Faecal proteomics links neutrophil degranulation with mortality in patients with alcohol-associated hepatitis

Abstract

Objective: Patients with alcohol-associated hepatitis (AH) have a high mortality. Alcohol exacerbates liver damage by inducing gut dysbiosis, bacterial translocation and inflammation, which is characterised by increased numbers of circulating and hepatic neutrophils.

Design: In this study, we performed tandem mass tag (TMT) proteomics to analyse proteins in the faeces of controls (n=19), patients with alcohol-use disorder (AUD; n=20) and AH (n=80) from a multicentre cohort (InTeam). To identify protein groups that are disproportionately represented, we conducted over-representation analysis using Reactome pathway analysis and Gene Ontology to determine the proteins with the most significant impact. A faecal biomarker and its prognostic effect were validated by ELISA in faecal samples from patients with AH (n=70), who were recruited in a second and independent multicentre cohort (AlcHepNet).

Result: Faecal proteomic profiles were overall significantly different between controls, patients with AUD and AH (principal component analysis p=0.001, dissimilarity index calculated by the method of Bray-Curtis). Proteins that showed notable differences across all three groups and displayed a progressive increase in accordance with the severity of alcohol-associated liver disease were predominantly those located in neutrophil granules. Over-representation and Reactome analyses confirmed that differentially regulated proteins are part of granules in neutrophils and the neutrophil degranulation pathway. Myeloperoxidase (MPO), the marker protein of neutrophil granules, correlates with disease severity and predicts 60-day mortality. Using an independent validation cohort, we confirmed that faecal MPO levels can predict short-term survival at 60 days.

Conclusions: We found an increased abundance of faecal proteins linked to neutrophil degranulation in patients with AH, which is predictive of short-term survival and could serve as a prognostic non-invasive marker.

Keywords: ALCOHOL-INDUCED INJURY; ALCOHOLIC LIVER DISEASE.

Figure 1.. Fecal protein patterns are different between controls, patients with alcohol use disorder (AUD) and alcohol-associated hepatitis (AH)

Proteomics was performed from fecal samples of controls (Ctrl, n=19), patients with alcohol use disorder (AUD; n=20) and alcohol-associated hepatitis (AH, InTeam; n=80). A: Human proteins in fecal samples in Ctrl, AUD, and AH (P = 0.001 PERMANOVA test [18]); AH vs. AUD P_adj = 0.132, AH vs. healthy P_adj = 0.003, AUD vs. control P_adj = 0.003. P-values were calculated using the dissimilarity index calculated by the method of Bray-Curtis and Permutational Multivariate Analysis of Variance Using Distance Matrices. B: Heatmap of human proteins in fecal samples was clustered using hierarchical clustering. C: All detected 680 human proteins were tested for significant differences between groups using Kruskal-Wallis test (n=261, blue). Afterwards they were tested for significant differences between AH and AUD patients using Kruskal Wallis Test (n=149, red). 101 proteins overlapped in both groups (violet) and were selected if they showed a progressive increase or decrease between all three groups (n=12, violet with violet background). P value adjustment by Benjamin-Hochberg. D: Protein abundance of the 12 detected proteins of Fig. 1C. Significant difference in between groups using Wilcoxon Test is marked in the plot (* P <0.05; ** P <0.01; *** P < 0.001, **** P<0.0001). Kruskal-Wallis-Test: MPO, myeloperoxidase (P = 9.41e-7); CTSG, cathepsin G (P = 7.52e-8); HBB, hemoglobin beta (P = 3.66e-7); HBD, hemoglobin delta (P = 7.20e-8); LTF, lactoferrin (P = 1.62e-8); LAMP1, lysosome-associated membrane glycoprotein 1 (P = 6.37e-5); ORM2, orsomucoid 2 (P = 1.12e-6); IGLV7–47, immunoglobulin light chain 7–47 (P = 5.47e-5); NEB, nebulin (P = 4.57e-4); MYL11, myosin light chain 11 (P = 1.92e-6); MYL1, myosin light chain 1 (P = 9.92e-6); ATP2A1, sarcoplasmic/endoplasmic reticulum calcium ATPase 1 (P = 7.59e-5)

Figure 2.. Fecal proteins are associated with disease severity and survival in patients with alcohol-associated hepatitis (AH)

A: Correlation heatmap of differentially expressed proteins from Fig. 1C clustered by hierarchical clustering. P values are calculated using Pearson correlation. * P <0.05; ** P <0.01; *** P < 0.001 B: Pearson correlation of the 12 differentially expressed proteins from Fig. 1C with histological, clinical and laboratory parameter. * P <0.05; ** P <0.01; *** P < 0.001 C and E: Model for End-Stage Liver Disease-Sodium (MELD-Na) scores of patients with AH divided into a low and high group according to their fecal MPO ((C), P = 0.012) or nebulin (NEB) protein level ((E), P = 0.044). Optimal cut-off values (cut-off for NEB = 4.33; cut-off for MPO = 6.38) were determined using maximally selected rank statistics (maxstat.test in R package maxstat). P-value was calculated using Wilcoxon-Test. D and F: Kaplan-Meier curve of 60-day mortality for 72 patients with AH. Patients were grouped according to their fecal level of MPO or nebulin, which was determined in (C) or (E), respectively.. Date of liver transplantation was regarded as death date for patients receiving transplantation (n=3), censored patients n=3. MPO: log rank P = 0.035; Nebulin: log rank P = 0.14.

Figure 3.. Fecal proteins are enriched in the neutrophil degranulation pathway in patients with alcohol-associated hepatitis (AH)

Difference in protein abundance was calculated between AH vs. controls, AH vs. AUD patients, and AUD vs. controls using the moderated t-test. A protein was regarded as significantly upregulated when the abundance in one group was 1.5 times higher than in the compared group and the P-value <0.05. Afterwards all proteins with a significantly different abundance >0.6 (p<0.05) were selected and over-representation analysis (ORA) using Reactome pathways was performed. P-values were adjusted using the Benjamin-Hochberg method.

Figure 4.. Overrepresentation Analysis using gene ontology enrichment terms associates increased fecal proteins with immune response and azurophilic granules in patients with alcohol-associated hepatitis (AH)

Difference in protein abundance was calculated between AH vs. controls, AH vs. AUD patients, and AUD vs. controls using the moderated t-test. A protein was regarded as significantly upregulated when the abundance in one group was 1.5 times higher than in the compared group and the P-value < 0.05. Afterwards all proteins with a significant different abundance of >0.6 (P <0.05) were selected and Overrepresentation Analysis (ORA) using gene ontology (GO) terms was performed. P-values were adjusted using the Benjamin-Hochberg method.

Figure 5.. The association between mortality and fecal MPO levels can be validated in an independent cohort of patients with alcohol-associated hepatitis (AH)

A: Difference in protein abundance was calculated between dead and alive patients with AH at day 60 using the moderated t-test. A protein was regarded as significantly upregulated when the abundance in the alive group was 1.5 times higher than in the compared group and the P-value < 0.05. A protein was regarded as significantly downregulated when the abundance in the alive group was 1.5 times lower than in the compared group and the P-value <0.05. Afterwards all proteins with a significant different abundance of >0.6 (P <0.05) were selected and Overrepresentation Analysis (ORA) using gene ontology (GO) terms was performed. P-values were adjusted using the Benjamin-Hochberg method. B: Fecal MPO was measured by ELISA and then normalized for total protein content in feces of patients with alcohol-associated hepatitis (AH) (n=70 in a validation cohort from AlcHepNet). An optimal cut-off value of 0.074 ng MPO/mg protein was calculated using maximally selected rank statistics (maxstat.test in R) to divide the cohort into two groups with low and high fecal MPO. Kaplan-Meier curve of 60-day mortality for patients with AH. (log rank P = 0.037).