702 NATURE MEDICINE • VOLUME 5 • NUMBER 6 • JUNE 1999

ARTICLES

Unmethylated CpG motifs are often found in bacterial DNA, andexert immunostimulatory effects on hematopoetic cells1–3.Bacteria produce severe joint inflammation in septic and reac-tive arthritides; bacterial DNA may be involved in this process.We injected bacterial DNA originating from Escherichia coli andStaphylococcus aureus and oligonucleotides containing CpGdirectly into the knee joints of mice of different strains. Arthritiswas seen by histopathology within 2 hours and lasted for atleast 14 days. Unmethylated CpG motifs were responsible forthis induction of arthritis, as oligonucleotides containing thesemotifs produced the arthritis. The arthritis was characterized byan influx of monocytic, Mac-1+ cells and by a lack of T lympho-cytes. Depletion of monocytes resulted in abrogation of the syn-ovial inflammation. Tumor necrosis factor (TNF)-α, a cytokineproduced by cells of the monocyte/macrophage lineage, is animportant mediator of this disease, as expression of mRNA forTNF-α was evident in the inflamed joints, and the CpG-mediatedinflammation was abrogated in mice genetically unable to pro-duce this cytokine. These findings demonstrate that bacterialDNA containing unmethylated CpG motifs induces arthritis, andindicate an important pathogenic role for bacterial DNA in sep-tic arthritis.

Bacterial infections can be localized to the joints, causing septicarthritis, the most rapidly progressing joint disease4. The outcomeof bacterial arthritis has not changed much in the past fewdecades, despite the use of more effective antibiotics and im-proved methods of joint drainage4. Permanent joint damage de-velops in most cases and mortality continues to be high4. To studythe pathogenesis of septic arthritis, a murine model of this diseasehas been developed; certain bacterial virulence factors as well ashost immune responses are important in the induction phase andprogression of this disease5. Here we report the effect of bacterialDNA in producing arthritis, with the pathology defined as the pro-liferation of synoviocytes in the synovial lining layer and the in-flux of inflammatory cells to the synovium. We injected bacterialDNA, vertebrate DNA or oligonucleotides containing CpG dinu-cleotides intra-articularly into knee joints of mice of differentstrains. Bacterial DNA and oligonucleotides containing unmethy-lated CpG dinucleotides (CpG-ODN) induced arthritis(Fig. 1a);vertebrate DNA did not (Fig. 1d). A requirement for the inductionof this type of arthritis is that the DNA be deposited locally in thejoint rather than administered systemically. Indeed, intraperi-toneal injection of mice with either 1 nmol (6 µg; n = 5) or 10nmol (60 µg; n = 5) of CpG-ODN produced no histological evi-dence of knee joint inflammation. In contrast, the CpG-ODNgiven intra-articularly at a dose of 1 or 10 nmol produced go-narthritis in almost 100% of the mice. Joint inflammation in-

duced by oligonucleotides containing unmethylated CpG wascharacterized by rapid onset (within 2 hours) and persistence forat least 14 days (Fig. 2a). To exclude the possibility of lipopolysac-charide (LPS) contamination of the bacterial DNA or syntheticoligonucleotides, we injected both types of DNA into the kneejoints of C3H/HeJ mice, whose lymphocytes fail to proliferate inresponse to LPS (LPS nonresponder mice). There was no differencein the frequency or severity of arthritis produced after the injec-tion of bacterial DNA and ODN1668 into C3H/HeJ mice or mice ofthe congeneic LPS-responder strain C3H/HeN (Fig. 2b), indicatingthat induction of arthritis was due to bacterial DNA rather than toLPS contamination. In contrast, intra-articular injection of LPS in-duced arthritis in C3H/HeN mice but not C3H/HeJ mice (Fig. 2b).

We determined the roles of different immune cells in arthritisinduced by CpG-ODN. First, we assessed neutrophils, as these arethe cells first to migrate into tissues in response to inflammatorystimuli. Indeed, in bacterial arthritis, neutrophil-depleted ani-mals show a significantly higher frequency of arthritis and mor-tality than do control animals6. However, neutrophil-depletedmice did not differ from control mice in their development ofCpG-ODN-mediated arthritis (Fig. 3a), indicating that neu-

Intra-articularly localized bacterial DNA containing CpGmotifs induces arthritis

GUO-MIN DENG, ING-MARIE NILSSON, MARGARETA VERDRENGH,L.VINCENT COLLINS & ANDREJ TARKOWSKI

Department of Rheumatology, University of Göteborg, Guldhedsgatan 10A, S-413 46 Göteborg, SwedenCorrespondence should be addressed to G.-M.D.; email: guo-min@immuno.gu.se

a b

c d

Fig. 1 The histopathological, immunohistochemical and in situ hy-bridization appearance of bacterial DNA-induced arthritis. JC, joint cavity;V, blood vessel; C, cartilage; ST, synovial tissue; arrows, inflammatory cellsin synovia. a, Histopathology of an arthritic knee joint of a mouse killed 14d after intra-articular inoculation of ODN 1668. Infiltration of mononuclearcells around blood vessels is apparent. b, Immunohistochemistry of anarthritic knee joint, showing synovial expansion of cells expressing Mac-1(brown) in a mouse killed 14 d after intra-articular inoculation of ODN1668. c, In situ hybridization of an arthritic synovium showing TNF-α mes-senger RNA expression 3 d after intra-articular inoculation of ODN 1668.d, Normal histopathological appearance of mouse knee joint after injec-tion with calf thymus DNA.

© 1999 Nature America Inc. • http://medicine.nature.com©

199

9 N

atu

re A

mer

ica

Inc.

• h

ttp

://m

edic

ine.

nat

ure

.co

m

NATURE MEDICINE • VOLUME 5 • NUMBER 6 • JUNE 1999 703

ARTICLES

trophils are not mandatory in the development of this condition.We next used severe combined immunodeficiency (SCID) mice,which lack B lymphocytes and T lymphocytes but have an intactmonocyte/macrophage population, to explore the effect of CpG-ODN on producing arthritis. The incidence of arthritis inducedby CpG-ODN did not differ between SCID mice and theircongeneic controls (Fig. 3a), indicating that T cells and B cells arenot required for development of arthritis, but thatmacrophages/monocytes might be required. To further assess therole of macrophages in arthritis induced by CpG-ODN, we used acytotoxic drug (etoposide) known to selectively deplete mono-cyte/macrophage population7. Pre-treatment of mice with etopo-side led to a 95% depletion of monocytes and a total abrogationof arthritis (Fig. 3a). In addition, the in vitro splenocyte prolifera-tive responses induced by CpG-ODN and bacterial DNA werestrongly suppressed by etoposide (Fig. 3b). The importance ofmonocytes/macrophages in arthritis induced by CpG-ODN wassupported by the abundance of Mac-1+ mononuclear cells(macrophages) and the scarcity of T lymphocytes, both in thesynovial lining cell layer, deep in the sublining space, as well as insurrounding synovial vessels during synovitis (Fig. 1b).

We next assessed the role of factors released from mono-cytes/macrophages that might be important in inducing synovi-tis. As TNF-α, produced by monocytes/macrophages, is one ofthe chief cytokines mediating both septic8 and autoimmune9

arthritis, we analyzed its effect on joint disease induced by CpG-ODN. In situ hybridization analysis demonstrated the presenceof cells expressing TNF-α mRNA in synovial tissue of arthriticmice (Fig. 1c), whereas such cells were not found in synovial tis-

sue of control mice inoculated with calf thymus DNA or withPBS. These results indicate that TNF-α might function even inarthritis produced by bacterial DNA. TNF-α knockout mice werealmost unable to produce joint inflammation after being in-jected with CpG-ODN. Thus, the incidence of arthritis was sig-nificantly less in TNF-α knockout mice than in congeneiclittermates (Table 1; P < 0.05). These results indicate that TNF-α isfunctional in CpG-ODN-induced arthritis. Our data corroboratea recent report that bacterial DNA or oligonucleotides contain-ing CpG motifs stimulate macrophages in vitro and in vivo to pro-duce TNF-α (ref. 3). Another cytokine of interest in arthritis isinterleukin-12 (IL-12), which is produced by T-helper 1 lympho-cytes and monocytes/macrophages and is involved in the in-flammatory cascade during the development of synovitis.Indeed, mice genetically unable to produce IL-12 show a de-creased incidence of collagen-induced arthritis10 and septicarthritis (O. Hultgren, manuscript in preparation). Even inarthritis induced by CpG DNA, the prevalence of synovial in-flammation in IL-12 knockout mice was less than that in thecongeneic mice (Table 1). IL-12 is known to stimulatemacrophage production of interferon-γ, a cytokine known topromote the development of both autoimmune and septicarthritis11,12.

Why is bacterial DNA but not eukaryotic DNA arthritogenic?Unmethylated CpG motifs are present at the expected frequencyof one per sixteen dinucleotides in the bacterial DNA, but areunder-represented (CpG ‘suppression’) and predominantlymethylated in eukaryotic DNA (ref. 13). This structural differ-ence seems to explain the finding that bacterial DNA and certainsynthetic oligonucleotides containing unmethylated CpG dinu-cleotides activate immune cells1–3. To determine whether themethylation of CpG dinucleotides protects against the inductionof arthritis, we used oligonucleotides containing either an em-bedded unmethylated CpG dinucleotide (1668), a methylatedCpG dinucleotide of the same sequence (1668m), or a CpG in-verted to non-methylated GpC dinucleotide (1720). Only ODN1668 was able to cause arthritis (Fig. 2c). Thus, oligonucleotidescontaining unmethylated CpG dinucleotides are responsible forthe induction of arthritis. As phosphodiester oligonucleotidesare relatively sensitive to nucleases, we synthesized CpGoligonucleotides with a nuclease-resistant, phosphorothioate-

Fig. 2 Effect of bacterial DNA and ODN 1668 on the development of arthritis.a, Kinetics of ODN 1668-induced arthritis in C57BL/6 mice (n = 4 per group).Data represent mean +/– s.d. b, Incidence of arthritis in C3H/HeN mice (left,HeN; n = 5) or C3H/HeJ mice (right, HeJ; n = 5), killed 3 d after inoculation of ei-

ther E. coli DNA, ODN 1668 or LPS. c, Incidence of arthritis in C57BL/6 mice afterinoculation of genomic E. coli DNA (30 µg), calf thymus DNA (30 µg), ODN1668, ODN 1668m, ODN 1720 (all 6 µg) or LPS (1 µg) (n = 5 per group), S. au-reus DNA (20 µg; n = 4) or PBS. All mice were killed 3 d after the inoculation.

Table 1 Incidence and severity of arthritis 3 d after intra-articularinoculation of ODN 1668

Genotype Number of Incidence of Severity of mice arthritis arthritis

TNF-α +/+ 9 88.9% 1.1± 0.6TNF-α –/– 9 33.3%* 0.33± 0.5IL-12 +/+ 9 100% 1.33± 0.5IL-12 –/– 8 62.5%NS 0.75± 0.75NS

TNF-α and IL-12 knockout mice, as well as their wild-type littermates, were assessed.NS, not significant; *, P < 0.05.

a

Seve

rity

(mea

n+/-

SD)

Days after injection of ODN 1668

Inci

den

ce o

f art

hriti

s (%

)

Inci

den

ce o

f art

hriti

s (%

)

E. c

oliD

NA

E. c

oliD

NA

E. c

oliD

NA

S. a

ureu

s D

NA

OD

N 1

668

OD

N 1

668

OD

N 1

668

OD

N 1

720

OD

N 1

668mLP

S

LPS

PBS

Cal

f thy

mus

DN

ALPS

HeN HeJ

b c

© 1999 Nature America Inc. • http://medicine.nature.com©

199

9 N

atu

re A

mer

ica

Inc.

• h

ttp

://m

edic

ine.

nat

ure

.co

m

704 NATURE MEDICINE • VOLUME 5 • NUMBER 6 • JUNE 1999

ARTICLES

Fig. 3 Cellular requirements for induction ofCpG ODN arthritis: a, Incidence of arthritis inBALB/C mice depleted of neutrophils using mon-oclonal antibody against RBG-8C5 (n = 5 pergroup), SCID mice deficient in T cells and B cells(n = 7) and congeneic CB17 control mice (n = 7),and C57BL/6 mice depleted of macrophagesusing etoposide treatment (n = 5 per group).Brackets connect bars compared for statisticalsignificance; NS, not significant. b, Macrophagedepletion downregulates the proliferative re-sponses of spleen mononuclear cells to ODN1668 and to E. coli DNA. Data represent mean+/– s.e.m. Brackets connect bars compared forstatistical significance

modified backbone to determine whether they would also causearthritis. Indeed, phosphorothioate-modified, unmethylatedCpG oligonucleotides induced arthritis 3 days after inoculationand did so at a magnitude similar to that of phosphodiesteroligonucleotides (data not shown). Bacterial DNA and syntheticoligonucleotides are taken up by monocytes in a saturable,sequence-independent, temperature-dependent and energy-dependent manner14 into an acidified intracellular compart-ment, where DNA is degraded to oligodeoxynucleotides15. Oncethere, unmethylated CpG dinucleotides induce reactive oxygenspecies, thereby activating the nuclear translocation of transcrip-tion factor NF-κB (ref. 16), which controls the induction of TNF-α production17.

As LPS and bacterial DNA stimulate macrophages by differentsignal pathways1,18, and both of these compounds may be presentduring septic arthritis, we determined whether LPS and ODN1668 act synergistically in inducing arthritis. There was a high in-cidence of arthritis when suboptimal amounts of both agentswere injected simultaneously, and a lower incidence after injec-tion of these substances individually (Table 2).

Our data demonstrate that bacterial DNA containing unmethy-lated CpG motifs induces arthritis, and indicate an importantpathogenic role for this molecule in septic arthritis. These find-ings may have implications for the treatment of bacterial arthri-tides. Thus, killing bacteria with antibiotics might not besufficient, as bacterial DNA released from resident bacteria is pro-inflammatory and, acting in concert with LPS, might promote acontinuous inflammatory response. Indeed, epidemiologic stud-ies have indicated that despite antibiotic treatment and, thus,eradication of all live bacteria, inflammation and destruction of

joints continues4. Future treatment strategies might involveblocking bacterial growth in combination with efficient irrigationof joints to remove bacteria and their components, together withanti-inflammatory therapy.

MethodsMice and reagents. BALB/C, C57BL/6, and NMRI mice were purchasedfrom ALAB ( Stockholm, Sweden). C3H/HeJ and C3H/HeN mice, and SCID-mice and their congeneic strain C.B.17, were purchased from M&B(Bomholtvej, Denmark). TNF-α and IL-12 knockout mice were provided byJ.D. Sedgwick (Centenary Institute of Cancer Medicine and Cell Biology,Sydney, Australia) and Hoffmann-La Roche, respectively. All mice werehoused in the animal facility of the Department of Rheumatology(University of Göteborg). Male mice 6–8 weeks of age were used in all theexperiments. The hybridoma cells secreting RB6-8C5 were a gift from R.Coffman (DNAX Research Institute, Palo Alto, California). Etoposide wassupplied by Bristol-Myers Squibb (Bromma, Sweden).

Genomic DNA and oligonucleotides. E. coli (strain B) DNA and calf thy-mus DNA were purchased from Sigma and were further purified by extrac-tion with phenol:chloroform:isoamyl alcohol (25:24:1) followed by ethanolprecipitation. The purity and concentration of DNA were determined usinga spectrophotometer (Molecular Devices, Sunnyvale, California). GenomicDNA from S. aureus strain LS-1 was prepared by lysing the bacteria with acombination of lysostaphin, lysozyme and proteinase K (all from Sigma).The DNA was extracted twice with TE-equilibrated phenol and once withchloroform followed by precipitation with 0.1 volumes of 3 M sodium ac-etate and 2 volumes of 95% ethanol. Genomic DNA was spooled on a glassrod and air-dried before being re-solubilized in sterile water. Heat-dena-tured (single-stranded) genomic DNA was used in all experiments.Oligonucleotides 1668, 1668m and 1720 were synthesized byScandinavian Gene Synthesis AB (Köping, Sweden). The phosphorothioate-modified oligonucleotide 1668 was synthesized by GenosysBiotechnologies (Cambridge, UK). The sequences of ODN 1668 (contain-ing the CpG motif) and its methylated derivative 1668m as well as ODN1720 (with an inverted CG motif) were as described3: 1668, 5’–TCCAT-GACGTTCCTGATGCT–3’; 1668m, 5’–TCCATGAXGTTCCTGATGCT–3’ (X, 5-methyl-deoxycytidine); 1720, 5’–TCCATGAGCTTCCTGATGCT–3’.

Injection protocol and cell depletion procedures. DNA, oligonucleotidesor LPS were injected intra-articularly in the knee joints of mice of differentstrains. The contralateral knee joint was always used as a negative control,injected with either vertebrate DNA, oligonucleotides lacking CpG motifs,oligonucleotides containing methylated CpG motifs or phosphate-bufferedsaline (PBS).

For neutrophil depletion, BALB/c mice were injected intraperitonally with1 mg of the monoclonal antibody RB6-8C5 or the IgG rat anti-ovalbumin

Table 2 LPS and oligonucleotides containing CpG actsynergistically in the induction of arthritis.

Treatment Incidence of arthrtitis Severity of arthritis

none 0% 0 ± 0LPS (0.1 µg) 75 % 0.8 ± 0.51668 (0.6 µg) 75% 1.0 ± 0.8LPS(0.1µg)+1668(0.6 µg) 100% 2.0 ± 0.8

BALB/c mice (n = 4 per group) were injected intra-articularly with LPS, ODN 1668 or acombination of LPS and ODN 1668. After 3 d, the mice were killed and knee joints wereexamined by histopathology.

b

RB6

Con

trol

Con

trol

Etop

osid

e

Etop

osid

e

OD

N16

68

E. c

oliD

NA

Etop

osid

e +E

. col

iDN

A

Etop

osid

e +

OD

N16

68

Non

e

SCID

CB1

7

Neutrophildeletion

Macrophagedeletion

T + B cellabsence

Inci

den

ce o

f art

hriti

s (%

)

Up

take

of 3 H

-thy

mid

ine

(cp

m)

NS NS P < 0.01

P < 0.01 P < 0.01

a

© 1999 Nature America Inc. • http://medicine.nature.com©

199

9 N

atu

re A

mer

ica

Inc.

• h

ttp

://m

edic

ine.

nat

ure

.co

m

NATURE MEDICINE • VOLUME 5 • NUMBER 6 • JUNE 1999 705

ARTICLES

monoclonal antibody (control) 2 h before being injected with ODN 1668.Analysis of blood smears showed that RB6-8C5 depletes the granulocytepopulation by more than 90% (ref. 6).

Macrophage depletion was induced by subcutaneous injection of etopo-side (12.5 mg per kg body weight, in a volume of 100 µl) into the nuchalregion on 3 consecutive days before and 2 consecutive days after injectionof ODN 1668; control mice received the same volume of the vehicle dilutedin PBS (ref. 7). FACS analysis showed that the etoposide depleted themonocyte/macrophage population by more than 90%.

Histopathological and immunohistochemical examination of joints.Histopathological examination of joints was done after routine fixation, de-calcification and paraffin embedding of the tissue. Tissue sections fromknee joints were cut and stained with hematoxylin and eosin. All the slideswere coded and evaluated by experimenters ‘blinded’ to sample identity.The specimens were evaluated for synovial hypertrophy, occurrence of in-flammatory cells in synovial sublining compartment, pannus formation,and cartilage and subchondral bone destruction.

For immunohistochemical examination, the knee joints were removedand demineralized by an enzymatic procedure that has been described indetail19. The demineralized specimens were mounted, frozen in isopentaneprechilled in liquid nitrogen, and stored at –70 °C. Serial cryosections 6 µmin thickness were stained with rat monoclonal antibodies directed againstmouse CD11b (Mac-1), CD4+ (GK1.5) or CD8+ (53.6.7), then incubatedwith biotinylated secondary antibodies and avidin-biotin-peroxidase com-plexes (ABC) and 3-amino-9-ethyl-carbazole containing H2O2. All sectionswere counterstained with Mayer’s hematoxylin.

In situ hybridization for cytokine mRNA detection. In situ hybridizationwas done to detect TNF-α mRNA expression as described20. Syntheticoligonucleotide probe TNF-α (a gift from T. Olsson, Karolinska Institute,Stockholm, Sweden), was labeled at the 3´ end using terminal deoxynu-cleotidyl transferase (Advanced Biotechnologies, Leatherhead, UK) and 35S-ATP (Dupont Scandinavia, Stockholm, Sweden). Sections of freshlyfrozen spleen and synovial tissues 4 µm in thickness were thaw-mountedonto slides and were hybridized with 1 × 106 cpm of labeled probe per 100µl hybridization mixture. After emulsion autoradiography, developmentand fixation, the coded slides were examined by darkfield microscopy forpositive cells containing more than 15 silver grains per cell in a ‘starlike’distribution.

Lymphocyte proliferation. Triplicate aliquots (200 µl) of MNC suspensionswere added to 96-well round-bottomed microtiter plates (Nunc,Copenhagen, Denmark) at a cell density of 2 × 106/ml, with 10-µl aliquotsof either 1668, 1668m or 1720 at a final concentration of 1 µM or E.coliDNA at a final concentration of 100 µg/ml. Etoposide was added at the ini-tiation of the culture at 6 µg/ml, a concentration known to deplete mono-cytes/macrophages7. After being incubated for 72 h, the cells were pulsedfor 12 h with 1 µCi of 3H-methylthymidine (specific activity, 42 Ci/mmol;Amersham). Cells were collected onto glass fiber filters and thymidine in-

corporation was measured. The results are expressed as the stimulationvalue (cpm, mean ± s.e.m.).

Statistical analysis. The difference of incidence and severity in groups wereanalyzed by use of the Fisher’s test and Mann-Whitney U test, respectively.

RECEIVED 18 MARCH; ACCEPTED 14 APRIL1999

1. Krieg, A.M. et al. CpG motifs in bacterial DNA trigger direct B-cell activation. Nature374, 546–549 (1995).

2. Klinman, D.M., Yi, A.K., Beaucage, S.L., Conover, J. & Krieg, A.M. CpG motifs pre-sent in bacterial DNA rapidly induce lymphocytes to secrete interleukin 6, inter-leukin 12, and interferon gamma. Proc. Natl. Acad. Sci. USA 93, 2879–2883 (1996).

3. Sparwasser, T. et al. Bacterial DNA causes septic shock. Nature 386, 336–337(1997).

4. Goldenberg, D.L. Septic arthritis. Lancet 351, 197–202 (1998).5. Tarkowski, A. & Wagner, H. Arthritis and sepsis caused by Staphylococcus aureus: can

the tissue injury be reduced by modulating the host’s immune system? Mol. Med.Today 4, 15–18 (1998)

6. Verdrengh, M. & Tarkowski, A. Role of neutrophils in experimental septicemia andseptic arthritis induced by Staphylococcus aureus. Infect. Immun. 65, 2517–2521(1997).

7. Calame, W., Douwas-idema, A.E., van den Barselaar, M.T., Furth, R.V. & Mattie, H.Influence of cytostatic agents on the pulmonany defence of mice infected withKlebsiella pneumoniae and on the efficacy of treatment with ceftriaxone. J. Infect. 29,53–66 (1994).

8. Hultgren, O., Eugster, H.P., Sedgwick, J.D., Körner, H. & Tarkowski, A. TNF/LTαdouble mutant mice resist septic arthritis but display increased mortality in responseto Staphylococcus aureus. J. Immunol. 161, 5937–5942 (1998).

9. Williams, R.O., Feldmann, M. & Maini, R.N. Anti-tumor necrosis factor amelioratesjoint disease in murine collagen-induced arthritis. Proc. Natl. Acad. Sci. USA 89, 9784–9788 (1992).

10. Parks, E. et al. Transient gene transfer of IL-12 regulates chemokine expression anddisease severity in experimental arthritis. J. Immunol. 160, 4615–4619 (1998).

11. Mauritz, N.J. et al. Treatment with IFN-gamma triggers the onset of collagen arthri-tis in mice. Arthritis Rheum. 31, 1297–1304 (1988).

12. Zhao, Y.X., Nilsson, I.M. & Tarkowski, A. The dual role of interferon-gamma in ex-perimental Staphylococcus aureus septicaemia versus arthritis. Immunology 93, 80–85 (1998).

13. Bird, A.P. CpG-rich islands and function of DNA methylation. Nature 321, 209–213(1986).

14. Geselowitz, D.A. & Neckers, L.M. Analysis of oligonucleotide binding, internaliza-tion, and intracellular trafficking utilizing a novel radiolabeled crosslinker. AntisenseRes. Dev. 2, 17–26 (1992).

15. Bennett, R.M., Gabor, G.T. & Merritt, M.M. DNA binding to human leukocytes. J.Clin. Invest. 76, 2182–2190 (1985).

16. Yi, A.K. et al. CpG motifs in bacterial DNA activate leukocytes through the pH-de-pendent generation of reactive oxygen species. J. Immunol. 160, 4755–4761(1998).

17. Baeuerle, P.A. & Henkel, T. Function and activation of NF-kappa B in the immunesystem. Annu. Rev. Immunol. 12, 141–179 (1994).

18. Wright, S.D., Ramos, R.A., Tobias, P.S., Vlevitch, R.J. & Mathison, J.C. CD 14, a re-ceptor for complexes of Lipopolysaccharide (LPS) and LPS binding protein. Science249, 1431–1433 (1990).

19. Jonsson, R., Tarkowski, A. & Klareskog, L.A. Demineralization procedure for im-munohistopathological use. J. Immunol. Methods 88, 109–114 ( 1986).

20. Zhao, Y.X., Ljungdahl, A., Olsson, T. & Tarkowski, A. In situ hybridization analysis ofsynovial and systemic cytokine messenger RNA expression in superantigen-medi-ated Staphylococcus aureus arthritis. Arthritis Rheum. 39, 959–967 (1996).

© 1999 Nature America Inc. • http://medicine.nature.com©

199

9 N

atu

re A

mer

ica

Inc.

• h

ttp

://m

edic

ine.

nat

ure

.co

m