Diagnostics of Autoimmune Hepatitis Enabled by Non-Invasive Clinical Proteomics

Abstract

Background: Autoimmune hepatitis (AIH) may be difficult to diagnose and distinguish clinically and biochemically from other chronic liver diseases like metabolic dysfunction-associated steatotic liver disease (MASLD).

Aims: To identify pathways involved in the pathogenesis and identify disease-specific biomarkers of AIH.

Methods: We recruited 19 newly diagnosed patients with AIH, 17 with MASLD, and 19 healthy controls. Liver tissue and plasma were collected, and untargeted mass-spectrometry-based proteomics was performed. For classification of AIH versus MASLD and healthy, machine learning analyses were performed employing logistic regression models on liver and plasma proteome data. Findings were validated using data from the United Kingdom Biobank (UKB).

Results: We identified 7632 liver and 556 plasma proteins with 2521 liver and 227 plasma proteins differing between AIH and healthy, including 56 overlapping. Metabolic dysregulation and systemic immune activation characterised the AIH liver and plasma proteome, respectively. Plasma proteome profiling enabled classification of AIH from MASLD and healthy with an area under the receiver operating characteristic curve of 0.91 (0.09 SD). Validation in the UKB was possible for 8 of 20 diagnostic proteins and showed consistent directional changes. Three proteins (C7, ICAM1, cAST) were significantly different between AIH and MASLD/healthy in unadjusted analyses, and 6 of 8 proteins (C7, ICAM1, cAST, IGFBP3, TIMP1, TTR) were significantly different when adjusting for age and sex.

Conclusions: Clinical proteomic analyses of paired liver-plasma samples from patients with AIH enabled high diagnostic potential. Proteomics may constitute a novel non-invasive diagnostic tool for AIH if validated in larger, age- and sex-matched cohorts.

Clinical trial number: NCT05335603.

Keywords: MASLD; Olink; United Kingdom biobank; artificial intelligence; biomarker; machine learning; mass spectrometry.