Download Prevention of collagen-induced arthritis by gene delivery of

Gene Therapy (1998) 5, 1584–1592
 1998 Stockton Press All rights reserved 0969-7128/98 $12.00
http://www.stockton-press.co.uk/gt
Prevention of collagen-induced arthritis by gene
delivery of soluble p75 tumour necrosis factor receptor
RA Mageed1, G Adams1, D Woodrow2, OL Podhajcer3 and Y Chernajovsky1
1
Kennedy Institute of Rheumatology and 2Department of Histopathology, Charing Cross Hospital, Hammersmith, London, UK; and
Fundación Campomar, Universidad de Buenos Aires, Argentina
3
Collagen type II-induced arthritis (CIA) in DBA/1 mice can
be passively transferred to SCID mice with spleen B- and
T-lymphocytes. In the present study, we show that infection
ex vivo of splenocytes from arthritic DBA/1 mice with a
retroviral vector, containing cDNA for the soluble form of
human p75 receptor of tumour necrosis factor (TNF-R)
before transfer, prevents the development of arthritis, bone
erosion and joint inflammation in the SCID recipients.
Assessment of IgG subclass levels and studies of synovial
histology suggest that down-regulating the effector functions of T helper-type 1 (Th1) cells may, at least in part,
explain the inhibition of arthritis in the SCID recipients. In
contrast, the transfer of splenocytes infected with mouse
TNF-␣ gene construct resulted in exacerbated arthritis and
enhancement of IgG2a antibody levels. Intriguingly, infection of splenocytes from arthritic DBA/1 mice with a construct for mouse IL-10 had no modulating effect on the
transfer of arthritis. The data suggest that manipulation of
the immune system with cytokines, or cytokine inhibitors
using gene transfer protocols can be an effective approach
to ameliorate arthritis.
Keywords: collagen-induced arthritis; gene therapy; tumour necrosis factor p75 receptor
Introduction
The properties and outcome of T cell-dependent (TD)
immune responses can be predicted by the cytokine profile and phenotype of the responding T helper (Th) lymphocytes.1 This tenet is based primarily on the paradigm
that Th cells comprise functionally distinct populations
characterised by the pattern of cytokines they produce
and their effector functions.2 Furthermore, there is emerging evidence that the inappropriate induction of functionally distinct T cell populations may contribute
to many autoimmune, allergic and chronic infectious
diseases.3
In inflammatory joint diseases, such as rheumatoid
arthritis (RA) and CIA, there is increasing evidence for
the involvement of pro-inflammatory cytokines induced,
or produced by Th1-type cells in disease pathogenesis.
For example, TNF-␣ and IL-1 have been identified at sites
of disease and treatment with antibodies to these cytokines ameliorates arthritis.4–6 Furthermore, IFN-␥ has
been detected in culture supernatants of draining lymph
node cells taken during the early phases of CIA in DBA/1
mice suggesting a dominant role for Th1-type cells at the
onset of arthritis.7 In addition to the stimulatory effect of
cytokines produced or induced by Th1-type cells on the
inflammatory arm of the immune system, these cytokines
can directly cause cartilage and bone resorption, inhibit
the synthesis of proteoglycan and collagen in cartilage
and cause mild arthritis when injected directly into the
Correspondence: RA Mageed, Kennedy Institute of Rheumatology, 1
Aspenlea Road, Hammersmith, London, W6 8LH, UK
Received 18 December 1997; accepted 12 August 1998
joint space.8,9 Furthermore, TNF-␣ induces the release of
prostaglandin E2 and collagenase by synovial cells, contributes to fibrosis and facilitates inflammatory cell infiltration by promoting adhesion of neutrophils and lymphocytes to endothelial cells. This has led to the
hypothesis that excessive production of TNF-␣ may be
important in the pathogenesis of arthritis.10 One of the
early studies to validate this concept was performed
using in vitro cultures of mononuclear cells from RA synovial membranes.11 This study revealed that anti-TNF
antibodies diminished the production of other proinflammatory cytokines such as IL-1 and GM-CSF.
Consequently, this was followed by treatment of mice
with CIA and RA patients with antibodies to TNF. These
studies confirmed TNF-␣ as a useful target for therapeutic intervention.6 However, the precise way by which
blocking TNF-␣ ameliorated arthritis is not clear. Furthermore, there can be problems associated with antibodybased therapies, such as repeated intravenous administration to overcome the problem of absorption and clearance and immunogenicity of the antibody itself.
In the present study, we use a transfer model of CIA
to define how TNF-␣ influences arthritis, and assess the
value of a gene delivery system to block its actions. CIA
has been well characterised and the model is regarded
as useful to elucidate pathogenic mechanisms relevant to
human RA and identify potential targets for therapeutic
intervention.4 To examine the processes that underlie
CIA, a previous study showed that the disease can be
transferred from arthritic DBA/1 mice to SCID recipients
with spleen cells.12 However, the nature of the cellular
interactions and molecules that regulate disease
remained largely undefined. Furthermore, the effect of
manipulating the cytokine network by cytokine, or cyto-
Gene therapy of murine arthritis
RA Mageed et al
kine inhibitor administration have yet to be adequately
assessed. This has primarily been because of the need to
administer large amounts of recombinant cytokines or
cytokine inhibitors which have a short half-life, in
repeated doses. Gene delivery obviates the need to
express, purify and administer multiple doses of cytokine
or cytokine inhibitory proteins.
Using IgG2a and IgG1 antibodies as serological markers of Th1- and Th2-type effector functions, data from the
present study suggest that TNF-␣ may be involved in
CIA by directly inducing and/or enhancing Th1-type cell
effector functions. These findings imply that knowledge
of the immune response mode may be an important first
step in designing new immunotherapeutic strategies for
human diseases.
Results
Transfer of collagen-induced arthritis from DBA/1 to
SCID mice
All SCID recipients of unmodified splenocytes from
arthritic DBA/1 mice developed arthritis within 11–13
days after intraperitoneal injection. The arthritis was
characterised by variable synovitis mainly in the distal
interphalangeal joints (Figure 1). Splenocytes from naı̈ve
non-arthritic DBA/1 mice with, or without preincubation with CII in vitro, did not cause arthritis nor
produce anti-CII antibodies in the SCID recipients. Furthermore, infection of splenocytes from arthritic DBA/1
mice with a retroviral vector expressing the herpes simplex virus thymidine kinase (HSVtk) did not affect the
transfer of arthritis, paw thickness or the level of antiCII antibodies (data published in detail in Ref. 13). SCID
recipients that developed arthritis appeared otherwise
healthy and without clinical signs of graft-versus-host
disease (eg ruffled fur, hunched posture or diarrhoea).
Figure 1 Photomicrograph of a synovial section from a SCID recipient of
unmodified spleen cells from arthritic DBA/1 mice 15 days after cell transfer. The section was stained with haematoxylin and eosin and photographed at an original magnification of ×80. The Figure depicts a joint
showing synovial hyperplasia (some indicated by an arrow) and an underlying lymphocytic infiltrate with changes mainly in the distal interphalangeal joint (indicated by arrow heads).
The effect of ex vivo infection of splenocytes from
arthritic DBA/1 mice with retroviral constructs expressing
soluble human p75 TNF-R, mTNF-␣ or mIL-10 on the
transfer of arthritis to SCID recipients
Seventeen of the 20 SCID recipients of arthritic DBA/1
splenocytes infected with the retroviral construct for the
soluble TNF-R failed to develop clinical features of
arthritis although on histological examination they
showed a mild and acute form of synovitis (Table 1, pathology in Figure 2). These SCID mice had measurable levels of TNF-R in the serum until termination of the experiments (15 days after transfer). However, serum levels of
the p75 TNF-R did not correlate with paw swelling and
there was no consequent correlation with the clinical
response seen (Spearman’s non-parametric correlation,
r = 0.4362, P = 0.0545, Table 2). The decrease in paw swelling, however, correlated with lower levels of serum antiCII antibodies. In contrast, SCID recipients of DBA/1
splenocytes infected with the mTNF-␣ or the mIL-10 construct had more severe disease affecting many joints and
higher levels of anti-CII antibodies compared with SCID
recipients of unmodified splenocytes from arthritic
DBA/1 mice (Tables 1 and 3).
Pathological indices of arthritis in the SCID recipients
Results of histological examination of joints from arthritic
SCID recipients of unmodified splenocytes, or SCID
recipients of splenocytes infected with the soluble p75
TNF-R, mTNF-␣, or mIL-10 construct are summarised in
Table 1. Arthritic SCID recipients of unmodified DBA/1
splenocytes showed variable synovitis, which was mainly
present in the distal interphalangeal joints (Figure 1). This
synovitis was characterised by an increase in the number
of synovial lining layer cells and an intraluminal accumulation of fibrin and neutrophils. In contrast, mice receiving splenocytes infected with the soluble p75 TNF-R construct had a mild synovitis that consisted of few focal
increases in the number of synovial lining layer cells
(Figure 2). Mice given arthritic DBA/1 splenocytes
infected with mTNF-␣ or mIL-10 construct had severe
arthritis characterised by chondrocyte necrosis with focal
fragmentation of the articular cartilage together with
pannus formation, which particularly eroded through the
bone beneath the lateral edge of the cartilage (Figures 3
and 4, respectively). The synovial lining layer cells were
increased in numbers and fibrin and neutrophils accumulated both in the lining layer and the joint space. The surrounding tissue contained a prominent infiltrate of lymphocytes, macrophages, neutrophils and fibroblasts with
many dilated blood vessels engorged with red blood
cells. The number of macrophages was more evident in
the SCID recipients of splenocytes infected with mTNF␣ or mIL-10 construct than recipients of unmodified
splenocytes. In contrast, these changes were not seen in
the SCID recipients of splenocytes infected with the
soluble p75 TNF-R construct.
Levels of total IgG1 and IgG2a subclasses and IgG
antibody subclasses to CII in the SCID recipients
Levels of total IgG1 and IgG2a isotypes, and bovine CIIspecific IgG, IgG1 and IgG2a antibody subclasses were
measured in all SCID recipients. Levels of total IgG2a
were consistently higher than those of IgG1 in the
arthritic SCID recipients of unmodified DBA/1 splenocytes, or splenocytes infected with the mTNF-␣ or mIL-
1585
Gene therapy of murine arthritis
RA Mageed et al
1586
Table 1 Comparison of the pathological indices of arthritis in SCID recipients of splenocytes from arthritic DBA/1 mice without infection
with retroviral vectors or after infection with retroviral vector constructs for the soluble human p75 TNF-R, mTNF-␣ or mIL-10
Disease indices
Synovial lining cell
hyperplasia
Fibrin deposition
Chondrocyte necrosis
Fragmented cartilage
Neutrophils
Lymphocytes
Macrophages
Fibroblasts
Pannus
Location of observed changes
SCID recipients of
unmodified splenocytes
SCID recipients of
SCID recipients of
SCID recipients of
splenocytes infected with splenocytes infected with splenocytes infected with
mIL-10 construct
the soluble p75 TNF-R
mTNF-␣ construct
construct
++
+
−
−
+
++
++
++
−
Mainly in the distal
interphalangeal joints
+
+++
+++
−
−
−
−
−
−
−
−
+++
+++
+++
+++
+++
+++
+++
+++
Most joints
+++
+++
+++
+++
+++
+++
+++
+++
Most joints
Paws from arthritic mice (one or two paws from each of 20 mice) were removed and fixed in buffered formalin. For the purpose of
the microscopic examination, the paws were decalcified with EDTA in buffered formalin and embedded in paraffin. Six micrometrethick sections were stained with haematoxylin and eosin and disease parameters assessed microscopically in comparison with sections
obtained from control untreated SCID mice. No histopathological changes are indicated by the − sign. Occasional changes are indicated
as +; changes seen frequently are indicated as ++; widespread changes are indicated as +++.
Figure 2 Photomicrograph of a synovial section from a SCID recipient of
arthritic DBA/1 mice splenocytes infected with the soluble human p75
TNF-R construct. The section, from a SCID mouse killed 15 days after
cell transfer, shows mild synovial lining layer hyperplasia (indicated by
arrows). The section was stained with haematoxylin-eosin and photographed at a magnification of ×80.
10 constructs (Table 3), with two exceptions. In the
experiments presented in this manuscript, two SCID
recipients of non-infected splenocytes had very high levels of IgG1 and to a lesser extent, IgG2a. The increase in
the level of total IgG in these two SCID recipients was
also reflected in the level of specific anti-CII antibodies.
However, the increase in the level of anti-CII antibodies
did not strictly reflect the change in the ratio of total
IgG2a to IgG1. The reason for the disparity in the level
of IgG1 and IgG2a in these two SCID recipients compared with the rest is unclear, but is not likely to be due
to the mice being ‘leaky’. All possible leaky SCID mice
were excluded from the transfer experiments before
initiation of the study. The levels of IgG2a anti-bovine
CII antibodies were significantly higher than IgG1 in all
the arthritic SCID recipients, with one exception (one of
the aforementioned two mice). When the data were
expressed as ratios of IgG2a to IgG1 (total and CIIspecific) the values were always disproportionately high
in the three arthritic groups. These values and ratios were
reversed in favour of IgG1 (total and CII-specific) in the
SCID recipients of splenocytes infected with the soluble
p75 TNF-R construct (Table 3). This coincided with the
prevention of arthritis transfer and amelioration of histopathological features of the disease. The experiments
were repeated on two other occasions confirming the
reversal in the ratio of IgG2a:IgG1 in favour of IgG1 in
SCID recipients of DBA/1 splenocytes infected with the
soluble TNF-R construct. The data suggested that there
was a dominance of a Th1-driven IgG2a antibody production in SCID recipients of unmodified arthritic
DBA/1 splenocytes, or splenocytes infected with mTNF␣ or mIL-10. The latter observation was somewhat surprising in view of the effect of IL-10 on down-regulating
macrophage functions, including T cell activation and
Th1-type cytokine production (eg TNF-␣).14 The explanation for this finding may lie in the relatively low quantity of IL-10 produced by cells infected with the construct
compared with levels required for an effective action on
arthritis.15 Alternatively, it is possible that the exacerbated arthritis may be due to the effect of mIL-10 on B
cell differentiation and antibody production. It is noteworthy in this respect that levels of both total IgG1 and
IgG2a, and of anti-CII antibodies were generally higher
in the SCID recipients of splenocytes infected with the
mIL-10 construct compared with the other groups.
Discussion
Collagen-induced arthritis in DBA/1 mice has been
extensively used to study clinical and therapeutic aspects
of human RA.4,16 Clinically, both diseases involve sustained proliferative synovitis with cartilage and bone
destruction.17 Immunologically, CIA has a number of
Gene therapy of murine arthritis
RA Mageed et al
Table 2 Lack of correlation between serum levels of the soluble
human p75 TNF-R, IgG anti-CII antibodies and paw thickness
in SCID recipients of splenocytes from DBA/1 mice infected
with the retroviral construct for soluble p75 TNF-R
Serum concentration of
soluble p75 TNF-R
(pg/ml)
3832
3344
4042
564
2409
1203
7701
4400
5450
⬍20
6650
565
300
515
590
⬍20
5200
⬍20
5610
1675
Mean ± s.d.
(median)
Level of anti-bovine
CII in SCID
recipients (mg/ml)
Paw thickness of
SCID recipients
(mm)
0.04
0
0
0.01
0
0
0
7.7
18.6
1.8
13.3
2.0
1.6
1.7
1.8
0.5
0.9
0
0.6
0.3
2.54 ± 4.98
(0.550)
2.0
2.1
2.2
2.1
1.9
2.0
1.9
2.7a
2.9a
2.0
3.2a
2.2
1.9
1.7
1.9
2.0
2.1
1.9
2.0
1.9
2.13 ± 0.37
(2.0)
Each of the 20 SCID recipients included in the Table (from three
separate experiments) received 5 × 107 splenocytes, in single cell
suspension, from arthritic DBA/1 mice intraperitoneally after
48 h incubation in vitro with 100 ␮g/ml bovine CII. Four hours
before culture termination, the cells were centrifuged and resuspended in culture medium containing supernatants from the
packaging cell line.
The mean ± standard deviation of paw thickness in control
SCID mice receiving non-infected splenocytes and mice receiving splenocytes infected with retroviral constructs containing
HSVtk, mTNF-␣ or mIL-10 genes were: eight mice receiving
non-infected splenocytes, 2.69 ± 0.2 (median 2.7; range 2.3–3.0);
four mice receiving splenocytes infected with a retroviral gene
construct for HSVtk, 2.76 ± 0.2 (median 2.75; range 2.6–3.0); four
mice receiving splenocytes infected with the retroviral construct
for mTNF-␣, 2.63 ± 0.16 (median 2.6; range 2.4–2.8); 10 mice
receiving the retroviral construct containing mIL-10 gene,
2.75 ± 0.21 (median 2.7; range 2.3–3.1). Paw thickness was significantly different between control mice receiving non-infected
splenocytes from arthritic DBA/1 mice and those receiving similar splenocytes infected with the soluble p75 TNF-R (Mann–
Whitney test; P ⬍ 0.001).
Spearman’s non-parametric correlation coefficient for serum
levels of soluble p75 TNF-R with anti-CII antibodies and paw
thickness were: r = 0.0702 (P = 0.769) and r = 0.4362
(P = 0.0545), respectively.
a
These three mice did not show beneficial therapeutic effect and
developed arthritis.
similarities with RA, eg the requirement for genetic predisposition factors involving both MHC- and non-MHCrelated gene loci. Furthermore, attainment of reactivity
with self antigens, native CII in CIA, is important to the
development of both diseases.18 More recently, passive
transfer of CIA from DBA/1 to SCID recipients has
shown that interactions between antigen presenting cells,
B and T cells are crucial for establishing disease.12
In the present study, we demonstrate that the transfer
of arthritis from arthritic DBA/1 mice to SCID recipients
can be blocked by modifying some of the effector functions of Th1- and/or Th2-type cells.1,2 This apparent
modification of the immune response was achieved by
infecting lymphocytes with a retroviral construct carrying cDNA for the soluble form of human p75 TNF-R.
Alteration in the ratio of total IgG subclasses and of antiCII antibodies in favour of IgG1 compared with IgG2a
levels and histological changes within the synovium suggest that the prevention of arthritis was associated with
down-regulating the effector function of Th1-type cells.
A link between TNF-␣ and Th1-type responses was also
observed in TNF-␣ knock-out mice, which were shown
to be incapable of mounting IgG2a antibody responses
when challenged with antigens.19 However, it is not yet
clear whether down-regulation of IgG2a production
shown in our study resulted from a direct inhibitory
effect on Th1 cells or indirectly through the induction of
Th2 cells.
In immune responses to exogenous antigens, Th lymphocytes mature along two alternative pathways into
Th1-type cells, which play the major role in cell-mediated
immunity by producing IL-2, TNF-␣ and IFN-␥, or Th2type cells which are involved in antibody mediated
responses by producing IL-4, IL-5 and IL-10.1,2 Both Th
subpopulations cross-regulate each other through the
production of cytokines. For example, IFN-␥ induces
Th1-type cells and inhibits the differentiation and functions of Th2-type cells.1 Conversely, IL-4 induces Th2type cells and inhibits Th1-type cell proliferation and
opposes the effects of IFN-␥.19 Therefore, reciprocal regulation occurs between Th1/Th2-type cells. The extent of
the reciprocal action of Th1- and Th2-type cells in the
CIA on each other and influence on the disease process
is only emerging. A recent study of in vitro cultured
lymph node cells from immunised DBA/1 mice showed
increased IFN-␥ production during the first 6 days of
immunisation.7 However, a role for imbalances in the
cytokine network in CIA in favour of cytokines induced,
or produced by Th1-type cells has long been suggested.20,21 For example, it was shown that mice transgenic for human TNF-␣ develop arthritis.20 Furthermore,
natural levels of IL-1 are chronically increased in mice
with CIA and lymphocytes from these mice produce significantly increased amounts of IL-2.21 The contribution
of TNF-␣ and IL-1 (induced by Th1-type cells) to cartilage
and bone destruction and the recruitment of immune
cells into the synovium are also well known.4,8 Most of
these studies stressed that the role of TNF-␣ and IL-1 in
arthritis was due to the pro-inflammatory nature of these
two cytokines, their ability to induce prostaglandin E2
in culture and enhancement of bone and cartilage
destruction.
The experiments described here suggest that TNF-␣
also contributes to the immune-mediated arthritis
through the induction of IgG2a production, an isotype
known to be induced by Th1-type cells. The data also
showed that systemic levels of soluble p75 TNF-R did not
reflect a positive therapeutic effect. Studies over the last
few years have shown that two distinct forms of TNF
exist, a membrane-expressed form (mTNF) and a soluble
form which is derived from proteolytic cleavage of the
mTNF.22 Despite this relationship, however, it has been
established that mTNF is significantly more bioactive
than soluble TNF and that engagement of mTNF confers
intracellular signalling typical of TNF responses such as
1587
Gene therapy of murine arthritis
RA Mageed et al
1588
Table 3 Levels of total serum IgG1 and IgG2a subclasses and IgG1 and IgG2a antibodies to bovine CII in the serum of SCID recipients
of arthritic DBA/1 splenocytes
SCID recipient of
A. Unmodified splenocytes
Mouse
1
2
3
4
5
mean ± s.e.m.
B. Splenocytes with mTNF-␣
construct
Mouse
1
2
3
4
5
mean ± s.e.m.
C. Splenocytes with mIL-10
construct
Mouse
1
2
3
4
5
mean ± s.e.m.
D. Splenocytes with p75
TNF-R construct
Mouse
1
2
3
4
5
mean ± s.e.m.
Total IgG1
(␮g/ml)
Total IgG2a
(␮g/ml)
IgG1 anti-CII
(␮g/ml)
IgG2a antiCII (␮g/ml)
Ratio of total
IgG2a:IgG1
Ratio of anti-CII
IgG2a:IgG1
16
17
19
8318
8260
928
771
876
1882
2854
0.4
2.2
0.2
182
380
14
3
45
210
236
58.0
45.4
46.1
0.2
0.4
30.0 ± 27.6
35.0
1.4
225.0
1.2
0.6
52.6 ± 97.5
35
49
31
166
135
786
992
1613
2214
926
11
2
2
98
51
35
23
39
472
312
22.4
20.4
52.4
13.3
6.9
23.0 ± 17.3
3.4
15.5
19.4
4.8
6.1
9.2 ± 6.6
38
112
258
179
56
1130
2278
1894
1997
1164
11
24
92
62
2
57
223
417
287
23
30.0
20.3
7.3
11.1
21.0
18.0 ± 8.9
5.1
9.5
4.5
4.7
15.1
7.02 ± 3.2
886
732
585
870
1032
241
316
127
174
357
110
209
48
64
95
5
14
11
23
29
0.3
0.4
0.2
0.2
0.4
0.3 ± 0.1
0.05
0.06
0.22
0.35
0.30
0.2 ± 0.14
The level of total IgG1 and IgG2a subclasses and bovine CII-specific IgG1 and IgG2a were determined in ELISA. For total IgG1 and
IgG2a, the plates were sensitised with anti-mouse ␥-chain subclass-specific antibodies while for CII-specific antibodies the plates were
sensitised with bovine CII. Concentration of bound IgG isotypes was determined by extrapolation from standard curves constructed
from known inputs of purified IgG1, IgG2a or bovine CII-specific antibodies. The ratios of IgG2a anti-CII to IgG1 anti-CII antibodies
in SCID recipients of unmodified (non-infected) splenocytes (group A) or splenocytes infected with a construct for mTNF-␣ or mIL-10
gene were significantly higher than those obtained for SCID recipients of splenocytes infected with the soluble p75 TNF-R gene construct
(group D) (Mann–Whitney test; P ⬍ 0.05). The ratios of total IgG2a to IgG1 were different in the different groups but the difference
between group A and D was not significant in the Mann–Whitney test (P = 0.13). However, multivariate analysis using Pillai’s trace
showed that the four groups were significantly different in the level of total and anti-CII antibodies (F = 2.652, P = 0.009). The mean
and standard error of the mean (mean ± s.e.m.) for the ratio of IgG2a:IgG1 for each group is given. This experiment was repeated on
two other occasions with similar outcome (data not shown).
cytotoxicity and B cell activation.23 The lack of a direct
relationship between serum levels of soluble p75 TNF-R
and the therapeutic effects seen may suggest that the soluble receptor functions by controlling the bioavailability
of mTNF in the microenvironment of activated T cells.
This implies that manipulation of local TNF production
may be required to achieve a therapeutic effect. It is, however, unclear at this stage whether the soluble form of
the p75 TNF-R is modulating the biological effects of
mTNF (and lymphotoxin) by blocking signaling through
p75, p55 or both receptors. Earlier studies have suggested
that the contribution of the soluble form of p75 TNF-R to
cellular responses was to act as a buffer for excessive soluble TNF, while the membrane form of the p75 TNF-R
enhanced p55 TNF-R signaling after binding to TNF
(ligand passing).24 However, recent data suggest that
mTNF can directly induce cellular responses through the
membrane p75 receptor and independent of the p55 TNFR.25 The precise pathway through which the soluble p75
TNF-R inhibits Th1-type cells, therefore, requires further
investigation. The data, nevertheless, suggest that the
observed effect of soluble p75 TNF-R may possibly be
through blocking cognate immune cell interactions. This
suggestion is in line with previous findings that antibodies to the CD40 ligand (gp39) and anti-CD4 antibodies, which down-regulate Th1 cells, are effective in
treating arthritic DBA/1 mice.26,27
Histological examination of joints from arthritic and
treated mice revealed a reduction in cellularity of the
synovium of mice receiving splenocytes infected with the
soluble p75 TNF-R construct suggesting that a factor in
regulating arthritis may be the prevention of extrava-
Gene therapy of murine arthritis
RA Mageed et al
Figure 3 Photomicrograph of a synovial section from a SCID recipient of
arthritic DBA/1 mice splenocytes infected with the mTNF-␣ construct.
The section is from a SCID mouse killed 15 days after cell transfer. The
articular cartilage contains necrotic chondrocytes (indicated by arrow
heads) and there is pannus eroding the cartilage edge (indicated by
arrows). The joint space contained fibrin and neutrophils with a surrounding chronic synovitis. The section was stained with haematoxylin-eosin
(×80 magnification).
Figure 4 Photomicrograph of a synovial section from a SCID recipient of
arthritic DBA/1 mice splenocytes infected with the mIL-10 construct 15
days after cell transfer. The fragmented articular cartilage contains many
necrotic chondrocytes (indicated by arrow heads). There is also pannus
eroding on both sides of the joint (indicated by arrows). The section was
stained and photographed as for the other Figures.
sation of immune cells to the synovium. It was not clear
from previous studies whether leucocyte migration was
inhibited through manipulation of the cytokine network.
It was, however, suggested that TNF-␣ can directly
influence the migration of leucocytes through its effect
on endothelial cell expression of intercellular adhesion
molecule-1 (ICAM-1), vascular cell adhesion molecule-1
(VCAM-1) and E-selectin and indirectly through the
induction of chemokines such as IL-8.28 Furthermore,
studies of human RA patients treated with a chimeric
monoclonal anti-TNF antibody decreased serum levels of
E-selectin and ICAM-1 levels suggesting that this may
reflect diminished activation of endothelial cells in the
synovial microvasculature leading to reduced migration
of leucocytes into the synovial joints.29 These studies all
point to the possibility that TNF-␣ is a potent cytokine
in inducing chemotactic activity and leucocyte recruitment into the joint. Our study provides some evidence
for these suggestions and suggests that migration of
immune cells into the joint and cartilage and bone
destruction may have been prevented by suppressing
immune activation through the TNF-␣ pathway. It is
important, however, to point out that the study can not
exclude the possibility that the reduction in synovial
cellularity may be due to an increase in apoptosis. This
is clearly an area that needs further investigation.
In view of these data and of previous studies, we
hypothesise that the pathogenesis of CIA in SCID recipients may be dependent, at least during the initial phases,
on a Th1-type response in which TNF-␣ is involved. Furthermore, it appears that the transfer of arthritis can be
inhibited when the immune response is deviated to a
Th2-type response. It remains, however, to be conclusively determined whether the biological effect of the p75
TNF-R was secondary to its proposed inhibitory effect on
Th1 cells and cytokine production, or to the blocking of
other TNF-␣ pathways. Although manipulation of the
cytokine network appears critical in disease prevention,
it is unlikely that cytokines alone can precipitate diseases.
It is known, for example, that cytokines are incapable on
their own of overcoming genetic resistance to CIA in
mouse strains that do not express H-2q or H-2r MHC
haplotypes, and that cytokines failed to facilitate the
development of arthritis in CD4-deficient mice.30 This is,
however, contrasted by the recent finding that mice
transgenic for human TNF-␣ develop arthritis, although
this arthritis does not appear to be mediated by the
immune system.20 Thus it is possible that in CIA in the
DBA/1 mouse deregulated Th1-type cytokine production
may lead to inappropriate induction of autoreactive Th1type cells in susceptible mice. The induction of autoimmune diseases by these cells is likely to be contingent
on excessive IFN-␥ production, which can lead to
changes in the cleavage pattern of self-antigens, upregulation of MHC antigens and increase presentation of
self-peptides.3 Th2 cells, on the other hand, can act as
regulatory T cells by limiting the response to self-antigens. It is interesting in this respect to note that administration of IFN-␥ at the time of immunisation with CII has
been shown to augment the incidence of arthritis and
accelerates its onset.31 Furthermore, there is evidence for
an association between the early stages of CIA in DBA/1
mice and IFN-␥ production by lymph node cells.7 These
findings suggest that the role of IFN-␥ in enhancing Th1
responses to self-peptides is perhaps an important aspect
of arthritis and that other Th1-associated effector functions, such as extravasation of inflammatory cells, are
also crucial for the inflammation and tissue destruction.
Materials and methods
Animals
Eight- to 12-week-old DBA/1 and CB-17 scid/scid (SCID)
mice used in the study were bred and maintained under
germ-free conditions at the Kennedy Institute of Rheumatology. ‘Leaky’ SCID mice, producing detectable levels of
immunoglobulin (⬎5 ng/ml), were identified before the
experiments using ELISA and excluded from the study.
1589
Gene therapy of murine arthritis
RA Mageed et al
1590
Induction and transfer of arthritis
Type II collagen (CII) was purified from bovine articular
cartilage as previously described.4 Arthritis was induced
in male DBA/1 mice by intradermal injection of 100 ␮g of
the purified CII emulsified in Freund’s complete adjuvant
(Difco Laboratories, East Mosley, UK) at the base of the
tail. Onset of arthritis was macroscopically visible as paw
swelling 3 weeks after immunisation. Spleens from
arthritic DBA/1 mice were aseptically removed and used
to transfer arthritis to SCID mice. Separated splenocytes
were washed with Dulbecco’s modified Eagle’s medium
(DMEM; Gibco-BRL, Paisley, UK) containing 5% heat
inactivated foetal calf serum and a mixture of penicillin
(100 units/ml) and streptomycin (100 ␮g/ml) before
incubation with bovine CII for 48 h in culture before use
for the transfer of arthritis. Supernatants from the packaging cell line CRIP transfected with the retroviral gene
constructs were added to the culture 4 h before their
inoculation into the SCID recipients.
Clinical features of arthritis were monitored by
assessing paw swelling and the number of swollen paws
using callipers once every 2 days. Onset of arthritis in the
SCID recipients was generally observed at day 11. Severity of arthritis was assessed according to established clinical scores.4,12 The results were expressed as paw thickness
in millimetres (mm) and compared with paw thickness
before disease onset.
Histopathology
Paws (one to two per mouse) were removed after death,
fixed in 10% (wt/vol) buffered formalin, and decalcified
with EDTA in buffered formalin (5.5%). The paws were
embedded in paraffin, sectioned and stained with haematoxylin and eosin for microscopic evaluation of cellular
and architectural changes in a blinded fashion. The severity of arthritis in each joint was classified as mild: mild,
minimal synovitis, cartilage loss and bone erosions limited to discrete foci; moderate, synovitis and erosions
present but normal joint architecture intact; severe, synovitis, extensive erosions and joint architecture disrupted.
Antibody detection
Anti-CII antibody levels were determined by ELISA.4,12,16
Microtitre ELISA plates (ICN Biomedicals, Thame, UK)
were sensitised with native bovine CII (at 2 ␮g/ml) in
Tris-buffered saline (TBS) pH 7.4 by incubation at 4°C
overnight. Non-occupied sites were quenched with a
solution of 2% (wt/vol) casein in TBS for 1 h at room
temperature and then serial dilutions of test sera in
phosphate-buffered saline (PBS) containing 0.05% Tween
20 (PBS/Tween) were added to the wells and incubated
for 1 h at 37°C. Bound total IgG, or IgG subclasses 1 and
2a, were revealed with horseradish peroxidase (HRP)
conjugated sheep anti-mouse IgG-Fc, IgG1 or IgG2a antibodies, respectively (Binding Site, Birmingham, UK). IgG
or IgG subclass antibodies were quantified by extrapolation using standard curves constructed from optical
density values obtained for known inputs of purified
total IgG, IgG1 or IgG2a anti-collagen antibodies.
Total IgG1 and IgG2a levels were determined using a
capture ELISA with sheep anti-mouse IgG1 or IgG2a (at
2 ␮g/ml in PBS; Binding Site) sensitised plates. Doublefolding dilutions of serum samples starting from 1:200
were added to each well and incubated for 2 h at 37°C.
A standard control containing known inputs of affinity-
purified mouse IgG1 or IgG2a proteins (Binding Site)
were added and titrated to give known concentration of
IgG content to construct a standard curve. Bound IgG1
or IgG2a were detected using HRP-conjugated antimouse IgG1 or IgG2a. Quantification was achieved by
extrapolation from standard curves constructed as above.
The sensitivity limits of the assays used in the study were
about 5 ng/ml of IgG1 or IgG2a.
Retroviral vectors and constructs
Construction of the retroviral vector pBabe-pAGO-Neo,
expressing the Herpes simplex virus thymidine kinase
gene has been described before.13 The soluble human p75
TNF-R construct was prepared using cDNA for the extracellular domain of human p75 TNF-R obtained from a
baculovirus clone (pVL139 hp75).13 The fragment was
digested with XbaI, blunted with Klenow fragment of
DNA polymerase, ligated to BamHI linkers and then
digested with NcoI and BamHI and cloned into the NcoIBamHI sites of the retroviral vector MFG. The packaging
cell line CRIP was transfected with this retroviral vector
and the selectable marker pSV2neo at a molar ratio of
20:1. Clones resistant to geneticin (G418; Gibco-BRL) at 1
mg/ml were isolated and a high producer clone obtained
after multiple infections with transient ecotropic virus
obtained from infection of BOSC23 cells with this virus.
This construct expressed 260 ng/ml p75-TNF-R per 106
cells in 24 h, as determined by a previously described
ELISA.32 Subsequent measurement of serum levels of
human TNF-R in the SCID recipient of splenocytes
infected with the p75 TNF-R were determined using the
same ELISA protocol.
The mIL-10 construct was prepared using mouse IL-10
(mIL-10) cDNA (kindly provided by Dr K Moore, DNAX,
Palo Alto, CA, USA) obtained from a 1.2 kb XhoI fragment containing the whole of the open reading frame of
the IL-10 cDNA, and cloned into the SalI site of the retroviral vector pBabe puro.33 The resulting retroviral construct (named mIL-10 Babe puro) was packaged after calcium phosphate transfection into GPenvAM12 cells34 and
cell clones resistant to 1.5 ␮g/ml puromycin (Sigma) in
culture medium were pooled and used as a source of
viral supernatant. These cells produced 2–4 ng/ml mouse
IL-10 per 106 cells in 24 h, as determined by ELISA using
monoclonal antibodies with specificity for mIL-10
(Pharmingen, San Diego, CA, USA).
Mouse TNF-␣ (mTNF-␣) construct was prepared using
a 2.8 kb genomic clone of mouse TNF-␣, kindly donated
by Drs N Sarvetnick (Scripps Research Institute, La Jolla,
CA, USA) and S Thompson (Cantab Pharmaceuticals,
Cambridge, UK), isolated from Bluescript (IIKS) phagemid as an XhoI to BamHI segment. The fragment also contained a hepatitis B virus polyadenylation signal. After
filling the ends with Klenow, this fragment was cloned
into the ‘filled-in’ EcoRI site of pBabe bleo. The hepatitis
polyadenylation signal was removed by digestion with
EcoRI. The vector was named mTNF-Babe bleo. Packaged
retrovirus was obtained as above after selection in
25 ␮g/ml pleomycin (Sigma).
Tissue culture and infection of spleen cells
The packaging cell line GPenvAM1234 was maintained in
DMEM, 10% new-born calf serum with glutamine, penicillin and streptomycin. One day before transfection with
the retroviral constructs, cells were trypsinised and
Gene therapy of murine arthritis
RA Mageed et al
plated out at 5 × 105 cells per 9-cm diameter dish. Four
hours before DNA was added as a calcium phosphate
precipitate, the medium was changed and fresh medium
added. The transfection was carried out as previously
described.13 Transfected cells were selected in 1 mg/ml
G418 and more than 30 clones were pooled as a population and maintained in G418 at 0.5 mg/ml. Viral supernatants were collected overnight from confluent 15-cm
plates each containing 12 ml medium with 10% foetal
bovine serum (FBS) or 10% new-born calf serum without
the selective drug. The supernatants were centrifuged at
400 g for 10 min to remove cellular debris, transferred to
sterile tubes and snap frozen on dry ice. These supernatants were stored at −70°C before use.
Spleens were removed and dissected aseptically and
put into DMEM medium containing 10% FBS. The
spleens were sieved through a cell strainer and washed
with Hank’s solution containing 2% FBS. The cells were
centrifuged and the pellet resuspended in 2 ml of Tris
buffer containing 0.15 m ammonium chloride to lyse red
blood cells. The cells were centrifuged and washed once
in the same medium and passed through a sieve to
remove cell clumps. The cells were plated at 5 × 106
cells/ml in DMEM with 10% FBS and CII at 100 ␮g/ml
and incubated for 48 h at 37°C in a humidified incubator
with 5% CO2 in the air. When necessary, cells were
infected with retroviruses as follows: at the start of the
last 4 h of incubation with CII, the cells were centrifuged
at 500 g for 5 min, resuspended at 2 × 106 in viral supernatant supplemented with DEAE dextran at a final concentration of 70 ␮g/ml. After 4 h, the cells were centrifuged, washed twice and resuspended in DMEM without
serum at 5 × 107/ml and 1 ml injected intraperitoneally
into each SCID recipient.
Statistical analyses
Differences between control mice receiving non-infected
splenocytes and the treated group receiving soluble p75
TNF-R gene construct in paw thickness and ratio of
IgG2a:IgG1 were determined using the two-sided nonparametric Mann–Whitney test for non-paired samples.
Correlation coefficients were determined using Spearman’s non-parametric correlation. Both analyses were
performed using the GraphPad Instat software
(GraphPad Software, San Diego, CA, USA). The effect of
the different treatments on the level of total IgG1, total
IgG2a, IgG1 and IgG2a anti-CII was compared using Pillai’s trace of multivariate analysis of variance, which is
the most robust and powerful test. The test was performed using the Statistical Package for Social Sciences
(SPSS) software for Windows.
Acknowledgements
We would like to thank Mr P Connolly for preparing the
histology sections, Dr Kevin Moore, for providing mIL10 cDNA, Drs Sarvetnick and SJ Thompson for providing
the genomic clone of murine TNF-␣ used in the study.
The study was supported by the Arthritis and Rheumatism Council of Great Britain. Dr O Podhajcer was supported by a grant from the Antorchas Foundation and
the British Council (Argentina).
References
1 Paul WE, Seder RA. Lymphocyte responses and cytokines. Cell
1994; 76: 241–251.
2 Mosmann TR, Coffman RL. TH1 and TH2 cells: different patterns of lymphokine secretion lead to different functional
properties. Ann Rev Immunol 1989; 7: 145–173.
3 Liblau RS, Singer SM, McDevitt HO. Th1 and Th2 CD4+ T cells
in the pathogenesis of organ-specific autoimmune diseases.
Immunol Today 1995; 16: 34–38.
4 Williams RO, Feldmann M, Maini RN. Anti-tumor necrosis factor ameliorates joint disease in murine collagen-induced
arthritis. Proc Natl Acad Sci USA 1992; 89: 9784–9788.
5 Chu CQ et al. Detection of cytokines at the cartilage/pannus
junction in patients with rheumatoid arthritis: implications for
the role of cytokines in cartilage destruction and repair. Br J
Rheumatol 1992; 31: 653–661.
6 Elliott M J et al. Treatment of rheumatoid arthritis with chimeric
monoclonal antibodies to tumor necrosis factor alpha. Arthritis
Rheum 1993; 36: 1681–1690.
7 Mauri C, Williams RO, Walmsley M, Feldmann M. Relationship
between Th1/Th2 cytokine patterns and the arthritogenic
response in collagen-induced arthritis. Eur J Immunol 1996; 26:
1511–1518.
8 Bertolini DR et al. Stimulation of bone resorption and inhibition
of bone formation in vitro by human tumour necrosis factors.
Nature 1986; 319: 516–518.
9 Pettipher ER, Higgs GA, Henderson B. Interleukin 1 induces
leukocyte infiltration and cartilage proteoglycan degradation in
the synovial joint. Proc Natl Acad Sci USA 1986; 83: 8749–8753.
10 Feldmann M et al. Are imbalances in cytokine function of importance in the pathogenesis of rheumatoid arthritis. In: Davies ME,
Dingle JT (eds). Immunopharmacology of Joints and Connective
Tissue. Academic Press: London, 1994, pp 119–135.
11 Brennan FM et al. Inhibitory effect of TNF alpha antibodies on
synovial cell interleukin-1 production in rheumatoid arthritis.
Lancet 1989; 2: 244–247.
12 Taylor PC, Plater-Zyberk C, Maini RN. The role of the B cells
in the adoptive transfer of collagen-induced arthritis from
DBA/1 (H-2q) to SCID (H-2d) mice. Eur J Immunol 1995; 25:
763–769.
13 Chernajovsky Y et al. Inhibition of transfer of collagen-induced
arthritis into SCID mice by ex vivo infection of spleen cells with
retroviruses expressing soluble tumor necrosis factor receptor.
Gene Therapy 1995; 2: 731–735.
14 Tripp CS, Wolf SF, Unanue ER. Interleukin 12 and tumor
necrosis factor alpha are costimulators of interferon gamma production by natural killer cells in severe combined immunodeficiency mice with listeriosis, and interleukin 10 is a physiologic
antagonist. Proc Natl Acad Sci USA 1993; 90: 3725–3729.
15 Walmsley M et al. Interleukin-10 inhibition of the progression
of established collagen-induced arthritis. Arthritis Rheum 1996;
39: 495–503.
16 Williams RO, Mason LJ, Feldmann M, Maini RN. Synergy
between anti-CD4 and anti-tumor necrosis factor in the amelioration of established collagen-induced arthritis. Proc Natl Acad Sci
USA 1994; 91: 2762–2766.
17 Kaklamanis PM. Experimental animal models resembling rheumatoid arthritis. Clin Rheumatol 1992; 11: 41–47.
18 Holmadhl R et al. Origin of the autoreactive anti-type II collagen
response. II. Specificities, antibody isotypes and usage of V gene
families of anti-type II collagen B cells. J Immunol 1989; 142:
1881–1886.
19 Hsieh CS et al. Differential regulation of T helper phenotype
development by interleukins 4 and 10 in an alpha beta T cell
receptor transgenic system. Proc Natl Acad Sci USA 1992; 89:
6065–6069.
20 Keffer J et al. Transgenic mice expressing human tumour
necrosis factor: a predictive genetic model of arthritis. EMBO J
1991; 10: 4025–4031.
21 Phadke K, Carlson DG, Gitter BD, Butler LD. Role of interleukin
1591
Gene therapy of murine arthritis
RA Mageed et al
1592
22
23
24
25
26
27
28
1 and interleukin 2 in rat and mouse arthritis models. J Immunol
1996; 136: 4085–4091.
Kriegler M et al. A novel form of TNF/cachectin is a cell surface
cytotoxic transmembrane protein: ramifications for the complex
physiology of TNF. Cell 1988; 53: 45–53.
Aversa G, Punnonen J, de Vries JE. The 26-kDa transmembrane
form of tumor necrosis factor alpha on activated CD4+ T cell
clones provides a costimulatory signal for human B cell activation. J Exp Med 1993; 177: 1575–1585.
Tartaglia LA et al. The two different receptors for tumor necrosis
factor mediate distinct cellular responses. Proc Natl Acad Sci USA
1991; 88: 9292–9296.
Grell M et al. The transmembrane form of tumor necrosis factor
is the prime activating ligand of the 80 kDa tumor necrosis factor receptor. Cell 1995; 83: 793–802.
Durie FH et al. Prevention of collagen-induced arthritis with an
antibody to gp39, the ligand for CD40. Science 1993; 261:
1328–1330.
Ranges GE, Sriram S, Cooper SM. Prevention of type II collageninduced arthritis by in vivo treatment with anti-L3T4. J Exp Med
1985; 162: 1105–1110.
Kelvin DJ et al. Chemokines and serpentines: the molecular
biology of chemokine receptors. J Leukoc Biol 1993; 54: 604–612.
29 Paleolog EM et al. Deactivation of vascular endothelium by
monoclonal anti-tumor necrosis factor alpha antibody in rheumatoid arthritis. Arthritis Rheum 1996; 39: 1082–1091.
30 Hom JT, Cole H, Estridge T, Gliszczynski VL. Interleukin-1
enhances the development of type II collagen-induced arthritis
only in susceptible and not in resistant mice. Clin Immunol
Immunopathol 1992; 62: 56–65.
31 Cooper SM, Sriram S, Ranges GE. Suppression of murine collagen-induced arthritis with monoclonal anti-Ia antibodies and
augmentation with IFN-gamma. J Immunol 1988; 141: 1958–1962.
32 Leeuwenberg JF, Jeunhomme TM, Buurman WA. Slow release
of soluble TNF receptors by monocytes in vitro. J Immunol 1994;
152: 4036–4043.
33 Morgenstern JP, Land H. Advanced mammalian gene transfer:
high titre retroviral vectors with multiple drug selection markers and a complementary helper-free packaging cell line. Nucleic
Acid Res 1990; 18: 3587–3596.
34 Markowitz D, Goff S, Bank A. Construction and use of a safe
and efficient amphotropic packaging cell line. Virology 1988; 167:
400–406.