Proteomics identifies complement protein signatures in patients with alcohol-associated hepatitis

Diagnostic challenges continue to impede development of effective therapies for successful management of alcohol-associated hepatitis (AH), creating an unmet need to identify noninvasive biomarkers for AH. In murine models, complement contributes to ethanol-induced liver injury. Therefore, we hypothesized that complement proteins could be rational diagnostic/prognostic biomarkers in AH. Here, we performed a comparative analysis of data derived from human hepatic and serum proteome to identify and characterize complement protein signatures in severe AH (sAH). The quantity of multiple complement proteins was perturbed in liver and serum proteome of patients with sAH. Multiple complement proteins differentiated patients with sAH from those with alcohol cirrhosis (AC) or alcohol use disorder (AUD) and healthy controls (HCs). Serum collectin 11 and C1q binding protein were strongly associated with sAH and exhibited good discriminatory performance among patients with sAH, AC, or AUD and HCs. Furthermore, complement component receptor 1-like protein was negatively associated with pro-inflammatory cytokines. Additionally, lower serum MBL associated serine protease 1 and coagulation factor II independently predicted 90-day mortality. In summary, meta-analysis of proteomic profiles from liver and circulation revealed complement protein signatures of sAH, highlighting a complex perturbation of complement and identifying potential diagnostic and prognostic biomarkers for patients with sAH.


INTRODUCTION
Alcohol-associated hepatitis (AH) is an acute, inflammatory clinical syndrome in active and chronic heavy drinkers, and is characterized by rapid onset of jaundice and hepatic decompensation (1-3) with 28-day short-term mortality of up to 30-40% from first presentation in severe cases (2); mortality increasing with concomitant bacterial infections and multi-organ failure (4)(5)(6).Despite the progressive nature and high mortality rate of AH, no available medical therapies provide complete and sustained clinical benefit (7,8).This is in part due to the complex and incompletely understood pathogenesis and diagnostic challenges associated with the disease.
Corticosteroids, the recommended therapy for management of AH (9)(10)(11), only have therapeutic value in a subset of patients (8,12).Due to the diagnostic challenges in patients with AH (13), including the risk, inconvenience, and cost associated with invasive liver biopsy, there is a strong need for the identification and development of biomarkers to effectively diagnose, manage, and predict clinical outcomes in patients with AH (14).Complement, a system of over 50 circulating and membrane-bound proteins, is a vital part of the innate immune system, involved in immune surveillance, clearance of cellular debris and pathogens, and tissue repair, thus contributing to maintaining homeostasis.However, if not properly regulated, it can contribute to uncontrolled inflammation.The complement system is activated by one or more of three pathways: the classical, lectin or the alternative pathway (AP) (15,16).
Accumulating evidence from murine models indicates that complement activation and release of anaphylatoxins contributes to liver inflammation and drives progression of ethanol-induced liver injury (17,18).For example, we and others have shown the involvement of complement in initiation and progression of hepatic inflammation and injury in response to chronic ethanol feeding in mice.Mice with a deficiency in C1q (19) or those treated with human C1 inhibitor, Cinryze, (20) were protected from chronic ethanol-induced liver injury.Further, inhibition of complement receptor 2 (CR2)-Crry-mediated activation of C3 decreased inflammatory responses and hepatic steatosis in ethanol-exposed mice (21).In contrast, complement Factor D (CFD), a component of the alternative pathway, protects mice from chronic ethanol-induced injury (22).Importantly, studies evaluating the impact of AH on the quantity and activation of circulating and hepatic complement provide insights into the involvement of complement in AH.Analysis of complement in liver explants from a small number of patients with AH (n=3-5) revealed increased quantity of immunoreactive C1q, C3, C5, and C5aR (23).In a separate analysis of liver explants from 5 patients, expression of C1qR, C3aR, C5aR, C5aR2 mRNA and C3b, iC3b, C3c protein were also higher than in healthy controls (HC) (22).In patients with AH enrolled by the Defeat Alcoholic Steatohepatitis (DASH) consortium, the concentration of both circulating C5a and CFBa are increased, while CFI and sC5b9 are decreased compared to HCs.Importantly, both CFI and sC5b9 are negatively associated with 90-day mortality in this cohort of patients with AH (24).
Given the dynamic dysregulation of complement in both murine models of ALD and in patients with AH, we hypothesized that the complement system may provide useful biomarkers in this disease process.In the current meta-analysis of circulating and hepatic proteome data from patients with sAH, alcohol cirrhosis (AC) or alcohol use disorder (AUD) and HCs, we identified complement protein signatures of sAH with potential to serve as biomarkers for diagnostic and prognostic use in sAH.
In the serum proteome dataset, a total of 36 complement proteins were found from the total of 1305 identified proteins including the heterotrimers C8αβγ and C1qABC, heterodimer Integrin α2/β3, and the activated form of C4A (or C4Ab) (Supplementary Table 3).Of these, 23 complement proteins were significantly different (Benjamini-Hochberg FDR adjusted p-value < 0.05) across groups: sAH, AUD, and HCs (Table 2).These differences were predominantly between sAH and AUD, except for C1R, and between sAH and HCs, except for CFB, C9, C6, and ELANE.In contrast, only COLEC11, C8αβγ, CFB, C9, and C1R were different between AUD and HCs (Table 2).

Overlap in the complement proteome signatures between liver and serum
Thirteen complement proteins were identified in both liver and serum (Table 3).Eleven of these overlapping complement proteins were regulated in similar directions in sAH, with 8 down-and 3 up-regulated, but the extent of up-and down-regulation varied between liver and serum, except for FCN2, which was downregulated to 0.6 fold in both compartments (Table 3).Of note, 2 proteins, ficolin 1 (FCN1) and vitronectin (VTN), were differentially regulated between liver and serum, with both upregulated in the liver and downregulated in serum (Table 3).Further comparative analysis between the liver and serum complement proteome revealed that of these 11 overlapping proteins, only 5 were significantly different in both compartments between patients with sAH and HCs, including decreases in FCN2, CLU, and CALR and increases in CD59 and CD93 (C1qR1) (Table 3).

Validation of select complement proteins
Results from both liver and serum proteome data were confirmed by performing western blot analysis and ELISA quantification of the complement proteins CD59 and COLEC11, respectively.These proteins were the top up-regulated complement proteins in sAH relative to HC in the liver and serum proteomic datasets, respectively (Tables 1 and 2).In agreement with the liver proteome (Table 1), patients with sAH undergoing liver transplant showed increased expression of hepatic CD59 protein compared to HCs (validation cohort 1) (Figure 2A).Further, in our validation cohort 2, patients with sAH had elevated plasma COLEC11 protein concentration compared to HCs (Figure 2B) in agreement with the serum proteome (Table 2).

Differentially expressed pathways in patients with severe AH compared to healthy controls
Using Reactome molecular pathway analysis (https://reactome.org/), we identified differentially expressed pathways connected to the complement proteins that were impacted by disease state.
One hundred fourteen pathways (57 up-and 57 down-regulated; Supplementary Table 4) were identified in the liver with the top-regulated pathways being Endoplasmic Reticulum (ER) to Golgi anterograde transport; Membrane Trafficking; Coat Protein Complex II (COPII)-mediated vesicle transport; Cargo concentration in the ER; COPI-mediated anterograde transport; and Transport to the Golgi and subsequent modification.CD59 gene was seen to be common with these topregulated pathways in the liver (Table 4).
A total of 177 (89 up-and 88 down-regulated; Supplementary Table 5) pathways were identified in the serum.The top-regulated pathways are those involved in: Binding and uptake of ligands by scavenger receptors; Scavenging by Class A Receptors; and Vesicle-mediated transport.
Interestingly, MASP1, CALR, and COLEC11, all associated with the lectin pathway of complement activation, were common to these top-regulated pathways in the serum (Table 4).

Circulating complement proteins correlate with clinical variables of interest
Since hepatocytes are the main source of most of the circulating complement proteins (27), we sought to understand the relationship between these complement proteins and sAH by doing a correlation analysis.Spearman's correlation analysis between complement proteins and clinical variables of interest showed that serum C1QBP, C8αβγ, F2, FCN1, ITGα2/β3, and MASP1 negatively correlated with MELD, while CD55, CD59, CD93, CFD, CTSG, and FCN3 positively correlated with MELD (Supplementary Table 6).
Further, as albumin can be used as an indicator of the capacity of the liver to synthesize and secrete circulating proteins, we examined the correlation of complement proteins with albumin.
Of the 18 patients with sAH, 10 (55.5%) had received albumin infusions at least once within the 14 days leading up to sample collection (Supplementary Table 7).However, there was no significant difference in the reported index albumin values between those who received albumin infusions and those who did not (Supplementary Figure 1).Of the 36 circulating complement proteins, only 4 correlated with albumin (Supplementary Table 6).Interestingly, of these, serum C1QBP and ITGα2/β3, which were negatively correlated, and CD93, which was positively correlated, are not secretory proteins.FCN1, which is a secreted protein, was negatively correlated with albumin.Taken together, these data indicate that changes in complement in sAH were not associated with decreased secretory capacity of the liver.

Serum complement proteins are associated with 90-day mortality in sAH
In the cohort of patients (test cohort 2) used for the serum analysis, 11 (61%) of the 18 patients with sAH died within 90 days.Amongst all clinical parameters at diagnosis, only MELD was different between those that were dead and alive at 90 days (Supplementary Table 9).However, we found that 8 complement proteins had potential prognostic value.Patients who died had significantly lower relative serum concentrations of MASP1, C1QBP, CALR, F2, C8αβγ, and CFB (Figure 3A) and higher relative serum concentrations of FCN3 and CD93 (Figure 3B).All 8 of these complement proteins were significantly associated (Benjamini-Hochberg FDR adjusted pvalue < 0.05) with 90-day mortality (Figure 3C).Next, we asked whether serum levels of these complement proteins could potentially predict 90-day mortality using multiple regression analysis.
To select the suitable variables to be included in the regression model, a stepwise selection approach was used to identify possible predictors of 90-day mortality in patients with sAH from the following potential risk factor variables: age, sex, creatinine, albumin, MELD, and the 8 complement proteins (MASP1, C1QBP, CALR, F2, C8αβγ, CFB, FCN3 and CD93) with potential prognostic value.Results of this stepwise multivariate analysis identified complement proteins MASP1 and F2 as the only independent predictors of mortality; these proteins were therefore included in the regression model.None of the remaining variables outside the model met the entry criteria.Hosmer and Lemeshow test showed that there was no evidence of a lack of fit in the selected model (p = 0.7854).
To validate this result, we measured circulating levels of MASP1 and F2 by ELISA in validation cohort 3, comparing patients with sAH who were alive to those who died during a follow-up period of 90 days.While our initial serum proteome findings revealed that deceased patients had significantly lower relative serum concentrations of both MASP1 and F2 (Figure 3A), our validation results yielded some nuanced differences.Notably, plasma concentrations of MASP1 were not significantly different between patients who were alive or deceased (Supplementary Figure 2A).
In contrast, for F2, plasma concentrations were significantly lower in deceased patients (Figure 3E), consistent with the results from the serum proteome dataset (Figure 3A).Furthermore, the AUC value of 0.74 from our validation cohort (Figure 3F) closely mirrors the AUC of 0.77 from the serum proteome analysis (Figure 3D), and is higher than AUC of 0.70 for MELD.

Multiple complement proteins discriminate severe AH from HC, AUD, and AC
We investigated whether liver and serum complement proteins could discriminate sAH from HC, AUD, and AC by analyzing ROC curves for these proteins.Thirteen hepatic complement proteins discriminated sAH from HCs with AUC values ranging from 0.76-1.00(Supplementary Table 10).
The top five discriminatory complement proteins were CD59 (AUC: 1.00; sensitivity 100% and specificity 100%), C1QR1 (AUC: 1.00; sensitivity 100% and specificity 100%), FCN1 (AUC: 0.94; sensitivity 100% and specificity 83.3%), MBL2 (AUC: 0.92; sensitivity 100% and specificity 91.7%), and C4BPB (AUC: 0.92; sensitivity 100% and specificity 91.7%) (Figure 4).Also, multiple serum complement proteins discriminated sAH from either HCs or patients with AUD with no clinical evidence of AH (Supplementary Table 11).COLEC11 and C1QBP discriminated sAH from HCs and AUD both with AUC curves of 1.00 (sensitivity 100% and specificity 100%) and 1.00 (sensitivity 100% and specificity 100%), respectively (Figure 5).Importantly, we sought to compare the complement proteome signature between patients with sAH and AC, given the complexities involved in distinguishing these two stages of ALD in the setting of chronic liver disease.First, we compared complement proteins in compensated versus decompensated AC; no significant differences (Benjamini-Hochberg FDR adjusted p-value < 0.05) were seen between these two groups of AC (Supplementary Table 12).Since no differences were seen between compensated and decompensated AC and the limited sample size, we combined both groups to investigate the potential of serum complement proteins as discriminative markers between sAH and AC.Our results showed that 17 complement proteins exhibited significant differences (Benjamini-Hochberg FDR adjusted p-value < 0.05) between these two groups (Table 5).The diagnostic potential of these proteins was further substantiated by AUC values ranging from 0.69 to 0.91.Notably, C1QBP, CD59, C8αβγ, and CD55 emerged as top discriminatory markers, demonstrating both high sensitivity and specificity (Figure 6A & B; Table 5).
Next, in our validation cohort 3, we measured circulating concentrations of MASP1, F2, and COLEC11 by ELISA in patients with sAH, AC and HC.For MASP1, in contrast to the higher relative concentrations in AC compared to sAH in the serum proteome (Table 5), plasma concentrations in the validation cohort were not significantly different between patients with AC compared to sAH or HC (Figure 6C).However, plasma concentrations of F2 were significantly lower in sAH compared to both HC and AC (Figure 6D).Further, plasma concentrations of COLEC11 in sAH were significantly higher compared to both HC and AC (Figure 6E), with a 1.8fold elevation compared to AC, consistent with the differences observed in the serum proteome (Table 5).

Dysregulation of complement proteome dysregulation correlates with pro-inflammatory cytokines in AH
To explore a potential mechanism by which dysregulated complement could drive the progression of AH, we delved into the interplay between complement proteins and inflammatory cytokines using a matched, independent dataset encompassing proteomics and multiplex analysis from patients with sAH (test cohort 3).We identified 45 complement proteins from proteomics and 17 pro-inflammatory cytokines (inclusive of both classical and non-classical markers of inflammation) from the multiplex data.Spearman's correlation analysis revealed both significant positive and negative associations between complement proteins and pro-inflammatory cytokines (Table 6).

DISCUSSION
In this meta-analysis, we sought to identify complement protein signatures in patients with ALD.
Complement protein signatures were manually identified by comparing hepatic and serum proteomic profile of patients with ALD.Expression of multiple components of complement activation pathways, as well as receptors, proteases, regulators, and complement-associated proteins were perturbed in both liver and serum of patients with ALD.Analysis of RNA-seq dataset from livers of patients with sAH and HCs also revealed that multiple components of the complement system were perturbed in patients with sAH.These data provide insights into the pathophysiology of ALD and identify potential therapeutic targets.Importantly, serum complement signatures demonstrated a strong discriminatory ability to distinguish patients with sAH from individuals with AC, AUD and HCs.Furthermore, these signatures were associated with and predictive of 90-day mortality in patients with sAH.
One of the important findings of our proteomic analysis was identification of perturbations in classical and lectin, but not alternative, complement activation pathways.In the classical pathway, multiple complement related proteins involved in the precise regulation of this pathway were impacted by sAH in both the liver and serum.C1R and C1S, serine proteases usually found in complex with inactive circulating C1q (the C1 complex), are required to initiate and complete the classical pathway activation (28,29).Here, hepatic and serum C1R and hepatic C1S (p = 0.09) were increased in patients with sAH.Binding of C1q to its receptor mediates several cellular processes including phagocytic uptake of apoptotic cells and pathogens; interaction of C1q with its receptors and binding proteins can either enhance or inhibit its function (30,31).Expression of C1q, as well as multiple C1q receptors and C1q-binding proteins, were detected in liver and serum and were differentially impacted by sAH.For example, interaction between C1q and C1qR1, also called CD93, a type I membrane glycoprotein, enhances phagocytic function (32).Expression of CD93 was elevated in both liver and serum of patients with sAH compared to HCs.In contrast, the C1q binding proteins, C1QBP and CALR limit binding of C1q to immune complexes thus inhibiting the classical pathway of complement activation (33,34).The interaction of C1QBP and C1q also inhibits complement hemolytic activity (34).Here, both these inhibitory C1q binding proteins were lower in the serum of patients with sAH and hepatic expression of CALR was lower in patients with sAH.Another C1q binding protein, CD209, a C-type lectin receptor, also inhibits the classical pathway via interactions with the IgG binding site of C1q (35).Expression of CD209 was decreased in liver of patients with sAH.Collectively, increased expression of C1R and C1S, involved in activation of the classical pathway, in parallel with decreased expression of multiple C1q binding proteins that inhibit activation of the classical pathway, suggest that patients with sAH may have enhanced classical pathway activity, while increased expression of CD93 may provide enhanced phagocytic function in sAH.
In the lectin pathway, multiple components were also perturbed by sAH in both liver and serum.
Our validation cohort also showed increased plasma COLEC11 concentrations in patients with sAH compared to HCs; very low to undetectable concentrations were found in HCs, consistent with previous reports (39,40).Complex perturbations in the expression of lectin components make it difficult to determine the impact of AH on this pathway.However, pathway analysis of both the liver and serum proteome suggested that the lectin pathway was activated.
This study also focused on complement regulators which are vital for sustaining the balance between complement activation and control in order to protect cells and tissues from unwanted inflammation and complement-mediated damage (41,42).For example, SERPING1 (C1INH), regulates the classical pathway by binding and inactivating C1R and C1S proteases, leading to C1 complex dissociation (43,44).C4BP negatively regulates classical and lectin pathway by preventing both formation of C3 convertase and facilitating its dissociation (45,46).Also, CD55 inhibits both classical and the alternative pathway by inhibiting both the formation of new C3-and C5-convertases and accelerating their dissociation (47,48).Additionally, CFI inhibits all 3 pathways by cleaving C3b and C4b while requiring the presence of other several cofactor proteins including C4BP to function maximally (49,50).CLU, CD59, and VTN are regulators preventing MAC assembly; while CLU and VTN binds to C5b-7, CD59 binds to C5b-8 or C9 (28).Here, these regulators were differentially impacted by sAH.Serum SERPING1 and CD55 were increased while VTN (p = 0.1) and CFI were decreased in patients with sAH compared to HCs, consistent with our previous studies where plasma CFI was decreased in both moderate and sAH compared to HCs (24).Hepatic C4BPA and C4BPB were decreased while hepatic and serum CLU were decreased and CD59 was increased in patients with sAH.Our validation cohort also showed increased expression of hepatic CD59 in patients with sAH compared to HCs.Pathway analysis revealed CD59 to be associated with vesicle trafficking pathways in both liver and serum proteome consistent with previous reports where CD59 aids vesicle signaling in pancreatic β-cells (51).Collectively, decreased expression of CFI coupled with decreased expression of the C4BP's would result in increased C3 cleavage, potentiating complement activation, while increased CD59 expression would likely protect damaged or dying liver cells from complement-mediated cell lysis.
The diagnosis and prognosis of AH in chronic liver disease presents major challenges (13,14), particularly due to the similarities in clinical and laboratory findings in patients with sAH and AC.This complexity underscores the need for developing non-invasive biomarkers for AH.Liver biopsy, not commonly advocated for suspected AH cases, rarely changes the clinical diagnosis, except when atypical clinical features are present, further underscoring the importance of noninvasive markers.In our study, we identified multiple hepatic and serum complement proteins with potential diagnostic abilities to distinguish patients with sAH from those with AC, AUD and HCs.Generally, there were few significant differences in complement proteins between AUD and HC groups, suggesting that heavy drinking alone does not lead to complement dysregulation.Of particular efficacy, COLEC11 in both serum proteome and plasma validation was able to discriminate between sAH and AC.COLEC11 was low to undetectable in plasma of HCs, but was up to 265-and 2-fold higher in patients with sAH and AC, respectively.Thus, determining circulating levels of COLEC11 may offer valuable diagnostic insights for AH.
Importantly, MASP1 and F2 independently predicted 90-day mortality in patients with sAH, with MASP1 outperforming MELD in predictive accuracy, while F2 was just as good as MELD.MASP1 also performed better than the previously reported concordance-statistic (equivalent to AUC) for MELD of 0.86 (52).Despite some differences observed in the mortality prediction of MASP1 between serum proteome and plasma, F2 showed consistent predictive ability in both datasets.
Considering the liver's crucial role in synthesizing coagulation factors like F2, our study's finding of altered F2 in patients with sAH aligns with expectations.Notably, F2 emerged as a significant predictor of mortality in these patients.This is consistent with prior research indicating reduced thrombin production in chronic liver disease, worsening with disease progression (53).Moreover, studies have associated coagulation abnormalities with increased morbidity and mortality in chronic liver disease (54).In our study, the perturbed levels of F2, particularly the significantly lower concentrations in deceased patients at 90 days, underscore its prognostic value in AH.
Delving into the interplay between complement dysregulation and inflammation, we found that CR1L, a negative regulator of complement activation, was negatively associated with proinflammatory cytokines in sAH.Its established role in regulating complement-dependent cytotoxicity is consistent with an involvement in restraining excessive inflammation.The negative correlation with key inflammatory cytokines, notably those involved in acute and chronic inflammatory responses and liver diseases, like IL1β, IL-18, and TNFα (55), reinforces the hypothesis that CR1L might have a protective role in AH.This protective mechanism could be crucial in mitigating liver damage and progression of AH, highlighting the potential therapeutic value of targeting CR1L in managing AH-related inflammation.However, further research is needed to elucidate the exact mechanisms by which CR1L interacts with these cytokines and to explore its potential as a therapeutic target in AH.
The strength of this meta-analysis study include the pooled estimate of effect from multiple data sets which may reduce the probability of false negative results and the use of MELD score in both tests and validation cohorts to categorize the severity of AH.The major limitation is that data from other metabolic dysfunction-associated liver diseases were not included.For instance, some of the complement proteins identified in this study, including C1QBP, C4BPA, CLU, and SERPING1, have also been identified in proteomic analyses as potential biomarkers in metabolic dysfunctionassociated steatotic liver disease/metabolic dysfunction-associated steatohepatitis (MASLD/MASH) (56)(57)(58).However, to the best of our knowledge, no studies to date have reported COLEC11 to be involved in any chronic liver disease.Thus, comparing complement proteome between alcohol-and metabolic-associated liver diseases would provide valuable insights as to the diagnostic specificity of complement proteins in liver diseases.Other limitations are the heterogeneity of study demographics, methods utilized, and data quality of test cohorts.
Integration of additional proteomic studies would provide more robust results and help in identifying potentially novel biomarkers for use in ALD.
In summary, this study provides insights into complement protein signatures and the complex, dynamic perturbation of the complement system in ALD.We show, to the best of our knowledge, for the first time the involvement of the lectin pathway of complement activation in AH.Circulating COLEC11 was positively and MASP1 negatively associated with sAH, both with good discriminatory performance distinguishing sAH from those with AC, AUD and HCs.Interestingly, CR1L was negatively associated with pro-inflammatory cytokines, adding another layer of understanding to the complex interplay of immune responses in ALD.Additionally, circulating

Human Subjects:
Sex as a Biological Variable: Both sexes were included in all Human Subjects cohorts.

Test cohort 1: Liver proteomics data
The hepatic proteomics data were generated by liquid chromatography-mass spectrometry (LC-  (25).Demographic and clinical parameters of this study cohort have been reported (25).

Test cohort 2: Serum proteomics data
The circulating proteomic data were acquired using the aptamer-based, proteomic SomaScan

Test cohort 3: Serum proteomics and multiplex data
Multiplex data from a Luminex assay matched to proteomics data from plasma samples of patients with sAH (n=10) were used.These patients were recruited into the observational arm of the Alcohol Hepatitis Network (NCT03850899).Plasma samples for proteomics were processed using Pierce™ Top 12 (T12) Abundant Protein Depletion Spin Columns (Thermo Scientific #85164).Depleted samples were then analyzed using LC-MS/MS, and the resultant data processed with Progenesis QI and the MASCOT Search Engine.The subset of proteins with a MASCOT score greater than 30, with an emphasis on identifying complement proteins, were analyzed utilizing both the multiplex and proteomics data, we conducted Spearman's correlation analysis to understand the association between the normalized abundance values of these complement proteins and the concentrations of pro-inflammatory cytokines.

Validation Cohorts: Western blot and ELISA
Additional cohorts of patients were used as validation cohorts for Western blots and ELISAs.
Demographic and clinical characteristics of all validation cohorts are provided in Supplementary Table 13.
Validation cohort 1: Immunoblotting of liver tissue was conducted to validate expression of CD59.
Samples were obtained from explanted livers in patients with sAH (n=5) and wedge biopsies from Validation cohort 3: MASP1, F2 and COLEC11 were measured in 115 subjects: 98 and 17 patients with sAH and AC, respectively.Samples from patients with AC together with their clinical and demographic data were obtained from the NOAC biorepository.Samples from patients with sAH together with their clinical and demographic data were obtained from the Alcoholic Hepatitis Network (AlcHepNet) observational study biorepository (NCT03850899).

RNA-seq Analysis
Raw Next, a meta-analysis was performed across the separate RNA-seq studies using the 'metafor' R package (63).Summarized log2 fold change values and a meta-analyzed P value were obtained using a random-effects model with original study log2 fold change values.The summaries for each individual complement protein gene were then plotted using the 'forestplot' package in R (64).

Liver homogenates and Immunoblotting
Frozen liver tissue from human subjects were homogenized in lysis buffer and protein concentration measured using the DC Protein Assay (Bio-Rad, Hercules, CA).Samples were denatured at 37°C in Laemmli buffer for 15 minutes.Samples were separated on 8-16% SDS-PAGE gels (Bio-Rad, Hercules, CA), transferred to polyvinylidene fluoride membranes with a wet transfer apparatus (Bio-Rad), and blocked in 5% milk in TBS-T.Membrane was probed with antibody specific for CD59 (Cell Signaling Technologies, Danvers, MA; #65055) and HSC70 (Santa Cruz Biotechnologies, Carlsbad, CA; sc-7298) was used as loading control.Membrane was developed using Immobilon western developing reagents (Millipore).Chemiluminescence was visualized using iBright FL1500 imaging system (ThermoFisher, Waltham, MA).Arbitrary density of immune-positive bands was quantified using ImageJ software.

Sample collection and ELISA measurement
Blood samples were obtained within 48 hours of patient enrollment.Plasma was then separated, aliquoted, and stored at -80°C until use.Aliquoted samples were thawed on ice prior to measurement.For COLEC11 (Catalog# MBS2883570) and MASP1 (Catalog# MBS2507077), samples were diluted 20-fold in sample diluent provided by the manufacturer, while for F2 (Catalog# MBS2019898), samples were diluted 2000-fold in PBS.All assays were performed according to manufacturer's instruction (MyBioSource, CA, USA).

Reactome Analysis
Molecular pathway analysis was conducted with the analysis tools of Reactome, version 83 (https://reactome.org/).Pathway identifier mapping, over-representation, and expression analysis were merged using the analysis tools.The Pathway Analysis with Down-weighting of Overlapping Genes (PADOG) method was used as the gene set analysis method.This method computes a gene set score by taking the weighted average of the absolute values of moderated gene t-scores (65).The Reactome analysis generated the following parameters: log FC (log2 fold change for AH over HC), AveExpr (average expression in log2 counts per million reads, across all samples), t (moderated t-statistic from the test that the logFC differs from 0), p-value (raw p-value from the test that the logFC differs from 0), adj.P.Val (Benjamini-Hochberg false discovery rather adjusted p-value) and B (log odds that the gene is differentially expressed).An FDR P value < 0.05 was used to determine pathways considered to be significantly overrepresented.

Statistical Analysis
Continuous variables are presented as means ± SEM.Spearman's correlation was used between complement proteins and clinical variables, as well as between these proteins and pro-inflammatory cytokines, point biserial correlation was used for the association between complement proteins and sAH as well as between these proteins and 90-day mortality.Correlation coefficient (r) was used to measure the strength and direction of association and P value to determine the significance of the correlation coefficient.Receiver operating characteristic (ROC) curves were generated to evaluate the diagnostic and 90-day mortality prognostic potentials of complement proteins in AH.Only the complement proteins that were significantly different between groups were used for ROC analysis.To limit overfitting, the leave-one-out cross validation (LOOCV) method was used while fitting in a general linear regression model.In order to select the suitable variables to establish the prognostic model, the stepwise multiple regression selection approach was used.A significance level of 0.3 (SLENTRY=0.3) and 0.35 (SLSTAY=0.35) is required to allow a variable into the model and to stay in the regression model respectively.SAS (SAS ® Enterprise Guide ® 8.2) was used for stepwise regression analysis and ANOVA analysis using the GLM procedure.The R package 'Corrplot' was used to generate the correlation matrix while the package 'pROC' was used to generate the ROC plots.Youden's index from the LOOCV procedure was used to assess the performance of the complement proteins; Area under the curve (AUC), optimal threshold and corresponding sensitivity and specificity values are reported.All other statistical tests, as well as volcano plots, box-and-whiskers plots, were performed with GraphPad prism 9. Box-and-whiskers plots show each individual value as a point superimposed on the graph, minimum and maximum values, lower and upper quartile as well as median.A false discovery rate (FDR) approach was used to control for multiple testing, and Benjamini-Hochberg FDR adjusted P values < 0.05 were considered statistically significant except otherwise stated.

Study Approval:
All studies were approved by the Institutional Review Board of all participating institutions and all study participants consented prior to data and sample collection.
healthy donors (n=5) during liver transplantation from the Clinical Resource for Alcoholic Hepatitis Investigations at JHU (R24AA025017).Validation cohort 2: COLEC11 was measured by ELISA in 134 subjects: 24 HCs and 110 patients with sAH.HCs together with their clinical and demographic data were obtained from the Northern Ohio Alcohol Center (NOAC) biorepository (NCT 03224949).Patients with a clinical diagnosis of severe AH (MELD score ≥ 20) at admission were recruited as part of the Defeat Alcoholic Steatohepatitis (DASH) consortium, a multicenter (Cleveland Clinic, University of Texas Southwestern, University of Massachusetts and U of L), randomized, double-blind controlled trial (59) (NCT01809132 and NCT03224949).Details of patient recruitment, as well as the inclusion and exclusion criteria for the DASH consortium have been previously reported by Dasarathy et al. (59).
counts were obtained from three different bulk RNA-seq data; Massey et al. (2021) (11 HC and 10 AH from liver explant tissue) (60), Johns Hopkins (7 HC and 13 AH from liver explant tissue) (61) and Argemi et al. (2019) (divided into 2 subsets -10 HC and 11AH liver explant tissue and 10 HC and 18 AH non-explant liver tissue) (62).The R package DESeq2 was used to analyze the log2 fold change for AH vs HC for each of the 4 comparisons (3 datasets of explants and the 1 non-explant).

Figure 6 Figure 6 .
Figure 6 Two group comparison of continuous variables was done either by unpaired t-test or Mann-Whitney U test based on the results of Shapiro-Wilk normality test (except when stated otherwise).For multiple group comparison of continuous variables, analysis of variance (ANOVA) using the general linear models (GLM) procedure was done.Data were log transformed as necessary to obtain a normal distribution, with follow-up comparisons done by lease square means testing.Categorical variables are presented as counts and percentages with Fisher's exact tests used for comparison.Correlation analysis was used to assess the association of complement proteins, continuous clinical variables, sAH, 90-day mortality, and pro-inflammatory cytokines.While

Table 2 .
Serum complement proteome: relative abundance compared to HC: Test cohort 2

Table 4 .
Top-regulated pathways in liver and serum proteome: Test cohorts 1 and 2