Dimerization of ADAR1 modulates site-specificity of RNA editing

Abstract

Adenosine-to-inosine editing is catalyzed by adenosine deaminases acting on RNA (ADARs) in double-stranded RNA (dsRNA) regions. Although three ADARs exist in mammals, ADAR1 is responsible for the vast majority of the editing events and acts on thousands of sites in the human transcriptome. ADAR1 has been proposed to form a stable homodimer and dimerization is suggested to be important for editing activity. In the absence of a structural basis for the dimerization of ADAR1, and without a way to prevent dimer formation, the effect of dimerization on enzyme activity or site specificity has remained elusive. Here, we report on the structural analysis of the third double-stranded RNA-binding domain of ADAR1 (dsRBD3), which reveals stable dimer formation through a large inter-domain interface. Exploiting these structural insights, we engineered an interface-mutant disrupting ADAR1-dsRBD3 dimerization. Notably, dimerization disruption did not abrogate ADAR1 editing activity but intricately affected editing efficiency at selected sites. This suggests a complex role for dimerization in the selection of editing sites by ADARs, and makes dimerization a potential target for modulating ADAR1 editing activity.

Fig. 1. ADAR1-dsRBD3 forms a stable symmetric dimer.

A Overall organization of ADAR1-dsRBD3 dimer. Monomers A and B are displayed in cartoon mode and coloured in blue and in yellow, respectively. Secondary structure elements are labelled on each monomer. B ADAR1-dsRBD3 forms a symmetric dimer. Top view along the C₂ symmetry axe. C Contacts at the dimer interface between monomer A (in blue) and monomer B (in yellow) on one edge of the β-sheet. Important residues are shown as sticks. Polar contacts are shown as dashed lines. D Contacts at the dimer interface between monomer A (in blue) and monomer B (in yellow) on the other edge of the β-sheet. Important residues are shown as sticks. Polar contacts are shown as dashed lines. Data have been deposited to the PDB (accession code 7ZJ1).

Fig. 2. The ADAR1-dsRBD3 dimer observed in the crystal also exists in solution.

SAXS characterization of ADAR1-dsRBD3 (dsRBD3-short construct, residues 716–797). Upper panel: Surface representation of ADAR1-dsRBD3 monomer (left) or dimer (right) structure determined by crystallography (see Fig. 1). Lower panel: Experimental data are shown as black dots with estimated experimental errors as grey bars. The theoretical scattering fits, as reported by CRYSOL from the crystal structure atomic coordinates of dimeric and monomeric ADAR1-dsRBD3, are shown as red dots. The error-normalized fit residuals are reported below as small black dots. Goodness of fit (χ²) are reported in each case. Data have been deposited to the SASBDB (accession code SASDVH7).

Fig. 3. Dimerization of ADAR1-dsRBD3 is compatible with its binding to dsRNA.

A Overall organization of the asymmetric unit of ADAR1-dsRBD3:dsRNA crystal structure. Monomers A and B are displayed in cartoon mode and coloured in blue and in yellow, respectively. Secondary structure elements are labelled on each monomer. dsRNA helices are displayed in green and in grey, for RNA strands within the asymmetric unit or symmetry-related RNA strands, respectively. B Schematic representation of the RNA self-assembly. The RNA sequence (5’-CGAAGCCUUCGCG-3’) contains two-nucleotides 3’-overhangs used to generate a dsRNA by self-hybridization. The central dsRNA block is shown in green, and the flanking blocks in grey. Arrows indicate the direction of self-assembly that leads to a pseudo-A-form dsRNA helix. C Schematic representation of ADAR1-dsRBD3:dsRNA contacts. Dotted lines indicate contacts between ADAR1-dsRBD3 residues and the RNA. Residues from various dsRBD regions are shown with different colours, with residues from helix αN in blue, from helix α1 in yellow, from the β1-β2 loop in orange, and from the N-terminal tip of helix α2 in red. D–F Detailed views of the three regions of interactions. Polar contacts are displayed as dotted lines. The same colour code is used as in panel C. Data have been deposited to the PDB (accession code 7ZLQ).

Fig. 4. ADAR1-dsRBD3 can mediate ADAR1 dimerization in vivo.

A Co-immunoprecipitation of FLAG-tagged and HA-tagged full length ADAR1 p110 (FLAG-ADAR1, HA-ADAR1). Both protein versions were transfected as indicated by + or – signs. After precipitating the FLAG-tagged version, the precipitate was tested for the presence of the HA-tagged protein in the presence of RNases. Mutations in dsRBD3 that prevent dimer formation (i.e. V747A, D748Q, W768V and C773S – FLAG/HA-ADAR1 mut) still allow dimer formation of full length ADAR1 p110, likely via the deaminase domain. B Quantification of full-length Co-IP/IP ratios of A and Supplementary Fig S11. Three independent blots were used to quantify individual bands of immunoprecipitated proteins. Data height correspond to mean values, and error bars indicate standard deviation (n = 3). C In the absence of the deaminase domain (Δdeaminase), the dimer forming surface on dsRBD3 becomes essential for dimer formation. FLAG-tagged full length ADAR1 (FLAG-ADAR1) interacts with HA-tagged ADAR1 without a deaminase domain (HA-ADAR1 Δdeaminase). However, when the dsRBD3 is mutated to prevent dimer formation (FLAG-ADAR1 mut) the interaction with a deaminase deficient ADAR1 is disrupted (HA-ADAR1 mut Δdeaminase). D Quantification of Co-IP/IP ratios of C and Supplementary Fig. S11. Three independent blots were used to quantify individual bands of immunoprecipitated proteins. Data height correspond to mean values, and error bars indicate standard deviation (n = 3). The ratio of HA-tagged/FLAG-tagged precipitated proteins were quantified using ImageJ (lys: lysate; IP: immunoprecipitation). Blots are probed with anti-FLAG (α-FLAG) and anti-HA (α-HA) antibodies. Source data are provided as a Source Data file.

Fig. 5. dsRBD3 dimer interface mutations affect editing activity of ADAR1 in vivo.

HEK293T were transfected with full length ADAR1 p150 (p150), ADAR1 p110 (p110), ADAR1 p150 dsRBD3 mutant (p150 mut) and ADAR1 p110 dsRBD3 mutant (p110 mut). After transfection, total RNAs were extracted and different targets were amplified from cDNA. Amplicons were sequenced with Sanger sequencing and editing levels were determined by the relative height of G over A peaks. A Azin1 site 1 is exclusively edited by ADAR1 p150 and a dimerization mutation abolishes editing. B Editing at site 2 in the same Azin1 transcript is strongly reduced upon inhibition of dimer formation. C Editing of Gli1 is reduced by the dimerization mutant but only in the context of ADAR1 p150. D, E Editing in the Alu element embedded within the Cflar transcript show that sites 1 and 2 are not affected by the loss of dimerization while editing of sites 3, 4, 5, and 6 is strongly reduced upon loss of dsRBD3 dimerization. PEI serves as a negative control, in which cells were transfected with transfection reagent only. Asterisks (*) denote no measurable editing. Data height correspond to mean values, and error bars indicate standard deviation from 3 independent experiments (n = 3). F Graphical depiction of changes in editing observed in the dimerization mutants for ADAR1 p110 and the corresponding mutant isoform (top row) and ADAR1 p150 and the corresponding dimerization mutant (bottom row) in the editing sites of the 3’-UTR of Nicn1. Sites that are not affected by editing are depicted in green, sites that show less than 50% reduction in editing are highlighted in orange, while sites showing a reduction >50% of editing in the mutant are labelled in red. See also Supplementary Fig. S12. Source data are provided as a Source Data file.