Mitogen-activated protein kinase kinase 3is a pivotal pathway regulating p38activation in inflammatory arthritisTomoyuki Inoue*, David L. Boyle*, Maripat Corr*, Deepa Hammaker*, Roger J. Davis†, Richard A. Flavell‡,and Gary S. Firestein*§

*Division of Rheumatology, Allergy, and Immunology, University of California at San Diego School of Medicine, La Jolla, CA 92093-0656; †Department ofBiochemistry and Molecular Pharmacology, University of Massachusetts School of Medicine, Worcester, MA 01605; and ‡Department of Immunobiology,Yale University School of Medicine, New Haven, CT 06520

Edited by Dennis A. Carson, University of California at San Diego School of Medicine, La Jolla, CA, and approved February 10, 2006 (received for reviewOctober 20, 2005)

p38 mitogen-activated protein kinase (MAPK) regulates cytokinesin arthritis and is, in turn, regulated by MAPK kinase (MKK) 3 andMKK6. To modulate p38 function but potentially minimize toxicity,we evaluated the utility of targeting MKK3 by using MKK3�/�

mice. These studies showed that TNF-� increased phosphorylationof p38 in WT cultured synoviocytes but that p38 activation, IL-1�,and IL-6 expression were markedly lower in MKK3�/� synovio-cytes. In contrast, IL-1� or LPS-stimulated p38 phosphorylation andIL-6 production by MKK3�/� synoviocytes were normal. Detailedsignaling studies showed that NF-�B also contributes to IL-6 pro-duction and that TNF-�-induced NF-�B activation is MKK3-depen-dent. In contrast, LPS-mediated activation of NF-�B does notrequire MKK3. To determine whether this dichotomy occurs in vivo,two inflammation models were studied. In K�BxN passive arthritis,the severity of arthritis was dramatically lower in MKK3�/� mice.Phospho-p38, phospho-MAPK activator protein kinase 2, IL-1�, CXCligand 1, IL-6, and matrix metalloproteinase (MMP) 3 levels in thejoints of MKK3�/� mice were significantly lower than in controls.Exogenous IL-1� administered during the first 4 days of the passivemodel restored arthritis to the same severity as in WT mice. In thesecond model, IL-6 production after systemic LPS administrationwas similar in WT and MKK3�/� mice. Therefore, selective MKK3deficiency can suppress inflammatory arthritis and cytokine pro-duction while Toll-like receptor 4-mediated host defense remainsintact.

cytokines � inflammation � rheumatoid arthritis � fibroblast �signal transduction

The mitogen-activated protein kinases (MAPKs) are signaltransduction enzymes that integrate cellular responses to the

environmental stresses. The pathways include parallel and in-teracting cascades that are initiated by activation of upstreamMAPK kinase (MKK) kinases, which, in turn, activate MKKs.These enzymes then phosphorylate the three main MAPKfamilies, extracellular signal-regulated kinase (ERK), c-Jun N-terminal kinase (JNK), and p38 MAPK. In addition to partici-pating in many normal host functions, the MAPKs have beenimplicated in inflammatory diseases (1). p38, in particular, isthought to play a role in rheumatoid arthritis (RA). It isexpressed and activated in the rheumatoid synovial intimal lining(2), and p38 inhibitors are effective in rodent models of arthritis(3). Therefore, p38 could serve as a therapeutic target (4, 5).Inhibition of this kinase in humans has been complicated byseveral problems, such as hepatotoxicity and skin rash (6).Additional concerns related to impaired resistance to infectionhave been raised with p38 inhibitors because of effects on hostdefense, including suppression of neutrophil killing and en-hanced colony counts in bacterial infections (7, 8).

An alternative approach is to target upstream MKKs, such asMKK3 or MKK6, which could potentially regulate some p38

functions while leaving others intact. MKK3 and MKK6 are twoclosely related kinases that phosphorylate p38 at the Thr-Gly-Tyr site but do not activate ERK1�2 or JNK (9, 10). MKK4 is aJNK-activating kinase but is also able to phosphorylate p38 whenit is overexpressed (11). The hierarchy of MKK3 and MKK6varies considerably depending on the cell lineage and the type ofenvironmental stimulus (12), raising the possibility that onecould block an upstream kinase that will have an impact on onlya subset of functions that are relevant to disease while leavingother p38-mediated responses intact.

We previously demonstrated that both MKK3 and MKK6 canphosphorylate p38 in cultured fibroblast-like synoviocytes (FLS)and are activated in RA synovium (13). The relative contribu-tions of these two kinases to p38 activation in synoviocytes havealso been examined by using dominant negative (DN) MKK3and MKK6 constructs (14). Among the different MKK subtypes,these data suggested that MKK3 might play a more prominentrole in p38-mediated cytokine and MMP expression in FLS. Thepresent study was performed to determine the hierarchy of thep38 MAP kinases in arthritis and assess the potential for MKK3as a therapeutic target. These experiments demonstrate thatMKK3 deficiency markedly decreases synovial inflammation inthe passive K�BxN model of arthritis, in part through defectiveIL-1� expression, while still permitting normal responses to LPSthrough a NF-�B-dependent mechanism.

ResultsMAPK Signaling in TNF-�-Activated FLS. Expression of phospho-p38, phospho-JNK, and phospho-ERK in medium and TNF-�-stimulated cells was initially determined by Western blot anal-ysis. TNF-� increased phosphorylation of p38 in the WT cells ina dose-dependent fashion, with maximum activation occurring at�10 ng�ml (data not shown). JNK and ERK phosphorylationwere normal in MKK3�/� FLS, although phosphorylation of p38was significantly lower in MKK3�/� cells (66% inhibition com-pared with WT; P � 0.05, n � 3) (Fig. 1A). In vitro kinase assayson WT and MKK3�/� FLS lysates were then performed by usingGST-activating transcription factor (ATF) 2 as the substrate. Asshown in Fig. 1B, marked increases of kinase activity were

Conflict of interest statement: No conflicts declared.

This paper was submitted directly (Track II) to the PNAS office.

Abbreviations: RA, rheumatoid arthritis; FLS, fibroblast-like synoviocytes; MAPK, mitogen-activated protein kinase; MKK, MAPK kinase; ATF, activating transcription factor; MMP,matrix metalloproteinase; TLR, Toll-like receptor; ERK, extracellular signal-regulated ki-nase; JNK, c-Jun N-terminal kinase; DN, dominant negative; siRNA, small interfering RNA;IKK, I�B kinase; APK, activator protein kinase; rhTNF-�, recombinant human TNF-�.

§To whom correspondence should be addressed at: Division of Rheumatology, Allergy, andImmunology, University of California at San Diego School of Medicine, 9500 Gilman Drive,La Jolla, CA 92093-0656. E-mail: gfirestein@ucsd.edu.

© 2006 by The National Academy of Sciences of the USA

5484–5489 � PNAS � April 4, 2006 � vol. 103 � no. 14 www.pnas.org�cgi�doi�10.1073�pnas.0509188103

Dow

nloa

ded

by g

uest

on

Mar

ch 3

, 202

0

observed in TNF-�-stimulated FLS. Kinase activity was signif-icantly lower in MKK3�/� cells compared with WT cells (42%inhibition; P � 0.05, n � 3).

MKK3-Dependent IL-6 and IL-1� Production After TNF-� Stimulation.To determine whether MKK3 regulates production of p38-driven cytokines in FLS, WT and MKK3�/� cells were treatedwith medium or TNF-�, and culture supernatants were col-lected after 24 h. As shown in Fig. 2A, TNF-� significantlyincreased IL-6 production by WT FLS. However, IL-6 pro-duction by the MKK3�/� FLS after TNF-� stimulation wasmarkedly lower than by WT cells (P � 0.01). The p38 inhibitorSB203580 (10 �M) also blocked IL-6 production by the WTcells, confirming p38 dependence of the TNF-� pathway.SB203580 further decreased IL-6 production in the MKK3�/�

cells, suggesting that a small amount of residual p38 activationmight still contribute. Quantitative real-time PCR showed thatTNF-�-induced IL-6 mRNA accumulation was inhibited in theMKK3-deficient cells (Fig. 2B). As with IL-6, TNF-�-inducedIL-1� mRNA induction was suppressed in the MKK3�/� FLSas well as by SB203580 (for WT FLS, TNF-� stimulationresulted in 16.2 � 5.2-fold change in IL-1��hypoxanthinephosphoribosyltransferase mRNA expression, and TNF-� plusSB203580 resulted in 0.9 � 0.5-fold change; for MKK3�/�

cells, TNF-� stimulation resulted in 3.0 � 1.2-fold change, andTNF-� plus SB203580 resulted in 1.1 � 1.0-fold change; P �0.01 for WT compared with MKK3�/�, n � 4).

MKK3-Independent IL-6 and IL-1� Production After Toll-Like Receptor(TLR) 4 and IL-1� Stimulation. To determine the effect of MKK3deficiency on TLR4-dependent innate immune responses, weexamined the contribution of MKK3 to LPS-induced cytokineproduction by FLS. As shown in Fig. 2C, LPS-induced IL-6protein production from MKK3�/� FLS was similar to WT FLSand was blocked by SB203580. Similar results were obtained byusing another selective p38 inhibitor, SB202190 (data notshown). In addition, p38� small interfering RNA (siRNA)

knock-down also inhibited LPS- and TNF-�-induced IL-6 pro-duction (Fig. 2D). LPS-induced IL-1� mRNA expression wasalso similar in WT and MKK3�/� FLS (IL-1��hypoxanthinephosphoribosyltransferase mRNA induction was 5.9 � 0.4-fold

Fig. 1. MAPK signaling in TNF-�-stimulated WT and MKK3�/� FLS. CulturedFLS from WT and MKK3�/� mice were stimulated with medium or TNF-� for 15min. Cell lysates were evaluated by Western blot analysis. (A) Representativeblot shows inhibition of p38 but not JNK or ERK phosphorylation (n � 3). (B)Cell lysates were immunoprecipitated with anti-p38, and kinase activity wasmeasured, using GST-ATF-2 as the substrate (n � 3).

Fig. 2. Regulation of TNF-�-, IL-1�-, or LPS-induced IL-6 expression by WT andMKK3�/� FLS. Cultured FLS from WT and MKK3�/� mice were stimulated withor without TNF-�, IL-1�, or LPS for 24 h. Culture supernatants and total RNAwere collected. IL-6 protein and mRNA levels were measured by ELISA andreal-time PCR, respectively. (A and B) TNF-�-induced IL-6 protein and mRNAwere significantly decreased in MKK3�/� FLS. (C) LPS-induced IL-6 levels weresimilar in WT and MKK3�/� FLS. However, SB203580 (10 mM) blocked IL-6production. (D) Effect of p38� siRNA knock-down. FLS were transfected witheither p38� siRNA or scrambled control (sc) and stimulated with TNF-� or LPS,and supernatants were assayed for IL-6 by ELISA (n � 4). Knock-down wasconfirmed by Western blot analysis (not shown). siRNA blocked IL-6 produc-tion, although the presence of other p38 isoforms or small amounts of residualp38� could still affect the results. (E) IL-1�-induced IL-6 levels were notchanged in WT and MKK3�/� FLS. However, both are blocked by SB203580 (10mM). *, P � 0.05; **, P � 0.01 compared with WT, TNF-�, IL-1�, or LPS control.(F) Cultured FLS from WT and MKK3�/� mice were stimulated with TNF-� or LPSfor 15 min. Western blot analysis was performed to evaluate p38 activation.TNF-� was a more potent inducer of phospho-p38 than LPS in WT FLS.However, LPS-stimulated p38 phosphorylation was independent of MKK3.

Inoue et al. PNAS � April 4, 2006 � vol. 103 � no. 14 � 5485

MED

ICA

LSC

IEN

CES

Dow

nloa

ded

by g

uest

on

Mar

ch 3

, 202

0

and 4.7 � 1.6-fold, respectively; P � 0.10). As shown in Fig. 2E,MKK3 deficiency had no effect on IL-1�-induced IL-6 produc-tion, which is consistent with similarity between LPS and IL-1signaling pathways (15).

Activation of p38 by LPS in MKK3�/� FLS. Because of the differentialeffects of TNF-� and LPS in the MKK3�/� FLS, we examinedthe ability of each ligand to activate p38. Fig. 2F shows thatTNF-� is a more potent inducer of p38 phosphorylation thanLPS in WT FLS. In MKK3�/� FLS, however, TNF-� was muchless effective compared with WT FLS. Although LPS was lesspotent in the WT cells, there was no difference in p38 activationbetween WT and MKK3�/� FLS. Therefore, unlike TNF-�,LPS-mediated p38 phosphorylation is independent of MKK3.

Activation of NF-�B by LPS and TNF-� in MKK3�/� FLS. The obser-vation that LPS is a less potent activator of p38 (Fig. 2F) suggeststhat another pathway might also contribute to the high IL-6production despite modest LPS-induced phospho-p38 levels. Toevaluate a potential role for NF-�B, which has been implicatedin IL-6 regulation by FLS, WT and MKK3�/� cells werestimulated with LPS or TNF-� for 1 h, and EMSAs wereperformed. As shown in Fig. 3A, NF-�B activation was robust inboth TNF-�- and LPS-stimulated WT FLS. However, LPSstimulation in MKK3�/� cells had normal NF-�B responses,whereas MKK3 deficiency significantly decreased TNF-�-mediated NF-�B activation (63% inhibition; P � 0.01, n � 3).For comparison, activator protein 1 activation was not affectedby the MKK3 status of cells (Fig. 3B). These data suggest thatNF-�B is not only required for IL-6 but also that it might accountfor the ability of LPS to overcome limited p38 activation in WTand MKK3�/� FLS. To prove the role of NF-�B, WT andMKK3�/� FLS were infected with an I�B kinase (IKK) 2 DNadenovirus. As shown in Fig. 3C, LPS-induced IL-6 productionwas blocked by IKK2 DN treatment in both WT and MKK3�/�

FLS.

Role of MKK3 in Passive K�BxN Arthritis. Because p38 activation andcytokine production in TNF-�-stimulated FLS depend onMKK3, passive K�BxN arthritis was studied in WT andMKK3�/� mice. The severity of arthritis was dramatically lowerin the MKK3�/� mice (Fig. 4A). Histopathological analysisshowed markedly reduced inflammatory cell infiltration andjoint destruction in MKK3�/� mice (histologic scores were0.43 � 0.28 and 6.0 � 0.65 for MKK3�/� and WT, respectively;P � 0.01) (see Fig. 4B for representative sections). Western blot

analysis of the joint extracts showed that phospho-p38 andphospho-MAPK activator protein kinase (APK) 2 induction wassignificantly inhibited in MKK3�/� mice (P � 0.01; Fig. 4D), aswere joint levels of IL-6 and MMP-3 at the conclusion of theexperiment (P � 0.05; Fig. 4C). Because the mechanisms ofdisease initiation could be distinct from chronic inflammation,we also looked at relevant mediators 2 days after the secondserum injection. CXC ligand 1 and IL-1� expression weresuppressed in early arthritis (P � 0.05; Fig. 4E). Because thepassive K�BxN model is IL-1-dependent (16, 17), we determinedwhether administration of exogenous IL-1 restores arthritisseverity in the MKK3�/� mice. As shown in Fig. 4F, IL-1�treatment led to arthritis severity that was identical to WT micein the early phases of the disease. Arthritis rapidly decreased tocontrol levels after the IL-1� was discontinued.

Role of MKK3 in LPS-Induced Cytokine Production in Vivo. In vitrostudies suggested that activation of TLR4 did not require MKK3for IL-6 expression. Therefore, WT and MKK3�/� mice wereadministered LPS, and cytokine levels in the serum were deter-mined. IL-6 levels in the MKK3�/� mice were only slightly lowerthan in WT mice, demonstrating that the TLR4-mediated innateimmune responses are largely intact in vivo. In contrast, pre-treatment with SB203580 (50 mg�kg) blocked LPS-stimulatedcytokine production (Fig. 5).

DiscussionRA is a chronic inflammatory disease marked by synovialhyperplasia and local invasion into the extracellular matrix.Synovitis is mediated, in part, by cytokines that activate a broadarray of cell signaling mechanisms and lead to the release ofdestructive enzymes (18). MAPKs are especially importantbecause they control the production of cytokines such as IL-1,IL-6, IL-8, and MMPs that participate in destruction of theextracellular matrix. The three MAPK families, ERK, JNK, andp38, modulate gene expression by phosphorylating numerouskey transcription factors, such as activator protein 1 and ATF-1�2, as well as the downstream kinase MAPKAPK-2 (1, 19).

Several MAPK members are activated in the rheumatoidsynovium and have been implicated in the pathogenesis of RA.Among them, p38 is thought to be crucial because selective p38inhibitors block joint inflammation and destruction in severalanimal models of arthritis (3). Of the p38 isoforms, p38� isparticularly relevant for cytokine regulation in macrophages andis blocked by pyridinylimidazole compounds, such as SB203580,which compete for the ATP-binding pocket (20). Phospho-p38�

Fig. 3. Regulation of NF-�B activation in WT and MKK3�/� FLS. (A) Cultured FLS from WT and MKK3�/� mice were stimulated with TNF-� or LPS for 1 h. EMSAwas performed to evaluate NF-�B binding. In WT cells, TNF-� and LPS stimulation led to similar levels of NF-�B activation. However, NF-�B induction was defectivein the MKK3�/� FLS after TNF-� stimulation. (B) For comparison, activator protein 1 (AP-1) binding was not affected by MKK3. (C) Cultured FLS from WT andMKK3�/� mice were infected by IKK2 DN adenovirus at a multiplicity of infection of 5,000 or control virus encoding �-gal and were stimulated with LPS for 24 h.IL-6 was measured by ELISA. LPS-induced IL-6 production was blocked by IKK2 DN treatment in both WT and MKK3�/� FLS (n � 3). *, P � 0.05; **, P � 0.01 comparedwith �-gal control.

5486 � www.pnas.org�cgi�doi�10.1073�pnas.0509188103 Inoue et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 3

, 202

0

is present in rheumatoid synovial blood vessels, perivascularmononuclear cell infiltrates, and the intimal lining (2). p38� isalso activated in RA, especially at sites of synovial invasion intothe extracellular matrix (21).

The two main upstream kinases that regulate p38, MKK3 andMKK6, are also expressed and activated in RA synovium,especially in the intimal lining FLS (13). Using DNA constructs,we showed that both MKK3 and MKK6 contribute to p38phosphorylation in cytokine-stimulated FLS (14). However,MKK3 blockade more effectively inhibited p38-mediated induc-tion of IL-6, IL-8, and MMP-3. p38 can also serve as a substratefor a third MKK, MKK4, which is normally considered a JNKkinase. However, siRNA knock-down studies suggest that MKK4does not play a significant role in p38 activation or IL-6production in FLS after exposure to TNF-� (data not shown).These data suggest that MKK3 might be the dominant MKK thatcontrols p38 in FLS after TNF-� stimulation.

To test this hypothesis, we evaluated p38 function in culturedFLS from MKK3�/� mice. These studies showed that p38requires MKK3 in primary murine FLS for normal phosphory-lation after TNF-� stimulation. In contrast, MKK3 deficiencyhad no effect on JNK or ERK activation. Although p38 phos-

phorylation was modestly decreased in MKK3�/� FLS, muchmore profound suppression of IL-6 production was observed.The impact of MKK3 deficiency on IL-6 production was mim-icked by SB203580 or p38� knock-down in WT cells and issimilar to MKK3�/� murine embryonic fibroblasts (22). FLS arealso a primary source of IL-6 in RA synovium (23); these studiessuggest that selective MKK3 blockade might effectively modu-late the cytokine profile in inflammatory arthritis. The mecha-nism of IL-6 inhibition is complex and can be caused by changesin transcriptional, translational, and mRNA stability effects aspreviously demonstrated for p38 in cultured FLS (24).

The possibility that TLR responses or other pathways to p38activation and cytokine production remain intact despite MKK3deficiency was tested in vitro. First, the TLR4 ligand LPS wasused to stimulate the murine FLS to assess cytokine production.Even though LPS is not a particularly potent inducer of phospho-p38 compared with TNF-�, the absence of MKK3 had no effecton its ability to activate p38. Surprisingly, MKK3 plays no rolein this pathway, even though TNF-�-mediated IL-6 productionis almost completely dependent on this kinase. Similarly, IL-1�stimulation, which shares signaling pathways with TLR4 (15),also was MKK3-independent. Of particular interest, SB203580

Fig. 4. Effect of MKK3 on passive K�BxN arthritis in vivo. (A) WT or MKK3�/� mice (n � 10 WT; n � 9 MKK3�/�) were administered K�BxN serum i.p. on days0 and 2, and arthritis scores were evaluated until day 12. (B) Effect of MKK3 on histopathological evaluation in passive K�BxN arthritis on day 12. Representativeexamples of ankle histology demonstrate markedly less inflammation in the MKK3�/� mice compared with WT. (C and E) Effect of MKK3 deficiency on synovialIL-6 and MMP-3 (day 12) expression and IL-1� and CXC ligand 1 (day 4) expression. Real-time PCR or ELISAs were performed on joint extracts of WT and MKK3�/�

mice. *, P � 0.05; **, P � 0.01 compared with WT arthritis control (n � 4). (D) Effect of MKK3 on synovial p38 and MAPKAPK-2 activation. Western blot analysiswas performed on the joint. Note the lower levels of phospho-p38 and phospho-MAPKAPK-2 (P-MK2) in the MKK3�/� mice. (F) WT or MKK3�/� mice (n � 5 exceptfor WT plus IL-1, where n � 3) were administered K�BxN serum i.p. on days 0 and 2. They were treated with IL-1� (1 �g i.p.) or vehicle on days 0, 2, and 4. IL-1treatment (see arrows) restored arthritis to normal levels in the MKK3�/� mice.

Inoue et al. PNAS � April 4, 2006 � vol. 103 � no. 14 � 5487

MED

ICA

LSC

IEN

CES

Dow

nloa

ded

by g

uest

on

Mar

ch 3

, 202

0

blocked LPS as well as IL-1�-induced IL-6 production, suggest-ing that p38 activation still participates through other upstreamkinases.

The relatively low level of phospho-p38 after LPS stimulationcompared with TNF-� also suggested that the TLR ligand mightactivate alternative pathways that contribute to IL-6 production.NF-�B is an attractive candidate because it can induce IL-6expression in cultured FLS (25). Both LPS and TNF-� increasedNF-�B binding in FLS, but MKK3�/� cells had a significantdefect in NF-�B activation only after TNF-� stimulation. There-fore, both MKK3 and p38 are required for full NF-�B activationafter TNF receptor ligation, but MKK3 plays no role in NF-�Bactivation induced by LPS. The requirement for NF-�B inLPS-mediated cytokine expression was confirmed by using a DNIKK2 construct, which significantly decreased IL-6 production.

These data suggest that LPS-induced IL-6 production dependson NF-�B and p38, without a contribution from MKK3. TNF-�-induced IL-6 production requires NF-�B, MKK3, and p38.Furthermore, NF-�B induction by TNF-� also requires MKK3,and, as a result, NF-�B cannot rescue IL-6 production inMKK3�/� FLS. The possibility that p38 regulates NF-�B hasbeen examined in a few other systems with variable results. Forinstance, LPS-stimulated NF-�B activation in RAW264.7 cellsand cytokine-stimulated FLS was not inhibited by SB203580 (26,27). In contrast, p38 was required for NF-�B p65 transactivationin murine L929 cells and murine embryonic fibroblasts (28, 29).Our data strongly suggest that p38 or MKKs play a key rolein NF-�B-mediated DNA binding and IL-6 production insynoviocytes.

The function of MKK3 was then evaluated in the passiveK�BxN model of arthritis (30). MKK3 deficiency profoundlydecreased both the incidence and severity of arthritis. In par-ticular, the increase in synovial p38 phosphorylation observed inthe WT mice was completely prevented in the MKK3�/� ani-mals, as were MAPKAPK-2 phosphorylation and IL-6 andMMP-3 production. Early pathogenic events in the arthritismodel are also modulated by MKK3. For instance, CXC ligand1 expression in MKK3�/� joints in the early stages of arthritis issignificantly decreased and could suppress accumulation ofneutrophils to the synovium. Perhaps even more important,expression of IL-1� was also suppressed in the MKK3�/� mice.

The critical role of defective IL-1� production in early diseasewas proven by administering exogenous cytokine to MKK3�/�

mice, which restored arthritis to WT levels. Arthritis rapidlydecreased after IL-1 treatment was discontinued, suggesting thatcytokine replacement at the onset is not enough to establishchronic synovitis. Of interest, WT mice also had a small butpersistent increase in arthritis severity.

In vivo studies confirm that MKK3 is not required for TLR4responses. LPS administration to mice leads to a dramaticincrease in IL-6 production, which was only slightly lower in theMKK3�/� animals. Mice pretreated with SB203580 had a sig-nificant decrease in cytokine levels after LPS injection, indicat-ing that p38 still participates in the model. These results contrastwith the arthritis studies, where cytokine production decreasedto normal levels in the absence of MKK3. Hence, MKK3deficiency dramatically decreases cytokine production and clin-ical arthritis while having a minimal impact on in vivo responsesto TLR4 stimulation. The role of p38 in TLR4 responses may besimilar in humans, where administration of a p38 inhibitor blocksIL-6 and TNF-� production after LPS challenge in normalvolunteers (31). Although formal evaluation of host defense hasnot been performed, the fact that TLR4 signaling as well asIL-1-mediated events can remain intact with MKK3 deficiencycompared with p38 blockade suggests an improved the safetyprofile of MKK3 inhibition compared with p38.

Our studies demonstrate that selective deficiency in MKK3can suppress inflammation in an animal model of arthritis, inpart through an IL-1�-dependent mechanism in early disease.More striking, the mice retain the ability to respond to TLR4ligation in a p38-dependent fashion. The mechanism of TLR4responses is probably related to preservation of p38 responsesthrough other MKKs in combination with preserved NF-�Bsignaling. It is precisely this signaling diversity within the p38pathway that provides an opportunity to block components ofthe p38 cascade that are implicated in pathogenic responsesrather than host defense. These studies suggest that targetingupstream kinases such as MKK3 can offer a more selectiveapproach to p38 inhibition than targeting p38 itself.

Materials and MethodsFLS. Multiple FLS lines were derived from MKK3�/� mice (22)and matched WT littermates on the C57BL�6 background. FLSwere grown in DMEM containing 10% FCS and used frompassages three to four, during which time they are a homoge-neous population of cells (�1% CD11b-positive, �1% phago-cytic, and �1% Fc�RII-positive) (32). All cell cultures weresynchronized in 0.1% FCS in DMEM for 24 h before stimulationand maintained in low serum conditions throughout theexperiment.

Antibodies and Reagents. Anti-MKK3, anti-MKK6, anti-�-actin,and secondary antibodies (Santa Cruz Biotechnology); anti-phosho-p38 (Thr-180�Tyr-182), anti-p38, anti-phospho-MAPKAPK-2, anti-phospho-JNK1�2, anti-JNK1�2, anti-phospho-ERK1�2, and anti-ERK1�2 antibodies; GST-ATF-2(Cell Signaling Technology, Beverly, MA); recombinant humanTNF-� (rhTNF-�); recombinant mouse IL-1� (R & D Systems);and SB203580 (Calbiochem) were used.

Western Blot Analysis. FLS were treated with medium, rhTNF-�(50 ng�ml), or LPS (1 �g�ml) for 15 min. Protein was extractedand fractionated on 10% SDS�PAGE, and then Western blottingwas performed (14). The densitometry analysis was done byusing NIH IMAGE 1.61.

Immunoprecipitation and Kinase Assays. FLS were treated withmedium or rhTNF-� for 15 min, lysed in modified RIPA buffer(50 mM HEPES, pH 7.4�150 mM NaCl�1% Triton X-100�10%

Fig. 5. Effect of MKK3 on LPS-induced cytokines production in vivo. LPS (400�g per mouse) was injected i.p., and plasma was collected 3 h later. IL-6 wasmeasured by ELISA. Note that LPS-induced IL-6 production was only modestlylower in MKK3�/� mouse sera compared with WT. However, SB203580 (50mg�kg orally) inhibited IL-6 production in both genotypes. *, P � 0.05; **, P �0.01 compared with WT or LPS control (n � 5).

5488 � www.pnas.org�cgi�doi�10.1073�pnas.0509188103 Inoue et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 3

, 202

0

glycerol�2.5 mM MgCl2�1.0 mM EDTA, pH 8.0�20 mM �-glycerophosphate�10 mM NaF�1 mM Na2VO4�10 �g/ml apro-tinin�1 �M pepstatin A�1 mM PMSF), and incubated withanti-p38 mAb, followed by protein G-Sepharose overnight. Theimmunoprecipitates were washed and incubated with mediumcontaining 5 �Ci (1 Ci � 37 GBq) of [�-32P]ATP, 100 mM ATP,and 4 �g of GST-fusioned-ATF-2 at 37°C for 30 min. Reactionswere fractionated on 12% SDS�PAGE and visualized by auto-radiography (14).

Cell Transfection. For virus infection, FLS were treated withpurified adenovirus IKK2 DN overnight at a multiplicity ofinfection of 5,000 as described in ref. 25. Adenovirus encodingfor �-gal (CMV-LacZ) was used as a control. For siRNAtransfection, using a mouse embryonic fibroblast nucleofectorkit (MEF2) with program T-20 (Amaxa, Gaithersburg, MD),FLS were transfected with 1 �g of p38� siRNA or control siRNA(Dharmacon) (14).

ELISA and mRNA Assay. FLS were treated with medium, rhTNF-�(50 ng�ml), IL-1� (2 ng�ml), or LPS (1 �g�ml) for 24 h.Supernatants were harvested and assayed for IL-6 protein byELISA (R & D Systems). Total RNA was harvested to assay IL-1and IL-6 mRNA expression. Quantitative real-time PCR wasperformed by using the GeneAmp 5700 sequence detectionsystem (Applied Biosystems) as described in ref. 33.

EMSA. FLS were treated with medium, rhTNF-� (50 ng�ml), orLPS (1 �g�ml) for 60 min, and EMSA was performed by using

the gel-shift assay system (Promega). Protein concentration-normalized nuclear extracts were mixed with the purified �-32P-labeled oligonucleotides, resolved by 5% PAGE, and visualizedby autoradiography.

Passive K�BxN Mice Serum Transfer Model of Arthritis. RecipientMKK3�/� and matched WT mice were injected i.p. with 100 �lof pooled adult K�BxN mice serum on days 0 and 2. Arthritis wasevaluated for each paw as described in ref. 34. For tissue proteinassays, snap-frozen ankle joints were homogenized in lysis bufferand assayed by ELISA or immunoblotting. In some cases, RNAwas isolated, and quantitative real-time PCR was performed.The other hind paws were fixed in 10% formalin and decalcifiedin EDTA. A semiquantitative scoring system was used to assesssynovial inflammation, extraarticular inflammation, erosion,and proteoglycan loss (34).

Cytokine Levels in Endotoxic Shock. MKK3�/� mice or matchedWT mice (five animals per group) were injected i.p. with 400 �gof Escherichia coli 0111:B4 LPS (Sigma) (35). In some mice,SB203580 (50 mg�kg) in 0.03 M HCl–0.5% tragacanth (Sigma)was orally administered 30 min before LPS injection. Plasma wascollected 3 h later and assayed for IL-6 by ELISA.

Statistical Analysis. Data are expressed as mean � SEM. Com-parisons were performed by Student’s t test with Bonferroni’scorrection where appropriate.

This work was supported by grants from the National Institutes of Health.

1. Morel, J. & Berenbaum, F. (2004) Joint Bone Spine 71, 503–510.2. Schett, G., Tohidast-Akrad, M., Smolen, J. S., Schmid, B. J., Steiner, C. W.,

Bitzan, P., Zenz, P., Redlich, K., Xu, Q. & Steiner, G. (2000) Arthritis Rheum.43, 2501–2512.

3. Mclay, L. M., Halley, F., Souness, J. E., McKenna, J., Benning, V., Birrell, M.,Burton, B., Belvisi, M., Collis, A., Constan, A., et al. (2001) Bioorg. Med. Chem.9, 537–554.

4. Pargellis, C. & Regan, J. (2003) Curr. Opin. Investig. Drugs 4, 566–571.5. Saklatvala, J. (2004) Curr. Opin. Pharmacol. 4, 372–377.6. Kumar, S., Boehm, J. & Lee, J. C. (2003) Nat. Rev. Drug Discov. 2, 717–726.7. Schnyder, B., Meunier, P. C. & Car, B. D. (1998) Biochem. J. 331, 489–495.8. van den Blink, B., Juffermans, N. P., ten Hove, T., Schultz, M. J., van Deventer,

S. J., van der Poll, T. & Peppelenbosch, M. P. (2001) J. Immunol. 166, 582–587.9. Raingeaud, J., Whitmarsh, A. J., Barrett, T., Derijard, B. & Davis, R. J. (1996)

Mol. Cell. Biol. 16, 1247–1255.10. Moriguchi, T., Kuroyanagi, N., Yamaguchi, K., Gotoh, Y., Irie, K., Kano, T.,

Shirakabe, K., Muro, Y., Shibuya, H., Matsumoto, K., et al. (1996) J. Biol.Chem. 271, 13675–13679.

11. Brancho, D., Tanaka, N., Jaeschke, A., Ventura, J. J., Kelkar, N., Tanaka, Y.,Kyuuma, M., Takeshita, T., Flavell, R. A. & Davis, R. J. (2003) Genes Dev. 17,1969–1978.

12. Han, J., Lee, J. D., Jiang, Y., Li, Z., Feng, L. & Ulevitch, R. J. (1996) J. Biol.Chem. 271, 2886–2891.

13. Chabaud-Riou, M. & Firestein, G. S. (2004) Am. J. Phathol. 164, 177–184.14. Inoue, T., Hammaker, D., Boyle, D. L. & Firestein, G. S. (2005) J. Immunol.

174, 4301–4306.15. O’Neill, L. A. (2005) Curr. Opin. Immunol. 17, 1–7.16. Kyburz, D. & Corr, M. (2003) Springer Semin. Immunopathol. 25, 79–90.17. Ohmura, K., Johnsen, A., Ortiz-Lopez, A., Desany, P., Roy, M., Besse, W.,

Rogus, J., Bogue, M., Puech, A., Lathrop, M., et al. (2005) Proc. Natl. Acad.Sci. USA 102, 12489–12494.

18. Firestein, G. S. (2003) Nature 423, 356–361.19. Neininger, A., Kontoyiannis, D., Kotlyarov, A., Winzen, R., Eckert, R., Volk,

H. D., Holtmann, H., Kollias, G. & Gaestel, M. (2002) J. Biol. Chem. 277,3065–3068.

20. Hale, K. K., Trollinger, D., Rihanek, M. & Manthey, C. L. (1999) J. Immunol.162, 4246–4252.

21. Kuchen, S., Seemayer, C. A., Rethage, J., von Knoch, R., Kuenzler, P., Michel,B. A., Gay, R. E., Gay, S. & Neidhart, M. (2004) Autoimmunity 37, 57–65.

22. Wysk, M., Yang, D. D., Lu, H. T., Flavell, R. A. & Davis, R. J. (1999) Proc. Natl.Acad. Sci. USA 96, 3763–3768.

23. Firestein, G. S., Alvaro-Garcia, J. M. & Maki, R. (1990) J. Immunol. 144,3347–3353.

24. Miyazawa, K., Mori, A., Miyata, H., Akahane, M., Ajisawa, Y. & Okudaira, H.(1998) J. Biol. Chem. 273, 24832–24838.

25. Aupperle, K. R., Bennett, B. L., Han, Z., Boyle, D. L., Manning, A. M. &Firestein, G. S. (2001) J. Immunol. 166, 2705–2711.

26. Baldassare, J. J., Bi, Y. & Bellone, C. J. (1999) J. Immunol. 162, 5367–5373.

27. Suzuki, M., Tetsuka, T., Yoshida, S., Watanabe, N., Kobayashi, M., Matsui, N.& Okamoto, T. (2000) FEBS Lett. 465, 23–27.

28. Jijon, H., Allard, B. & Jobin, C. (2004) Cell. Signalling 16, 1023–1032.29. Vanden Berghe, W., Plaisance, S., Boone, E., De Bosscher, K., Schmitz, M. L.,

Fiers, W. & Haegeman, G. (1998) J. Biol. Chem. 273, 3285–3290.30. Ji, H., Pettit, A., Ohmura, K., Ortiz-Lopez, A., Duchatelle, V., Degott, C.,

Gravallese, E., Mathis, D. & Benoist, C. (2002) J. Exp. Med. 196, 77–85.31. Branger, J., van den Blink, B., Weijer, S., Madwed, J., Bos, C. L., Gupta, A.,

Yong, C. L., Polmar, S. H., Olszyna, D. P., Hack, C. E., et al. (2002) J. Immunol.168, 4070–4077.

32. Yamanishi, Y., Boyle, D. L., Pinkoski, M. J., Mahboubi, A., Lin, T., Han, Z.,Zvaif ler, N. J., Green, D. R. & Firestein, G. S. (2002) Am. J. Pathol. 160,123–130.

33. Boyle, D. L., Rosengren, S., Bugbee, W., Kavanaugh, A. & Firestein, G. S.(2003) Arthritis Res. Ther. 5, R352–R360.

34. Simelyte, E., Rosengren, S., Boyle, D. L., Corr, M., Green, D. R. & Firestein,G. S. (2005) Arthritis Rheum. 52, 1876–1884.

35. Firestein, G. S., Boyle, D. L., Bullough, D. A., Gruber, H. E., Sajjadi, F. G.,Montag, A., Sambol, B. & Mullane, M. (1994) J. Immunol. 152, 5853–5859.

Inoue et al. PNAS � April 4, 2006 � vol. 103 � no. 14 � 5489

MED

ICA

LSC

IEN

CES

Dow

nloa

ded

by g

uest

on

Mar

ch 3

, 202

0