A GpC-Rich Oligonucleotide Acts on Plasmacytoid Dendritic Cells To Promote Immune Suppression

One of the strategies used by mammalian cells in innate and adaptive immunity to sense pathogens implies selective recognition of unmethylated oligonucleotide sequences containing CpG motifs, particularly abundant in bacterial and viral but not mammalian genomic DNA (1). Surveillance mechanisms in the host involve, indeed, pattern recognition receptors occurring on the cell surface and in the cytoplasm. TLR7, TLR8, and TLR9, which are present in intracellular compartments, respond to single-stranded nucleic acids, activating innate and adaptive immune responses. TLR9, expressed by dendritic cells (DCs), B cells, and macrophages, not only reacts to pathogen-associated molecular patterns but also binds short synthetic CpG-containing oligodeoxynucleotides (ODNs), resulting in lymphocyte maturation, enhanced APC function, and the release of inflammatory cytokines and type I IFNs (2). These remarkable immunostimulatory properties of short CpG-ODNs, and mostly of those with a phosphorothioate-stabilized backbone (3), have suggested a possible therapeutic use as vaccine adjuvants for prevention of viral, bacterial, or parasitic diseases. CpG-ODNs, when used in vaccination trials in healthy human volunteers (4, 5) and in immunocompromised, HIV-infected patients (6), manifested strongly enhanced vaccination efficiency (7, 8). An antitumor activity for CpG-ODNs has also been advocated when CpG-ODNs are used either alone or in combination with chemotherapy (9, 10).

In addition to their immunostimulant activity and by virtue of their ability to induce Th1 and regulatory T cell responses that suppress pathogenic immunity (11), CpG-containing DNA sequences manifest beneficial activity in experimental settings of allergy (12) and autoimmunity (13). Type I allergic diseases and systemic autoimmune diseases depend on pathogenic T cell responses. These harmful immune responses could thus be modified by engaging TLR9 and TLR7, paving the way to new forms of treatment. However, data from mouse models suggest a role for abnormal TLR7- and/or TLR9-mediated immune activation in systemic autoimmune diseases, DNA-mediated sepsis, or other conditions dominated by chronic inflammation (12). A pathogenic role for TLR7 and a protective role for TLR9 in the pathogenesis of systemic lupus erythematosus has recently been documented (14). This is in line with previous observations that inhibitory ODNs, including a GpC-ODN, have a potential for treating experimental lupus nephritis (15).

Overall, TLR9- and TLR7-dependent signaling manifests both potentially therapeutic and detrimental effects on host immunity as a function of the experimental setting and/or TLR activation modalities (16, 17). In most experimental settings of chronic inflammation and autoimmunity, the protective effects of TLR9 ligation are associated with the induction of IDO, an enzyme that—initiating the suppressive mechanisms of tryptophan depletion and kynurenine production (18)—is operative in both the innate and adaptive forms of immunity (13, 19).

Although not investigated in detail as yet, GpG-ODNs, which have been demonstrated to be nonstimulatory ligands of C-terminal TLR9 (20), appear to be beneficial in at least two experimental models of autoimmunity. In one study, a GpG-ODN suppressed the severity of experimental autoimmune encephalomyelitis, downregulating autoreactive Th1 and B cell responses (21). In another study, GpG-ODN delayed the onset and attenuated the severity of lupus nephritis by antagonizing and blocking the activation of multiple TLRs (22). Dual or triple activation of TLR7, TLR8, and/or TLR9 by single-stranded ODNs may, in fact, occur (23). Of interest, plasmacytoid dendritic cells (pDCs) contribute to the protection of the lung against influenza A virus infections, mainly via TLR7 signals (24).

In this study, we investigated the possible immunoregulatory effects of GpC-ODNs. We found that 1) a GpC-ODN selectively conferred suppressive properties on pDCs, contingent on functional IDO; 2) the induction of IDO by GpC depended on autocrine TGF-β and noncanonical NF-κB transcriptional activity; and 3) the downstream response culminating in IDO induction required TLR7/TIR domain-containing adapter inducing IFN-β (TRIF)-mediated, but not TLR9-mediated, signaling events.

Female C57BL/6 mice were obtained from Charles River. Mice homozygous for the TLR9 (Tlr9^−/−), MyD88 (Myd88^−/−), or TRIF (Ticam1^−/−) targeted mutation raised on the C57BL/6 background were generated as described (13) and bred at the animal facility of the University of Perugia. Mice homozygous for the TLR7 (Tlr7^−/−) mutation were obtained from The Jackson Laboratory. The HY peptide (amino acid sequence, WMHHNMDLI) was synthesized and purified as described (25). Endotoxin-free GpC-ODN 1826 (5′-TCCATGAGCTTCCTAAGCTT-3′) and the negative control thereof (5′-TCCATGAATTTCCTAAATTT-3′), both on a phosphorothioate backbone, were purchased from Invitrogen Life Technologies. In selected experiments (Fig. 1), a second ODN was used (1668; 5′-TCCATGAGCTTCCTGATGCT-3′) after replacing the CpG motif with a GpC sequence (negative control ODN, 5′-TCCATGAATTTCCTGATATT-3′). Endotoxin contamination, when measured by the Limulus assay (Sigma-Aldrich), was negligible in all ODN preparations. Imiquimod (Sequoia Research Products, Pangbourne, U.K.) was used at the concentration of 40 ng/ml. All in vivo studies were done in compliance with national (Italian Parliament DL 116/92) and Perugia University Animal Care and Use Committee guidelines.

Splenic DCs were prepared and fractionated according to CD11c/CD8 expression using positive selection columns in combination with CD11c and CD8 MicroBeads (Miltenyi Biotec) and in the presence of EDTA to disrupt DC–T cell complexes, as previously described (26). The recovered cells were >99% CD11c⁺, >99% MHC I-A⁺, >98% B7-2⁺, <0.1% CD3⁺, and appeared to consist of 90–95% CD8^–, 5–10% CD8⁺, and ∼5% mPDCA-1⁺ cells. After cell fractionation, the recovered CD8⁻ cells were ∼45% CD4⁺ and typically contained <0.5% contaminating CD8⁺ DCs, whereas the CD8⁺ fraction was made up of >95% CD8⁺ DCs. For positive selection of mPDCA-1⁺ pDCs, we fractionated CD11c⁺ cells using mPDCA-1 MicroBeads (Miltenyi Biotec). More than 95% of the mPDCA-1⁺ cells were stained by the 120G8 marker. pDCs were exposed for the indicated times to 1 μg/ml ODN (GpC or negative control ODN) in the presence or absence of 4 μM 1-methyl-dl-tryptophan (1-MT) at 37°C.

For immunization, cells were washed between and after incubations before peptide loading (5 μM, 2 h at 37°C), irradiation, and i.v. injection into recipient hosts. A total of 3 × 10⁵ peptide-pulsed CD8^– DCs (majority fraction) were injected either alone or in combination with a minority fraction (5%) of pDCs from wild-type, Tlr7^−/−, Tlr9^−/−, Myd88^−/−, or Ticam1^−/− mice that had been stimulated overnight with 1 μg/ml ODN. A skin test assay was used for measuring class I-restricted delayed-type hypersensitivity responses to synthetic peptides (25). At 2 wk, we measured the response to intrafootpad challenge with the eliciting peptide. Results were expressed as the increase in footpad weight of peptide-injected footpads over that of vehicle-injected counterparts. Data are the mean ± SD for at least six mice per group. Statistical analysis was performed using Student paired t test by comparing the mean weight of experimental footpads with that of control counterparts.

Human mature FLT3 ligand-cultured dendritic cells (FL-DCs) were obtained from PBMCs. Briefly, PBMCs were separated from whole blood of healthy donors by centrifugation through a density gradient of Ficoll-Paque (GE Healthcare). Whole blood was layered onto a sterile aqueous medium containing Ficoll at a predetermined density of 1.077 g/ml at 25°C. Gentle centrifugation at room temperature resulted in the separation of PBMCs at the blood–Ficoll interface, with the other WBCs and RBCs passing through the interface and accumulating at the bottom of the tube. Human FL-DCs were obtained from purified CD14⁺ monocytes from healthy donors by culturing cells in Iscove modified medium for 7 d in the presence of 50 ng/ml human FLT3L (Peprotech). The resulting FL-DCs were washed and cultured for 18 h with 1 μg/ml of a GpC-containing ODN 2006 (5′-TGCTGCTTTTGTGCTTTTGTGCTT-3′) or negative control ODN (5′-TATTATTTTTGTATTTTTGTATTT-3′). FACS analysis revealed that FL-DCs were CD123⁺ CD11c ^low BDCA2⁺ and BDCA4⁺. ODN-treated FL-DCs were cocultured with CD4⁺ T cells isolated from the same donor for 4 d, and FOXP3 expression was revealed by intracellular FACS analysis by means of a specific Alexa Fluor 488-labeled Ab (BioLegend).

Small interfering RNA (siRNA) sequences targeting Ido1 (sense, 5′-GGGCUUCUUCCUCGUCUCUtt-3′; antisense, 5′-AGAGACGAGGAAGAAGCCCtt-3′), Chuk (sense, 5′-GGAAUAAAUACAGGUUCUCtt-3′; antisense, 5′-GAGAACCUGUAUUUAUUCCtg-3′), and Ikbkb (sense, 5′-GGUGCAUUCAUUAUCUUAAtt-3′; antisense, 5′-UUAAGAUAAUGAAUGCACCtg-3′) and control, scrambled sequences were synthesized as described (26). For transfection, siRNAs (6.7 μg) in 30 μl transfection buffer (20 mM HEPES, 150 mM NaCl; pH 7.4) were pipetted into a sterile Eppendorf tube. In a separate tube, 6.7 μg 1,2-dioleoyl-3-trimethylammonium-propane was mixed with 30 μl transfection buffer, and then both solutions were gently mixed. After incubation at room temperature for 20 min, the mixture was added to 1 ml complete medium containing 10⁶ pDCs and incubated for 24 h at 37°C. 1-MT, GpC-ODN, or negative control ODN was added 5 h after the specific silencing procedure. Cells were then recovered, washed, and immediately used for in vitro or in vivo experiments. Nuclear localization of NF-κB subunits was assessed by immunoblotting using anti-p100/52 (Cell Signaling) and anti-p65 Abs (Santa Cruz Biotechnology). Anti-nucleolin Ab (Santa Cruz Biotechnology) was used as a normalizer. IDO expression was investigated in cells cultured for 24 h with or without 1 μg/ml GpC-ODN (in specific experiments in the presence of neutralizing anti-TGF-β, 50 μg/ml, or an isotype control) by immunoblotting with rabbit anti-mouse IDO mAb raised in our laboratory (27). Anti–β-actin Ab (Sigma-Aldrich) was used as a normalizer. For measurement of IDO functional activity, pDCs were stimulated with 1 μg/ml GpC-ODN and, after 18 h, l-kynurenine, the initial tryptophan catabolite, was measured in culture supernatants by HPLC, as described (28).

Unmethylated synthetic ODNs affect immune responsiveness in different ways (29). A single base substitution of guanine for cytosine in the CpG motif converts this mostly stimulatory ODN into an autoimmune-suppressive GpG-ODN (21). We asked whether a guanine/cytosine inversion from CpG to GpC in ODN 1826 (hereafter referred to as GpC-ODN) would also affect the qualitative nature of ODN effects. Control and GpC-ODNs—on a phosphorothioate backbone to improve stability to nucleases—were tested for effects on pDCs, a DC type highly efficient in sensing microbial DNA and mostly acting as an immunogenic APC. The GpC-conditioned pDCs were tested for modulatory activity in a skin test assay for measuring the induction of class I-restricted reactivity to intrafootpad challenge with the synthetic peptide HY in female mice (30). In the afferent induction of the response, and according to an established protocol, we injected immunogenic, peptide-pulsed splenic CD8^– DCs either alone or in combination with pDCs (a fraction constituting 5% of the final cell mixture), which were either untreated or treated with 1 μg/ml GpC-ODN or the negative control ODN (Fig. 1). In control groups (i.e., in mice cotransferred with untreated or control ODN-treated pDCs), significant skin test reactivity was detected, indicating that the pDCs, as such, were ineffective in opposing host priming by immunogenic CD8^– DCs in our setting. However, cotransfer of recipient mice with pDCs treated with GpC-ODN negated the onset of HY-specific immunity at 2 wk of cotransfer. Of interest, the tolerogenic effect of GpC-ODN 1826 with a CpG to GpC inversion was shared by another ODN (1668) in which the CpG sequence had likewise been replaced by a GpC motif (Fig. 1).

Therefore, GpC-ODNs, used in vitro for conditioning pDCs, conferred strong suppressive activity on those cells, preventing the induction of skin test reactivity.

In previous studies, we have found that most molecules acting as tolerogenic stimuli in the skin test assay require functional IDO expression by DCs, used for sensitization either as a single population or as a minority fraction on cotransfer with immunogenic (IDO-negative) cells (31). To investigate a possible role for IDO in the suppressive activity of GpC-ODN–conditioned pDCs, we used immunogenic CD8^– DCs admixed with 5% pDCs treated with GpC-ODN (or negative control ODN) for skin test sensitization. The conditioned pDCs were used either as such or after treatment with the IDO inhibitor 1-MT or after silencing of Ido1 by specific siRNA (or control, scrambled siRNA) (Fig. 2A). The suppressive effect associated with coinjection of GpC-ODN–treated pDCs was negated by either maneuver, namely enzyme inhibition (via 1-MT treatment) or Ido1 silencing. Moreover, immunoblot (at 24 h; Fig. 2B) and l-kynurenine level assessment (at 18 h; Fig. 2C) revealed increased IDO expression and function in pDCs exposed to GpC-ODN but not to negative control ODN. These data suggest that the suppressive effect of GpC-ODN in fostering a tolerogenic phenotype in pDCs was contingent on IDO expression and function when GpC-ODN was used at concentrations at which CpG-ODN is instead proinflammatory (32). GpC was invariably tolerogenic when tested over a wide range of concentrations, that is, up to 10 μg/ml (data not shown).

In most experimental settings, IDO induction requires either type I/type II IFNs (28) or TGF-β (33). We investigated whether IDO induction by GpC-ODN is a direct or, rather, an indirect effect, requiring specific cytokine production. Purified pDCs were exposed to GpC-ODN or negative control ODN (1 μg/ml, overnight), and supernatants were assayed for production of IFN-α, IFN-γ, TGF-β, IL-10, IL-6, and IL-23 (Fig. 3A). TGF-β was the only cytokine significantly induced by the GpC-ODN, and its production appeared to be necessary for IDO protein expression (Fig. 3B). Neutralization of TGF-β by specific Ab negated the tolerogenic effect conferred by GpC-ODN on pDCs, as measured in a skin test assay (Fig. 3C).

Besides their potent stimulatory role in innate immunity, pDCs also provide suppressive function—detectable by skin test assay—once cells are primed with tolerogenic signals (including high-dose CpG-ODN, CTLA-4–Ig, CD200–Ig, 4-1BB ligand, CD40L, and GITR–Ig) that tip the balance of canonical and noncanonical NF-κB activation in favor of the noncanonical pathway, which is necessary for Ido1 transcription (31, 32). To ascertain whether GpC-ODN would affect noncanonical NF-κB activation in skin test priming, we used otherwise tolerogenic, GpC-conditioned pDCs after silencing the expression of the kinases—IKKα, encoded by Chuk, and IKKβ, encoded by Ikbkb—indispensable in noncanonical and canonical NF-κB activation, respectively. Using the same experimental design as in Fig. 2A—and using siRNA-treated pDCs as a minority fraction combined with an immunogenic, majority counterpart—we found that IKKα, but not IKKβ, was required for the induction of the suppressive phenotype by GpC-ODN in pDCs (Fig. 4A).

To gain insight into the events upstream of noncanonical NF-κB activation, we investigated the role of the adapters MyD88 and TRIF, which are part of distinct intracellular TLR-associated transduction pathways. Thus, nuclear localization of the noncanonical NF-κB subunit p52—and of p65, belonging in the canonical pathway—was quantified by means of an immunoblot assay, using nuclear extracts from wild-type, Ticam1^−/−, or Myd88^−/− pDCs, incubated for 30 or 60 min with GpC-ODN (Fig. 4B). In wild-type C57BL/6 and in Myd88^−/− cells, we found that GpC-ODN activated nuclear translocation of p52, but not p65, at as early as 30 min of exposure. In contrast, in Ticam1^−/− cells, nuclear translocation of neither subunit occurred, suggesting a specific activation of noncanonical NF-κB by GpC-ODN in response to TRIF signaling.

The TRIF requirement for GpC-ODN activity was further investigated in vivo by a skin test assay, showing that appropriately conditioned pDCs from Ticam1^−/− (but neither wild-type nor Myd88^−/−) mice would allow the induction of skin test reactivity by cotransferred CD8^– DCs (Fig. 4C). Therefore, GpC-ODN did not activate the proinflammatory, canonical pathway of NF-κB, and, in fact, rendered pDCs tolerogenic, through TRIF and noncanonical NF-κB–dependent transcriptional expression of Ido1.

Among various TLRs, TLR9 is an intracellular receptor recognizing microbial DNA, whose ligation by ODNs containing CpG or GpG motifs has been found to trigger MyD88-mediated signaling that may either activate or suppress immune responses, depending on a specific setting (13, 22). The observation that GpC-ODN signaling in our experimental paradigm required TRIF as do TLR7 ligands (34) prompted us to investigate any role of TLR7 in GpC effects. In the afferent induction of skin test reactivity, we injected peptide-pulsed immunogenic CD8^– DCs admixed with Tlr7^−/−or Tlr9^−/− pDCs, treated with either negative control ODN or with GpC-ODN (Fig. 5A). The results showed that Tlr7^−/− pDCs, but not Tlr9^−/− pDCs, were unresponsive to tolerogenic GpC-ODN treatment. These results suggest that GpC-ODN activated a TLR7/TRIF-dependent pathway in pDCs, which was independent of TLR9 signaling. Lack of Tlr7 in pDCs also ablated induction of IDO protein (Fig. 5B) and function by GpC-ODN, as measured by kynurenine production (Fig. 5C). Of interest, the GpC-ODN was able to negate the induction of inflammatory cytokines (IFN-α, TNF-α, and IL-6) by an equimolar concentration of imiquimod, a reference activating ligand of TLR7. Under the same experimental conditions, however, the copresence of GpC-ODN and imiquimod allowed for the production of TGF-β (Fig. 5D).

We investigated the possible effects of GpC conditioning on human FL-DCs by analyzing IDO expression and FL-DC ability to promote a regulatory phenotype in naive T cells. Purified FL-DCs were exposed to modified GpC-ODN 2006 (1 μg/ml, overnight) and were then used in immunoblotting experiments (Fig. 6A). Unlike the negative control ODN, the human TLR9-recognizing sequence containing a GpC motif strongly induced IDO expression. When the GpC-conditioned FL-DCs were cocultured in vitro with human purified CD4⁺ T cells for 4 d, the recovered T cells manifested strong FOXP3 induction—a marker of regulatory T cell activity—on FACS analysis (Fig. 6B).

Over the past few years, unmethylated synthetic DNA sequences containing CpG motifs have been advocated as potential vaccine adjuvants capable of evoking innate and adaptive anti-microbial as well as anti-tumor responses through ligation of TLR9, an intracellular receptor that—expressed by sentinel cells such as pDCs and macrophages—is involved in immune surveillance (29). However, some studies have documented that CpG-ODNs may also lead to IDO-dependent activation of suppressive pathways as a physiological means of restoring the homeostatic balance and avoiding hyperinflammatory responses (13, 16, 32, 35). Thus CpG-ODNs are endowed with dual activity, and their adjuvant effects might, at least in part, be blunted by the activation of suppressive mechanisms, rendering their overall immunotherapeutic efficacy either suboptimal or unpredictable.

Besides CpG-rich stimulatory sequences, nonstimulatory and/or inhibitory TLR9 ligands have also been searched, reflecting the need for identifying TLR9-targeting sequences capable of opposing hyperinflammation and autoimmunity. Studies have revealed a complex relationship between ODN structure, TLR9 binding, and functional activity of several oligonucleotides. Structurally different sequences can differently modulate TLR9 cleavage—strictly required for ODN binding to C-terminal TLR9 and activation—and thus bias functional outcomes of TLR9 ligation (12, 36). In particular, an ODN containing GpC motifs was found to act as a nonstimulatory ligand because it failed to bind C-terminal TLR9 efficiently (20). In contrast, an ODN containing CpG inhibitory sequences was an effective antagonist of CpG responses by virtue of a competitive mechanism on C-terminal TLR9 (12, 20).

In this study, we demonstrate that a single base pair switch, converting the CpG into a GpC motif, affects TLR binding (20) and the qualitative type of the associated immune response. In particular, we found that GpC-ODN, sensed by pDCs as an exquisitely tolerogenic signal, invariably activated the suppressive program of noncanonical NF-κB–dependent transcriptional expression of the IDO-encoding gene (31). We also found that noncanonical NF-κB activation would not result in our setting from TLR9 ligation and MyD88-dependent transduction, as is the case for CpG, but from TLR7/TRIF-dependent events.

Among TLRs, TLR3 is typically associated with the TRIF transduction pathway, yet TLR3 is poorly expressed by pDCs (37). Sustained IDO induction depends on noncanonical NF-κB activation (19). Because TRIF-mediated signaling may involve noncanonical NF-κB family members (38, 39), we hypothesized that a TLR7/TRIF axis could be at work in IDO induction by GpC-ODN. Our data revealed that GpC-ODN acts as a tolerogenic signal for pDCs, which are also capable of initiating IDO-dependent long-term suppression independent of IDO’s catalytic function (40). Notably, imiquimod is also a TLR7 ligand, but in contrast to GpC-ODN in our setting, it is considered to be an immunostimulant molecule when used at therapeutic concentrations (41). Nevertheless, contrasting results have been reported in the literature as resulting from TLR7 ligation, which either inhibited Th17 responses via induction of IL-10 (42) or triggered Th17-type responses in humans (43). It is possible that different ODN sequences activate distinct transduction pathways upon TLR7 ligation, leading to opposite functional outcomes. Also, it is well documented that TLR7 activation by ssRNA and small molecules such as imidazoquinolines is mainly MyD88 dependent. In response to the imidazoquinolines, neither MyD88- nor TLR7-deficient mice show any inflammatory cytokine production by macrophages, proliferation of splenocytes, or maturation of DCs (44). Therefore, our finding that GpC-induced signaling events are abolished in pDCs from both TRIF- and TLR7-deficient mice points to a previously undescribed pathway of TLR7 signaling.

Physiologically, TLR7 recognizes and responds to viral ssRNA through a signal transduction pathway leading to both induction of type I IFNs—typically involved in virus elimination—and differentiation of DCs (45). Imiquimod, resiquimod (R-848), and loxoribine are RNA homologs widely used as TLR7 agonists to mimic viral infection and thus evoke functional responses in DCs, resulting in immune activation. Abnormal TLR7-mediated immune activation has been advocated in the pathogenesis of systemic lupus erythematosus (14). Remarkably, the GpC-ODN used in this study acted as an exquisitely tolerogenic agonist of TLR7, initiating signaling events that culminated in tolerance induction. Therefore, TLR7 appears to be capable of mediating opposite functional effects, depending on the ligand nature and experimental setting. These observations may pave the way to the use of TLR7 as a new potential target of chemically modified ODNs for induction of immunosuppression in negative vaccination strategies that would target autoimmune conditions.

We thank G. Andrielli for technical assistance and digital art.

The authors have no financial conflicts of interest.