Breaching Self-Tolerance to Alu Duplex RNA Underlies MDA5-Mediated Inflammation

Abstract

Aberrant activation of innate immune receptors can cause a spectrum of immune disorders, such as Aicardi-Goutières syndrome (AGS). One such receptor is MDA5, a viral dsRNA sensor that induces antiviral immune response. Using a newly developed RNase-protection/RNA-seq approach, we demonstrate here that constitutive activation of MDA5 in AGS results from the loss of tolerance to cellular dsRNAs formed by Alu retroelements. While wild-type MDA5 cannot efficiently recognize Alu-dsRNAs because of its limited filament formation on imperfect duplexes, AGS variants of MDA5 display reduced sensitivity to duplex structural irregularities, assembling signaling-competent filaments on Alu-dsRNAs. Moreover, we identified an unexpected role of an RNA-rich cellular environment in suppressing aberrant MDA5 oligomerization, highlighting context dependence of self versus non-self discrimination. Overall, our work demonstrates that the increased efficiency of MDA5 in recognizing dsRNA comes at a cost of self-recognition and implicates a unique role of Alu-dsRNAs as virus-like elements that shape the primate immune system.

Keywords: Aicardi-Goutières syndrome; Alu; IFIH1; MDA5; auto-inflammation; innate immunity; retroelement.

Figure 1. Basal signaling activity of gain-of-function (GOF) variants of MDA5 is mediated by endogenous RNA recognition

(A) IFNβ reporter activity of wild-type (WT) and GOF MDA5 (R337G, D393V and G495R), with and without premature truncation mutations. Right: a western blot of ectopically expressed, FLAG-tagged MDA5. (B) Schematic of MDA5 2CARD fused to zinc finger domain (ZF) of Zif268 and DNA oligos containing tandem repeats of the ZF binding site (ZBS). (C) IFNβ reporter activity of 2CARD-ZF and oligomerization-deficient mutant (m2CARD-ZF, A20K/R21E) in the presence and absence of DNA with 1–4 repeats of ZF-binding sequence (ZBS, GCGTGGGCG) (Jamieson et al., 1996). Right: a western blot of indicated proteins. (D) Effect of co-expression of dsRNA binding proteins, NS1 and E3L, on the IFNβ reporter activity of G495R (a GOF MDA5 variant) and GST-2CARD. The dsRNA binding-deficient mutants, mNS1 and mE3L, were used as controls. Right: a western blot of indicated proteins. All cellular assays were performed in 293T cells. Data are represented as mean ± SD (n=3) for (A), (C) and (D). See also Figure S1.

Figure 2. RNA-rich cellular environment necessitates MDA5 filament assembly on dsRNA for signaling

(A–C) Native gel shift assay (A), representative electron micrographs (B) and ATPase activity (C) of G495R in complex with 512-ssRNA and 512-dsRNA (see Methods for the RNA sequence). 2CARD deletion mutant (Δ2CARD) of MDA5, which is both necessary and sufficient for RNA binding, was used for all EM, native gel shift assays and ATPase assays in this study. In all comparisons, same mass concentrations (2.5, 0.6 and 0.4 ng/µl for native gel, EM and ATPase assays, respectively) of RNAs were used. (D) Schematic of the cell-free IRF3 dimerization assay using S1 extract from 293T cells. Right: native gel showing ³⁵S-IRF3 dimerization in the presence and absence of MDA5, dsRNA and K63-Ub_n. (E and F) IRF3 stimulatory activity of G495R in complex with an increasing concentration (0.5–50 ng/µl) of 512-ssRNA and 512-dsRNA (E), and yeast tRNA and 42-ssRNA (F). (G) A model to explain the observed difference between dsRNA and ssRNA in stimulating the MDA5 signaling activity. On dsRNA (upper panel), cooperative filament formation allows proximity-induced oligomerization of 2CARD regardless of the level of RNA. On ssRNA (lower panel), however, receptor oligomerization occurs only when RNA is present at substoichiometric concentrations (left), but not in excess (right). (H) IRF3 stimulatory activity of G495R in complex with stimulatory RNA (sRNA), 512-ssRNA and 512-dsRNA (both at 0.5 ng/µl), in the presence of an increasing concentration (0.5–50 ng/µl) of non-stimulatory, competitor tRNA (cRNA). See also Figure S2.

Figure 3. Alu:Alu hybrids formed by IR-Alus are the primary ligands for G495R MDA5

(A) Schematic of the RNase A protection assay. Cytosolic RNA (5 ng/µl) from 293T cells was pre-incubated with purified MDA5 G495R (150 nM), treated with RNase A, and recovered for subsequent biochemical and functional analyses (See Methods). (B) Western blot analysis of the 293T cytosolic fraction, from which cytosolic RNA was purified. (C and D) IRF3 dimerization (C), filament formation (D) assays with RNAs recovered from the G495R-protected digestion. CytoRNA-0.0, −0.5 and −2.0 indicate RNAs recovered after digestion with 0.0, 0.5, and 2.0 ng/µl of RNase A, respectively. Same mass concentrations (0.5 ng/µl for IRF3 dimerization and 2.0 ng/µl for EM) of RNAs were used. IRF3 dimerization was measured in the presence of an increasing concentration of competitor tRNA (cRNA, 0–8 ng/µl). (E) RNA-seq followed by RepeatMasker analysis of cytoRNA-0.0 and cytoRNA-2.0. The table below shows averages (standard deviations in parenthesis) of two independent biological repeats. (F) Normalized gene counts of cytoRNA-2.0 plotted against cytoRNA-0.0. (G) Distribution of sequencing reads of cytoRNA-0.0 and cytoRNA-2.0. Two representative genes (BRI3BP and CXorf56) from the top enriched genes are shown. Thin, medium thick and thick lines represent intron, UTR and CDS, respectively, according to the GENCODE v24 annotation. Red arrows represent Alu elements according to the RepeatMasker annotation. Y-axis represents read count. (H) Schematic of Alus in the inverted repeat (IR) configuration. (I) Histograms of the enrichment factors of IR-Alus (gap between Alus < 1 kb) (grey) and other Alus that do not meet the IR-Alu criteria (red). (J) Left: schematic of the in-cellulo RNase A protection assay. G495R Δ2CARD (or empty vector, EV) was ectopically expressed in 293T cells and RNase A was transiently introduced through the pore forming protein, SLO (See Methods). Right: the level of Alu RNA relative to GAPDH after in-cellulo RNase A digestion. Data represent mean ± SD (n=3). See also Figure S3.

Figure 4. Alu:Alu hybrids, unlike other cellular dsRNA, can robustly stimulate MDA5 filament formation and signaling

(A) IRF3 stimulatory activity of G495R in complex with stimulatory RNAs (sRNA, 0.5 ng/µl) that represent five types of cellular dsRNAs. The activity was measured with an increasing concentration of competitor tRNA (cRNA, 0–8 ng/µl). (B and C) ATPase activity (B) and representative electron micrographs (C) of G495R in complex with cellular dsRNAs (0.4 ng/µl for ATPase assay and 0.6 ng/µl for EM) used in (A). Data are represented as mean ± SD (n=3) for (B). (D) Representative electron micrographs of G495R filaments formed on Alu:Alu hybrids from BPNT1 and DESI1 3’UTRs. (E) Alignment scores between adjacent inverted repeats. Pairs of inverted repeats were ranked by the alignment scores (Y-axis), and listed in the descending order (X-axis). Two scoring systems were used for sequence alignment: default from the program Exonerate (top) and modified parameters with higher penalty for gaps (bottom). Red bars represent pairs of inverted repeats that are not Alu elements; there are 11 (top) and 5 (bottom) non-Alu pairs out of ~1500 total inverted repeat pairs. All the remaining pairs (yellow) are IR-Alus. (F) Sequencing chromatograms of Alu:Alu hybrids (NICN1 3’UTR) before and after in vitro A-to-I modification by ADAR1. Yellow highlights indicate the positions of the modification. Note that an A-to-I-modified base is reverse transcribed as G. (G and H) Representative electron micrographs (G) and IRF3 stimulatory activity (H) of G495R in complex with Alu:Alu hybrids (NICN1 3’UTR) before and after A-to-I modification. See also Figure S4.

Figure 5. Paired Alus, not unpaired Alus, stimulate GOF MDA5 and are abundant in cytosol

(A–C) Representative electron micrographs (A), IRF3 stimulatory activity (B), and ATPase activity (C) of G495R in complex with the sense (+) or antisense (−) strand of Alu from the NICN1 3’UTR. (D) Schematic of the RNase H-based method to selectively cleave unpaired, but not paired Alu RNAs. (E) Gel analysis of the RNase H assay. In vitro transcribed Alu RNAs (from NICN1 3’UTR) were subjected to the RNase H assay as described in (D). An oligo targeting GAPDH (αGAPDH) was used for negative controls. (F) Quantitation of Alu:Alu hybrids in cytosolic RNA. The RNase H assay in (D) was performed using purified cytosolic RNA from 293T cells, and remaining Alu(+) and Alu(−) were quantitated relative to the spike-in control. (G) The levels of Alu(+), Alu(−), GAPDH and ACTB (right) relative to the spike-in control before and after the RNase A protection assay. CytoRNA-0.0, −0.5 and −2.0 indicate RNAs recovered after digestion with 0.0, 0.5 and 2.0 ng/µl RNase A in the presence or absence of G495R. Data represent mean ± SD (n=3) for (C), (F) and (G). See also Figure S5.

Figure 6. WT MDA5 is sensitive to dsRNA structural irregularities and is thus inefficient in recognizing an imperfect duplex of Alu:Alu hybrid

(A) Schematic of Alu:Alu hybrids formed by IR-Alus in NICN1 3’UTR. Red and white half arrows indicate sense (+) and antisense (−) Alus, respectively. Below is the sequence alignment of Alu(+) (top strand) and the reverse complement of Alu(−) (bottom strand). Red # and space indicate mismatch and bulge, respectively. (B) Representative electron micrographs of WT and G495R in complex with the naturally occurring Alu:Alu hybrids from NICN1 3’UTR (red:white arrow, top) or with an artificial perfect duplex formed by Alu(+) and its reverse complement (red:red arrow, bottom). (C) ATPase activity of WT and G495R when bound by unpaired or paired Alus from NICN1 3’UTR. Arrows are as defined in (A and B). (D) Representative electron micrographs of WT and G495R in complex with 512 bp dsRNA with or without 6 nt mismatch at the center. (E) RNase I footprinting assay to examine the occupancy of the 6 nt mismatched site by WT or G495R molecules. The RNase I sensitivity was examined with an increasing concentration of MDA5 (top) or RNase I (bottom). The saturating concentration (1 µM) of MDA5 was used in the bottom to compare WT and G495R independent of their differential affinities for dsRNA. (F and G) Representative electron micrographs (F) and ATPase activity (G) of WT and G495R in complex with A-to-I modified 512-dsRNA. Data are mean ± SD (n=3) for (C and H). See also Figure S6.

Figure 7. Unmodified Alu:Alu hybrids activate wild-type MDA5 under the ADAR1-deficiency

(A) The IFNβ reporter activity of WT MDA5 and GST-2CARD in the presence or absence of NS1/mNS1 in ADAR1-WT and -KO cells. (B) RNA-seq followed by RepeatMasker analysis of cytoRNA-0.0 and cytoRNA-2.0, which were generated by WT MDA5-protected digestion of ADAR1-KO cytosolic RNAs. The numbers represent averages (standard deviations in parenthesis) of the two independent biological repeats. (C) Distribution of sequencing reads of cytoRNA-0.0 and cytoRNA-2.0. Two representative genes (BRI3BP and CXorf56) from the top enriched genes are shown as in Figure 3G. (D) The level of Alu RNA relative to GAPDH after in-cellulo RNase A protection assay. WT MDA5Δ2CARD (or empty vector, EV) was ectopically expressed in ADAR1-WT or -KO cells and RNase A was introduced into the cells using SLO pores as in Figure 3J. (E) The level of IFNβ or MDA5 (IFIH) mRNA induction upon treatment with the IFNβ protein in ADAR1-WT and -KO cells. (F) Sequencing chromatograms of representative IR-Alus (in PHAX 3’UTR) from ADAR1-WT and -KO cytosol before and after (24 hr post) IFNβ treatment. (G) IRF3 stimulatory activity of WT MDA5 in complex with increasing concentration of cytosolic RNA (0.5–15 ng/µl) from ADAR1-WT and -KO cells before and after IFNβ treatment. (H) A model of how Alu:Alu hybrids activate GOF MDA5 in ADAR1-sufficient cells (left) and WT MDA5 in ADAR1-deficient cells (right). In ADAR1-sufficient cells, WT MDA5 does not recognize cellular RNAs due to its sensitivity to structural irregularities caused by mismatches, bulges and A-to-I modifications. GOF mutations, however, make MDA5 less sensitive to such dsRNA structural irregularities, allowing recognition of imperfect duplexes such as Alu:Alu hybrids. In ADAR1-deficient cells, on the other hand, the lack of A-to-I modification increases the structural integrity of Alu:Alu hybrids, allowing aberrant recognition by WT MDA5. Data represent mean ± SD (n=3) for (A), (D) and (E). See also Figure S7.