Porous Silicon Particle-Assisted Mass Spectrometry Technology Unlocks Serum Metabolic Fingerprints in the Progression From Chronic Hepatitis B to Hepatocellular Carcinoma

• Published in: ACS applied materials & interfaces Vol. 17; no. 4; pp. 5893 - 5908
• Main Authors: Jiang, Xinrong, Tao, Liye, Cao, Shuo, Xu, Zhengao, Zheng, Shuang, Zhang, Huafang, Xu, Xinran, Qu, Xuetong, Liu, Xingyue, Yu, Jiekai, Chen, Xiaoming, Wu, Jianmin, Liang, Xiao
• Format: Journal Article
• Language: English
• Published: United States American Chemical Society 29.01.2025
• Subjects: Adult; algorithms; Biological and Medical Applications of Materials and Interfaces; biopsy; blood serum; Carcinoma, Hepatocellular - blood; Carcinoma, Hepatocellular - diagnosis; Carcinoma, Hepatocellular - metabolism; Carcinoma, Hepatocellular - pathology; chemometrics; chronic hepatitis B; cohort studies; desorption; Disease Progression; Female; Hepatitis B, Chronic - blood; Hepatitis B, Chronic - metabolism; hepatoma; Humans; liver cirrhosis; Liver Neoplasms - blood; Liver Neoplasms - diagnosis; Liver Neoplasms - metabolism; Liver Neoplasms - pathology; Male; mass spectrometry; metabolites; Metabolomics; Middle Aged; Porosity; silicon; Silicon - chemistry; Spectrometry, Mass, Matrix-Assisted Laser Desorption-Ionization - methods; transportation; viruses; laser desorption ionization; serum metabolic fingerprinting; liver disease diagnosis; porous silicon; machine learning algorithm
• ISSN: 1944-8244; 1944-8252; 1944-8252
• DOI: 10.1021/acsami.4c17563

Hepatocellular carcinoma (HCC) is a common malignancy and generally develops from liver cirrhosis (LC), which is primarily caused by the chronic hepatitis B (CHB) virus. Reliable liquid biopsy methods for HCC screening in high-risk populations are urgently needed. Here, we establish a porous silicon-assisted laser desorption ionization mass spectrometry (PSALDI-MS) technology to profile metabolite information hidden in human serum in a high throughput manner. Serum metabolites can be captured in the pore channel of APTES-modified porous silicon (pSi) particles and well-preserved during storage or transportation. Furthermore, serum metabolites captured in the APTES-pSi particles can be directly detected on the LDI-MS without the addition of an organic matrix, thus greatly accelerating the acquisition of metabolic fingerprints of serum samples. The PSALDI-MS displays the capability of high throughput (5 min per 96 samples), high reproducibility (coefficient of variation <15%), high sensitivity (LOD ∼ 1 pmol), and high tolerance to background salt and proteins. In a multicenter cohort study, 1433 subjects including healthy controls (HC), CHB, LC, and HCC volunteers were enrolled and nontargeted serum metabolomic analysis was performed on the PSALDI-MS platform. After the selection of feature metabolites, a stepwise diagnostic model for the classification of different liver disease stages was constructed by the machine learning algorithm. In external testing, the accuracy of 91.2% for HC, 71.4% for CHB, 70.0% for LC, and 95.3% for HCC was achieved by chemometrics. Preliminary studies indicated that the diagnostic model constructed from serum metabolic fingerprint also displays good predictive performance in a prospective observation. We believe that the combination of PSALDI-MS technology and machine learning may serve as an efficient tool in clinical practice.