Toll Pathway–Dependent Blockade
of CD4?CD25?T Cell–Mediated
Suppression by Dendritic Cells
Chandrashekhar Pasare and Ruslan Medzhitov*
Toll-like receptors (TLRs) control activation of adaptive immune responses by
antigen-presenting cells (APCs). However, initiation of adaptive immune re-
sponses is also controlled by regulatory T cells (TRcells), which act to prevent
activation of autoreactive T cells. Here we describe a second mechanism of
immune induction by TLRs, which is independent of effects on costimulation.
Microbial induction of the Toll pathway blocked the suppressive effect of
CD4?CD25?TRcells, allowing activation of pathogen-specific adaptive
immune responses. This block of suppressor activity was dependent in part
on interleukin-6, which was induced by TLRs upon recognition of microbial
TLRs play an essential role in innate host
defense as well as in the control of adaptive
immune responses (1, 2). These receptors
evolved to detect the presence of infection
through recognition of conserved pathogen-
associated molecular patterns (PAMPs) (3).
Because PAMPs are produced by pathogens,
but not by the host cells, their recognition by
TLRs allows for self-nonself discrimination
(4). TLRs control activation of the adaptive
immune responses by inducing dendritic cell
(DC) maturation (1, 5). Maturation of DCs is
characterized by the up-regulation of major
histocompatibility complex (MHC) class II
and costimulatory molecules (CD80/CD86),
which provide the two requisite signals for
naı ¨ve T cell activation (6).
Induction of CD80/CD86 expression on
APCs is not the only mechanism that controls
T cell activation. T cell responses are also
regulated by CD4?CD25?suppressor or reg-
ulatory T cells (TRcells) (7–10). These cells
are critical for the maintenance of peripheral
T cell tolerance, because their depletion leads
to organ-specific autoimmune diseases (11,
12). Although the molecular mechanism of
TRcell–mediated suppression is currently un-
known, it has been shown to be cell-contact
dependent and to result in the inhibition of
interleukin-2 (IL-2) transcription in respond-
er T cells (13, 14). The peripheral T cell
repertoire consists of both self-reactive T
cells and a majority of T cells with potentially
useful specificities for pathogen-derived an-
tigens. Although inhibition of the former by
TRcells is clearly beneficial, suppression of
T cells specific for pathogens could be detri-
mental to effective protective immunity to
microbial infection. Because microbial infec-
tion is first detected and subsequently regu-
lated by TLRs (1, 2), we hypothesized that
the suppressor activity of TRcells may also
be regulated by TLRs themselves.
Ligation of TLRs on DCs overcomes
CD4?CD25?T cell–mediated suppres-
sion. To test the above hypothesis, we first
established that CD4?CD25?TRcells could
effectively suppress CD4?CD25–T cell ac-
tivation when freshly isolated splenic DCs
(15) were used as APCs (fig. S1). These cells
express low levels of CD80 and CD86, which
is characteristic of immature DCs (16), and
stimulation with TLR ligands such as lipo-
polysaccharide (LPS) or CpG leads to their
maturation, characterized by up-regulation of
cell-surface MHC class II and the costimula-
tory molecules CD80 and CD86, as well as in
the secretion of cytokines, chemokines, and
other soluble factors (6). Stimulation of DCs
with TLR4 and TLR9 ligands [LPS (17, 18)
and CpG (19), respectively] reversed the TR
cell–mediated suppression, restoring T cell
proliferation to near normal levels (15) (Fig.
1A). This effect was observed even at a high
(1:1) ratio of TRcells to responder T cells
(Fig. 1B). Further, when DCs from TLR4-
deficient mice were used as APCs, suppres-
sion was blocked in the presence of CpG but
not LPS (16). These results suggest that mi-
crobial induction of DC maturation by TLRs
is important for abrogating the suppressive
effects of TRcells.
Block of TRcell–mediated suppres-
sion is independent of costimulation.
We next determined whether the TLR-
induced block of suppression was dependent on
induction of costimulatory molecules or secret-
ed cytokines. TLR4 signals through at least two
distinct pathways: a MyD88-dependent path-
way, which is essential for production of in-
flammatory cytokines, and a MyD88-indepen-
dent pathway, which is sufficient for the up-
regulation of MHC class II and costimulatory
molecules in DCs (20). Thus, when MyD88-
deficient DCs (15) are stimulated with LPS,
they up-regulate normal levels of costimulatory
molecules. When these DCs were used as
APCs, neither LPS nor CpG treatment was able
to interfere with TRcell–mediated suppression
(Fig. 2A) despite normal up-regulation of
CD80, CD86, and CD40 in MyD88-deficient
DCs by LPS (fig. S2). This result demonstrates
that the block of TRcell–mediated suppression
was independent of expression of costimulatory
molecules on DCs.
Cytokines(s) secreted by DCs are re-
quired for overcoming suppression. To
explore whether secreted cytokines might be
responsible for the observed effects of LPS and
CpG, we added conditioned medium (CM)
from DCs treated with LPS (LPS CM) (Fig.
2B) or CpG (16) to the TRcell assays. Whether
Howard Hughes Medical Institute and Section of Im-
munobiology, Yale University School of Medicine,
New Haven, CT 06520, USA.
*To whom correspondence should be addressed. E-
Fig. 1. CD25?T cell–mediated suppression of
CD4?CD25–T cells is blocked by activation of
TLRs on DCs. (A and B) T cell proliferation
assays were set up as described (15) in the
CD4?CD25?TRcells, with or without LPS (100
ng/ml) (left) or CpG (1 ?M) (right). T cell
proliferative responses are expressed as the
mean ? SE of triplicate cultures (A) or as the
percentage of total counts (B), with 100% rep-
resenting counts in the absence of TRcells
separately for each condition. Standard errors
were less than 10%.
www.sciencemag.orgSCIENCE VOL 29914 FEBRUARY 2003
we used wild-type (WT) DCs or MyD88-defi-
cient DCs as APCs, the addition of WT LPS
CM led to a block of TRcell–mediated suppres-
sion (Fig. 2B). In contrast, CM from MyD88-
deficient or TLR4-deficient DCs stimulated
with LPS had no effect on block of suppression
(Fig. 2C). CM from WT DCs not treated with
TLR ligands had no effect on suppression (16).
Stimulation of the TLR/MyD88 pathway in
DCs therefore leads to induction of a secreted
factor(s) that is responsible for the block of
suppression, which is independent of the co-
stimulatory pathway of T cell activation.
TLRs are also expressed in many cell
types other than DCs, where they can induce
expression of distinct subsets of genes (1).
We therefore sought to determine whether the
ability of TLRs to induce factor(s) responsi-
ble for blocking of suppression was restricted
to professional APCs. In assays with CM
from LPS-stimulated DCs, macrophages, B
cells, and fibroblasts, block of TRcell–medi-
ated suppression was observed only in LPS
CM from DCs and macrophages (fig. S3A).
CM from B cells and fibroblasts stimulated
with LPS failed to block the suppressive ac-
tivity of TRcells (fig. S3B).
We next tested whether the soluble factor(s)
present within LPS and CpG CM acted directly
on TRcells to turn them off or rendered re-
sponder T cells refractory to suppression by TR
cells. TRcells were incubated for 36 hours with
antibodies to CD3 (anti-CD3) and MyD88–/–
DCs in the presence or absence of CM from
LPS-treated WT DCs and then used in a sup-
pression assay. TRcells preincubated with LPS
CM were capable of suppressing responder T
cell activation as efficiently as those that were
not exposed to DC CM and freshly isolated TR
cells (Fig. 3A). These data suggest that a cyto-
kine(s) present in LPS-stimulated DC CM acts
on responder T cells and makes them refractory
to TRcell–mediated suppression.
To demonstrate that LPS CM recovers
proliferation of responder T cells rather than
induces proliferation of TRcells, we per-
formed suppression assays with 5- (and 6-)
carboxyfluorescein diacetate succininyl ester
(CFSE)–labeled responders or suppressors.
LPS CM had no effect on TRcell prolifera-
tion but recovered proliferation of responder
T cells (fig. S4).
IL-6 is critical for overcoming suppres-
sion. We next attempted to identify the cy-
tokine(s) responsible for block of suppres-
sion. Exogenous IL-2 has been shown to
block suppression mediated by TRcells (13,
14). LPS- or CpG-treated CM from IL-2–/–
DCs blocked suppression mediated by TR
cells at similar levels to WT CM (16), dem-
onstrating that IL-2 plays no role in the TLR-
mediated block of suppression. Moreover,
release of suppression was not abrogated by
blocking antibodies to common ? chain (?c),
ruling out involvement of all the cytokines
that signal through this receptor, including
IL-7 and IL-15 (21) (Fig. 3B and table S1).
Several additional candidate cytokines were
tested, and their involvement in the blockade
of suppression was ruled out by various
methods (see table S1 for a complete listing
of the cytokines tested).
We next sought to identify the cytokine(s)
responsible for block of suppression, using size
exclusion chromatography to fractionate LPS-
tested in TRcell suppression assays, the activity
was found in fractions 10 through 14 (Fig. 4A).
DCs secrete proinflammatory cytokines such as
as well as a variety of chemokines, when stim-
ulated by TLR ligands. To determine whether
the activity in the fractions correlated with the
presence of any of the cytokines known to be
induced by TLRs in DCs, we assayed all the
fractions for these cytokines by ELISA (en-
zyme-linked immunosorbent assay). We found
that all the active fractions contained IL-6 (Fig.
4A), whereas TNF-? was present in inactive
fractions and IL-12 was found in some, but not
all, active fractions (16). We used antibodies
against each of these cytokines individually to
neutralize their activity in LPS-treated WT DC
CM. We found that neutralization of IL-12 or
TNF-? had no effect (16), but neutralization of
IL-6 almost completely abrogated the ability of
DC CM to block suppression (Fig. 4B). More-
over, LPS- or CpG-treated DC CM from IL-12
and TNF-?–deficient DCs blocked suppression
Fig. 2. Block of suppression is independent of costimulation but dependent on cytokine(s) secreted
by DCs in response to TLR ligation. (A to C) TRcell suppression assays were done with either WT
(A and B) or MyD88–/–splenic DCs (A to C) as APCs, with or without LPS or CpG (A). LPS DC CM
from either WT (B and C) or the indicated TLR4–/–mutant mice (C) was added at the beginning of
cultures. T cell responses are shown as the mean ? SE of triplicate cultures (A) or as the percentage
total counts, with 100% representing the counts in the absence of TRcells separately for each
condition (B and C). Standard errors were less than 10%.
Fig. 3. The cytokine(s) that is responsible for
block of suppression acts on responder T cells
and does not signal through the common gam-
ma chain. (A) CD4?CD25?T cells were cul-
tured in the presence of MyD88–/–DCs for 36
hours with anti-CD3 (0.1 ?g/ml) alone or with
anti-CD3 and LPS-treated WT DC CM. After 36
hours, the indicated number of freshly isolated
TRcells or cells as treated above were used in a
TRcell assay with MyD88–/–DCs as APCs. T cell
responses are shown as the mean ? SE of
triplicate cultures. (B) TRcell suppression assay
was set up with MyD88–/–splenic DCs as APCs.
CD4?CD25–T cells were first stained with a
mixture of anti-?cantibodies (10 ?g/ml) for 2
hours wherever indicated. The anti-?cantibod-
ies were present in the culture throughout the
entire period of the assay. LPS-treated WT DC
CM was added at the beginning of the cultures.
T cell responses are shown as the mean ? SE of
R E S E A R C H A R T I C L E
14 FEBRUARY 2003 VOL 299SCIENCE www.sciencemag.org
mediated by TRcells comparably to that medi-
ated by WT DC CM (16) (table S1). By con-
trast, CM from LPS-treated IL-6–/–DCs did not
lead to block of suppression (Fig. 4C). Recom-
binant IL-6 alone or together with MyD88–/–
DC CM did not lead to block of suppression
(16), although the addition of recombinant IL-6
restored the ability of DC CM from IL-6–/–DCs
to block suppression (Fig. 4D). This result sug-
gests that IL-6 needs to act synergistically
with another TLR-induced cytokine(s) for
efficient blockade of TRcell function.
These data, combined with the presence of
IL-6 in all the active fractions, indicated
that IL-6 has a major role in the TLR-
mediated block of suppression. This conclu-
sion was further confirmed when suppres-
sion assays were performed with DCs from
WT, MyD88–/–, and IL-6–/–mice as APCs in
the presence or absence of LPS (Fig. 4E).
To further test the contribution of IL-6 to
T cell activation in the presence of TRcells,
we set up in vitro priming assays using total
CD4 T cells from WT mice with titrating
doses of anti-CD3. DCs from WT or IL-6–
deficient mice were used as APCs either in
the presence or absence of LPS. Although
LPS significantly increased the ability of WT
DCs to prime naı ¨ve T cells, it had minimal
effect on the ability of IL-6–deficient DCs to
prime T cells in assays that contain physio-
logical ratios of responder and TRcells (Fig.
5A). However, in the absence of TRcells,
IL-6–/–DCs induce T cell activation similar
to that induced by WT DCs (Fig. 5B). Col-
lectively, these results demonstrate that IL-6
plays a critical role in T cell activation by
overcoming TRcell–mediated suppression.
IL-6 is required for T cell activation in
vivo. If IL-6 induced during infection allows
effector T cells to overcome suppression by TR
cells, then IL-6–deficient mice should have a
defect in T cell activation in vivo. To test this
hypothesis, we immunized WT and IL-6–defi-
cient mice with ovalbumin, using LPS as an
adjuvant (15). We found that IL-6–deficient
mice were severely compromised in their in-
Fig. 4. The presence of IL-6 in DC CM is critical for block of TRcell–mediated suppression. (A)
Fractions from size exclusion chromatography were tested for their activity to block suppression.
MyD88–/–splenic DCs were used as APCs to set up TRcell suppression assays. T cell proliferative
responses in the presence of TRcells are shown. T cell proliferative responses in the absence of TR
cells were ?235,000 counts per minute (cpm) for fractions 10 to 14 and for unfractionated DC CM,
and ?180,000 cpm for all the other fractions. Fractions were analyzed for the presence of various
cytokines. Data are presented as relative units of IL-6 found in these fractions. The apparent
discrepancy between activity of fractions and distribution of IL-6 amounts in the fractions is due
to differences in sensitivities of the two assays. (B to D) TRcell suppression assays were set up as
described, and MyD88–/–mice were used as APCs. LPS-treated DC CM from WT (B, C, and D),
MyD88–/–(C), or IL-6–/–mice (C and D) was added at the beginning of the cultures. To neutralize
IL-6, DC CM was preincubated with anti–IL-6 neutralizing antibody (10 ?g/ml) (B). Recombinant
mouse IL-6 (20 ng/ml) was used to replenish DC CM from IL-6–deficient mice (D). T cell responses
are shown as the percentage of total counts, with 100% representing the counts in the absence of
TRcells separately for each condition. Standard errors were ?5%. (E) A T cell proliferation assay
was set up in the presence or absence of TRcells with splenic DCs from indicated mice. LPS was
used at 100 ng/ml. T cell responses are shown as the mean ? SE of triplicate cultures.
Fig. 5. Production of IL-6 by DCs stimulated
with TLR ligands is important for in vitro T cell
priming. (A) Total CD4 T cells from WT mice
were stimulated with titrating doses of anti-
CD3, with DCs from either WT or IL-6–defi-
cient mice as APCs. LPS was used at 100 ng/ml.
T cell proliferative responses are shown as the
CD4?CD25–T cells from WT mice were stim-
ulated with titrating doses of anti-CD3, using
either WT or IL-6–deficient DCs. LPS (100 ng/
ml) was used to stimulate DCs. T cell prolifer-
ative responses are shown as the mean ? SE of
R E S E A R C H A R T I C L E
www.sciencemag.org SCIENCEVOL 299 14 FEBRUARY 2003
duction of T cell responses, as measured by
proliferation (Fig. 6A) and secretion of IL-2
(Fig. 6B). IL-6–/–DCs matured and produced
cytokines, in response to LPS, similar to WT
be primed in the absence of TRcells in IL-6–/–
mice, we set up monoclonal antibody (mAb)–
a previously published protocol (22) (fig. S6).
Depletion of TRcells followed by immuniza-
tion led to T cell priming in IL-6–/–mice (Fig.
6C), suggesting that the defect in T cell activa-
tion in IL-6–/–mice is due to the inability to
ence in T cell responses between WT mice and
TR-depleted IL-6–/–mice is probably due to the
fact that in vivo depletion is transient (22) (fig.
S6B) and newly generated TRcells may not
allow the same level of T cell responses in
IL-6–/–mice. These results do demonstrate,
however, that induction of IL-6 by microbial
products is required to induce T cell activation
in the presence of TRcells. When TRcells
are depleted, induction of costimulation
seems to be sufficient to induce T cell
priming. Depletion of TRcells in WT mice
during priming results in enhanced T cell
responses compared with control mice (Fig.
6D). This result demonstrates that even T
cells specific for foreign antigens are sub-
ject to TRcell–mediated suppression in
vivo. Taken together, these data demon-
strate that IL-6 plays a major role in T cell
activation both in vitro and in vivo because
of its ability to overcome suppression me-
diated by TRcells.
Conclusions. The results presented in
this study suggest that induction of T cell
responses is controlled by TLRs at least at
two different levels. TLR-mediated recogni-
tion of PAMPs and induction of costimula-
tion on DCs is required to direct T cell re-
sponses against pathogen-derived antigens in
a cognate T cell–APC interaction. However,
induction of costimulation alone does not
seem to be sufficient to induce T cell activa-
tion in vivo. Indeed we find that production
of IL-6 by DCs in response to TLR ligation
during infection is critical for T cell activa-
tion, because it allows pathogen-specific T
cells to overcome the suppressive effect of
CD4?CD25?TRcells. Removal of TRcells,
however, allows T cell activation even in the
absence of IL-6 (in IL-6–deficient mice),
suggesting that induction of costimulation on
DCs is sufficient for T cell activation in the
absence of TRcells.
Although IL-6 can act systemically, pro-
duction of high levels of IL-6 during infec-
tion will not normally result in a nonspecific
block of suppression, because activation of
antigen-specific T cells still requires the co-
stimulatory signals provided in a cognate T
cell–APC interaction. However, it is tempting
to speculate that during chronic infections,
conditions may arise that lead to IL-6–medi-
ated release of suppression of self-reactive T
cells, which may explain the link between
infection and some autoimmune diseases
(23). IL-6–deficient mice, on the other hand,
are resistant to autoimmune diseases such as
experimental autoimmune encephalitis and
rheumatoid arthritis (24–27). In addition, ad-
ministration of IL-6R mAb to mice with severe
combined immunodeficiency (SCID), after
transfer of CD45 RBhighT cells, confers pro-
tection from colitis (28), and IL-6–deficient
mice have been shown to be less susceptible to
colitis (29). These reports support our findings,
and our results seem to provide a mechanistic
explanation for these observations.
IL-6 has multiple well-characterized
functions, particularly in the context of the
acute phase response and B cell differenti-
ation. Our study suggests that the failure to
overcome TR-mediated suppression con-
tributes to the phenotype of IL-6–/–mice,
including their susceptibility to infection
and resistance to autoimmunity.
In conclusion, the present study demon-
strates that innate immune recognition by
TLRs controls the activation of adaptive im-
mune responses by at least two distinct mech-
anisms: the induction of costimulatory mole-
cules on DCs and the production of IL-6,
which renders pathogen-specific T cells re-
fractory tothe suppressive
References and Notes
1. S. Akira, Nature Immunol. 2, 675 (2001).
2. R. Medzhitov, Nature Rev. Immunol. 1, 135 (2001).
3. ????, C. A. Janeway Jr., Cell 91, 295 (1997).
4. C. A. Janeway Jr., Cold Spring Harbor Symp. Quant.
Biol. 54, 1 (1989).
5. M. Schnare et al., Nature Immunol. 2, 947 (2001).
6. J. Banchereau, R. M. Steinman, Nature 392, 245
7. K. J. Maloy, F. Powrie, Nature Immunol. 2, 816 (2001).
8. C. A. Piccirillo, E. M. Shevach, J. Immunol. 167, 1137
9. S. Sakaguchi, Cell 101, 455 (2000).
10. E. M. Shevach, Annu. Rev. Immunol. 18, 423 (2000).
11. M. Asano, J. Exp. Med. 184, 387 (1996).
12. O. Taguchi, T. Takahashi, Eur. J. Immunol. 26, 1608
13. A. M. Thornton, E. M. Shevach, J. Exp. Med. 188, 287
14. T. Takahashi et al., Int. Immunol. 10, 1969 (1998).
15. Materials and methods are available as supporting
material on Science Online.
16. C. Pasare, R. Medzhitov, unpublished observations.
17. S. T. Qureshi et al., J. Exp. Med. 189, 615 (1999).
18. A. Poltorak et al., Science 282, 2085 (1998).
19. H. Hemmi et al., Nature 408, 740 (2000).
20. T. Kaisho, J. Immunol. 166, 5688 (2001).
21. K. Sugamura et al., Annu. Rev. Immunol. 14, 179
22. S. Onizuka et al., Cancer Res. 59, 3128 (1999).
23. N. R. Rose, Semin. Immunol. 10, 5 (1998).
24. T. Kishimoto, J. Autoimmun. 5 Suppl A, 123 (1992).
25. E. B. Samoilova, J. Immunol. 161, 6480 (1998).
26. S. Ohshima et al., Proc. Natl. Acad. Sci. U.S.A. 95,
27. Y. Okuda et al., Int. Immunol. 10, 703 (1998).
28. M. Yamamoto, J. Immunol. 164, 4878 (2000).
29. A. Suzuki et al., J. Exp. Med. 193, 471 (2001).
30. We thank A. Rudensky, G. Barton, E. Kopp, and A.
Unni for helpful discussions; S. Akira for the gift of
MyD88, TLR2, and TLR4 knockout mice; and G.
Dranoff for the gift of Flt3 ligand–expressing cells.
This research was supported by the Howard Hughes
Medical Institute, grants from NIH, and a Searle
Scholarship to R.M.
Supporting Online Material
Materials and Methods
Figs. S1 to S6
9 September 2002; accepted 11 December 2002
Published online 16 January 2003;
Include this information when citing this paper.
Fig. 6. T cell priming is defective
in vivo in IL-6–/–mice. (A and B)
Purified CD4 T cells from drain-
ing lymph nodes of WT or IL-6–
deficient mice, immunized with
ovalbumin and LPS emulsified in
(IFA), were restimulated with ti-
trating doses of ovalbumin. Irra-
diated (2000 rads) B cells from
WT mice were used as APCs.
Data are shown as proliferative
responses (A) and IL-2 produc-
tion (B) by responding T cells. (C
and D) Anti-CD25 (Clone PC61)
or control rat immunoglobulin G
was injected intravenously into
mice on day 0. Depletion of
CD25?T cells was confirmed by
staining the peripheral blood
mononuclear cells (PBMCs) on
day 3 (fig. S6). Mice were then
ovalbumin and LPS emulsified in
IFA. Purified CD4 T cells from draining lymph nodes of WT or IL-6–deficient mice were restimu-
lated on day 7 of immunization with titrating doses of ovalbumin. Irradiated (2000 rads) B cells
from WT mice were used as APCs. Data are shown as proliferative responses and are representative
of three independent experiments.
R E S E A R C H A R T I C L E
14 FEBRUARY 2003VOL 299SCIENCEwww.sciencemag.org