Structural and biochemical impact of C8-aryl-guanine adducts within the NarI recognition DNA sequence: influence of aryl ring size on targeted and semi-targeted mutagenicity

Chemical mutagens with an aromatic ring system may be enzymatically transformed to afford aryl radical species that preferentially react at the C8-site of 2′-deoxyguanosine (dG). The resulting carbon-linked C8-aryl-dG adduct possesses altered biophysical and genetic coding properties compared to the precursor nucleoside. Described herein are structural and in vitro mutagenicity studies of a series of fluorescent C8-aryl-dG analogues that differ in aryl ring size and are representative of authentic DNA adducts. These structural mimics have been inserted into a hotspot sequence for frameshift mutations, namely, the reiterated G3-position of the NarI sequence within 12mer (NarI(12)) and 22mer (NarI(22)) oligonucleotides. In the NarI(12) duplexes, the C8-aryl-dG adducts display a preference for adopting an anti-conformation opposite C, despite the strong syn preference of the free nucleoside. Using the NarI(22) sequence as a template for DNA synthesis in vitro, mutagenicity of the C8-aryl-dG adducts was assayed with representative high-fidelity replicative versus lesion bypass Y-family DNA polymerases, namely, Escherichia coli pol I Klenow fragment exo− (Kf−) and Sulfolobus solfataricus P2 DNA polymerase IV (Dpo4). Our experiments provide a basis for a model involving a two-base slippage and subsequent realignment process to relate the miscoding properties of C-linked C8-aryl-dG adducts with their chemical structures.


S2
. Selected B3LYP/6-31G(d) bond lengths (Å) and dihedral angles (χ and θ, deg.) for the minima and transition states of the PhG adduct. (ZPE corrected relative energies from B3LYP/6-311+G(2df,p) single-point calculations provided in parenthesis, kJ mol -1 ). Figure S2. Selected B3LYP/6-31G(d) bond lengths (Å) and dihedral angles (χ and θ, deg.) for the minima and transition states of the QG adduct. (ZPE corrected relative energies from B3LYP/6-311+G(2df,p) single-point calculations provided in parenthesis, kJ mol -1 ). a Upon introduction of the C8-quinoyl moiety to DNA to incorporate anti  QG ( ~ 0) against C or G using Gaussview, large steric clashes occurred, which yielded highly distorted structures after energy minimization. For this reason, simulations for these conformations were carried out by rebuilding the initial structure by changing  ~ 180 to  ~ 0 in the structure from the last frame of the simulation of the respective anti  QG ( ~ 180) conformation, while retaining the  dihedral angle.  Figure S3: Initial (energy minimized) structures used for MD simulations of the FurG adduct in the NarI helix paired against (a) the complementary C and (b) a G mismatch.

S4
S7 Figure S4: Initial (energy minimized) structures used for MD simulations of the PhG adduct in the NarI helix paired against (a) the complementary C and (b) a G mismatch.
S8 Figure S5: Initial (energy minimized) structures used for MD simulations of the CNPhG adduct in the NarI helix paired against (a) the complementary C and (b) a G mismatch. Figure S6: Initial (energy minimized) structures used for MD simulations of the QG adduct in the NarI helix paired against (a) the complementary C and (b) a G mismatch. S10 Figure S7: Initial (energy minimized) structures used for MD simulations of the (a-d) FurG, (e and f) PhG and (g and h) CNPhG adducts in the NarI helix paired against THF. S11 Figure S8: Initial (energy minimized) structures used for MD simulations of the (a-d) FurG, (e and f) PhG and (g and h) CNPhG adducts in the NarI helix paired against -2 base deletion. S12 Figure S9: Initial (energy minimized) structures used for MD simulations of the QG adduct in the NarI helix paired against (a) THF and (b) -2 base deletion. S13 Figure S10: Radar plots for the percent distribution of the χ (deg.) and θ (deg.) dihedral angles throughout the 40 ns trajectories for the FurG adduct paired in the syn (red) and anti (blue) conformations against complementary C (a-d) and G mismatch (e-h). Figures a, b, e and f correspond to simulations with  ~ 0, whereas figures c, d, g and h correspond to simulations with  ~ 180. S14 Figure S11: Radar plots for the percent distribution of the χ (deg.) and θ (deg.) dihedral angles throughout the 40 ns trajectories for the FurG adduct paired in the syn (red) and anti (blue) conformations against complementary THF (a-d) and -2 base deletion (e-h). Figures a, b, e and f correspond to simulations with  ~ 0, whereas figures c, d, g and h correspond to simulations with  ~ 180. S15 Figure S12: Radar plots for the percent distribution of the χ (deg.) and θ (deg.) dihedral angles throughout the 40 ns trajectories for the PhG adduct paired in its syn (red) and anti (blue) conformations against complementary C (a and b) and G mismatch (c and d). S16 Figure S13: Radar plots for the percent distribution of the χ (deg.) and θ (deg.) dihedral angles throughout the 40 ns trajectories for the PhG adduct paired in its syn (red) and anti (blue) conformations against THF (a and b) and -2 base deletion (c and d). S17 Figure S14: Radar plots for the percent distribution of the χ (deg.) and θ (deg.) dihedral angles throughout the 40 ns trajectories for the CNPhG adduct paired in its syn (red) and anti (blue) conformations against complementary C (a and b) and G mismatch (c and d).

S9
S18 Figure S15: Radar plots for the percent distribution of the χ (deg.) and θ (deg.) dihedral angles throughout the 40 ns trajectories for the CNPhG adduct paired in its syn (red) and anti (blue) conformations against THF (a and b) and -2 base deletion (c and d). S19 Figure S16: Radar plots for the percent distribution of the χ (deg.) and θ (deg.) dihedral angles throughout the 40 ns trajectories for the QG adduct paired in its syn (red) and anti (blue) conformations against complementary C (a-d) and G mismatch (e-h). Figures a, b, e and f correspond to simulations with  ~ 0, whereas figures c, d, g and h correspond to simulations with  ~ 180. S20 Figure S17: Radar plots for the percent distribution of the χ (deg.) and θ (deg.) dihedral angles throughout the 40 ns trajectories for the QG adduct paired in its syn (red) and anti (blue) conformations against THF (a-d) and -2 base deletion (e-h). Figures a, b, e and f correspond to simulations with  ~ 0, whereas figures c, d, g and h correspond to simulations with  ~ 180. Table S2. Occupancies for the hydrogen bonds between the adduct and the opposing base for the studied adduct conformation(s) over the duration of the MD simulations a adduct conformation opposite base conformation bond % occupancy antiFurG (~0) anti-C N2-H(FurG)…O2 (C) 99.9 N1-H(FurG)…N3 (C) 99.9

S21
anti-C N2-H(QG)…O2(C) 99.9 N1-H(QG)…N3 (C) 99.9 N4-H(C)…O6(QG) 98.4 antiQG (~180) anti-C N2-H(QG)…O2(C) 99.9 N1-H(QG)…N3 (C) 99.9  Figure S18. Portions of representative structures showing the orientation of the FurG adduct placed at G 3 position and paired against cytosine in the NarI recognition sequence. The hydrogen bonding pattern of the adduct is also shown explicitly for each of the structures. The corresponding free energy rankings are given in parentheses (bold) against each structure. Figure S19. Portions of representative structures showing the orientation of the FurG adduct placed at G 3 position and paired against guanine in the NarI recognition sequence. The hydrogen bonding pattern of the adduct is also shown explicitly for each of the structures. The corresponding free energy rankings are given in parentheses (bold) against each structure. Figure S20. Portions of representative structures showing the orientation of the FurG adduct placed at G 3 position and paired against THF in the NarI recognition sequence. The corresponding free energy rankings are given in parentheses (bold) against each structure. Figure S21. Portions of representative structures showing the orientation of the FurG adduct placed at G 3 position and paired against -2 base deletion in the NarI recognition sequence. The corresponding free energy rankings are given in parentheses (bold) against each structure. Figure S22. Portions of representative structures showing the orientation of the PhG adduct placed at G 3 position and paired against (a) cytosine and (b) guanine in the NarI recognition sequence. The hydrogen bonding pattern of the adduct is also shown explicitly for each of the structures. The corresponding free energy rankings are given in parentheses (bold) against each structure.

S27
S28 Figure S23. Portions of representative structures showing the orientation of the PhG adduct placed at G 3 position and paired against THF in the NarI recognition sequence. The corresponding free energy rankings are given in parentheses (bold) against each structure.
S29 Figure S24. Portions of representative structures showing the orientation of the CNPhG adduct placed at G 3 position and paired against (a) C and (b) G in the NarI recognition sequence. The hydrogen bonding pattern of the adduct is also shown explicitly for each of the structures. The corresponding free energy rankings are given in parentheses (bold) against each structure.
S30 Figure S25. Portions of representative structures showing the orientation of the CNPhG adduct placed at G 3 position and paired against THF in the NarI recognition sequence. The corresponding free energy rankings are given in parentheses (bold) against each structure. Figure S26. Portions of representative structures showing the orientation of the QG adduct placed at G 3 position and paired against C in the NarI recognition sequence. The hydrogen bonding pattern of the adduct is also shown explicitly for each of the structures. The corresponding free energy rankings are given in parentheses (bold) against each structure. Figure S27. Portions of representative structures showing the orientation of the QG adduct placed at G 3 position and paired against G in the NarI recognition sequence. The hydrogen bonding pattern of the adduct is also shown explicitly for each of the structures. The corresponding free energy rankings are given in parentheses (bold) against each structure. Figure S28. Portions of representative structures showing the orientation of the QG adduct placed at G 3 position and paired against THF in the NarI recognition sequence. The corresponding free energy rankings are given in parentheses (bold) against each structure.

S33
S34 Figure S29. Portions of representative structures showing the orientation of the QG adduct placed at G 3 position and paired against -2 base deletion in the NarI recognition sequence. The corresponding free energy rankings are given in parentheses (bold) against each structure.  a total energy calculated using the molecular mechanics force field and Poisson-Boltzmann's approach. b entropy term calculated using normal mode analysis. c Total free energy, G= E  TS. d Relative free energy indicating the lowest energy structure from simulations with the particular adduct paired opposite a particular nucleobase (C or G).

S35
S37 Figure S30. Backbone RMSD for MD simulations on the studied adducts in different conformations in the NarI duplex against cytosine and guanine.
S38 Figure S31. Backbone RMSD for MD simulations on the studied adducts in different conformations in the NarI duplex against THF and -2 base deletion.

S33. Detailed Results from MD Simulations
Adducts paired against C. Against C, initial structures for the anti-conformations of the adducts were built to allow the Watson-Crick base pairing. Two possible orientations of the C8moiety with respect to the nucleobase were considered for FurG and QG, which roughly correspond to  ~ 0 or 180 and exhibit deviations from planarity of up to 20 to eliminate steric clashes with the DNA backbone (Table S1 and Figures S3S6). Thus, the starting antiorientations position the C8-moiety in the major groove, but differ in solvent exposure of the bulky substituent. Since the C8-moiety is symmetrical in the case of PhG and CNPhG (i.e.,  ~ 0 and  ~180 correspond to the same structure), only one  orientation was considered.
The initial models for the syn conformations of the adducts were also built by considering two orientations of the C8-moiety ( ~0 and ~180), with deviations from planarity ranging between 0 and 41 in order to eliminate steric clashes with the backbone (Table S1 and Figures S5-S8).
Adducts in the syn orientation can give rise to two major conformations, namely the stacked (S) and the wedged (W) form. The W-type conformation positions the C8-moiety in the minor groove, while the S-type conformation intercalates the C8-moiety between the neighboring base pairs. The two -orientations used in the initial models of the syn conformations ensured that both W-and S-type structures were considered.

Analysis of the MD trajectories reveals that the anti-orientation of all adducts maintains
Watson-Crick H-bonding with the complementary C (more than 98% occupancy of H-bonds; S41   Table S2), which places the C8-moiety in the major groove (Figures S18a,b,S22a,b,S24a,b and S26a,b). Over the course of the simulations, the  dihedral angle of the adducts remains in the range of an anti conformation ( ~ 220  250 for FurG, PhG and CNPhG, and  ~ 200  250 for QG), whereas  deviates from planarity by a maximum of 40 (Figures S10, S12, S14 and S16). However, depending on the size of the C8-moiety and the orientation of the -dihedral angle, the syn-conformations adopt different structural characteristics.
In the case of FurG, the ( ~ 0) syn conformation displays a bifurcated Hoogsteen H-bond involving the amino group of the opposing C (more than 70% occupancy, Table S2), which locates the C8-moiety in the minor groove ( Figure S18c). The  value for this syn conformation remains in the range of 40 -80, and  ranges between 0 and 10 ( Figure S10). In the second ( ~ 180) syn conformation, such Hoogsteen H-bonds are prevented by H-bonding between N4 of C and the oxygen atom of the furyl ring (~32% occupancy; Table S2). Despite the loss of Hoogsteen H-bonding, the C8-moiety stacks poorly with the neighboring bases, and remains in an extrahelical position in the minor groove ( Figure S18d). The  values remain between 30 and 50, and  deviates from planarity by a maximum of 20 over the course of the simulation ( Figure S10d).
In contrast to FurG, the syn-conformations of PhG and CNPhG exhibit bimodal distributions with respect to , with two peaks at ~ 40 and ~ 80 (Figures S12a and S14a  Table S2). In addition, the structure mostly remains in a W-type conformation with poor stacking between the phenyl or cyanophenyl moiety and the neighboring base pairs.
The bulkier QG adduct can adopt both W-and S-type structures in the syn orientation. When syn  QdG adopts  ~ 0, a W-type structure is found that contains two Hoogsteen H-bonds between N7 and O6 of QG and the amino group of the opposing C (77% and 53% occupancy respectively; Table S2 and Figure S26c). On the other hand, syn  QG with  ~ 180 yields a basedisplaced intercalated structure ( Figure S26d). Both the W-and S-structures exhibit a unimodal distribution with respect to the  and  dihedral angles of QG, where the W-structure mostly adopts  values ranging from 60 -80 and the S-type structures adopt  values between 30 and 60 throughout the simulation ( Figure S16). However, the C8-moiety is more planar in the Sconformation ( deviates from planarity by less than 10; Figure S16d) in order to maximize stacking with the neighboring bases. In contrast,  deviates from planarity by up to 20 -30 in the W-structure ( Figure S16b), where the increased flexibility arises since the C8-moiety is located in the minor groove.
The calculated free energies ( Adducts paired against G. In order to understand the structural characteristics of guanine mismatch stabilization by the C-linked adducts, adducts in the syn orientation were paired against S43 anti-G, while the antiorientations of the adducts were paired against syn-G. Initial structures of the adducts paired against G were prepared from the corresponding initial structures of the adducts paired against C by replacing C with G and slightly adjusting  and  of the adduct to allow H-bonding interactions with the opposing G. Details of the initial structures used in the simulations are provided (Table S1 and Figures S3b, S4b, S5b and S6b).
In the ( ~ 0 and 180) anti conformations, the FurG adduct forms Hoogsteen H-bonds with the opposing syn  G, where the N2-H(FurG)...N7 H-bond persists for ~97% of the simulation time ( Figure S19a,b and Table S2). In addition, two relatively flexible H-bonds between N1 of FurG and the O6 and N7 atoms of G (45-62% occupancy) stabilize these structures. Overall, the  dihedral angle remains in the range of an anti conformation, while  deviates from planarity by less than 10 in both simulations ( Figures S10e-h). Similarly, the anti conformations of the PhG and CNPhG adducts also displays Hoogsteen H-bonding (see Table S2 for occupancies and Figures S22c, S23c for representative structures), where  ranges from 220 -240, and  deviates from planarity by up to 20 (Figures S12c,d and S14c,d). However, in its anti conformation, the QG adduct exhibits a bimodal distribution with respect to  and  when  ~ 0, and a unimodal distribution with respect to both dihedral angles when  ~ 180 ( Figure S16).
The representative structures from the bimodal distribution at  ~ 0 are related by a 20 flip of  from planarity in opposite orientations, and differ with respect to  by ~50 (Figures S16e,f and S27a). However, both structures display only one interbase H-bond involving N2 of QG and O6 of G, which persists for ~90% of the total simulation time (Table S2). On the other hand, the structure at  ~ 180 displays a bifurcated Hoogsteen H-bonding between O6 of G, and the N2 and N1 donors of QG (~93% and ~56% occupancy respectively, Table S2). In this structure,  ranges from 260 -280, and  ranges from 160 -180 (Figures S16g,h and S27b).

S44
The syn conformation of FurG with initial  ~ 0 displays a unimodal distribution with respect to  and , while the conformation with initial  ~ 180 yields a bimodal distribution with respect to  (where one of the peaks is dominant) and a unimodal distribution with respect to  ( Figure S10e-h). The  ~ 0 structure possesses (two) bifurcated Hoogsteen H-bonds involving O6 of guanine and N1 and N2 donors of FurG (~72% occupancy, Table S2) and a weaker interaction between the amino group of G and N7 of FurG (~55% occupancy, Table S2) due to steric hindrance of the CH groups of the furyl moiety and the amino group of opposing G ( Figure   S19c). On the other hand, the simulation with initial  ~ 180 displays two persistent Hoogsteen H-bonds (more than 81% occupancy, Table S2). One of the clusters corresponding to bimodal  distribution obtained from this simulation exhibits greater buckling of the FurG:G Hoogsteen base pair, which is responsible for the difference in  between the two clusters ( Figure S19d).
On the other hand, PhG adopts two syn conformations that differ in their glycosidic torsion angle ( ~ 40 and  ~ 80) and H-bonding interactions. The conformation at  ~ 80 maintains Hoogsteen H-bonding between PhG or CNPhG and the opposing anti  G. In contrast, the orientation at  ~ 40 weakens H-bonding between the amino group of G and N7 of PhG or CNPhG, which locates the C8-phenyl group towards the interior of the helix and enhances stacking interactions with the 5-neighboring base pair ( Figure S22d). Whereas the conformation at  ~ 80 is dominant in case of PhG, CNPhG prefers the conformation at  ~ 40 (Figures S12c and S14c). This is the reason why the occupancies of Hoogsteen H-bonds are greater in PhG than CNPhG (Table S2), which directly affects their anti/syn free energy differences (Table   S3).
The QG adduct also acquires two syn conformations (corresponding to  ~ 0 and  ~180), which have different H-bonding patterns ( Figure S27), but a similar range of  values and S45 unimodal distributions with respect to  and  ( Figure S16e-h). The structure at  ~ 0 displays a similar Hoogsteen H-bonding pattern as FurG, where the bifurcated H-bonding between O6 of the QG and N2 and N1 of G persists for more than 72% of the simulation time and the H-bond between the amino group of G and N7 of the adduct persists for ~56% of the simulation time (Table S2 and Figure S27c). However, an additional H-bond between the nitrogen atom of the quinolyl ring and the amino group of the opposing G is formed when  ~ 180 ( Figure S27d), which persists for ~93% of the simulation time (Table S2). This weakens the N1-H(G)...O6(QG) H-bond (~31% occupancy), and leads to a bifurcated H-bonding pattern between the amino group of the opposing G (~93% and ~99% occupancy of H-bonds involving O6 and nitrogen atom of quinolyl ring of QG, respectively). The calculated free energies indicate all four adducts prefer a syn conformation when paired opposite a G mismatch by 24 to 113 kJ mol 1 depending on the identity of the C8-substituent (Table S3).
Adducts paired against THF. In order to understand the tendency of the adducts to intercalate within the DNA helices, the adducts in their syn and anti conformations were paired against an abasic site. The tetrahydrofuran moiety, which excludes the -OH group present in the abasic site, was used as a model for mimicking the abasic site residue. Initial structures of the adducts paired against abasic site were prepared from the corresponding initial structures of the adducts paired against C by removing cytosine nucleobase moiety from the nucleotide opposing the adduct and substituting it with a hydrogen atom. In addition, wherever required,  and  dihedral angles were adjusted to allow optimal interactions. Details of the initial structures used in the simulations are provided in Tables S1 and Figures S7 and S9a.
In the anti conformations of all the adducts (which include both  ~ 0 and 180 conformations of FurG and QG), the C8-subsituent remains in the major groove, and has little possibility of S46 forming additional stacking interactions with the flanking bases (Figures S20a,b,S23a,S25a and S28a,b). Over the course of simulations, all the adducts acquire unimodal distributions with respect to  and , indicating that only one conformation is dominant for all the adducts ( Figures   S11a-d, S13a,b, S15a,b and S17a-d). On the other hand, when paired in their syn conformations against THF, the adducts have greater propensity to intercalate their C8-substituents between the flanking base pairs within the DNA helix (Figures S20c,d,S23b,S25b and S28a,b). All the adducts acquire unimodal distribution with respect to  and  over the course of simulations (Figures S11a-d, S13a,b, S15a,b and S17a-d). The calculated relative free energies for different conformations indicate that all the adducts prefer syn conformations against THF, where the synpreference mainly arises due to additional stacking interactions with the flanking bases.
Adducts paired against −2. In order to analyse the tendency of the adducts to stabilize -2 base mutations within the DNA helices, the DNA strands containing the adducts in their syn and anti conformations were paired against 10-mer strands, thereby introducing a 2-base bulge. The initial structures for the simulations were built based on the NMR structure of C8-bonded aminofluorene-dG adduct (PDB code: 1AX6) by removing the C8-moiety are replacing it with Fur, Ph, CNPh or Q. Details of the initial structures used in the simulations are provided in Table   S1 and Figures S8 and S9b.
In one of its anti-conformations ( ~ 0), FurG is stabilized through H-bonding between the cytosine present on the opposite strand and 5 to the bulge. However, this interaction disrupts the H-bonding between this cytosine and its complementary guanine (orange bases in Figure S21a).
On the other hand, the structure at  ~ 180 retains the base pair 5 to the bulge (orange base pair in Figure S21b), and stabilizes the adduct through additional stacking interactions. Similarly, stabilizing stacking interactions are also observed in one of the syn conformations ( ~ 0) of the S47 adduct ( Figure S21c). However, the adducted nucleotide stacks poorly in the syn conformation at  ~ 180 ( Figure S21d). Overall, the anti conformation ( ~ 0) which stabilize the adduct through H-bonding interactions rank lower in energy than the anti-( ~ 180) and syn-( ~ 0) conformations (by 4-14 kJ mol -1 ), where the adduct is stabilized through stacking interactions.
However, with poor stacking interactions, the syn conformation at  ~ 180 ranks highest in energy (Table S3).
The anti conformation of PhG adduct against -2 deletion is similar to the anti conformation of where the adduct is stabilized through stacking interaction with the flanking base pairs. Overall, the anti conformation of PhG is more stable against the -2 deletion, where anti/syn energy difference is smaller than FurG (Table S3). Although the anti conformation of the CNPhG adduct is also stabilized through H-bonding with the cytosine at the same position, the syn conformation attains greater stability due to penetration of the unpaired cytosine present 5 to the adduct into the helix. This provides additional stacking stabilization to the adduct, which results in the syn conformation being 14 kJ mol -1 more stable than the corresponding anti conformation.
In case of QG, the anti conformation of the adduct at  ~ 0 neither forms H-bonding interactions with any of the base on the opposite strand nor stacks with the bases on the same strand ( Figure S29a). On the other hand, the anti conformation at  ~ 180 is stabilized through Watson-Crick H-bonding between the adduct and the cytosine present on the opposite strand and 5 to the bulge ( Figure S29b). However, such a pairing also disrupts the other base pairing interaction 5 to the lesion site (e.g. see interaction between orange and pink bases in Figure   S48 S29b, which disrupts the normal Watson-Crick pairing between two pink bases). In contrast, both the syn conformations ( ~ 0 and  ~ 180) stabilize the adduct by stacking the quinolyl moiety within the helix. Whereas the simulation at  ~ 0 exhibit a unimodal distribution with respect to  and  dihedral angles, the structure at  ~ 180 exhibits a bimodal distribution with respect to  and  ( Figure S17). The two clusters of the bimodal distribution differ in the extent of stacking between the quinolyl moiety and the flanking bases. Overall, the syn conformations of QG are lower in energy than the anti conformations due to additional stacking interactions provided by the bulky C8-substituent. Specifically, the syn conformation at  ~ 0 is the overall lowest energy conformation, due to most optimal stacking interaction between the adduct and the DNA helix. populated. This CT state is not emissive in water as it undergoes dynamic quenching. Therefore, locally excited (LE) emission is observed at 407 nm, which is attributed to the relatively small population of excited state QG that did not undergo CT processes and equilibrate to the lower energy CT state ( Figure S34b). In polar aprotic CH 3 CN the perpendicular geometry of D •+ -A •is stabilized in the polar solvent; the CT emission is not quenched in aprotic solvents as it was in water, and dual emission is observed from LE and CT ( Figure S34c). When QG was excited in CHCl 3 an emissive CT state at 468 nm was observed (dotted green trace in Figure S34a). We

S50
propose that the emission resulted from a charge-separated planar D + =A -CT state that is lower in energy than either the perpendicular CT or the LE state. Figure S35. Fluorescence emission spectra of C8-aryl-G modified NarI (12)

S35. Crystal structure analysis of QG monohydrate
A crystal of QG (pale-yellow thick plate, 0.4x0.4x0.25 mm) was selected from a bulk crystalline product isolated from water-acetonitrile. The crystal was mounted on a Mitegen probe and studied at 150K on a SuperNova Agilent single-crystal diffractometer equipped with a microfocus CuK  ( = 1.54184 Å) radiation source and Atlas CCD detector. Diffraction intensity data were collected using -scan to the maximum 2 angle of 148.7 o (resolution of 0.8Å), with the redundancy factor of 16. The unit cell parameters were refined using the entire data set.
The data were processed using CrysAlisPro software [A]. Absorption corrections were applied using the multiscan method. The structure was solved (direct methods) and refined (full-matrix least-squares on F 2 ) using SHELXL-97 [B]. Non-hydrogen atoms were refined anisotropically. Hydrogen atoms were refined isotropically; initially they were introduced at calculated positions but later their coordinates were refined. Geometric calculations were carried out using the WinGX [C] and Olex [D] software packages.
Crystal data and experimental parameters are summarized in Table S4. Full data for the studied structure, including a CIF-file, have been deposited with the Cambridge Crystallographic Data Centre (no. 1013599) and a copy of these data are available free of charge upon request from the CCDC web-site: http://www.ccdc.cam.ac.uk/data_request/cif or by e-mail: deposit@ccdc.cam.ac.uk.