Content uploaded by Johan Bäcklund
Author content
All content in this area was uploaded by Johan Bäcklund
Content may be subject to copyright.
Predominant selection of T cells specific for the
glycosylated collagen type II epitope (263–270)
in humanized transgenic mice and in
rheumatoid arthritis
Johan Ba
¨cklund
†‡
, Stefan Carlsen
†‡
, Torsten Ho
¨ger
§
, Bjo
¨rn Holm
¶
, Lars Fugger
储
, Jan Kihlberg
¶
, Harald Burkhardt
§
,
and Rikard Holmdahl
†
**
†Section of Medical Inflammation Research, So¨ lvegatan 19, I11 BMC, Lund University, SE-221 84 Lund, Sweden; §Department of Internal Medicine III and
Institute of Clinical Immunology, Friedrich-Alexander-University Erlangen-Nu¨ rnberg, D-91054 Erlangen, Germany; ¶Department of Chemistry, Umeå
University, SE-901 87 Umeå, Sweden; and 储Department of Clinical Immunology, Aarhus University Hospital, Skejby Sygehus, 8200N Aarhus, Denmark
Communicated by N. Avrion Mitchison, University College London, London, United Kingdom, April 29, 2002 (received for review February 25, 2002)
Rheumatoid arthritis (RA) is associated with certain MHC class II
alleles and is characterized by a chronic autoimmune response
in the joints. Using transgenic mice expressing human DR4
(DRB1*0401) and human CD4, but lacking endogenous MHC class
II, we show that posttranslational glycosylation of type II collagen
(CII) influences the level of T cell tolerance to this candidate
cartilage-specific autoantigen. In such mice, the expression of
human CII resulted in a tolerized murine T cell response to human
CII. However, tolerance induction remained incomplete, preferen-
tially deleting responses to the nonmodified CII 263–270 epitope,
whereas T cell recognition of a glycosylated variant of this epitope
was affected to a lesser degree. A similar dominance of T cell
responses to CII-glycopeptides was recorded in a cohort of severely
affected RA-patients (nⴝ14). Thus, RA T cells predominantly
recognize the immunodominant CII peptide in its glycosylated form
and may explain why previously it has been difficult to detect T cell
responses to CII in RA patients.
Rheumatoid arthritis (RA) is an autoimmune disease that
primarily affects peripheral joints with cartilage destruction
and subsequent bone erosion. The role of T cells in RA is
supported by the large number of activated CD4
⫹
cells reported
in the synovium of affected joints and by the association of RA
to certain MHC (HLA) class II genes [e.g., DRB1*0401 (in DR4)
and DRB1*0101 (in DR1)] that encode a specific peptide
binding pocket, the so-called shared epitope.
Collagen type II (CII), the main constituent of hyaline carti-
lage, has been proposed as one possible autoantigen in RA
because CII-specific antibodies are frequently found in RA
patients and because an RA-like disease can be induced in
certain mouse strains after immunization with CII. However,
reports on T cell immunity to CII in RA patients as well as in
healthy individuals are inconclusive, and the role of CII, and
even T cells, in RA is still argued and remains to be proven.
Previous studies perfor med in DR4- and DR1-expressing mice
have located the immunodominant T cell epitope to position
263–270 in CII by using synthetic peptides (1–3), but studies in
RA patients have in general failed to identify DR4兾DR1-
restricted T cells specific for the same epitope (4, 5). One
potential reason for this failure is that these studies did not
directly address that CII can become posttranslationally modi-
fied. Within CII263–270, the lysines at positions 264 and 270 can
be hydroxylated and further glycosylated with mono- or disac-
charides, i.e., with a

-D-galactopyranosyl or an
␣
-D-
glucopyranosyl-(1-2)-

-D-galactopyranosyl residue. Such modi-
fications have previously been shown to be of importance in the
development of collagen-induced arthritis (CIA) in A
q
-
expressing mice (6, 7). Although the A
q
-molecule is of a DQ
isotype, its peptide-binding groove shows more similarity to the
shared epitope variant of the DR4-molecule, because it binds
almost the same CII epitope and presents the same amino acid
side chains to T cells (3, 8–11). Furthermore, CIA is most
commonly induced with heterologous CII, which is believed to
induce a heteroreactive T cell response, followed by a B cell
response, which in contrast to the T cell response is highly
crossreactive to mouse CII (12, 13). Thus, earlier CIA experi-
ments in DR4-expressing mice have not addressed T cell toler-
ance to self-CII. Therefore, to get a better understanding of
autoimmunity to cartilage-derived proteins, we need to consider
both posttranslational modifications and the aspect of T cell
tolerance in these models. To achieve this result, we used a
humanized mouse model expressing HLA-DRB1*0401兾
DRA1*0101, human CD4, and human CII (huCII) on a back-
ground deficient of murine class II expression (14, 15). In these
mice, T cell responses to huCII peptides and the impact of
posttranslational modification on the induction of CII-specific T
cell tolerance was investigated.
Materials and Methods
Mice. Crossing of different transgenic mice generated two groups of
mice that were used for experiments: (i) DR4, expressing transgenic
DRB*0401 and human CD4 but no murine class II molecule; and
(ii) huCII兾DR4, expressing DRB*0401, human CD4 and huCII but
no murine class II molecule. To get these groups, huCII transgenic
mice (15), mice expressing HLA-DR4 together with human CD4
(14), and mice lacking H-2 class II (16) were crossed as follows.
First, HLA-DR4兾huCD4 on a B10 background was introduced
with H-2
plus/⫺
on a B6 background and backcrossed twice to B10
and subsequently intercrossed. HLA-DR4兾huCD4, H-2
⫺/⫺
mice
were then crossed with huCII mice on a C3H background. Off-
spring were backcrossed three generations to HLA-DR4兾huCD4,
H-2
⫺/⫺
on the B10 background. Finally, mice were intercrossed
twice and selected for homozygosity in H-2
⫺/⫺
. Mice were bred and
kept in the animal facility of Medical Inflammation Research
(http:兾兾net.inflam.lu.se).
Antigens and
in Vitro
T Cell Assays. HuCII was extracted from hip
joints (obtained from replacement surgery) after pepsin diges-
tion and purified as described (17). The following nonmodified
and glycosylated CII peptides were synthesized as described (8,
18, 19): K264兾270 (nonmodified CII261–275 with a lysine resi-
due at position 264 and 270); Gal264 (CII261–278, glycosylated
Abbreviations: CII, type II collagen; CIA, collagen-induced arthritis; huCII, human type II
collagen; PBMC, peripheral blood mononucleated cells; RA, rheumatoid arthritis; mDR, T
cell clones from mouse CII and DR4; hDR, T cell clones from huCII兾DR4.
See commentary on page 9611.
‡J.B. and S.C. contributed equally to the present work.
**To whom reprint requests should be addressed. E-mail: rikard.holmdahl@inflam.lu.se.
9960–9965
兩
PNAS
兩
July 23, 2002
兩
vol. 99
兩
no. 15 www.pnas.org兾cgi兾doi兾10.1073兾pnas.132254199
with a

-D-galactopyranose residue on L-hydroxylysine exclu-
sively at position 264); Gal270 (CII261–278, glycosylated exclu-
sively at position 270); and Gal264兾270 (CII259–278, glycosy-
lated at position 264 and 270). The following modified forms of
the Gal264-glycopeptide, named deoxy-glycopeptides, were also
synthesized (B.H., unpublished data): 2-deoxyGal (CII259–273,
where the hydroxy-group at position 2 on the

-D-galactopyr-
anosyl moiety has been eliminated); 3-deoxyGal (CII259 –273,
missing the hydroxy-group at position 3 on galactose); and
4-deoxyGal (CII259–273, missing the hydroxy-group at position
4 on galactose). As control for the deoxy-peptides, a shortened
form of the Gal264-peptide (CII259–273) was used. Mice were
immunized with huCII in complete Freund’s adjuvant, and 10
days later cells from draining lymph nodes were stimulated in
vitro for determination of antigen-specific proliferation and
IFN-
␥
production as described (20). Establishment of T cell
hybridoma clones and determination of antigen specificity was
performed as described (8), with the exception that Gal264兾
270-peptide was used for immunization and for the first in vitro
restimulation (20
g兾ml).
Analysis of Human T Cell Responses. Fourteen patients fulfilling the
RA-classification criteria of the American College of Rheuma-
tology (21) were recruited for the study. The study protocol was
approved by the review board of the Friedrich-Alexander-
University Erlangen-Nuremberg, and informed consent was
obtained from all individuals before entering the study. All
patients had a severe course of the disease so that insufficient
response to conventional therapy required an intensified treat-
ment with TNF
␣
-blocking antibodies [D2E7 (Knoll AG-BASF
Pharma, Ludwigshafen, Germany) or infliximab (Essex Pharma,
Munich, Germany]. The mean age of the patients (2 male and
12 female) was 63.6 ⫾6.5 (SD) years, and the disease duration
10.9 ⫾8.1 (SD) years. In 9 patients, information about the
expression of HLA-DRB1*-alleles was available and is included
in the result section. Peripheral blood mononucleated cells
(PBMC) from the patients were separated by using Histoprep
(BAG, Lich, Germany). For antigenic stimulation of 10
6
PBMC,
10
g of CII-peptide and 1
g anti-human CD28 (Becton
Dickinson) were added per ml of culture medium. T cell
receptor-specific responses were controlled in parallel by using
culture conditions that either omitted any stimulation or only
exposed the cells to the costimulatory anti-CD28 antibody
overnight in the absence of antigen. T cell responsiveness to a
common recall antigen was tested in parallel cultures of PBMC
by using 10
g兾ml tetanus toxoid (Calbiochem-Nova Biochem)
and anti-CD28 for stimulation. Monensin (2.5 mM; Sigma-
Aldrich) was added to the overnight cultures, and the cells were
incubated for additional 4 h before harvesting. Subsequently, the
cells were washed twice in PBS and fixed in 4% paraformalde-
hyde兾PBS solution for 7 min at 37°C, followed by a repeated
washing procedure in PBS. A permeabilization step was per-
formed for 10 min with 0.5% saponin兾1% BSA兾0.1% NaN
3
in
PBS; afterwards, the cells were washed twice with PBS兾1% BSA.
Cells were stained with 0.2
g rat anti-human IL-2-PE (Becton
Dickinson) and 3
l CD4-FITC or CD3-FITC (Beckman
Coulter) for 20 min at 4°C. Fluorescence intensities were deter-
mined by using a Coulter Epics XL-MCL flow cytometer and
SYSTEM-II software. Large activated lymphocytes (blasts) were
gated according to forward and side scatter as described previ-
ously (22, 23). Cells not treated with saponin were used to
exclude background staining of anti-IL-2 antibody.
Results
Strong, but Incomplete, Negative Selection of the Autoreactive T Cell
Population. Previous studies have identified the immunodomi-
nant T cell epitope in bovine and human CII in DR4-transgenic
mice, and they have also shown that these mice are susceptible
to CIA (1–3). However, because bovine and human CII both
differ from mouse CII within the identified epitope (at position
266, glutamic acid in heterologous CII, compared with aspartic
acid in mouse CII), these reports did not include the aspect of
T cell tolerance to self-CII. To generate an animal model of RA
that would reflect the situation in humans more accurately, we
crossed DR4兾human CD4-transgenic and mouse MHC class
II-deficient mice with huCII-transgenic mice to generate two
lines of mice: DR4 mice (DR4兾human CD4-transgenic and
mouse MHC class II-deficient mice, expressing mouse CII in
cartilage) and huCII兾DR4 mice (DR4兾human CD4-transgenic
and mouse MHC class II-deficient mice, expressing huCII in
cartilage). These mice were then immunized with huCII to
examine the degree of T cell tolerance to endogenously ex-
pressed CII.
The heterologous immune response in DR4 mice to CII was
biased toward the nonmodified peptide (K264兾270), followed by
a weaker response to peptides glycosylated with

-D-
galactopyranosyl moieties on hydroxylysine residues at either
position 264 (Gal264) or 270 (Gal270) (Fig. 1). The response to
huCII protein was also weak but significantly above background
level when measuring the production of IFN-
␥
. In sharp contrast,
the response in huCII兾DR4 mice was severely reduced against all
CII antigens as compared with the response in DR4 mice,
showing that strong tolerance to self-CII is present (Pvalue less
than 0.05 against all CII antigens except the Gal264兾270-specific
IFN-
␥
production). Most importantly, however, tolerance was
not complete because huCII兾DR4 mice were able to mount a
significant response above background level against the Gal264-
peptide, when the proliferation (P⫽0.023) as well as IFN-
␥
production (P⫽0.041) was measured (Fig. 1). In contrast, the
response against the nonmodified K-peptide in huCII兾DR4 mice
was not significant above the background response (P⬎0.05,
Fig. 1).
To further evaluate the shift in epitope selection between DR4
mice (with a dominant response to the K264兾270-peptide) and
huCII兾DR4 mice (with a significant response against the
Fig. 1. Strong but incomplete tolerance to glycosylated CII in humanized
mice. Recall in vitro response of lymph node cells from DR4 and huCII兾DR4
mice immunized 10 days earlier with human CII (huCII). Cells were restimu-
lated with huCII and the following CII-peptides: nonglycosylated (K264兾270),
glycosylated at hydroxylysine 264 (Gal264), glycosylated at hydroxylysine 270
(Gal270), or glycosylated at both hydroxylysine residues (Gal264兾270) with

-D-galactopyranosyl residues. Ten animals (5 males and 5 females) of each
mouse line were investigated. As a positive control, the recall response to
mycobacteria antigen PPD (purified protein derivate, present in complete
Freund’s adjuvant) was measured. The response in huCII兾DR4 mice was sig-
nificantly reduced against all CII antigens (P⬍0.05) as compared with DR4
mice, except for Gal264兾270-specific IFN-
␥
production. *, Significant responses
above background levels for huCII兾DR4 mice. *,P⬍0.05, Mann–Whitney U
test; bars represent mean ⫾SEM. Antigen-specific proliferation and IFN-
␥
production: ⌬cpm and ⌬U兾ml respectively (response with antigen—response
in the absence of antigen).
Ba¨cklund et al. PNAS
兩
July 23, 2002
兩
vol. 99
兩
no. 15
兩
9961
IMMUNOLOGY
Gal264-peptide alone), we made a pairwise comparison of the
response to the K- and the Gal264-peptide (Fig. 2). The response
to the K-peptide was stronger than to the Gal264-peptide in all
DR4 mice at both concentrations tested (10 and 50
g兾ml, P⫽
0.0051). In contrast, a relatively stronger response against the
Gal264-peptide than to the K-peptide was noted in 8 of 10
huCII兾DR4 mice when cells were stimulated with the higher
antigen concentration (P⫽0.0218) and in all mice when
stimulated with the lower antigen concentration (P⫽0.0051).
Hence, CII-specific T cells are strongly tolerized in DR4 mice
expressing huCII in cartilage. However, glycopeptide-specific T
cells appear to be less affected by tolerance than T cells specific
for the nonmodified epitope.
Different T Cell Fine Specificity of Auto- and Heteroreactive T Cell
Clones. To confirm our finding that DR4-restricted CII recog-
nition includes T cells specific for glycosylated epitopes, hybrid-
oma clones from Gal264兾270-peptide-immunized DR4 and hu-
CII兾DR4 mice were established and characterized for
comparison. Despite the weak in vitro response to glycosylated
CII antigens seen earlier in DR4 and huCII兾DR4 mice after CII
immunization, glycopeptide-specific hybridomas were success-
fully established from both lines of transgenic mice (Table 1).
Specific recognition of the glycopeptide was evident because
all clones responded to the Gal264兾270-peptide but not to the
nonglycosylated peptide. The majority of T cell clones from DR4
mice (named mDR for mouse CII and DR4) recognized the
Gal264-peptide (8 of 10, group I and II), whereas only 2 clones,
originally collected from the same subcloning, specifically re-
sponded to the Gal270-peptide (group III). Gal264-specific
clones could further be divided in two groups where clones that
displayed a higher sensitivity to the glycopeptide also had a
strong response toward huCII (group I), whereas clones that
responded weakly to the Gal264-peptide also responded weakly
to huCII (group II). In line with this finding, the Gal270-specific
clone(s) responded weakly to the Gal270-peptide and not at all
to huCII.
Interestingly, T cell clones from huCII兾DR4 mice (named
hDR, for huCII and DR4) differed from mDR clones in their
response to CII-peptides: (i) The biased recognition of the
Gal264-peptide was not observed for hDR-clones (Table 1).
(ii) Among the Gal264-specific clones (group I and II), the
response was somewhat weaker, compared with the correspond-
ing mDR clones. This difference in antigen sensitivity was even
more evident in the response to intact huCII (Fig. 3 and Table
1). (iii) Gal270-specific hDR clones also responded to huCII
(group III), although high antigen concentration was required
(Table 1).
To confirm that it was the posttranslational modification of
Fig. 2. Comparative analysis of the individual response to the nonmodified
and the galactosylated T cell epitope in DR4 and huCII兾DR4 mice. Recall in vitro
response of lymph node cells from DR4 and huCII兾DR4 mice immunized 10
days earlier with human CII. Cells were restimulated with either 10 or 50
g兾ml
of the K264兾270- or the Gal264-peptide. A ratio above 1 indicates a stronger
response against the K264兾270-peptide whereas a ratio below 1 indicates a
stronger response to the Gal264-peptide. The response of individual DR4 mice
was stronger to the K264-peptide whereas the responses of huCII兾DR mice
were stronger to the Gal264-peptide (Wilcoxon signed rank test).
Table 1. DR4-restricted T cell hybridoma responses to the CII259–273 epitope and human CII
Group Clone
Glycopeptide specificity Galactose fine-specificity
K264兾270* Gal264* Gal270* Gal264兾270* huCII
†
Gal264* 2-deoxyGal* 3-deoxyGal* 4-deoxyGal*
I mDR-1.1 0 5 0 5 5 5 0 4 0
mDR-2.2 0 4 0 4 5 4 0 3 0
mDR-8.4 0 4 0 4 5 4 0 4 0
mDR-14.1 0 5 0 5 6 5 0 4 0
mDR-16.2 0 5 0 5 6 4 0 3 0
II mDR-6.3 0 3 0 4 1 3 0 0 0
mDR-15.2 0 3 0 3 1 2 0 0 0
mDR-17.1 0 5 0 3 3 5 0 0 0
III mDR-4.1 0 0 3 3 0
mDR-4.2 0 0 3 3 0
I hDR-2.3 0 4 0 2 1 4 0 0 4
hDR-9.1 0 4 0 2 2 3 0 1 3
hDR-11.2 0 4 0 2 1 4 0 1 4
II hDR-13.1 0 3 0 4 0
hDR-13.2 0 3 0 3 0 2 0 4 0
III hDR-1.1 0 0 4 4 2
hDR-4.5 0 0 4 4 2
hDR-6.1 0 0 4 3 1
hDR-12.3 0 0 4 4 1
IV hDR-3.3 0 0 0 4 0
Semiquantitative scoring of T cell hybridoma response following stimulation with glycosylated or nonglycosylated collagen peptides (see Fig. 1). Shaded boxes
highlight specific responses for the different groups of clones.
*Sensitivity of the T cell hybridomas was determined by the amount of peptide required for a CTLL response ⬎1000 CPM above background; 0, no response; 1 ⫽
50
g兾ml; 2 ⫽10
g兾ml; 3 ⫽2
g兾ml; 4 ⫽0.4
g兾ml; 5 ⫽0.08
g兾ml; 6 ⫽0.016
g兾ml.
†Sensitivity of the T cell hybridomas to CII protein was determined as for the CII-peptides, but number indicates a 5 times higher concentration of antigen, i.e.,
1⫽250
g兾ml; 2 ⫽50
g兾ml; 3 ⫽10
g兾ml; 4 ⫽2
g兾ml; 5 ⫽0.4
g兾ml; 6 ⫽0.08
g兾ml.
9962
兩
www.pnas.org兾cgi兾doi兾10.1073兾pnas.132254199 Ba¨cklund et al.
CII that was specifically recognized by the obtained clones,
Gal264-specific hybridomas were also tested against three dif-
ferently modified glycopeptides, where one of the hydroxy-
groups on the galactose moiety had been selectively removed at
either position 2, 3, or 4, thereby generating three mono-deoxy
glycopeptides. The two groups among the Gal264-specific mDR
clones were confirmed in that they, apart from having different
sensitivity to the Gal264-peptide, also recognized the sugar
residue differently (Table 1). Similarly, the two groups of
Gal264-specific hDR clones were also confirmed by the use of
the deoxy-glycopeptides. Notably, differences in deoxy-
glycopeptide responses between mDR and hDR clones indicated
that T cells were differently selected if huCII was expressed. For
example, all mDR clones were stringently dependent on the
hydroxy-group at the C4-carbon of galactose, whereas the re-
sponses of 3 of 4 hDR clones did not depend on the C4-hydroxy-
group.
Glycopeptide-Specific T Cells Dominate the CII Response in RA Pa-
tients. Thus far, results from our humanized animal model
showed that T cells specifically recognizing the glycostructure
remain after tolerance induction and become dominant by the
introduction of autologous huCII, whereas T cells with other fine
specificities to the CII epitope seem to be functionally impaired
or deleted to a greater extent. We argued that these findings
could explain earlier reported difficulties in identifying CII
specific T cell clones from joints or blood of RA patients, and
therefore we tested a total number of 14 RA patients for T cell
responses against the different glycosylated peptides. In the
investigated cohort of RA patients, cytokine flow cytometry of
in vitro-stimulated PBMC revealed specific IL-2 responses to the
glycosylated CII-peptides. Using the increase in the percentage
of IL-2-producing cells within the CD3 or CD4 positive T cell
population above baseline of negative control as a parameter of
specific T cell stimulation, 7 patients could be identified as
CII-peptide responsive (Table 2). Interestingly, all CII-
responding patients recognized the glycopeptides, but only 2 of
these responded also to the nonglycosylated peptide. This result
shows that T cell recognition was confined to the glycosylated
CII-variants in 5 patients, representing more than 30% of the
entire cohort. With one exception (patient 10) the baseline
population of IL-2-producing T cells was low in all patients
investigated (Table 2). However, also in patient 10, a specific rise
in the percentage of IL-2-producing CD4
⫹
T cells from 0.20 to
0.92% on stimulation with Gal264兾270 was detectable. The
individual glycospecific responses of patients were heteroge-
nous. In some patients, the response was biased to one of the
glycosylation sites, whereas other patients had equal responses to
both sites. In addition, in some patients, the response was
restricted to one of the glycosylation sites, whereas the response
to the peptide with glycosylation at both positions was weak or
nonexistent. This latter response pattern was also seen in mice
expressing both human DR4 and human CII (see Fig. 1). We also
Fig. 3. An example of different responses to glycosylated CII between DR4
and huCII兾DR4 mice. T cell hybridomas clones mDR-16.2 and hDR-9.1 were
obtained from DR4 and huCII兾DR4 mice, respectively, after immunization with
the Gal264兾270-peptide. Cells were stimulated with titrated amounts of huCII
or CII-peptides (see Fig. 1) and investigated for production of IL-2.
Table 2. Analysis of T cell recognition of CII-peptide 259–273 in RA patients
Patient
no. No antigen* TT* K264兾270* Gal264* Gal270* Gal264兾270* HLA
†
CII-peptide
response
‡
FACS staining
1 0.04 0.42 0.05 0.02 0.03 0 ND —CD3兾IL-2
2 0.03 0.07 0.06 0.04 0.04 0.03 ND —CD3兾IL-2
3
§
0.07 1.35 0.09 0.24 0 0.12 DRB1*0404;11 (⫹) CD3兾IL-2
4 0.02 0.04 0.02 0.79 0.99 0.89 DRB1*0401;14 ⫹⫹⫹ CD3兾IL-2
5 0.02 0.07 0.29 0.29 0.65 0 DRB1*0401;08 ⫹⫹⫹ CD3兾IL-2
6 0.02 0.09 0.03 0.02 0.02 0.02 ND —CD3兾IL-2
7 0.04 0.20 0.02 0.03 0.01 0.03 DRB1*08; 14 —CD3兾IL-2
8 0.01 0.02 0.01 0.01 0 0.01 DRB1*07; 11 —CD4兾IL-2
9
§
0.04 0.37 0.01 0.12 0.04 0.05 DRB1*0401;11 (⫹) CD4兾IL-2
10
§
0.20 0.28 0.26 0.14 0.25 0.92 DRB1*0401;03 ⫹⫹ CD4兾IL-2
11 0.01 0.38 0.31 0.17 0.18 0.14 DRB1*0102;13 ⫹CD4兾IL-2
12 0.05 0.61 0.07 0.41 0.48 0.07 DRB1*0404;15 ⫹CD4兾IL-2
13 0.04 0.21 0.02 0.04 0.02 0.03 ND —CD4兾IL-2
14 0.02 0.11 0.06 0.04 0.01 0.01 ND —CD4兾IL-2
Percentage of IL-2-producing T cells after in vitro culture of PBMC without antigen (No antigen) or following stimulation with tetanus toxoid (TT1), or
following stimulation with CII-peptides (see Fig. 1). IL-2 production was analyzed for the entire CD3⫹T cell population in patients 1 to 7 or within the CD4⫹subset
in patients 8 to 14.
*Percentage of double-positive T cells.
†DRB1-alleles containing the ‘‘shared epitope’’ QKRAA or QRRAA in amino acid positions 70–74 are highlighted in shaded cells and bold italics; subtyping of
DRB1* was performed only on 04 and 01 haplotypes. ND, Not determined.
‡Semiquantitative scoring of T cell response to any of the CII-peptides (% of IL-2-producing T cells): —,⬍0.1% or ⬍3⫻(no antigen control); (⫹), ⱖ3⫻(no antigen
control); ⫹,ⱖ4⫻(no antigen control); ⫹⫹,ⱖ5⫻(no antigen control); ⫹⫹⫹,ⱖ10⫻(no antigen control). Shaded cells and bold figures in the table highlight
CII-peptide-specific responses fulfilling at least the (⫹) criterion.
§Patients where repetitive analyses were performed. Patient 3 elicited a similar response exclusively to the Gal264-peptide in a second analysis performed 1 wk
later (see Fig. 4). For patient 9, in a second study performed 1 mo later, 0.1% of CD3⫹cells produced IL-2 upon stimulation with the Gal264-peptide, compared
with 0.04% in the absence of antigen and 0.02% after stimulation with the K264兾270- or Gal270-peptide. Patient 10 failed to respond to any CII-peptide when
investigated 6 mo later.
Ba¨cklund et al. PNAS
兩
July 23, 2002
兩
vol. 99
兩
no. 15
兩
9963
IMMUNOLOGY
performed repetitive analysis of T cell responses to CII in 3 of
the 14 patients. In the initial investigation, patient 3 responded
exclusively to the Gal264-peptide because 0.24% of CD3
⫹
cells
produced IL-2 on stimulation with this peptide, compared with
0.07% in the absence of antigen. One week later, the same
patient elicited an almost identical response (Fig. 4) where the
two-color staining for CD4 and IL-2 show an increase in
the double-positive T cell population on stimulation with the
Gal264-peptide to 0.22% whereas exposure to the K264兾270-
peptide did not result in an increased IL-2 production. Similarly,
patient 9 initially responded to the Gal264-peptide (Table 2),
and this exclusive response was also noted when the patient was
investigated 1 mo later (see footnotes in Table 2). However, in
patient 10, the initially recorded response to the Gal264兾270-
peptide (Table 2) was undetectable 6 mo after the first assess-
ment (see footnotes in Table 2). Thus, although CII-specific T
cell responses may vary with time in some patients, possibly
because of complex influences by treatment or spontaneous
variations in the immune response of a characteristically chronic
relapsing disease, the results obtained in a small cohort of RA
patients with established diseases and severe course show the
dominant targeting of glycosylated variants of the CII-peptide by
the autoreactive T cell response. Finally, the available HLA-
typing information on the patients reveals that all patients
exhibiting T cell responses to the respective CII-peptide variants
express at least one of the HLA-DRB1*-alleles containing the
amino acid consensus motif QK(R)RAA in position 70–74. This
motif constitutes the so-called ‘‘shared epitope’’ (24, 25), and is
crucial for the binding of the CII-peptides to the respective class
II molecules during antigen presentation (26).
Discussion
Posttranslationally glycosylated peptides from CII are presented
by the RA-associated DR4 molecules as shown in human
DR4-transgenic mice. In DR4-transgenic mice expressing huCII,
and most likely also in humans, T cells recognizing the nongly-
cosylated 263–270 epitope are strongly tolerized, or even de-
leted, whereas T cells specific for the different glycosylated
peptides persist. This is a potentially crucial finding because it
provides an explanation not only for the role of the DR4 in the
presentation of joint derived peptides, but also for the difficulties
in providing evidence for T cells recognizing autologous CII.
The relevance of the findings made in mice expressing huCII
and DR4 is evident because 30% of the investigated RA patients
exhibited a predominant response to the glycosylated forms of
the CII263–270 epitope. Thus, our data show that the physio-
logical posttranslational modification of variable carbohydrate
attachment converts the immunodominant naked self-peptide,
which is tolerogenic, into several cryptic self-determinants that
remain immunogenic.
In addition to reports showing that transgenic expression of
DR4 and DR1 in mice permits development of CIA (2, 3, 27),
there are several investigations describing T cell immunity to CII
in either healthy individuals or RA patients or both (5, 28–36).
However, less is known regarding the determinants recognized
by CII-specific T cells in humans, and the interpretations of the
reported findings are not all in agreement (4, 5, 34, 35). For
example, one recent study failed to identify CII-specific T cells
in RA patients when CII259–272 loaded DR4-tetramers were
used (4), despite the earlier defined immunodominance of this
peptide in DR4-transgenic mice (1).
By including the aspect of T cell tolerance to self-CII and
recognition of glycosylated CII-epitopes in our humanized an-
imal model of RA, we can now provide a possible explanation
for the ambiguous results concerning CII recognition in RA
patients. Our finding that T cells specific for glycosylated CII
appeared less tolerized by endogenous expression of huCII than
T cells specific for the nonmodified epitope is predictable from
an earlier finding in A
q
-expressing mice (15, 37). Mice with
transgenic expression of either rat or human CII displayed strong
T cell tolerance to the immunodominant CII epitope and were
also partially protected from CIA when immunized with rat and
human CII, respectively (15, 37). Moreover, mice transgenic for
the A
q
-restricted immunodominant T cell epitope CII256–270,
present on heterologous CII, displayed relatively stronger tol-
erance against the nonglycosylated variant (20). Whether per-
sistence of glycopeptide-specific T cells in huCII兾DR4 mice
depends on the possibility that the glycopeptides bind with lower
affinity to DR4 than the nonmodified peptide, as was observed
for A
q
(10), was not addressed, but could be considered as a
reasonable explanation. It should be emphasized, however, that
the glycosylated side chain is more likely to be oriented toward
the T cell receptor rather than the MHC, as we show here that
glycopeptide-specific T cell responses are critically dependent on
the galactose moiety. This assumption is further supported from
earlier studies using the A
q
-molecule (10) and is also indicated
from other studies using glycopeptide-specific T cells (38). As an
alternative or complementary explanation for a biased toler-
ance, it has been shown that relatively small differences in
availability or levels of autoantigens in vivo have great impact on
the size or status of the autoreactive T cell repertoire (39, 40).
Thus, different expression levels of the posttranslational forms
of the CII263–270 epitope in endogenous CII might modulate
the repertoire selection of CII-specific T cells. It should be
emphasized, however, that also glycopeptide-specific T cells
were influenced by endogenous huCII, although to a lesser
degree than nonglycosylated CII-specific T cells. The response to
the individual glycopeptides was only slightly weaker in huCII兾
DR4 mice than in DR4 mice, but the reduction was more obvious
when comparing the response to the intact CII protein, indicat-
ing anergized or low affinity T cells. In addition, with the use of
deoxy-peptides, we found that T cells from huCII兾DR4 mice
recognized the glycopeptide differently, indicating also that
huCII influences T cell repertoire selection.
Fig. 4. Specific recognition of CII-glycopeptides by human T cells. The
two-color flow-cytometry of in vitro-stimulated T cells from an RA-patient
shows fluorescence intensities for surface-binding of a FITC-labeled anti-CD4
antibody on the xaxis whereas the yaxis represents signal intensities for
intracellular staining with a phycoerythrin-labeled anti-IL-2 antibody. PBMCs
were cultured overnight without antigen, with tetanus toxoid, the K264兾270-
peptide, or the Gal264-peptide. The figures in the upper right quadrant of the
different panels represent the percentage of double-positive cells.
9964
兩
www.pnas.org兾cgi兾doi兾10.1073兾pnas.132254199 Ba¨cklund et al.
The mice used in this report are complex, because they carry
multiple genetic modifications on a mixed genetic background,
and translation of our in vitro findings to the human system
should be done with some caution. Nevertheless, despite the
complexity of the experimental set-up, our data on T cell
tolerance in huCII兾DR4 mice are intact and support earlier
findings (15, 20, 37, 41). Collectively, these data strongly suggest
that endogenous CII is physiologically exposed to the immune
system but does not lead to complete tolerance of a subset of
CII-specific T cells.
In summary, the results in a humanized animal model of RA
and the functional assays on cellular autoimmunity in RA
patients provide convincing evidence for incomplete T cell
tolerance to a set of closely related self-determinants that are
physiologically modified by posttranslational glycosylation. It
should be stressed, however, that our data do not prove that the
remaining CII-specific T cells are causative or even involved in
the pathogenesis of RA. Despite this fact, our findings are
interesting because they may constitute the missing link between
the CII-specific humoral response, observed in a substantial
fraction of RA patients, and the difficulties in the identification
of CII-specific T cells in RA patients. Furthermore, T cell
recognition of glycosylated CII in RA may be useful as a disease
progression marker and for classification of this heterogenous
disease.
We thank Carlos Palestro for taking good care of the animals, as well as
Alexandra and Caroline Treschow for critically reading the manuscript.
The work was supported by grants from the King Gustaf V’s 80-Year
Foundation, Kock’s Foundations, O
¨sterlund’s foundation, the Swedish
Association against Rheumatism, the Swedish Medical Research Coun-
cil, the Swedish Foundation for Strategic Research, the Go¨ran Gustafs-
son Foundation, the Deutsche Forschungsgemeinschaft (SFB 263,
project C3), the Bundesministerium fu¨r Bildung und Forschung (Med-
Net Entzu¨ndlich rheumatische Erkrankungen, project C 2.1; BMBF,
01GI9948), the European Commission (Bio4-98-0479), and the Karen
Elise Jensen and Novo Nordisk Foundations in Denmark.
1. Fugger, L., Rothbard, J. B. & Sonderstrup McDevitt, G. (1996) Eur. J. Im-
munol. 26, 928–933.
2. Rosloniec, E. F., Brand, D. D., Myers, L. K., Esaki, Y., Whittington, K. B.,
Zaller, D. M., Woods, A., Stuart, J. M. & Kang, A. H. (1998) J. Immunol. 160,
2573–2578.
3. Andersson, E. C., Hansen, B. E., Jacobsen, H., Madsen, L. S., Andersen, C. B.,
Engberg, J., Rothbard, J. B., McDevitt, G. S., Malmstro¨m, V., Holmdahl, R.,
et al. (1998) Proc. Natl. Acad. Sci. USA 95, 7574–7579.
4. Kotzin, B. L., Falta, M. T., Crawford, F., Rosloniec, E. F., Bill, J., Marrack, P.
& Kappler, J. (2000) Proc. Natl. Acad. Sci. USA 97, 291–296.
5. Kim, H. Y., Kim, W. U., Cho, M. L., Lee, S. K., Youn, J., Kim, S. I., Yoo,
W. H., Park, J. H., Min, J. K., Lee, S. H., et al. (1999) Arthritis Rheum. 42,
2085–2093.
6. Michae¨lsson, E., Malmstro¨m, V., Reis, S., Engstro¨m, Å., Burkhardt, H. &
Holmdahl, R. (1994) J. Exp. Med. 180, 745–749.
7. Corthay, A., Ba¨cklund, J., Broddefalk, J., Michae¨lsson, E., Goldschmidt, T. J.,
Kihlberg, J. & Holmdahl, R. (1998) Eur. J. Immunol. 28, 2580–2590.
8. Michae¨lsson, E., Andersson, M., Engstrom, Å. & Holmdahl, R. (1992) Eur.
J. Immunol. 22, 1819–1825.
9. Brunsberg, U., Gustafsson, K., Jansson, L., Michae¨lsson, E., Ahrlund Richter,
L., Pettersson, S., Mattsson, R. & Holmdahl, R. (1994) Eur. J. Immunol. 24,
1698–1702.
10. Kjellen, P., Brunsberg, U., Broddefalk, J., Hansen, B., Vestberg, M., Ivarsson,
I., Engstrom, Å., Svejgaard, A., Kihlberg, J., Fugger, L. & Holmdahl, R. (1998)
Eur. J. Immunol. 28, 755–767.
11. Dessen, A., Lawrence, C. M., Cupo, S., Zaller, D. M. & Wiley, D. C. (1997)
Immunity 7, 473–481.
12. Holmdahl, R., Andersson, M., Goldschmidt, T. J., Gustafsson, K., Jansson, L.
& Mo, J. A. (1990) Immunol. Rev. 118, 193–232.
13. Holmdahl, R., Vingsbo, C., Mo, J. A., Michae¨lsson, E., Malmstro¨m, V.,
Jansson, L. & Brunsberg, U. (1995) Immunol. Rev. 144, 109–135.
14. Fugger, L., Michie, S. A., Rulifson, I., Lock, C. B. & McDevitt, G. S. (1994)
Proc. Natl. Acad. Sci. USA 91, 6151–6155.
15. Malmstro¨m, V., Ho, K. K., Lun, J., Tam, P. P., Cheah, K. S. & Holmdahl, R.
(1997) Scand. J. Immunol. 45, 670–677.
16. Cosgrove, D., Gray, D., Dierich, A., Kaufman, J., Lemeur, M., Benoist, C. &
Mathis, D. (1991) Cell 66, 1051–1066.
17. Andersson, M. & Holmdahl, R. (1990) Eur. J. Immunol. 20, 1061–1066.
18. Broddefalk, J., Ba¨cklund, J., Almqvist, F., Johansson, M., Holmdahl, R. &
Kihlberg, J. (1998) J. Am. Chem. Soc. 120, 7676–7683.
19. Holm, B., Broddefalk, J., Flodell, S., Wellner, E. & Kihlberg, J. (2000)
Tetrahedron 56, 1579–1586.
20. Malmstro¨m, V., Ba¨cklund, J., Jansson, L., Kihlberg, J. & Holmdahl, R. (2000)
Arthritis Res. 2, 315–326.
21. Arnett, F. C., Edworthy, S. M., Bloch, D. A., McShane, D. J., Fries, J. F.,
Cooper, N. S., Healey, L. A., Kaplan, S. R., Liang, M. H., Luthra, H. S., et al.
(1988) Arthritis Rheum. 31, 315–324.
22. Assenmacher, M., Schmitz, J. & Radbruch, A. (1994) Eur. J. Immunol. 24,
1097–1101.
23. Assenmacher, M., Lohning, M., Scheffold, A., Manz, R. A., Schmitz, J. &
Radbruch, A. (1998) Eur. J. Immunol. 28, 1534–1543.
24. Stastny, P. (1978) N. Engl. J. Med. 298, 869–871.
25. Gregersen, P. K., Silver, J. & Winchester, R. J. (1987) Arthritis Rheum. 30,
1205–1213.
26. Diab, B. Y., Lambert, N. C., L’Faqihi, F. E., Loubet-Lescoulie, P., de Preval,
C. & Coppin, H. (1999) Immunogenetics 49, 36–44.
27. Rosloniec, E. F., Brand, D. D., Myers, L. K., Whittington, K. B., Gu-
manovskaya, M., Zaller, D. M., Woods, A., Altmann, D. M., Stuart, J. M. &
Kang, A. H. (1997) J. Exp. Med. 185, 1113–1122.
28. Trentham, D. E., Dynesius, R. A., Rocklin, R. E. & David, J. R. (1978) N. Engl.
J. Med. 299, 327–332.
29. Stuart, J. M., Postlethwaite, A. E., Townes, A. S. & Kang, A. H. (1980) Am. J.
Med. 69, 13–18.
30. Solinger, A. M., Bhatnagar, R. & Stobo, J. D. (1981) Proc. Natl. Acad. Sci . USA
78, 3877–3881.
31. Londei, M., Savill, C. M., Verhoef, A., Brennan, F., Leech, Z. A., Duance,
V., Maini, R. N. & Feldmann, M. (1989) Proc. Natl. Acad. Sci. USA 86,
636–640.
32. Ofosu-Appiah, W. A., Warrington, R. J. & Wilkins, J. A. (1989) Clin. Immunol.
Immunopathol. 50, 264–271.
33. Lacour, M., Rudolphi, U., Schlesier, M. & Peter, H. H. (1990) Eur. J. Immunol.
20, 931–934.
34. Yan, T., Burkhardt, H., Ritter, T., Broker, B., Mann, K. H., Bertling, W. M.,
von der Mark, K. & Emmrich, F. (1992) Eur. J. Immunol. 22, 51–56.
35. Snowden, N., Reynolds, I., Morgan, K. & Holt, L. (1997) Arthritis Rheum. 40,
1210–1218.
36. Berg, L., Ronnelid, J., Sanjeevi, C. B., Lampa, J. & Klareskog, L. (2000)
Arthritis Res. 2, 75–84.
37. Malmstro¨m, V., Michae¨lsson, E., Burkhardt, H., Mattsson, R., Vuorio, E. &
Holmdahl, R. (1996) Proc. Natl. Acad. Sci. USA 93, 4480–4485.
38. Rudd, P. M., Elliott, T., Cresswell, P., Wilson, I. A. & Dwek, R. A. (2001)
Science 291, 2370–2376.
39. Cibotti, R., Kanellopoulos, J. M., Cabaniols, J. P., Halle-Panenko, O., Kos-
matopoulos, K., Sercarz, E. & Kourilsky, P. (1992) Proc. Natl. Acad. Sci. USA
89, 416–420.
40. Akkaraju, S., Ho, W. Y., Leong, D., Canaan, K., Davis, M. M. & Goodnow,
C. C. (1997) Immunity 7, 255–271.
41. Malmstro¨m, V., Kjellen, P. & Holmdahl, R. (1998) J. Autoimmun. 11, 213–221.
Ba¨cklund et al. PNAS
兩
July 23, 2002
兩
vol. 99
兩
no. 15
兩
9965
IMMUNOLOGY
