Diagnostic Signatures for Lung Cancer by Gut Microbiome and Urine Metabolomics Profiling

Abstract

<h3>Purpose/Objective(s)</h3> To develop the diagnostic signatures for lung cancer (LC) by gut microbiome and urine metabolomics profiling with multi-omics approach. Additionally, to investigate the predictive role of gut microbiome on the prognosis of locally advanced non-small cell lung cancer (LANSCLC) patients after concurrent chemoradiotherapy (CCRT). <h3>Materials/Methods</h3> Fecal and urine samples were collected in LC patients and healthy individuals. The total cohort had been divided into the training set (47 patients and 20 healthy individuals) and the validation set (48 patients and 30 healthy individuals) for multi-omics analysis. Gut microbiota was analyzed through the 16S ribosomal RNA gene sequencing and shotgun metagenomics. Urine untargeted metabolomics was analyzed by liquid chromatography-mass spectrometry. Multi-omics diagnostic model incorporating features from fecal microbiome and urine metabolites was developed in the training set, and then validated in the validation set. Among them, 36 patients with LANSCLC were selected for prognosis analysis, with fecal samples collected at three longitudinal time points (baseline, 2 weeks after the start of CCRT and the end of CCRT). These patients were divided into 2 groups according to progression-free survival (PFS) (PFS > = 12 months [LP], n = 14; PFS < 12 months [SP], n = 22). The microbiome profiling of LP and SP were compared using Wilcoxon rank-sum test. <h3>Results</h3> LC patients had a significant shift of microbiota composition and functional genes distribution compared with healthy individuals. Diagnostic model for LC had been achieved by combining gut microbiome and urine metabolomics profiling, the area under curve (AUC) of the training set was 0.9997, and 0.9769 for the validation set. We also found a significant decrease of Prevotella in LC patients and it was negatively correlated with multiple functions in KEGG level 2. For the 36 patients with LANSCLC who had underwent CCRT, the alpha diversity was significantly increased in LP group compared with SP group at 2 weeks after CCRT (Chao1: <b>P</b> = 0.013, PD: <b>P</b> = 0.037, Shannon: <b>P</b> = 0.011). Further analysis of microbiota composition identified 21 differentially abundant species (<b>P</b> < 0.05). 19 and 2 species were LP-enriched and SP-enriched, respectively. 5 Lachnospiraceae genus, 3 Eubacterium genus were significantly enriched in LP group, while Prevotellaceae and Lactobacillus were significantly enriched in SP group. <h3>Conclusion</h3> The current study employed a multi-omics approach, analyzed the fecal microbiome and urine metabolites from LC patients using a standardized pipeline, and identified disease-associated microbiome shifts across datasets using statistical analysis. We inferred a microbiome-metabolite catalog and its molecular pathway and functional link to host targets, which could be further utilized to aid lung cancer diagnosis and treatment decision. Furthermore, gut microbiota at 2 weeks after the start of CCRT might have a predictive role on the prognosis of NSCLC after CCRT.