Fine phenotypic and functional characterization of effector cluster of differentiation 8 positive T cells in human patients with primary biliary cirrhosis.

Abstract

In primary biliary cirrhosis (PBC), patients develop a multilineage response to a highly restricted peptide of the E2 component of pyruvate dehydrogenase (PDC-E2) involving autoantibody and autoreactive cluster of differentiation (CD)4(+) and CD8(+) T-cell responses. Recent data from murine models have suggested that liver-infiltrating CD8(+) cells play a critical role in biliary destruction in PBC. We hypothesized that chronic antigen stimulation of CD8(+) T cells alters effector memory T cell (T(EM) ) frequency and function similar to that seen with chronic viral infections, including failure to terminally differentiate and relative resistance to apoptosis. We have rigorously phenotyped CD8(+) T-cell subpopulations from 132 subjects, including 76 patients with PBC and 56 controls, and report a higher frequency of T(EM) cells characterized as CD45RO(high) CD57(+) CD8(high), but expressing the gut homing integrin, α4β7, in peripheral blood mononuclear cells of PBC. These CD8(high) T(EM) cells have reduced expression of Annexin V after TCR stimulation. Consistent with a T(EM) phenotype, CD45RO(high) CD57(+) CD8(high) T cells express higher levels of granzyme A, granzyme B, perforin, CCR5 and α4β7, and lower levels of CCR7 and CD28 than other CD8(high) T cells. Furthermore, interleukin (IL)-5 produced by CD8(+) CD57(+) T lymphocytes upon in vitro T-cell receptor stimulation are increased in PBC. Histologically, CD8(+) CD57(+) T cells accumulate around the portal area in PBC. Moreover, CD8(+) CD57(+) T cells respond specifically to the major histocompatibility class I epitope of PDC-E2. CONCLUSION: In conclusion, our data demonstrate that CD45RO(high) CD57(+) CD8(high) T cells are a subset of terminally differentiated cytotoxic T(EM) cells, which could play a critical role in the progressive destruction of biliary epithelial cells.

Full text

Page 1

Fine Phenotypic and Functional Characterization of
Effector Cluster of Differentiation 8 Positive T Cells in
Human Patients With Primary Biliary Cirrhosis
Masanobu Tsuda,1,2Yoko M. Ambrosini,1Weici Zhang,1Guo-Xiang Yang,1Yugo Ando,1Guanghua Rong,1
Koichi Tsuneyama,1,3Kosuke Sumida,4Shinji Shimoda,4Christopher L. Bowlus,5Patrick S.C. Leung,1
Xiao-Song He,1Ross L. Coppel,6Aftab A. Ansari,7Zhe-Xiong Lian,1,8and M. Eric Gershwin1
In primary biliary cirrhosis (PBC), patients develop a multilineage response to a highly
restricted peptide of the E2 component of pyruvate dehydrogenase (PDC-E2) involving
autoantibody and autoreactive cluster of differentiation (CD)41and CD81T-cell responses.
Recent data from murine models have suggested that liver-infiltrating CD81cells play a crit-
ical role in biliary destruction in PBC. We hypothesized that chronic antigen stimulation of
CD81T cells alters effector memory T cell (TEM) frequency and function similar to that
seen with chronic viral infections, including failure to terminally differentiate and relative
resistance to apoptosis. We have rigorously phenotyped CD81T-cell subpopulations from
132 subjects, including 76 patients with PBC and 56 controls, and report a higher frequency
of TEMcells characterized as CD45ROhighCD571CD8high, but expressing the gut homing
integrin, a4b7, in peripheral blood mononuclear cells of PBC. These CD8highTEMcells
have reduced expression of Annexin Vafter TCR stimulation. Consistent with a TEMpheno-
type, CD45ROhighCD571CD8highT cells express higher levels of granzyme A, granzyme B,
perforin, CCR5 and a4b7, and lower levels of CCR7 and CD28 than other CD8highT cells.
Furthermore, interleukin (IL)-5 produced by CD81CD571T lymphocytes upon in vitro
T-cell receptor stimulation are increased in PBC. Histologically, CD81CD571T cells accu-
mulate around the portal area in PBC. Moreover, CD81CD571T cells respond specifically
to the major histocompatibility class I epitope of PDC-E2. Conclusion: In conclusion, our
data demonstrate that CD45ROhighCD571CD8highT cells are a subset of terminally differ-
entiated cytotoxic TEMcells, which could play a critical role in the progressive destruction of
biliary epithelial cells. (HEPATOLOGY 2011;54:1293-1302)
P
bile duct biliary epithelial cells (BECs).1The serologi-
rimary biliary cirrhosis (PBC) is a female-pre-
dominant, organ-specific autoimmune disease
characterized by destruction of intrahepatic small
cal hallmark of PBC is the presence of antimitochon-
drial autoantibodies (AMAs) directed against the pyru-
vate dehydrogenase E2 complex (PDC-E2) located in
theinnermembraneofmitochondria.2-4
High
Abbreviations: AMA, antimitochondrial antibodies; APC, allophycocyanin; BEC, bile duct biliary epithelial cell; BSA, bovine serum albumin; CD, cluster of
differentiation; CVH, chronic viral hepatitis; EDTA, ethylenediaminetetraacetic acid; ELC, EBI1-ligand chemokine; ELISA, enzyme-linked immunosorbent assay;
FcR, Fc receptor; FBS, fetal bovine serum; FITC, fluorescein isothiocyanate; HLA, human leukocyte antigen; HNK-1, human natural killer-1; HPLC, high-
performance liquid chromatography; IFN-c, interferon gamma; IP-10, IFN-c-inducible protein 10; IL, interleukin; IgG, immunoglobulin G; LGL, large granular
lymphocyte leukemia; mAb, monoclonal antibody; MAdCAM-1, mucosal addressin cell adhesion molecule-1; MFI, mean fluorescent intensity; NASH, nonalcoholic
steatohepatitis; NK, natural killer; PBC, primary biliary cirrhosis; PBMCs, peripheral blood mononuclear cells; PBS, phosphate-buffered saline; PD-1, programmed
death 1; PDC-E2, E2 component of pyruvate dehydrogenase; PE, phycoerythrin; RPMI, Roswell Park Memorial Institute; SEM, standard error of the mean; SLC,
secondary lymphoid tissue chemokine; TCR, T-cell receptor; TECK, thymus-expressed chemokine; TEM, effector memory T cell; thymus-expressed chemokine CCL25,
CC chemokine ligand 25; TIM-3, T-cell immunoglobulin mucin-3.
From the1Division of Rheumatology, Allergy and Clinical Immunology, University of California at Davis, Davis, CA;2Department of Emergency and Critical Care
Medicine, Kansai Medical University, Osaka, Japan;3Department of Diagnostic Pathology, Graduate School of Medicine and Pharmaceutical Science for Research,
University of Toyama, Toyama 930-0194, Japan;4Medicine and Biosystemic Science, Kyushu University Graduate School of Medical Sciences, Fukuoka 812-8582,
Japan;5Division of Gastroenterology and Hepatology, University of California at Davis, Sacramento, CA;6Department of Microbiology, Monash University, Clayton,
Australia;7Department of Pathology, Emory University School of Medicine, Atlanta, GA; and8Institute of Immunology and School of Life Sciences, University of Science
and Technology of China, Hefei 230027, China.
Received January 31, 2011; accepted June 18, 2011.
Financial support for this work was provided by the National Institutes of Health (grant DK39588; Bethesda, MD).
1293

Page 2

frequency of cluster of differentiation (CD)4þand
CD8þT-cell infiltrates have been noted within the
portal tracts of the PBC liver, which strongly suggests
that these cells are involved in the pathogenesis of
PBC.5Indeed, PDC-E2-specific autoreactive CD4þT
and CD8þT cells have been identified both in periph-
eral blood and, at much higher levels, in the liver of
PBC patients.6-8The dominant CD4þand CD8þ
T-cell epitopes on PDC-E2 have been mapped.6-8
Although both CD4þand CD8þT cells are present
within portal tract infiltrates, there is a growing body
of data that suggests a more direct role of cytotoxic
CD8þT cells in biliary destruction.9-11The study of
effector pathways is a particularly challenging problem
in human autoimmunity. For one thing, the majority
of effector pathways are likely to be mediated by non-
specific bystander cells recruited during inflammation.
For another, it has been difficult to identify subpopula-
tions of cells by phenotype and thence link such data
to functionality. Our laboratory has focused attention
on effector T-cell populations, using a variety of tech-
nologies, and has highlighted the important role of T
cells in this and similar pathways.
In this study, we took advantage of newer reagents,
including cell-surface markers, that are associated with
CD8higheffector memory T cells (TEM), organ- and tis-
sue-specific homing, and alterations in susceptibility to
apoptosis. Indeed, we report that patients with PBC not
only haveanincreased
highCD57þCD8highT cells, compared to controls, but
also that such cells have increased a4b7 expression with
concurrentlyincreasedexpression
decreased expression of CCR7 and CD28, compared to
other CD8highT cells. Furthermore, this T-cell subset
has increased the production of granzyme A, granzyme
B, and perforin, compared with other CD8highT cells,
and, interestingly, have decreased stimulation-induced
apoptosis. Furthermore, interferon gamma (IFN-c) and
interleukin (IL)-5 produced by CD8þCD57þT lym-
phocytes upon in vitro T-cell-receptor (TCR) stimulation
areincreased in PBC
CD8þCD57þT cells accumulate around the portal area
in the liver of PBC patients. Moreover, purified
CD8þCD57þT cells from PBC patients specifically
respond to the major histocompatibility class I restricted
epitope of PDC-E2. These data have implications for
frequencyof CD45RO-
of CCR5and
patients.Histologically,
understanding CD8 effector pathways in this autoim-
mune disease.We submit
highCD57þCD8highT cells are a subset of cytotoxic
memory cells, which play a critical role in the chronic,
progressive destruction of BECs in PBC.
that CD45RO-
Materials and Methods
Subjects. Heparinized (Vacutainer; BD Biosciences,
Franklin Lakes, NJ) peripheral blood samples were
obtained from 76 PBC patients (59.0 6 1.0 years; mean
6 standard error of the mean [SEM]) and 56 age-
matched healthy controls (54.8 6 1.5 years). The diag-
nosis of PBC was based on internationally accepted crite-
ria.12Stage of disease was established according to Lud-
wig et al..13In the present study, 50 of 76 (65.8%)
patients with PBC were stage I or II and 22 of 76
(28.9%) were III or IV, whereas 5 of 76 (6.6%) patients
were AMA negative (Table 1). We did not observe any
difference between AMA-positive and -negative patients;
hence, the data are combined herein. The study was
approved by the Institutional Review Board of the Uni-
versity of California at Davis (Davis, CA), and all subjects
provided written, informed consent prior to enrollment.
Peripheral Blood Mononuclear Cell Isolation. Per-
ipheral blood mononuclear cells (PBMCs) from all
subjects were isolated by density gradient using Histo-
paque-1077 (Sigma Chemical Co., St. Louis, MO)
under endotoxin-free conditions. PBMCs were resus-
pended in phosphate-buffered saline (PBS) (Mediatech
Inc., Herndon, VA), containing 0.5% bovine serum al-
bumin (BSA) (Fraction V, OmniPur; EMD Chemicals
Inc., Gibbstown, NJ) and 0.05% ethylenediaminetetra-
acetic acid (EDTA). The viability of cells was >98%,
which was confirmed using trypan blue dye exclusion.
Evaluation of Cell Phenotypes. The polychromatic
phenotypic analysis of PBMCs was carried out on a
FACScan flow cytometer (BD Immunocytometry Sys-
tems, San Jose, CA) upgraded for the detection of five
colors by Cytek Development (Fremont, CA). Cells
were stained with different combinations of fluoro-
chrome-conjugated monoclonal antibodies (mAbs),
including CCR5, CD8b, CCR7, and CD45RO (BD
Pharmingen, San Diego, CA), CCR9 (R&D Systems,
Minneapolis, MN), CD56, CXCR3, CD57, CD8a,
CD45RO, CD28, and CD16 (BioLegend, San Diego,
Address reprint requests to: M. Eric Gershwin, M.D., Division of Rheumatology, Allergy and Clinical Immunology, University of California at Davis School of
Medicine, 451 Health Sciences Drive, Suite 6510, Davis, CA 95616. E-mail: megershwin@ucdavis.edu; fax: 530-752-4669.
Copyright V
View this article online at wileyonlinelibrary.com.
DOI 10.1002/hep.24526
Potential conflict of interest: Nothing to report.
C 2011 by the American Association for the Study of Liver Diseases.
1294TSUDA ET AL.HEPATOLOGY, October 2011

Page 3

CA), and CCR7 (eBioscience, San Diego, CA). The
allophycocyanin (APC)-conjugated anti-a4b7 was pro-
duced in our laboratory. Immunoglobulin G (IgG) iso-
type controls with matching conjugates for each anti-
body were used as negative controls. PBMCs were
resuspended in staining buffer (0.2% BSA, 0.04%
EDTA, and 0.05% sodium azide in PBS), divided into
25-lL aliquots, and incubated with antihuman Fc re-
ceptor (FcR) blocking reagent (eBioscience) for 15
minutes at 4?C. Cells were then washed and stained
with the antibody cocktails for 30 minutes at 4?C.
Cells were washed once with PBS containing 0.2%
BSA. For intracellular staining, cells were first stained
withphycoerythrin (PE)-anti-CD57
PerCP-anti-CD8a (BioLegend), APCe780-anti-CCR7
(eBioscience), and APC-anti-CD45RO (BioLegend),
then fixed and permeabilized with BD Cytofix/Cyto-
perm solution (BD Pharmingen) for 15 minutes at
4?C. Subsequently, intracellular staining was performed
with AF488-labeledantigranzyme
AF488-labeled anti-Perforin (BioLegend), or fluorescein
isothiocyanate (FITC)-labeled antigranzyme B (BD
Pharmingen) or IgG isotype controls. After staining,
cells were washed and fixed with 1% paraformaldehyde
in PBS. Acquired data were analyzed with Cellquest
PRO software (BD Immunocytometry Systems).
Isolationand Culturing
Cells. CD8þT cells were isolated with RosetteSepTM
human CD8þT cell enrichment cocktails (StemCell
Technologies, Vancouver, British Columbia, Canada),
following the manufacturer’s instructions, then resus-
pended in PBS containing 2% fetal bovine serum
(FBS). The CD57þCD8þT-cell subset was isolated
from the enriched CD8þT cells using human CD57
MicroBeads (Miltenyi Biotec, Auburn, CA). An aliquot
of the isolated CD57þpopulation was analyzed for
purity with flow cytometry, which was always >92%.
Aliquots of CD57þCD8þT cells (2 ? 105) were cul-
tured in 96-well, round-bottomed plates in 200 lL of
Roswell Park Memorial Institute (RPMI) medium with
10% heat-inactivated FBS (Gibco-Invitrogen Corp.,
Grand Island, NY), 100 lg/mL of streptomycin, 100
(BioLegend),
A (BioLegend),
ofCD571CD81
T
U/mL of penicillin, and 0.5 lg/mL each of anti-CD3
(BioLegend) and anti-CD28 (BioLegend). Cells were
incubated for 5 days at 37?C in a humidified 5% CO2
incubator, then centrifuged. The supernatant was col-
lected for cytokine analysis.
CytokineDetection. Supernatant
CD57þCD8þT cells was analyzed with enzyme-
linked immunosorbent assay (ELISA) kits for IFN-c
(R&D Systems), granzyme A (Bender MedSystems,
Vienna, Austria), and IL-5 (BioLegend).
Apoptosis. To assess the relative susceptibility of in
vitro stimulated CD45ROhighCD57þCD8highT cells
to apoptosis, 1 ? 106PBMCs were cultured in 48-
well, flat-bottomed plates in 1 mL of RPMI 1640
(Gibco-Invitrogen Corp.), supplemented with 10%
heat-inactivated FBS, 100 lg/mL of streptomycin, 100
U/mL of penicillin, 5 lg/mL of anti-CD3 (Bio-
Legend), and 5 lg/mL of anti-CD28 (BioLegend).
Cultures were incubated at 37?C in 5% CO2. After
48 hours of culturing, cells were washed twice with
0.2% BSA in PBS, and the frequency of cells under-
going apoptosis was determined with flow cytometry,
using FITC-conjugated anti-Annexin-V (BD Pharmin-
gen), following the manufacturer’s instructions. Cells
were also stained with Fas (CD95), programmed death
1 (PD-1) (Biolegend), and T-cell immunoglobulin
mucin-3 (TIM-3) (eBioscience) after culture.
Synthetic Peptide Assay. PBMC from a nested
study, including 3 patients with PBC (PBC 1-3) who
were human leukocyte antigen (HLA) A2.1 and 6
other patients with PBC (PBC 4-9) who had other
class I alleles, were isolated. As controls, PBMCs from
4 healthy HLA A2.1 controls and 5 HLA A2.1 nega-
tive were collected. The peptide, 159-167 of PDC-E2
(KLSEGDLLA), was synthesized by F-moc chemistry
(Model Synergy; Applied Biosystems Inc., Foster City,
CA). This peptide was purified by reverse-phase high-
performance liquid chromatography (HPLC), and the
purity was more than 80% as determined by HPLC
analysis. Aliquots ofCD57þCD8þ
CD57?CD8þT cells (2 ? 105) were cultured in the
presence of autologous irradiated (3,000 rad) APC (2
? 105), with or without the 159-167 synthetic pep-
tide, for 5 days. The supernatant was collected for
cytokine analysis. Controls were used throughout all
assays. Supernatant from the cultured cells was ana-
lyzed by ELISA for IFN-c (R&D Systems).
Immunohistochemistry. Liver sections were immu-
nostained using our standard microwave protocol, as
previously described.14,15All tissues were fixed in 10%
neutral buffered formalin and embedded in paraffin,
then 4-lm-thick sections were cut from each paraffin
fromcultured
Tcells or
Table 1. Clinical Characteristics of PBC Patients*
PBC (n 5 76)AMA PositiveAMA Negative Age: 59.0 6 1.0
Stage I
Stage II
Stage III
Stage IV
Unknown
n ¼ 32
n ¼ 15
n ¼ 9
n ¼ 11
n ¼ 4
n ¼ 1
n ¼ 2
n ¼ 1
n ¼ 1
Early stage (65.8%)
Late stage (28.9%)
*There were 56 controls (age, 54.8 6 1.5 years).
Abbreviations: PBC, primary biliary cirrhosis; AMA, antimitochondrial antibodies.
HEPATOLOGY, Vol. 54, No. 4, 2011TSUDA ET AL.1295

Page 4

block from 4 patients with PBC, 3 with chronic viral
hepatitis (CVH), and 1 with nonalcoholic steatohepati-
tis (NASH). The following antibodies were used for the
detection of CD8 and CD57 in human liver specimens:
rabbit polyclonal antibody against CD57 (Novus Bio-
logicals, Littleton, CO), antihuman CD8 antibody
(mAb; DAKO, Glostrup, Denmark) and Envision-
peroxidase (DAKO). In all samples, predetermined opti-
mal dilutions were used, and positive and negative sam-
ples were included with each assay, and the data were
interpreted by a ‘‘blinded’’ pathologist (K.T.).
Statistical Analysis. The percentages of CD8high
T-cell subsets that express individual cell markers in
PBC patients and healthy controls were expressed as
mean 6 SEM and compared with the two-tailed
Mann-Whitney U test. A P value <0.05 was consid-
ered statisticallysignificant.
CD45ROhighCD57þCD8highT-cell subset and other
CD8highT-cell subsets were analyzed using a two-tailed
Wilcoxon matched-pairs test.
Percentages ofthe
Results
Increased Frequency of CD45ROhighCD571CD8high
T Cells in Peripheral Blood of PBC Patients. There
were no significant differences observed in the mean
frequencies of CD8highT cells in the PBMC and
CD57þcells in CD8highT cells of PBC patients, com-
pared with control subjects (Fig. 1). However, the fre-
quency of CD45ROhighCD57þcells were significantly
higher in CD8highT cells of PBC patients (7.15% 6
0.77%), in particular patients at earlier disease stages
(8.25% 6 1.16%), compared with healthy controls
(4.10% 6 0.37%; P < 0.0005) (Fig. 2). The
CD45ROhighCD57þCD8high
include natural killer (NK) cells, as the vast majority
of these cells were CD3þ(99.9% 6 0.14%) and
CD16?(98.69% 6 0.69%) (data not shown).
Increased Expression of a4b7highand Decreased
CD28 on CD45ROhighCD571CD8highT Cells in
PBC Patient. To further characterize the CD45RO-
highCD57þCD8highT-cell subset, these cells were ana-
lyzed for their expression of a panel of phenotypic
markers,including multiple chemokine
a4b7 integrin, and the costimulatory molecule, CD28.
Results of these studies (Fig. 3) demonstrate that
whereasthe frequency
highCD57þCD8highsubset expressing the gut homing
a4b7highintegrin was significantly higher in PBC
patients (18.51% 6 1.94%) than that in controls
(11.69% 6 1.41%; P < 0.03) and decreased the
populationdid not
receptors,
oftheCD45RO-
Fig. 1. Analysis of the frequencies of CD8highCD57þcells in lym-
phocyte population from PBC patients and healthy subjects. Left pan-
els: representative dot plot of gated lymphocyte population. Right
panels: percentage of the gated CD8highin PBMC (upper panel) or
CD57þin CD8highlymphocytes (lower panel) populations in healthy
controls (cont), all PBC patients (PBC), and PBC patients at early
stages (I/II) and late stages (III/IV). Bars denote mean percentages.
Fig. 2. Analysis of the frequency of the CD57þand CD45ROhigh
cells in CD8highlymphocytes from PBC patients and healthy subjects.
Top panels: representative dot plots of CD8highgated lymphocyte pop-
ulation from a healthy control and a patient with PBC. Bottom panel:
comparison of the mean percentages of the gated CD45ROhighCD57þ
population in CD8highlymphocytes, among healthy controls (cont), all
PBC patients (PBC), and PBC patients at early stages (I/II) and late
stages (III/IV). Bars denote mean percentages. ***P < 0.001.
1296 TSUDA ET AL. HEPATOLOGY, October 2011

Page 5

expression of CD28 in early-stage PBC patients
(25.01% 6 5.81%) than controls (58.76% 6 9.36%;
P < 0.03), there were no significant differences in the
frequencies of the cells expressing the chemokine
receptors, CXCR3 (a receptor for IFN-c-inducible pro-
tein 10 [IP-10] and the monokine, Mig), CCR5 (a
receptor for RANTES and MIP-1a,b), and CCR7 (a
receptor for EBI1-ligand chemokine [ELC] and sec-
ondary lymphoid tissue chemokine [SLC]), compared
with controls. Of interest, there was no difference in
the frequencies of this CD8highsubset that expressed
the gut-homing chemokine receptor, CCR9, a receptor
for thymus-expressed chemokine (TECK)/CC chemo-
kine ligand 25 (thymus-expressed chemokine CCL25).
Altered Granzyme and Perforin Expression in
CD45ROhighCD571CD8high
patients, we have compared phenotypes of CD45RO-
highCD57þCD8highT cells to other CD8highT cells. In
addition to the analysis of homing and chemokine
receptors, we also studied the cytotoxic potential of the
CD45ROhighCD57þCD8highsubset of T cells. Interest-
ingly, though there was a significant increased expression
of CCR5 and a4b7high, and significantly decreased
expressionof CCR7and
highCD57þCD8highT cells, compared to other CD8high
T cells in PBC patients, there was no difference
observed in a4b7highand CD28 expressions in healthy
controls (Fig. 4A,C). CCR5, CCR7, granzyme A, gran-
zyme B, and perforin demonstrated a similar pheno-
typic pattern in PBC and healthy controls (data not
shown). In addition, relative to other CD8highT cells,
T Cells. In PBC
CD28 inCD45RO-
CD45ROhighCD57þCD8highT cells had increased pro-
duction of granzyme A (P < 0.001), granzyme B (P <
0.001), and perforin (P < 0.001), suggesting their
strong cytotoxic effector functions (Fig. 4B).
The CD45ROhighCD571CD8highT Cells Are Not
Susceptible to Apoptosis Upon Anti-CD3 Stim-
ulation. CD57þT cells have previously been demon-
strated to be susceptible to apoptosis during chronic anti-
genic stimulation.16
Therefore, we determined the
relative susceptibility of CD45ROhighCD57þCD8highT
cells from PBC patients to undergo apoptosis. PBMCs
from PBC patients and controls were stimulated by anti-
CD3/28 for 48 hours, then examined for Annexin V
expression by the CD45ROhighCD57þCD8highT cells.
Interestingly, Annexin V expression was decreased in
CD45ROhighCD57þCD8highT cells from PBC patients
(35.23% 6 3.07%), specifically those with early disease
stages (32.68% 6 4.02%), compared with healthy con-
trols (46.18% 6 1.51%; P < 0.03) (Fig. 5A). We further
investigated the expression of Fas, TIM-3, and PD-1, and
found that PD-1 expression was significantly decreased in
PBC patients (51.8% 6 6.54%), compared to that in
healthy controls (75.03% 6 7.12%; P < 0.02) (Fig. 5A).
This reduced susceptibility to stimulation-induced apo-
ptosis was not observed in other CD8highT cells, suggest-
ing a unique apoptosis resistance in the CD45RO-
highCD57þCD8highT cells from PBC patients.
The CD571CD81
T-Cell Subset From PBC
Patients Produce Increased Levels of IL-5, IFN-c,
and Granzyme A. To investigate the cytokine profile
of CD57þCD8þ
T cells, CD57þCD8þ
T cells
Fig. 3. Analysis of the expres-
sion ofCXCR3,
CCR9,CD28,
CD45ROhighCD57þCD8highgated T
lymphocytes from healthy controls
(cont), all PBC patients (PBC), and
PBC patients at early stages (I/II)
andlate stages
denote the mean percentages. *P
< 0.05.
CCR5,
and
CCR7,
a4b7 on
(III/IV). Bars
HEPATOLOGY, Vol. 54, No. 4, 2011 TSUDA ET AL.1297

Page 6

isolated from PBC patients and healthy controls were
stimulated in vitro with anti-CD3/28 for 5 days. Super-
natants were collected and analyzed for levels of secreted
IFN-c, IL-5, and granzyme A. As shown in Fig. 6,
CD57þCD8þT cells from PBC patients secreted
increased levels of IL-5 (157.4 6 47.7 pg/mL), com-
pared with controls (30.0 6 7.7 pg/mL; P < 0.001).
No significant difference was observed in the levels of
IFN-c and granzyme A.
CD81CD571Cells Were Infiltrated in Hepatic
Portal Track and Respond to the HLA-A2-Restricted
CTL Epitope, PDC-E2. CD57þcells coexisted in the
area of CD8þinfiltration (i.e., were CD8þCD57þ
double positive). In contrast, it was uncommon to
have CD57þcoexpression detected in CD8þ-infiltrat-
ing control livers (Fig. 7A). To further assess the preva-
lence of autoreactive T cells, we investigated the differ-
ence in response with the HLA-A2-restricted cytotoxic
T lymphocyte (CTL) epitope. CD8þCD57þcells from
HLA-A2.1-positive PBC patients (124.3 6 16.5 pg/ml)
had increased production of IFN-c upon PDC-E2 stim-
ulation, compared to CD8þCD57þcells from HLA-
A2.1-positive healthy controls (39.7 6 10.1 pg/ml); as
expected, there was no significant difference between
PBC non-HLA-A2.1 patients versus controls (Fig. 7B).
Discussion
In this study, we have carried out a comprehensive
phenotypic and functional characterization of CD8þT
cells, which is believed to be directly responsible for
thedestructionof BECs in PBC.Ourresults
Fig. 4. Comparison of phenotypes between CD45ROhighCD57þCD8highT cells and other CD8highT cells. The other CD8highT cells include
CD45RO? to lowCD57þCD8highand CD57?CD8highcells. Bars denote mean percentages. *P < 0.05; **P < 0.01; ***P < 0.001. (A) Cell-sur-
face markers in PBC. (B) Intracellular markers in PBC. (C) Cell-surface markers in healthy control.
1298 TSUDA ET AL. HEPATOLOGY, October 2011

Page 7

demonstrate the following: (1) PBC patients had an
increased frequency of CD45ROhighCD57þcells in
CD8highT cells, compared with age-matched healthy
controls; (2) CD45ROhighCD57þCD8high
from PBC patients more frequently expressed a4b7
and demonstrated reduced CD28 expression, com-
T cells
pared
CD45ROhighCD57þCD8highsubset had increased fre-
quency of CCR5þand a4b7highcells, decreased fre-
quency of CCR7þand CD28þcells, and expressed
increased levels of granzyme A, B, and perforin, in
comparison to other CD8highT cells, consistent with
an effector memory phenotype; (4) upon CD3 stimu-
lation, CD57þCD8þT cells from PBC patients were
less prone to apoptosis while having secreted increased
levelsofIL-5 than healthy
CD57þCD8þT cells infiltrate the PBC liver portal
area; this cell population demonstrates autoreactivity
against the HLA-A2.1, the restricted epitope.
It is of interest to note that the CD57þCD8þT-cell
subset has been previously described as possessing both
cytotoxic and regulatory functions.17-21In our present
study, the results suggest that a subpopulation of the
CD57þCD8þ
T cells,
highCD57þCD8highT cells, is a subset of cytotoxic
effector memory cells that could be critical in cell-
mediated immune response in PBC. The CD57 anti-
gen is a glycoepitope that was first described on
human NK 1 (HNK-1) cells.22An increase in the fre-
quency of CD57þT cells has been reported in
patients after bone marrow and solid organ trans-
plants,23,24in rheumatoid arthritis,25,26and acquired
immune deficiency patients.27These studies have sug-
gested a role for such CD57þT cells in the immuno-
logical abnormalities manifested in such diseases.
Although the frequency of the CD8þCD57þT cells
in normal hosts ranges from 5% to 20%,28the fre-
quencyofthis subsetincreases
Although several studies have suggested an augmented
cytotoxic ability of the CD8þCD57þT cell,19-21there
is a paucity of data on this CD8þCD57þT-cell popu-
lation in PBC.
The present results provide further insights into the
potential mechanisms by which CD8þcytotoxic T
cells serve as effector cells in the pathogenesis of
PBC.7,9,11We demonstrate herein that the CD45RO-
highsubset of CD57þCD8highcells were more resistant
to stimulation-induced apoptosis, as compared to their
counterparts, in the control subjects, which is similar
to the finding that the CD57þCD8þT-cell population
in PBMCs from patients with large granular lympho-
cyte leukemia (LGL) were resistant to Fas-stimulated
apoptosis (31).31This resistance to apoptosis is dem-
onstrated by lower expression of Annexin V and PD-1.
PD-1/PD-L interaction plays a critical role in CD8þ
T-cell tolerance32; previous work has demonstrated
that a decrease in PD-1 signaling can generate murine
autoimmune hepatitis.33
withcontrols; (3)inPBCpatients,the
controls;and (5)
namelyCD45RO-
withaging.29,30
Fig. 5. Comparison of the expression of Annexin V, Fas, TIM-3, and
PD-1 on CD45ROhighCD57þCD8highT cells and other CD8highT lym-
phocytes between PBC patients and healthy subjects. PBMCs were
stimulated with anti-CD3/28, then stained for Annexin V, Fas, TIM-3,
and PD-1. (A) Percentage of Annexin V, TIM-3, and PD-1-positive cells
or mean fluorescent intensity (MFI) of Fas in CD45ROhighCD57þCD8high
T cells from healthy controls (cont), all PBC patients (PBC), and PBC
patients at early stages (I/II) and late stages (III/IV). (B) Percentage
of Annexin V, TIM-3, and PD-1 positive cells or MFI of Fas in other
CD8highT cells. See Fig. 4 legend for the definition of other CD8high
T-cell populations. Bars denote mean percentages. *P < 0.05.
HEPATOLOGY, Vol. 54, No. 4, 2011 TSUDA ET AL.1299

Page 8

We reasoned that examination of chemokine recep-
tors and the integrin, a4b7, which provide homing
signals for circulating leukocytes to migrate to disease-
specific tissues, would provide evidence that these cir-
culating cells reflect the immune response in the target
organ.34T-cell recruitment to the liver is orchestrated
by a series of adhesion molecules and homing chemo-
kines.35We demonstrate herein that the CD45ROhigh
subsetof CD57þCD8high
expressed the homing integrin, a4b7. Although the
integrin, a4b7, and chemokine receptor, CCR9, are
typically associated with gut-homing phenotypes, they
have also been shown to mediate the adhesion of liver-
infiltrating lymphocytes through the expression of their
cognate ligands. Specifically, hepatic expression of the
a4b7 ligand, mucosal addressin cell adhesion molecule-1
(MAdCAM-1), has been demonstrated in a variety of
liver diseases, including PBC.35The CCR9 ligand and
the presumedgut-specific
expressed primarily by epithelial cells that line the small
intestine and has also been shown to be expressed on
hepatic endothelial cells of patients with primary scleros-
ing cholangitis, but, in contrast to MAdCAM-1, not in
PBC.36Thus, it is not surprising that, in our study, there
was an increased frequency of CD45ROhighCD57þCD8-
highT cells expressing a4b7, but not CCR9.
In a murine model of graft-versus-host disease,
which develops both portal hepatitis and nonsuppura-
tive destructive cholangitis similar to PBC, CCR5-
expressing CD8þT cells migrate into the portal areas
of the liver and play a significant role in causing liver
injury.37
The increased expression of CCR5 by
CD45ROhighCD57þCD8highT cells, but not other
CD8high
T cells,suggests
highCD57þCD8highT cells play an active role as effec-
tor cells during bile duct destruction in PBC. The
expression of CD28 decreases when CD8þT cells dif-
ferentiate from memory to effector CD8þT cells.38
The decrease in CD28 expression observed in PBC,
especially in early-stage PBC patients, implicates strong
effector functionwithautoreactive
CD45ROhighCD57þCD8highT cells.
Also, our observation of decreased CCR7 expression
on CD45ROhighCD57þCD8highT cells, compared
with other CD8highT cells, is consistent with the
cells morefrequently
chemokine, CCL25, is
thatCD45RO-
properties of
Fig. 7. (A) Liver immunohistochemistry from representative PBC and
CVH patients. Top and middle rows demonstrate 100% near complete
(top row) to 70% (middle row) CD57-positive staining within the CD8-
positive area in PBC; only minor CD57-positive staining was detected
within the CD8-positive area in CVH, as shown in the bottom row. (B)
IFN-c production upon PDC-E2 peptide stimulation. Data from HLA-
A2.1-positive PBC patients (n ¼ 3) and healthy controls (n ¼ 4) are
shown. Left bar graphs are without peptide stimulation, and right bar
graphs are with peptide stimulation.
Fig. 6. Production of IFN-c, IL-5,
and granzyme A by CD57þCD8þT
cells fromPBC
healthy controls. Aliquots of iso-
lated CD57þCD8þT cells from
PBC patients (n ¼ 7) and healthy
controls (n ¼ 8) were cultured in
duplicates in the presence of anti-
CD3/28 (0.5 lg/mL) for 5 days.
***P < 0.001.
patientsand
1300 TSUDA ET AL.HEPATOLOGY, October 2011

Page 9

theory that these cells are effector memory cells, as
opposed to CCR7þcentral memory cells, which
express lymph-node homing receptors and lack imme-
diate effector function.39It is reasoned that the
lymph-node homing CD8highT cells may become
mobilized to the periphery and acquire a different
spectrum of cell-surface molecules while decreasing the
levels of CCR7 expression during this process.
Our data demonstrate an increased secretion of IL-5
by CD57þCD8þT cells, compared with a similar
population from controls. Increased IL-5 has also been
found in other studies of CD8þCD57þT lympho-
cytes.40The transcripts for both Th1- and Th2-type
cytokines, such as INF-c, IL-2, and IL-5, are up-
regulated in the blood and liver of PBC,41,42and IL-5
promotes the differentiation of activated B cells into
Ig-producing cells and augments both IgM and IgA
production.43,44Moreover, IL-5 has potent, specific
effects on eosinophil activation and degranulation.45,46
Eosinophilia has been demonstrated in PBC patients,
and eosinophil cytotoxic products, such as major basic
protein, have been localized to the periportal regions
of the patient liver.42,47Our data demonstrate that
CD57þCD8þT cells are a potential source of IL-5 dur-
ing the chronic stages of PBC, exacerbating the destruc-
tion of BECs. CD57þCD8þ, in particular CD45RO-
highCD57þCD8high, T cells may also contribute to
continuous AMA production in PBC.
Collectively, our data demonstrate that CD57þ
CD8highT cells are a subset of cytotoxic memory T
cells that include specific autoreactive CD8þT cells.
Our results demonstrate, for the first time, the
increased frequency of CD45ROhighCD57þCD8highT
cells with the unique increased expression of a4b7
integrin and CCR5 as well as resistance to apoptosis
in PBC PBMCs; this reflects a role of CD45RO-
highCD57þCD8highT cells as a CD8þsubpopulation
contributing to the progressive destruction of small
bile ducts. We do not imply that the data herein will
be unique only to patients with PBC and, indeed,
may well be a property of multiple other autoimmune
diseases, obviously with different antigenic specificity
and tissue-specific homing receptors. We do suggest,
however, that further studies focused on these effector
mechanisms will enable the dissection of the role of
CD8þsubpopulations in PBC.
Acknowledgments:
Kawata, Dr. Katsunori Yoshida, and Dr. Yuki Moritoki
for technical support in this experiment. We also
thank Ms. Nikki Phipps for support in preparing this
article.
The authors thank Dr. Kazuhito
References
1. Gershwin ME, Mackay IR. The causes of primary biliary cirrhosis: con-
venient and inconvenient truths. HEPATOLOGY 2008;47:737-745.
2. Mao TK, Davis PA, Odin JA, Coppel RL, Gershwin ME. Sidechain
biology and the immunogenicity of PDC-E2, the major autoantigen of
primary biliary cirrhosis. HEPATOLOGY 2004;40:1241-1248.
3. Gershwin ME, Mackay IR, Sturgess A, Coppel RL. Identification and
specificity of a cDNA encoding the 70 kd mitochondrial antigen recog-
nized in primary biliary cirrhosis. J Immunol 1987;138:3525-3531.
4. Moteki S, Leung PS, Coppel RL, Dickson ER, Kaplan MM, Munoz S,
Gershwin ME. Use of a designer triple expression hybrid clone for
three different lipoyl domain for the detection of antimitochondrial
autoantibodies. HEPATOLOGY 1996;24:97-103.
5. Krams SM, Van de Water J, Coppel RL, Esquivel C, Roberts J, Ansari
A, Gershwin ME. Analysis of hepatic T lymphocyte and immunoglob-
ulin deposits in patients with primary biliary cirrhosis. HEPATOLOGY
1990;12:306-313.
6. Shimoda S, Nakamura M, Ishibashi H, Hayashida K, Niho Y. HLA
DRB4 0101-restricted immunodominant T cell autoepitope of pyru-
vate dehydrogenase complex in primary biliary cirrhosis: evidence of
molecular mimicry in human autoimmune diseases. J Exp Med 1995;
181:1835-1845.
7. Kita H, Lian ZX, Van de Water J, He XS, Matsumura S, Kaplan M,
et al. Identification of HLA-A2-restricted CD8(þ) cytotoxic T cell
responses in primary biliary cirrhosis: T cell activation is augmented by
immune complexes cross-presented by dendritic cells. J Exp Med 2002;
195:113-123.
8. Kita H, Matsumura S, He XS, Ansari AA, Lian ZX, Van de Water J,
et al. Quantitative and functional analysis of PDC-E2-specific autoreac-
tive cytotoxic T lymphocytes in primary biliary cirrhosis. J Clin Invest
2002;109:1231-1240.
9. Yang GX, Lian ZX, Chuang YH, Moritoki Y, Lan RY, Wakabayashi K,
et al. Adoptive transfer of CD8(þ) T cells from transforming growth
factor beta receptor type II (dominant negative form) induces autoim-
mune cholangitis in mice. HEPATOLOGY 2008;47:1974-1982.
10. Ueno Y, Ambrosini YM, Moritoki Y, Ridgway WM, Gershwin ME.
Murine models of autoimmune cholangitis. Curr Opin Gastroenterol
2010;26:274-279.
11. Wakabayashi K, Lian ZX, Moritoki Y, Lan RY, Tsuneyama K, Chuang
YH, et al. IL-2 receptor alpha(-/-) mice and the development of pri-
mary biliary cirrhosis. HEPATOLOGY 2006;44:1240-1249.
12. Talwalkar JA, Lindor KD. Primary biliary cirrhosis. Lancet 2003;362:
53-61.
13. Ludwig J, Dickson ER, McDonald GS. Staging of chronic nonsuppura-
tive destructive cholangitis (syndrome of primary biliary cirrhosis).
Virchows Arch A Pathol Anat Histol 1978;379:103-112.
14. Kumada T, Tsuneyama K, Hatta H, Ishizawa S, Takano Y. Improved
1-h rapid immunostaining method using intermittent microwave
irradiation: practicability based on 5 years application in Toyama
Medical and Pharmaceutical University Hospital. Mod Pathol 2004;17:
1141-1149.
15. Lan RY, Salunga TL, Tsuneyama K, Lian ZX, Yang GX, Hsu W, et al.
Hepatic IL-17 responses in human and murine primary biliary cirrho-
sis. J Autoimmun 2009;32:43-51.
16. Ohkawa T, Seki S, Dobashi H, Koike Y, Habu Y, Ami K, et al. Sys-
tematic characterization of human CD8þ T cells with natural killer
cell markers in comparison with natural killer cells and normal CD8þ
T cells. Immunology 2001;103:281-290.
17. Autran B, Leblond V, Sadat-Sowti B, Lefranc E, Got P, Sutton L, et al.
A soluble factor released by CD8þCD57þ lymphocytes from bone
marrow transplanted patients inhibits cell-mediated cytolysis. Blood
1991;77:2237-2241.
18. Wang EC, Moss PA, Frodsham P, Lehner PJ, Bell JI, Borysiewicz LK.
CD8highCD57þ T lymphocytes in normal, healthy individuals are oli-
goclonal and respond to human cytomegalovirus. J Immunol 1995;
155:5046-5056.
HEPATOLOGY, Vol. 54, No. 4, 2011TSUDA ET AL.1301

Page 10

19. Pedroza-Seres M, Linares M, Voorduin S, Enrique RR, Lascurain R,
Jimenez-Martinez MC. Pars planitis is associated with an increased fre-
quency of effector-memory CD57þ T cells. Br J Ophthalmol 2007;91:
1393-1398.
20. Sada-Ovalle I, Torre-Bouscoulet L, Valdez-Vazquez R, Martinez-Cairo
S, Zenteno E, Lascurain R. Characterization of a cytotoxic CD57þ T
cell subset from patients with pulmonary tuberculosis. Clin Immunol
2006;121:314-323.
21. Chattopadhyay PK, Betts MR, Price DA, Gostick E, Horton H, Roe-
derer M, De Rosa SC. The cytolytic enzymes granyzme A, granzyme
B, and perforin: expression patterns, cell distribution, and their rela-
tionship to cell maturity and bright CD57 expression. J Leukoc Biol
2009;85:88-97.
22. Abo T, Balch CM. A differentiation antigen of human NK and K cells
identified by a monoclonal antibody (HNK-1). J Immunol 1981;127:
1024-1029.
23. Leroy E, Calvo CF, Divine M, Gourdin MF, Baujean F, Ben Aribia
MH, et al. Persistence of T8þ/HNK-1þ suppressor lymphocytes in
the blood of long-term surviving patients after allogeneic bone marrow
transplantation. J Immunol 1986;137:2180-2189.
24. Fregona I, Guttmann RD, Jean R. HNK-1þ (Leu-7) and other lym-
phocyte subsets in long-term survivors with renal allotransplants. Trans-
plantation 1985;39:25-29.
25. Dupuy d’Angeac A, Monier S, Jorgensen C, Gao Q, Travaglio-Enci-
noza A, Bologna C, et al. Increased percentage of CD3þ, CD57þ
lymphocytes in patients with rheumatoid arthritis. Correlation with
duration of disease. Arthritis Rheum 1993;36:608-612.
26. Arai K, Yamamura S, Seki S, Hanyu T, Takahashi HE, Abo T. Increase
of CD57þ T cells in knee joints and adjacent bone marrow of rheu-
matoid arthritis (RA) patients: implication for an anti-inflammatory
role. Clin Exp Immunol 1998;111:345-352.
27. Sadat-Sowti B, Debre P, Mollet L, Quint L, Hadida F, Leblond V,
et al. An inhibitor of cytotoxic functions produced by CD8þCD57þ
T lymphocytes from patients suffering from AIDS and immunosup-
pressed bone marrow recipients. Eur J Immunol 1994;24:2882-2888.
28. Morley JK, Batliwalla FM, Hingorani R, Gregersen PK. Oligoclonal
CD8þ T cells are preferentially expanded in the CD57þ subset.
J Immunol 1995;154:6182-6190.
29. Ligthart GJ, van Vlokhoven PC, Schuit HR, Hijmans W. The
expanded null cell compartment in ageing: increase in the number of
natural killer cells and changes in T-cell and NK-cell subsets in human
blood. Immunology 1986;59:353-357.
30. McNerlan SE, Rea IM, Alexander HD, Morris TC. Changes in natural
killer cells, the CD57CD8 subset, and related cytokines in healthy
aging. J Clin Immunol 1998;18:31-38.
31. Lamy T, Liu JH, Landowski TH, Dalton WS, Loughran TP, Jr. Dysre-
gulation of CD95/CD95 ligand-apoptotic pathway in CD3(þ) large
granular lymphocyte leukemia. Blood 1998;92:4771-4777.
32. Martin-Orozco N, Wang YH, Yagita H, Dong C. Cutting edge: pro-
grammed death (PD) ligand-1/PD-1 interaction is required for CD8þ
T cell tolerance to tissue antigens. J Immunol 2006;177:8291-8295.
33. Kido M, Watanabe N, Okazaki T, Akamatsu T, Tanaka J, Saga K,
et al. Fatal autoimmune hepatitis induced by concurrent loss of natu-
rally arising regulatory T cells and PD-1-mediated signaling. Gastroen-
terology 2008;135:1333-1343.
34. Aoki CA, Bowlus CL, Gershwin ME. The immunobiology of primary
sclerosing cholangitis. Autoimmun Rev 2005;4:137-143.
35. Grant AJ, Lalor PF, Hubscher SG, Briskin M, Adams DH. MAdCAM-
1 expressed in chronic inflammatory liver disease supports mucosal
lymphocyte adhesion to hepatic endothelium (MAdCAM-1 in chronic
inflammatory liver disease). HEPATOLOGY 2001;33:1065-1072.
36. Miles A, Liaskou E, Eksteen B, Lalor PF, Adams DH. CCL25 and
CCL28 promote alpha4 beta7-integrin-dependent adhesion of lympho-
cytes to MAdCAM-1 under shear flow. Am J Physiol Gastrointest Liver
Physiol 2008;294:G1257-G1267.
37. Murai M, Yoneyama H, Harada A, Yi Z, Vestergaard C, Guo B, et al.
Active participation of CCR5(þ)CD8(þ) T lymphocytes in the patho-
genesis of liver injury in graft-versus-host disease. J Clin Invest 1999;
104:49-57.
38. Tomiyama H, Matsuda T, Takiguchi M. Differentiation of human
CD8(þ) T cells from a memory to memory/effector phenotype.
J Immunol 2002;168:5538-5550.
39. Sallusto F, Lenig D, Forster R, Lipp M, Lanzavecchia A. Two subsets
of memory T lymphocytes with distinct homing potentials and effector
functions. Nature 1999;401:708-712.
40. Chong LK, Aicheler RJ, Llewellyn-Lacey S, Tomasec P, Brennan P,
Wang EC. Proliferation and interleukin 5 production by CD8hi
CD57þ T cells. Eur J Immunol 2008;38:995-1000.
41. Martinez OM, Villanueva JC, Gershwin ME, Krams SM. Cytokine pat-
terns and cytotoxic mediators in primary biliary cirrhosis. HEPATOLOGY
1995;21:113-119.
42. Krams SM, Cao S, Hayashi M, Villanueva JC, Martinez OM. Eleva-
tions in IFN-gamma, IL-5, and IL-10 in patients with the autoimmune
disease primary biliary cirrhosis: association with autoantibodies and
soluble CD30. Clin Immunol Immunopathol 1996;80:311-320.
43. Yokota T, Coffman RL, Hagiwara H, Rennick DM, Takebe Y, Yokota
K, et al. Isolation and characterization of lymphokine cDNA clones
encoding mouse and human IgA-enhancing factor and eosinophil col-
ony-stimulating factor activities: relationship to interleukin 5. Proc Natl
Acad Sci U S A 1987;84:7388-7392.
44. Morikawa K, Oseko F, Morikawa S, Imai K, Sawada M. Recombinant
human IL-5 augments immunoglobulin generation by human B lym-
phocytes in the presence of IL-2. Cell Immunol 1993;149:390-401.
45. Fujisawa T, Abu-Ghazaleh R, Kita H, Sanderson CJ, Gleich GJ. Regu-
latory effect of cytokines on eosinophil degranulation. J Immunol
1990;144:642-646.
46. Lopez AF, Sanderson CJ, Gamble JR, Campbell HD, Young IG, Vadas
MA. Recombinant human interleukin 5 is a selective activator of
human eosinophil function. J Exp Med 1988;167:219-224.
47. Terasaki S, Nakanuma Y, Yamazaki M, Unoura M. Eosinophilic infil-
tration of the liver in primary biliary cirrhosis: a morphological study.
HEPATOLOGY 1993;17:206-212.
1302TSUDA ET AL.HEPATOLOGY, October 2011