(Figure 4C). Model development results. A ROC curves for 2-Methylbutyl beta-D-glucopyranoside, 4-(2,3-dihydro-1,4-benzodioxin-6-yl)butanoic acid, and 4-(octyloxy)benzoic acid, used to distinguish NKTCL from HC, demonstrating the diagnostic potential of these metabolites. B Sensitivity and specificity evaluation of five computational algorithms used in the diagnostic model. C ROC curve of the five computational algorithms, showing the overall performance of the model in distinguishing NKTCL from HC Discussion Metabolic reprogramming is considered a crucial hallmark of tumors, associated with tumor initiation and progression, and aids cancer cells in rapidly acquiring energy and nutrients to adapt to various environmental stimuli and stresses [10, 11]. In NKTCL, metabolic pathways are also rewired [12–14]. We have identified this characteristic and its sensitivity to asparaginase-based therapies, leading to the development of the DDGP regimen for treating NKTCL, which has demonstrated promising clinical efficacy [3]. The recent studies have shown that both the gut microbiome and refined prognostic stratification systems can shape immune responses and clinical outcomes in NKTCL, indicating that metabolic remodeling may represent a key biological layer linked to treatment sensitivity and patient prognosis [15, 16]. This emphasizes the need for a more comprehensive exploration of the metabolic landscape in NKTCL. Abnormalities in metabolomics reflect downstream effects of genomics, transcriptomics, and proteomics, directly indicating biological dysfunction [17, 18]. This understanding helps elucidate the pathological mechanisms of tumors and aids in comprehending the metabolic reprogramming and biological features of NKTCL, facilitating the identification of key metabolites and their changes, thereby providing new therapeutic perspectives. Tumor-derived metabolites detected in plasma represent major cancer byproducts [19]. We used untargeted metabolomics to analyze plasma samples from NKTCL patients and healthy controls. Our findings confirmed that the plasma metabolic profiles of NKTCL patients significantly differ from those of healthy individuals, with 72 metabolites showing significant changes. The differential metabolites were enriched in pathways related to purine metabolism, arginine and proline synthesis, pyrimidine metabolism, and arginine biosynthesis. Purine and pyrimidine metabolism are crucial for tumor cell proliferation, providing nucleotides for DNA and RNA synthesis. NKTCL, as a highly proliferative lymphoma, relies on enhanced purine and pyrimidine metabolism to support its rapid cell division, driving aggressive tumor growth. Targeting these metabolic pathways could offer a promising strategy for future NKTCL treatments [20]. In NKTCL, both arginine biosynthesis and arginine/proline synthesis pathways are significantly altered, supporting tumor growth and immune evasion. Arginine biosynthesis provides precursors for cell proliferation, while proline synthesis aids in tumor survival and immune modulation [21, 22]. For example, in solid tumors, arginine deprivation therapies, such as ADI-PEG 20, have shown efficacy by depleting extracellular arginine, inhibiting tumor growth and enhancing chemotherapy sensitivity. These findings suggest that targeting arginine and proline metabolism in NKTCL could offer a promising therapeutic approach [23–25]. With the rapid advancements in metabolomics technology, metabolomic analysis has become a powerful tool for discovering biomarkers that aid in early cancer screening. Considering interpretability and linear correlations, we selected three significant metabolites from the top ten differentially expressed compounds to construct a diagnostic model: 2-Methylbutyl beta-D-glucopyranoside, 4-(2,3-dihydro-1,4-benzodioxin-6-yl)butanoic acid, and 4-(octyloxy)benzoic acid, all of which exhibited significant reductions in the peripheral blood of NKTCL patients. Model results showed that random forest demonstrated strong diagnostic efficacy, with the area under the ROC curve reaching impressive levels. 2-Methylbutyl beta-D-glucopyranoside, a carbohydrate derivative belonging to β-glucosides, reflects a potential change in glucose metabolism dependency in NKTCL patients [26]. 4-(2,3-dihydro-1,4-benzodioxin-6-yl)butanoic acid, an organic acid involved in antioxidant and cell signaling processes, indicates weakened resistance to oxidative stress in NKTCL cells [27]. Meanwhile, 4-(octyloxy)benzoic acid, a phenolic acid with antioxidant and anti-inflammatory properties, suggests an immunosuppressive state within the tumor microenvironment. Together, these three metabolites provide insights into the unique characteristics of NKTCL, aiding in early clinical screening and diagnosis [28]. We acknowledge the sample size difference (n = 92 for NKTCL patients vs. 28 healthy controls) as a limitation. The smaller control group may introduce bias, though we minimized this by matching controls with NKTCL patients based on age and sex. Additionally, the single-center design and potential confounders (such as diet and medication) may influence the results. Future studies will expand the control group, increase sample size, and include multicenter recruitment to improve generalizability. Conclusions In summary, we identified key metabolic changes in NKTCL patients, including alterations in purine, arginine, and pyrimidine metabolism. A machine learning-based diagnostic model, using three metabolites, showed potential for screening NKTCL. These findings suggest that noninvasive metabolic profiling could aid early detection and provide a foundation for future NKTCL treatment research. Supplementary Information Additional file 1.Additional file 2.Additional file 3. Additional file 1. Additional file 2. Additional file 3.