Patient Cohort

The study population consisted of 65 patients with alcohol-associated hepatitis (Table 1). Of these, 60% (39/65) were male, with a median age of 48.5 years and body-mass index (BMI) of 26.9 kg/m². 19 out of 29 (65.5%) biopsied patients had cirrhosis. The median value of ALT and AST was 49 U/L and 146 U/L, respectively. See additional serum disease markers and scores in Table 1.

Presence of Fecal Gelatinase does not Predict 30-day and 90-day Mortality in Patients with Alcohol-Associated Hepatitis

Of the total, 30.8% (20/65) patients were positive for gelatinase in their stool samples (Table 1). The contribution of each center for gelatinase-positive and gelatinase-negative patients is shown in Table 2. However, because of the limited number of patients from each country, matching of the area using statistical techniques could not be carried out in a meaningful way. Twelve out of 48 (25%) patients were cytolysin positive, and no difference regarding gelatinase-positivity between the cytolysin-positive and cytolysin-negative groups (33.3% vs. 33.3%) was observed. Six gelatinase-positive and 13 gelatinase-negative patients of the entire population had cirrhosis. For five out of the 65 subjects, no mortality data was available and therefore there were 60 patients remaining to study the association of gelatinase with outcomes. There were no significant differences in 30-day survival and 90-day survival between gelatinase-positive (n=17) and gelatinase-negative patients with alcohol-associated hepatitis (n=43) (Figure 1A, 1B). Overall, seven out of 60 (11.7%) and 16 out of 60 patients with alcohol-associated hepatitis (26.7%) were confirmed dead within 30 and 90 days, respectively. Two out of 17 gelatinase-positive patients (11.8%) and five out of 43 gelatinase-negative patients (11.6%) were confirmed dead within 30 days, whereas three out of 17 gelatinase-positive patients (17.6%) and 13 out of 43 gelatinase-negative patients (30.2%) were confirmed dead within 90 days. We subsequently assessed whether the presence of fecal gelatinase could predict mortality of patients with alcohol-associated hepatitis. In our cohort, fecal gelatinase demonstrated poorer performance in predicting 30-day mortality compared with all other tested liver disease markers (FIB-4, ABIC score, Child-Pugh class, MELD, MELD-Na, and Maddrey’s DF), with a low AUC of 0.50 and low best Youden Index of 0.00 compared to the best performer FIB-4, which exhibited an AUC of 0.75 and best Youden Index of 0.54 (Figure 1C, Table 3). Fecal gelatinase also had a sensitivity of 0.29, specificity of 0.72, accuracy of 0.67, a positive predictive value (PPV) of 0.12, and negative predictive value (NPV) of 0.88 for 30-day mortality (Table 3). Similarly, its predictive performance was poor for 90-day mortality, with an AUC of 0.57 and best Youden Index of 0.13, whereas the FIB-4 and ABIC score displayed a significantly better AUC of 0.79 and 0.78 and higher best Youden Index of 0.49 and 0.63, respectively (Figure 1D, Table 4). Additionally, the Cox hazard ratio of death for gelatinase-positive vs gelatinase-negative patients was 0.6152, 95% confidence interval (0.1753, 2.159), p=0.448 for 30 days and 0.6389, 95% confidence interval (0.1818, 2.245), p=0.485 for 90 days.

–

FIGURE 1: Fecal gelatinase does not predict 30-day or 90-day mortality in patients with alcohol-associated hepatitis. (A) Kaplan-Meier curve for 30-day survival in gelatinase positive and gelatinase negative patients with alcohol-associated hepatitis. (B) Kaplan-Meier curve for 90-day survival in gelatinase positive and gelatinase negative patients with alcohol-associated hepatitis. (C) ROC curves of liver disease markers for 30-day mortality in patients with alcohol-associated hepatitis. (D) ROC curves of liver disease markers for 90-day mortality in patients with alcohol-associated hepatitis (Gelatinase n=60, Child-Pugh n=57, FIB-4 n=57, Maddrey’s DF n=52, ABIC n=59, MELD score n=59, and MELD-Na score n=59). ABIC, ‘Age, serum bilirubin, INR, and serum creatinine score’; FIB-4, Fibrosis-4 Index; INR, international normalized ratio; Maddrey’s DF, Maddrey’s Discriminant Function; MELD, model for end-stage liver disease; MELD-Na, sodium-adjusted model for end-stage liver disease; ROC, receiver operating characteristic.

–

Consistent with the findings above, the presence of fecal gelatinase exhibited low feature importance for predicting both 30-day and 90-day mortality according to random forest analysis, with very low mean decrease Gini score and mean decrease accuracy when compared with other established liver disease markers (Figure 2). The FIB-4 and ABIC scores were the best predictors for 30-day mortality per mean decrease Gini (Figure 2A) and mean decrease accuracy (Figure 2B). FIB-4 also had the highest predictive value for 90-day mortality by mean decrease GINI (Figure 2C), whereas the ABIC score had the highest predictive value for 90-day mortality by mean decrease accuracy (Figure 2D).

–

FIGURE 2: Fecal gelatinase is not an important feature for 30-day or 90-day survival in patients with alcohol-associated hepatitis per random forest analysis. (A) Mean decrease Gini score and (B) mean decrease accuracy for 30-day mortality by random forest analysis were quantitated for presence of gelatinase and multiple liver disease markers to determine their respective feature importance for mortality. (C) Mean decrease Gini score and (D) mean decrease accuracy for 90-day mortality by random forest analysis were quantitated for gelatinase and multiple liver disease markers to determine their respective feature importance for mortality (Gelatinase n=60, Child-Pugh n=57, FIB-4 n=57, Maddrey’s DF n=52, ABIC n=59, MELD score n=59, and MELD- Na score n=59). ABIC, ‘Age, serum bilirubin, INR, and serum creatinine score’; FIB-4, Fibrosis-4 Index; INR, international normalized ratio; Maddrey’s DF, Maddrey’s Discriminant Function; MELD, model for end-stage liver disease; MELD-Na, sodium-adjusted model for end-stage liver disease.

Fecal Gelatinase-Positive Patients do not Exhibit More Severe Liver Disease than Gelatinase-Negative Patients

Next, we conducted a comprehensive comparison of various liver disease indicators between the gelatinase-positive and gelatinase-negative groups. The markers assessed included AST, ALT, GGT, AP, bilirubin, FIB-4, Maddrey’s DF, ABIC, MELD score, and MELD-Na score. Our analysis revealed no significant differences between the gelatinase-positive and gelatinase-negative groups (Figure 3). AST levels in gelatinase-positive patients were higher compared with gelatinase-negative patients but the difference was not statistically significant (medians of 167 U/L vs. 128 U/L, P = 0.051) (Figure 3A). The levels of ALT showed similarity between gelatinase-positive and gelatinase-negative patients (medians of 55.0 U/L vs. 42.5 U/L, P = 0.24) (Figure 3B). GGT levels were 189 U/L in gelatinase-positive patients and 162 U/L in gelatinase-negative patients (Figure 3C). The alkaline phosphatase and bilirubin levels between gelatinase-positive and gelatinase-negative patients were very similar (medians of 152 mg/dL vs. 181 mg/dL, P = 0.76, and medians of 13.7 mg/dL vs. 14.1 mg/dL, P = 0.80, respectively) (Figure 3D, 3E). The median value of FIB-4 was 8.64 in gelatinase-positive patients and 6.28 in gelatinase-negative patients. Additionally, the median values of Maddrey’s DF (67.1 vs. 71.7) and ABIC (7.84 vs. 8.12) were lower in gelatinase-positive patients compared with gelatinase-negative patients. Furthermore, the median values of MELD score and MELD-Na score between gelatinase-positive and gelatinase-negative patients were nearly identical (Figure 3F-J). Additionally, we found that seven out of 56 patients (30.4%) with MELD>20 were gelatinase-positive, compared with two out of eight patients (25.0%) with MELD of 20 or lower were gelatinase-positive (P = 0.76, Pearson’s Chi-squared test). Furthermore, ten out of 28 patients (35.7%) on steroids were gelatinase-positive, whereas nine out of 36 patients (25.0%) were not on steroids (P= 0.35, Pearson’s Chi-squared test). Additionally, four out of 15 patients (26.7%) on antibiotics were gelatinase-positive, compared with 15 out of 49 patients (30.6%) not on antibiotics (P = 0.77, Pearson’s Chi-squared test). Finally, we found that neither gender nor overweight or obesity were associated with gelatinase positivity (data not shown).

–

FIGURE 3: Fecal gelatinase positive patients with alcohol-associated hepatitis do not have more severe liver disease than gelatinase negative patients. (A) AST (n=64). (B) ALT (n=63). (C) GGT (n=38). (D) AP (n=63). (E) Bilirubin (n=64). (F) FIB-4 (n=62). (G) Maddrey’s DF (n=56). (H) ABIC (n=64). (I) MELD score (n=64). (J) MELD-Na score (n=64). ABIC, ‘Age, serum bilirubin, INR, and serum creatinine score’; ALT, alanine aminotransferase; AP, alkaline phosphatase; AST, aspartate aminotransferase; FIB-4, Fibrosis-4 Index; GGT, gamma-glutamyltransferase; HE, hepatic encephalopathy; INR, international normalized ratio; Maddrey’s DF, Maddrey’s Discriminant Function; MELD, model for end-stage liver disease; MELD-Na, sodium-adjusted model for end- stage liver disease. P value of equal or less than 0.05 was considered as statistically significant.

Patient Cohorts

A total of 65 patients with alcohol-associated hepatitis were enrolled between June 2014 and January 2019 (Figure S1). Patients were recruited from the Gastroenterology/Hepatology clinics (if outpatient) and hospitals (if inpatient) of the InTeam Consortium across various countries including the USA, Mexico, Canada, UK, France, and Spain. The clinical data and follow-up data of patients (including dates of death) were collected and submitted separately by each hospital/research center using a secure clinical data submission system 8, 37, 38. The patients were followed up until the last time they were seen in clinic or the hospital before being censored or until each center was notified of the confirmed death (with a specific date of death). The inclusion criteria were as follows: recent active alcohol abuse (more than 50 g/day for men and more than 40 g/day for women) within the past three months, coupled with elevated aspartate aminotransferase (AST) levels exceeding alanine aminotransferase (ALT) and total bilirubin >3 mg/dl over the last three months (ClinicalTrials.gov identifier number NCT02075918) 8, 38, 35. Either a clinically indicated liver biopsy or clinical presentation aligning with the diagnosis of alcohol-associated hepatitis was also required. Exclusion criteria included autoimmune liver disease (antinuclear antibody greater than 1:320), chronic viral hepatitis, hepatocellular carcinoma, complete portal vein thrombosis, terminal extrahepatic illness, pregnancy, and absence of signed informed consent 8, 38, 35.

The protocol was approved by the Ethics Committees at various institutions including Hôpital Huriez (Lille, France), Universidad Autonoma de Nuevo Leon (Monterrey, Mexico), Hospital Universitario Vall d’Hebron (Barcelona, Spain), King’s College London (London, UK), University of Alberta (Edmonton, Canada), Yale University (New Haven, USA), University of North Carolina at Chapel Hill (Chapel Hill, USA), Columbia University (New York, USA), University of Pittsburgh (Pittsburgh, USA), University of Wisconsin (Madison, USA), VA San Diego Healthcare System (San Diego, USA), and University of California San Diego (La Jolla, USA). Written informed consent was collected from each patient upon enrollment.

The body-mass index (BMI) was calculated during the clinical office visit. The laboratory markers, including AST, ALT, gamma-glutamyltransferase (GGT), alkaline phosphatase (AP), bilirubin, albumin, international normalized ratio (INR), creatinine, white blood count (WBC), hemoglobin, and platelets were quantified in the hospital laboratories of the respective centers. The Fibrosis-4 Index (FIB-4) equals (age*AST)/(platelets*√ALT) 39. The ‘Age, serum bilirubin, INR, and serum creatinine score’ (ABIC score) corresponds to [(age*0.1) + (bilirubin*0.08) + (creatinine*0.3) + (INR*0.8)] 40. The Child-Pugh class as a severity marker of cirrhosis can be A (5-6 points), B (7-9 points), or C (10-15 points), C being the most severe class. The respective Child-Pugh class is defined by the point sum of the various severity markers: encephalopathy (none = 1 point, grade 1 and 2 = 2 points, grade 3 and 4 = 3 points), ascites (none = 1 point, slight = 2 points, moderate = 3 points), bilirubin (<2 mg/dL = 1 point, 2-3 mg/dL = 2 points, > 3 mg/dL = 3 points), albumin (>3.5 g/dL = 1 point, 2.8-3.5 g/dL = 2 points, <2.8 g/dL = 3 points) and prothrombin time (<4 sec = 1 point, 4-6 sec = 2 points, >6 sec = 3 points)/INR (<1.7 = 1 point, 1.7-2.2 = 2 points, >2.2 = 3 points) 41. Maddrey’s discriminant function (Maddrey’s DF) is defined by [bilirubin + 4.6*(patient’s prothrombin time – control prothrombin time)] 42. The model for end-stage liver disease (MELD) equals [9.57*ln(creatinine) + 3.78*ln(bilirubin) + 11.20*ln(INR) + 6.43] 43. MELD-Na was calculated by MELD + (140 – sodium) − [0.025 × MELD × (140 − sodium)] 44.

Fecal DNA Extraction and Real-time Quantitative Polymerase Chain Reaction (RT-qPCR)

Human stool samples were collected from the patients and immediately stored at -80°C. For DNA extraction, 250 mg of stool were placed in 2 mL screw-cap tubes (Stellar scientific, #T20-C3220-SG). 500 μL of InhibitEX Buffer were added to each stool sample and kept on ice throughout. Samples were homogenized using a bead beater. Subsequently, genomic DNA from human stool was extracted using the QIAamp Fast DNA Stool Mini Kit (Qiagen, #51604) following the provided instructions 45. RT-qPCR was conducted using SYBR qPCR with a touch-down protocol and KAPA-SYBR FAST qPCR master mix (Roche Diagnostics, #07959389001) 46. The qPCR reaction mix for each sample consisted of 10 μL 2x Kapa SYBR, 0.4 μL 100 μM forward primer, 0.4 μL 100 μM reverse primer, 2 μL DNA template, and 7.2 μL water. The PCR conditions were as follows: initial denaturation at 95°C for 30s, followed by 6 cycles of denaturation at 95°C for 3s, annealing and extension at 65°C for 20s, and 72°C for 1s; subsequently, 33 cycles of denaturation at 95°C for 3s, annealing and extension at 58°C for 20s, and 72°C for 1s; finally a melting curve from 65 to 95°C with an increase of 0.5°C per cycle for 5s. The following primer sequences were used: 16S rRNA 47 forward: 5′-GTGSTGCAYGGYTGTCGTCA-3′; reverse: 5′-ACGTCRTCCMCACCTTCCTC-3′. E. faecalis gelE 48 forward: 5′-TATGACAATGCTTTTTGGGAT-3′; reverse: 5′-AGATGCACCCGAAATAATATA-3′. The wild-type E. faecalis V583 and the E. faecalis V583 ∆gelE served as positive and negative controls, while nuclease-free water served as a no-template control. Once amplification was complete, the system automatically generated Ct (cycle threshold) values for each reaction. The average Ct values of E. faecalis V583, E. faecalis V583 ∆gelE, and water were 11.23, 30.53, and 37.26, respectively. A cut-off Ct value of <30 was applied for the presence of gelatinase.

Statistical Analysis

R statistical software (R version 1.3.1093, 2020 the R Foundation for Statistical Computing) was used to perform statistical analyses. Results are described as median and interquartile range for each continuous outcome and as number and percentage for factor variables. We used the pROC library in R to calculate Area under the receiver operating characteristic (ROC) curve (AUC) and best threshold to maximize the Youden Index. We also used the pROC library to calculate sensitivity, specificity, accuracy, positive predictive value (PPV), negative predictive value (NPV) – all at the described respective threshold – and P values between two AUCs per Delong method 49. Kaplan-Meier curves were created and the statistical significance of a difference was calculated to demonstrate the transplant-free survival of gelatinase-positive vs gelatinase-negative groups using the survival and survminer libraries in R; censoring out of participants was indicated by vertical ticks. The random forest library was carried out to calculate mean decrease Gini score and mean decrease accuracy 50. Continuous variables were compared using the Mann-Whitney U test. A two-tailed P value < 0.05 was considered statistically significant.