Evolutionary metabolic landscape from preneoplasia to invasive lung adenocarcinoma

Abstract

Metabolic reprogramming evolves during cancer initiation and progression. However, thorough understanding of metabolic evolution from preneoplasia to lung adenocarcinoma (LUAD) is still limited. Here, we perform large-scale targeted metabolomics on resected lesions and plasma obtained from invasive LUAD and its precursors, and decipher the metabolic trajectories from atypical adenomatous hyperplasia (AAH) to adenocarcinoma in situ (AIS), minimally invasive adenocarcinoma (MIA) and invasive adenocarcinoma (IAC), revealing that perturbed metabolic pathways emerge early in premalignant lesions. Furthermore, three panels of plasma metabolites are identified as non-invasive predictive biomarkers to distinguish IAC and its precursors with benign diseases. Strikingly, metabolomics clustering defines three metabolic subtypes of IAC patients with distinct clinical characteristics. We identify correlation between aberrant bile acid metabolism in subtype III with poor clinical features and demonstrate dysregulated bile acid metabolism promotes migration of LUAD, which could be exploited as potential targetable vulnerability and for stratifying patients. Collectively, the comprehensive landscape of the metabolic evolution along the development of LUAD will improve early detection and provide impactful therapeutic strategies.

Fig. 1. Schematic overview of the study.

a Overview of the study design. The illustration was created with BioRender.com. b Eight morphological stages from lung preneoplasia to invasive adenocarcinoma. c, d Clinical parameters of the study cohort 1 (c) and cohort 2 (d) were indicated in the heatmap. DFS, disease-free survival; OS, overall survival; CTC, circulating tumor cell (FU/3 mL). e, f Classes and counts of metabolites detected in cohort 1 (tissue samples) and cohort 2 (plasma samples).

Fig. 2. Progressive metabolic evolution from AAH to IAC.

a Partial least squares discriminant analysis (PLS-DA) of the metabolomics data from AHH, AIS, MIA, and IAC patients. b Volcano plots of the significantly differential metabolites in pre-invasive group (AAH/AIS) versus IAC or in MIA versus IAC were shown. Two-sided Wilcoxon rank-sum tests followed by Benjamini-Hochberg (BH) multiple comparison test with false discovery rate (FDR) < 0.05 and fold change > 1.25 or <0.8. Metabolites significantly increased or decreased were colored in purple and green, respectively. c Kyoto Encyclopedia of Genes and Genomes (KEGG) metabolic pathways enriched by significantly differential metabolites in pre-invasive group (AAH/AIS) versus IAC or in MIA versus IAC group. One-sided Fisher’s exact test followed by BH multiple comparison test with FDR < 0.05. d Clustering of metabolic trajectories using differential metabolites among AAH, AIS, MIA, and IAC. Two-sided Kruskal-Wallis tests followed by BH multiple comparison test with FDR < 0.1. e, f Dynamic alterations of metabolites in cluster 1 and cluster 2. The dots represent the mean log₂ relative abundance. Two-sided Kruskal-Wallis tests followed by BH multiple comparison test with FDR < 0.1. g Violin plots of metabolites in cluster 3 and cluster 4. The difference of metabolites among AHH, AIS, MIA, and IAC were evaluated using a two-sided Kruskal-Wallis test. Black dots represent population medians.

Fig. 3. The alteration of circulating metabolites in plasma from patients with different stages.

a Heatmap of metabolite alterations from benign diseases, AIS, MIA, and IAC. Two-sided Kruskal-Wallis tests with P < 0.1. b The receiver operating characteristic (ROC) curve and log₁₀ relative abundance of 3-chlorotyrosine, 12:0-carnitine, glutamate, and phosphocholine (benign diseases, n = 10; AIS/MIA/IAC, n = 82). A two-sided Wilcoxon rank-sum test was used. c The ROC curve and log₁₀ relative abundance of cystine and valine (benign diseases, n = 10; AIS/MIA, n = 54). A two-sided Wilcoxon rank-sum test was used. d The ROC curve and log₁₀ relative abundance of asparagine and cystine (benign diseases, n = 10; AIS, n = 32). A two-sided Wilcoxon rank-sum test was used. In the box plots b–d, the center line represents the median, and the box bounds represents the inter-quartile range. The whiskers span 1.5-fold the inter-quartile range. AUC, Area Under Curve.

Fig. 4. Metabolic stratification of IAC patients and their clinicopathologic correlations.

a Heatmap indicating the relative abundance of metabolites in the identified three metabolomics subtypes. b Clinical parameters of each metabolic subtype were presented. Fisher’s exact test was used. c, d Association of three metabolic subtypes with clinical outcomes including disease-free survival and overall survival in IAC patients. A two-sided log-rank test was used.

Fig. 5. Accumulation of bile acids in metabolic subtype III and its association of with clinical outcome.

a KEGG metabolic pathways enriched by significantly differential metabolites (Two-sided Wilcoxon rank-sum test, FDR < 0.05) in each subtype relative to the other two metabolic subtypes. One-sided Fisher’s exact test followed by Benjamini-Hochberg (BH) multiple comparison test with FDR < 0.05. b Boxplots of the log₂ relative abundance of bile acids in three metabolic subtypes (S-I, n = 33; S-II, n = 63; S-III, n = 30). Two-sided Kruskal-Wallis tests were used. The centerline represents the median, and the box bounds represent the inter-quartile range. The whiskers span 1.5-fold the inter-quartile range. c, d Kaplan–Meier curves predicting the disease-free survival and overall survival of IAC patients stratified by bile acid level with two-sided log-rank P value. The patients were divided into high and low groups by 0.4 quantile of the bile acids levels in IAC patients.

Fig. 6. Aberrant bile acid metabolism promotes migration of invasive LUAD.

a Representative image of immunohistochemical staining of TGR5 and vimentin on 126 IAC patient tumor from S-I (n = 33), S-II (n = 63), and S-III (n = 30) subtypes. Scale bar, 100 μM. b Immunohistochemical (IHC) score of TGR5 and vimentin from patients in S-I (n = 33), S-II (n = 63), and S-III (n = 30) subtypes. A two-sided One-way ANOVA test was used. The centerline represents the median, and the box bounds represent the inter-quartile range. The whiskers span 1.5-fold the inter-quartile range. c Kaplan-Meier curves comparing the disease-free survival and overall survival in IAC patients with a high group (histoscore > 80 in TGR5 expression) versus low group (histoscore ≤ 80 in TGR5 expression). A two-sided log-rank test was used. d Kaplan-Meier curves comparing the disease-free survival or overall survival in IAC patients with a high group (histoscore > 20 in vimentin expression) versus low group (histoscore ≤ 20 in vimentin expression). A two-sided log-rank test was used. e The correlation plot of TGR5 with vimentin expression with significant Pearson’s correlation in TCGA LUAD dataset (n = 525). R, Pearson’s correlation coefficient. f, g Transwell migration assays were performed on H1299 cells treated with CA (100 μM), TCDCA (100 μM), and GCDCA (100 μM) or treated with CA (100 μM) and transfected with or without siTGR5. Representative images (left, scale bar, 100 μm) for three biological repeats and statistical analyses (right, n = 5) of the migrated cells are shown. siCtrl, siControl. Data represent the mean ± s.e.m. and One-way ANOVA followed by Tukey’s multiple comparison test was used. h Diagram showing the experimental design for in vivo metastasis assay (see Methods). i Lung metastasis assay of A549-luciferase cells treated with indicated bile acids (Vehicle, n = 7; CA, n = 8; GCDCA, n = 7; TCDCA, n = 5, example mice shown to left). j Quantification of the metastasis nodules on the pulmonary surface of each groups (Vehicle, n = 7; CA, n = 8; GCDCA, n = 7; TCDCA, n = 5). In i and j, data represent the mean ± s.e.m. and two-tailed Student’s t-test was used.