Differential Structural Features of Two Mutant ADAR1p150 Zα Domains Associated with Aicardi-Goutières Syndrome

Abstract

The Zα domain of ADARp150 is critical for proper Z-RNA substrate binding and is a key factor in the type-I interferon response pathway. Two point-mutations in this domain (N173S and P193A), which cause neurodegenerative disorders, are linked to decreased A-to-I editing in disease models. To understand this phenomenon at the molecular level, we biophysically and structurally characterized these two mutated domains, revealing that they bind Z-RNA with a decreased affinity. Less efficient binding to Z-RNA can be explained by structural changes in beta-wing, part of the Z-RNA-protein interface, and alteration of conformational dynamics of the proteins.

Graphical Abstract

A well-functioning immune system can distinguish non-self from self RNA. This task is in part carried out by the adenosine deaminase acting on RNA (ADAR1), which catalyzes the conversion of some adenosines in self RNA to inosines^1–3. In humans, this ‘A-to-I’ editing is augmented upon infection, primarily through the interferon-induced longer isoform of ADAR1 (p150)^1,2. p150 contains a ~65-amino acid long Z-RNA binding domain (Zα) at its N-terminus (Fig. 1A). Zα enhances substrate specificity and enzymatic activity in vitro³, and acts synergistically with the downstream Zβ domain^4,5. The mechanism that enables ADAR1p150 to achieve this level of specificity in vivo remains unknown.

Figure 1: Both AGS mutants of Zα have decreased binding affinity for substrate dsRNA and fail to convert it to the Z-form.

A) Domain schematic of human ADAR1p150 showing known AGS associated mutations⁷ B) Circular dichroism spectra of a (CpG)₃ dimer in the presence and absence of wild type and mutant Zα. NaClO₄ is used to have a reference Z-form. C) Binding isotherms derived from ITC of (CpG)₃ dimer and wild type or mutant Zα and extracted biophysical values.

Inheritable mutations within Zα highlight its importance for self-RNA editing. In particular, the point mutants Asn173Ser (Zα_N173S) and Pro193Ala (Zα_P193A) decrease editing levels, causing autoimmune diseases^6–8. Both mutations are frequently found in patients suffering from Aicardi-Goutières syndrome (AGS) and Bilateral Striatal Necrosis/Dystonia (BSN)^7,9,10. AGS is a Mendelian genetic disorder primarily affecting the nervous system in children, leading to continual activation of the innate immunity response in the absence of a viral infection^1,7,11. BSN is an early-onset disease characterized by developmental regression, dystonia, and cerebral calcification⁸. Currently, therapies to treat AGS and BSN are of limited efficacy^11,12, although small-molecule modulators could be therapeutically useful¹³.

The P193A ADAR1p150 mutant results in decreased A-to-I editing as measured through transcriptome sequencing⁶, likely due to decreased affinity for substrate dsRNA. Similar studies have not been undertaken for the Zα_N173S mutant, though the effects are likely analogous⁹. Because N173 and P193 in Zα are involved in the recognition of Z-form nucleic acids ^14,15, we sought to investigate the effect of the AGS- and BSN-related N173S and P193A mutations on the three-dimensional structure of the Zα domain and its ability to bind RNA, the former a critical binding residue while the latter is involved in binding as well as the conversion to Z-form¹⁶. Therefore, we aimed to understand structurally if this loss of binding was simply a result of the loss of key binding residues or if there are unforeseen consequences of these mutations.

Previous in vivo analyses of AGS-causing mutants ADAR1 isoforms suggest many of the identified mutations within the binding domains result in a decrease or loss of binding affinity to substrate dsRNA¹⁷, though many of these mutants lack robust biophysical or biochemical characterization. However, limited biophysical data has shown a loss or decrease in binding as measured by CD or NMR of the Zα double mutant N173A, Y177A¹⁸ as well as K169A, K170A, and R174A¹⁹ mutants. Because studies of N173A (not the clinically observed mutation) and P193A suggest a perturbation to Z-form RNA binding as the means by which these two mutants result in disease, we sought to biophysically characterize dsRNA binding to Zα_N173S and Zα_P193A. Circular dichroism (CD) provides a spectral evaluation of A-form and Z-form character of RNA. We used this method to evaluate the ability of Zα to induce the conversion from A-form unbound RNA to bound Z-form. Using CD, we are also able to discriminate the ability of the mutant Zα domains to bind an idealized substrate dsRNA. By measuring the CD spectra of an RNA hexamer (CpG)₃ dimer in the presence of Zα_WT or the mutants, we determined that Zα_WT can induce Z-form, as noted by a decrease in the peak at 266 nm, the presence of a peak at 285 nm, and the 295 nm peak changing from negative to positive. However, neither of the mutants were able to induce Z-form in the RNA, retaining their characteristic A-form peaks (Fig. 1B).

As an orthogonal method to quantitatively probe the mutant Zα domains’ dsRNA binding, we used isothermal titration calorimetry (ITC) to determine binding affinities for Zα_WT and the two AGS mutants. Zα_WT, when challenged with a (CpG)₃ homodimer, yielded a K_d of 240 nM with a stoichiometry of 0.4 (Fig. 1C), indicating the binding of two Zα proteins per (CpG)₃ dimer (as observed in the crystal structure¹⁴). Conversely, both mutants demonstrated increased K_d values, thus decreased substrate affinity, 3.1 μM for Zα_N173S and 2.4 μM for Zα_P193A with stoichiometries of 0.7 and 0.9 respectively (Fig. 1C). This observed decrease in substrate affinity is consistent with the CD measurements and potentially explains the decreased A-to-I editing observed in Zα_P193A of ADAR1p150⁶, as well as suggesting a decrease in editing would also be observed for Zα_N173S. Because the mutations causing this decreased affinity occur outside of the catalytic deaminase domain, we speculate that while the k_cat may be unchanged, the weaker binding affinity mandates a faster k_off, slower k_on, or a combination. The result of this would be relatively elevated K_M and decreased catalytic efficiency, though further experiments are needed to confirm this. Interestingly, while both Zα_WT and Zα_N173S demonstrated negative −TΔS terms, the Zα_P193A mutant yielded a positive term. This may reflect an increase of microstates Zα_P193A may occupy, though additional work is required to confirm this.

We then purified human ¹⁵N/¹³C-labeled proteins from Escherichia coli and characterized them by NMR. Chemical shift perturbations (CSP) derived from the Zα mutants in reference to the wild type spectrum reveal two different patterns of perturbations. While chemical shifts within Zα_N173S are distributed throughout the domain, for Zα_P193A they localize to the mutation-containing binding loop and the N-terminal region of helix 2 (Fig. 2A, S1, S2).

Figure 2: The structures of two Aicardi-Goutières Syndrome mutants of the ADARp150 Zα domain.

A) Chemical shift perturbations (CSP) of both AGS mutants of the Zα domain in reference to the wild type mapped onto the crystal structure of Zα (PDB 1QGP²⁹). The site of the mutation is colored in green and the CSP average is denoted with a dashed line. B) Superposition of the twenty lowest energy structures of Zα_N173S and the crystal structure of Zα. C) Superposition of the twenty lowest energy structures of Zα_P193A and the crystal structure of Zα.

To better assess local single residue and global alterations of protein dynamics for the Zα_N173S and Zα_P193A, we measured NMR relaxation experiments sensitive to different time scales. Both local and global dynamics can be assessed by various NMR relaxation experiments at various timescales. Picosecond to nanosecond dynamics can be probed by R₁ and cross-correlation relaxation (CCR) experiments, while μs to ms dynamics, often indicative of domain motion, are probed by R_1ρ relaxation experiments. The average ¹⁵N-R₁ relaxation rate was essentially unaltered between the two Zα mutants (1.7 s⁻¹ for Zα_N173S; 1.6 s⁻¹ for Zα_P193A), compared to Zα_WT (1.5 s⁻¹; Fig. S3). ¹⁵N CSA/¹H-¹⁵N dipole CCR is also similar for the two mutants (4.8 s⁻¹ and 3.8 s⁻¹ for Zα_N173S and Zα_P193A, respectively; Fig. S4). Because both R₁ and CCR probe molecular tumbling on the fast timescale (ps to ns), these data imply little change to molecular tumbling. However, average R_1ρ relaxation rates that are also sensitive to μs to low ms dynamics, were faster than expected for Zα_N173S (11.4 s⁻¹) than for Zα_P193A (7.8 s⁻¹), which is similar to Zα_WT (7.6 s⁻¹; Fig. S4, S5). Based on the slightly elevated R₁ and CCR values for Zα_N173S, only half of the difference in the R_1ρ relaxation rates can be explained by dynamics faster than ~10 ns. In support of these findings, analytical ultracentrifugation and size exclusion experiments yielded results consistent with a monomeric state, thus excluding the possibility of substantial dimerization (data not shown). Comparing the dynamic trends for residue 173 shows similar ¹⁵N-R₁ values (1.6 for N173; 1.7 for S173; Fig. S3) while the ¹⁵N-R1ρ value was elevated for S173 (9.2 for N173; 14.0 for S173; Fig. S5). This disparity between the Zα_WT domain and Zα_N173S is a reflection of altered global dynamics of Zα_N173S, arising from a single solvent facing residue. Overall, our findings suggest that Zα_P193A has similar dynamics to Zα_WT, while Zα_N173S has altered domain dynamics.

To gain further insight, we calculated the complete structure of the mutant Zα domains using exact NOEs (eNOE) derived from NOESY spectra (Fig. 2B, C, S6). This recently described method^20,21 makes use of NOESY spectra with increasingly longer mixing times to extract high precision proton pair distances to yield precise structures. In accordance with the CSP and relaxation data, Zα_P193A demonstrates overall minimal deviation from Zα_WT (heavy atom RMSD = 1.1 Å). Similarly, Zα_N173S overlays well with Zα_WT (heavy atom RMSD = 1.3 Å) as compared to the structure of the Zα_N173A,Y197A double mutant¹⁸ relative to Zα_WT (heavy atom RMSD = 0.6 Å).

In Zα_P193A, the binding loop containing the point mutation undergoes a structural displacement of 5.8 Å away at its apex from its position in Zα_WT (Fig. 2C), similar to MD simulations of this mutant²². An overlay with dsRNA-bound Zα_WT reveals that this loop reorganization displaces several nucleic acid binding residues from the RNA binding interface which normally make key hydrogen bonding and Van der Waals contacts with Z-RNA and Z-DNA, specifically; Trp191, Pro192, Pro193, and Trp195. Similarly, the structural disruption of the beta-wing likely compounds the mutant’s failure to properly bind substrate. This disruption of the beta-wing also likely impacts proper Z-form conversion as this motif has been shown to be important for this role in other Zα domains¹⁶.

Although Zα_N173S overlays with the Zα_WT crystal structure, one feature of note is a minor perturbation in the beta-wing which shows small structural perturbations from the known crystal structure (backbone RMSD 1.9 Å) (Fig. 2B, S7). This is interesting, as the structure ensemble for the Zα_N173A,Y197A double mutant shows similar displacements¹⁸, suggesting that mutations within helix 3 may affect beta wing dynamics. However, the loops may also adopt different conformations because the structures are not in complex with Z-form nucleic acids, as reported before for Zα of PKZ²³. Quite possibly, differences in loop conformations may be due to the more sparce eNOEs observed in this region of the spectra (Fig. S6). In any case, our analysis reveals that a loss of contacts at position 173 is not the sole reason why Zα_N173S shows decreased binding to RNA. Replacing the asparagine at position 173 with a serine leads to a global perturbation of the domain dynamics while the global structure remains largely unaltered. The observed effect is thus different from that seen in Zα_P193A, in which the change of amino acid at position 193 leads to disruption of protein-RNA contacts through the reorganization of the beta-wing motif rather than altered dynamics.

The data presented here provide structural insight into how these AGS mutants potentially result in diseases which have remained poorly understood. In the case of Zα_N173S, the globally altered dynamics we observe may explain the disease phenotype observed resulting from this mutation. This result expands upon the previous postulate that the loss of function is solely the result of the disruption of a single binding residue. Additionally, our findings reinforce the idea that in some cases the observed disfunction of protein resulting from a point mutation cannot be explained by the loss of an important hydrogen bond or ionic interaction, but rather there may be unforeseen consequences of these mutations, as is seen here where structure predictions of the Zα_P193A domain fail to recapitulate the local structural perturbation (Fig. S8). Our findings serve as a reminder of the work of Brian Matthews, who was among the first to reveal how point mutations within proteins lead to conformation variability²⁴. Systematically studying the structures of proteins with point mutations causing disorders provides a more robust molecular understanding of the associated dysfunctions, while ultimately also helping to improve our tools for predicting the effect of mutations on protein structures²⁵.

METHODS

Expression of unlabeled human ADAR1p150’s Zα domain:

The Zα domain of hADAR1p150 (residues 140–202) (UniProt ID: P55265) cloned in the pET28a(+) plasmid (N-terminal 6x His-tag and thrombin cleavage site between His-tag and the Zα sequence) was a gift from Drs. Peter Dröge and Alekos Athanasiadis. The Zα_N173S and Zα_P193A mutants were synthesized and cloned into pET28a(+) vectors by Genscript and confirmed through sequencing. Different Zα constructs were expressed and purified similarly as described in^4,26. The proteins were recombinantly expressed in BL21(DE3) E.coli cells. Cells were grown in LB to an OD₆₀₀ of 0.6 and then induced with 0.5 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) overnight at 18°C. After pelleted cells were resuspended in lysis buffer containing 50 mM Tris-HCl (pH 8.0), 300 mM NaCl, 10 mM imidazole, and 5 mM β-mercaptoethanol (BME), they were chemically lysed by deoxycholic acid at 2 mg mL⁻¹ for 30 minutes on ice. Cell lysate was then sonicated for 15 rounds of 15 seconds on, 30 seconds off at 50 W on ice. Cell lysate was clarified by centrifugation at 30,000×g for 30 minutes. The soluble fraction was purified by nickel affinity chromatography (Histrap column) using a wash buffer containing 1 M NaCl, 50 mM Tris (pH 8.0), 10 mM imidazole, and 5 mM β-mercaptoethanol, followed by elution in 300 mM NaCl, 50 mM Tris (pH 8.0), 500 mM imidazole, and 1 mM BME. To further purify the proteins, size exclusion chromatography was performed using a Sepax 300 SEC column (GE Life Sciences) in 100 mM NaCl and 50 mM sodium phosphate (pH 6.4). Protein stocks were stored at −80°C.

Expression of ¹³C and ¹⁵N labeled human ADAR1p150’s Zα domain:

Proteins were prepared as described above with the following modifications to the protocol. Cell growth was carried out in M9 minimal media containing the following components: 100 mL of 10x M9 salts (60g/L Na₂HPO₄, 30g/L KH₂PO₄, 5 g/L NaCl, pH 7.4), 10 mL of 100g/L ¹⁵NH₄Cl (pH 7.4), 2 mL 1 M MgSO₄, 12.5 mL 20% (w/v) ¹³C-glucose, 0.2 mL 0.5 M CaCl₂, 1 mg biotin, 0.5 mL 2 mg/mL thiamine hydrochloride, 1 mL 15 mg/mL FeCl₂ in 1M HCl, 1 mL 15 mg/mL ZnCl₂, 2 mL 10% (w/v) yeast extract. The media was brought to 1 L with autoclaved milli-Q filtered water and then passed through a 0.22 μM filter. Cells were grown to an OD₆₀₀ of 0.4 and induced with 0.5 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) for 3 hours at 37°C.

Nucleic acid constructs:

(CpG)₃ RNA oligo constructs were purchased from Dharmacon.

RNA circular dichroism:

All CD experiments were performed at 100 μM duplex oligo (0.38 mg/mL) and 600 μM protein (5.4 mg/mL) in a buffer containing 25 mM NaCl and 20 mM sodium phosphate (pH 6.4) unless otherwise indicated. Oligoribonucleotides were briefly heated to 90°C and cooled to room temperature. Spectra were acquired in a JASCO J-815 CD spectrometer using a 1 mm quartz cuvette. Spectra were an average of two scans measured from 320 to 220 nm with a 1 nm step using a scanning speed of 50 nm/min and a digital integration time of 4 seconds. Then protein was added to the indicated concentration and incubated at room temperature for 10 minutes.

Isothermal titration calorimetry:

Nucleic acid oligoribonucleotides and protein for ITC were dialyzed overnight at 4°C in the same beaker against a buffer containing 25 mM NaCl and 20 mM sodium phosphate (pH 6.4). Binding heat was measured using a Malvern ITC200 calorimeter at 25°C and mixing speed of 750 RPM, with 180 s injection delays and a reference power of 10 μcals⁻¹. For Zα_WT, the titration was measured with twenty 2 μL consecutive injections of 500 μM r(CpG)₃ RNA into 50 μM protein. Titrations with Zα_N173S and Zα_P193A mutants were measured with twenty 2 μL consecutive injections of 500 μM RNA into 50 μM protein followed by an additional twenty 2 μL consecutive injections of 500 μM RNA into the cell and concatenation of the two datasets using MicroCal Concat ITC version 1 (Malvern). All ITC thermograms were analyzed and fit using Microcal Analysis version 7 SR4 (Origin).

AlphaFold2 structure prediction:

The sequence of either mutant domain was submitted to the ColabFold²⁷ server running AlphaFold2 for structure prediction. The primary sequence of each mutant domain was input and run using MMseqs2 and AlphaFold2-ptm mode. The top 5 confidence models were returned and the highest scoring model was used for comparison with the experimentally recovered structures.

NMR Spectroscopy:

All NMR experiments were carried out on a Varian 900 MHz (run using VNMRJ version 4.2 Revision A (Agilent)) equipped with a 5 mm triple resonance ¹H/¹³C/¹⁵N cold probe with a Z-axis gradient and a Bruker 600 MHz spectrometers (run using TopSpin version 7 (Bruker)) equipped with a 5/3 mm triple resonance ¹H/¹³C/¹⁵N/¹⁹F cryoprobe (CP2.1 TCI) in 100 mM NaCl and 20 mM potassium phosphate (pH 6.4) with 5% D₂O. For all non-uniformly sampled (NUS) experiments, schedules were generated using Poisson-Gap sampling from Gerhard Wagner’s lab website: http://gwagner.med.harvard.edu/intranet/hmsIST/gensched_new.html²⁸.

Wild-type Zα:

The NMR resonance assignment and structure calculation of wild-type Zα have been carried out previously²⁹ and backbone chemical shifts can be found under BMRB accession code 50714³⁰. The ¹⁵N-HSQC spectrum of wild-type Zα was collected on the Varian 900 MHz spectrometer with 1048 (¹H) × 120 (¹⁵N) complex points, a 2 s recycle delay, 8 scans, and spectral widths of 16 and 35 ppm for the ¹H and ¹⁵N dimensions, respectively. The ¹⁵N CSA/¹⁵N-¹H dipole-dipole cross-correlated relaxation (CCR) experiment³¹ was run on the Varian 900 MHz spectrometer with 1024 (¹H) × 96 (¹⁵N) complex points, a 1.7 s recycle delay, 20 scans, and spectral widths of 15.6 and 35 ppm for the ¹H and ¹⁵N dimensions, respectively. The effective periods during which CCR was active were 0, 20, 40, 60, 80, 100, 120, and 150 ms. The ¹⁵N R₁ relaxation experiment was collected on the 900 MHz Varian spectrometer and run with 1048 (¹H) × 80 (¹⁵N) complex points, a recycle delay of 2 s, 8 scans, and spectral widths of 16 and 35 ppm for the ¹H and ¹⁵N dimensions, respectively. The relaxation delays were 0, 100, 200, 300, 400, 500, 600, 700, 800, and 900 ms. The ¹⁵N R_1ρ relaxation experiment was run with 1048 (¹H) × 80 (¹⁵N) complex points, a recycle delay of 2 s, 8 scans, and spectral widths of 16 and 35 ppm for the ¹H and ¹⁵N dimensions, respectively. The relaxation delays under a spin-locking field strength of 1500 Hz were 0, 20, 40, 60, 80, 100, 120, 140, 160, and 180 ms.

Zα_N173S mutant.

The ¹⁵N-HSQC spectrum of the Zα_N173S mutant was collected on the Bruker 600 MHz spectrometer with 1024 (¹H) × 160 (¹⁵N) complex points, a 1.6 s recycle delay, 32 scans, and spectral widths were 16 and 35 ppm for the ¹H and ¹⁵N dimensions, respectively. The constant time ¹³C-HSQC spectrum was collected on the Bruker 600 MHz spectrometer with 1024 (¹H) × 128 (¹³C) complex points, a 1.6 s recycle delay, 32 scans, and spectral widths of 16 and 80 ppm for the ¹H and ¹³C dimensions, respectively. The [¹⁵N,¹³C]-HNCACB was collected on the Bruker 600 MHz spectrometer with 1024 (¹H) × 40 (¹⁵N) × 64 (¹³C) complex points (1268 of the total 2560 indirect points were collected following a 50% NUS sampling scheme), a 1 s recycle delay, 16 scans, and spectral widths of 13.6 (¹H), 35 (¹⁵N), and 80 (¹³C) ppm. The ¹⁵N-HBHANH experiment was collected on the Bruker 600 MHz spectrometer with 1024 (¹H) × 50 (¹⁵N) × 64 (¹H) complex points (1275 of the total 3200 points were collected following a 40% NUS sampling scheme), a 1 s recycle delay, 16 scans, and spectral widths of 16 (¹H), 35 (¹⁵N), and 16 (¹H) ppm. The [¹³C]-HCCH TOCSY was collected on the Bruker 600 MHz spectrometer with 1024 (¹H) × 40 (¹³C) × 120 (¹H) complex points (1905 of the total 4800 points were collected following a 40% NUS sampling scheme), a 1 s recycle delay, 16 scans, and spectral widths of 13.6 (¹H), 80 (¹³C), and 13.6 (¹H) ppm. A uniformly-sampled 3D simultaneously [¹⁵N,¹³C]-resolved [¹H,¹H,X^13C,15N]-NOESY²⁰ was collected on the Varian 900 MHz spectrometer with 1024 (¹H) × 160 (¹H) × 50 (X^13C,15N) complex points, a recycle delay of 1.2 s, 4 scans, spectral widths of spectral widths 15.6 (¹H), 15.6 (¹H), 34/30 ppm (¹⁵N/¹³C), and a NOESY mixing time of 60 ms. The NUS NOESY buildup series with NOESY mixing times of 20, 30, 40, 50, and 60 ms was collected in the same manner as the uniformly-sampled case but following a 50% NUS sampling scheme (4010 out of the total 8000 points were collected)²¹. The ¹⁵N CSA/¹⁵N-¹H dipole-dipole cross-correlated relaxation (CCR) experiment³¹ was run on the Varian 900 MHz spectrometer with 1024 (¹H) × 96 (¹⁵N) complex points, a 1.7 s recycle delay, 20 scans, and spectral widths of 15.6 and 35 ppm for the ¹H and ¹⁵N dimensions, respectively. The effective periods during which CCR was active were 0, 20, 40, 60, 80, 100, 120, and 150 ms. The ¹⁵N R₁ relaxation experiment was run on the Varian 900 MHz spectrometer with 1048 (¹H) × 32 (¹⁵N) complex points, a recycle delay of 2 s, 8 scans, spectral widths of 16 and 35 ppm for the ¹H and ¹⁵N dimensions, respectively, and relaxation delays of 0, 100, 200, 300, 400, 500, 600, 700, 800, and 900 ms. The ¹⁵N R_1ρ relaxation experiment was run on the Varian 900 MHz spectrometer with 1048 (¹H) × 32 (¹⁵N) complex points, a recycle delay of 2 s, 8 scans, spectral widths of 16 and 35 ppm for the ¹H and ¹⁵N dimensions, respectively, and relaxation delays of 0, 20, 40, 60, 80, 100, 120, 140, 160, and 180 ms under a spin-locking field strength of 1500 Hz. The [¹H-¹⁵N]-heteronuclear NOE enhancement experiment was run on the Varian 900 MHz spectrometer with 1024 (¹H) × 80 (¹⁵N) complex points, a 1 s recycle delay, 20 scans, and spectral widths of 15.6 and 33 ppm for the ¹H and ¹⁵N dimensions, respectively. Spectra in the presence and absence of ¹H saturation were recorded in an interleaved manner.

Zα_P193A mutant.

All experiment run on the Zα_N173S mutant were repeated for the Zα_P193A mutant. Slightly different parameters were used for the following experiments: The ¹⁵N-HSQC spectrum was collected with 1024 (¹H) × 94 (¹⁵N) complex points. The [¹⁵N,¹³C]-HNCACB was collected with 1024 (¹H) × 40 (¹⁵N) × 44 (¹³C) complex points (937 of the total 1760 points were collected following a 50% NUS sampling scheme). The [¹H-¹⁵N]-heteronuclear NOE enhancement experiment was run with 48 scans, and spectral widths of 16 and 33 ppm for the ¹H and ¹⁵N dimensions, respectively.

Data processing.

All spectra were processed with the NMRPipe/NMRDraw/NlinLS package³². The time-domain data were multiplied with a squared cosine function in the direct dimension and cosine functions in the indirect dimensions and the number of complex points were doubled by zero-filling once. A polynomial function was used for solvent suppression. The 3D NUS-spectra were constructed using the hmsIST software²⁸.

Resonance assignment.

Resonance assignment was performed using the CCPNmr analysis software version 2.4.2t³³. Chemical shift assignments for the Zα_N173S and Zα_P193A constructs have been submitted to the Biological Magnetic Resonance Data Bank (BMRB) under entry codes 51833 and 51834, respectively.

Calculation of R₂ and τ_corr from R₁ and R_1ρ.

Transverse ${\mathrm{R}}_2$ relaxation rates were calculated from longitudinal ${\mathrm{R}}_1$ and spin-locked longitudinal ${\mathrm{R}}_{1\mathrm{\rho }}$ relaxation rates using the following equation:

$$R_2=R_{1\rho }+(R_{1\rho }-R_1){\tan }^2(\mathrm{\theta })$$

where $\mathrm{\theta }=\tan ({\mathrm{\gamma }}_{\mathrm{N}}{B}_1/2\mathrm{\pi }\mathrm{\Delta }\mathrm{v}),\mathrm{\Delta }\mathrm{v}$ is the resonance offset, $|{\mathrm{\gamma }}_{\mathrm{N}}{B}_1/2\mathrm{\pi }|$ is the strength of the spin-lock field ${\mathrm{B}}_1$, and ${\mathrm{\gamma }}_{\mathrm{N}}$ is the gyromagnetic ratio of the ¹⁵N spin. Effective overall tumbling times ${\tau }_{\mathit{corr}}$ were then calculated from the $R_2/R_1$ ratio³⁴.

Chemical shift perturbation determination.

Chemical shift perturbations were determined between the wild type protein and either mutant protein using the according ¹H- and ¹⁵N-chemical shifts δ in the following equation:

$$CSP=\sqrt{{({\delta }_{H,WT}-{\delta }_{H,mut})}^2+0.2{({\delta }_{N,WT}-{\delta }_{N,mut})}^2}$$

NOESY buildup fitting and extraction of distance restraints:

The uniformly-sampled 3D simultaneously [¹⁵N,¹³C]-resolved [¹H,¹H,X^13C,15N]-NOESY²⁰ spectra measured on Zα_N173S and Zα_P193A constructs with 60 ms mixing times were assigned in CCPNmr³³. The peaks were then transferred to the corresponding 60 ms NUS NOESY spectra and peaks which decreased significantly in quality due to NUS²¹ were removed. The peak lists were then exported to NMRPipe format, and then cross- and diagonal-peak intensities at all mixing times (20, 30, 40, 50, and 60 ms mixing times) were extracted using the NlinLS autofit script in NMRPipe. Fitted auto-relaxation rate constant (ρ) and initial magnetization (M₀) values were used to determine cross-relaxation rate constants (σ) using the full-matrix approach³⁵ package implemented in CYANA³⁶ version 3.98. Spin-diffusion corrections were calculated using the previously solved NMR structure of the wild-type Zα from H. sapiens ADAR1p150 (PDB ID: 1QGP²⁹) and applied to the intensities of the cross-peak buildup curves. We used the average τ_corr values calculated from R₂/R₁ of the respective Zα_N173S and Zα_P193A constructs as inputs for the spin-diffusion corrections. The quality of the fits was inspected visually, and subpar buildups were discarded. Previously determined error tolerances for bi- and uni-directional eNOEs³⁷ were automatically applied by CYANA.

Structure calculations:

Structure calculations of the Zα constructs were carried out in CYANA 3.98³⁶ using 275 bi-directional eNOEs, and 489 uni-directional eNOEs as input for Zα_P193A and 198 bi-direcitonal eNOEs, and 513 uni-directional eNOEs as input for Zα_N173S. The calculations started with 100 initial structures with random torsion angle values using the standard simulated annealing protocol with 50,000 torsion angle dynamics steps. The 20 structures with the lowest target function values were selected for the ensembles. Distance restraints causing violations larger than 1 Å were discarded and the structures were re-calculated. Backbone RMSD values of the two constructs are reported for all residues excluding the flexible termini (resides: 140–198). The final structures of Zα_N173S and Zα_P193A were deposited in the Protein Data Bank with codes 8GBC and 8GBD, respectively.

Highlights:

• The Zα domain of the ADAR1 isoform ADARp150 is critical for the innate immune response

• Zα point-mutations N173S and P193A cause Aicardi-Goutières Syndrome

• Zα mutants bind Z-RNA with decreased affinity

• Mutations change Z-RNA-protein interface structure and conformational dynamics