Research Article

High prevalence of IgA class anti-neutrophil cytoplasmicantibodies (ANCA) is associated with increased risk of

bacterial infection in patients with cirrhosis

Maria Papp1,⇑, Nora Sipeki1, Zsuzsanna Vitalis1, Tamas Tornai1, Istvan Altorjay1, Istvan Tornai1,Miklos Udvardy2, Kai Fechner3, Silvia Jacobsen3, Bianca Teegen3, Andrea Sumegi4, Gabor Veres5,

Peter Laszlo Lakatos6, Janos Kappelmayer7, Peter Antal-Szalmas7

12nd Department of Medicine, Division of Gastroenterology, University of Debrecen, Debrecen, Hungary; 22nd Department of Medicine, Divisionof Hematology, University of Debrecen, Debrecen, Hungary; 3EUROIMMUN Medizinische Labordiagnostika AG, Lübeck, Germany; 4VascularBiology, Thrombosis and Hemostasis Research Group, Hungarian Academy of Sciences, Debrecen, Hungary; 51st Department of Pediatrics,

Semmelweis University, Budapest, Hungary; 61st Department of Medicine, Semmelweis University, Budapest, Hungary; 7Department ofLaboratory Medicine, University of Debrecen, Debrecen, Hungary

Background & Aims: Anti-neutrophil cytoplasmic antibodies(ANCA) are a non-uniform family of antibodies recognizingdiverse components of neutrophil granulocytes. ANCA formationmight be induced by protracted bacterial infections or probablyreflect an abnormal immune response to commensal microorgan-isms. Bacterial infections are common complications in cirrhosiswith high incidence of episodes caused by enteric organisms,therefore, we sought to study the presence and clinical impor-tance of ANCA in cirrhosis.Methods: Sera of 385 patients with cirrhosis of different etiolo-gies were assayed for ANCA of IgG, IgA, IgA1, IgA2, and secretoryIgA subtypes by indirect immunofluorescence and ELISAs. Thecontrol group comprised 202 patients with chronic liver diseases

Journal of Hepatology 20

Keywords: Anti-neutrophil cytoplasmic antibodies; Cirrhosis; Bacterial infection.Received 26 June 2012; received in revised form 12 April 2013; accepted 16 April2013; available online 29 April 2013⇑ Corresponding author. Address: 2nd Department of Medicine, Division ofGastroenterology, University of Debrecen, Nagyerdei krt. 98, H-4032 Debrecen,Hungary. Tel./fax: +36 52 255 152.E-mail address: papp.maria@med.unideb.hu (M. Papp).Abbreviations: ANCA, anti-neutrophil cytoplasmic antibody; IIF, indirect immu-nofluorescence; C-ANCA, cytoplasmic anti-neutrophil cytoplasmic antibody; P-ANCA, perinuclear anti-neutrophil cytoplasmic antibody; PR-3, proteinase-3;MPO, myeloperoxidase; TBB-5, human b tubulin isotype-5; BT, bacterial trans-location; MELD, model for end-stage liver disease; SBP, spontaneous bacterialperitonitis; PPI, proton pump inhibitor; chronic HCV, chronic hepatitis C virus;CSI, clinically significant bacterial infection; CRP, high-sensitivity C reactive pr-otein; PCT, procalcitonin; PMN, polymorphonuclear cell; PBC, primary biliary c-irrhosis; AMA, antimitochondrial antibodies; sIgA, secretory IgA; SC, secretorycomponent; ELISA, enzyme-linked immunosorbent assay; ASCA, anti-Saccharo-myces cerevisiae antibodies; OMP, outer membrane protein; SD, standard devi-ation; IQR, interquartile range; ANOVA, one-way analysis of variance; OR, oddsratio; HR, hazard ratio; 95% CI, 95% confidence intervals; CLD, chronic liver dis-eases; BAFF, B cell activating factor; APRIL, a proliferation-inducing ligand; TGF,transforming growth factor; IL, interleukin; TLR, toll-like receptor; TNFa, tumornecrosis factor alpha; iNOS, inducible nitric oxide synthase; sCD89, soluble CD89;mIgA, monomeric immunoglobulin A; pIgA, polymeric immunoglobulin A; LBP,serum lipopolysaccharide-binding protein; anti-PPS, anti-pneumococcal polysac-charide antibody.

without cirrhosis and 100 healthy subjects. In cirrhosis, a 2-yearfollow-up, observational study was conducted to assess a possi-ble association between the presence of ANCA and clinically sig-nificant bacterial infections.Results: Prevalence of ANCA IgA was significantly higher in cir-rhosis (52.2%) compared to chronic liver diseases (18.6%) orhealthy controls (0%, p <0.001 for both). ANCA IgA subtypingassays revealed marked increase in the proportion of IgA2 sub-type (46% of total ANCA IgA) and presence of the secretory com-ponent concurrently. Presence of ANCA IgA was associated withdisease-specific clinical characteristics (Child-Pugh stage andpresence of ascites, p <0.001). During a 2-year follow-up period,risk of infections was higher among patients with ANCA IgA com-pared to those without (41.8% vs. 23.4%, p <0.001). ANCA IgA pos-itivity was associated with a shorter time to the first infectiouscomplication (pLogRank <0.001) in Kaplan–Meier analysis andwas identified as an independent predictor in multivariate Cox-regression analysis (HR:1.74, 95% CI: 1.18–2.56, p = 0.006).Conclusions: Presence of IgA type ANCA is common in cirrhosis.Involvement of gut mucosal immune system is in center of theirformation and probably reflects sustained exposure to bacterialconstituents.� 2013 European Association for the Study of the Liver. Publishedby Elsevier B.V. All rights reserved.

Introduction

Anti-neutrophil cytoplasmic antibodies (ANCA) are a group ofheterogeneous antibodies. Two basic ANCA patterns are detect-able in serum samples by indirect immunofluorescence (IIF) onnormal peripheral blood neutrophils: the cytoplasmic (C-ANCA)and the perinuclear (P-ANCA). These antibodies are used to diag-nose and monitor the inflammatory activity in primary small ves-sel vasculitides. Their target antigens are well characterized:proteinase-3 (PR-3) and myeloperoxidase (MPO) for most

13 vol. 59 j 457–466

Research Article

C-ANCA and P-ANCA, respectively. In vasculitis, it has been sug-gested that ANCA may have a pathogenic role by driving theautoimmune process [1].

ANCAs are also found in a variety of non-vasculitic clinicalconditions, namely inflammatory bowel disease [2], rheumatoidarthritis [3], chronic autoimmune liver diseases [4,5], and variousinfections [6]. P-ANCA in these diseases is different from classicP-ANCA with regards to both antigen specificity and staining pat-tern [7]. For distinction, it was called atypical P-ANCA, however,various different names, like x-ANCA, pANNA or DNA-ANCA alsoexist in published literature. Atypical P-ANCAs directed against avariety of still ill defined (nuclear or nuclear associated cytoplas-mic) antigens of neutrophil granulocytes. Their clinical or patho-physiological significance has not been clarified yet.

Formation of ANCA may be related to the inflammatory pro-cesses. It was suggested that the death of neutrophils involvedin the inflammatory response might exceed their scavengingcapacity resulting in the release of cytosolic proteins of neutro-phils locally at the site of inflammation, and thereby, initiatingan autoimmune response [8]. Another hypothesis is that the gen-eration of ANCA is linked to the cross-reaction between certainbacterial proteins and host antigens [9]. It is also possible thatprolonged infections can trigger the development of ANCA bymolecular mimicry [6]. Interestingly, it has been reported thatthe presence of atypical P-ANCAs may reflect an abnormalimmune response to intestinal microorganisms [10]. In autoim-mune liver disorders, atypical P-ANCA is directed against humanb tubulin isotype-5 (TBB-5) and cross-react with the bacterialprotein FtsZ. This results from the fact that TBB-5 shares anextraordinarily high structural homology with this microbial celldivision protein, which is present in almost all bacteria of theintestinal microflora [10]. Occurrence of P-ANCAs has also beenconsidered as a sign of immunological response to enteric bacte-rial antigens in other diseases [11,12]. Moreover, there is lack ofP-ANCAs in animal models in a germ-free environment [13].

Our group previously reported [14] that the presence of anti-microbial antibodies was common in patients with cirrhosismainly in those with advanced diseases and portal hypertension,suggesting that serological response to various microbial compo-nents might be the consequence of sustained exposure to micro-bial antigens. Multiple levels of immune dysfunction have beenreported in patients with cirrhosis rendering them susceptibleto bacterial infections [15], which aggravated the course of illnessand was associated with significant morbidity and mortality [16].An important feature of infections in cirrhosis is the high inci-dence of episodes caused by enteric organisms [17]. Bacterialtranslocation (BT), or in other words, the passage of bacteria orbacterial products from the gut into the circulation, is a majormechanism in the development of these infections and has aremarkable impact on the course of the underlying disease[18,19]. Thus, we can hypothesize that ANCAs are also frequentlypresent in patients with cirrhosis and their presence may be asso-ciated with the clinical course of the disease and infectious com-plications caused by bacteria. At present however, there are nocomprehensive data concerning ANCAs in cirrhosis and itscomplications.

The aims of this study were to investigate (1) the prevalenceand type of ANCA in a large cohort of patients with cirrhosis;(2) associations between the presence of ANCA and disease sever-ity or existence of portal hypertension; (3) whether presence ofANCA is associated with the development of clinically significant

458 Journal of Hepatology 201

bacterial infections; and (4) the possible origin of cirrhosis-asso-ciated ANCA.

Patients and methods

Patients

Sera of 385 consecutive patients with cirrhosis of different etiologies (male/female: 206/179, age: 56.6 ± 11.0 years) were collected at the GastroenterologyDivision of the 2nd Department of Medicine (Debrecen University) betweenMay 2006 and April 2009. Mean disease duration from diagnosis of cirrhosiswas 3.9 ± 4.2 years among patients with cirrhosis at the time of the inclusion.Diagnosis of cirrhosis was based on clinical, biochemical, imaging [20], and, whenavailable, histological data. Blood samples and clinical data, including severity ofcirrhosis graded according to the Child–Pugh classification and the model for end-stage liver disease (MELD) score [21], presence and grade of ascites and enceph-alopathy were captured at inclusion. Clinical data, including age at onset, etiol-ogy, presence of esophageal varices, previous episodes of variceal bleeding andhepatic encephalopathy, prior spontaneous bacterial peritonitis (SBP) events,co-morbidities (myocardial infarction, congestive heart failure, peripheral arterialdisease, cerebrovascular disease, chronic pulmonary disease, chronic renal failure,gastrointestinal ulcer disease, diabetes mellitus, and non-metastatic and meta-static cancer, including hepatocellular carcinoma, vasculitis) and current medica-tion were collected by in-depth review of the patients’ medical charts. During thestudy period, indications for non-selective beta-blockers and non-absorbableantibiotics either in primary or secondary prophylaxis of variceal bleeding andSBP were considered in the recommendations of current guidelines [22,23]. Indi-cations for proton pump inhibitors (PPIs) were gastroesophageal reflux disease,erosive gastritis, peptic ulcer disease or treatment for Helicobacter pylori infection[24]. Etiology of cirrhosis was alcoholic in 246 (63.9%), and non-alcoholic in 139(36.1%) cases. This latter group comprised 115 patients with hepatitis C virus(HCV)-related cirrhosis and 24 with other causes. Clinical data of patients is sum-marized in Table 1.

345 of the 385 patients were available to be enrolled into a prospective fol-low-up study, where we registered adverse outcomes including death or develop-ment of clinically significant bacterial infections (CSI). Data were collected duringregular and extraordinary outpatient follow-up visits and inpatient stays. In Hun-gary, a follow-up visit is usually scheduled every 3 months at a specialized gastro-enterology center (a follow-up between 1 and 3 months may be scheduled ifdictated by disease severity or presence of disease specific complications). Thefollow-up period lasted 24 months or death/loss of follow-up (median follow-up: 729 days [range: 5–730]). An infectious episode was defined clinically signif-icant if it warranted hospitalization, and patients were admitted due to deterio-ration of general conditions or liver function, or other infection-relatedcomplications, such as variceal bleeding, hepatic encephalopathy, diuretic-resis-tant ascites formation and renal failure. Infectious episodes were identified byreviewing medical records, including clinical symptoms, laboratory data andimaging findings, and use and efficacy of antibiotic therapy by two independentgastroenterologists (M.P. and Zs.V.). The following laboratory data were consid-ered: elevation of the white blood cell count (absolute: >10.8 � 109/L or relative[in patients with leukopenia]: double of count at former visits) with an elevatedneutrophil rate (>76%) and elevated serum levels of high-sensitivity C reactiveprotein (CRP) (>10.0 mg/L) and/or procalcitonin (PCT) (>0.15 lg/L) [25], includingmicrobiological culture results, where available. Autopsy records (n = 77) werealso assessed in cases of death. The following bacterial infections were consideredbased on conventional criteria: infections of skin and soft tissue, orocavitalregion, upper and lower respiratory tract (acute bronchitis, pneumonia), biliarytract (cholecystitis, cholangitis, liver abscess), intestinal tract (gastroenteritis),urinary tract (uncomplicated cystitis were excluded), osteomyelitis, and endocar-ditis. Spontaneous bacterial peritonitis was diagnosed if ascitic fluid polymorpho-nuclear cell (PMN) count was greater than 250/mm3, with or without positiveculture, in the absence of an intra-abdominal source of infection. Bacteriaemiawas considered when clinical symptoms and signs of infection were presentand confirmed by microbiological demonstration of the causative organism fromblood culture in the absence of site-specific infection.

Serum samples were also obtained from patients with chronic hepatitis C(chronic HCV, n = 119, male/female: 50/69, age: 54.6 ± 11.7 years) and primarybiliary cirrhosis (PBC, n = 102, male/female: 4/98, age: 58.8 ± 11.8 years) withoutcirrhosis as disease controls. METAVIR scoring system was used to rank liverfibrosis and necroinflammatory activity [26–28]. Patients with fibrosis stage 4(F4) were considered having cirrhosis and excluded from the disease controlgroup. The diagnosis of PBC was based on biochemical evidence of cholestasis,

3 vol. 59 j 457–466

Table 1. Clinical characteristics of patients with cirrhosis.

TOTAL(N = 385)

Patients with alcoholic cirrhosis(n = 246)

Patients with non-alcoholic cirrhosis(n = 139)

Sex (male/female) 206/179 144/102 62/77Age (mean ± SD, yr) 56.6 ± 11.0 57.6 ± 9.7 54.8 ± 12.8Child-Pugh stage, n (%)*

A 144 (37.4) 58 (23.6) 86 (61.9)B 136 (35.3) 101 (41.1) 35 (25.2)C 105 (27.3) 87 (35.4) 18 (12.9)

MELD score (points)§ 14.3 ± 6.0 15.1 ± 6.4 12.8 ± 4.8

Serum bilirubin (μmol/L)* 67.4 ± 96.3 44.0 ± 70.4 79.0 ± 105.1Serum albumin (g/L)* 33.0 ± 8.0 35.6 ± 7.5 31.9 ± 7.9Co-morbidities present, n (%) 177 (46.0) 114 (46.3) 63 (45.3)HCC, n (%)* 34 (8.8) 11 (4.5) 23 (16.5)

SD, standard deviation; HCC, hepatocellular carcinoma.⁄p <0.001 and §p <0.01 between patients with alcoholic vs. non-alcoholic cirrhosis.

JOURNAL OF HEPATOLOGY

serum anti-mitochondrial antibodies (AMA) and/or PBC-specific AMA-M2 positiv-ity, compatible histology, and the exclusion of extrahepatic cholestasis [29]. Thediagnosis of chronic HCV was based on positive HCV ribonucleic acid, elevatedliver function tests, and compatible liver biopsy.

The healthy control group consisted of 100 age- and gender-matched individ-uals (male/female: 45/55, age: 50.5 ± 16.7 years) selected from consecutive blooddonors in Debrecen. The control subjects did not have any known gastrointestinalor liver diseases.

Methods

ANCA indirect immunofluorescence assayCollected sera were frozen at �80 �C until testing. Detection of ANCA was per-formed by a semiquantitative indirect immunofluorescence (IIF) technique usingboth ethanol- and formalin-fixed human peripheral blood neutrophil substrates(EUROIMMUN Medizinische Labordiagnostika AG, Lübeck, Germany). Specimenswere incubated at a 1:10 dilution in phosphate-buffered saline and the assayswere performed according to the manufacturers’ instructions. The presence ofANCAs was detected with fluorescein-labeled goat antihuman IgA and IgG anti-body (EUROIMMUN Medizinische Labordiagnostika AG), separately. Examinationand classification were performed under ultraviolet light using EUROStar II Plusmicroscope (EUROIMMUN Medizinische Labordiagnostika AG) at a magnificationof 400�. Interpretation of the immunofluorescence results was based on thebehavior of the specimens on ethanol- and formalin-fixed slides and includedthe following patterns [7,30]: C-ANCA, typical P-ANCA, and atypical P-ANCA.Serum end point titers of ANCA equal to or greater than 1:10 are considered posi-tive. To quantify the ANCA titer, dilutions (dilution factor 3.2 [square root of 10])of each specimen were made starting at 1:10.

Characterizations of IgA type ANCAs in cirrhosisWe used the IIF ANCA testing system (ethanol-fixed granulocytes) describedabove, but applied specific monoclonal mouse anti-human antibodies (in 1:30dilution) against each IgA subtype in parallel (Acris Antibodies Gmbh, Herford,Germany; anti human IgA1 – clone: NI 69-11, anti human IgA2 – clone: NI 512,anti human secretory (s) IgA – clone: NI 194-4). A secondary, FITC-labeled, poly-clonal goat anti-mouse antibody (1:10 dilution; DAKO, Glostrup, Denmark) wasalso used to augment the fluorescence signal. To maximize the comparison poten-tial of different subtyping assay results, we applied an image analysis based quan-tification method using a single dilution of the samples (1:10).

Samples were analyzed by a MicroOptix MX 300 TF microscope (West MedicaAustria, Perchtoldsdorf, Austria) using 400� magnification. Fluorescence imageswere recorded in 32 bit ‘‘.bmp’’ format using a MicroOptix/Vision CAM V330 cam-era and the SB Video&Audio AV Grabber (West Medica). A green background slide(Chroma Technology Corp., Rockingham, VT, USA) was also captured at everyexperimental day and used to correct for illumination strength and inhomogene-ity. The images were processed using the ‘‘ImageJ 1.46r’’ Software [31]. Fluores-cence intensities in the green channel – expressed in arbitrary units (A.U.) – of20 individual granulocytes per field of view were determined after segmenting

Journal of Hepatology 201

the thresholded image, and mean fluorescence intensity/cell ± SD values werecalculated. For individual sera, three mean A.U./cell values were reported accord-ing to three IgA types (IgA1, IgA2, and sIgA) and used for further calculations.

In order to standardize the fluorescence signal measured with the anti-IgA1and anti-IgA2 antibodies, Quantum™ Simply Cellular� (QSC) microspheres (BangsLaboratories, Inc., Fishers, IN, USA) were used. Based on these experiments, weconcluded that an equal fluorescence intensity of IgA1-type and IgA2-type ANCAin our system shows an equal number of antibodies bound, and therefore, thefluorescence intensity values we obtained can be used directly for calculatingthe IgA2/IgA1 ANCA rate of our samples. Detailed description of these experi-ments such as analytical properties and the comparison of the quantitative andsemiquantitative IIF methods are presented in the Supplementary Material.

The IgA subtype analysis was performed in the case of 142 ANCA IgA-positive,20 IgA-negative cirrhotic patients, and 20 healthy controls. The fluorescenceintensity values for IgA1, IgA2, and sIgA did not differ significantly betweenhealthy controls and disease controls (healthy controls: IgA1 = 12.8 ± 3.3,IgA2 = 14.1 ± 4.5, sIgA = 13.0 ± 2.7; disease controls: IgA1 = 16.0 ± 4.9, p = 0.08,IgA2 = 14.6 ± 3.4, p = 0.98, sIgA = 14.9 ± 3.1, p = 0.06), therefore, we merged thesetwo populations in a unified control group (n = 40). IgA2/IgA1 rate was calculatedby dividing the fluorescence intensity of ANCA IgA1 or ANCA IgA2 by the sum ofthe fluorescence intensities of ANCA IgA1 and IgA2 in each serum sample. Todefine the individual positivity of each sample for the presence of secretory com-ponent (SC), a cut-off (23.4) was defined based on the mean + 3 SD fluorescenceintensity values of the control group.

Determination of antigen specificity of ANCA IgAThe presence of anti-MPO IgA and anti-PR-3 IgA antibodies in sera previouslypositive for ANCA IgA in the IIF test (n = 162) was determined by enzyme-linkedimmunosorbent assays (ELISA) (QUANTA Lite™ MPO and QUANTA Lite™ PR-3INOVA Diagnostics, San Diego, CA) according to the manufacturers’ instructions.The results are presented as OD. Values above the OD cut-off 0.159 and 0.140were considered positive for anti-MPO and anti-PR-3, respectively. In our labora-tory, these cut-off OD values represented the mean + 3 SD values of the healthycontrols (n = 92). The results were documented in absolute OD values and in fre-quency of positivity.

Anti-microbial antibody assaysASCAs are antibodies directed primarily against a 200 kDa phosphopeptidoman-nan cell wall component of the common baker’s or brewer’s yeast Saccharomycescerevisiae [32], while anti-OMP Plus™ antibodies are against multiple bacterialproteins derived from two species of intestinal bacteria (one Gram-positive andone Gram-negative). None of them are from the phylum proteobacteria, of whichEscherichia coli is a member [14].

The presence of anti-Saccharomyces cerevisiae antibodies (ASCA) IgA, ASCAIgG, and anti-OMP Plus™ IgA antibodies in serum was determined by ELISA(QUANTA Lite™, INOVA Diagnostics, San Diego, CA) according to the manufactur-ers’ instructions. The results are presented as arbitrary units, and values above thecut-off of 25 units were considered as positive. The results were documented in

3 vol. 59 j 457–466 459

Table 2. Prevalence of ANCA in patients with chronic liver diseases and healthy controls.

N Total ANCA positive*

ANCA IgA# ANCA IgG§ ANCA IgAonly

ANCA IgG only

ANCA IgA and IgG

Cirrhosis with various etiology 385 281 (73.0%) 201 (52.2%) 140 (36.4%) 141 (36.6%) 80 (20.8%) 60 (15.6%)Chronic HCV 119 69 (58.0%) 13 (10.9%) 62 (52.1%) 7 (5.9%) 56 (47.1%) 6 (5.0%)PBC 102 39 (38.2%) 28 (27.4%) 15 (14.7%) 24 (23.5%) 11 (10.8%) 4 (3.9%)Healthy controls 100 4 (4.0%) 0 (0.0%) 4 (4.0%) 0 (0.0%) 4 (4.0%) 0 (0.0%)

HCV, viral hepatitis C, PBC, primary biliary cirrhosis.⁄p <0.01 between chronic HCV and cirrhosis or PBC; p <0.001 between any other two groups for total ANCA.#p <0.01 between chronic HCV and PBC; p <0.001 between any other two groups for ANCA IgA.§p = 0.014 between PBC and HC; p <0.01 between chronic HCV and cirrhosis; p <0.001 between any other two groups for ANCA IgG by using v2-test or v2-test with Yatescorrection if appropriate.

Research Article

absolute values and in frequency of positivity. ASCA and anti-OMP Plus™antibody evaluation was performed in our previous study [14]. Of the 385patients involved in the present study, 250 took part in the previous one.

All the serologic assays were performed in a blinded fashion without priorknowledge of the patients’ diagnosis or other clinical information. Semiquantita-tive ANCA evaluation and anti-microbial serologic examination were completedin the laboratory of EUROIMMUN Medizinische Labordiagnostika AG by BT andNW and in the laboratory of INOVA Diagnostics by ZS, respectively. ANCA IgAcharacterization assays were performed by NS, MP, and PASz in the Departmentof Laboratory Medicine, University of Debrecen.

Ethical considerations

The study protocol was approved by the Ethical and Science Committee of theUniversity of Debrecen. Each patient was informed of the nature of the studyand signed an informed consent form.

Statistical methods

Variables were tested for normality using Shapiro Wilk’s W test. Continuous vari-ables were summarized as means [standard deviation (SD)] or medians [inter-quartile range (IQR)], according to their homogeneity. To evaluate differencesbetween cirrhosis and disease or healthy control groups, as well as within sub-groups of patients with cirrhosis, the following statistical methods were used.Categorical variables were compared with the Fisher’s exact test or v2-test withYates correction, as appropriate. Continuous variables were compared using theStudent’s t test, one-way analysis of variance (ANOVA), or Mann-Whitney’s U test.Spearman SRO tests were used to analyze association between continuous vari-ables. Kaplan-Meier survival curves were plotted to analyze the associationbetween categorical clinical variables or ANCAs and CSI during follow-up withLogRank and Breslow tests. Cox-regression analysis was used to assess the asso-ciation between categorical clinical variables or ANCAs and time to CSI. Variableswith a p <0.1 in univariate tests were selected for the multivariate testing. Asso-ciations are given as Spearman’s r values, odds ratio (OR), and hazard ratio (HR)with a 95% confidence intervals (CI). A 2-sided probability value <0.05 was con-sidered to be significant. For statistical analysis, GraphPadPrism 7 (San Diego,CA) and SPSS 15.0 (SPSS Inc, Chicago, IL) programs were used.

Results

Anti-neutrophil cytoplasmic antibodies (ANCA) in patients withchronic liver diseases

The prevalence of ANCA IgA and IgG in patients with variouschronic liver diseases (CLD) is shown in Table 2. The rates of totalANCA seropositivity, including ANCA IgA and/or IgG positivity,were significantly higher in all of the CLDs as compared tohealthy controls. Among CLDs, the frequency of total ANCAseropositivity was significantly higher in patients with cirrhosiscompared to those without cirrhosis (ORANCATotal 1.96, 95% CI:1.28–3.00, p <0.01 for chronic HCV and ORANCATotal: 4.37, 95%CI: 2.76–6.90, p <0.001 for PBC). IgA class ANCA was detectedsignificantly more frequently in patients with cirrhosis (52.2%)

460 Journal of Hepatology 201

compared to patients with chronic HCV (10.9%) or PBC (27.4%,p <0.001, for both). However, IgG class ANCA occurred more oftenin patients with chronic HCV (52.1%) than cirrhosis (36.4%,p <0.01) or PBC (14.7%, p <0.001). None of the patients withHCV had ANCA associated vasculitis.

Fluorescence staining patterns of ANCA isotypes differedamong CLD groups as well. In patients with cirrhosis, the C-ANCAstaining emerged as the predominant fluorescence pattern of IgAclass ANCA (46.3%) and atypical P-ANCA of IgG class ANCA(51.1%). The other two staining patterns were equally distributed(atypical P-ANCA: 22.4% and P-ANCA: 31.3% for IgA class ANCA;P-ANCA: 27.3% and C-ANCA: 21.6% for IgG class ANCA). In con-trast, the majority of ANCA patterns – for both IgA and IgG classANCA – were atypical P-ANCA in patients with chronic HCV(84.6% and 82.5%) and PBC (82.5% and 86.7%). 15.6%, 5.0%, and3.9% of ANCA positive sera from patients with cirrhosis, chronicHCV, and PBC simultaneously contained both IgA and IgG classANCA, respectively. The concordance between ANCA patterns ofdifferent Ig classes varied in double positive sera. Complete con-cordance was found in chronic HCV (100.0%) but in PBC and incirrhosis only 50.0% and 35.0% of sera showed the same ANCApattern using anti-IgA and IgG secondary antibodies, respectively.

Characterizations of IgA type ANCAs in cirrhosis

In order to characterize the ANCA IgA antibodies in cirrhosis, ANCAIgA positive samples in the semi quantitative IIF assay were evalu-ated further. Proportion of IgA subclasses and presence of SC of eachtested samples were measured simultaneously. The proportion ofANCA IgA2 subtype was markedly elevated: 46.2 ± 8.5% and wasonly slightly lower than ANCA IgA1 subtype (53.8 ± 8.5%). The distri-bution of IgA2 proportion was similar among the three differentANCA patterns (C-ANCA: 45.8% ± 8.1%, P-ANCA: 48.0 ± 6.8% andatypical P-ANCA: 44.9 ± 8.6%). The presence of SC was high, 86.8%of ANCA IgA positive samples showed positivity for this component.

We also evaluated the antigen specificity of ANCA IgA positivesera for MPO and PR-3. The frequency of anti-PR3 and anti-MPO pos-itivity was low. 4.9% of P-ANCA positive and 16.5% of C-ANCA positivesamples showed anti-MPO and anti-PR3 positivity, respectively.

Association between ANCA and severity of disease, presence ofdisease specific complications or anti-microbial serology response inpatients with cirrhosis.

In patients with cirrhosis, ANCA IgA positivity (cases with IgA+/IgG� or IgA+/IgG+) increased gradually according to disease sever-ity as rated by the Child-Pugh stage (Table 3A). Presence of ANCAIgA (cases with IgA+/IgG� or IgA+/IgG+) was also positively

3 vol. 59 j 457–466

Table 3. Prevalence of ANCA positivity in patients with cirrhosis. (A) Impact of disease severity according to Child-Pugh stages. (B) Impact of ascites.

TotalNon-alcoholicsAlcoholics

ANCA IgA and/or IgGNon-alcoholicsAlcoholics ANCA IgANon-alcoholicsAlcoholicsANCA IgG onlyNon-alcoholicsAlcoholics

Child-Pugh stagesA B C1448658

13635101

1051887

100 (69.4%)60 (69.8%)40 (68.9%)

100 (73.5%)24 (68.6%)76 (75.2%)

81 (77.1%)10 (55.6%)71 (81.6%)

55 (38.2%)24 (27.9%)§

31 (53.4%)

78 (57.4%)14 (40.0%)‡

64 (63.4%)

68 (64.8%)*8 (44.4%)¶

60 (69.0%)45 (31.2%)36 (41.9%)¢

9 (15.5%)

22 (16.2%)10 (28.6%)12 (11.9%)

13 (12.4%)*2 (11.1%)11 (12.6%)

A B Patients without ascites

Patients with ascites

20493111

18146135

141 (69.1%)64 (68.8%)77 (69.4%)

140 (77.3%)30 (65.2%)110 (81.5%)

86 (42.2%)24 (25.8%)x

62 (55.9%)

115 (63.5%)*22 (47.8%)§

93 (68.9%)55 (27.0%)40 (43.0%)¶

15 (13.5%)

25 (13.8%)**8 (17.4%)*17 (12.6%)

(A) ANCA IgA = cases with IgA+/IgG� or IgA+/IgG+.ANCA IgG only = cases with IgA�/IgG+.⁄p <0.001 and p = 0.031 between three different Child groups.§p <0.01, �p = 0.016, –p = 0.047 and ¢p <0.001 between non-alcoholics vs. alcoholics, by using Fisher’s exact test or v2-test with Yates correction.(B) ANCA IgA, cases with IgA+/IgG� or IgA+/IgG+.ANCA IgG only, cases with IgA�/IgG+.⁄p <0.001 and ⁄⁄p <0.01 between patients with and without ascites.xp = 0.035 and §p <0.01 and –p <0.001 between non-alcoholics vs. alcoholics ANCA IgA in non-alcoholics vs. alcoholics by using Fisher’s exact test or v2-test with Yatescorrection.

JOURNAL OF HEPATOLOGY

(ORANCAIgA: 3.0, 95% CI: 1.90–4.71) associated with the presence ofascites (Table 3B). Similarly to seropositivity rates, the more severethe disease, the higher the titers of the ANCA IgA (Fig. 1). ANCA IgAtiters were also significantly higher in patients with ascites as com-pared to those ones without (Fig. 1). These associations were alsoverified if we re-analyzed them for alcoholic and non-alcoholicsubgroups separately (Table 3). Some etiological differences inthe ANCA IgA response, however, were revealed. More enhancedANCA IgA formation was found in alcoholic patients as comparedto non-alcoholic ones. ANCA IgA positivity rates were significantlyhigher (by 20%) in alcoholic patients in each severity group as com-pared to rates of the corresponding severity group in non-alcoholicpatients. Furthermore, alcoholic patients with the less severe dis-ease (Child A or no ascites) had already showed marked ANCAIgA positivity rate (53.4% or 55.9%) (Table 3).

On the contrary, ANCA IgG response behaved oppositely toANCA IgA in non-alcoholic patients. Presence of ANCA IgG only(cases with IgA�/IgG+) was decreased gradually according to dis-ease severity as rated by Child-Pugh score (Table 3A) and nega-tively associated with the presence of ascites (ORANCAIgGonly:0.28, 95% CI: 0.12–0.66) (Table 3B). In alcoholic patients, how-ever, the positivity rate for ANCA IgG only (cases with IgA�/IgG+) was low independently of disease severity or presence ofascites. Alcoholic patients with the least severe disease (Child Aor no ascites) have already shown markedly decreased ANCAIgG only positivity rates (15.5% or 13.5%) and these were signifi-cantly lower than the rates of patients with non-alcoholic diseasewith the same severity (41.9% or 43.0%, p <0.001 for both). Thesedifferences were not present in the Child C group (Table 3).

ASCA IgA and anti-OMPPlus IgA antibodies were more preva-lent in patients with ANCA IgA positivity as compared to thosewithout (59.0% vs. 23.4%, p <0.001 and 75.5% vs. 50.5%,p <0.001). A similar association was found between the presenceof ASCA IgG and ANCA IgG antibodies. The presence of ASCA IgG

Journal of Hepatology 201

was 23.6% in ANCA IgG positive and 11.5% in ANCA IgG negativepatient group (p <0.01).

Bacterial infections: general characteristics

A total of 187 clinically significant infectious episodes wereidentified in the 345 patients with cirrhosis during the 2-year fol-low-up. 110 patients (31.9%) developed some type of clinicallysignificant infections of which 39.4% had more than one episode.The distribution of different severe infections was as following:27.5% spontaneous bacterial peritonitis, 17.5% pneumonia,12.5% urinary tract infection, 8.9% skin and soft tissue infections,and 16.5% miscellaneous. The origin of the infection could not beidentified in 17.1% of the cases. Bacteria were Gram-negative in60.9% and Gram-positive in 39.1% of positive cases. The propor-tions of different types of infection regarding either their locationor Gram specificity were similar among patients with or withoutANCA (data not shown).

In the study population, the clinical factors known to affectrisk of infections were also evaluated. Eighty-four (24.3%)patients had advanced disease (Child C) and 158 (45.8%) hadascites. Among patients with ascites, 36 (22.8%) had a priorSBP episode of whom 14 (38.9%) had more than one episode.Patients with history of prior episodes of variceal bleeding orhepatic encephalopathy were 76 (22.0%) and 57 (16.5%), respec-tively, at the time of entry into the follow-up study. Onehundred and eighty-three patients (53.0%) got non-selectivebeta-blockers, which rate was significantly higher in ANCA IgApositive patients as compared to ANCA IgA negative ones(60.0% vs. 45.6%, p <0.01). This corresponds to the finding thatANCA IgA positive patients had more advanced disease. Use ofproton pump inhibitors represented 38.2% of cases and was dis-tributed equally among ANCA IgA positive and negative patients(37.4% vs. 39.1%).

3 vol. 59 j 457–466 461

Child-Pugh

A B C

100

80

60

40

20

0

ANC

A Ig

A le

vels

(%)

01:10≥1:32

01:10≥1:32

A B

31.3

6.9

61.8

49.3

7.4

43.4

60.0

5.7

34.3

Ascites

No Yes

100

80

60

40

20

0AN

CA

IgA

leve

ls (%

)

37.3

4.9

57.8

54.7

8.8

36.5

Fig. 1. ANCA IgA levels in patients with cirrhosis with various levels ofseverity. Depicted by (A) Child-Pugh stages or (B) presence of ascites. p <0.001between all Child-Pugh groups; p <0.001 between patients with or withoutascites by v2-test.

Research Article

Clinical and laboratory predictors of clinically significant bacterialinfection

Among the clinical factors, the disease severity according to theChild–Pugh stage (p <0.001), presence of ascites (OR: 3.02; 95%CI: 1.89–4.84, p <0.001), history of hepatic encephalopathy (OR:2.15; 95% CI: 1.20–3.86, p <0.01) and co-morbidities (OR: 2.03;95% CI: 1.28–3.22, p <0.01) were identified as risk factors forthe development of CSI in univariate analysis (v2-test or v2-testwith Yates correction). History of a prior SBP episode was a riskfactor for the development of further SBP episodes (OR: 2.87;95% CI: 1.20–6.89, p <0.01).

Patients with ANCA isotype IgA presented with infectious epi-sodes significantly more frequently compared to patients withoutANCA isotype IgA (38.9% vs. 24.1%, p <0.01 at the positivity cut-offtiter 1:10 and 41.8% vs. 23.4%, p <0.001 at titer 1:32). The infec-tion rate was associated with the magnitude of the ANCA IgAserologic response (negative: 24.1%, titer of 1:10: 18.2%, 1:32:42.9% and P1:100: 54.7%, p <0.001). The highest infection ratewas observed in ANCA IgA positive patients with a titer of1:100 or higher (P100). In this subgroup, the occurrence of CSIwas 54.7%. 51.1% of the ANCA IgA positive patients (90/176)belonged to this subgroup. It represented 26.1% (90/345) of thewhole cohort, including ANCA IgA negative patients as well. Sim-ilarly, ANCA IgA patterns were associated with the developmentof CSI. Among the different ANCA IgA types, the presence of C-ANCA pattern was associated with the highest risk for CSI (OR:2.71; 95% CI: 1.54–4.75, p <0.001). Of note, 80.2% of sera with aC-ANCA pattern showed an ANCA IgA titer of P100 comparedto those with non-C-ANCA pattern (25.8%, p <0.001).

Clinical and laboratory parameters associated with time to firstclinically significant bacterial infection

In a Kaplan-Meier analysis, Child-Pugh stages, presence of asci-tes, co-morbidities (Fig. 2), history of hepatic encephalopathy(p <0.01), and variceal bleeding p = 0.046) were associated withtime to first CSI. In addition, history of prior SBP episode wasassociated with time to further SBP episode (p <0.01).

A shorter time to first infection was found for patients withANCA IgA, compared with those without infection, during the

462 Journal of Hepatology 201

2-year follow-up period (HR: 1.83 95% CI: 1.26–2.66, p <0.01 atpositivity cut-off titer 1:10 and HR: 2.15 95% CI: 1.46–3.15,p <0.001 at the positivity cut-off titer 1:32) (Fig. 2). Of the differ-ent ANCA IgA types, the presence of C-ANCA pattern was associ-ated with the highest risk for CSI (HR: 3.04 95% CI: 1.83–5.04,p <0.001).

A Cox-regression model was also used to investigate the influ-ence of ANCA IgA positivity on the development of CSI. Afteradjusting for gender, co-morbidities, and disease severity accord-ing to Child–Pugh stage, ANCA IgA positivity at the cut-off titer1:32 was an independent variable associated with shorter timeto first infection (p = 0.006) (Table 4). Our choice of using Child-Pugh stage to adjust the disease severity in our multivariatemodel was based on the fact that it is the best-known variablebearing significant impact on the development of bacterial infec-tion in cirrhosis and encompasses parameters indicating paren-chymal insufficiency and portal hypertension in the mostcomplex way. To avoid redundancy in our multivariable model,individual components or related factors to Child-Pugh scorewere not involved in parallel even if they were significantly asso-ciated with the development of CSI in univariate analysis.

Survival analysis

In total, 77 patients (22.3%) died during the 2-year follow-up.Kaplan-Meier survival analysis demonstrated a significantlyworse survival in patients with advanced disease according toChild-Pugh stage (p <0.001), ascites (p <0.001) or co-morbidity(p <0.01). The presence of CSI (p <0.001), but not the ANCA IgApositivity (p = 0.117), was associated with a significantly highermortality rate.

Discussion

To our knowledge, this is to date the largest study to investigatethe prevalence, type, and clinical significance of ANCA in patientswith cirrhosis of different etiology. We used IIF technique andboth ethanol- and formalin-fixed human neutrophil substratesfor the detection of ANCA. Contrary to routine laboratory practice,ANCA was identified by anti-IgA secondary antibody in addition toanti-IgG one. Thereafter, IgA type ANCAs were subtyped as well.Prevalence and characteristics of IgG class ANCA in CLD werestudied extensively; in contrast, data on ANCA IgA are limited.

In the present study, we demonstrated – for the first time –that enhanced ANCA IgA formation is a feature of cirrhosisregardless of its etiology and is associated with disease severityas well. The presence of ANCA IgA was significantly higher in cir-rhosis (in alcoholics and non-alcoholics) compared to eitherchronic HCV or healthy controls and increased with the progres-sion of the disease. Of the CLDs, the highest IgA ANCA prevalencewas found in PBC, however, this elevation was significantly lowercompared to the cirrhotic group. Marked elevation of serum IgAconcentration in alcoholic cirrhosis has been long known [33].Correspondingly, within the cirrhotic group, higher ANCA IgApositivity rate and enhanced titers were found in patients withalcoholic compared to those with non-alcoholic disease.

The cause of enhanced serum IgA formation in cirrhosis hasnot been fully understood yet. The involvement of the intestinaltract, however, is very probable [33]. Disruption of the gut-bar-rier integrity at both the mechanical and immunological level isa well-known feature of cirrhosis and becomes more pronounced

3 vol. 59 j 457–466

Child AChild BChild C

0 150 300 450 600 750Days

0 150 300 450 600 750Days

0 150 300 450 600 750Days

0 150 300 450 600 750Days

0

20

40

60

80

100

Infe

ctio

n-fre

e su

rviv

al (%

)

0

20

40

60

80

100

Infe

ctio

n-fre

e su

rviv

al (%

)

0

20

40

60

80

100

Infe

ctio

n-fre

e su

rviv

al (%

)

0

20

40

60

80

100

Infe

ctio

n-f re

e su

rviv

al (%

)

A B C D

Ascites: noAscites: yes

Co-morbidity: noCo-morbidity: yes

ANCA IgA negativeANCA IgA positive

Fig. 2. Association between Child-Pugh stage, ascites, co-morbidities, ANCA isotype IgA, and development of clinically significant bacterial infections in patientswith cirrhosis. Infection-free survival refers to the percentage of patients in the cohort without infection at a given time during the follow-up. (A) Patients with Child Cstage cirrhosis were at higher risk of developing infections compared to patients with Child A or B disease. pBreslow: <0.001, pLogRank: <0.001. (B) Patients with asciteswere at higher risk of developing infections compared to patients without ascites. pBreslow: <0.001, pLogRank: <0.001. (C) Patients with co-morbidity were at higher risk ofdeveloping infections compared to patients without co-morbidity. pBreslow: <0.01, pLogRank: 0.034. (D) Patients with ANCA isotype IgA (P1:32 titer) were at higher risk ofdeveloping infections compared to those without ANCA isotype IgA. pBreslow: <0.001, pLogRank: <0.001.

Table 4. Summary of Cox model: factors affecting time to first severe bacterialinfection in patients with cirrhosis.

p value Hazard ratio

95% CI

Sex 0.299 1.22 0.83-1.77Co-morbidities 0.008 1.68 1.14-2.47Child-Pugh stage <0.001

A ReferenceB 0.016 1.81 1.12-2.95C <0.001 3.42 2.07-5.64

Seropositivity to ANCA IgA ≥1:32 titer

0.006 1.74 1.18-2.56

p value: level of significance; 95% CI: 95% confidence interval.

JOURNAL OF HEPATOLOGY

with disease progression [18]. Two-third of patients with cirrho-sis who underwent capsule endoscopy showed mucosal inflam-matory like abnormalities and corresponding histologicalalterations in the small bowel [34,35]. Alcohol has further directtoxic effects on intestinal epithelial cells [33]. This disturbedintegrity of the gut barrier with the small bowel bacterial over-growth [36] may in turn enable a sustained local invasion of bac-terial constituents from the gut lumen (BT), which thenstimulates the secretory immune system and is also involved inthe pathogenic processes of disease specific complications in cir-rhosis [18]. Supporting gut involvement in IgA production in cir-rhosis, increased concentration of different serum antibodies togut bacterial proteins [37–40], or host proteins having cross-reac-tive epitopes with bacterial constituents [41,33], and high levelsof serum sIgA, were reported in early publications [42]. In cirrho-sis, the IgA class anti-microbial antibodies were also more preva-lent in our previous study [14].

IgA has long been accepted as an important factor in mucosalimmunity, supported by basic research data, too. Based on animalexperiments, the development of intestinal IgA and bacterial colo-nization showed strong interactions: (1) germ free mice after oraladministration of commensal bacteria produced sIgA, that – afterrepeated doses – reduced penetration of bacteria to mesenteriallymph nodes; (2) in immundeficient mice, the induction of innateimmune genes of the gut by commensal bacteria was diminished

Journal of Hepatology 201

after injection of IgA secreting hybridoma cells; (3) in polymericIg receptor deficient animals, the lack of transport of IgA into theintestinal lumen resulted in a low-grade intestinal inflammation;and (4) in activation-induced cytidinedeaminase (AID) deficientmice, where proper IgA production is absent, bacterial overgrowthis present in the small intestine. These experiments could alsoshow that large doses of bacteria are required for stimulation ofmucosal IgA, the extent of which is determined by the total doseof bacteria; and the specific IgA response lasted quite long unlessintroduction of other bacteria into the gut [43–46].

There might be several modes of action via IgA interactionwith the commensal bacteria. IgA molecules can cover thesemicroorganisms and inhibit their penetration through the gutepithelial cells. Secretory IgA can facilitate the uptake and pre-sentation of certain bacterial antigens as well and – by inducingaltered signaling, e.g., via CD89 – drive the local adaptive immunesystem into a more tolerogenic state. In this way, sIgAs can par-ticipate in the regulation of bacterial communities in the gutlumen. On the other hand, the presence of gut microbes has animportant role in the regulation of local IgA production. The IgAsecretion of B cells in the germinal centre of Peyer’s patches, inisolated lymphoid follicules and also in the case of sparse lym-phoid elements in the lamina propria of the gut, clearly requiresthe presence of microbial antigens. Dendritic cells constantlysample intestinal bacteria either by the help of M cells coveringthe Peyer’s patches or by extending processes between the junc-tions of epithelial cells, and create a special cytokine microenvi-ronment with the help of special T cells that stimulate IgA classswitch recombination. The T-independent activation is mediatedby B-cell activating factor (BAFF), a proliferation-inducing ligand(APRIL), transforming growth factor (TGF) b, interleukin (IL) 6,and retinoic acid expression/secretion of dendritic, stromal andepithelial cells accompanied by cross-linking of target B cellreceptors by dendritic cells [43–49]. Furthermore, conservedmolecular structures of bacteria like CpG oligonucleotides canactivate B cells’ IgA production directly via Toll-like receptor(TLR) 9 [50]. Activated and IgA-switched B cells migrate to thedraining mesenterial lymphnodes, via the thoracic duct can reachthe blood and then – guided by special homing receptors – returnback to the mucosal surfaces. These IgA-producing B and plasmacells show a more ‘‘monocytic’’ phenotype as they express tumor

3 vol. 59 j 457–466 463

Research Article

necrosis factor (TNF) a, inducible nitric oxide synthase (iNOS),and a few myeloid surface markers, furthermore they possessnon-conventional myeloid/dendritic cell functions [51]. As acti-vated mucosal B cells travel through the systemic circulation,an enhanced local IgA production stimulated by enhanced BTcan be noticed in peripheral blood, too.

Since the composition and extent of bacterial load in gut have avery clear effect on IgA production and enhanced BT is a specialfeature of cirrhosis, we hypothesized that bacterial antigensderived from the mucosal compartment and their cross-reactivitywith granulocyte cytosolic or granular proteins might play a cen-tral role in the enhanced IgA class ANCA formation in cirrhosis. Weperformed a detailed characterization of IgA type ANCAs. Deter-mination of the ratio of IgA1 and IgA2 subtypes and detection ofthe SC on IgA molecules in sera can help identify the location oftheir formation (bone marrow or mucosal compartment). In pre-vious publications, an increase in the proportion of IgA2 subtypeand the presence of the SC are concurrently considered as a con-firmatory evidence for the mucosal origin of IgA secretion[52,53]. The proportion of IgA2 is about 10% of total IgA in humansera, while IgA1 is 90%, and they largely exist in monomeric (m)form. Sharing of SC containing IgA antibodies from the total IgApool is no more than 1%, since SC is attached to di- or polymericIgA (pIgA) via its transport through the epithelial cells into thegut lumen or to other mucosal surfaces [54]. The proportion ofANCA IgA2 subtype was markedly elevated (46%) and the pres-ence of SC was high as well (87%) in ANCA IgA positive samplesof our patients with cirrhosis irrespective of their IIF ANCA pat-terns. We could not address directly the structural form of ANCAIgA, but it is very probable that these antibodies are in their di- orpolymeric form in the majority of ANCA IgA due to the presence ofSC. ANCA IgA antibodies were not assessed directly in the differentmucosal compartments or in the organs to which bacteria aretranslocated (e.g., liver or ascites), which is a limitation of thepresent study, though our serological findings in the ANCA sub-typing assays highly support our hypothesis about the ANCA IgAformation. Similarly, gut mucosal compartment origin of diseasespecific IgA type serum auto-antibodies was confirmed (in celiacdisease [55,56]) or excluded (in Guillain-Barre syndrome [57])based on the results of various IgA subtyping assays in previousstudies. Recent data [58] indicated that elevated concentrationsof IgA/sIgA in cirrhotic patients might be explained partly by thealtered expression/concentration, glycosylation, and function ofmonocytic CD89 (FcaRI) and its soluble form (sCD89). Further-more, sCD89 molecules showed a higher affinity to pIgA com-pared to mIgA. In this way, the enhanced binding of sCD89 toIgA might provide an indirect evidence for the increased poly-meric state of this molecule. Unfortunately, our biochip techniqueusing ethanol-fixed neutrophil substrates is not suitable for exam-ining whether soluble CD89 is complexed with ANCA IgA. CD89 isconstitutively expressed on neutrophils [59] and so interferedwith the examination of CD89 expression on ANCA IgA molecules.Namely, incubation of neutrophil substrates with either ANCAIgA-positive or negative sera or buffer, resulted in an abundantand similar CD89 expression detected by IIF (data not shown).

The gut mucosal origin of IgA postulated either in previous orin present study was not necessarily supported by all earlyreports. Despite significantly elevated serum mIgA and pIgA lev-els, normal mIgA, pIgA, and SC secretion rates were detected inthe jejunal fluid of patients with alcoholic cirrhosis by Colombelet al. [60]. Free dimeric IgA was not found in the jejunal fluid as

464 Journal of Hepatology 201

well. Authors assumed that an abnormal synthesis and transportat the jejunal level do not play a role in the mechanism resultingin increased circulating pIgA in patients with alcoholic cirrhosis.Of note, in the study of Colombel et al., other parts of the intesti-nal tract than jejunum were not studied; presumably thesemucosal areas behave differently. Moreover, the total mIgA andpIgA were measured and one cannot exclude that the small, anti-gen-specific subfractions of the total IgA (e.g., ANCA IgA) mightact differently than the bulk IgA mass.

The role of bacterial antigens in the induction of ANCA forma-tion is also well established in some disorders and experimentalanimal models [6] on the basis of cross-reactivity of certain bacte-rial constituents with host proteins due to sequence or structuralsimilarities [61]. In addition, the presence of classical and atypicalP-ANCA in inflammatory bowel diseases or autoimmune liver dis-eases was suggested to reflect an abnormal immune response tointestinal microorganisms [10–12]. Yet, ANCA is detectable inother chronic conditions complicated by systemic infections[62,6]. Moreover, the genetic sequence of Staphylococcus aureusis complementary to critical sequences of PR-3, the main targetantigen of C-ANCA [63]. In an experimental system, the immuniza-tion of rats with the pasteurized protein from E. coli and Staphylo-coccus resulted in circulating ANCA [64]. It has also been shownthat B cells, stimulated by bacterial unmethlylated CpG oligode-oxynucleotides via TLR-9, produce ANCA [65]. Our findings in theclinical part of the study can further support the link between bac-terial infections and induction of ANCA formation. The occurrenceof ANCA IgA positivity was significantly higher in patients with his-tory of CSI in the past as compared to those without (45.8% vs.30.4%, p <0.01) assessed retrospectively. Moreover, in a 2-year fol-low-up study, the presence of IgA class ANCA has been proved anindependent risk factor for the development of upcoming CSI aswell. The pathogenetic role of ANCA in disorders other than vascu-litis is controversial. In our patient cohort, the presence of IgA classANCA was not associated with either the overall or infection-related mortality. Similarly, in a very recent study of Reiberger[66], an elevated serum lipopolysaccharide-binding protein (LBP)level – as an indirect marker of BT – was not associated with ahigher mortality rate in cirrhosis. Our findings suggest that ANCAIgA has no pathogenetic role in the progression (injury, fibrogene-sis) of cirrhosis or outcome of bacterial complications. Even so,enhanced IgA class ANCA formation is probably not just an epiphe-nomenon of chronic inflammation, but an indicator of the underly-ing mechanism that ultimately leads to CSI. Further long-termprospective studies are needed to determine whether assessmentof IgA class ANCA, as an indirect marker of BT, has an additionalbenefit over the currently used factors (disease severity stage,presence of ascites) for the prediction of infections in patients withcirrhosis.

To determine the antigen specificities of cirrhosis relatedANCA IgA antibodies would be very important, but was beyondthe scope of the present study. Antigens of IgA type ANCAs aresupposedly not the classical cytosolic/granular neutrophil pro-teins, which were reported mainly as antigens of IgG type ANCAsin vasculitis. In our cohort, occurrence of anti-PR3 and anti-MPOpositivity was low. Similarly, Schwarze et al. [67] could not dem-onstrate any reactivity of IgA class ANCA with the known cyto-solic and granular proteins of neutrophils in autoimmune liverdiseases using ELISAs. Furthermore, Western blot analysis per-formed with whole cell lysates from neutrophils revealed a het-erogeneous pattern of reactive proteins of IgA class ANCA. Our

3 vol. 59 j 457–466

JOURNAL OF HEPATOLOGY

findings that all the three classical ANCA patterns were foundamong IgA type ANCAs, imply that multiple antigen might existat the background of their formation as well.

Only a few previous studies [4,68,69] assessed the potentialrelationship between the presence of IgG ANCA and existence orseverity of cirrhosis, and the results of these works are rather con-troversial. In the present study, ANCA IgG response and itschanges did not parallel the ANCA IgA response. In alcoholicpatients, occurrence of IgG ANCA was much lower, approximatelyhalf of the IgA levels, already in the less severe disease stage. Innon-alcoholic patients, the ANCA IgG positivity rate decreasedgradually according to disease severity and reached this markedlyreduced level compared to IgA ANCA only in advanced disease.These alterations in ANCA IgA and IgG response clearly reflectthose tendencies known from vaccination studies in this patientpopulation and presumably reflect the impaired adaptive immunesystem in cirrhosis, mainly in advanced stage and direct inhibitoryeffect of alcohol on T-cell-mediated immunity [70]. After pneu-mococcal vaccination, anti-PPS (pneumococcal polysaccharide)IgA antibody levels were significantly higher than in control sub-jects, whereas IgG levels were reduced [71]. Considerably lowerimmunogenicity and faster decline of specific, protective IgGresponse were reported in individuals with cirrhosis – particularlyin the alcohol-induced form – after hepatitis B vaccination com-pared to CLD [72]. Patients with compensated cirrhosis were fivetimes more likely to respond to hepatitis A vaccination comparedwith cirrhotic patients in decompensated stage [73].

In summary, the presence of IgA class ANCA was common inpatients with cirrhosis and was associated with advanced diseaseand portal hypertension. Furthermore, presence of ANCA IgA wasindependently associated with the development of clinically sig-nificant bacterial infections during a 2-year follow-up. Markedincrease in the proportion of IgA2 type of ANCA with the presenceof secretory component supports the involvement of gut mucosalimmune system and sustained exposure to bacterial constituentsas a trigger of ANCA formation in liver cirrhosis.

Financial support

Maria Papp was supported by the Janos Bolyai Research Scholar-ship of the Hungarian Academy of Sciences. Peter Antal-Szalmáswas supported by the National Scientific Research Programs ofthe Hungarian Ministry of Social Welfare (OTKA; T046694), andby the TÁMOP 4.2.1./B-09/1/KONV-2010-0007 and TÁMOP-4.2.2.A-11/1/KONV-2012-0025 projects. These latter projectsare implemented through the New Hungary Development Plan,co-financed by the European Social Fund.

Conflict of interest

Kai Fechner, Silvia Jacobsen and Bianca Teegen are employed atEUROIMMUN Medizinische Labordiagnostika AG, a manufacturerof in vitro diagnostics.

Acknowledgements

The authors are very much grateful to György Vereb jr. M.D.,Ph.D., D.sc. for his kind help in the introduction of the Image Jsoftware and to Róza Földesi and Éva Tömöri for their technicalassistance.

Journal of Hepatology 201

Supplementary data

Supplementary data associated with this article can be found,in the online version, at http://dx.doi.org/10.1016/j.jhep.2013.04.018.

References

[1] Kallenberg CG, Mulder AH, Tervaert JW. Antineutrophil cytoplasmic anti-bodies: a still-growing class of autoantibodies in inflammatory disorders.Am J Med 1992;93:675–682.

[2] Papp M, Altorjay I, Norman GL, Shums Z, Palatka K, Vitalis Z, et al.Seroreactivity to microbial components in Crohn’s disease is associated withileal involvement, noninflammatory disease behavior and NOD2/CARD15genotype, but not with risk for surgery in a Hungarian cohort of IBD patients.Inflamm Bowel Dis 2007;13:984–992.

[3] Mulder AH, Horst G, van Leeuwen MA, Limburg PC, Kallenberg CG.Antineutrophil cytoplasmic antibodies in rheumatoid arthritis. Character-ization and clinical correlations. Arthritis Rheum 1993;36:1054–1060.

[4] Mulder AH, Horst G, Haagsma EB, Limburg PC, Kleibeuker JH, Kallenberg CG.Prevalence and characterization of neutrophil cytoplasmic antibodies inautoimmune liver diseases. Hepatology 1993;17:411–417.

[5] Bansi D, Chapman R, Fleming K. Antineutrophil cytoplasmic antibodies inchronic liver diseases: prevalence, titre, specificity and IgG subclass. JHepatol 1996;24:581–586.

[6] Csernok E, Lamprecht P, Gross WL. Clinical and immunological features ofdrug-induced and infection-induced proteinase 3-antineutrophil cytoplas-mic antibodies and myeloperoxidase-antineutrophil cytoplasmic antibodiesand vasculitis. Curr Opin Rheumatol 2010;22:43–48.

[7] Terjung B, Worman HJ, Herzog V, Sauerbruch T, Spengler U. Differentiation ofantineutrophil nuclear antibodies in inflammatory bowel and autoimmuneliver diseases from antineutrophil cytoplasmic antibodies (p-ANCA) usingimmunofluorescence microscopy. Clin Exp Immunol 2001;126:37–46.

[8] Fulcher DA. Anti-neutrophil cytoplasmic antibodies in hepatobiliary disease.J Gastroenterol Hepatol 2000;15:344–345.

[9] Falk RJ, Jennette JC. ANCA disease: where is this field heading? J Am SocNephrol 2010;21:745–752.

[10] Terjung B, Sohne J, Lechtenberg B, Gottwein J, Muennich M, Herzog V, et al.P-ANCAs in autoimmune liver disorders recognise human beta-tubulinisotype 5 and cross-react with microbial protein FtsZ. Gut 2010;59:808–816.

[11] Yang P, Danielsson D, Jarnerot G. Escherichia coli and Proteus mirabilis inhibitthe perinuclear but not the circulating antineutrophil cytoplasmic antibodyreaction. Scand J Gastroenterol 1998;33:529–534.

[12] Seibold F, Brandwein S, Simpson S, Terhorst C, Elson CO. PANCA represents across-reactivity to enteric bacterial antigens. J Clin Immunol1998;18:153–160.

[13] Selmi C, Gershwin ME. Autoantibodies in autoimmune liver disease:biomarkers versus epiphenomena. Gut 2010;59:712–713.

[14] Papp M, Norman GL, Vitalis Z, Tornai I, Altorjay I, Foldi I, et al. Presence ofanti-microbial antibodies in liver cirrhosis – a tell-tale sign of compromisedimmunity? PloS One 2010;5:e12957.

[15] Christou L, Pappas G, Falagas ME. Bacterial infection-related morbidity andmortality in cirrhosis. Am J Gastroenterol 2007;102:1510–1517.

[16] Tandon P, Garcia-Tsao G. Bacterial infections, sepsis, and multiorgan failurein cirrhosis. Semin Liver Dis 2008;28:26–42.

[17] Borzio M, Salerno F, Piantoni L, Cazzaniga M, Angeli P, Bissoli F, et al.Bacterial infection in patients with advanced cirrhosis: a multicentreprospective study. Dig Liver Dis 2001;33:41–48.

[18] Wiest R, Garcia-Tsao G. Bacterial translocation (BT) in cirrhosis. Hepatology2005;41:422–433.

[19] Saab S, Hernandez JC, Chi AC, Tong MJ. Oral antibiotic prophylaxis reducesspontaneous bacterial peritonitis occurrence and improves short-termsurvival in cirrhosis: a meta-analysis. Am J Gastroenterol2009;104:993–1001, quiz 1002.

[20] Carey E, Carey WD. Noninvasive tests for liver disease, fibrosis, and cirrhosis:is liver biopsy obsolete? Cleve Clin J Med 2010;77:519–527.

[21] Durand F, Valla D. Assessment of the prognosis of cirrhosis: Child-Pughversus MELD. J Hepatol 2005;42:S100–S107.

[22] Garcia-Tsao G, Sanyal AJ, Grace ND, Carey W. Prevention and management ofgastroesophageal varices and variceal hemorrhage in cirrhosis. Hepatology2007;46:922–938.

3 vol. 59 j 457–466 465

Research Article

[23] Wong F, Bernardi M, Balk R, Christman B, Moreau R, Garcia-Tsao G, et al.

Sepsis in cirrhosis: report on the 7th meeting of the International AscitesClub. Gut 2005;54:718–725.

[24] Lodato F, Azzaroli F, Di Girolamo M, Feletti V, Cecinato P, Lisotti A, et al.Proton pump inhibitors in cirrhosis: tradition or evidence based practice?World J Gastroenterol 2008;14:2980–2985.

[25] Papp M, Vitalis Z, Altorjay I, Tornai I, Udvardy M, Harsfalvi J, et al. Acutephase proteins in the diagnosis and prediction of cirrhosis associatedbacterial infections. Liver Int 2012;32:603–611.

[26] Bedossa P, Poynard T. An algorithm for the grading of activity in chronichepatitis C. The METAVIR Cooperative Study Group. Hepatology1996;24:289–293.

[27] Poupon R. Primary biliary cirrhosis: a 2010 update. J Hepatol2010;52:745–758.

[28] Goodman ZD. Grading and staging systems for inflammation and fibrosis inchronic liver diseases. J Hepatol 2007;47:598–607.

[29] Kaplan MM, Gershwin ME. Primary biliary cirrhosis. N Engl J Med2005;353:1261–1273.

[30] Savige J, Gillis D, Benson E, Davies D, Esnault V, Falk RJ, et al. InternationalConsensus Statement on testing and reporting of antineutrophil cytoplasmicantibodies (ANCA). Am J Clin Pathol 1999;111:507–513.

[31] Collins TJ. ImageJ for microscopy. BioTechniques 2007;43:25–30.[32] Papp M, Norman GL, Altorjay I, Lakatos PL. Utility of serological markers in

inflammatory bowel diseases: gadget or magic? World J Gastroenterol2007;13:2028–2036.

[33] van de Wiel A, Schuurman HJ, Kater L. Alcoholic liver disease: an IgA-associated disorder. Scand J Gastroenterol 1987;22:1025–1030.

[34] De Palma GD, Rega M, Masone S, Persico F, Siciliano S, Patrone F, et al.Mucosal abnormalities of the small bowel in patients with cirrhosis andportal hypertension: a capsule endoscopy study. Gastrointest Endosc2005;62:529–534.

[35] Bhonchal S, Nain CK, Prasad KK, Nada R, Sharma AK, Sinha SK, et al.Functional and morphological alterations in small intestine mucosa ofchronic alcoholics. J Gastroenterol Hepatol 2008;23:e43–e48.

[36] Bauer TM, Steinbruckner B, Brinkmann FE, Ditzen AK, Schwacha H, Aponte JJ,et al. Small intestinal bacterial overgrowth in patients with cirrhosis:prevalence and relation with spontaneous bacterial peritonitis. Am JGastroenterol 2001;96:2962–2967.

[37] Staun-Olsen P, Bjorneboe M, Prytz H, Thomsen AC, Orskov F. Escherichia coliantibodies in alcoholic liver disease. Correlation to alcohol consumption,alcoholic hepatitis, and serum IgA. Scand J Gastroenterol 1983;18:889–896.

[38] Bjorneboe M, Prytz H, Orskov F. Antibodies to intestinal microbes in serumof patients with cirrhosis of the liver. Lancet 1972;1:58–60.

[39] Protell RL, Soloway RD, Martin WJ, Schoenfield LJ, Summerskill WH. Anti-Salmonella agglutinins in chronic active liver disease. Lancet1971;2:330–332.

[40] Nolan JP, DeLissio MG, Camara DS, Feind DM, Gagliardi NC. IgA antibody tolipid A in alcoholic liver disease. Lancet 1986;1:176–179.

[41] Kreisel W, Siegel A, Bahler A, Spamer C, Schiltz E, Kist M, et al. Highprevalence of antibodies to calreticulin of the IgA class in primary biliarycirrhosis: a possible role of gut-derived bacterial antigens in its aetiology?Scand J Gastroenterol 1999;34:623–628.

[42] Pelletier G, Briantais MJ, Buffet C, Pillot J, Etienne JP. Serum and intestinalsecretory IgA in alcoholic cirrhosis of the liver. Gut 1982;23:475–480.

[43] Macpherson AJ, Geuking MB, McCoy KD. Homeland security: IgA immunityat the frontiers of the body. Trends Immunol 2012;33:160–167.

[44] Hooper LV, Macpherson AJ. Immune adaptations that maintain homeostasiswith the intestinal microbiota. Nat Rev Immunol 2010;10:159–169.

[45] Slack E, Balmer ML, Fritz JH, Hapfelmeier S. Functional flexibility of intestinalIgA – broadening the fine line. Front Immunol 2012;3:100.

[46] Macpherson AJ, Geuking MB, Slack E, Hapfelmeier S, McCoy KD. The habitat,double life, citizenship, and forgetfulness of IgA. Immunol Rev2012;245:132–146.

[47] Suzuki K, Fagarasan S. How host-bacterial interactions lead to IgA synthesisin the gut. Trends Immunol 2008;29:523–531.

[48] Corthesy B. Secretory immunoglobulin A: well beyond immune exclusion atmucosal surfaces. Immunopharmacol Immunotoxicol 2009;31:174–179.

[49] Fagarasan S. Evolution, development, mechanism and function of IgA in thegut. Curr Opin Immunol 2008;20:170–177.

[50] Blaas SH, Stieber-Gunckel M, Falk W, Obermeier F, Rogler G. CpG-oligode-oxynucleotides stimulate immunoglobulin A secretion in intestinal mucosalB cells. Clin Exp Immunol 2009;155:534–540.

466 Journal of Hepatology 201

[51] Fritz JH, Rojas OL, Simard N, McCarthy DD, Hapfelmeier S, Rubino S, et al.Acquisition of a multifunctional IgA+ plasma cell phenotype in the gut.Nature 2012;481:199–203.

[52] Pabst O. New concepts in the generation and functions of IgA. Nat RevImmunol 2012;12:821–832.

[53] Brandtzaeg P. Update on mucosal immunoglobulin A in gastrointestinaldisease. Curr Opin Gastroenterol 2010;26:554–563.

[54] Woof JM, Kerr MA. The function of immunoglobulin A in immunity. J Pathol2006;208:270–282.

[55] Osman AA, Richter T, Stern M, Mothes T. The IgA subclass distributions ofendomysium and gliadin antibodies in human sera are different. Clin ChimActa 1996;255:145–152.

[56] Colombel JF, Mascart-Lemone F, Nemeth J, Vaerman JP, Dive C, Rambaud JC.Jejunal immunoglobulin and antigliadin antibody secretion in adult coeliacdisease. Gut 1990;31:1345–1349.

[57] Koga M, Yuki N, Hirata K. Subclass distribution and the secretory componentof serum IgA anti-ganglioside antibodies in Guillain-Barre syndrome afterCampylobacter jejuni enteritis. J Neuroimmunol 1999;96:245–250.

[58] Tissandie E, Morelle W, Berthelot L, Vrtovsnik F, Daugas E, Walker F, et al.Both IgA nephropathy and alcoholic cirrhosis feature abnormally glycosyl-ated IgA1 and soluble CD89-IgA and IgG-IgA complexes: common mecha-nisms for distinct diseases. Kidney Int 2011;80:1352–1363.

[59] Hamre R, Farstad IN, Brandtzaeg P, Morton HC. Expression and modulationof the human immunoglobulin A Fc receptor (CD89) and the FcR gammachain on myeloid cells in blood and tissue. Scand J Immunol2003;57:506–516.

[60] Colombel JF, Vaerman JP, Mesnard B, Dehennin JP, Dive C, Rambaud JC.Jejunal immunoglobulin secretion in alcoholic patients with and withoutcirrhosis. J Hepatol 1991;12:145–149.

[61] Willcocks LC, Lyons PA, Rees AJ, Smith KG. The contribution of geneticvariation and infection to the pathogenesis of ANCA-associated systemicvasculitis. Arthritis Res Ther 2010;12:202.

[62] Bonaci-Nikolic B, Andrejevic S, Pavlovic M, Dimcic Z, Ivanovic B, Nikolic M.Prolonged infections associated with antineutrophil cytoplasmic antibodiesspecific to proteinase 3 and myeloperoxidase: diagnostic and therapeuticchallenge. Clin Rheumatol 2010;29:893–904.

[63] Pendergraft 3rd WF, Preston GA, Shah RR, Tropsha A, Carter Jr CW, JennetteJC, et al. Autoimmunity is triggered by cPR-3(105–201), a protein comple-mentary to human autoantigen proteinase-3. Nat Med 2004;10:72–79.

[64] Savige J, Nassis L, Cooper T, Paspaliaris B, Martinello P, MacGregor D.Antineutrophil cytoplasmic antibody (ANCA)-associated systemic vasculitisafter immunisation with bacterial proteins. Clin Exp Rheumatol2002;20:783–789.

[65] Hurtado PR, Jeffs L, Nitschke J, Patel M, Sarvestani G, Cassidy J, et al. CpGoligodeoxynucleotide stimulates production of anti-neutrophil cytoplasmicantibodies in ANCA associated vasculitis. BMC Immunol 2008;9:34.

[66] Reiberger T, Ferlitsch A, Payer BA, Mandorfer M, Heinisch BB, Hayden H,et al. Non-selective betablocker therapy decreases intestinal permeabilityand serum levels of LBP and IL-6 in patients with cirrhosis. Journal ofhepatology 2013;58(5):911–921.

[67] Schwarze C, Terjung B, Lilienweiss P, Beuers U, Herzog V, Sauerbruch T, et al.IgA class antineutrophil cytoplasmic antibodies in primary sclerosingcholangitis and autoimmune hepatitis. Clin Exp Immunol2003;133:283–289.

[68] Roozendaal C, de Jong MA, van den Berg AP, van Wijk RT, Limburg PC,Kallenberg CG. Clinical significance of anti-neutrophil cytoplasmic antibod-ies (ANCA) in autoimmune liver diseases. J Hepatol 2000;32:734–741.

[69] De Riva V, Celadin M, Pittoni M, Plebani M, Angeli P. What is behind thepresence of anti-neutrophil cytoplasmatic antibodies in chronic liverdisease? Liver Int 2009;29:865–870.

[70] Leber B, Mayrhauser U, Rybczynski M, Stadlbauer V. Innate immunedysfunction in acute and chronic liver disease. Wien Klin Wochenschr2009;121:732–744.

[71] McCashland TM, Preheim LC, Gentry MJ. Pneumococcal vaccine response incirrhosis and liver transplantation. J Infect Dis 2000;181:757–760.

[72] De Maria N, Idilman R, Colantoni A, Van Thiel DH. Increased effectiveimmunogenicity to high-dose and short-interval hepatitis B virus vaccina-tion in individuals with chronic hepatitis without cirrhosis. J Viral Hepat2001;8:372–376.

[73] Arguedas MR, Johnson A, Eloubeidi MA, Fallon MB. Immunogenicity ofhepatitis A vaccination in decompensated cirrhotic patients. Hepatology2001;34:28–31.

3 vol. 59 j 457–466