PBC Screen: An IgG/IgA dual isotype ELISA detecting multiple mitochondrial and nuclear autoantibodies specific for primary biliary cirrhosis

lable at ScienceDirect

Journal of Autoimmunity 35 (2010) 436e442

Contents lists avai

Journal of Autoimmunity

journal homepage: www.elsevier .com/locate/ jaut imm

PBC Screen: An IgG/IgA dual isotype ELISA detecting multiple mitochondrialand nuclear autoantibodies specific for primary biliary cirrhosis

Haiying Liu a,b,1, Gary L. Norman c,1, Zakera Shums c, Howard J. Worman d, Edward L. Krawitt e,Nicola Bizzaro f, Diego Vergani g, Dimitrios P. Bogdanos g, George N. Dalekos h, Piotr Milkiewicz i,Albert J. Czaja j, E. Jenny Heathcote k, Gideon M. Hirschfield k, Eng M. Tan l, Kiyomitsu Miyachim,Monica Bignotto n, Pier Maria Battezzati n, Ana Lleo a,o, Patrick S. Leung p, Mauro Podda a,o,M. Eric Gershwin p, Pietro Invernizzi a,p,*aDivision of Internal Medicine and Center for Autoimmune Liver Diseases, IRCCS Istituto Clinico Humanitas, Rozzano, ItalybClinical Laboratory, Guangzhou Women and Children’s Medical Center, Guangzhou, Chinac INOVA Diagnostics, San Diego, CA, USAdDepartment of Medicine and Department of Pathology and Cell Biology, College of Physicians and Surgeons, Columbia University, New York, NY, USAeDepartment of Medicine, University of Vermont, Burlington, VT, USAf Laboratorio di Patologia Clinica, Ospedale Civile, Tolmezzo, UD, Italyg Institute of Liver Studies, King’s College London School of Medicine at King’s College Hospital, London, UKhDepartment of Medicine, Academic Liver Unit and Research Laboratory of Internal Medicine, Larissa Medical School, University of Thessaly, Larissa, Greecei Liver Unit, Pomeranian Medical School, Szczecin, PolandjDivision of Gastroenterology and Hepatology, Mayo Clinic College of Medicine, Rochester, MN, USAk Toronto Western Hospital, University of Toronto, Toronto, Ontario, CanadalDepartment of Molecular and Experimental Medicine, The Scripps Research Institute, La Jolla, CA, USAmKeigu Clinic, Consolidated Diagnostic Divisions, Health Sciences Research Institute, Inc., Yokohama Kanagawa-ken, Japann Liver and Gastroenterology Unit, San Paolo School of Medicine, Università degli Studi di Milano, Milan, ItalyoDepartment of Translational Medicine, Università degli Studi di Milano, Rozzano, ItalypDivision of Rheumatology, Allergy and Clinical Immunology, University of California at Davis, Davis, CA, USA

a r t i c l e i n f o

Article history:Received 9 August 2010Received in revised form14 September 2010Accepted 14 September 2010

Keywords:BiomarkersAutoantibodiesenzyme-linked immunosorbent assay(ELISA)Antimitochondrial antibodysp100gp210

Abbreviations: PBC, Primary biliary cirrhosis; AMmicroscopy; PDC-E2, E2 subunits of the pyruvate dehysubunit of the 2-oxo-glutarate dehydrogenase compldiseases; ROC, Receiver operator characteristic; PLR, P* Corresponding author. Division of Internal Medici

Italy. Tel.: þ39 02 8224 5128; fax: þ39 02 8224 5191E-mail address: [email protected] (P.

1 These authors contributed equally to this study.

0896-8411/$ e see front matter � 2010 Elsevier Ltd.doi:10.1016/j.jaut.2010.09.005

a b s t r a c t

A dual isotype (IgG, IgA) enzyme-linked immunosorbent assay (ELISA) designed to provide enhanceddetection of primary biliary cirrhosis (PBC)-specific autoantibodies against both major mitochondrial andnuclear antigens has been developed and recently become commercially available. The assay (PBCScreen) simultaneously detects IgG and IgA autoantibodies to the immunodominant portions of the 3major mitochondrial (MIT3) and nuclear (gp210, and sp100) antigens. The aim of this study was tocompare the performance of the PBC Screen to the combined performance obtained with individual IgGELISAs to MIT3, gp210, and sp100 on a large group of selected patients from multiple centers. A total of1175 patients with PBC and 1232 subjects without PBC were evaluated. Non-PBC groups included healthycontrols (624) as well as individuals with autoimmune hepatitis (281), primary sclerosing cholangitis(77), viral hepatitis (91 hepatitis B and 98 hepatitis C), other liver diseases (31), and other infectious orautoimmune diseases (30). The PBC Screen at the receiver operator characteristic optimized cutoff of 27.8units, had an overall sensitivity of 83.8%, specificity of 94.7% and area under curve of 0.9212. This wassimilar to the specificity of 96.1% obtained by the combined results of individual MIT3, sp100, and gp210IgG ELISAs (kappa index at 0.898). Of the 253 PBC patients without AMA detectable by immunofluo-rescence, 113 (44.7%) were interpreted as positive for PBC-specific autoantibodies. In conclusion, the PBC

A, Antimitochondrial antibodies; ELISA, Enzyme-linked immunosorbent assay; IIF, Indirect immunofluorescencedrogenase complex; BCOADC-E2, E2 subunit of the branched-chain 2-oxo-acid dehydrogenase complex; OGDC-E2, E2ex; ANA, Antinuclear antibodies; AIH, Autoimmune hepatitis; PSC, Primary sclerosing cholangitis; AID, Autoimmuneositive likelihood ratio; NLR, Negative likelihood ratio.ne and Center for Autoimmune Liver Diseases, IRCCS Istituto Clinico Humanitas, Via A. Manzoni 113, 20089 Rozzano,.Invernizzi).

All rights reserved.

H. Liu et al. / Journal of Autoimmunity 35 (2010) 436e442 437

Table 1Sources of samples enrolled in this study.

Sources PBC patients

No. AMAþ

Gershwin 55 55Heathcote 320 253Bogdanos/Vergani/Dalekos 132 74Worman 52 52INOVAProMedX

Phase I 559 415Czaja 7 7Krawitt 12 12Bizzaro 243 153Invernizzi 194 178Miyachi/Tan 160 138

Phase II 616 488

Total 1175 922

Screen is an appropriate first-line test for the diagnosis of PBC, including for patients negative formarkers assessed using conventional methods.

� 2010 Elsevier Ltd. All rights reserved.

1. Introduction

Primary biliary cirrhosis (PBC) is a chronic liver disease char-acterized by the progressive destruction of the small intra-hepaticbile ducts, increasing functional impairment of the liver and, overtime, to liver failure and the necessity of liver transplantation [1].Serological assays are important for the diagnosis of PBC [2e4].Anti-mitochondrial antibodies (AMA), detected by indirect immu-nofluorescence microscopy (IIF) on rodent kidney/stomach/liversections or HEp-2 cell line substrates, are the classic serologicalmarkers of PBC and are found in up to 90e95% of PBC patients [5,6].IIF, however, is labor-intensive and highly dependent on skill andexperience of the observer for interpretation of the diagnosticlabeling patterns. Moreover, patients with liver or extra-hepaticautoimmune diseases often have antibodies to multiple targetsand this can make interpretation of the resulting complex patternsdifficult, leading to both false positive and negative results [7].ELISAs developed using a purified mitochondrial antigen fractionoriginally designated “M2” have permitted more sensitive andobjective detection of AMA antibodies. More recently, Gershwin,Leung, and colleagues created a recombinant fusion protein (MIT3),which includes the immunodominant portions of the 3 primarytargets of AMA, namely the E2 subunits of the pyruvate dehydro-genase complex (PDC-E2), the branched-chain 2-oxo-acid dehy-drogenase complex (BCOADC-E2) and the 2-oxo-glutaratedehydrogenase complex (OGDC-E2) [8]. This antigen has demon-strated a higher diagnostic sensitivity than conventional ELISAs forAMA [8]. Lastly, it has been suggested that Luminex and an ELISAusing a mixture of purified PDC and MIT3 as antigenic targets mayincrease sensitivity of the AMA detection systems [9,10].

AMA are considered to be the hallmark of PBC, but they are notthe only disease-specific autoantibodies. Antinuclear antibodies(ANA) from patients with PBC recognize 2 highly specific nuclearantigens. ANA in patients with PBC that label the nuclear peripheryin a punctate rim-like pattern most often recognize gp210, anintegral membrane protein associated with the nuclear porecomplex [11]. ANA that give a multiple nuclear dot pattern recog-nize the nuclear body-associated protein sp100 [12]. Autoanti-bodies against gp210 and sp100 are detected in approximately 25%

Controls

AMA- PBC/AIH No.

2067 10 9858 165

429

500

125 10 8167 1952 38

9016 2 9922 84

128 11 416

253 21 1232

of patients with PBC using anti-total IgG antiserum but up to65% when anti-IgG isotype specific antisera was used [13], some ofwhom do not have detectable AMA, and are highly specific for thedisease [3,4]. Hence, tests to detect gp210 and sp100 antibodiesincrease the specificity and sensitivity in diagnosing PBC, especiallyin cases where the clinical presentation and/or serological picturemay be unclear. ELISAs utilizing recombinant fusion proteins orsynthetic polypeptides have been developed to detect gp210 andsp100 antibodies in patients with PBC [12,14,15]. In addition, weand others showed that these nuclear antibodies are associatedwith more severe disease and have prognostic as well as diagnosticvalue in PBC (in contrast to AMA alone) [16,17].

Optimal detection of antibodies against the immunodominantportions of the major mitochondrial as well as the nuclear antigenswould require performing 3 separate assays. Practically, however,the costs and logistics involved in running 3 tests may discouragethis approach and, as a result, some individuals with PBC mayremain undiagnosed or present clinicianswith diagnostic dilemmas.To address this problem, a new multi-analyte, dual isotype ELISA,which simultaneously screens for IgG and IgA antibodies to the 3immunodominant epitopes of AMA, gp210, and sp100 (PBC Screen)in each test serum has been developed. The present multi-centerstudywas performed to assess this newly developed ELISA in a largecohort of well-characterized PBC patients and patients with otherdiseases assembled from centers in North America, Europe, andAsia,and to test the concordance of the method with results obtainedusing the individual MIT3, gp210, and sp100 ELISAs.

2. Material and methods

2.1. Subjects

The study was performed in 2 phases. The first phase includedsubjects from the United States, Canada, United Kingdom, andGreece from 2005e2006, while the second phase was performedfrom 2007 to 2009 on an independent collection of sera fromsubjects in the United States, Italy, and Japan. We enrolled a total of1175well-defined PBC patients, including 922with detectable AMAand 253 without detectable AMA as determined by IIF, and 1232

AIH PSC Viral hepatitis Other liverdiseases

Otherinfectious/AID

Healthycontrols

2042 47 97 147 11

1 2 129

500

49 48 149 20 30 520177 8 1036 1 1

19 20 6040 44

232 29 40 11 104

281 77 189 31 30 624

Fig. 1. ROC curve of the PBC Screen showing sensitivity and specificity. The area undercurve (AUC) value for the PBC Screen test was 0.9212 on the well-defined 1175 patientswith PBC and 1232 controls with non-PBC. ROC analysis resulted in a sensitivity of83.8% (985/1175) for the complete cohort of PBC patients (including both AMA-positiveand AMA-negative patients), with a specificity of 94.7% at an optimal cutoff value of27.8 units for diagnosis of PBC. At the optimal cutoff, positive likelihood ratio, negativelikelihood ratio and diagnostic accuracy were 15.81, 0.17, and 0.894, respectively.

H. Liu et al. / Journal of Autoimmunity 35 (2010) 436e442438

controls (Table 1). A specific effort was made to enroll AMA-nega-tive PBC cases. In order to evaluate the performance of a noveltesting system in the “real” clinical setting, all the antibody testingswere performed in different laboratories by using the same ELISAassays provided by INOVA (INOVA Diagnostics, San Diego, CA).

The diagnosis of PBC was based upon internationally acceptedcriteria [1,5,6,18]. Briefly, 2 of the following 3 criteria had to be metfor a diagnosis of PBC: 1) serum AMA (titer �1:40), 2) a cholestaticpattern of liver biochemistry with at least 1 of serum bilirubin,alkaline phosphatase, or gamma-glutamyl transpeptidase abovethe reference range, and 3) diagnostic or compatible liver histology.

Disease controls consisted of 5 groups: autoimmune hepatitis(AIH) (281), primary sclerosing cholangitis (PSC) (77), viral hepa-titis (189), other liver diseases (31), and other infectious or auto-immune diseases (AID) (30). The diagnosis of AIH was madeaccording to internationally recognized criteria [19]; the diagnosisof PSC was established based on diagnostic features shown byendoscopic retrograde cholangiography or magnetic resonancecholangiopancreatograpgy [5,20]. There were 21 cases with bothfeatures and a diagnosis of PBC [1,5,6,18] and AIH [19] that wereconsidered affected by an PBC/AIH overlap syndrome. The viralhepatitis group included 91 patients with hepatitis B virus infec-tion, and 98 with hepatitis C virus infection. The “other liverdiseases” group included 11 patients with alcoholic liver disease, 9patients with vanishing bile duct syndrome, 10 with cryptogenichepatitis, and 1 with drug-induced hepatitis. The “other infectiousor AID” group included sera from 7 cases with H. pylori infection,6 with cytomegalovirus infection, 2 with herpes simplex virusinfection, and 15 cases with a diagnosis of systemic lupus eryth-ematosus, systemic sclerosis, or with ANA positive serology.

2.2. PBC Screen

A total of 2407 serum samples were screened for IgG and IgAautoantibodies to MIT3, gp210 and sp100 using the Quanta Lite�PBC Screen IgG/IgA ELISA (INOVA Diagnostics). Briefly, 100 mL of1:100 diluted serum was added to polystyrene microwells coatedwith a mixture of affinity-purified recombinant MIT3 antigen,which includes immunodominant portions of PDC-E2, BCOADC-E2,OGDC-E2 [21], and synthetic peptides corresponding to theimmunodominant regions of the sp100 [22,23] and gp210 [24].Reactions were allowed to proceed for 30 min at room tempera-ture. After washing 3 times, 100 mL of peroxidase-conjugated goatanti-human immunoglobulin IgG/IgA was added to each well andincubated for 30 min. Following washing, 100 mL of 3,30,5,50-tet-ramethylbenzidine chromogen was added to the wells, incubatedfor 30 min and 100 mL of 0.344 M H2SO4 stop solutionwas added toterminate the reaction. Absorbance was read at 450/620 nm andresults, expressed in arbitrary units, were calculated in reference toa kit-provided calibrator. The manufacturer’s cutoff was estab-lished at 25 units. Samples were considered as negative, (�20.0units), equivocal (20.1e24.9 units), or positive (�25.0 units). Apositive result indicated the presence of antibodies to 1 or more ofthe antigens in the PBC Screen. A negative result indicated theabsence of antibodies to MIT3, gp210, or sp100, or levels below thenegative cutoff of the assay.

2.3. Individual MIT3, gp210 and sp100-specific ELISAs

Sera from patients with PBC and controls were tested on theindividual M2 EP (MIT3), gp210 and sp100 IgG Quanta LIte�ELISAs (INOVA Diagnostics) in order to assess the concordance ofthe PBC Screen ELISA with the combined results of individualassays. All the ELISAs are FDA-cleared for in vitro diagnostic useand were performed according to the manufacturer’s instructions.

All steps were similar to the procedure of the PBC Screen ELISA,with the exception of the detection step. The individual ELISAs useperoxidase-conjugated goat anti-human immunoglobulin IgG fordetection, while the PBC Screen ELISA uses a peroxidase-conju-gated goat anti-human immunoglobulin IgG/IgA polyvalentconjugate.

2.4. Statistical analysis

An optimized cutoff value for the PBC Screenwas determined byreceiver operator characteristic (ROC) curve analysis with SPSS 16.0for Windows (SPSS Inc., Chicago, IL). The best cutoff point wasdefined as the highest value of sensitivity plus specificity. A positivelikelihood ratio (PLR) was calculated using the formula:sensitivityO(1-specificity). Similarly, a negative likelihood ratio(NLR) was calculated as follows: (1-sensitivity)/specificity. Sensi-tivity and specificity of the PBC Screen for diagnosis of PBC werecalculated both using the manufacturers’ cutoff and the cutoffobtained by ROC curve analysis of the study cohort described in thisstudy. Data were analyzed by KruskaleWallis H test and c2 test;Coincidence of two tests was compared with Kappa index. Two-tailed P values less than 0.05 were considered statistically signifi-cant. Scattergram was plotted using GraphPad prism (GraphPadSoftware, Inc., San Diego, CA).

3. Results

3.1. Sensitivity and specificity of the PBC Screen for diagnosis of PBC

ROC analysis resulted in a sensitivity of 83.8% (985/1175) for thecomplete cohort of PBC patients, including those with and withoutAMA detectable by IIF, with a specificity of 94.7% at an optimalcutoff value of 27.8 units for diagnosis of PBC (Fig. 1). The areaunder curve value for this assay was 0.9212. At the optimal cutoff,

Table 2Diagnostic characteristic of the PBC Screen with cutoff at 25.0 or 27.8 units.

Cutoff(units)

Sensitivity (%) Specificity(%)

PLR NLR Accuracy(%)

All AMAþ IIF AMA� IIF

25.0 84.3 94.9 45.5 93.8 13.38 0.17 89.127.8 83.8 94.6 44.7 94.7a 15.81 0.17 89.4Diff. �0.5 �0.3 �0.8 þ0.9 þ2.43 0 þ0.3

a Compared with cutoff at 25.0 units, P ¼ 0.2996.

A AMA-negative (by IIF) PBC samples (n=253)

46 13

20 1

3 11

MIT3=71 gp210=37

H. Liu et al. / Journal of Autoimmunity 35 (2010) 436e442 439

PLR, NLR, and diagnostic accuracy were 15.81, 0.17, and 0.894,respectively. The diagnostic characteristics of the new cutoff (27.8units) determined from the data in the present study werecompared with the manufacturer’s recommended cutoff (25.0units) (Table 2). Sensitivity and specificity at cutoffs of 25 unitswere 84.3%, and 93.8% compared to 83.8%, and 94.7% with theadjusted cutoff. The new cutoff did not affect the sensitivity forspecimens with detectable AMA by IIF, but did result in a 0.8%reduction in sensitivity for the AMA-negative specimens by IIF.Compared to the 25 unit cutoff, however, the new cutoff resulted inincreases of 0.9%, 2.43, and 0.3% for specificity, PLR, and diagnosticaccuracy, respectively.

The PBC Screen was positive for 94.6% (872/922) of the patientswith PBC and AMA detectable by IIF and for 44.7% (113/253) ofthose without detectable AMA by IIF. PBC-specific autoantibodieswere detected in 20 of 21 patients carrying a diagnosis of both PBCand AIH, and in 36 of 281 (12.8%) patients with only a diagnosis ofAIH. In patients with PSC, viral hepatitis, other liver diseases, otherinfectious or AID, and healthy controls, PBC-specific antibodieswere detected in 2/77 (2.6%), 13/189 (6.9%), 2/31 (6.5%), 1/30 (3.3%),and 11/624 (1.8%) of the specimens, respectively (Fig. 2). Themedian value of the PBC Screenwas 4.8 units in healthy subjects. Ofthe 65 control specimens that tested positive by the PBC Screen,only 17 (1 healthy control, 2 hepatitis B virus infection,14 AIH) werepositive with values of more than 100 units.

3.2. Antibody response to individual MIT3, gp210,and sp100 in PBC patients

Of the 1175 PBC patients, 78.6% (923), 17.9% (210), and 18.6%(219) specimens were respectively found to be positive for MIT3,

PBC Screen IgG/IgA

BP

C

(

11

=n

75)

A

Pe

v+

AM

BC

(=

n

9

)2

2

)3

52

=n(

CB

Pe

v-A

MA

A/C

BP

IH

(

12

=n

)

HIA

(n=2

18

) )7

7=

n(

CS

P

Vi

eh

lar

ap

titi

(s

n

)9

81

=

Ot

eh

rilv

re

esi

d

a

es

(=

n

3

)1

Ot

eh

r

fni

ce

tiu

o

A/s

ID

(=

n

30)

He

htla

yc

tn

o

r

lo

s(

26

=n

4)

0

100

200

300

400

500

N=2407

27.8

sti

nU

Fig. 2. Distribution of PBC specific autoantibodies measured by the PBC Screen invarious control groups. The long dashed line shows the new cutoff (27.8 units) recal-culated by ROC analysis on the basis of 2407 subjects in this study, and the medianlevels of the PBC Screen for each group are indicated by short real lines. Note.Abbreviations are same as that in the text.

gp210, and sp100 by individual ELISA tests (Fig. 3). Among them,155 specimens showed dual reactivity to MIT3 and sp100, 146showed dual reactivity to MIT3 and gp210, 5 to sp100 and gp210,and 31 to all of these. Combining the results of the 3 individual testsresulted in a sensitivity of 83.8% (985/1175).

3.3. PBC patients without AMA detected by IIF

A key strength of the PBC Screen is that it could identifypatients without AMA detected by routine IIF. From the multiplecenters in this study, we assembled a cohort of 253 subjectswith a diagnosis of PBC but without AMA detected by IIF. ThePBC Screen detected PBC-specific antibodies in 113 of the 253(44.7%) serum samples from the patients without AMA detect-able by IIF. When tested by individual assays, serum samplesfrom 117 of these patients had antibodies that reacted with atleast 1 of the 3 PBC-specific antigens. Overall, of these 117patients, 71 (60.7%) were positive for MIT3, 38 (32.5%) positivefor sp100, and 37 (30.8%) positive for gp210 (Fig. 3a). Twenty-eight (23.9%) of them had antibodies against more than 1 ofthese 3 PBC antigens. Inclusion of gp210 in the PBC Screenallowed for the detection of 20 patients exclusively positive forthis marker. Similarly, the inclusion of sp100 in the PBC Screenallowed the recognition of 23 patients with only sp100 anti-bodies. Three specimens were recognized by both gp210 andsp100, but not by MIT3. Of 922 PBC patients that had AMAdetected by IIF, 872 (94.6%) were considered to be positive usingthe PBC Screen and 868 (94.1%) were positive by the combinedindividual ELISAs (Fig. 3b). Of these, 16 specimens showedreactivity to sp100, and/or gp210.

B AMA-positive (by IIF) PBC samples (n=922)

23 sp100=38

545 133

9

30 144 2

5

MIT3=852

sp100=181

gp210=174

Fig. 3. Overlap of PBC-specific antibodies in PBC samples. The PBC Screen identified aspositive 113 of the 253 (44.7%) serum samples from patients with AMA not detected byIIF. When tested by individual assays, 117 AMA-negative PBC patients reacted with atleast 1 of the 3 PBC-specific antigens. Overall, of the 117 AMA-negative PBC patients, 71(60.7%) were positive for MIT3, 38 (32.5%) positive for sp100, and 37 (30.8%) positivefor gp210. The PBC Screen was positive in 46 patients with only gp210 and/or sp100antibodies. Of the 922 patients with AMA-positive PBC, 868 (94.1%) were positive bythe combined individual ELISAs and 16 specimens showed reactivity to only sp100and/or gp210.

Table 4Discrepant samples between the PBC Screen and combined individual tests.

No. AMA PBC Screen (units) MIT3 (units) gp210 (units) sp100 (units)

1 � 8.0 1.0 45.0 1.02 � 12.1 2.6 50.9 4.33 � 13.0 2.0 27.0 1.04 � 13.9 6.2 45.1 4.35 � 23.0 2.0 90.0 1.06 � 24.2 17.6 53.2 2.97 � 24.9 10.3 4.0 29.88 � 27.0 29.0 2.0 3.09 � 28.1 15.0 5.2 5.710 � 29.0 9.0 1.0 1.011 � 31.5 21.8 1.9 7.812 � 47.4 16.9 4.0 18.613 � 51.6 14.6 5.1 3.414 þ 4.0 66.7 3.1 6.515 þ 9.3 74.3 3.4 7.816 þ 11.4 22.2 30.1 6.217 þ 13.6 15.8 53.3 2.218 þ 25.5 9.2 75.1 4.319 þ 25.7 121.7 1.4 5.220 þ 27.9 8.8 5.3 24.421 þ 27.9 15.6 2.8 2.122 þ 29.9 6.5 1.9 3.223 þ 30.0 20.5 3.1 15.024 þ 30.5 11.2 3.9 3.525 þ 31.6 10.6 8.3 5.026 þ 34.6 23.5 5.0 5.527 þ 35.0 9.7 3.5 11.028 þ 36.5 4.9 2 3.329 þ 41.4 21.0 1.5 8.830 þ 41.5 24.2 10.1 5.131 þ 45.2 21.3 2.9 10.332 þ 59.6 20.2 5.4 4.033 þ 60.2 18.7 1.6 4.334 þ 71.1 22.7 4.9 5.435 þ 75.6 19.1 4.7 3.6

Note. Equivocal results (20.1e27.7 units) are interpreted as negative for the PBCScreen. Samples are interpreted as positive (�25.0 units) for individual MIT3, gp210or sp100 according to the manufacturer’s instructions.

H. Liu et al. / Journal of Autoimmunity 35 (2010) 436e442440

3.4. Coincidence of the PBC Screen with combinedindividual tests and IIF

The MIT3, gp210 and sp100 IgG individual ELISAs were per-formed on all 1175 PBC sera and 1179 controls. Sera from the 1175PBC patients were also all tested by IIF, although in multiple routineclinical laboratories. The sensitivity and specificity of IIF for thetotal PBC cohort was 78.5% and 95.9%, respectively. In comparison,the PBC Screen demonstrated a sensitivity of 83.8%, significantlyhigher than that of IIF (P ¼ 0.006), with a similar specificity(P ¼ 0.232). Of the 253 PBC patients who were AMA negative by IIF,113 (44.7%) were positive and 12 equivocal by the PBC Screen. Therewere 50 PBC patients who had detectable AMA by IIF, but werenegative or equivocal by the PBC Screen. Concordance of the PBCScreen with IIF was 86.1% at kappa index 0.549 (Table 3).

Concordance of the PBC Screen with the combined individualtests (MIT3, gp210 and sp100) was 97.6% with kappa index 0.912(Table 3). Of the 1175 PBC patients, 35 showed discrepant resultsbetween the PBC Screen and the combined tests. Specifically, 21sera tested positive by the PBC Screen were negative (10) orequivocal (9) by the combined tests, while 14 sera were positive byindividual tests, but negative, (8) or equivocal (6) by the PBC Screen(Table 4).

The sensitivity of the PBC Screen to detect PBC in patientswithout AMA detectable by IIF (44.7%, 113/253) was comparable tothat of combined individual tests (46.2%, 117/253), and its speci-ficity did not change significantly (94.7% compared to 94.5%)(Table 5). The 13 PBC patients with AMA detectable by IIF that hadequivocal results (20.1e27.7 units) on the PBC Screen wereconsidered to have negative results to perform these calculations.The 44.7% (113/253) sensitivity of the PBC Screen for PBC in patientswithout detectable AMA by IIF was strikingly higher than the 28.1%(71/253) sensitivity of the MIT3 IgG ELISA alone (P < 0.0001).The sensitivity of the PBC Screen for patients with PBC and AMAdetectable by IIF (94.6%, 872/922) was higher, but not significantlydifferent than that of the combined individual tests (94.1%, 868/922), or that of MIT3 IgG alone (92.4%, 852/922).

3.5. PBC Screen in Phase I and Phase II studies

Because the distributions of the PBC Screen results in PBCpatients, disease controls, and healthy controls were not normalwhen evaluated by KolmogoroveSmirnov test, KruskaleWallis Htests were applied to compare antibody levels for the 3 groups ofdifferent phases (Table 6). There were no differences between PBCpatients and disease controls for Phase I and Phase II, while therewas a significant statistical difference in the measurements ofhealthy controls (P < 0.001). The 816 controls involved in Phase Istudy with 520 healthy donors had a median level of 4.6 units, and104 controls in Phase II had a median level of 6.4 units. Despite thestatistical difference, however, there was no practical differencebetween 4.6 and 6.4 units. Both were unequivocally negative.

Table 3Coincidence of the PBC Screen with combined individual tests (MIT3, gp210 andsp100) and indirect immunofluorescence assay in PBC patients.a

Combined individual tests IIF

positive Equivocalb negative positive negative

PBC Screen positive 970 9 5 872 113equivocal 6 2 11 6 12negative 8 4 160 44 128

Coincidence 97.6% 86.1%Kappa index 0.912 0.549

a See text for details.b Equivocal: 20.1e27.7 units.

The frequencies of positive PBC Screen results in PBC patients fromthe various centers in our study are shown in Table 7. The positiverates of the PBC Screen for patients with AMA detectable by IIF indifferent laboratories worldwide ranged from 83.3 to 98.8% and thepositivity for patients with PBC without AMA detectable by IIFvaried from 28.4% to 62.5%; The high proportion of AMA-negativepatients reflects our specific case-finding effort.

4. Discussion

IIF microscopy is the standard for detection of AMA in manyclinical laboratories. However, ELISAs, especially those usingrecombinant proteins such as MIT3, are significantly more sensitive[8]. In addition, ELISAs make detection of antibodies against nuclearpore protein gp210 and the nuclear body protein sp100 more

Table 5Sensitivity and specificity of the PBC Screen, combined individual tests of MIT3,gp210 and sp100, and MIT3-based ELISA in PBC patients.

Assay Sensitivity Specificity

PBC AMA� by IIF PBC AMAþ by IIF

PBC Screen ELISA 44.7% (113/253) 94.6% (872/922) 94.7% (1167/1232)Combined

individual ELISAs46.2%a (117/253) 94.1%b (868/922) 94.5%c (1115/1179)

MIT3-based ELISA 28.1%d (71/253) 92.4%e (852/922) 95.6%f (1127/1179)

Compared with the PBC Screen: aP ¼ 0.8583, bP ¼ 0.6863, cP ¼ 0.8680, dP < 0.0001,eP ¼ 0.0590, fP ¼ 0.3229.

Table 6Comparison of the PBC Screen results in Phase I and Phase II studies.

Group Phase No. Median(units)

Range(25e75%)

KrukaleWallisH test

PBC patients I 559 160.6 94.9e198.0 0.574II 616 153.1 73.6e200.2

Disease controls I 296 7.7 4.9e12.8 0.403II 312 6.9 4.3e13.5

Healthy controls I 520 4.6 3.2e7.1 0.000II 104 6.4 4.7e8.7

H. Liu et al. / Journal of Autoimmunity 35 (2010) 436e442 441

accurate and objective [2e4]. Close to half of the specimens in ourstudy that were originally classified as “AMA-negative” PBC byexperienced hepatologists were shown to have PBC-specific anti-bodies using 1 or more of the individual MIT3, gp210, or sp100ELISAs. While this clearly indicates that effective serological diag-nosis of PBC requires testing for PBC-specific antibodies against all 3of the major mitochondrial antigens and the 2 major nuclear anti-gens, the costs and labor required for 3 separate tests make somelaboratories reluctant to incorporate testing for gp210 and sp100into their protocols. In addition, the simultaneous detection ofPBC-specific ANA is a major advantage of the PBC Screen in light oftheir prognostic value [16,17].

In the present study, we evaluated a new PBC-screening assaythat allows dual IgG and IgA isotype testing for MIT3, gp210,and sp100 antibodies in a single ELISA and compared the results tothose obtained with individual IgG ELISAs. The strength of thisstudy was that it utilized 3 standardized individual ELISAs anda screening assay, which incorporated the same antigens (3 mito-chondrial and 2 nuclear ones), to test a large group of PBC patientsfrom diverse geographic locations into a single study. Furthermore,the participation of many international centers allowed us to assessthe utility of these assays on well-documented, unequivocal PBCpatients who had been classified as “AMA-negative”.

Our results showed that the PBC Screen had an overall sensi-tivity of 83.8% on a cohort of PBC patients of which 21%were knownto be AMA negative by IIF. Note that such a large proportion ofAMA-negative patients reflects our case-finding strategy, not theexpected prevalence of these cases in PBC population [3,4]. The PBCScreen had a specificity of 94.7% based on testing a group of 1232disease and healthy controls. Although the PBC Screen determinedboth IgG and IgA isotypes to an antigenic mixture of mitochondrialand nuclear antigens, the test offered equivalent diagnostic reli-ability in our study compared to the combined results obtainedwith ELISAs based on individual autoantigens. The PBC Screen hada superior performance for the diagnosis of PBC in patients withoutdetectable AMA by IIF, with almost half of such patients identifiedby thismethod. Although theMIT3-based ELISAwas able to identify28.1% of the PBC patients with AMA detectable by IIF, the additionalantigens contained in the PBC Screen permitted detection of

Table 7Frequencies of the PBC Screen results in different laboratories.

Labs. PBC AMAþ by IIF PBC AMA� by IIF

No. positive % No. positive %

Gershwin, USA 55 53 96.4 e e e

Heathcote, Canada 251 250 98.8 67 19 28.4Bogdanos/Vergani/Dalekos, UK/Greece 74 65 87.8 58 32 55.2Worman, USA 52 47 90.4 e e e

Krawitt, USA 12 10 83.3 e e e

Bizzaro, Italy 153 150 98.0 90 40 44.4Invernizzi, Italy 178 154 86.5 16 10 62.5Miyachi/Tan, Japan 138 136 98.6 22 12 54.5Total 922 872 94.6 253 113 44.7

patients who were only positive for gp210 and/or sp100, or the fewspecimens that only had PBC-specific IgA antibodies.

Since the present study expanded the dataset used to originallyset of the PBC Screen cutoff, we recalculated the cutoff based on thenew data set of 2407 PBC patients and controls. A new optimal cutoff(27.8 units) was suggested by ROC analysis of the PBC Screen multi-center study data, with an area under the curve greater than 0.9indicating a superior test performance. Although ROC analysis vali-dated the cutoff recommended by manufacturer (25.0 units) basedon the phase I specimens, the higher specificity and PLR obtained atthe new cutoff (27.8 units) may be more useful for diagnosis of PBC.

The higher sensitivity of the PBC Screen over IIF and conven-tional M2 assays is expected since it detects antibodies to gp210,sp100 and AMA utilizing the enhanced sensitivity to MIT3 antigen.The assay therefore significantly reduces the number of PBCpatients without serological evidence of PBC. In addition, this newELISA can detect MIT3, gp210, and sp100 antibodies of not only IgG,but also the IgA isotype. IgA AMA may have a specific pathogeneticrole in PBC, because secretory IgA is the predominant immuno-globulin of the mucosal immune system and the major proteinin bile [25,26]. The clinical significance of IgA AMA, however, isuncertain. Although less frequent than IgG AMA, IgA AMA canoccasionally be the only isotype detected in PBC patients and can bepresent at high serum concentrations [27]. In addition, IgA AMA ismore frequently observed in the early stage of PBC [27,28].

The specificity of the PBC Screen has not been sacrificed fordetection of dual isotypes of antibodies to multiple antigens. In 65of 1232 controls positive on the PBC Screen, over 55% (36/65) werefound to be specimens from patients with AIH and 14 of these 36screen-positive AIH patients were strongly positive. The reportedfrequency of AMA detection in patients with AIH without thesimultaneous presence of PBC ranges from 5% to 34% [27,29e32].The frequency in our study was 12.8% (36/281). The clinical impactof the presence of AMA in AIH patients is uncertain since studieshave reported that the clinical features and histological findingsof AIH are not influenced by the AMA status over many years offollow-up [31,33]. We can provide 3 possible explanations for thisphenomenon: 1) AIH with AMA is a unique AIH/PBC “overlapsyndrome”, 2) early treatment of subjects with AIH prevents thedevelopment of subsequent PBC, or perhaps most likely 3) theseAIH patients with detectable AMAwill develop PBC or actually havePBC. Note that the AIH patients without a definite diagnosis of PBCat inclusion into the study were grouped into the non-PBC controlgroup. This means that a positive PBC Screen result on thesesubjects had been counted as a “false positive” and contributed toa decrease in assay specificity, despite the fact that the serologicalresult may actually be correct. The specificity of the PBC Screenmaytherefore have been underestimated, as individuals with positiveresults despite not carrying a diagnosis of PBC may actually haveundiagnosed or evolving PBC.

At present, there is no “gold standard” assay with 100%sensitivity and specificity for the diagnosis of PBC [34e37]. AMAimmunoblotting has been reported to have almost 100% sensitivityand can detect individual reactivity of 2-OADC enzymes [38];however, this assay is labor intensive. An enzyme inhibition assayhas been reported to have almost 100% specificity, but it has lowersensitivity compared to IIF [39]. Our results indicate that the PBCScreen, which is an ELISA based on an antigenic mixture of themajormitochondrial and nuclear antigenic targets, performs aswellas ELISAs based on individual antigens. Out of the 2407 specimenstested, only 35 specimens (1.5%) were discordant between the PBCScreen and the individual ELISAs. The combined use of the antigensMIT3, sp100, and gp210 resulted in detection of PBC patients whomight have beenmissed or fallen into the clinically uncertain groupof “AMA-negative PBC.” The PBC Screen may therefore aid in

H. Liu et al. / Journal of Autoimmunity 35 (2010) 436e442442

accelerating their diagnosis and institution of treatment. In addi-tion, the PBC Screen appears to be more rapid and cost-effectivethan IIF, which is time-consuming, labor-intensive, subjective, andrequires highly trained personnel for interpretation.

Follow-up of specimens positive on the PBC Screenwith IIF maybe useful if it confirms the presence of a PBC-specific IIF pattern.A negative result, however, may not offer diagnostic clarificationsince it may be due to the lower sensitivity of IIF or the true lack ofPBC-specific antibodies. Positive PBC Screen results can be furtherfollowed-up with specific ELISA testing for MIT3, gp210, and sp100antibodies. Determination of the antibody specificity responsiblefor the positive screen test may become increasingly important asthe clinical significance of specific PBC-specific antibodies is moreclearly established [3,4]. For example, the presence of gp210 anti-bodies in patients with PBC is known to be associated with moresevere liver disease and a poor prognosis [16,17,40].

In conclusion, our findings suggest that the PBC Screen can beused as first-line assay for diagnosing PBC. It can also be useful forroutine clinical laboratory testing because of its semi-automatedand non-subjective readout. The PBC Screen provides a multi-analyte, dual isotype assay for the detection of serological markersof PBC. This assay should help in the diagnosis of PBC patients,especially those negative for serological markers when usingconventional methods such as IIF.

Acknowledgements

DPB was supported by a CSL award from the Higher EducationFunding Council for England.

References

[1] Kaplan MM, Gershwin ME. Primary biliary cirrhosis. N Engl J Med 2005;353:1261e73.

[2] Invernizzi P, Selmi C, Ranftler C, Podda M, Wesierska-Gadek J. Antinuclearantibodies in primary biliary cirrhosis. Semin Liver Dis 2005;25:298e310.

[3] Invernizzi P, Lleo A, Podda M. Interpreting serological tests in diagnosingautoimmune liver diseases. Semin Liver Dis 2007;27:161e72.

[4] Bogdanos DP, Invernizzi P, Mackay IR, Vergani D. Autoimmune liver serology:current diagnostic and clinical challenges. World J Gastroenterol 2008;14:3374e87.

[5] EASL clinical practice guidelines: management of cholestatic liver diseases.J Hepatol 2009;51:237e67.

[6] Lindor KD, Gershwin ME, Poupon R, Kaplan M, Bergasa NV, Heathcote EJ.Primary biliary cirrhosis. Hepatology 2009;50:291e308.

[7] Lleo A, Invernizzi P, Gao B, Podda M, Gershwin ME. Definition of humanautoimmunityeautoantibodies versus autoimmune disease. Autoimmun Rev2010;9:A259e66.

[8] Moteki S, Leung PS, Coppel RL, Dickson ER, Kaplan MM, Munoz S, et al. Use ofa designer triple expression hybrid clone for three different lipoyl domain for thedetection of antimitochondrial autoantibodies. Hepatology 1996;24:97e103.

[9] Oertelt S, Rieger R, Selmi C, Invernizzi P, Ansari AA, Coppel RL, et al. A sensitivebead assay for antimitochondrial antibodies: chipping away at AMA-negativeprimary biliary cirrhosis. Hepatology 2007;45:659e65.

[10] Dahnrich C, Pares A, Caballeria L, Rosemann A, Schlumberger W, Probst C,et al. New ELISA for detecting primary biliary cirrhosis-specific anti-mitochondrial antibodies. Clin Chem 2009;55:978e85.

[11] Courvalin JC, Lassoued K, Bartnik E, Blobel G, Wozniak RW. The 210-kDnuclear envelope polypeptide recognized by human autoantibodies inprimary biliary cirrhosis is the major glycoprotein of the nuclear pore. J ClinInvest 1990;86:279e85.

[12] Szostecki C, Guldner HH, Netter HJ, Will H. Isolation and characterization ofcDNA encoding a human nuclear antigen predominantly recognized byautoantibodies from patients with primary biliary cirrhosis. J Immunol1990;145:4338e47.

[13] Rigopoulou EI, Davies ET, Pares A, Zachou K, Liaskos C, Bogdanos DP, et al.Prevalence and clinical significance of isotype specific antinuclear antibodiesin primary biliary cirrhosis. Gut 2005;54:528e32.

[14] Tartakovsky F, Worman HJ. Detection of Gp210 autoantibodies in primarybiliary cirrhosis using a recombinant protein containing the predominantautoepitope. Hepatology 1995;21:495e500.

[15] Bandin O, Courvalin JC, Poupon R, Dubel L, Homberg JC, Johanet C. Specificityand sensitivity of gp210 autoantibodies detected using an enzyme-linked

immunosorbent assay and a synthetic polypeptide in the diagnosis of primarybiliary cirrhosis. Hepatology 1996;23:1020e4.

[16] Wesierska-Gadek J, Penner E, Battezzati PM, Selmi C, Zuin M, Hitchman E,et al. Correlation of initial autoantibody profile and clinical outcome inprimary biliary cirrhosis. Hepatology 2006;43:1135e44.

[17] Nakamura M, Kondo H, Mori T, Komori A, Matsuyama M, Ito M, et al. Anti-gp210 and anti-centromere antibodies are different risk factors for theprogression of primary biliary cirrhosis. Hepatology 2007;45:118e27.

[18] Ludwig J, Dickson ER, McDonald GS. Staging of chronic nonsuppurativedestructive cholangitis (syndrome of primary biliary cirrhosis). Virchows ArchA Pathol Anat Histol 1978;379:103e12.

[19] Krawitt EL. Autoimmune hepatitis. N Engl J Med 2006;354:54e66.[20] Lee YM, Kaplan MM. Management of primary sclerosing cholangitis. Am

J Gastroenterol 2002;97:528e34.[21] Harada K, Sudo Y, Kono N, Ozaki S, Tsuneyama K, Gershwin ME, et al. In situ

nucleic acid detection of PDC-E2, BCOADC-E2, OGDC-E2, PDC-E1alpha,BCOADC-E1alpha, OGDC-E1, and the E3 binding protein (protein X) inprimary biliary cirrhosis. Hepatology 1999;30:36e45.

[22] Szostecki C, Will H, Netter HJ, Guldner HH. Autoantibodies to the nuclearSp100 protein in primary biliary cirrhosis and associated diseases: epitopespecificity and immunoglobulin class distribution. Scand J Immunol1992;36:555e64.

[23] Bluthner M, Schafer C, Schneider C, Bautz FA. Identification of major linearepitopes on the sp100 nuclear PBC autoantigen by the gene-fragment phage-display technology. Autoimmunity 1999;29:33e42.

[24] Nickowitz RE, Worman HJ. Autoantibodies from patients with primary biliarycirrhosis recognize a restricted region within the cytoplasmic tail of nuclearpore membrane glycoprotein Gp210. J Exp Med 1993;178:2237e42.

[25] Nishio A, Van de Water J, Leung PS, Joplin R, Neuberger JM, Lake J, et al.Comparative studies of antimitochondrial autoantibodies in sera and bile inprimary biliary cirrhosis. Hepatology 1997;25:1085e9.

[26] Malmborg AC, Shultz DB, Luton F, Mostov KE, Richly E, Leung PS, et al.Penetration and co-localization in MDCK cell mitochondria of IgA derivedfrom patients with primary biliary cirrhosis. J Autoimmun 1998;11:573e80.

[27] Gabeta S, NormanGL, Liaskos C, Papamichalis PA, Zografos T, Garagounis A, et al.Diagnostic relevance and clinical significance of the new enhanced performanceM2 (MIT3) ELISA for the detection of IgA and IgGantimitochondrial antibodies inprimary biliary cirrhosis. J Clin Immunol 2007;27:378e87.

[28] Nakajima M, Shimizu H, Miyazaki A, Watanabe S, Kitami N, Sato N. Detectionof IgA, IgM, and IgG subclasses of anti-M2 antibody by immunoblotting inautoimmune cholangitis: is autoimmune cholangitis an early stage of primarybiliary cirrhosis? J Gastroenterol 1999;34:607e12.

[29] Liaskos C, Bogdanos DP, Rigopoulou EI, Dalekos GN. Development of anti-mitochondrial antibodies in patients with autoimmune hepatitis: art of factsor an artifact? J Gastroenterol Hepatol 2007;22:454e5.

[30] Alvarez F, Berg PA, Bianchi FB, Bianchi L, Burroughs AK, Cancado EL, et al.International autoimmune hepatitis group report: review of criteria fordiagnosis of autoimmune hepatitis. J Hepatol 1999;31:929e38.

[31] Nezu S, Tanaka A, Yasui H, Imamura M, Nakajima H, Ishida H, et al. Presence ofantimitochondrial autoantibodies in patients with autoimmune hepatitis.J Gastroenterol Hepatol 2006;21:1448e54.

[32] Vergani D, Alvarez F, Bianchi FB, Cancado EL, Mackay IR, Manns MP, et al. Liverautoimmune serology: a consensus statement from the committee for auto-immune serology of the International Autoimmune Hepatitis Group. J Hepatol2004;41:677e83.

[33] O’Brien C, Joshi S, Feld JJ, Guindi M, Dienes HP, Heathcote EJ. Long-termfollow-up of antimitochondrial antibody-positive autoimmune hepatitis.Hepatology 2008;48:550e6.

[34] Kadokawa Y, Omagari K, Hazama H, Ohba K, Masuda J, Kinoshita H, et al.Evaluation of newly developed ELISA using "MESACUP-2 test mitochondrialM2" kit for the diagnosis of primary biliary cirrhosis. Clin Biochem 2003;36:203e10.

[35] Miyakawa H, Tanaka A, Kikuchi K, Matsushita M, Kitazawa E, Kawaguchi N,et al. Detection of antimitochondrial autoantibodies in immunofluorescentAMA-negative patients with primary biliary cirrhosis using recombinantautoantigens. Hepatology 2001;34:243e8.

[36] Muratori P, Muratori L, Gershwin ME, Czaja AJ, Pappas G, MacCariello S, et al.‘True’ antimitochondrial antibody-negative primary biliary cirrhosis, lowsensitivity of the routine assays, or both? Clin Exp Immunol 2004;135:154e8.

[37] Tanaka A, Miyakawa H, Luketic VA, Kaplan M, Storch WB, Gershwin ME. Thediagnostic value of anti-mitochondrial antibodies, especially in primarybiliary cirrhosis. Cell Mol Biol (Noisy-le-grand) 2002;48:295e9.

[38] Kitami N, Komada T, Ishii H, Shimizu H, Adachi H, Yamaguchi Y, et al.Immunological study of anti-M2 in antimitochondrial antibody-negativeprimary biliary cirrhosis. Intern Med 1995;34:496e501.

[39] Hazama H, Omagari K, Masuda J, Ohba K, Kinoshita H, Matsuo I, et al. Auto-mated enzymatic mitochondrial antibody assay for the diagnosis of primarybiliary cirrhosis: applications of a routine diagnostic tool for the detection ofantimitochondrial antibodies. J Gastroenterol Hepatol 2002;17:316e23.

[40] Invernizzi P, Podda M, Battezzati PM, Crosignani A, Zuin M, Hitchman E, et al.Autoantibodies against nuclear pore complexes are associated with moreactive and severe liver disease in primary biliary cirrhosis. J Hepatol 2001;34:366e72.