It was first observed by Pisetsky and his group [51
] that several G-rich ODNs made with the PS, but not PO, backbone had inhibitory activity. We and others [39
] have extended these early observations to define structural requirements for TLR9 stimulation and inhibition in the nanomolar range. Strikingly, changes at the 5' end of an ODN (either inhibitory or stimulatory), particularly those affecting the pyrimidine-rich CCT triplet, diminished both the stimulatory and inhibitory activity for the TLR9 pathway [39
]. We have also identified a stretch of three (four) consecutive Gs necessary for optimal inhibition in the nanomolar range [39
Could DNA-like therapeutics offer a pathway-specific tool for treating systemic lupus? Variants of our prototypic INH-ODN 2114 (TCCTGGAGGGGAAGT) [44
] have already been tested in lupus-prone animals. For example, Patole and colleagues [36
] found that INH-ODN 2114 was active in the MRL-lpr/lpr
strain, whereas Klinman's group [35
] discovered that an ODN containing multiple TTAGGG repeats, like those in telomeric DNA, was inhibitory in lupus-prone NZB/NZW F1 mice. Barrat and colleagues [37
] recently obtained similar results in the NZB/W-F1 strain, with an INH-ODN that combined the canonical TLR9-inhibitory motif with a TLR7-specific TGC triplet at the 5' end. One dsDNA-based biocompound LJP 394 (Abetimus) developed by the La Jolla Pharmaceutical Company (San Diego, CA, USA) has already entered clinical trials in human lupus and showed some promising results (for example, ability to decrease anti-dsDNA antibody production in vivo
The current report addresses the relevance of secondary structure to INH-ODN activity. We created class R INH-ODNs by starting with the shortest strongly active 12-mer linear INH-ODN 4084-F (CCTGGATGGGAA) (Table ) and then extending it with 12 more bases, resulting in complete palindromes (class R) or in non-palindromic linear sequences (class B). Class R INH-ODNs were between 10- to 30-fold less potent as inhibitors in resting mouse B cells than class B INH-ODNs when synthetic linear TLR9 agonists were used for stimulation (Figure ). Similar findings were obtained in human primary B cells/B-cell lines. In contrast to resting B cells, when 'professional antigen-presenting cells' (for example, primary macrophages and bone marrow-derived DCs) were stimulated either with complex CpG-class A(D) ODNs or with DNA-containing immune complexes, classes R and B INH-ODNs showed very similar inhibitory potencies (Figure ). Stimulation with these complex TLR9 ligands is believed to occur in DCs in an early endosomal compartment [7
This difference in inhibitory activity between class B and class R INH-ODNs in resting (follicular) B cells was clearly dependent on the ability of INH-ODNs to make double-stranded secondary structures. The biggest decrease in potency was observed with class R INH-ODNs containing either complete palindromes or short (up to three nucleotides) single-stranded 5' or 3' linear overhangs (Figure ). Further increase in the length of the single-stranded overhang progressively reduced the difference between linear and class R INH-ODNs.
These results suggest a hypothesis that, in B cells (though not in macrophages), dsDNA (including class R INH-ODNs) has limited access to endosomal TLR9 relative to single-stranded DNA. Single-stranded CpG-ODNs have been shown to encounter TLR9 in LAMP-1-positive endosomes [8
]. We postulated that, if dsDNA gains access to these endosomes by BCR-mediated entry, the difference in potencies between double-stranded and single-stranded versions of the same inhibitory sequence would disappear. This hypothesis could be tested in transgenic B cells expressing a BCR either for dsDNA or for Ig by stimulating the cells with one or both of these components.
Thus, important proofs of concept emerged from studies in AM14 B cells (Figures and ). These rheumatoid factor-specific B cells express a transgenic BCR derived from hybridoma-secreting anti-Ig-specific autoantibodies [43
]. Their BCR recognizes the Fc-fragment of IgG2a antibodies expressing 'a' but not 'b' allotype [25
]. In vitro
, AM14 B cells proliferate when anti-nucleosome antibodies (PL2-3) are added to spent cultures (Figure ). This proliferation is sensitive to DNAse treatment and requires an intact MyD88/TLR9 pathway [25
], clearly arguing for a role of endogenous DNA in PL2-3-induced AM14 B-cell activation. Similar to other TLR9-expressing cells, AM14 B cells proliferate vigorously to stimulation with linear CpG ligands (Figure ).
These results suggest that PL2-3 binds nucleosome-associated chromatin released into the cultures from dying cells and that these immune complexes engage the anti-Ig BCR, deliver 'signal 1', and enter the cell [25
]. The dsDNA they contain then reaches TLR9 in the endosome ('autophagosomes' [18
]), otherwise accessible only to single-stranded DNA. Therefore, AM14 B cells represent an excellent model that can be used to contrast data obtained through BCR-dependent and BCR-independent TLR9-mediated signaling and to test some predictions of this hypothesis. As in non-autoreactive B cells, our results show that 'BCR-independent' stimulation with linear CpG-ODN 1826 was differentially inhibited by class R and class B INH-ODNs, with class B 10-fold more potent than class R (Figure ). However, in 'BCR-dependent' immune complex-induced B-cell proliferation, classes B and R INH-ODNs showed similar potencies for inhibition, primarily because of the increased potency of class R INH-ODNs for BCR-mediated activation (Figure ).
While the exact mechanism of class R INH-ODNs action in autoreactive B cells remains to be determined, several possibilities may be considered: (a) competition between TLR9-inhibiting ODNs and endogenous DNA for binding to PL2-3 antibodies/nucleosome/HMGB1 complex, (b) competition for binding to membrane RAGE (receptor for advanced glycation endproducts)/BCR complex, (c) preferential recruitment of TLR9 into early versus late endosomes depending on the nature of the TLR9 ligand used, (d) BCR-dependent increase in passive endocytosis of class R INH-ODNs, and (e) BCR-dependent recruitment of specific helicases (DNA-unzipping enzymes) into an early endosomal compartment. These possibilities are not mutually exclusive. For example, inhibition of BCR signaling might result from both better uptake/trafficking of class R INH-ODNs into endosomes and better recruitment of helicases into TLR9-containing compartments. Helicases must have an important role in the class R INH-ODN-mediated inhibition because recent affinity studies revealed superior binding of single-stranded TLR9 ligands to chip-immobilized TLR9 [3
], an event that may depend heavily on the sugar backbone of the TLR9 ligands used for stimulation [4
Another possibility is a direct competition between endogenous DNA and INH-ODNs for binding to anti-dsDNA antibodies. While this possibility may account for some competition observed in the PL2-3 model, there are no data in the literature to suggest that natural anti-dsDNA antibodies (or anti-chromatin antibodies) preferentially bind DNA segments carrying the canonical inhibitory motif over other DNA sequences [33
The possibility that the double-stranded structure of class R INH-ODNs may direct these ODNs preferentially to an early endosomal compartment in macrophages and DCs and into BCR-related 'autophagosomes' in autoreactive B cells has a precedent in a work by Guiducci and colleagues [8
]. These authors have found that IFN-α secretion from human plasmacytoid DCs induced by multimeric class A(D) CpG-ODNs occurs in transferring receptor-positive endosomes whereas linear monomeric CpG-ODNs (like CpG-1826 used in our experiments) preferentially localize to LAMP-1-positive endosomes and are poor stimulators of IFN-α secretion. However, when linear CpG ligands are complexed into microparticles, they now gain the ability to move to transferrin receptor-positive endosomes, inducing robust IFN-α production. Therefore, the most important determinant of TLR9 signaling may be its endosomal localization [7
] or alternatively CXCL16 engagement as suggested by the Klinman's group [9
]. For one or more of these reasons, class R INH-ODNs have higher potency for BCR-dependent TLR9 activation in autoreactive B cells than for BCR-independent activation of normal B cells.
In this article, we further question the ability of 'TLR9-specific' INH-ODNs to target additional TLR signaling pathways. While our previous data [44
] and data from several other groups [60
] have clearly identified that INH-ODNs at concentrations of up to 1 μM fail to inhibit signaling through the TLR2, 3, and 4 pathways, recent literature has suggested a possibility that PS-ODNs, including INH-ODNs, may also block the TLR7 signaling pathway [28
]. It is believed that, in mice (in contrast to humans), the TLR8 pathway is non-functional since cells from TLR7-/-
mice fail to respond to stimulation with synthetic TLR7/8 ligands [2
]. Here, we show that our INH-ODNs can block TLR7-dependent activation of primary macrophages, macrophage cell lines, RNA immune complex-activated DCs, and AM14 B cells in a dose-dependent but sequence-independent manner (Figure ). However, in spite of these sequence-independent inhibitory effects, palindromic INH-ODNs still have lower potency for resting B cells (Figure ), but classes B and R are equally potent and effective in RNA/anti-RNA-stimulated AM14 B cells (Figure ). Here again, BCR engagement is associated with greater potency of class R INH-ODNs.
Finally, in vivo
, in the MRL-Faslpr/lpr
strain, class R INH-ODNs were better than class B INH-ODNs as inhibitors of both IgG anti-dsDNA and anti-Sm/RNP antibody secretion (Figure ). These results may have significant impact for developing novel DNA-like therapeutics for treating lupus. While TLR7-/-
lupus mice have better survival, they still succumb to autoimmune disease [49
], thus suggesting a possible role for additional DNA/RNA-triggered intracellular signaling pathways (for example, DAI and RIG-I). Interestingly, in this strain, complete lack of TLR9 results in two opposite outcomes: diminished anti-dsDNA secretion but increased anti-Sm/RNP antibody production [49
]. Moreover, the survival of these mice is reduced, likely due to insufficient number/function of regulatory T cells [66
]. One can wonder whether the ability of class R INH-ODNs to block both TLR9 and TLR7 activation in autoreactive B cells may explain the beneficial effect of these INH-ODNs on autoantibody secretion in vivo
. The inferior efficacy of more in vitro
potent class B INH-ODNs in lupus mice came as a surprise. While we lack a logical explanation for this result, recent literature suggests that PS-ODNs (including TLR9-specific INH-ODNs) may redirect TLR7/8 ligand-induced activation away from the TLR7 toward the TLR8 pathway, as observed in HEK cells expressing TLR7 and TLR8 and in primary human peripheral blood mononuclear cells [67
] and mouse cells [68
]. This redirection of TLR activation may result in decreased IFN-α secretion from plasmacytoid DCs but a corresponding increase in IL-12, TNF-α, and IFN-γ secretion resulting from TLR8-expressing cells (for example, human monocytes) [67
Future studies should explore the efficacy of class R INH-ODNs in different strains of lupus mice, including the BXSB male mice. This lupus strain harbors a duplication of the TLR7 gene which appears to be responsible for the phenotype [47
]. Moreover, refinement of the palindromic structure to generate combined selective TLR7/TLR9 inhibitors [37
] together with anti-B-cell-depleting protocols [23
] to re-establish critical B-cell differentiation checkpoints [69
] may result in better treatments for human lupus.