Adenosine-to-inosine (A-to-I) editing of eukaryotic cellular RNAs is essential for protection against auto-immune disorders. Editing is carried out by ADAR1, whose innate immune response-specific cytoplasmic isoform possesses a Z-DNA binding domain (Zα) of unknown function. Zα also binds to CpG repeats in RNA, which are a hallmark of Z-RNA formation. Unexpectedly, Zα has been predicted — and in some cases even shown — to bind to specific regions within mRNA and rRNA devoid of such repeats. Here, we use NMR, circular dichroism, and other biophysical approaches to demonstrate and characterize the binding of Zα to mRNA and rRNA fragments. Our results reveal a broad range of RNA sequences that bind to Zα and adopt Z-RNA conformations. Binding is accompanied by destabilization of neighboring A-form regions which is similar in character to what has been observed for B-Z-DNA junctions. The binding of Zα to non-CpG sequences is specific, cooperative and occurs with an affinity in the low micromolar range. This work allows us to propose a model for how Zα could influence the RNA binding specificity of ADAR1.

Circular Dichroism

All CD measurements were collected in 1-nm steps from 320 to 220 nm using a JASCO J-815 CD spectrometer (run using Spectra Manager version 2 (JASCO)) in a 0.1 cm quartz cuvette in 20 mM potassium phosphate (pH 6.4), 25 mM NaCl, 0.5 mM EDTA, 1 mM DTT at 25°C. Two scans were acquired and averaged. RNAs were heat denatured at 95°C for one minute followed by rapid cooling on ice and titration samples were prepared by incubating 50 µM of the RNA constructs with the specified amount of Zα or Zα_Tyr177Ala at 42°C for 30 minutes and then bringing the samples down to 25°C over a period of 20 minutes. Control experiments were run to ensure that the absorbance of Zα and Za_Tyr177Ala alone at the same concentrations was minimal in the 250-320 nm range (the range which reports on RNA secondary structure). This indicated that changes in the CD spectra were due to conformational changes in the nucleic acid and not from the superposition of the RNA and Zα spectra. We tested our RNAs with 6 M sodium perchlorate which was shown to fully induce the Z-conformation in the (CpG)_n RNAs.

Extent of Z-formation Calculation

To experimentally quantify the extent of Z-conformation present in an RNA construct, we introduced a Z-conformation score derived from CD spectra (E_Z). E_Z is based on the CD intensities at wavelengths 285 and 295 nm (Int²⁸⁵ and Int²⁹⁵), both of which have been shown to be characteristic of Z-conformation, and the decrease in intensity at 266 nm (Int²⁶⁶), which reports on the reduction of the A-form. We assumed that the growth of the intensities at 285 and 295 nm and decrease at 266 nm are proportional to the presence of Z-conformation with an offset accounting for intensities measured when no Z-conformation is present. We calibrated the following equation from CD intensities measured from (CpG)₃, where the intensities reach plateaus at 100% Z-conformation:

 

(1) E_Z = (1.800*decay²⁶⁶ + 0.718*growth²⁸⁵ + 1.109*growth²⁹⁵)/3

 

where

 

decay²⁶⁶ = (Int²⁶⁶_free – Int²⁶⁶_bound)/Int²⁶⁶_free

growth²⁸⁵ = (Int²⁸⁵_bound – Int²⁸⁵_free)/Int²⁶⁶_free

growth²⁹⁵ = (Int²⁹⁵_bound – Int²⁹⁵_free)/Int²⁶⁶_free

and the prefactors were chosen so that the E_Z score of the (CpG)₃ RNA would be equal to one. 

 

The fractional decrease in the E_Z score was calculated by the following equation:

 

(2) Fractional decrease in the E_Z score = (E_{Z ZαTyr177Ala} - E_{Z ZαWT})/E_{Z ZαWT}

 

where E_{Z ZαWT} and E_{Z ZαTyr177Ala} are the E_Z scores determined from the CD measurements with the wild-type Zα and mutant Zα (Tyr177Ala), respectively.

 

Isothermal Titration Calorimetry

RNA constructs for ITC and the Zα protein were dialyzed overnight into 20 mM potassium phosphate (pH 7.0), 25 mM NaCl, 0.5 mM EDTA, and 1 mM DTT (in the same beaker to match buffers) and concentrated to ~500 µM using Amicon 3 kDa cutoff centrifugal filters. Binding heat was measured on a Malvern ITC200 instrument (run using ITC 200 version 1.26.1 (Malvern)) at 25°C and 750 RPM, with 180 s injection delays and a reference power of 10 µcals^-1. The titrations of (CpG)₃ into Zα were measured with twenty 2 µL injections of 200 µM RNA into 50 µM of protein. The titration of Zα into h43 E. coli was measured with eighty consecutive 0.5 µL injections of 1 mM Zα into 50 µM of RNA. The titration of h43 E. coli into Zα was measured with twenty 2 µL injections of 400 µM RNA into 20 µM of protein. The titration of AluSx1Jo into Zα was measured with twenty 2 µL injections of 200 µM RNA into 20 µM of protein. All ITC thermograms were analyzed and fit using Microcal Analysis version 7 SR4 (Origin).

Analytical Ultracentrifugation

For the interactions between Zα and the different RNAs, the concentrations of Zα and the RNA tested were 2 μM and 12 μM, respectively, corresponding to a 1:6 ratio of RNA:protein, except for h43 which was measured at three different concentrations of Zα (4, 8, and 12 µM). Samples were loaded into a cell composed of standard 12 mm EPON centerpieces with quartz windows and sedimented at 50,000 RPM at 25°C using an XL-I (Beckman Coulter) AUC instrument (run using ProteomeLab version 6.0 (Beckman)). UV absorbance was monitored at 260 nm for 16 hrs. Data were analyzed with SEDFIT (Version 14.7g, NIH) using a specific volume which was normalized to the weight-average of RNA and protein. The complex stoichiometry was chosen according to which theoretical weight was the closest to the measured weight. Error in the measured molecular weight by AUC can be caused by deviation of the predicted specific volume or viscosity of the sample from the actual values. Additionally, if one of the binding sites is weaker than the others resulting in decreased site occupancy, this may cause a deviation in the observed molecular weight from the predicted complex size.

NMR Experiments

All NMR experiments were carried out on Varian 600 and 900 MHz spectrometers (run using VNMRJ version 4.2 Revision A (Agilent)) equipped with 5 mm triple resonance ¹H/¹³C/¹⁵N cold probes with a Z-axis gradient as well as a Bruker 600 MHz spectrometer (run using TopSpin version 7 (Bruker))_equipped with a 5/3 mm triple resonance ¹H/¹³C/¹⁵N/¹⁹F cryoprobe (CP2.1 TCI). 1D ¹H NMR titrations for h43 E. coli, H66 H. sapiens, and H25 E. coli with Zα were carried out on the Varian 600 MHz spectrometer, while the titration for the AluSx1Jo RNA was done on the Bruker 600 MHz spectrometer using a W5 scheme for water suppression (RNA concentration was 500 μM for all). The number of scans for all titration points was 128 with a relaxation delay of 1.6 s, and the spectral width was 24 ppm. 2D ¹H-¹H NOESY spectra were recorded on the Varian 900 MHz spectrometer for h43 E. coli, the Varian 600 MHz spectrometer for H66 H. sapiens and H25 E. coli, and the Bruker 600 MHz spectrometer for the AluSx1Jo RNA (RNA concentration was 1 mM for all). The 2D ¹H-¹H NOESY recorded for h43 E. coli was acquired with a mixing time of 200 ms, 1470 x 800 complex points (399 of the points were actually collected following a 50% NUS sampling scheme generated using the Schedule Generator from the Wagner group: http://gwagner.med.harvard.edu/intranet/hmsIST/gensched_new.html), a 1.3 s recycle delay, and 32 scans. The spectral widths were 22 ppm x 22 ppm for both ¹H dimensions. The 2D ¹H-¹H NOESY recorded for H66 H. sapiens was acquired with a mixing time of 320 ms, 1386 x 400 complex points (162 of the points were collected following a 40% NUS sampling schedule), a 1.3 s recycle delay, and 32 scans. The spectral widths were 20 ppm x 20 ppm for both ¹H dimensions. The 2D ¹H-¹H NOESY recorded for H25 E. coli was acquired with a mixing time of 200 ms, 1396 x 400 complex points (162 of the points were collected following a 40% NUS sampling schedule), a 1.3 recycle delay, and 32 scans. The spectral widths were 20 ppm x 20 ppm for both ¹H dimensions. The 2D ¹H-¹H NOESY recorded for AluSx1Jo was acquired with a mixing time of 300 ms, 1024 x 400 complex points, a 2 s recycle delay, and 32 scans. The spectral widths were 24.5 ppm x 24.5 ppm for both ¹H dimensions. 1D spectra from the Bruker spectrometer were processed using TopSpin; all other data were processed using the NmrPipe/NmrDraw/NlinLS package version 10.9. All NUS data was reconstructed using the hmsIST software (a part of NMRPipe). All assignments were done in ccpNmr Analysis version 2.4.2.

For the titration of AluSx1Jo into Zα, all NMR measurements were carried out on the Varian 900 MHz spectrometer. The ¹⁵N-HSQC spectra for the titration were measured with 1048 x 60 complex points with a 1.6 s recycle delay and 32 scans. The spectral widths were 16 ppm x 35 ppm for the ¹H and ¹⁵N dimensions. The concentration of Zα was 200 μM. The R₁ relaxation experiments were measured with 1048 x 64 complex points, a recycle delay of 2 s, 16 scans, and relaxation delays of 0, 100, 200, 300, 400, 500, 600, 700, 800, 900, 1000, and 1200 ms. The spectral widths were 16 ppm x 35 ppm for the ¹H and ¹⁵N dimensions. The R_1ρ relaxation experiments were measured with 1048 x 64 complex points, a recycle delay of 2 s, 32 scans, and relaxation delays of 0, 10, 20, 30, 40, 60, 80, 100, 120, and 160 ms. The spectral widths were 16 ppm x 35 ppm for the ¹H and ¹⁵N dimensions. During the R_1ρ relaxation time, a ¹⁵N spin-lock field of 1500-Hz strength was applied. The transverse relaxation rate R₂ was calculated from R₁ and R_1ρ using the following equation:

 

(3) R₂ = R_1ρ + (R_1ρ – R₁)*tan²(θ)

 

where θ = tan^-1(2πΔv/γ_NB₁), Δv is the resonance offset, |γ_NB₁/2π| is the strength of the spin-lock field B₁, and γ_N is the gyromagnetic ratio of the ¹⁵N spin.

 

τ_corr was calculated from the ratio of R₂/R₁. The τ_corr for free Zα was calculated from R₁/R_1ρ measurements done with a 500 μM Zα sample instead of the 200 μM sample used in the titration.

The Carr-Purcell-Meiboom-Gill (CPMG) experiment at 2:1 Zα:AluSx1Jo (1 mM Zα, 500 μM AluSx1Jo) was measured on the 600 MHz Bruker spectrometer with 1024 x 64 complex points, a recycle delay of 1.2 s, 32 scans, and 11 ν_CPMG values (T = 40 ms) ranging from 10-1000 Hz. The spectral widths were 16 ppm x 35 ppm for the ¹H and ¹⁵N dimensions. Dispersion profiles were fit to a two-state fast CPMG exchange model with the following equation:

 

(4) R_2eff = R_2a + p_a*(1-p_a)*(Δω²)/k_ex*{1-(4*ν_CPMG/k_ex)*(tanh(k_ex/ν_CPMG/4))}

 

 where R_2a and p_a are the R₂ relaxation rate and the population of state A, Δω is the difference in the chemical shift between states A and B, k_ex is the rate of exchange between states A and B, and ν_CPMG is the effective field strength of the refocusing pulse train.

 

The ¹⁵N-CEST experiment at 10:1 Zα:AluSx1Jo (1 mM Zα, 100 μM AluSx1Jo) was measured on the 600 MHz Bruker spectrometer with 1024 x 64 complex points, a 1 s recycle delay, and 16 scans. The spectral widths were 16 ppm x 35 ppm for the ¹H and ¹⁵N dimensions. A weak ¹⁵N B₁ field of 5 Hz-strength was applied during a 400 ms relaxation time. 66 data sets were acquired corresponding to a chemical shift range of 102-132 ppm with steps of 0.5 ppm. From 115-116 ppm, the steps were decreased to 0.25 ppm to acquire additional points around Tyr177. In all experiments, ¹H decoupling was achieved using a 90_x240_y90_x composite pulse.

K_D fitting of NMR titration points

          Chemical shift perturbations from the ¹⁵N-HSQC titration of AluSx1Jo into ¹⁵N-labelled Zα were calculated using the following equation:

 

(5) CSP = (δ_H,free- δ_H,bound)² + 0.2(δ_N,free- δ_N,bound)²

 

where δ_H is the chemical shift of a peak in the ¹H dimension, and δ_N is the chemical shift of a peak in the ¹⁵N dimension. The K_D was determined by fitting CSPs to the following equation:

 

(6) CSP = CSP_max*((K_D+[L]+([P]/2))-sqr(([P]/2)²-(4*[L]*([P]/2)))/2*([P]/2))

Where [P] is the concentration of protein in solution (divided by two because the ligand has two binding sites), [L] is the ligand concentration, and CSP_max is the maximum CSP measured over the series of titration points.