Unmethylated CpG-oligodeoxynucleotides (ODNs) are generally thought of as potent adjuvants with considerable therapeutic potential to enhance immune responses against microbes and tumors. Surprisingly, certain so-called stimulatory CpG-ODNs strongly inhibited the effector phase of inflammatory arthritis in the K/BxN serum transfer system, either preventively or therapeutically. Also unexpected was that the inhibitory influence did not depend on the adaptive immune system cells mobilized in an immunostimulatory context. Instead, they relied on cells of the innate immune system, specifically on cross talk between CD8α+ dendritic cells and natural killer cells, resulting in suppression of neutrophil recruitment to the joint, orchestrated through interleukin-12 and interferon-γ. These findings highlight potential applications of CpG-ODNs and downstream molecules as antiinflammatory agents.
The inflammatory arthritides are an important health problem—rheumatoid arthritis (RA) affects ∼1% of the world's population. Genetic factors are known to have a substantial impact on the incidence and progression of RA, but the low concordance rate of this disease in monozygotic twins (∼20%) raises the possibility of environmental factors also exerting a crucial influence (1). Indeed, there have been many reports of correlations between viral or bacterial infections and the onset of RA in humans (2). Furthermore, several animal models of inflammatory arthritis are initiated by one or another microbial constituents (3).
Cells of the innate immune system detect the intrusion of pathogens by recognizing conserved sequence motifs embedded within mole cules expressed by the intruders (4). Some examples of such motifs are LPS, peptidoglycans (PGNs), double- and single-stranded RNAs, and unmethylated DNA CpG motifs (CpGs). These “pathogen-associated molecules (PAMs)” are specifically recognized by sets of receptors, the best characterized of which are the toll-like receptor (TLR) family members (4, 5). PAMs and their receptors provide powerful molecular probes for assessing a potential impact of microbes on the initiation and progression of arthritis. Indeed, administration of PAMs promoted disease development in several mouse models of arthritis (6–8).
Given their potential importance, we decided to explore the roles of TLRs and their ligands in the K/BxN mouse model of inflammatory arthritis, which is a particularly convenient system for performing mechanistic dissection. This model has the additional advantage of not requiring adjuvants, which often contain TLR ligands, for disease induction, thereby avoiding potential problems of synergy. K/BxN mice express a transgene-encoded TCR that is reactive to a self-peptide derived from the ubiquitously expressed enzyme glucose-6-phosphate-isomerase (GPI), presented by the MHC class II molecule Ag7 (3). Beginning at 3–5 wk of age, K/BxN mice spontaneously develop an autoimmune/inflammatory disease with many features in common with those of human RA (3). Conveniently, transfer of serum from arthritic mice into healthy nontransgenic animals induces arthritis in a highly synchronized and reproducible fashion (3). Importantly, this system does not require LPS to be effective, unlike collagen antibody–induced arthritis (7), again, simplifying the interpretation of results. Arthritis is a very complex, polygenic, and multifactorial disorder, so a major attraction of this serum-transfer system is that it allows one to focus on the inflammatory effector phase of disease, without the complicating influences of the autoimmune initiation phase.
To delineate points of potential environmental impact specifically in the arthritis effector phase, we examined the influence of selected TLRs and TLR ligands in the K/BxN serum-transfer system. Results from these experiments confronted us with two surprises: a potent antiinflammatory effect exerted by so-called stimulatory TLR9 ligands, and a mechanism that relies solely on cells of the innate immune system, without the requirement for cells of the adaptive immune system seen with other CpG-mediated immunotherapeutic activities.
Certain CpGs inhibit K/BxN serum-transferred arthritis
As a first approach to exploring the role of TLRs and their ligands in the effector phase of inflammatory arthritis, we screened several TLRs for the ability to influence K/BxN serum-transferred disease. As might have been expected (7, 8), LPS exacerbated arthritis development in B6 mice when administered i.p. 2 d before K/BxN serum transfer (Fig. 1 A), whether assayed as ankle thickening (left) or as a clinical score (right). Preparations of PGNs, Pam3Cys4, and poly I:C had a similar effect when tested under the same protocol (unpublished data). In contrast, and surprisingly, injection of the CpG oligodeoxynucleotide (ODN) 1668 (Fig. 1 B) severely inhibited development of disease (Fig. 1 A). None of the TLR ligands was able to induce arthritis when injected alone i.p. (unpublished data), in contrast to what has been reported for intraarticular injection of unmethylated CpGs (6), for example.
Interestingly, although some CpG-ODNs were capable of inhibiting K/BxN serum-transferred arthritis, others were not (Fig. 1 B). ODNs 1826 and 1758 also had a suppressive effect, although in the latter case it was only a moderate one; ODNs 1585, D19 (unpublished data), 1643, and H154 had no detectable influence. The two CpG-ODNs with the strongest influence on disease development are the most similar in sequence, differing by only two nucleotides at the 3′ end. Both of these are so-called CpG-B-ODNs, which are characterized by having one or more CpG sequences on a fully phosphorothioate-modified backbone, and by being particularly stimulatory for B cells (9). The ODNs with no effect on arthritis included ones classified as CpG-B (1643); CpG-A, with their mixed phosphorothioate/phosphodiester backbone, central CpG-containing palindrome, and terminal polyG motifs (1585); and CpG-N, which can neutralize the effects of immunostimulatory CpGs, including their induction of arthritis in certain contexts (10). The one CpG-N assessed, H154, was also unable to suppress the negative effect of CpG-Bs on arthritis in our system (unpublished data).
ODN1668 was optimally effective at inhibiting K/BxN serum-transferred arthritis when introduced coincident with, or shortly before injection of, the arthritogenic serum. However, it also had efficacy in a therapeutic mode, when administered as long as 4 d after serum transfer (Fig. 1 C). It did not seem to revert symptoms to the predisease state, but rather to stop them in their tracks.
CpG-ODNs operate over a time window between the opening of the vasculature and the influx of inflammatory cells
Mechanisms of pathogenesis in the K/BxN serum-transfer system have been extensively dissected (3, 8, 11–14). Quite similar to a reverse-passive Arthus reaction, arthritis development entails the following, sequentially: GPI–anti-GPI immune complex (IC) formation; activation of inflammatory and probably endothelial cells; induction of macromolecular leakage in joint vessels; complement activation; inflammatory cell influx; an amplification loop resulting in massive cytokine production; leukocyte recruitment and synovial hyperplasia; and, ultimately, joint destruction. To delineate precisely when CpGs impinge on the arthritogenic process, we evaluated their effect on the different disease subphenotypes that had been instrumental in constructing this pathogenetic scenario.
Consistent with the observed dampening of inflammation signaled by a reduction in ankle swelling, hematoxylin and eosin (HE)–stained ankle sections taken from mice 8 d after coadministration of K/BxN serum and ODN1668 revealed an absence of synovial hyperplasia and leukocyte infiltration (Fig. 2 A). IC deposition and complement activation, detected as GPI/IgG/C3d costaining on cryostat sections, was also absent from the ankles of serum/CpG-cotreated animals, although there was weak GPI binding at the cartilage–articular cavity interface (Fig. 2 B), as expected (15). Neutrophil recruitment was assayed 1 d after serum ± CpG administration by staining ankle sections with HE and chromogenic chloroacetate esterase substrate (CAE), which specifically identifies granulocytes/neutrophils. At this time point, the disease is still at an early stage, so only 3 out of 6 mice injected with serum alone showed infiltration of neutrophils; in contrast, none of the animals coinjected with ODN1668 exhibited neutrophil recruitment (Fig. 2 C).
However, in serum/CpG-cotreated animals the typical vascular permeability to macromolecules that can be visualized by laser-scanning confocal microscopy of an immobilized heel in vivo in real time, subsequent to injection of a Cy5.5-labeled graft copolymer imaging reagent, was present (Fig. 2 D, left; compare iv and ii). Specifically, the maximum rate of leak was the same in K/BxN serum-transferred mice with or without coadministration of ODN1668 (Fig. 2 D, right). Injection of this CpG in the absence of serum did not provoke vessel leakage (Fig. 2 D, compare iii and i). Because the vascular permeability provoked by serum transfer depends on ICs (11), it would appear that ODN1668 did not compromise GPI/anti-GPI IC formation per se, which is consistent with the observation of indistinguishable anti-GPI antibody (Ab) titers in the blood in the presence or absence of this ODN (unpublished data).
Thus, the window within which CpGs impinge on the unfolding of arthritis in this system is delimited on one side by opening of the vasculature (>2 min after serum injection) and on the other side by inflammatory cell influx subsequent to IC deposition, and the consequent activation of complement (<24 h after injection).
The CpG-mediated inhibitory effect is signaled through TLR9
CpGs are known to exert their immunostimulatory effects through binding to TLR9 (16). Yet, because it has been hypothesized that other receptors might be involved in CpG-induced activation (17), and because binding of ODNs 1668 and 1826 to TLR9 has generally been associated with stimulatory rather than suppressive influences (9), we wished to verify that CpG-mediated inhibition of K/BxN serum-transferred arthritis actually does result from their signaling through TLR9. First, TLR9-transfected human embryonic kidney (HEK) 293 cells were stimulated with the various ODNs whose properties are depicted in Fig. 1 B, and induction of a luciferase reporter gene driven by NF-κB–responsive regulatory elements was measured. There was a clear correlation between the strength of the TLR9-transduced signal and its disease-dampening capacity, ODNs 1668 and 1826 being strongest according to both criteria, ODN1758 being intermediate, and ODN1643 and H154 being the weakest (Fig. S1, left). None of the CpG-ODNs elicited a signal in control TLR2-transfected HEK-293 cells (Fig. S1, right). Second, according to data on KO mice, TLR9 was not required for arthritis to develop in response to K/BxN serum transfer; however, it was needed for the ODN1668-induced inhibition of this process (Fig. 3 A).
CpG-mediated inhibition of arthritis requires IL-12 and IFN-γ
To begin dissecting the molecular mechanisms underlying the dampening of K/BxN serum-transferred arthritis by ODN1668, we screened disease inhibition in several mouse strains deficient in the expression of cytokines, particularly some of those previously reported to operate downstream of TLR9 signaling. These experiments were, of course, limited to an assessment of strains that could develop disease in the first place, i.e., not those devoid of IL-1 or TNF-α.
Both IL-12p35 and IFN-γ (Fig. 3 B) were required for ODN1668 to exert its suppressive effect, as transfer of serum plus CpG, versus serum alone, into the corresponding KO strains provoked arthritis with the same kinetics and severity, i.e., with no sign of a CpG-mediated inhibitory effect. In contrast, IL-6 was not needed for effective inhibition (unpublished data).
There was an excellent correlation between the ability of a given ODN to inhibit arthritis and its capacity to induce IFN-γ and IL-12p40 gene expression in spleens of recipient mice 4 h after injection (compare Fig. 1 B and Fig. 3 C). In contrast, splenocyte transcription of several other cytokine genes (e.g., TNF-α, IL-15, IL-2, and IL-1) demonstrated no such correlation (unpublished data). The increase in IL-12p40 expression was associated with a transient augmentation in secretion of the IL-12 cytokine, itself, rather than of IL-23, which also contains the p40 subunit, as indicated by ELISA measurements of serum-cytokine levels (Fig. 3 D).
IL-12p40 expression was important for induction of IFN-γ transcription, although some increase in transcript levels was still evident in its absence (Fig. 3 E); IFN-γ expression also promoted the induction of IL-12 transcription (Fig. 3 F). As discussed below, this mutual dependence might be taken as evidence of a feed-forward loop.
Strikingly, systemic administration of IFN-γ was an effective inhibitor, by itself, of serum-transferred arthritis, arguing that systemic mechanisms are probably in play (Fig. 4 A). Because CpG blocked neutrophil accumulation in the joints, and IFN-γ was identified as the downstream effector of CpG, we next examined whether IFN-γ prevented joint inflammation by acting directly on neutrophil migration. Using a novel laser-scanning, time-lapse, intravital imaging system, we found that IFN-γ administration did inhibit recruitment of WT neutrophils into inflamed joints; however, this inhibition was lost when neutrophils from IFN-γR KO animals were used (Fig. 4, B and C). Note that the absence of IFN-γR on neutrophils did not, in and of itself, influence their migration into the joint under the influence of serum alone (unpublished data). Therefore, IFN-γ directly blocks neutrophil migration by binding to their IFN-γ receptors. Without neutrophil migration, arthritis cannot set in.
The case for DC–NK cell crosstalk
The combined requirement for IL-12 and IFN-γ, and their interdependent expression, evoke recent findings on the importance of the interplay between dendritic and NK cells, and the molecular crosstalk between them, in initiating responses by the innate immune system (18, 19). On one hand, DCs can prime, further the activation of, augment the expansion of, and enhance the activities of NK cells through the production of cytokines like IL-2, -12, -15, IFN-α/β, and TNF-α. On the other hand, mostly through the intermediary of IFN-γ, NK cells can enhance the differentiation, survival, and function of DCs, especially their expression of costimulatory molecules and production of IL-12 and TNF-α. Because most of the studies responsible for these conclusions were performed with cultured cells, the in vivo significance remains somewhat controversial. Nevertheless, we hypothesized that such a scenario underlies the CpG-mediated inhibition of K/BxN serum-transferred arthritis, and set out to test this notion.
As a first step, we performed a cytofluorimetric comparison of splenocytes from mice injected with PBS, the control CpG 1643, or the arthritis-inhibitory CpG 1668. As detailed in Fig. 5 (A and B), most of the examined cell populations had increased in number by 8 d after ODN1668 injection, a time point when the expansion of spleen cells was maximal and the inhibitory effect on arthritis was clearly evident. In particular, there was a striking overrepresentation of NK cells (CD3−NK1.1+; top) augmenting 18-fold in number and rising fractionally from an average of 4.4–22%.
Next, we examined cytokine mRNA induction in the different splenocyte populations (Fig. 5 C). The dominant IL-12–producing cells 4 h after injection of ODN1668 were DCs; NK cells were the major producers of IFN-γ, although NKT and DCs also made low amounts.
In mice, splenic CD8α+ and CD8α− DCs both express TLR9, and both can be induced to secrete IL-12 after stimulation in culture with CpG (20). However, we found that in vivo, CD8α+, but not CD8α−, DCs were the major cell type responsible for CpG-induced IL-12 induction (Fig. 5 D and Fig. S2). Maximum expression of IL-12 occurred at 1 h after CpG injection; in contrast, induction of IFN-γ expression by NK cells maxed later at 4 h.
We next sought to determine where TLR9 needed to be expressed for DCs to turn up IL-12 expression and NK cells to up-regulate IFN-γ transcription. CD45.2 TLR9−/− spleno cytes were mixed 1:1 with splenocytes from WT CD45.1 congenic B6 mice, and the mixture was stimulated with CpG. TLR9−/− DCs were not induced to express IL-12 (Fig. 5 E, left); a similar result was obtained by staining for intracellular IL-12 (Fig. S3). There was no difference in IFN-γ induction in TLR9−/− versus WT NK cells (Fig. 5 E, right). Thus, CpG engagement on CD8α+DCs appears to be the gateway for this antiinflammatory effect, inducing IL-12 by the DCs and subsequent activation of NK cells to make IFN-γ.
Lastly, the requirement for DCs and NK cells was assessed. Punctual depletion of DCs was performed by exploiting a mouse strain wherein DCs can be killed off by diphtheria toxin (DT) treatment of transgenic animals expressing the simian DT receptor (DTR) primarily on DCs, under the dictates of the mouse CD11c promoter (21). These mice were found to lack all CD11c+ DC populations, which reduced their capacity to control microbial infections. Because it was reported that the DC compartment began to reemerge 3–5 d after DT treatment, we needed to modify our protocol as per Fig. 6 A (i) to enable us to assay arthritis induction and CpG-mediated inhibition of it during this short time frame. The DC populations were strongly reduced by this protocol, whereas other cell-types, e.g., NK, T, and B cells, were not (Fig. 6 A [ii] and not depicted). Loss of DCs during this time frame had no detectable effect on arthritis progression; however, it did strongly compromise the inhibitory effect of ODN1668 (Fig. 6 A, iii).
IL-15–deficient mice have a strong deficiency in NK cells, but also have deficits in some populations of T lymphocytes, particularly memory T and NKT cells (Fig. S4 A) (22). However, the CpG-induced, arthritis-inhibitory effect was independent of all α/β T and B cell populations, as was evident with mice carrying a mutation of the TCRα locus or in animals lacking B cells caused by the μMT mutation (Fig. S5). IL-15 KO mice were fully susceptible to K/BxN serum-transferred arthritis, but not to inhibition of disease by ODN1668 (Fig. 6 B, left). Because it was recently reported that mice lacking IL-15 also have defects in some DC populations (although we did not find a numerical deficit; Fig. S4) (23), we tried a second strategy for eliminating NK cells. The polyclonal Ab anti–asialo-GM1 will deplete NK cells, though there are some reports that a subset of CD8+ T cells may be removed as well. Again, however, we know that T cells are not required for the CpG-induced suppression of arthritis (Fig. S5, middle). Anti–asialo-GM1 treatment reduced NK cells to ∼30% of the normal level, although not depleting the T or NKT cell populations (Fig. S4 B), but did not affect serum-induced arthritis. Such treatment did, on the other hand, compromise the ability of ODN1668 to inhibit disease unfolding (Fig. 6 B, right). Together, these two sets of findings on NK cell depletion argue that this cell type has a critical role in CpG-mediated inhibition of serum-transferred arthritis. Again, consistent with the proposed scenario of DC–NK cell cross talk were results on cytokine gene expression in the DTR and IL-15 KO mice; a strong reduction in both IL-12 and IFN-γ expression in the former case, with residual levels probably being attributable to incomplete DC depletion (Fig. 6 C, left), but a drop in only IFN-γ transcripts in the latter case (Fig. 6 C, right).
The therapeutic potential of CpGs has excited great interest of late (24). Their potent adjuvant activity can be applied to eliciting more effective immune responses to infectious agents or tumors. Their ability to bias responses in a TH1 direction can be harnessed to redirect pathogenic TH2 responses, e.g., in the context of allergy or asthma. Given this backdrop, the observation that other PAMs are known to have an enhancing effect on diverse rodent models of inflammatory arthritis (7, 8), and the precedent that CpGs, themselves, have been reported to promote arthritis in more than one model (6, 25), we were surprised to find that ODN1668 could block arthritis development in the K/BxN serum-transfer system. Even more striking, ODN1668 is one of the few agents thus far found to be capable of reversing ongoing arthritis in this system.
Our mechanistic studies afforded a second surprise. The original scenario constructed to account for the immunostimulatory activities of CpG-B-ODNs began with TLR9-signaled activation of B and DCs and ended with a potent response by the adaptive immune system, entailing Abs, TH1 cells, and cytotoxic CD8+ T cells (Fig. 7 A) (9). More recently, this scenario was enriched to include DC-induced stimulation of NK cells, but it still culminated in and highlighted the promotion of more effective T cell and Ab responses (24, 26). Cases of CpG-ODNs dampening immune responses have been described in the past—autoimmune (27–29), asthmatic (30), or standard T cell (31)—in nature, but, in all of these studies, data, or sometimes just mere assumption, focused attention on the response of the adaptive immune system (Fig. 7 B). In contrast, the arthritis-inhibitory property of ODN1668 described in this study did not depend on either B or T cells, and thereby reflected an influence solely on the operation of the innate immune system (Fig. 7 C). The key turned out to be cross talk between CD8α+DCs and NK cells, which relied critically and interdependently on IL-12 produced by the former and IFN-γ made (mostly) by the latter. The eventual outcome of this “conversation” was suppression of the recruitment of neutrophils, a direct effect of IFN-γ on that cell type. At first glance, our data may also appear reminiscent of the recent observations of Katakura et al. on experimental colitis (32). However, these authors described only a protective effect; in fact, when administered therapeutically, ODN1668 exacerbated colitis (33). Moreover, in this case, IFN-γ was pro- rather than antiinflammatory (33).
If one considers the extensive body of existing information on events downstream of CpG-ODN engagement of TLR9 on immune cells, one is faced with a complex, often-conflicting set of data (18, 19). An example is the differential reliance on priming and facilitating cytokines, such as IL-2, IL-15, IFN-α/β, and TNF-α, to optimally activate NK cells. Part of the difficulty in assessing and comparing results is that many of them emanate from in vitro experiments on cells derived from various tissues according to diverse expansion protocols. For example, this factor could easily explain why we saw no influence of IL-2 in either gain-of-function or loss-of-function in vivo experiments (Fig. S6), although it appeared to have a critical role in certain in vitro contexts (34, 35). Another issue is that the various experiments did not always use the same CpG-ODN; sometimes CpG-A versus -B, sometimes a different one of the many available CpG-Bs. This element is a critical one, as we saw from Fig. 1 B that even CpGs of the same class can have very different effects in vivo. Third, the picture that emerges is complicated by the difficulty for any study to be comprehensive in light of a constantly increasing cast of players. In this respect, we have briefly explored the roles in CpG-mediated protection of several elements that have elicited interest of late, including the following: indoleamine 2,3-dioxygenase (IDO) (31), interferon-producing killer DCs (IKDCs) (36), and plasmacytoid DCs (pDCs). Under a dose of 1-methyl-tryptophan, which is an IDO inhibitor previously reported to reverse the immunoresistance of IDO-expressing tumors (37), we did not see any effect on CpG-mediated arthritis inhibition (Fig. S7). This result is consistent with the observation that the immunosuppressive effect of IDO is known to work through T cells (31, 38), whereas CpG-mediated protection from arthritis is T cell–independent (Fig. S5). Likewise, the recently identified IKDC population may not have a substantial role in our system, as the majority of NK1.1-positive cells expanded after CpG treatment were CD11c-negative (Fig. S8). In contrast, we found that pDCs did have a partial effect on CpG-mediated protection from arthritis (Fig. S9). Ostensibly, this effect does not operate through the NK–DC/IL-12–IFN-γ axis we have focused on in this study, as pDCs do not produce either of these two cytokines (Fig. 5 D). The role of pDC is instead through a secondary axis involving Type I IFN (unpublished data). Lastly, existing information comes from a mix of mouse and human studies and, although the general outline is very similar in the two cases, details of the scenario certainly diverge, reflecting species differences such as the precise cell types that express TLR9 (9).
One of the key elements in the CpG-mediated inhibitory effect was IL-12. This dependence may explain the divergent results with the other TLR ligands we tested, as CpG is a much stronger IL-12 inducer than LPS or polyI:C (unpublished data) (39, 40). IFN-γ is the other important molecule mediating the protective effect. Systemic administration of IFN-γ prevented arthritis, and this effect appeared to be via blockade of neutrophil recruitment to the joint. Neutrophils were the direct targets of the IFN-γ–mediated inhibitory effect, as migration of those devoid of IFN-γR was not inhibited. One attractive hypothesis for IFN-γ's downstream mode of action is through down-regulation of CXCR4, as this cytokine has been described to control neutrophil migration in different contexts by posttranslational down-regulation of this chemokine receptor (41, 42).
Why did CpG-ODNs dampen arthritis in the K/BxN serum-transfer system, whereas they promoted disease in other arthritis models (6, 25)? One explanation is that local CpG injection (6) may be proinflammatory, but systemic delivery antiinflammatory. Such behavior would parallel observations on the effect of administering IFN-γ in the collagen-induced arthritis model; injection into the paw promoted disease, although back delivery was inhibitory (43, 44). An explanation in another case (25) is that the influence of CpGs may differ according to when they are introduced, with the lymphocyte-mediated initiation phase being more prone to an enhancing effect. One strength of the K/BxN serum-transfer system is that it focuses uniquely on the effector phase mediated by cells of the innate immune system, which is the phase when most patients would present. Lastly, it may be relevant that the K/BxN system does not entail other potentially synergistic TLR ligands, such as adjuvant (25) or LPS (7).
We hope that our findings will be translatable to the human arthritis context. Differences in TLR9 expression (or other factors) between mice and humans may mean that translation will not be entirely linear, but there should be many commonalities in the molecules, cells, pathways, and networks involved, so that their identification and characterization should provide points of access to the inhibitory process beyond CpG-ODNs themselves. In this regard, it is encouraging that the antiinflammatory properties of IFN-γ are being increasingly appreciated and studied in both species, leading to the recognition that some of the critical molecular mechanisms are shared, at least as far as can be concluded from cell culture experiments (45). Also promising is that clinical trials of rhIFN-γ in small cohorts of RA patients in the 1980s and 1990s proved it safe, and usually (46–48), but not always (49), somewhat effective. It may be worth revisiting this strategy, or allied therapeutic strategies, especially if it becomes possible to phenotypically or genetically stratify the patients most likely to have a beneficial response.
Materials And Methods
C57BL/6 (B6), B6.SJL (CD45.1+ congenic B6 mice), and NOD/Lt (NOD) mice were purchased from The Jackson Laboratory. KRN TCR-transgenic mice, which were described elsewhere (50), were maintained on the B6 genetic background (K/B6). Crossing K/B6 animals with NOD mice generated arthritic K/BxN offspring. H-2g7 congenic mice on the B6 background (B6g7) were bred in our animal facility. The various gene-deficient mice are described in the Supplemental materials and methods . All mouse procedures were conducted in compliance with federal and institutional guidelines.
Reagents are described in the Supplemental materials and methods.
Serum-transfer system and evaluation of arthritis.
Optimal serum-transferred arthritis was induced by i.p. injection of 100–150 μl of pooled serum from 8-wk-old K/BxN mice on days 0 and 2, unless otherwise mentioned. Suboptimal arthritis was induced by i.p. injection of 100 μl K/BxN serum on day 0. Ankle thickness was measured with a caliper (J15 Blet micrometer). Each limb was scored on a scale from 0 (no observable swelling) to 3 (severe inflammation), and the four-limb scores were summed to yield a clinical index (maximum 12 points).
Ankle sections were stained with hematoxylin and eosin or with CAE plus hematoxylin as a counterstain, as previously described (50, 51). Neutrophils exhibited a red cytoplasmic staining with CAE, which is a marker of the myeloid and mast cell lineages that does not stain monocytes, macrophages, or basophils. Immunohistology methods are detailed in the Supplemental materials and methods, with procedures carried out basically as previously described (15).
Intravital confocal microscopy.
Microscopy was performed as recently published (11), and is detailed in the Supplemental materials and methods.
Neutrophil imaging: a laser-scanning, time-lapse intravital microscopy system.
The imaging system for neutrophils is described in detail in the Supplemental materials and methods.
Anti-GPI Ab titers were measured as previously described (52). In brief, recombinant mouse GPI was coated on ELISA plates at 5 mg/ml, and diluted mouse sera was added. Subsequently, alkaline phosphatase (AP)–conjugated anti–mouse total IgG, followed by AP-conjugated streptavidin, was applied. After substrate addition, titers were quantitated according to the optimal density (OD) values from an ELISA reader. Serum levels of IL-12 and -23 were measured by an ELISA kit, using the instructions provided by the company (eBioscience). Capture-Ab clones C18.2 (anti–IL-12p35) and G23-8 (anti–IL-23p19) were specific for their respective cytokines. Both detection systems used clone C17.8 as the detection Ab for the p40 subunit shared by the two cytokines. Standard curves were generated using the recombinant IL-12 or -23 provided by the kit.
Abs and flow cytometry techniques are described in the Supplemental materials and methods.
An equal ratio of splenocytes from TLR9−/− (CD45.2+) and B6.SJL mice (CD45.1+) mice were cultured at 2 × 106/well in 96-well plates with 1 μM CpG. Cells were harvested after 2 or 4 h and sorted into WT or TLR9−/− DC and NK cell populations based on the CD45 alleles expressed.
NK cell depletion.
Polyclonal rabbit anti–asialo-GM1 Ab (Cedarlane Laboratories) was used for in vivo depletion of NK cells. Mice were injected i.p. with 50 μl of Ab on days 1, 2, 5, and 9 via serum injection or with an equivalent amount of rabbit gamma-globulin (The Jackson Laboratory).
RNA was isolated from splenocytes via Trizol and reverse-transcribed using oligo dT priming and Superscript polymerase (Invitrogen). Quantitative RT-PCR was performed on an Mx3000p instrument (Stratagene), using gene-specific fluorogenic assays (TaqMan; Applied Biosystems). Primers and probes are listed in the Supplemental materials and methods. The cytokine transcripts in spleens were quantitated by RT-PCR using hypo xanthine guanine phosphoribosyl transferase mRNA as an internal standard. Fold induction represents relative expression levels of cytokine transcripts from CpG-ODN–treated mice standardized against the level of cytokine transcripts from untreated B6 mice.
Online supplemental material.
Fig. S1 shows various signal strengths induced by different CpG-ODNs. Fig. S2 is a FACS plot of both CD8α-positive and -negative DC population. Fig. S3 shows IL-12 staining in both WT and TLR9 KO DC after ODN1668 stimulation. Fig. S4 shows lack of NK cells in IL-15 KO mice and depletion of NK cells in anti–asialo-GM1-treated mice. Fig. S5 shows that T and B cells are not required for CpG-mediated protection. Fig. S6 demonstrates that IL-2 is not induced by CpG stimulation. Fig. S7 shows CpG-mediated protection is IDO independent. Fig. S8 shows that CpG only induces minor IKDC expansion. Fig. S9 shows that pDC is partially required for CpG-mediated protection.
We thank V. Tran and K. Hattori for assistance with the mouse colonies and serum collection; Dr. V. Rubin-Kelley for providing some of the IL-15−/− mice; Dr. A. Ortiz-Lopez for maintaining the RT-PCR primer database; P. Patel for imaging; J. Lavecchio and G. Buruzula for flow cytometry; T. Bowman and M. Donovan for histology; and C. Laplace for assistance with figure preparation.
This work was supported by a grant from the National Institutes of Health (PO1 AI54904) to D. Mathis, C. Benoist, and R. Weissleder, by a grant from Biogen Idec, Inc. to D. Mathis and C. Benoist, by a mouse imaging resource (R24 CA 92782), and by the Joslin Diabetes Center's National institute of Diabetes and Digestive and Kidney Diseases–funded Diabetes and Endocrinology Research Center cores. H.-J. Wu was partially supported by a Ruth L. Kirschstein National Research Service Award, and B. Binstadt by a Pfizer Postdoctoral Fellowship in Rheumatology/Immunology.
The authors have no conflicting financial interests.
Abbreviations used: Ab, antibody; AP, alkaline phosphatase; CAE, chloroacetate esterase substrate; DT, diphtheria toxin; DTR, DT receptor; GPI, glucose-6-phosphate-isomerase; HE, hematoxylin and eosin; HEK, human embryonic kidney; IC, immune complex; IDO, indoleamine 2,3-dioxygenase; IKDC, interferon-producing killer dendritic cell; OD, optimal density; ODN, oligodeoxynucleotide; PAM, pathogen-associated molecule; pDC, plasmacytoid DC; PGN, peptidoglycan; RA, rheumatoid arthritis; TLR, Toll-like receptor.