2′-Alkynyl spin-labelling is a minimally perturbing tool for DNA structural analysis

Abstract The determination of distances between specific points in nucleic acids is essential to understanding their behaviour at the molecular level. The ability to measure distances of 2–10 nm is particularly important: deformations arising from protein binding commonly fall within this range, but the reliable measurement of such distances for a conformational ensemble remains a significant challenge. Using several techniques, we show that electron paramagnetic resonance (EPR) spectroscopy of oligonucleotides spin-labelled with triazole-appended nitroxides at the 2′ position offers a robust and minimally perturbing tool for obtaining such measurements. For two nitroxides, we present results from EPR spectroscopy, X-ray crystal structures of B-form spin-labelled DNA duplexes, molecular dynamics simulations and nuclear magnetic resonance spectroscopy. These four methods are mutually supportive, and pinpoint the locations of the spin labels on the duplexes. In doing so, this work establishes 2′-alkynyl nitroxide spin-labelling as a minimally perturbing method for probing DNA conformation.


S49
Supplementary Tables   Table  Title  Page   1 Mass spec and UV melting data for modified oligonucleotides S6 2 Sequences of oligonucleotides used for crystallization S15 3 Conditions for oligonucleotide crystallization S15 4 X-ray data collection statistics S15 5 Twinning statistics for duplex 6 (6QJS) S16 6 Refinement statistics for duplexes 6 and 7 (6QJS and 6QJR) S16 7 Phase angles and glycosyl torsion angles for duplexes 6 and 7 S16 8 Sugar puckers and pseudorotation phase angles of crystal structures PDB 6QJS (containing 6-Me-labelled uridine), PDB 6QJR (containing 5-Melabelled uridine), and those of their unmodified counterparts PDBs 1S2R and 1D98, respectively S17 9 c torsion angles of crystal structures PDB 6QJS (containing 6-Me-labelled uridine), PDB 6QJR (containing 5-Me-labelled uridine), and those of their unmodified counterparts PDBs 1S2R and 1D98, respectively S18 10 Phase angles for modified nucleotides in the snapshot in Fig. 4 in the main text, and mean / standard deviations over 200.5-1000.5 ns S34 11 Glycosidic torsion angles for the modified nucleotides snapshot in Fig. 4 in the main text, and mean / standard deviations over 200. 5-

S6
stock solution of TEAB to 0.2 M, adjusting its pH to 7-8, then diluting this to 0.1 M with either water (buffer A) or acetonitrile (buffer B).
UPLC-MS characterisation of oligonucleotides was performed on an Acquity UPLC system with a BEH C18 1.7-µm column (Waters) in conjunction with a Waters Xevo G2-QTOF mass spectrometer (ESImode). A gradient of methanol in TEA and hexafluoroisopropanol (HFIP) was used (buffer A: 8.6 mM TEA, 200 mM HFIP in 5% methanol/water (vol/vol); buffer B: 20% buffer A in methanol). Buffer B was increased from 0-70% over 8 min, at a flow rate of 0.2 mL min -1 . Raw data were processed and deconvoluted with the MassLynx software package. Oligonucleotides were quantified based upon their absorption at 260 nm using the nearest-neighbour molar extinction coefficient.  (4,5) used p/2(observer) -t1 -p(observer) -t p(pump) -(t1 + t2 -t) -p(observer) with the observer sequence (32 ns pulses) at 34 GHz and on the maximum intensity of the nitroxyl echo-detected field sweep and the pump pulse (16 ns) at 80 MHz lower frequency. t1 was set at 400 ns and nuclear modulation averaging (5 sets with a difference of 24 ns each time) was used. t2 was either 2, 3 or 4 µs and t was stepped in 8 ns increments. DeerAnalysis2016(6) was used to process the data by removing the last 800 ns to ensure no artefacts were present and setting the zero-time to 328 ns before fitting the homogeneous background decay function.(5) Tikhonov Regularisation was used to extract distance distributions using the parameter determined by DeerAnalysis2016 and the L-curve method.
Distance distributions: Throughout the manuscript, distances have been measured from halfway between the N and O atoms of the nitroxide, as is standard in the analysis of DEER data.
Modelling was carried out to identify spin density on N and O using Gaussian 09, as follows:  There are small differences between duplexes 5 and 6, but very little change between the concentrations tested. The modulation depth of the DEER data was found to depend not only on oligonucleotide concentration, but also on the temperature of the sample prior to rapid freezing in liquid nitrogen for loading into the EPR spectrometer ( Figure S3); rapid cooling of DNA solutions from 293 K gave smaller modulation depths than when cooled from 253 K where the solution was near freezing, implying less complete duplex formation at the higher temperature (estimated at 54% and 69% for 5 and 6 respectively at 200 µM concentration of ssDNA, cooled from 293 K, compared to 100% when cooled from 253 K, see Figure S4). Importantly, the actual distance distributions remain almost equivalent regardless of the method used for freezing or the concentration of the sample ( Figure S5), showing that irrespective of the extent of duplex formation, the nature of the duplex is near identical, and is well-defined. .
To give a value for C we use the approximation that the modulation depth for the 200 µM ssDNA samples frozen from 253 K (pre-cooled) gives the maximum dsDNA obtainable under these measurement conditions, for each label (assigned value of 100% dsDNA). We also assume that each strand is fully spin labelled. The percentage of duplex DNA in the mixture, as shown in Figure   S4, is given from the above equation as (‹n›-1) x 100. The maximum dsDNA concentration at a given value is therefore 0.5 x [ssDNA].  Figure S6) is lower than that of the main peak.    Table S3.
Diffraction data collection. Crystals were flash-cooled in liquid nitrogen. Diffraction data were collected at 100 K at the Diamond Light Source synchrotron science facility, Harwell, Beamline I04, using Pilatus 6M hybrid pixel array detectors. For the 6NO U and 5NO U substituted duplexes, X-ray wavelengths of 0.979 and 0.916 Å were used, respectively. See Table S4.
Diffraction data processing, phase determination, model building, and refinement. Data were indexed and scaled with XDS(8) and AIMLESS. (9) The structures 6QJS (duplex 6) and 6QJR (duplex 7) were solved by molecular replacement with Phaser(10) using PDB codes 1S2R and 1D98 as search models. To avoid model bias, the residues corresponding to the modification sites were removed from the search model. Additionally, terminal base pairs from 1S2R had to be removed for molecular replacement of the 6-Me structure to be successful. After an initial stage of refinement with REFMAC5,(11) missing residues or atoms were rebuilt manually with COOT, (12) followed by successive rounds of further refinement.

X-ray wavelength (Å)
Space group and unit cell dimensions (Å)
Duplex / nt X-ray phase angle Sugar conformation X-ray glycosyl torsion angle        In the simulation where the label was attached to A-DNA as a starting structure, the DNA conformation immediately changed to the B form, as expected and previously reported (for a different system), (22) with the spin label positioned in the minor groove. In the simulation starting from B-DNA, the conformation of the DNA itself remained stable, and the spin label remained in the "antiparallel" conformation, which was more or less identical to the starting conformation. In summary, the same DNA conformation is consistently observed, but with two possible conformations for the spin labels (in the minor groove and antiparallel). These two conformations cannot interconvert in the MD simulations, as the spin labels are on different sides of the backbones in the two cases. Interconversion would require pulling the spin labels through the interior of the duplex, or duplex dissociation, both processes that cannot be achieved in conventional MD simulations.
The optimized structures were used as starting geometries for Langevin dynamics simulations using pmemd.cuda on Nvidia Tesla K40m graphics cards, using a 2 fs timestep, a collision frequency of 2 ps -1 , and SHAKE constraints on bonds involving hydrogen.(49) Periodic boundary conditions were used throughout and the distance cutoff for all nonbonding interactions was set to 10 Å. Long-range electrostatics were described by the particle-mesh Ewald method.(50,51) For van der Waals interactions beyond those included in the direct sum, a continuum model correction

S24
for energy and pressure was used, as implemented in Amber. Simulation snapshots were saved every 100 ps.
System heat-up to 298 K was performed during a 500 ps simulation with weak restraints (10 kcal mol -1 Å -2 ) on the spin labels/DNA in the NVT ensemble. After that, 1 µs NPT simulation was performed for each simulation system at 298 K and 1 bar (weak pressure coupling, isotropic position scaling, pressure relaxation time 2 ps) without any restraints. Analyses of the MD data were performed using cpptraj(52) from the Amber (18,19) suite, and vmd 1.9.3(53) and pymol (54) were used for visualization. Standard MD analyses (RMSD, RMSF, radius of gyration, Watson-Crick base pairing/fraying of termini) were performed using cpptraj from the AmberTools suite (52) and the simulations were found to be well equilibrated after 200.5 ns. This initial phase was considered as equilibration phase. Fraying of the terminal base pairs was practically not observed.  Analysis of the glycosidic torsion angles also revealed excellent agreement between the MD simulations and the X-ray structures. For the minor groove label conformation of duplex 6 as shown in Figure 4a in the main text, torsion angles of -145° (mean -144 ± 9°) and -155° (mean -143 ± 9°) were obtained for the two 6 U nucleotides, compared to values of -141° and -146° in the X-ray structure (Figure 3b). Angles of -139° (mean -144 ± 9°) and -144° (mean -145 ± 10°) for duplex 5 also match very well with the X-ray value of -146° in duplex 7. Figure S18. Mean value of the phase angles describing the sugar pucker of all residues vs. time.
Simulation of system 5 with the spin labels in the minor groove (first of three independent runs).  2G  3C  4A  5A  6A  7T  8T  9U  10G  11C  12G  13C  14G  15C  16A  17A  18A  19T  20T  21U  22G  23C  24G mean phase angle P [°] sequence Table S10. Phase angles for modified nucleotides in the snapshot in Figure 4 in the main text, and mean / standard deviations over 200.5 -1000.5 ns (first of three independent runs). a Phase angle in duplex 7.  Table S11. Glycosidic torsion angles for the modified nucleotides snapshot in Figure 4  to improve clarity.
dCT 6NO UACGCGTCATTG. Strand S1 is complementary to stands S2 and S3, which were modified at position 3 by substitution of a 2'-spin-labelled uridine residue (denoted 5NO U or 6NO U) for thymidine, as shown in Figure S24. Lyophilized oligonucleotides were dissolved in 20 mM sodium phosphate, 80 mM KCl pH 7 in 99.9% D2O + 23 µM DSS (2,2' dimethyl silapentane-5-sulfonate), and duplexes 9 and 10 were formed by mixing equimolar amounts of oligo S1 with oligos S2 and S3 respectively, followed by slow annealing from 60 °C. All NMR spectra were recorded at 14.1 T on an Agilent DD2 spectrometer using a 3 mm inverse triple resonance HCN cold probe. Chemical shifts of the unmodified 14-mer duplex were obtained from published assignments. (56) Sugar conformations were assessed as S/N mixtures using sums of coupling constants derived from DQF-COSY, NOESY and 1D spectra, and critical distance estimates from the 50 ms NOESY mixing time experiments (especially r(1'4')) as previously described. (57,58) S40

S45
Chemical shift differences between the two different spin-labelled duplexes in the diamagnetic and paramagnetic states. Data were taken from Tables S12, S13, S16, S17, and A B Figure S26. Chemical shift difference for 10 NOH -11 and 9 NOH -11 ( Figures A and B, respectively).
Data taken from Tables S12 and S13, and Chernatynskaya, A.V.  peaks than the reduced form (hydroxylamine) owing to paramagnetic broadening.

Data availability
The structure factors and coordinates of the tetramethylpiperidinoxyl and tetramethylpyrrolinoxyl