Background Chronic inflammation and elevated reactive oxygen species are key contributors to hepatocellular carcinoma (HCC) progression. Objective This study aims to investigate the role of the oxidative stress sensor protein Pirin (PIR) as a critical mediator of inflammation in HCC progression. Design We investigated PIR’s role in HCC tumourigenesis through RNA interference, genetic knockout and pharmaceutical inhibition in HCC cell lines and various mouse models. Furthermore, we used transcriptomics, quantitative reverse transcription PCR, western blot, immunofluorescence staining and immunohistochemistry analysis to elucidate the molecular details. Results This study reveals a novel redox-dependent mechanism governing PIR’s nuclear shuttling, contributing to liver inflammation and HCC progression. We identified a positive feedback axis where nuclear PIR amplifies inflammatory responses, leading to hepatitis and HCC advancement. Cytokines in this loop are regulated by PIR-enhanced v-rel reticuloendotheliosis viral oncogene homolog A (RELA) transcription, promoting PIR’s nuclear translocation, increasing proinflammatory cytokine levels, and disrupting redox balance. We confirmed that liver parenchymal cells produce autocrine cytokines supporting their growth and malignancy. Notably, PIR’s redox-mediated nuclear shift can be inhibited by N-acetyl cysteine or PIR inhibitors, reducing HCC promotion in mice. Conclusion We elucidate a novel redox-dependent regulatory mechanism governing the nuclear localisation of PIR and its role in promoting liver inflammation and HCC progression. Our findings underscore the significance of cellular redox status in regulating PIR’s activity and highlight the potential of targeting this pathway with antioxidants to mitigate HCC progression.

Testosterone production by testicular Leydig cells (LCs) in male mammals is energetically demanding and prone to mitochondrial damage. Despite these challenges, LCs exhibit remarkable longevity and minimal turnover, suggesting the existence of specialized mechanisms that maintain LC mitochondrial homeostasis under such constrains. Here we identify a mitochondrial transfer network between LCs and different testicular macrophage (tMac) subpopulations. Leydig cells release extracellular vesicles containing defective mitochondria, which are eliminated by CD206hi tMacs in a TREM2-dependent process. Deletion of Trem2 in tMacs disrupts this transfer, leading to impaired testosterone synthesis. Conversely, LCs acquire extracellular vesicles containing functional mitochondria from MHCIIhi tMacs through ITGβ1-VCAM1 interactions. Loss of Vcam1 in LCs hinders this mitochondrial transfer, thereby compromising testosterone production. Together, our findings reveal an unrecognized mitochondrial transfer network between LCs and tMacs that safeguards LC homeostasis and testosterone production, offering valuable insights into intercellular communication mechanisms that maintain tissue homeostasis.

BACKGROUND:Processing methods for fine-needle aspiration biopsy (FNA) samples mainly include conventional smears (CS) and liquid-based preparations (LBP). There is still debate as to which method is better and the diagnostic value and necessity of combining the two methods remains unclear. The objective of the current study was to compare the diagnostic performance of the two methods and their combined use in thyroid nodules. METHODS:We analyzed thyroid cytopathology data from 16 medical centers between June 2010 and November 2025, comparing nondiagnostic and indeterminate nodules rates across preparation methods. For histologically confirmed samples, diagnostic performance metrics were calculated. Cases with separate CS and LBP descriptions (multi-diagnoses group) were analyzed for diagnostic consistency and performance. RESULTS:In total, 89,392 thyroid FNA cases were included (49,309 CS, 13,161 LBP, and 26,922 combined). The rate of indeterminate nodules was 10.3% (CS), 10.9% (LBP), and 14.8% (combined), while nondiagnostic rate was lowest in the combined group (7.3% vs. 10.5% for CS and 17.9% for LBP). LBP demonstrated higher sensitivity (98.1% vs. 95.0%) and accuracy (97.0% vs. 93.7%) than CS, while combined use provided no significant advantage over LBP alone. In the multi-diagnoses group, CS-LBP concordance among diagnostic samples was 92.9%, with comparable diagnostic performance across all methods. CONCLUSIONS:LBP demonstrated superior diagnostic performance compared with CS, but combined use of both methods provided no significant advantage over LBP alone.

e16304 Background: Prognoses for patients with hepatocellular carcinoma (HCC) remain suboptimal due to the high recurrence rates following curative intended hepatectomy. In this study, we evaluated the safety and efficacy of neoadjuvant Cadonilimab (a bi-specific antibody targeting PD-1 and CTLA-4) in combination with transarterial chemoembolization (TACE) for patients at high risk of recurrence. Methods: In this phase II, single-arm trial, resectable HCC (BCLC stage A/B) participants were enrolled. Patients should meet at least one of the following high-risk criteria: a) Multiple tumors or satellite lesions (≥2); b) tumor diameter > 5cm (single tumor); c) AFP ≥ 400 ug/L; d) Predicted to be positive by the Microvascular invasion prediction model of our center according to preoperative imaging. All patients received two cycles of Cadonilimab(10mg/kg, Q3W) following TACE treatment. After resection, patients received additional 12 months of Cadonilimab (10mg/kg, Q3W). The primary endpoint was the major pathological response(MPR, defined as ≥90% necrosis of the resected tumour),while the secondary endpoints include objective response rate (ORR), one-year recurrence rate and safety. Results: Until December 13th, 2024, a total of 30 patients were enrolled and underwent successful resection. Among the patients, 25(83.3%) were male, with the mean age of 54 years. Fourteen (46.7%) patients had multiple tumors, and 22(73.3%) had tumors larger than 5cm. Nineteen (63.3%) achieved MPR, while26(86.7%) exhibited 70% or greater tumour necrosis. According to mRECIST , the ORR was 83.3% (25/30) and DCR was 100%(30/30). Of these patients, 2 patients achieved complete response (CR), and 23 patients achieved partial response (PR). Additionally, 27 patients maintained stable decease, while only 3 patients received PR as assessed by RECIST1.1. The 1-year recurrence-free survival rate of entire cohort was 90%. Notably, patients who did not achieve MPR had a higher recurrence rate(27.3% vs. 5.3%). Simultaneously, this subgroup also demonstrated a lower ORR (54.6 vs. 100%) per mRECIST. The most common TRAEs of any grade were ALT/AST increase (90%), abdominal pain (86.7%) and bilirubin increase (56.7%). Four patients experienced grade 3 adverse events, including infusion reaction (n = 2), and white blood cell decreased (n = 2). Notably, abdominal pain and fever were more frequently associated with TACE. No grade 4 or 5 events were observed. The incidence of immune-related adverse event (irAE) was 23.3%. Furthermore, it was observed that patients who experienced irAEs tended to demonstrate a better pathological response. Conclusions: Cadonilimab combined with TACE as a neoadjuvant therapy demonstrated acceptable safety profile and promising efficacy for high risk HCC patients. Clinical trial information: NCT05578430 .

BACKGROUND: This study aimed to explore the feasibility and histological basis of pancreatic two-dimensional shear wave elastography (2D-SWE) measurement in predicting post-operative pancreatic fistula (POPF) following partial pancreatectomy and to build a prediction model by pre-operative features. METHODS: 305 pancreatic or periampullary disease patients were enrolled, divided into a training (n = 203) and a validation group (n = 102). POPF was graded as A (or the so-called “biochemical leak”), B, and C. Spearman correlation analysis assessed the histological basis of 2D-SWE measurement value’s (Emean’s) relationship with POPF grades. Multivariate logistic regression analyzed risk factors of POPF occurrence in the training group, and a predictive model was built. The discrimination ability of the model to differentiate POPF grades was analyzed by the area under the receiver operating characteristic curve (AUC). RESULTS: Emean could discriminate between the occurrence and grades of POPF (all AUC > 0.75, p < 0.050). Pancreatic fibrosis and acinar atrophy correlated with both Emean and POPF (all p < 0.050). The higher BMI (p = 0.002), non-dilated pancreatic duct (p = 0.001), and lower Emean (p < 0.001) were the independent risk factors of POPF occurrence and could help discriminate the occurrence of different grades of POPF in both groups (all AUC > 0.80, p < 0.050). CONCLUSIONS: Lower pancreatic 2D-SWE measurements indicated a higher risk of POPF, and the histological basis may lie in 2D-SWE’s correlation with pancreatic fibrosis and acinar atrophy. A model integrating pancreatic 2D-SWE measurements, BMI, and the presence of pancreatic duct dilation could predict POPF risk across varying severity levels.

Purpose To observe liver function changes from pre-operative to post-operative 12 months in patients with treatment-naive hepatocellular carcinoma (HCC) and chronic hepatitis B (CHB) undergoing partial hepatectomy, and explore the utility of peri-operative metrics, including two-dimensional shear wave elastography (2D-SWE) measurements, in predicting the liver function recovery thereof. Methods Totally 253 patients were prospectively enrolled and followed for one year. Data were collected pre-operatively and at post-operative 5th day and 1st, 3rd, 6th, and 12th months. Variance analysis with Tukey's correction explored the changes in albumin-bilirubin (ALBI)-measured liver function. Logistic regression analyzed the relationship between peri-operative indicators and post-operative liver function recovery (ALBI grade 1). The receiver operating characteristic curve assessed the logistic computation's discrimination. Results ALBI score increased on post-operative 5th day compared to pre-operative levels (p < 0.001), then fell below pre-operative levels by the post-operative 1st month (p < 0.001) and stabilized by the 3rd month (all p > 0.050). Liver function recovered in 155 (61.3%) patients. The fewer lesions (p = 0.027), lower 2D-SWE measurement value (p = 0.010), lower pre-operative ALBI score (p < 0.001), and the absence of post-hepatectomy liver failure (PHLF; p = 0.030) were the predicting factors of liver function recovery. The derived computation could differentiate the likelihood of liver function recovery at post-operative 1st, 3rd, and 12th months (all p < 0.050). Conclusions Liver function may stabilize by the post-hepatectomy 3rd month. Pre-operative 2D-SWE measurements, ALBI score, the number of lesions, and PHLF could affect liver function recovery in patients with HCC and CHB.

Abstract Introduction: Ipilimumab (Ipi) blocks the cytotoxic T lymphocyte antigen-4 (CTLA-4) immune checkpoint expressed on T cells, thereby enhancing T cell antitumor activity. Ipi has improved outcomes in several tumor types; however, unfortunately, not all cancer patients benefit from Ipi, and many experience severe immune-related adverse events due to its peripheral effects. Therefore, more effective, less toxic CTLA-4 inhibitors are highly desirable. We previously reported the mechanism of action of a novel anti-CTLA-4 antibody, utilizing a non-fucosylated (NF) Fc region (anti-CTLA-4 NF) thought to augment T-cell priming and modulate regulatory T cells (Tregs). Here, we report the antitumor activity, efficacy, and impact on tumor immune infiltration of the anti-CTLA-4 NF agent in a mouse model of resectable non-small cell lung cancer (NSCLC), as well as the T-cell receptor repertoire changes in a superantigen mouse model. Methods: We injected 129/Sv male mice with luciferase-transfected 344SQ Kras G12D ;p53 R172HΔG mutant NSCLC (344SQ-Luc+) cells. Mice with established flank tumors were randomized to the following perioperative (neoadjuvant and adjuvant) treatment groups: Isotype (control), anti-CTLA-4 Fc-inert, anti-CTLA-4 Ipi-like (classical Ipi), and anti-CTLA-4 NF. Surgical resection of tumors was performed after the last neoadjuvant treatment. Antitumor activity was analyzed by ANOVA test. Lung cancer-specific survival was estimated by Kaplan-Meier method and compared between groups using a log-rank test. Before tumor resection, mice were imaged for bioluminescence assessment of tumors. Resected tumors were immune-profiled using flow cytometry. Surviving mice with no palpable tumors were rechallenged with 344SQ-Luc+ cells, and tumor volumes were measured for 31 days. Using a staphylococcal enterotoxin B-stimulated (SEB) human CTLA-4 knock-in (KI) model, T-cell receptor sequencing was performed in mice treated with anti-CTLA-4 NF and Ipi at pre- and post-treatment timepoints. Results: Neoadjuvant anti-CTLA-4 NF inhibited tumor growth and reduced tumor volumes compared to Ipi-like, Fc-inert and control (P<0.0001). Perioperative anti-CTLA-4 NF also prolonged survival compared to Ipi-like, Fc-inert, and control (P<0.05). Resected tumors from mice treated with neoadjuvant anti-CTLA-4 NF were characterized by numerically increased CD4+ effector memory T cells and decreased Tregs and naïve CD4+ T cells. Surviving mice that were rechallenged with 344SQ-Luc+ cells had impaired tumor growth compared to treatment-naïve mice (P<0.0001). In the SEB CTLA-4 KI model, anti-CTLA-4 NF treatment resulted in an enhanced T-cell repertoire compared to Ipi. Conclusions: Our results suggest that the non-fucosylated Fc of CTLA-4 can enhance antitumor activity, prolong survival, enrich the T-cell repertoire, and potentially impart immunological memory compared to Ipi. Citation Format: Michael Wang, Amy Jhatakia, Xin Sun, Cheuk H Leung, John T Le, Neha Gupta, Jack Lohre, Lili Chen, Heather Y Lin, Wei Hu, Nicholas Wilson, Tina Cascone. Preclinical antitumor activity and immune modulation of a novel non-fucosylated (NF) CTLA-4 inhibitor in a mouse model of resectable non-small cell lung cancer [abstract]. In: Proceedings of the AACR IO Conference: Discovery and Innovation in Cancer Immunology: Revolutionizing Treatment through Immunotherapy; 2025 Feb 23-26; Los Angeles, CA. Philadelphia (PA): AACR; Cancer Immunol Res 2025;13(2 Suppl):Abstract nr B041.

RATIONALE: Regional lymph node metastases are clinical hallmark of lung cancer, and often indicate more dismal prognosis. However, the effect of lung cancer cells on the microenvironment components of lymph nodes and spatial transcriptome atlas remains unclear. Spatially-resolved analysis of cellular and molecular heterogeneity in the tumor and lymph nodes, particularly at the invasive front, could enhance our understanding of the tumor ecosystem and inform more effective treatment strategies. METHODS: Spatial transcriptomics sequencing experiments (SalusSTSTM) and single-cell RNA sequencing (scRNA-seq) were performed on each set of matched primary lung tumor (pTumor), metastatic lymph nodes (mLN) and normal lymph nodes (nLN) from 4 lung cancer patients (Fig. A). Segment Anything and Cellbin algorithm were applied to the spatial transcriptomic data, followed by the unbiased clustering (UMAP). Niche based cluster was used to further analyze the spatial pattern of any molecular signatures, where niche was defined as a collection of each cell and its surrounding cells within a 125μm radius, which is considered to reflect the local microenvironment. Enrichment analysis was performed on string-db.org. RESULTS: Unique Molecular Identifier (UMI) counts and gene counts per 10x10 μm2 bin for each sample are shown in figure panel B, with a median value ranging from 150 to 1079, and 111 to 781 respectively. Hematoxylin and eosin (HE) staining of samples from patient SCLC2 are presented in panel C, with corresponding UMI spatial distribution shown in Fig. D. In nLN, circular germinal center structures are visible on the UMI density map (circles in Fig. D), and in the unsupervised clustering (circles Fig. E). While in mLN and tumor samples, proliferating basal cells and neuroendocrine cells are the most abundant cells. Niche based unsupervised clustering with UMAP algorithm across all samples resulted in 25 clusters and there is one distinct cluster 3 found to be located along the tumor boundary. Differentially expressed genes of this cluster compared to other clusters were found to be highly associated with DNA repair pathways (Fig. G). CONCLUSIONS: This study is the first to map microenvironment components and spatially resolved transcriptome atlas of regional LN in lung cancer based on paired specimens (tumor-nLN-mLN). Significant heterogeneity exists across samples in gene counts and expression patterns, providing rich insights into the unique cellular behaviors and interactions of the LN metastasis. Further spatially-resolved analysis is needed to elucidate tumor cell invasion into the lymph node and their immune evasion mechanisms.

N6-methyladenosine (m6A) stands as the predominant modification in eukaryotic mRNA and is involved in various biological functions. Aberrant m6A has been implicated in abnormal cellular phenotypes, including defects in stem cell differentiation and tumorigenesis. However, the precise effects of m6A on cell proliferation and the underlining mechanism of metabolic gene regulation remain incompletely understood. Here, we established a cellular environment with low-m6A levels and observed a severe impairment of cell proliferation. Mechanistic studies revealed that the depletion of m6A on TIGAR mRNA led to increased expression, subsequently inhibiting glycolysis while promoting the pentose phosphate pathway (PPP). A genome-wide CRISPR-Cas9 screen identified numerous genes involved in cell proliferation that are sensitive to m6A modification, with G6PD emerging as a key regulator. Integration of gene expression and survival data from cancer patients suggested that patients with elevated G6PD expression may exhibit enhanced responsiveness to tumor growth inhibition through m6A suppression. Our findings elucidate the critical role of m6A in cell proliferation, highlighting the therapeutic potential of targeting m6A-mediated metabolic pathways in cancer.

Purpose:Predicting the pathological response after neoadjuvant conversion therapy for initially unresectable hepatocellular carcinoma (HCC) is essential for surgical decision-making and survival outcomes but remains a challenge. We aimed to develop a radiomics model to predict pathological responses. Methods:We included 203 patients with HCC who underwent hepatectomy after neoadjuvant conversion therapy between 2015 and 2023 and separated them into a training set (100 patients from Center A) and a validation set (103 patients from Center B). Pathological complete response (pCR)-related radiomic features were extracted from the largest tumor layer in the arterial and portal vein phases of the CT. A synthetic minority oversampling technique (SMOTE) was used to balance the minority groups in the training set. The SMOTE radiomics model was constructed using a logistic regression model in the SMOTE training set and its performance was verified in the validation set. Results:The AUC of the preoperative modified response evaluation criteria in solid tumors (mRECIST) assessment for pCR was 0.656 and 0.589 in the training and validation sets, respectively. The SMOTE radiomics model was established based on ten radiomic features and showed good pCR-predictive performance in the SMOTE training set (AUC, 0.889; accuracy, 87.7%) and the validation set (AUC: 0.843, accuracy: 86.4%). The RFS of the radiomics-predicted-pCR group was significantly better than that of the predicted-non-pCR group in the training cohort (P = 0.001, 2-year RFS: 69.5% and 30.1% respectively) and the validation cohort (P = 0.012, 2-year RFS: 65.9% and 38.0% respectively). Conclusion:The SMOTE radiomics model has great potential for predicting pathological response and evaluating RFS in patients with unresectable HCC after neoadjuvant conversion therapy.

Artificial intelligence (AI)-based decision support systems have demonstrated value in predicting post-hepatectomy liver failure (PHLF) in hepatocellular carcinoma (HCC). However, they often lack transparency, and the impact of model explanations on clinicians' decisions has not been thoroughly evaluated. Building on prior research, we developed a variational autoencoder-multilayer perceptron (VAE-MLP) model for preoperative PHLF prediction. This model integrated counterfactuals and layerwise relevance propagation (LRP) to provide insights into its decision-making mechanism. Additionally, we proposed a methodological framework for evaluating the explainability of AI systems. This framework includes qualitative and quantitative assessments of explanations against recognized biomarkers, usability evaluations, and an in silico clinical trial. Our evaluations demonstrated that the model's explanation correlated with established biomarkers and exhibited high usability at both the case and system levels. Furthermore, results from the three-track in silico clinical trial showed that clinicians' prediction accuracy and confidence increased when AI explanations were provided.

Neoadjuvant ipilimumab + nivolumab (Ipi+Nivo) and nivolumab + chemotherapy (Nivo+CT) induce greater pathologic response rates than CT alone in patients with operable non-small cell lung cancer (NSCLC). The impact of adding ipilimumab to neoadjuvant Nivo+CT is unknown. Here we report the results and correlates of two arms of the phase 2 platform NEOSTAR trial testing neoadjuvant Nivo+CT and Ipi+Nivo+CT with major pathologic response (MPR) as the primary endpoint. MPR rates were 32.1% (7/22, 80% confidence interval (CI) 18.7-43.1%) in the Nivo+CT arm and 50% (11/22, 80% CI 34.6-61.1%) in the Ipi+Nivo+CT arm; the primary endpoint was met in both arms. In patients without known tumor EGFR/ALK alterations, MPR rates were 41.2% (7/17) and 62.5% (10/16) in the Nivo+CT and Ipi+Nivo+CT groups, respectively. No new safety signals were observed in either arm. Single-cell sequencing and multi-platform immune profiling (exploratory endpoints) underscored immune cell populations and phenotypes, including effector memory CD8+ T, B and myeloid cells and markers of tertiary lymphoid structures, that were preferentially increased in the Ipi+Nivo+CT cohort. Baseline fecal microbiota in patients with MPR were enriched with beneficial taxa, such as Akkermansia, and displayed reduced abundance of pro-inflammatory and pathogenic microbes. Neoadjuvant Ipi+Nivo+CT enhances pathologic responses and warrants further study in operable NSCLC. (ClinicalTrials.gov registration: NCT03158129 .).

Supplementary Table S1 shows the list of primary antibodies used for immunohistochemical staining and immunoblotting assays. Supplementary Table S2 shows the clinical characteristics of 626"driver-gene-negative" LUAD patients from four cohorts. Supplementary Table S3 shows the clinical characteristics of 52 "driver-gene-negative" LUAD patients whose samples were used for microarray analysis. Supplementary Table S4 shows the microarray results of the genes in the Wnt-signaling pathway. Supplementary Table S5 shows the univariate association of 41 candidate genes with overall survival in the training SYSUFH cohort. Supplementary Table S6 shows the C-index of the three enrolled factors in the nomogram in the SYSUFH training cohort. Supplementary Table S7 indicates the summary of relevant findings of ITH analysis of twenty-one patients with "driver-gene-negative" LUAD. Supplementary Figure S1 shows the workflow to determine "driver-gene-negative" status in LUAD patients. Supplementary Figure S2 shows the pathological diagnosis of enrolled LUAD patients. Supplementary Figure S3 shows the study flowchart. Supplementary Figure S4 shows the signaling GO analysis and cellular component classification analysis for the genome-wide microarray. Supplementary Figure S5 shows the β-Catenin expression characteristics in patients with "driver-gene-negative" LUAD from the SYSUFH training cohort. Supplementary Figure S6 shows the different expression profiles of β-catenin in EGFRmt, KRASmt, and ALKft LUAD. Supplementary Figure S7 shows the protein expression characteristics of 41 candidate genes in 8 "driver-gene-negative" LUAD patients. Supplementary Figure S8 shows these representative immunohistochemical staining for 41 candidate molecules in 626 "driver-gene-negative" LUAD patient tissue samples. Supplementary Figure S9 shows kaplan-Meier estimates of progression-free survival (PFS) based on the CSDW signature in the four cohorts. Supplementary Figure S10 shows that the CSDW signature predicts clinical outcomes in all 626 patients with "driver-gene-negative" LAUD. Supplementary Figure S11 shows that the optimum cutoff value of the four prognostic markers was determined using the X-tile program in the training set and Kaplan-Meier analysis of all 626 LUAD samples. Supplementary Figure S12 shows the correlation between the four genes in the CSDW signature and tumor size, lymphatic metastasis and distant organ metastasis. Supplementary Figure S13 shows that kaplan-Meier analysis was used to compare the performance of the CSDW signature in all 626 LUAD patients with different clinicopathological risk factors. Supplementary Figure S14 shows that the nomogram to predict the 3-year and 5-year overall survival (OS) of "driver-gene-negative" LUAD patients. Supplementary Figure S15 shows that time-dependent ROC curves to evaluate the sensitivity and specificity of the nomogram in predicting overall survival in "driver-gene-negative" LUAD patients. Supplementary Figure S16 shows that intra- and inter-sample variability of β-catenin, SOX9, DVL3 and Wnt2b protein levels detected by IHC.

To evaluate the performance of automatic deep learning (DL) algorithm for size, mass, and volume measurements in predicting prognosis of lung adenocarcinoma (LUAD) and compared with manual measurements. A total of 542 patients with clinical stage 0-I peripheral LUAD and with preoperative CT data of 1-mm slice thickness were included. Maximal solid size on axial image (MSSA) was evaluated by two chest radiologists. MSSA, volume of solid component (SV), and mass of solid component (SM) were evaluated by DL. Consolidation-to-tumor ratios (CTRs) were calculated. For ground glass nodules (GGNs), solid parts were extracted with different density level thresholds. The prognosis prediction efficacy of DL was compared with that of manual measurements. Multivariate Cox proportional hazards model was used to find independent risk factors. The prognosis prediction efficacy of T-staging (TS) measured by radiologists was inferior to that of DL. For GGNs, MSSA-based CTR measured by radiologists (RMSSA

Differential diagnosis of pulmonary nodules detected by computed tomography (CT) remains a challenge in clinical practice. Here, we characterize the global metabolomes of 480 serum samples including healthy controls, benign pulmonary nodules, and stage I lung adenocarcinoma. The adenocarcinoma demonstrates a distinct metabolomic signature, whereas benign nodules and healthy controls share major similarities in metabolomic profiles. A panel of 27 metabolites is identified in the discovery cohort ( n = 306) to distinguish between benign and malignant nodules. The discriminant model achieves an AUC of 0.915 and 0.945 in the internal validation ( n = 104) and external validation cohort ( n = 111), respectively. Pathway analysis reveals elevation in glycolytic metabolites associated with decreased tryptophan in serum of lung adenocarcinoma vs benign nodules and healthy controls, and demonstrates that uptake of tryptophan promotes glycolysis in lung cancer cells. Our study highlights the value of the serum metabolite biomarkers in risk assessment of pulmonary nodules detected by CT screening.