Why are Hoogsteen base pairs energetically disfavored in A-RNA compared to B-DNA?

Transcription

1 Published online 4 October 2018 ucleic cids Research, 2018, Vol. 46, o doi: /nar/gky885 Why are oogsteen base pairs energetically disfavored in -R compared to B-D? tul Rangadurai 1, uiqing Zhou 1,DawnK.Merriman 2, athalie Meiser 3, Bei Liu 1, onglue Shi 2, Eric S. Szymanski 1 and ashim M. l-ashimi 1,2,* 1 Department of Biochemistry, Duke niversity School of Medicine, Durham,, S, 2 Department of hemistry, Duke niversity, Durham,, S and 3 oethe niversity, Institute for Organic hemistry and hemical Biology, Frankfurt am Main, ermany Received July 14, 2018; Revised September 17, 2018; Editorial Decision September 19, 2018; ccepted October 02, 2018 BSR (syn)-/and(syn)- + oogsteen () base pairs (bps) are energetically more disfavored relative to Watson rick (W) bps in -R as compared to B-D by >1 kcal/mol for reasons that are not fully understood. ere, we used MR spectroscopy, optical melting experiments, molecular dynamics simulations and modified nucleotides to identify factors that contribute to this destabilization of bps in - R. Removing the 2 -hydroxyl at single purine nucleotides in -R duplexes did not stabilize bps relative to W. In contrast, loosening the -form geometry using a bulge in -R reduced the energy cost of forming bps at the flanking sites to B-D levels. structural and thermodynamic analysis of purine-purine mismatches reveals that compared to B-D, the -form geometry disfavors syn purines by kcal/mol due to sugar-backbone rearrangements needed to sterically accommodate the syn base. Based on MD simulations, an additional penalty of 3 4 kcal/mol applies for purine-pyrimidine bps due to the higher energetic cost associated with moving the bases to form hydrogen bonds in -R versus B-D. hese results provide insights into a fundamental difference between -R and B-D duplexes with important implications for how they respond to damage and post-transcriptional modifications. IRODIO -form R (-R) and B-form D (B-D) double helices have several important differences (Figure 1). In B-D, the deoxyribose moiety is flexible and exists in dynamic equilibrium between major 2 -endo and minor 3 -endo sugar pucker conformations (Figure 1) (1,2). Steric clashes between the 2 -hydroxyl of a 2 -endo ribose sugar and the backbone, in conjunction with electronic effects of the 2 -hydroxyl preclude the formation of a corresponding B-form R double helix (3 5). Rather, the ribose primarily adopts the 3 -endo conformation in - R. his brings into proximity adjacent nucleotides thus shortening the double helix and moving bps away from the helical axis (6) (Figure1). he resulting -R double helix is also more rigid than its B-D counterpart (7 9). hese differences between -R and B-D have important biological implications for their recognition by proteins (10,11) and ligands (12), the templated processes of replication, transcription and translation, the consequences of ribonucleotide and deoxyribonucleotide misincorporation (13,14), and the impact of damage (15) and chemical modifications. We recently reported markedly different propensities between B-D and -R duplexes to form (syn)- + and (syn)-/ oogsteen () base pairs (bps) (15). bp is created by rotating the purine base in a Watson rick (W) bp 180 about the glycosidic bond to adopt a syn (0 < < 90,where is the glycosidic O torsion angle for purines) rather than an anti ( 180 < < 90 ) conformation, followed by translation of the two bases by 2.0 Å, creating a unique set of hydrogen bonds (h-bonds) (Figure 1B) (16). In B-D, the free energy associated with converting or W bps into their counterparts ( -W ) is estimated to be 2 4 kcal/mol based on MR relaxation dispersion (RD) and optical melting experiments (15,17,18). In contrast, the free energy cost for forming bps in -R ( 4 7 kcal/mol) is higher by 1 5 kcal/mol (15). he difference in the propensities of -R and B- D to form bps has important biological implications. hemically modified nucleotides such as 1 - methyladenonisine (m 1 ) and 1 -methylguanosine (m 1 ), which are impaired from forming W bps (Figure 1), occur as a form of damage in both D (19) andr(20), Downloaded from by guest on 18 ovember 2018 * o whom correspondence should be addressed. el: ; hashim.al.hashimi@duke.edu Present address: uiqing Zhou, Institute for Biophysical Dynamics, niversity of hicago, hicago, IL, S. he uthor(s) Published by Oxford niversity Press on behalf of ucleic cids Research. his is an Open ccess article distributed under the terms of the reative ommons ttribution License ( which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited.

2 11100 ucleic cids Research, 2018, Vol. 46, o. 20 B O 3 O 3 ~17 s -1 ~3000 s -1 ~7 Å ~6 Å 1' anti O anti 1' syn anti 10.4 Å 1' 8.5 Å O 1' pop ~ 99.5% pop ~ 0.5% 2'-endo 3'-endo O O 2 ~11 s B-D -R -1 1' ~1300 s -1 3 anti Oanti 1' syn 1' O anti 1' O 3 O 10.6 Å 8.3 Å m 1 pop ~ 99.2% pop ~ 0.8% m1 3 Watson-rick oogsteen anti O anti 1' 1' syn O anti 1' D P OP1 E OP1 P O O 2 O O 3 3 m m 1 1 O4' 3 1' anti Oanti 1' syn O anti 1' 1' O2' 1' 3 Watson-rick 3 oogsteen 3 B-D syn -R syn -R anti Figure 1. oogsteen base pairs in D and R. () Structural comparison of B-form D and -form R. (B) W bps in B-D exist in dynamic equilibrium with bps. Rates and populations were obtained using MR relaxation dispersion methods as described previously (18). () 1 -methylated purines disrupt W bps via steric clashes (red dashes) and loss of hydrogen bonding interactions. hey are accommodated as bps in B-D. (D) Syn purine bases clash (red dashes) with the sugar and backbone in -R but not in B-D. (E) Water mediated interactions between the 2 -hydroxyl and the 3 atom of the purine base may stabilize purine bases in the anti conformation in -R (PDB ID: 315D). and as post-transcriptional modifications in R (21). B- D is able to absorb this form of alkylation damage through the formation of bps (Figure 1) (17,22), which can in turn be recognized by damage repair enzymes (19,23), safeguarding the integrity of the genome. In contrast, because of the instability of bps in -R, these modifications disrupt base pairing all together (15). his enables m 1 and m 1 to serve unique functional roles as posttranscriptional modifications that direct the proper folding (24,25) and aminoacylation of tr (26), and prevent frameshifting errors during translation (27). similar mechanism has also been proposed (15) to be responsible for the ability of m 1 to increase translation (28,29) by destabilizing the 5 R of mr. hese functions would not be realized if m 1 and m 1 could form bps in -R. deep understanding regarding the factors that dictate the differences in bp stability relative to W in B-D versus -R is important for understanding the basis of these biological processes and may also provide guidelines that could help identify other processes in which the suppression or enhancement of bps is important. Based on computational modeling and a survey of crystal structures, we proposed previously that bps are energetically more disfavored in -R as compared to B- D due to steric clashes between the syn base and ribose sugar (3 O4 ), and the phosphate backbone (3 O5, 3-OP1) (15) (Figure1D). hese clashes arise due to the 3 -endo sugar pucker in R, which brings the syn purine base and phosphate group closer together. similar mechanism has also been proposed to be responsible for the tendency of 3 -endo sugars to disfavour the syn conformation of the purine base in isolated Ps (30 32). In contrast to R, base-backbone steric clashes are not observed for syn purines in B-D owing to its altered 2 -endo sugar pucker (Figure 1D). lthough this simple steric model is appealing, there is reason to believe that it does not fully capture all the factors contributing to bp instability in -R. Based on X-ray crystallography (33 35) and solution state MR spectroscopy (36 38), purine-purine mismatches such as (syn)- and (syn)- + /-(syn) adopt bps in both -R and B-D duplexes. his indicates that any steric clashes with the syn purine base can be resolved through conformational adjustments. Moreover, the free energy associated with these adjustments likely does not exceed the energetic cost of base opening (39), otherwise well-formed purine-purine bps would not be observed in solution. his raises the possibility that there are other factors contributing to the instability of (syn)- and (syn)- + bps in -R. In addition, other contributions involving the 2 -hydroxyl need to be re-assessed because X-ray crystallography (40,41), MR spectroscopy (42,43) and computational simulations (44) provide evidence for water mediated h-bonds between the 2 -hydroxyl and the base atoms in the minor groove of R (Figure 1E) (45). hese h-bonds with W bps would be disrupted Downloaded from by guest on 18 ovember 2018

3 ucleic cids Research, 2018, Vol. 46, o when forming a syn base and could also account, at least in part, for the greater instability of bps relative to W bps in -R versus B-D. Finally, differences in stacking interactions of the syn purine base and the strengths of h-bonds it forms could also contribute to the differences in the propensities of R and D duplexes to form bps. ere, we used MR spectroscopy, melting experiments, molecular dynamics (MD) simulations as well as chemically modified nucleotides to dissect the origins of bp instability in -R. Our results indicate that the -form geometry destabilizes purine-pyrimidine and purine-purine bps due to the energetic cost associated with changing the sugar-backbone conformation to accommodate the syn purine base (by kcal/mol as compared to D). Furthermore, an additional penalty of 3 4 kcal/mol as compared to B-D applies for the formation of purinepyrimidine bps in -R that is associated with the translation of bases into hydrogen bonding proximity. hese results provide deeper insights into a fundamental difference between R and D duplexes and have important implications for how -R and B-D respond to damage and post-transcriptional modifications. MERILS D MEODS Sample preparation nmodified D and m 1 r containing oligonucleotides: ll unmodified D oligonucleotides were purchased from Integrated D echnologies with standard desalting purification, while all m 1 r containing single stranded oligonucleotides ( 6 -D m1r10, 6 -R m1r10, IV-2 R m1r26, 2 -R m1r10,gcr m1r4 ) were purchased from E ealthcare Dharmacon with PL purification. m 1 d and m 1 r containing D and R oligonucleotides: he 6 -D m1r16 single-strand was purchased from Yale Keck Oligonucleotide Synthesis Facility with len-pak R cartridge purification while the 6 - D m1d16 single-strand was purchased from Midland ertified Reagents with reverse-phase PL purification. he 6 -R m1r16 and 6 -R m1d16 single-stranded oligonucleotides were synthesized in-house using a Mer- Made six oligo synthesizer. ltramild BDMS R amidites (Pac-r, c-r, ipr-pac-r, r, len Research orporation), m 1 phosphoramidites (len Research orporation) and ltramild ap solution were used with a coupling time of 12 min, with the final DM group being cleaved during the synthesis. he R oligonucleotides were then cleaved and deprotected after support removal using a 2 ml solution of 2 M ammonia in methanol for 24 h at room temperature. he solutions were then centrifuged and the supernatant dried under airflow. he resulting oligonucleotide crystals were dissolved in 100 l DMSO and 125 l E.3F, and heated at 65 for2.5hfor2 -O deprotection. he oligonucleotides were then ethanol precipitated and dissolved in a formamide based loading dye for purification using PE. el bands corresponding to the pure product were identified by V-shadowing and subject to electroelution (Whatman, E ealthcare) followed by ethanol precipitation. R oligonucleotides used for comparison with m 1 r/m 1 d containing R duplexes in optical melting experiments: ll the unmethylated R single strands used to prepare 6 -R, 6 -R d16, 6 -R m1d16 and 6 -R m1r16 duplexes that were compared with the m 1 r/m 1 d containing R duplexes in the optical melting measurements were synthesized in-house using a MerMade 6 oligo synthesizer. In particular, acetyl protected BDMS R amidites (hemgenes) and standard D phosphoramidites (n-bz d, hemgenes) were used with a coupling time of 6 min for R and 1 min for D, with the final 5 -DM group removed during synthesis. he oligonucleotides were cleaved from the supports (1 mol) using 1 ml of M (1:1 ratio of ammonium hydroxide and methylamine) for 30 min and deprotected at room temperature for 2 hrs. ll subsequent purification steps were similar to those used for the m 1 d/m 1 r containing R oligonucleotides. m 1 d and rmp containing D oligonucleotides: he 6 -D m1d10 single strand was purchased from Midland ertified Reagents with cartridge purification. ll other D single strands containing m 1 d (gcd m1d4 and 2 -D m1d10 ) or rmp incorporations ( 6 -D r10 and 6 -D r16 ), were synthesized in-house using a Mer- Made 6 oligo synthesizer. In particular acetyl protected B- DMS R phosphoramidites (hemgenes) and standard D phosphoramidites (n-ibu-d, bz-d, ac-d, d, dmfm 1 d, hemgenes) were used with a coupling time of 6 min (R) and 1 min (D), with the final 5 -DM group retained during synthesis. he oligonucleotides were cleaved from the supports (1 mol) using 1 ml of M (1:1 ratio of ammonium hydroxide and methylamine) for 30 min and deprotected at room temperature for 2 h. he m 1 d containing D samples were then purified using len- Pak D cartridges and ethanol precipitated, while the rmp containing samples were dried under airflow to obtain oligonucleotide crystals. hey were then dissolved in 115 ldmso,60 l E and 75 l E.3F and heated at 65 for2.5hfor2 -O deprotection. he samples were then neutralized using 1.75 ml of len-pak R quenching buffer, loaded onto len-pak R cartridges for purification and were subsequently ethanol precipitated. Other R oligonucleotides: he remaining R single strands used for preparing 6 -R, 6 -R d10, 6 - R m1d10, 6 -R m1r10, IV-2 R, 2 -R, 2 -R, 2 -R m1r4, gcr, gcr, gcr m1r4 samples were synthesized in-house using a MerMade 6 oligo synthesizer. In particular acetyl protected BDMS R phosphoramidites (hemgenes) and standard D phosphoramidites (n-ibu-d, dmfm 1 d, hemgenes) were used with a coupling time of 6 min (R) and 1 min (D), with the final 5 -DM group retained during synthesis. he subsequent purification steps used were similar to those used for preparing the rmp containing D oligonucleotides. Sample annealing and buffer exchange: ll single strands (afterethanol precipitation/purchase) were re-suspended in water. Duplex samples were prepared by mixing equimolar amounts of the constituent single strands and annealed by heating at 95 for 5 min and cooling at room temperature for 1 h. ll hairpin samples (IV-2 R, IV- Downloaded from by guest on 18 ovember 2018

4 11102 ucleic cids Research, 2018, Vol. 46, o. 20 2R m1r26 ) were prepared by diluting the re-suspended oligonucleotides to a concentration of 50 M followed by heating at 95 for 5 min and cooling on ice for 1 h. Extinction coefficients for all single and double stranded species were estimated using the DBIO oligo calculator ( Following annealing, the samples were exchanged three times into the desired buffer using centrifugal concentrators (4 ml, Millipore Sigma). Buffer preparation: Sodium phosphate buffers for MR and optical melting measurements were prepared by the addition of equimolar solutions of sodium phosphate monobasic and dibasic salts, sodium chloride, ED and magnesium chloride to give final concentrations (unless mentioned otherwise) of 15 mm (phosphate), 25 mm, 0.1 mm and 3 mm, respectively. he p of the buffers was adjusted by the addition of phosphoric acid, after which they were brought up to the desired volume, and filtered and stored for further usage. Potassium phosphate buffers for optical melting measurements were also prepared in an analogous manner using equimolar solutions of potassium phosphate monobasic and dibasic salts, and potassium chloride. MR spectroscopy MR experiments were performed on an 800 Mz Varian DirectDrive2 spectrometer and a 700 Mz Bruker vance 3 spectrometer equipped with triple-resonance cryogenic probes. ll experiments were conducted in p 5.4, 25 mm al and at 25 unless stated otherwise. he MR data was processed and analyzed with MRpipe (46) and SPRKY (47). Resonances were assigned using 2D OESY, OSY and DQF-OSY experiments along with SOFS MQ experiments for aromatic (48) and imino (49) spins. he OSY and DQF-OSY experiments were performed in D 2 O following sample lyophilization. he coupling constant J 1-2 for r16 and r10 in 6 -D r16 and 6 -D r10,and 1 for d10 in 6 - R d10 were measured along the direct dimension ( 1 ) of the DQF-OSY spectra after phasing of the relevant cross peak. hemical shift perturbations ( ) for a pair of resonances (2-2/8-8/6-6/1-1 ) belonging to a given residue were calculated using the following equation ( ) 2 ω = ( ω ) 2 γ + ω γ where ω and ω are the chemical shift perturbations in the hydrogen and carbon dimensions of a 2D SQ spectrum, and γ and γ are the gyromagnetic ratios of hydrogen and carbon. chemical shift perturbation ( )was considered to be significant when 0.4 ppm. Optical melting experiments MR samples were diluted using MR buffer to a concentration of 3 M and used for optical melting experiments (extinction coefficients for the modified duplexes were assumed to be the same as that for the unmodified ones. Modified bases are estimated to affect the extinction coefficient for the oligos used here by < 10% based on reference values in Basanta-Sanchez et al. (50)). Measurements were performed using a Perkin Elmer Lamba 25 V Vis spectrophotometer with a Peltier temperature control unit and a sample/blank volume of 400 L. Samples were heated a rate of 1 /min with the absorbance at 260 nm ( 260 ) being recorded every 0.5 min. he absorbance curves were then fit to obtain the thermodynamic parameters using an in-house Mathematica script and the following equations 260 = (((m ds ) + b ds ) f) + (((m ss ) + b ss ) (1 f)) ( ) 1 + 4e (1/ m 1/) /R ( 1 + 8e ) (1/ m 1/) /R 1/2 f = 4e (1/ m 1/) /R where m ds,b ds,m ss and b ss are coefficients representing the temperature dependence of the extinction coefficients of the double and single stranded species, is the temperature in Kelvin, f is the fraction of the double stranded species at a given temperature, m is the melting temperature in Kelvin, is the enthalpy of the melting transition in kcal/mol and R is the universal gas constant in kcal/mol/k. he free energy of the melting transition was then obtained as follows S = / m Rln ( t /2) and = S where t is the total concentration of the duplex species at the start of the measurement and S and are the entropy and free energy of the melting process. While performing experiments on hairpin systems, the fraction of folded hairpin at a given temperature was defined by f = e (1/ m 1/) /R 1 + e (1/ m 1/) /R he m and from fitting the absorbance curve were then used to get the free energy as follows S = / m and = S Errors in the thermodynamic measurements (one standard deviation) were estimated by performing the experiments in triplicate. MD simulations ll MD simulations were performed using the ff99 M- BER force field (51) with bsc0 corrections for D (52) and OL3 corrections (53) for R, using periodic boundary conditions as implemented in the MBER MD package (54). Starting structures for 6 -D with the 16-9 base pair in a W//* conformation were obtained from the MR structures of 6 -D (PDB ID: 5ZF) and 6 -D m1d16 (PDB ID: 5ZI) (55). * (56) refers to a (syn)- bp conformation in which the syn adenine is not hydrogen bonded to the thymine as the 1 1 distance across the bp is not constricted. he deposited 5ZF structure was used as is, to model the W conformation of the 16-9 bp. he conformation was modeled by removing the 1 -methyl group from 5ZI while the * conformation was modeled by flipping the 16 base in 5ZF 180 about the glycosidic bond. he starting structure for 2 - D with the 16-9 bp in a W conformation was obtained from PDB ID 5ZD (MR structure of 2 -D) Downloaded from by guest on 18 ovember 2018

5 ucleic cids Research, 2018, Vol. 46, o while that in a * conformation was derived from 5ZD by flipping the 16 base by 180 about the glycosidic bond. he starting structure for 2 -D with a conformation of the 16-9 bp was obtained from Sathyamoorthy et al. (55). Starting structures for the 6 and 2 R systems with the 16-9 bp in a W//* conformation were generated by constructing idealized helices using 3D (57). he conformation was modeled by superimposing (using base heavy atoms) the m 1 - bp from 5ZI (after removing the thymine and 1 -methyl groups) onto the 16-9 bp (in the idealized structure, with 16 in the syn conformation) and replacing the atoms of both bases with those in the superimposed bp. he * conformation was modeled by flipping the base moiety of 16 in the idealized structure about the glycosidic bond by 180. Starting structures for 6 -D and 6 -R with the bp in a * conformation were generating by rotating about the glycosidic bond in 5ZF and an idealized helix respectively. he conformation for the bp was modeled in a manner similar to that for 16-9 bp in 6 -R for both systems, with the reference (syn)- + bp taken from PDB ID 1XVK (bp 1:8). Parameters for protonated cytosine (for the and * starting geometries of the bp) were obtained from oh et al. (58). Starting structures for the 6 and 2 D and R duplexes containing a - mismatch at position 10:15 were built by constructing idealized helices using 3D (57). In particular, the base atoms of the 10(syn)- 15(anti) bp (obtained by replacing 15 in the idealized structure with an anti guanine) were replaced with those of a superimposed (using the base heavy atoms of the anti guanine) - mismatch from PDB ID 1D80 (bp 9:16). ll helices were then solvated using a truncated octahedral box of SP/E (59) water molecules, with box size chosen such that the boundary was at least 10 Å away from any of the D atoms. a + ions treated using the Joung-heatham parameters (60) were then added to neutralize the charge of the system. he system was then energy minimized in two stages with the solute (except for the sugar moiety of the 16-9/ bps in the conformation, and the sugar atoms of the - mismatch) being fixed (with a restraint of 500 kcal/mol/å2 ) during the first stage. his was followed by gradual heating of the system using the Berendsen thermostat (61) to 298 K under constant volume conditions for 100 ps with harmonic restraints on the solute (10 kcal/mol/å2 ). he system was then allowed to equilibrate for 1 ns under constant pressure (1 bar, using the Berendsen barostat, = 2 ps) and temperature (at 298 K, using Langevin dynamics, = 3ps 1 ) conditions. non-bonded cutoff of 9 Å was used for treating short range non-bonded interactions while the Particle Mesh Ewald method (62)was used to treat long range electrostatic interactions. ovalent bonds involving hydrogen were constrained using the SKE algorithm (63) to enable the use of a 2 fs timestep. Simulations of 6 -D with and * starting geometries of the 16(syn)-9 bp, and a 10(syn)-15 mismatch were also performed using the ff99 MBER force field (51) with the recently developed parmbsc1 (64) and OL15 (65) corrections for D, using the protocol described above. Production runs of length 1 s were obtained for all the simulations. he settings used for the production runs were identical to those used during equilibration. set of evenly (5 ps) spaced snapshots was used for subsequent analysis using the PPRJ suite (66) of programs. Visual examination of the MD trajectories revealed the absence of terminal end fraying artifacts near the base pair of interest in all simulations apart from that of 6 -D containing a 10(syn)-15(anti) mismatch (bsc0 force field, after 650 ns). he terminal base pairs in this simulation were seen to interact with the mismatch via insertion in both the minor and major grooves. owever, this did not affect the stability of the - mismatch (Supplementary Figure S9D). Furthermore, examination of the RMSD of the heavy atoms (excluding the terminal residues) of D/R during the production runs for all simulations suggests that they are converged (Supplementary Figure S7). Survey of crystal structures with purine-purine mismatches ll crystal structures containing nucleic acids with a resolution < 3 Å as of 27 pril 2017 were downloaded from the Protein Data Bank (PDB) (67) and analyzed for the presence of purine-purine mismatches formed by the canonical bases and their modified derivatives, that are flanked by two W bps on both sides to mimic a duplex like environment, using an in-house Python script. he mismatches were then classified based on the angle of the constituent bases to obtain syn-anti/ conformations for all mismatch types (-/-/-). wo structures of MutS bound to D containing purine-purine mismatches (1O6 and 1O7) were manually excluded owing to the distorted/open geometry of the mismatch caused by the intercalation of aromatic amino acids from MutS in the D minor groove, as they are unlikely to be representative of the accommodation of a mismatch in a duplex context. total of 69 purine-purine bps belonging to 37 distinct structures were identified out of a total of 5906 deposited structures in the PDB containing nucleic acids. dditional statistics obtained from the survey along with the identified purine-purine mismatches are shown in ables S3 and S4 in the supporting material. he torsion angles of the mismatched bps were compared with a set of unmodified W bps from free (not bound to proteins/ligands) D/R structures placed in a similar structural context, with B/ form helical geometry as determined using DSSR (68). Purine-purine bps were also compared with a set of isolated purine-pyrimidine bps in D (flanked by W bps on both sides) identified earlier (56). o crystal structures of purine-pyrimidine bps in duplex environments in R were found, in line with a previous study (15). alculation of changes in stacking interactions accompanying the formation of purine-purine mismatches riplets of bps containing the purine-purine mismatches in Supplementary able S4 along with their neighboring bps were extracted from the respective PDB files. Modified bases such as inosine, 5-bromo uridine, 8- bromo guanosine were replaced with the canonical bases (////) by adding/removing the extra atoms. For each mismatched triplet, a corresponding W base paired triplet was created by constructing an idealized B/-form Downloaded from by guest on 18 ovember 2018

6 11104 ucleic cids Research, 2018, Vol. 46, o. 20 helix with 3-D (57), by using the sequence of the mismatched strand containing the syn base. For example, the W base paired triplet corresponding to the mismatched sequence 5 -(syn)-3 /5 -(anti)-3 would be / he stacking interactions in the triplet of matched and mismatched bps were calculated by computing the area of base overlap using the analyze utility of the 3-D suite (57). he change in overlap area between the mismatched and matched triplets is computed with the inclusion of exocyclic groups in Supplementary Figure S6F and Supplementary able S5 in the supporting material. omputation of thermodynamic parameters for mismatch formation hermodynamic parameters for - and -/- mismatch formation (relative to a - bp) were computed using MELI 5.0 (69) for all possible sequence contexts surrounding the mismatch for both D and R. Default options for nearest neighbor thermodynamic parameters and ion correction terms were used along with a sodium ion concentration of 150 mm. he energetic terms for helix initiation and symmetry were set to zero, in order to mimic the placement of a mismatch within the context of a nonpalindromic duplex. RESLS Removing the 2 -hydroxyl at a purine nucleotide does not rescue bp formation in -R Based on the steric model (Figure 1D) (15) bps are disfavored in -R as compared to B-D due to steric clashes that do not involve the 2 -hydroxyl group. hus, we predict that removal of the 2 -hydroxyl at a purine nucleotide in R should not result in a resurgence of stably formed bps on 1 -methylation, as long as the local conformation remains -form. On the contrary, if bps are destabilized relative to W bps in R solely due to favorable water mediated (Figure 1E) (45) or other interactions of the hydroxyl with the W bp, then we predict that its removal should result in the occurrence of stable bps on 1 -methylation. o test these predictions, we examined the consequence of removing the 2 -hydroxyl at a purine nucleotide (r16 or r10) in the 6 -R duplex (Figure 2). omparison of 2D SQ MR spectra for d16 or d10 substituted 6 -R duplexes with their unmodified counterparts indicates that deoxyribose substitution does not alter the global -form conformation and results in small chemical shift perturbations in and around the site of substitution (Supplementary Figures S1 and S1B). he aromatic carbon and proton chemical shifts of the dmp residues are consistent with a helical -form R conformation (Supplementary Figures S1 and S1B). In addition, the sugar carbon (1,3 and 4 ) chemical shifts (70,71) and scalar coupling constants ( 1 = J J 1-2 and 2 = J J J 2-3 )(2,72) indicate that the dmp residue adopts a 3 -endo sugar pucker as expected for an -form conformation (Figure 2B and ). s the conformation remains -form on hydroxyl removal, bp formation should still be disfavored at the dmp residue based on the steric model. Indeed, 1 1D spectra of the imino region of the 6 -R duplexes containing m 1 d and m 1 d show an absence of characteristic hydrogen bonding signatures corresponding to bp formation (17) and a lack of base pairing near the site of 1 -methylation (Figure 2D). Furthermore, characteristic MR signatures of bp formation such as downfield shifted 8 resonances and strong intra-nucleotide 1-8 cross peaks corresponding to the syn purine are not observed (Supplementary Figure S2-). his indicates that the duplexes adopt partially melted conformations just like their m 1 r and m 1 r counterparts (15), although we cannot rule out that hydrogen bonded bps are not formed transiently. he above experiments suggest that the effects of 1 -methylation on 6 -R are independent of the presence/absence of the 2 -hydroxyl. his was independently validated using optical melting measurements on modified 6 -R duplexes. he destabilization due to the incorporation of m 1 orm 1 ( 1methyl-W ), computed as the difference in free energies of melting between the 1 -methylated duplex and its unmethylated counterpart, was seen to be minimally affected (by kcal/mol) on removal of the 2 -hydroxyl under a variety of experimental conditions such as low (Supplementary Figure S2D) and moderate salt (Supplementary Figure S2E), presence of magnesium (Supplementary Figure S2F), neutral p (Supplementary Figure S2) and presence of potassium (Supplementary Figure S2). aken together, the above results argue against the loss of stabilizing interactions with the purine 2 -hydroxyl as being the primary source of instability of bps relative to W bps in -R compared to B-D, and are in accordance with the steric model. dding the 2 -hydroxyl at a purine nucleotide does not significantly destabilize bps relative to W in B-D We previously performed the inverse experiment and examined the consequence of introducing a 2 -hydroxyl group at a purine nucleotide in B-D on the stability of bps relative to W. MR RD measurements on rmp (r10 and r16) substituted 6 -D duplexes (Figure 3) revealed that addition of the 2 -hydroxyl group slightly increases the energetic cost of forming bps relative to W ( -W ) by kcal/mol (15). We corroborated these findings using optical melting experiments on the rmp substituted 6 -D duplexes with and without 1 -methylation at 10 and 16 (Figure 3B). hese experiments yielded changes in -W on addition of the 2 -hydroxyl that are <0.1 kcal/mol (Figure 3B). ddition of the hydroxyl to 6 -D resulted in destabilization of both the W and bps by kcal/mol (Supplementary Figure S3, Supplementary able S1). hese results can help explain the faster rate of W/ exchange observed in the prior RD measurements on rmp substituted 6 - D duplexes (15). In particular, this arises because addition of the 2 -hydroxyl destabilizes both the W and states while having a smaller destabilizing effect on the transition state (Supplementary Figure S3). he above results are consistent with the steric model provided the local conformation at the rmp residue remains B-form. esting this prediction was important given that Downloaded from by guest on 18 ovember 2018

7 ucleic cids Research, 2018, Vol. 46, o D R D m 1 d 6 -R m1d /22 B ω (p.p.m.) / /3 (p.p.m.) d16 1' d10 1' DF 1' d16 4' d10 4' DF 4' d16 3' d10 3' DF 3' ' (p.p.m.) /3 (p.p.m.) 1'=8.4 z 11 4/22 6/ ''=31.4 z d '' (p.p.m.) 11 m 1 d 6 -R m1d10 Figure 2. Removal of the 2 -hydroxyl at a purine nucleotide minimally affects the sugar geometry in 6 -R and does not rescue bp formation. () Secondary structure of 6 -R with the sites of dmp incorporation indicated using black circles, and 6 -D. (B) hemical shift perturbations of the sugar resonances of the dmps in 6 -R relative to the corresponding dmp residues in 6 -D. lso shown are DF predicted (70) chemical shift perturbations for the transition of deoxyribonucleosides from a 2 -endo ( = -100 )toa3 -endo conformation ( = 160 ). he discrepancy with the expected change in the 3 chemical shift is likely due to the exclusion of the 3 -phosphate group from the chemical shift calculations (70). () DQF-OSY spectrum of 6 -R d10 showing the sums of the scalar couplings ( 1 = J J 1-2 and 2 = J J J 2-3 )forthe1 and 2 protons at the d10 residue. he corresponding scalar couplings could not be measured for d16 in 6 -R d16 due to severe overlap of the 1-2 and 1-2 cross peaks in the DQF-OSY spectra. Ranges of 1 and 2 for 3 -endo and 2 -endo deoxyribose are 8 11 z and z, and z and z respectively (72). (D) omparison of 1 1D spectra of the imino region of 6 -R m1d16 (red) and 6 -R m1d10 (blue), with the spectra for 6 -R d16 and 6 -R d10 (black). ll MR spectra were collected in p 5.4, 25 mm al and at 25. prior MR (73,74) and X-ray crystallography (75,76) studies showed that single ribonucleotides in B-D can adopt a local -form conformation in sequence contexts that differ from 6 -D. owever, based on MR chemical shifts (70,77), and sugar coupling constants (J 1-2 )(78,79) we find that in the case of 6 -D, the rmps adopt a 2 - endo pucker, as expected if the local conformation remains B-form on hydroxyl addition (Figure 3 and D). aken together, these results argue against the direct involvement of interactions involving a single 2 -hydroxyl group in determining the bp forming propensity of -R and B- D, consistent with the steric model. Loosening the -form geometry stabilizes a (syn)- + bp in IV-2 R R ccording to the steric model, loosening the -form geometry should help resolve steric clashes involving the syn base and reduce -W in R. We tested this prediction by examining the bp forming propensity of a W bp that is adjacent to a dinucleotide bulge in human immunodeficiency virus type 2 (IV-2) transactivation response element (R) R (Figure 4). ere, the bulge is expected to relax the conformational restraints of the - form helix. In prior studies, we showed that m 1 orm 1 do not form detectable bps in -R duplexes under a wide variety of conditions (15). In contrast, MR analysis of IV-2 R R containing m 1 r26 shows that it forms an m 1 26(syn)-39 + bp. In particular, we observe a downfield shift of the m 1 r26-8 resonance (by 4 ppm) (Figure 4) (15,70) and a strong intra-nucleotide 1-8 OE cross peak (Figure 4D) which indicate that the m 1 r26 base adopts a syn conformation. Furthermore, we also observe a downfield shifted imino proton at 15 ppm corresponding to that is hydrogen bonded to the syn m 1 r26 (Figure 4B) (80,81) and downfield shifted amino protons belonging to protonated 39 (Supplementary Figure S4) (17,80). owever, the m 1 r26-1 resonance is not downfield shifted (Supplementary Figure S4) as would be expected for a bp in D (17), presumably due to the adoption of an altered sugar pucker/ angle in the m 1 r26(syn)-39 + bp (70,77,82,83). 10 Downloaded from by guest on 18 ovember 2018

8 11106 ucleic cids Research, 2018, Vol. 46, o. 20 B D 6 -D d16δ-w (kcal/mol) D r16 6 -D d10 6 -D r '-endo / 4' (p.p.m.) r16 r10 3'-endo / 1' (p.p.m.) D r ' (p.p.m.) 1' (p.p.m.) J 1'-2' =9.7 z r10 J 1'-2' =8.3 z ' (p.p.m.) Figure 3. ddition of a 2 -hydroxyl to purine nucleotides in 6 -D does not affect sugar geometry and minimally impacts bp formation. () Secondary structure of 6 -D with the dmp residues indicated using black circles. he sites of rmp incorporation are 10 and 16. (B) he energetic cost of accommodating bps relative to W ( -W ) in D estimated from melting experiments on 6 -D constructs with and without 1 -methylated purine nucleotides at p 5.4, 25 mm al and 25. Errors in -W were obtained by propagating the errors from triplicate measurements (see Materials and Methods ). () 1 and 4 chemical shifts of the rmps in 6 -D fall in the 2 -endo quadrant of a 1 /4 correlation plot (77). Black lines denote the average chemical shits of helical / residues in R as determined from a survey of the BMRB (77). (D) DQF-OSY spectra showing the 1-2 coupling constants of the rmp sugars in 6 -D (p 5.4, 25 mm al and 25 ). he J 1-2 coupling is 8 10 z and 0 2 z for a 2 -endo and 3 -endo ribose sugar respectively m 1 r IV-2 IV-2 R R m1r B /18/ / / / /3 (p.p.m.) D E 6/8 (p.p.m.) m 1 r /8 (p.p.m.) 5 (p.p.m.) 1' (p.p.m.) m r (p.p.m.) (p.p.m.) Δ -W / Δ 1methyl-W (kcal/mol) IV-2 R r26 6 -R r10 Figure 4. Relaxing the geometric restraints of the -form stabilizes bps relative to W in IV-2 R R. () Secondary structure of IV-2 R and IV-2 R m1r26.(b) Overlayof 1 1D SOFS imino spectra of IV-2 R (black) and IV-2 R m1r26 (red) showing a downfield shifted imino peak corresponding to protonated 39 + that is hydrogen bonded to the syn m 1 r26. () Overlay of the aromatic 2D SQ spectra of IV-2 R (black) and IV-2 R m1r26 (red) showing a downfield shift of 26-8 by 4 ppm on 1 -methylation. (D) he medium and weak intra-nucleotide 1-8 OE cross peaks for m 1 r26 and 43 in IV-2 R m1r26 that are indicative of a syn and anti conformation of the purine base respectively, compared to the 5-6 cross peak of a reference cytosine 19. (E) Energetic cost of forming bps relative to W ( -W ) at the bp in IV-2 R compared to the energetic cost of 1 -methylation ( 1methyl-W ) at the bp in 6 -R (p 5.4, 25 mm al and 25 ). Errors in -W were obtained by propagating the errors from triplicate measurements (see Materials and Methods). ll MR spectra were collected at p 5.8, 25 mm al and 25. Optical melting experiments on m 1 r26 modified, and unmodified IV-2 R yielded a -W value of 1.8 ± 0.5 kcal/mol for the bp (Figure 4E). his is comparable to -W values measured in B-D ( kcal/mol) (15) and is much lower than the 1methyl-W value (5.7 ± 0.1 kcal/mol) measured in 6 -R for the bp (Figure 4E). herefore, loosening the -form geometry leads to stabilization of bps relative to W bps, in line with the steric model. ccommodating a syn purine base is energetically more costly in -R compared to B-D Our results suggest that (syn)- and (syn)- + bps are energetically disfavored in R owing to -form dependent steric clashes of the syn base. owever, based on X- ray crystallography and solution-state MR, purine-purine mismatches form syn-anti bps in both B-D and -R duplexes (33 38). Indeed, based on an analysis of crystal structures in the PDB, we identified 69 syn-anti purine-purine mismatches in duplex D and R out of Downloaded from by guest on 18 ovember 2018

9 ucleic cids Research, 2018, Vol. 46, o a total of 10,000 purine-purine mismatches (Supplementary able S3) that are listed in Supplementary able S4. Examination of the torsion angles shows that in -R, steric clashes between the syn purine base and backbone in these mismatches are alleviated by changing backbone torsion angles and from canonical gauche to non-canonical trans values while retaining a 3 -endo sugar pucker (Figure 5). Such re-arrangements were not observed in B-D (Supplementary Figure S5). Furthermore, all other backbone torsion angles remained similar to those of canonical W bps in D and R helices (Supplementary Figure S5). Prior MR (84) and computational (85) studies have shown that isolated Ps energetically favor gauche and torsions as compared to trans. hus, the backbone conformational changes toward trans needed to accommodate syn purines could represent an additional energetic cost ( syn-anti ) for forming bps in -R compared to B-D. If this were true, replacing a W bp with a mismatch should be more destabilizing for - R as compared to B-D. We chose mismatches as a model system owing to their tendency to adopt well defined syn-anti or anti-syn base pairing geometries (34 37), as opposed to which can adopt alternative conformations such as anti-anti (38,86) or which are not stably hydrogen bonded (87,88). While there have been studies examining the effects of mismatches on duplex stability (89,90), none have systematically compared the effects for the same sequence contexts in D and R. hus, we tested this prediction using - mismatches placed in two different sequence contexts 2 and gc, that have been previously characterized by MR (15) (Figure5B). MR spectra show that the - mismatches form bps (Figure 5) in both gcd and gcr based on the observation of two upfield shifted imino protons characteristic of bps (Figure 5D) (36,37). Because the sequence around the mismatch is symmetric, the MR signals arise from both 4(syn) 13(anti)as well as 4(anti) 13(syn) conformations (Figure 5) that are degenerate with respect to chemical shifts (4(syn) = 13(syn) and 13(anti) = 4(anti)) and are in slow to intermediate exchange on the MR chemical shift timescale as evidenced by line broadening in the aromatic and OESY spectra (Supplementary Figures S6 and B). o simplify spectra and aid the thermodynamic analysis, we introduced m 1 4 so as to specifically stabilize the m 1 4(syn) 13(anti) bpingcrandgcd,as verified based on the m 1-8 chemical shifts (Supplementary Figures S6 and S6D) (70). s expected, in both gcr and gcd, methylation of 4 resulted in disappearance of the (syn)-1 imino resonance owing to its replacement with a methyl group (Figure 5D), while minimally perturbing the other imino resonances. aken together, these results show that it is possible to accommodate a syn guanine base in -R when it is base-paired with another guanine. Based on optical melting experiments, m 1 (syn)- bps destabilized -R duplexes relative to their W counterparts by kcal/mol, which can be compared to kcal/mol in the case of D duplexes (Figure 5E). he greater destabilization of R compared to D is robustly observed in the presence of magnesium D E Δ syn-anti (kcal/mol) P OP1 ά O γ 3' OP1 P O 5/ anti syn /3 (p.p.m.) ' O 1' syn ά-γ άγ flip 2 -D d10 2 -R r10 gcd d4 gcr r4 B / / gcd gcr / / D 2 -R O O 1' 1' anti anti gcd m 1 d anti syn /3 (p.p.m.) F Δ syn-anti (kcal/mol) gcd d4 gcr r4 Δ syn-anti (kcal/mol) D d10 2 -R r10 Δ syn-anti (kcal/mol) O syn 1' gcr m 1 r4 2 -D d10 2 -R r10 Figure 5. he energetic cost of accommodating the syn base contributes to bp instability in R. () onformational changes in the and backbone torsions involved in accommodating the steric clash (red dashes) with the syn purine in purine-purine bps in -R (PDB ID: 4Y6). (B) Secondary structures of D and R constructs (gc and 2 sequence contexts) containing mismatches. he residue that is mutated to a is indicated in red, while the position of m 1 substitution (in the context of a mismatch) is denoted in blue. he dmp residues are indicated using black circles. () Dynamic exchange of mismatches between syn anti and anti syn geometries in duplex D and R. (D) 1 1D MR spectra of the imino region of gcd and gcr duplexes with mismatches (p 5.4, 25 mm al and 10 ). For gcr, we observe multiple imino peaks corresponding to the bps neighboring the mismatch that are in slow exchange on the chemical shift timescale. (E ) he energetic cost of accommodating a syn base in R and D ( syn-anti ) estimated from optical melting measurements of constructs with m 1 - mismatches and - bps at (E) moderate salt (p 5.4, 150 mm al and 25 ), (F) in the presence of magnesium (p 5.4, 150 mm al, 3 mm Mgl 2 and 25 ), and () low salt conditions (p 5.4, 25 mm al and 25 ). Estimates of syn-anti for 2 -D and 2 - R that were obtained by comparing unmethylated mismatches with their counterparts are shown in (, p 5.4, 25 mm al and 25 ). Errors in syn-anti were obtained by propagating the errors for the individual samples from triplicate measurements (see Materials and Methods ). Downloaded from by guest on 18 ovember 2018

10 11108 ucleic cids Research, 2018, Vol. 46, o. 20 (Figure 5F), under low salt conditions (Figure 5), for unmodified mismatches (Figure 5), in the presence of potassium (Supplementary able S2) and for alternative sequence contexts (Supplementary Figure S6E). onsistent with the adoption of an energetically unfavorable backbone conformation, the increased destabilization in the case of R is seen to be enthalpically driven (Supplementary able S2). Furthermore, the energetic cost of the transition as obtained from calculations on model compounds kcal/mol (85) is in reasonable agreement with the differences in energetic stabilities of m 1 (syn)-(anti) mismatches relative to W bps ( syn anti = (5.8 7)-(3 4.3) = kcal/mol) in -R versus B-D measured using optical melting experiments (Figure 5E, Supplementary able S2). hese results are also consistent with the observation that purine nucleotides with 3 -endo sugars have a reduced tendency to adopt a syn conformation of the base (30 32). aken together, our results indicate that accommodation of thesyn purine accounts for kcal/mol destabilization of purine-pyrimidine and purine-purine bps in -R compared to B-D. Movement of bases to form h-bonds is energetically more costly in -R compared to B-D he above results show that the steric clashes involving syn purines in -R can be resolved through changes in the backbone and torsion angles in the context of purinepurine mismatches. In principle, similar conformational adjustments could be used to accommodate m 1 r(syn)- and m 1 r(syn)- + bps in -R. owever, unlike - mismatches, m 1 r- and m 1 r- do not form bps in -R, rather they form partially melted conformations generally with an anti-anti geometry (15). his suggests that there may be additional energetic costs in -R for forming purine-pyrimidine versus purine-purine bps. One important distinction between the purine-pyrimidine and purine-purine bps is that in the former, the two bases have to come into closer proximity by 2 Åtoenable the formation of h-bonds (Figure 1B) (16,56). In contrast, for the larger purine-purine mismatches, such a translation of bases is not required (Supplementary Figure S5). We recently visualized the conformational changes in D that drive movement of the bases to form (syn)- bps using MR (55,71). Interestingly, similar conformational changes were also observed in MD simulations of m 1 (syn)- bps in B-D duplexes (55). Since it is not feasible to experimentally measure the energetic cost of moving the bases, we used MD simulations to test whether the translation of the syn adenine and its base pairing partner uridine to form h-bonds is energetically more costly in -R as compared to B-D, and could contribute to bp instability in R. Simulations were initiated from B-D and -R duplexes ( 6 sequence context) containing a single 16(syn)- /9 bp using the MBER simulation package (see Materials and Methods ). We analyzed the 1-1 distance and the stability of the h-bonds throughout the course of the simulation. In particular, a 1 1 distance <9.5 Å, angle for the syn adenine between 0 and 90, and h-bond donor acceptor distances < 3.5 Å(56) were B Probability Probability OP1 O 3 P R '-1' Distance (Å) D O OP de 7-hy 3 Distance (Å) W * de 6-hy O4 Distance (Å) Figure 6. ranslation of the bases to form h-bonds is energetically more costly in -R as compared to B-D. () Steric clashes (red dashes) between the syn purine base and backbone in (syn)- bps are accommodated by changing the sugar pucker away from 3 -endo to 4 -exo, in the MD simulations. (B) istograms of the 1-1 and h-bond distances for W and 16(syn)-/9 bps in 6 -D (bsc0) and 6 -R obtained from MD simulations. and * refer to independent simulations of 6 -D and 6 -R with different starting geometries in which the 16(syn)-/9 bp forms a bp and is in a * conformation (in which the syn adenine and the complementary / are not in base pairing proximity) respectively. Shown in inset to the 1 1 distance panel are representative structural snapshots of the 16(syn)-/9 bp obtained from the simulations. used to define the formation of a bp. Similar results were also obtained when an angle cutoff (hydrogen donor acceptor angle < 30 ) was additionally used to define the formation of a hydrogen bond (Supplementary able S6). In the case of D, the bp remained stably formed (Figure 6B, Supplementary ables S6 and S7, Supplementary Figure S8) for >90% of the simulation time in all three force fields tested (bsc0, bsc1 and OL15). ransitions to W bps were not observed during the simulations, which were short ( 1 s) compared to the lifetime (0.1 1 ms) of the bp in duplex D (17,18). Interestingly, the bp in the R simulations did adopt hydrogen bonded conformations, in which the steric clashes with the syn purine base were resolved through changes in sugar pucker (Figure 6), unlike the changes in the and torsions observed in crystal structures of purine-purine mismatches (Figure 5). lthough we cannot rule out that this could be caused due to the tendency of the MBER force fields to destabilize trans - conformations (91), this suggests that there could be multiple ways of sterically accommodating the syn purine base in R. Movement of the bases to form h-bonds (Supplementary Figure S9) in 6 -R was accompanied by localized changes in sugar pucker (away from 3 -endo to 4 -exo), torsion angle (towards 180 )ofthesyn purine nucleotide, torsion angle of its 3 neighbor (Supplemen- Downloaded from by guest on 18 ovember 2018

11 ucleic cids Research, 2018, Vol. 46, o tary Figure S9B), and over-twisting of the helix about the bp (by 6 relative to the W bp, Supplementary Figure S9). hese characteristics of the bp in -R obtained from the MD simulations are subtly different from the changes accompanying bp formation in B-D (55,71), which involve changes in the sugar pucker, ε and torsion angles of the syn purine and its 5 neighbor, along with under-twisting and major groove directed kinking of thehelixatthebp. owever in 6 -R, only 40% of conformations sampled during the course of the simulation have bases with constricted 1-1 distances that are positioned for hydrogen bonding (Supplementary able S6). For the remainder of the time, the bases splayed apart to adopt conformations in which the 1-1 distance is no longer constricted, and the adenine 7-uracil 3 (7 3 3) h-bond is broken (Figure 6B inset, Supplementary ables S6 and S7). hese conformations no longer feature changes in the and torsions, and sugar pucker, that are required to move the bases/constrict the backbone for pairing in R (Supplementary Figure S9B). Based on the simulations, the energetic cost to move the bases to form (syn)-/type h-bonds is estimated to be higher for -R as compared to B-D by constrict = 0.3 ( 3.1) = 3.4 kcal/mol (Supplementary able S6). In contrast, the stability of the h-bonds was found to be similar for mismatches in B-D and -R (Supplementary Figure S9D), suggesting that the instability of (syn)- bps in the simulations of R is not an artifact purely caused due to the presenceof a syn base. Qualitatively similar results were also obtained with simulations of 6 -D and 6 -R containing (syn)- + bps and 2 -D and 2 -R containing (syn)- bps respectively; the constriction of the backbone was seen to be energetically more costly in R as compared to D by constrict = 2.1 ( 1.9) = 4kcal/mol and constrict = 0.2 ( 2.8) = 3kcal/mol (Supplementary able S6). If constriction of the backbone to form h-bonds is energetically more costly in -R as compared to B-D, one might expect that - wobble mismatches, which also require closer proximity of bases relative to W bps, would also be more destabilizing in -R as compared to - mismatches in B-D, assuming that the absence of the methyl group on racil in R does not contribute to the energetics of the process. Indeed, based on an analysis of nearest neighbor thermodynamic parameters (89,90), replacing a W bp with a wobble results in a greater degree of destabilization of -R ( kcal/mol) as compared to replacing the W bp with a wobble mismatch in B-D ( kcal/mol) (Supplementary Figure S9E). he difference in free energy (1.8 ± 0.6 kcal/mol) is in reasonable agreement with the added energetic cost of constricting the backbone to form h-bonds in -R compared to B-D calculated using MD ( kcal/mol). DISSSIO Our results indicate that the -form geometry and not the 2 -hydroxyl at the purine nucleotide, is primarily responsible for the greater instability of bps relative to W bps in -R as compared to B-D. lthough direct influences on bp stability due to a single hydroxyl group in R are small, the combined presence of multiple hydroxyl groups on both strands indirectly suppresses bps by enforcing the -form helical geometry. In sharp contrast, by loosening the helical geometry of the -form, we successfully induced a stable m 1 (syn)- + bp in R (Figure 4). o our knowledge, this represents the first observation of a purine-pyrimidine bp in helical R under solution conditions. Our results also suggest that there are two contributions to the energetic penalty ( -W ) for forming bps in -R as compared to B-D. he first is the added energy cost of kcal/mol ( syn anti ) associated with conformational changes in the and torsion angles (Figure 5) or sugar pucker (Figure 6) needed to resolve steric clashes with the syn purine base (Figure 7). We used differences in the relative stabilities of - W and m 1 (syn)- bps between R and D to estimate this energetic term. his assumes that the syn purine base adopts similar conformations in purine-purine and purine-pyrimidine bps (prior to backbone constriction) in both B-D and -R. comparison of angles of syn purines in purine-purine mismatches, and in (syn)- + as well as (syn)- bps in D shows that this is indeed the case (Supplementary Figure S5). Furthermore, it is also assumed that the effects of replacing the that is paired with the (in W bps) with a (in bps) are similar between R and D. Interestingly, an analysis of purine-purine mismatches in the PDB (see Materials and Methods ) suggests that this replacement might be accompanied by a loss of stacking interactions in the case of R compared to D (Supplementary Figure S6F) and may also contribute to their instability as compared to D. dditional studies are needed to dissect how differences in stacking interactions contribute to the energetic cost of accommodating a syn base. hese results generalize the finding that bps are energetically disfavored in - R compared to B-D by including purine-purine mismatches even though the differences in energetic cost are lower compared to purine-pyrimidine bps. Based on MD simulations, a second contribution is an added energetic penalty of 3 4 kcal/mol ( constrict )applicable only for purine-pyrimidine bps (in R as compared to D) that is associated with translation of the bases to form h-bonds (Figure 7). Purine-purine bps which do not require movements of the bases (Supplementary Figure S5) would not need to pay this energetic cost. oupled with the potentially higher energy of alternative conformations such as anti anti, this rationalizes why - mismatches form bps in -R whereas m 1 - and m 1 - adopt unstably paired conformations with lack of base pairing around the site of incorporation (15). lthough the instability of the bp was evident in the simulations of R, they were unable to capture the experimentally observed instability of W bps neighboring the site of 1 -methylation (15), potentially due to the occurrence of opening events on time scales longer than those employed in the simulations (Supplementary Figure S10). Interestingly, with the exception of a couple of recent MR studies (36,37), prior MR studies reported that Downloaded from by guest on 18 ovember 2018

12 11110 ucleic cids Research, 2018, Vol. 46, o. 20 χ ~ (+)3 (+)4 kcal/mol syn flip ζ ε ε ζ P P D ~ (-)3 (-)2 kcal/mol constrict P χ ~ (+)6 (+)7 kcal/mol syn flip β P α R ~ 0 (+)2 kcal/mol constrict Figure 7. Mechanism of bp accommodation in D and R. he energetic cost for forming bps can be decomposed into contributions from flipping the base from an anti (W) to a syn conformation followed by movements of the bases to form type h-bonds. Both steps are energetically more costly in R as compared to D. ote that the syn base in R could also be accommodated by changing the and torsions to trans. he geometric changes accompanying bp formation in D are based on MR studies (55,71), while those in -R are based on MD simulations in this study. that - mismatches adopt alternative anti-anti and trans hydrogen bonded syn-anti conformations (Supplementary Figure S6) (92 95). he alternative bp geometries proposed in the prior studies likely arose from misinterpretation of weak intra-nucleotide 1-8 cross peak intensities as being indicative of an anti conformation for the base (93,94) without accounting for broadening due to conformational exchange between syn-anti and anti-syn conformations. Furthermore, the two upfield shifted imino proton signals (Figure 5D) were misinterpreted as arising from a trans hydrogen bonded bp conformation (Supplementary Figure S6) in which both the imino protons are hydrogen bonded to the O6 base atoms (92,95), without considering the possibility that the imino peak from a syn base that is exposed to solvent can also be observed as reported for bps (96 98). In contrast, more recent studies that concluded a bp geometry consistent with crystallographic studies of mismatches (34,35) (Figure 5), employed sequence contexts that are less prone to line broadening due to conformational exchange (36) or utilized 8-bromo guanine substitutions to stabilize the syn conformation of the guanine base (37). By using carbon chemical shifts and m 1 substitutions to minimize conformational exchange and line broadening, we have also circumvented the abovementioned complications, and obtained definitive evidence for (syn) (anti) bps in dynamic equilibrium. Based on the total energetic cost for accommodating the syn purine and translation of the bases to form h-bonds, we estimate that (syn)- and (syn)- + bps are disfavored in -R compared to B-D by a combined amount ( -W ) of kcal/mol. his exceeds the value of 1methyl-W ( kcal/mol) obtained from optical melting experiments on 1 -methylated purines in -R and B-D (15). his discrepancy can be rationalized by the observation that the 1 -methylated purines in R do not form bps, but rather adopt partially melted anti-anti conformations at least in the case of m 1 (15). hus, melting measurements on these constructs likely only provide a lower limit for the destabilization of the state in -R compared to B-D. herefore, instead of paying the energetic costs of flipping the base to syn and constricting the backbone, 1 -methylated purines in R prefer to adopt alternative anti-anti conformations (15) possibly stabilized by a single h-bond and more optimal stacking interactions. dditional experiments are needed to examine the robustness of the magnitudes of the obtained energies under more extensive experimental conditions such as higher p and salt concentrations. he results from this study have important biological implications. Firstly, our results provide key insights into the origins of the destabilizing effects of 1 -methylated purines in R, which constitute an important form of post-transcriptional regulation (21). We find that m 1 and m 1 destabilize R duplexes in a context dependent manner. hey are most destabilizing for W bps within the interior of -form helices wherein base pairing is disrupted completely (15). he destabilizing effects are weaker for W bps near junctions (Figure 4E), and other mismatches (Figure 5, Supplementary able S2), where they can be accommodated via non-canonical base pairing modes. he low propensity for -R to accommodate syn purine bases has also been proposed to play key roles in governing the propensity of R to form -quadruplexes with parallel topologies that exclusively contain anti bases in the 4 tetrad region, as opposed to those that are anti-parallel which contain a mixture of syn and anti guanines (99). Secondly, they provide new insights into how dmp and rmp substitutions affect the structures of -R and B-D. Such substitutions occur in nature as D and R polymerases misincorporate rps and dps respectively (13,14). Prior MR studies indicated that isolated rmps in D duplexes adopt a 3 -endo sugar Downloaded from by guest on 18 ovember 2018