Constitutive MHC class I molecules negatively regulate TLR ... · nature immunology VOLUME 13 NUMBER 6 JUNE 2012 . 551. Articles. Toll-like receptors (TLRs) are key pattern-recognition

nature immunology  VOLUME 13 NUMBER 6 JUNE 2012 551

A rt i c l e s

Toll-like receptors (TLRs) are key pattern-recognition receptors used by cells of the innate immune system to detect the conserved components of pathogens, and they have critical roles in host defense against invading microbial pathogens1,2. After the recognition of pathogen-associated molecule patterns, TLRs initiate innate immune responses by activating signaling pathways that depend on the adaptor MyD88 or the adaptor TRIF and consequently induce the produc-tion of proinflammatory cytokines and type I interferon3. Many factors have been identified as being essential for full activation of TLR responses; however, inhibitors of TLR pathways need further investigation, as inappropriate activation or overactivation of TLR signaling may result in inflammatory disorders such as septic shock or autoimmune diseases4. The identification of the negative regulators and details of the mechanisms by which TLR signaling is fine tuned remain to be fully elucidated.

Major histocompatibility complex (MHC) class I molecules are expressed on all nucleated cells from vertebrates. Classical MHC class I molecules (HLA-I in humans; H-2 I in mice) are heterodimers composed of β2-microglobulin (β2m) and a membrane-bound heavy chain with three extracellular domains and a cytoplasmic domain of about 40 amino acids5. The component β2m is essential for the stable expression of MHC class I, and β2m-deficient mice lack most if not all MHC class I molecules6. The main function of MHC class I molecules seems to be the presentation on the cell surface of small-peptide frag-ments of endogenously produced antigens for recognition by CD8+ cytotoxic T lymphocytes6. MHC class I molecules also contribute to the development and selection of CD8+ T cells in the thymus7, as well as

to the ‘education’ and tolerance of natural killer (NK) cells8,9. Normal expression of MHC class I molecules by target cells inhibits the lysis of target cells by NK cells. However, chronic interaction between MHC class I molecules and cognate inhibitory receptors is essential for the licensing of NK cells and for the acquisition of their killing ability9,10. In these cases, MHC class I molecules act as ligands to stimulate relevant receptors to trigger downstream signaling.

In addition to the classical function of antigen presentation, MHC class I molecules can also mediate reverse signaling after ligation and have nonclassical functions11–13. The aggregation of MHC class I molecules on the cell surface with agonist antibodies, T cell antigen receptors (TCRs) or the coreceptor CD8 activates signaling pathways in T cells, B cells, tumor cells or endothelial cells and elicits various biological effects, such as cell apoptosis, activation or prolifera-tion13–15. Crosslinkage of MHC class I molecules on human NK cells induces intracellular tyrosine phosphorylation and inhibits NK cell cytotoxicity directed against tumor target cells16. Such data suggest that MHC class I molecules can be ligands and signaling receptors themselves to mediate reverse signaling via association with other receptors or directly through aggregation. The activation of tyrosine kinases, including Lck, Lyn, Syk, Zap70 and Tyk2, occurs immedi-ately after engagement of MHC class I molecules17,18. So far, almost all data about the nonclassical functions of MHC class I molecules have been obtained from studies of lymphoid cells; however, whether myeloid cells or antigen-presenting cells (APCs) display MHC class I molecules with nonclassical functions and reverse signaling has not been elucidated. Intracellular MHC class II molecules do promote

1National Key Laboratory of Medical Immunology & Institute of Immunology, Second Military Medical University, Shanghai, China. 2National Center of Biomedical Analysis, Beijing, China. 3National Key Laboratory of Molecular Biology, Chinese Academy of Medical Sciences, Beijing, China. 4These authors contributed equally to this work. Correspondence should be addressed to X.C. ([email protected]).

Received 28 October 2011; accepted 6 March 2012; published online 22 April 2012; corrected online 4 May 2012; doi:10.1038/ni.2283

Constitutive MHC class I molecules negatively regulate TLR-triggered inflammatory responses via the Fps–SHP-2 pathwaySheng Xu1,4, Xingguang Liu1,4, Yan Bao1, Xuhui Zhu1, Chaofeng Han1, Peng Zhang1, Xuemin Zhang2, Weihua Li2 & Xuetao Cao1,3

The molecular mechanisms that fine-tune Toll-like receptor (TLR)-triggered innate inflammatory responses remain to be fully elucidated. Major histocompatibility complex (MHC) molecules can mediate reverse signaling and have nonclassical functions. Here we found that constitutively expressed membrane MHC class I molecules attenuated TLR-triggered innate inflammatory responses via reverse signaling, which protected mice from sepsis. The intracellular domain of MHC class I molecules was phosphorylated by the kinase Src after TLR activation, then the tyrosine kinase Fps was recruited via its Src homology 2 domain to phosphorylated MHC class I molecules. This led to enhanced Fps activity and recruitment of the phosphatase SHP-2, which interfered with TLR signaling mediated by the signaling molecule TRAF6. Thus, constitutive MHC class I molecules engage in crosstalk with TLR signaling via the Fps–SHP-2 pathway and control TLR-triggered innate inflammatory responses.

npg

© 2

012

Nat

ure

Am

eric

a, In

c. A

ll rig

hts

rese

rved

.

552  VOLUME 13 NUMBER 6 JUNE 2012 nature immunology

A rt i c l e s

TLR-triggered innate immune responses by maintaining activation of the kinase Btk19. Whether and how MHC class I molecules inter-sect with innate TLR signaling pathways and regulate TLR-triggered innate inflammatory response remains unknown.

Here we demonstrate that deficiency in MHC class I resulted in enhanced TLR-triggered production of proinflammatory cytokines and type I interferon both in vitro and in vivo. Consistent with that, engagement of MHC class I molecules inhibited TLR-triggered signal transduction and production of cytokines. Constitutive membrane MHC class I molecules physiologically interacted with the tyrosine kinase Fps (Fes). After the stimulation of TLRs with ligands, MHC class I molecules became phosphorylated and recruited more Fps, which resulted in more potent activation of the phosphatase SHP-2 and, finally, suppressed TLR-triggered inflammatory responses. Therefore, our results demonstrate a nonclassical function for MHC class I molecules as negative regulators of a TLR pathway, which adds new insight into the fine tuning of TLR-triggered innate inflamma-tory responses.

RESULTSMHC class I deficiency exacerbates TLR-triggered responsesTo investigate the role of MHC class I molecules in TLR-triggered innate inflammatory responses, we challenged β2-microglobulin-deficient (B2m−/−) mice with various TLR ligands, including lipopolysaccharide (LPS), the synthetic RNA duplex poly(I:C) or an oligodeoxynucleotide based on the dinucleotide motif CpG (CpG ODN). MHC class I–deficient mice produced significantly more tumor necrosis factor (TNF), interleukin 6 (IL-6) and interferon-β (IFN-β) than did their littermates (control mice; Fig. 1a). We detected no substantial difference between B2m−/− and wild-type mice in the development of myeloid cells or the expression of TLRs (data not shown), which excluded the possibility that the phenomenon noted above was due to the abnormal development of myeloid cells or TLR expression. Furthermore, after lethal challenge with LPS, almost all MHC class I–deficient mice died within 24 h, whereas 40% of their lit-termates (control mice) were alive at that time and about 25% of these survived the challenge (Fig. 1b). Consistent with that, we observed more severe infiltration of polymorphonuclear cells and interstitial pneumonitis in the lungs of MHC class I–deficient mice 8 h after LPS challenge (Fig. 1c). Furthermore, we also confirmed the greater production of TNF, IL-6 and IFN-β in mice deficient in the MHC class I heavy chain (H2-K1−/−H2-D1−/−; called ‘KbDb−/−’ here) after TLR challenge (Supplementary Fig. 1a).

Given that MHC class I–deficient mice have considerably fewer CD8+ T cells in the thymus, spleen and lymph nodes, we further recon-stituted lethally irradiated wild-type mice with wild-type, B2m−/− or

KbDb−/− bone marrow cells and then assessed their responses to chal-lenge with the TLR ligands. When challenged with LPS, chimeric mice reconstituted with B2m−/− or KbDb−/− bone marrow produced more TNF, IL-6 and IFN-β than did chimeric mice reconstituted with wild-type bone marrow (Fig. 1d and data not shown). Thus, these data demonstrated that TLR-triggered inflammatory responses were greater in MHC class I–deficient mice, which indicated a suppressive role for MHC class I molecules in the TLR response.

To investigate the role of MHC class I molecules in host resist-ance to pathogen infection, we challenged B2m−/− mice and KbDb−/− mice with Gram-negative Escherichia coli or Gram-positive Listeria monocytogenes. After infection with E. coli, the production of TNF, IL-6 and IFN-β in MHC class I–deficient mice was significantly greater than that of their littermates (control mice; Fig. 2a and Supplementary Fig. 1b). Accordingly, MHC class I–deficient mice had a greater load of E. coli bacteria in the blood (Fig. 2b and Supplementary Fig. 1b), consistent with published findings that proinflammatory cytokines promote the dissemination of E. coli20. After infection with L. monocytogenes, MHC class I–deficient mice also produced more proinflammatory cytokines and had a smaller bacterial load in the spleen and liver (Fig. 2c,d and Supplementary Fig. 1c). Furthermore, consistent with published reports21, NK cells from MHC class I–deficient mice showed less cytotoxicity directed against target cells and produced less IFN-γ after infection with L. monocytogenes (data not shown), which excluded the possibility that the smaller bacterial load in MHC class I–deficient mice was a result of enhanced killing by NK cells. These data indicated that MHC class I molecules may help the host restrict inflammatory responses after bacterial infection and protect the host from inflammatory inju-ries mediated by innate immune responses.

More cytokine production in MHC class I–deficient APCsAPCs, especially macrophages, are the main mediators of TLR- triggered innate inflammatory responses in vivo. Consistent with the in vivo data presented above, peritoneal macrophages from B2m−/− mice produced more TNF, IL-6 and IFN-β than did those from their littermates (control mice) in response to stimulation with LPS, poly(I:C) or CpG ODN (Fig. 3a and Supplementary Fig. 2a). We obtained similar results with KbDb−/− macrophages (Supplementary Fig. 1d). We also found exacerbated cytokine production in

4a

b

d

c

20 2.0

1.5

0.5

0

1.0

16

12

8

4

0

B2m+/+

B2m–/–

B2m+/+

B2m+/+

B2m+/+→B2m+/+

B2m–/–

B2m–/–

B2m–/–→B2m+/+

** **

** ** **

**

**

****

**

**3

2

TN

F (

ng/m

l)T

NF

(ng

/ml)

IFN

-β (

ng/m

l)

IL-6

(ng

/ml)

IL-6

(ng

/ml)

IFN

-β (

ng/m

l)

1

0

100

80

60

40

Sur

viva

l (%

)

20

0

4

3

2

1

0

122.5

2.0

1.5

1.0

0.5

0

8

4

0

0 8 16 24Time after LPS challenge (h) PBS

PBS

LPS

LPS PBS LPS PBS LPS

32 40 48 56

PBSLP

S

Poly(I:

C)CpG PBS

LPS

Poly(I:

C)CpG PBS

LPS

Poly(I:

C)CpG

Figure 1 MHC class I–deficient mice are more susceptible to TLR challenge. (a) Enzyme-linked immunosorbent assay (ELISA) of TNF, IL-6 and IFN-β in serum from B2m+/+ and B2m−/− mice (n = 5 per genotype) 2 h after intraperitoneal challenge with PBS, LPS, poly(I:C) or CpG-ODN. (b) Survival of B2m+/+ and B2m−/− mice (n = 10 per genotype) after lethal challenge with LPS (10 mg per kg body weight). P < 0.01 (Wilcoxon test). (c) Hematoxylin-and-eosin staining of lungs from B2m+/+ and B2m−/− mice 8 h after challenge with PBS or LPS. Original magnification, ×100. (d) ELISA of TNF, IL-6 and IFN-β in serum from wild-type mice (n = 6 per group) reconstituted for 8 weeks with B2m+/+ bone marrow cells (B2m+/+→B2m+/+) or B2m−/− bone marrow cells (B2m−/−→B2m+/+), followed by challenge with PBS or LPS and analysis 2 h later. *P < 0.05 and **P < 0.01 (Student’s t-test). Data are from three independent experiments (a,d; mean ± s.e.m.) or are representative of three independent experiments with similar results (b,c).

npg

© 2

012

Nat

ure

Am

eric

a, In

c. A

ll rig

hts

rese

rved

.

nature immunology  VOLUME 13 NUMBER 6 JUNE 2012 553

A rt i c l e s

MHC class I–deficient bone marrow–derived dendritic cells (Supplementary Fig. 2b). We then investigated the effect of knock-down of MHC class I on cytokine expression in macrophages in which TLRs were triggered. Macrophages in which the gene encod-ing H-2Kb was silenced by H-2Kb-specific small interfering RNA (siRNA) produced significantly more proinflammatory cytokines and IFN-β in response to stimulation with LPS, poly(I:C) or CpG ODN than did those transfected with control siRNA (Fig. 3b,c). To further demonstrate that the enhanced TLR-triggered inflamma-tory response in MHC class I–deficient mice in vivo was due to the MHC class I deficiency in macrophages, we adoptively transferred MHC class I–deficient macrophages into wild-type mice depleted of endogenous macrophages via pretreatment with clodronate liposomes. After LPS challenge, mice reconstituted with MHC class I–deficient (B2m−/− or KbDb−/−) macrophages produced more proin-flammatory cytokines and IFN-β than did those reconstituted with wild-type macrophages (Fig. 3d and data not shown). Therefore, these data suggested that MHC class I deficiency enhanced TLR-triggered inflammatory responses in macrophages and that this may have resulted in the greater susceptibility of MHC class I–deficient mice to lethal challenge with LPS observed above.

Ligation of MHC class I molecules via monoclonal antibodies induces downstream signals in lymphocytes and elicits biological func-tions13–15. After stimulation with LPS, poly(I:C) or CpG ODN, macro-phages crosslinked with antibody to H-2Kb (anti-H-2Kb) secreted less proinflammatory cytokines and IFN-β than did macrophages treated with control antibody (Fig. 3e). Crosslinkage of another MHC class I molecule, H2-Db, also suppressed the TLR-triggered inflamma-tory responses in macrophages (Supplementary Fig. 2c), whereas crosslinkage of both H-2Kb and H2-Db resulted in even less produc-tion of these cytokines (data not shown). These data suggested that crosslinkage of MHC class I molecules exerted an inhibitory effect on TLR-triggered inflammatory responses in macrophages.

CD8+ T cells attenuate TLR responses via MHC class ITo determine the physiological relevance of our observation that reverse signals from MHC class I molecules suppressed TLR-triggered immune responses, we looked for cells in vivo that provided ligands for MHC class I molecules on APCs. The known natural ligands for MHC class I in vivo are TCRs, the coreceptor CD8, the lectin-like receptor CD94-NKG2 and killer immunoglobulin-like receptors expressed mainly on NK cells and CD8+ T cells, which made these

Figure 3 MHC class I reverse signaling inhibits TLR-triggered production of proinflammatory cytokines and type I interferon in macrophages. (a) ELISA of TNF, IL-6 and IFN-β in supernatants of B2m+/+ and B2m−/− macrophages stimulated for 4 h with PBS, LPS, poly(I:C) or CpG ODN. (b) Immunoblot analysis of MHC class I (MHCI) in macrophages 48 h after transfection with control (Ctrl) siRNA or siRNA specific for MHC class I. β-actin serves as a loading control throughout. (c) ELISA of TNF, IL-6 and IFN-β in macrophages transected with siRNA as in b (key), then stimulated for 4 h with LPS, poly(I:C) or CpG. (d) ELISA of TNF, IL-6 and IFN-β in serum from wild-type mice (n = 4 per group) depleted of endogenous macrophages and then given adoptive transfer of wild-type macrophages (B2m+/+→B2m+/+) or MHC class I–deficient macrophages (B2m−/−→B2m+/+), followed by LPS challenge and analysis 2 h later. (e) ELISA of TNF, IL-6 and IFN-β in macrophages after ligation with control antibody (immunoglobulin G (IgG)) or monoclonal antibody to MHC class I (H-2Kb) and stimulation for 4 h with PBS, LPS, poly(I:C) or CpG ODN. *P < 0.05 and **P < 0.01 (Student’s t-test). Data are from three independent experiments (a,c–e; mean ± s.e.m.) or are representative of three independent experiments with similar results (b).

B2m+/+ B2m–/– B2m+/+ B2m+/+

B2m–/–

B2m–/–

5a

4* *

****

**

**

**

3

10 2.5 8

6

4

2

0

10

TN

F (

ng/m

l)

IL-6

(ng

/ml)

IL-6

(ng

/ml)

IFN

-β (

ng/m

l)

IFN

-β (

ng/m

l)

Blo

od E

. col

i(L

og10

CF

U/m

l)

Spl

een

LM (

log 10

CF

U)

Live

r LM

(lo

g 10 C

FU

)

2

1

0

5

4

3

2

1

02 h 4 h

B2m+/+ B2m–/–

4

* * **

*****3

TN

F (

ng/m

l)

2

1

0PBS LM PBS LM PBS LM 1 d 3 d 1 d 3 d

8

6

4

2

02 h 4 h 2 h 4 h 4 h 8 h

2.08

10

1.5

1.5

1.0

1.0

0.50.5

0 0

6

4

2

0

b c d

8

6

4

2

0

Figure 2 MHC class I–deficient mice are more susceptible to infection with E. coli but are more resistant to infection with L. monocytogenes (LM). (a,b) ELISA of TNF, IL-6 and IFN-β in serum (a) and analysis of bacterial loads in blood (b) of B2m+/+ and B2m−/− mice (n = 4 per genotype) 2, 4 or 8 h (horizontal axes) after intraperitoneal infection with 1 × 108 E. coli. CFU, colony-forming units. (c,d) ELISA of TNF, IL-6 and IFN-β in serum (c) and analysis of bacterial loads in spleen and liver (d) of B2m+/+ and B2m−/− mice (n = 4 per genotype) after treatment with PBS or infection with 1 × 104 L. monocytogenes; serum was obtained 4 h after infection. In b,d, each symbol represents an individual mouse; small horizontal lines indicate the mean. *P < 0.05 and **P < 0.01 (Student’s t-test). Data are from three independent experiments with similar results (mean ± s.e.m.).

a

b

c

ed

5 8 3B2m+/+ B2m–/–

4*

****

*

**3

2

TN

F (

ng/m

l)

1

0

CtrlsiRNA:

5 8 3

***

****

**

****

**

***

*

*

**

**

*

**

MHCI5

B2m+/+→B2m+/+ B2m–/–→B2m+/+

8 2.0* **

***

PBSLP

S

Poly(I:

C)CpG PBS

LPS

Poly(I:

C)CpG

PBSLP

S

Poly(I:

C)CpG PBS

LPS

Poly(I:

C)CpG PBS

LPS

Poly(I:

C)CpG

PBSLP

S

Poly(I:

C)CpG PBS

LPS

Poly(I:

C)CpGPBS

LPS

Poly(I:

C)CpG PBS

LPS

Poly(I:

C)CpG

6

4

IL-6

(ng

/ml)

IFN

-β (

ng/m

l)

IFN

-β (

ng/m

l)

IFN

-β (

ng/m

l)IF

N-β

(ng

/ml)

2

0

2

1

0

432

TN

F (

ng/m

l)

IL-6

(ng

/ml)

IL-6

(ng

/ml)

IL-6

(ng

/ml)

10

4Time (h) Time (h) Time (h)

6 4 6 4 6

6

4

2

0

1.5

1.0

0.5

0

4

3

2

TN

F (

ng/m

l)T

NF

(ng

/ml)

1

0

3 5 3IgG H-2Kb

6

4

2

0

2

1

0

2

1

0

4

3

2

1

0

2

1

0

***

*

MHCICtrl

MHCI

β-actin

npg

© 2

012

Nat

ure

Am

eric

a, In

c. A

ll rig

hts

rese

rved

.

554  VOLUME 13 NUMBER 6 JUNE 2012 nature immunology

A rt i c l e s

cells the most likely candidates. In vivo depletion of NK cells or CD8+ T cells resulted in significantly enhanced production of TNF, IL-6 and IFN-β in mice after LPS challenge (Fig. 4a), consistent with a published report22. We further cultured macrophages together with NK cells or CD8+ T cells in vitro and found that culture with CD8+ T cells resulted in significantly impaired production of TNF, IL-6, IFN-β by LPS-stimulated macrophages, whereas culture with NK cells did not (Fig. 4b,c).

To demonstrate the underlying mechanisms by which CD8+ T cells inhibited macrophage inflammatory responses to TLR ligands, we used a Transwell system and found that the inhibitory effect of CD8+ T cells was attenuated when the cells were physically separated (Fig. 4d). In addition, blockade of the inhibitory cytokines IL-10 or TGF-β did not relieve the suppressive effect of CD8+ T cells (Fig. 4d). Therefore, CD8+ T cell–mediated suppression of the innate inflammatory responses of macrophages was dependent on cell-cell contact. To elucidate whether MHC class I molecules were involved in the process, we cultured CD8+ T cells together with MHC class I–deficient macrophages. Notably, the suppressive effect of CD8+ T cells was completely abrogated (Fig. 4b). Additionally, when we adoptively transferred wild-type CD8+ T cells into B2m−/− mice whose macrophages lack expression of MHC class I, we observed no significant attenuation of inflammatory responses after LPS stimulation (Fig. 4e). These data suggested a nonredundant role for MHC class I expression on macrophages in the suppression of innate responses by CD8+ T cells.

We further determined whether engagement of TCRs by peptide–MHC class I complexes was required. OT-I cells (which transgenically express an ovalbumin-specific TCR) tempered cytokine production by macrophages just as nontransgenic CD8+ T cells did in the absence of cognate peptide (Fig. 4f). However, after the addition of ovalbumin peptide, the suppressive effect of OT-I cells was abrogated, whereas the inhibitory effect of nontransgenic CD8+ T cells was not influenced by the addition of ovalbumin peptide (Fig. 4f and data not shown). We also found that naive CD8+ T cells had a greater effect than memory T cells had in suppressing cytokine production by macrophages in which TLRs were triggered (Fig. 4g). Together these data suggested

that naive CD8+ T cells suppress innate inflammatory responses in macrophages in an MHC class I–dependent way.

Enhanced TLR signaling in MHC class I–deficient macrophagesTo investigate whether MHC class I deficiency intersected with the TLR signaling pathways in macrophages, we examined the activa-tion kinetics of the mitogen-activated protein kinase (MAPK) and transcription factor NF-κB pathways, which are both downstream of TLR signaling. Activation of the MAPKs Jnk, Erk and p38, the kinases IKKα and IKKβ and the NF-κB inhibitor IκBα was enhanced in LPS-stimulated MHC class I–deficient macrophages (Fig. 5a and data not shown). We also observed more phosphorylation of the transcription factor IRF3 (Fig. 5a). We obtained similar results with MHC class I–deficient macrophages stimulated with poly(I:C) or CpG ODN (Supplementary Fig. 3a). These data suggested that deficiency in MHC class I enhances TLR signaling in macrophages.

a b d

e f g

c** **

** *

* *

* *** **

**

**

TN

F (

ng/m

l)

TN

F (

ng/m

l)

IL-6

(ng

/ml)

MΦ

MΦ

MΦ

MΦ

MΦ +

CD8

+

MΦ +

naiv

e

MΦ +

mem

MΦ +

OT-I

MΦ +

OT-I

+ OVA

CD8+ T cellTranswell

Anti–IL-10Anti–TGF-β

+–– – – –

––––– – – –

++

+++

++

+

++

+

MΦ

MΦ +

NK

MΦ + CD8+ T cells5 15

4

3

2

1

0

2.0 4

PBS

NS NS NS NS

B2m+/

+

B2m+/

+

B2m–/

–

B2m–/

–

B2m–/

– + CD8

+

B2m+/

+

B2m–/

–

B2m+/

+

B2m–/

–

LPS PBS LPS PBS LPS

10

5

0

IFN

-β (

ng/m

l)

IFN

-β (

ng/m

l)

TN

F (

ng/m

l)

TN

F (

ng/m

l)

TN

F (

ng/m

l)

TN

F (

ng/m

l)

TN

F (

ng/m

l)

IL-6

(ng

/ml)

1.5

1.0

0.5

0

1.5

1.0

0.5

0

1.0

0.5

0

1.5

1.0

0.5

0 0

1

2

8 2 3

2

1

0

1

0

NS6

4

2

0

3

2

1

0

WT CD8+ T cell depletion NK cell depletion

Figure 4 CD8+ T cells suppress TLR-triggered production of inflammatory cytokines in macrophages in an MHC class I–dependent manner. (a) ELISA of TNF, IL-6 and IFN-β in the serum of wild-type mice (n = 4 per group) left undepleted (WT) or depleted of CD8+ T cells or NK cells, followed by LPS challenge and analysis 2 h later. (b) ELISA of TNF, IL-6 and IFN-β in supernatants of B2m+/+ or B2m−/− macrophages cultured alone (MΦ) or at a ratio of 1:1 with CD8+ T cells (MΦ + CD8+ T cells) and simulated for 4 h with LPS. (c) ELISA of TNF in supernatants of macrophages cultured alone (MΦ) or with NK cells (MΦ + NK) and simulated for 4 h with LPS. (d) ELISA of TNF in wild-type macrophages cultured in the presence (+) or absence (−) of CD8+ T cells, anti–IL-10 and anti–TGF-β, with (+) or without (−) a Transwell system (to separate CD8+ T cells in the upper chamber), then stimulated for 4 h with LPS (100 ng/ml). (e) TNF in serum from B2m+/+ or B2m−/− mice (left) or B2m−/− mice given transfer of wild-type CD8+ T cells (B2m−/− + CD8+; n = 4 mice per group), assessed 2 h after LPS challenge. (f) ELISA of TNF in macrophages cultured alone (MΦ) or with CD8+ T cells (MΦ + CD8+) or OT-I cells (MΦ + OT-I) in the presence (+ OVA) or absence of ovalbumin peptide (amino acids 257–264). (g) ELISA of TNF in wild-type macrophages cultured alone (MΦ) or with naive (MΦ + naive) or memory (MΦ + mem) CD8+ T cells (sorted by flow cytometry as CD8+CD44lo or CD8+CD44hi cells, respectively) and stimulated for 2 h with LPS (100 ng/ml). NS, not significant; *P < 0.05 and **P < 0.01 (Student’s t-test). Data are from three independent experiments with similar results (mean ± s.e.m.).

aLPS (min) 0 15 30 45 60 0 15 30 45 60

B2m+/+ B2m–/–

p-Erk

p-Jnk

p-p38

p-IKKα-IKKβ

p-IκBα

β-actin

0 30 60 90120 0 30 60 90120

B2m+/+ B2m–/–

LPS (min)

p-IRF3

β-actin

0 15 30 45 60

IgG H-2Kb

0 15 30 45 60LPS (min)

p-Erk

p-Jnk

p-p38

p-IKKα-IKKβ

p-IκBα

β-actin

0 30 60 90 120 0 30 60 90 120

IgG H-2Kb

LPS (min)p-IRF3

β-actin

b

Figure 5 MHC class I reverse signaling impairs TLR pathways in macrophages. (a) Immunoblot analysis of phosphorylated (p-) signaling molecules in lysates of B2m+/+ and B2m−/− macrophages stimulated for 0–120 min (above lanes) with LPS. (b) Immunoblot analysis of phosphorylated signaling molecules in lysates of macrophages treated with control antibody (IgG) or crosslinked with monoclonal antibody to MHC class I (H-2Kb) and stimulated with LPS. Data are from one experiment representative of three independent experiments with similar results.

npg

© 2

012

Nat

ure

Am

eric

a, In

c. A

ll rig

hts

rese

rved

.

nature immunology  VOLUME 13 NUMBER 6 JUNE 2012 555

A rt i c l e s

As ligation of MHC class I molecules attenuated TLR-triggered inflammatory responses, we further investigated the effect of MHC class I ligation on TLR pathways. Crosslinkage of MHC class I mole-cules on macrophages resulted in less activation of the MAPK, NF-κB and IRF3 pathways in response to LPS stimulation (Fig. 5b), whereas crosslinkage alone did not activate any of these pathways (data not shown). We obtained similar results after stimulation with poly(I:C) or CpG ODN (Supplementary Fig. 3b). Collectively, these data indi-cated that ligation of MHC class I molecules may have inhibited the TLR-triggered production of proinflammatory cytokines and type I interferon by impairing activation of MyD88- and TRIF-dependent pathways in macrophages.

Membrane MHC class I molecules bind FpsWe further explored the underlying mechanisms of the suppres-sion of TLR signaling by MHC class I molecules. First, we sought to determine whether MHC class I molecules interacted directly with TLRs and served as cofactors in inhibiting TLR signaling by forming a complex. Immunoprecipitation with anti–MHC class I demonstrated no such interaction between TLRs and MHC class I molecules, and we obtained similar results by coimmuno-precipitation with anti-TLR as well (data not shown). Confocal microscopy also showed that the distribution of TLR3, TLR4 and TLR9 in endosomes in B2m−/− macrophages was similar to that in wild-type macrophages after stimulation with the ligands for these TLRs (Supplementary Fig. 4a–c). The downregulation of surface TLR4 in response to stimulation with LPS was also similar in wild-type and MHC class I–deficient macrophages (Supplementary Fig. 4d). These results demonstrated that deficiency in MHC class I did not affect the endosomal distribution of TLR4, TLR3 or TLR9 in macrophages after activation of the TLR, which suggested the existence of other molecules that interact with MHC class I and temper TLR signaling.

Given that activation of many tyrosine kinases is involved in the control of TLR signaling, we immunoprecipitated proteins from lysates of LPS-stimulated macrophages with anti–MHC class I and

then used reverse-phase nanospray liquid chromatography–tandem mass spectrometry to identify possible MHC class I–associated tyro-sine kinases or other molecules that may be involved in the suppres-sion of TLR pathways. Among the kinases shown to be associated with MHC class I, the non-receptor kinase Fps attracted our attention, as Fps-deficient mice have been shown to be more susceptible to LPS-induced sepsis23. An immunoprecipitation assay confirmed that Fps did indeed interact with MHC class I molecules at steady state, and this interaction increased after LPS stimulation (Fig. 6a,b). In addi-tion, plasma-membrane MHC class I molecules associated with Fps, but intracellular MHC class I molecules did not (Fig. 6a). Confocal microscopy showed a dispersed pattern of Fps in the cytosol of unstimulated macrophages, although small amounts associated with MHC class I. After stimulation with TLR ligands, Fps was activated and aggregated as clusters along the cytoplasmic side of the membrane and showed substantially enhanced colocalization with membrane MHC class I molecules (Fig. 6c).

MHC class I binds Fps via intracellular phosphorylated tyrosinesWe went further to determine which domains of MHC class I and Fps were required for their interaction and for the suppression of TLR-triggered responses. Distinct from MHC class II molecules, MHC class I molecules have a relatively long tail and a tyrosine-phosphorylation site in the cytoplamic domain (Tyr320 in human and Tyr321 in mice), whereas Fps contains a Src-homology 2 (SH2) domain. Thus, we hypothesized that these two molecules may interact directly. We transfected mouse macrophages to express Myc-tagged Fps together with Flag-tagged wild-type H-2K, H-2K lacking the intracellular domain or H-2K with substitution of the intracellular tyrosine site. Coimmunoprecipitation showed that Fps interacted with wild-type MHC class I molecules but not with those with a mutant intracellular domain or tyrosine site (Fig. 6d). Mutant Fps with deletion of the SH2 domain did not interact with MHC class I either (Fig. 6e), which suggested a nonredundant role for the SH2 domain in the interaction of Fps and MHC class I. Gain-of-func-tion experiments also showed that overexpression of wild-type MHC

PMa cb d e

f

Cyt

15LPS (min) LPS (min)LPS

(0 min)

Fps MHCI Merge MockMockFlag-H2K(WT)

Flag-H2K(del)

Flag-H2KFlag-H2K(mut)

Moc

k

H2K(W

T)

H2K(d

el)

H2K(m

ut)

Moc

k

H2K(W

T)

H2K(d

el)

H2K(m

ut)

Moc

k

H2K(W

T)

H2K(d

el)

H2K(m

ut)

Myc-Fps

Myc-FpsMyc-Fps∆SH2

LPS(5 min)

543210

8* * * * * * ***6420

3

2

IFN

-β (

ng/m

l)

IL-6

(ng

/ml)

TN

F (

ng/m

l)

1

0

IP: MHCI

IB: MHCI

IB: MHCI

IP: IgG IP: IgG

IP: Flag

IB: Myc

IB: Myc

KbDb+/+

KbDb–/–

IP: Myc

IB: Flag

IB: Flag

IB: Fps

IB: Fps

IP: Fps

0 5 15

–

– – –

+

+

+ +

15 0 5 15 15 0 5 15

–

– – –

–

– ––

–– – –

–––

––

– –– –

–– ++

+

+ +++ + +

– – –

+

+

+ + + + +

+

++

+++ +++

+ + + +

Figure 6 Membrane MHC class I molecules interact directly with Fps. (a) Immunoassay of lysates of peritoneal macrophages stimulated for 0, 5 or 15 min with LPS, followed by immunoprecipitation (IP +) of plasma-membrane (PM) or cytoplasmic (Cyt) proteins with IgG or anti–MHC class I and immunoblot analysis (IB) with anti-Fps or anti–MHC class I. (b) Immunoassay of lysates of macrophages stimulated for 0, 5 or 15 min with LPS, followed by immunoprecipitation with IgG or anti-Fps and immunoblot analysis with anti–MHC class I or anti-Fps. (c) Confocal microscopy of macrophages stimulated for 0 or 5 min with LPS and labeled with anti–MHC class I and anti-Fps. Original magnification, ×600. (d) Immunoassay of lysates of macrophages mock transfected (Mock +) or transfected to express Flag-tagged wild-type H-2K (Flag-H2K(WT)), H-2K lacking the intracellular domain (Flag-H2K(del)) or H-2K with substitution of the intracellular tyrosine site (Flag-H2K(mut)), plus Myc-tagged Fps (Myc-Fps), then stimulated with LPS, followed by immunoprecipitation with anti-Flag and immunoblot analysis with anti-Myc or anti-Flag. (e) Immunoassay of lysates of macrophages mock transfected or transfected to express Myc-tagged wild-type Fps (Myc-Fps) or mutant Fps with deletion of the SH2 domain (Myc-Fps∆SH2), plus Flag-tagged wild-type H-2K, then stimulated with LPS, followed by immunoprecipitation with anti-Myc and immunoblot analysis as in d. (f) ELISA of TNF, IL-6 and IFN-β in supernatants of KbDb+/+ and KbDb−/− macrophages mock transfected or transfected to express wild-type H-2K or the H-2K mutants in d. *P < 0.01 (Student’s t-test). Data are representative of three independent experiments with similar results (a–e) or are from three independent experiments (f; mean ± s.e.m.).

npg

© 2

012

Nat

ure

Am

eric

a, In

c. A

ll rig

hts

rese

rved

.

556  VOLUME 13 NUMBER 6 JUNE 2012 nature immunology

A rt i c l e s

class I significantly suppressed the production of TNF, IL-6 and IFN-β in KbDb−/− macrophages, whereas overexpression of mutant MHC class I did not have this effect (Fig. 6f). Therefore, the tyrosine site of MHC class I and the SH2 domain of Fps were required for their interaction and for the suppressive effect of MHC class I molecules on TLR-triggered responses.

MHC class I suppresses TLR signaling via Fps activationAs tyrosine-phosphorylation of the intracellular domain of MHC class I was required for binding to the SH2 domain of Fps, we further determined which signaling molecules contributed to this phosphor-ylation. The kinase Src is suggested to phosphorylate MHC class I molecules in Jurkat human T lymphocyte cells24. It has been shown that Src is activated in macrophages after stimulation with TLR ligands20 and is able to restrain TLR-triggered cytokine produc-tion. Suppression of Src activity with its inhibitor PP1 substantially suppressed the phosphorylation of MHC class I and resulted in less Fps–MHC class I association in macrophages after stimulation with LPS (Supplementary Fig. 5a,b). Fps activation was also much lower when Src activity was inhibited (Supplementary Fig. 5c), which sug-gested a nonredundent role for Src in the interaction of MHC class I and Fps and also in the activation of Fps. Thus, Src may be involved in phosphorylation of the cytoplamic tail of MHC class I, allow-ing the association of MHC class I with Fps, which then transduces downstream signals.

To investigate the exact role of Fps in the negative regulation of TLR signaling by MHC class I molecules, we examined the activa-tion of Fps after stimulation with TLR ligands. Indeed, there was much less TLR ligand–induced phosphorylation of Fps in B2m−/− or KbDb−/− macrophages (Fig. 7a and data not shown). Crosslinkage of MHC class I molecules further enhanced their interaction with Fps and phosphorylation of Fps (Fig. 7b and Supplementary Fig. 6), which indicated that MHC class I molecules may have increased or maintained the TLR-triggered activation of Fps. Silencing of Fps resulted in more production of TNF, IL-6 and IFN-β in macrophages stimulated with LPS, poly(I:C) or CpG ODN (Fig. 7c,d), which sug-gested a negative role for Fps in TLR responses. We also found that knockdown of Fps was less potent in suppressing the TLR-triggered production of cytokines in B2m−/− macrophages than it was in wild-type macrophages (Fig. 7e). Accordingly, the inhibitory effect of crosslinkage of MHC class I molecules on TLR signaling was also substantially attenuated in macrophages in which the gene encoding Fps was silenced (data not shown). Together these data suggested

that membrane MHC class I molecules suppressed TLR-triggered inflammatory responses by associating with Fps and maintaining its activation.

Fps suppresses TLR-triggered response by activating SHP-2Fps activates the cell-adhesion molecule PECAM-1 (CD31) to medi-ate the suppressive effect of the receptor FcεRI on mast-cell activa-tion25, and published data have suggested that PECAM-1 can recruit SHP-2 to suppress TLR4 signaling26. Accordingly, we found that acti-vation of SHP-2 in B2m−/− or KbDb−/− macrophages was considerably attenuated after stimulation with LPS (Fig. 8a and data not shown), which indicated that SHP-2 may have been involved in the suppres-sion of TLR signaling by MHC class I molecules. Deficiency in SHP-2 enhanced the production of proinflammatory cytokines and IFN-β in macrophages stimulated with LPS, poly(I:C) or CpG ODN (Fig. 8b). However, the inhibitory effect of crosslinkage of MHC class I on TLR-triggered cytokine production was considerably attenuated in SHP-2-deficient macrophages (Fig. 8c). These data indicated that activation of SHP-2 might have contributed to the suppression of TLR signaling in macrophages by MHC class I molecules.

Next we investigated whether Fps contributed to the activation of SHP-2 enhanced by MHC class I molecules. After stimulation with TLR ligands, the phosphorylation of SHP-2 was considerably attenu-ated in macrophages in which the gene encoding Fps was silenced (Fig. 8d), which suggested that SHP-2 functions downstream of Fps. To more convincingly demonstrate the link between SHP-2 and Fps, we did coimmunoprecipitation assays and found an association between SHP-2 and Fps in resting macrophages, which was enhanced by stimulation with LPS (Fig. 8e). However, MHC class I molecules did not directly interact with SHP-2 (data not shown).

SHP-2 suppresses TLR-triggered TRIF-dependent signal path-ways by binding to and inhibiting activation of the kinase TBK1 (ref. 27). However, the molecules that interact with SHP-2 in the MyD88-dependent pathway have not been identified so far. As activation of both NF-κB and MAPKs was enhanced in MHC class I–deficient macrophages, we investigated whether SHP-2 interacted with MyD88, the signaling molecule TRAF6, the kinase IRAK1 and other molecules involved in early events in TLR signaling. Among those, only TRAF6 interacted with SHP-2, a result con-firmed by reverse coimmunoprecipitation (Fig. 8f). The carboxyl terminus of SHP-2 was responsible for its interaction with TRAF6 (Supplementary Fig. 7a). A luciferase reporter assay also showed that SHP-2 suppressed TRAF6-triggered activation of NF-κB in a

LPS (min)

a c

e

db

0 15 30

B2m+/+ B2m–/–

B2m+/+ B2m–/–

45 60 0 15 30 45 60 0 15

lgG H-2Kb

30 45 60 0 15 30 45 60

p-FpsFpsβ-actinIP: Fps IP: Fps

Fps

Ctrl FpssiRNA:

Ctrl Fps

TN

F (

ng/m

l)

IL-6

(ng

/ml)

IL-6

(ng

/ml)

IFN

-β (

ng/m

l)3 4 2.52.01.51.00.5

0

3210

**

** **** ****

**

**

** **

****

**

**** **

***2

1

PBSLP

SCpG

Poly(I:

C)PBS

LPS

CpG

Poly(I:

C)PBS

LPS

CpG

Poly(I:

C)0

siRNA: Ctrl Fps

LPS Poly(I:C)

CpG

Ctrl Fps Ctrl Fps siRNA: Ctrl Fps

LPS Poly(I:C)

CpGCtrl Fps Ctrl Fps

TN

F (

ng/m

l)

3 6543210

2

1

0

LPS (min)p-Fps

Fps

IFN

-β (

ng/m

l) 2.52.01.51.00.5

0

** **

**

* *

siRNA: Ctrl Fps

LPS Poly(I:C)

CpG

Ctrl Fps Ctrl Fps

Figure 7 MHC class I molecules suppress TLR-triggered response by maintaining Fps activation. (a,b) Immunoassay of B2m+/+ and B2m−/− macrophages (a) or wild-type macrophages crosslinked with IgG or anti–H-2Kb (b), stimulated for 0–60 min (above lanes) with LPS, followed by immunoprecipitation with anti-Fps and immunoblot analysis of phosphorylated Fps (p-Fps; detected by probing for phosphorylated tyrosine) or total Fps. (c) Immunoblot analysis of Fps in lysates of macrophages transfected for 48 h with control or Fps-specific siRNA. (d,e) ELISA of TNF, IL-6 and IFN-β in supernatants of wild-type C57BL/6 macrophages (d) or B2m+/+ and B2m−/− macrophages (e) transfected with control or Fps-specific siRNA and stimulated for 4 h with PBS, LPS, poly(I:C) or CpG ODN. *P < 0.05 and **P < 0.01 (Student’s t-test). Data are representative of three independent experiments with similar results (a–c) or are from three independent experiments (d,e; mean ± s.e.m.).

npg

© 2

012

Nat

ure

Am

eric

a, In

c. A

ll rig

hts

rese

rved

.

nature immunology  VOLUME 13 NUMBER 6 JUNE 2012 557

A rt i c l e s

dose-dependent manner (Supplementary Fig. 7b). As ubiquiti-nation of TRAF6 is required in MyD88-dependent signaling, we further compared the ubiquitination of TRAF6 in wild-type and SHP-2-deficient macrophages. SHP-2 deficiency resulted in more LPS-induced polyubiquitination of TRAF6 (Fig. 8g), which indi-cated that SHP-2 suppressed the MyD88 pathway by interacting with TRAF6 and controlling its ubiquitination. Collectively, these data suggested that MHC class I molecules negatively regulated TLR-triggered inflammatory responses by enhancing activation of the Fps–SHP-2 pathway (Supplementary Fig. 8).

DISCUSSIONReverse signaling by members of the TNF family28 (including FasL, 4-1BBL, CD40L, LIGHT, membrane TNF and so on) and the B7 family29,30 (including CD80, CD86, B7-H1, B7-DC and so on) has been under extensive exploration. Reverse signaling mediated by MHC class I molecules on the cell surface has also been demonstrated11–15, but there have been no reports on this signaling and its subsequent effects on myeloid cells. Here we have provided evidence that MHC class I molecules on macrophages stimulated with TLR ligands and/or after ligation with certain monoclonal antibodies may interact directly with Fps, maintaining its phosphorylation. Fps then recruits SHP-2, and SHP-2 participates at least partially in the suppression of TLR sig-naling by MHC class I molecules. Therefore, we have demonstrated a previously unknown nonclassical function for MHC class I molecules in which they are involved, via reverse signaling, in the negative regu-lation of TLR-triggered innate inflammatory responses.

Adaptive T cells are able to temper innate immune responses; however, details of the mechanisms involved are unknown31. In the

clinical context, patients with type I bare lymphocyte syndrome (also known as TAP-deficiency syndrome) lack almost all surface expression of MHC class I molecules and CD8+ T cells32,33. These patients have high concentrations of C-reactive protein and a high erythrocyte-sedimentation rate during acute infection, which sug-gests they have enhanced innate inflammatory responses33. We indeed found that CD8+ T cells suppressed the macrophage response to TLRs both in vitro and in vivo via MHC class I molecules on macrophages. However, we did not obtain details about the exact ligands of MHC class I on CD8+ T cells, which may be CD8, TCRαβ or one of the NK cell–related inhibitory receptors. Furthermore, other ligands or receptors for MHC class I molecules expressed on monocytes and macrophages have been suggested, such as the inhibitory receptors ILT-2 and ILT-4 (ref. 34). During activation in vitro or in vivo, those MHC class I ligands, and even unknown ligands, may interact in cis or in trans with MHC class I molecules on the membrane, providing the reverse signal to MHC class I molecules, and, finally, may orchestrate the inhibitory effect of MHC class I on TLR signaling.

Accumulating evidence suggests that a pool of MHC class I mol-ecules present at the plasma membrane can dissociate from the light-chain β2m and open their structure, resulting in MHC class I molecules known as ‘open conformers’ or ‘misfolding conformers’35. These are sparse in quiescent cells but increase considerably in abun-dance when the cells are metabolically activated. The open conformers are not stable and tend to associate with certain receptors or even with themselves, which is believed to be the basis for the reverse signaling of the biological effect of MHC class I molecules35. Notably, the cyto-plasmic tail of MHC class I, which is somewhat longer than that of MHC class II, has one conserved tyrosine residue, which provides a

Figure 8 Binding of SHP-2 to Fps and enhanced activation of SHP-2 are required for negative regulation of TLR responses by MHC class I molecules. (a) Immunoblot analysis of total and phosphorylated SHP-2 in lysates of B2m+/+ and B2m−/− peritoneal macrophages stimulated for 0–60 min (above lanes) with LPS. (b) ELISA of TNF, IL-6 and IFN-β in supernatants of macrophages with loxP-flanked alleles encoding SHP-2 (Ptpn11fl/fl; called ‘Shp2fl/fl’ here) and SHP-2-deficient macrophages (Ptpn11−/−; called ‘Shp2−/−’ here) stimulated for 4 h with PBS, LPS, poly(I:C) or CpG ODN. (c) ELISA of TNF, IL-6, IFN-β in supernatants of macrophages of the genotypes in b, treated with IgG or anti–H-2Kb (for crosslinking of MHC class I) and stimulated as in b. (d) Immunoblot analysis of phosphorylated SHP-2 in wild-type macrophages transfected with control or Fps-specific siRNA and then stimulated for 0–60 min (above lanes) with LPS. Total SHP-2 serves as loading control. (e) Immunoblot analysis of Fps and SHP-2 in lysates of macrophages stimulated for 0, 5 or 15 min with LPS and subjected to immunoprecipitation with IgG or anti–SHP-2 (left) or anti-Fps (right). (f) Immunoblot analysis of TRAF6 and SHP-2 in lysates of LPS-stimulated macrophages subjected to immunoprecipitation with IgG or anti-TRAF6 (left) or anti–SHP-2 (right). (g) Immunoblot analysis of lysates of LPS-stimulated macrophages (genotypes, as in b) subjected to immunoprecipitation with anti-TRAF6 and probed with anti-ubiquitin (Ub; above) or anti-TRAF6 (below). kDa, kilodaltons. *P < 0.05 and **P < 0.01 (Student’s t-test). Data are representative of three independent experiments with similar results (a,d–g) or are from three independent experiments (b,c; mean ± s.e.m.).

a

c

d

f g

e

bB2m+/+

Shp2fl/fl Shp2–/–

Shp2 Shp2

B2m–/–

Shp2fl/fl Shp2–/–

LPS (min)

p-SHP-2

0 15 30 45 60 0 15 30 45 60

0 15 30 45 60 0 15 30 45 60

3 5 2.5

4

3

2

1

0

5

6

4

3

2

1

0

3

2

1

0

4

3

2

1

0

2

1

0

PBSPBS

LPS

LPS

Poly(I:

C)

Poly(I:

C)CpG

LPS LPSPoly(I:C) Poly(I:C)CpG CpG LPS Poly(I:C) CpG

IgG

fl/fl fl/fl fl/fl–/– –/– fl/fl –/– fl/fl –/– fl/fl –/– Shp2 fl/fl –/– fl/fl –/– fl/fl –/––/–

H-2Kb

CpG PBSLP

S

Poly(I:

C)CpG

SHP-2

LPS (min)

Ctrl siRNA Fps siRNA

p-SHP-2

SHP-2

LPS (min)

IP: TRAF6

IB: TRAF6

IB: SHP-2

IP: IgG

LPS (min) LPS (min)

IP: SHP-2IP: IgG IP: IgG

15 0 5 15

++++–

– – –

15 0 15 60

++++–

– –

30

+– –

LPS (min)LPS (min)

IP: SHP-2

IB: TRAF6

IP: TRAF6

(kDa)170130

9572

IB: TRAF6IP: TRAF6

IB: Ub

IB: SHP-2

IP: IgG

15 0 15 60

++++–

– –

306030

+– –

15

15

0

0 6030150

5 15

++++–

– – –IB: Fps

IB: Fps

IP: Fps

IB: SHP-2

IB: SHP-2

TN

F (

ng/m

l)

TN

F (

ng/m

l)

IL-6

(ng

/ml)

IL-6

(ng

/ml)

IFN

-β (

ng/m

l)

IFN

-β (

ng/m

l)

**

**** **

**** **

**

***

****

****

** **

**

**2.0

1.5

1.0

0.5

0

npg

© 2

012

Nat

ure

Am

eric

a, In

c. A

ll rig

hts

rese

rved

.

558  VOLUME 13 NUMBER 6 JUNE 2012 nature immunology

A rt i c l e s

recognition site for the SH2 domain of kinases11,24. In addition, phos-phorylation of the tyrosine residue in the cytoplasmic domain of MHC class I molecules is also related to the formation of open conform-ers and may influence or reflect interaction with certain molecules and the initiation of MHC class I reverse-signaling pathways24,35,36. Moreover, Fps can be activated after stimulation by certain cytokines, such as GM-CSF, IL-4, IL-6, erythropoietin and so on, and has been detected as being associated with phosphorylated receptors of these cytokines, including IL-4Rα, gp130 and the common β-chain37–39. Therefore, we predicted and have proven that phosphorylation of Tyr321 of intracellular MHC class I was necessary for its association with the SH2 domain of Fps and that this interaction may contribute to maintenance of Fps activation and the suppressive function of Fps on TLR-triggered responses40.

Fps is a member of the non-receptor tyrosine kinase family, which has critical roles in regulating cytoskeletal rearrangements and the inside-out signaling involved in receptor-ligand, cell-matrix and cell-cell interactions41,42. Fps is expressed in hematopoietic cells of the myeloid lineage, including macrophages, neutrophils and so on, and is essential for the survival and terminal differentiation of myeloid-progenitor cells. Mice deficient in Fps expression are hyper-responsive to endotoxin challenge23, although the underlying mechanisms have not been fully elucidated. Here we have shown that Fps associated with SHP-2 and that this association had a nonredundant role in the suppression of TLR responses by MHC class I molecules.

SHP-2 is a non-receptor tyrosine phosphatase whose ubiquitous expression pattern is similar to that of MHC class I molecules43. SHP-2 functions as a negative regulator in various signaling path-ways through its phosphatase activity. Here, SHP-2 inhibited TLR signaling by binding to and controlling the activation of TRAF6 and TBK1 of the MyD88-dependent pathway and TRIF-dependent pathway27, respectively. In addition to SHP-2, Fps may also suppress TLR signaling via other mechanisms. It has been reported that Fps-deficient macrophages have impaired internalization of TLR4 when stimulated with LPS44, which may provide another explanation for the Fps-mediated suppression.

In conclusion, our results have demonstrated that membrane MHC class I molecules interacted with Fps and subsequently induced SHP-2 activation in response to stimulation with TLR ligands and, finally, suppressed innate inflammatory responses. Therefore, constitutively expressed MHC class I molecules are required for maintenance of the quiescent state and fine tuning of TLR-triggered innate inflammatory responses in macrophages. Our findings provide new insight into the negative regulation of TLR signaling and indicate a previously unidentified nonclassical function for MHC class I molecules in the regulation of innate inflammatory responses.

METHODSMethods and any associated references are available in the online version of the paper.

Note: Supplementary information is available in the online version of the paper.

ACknoWLedgmentSWe thank G. Feng (University of California, San Diego) for mice with loxP-flanked alleles encoding SHP-2; H. Shen (University of Pennsylvania School of Medicine) for L. monocytogenes; N. van Rooijen (Free University of Amsterdam) for liposomes; J. Long, X. Zuo and P. Ma for technical assistance; and T. Chen and Y. Han for discussions. Supported by the National Key Basic Research Program of China (2012CB910202 and 2010CB911903), the National 125 Key Project (2012ZX10002-014 and 2012AA020901), the National Natural Science Foundation of China (81123006) and the Shanghai Committee of Science and Technology (10DZ1910300).

AUtHoR ContRIBUtIonSX.C. and S.X. designed the experiments; S.X., X.L., Y.B., C.H., X.Zhu, P.Z., W.L. and X.Zha. did the experiments; X.C., S.X. and X.L. analyzed data and wrote the paper; and X.C. was responsible for research supervision, coordination and strategy.

ComPetIng FInAnCIAL InteReStSThe authors declare no competing financial interests.

Published online at http://www.nature.com/doifinder/10.1038/ni.2283. reprints and permissions information is available online at http://www.nature.com/reprints/index.html.

1. Kawai, T. & Akira, S. The role of pattern-recognition receptors in innate immunity: update on Toll-like receptors. Nat. Immunol. 11, 373–384 (2010).

2. Iwasaki, A. & Medzhitov, R. Regulation of adaptive immunity by the innate immune system. Science 327, 291–295 (2010).

3. O’Neill, L.A. & Bowie, A.G. The family of five: TIR-domain-containing adaptors in Toll-like receptor signalling. Nat. Rev. Immunol. 7, 353–364 (2007).

4. Liew, F.Y., Xu, D., Brint, E.K. & O’Neill, L.A. Negative regulation of toll-like receptor-mediated immune responses. Nat. Rev. Immunol. 5, 446–458 (2005).

5. Antoniou, A.N., Powis, S.J. & Elliott, T. Assembly and export of MHC class I peptide ligands. Curr. Opin. Immunol. 15, 75–81 (2003).

6. Raulet, D.H. MHC class I-deficient mice. Adv. Immunol. 55, 381–421 (1994).7. Nitta, T. et al. Thymoproteasome shapes immunocompetent repertoire of CD8+

T cells. Immunity 32, 29–40 (2010).8. Orr, M.T. & Lanier, L.L. Natural killer cell education and tolerance. Cell 142,

847–856 (2010).9. Jonsson, A.H. & Yokoyama, W.M. Natural killer cell tolerance licensing and other

mechanisms. Adv. Immunol. 101, 27–79 (2009).10. Elliott, J.M., Wahle, J.A. & Yokoyama, W.M. MHC class I-deficient natural killer

cells acquire a licensed phenotype after transfer into an MHC class I-sufficient environment. J. Exp. Med. 207, 2073–2079 (2010).

11. Tscherning, T. & Claesson, M.H. Signal transduction via MHC class-I molecules in T cells. Scand. J. Immunol. 39, 117–121 (1994).

12. Skov, S. lntracellular signal transduction mediated by ligation of MHC class I molecules. Tissue Antigens 51, 215–223 (1998).

13. Pedersen, A.E., Skov, S., Bregenholt, S., Ruhwald, M. & Claesson, M.H. Signal transduction by the major histocompatibility complex class I molecule. APMIS 107, 887–895 (1999).

14. Yang, J. et al. Targeting β2-microglobulin for induction of tumor apoptosis in human hematological malignancies. Cancer Cell 10, 295–307 (2006).

15. Zhang, X., Rozengurt, E. & Reed, E.F. HLA class I molecules partner with integrin β4 to stimulate endothelial cell proliferation and migration. Sci. Signal. 3, ra85 (2010).

16. Rubio, G. et al. Cross-linking of MHC class I molecules on human NK cells inhibits NK cell function, segregates MHC I from the NK cell synapse, and induces intracellular phosphotyrosines. J. Leukoc. Biol. 76, 116–124 (2004).

17. Skov, S., Odum, N. & Claesson, M.H. MHC class I signaling in T cells leads to tyrosine kinase activity and PLC-γ1 phosphorylation. J. Immunol. 154, 1167–1176 (1995).

18. Pedersen, A.E., Bregenholt, S., Skov, S., Vrang, M.L. & Claesson, M.H. Protein tyrosine kinases p53/56lyn and p72syk in MHC class I-mediated signal transduction in B lymphoma cells. Exp. Cell Res. 240, 144–150 (1998).

19. Liu, X. et al. Intracellular MHC class II molecules promote TLR-triggered innate immune responses by maintaining activation of the kinase Btk. Nat. Immunol. 12, 416–424 (2011).

20. Han, C. et al. Integrin CD11b negatively regulates TLR-triggered inflammatory responses by activating Syk and promoting degradation of MyD88 and TRIF via Cbl-b. Nat. Immunol. 11, 734–742 (2010).

21. Sun, J.C. & Lanier, L.L. Cutting edge: viral infection breaks NK cell tolerance to “missing self”. J. Immunol. 181, 7453–7457 (2008).

22. Perona-Wright, G. et al. Systemic but not local infections elicit immunosuppressive IL-10 production by natural killer cells. Cell Host Microbe 6, 503–512 (2009).

23. Zirngibl, R.A., Senis, Y. & Greer, P.A. Enhanced endotoxin sensitivity in fps/fes-null mice with minimal defects in hematopoietic homeostasis. Mol. Cell. Biol. 22, 2472–2486 (2002).

24. Santos, S.G., Powis, S.J. & Arosa, F.A. Misfolding of major histocompatibility complex class I molecules in activated T cells allows cis-interactions with receptors and signaling molecules and is associated with tyrosine phosphorylation. J. Biol. Chem. 279, 53062–53070 (2004).

25. Udell, C.M., Samayawardhena, L.A., Kawakami, Y., Kawakami, T. & Craig, A.W. Fer and Fps/Fes participate in a Lyn-dependent pathway from FcεRI to platelet-endothelial cell adhesion molecule 1 to limit mast cell activation. J. Biol. Chem. 281, 20949–20957 (2006).

26. Rui, Y. et al. PECAM-1 ligation negatively regulates TLR4 signaling in macrophages. J. Immunol. 179, 7344–7351 (2007).

27. An, H. et al. SHP-2 phosphatase negatively regulates the TRIF adaptor protein-dependent type I interferon and proinflammatory cytokine production. Immunity 25, 919–928 (2006).

28. Sun, M. & Fink, P.J. A new class of reverse signaling costimulators belongs to the TNF family. J. Immunol. 179, 4307–4312 (2007).

29. Greenwald, R.J., Freeman, G.J. & Sharpe, A.H. The B7 family revisited. Annu. Rev. Immunol. 23, 515–548 (2005).

npg

© 2

012

Nat

ure

Am

eric

a, In

c. A

ll rig

hts

rese

rved

.

nature immunology  VOLUME 13 NUMBER 6 JUNE 2012 559

A rt i c l e s

30. Chen, L. Co-inhibitory molecules of the B7–CD28 family in the control of T-cell immunity. Nat. Rev. Immunol. 4, 336–347 (2004).

31. Kim, K.D. et al. Adaptive immune cells temper initial innate responses. Nat. Med. 13, 1248–1252 (2007).

32. Zimmer, J. et al. Clinical and immunological aspects of HLA class I deficiency. QJM 98, 719–727 (2005).

33. Gadola, S.D., Moins-Teisserenc, H.T., Trowsdale, J., Gross, W.L. & Cerundolo, V. TAP deficiency syndrome. Clin. Exp. Immunol. 121, 173–178 (2000).

34. Colonna, M. et al. Human myelomonocytic cells express an inhibitory receptor for classical and nonclassical MHC class I molecules. J. Immunol. 160, 3096–3100 (1998).

35. Arosa, F.A., Santos, S.G. & Powis, S.J. Open conformers: the hidden face of MHC-I molecules. Trends Immunol. 28, 115–123 (2007).

36. Little, A.M., Nossner, E. & Parham, P. Dissociation of β2-microglobulin from HLA class I heavy chains correlates with acquisition of epitopes in the cytoplasmic tail. J. Immunol. 154, 5205–5215 (1995).

37. Izuhara, K., Feldman, R.A., Greer, P. & Harada, N. Interaction of the c-fes proto-oncogene product with the interleukin-4 receptor. J. Biol. Chem. 269, 18623–18629 (1994).

38. Matsuda, T. et al. Activation of Fes tyrosine kinase by gp130, an interleukin-6 family cytokine signal transducer, and their association. J. Biol. Chem. 270, 11037–11039 (1995).

39. Brizzi, M.F. et al. Granulocyte-macrophage colony-stimulating factor stimulates JAK2 signaling pathway and rapidly activates p93fes, STAT1 p91, and STAT3 p92 in polymorphonuclear leukocytes. J. Biol. Chem. 271, 3562–3567 (1996).

40. Filippakopoulos, P. et al. Structural coupling of SH2-kinase domains links Fes and Abl substrate recognition and kinase activation. Cell 134, 793–803 (2008).

41. Hellwig, S. & Smithgall, T.E. Structure and regulation of the c-Fes protein-tyrosine kinase. Front. Biosci. 17, 3146–3155 (2011).

42. Greer, P. Closing in on the biological functions of Fps/Fes and Fer. Nat. Rev. Mol. Cell Biol. 3, 278–289 (2002).

43. Lorenz, U. SHP-1 and SHP-2 in T cells: two phosphatases functioning at many levels. Immunol. Rev. 228, 342–359 (2009).

44. Parsons, S.A. & Greer, P.A. The Fps/Fes kinase regulates the inflammatory response to endotoxin through down-regulation of TLR4, NF-κB activation, and TNF-α secretion in macrophages. J. Leukoc. Biol. 80, 1522–1528 (2006).

npg

© 2

012

Nat

ure

Am

eric

a, In

c. A

ll rig

hts

rese

rved

.

nature immunology doi:10.1038/ni.2283

ONLINE METHODSMice and reagents. C57BL/6J mice were from Joint Ventures Sipper BK Experimental Animals. Mice deficient in β2m (B6.129P2-B2mtm1Unc/J; 002087), OT-I mice (Tg(TcraTcrb)1100Mjb/J; 003831) and Mx-Cre mice (B6.Cg-Tg(Mx1-cre)1Cgn/J; 003556) were from Jackson Laboratories. Mice deficient in H-2Kb and H-2Db were from Taconic Farms. Shp2fl mice (provided by G. Feng) were crossed with Mx-Cre mice. Shp2fl/fl mice and Mx-Cre × Shp2fl/fl mice were injected intraperitoneally with 250 µg poly(I:C) every other day for a total of five doses before use. All mice were bred in specific pathogen–free conditions. All animal experiments were in accordance with the National Institute of Health Guide for the Care and Use of Laboratory Animals, with the approval of the Scientific Investigation Board of the Second Military Medical University, Shanghai. LPS (from E. coli serotype 0111:B4), CpG ODN and poly(I:C) have been described19. E. coli serotype 0111:B4 was from the China Center for Type Culture Collection. L. monocytogenes was provided by H. Shen. Antibody to MHC class I (anti-H-2Kb; AF6-88.5) was from Biolegend. Anti-TLR4 (ab22048), anti-TLR3 (ab62566) and anti-TLR9 (ab52967) were from Abcam. Antibody to Erk phosphorylated at Thr202-Tyr204 (E10), to Jnk phosphorylated at Thr183-Tyr185 (G9), to p38 phosphor-ylated at Thr180-Tyr182 (9211), to IRF3 phosphorylated at Ser396 (4D4G), to IκBα phosphorylated at Ser32-36 (5A5), to IKKα-IKKβ phosphorylated at Ser176-Ser180 (16A6), to SHP-2 phosphorylated at Tyr580 (3703), to SHP-2 (D50F2) and to the Myc tag (2272) were from Cell Signaling Technology. Anti-Fps (sc-25415), anti-TRAF6 (sc-7221, sc-8409), anti-ubiquitin (P4D1), anti-IRAK1 (sc-5288) and anti-β-actin (sc-130656) were from Santa Cruz. Anti-Flag (M2) was from Sigma.

Cell culture and RNA-mediated interference. Thioglycollate-elicited mouse peritoneal macrophages were prepared and cultured in endotoxin-free RPMI-1640 medium with 10% FCS (Invitrogen) as described19. The siRNA targeting MHC class I α-chain or Fps was from Dharmacon. Mouse perito-neal macrophages were transfected with siRNA duplexes through the use of INTERFERin reagent (Polyplus) according to a standard protocol.

Cytokine detection. TNF, IL-6 and IFN-β in supernatants and serum were measured with ELISA kits (R&D Systems).

RNA quantification. A LightCycler (Roche) and SYBR RT-PCR kit (Takara) were used for quantitative real-time RT-PCR analysis as described19. Data were normalized to β-actin expression.

Immunoprecipitation and immunoblot analysis. Cells were lysed with cell-lysis buffer (Cell Signaling Technology) supplemented with protease inhibitor ‘cocktail’ (Calbiochem). Protein concentrations of the extracts were measured with a BCA assay (Pierce). The immunoprecipitation assays and immunoblot assays were done as described19,20.

Nanospray liquid chromatography–tandem mass spectrometry. Macrophages (3 × 108) were stimulated for 15 min with LPS with or without crosslinkage of MHC class I, then were lysed for preparation of immunoprecip-itates with anti–MHC class I. Proteins were eluted and digested. Digests were analyzed by nano-ultra-performance liquid chromatography–electrospray ionization tandem mass spectrometry19. Data from liquid chromatography–tandem mass spectrometry were processed through the use of ProteinLynx

Global Serverversion 2.4 (PLGS 2.4); the resulting peak lists were used for searching the NCBI protein database with the Mascot search engine.

Cell isolation and in vivo depletion. CD8+ T cells and NK cells were sorted from C57BL/6 or OT-I mice with a MoFlo XDP (DakoCytomation). Naive CD8+ T cells were sorted as CD44loCD62Lhi and memory CD8+ T cells were sorted as CD44hiCD62Llo. Cell purity was >97%. For depletion of autologous CD8+ T cells or NK cells in vivo, mice were given monoclonal anti-CD8 (2.43; American Type Culture Collection) or anti-NK1.1 (PK136; American Type Culture Collection), respectively (each at a dose of 200 µg per mouse), 3 d before infection. Depletion frequency was confirmed as being >90%.

Confocal microscopy. Macrophages, plated on glass coverslips in six-well plates, were left unstimulated or stimulated with LPS and then labeled with anti–MHC class I, anti-Fps, anti-TLR4, anti-TLR3, anti-TLR9 or anti-EEA1 (C45B10; Cell Signaling Technology). Cells were observed with a Leica TCS SP2 confocal laser microscope.

Establishment of the endotoxin-shock model and bacterial infection. The endotoxin-shock mouse model was established by intraperitoneal injection of LPS (15 mg per kg body weight (mg/kg)) as described20. Serum TNF, IL-6 and IFN-β were measured by ELISA in mice given intraperitoneal injection of LPS (5 mg/kg), poly(I:C) (20 mg/kg) or CpG ODN (20 mg/kg). For bacterial infec-tion, E. coli serotype 0111:B4 and L. monocytogenes in mid-logarithmic growth were collected, counted on agar plates and then resuspended in PBS. Mice were given intraperitoneal injection of 1 × 107 E. coli or intravenous injection of 1 × 104 L. monocytogenes. Serum was collected for measurement of cytokines (or colony-forming units for E. coli) and spleens or livers were lysed for mea-surement of colony-forming units (L. monocytogenes) as described20,45.

Bone marrow transplantation. Bone marrow cells (1 × 107) from B2m−/− mice, KbDb−/− mice or their littermates (control mice) were transplanted via tail-vein injection into lethally irradiated wild-type mice (10 Gy). After 8 weeks, CD8+ T cells in the spleen and lymph nodes were counted and MHC class I expression on macrophages was analyzed by flow cytometry.

Macrophage reconstitution. Bone marrow cells from B2m−/− mice, KbDb−/− mice or their littermates (control mice) were cultured for 7 d in mouse macro-phage colony-stimulating factor (50 ng/ml; PeproTech) for the preparation of bone marrow–derived macrophages. Clodronate liposomes were injected intraperitoneally into wild-type recipient mice (50 mg in 200 µl per mouse) for depletion of endogenous macrophages. Then 2 d later, B2m+/+ or B2m−/− bone marrow–derived macrophages (1 × 107) were transplanted into the recipient mice by tail-vein injection 6 h before challenge with LPS.

Statistical analysis. The statistical significance of comparisons between two groups was determined with Student’s t-test. The statistical significance of survival curves was estimated according to the method of Kaplan and Meier, and the curves were compared with the generalized Wilcoxon’s test. P values of less than 0.05 were considered statistically significant.

45. Xu, S. et al. IL-17A-producing γδT cells promote CTL responses against Listeria monocytogenes infection by enhancing dendritic cell cross-presentation. J. Immunol. 185, 5879–5887 (2010).

npg

© 2

012

Nat

ure

Am

eric

a, In

c. A

ll rig

hts

rese

rved

.