Sequence-dependent off-target inhibition of TLR7/8 sensing by synthetic microRNA inhibitors

Abstract

Anti-microRNA (miRNA) oligonucleotides (AMOs) with 2'-O-Methyl (2'OMe) residues are commonly used to study miRNA function and can achieve high potency, with low cytotoxicity. Not withstanding this, we demonstrate the sequence-dependent capacity of 2'OMe AMOs to inhibit Toll-like receptor (TLR) 7 and 8 sensing of immunostimulatory RNA, independent of their miRNA-targeting function. Through a screen of 29 AMOs targeting common miRNAs, we found a subset of sequences highly inhibitory to TLR7 sensing in mouse macrophages. Interspecies conservation of this inhibitory activity was confirmed on TLR7/8 activity in human peripheral blood mononuclear cells. Significantly, we identified a core motif governing the inhibitory activity of these AMOs, which is present in more than 50 AMOs targeted to human miRNAs in miRBaseV20. DNA/locked nucleic acids (LNA) AMOs synthesized with a phosphorothioate backbone also inhibited TLR7 sensing in a sequence-dependent manner, demonstrating that the off-target effects of AMOs are not restricted to 2'OMe modification. Taken together, our work establishes the potential for off-target effects of AMOs on TLR7/8 function, which should be taken into account in their therapeutic development and in vivo application.

Figure 1.

Off-target effect of 2′OMe AMOs on immunostimulatory RNA sensing. (A) BMMs from WT mice or miR-17∼92^flox/flox × LysMCre mice pre-treated for 45 min with 40 nM of the indicated 2′OMe AMO or non-2′OMe control RNA (RD) were stimulated overnight with 180 nM of immunostimulatory ssRNA (B-406AS-1), and TNF-α levels were measured in supernatants by ELISA. Percentages of TNF-α production compared to the RD+B-406AS-1 condition are given. The data are averaged from three independent experiments, where each treatment was conducted in biological triplicate. R848 was used as a positive control for induction of TNF-α. (B) Total RNA was extracted from non-stimulated BMMs, and miR-17-5p, miR-19a-3p and miR-92a-3p levels were measured by RT-qPCR. Expression was normalized to U6 RNA and is shown relative to WT mice. Data are averaged from three independent experiments in biological duplicate. (C) Human PBMCs were pre-treated for 45 min with the indicated dose of AMO or control RNA (RD), and stimulated overnight with 180 nM of B-406AS-1, and IFN-α levels were measured in supernatants by ELISA. The data are averaged from three independent experiments in different blood donors, where each treatment was conducted in biological triplicate. Percentages of cytokine production compared to the 0.6 nM RD+B-406AS-1 condition are given. (A and C) Ordinary two-way ANOVA with Dunnett's multiple comparison tests to the RD+B-406AS-1 condition are shown. Unless otherwise indicated, differences were not significant. (A–C) SEM is shown.

Figure 2.

Inhibition of immunostimulatory RNA sensing by 2′OMe AMOs is sequence dependent. (A) Total RNA from BMMs on day 8 post-isolation was analysed by TaqMan^® Array Rodent MicroRNAs in biological duplicate, and miRNA abundance relative to that of the most abundant miRNA—miR-222—calculated as detailed in the Materials and Methods section. ND, not detected. BMMs from WT mice pre-treated with 40 nM of the indicated 2′OMe AMO (B) or LNA/DNA AMO (C) for 45 min were stimulated overnight with 180 nM of immunostimulatory ssRNA (B-406AS-1), and TNF-α levels were measured in supernatants by ELISA. The RD+B-406AS-1 condition was used as control. The data are averaged from two independent experiments in biological triplicate. (B and C) Ordinary one-way ANOVA with Dunnett's multiple comparison tests to the RD+B-406AS-1 condition are shown. Unless otherwise indicated, differences were not significant. Class 1: AMOs with no inhibitory effect (black bars). Class 2: AMOs with inhibitory effect with 2′OMe chemistry only (light grey bars). Class 3: AMOs with inhibitory effect with LNA/DNA chemistry only (dark grey bars). Class 4: AMOs with inhibitory effect with both chemistries (white bars). (A–C) SEM is shown.

Figure 3.

Sequence-dependent inhibition of RNA sensing by 2′OMe AMOs is conserved in human PBMCs. Human PBMCs pre-treated with 40 nM of the indicated 2′OMe AMO (A) or LNA/DNA AMO (B) for 45 min were stimulated overnight with 180 nM of B-406AS-1. TNF-α levels were measured in supernatants by ELISA. Correlation of relative TNF-α production to RD control condition (i.e. inhibitory index) in both human PBMCs and mouse BMMs is shown for each AMO chemistry (right panels). 2′OMe AMOs of ‘Class 2’ are shown in grey. Correlation for 2′OMe AMOs was highly significant (P < 0.0001). (C and D) Human PBMCs were pre-treated for 45 min with the indicated dose of AMO or RD, and stimulated overnight with 180 nM of B-406AS-1. (C and D) TNF-α and IFN-α levels were measured in supernatants by ELISA. (A–D) The data are averaged from a minimum of three independent experiments in different blood donors, in biological triplicate. Percentages of cytokine production compared to the RD+B-406AS-1 condition (A and B) or the 0.6 nM RD+B-406AS-1 condition (C and D) are given. Ordinary one-way (A and B) or two-way (C and D) ANOVA with Dunnett's multiple comparison tests to the RD+B-406AS-1 (A and B), the miR-224-5p+B-406AS-1 (C) or the miR-195-5p+B-406AS-1 (D) conditions are shown. (A–D) Unless otherwise indicated, differences were not significant. SEM is shown.

Figure 4.

Inhibition of RNA sensing by 2′OMe is motif dependent. (A) BMMs from WT mice pre-treated with the indicated dose of AMO or RD for 45 min were stimulated overnight with 180 nM of B-406AS-1, and TNF-α levels were measured in supernatants by ELISA. The data are averaged from three independent experiments in biological triplicate. Percentages of cytokine production compared to the 0.6 nM RD+B-406AS-1 condition are given. Ordinary two-way ANOVA with Dunnett's multiple comparison tests to miR-200c-3p+B-406AS-1 are shown. (B) Human PBMCs were pre-treated for 45 min with the indicated dose of AMO or RD, and stimulated overnight with 180 nM of B-406AS-1. TNF-α levels were measured in supernatants by ELISA. The data are averaged from three independent experiments in different blood donors, in biological triplicate. Unpaired t-tests comparing miR-200a-3p+B-406AS-1 and indicated conditions are shown. (C) Top: alignment of miR-200a/b/c-3p AMOs with the core inhibitory motif highlighted in grey. Bottom: MEME pictogram of the relative frequency of bases constituting the inhibitory motif (40). (D) Alignment of miR-200a-3p, miR-200c-3p, miR-122-5p and NC1 AMO variants with altered/truncated inhibitory motifs (in 5′-3′ orientation). Native inhibitory motifs are underlined and point mutations are in red bold. BMMs from WT mice (E, F, G, I) or human PBMCs (F, H) pre-treated with 40 nM of the indicated 2′OMe AMO or RD were stimulated overnight with 180 nM of B-406AS-1, and TNF-α levels were measured in supernatants by ELISA. Percentages of cytokine production compared to the RD+B-406AS-1 condition are given. The data are averaged from a minimum of two independent experiments, in biological triplicate. Ordinary one-way ANOVA with Dunnett's multiple comparison tests to miR-200a-3p+B-406AS-1 (E) or NC1+B-406AS-1 (H and I) are shown. Unpaired t-tests comparing miR-200a-3p+B-406AS-1 (F) or RD+B-406AS-1 (G) and indicated conditions are shown (ns, not significant). (A, E, H, I) Unless otherwise indicated, differences were not significant. (A, B and E–I) SEM is shown.

Figure 5.

Motif-dependent activity of 2′OMe AMO is TLR7 specific. (A) mFOLD-predicted structure and free energy (ΔG, in kcal/mol) of miR-224–5p 2′OMe AMO used at 37°C (31). The bases of the consensus inhibitory motif are shown in red. Arrows point to mutations introduced in miR-224–5p AMOs to create the inhibitory motif in miR-224–5p mut2. BMMs from WT mice (B) and human PBMCs (C) pre-treated with 40 nM of the indicated 2′OMe AMO or RD, were stimulated overnight with 180 nM of B-406AS-1 (B and C) or β-Gal-656-REV (B), and TNF-α levels were measured in supernatants by ELISA. (B) The data are averaged from two (for β-Gal-656-REV) or three (for B-406AS-1) independent experiments, in biological triplicate. Percentages of cytokine production compared to the RD+B-406AS-1 (black) or β-Gal-656-REV condition (white) are given. (C) The data are averaged from three independent experiments in different blood donors, in biological triplicate. (D) HEK293 cells stably expressing TLR7 or TLR9 and an NF-κB-luciferase reporter, were transfected with 200 nM of the indicated AMO for 45 min prior to stimulation with R848 (200 ng/ml) (white) or polyI:C (20 μg/ml) (grey) for TLR7 cells, and ODN 2006 (400 nM) (black) for TLR9 cells. Firefly luciferase was assayed after overnight incubation as described in the Materials and Methods section. Luminescence reads normalized to unstimulated cells (NT) are shown as percentages of the RD+TLR ligand condition. The data are averaged from a minimum of three independent experiments in biological triplicate. (B–D) Ordinary one-way ANOVA with Dunnett's multiple comparison tests to the miR-224–5p+B-406AS-1 condition are shown. Unless otherwise indicated, differences were not significant. SEM is shown.