What makes thienoguanosine an outstanding fluorescent DNA probe?

Abstract

Thienoguanosine (^thG) is an isomorphic guanosine (G) surrogate that almost perfectly mimics G in nucleic acids. To exploit its full potential and lay the foundation for future applications, twenty DNA duplexes, where the bases facing and neighboring ^thG were systematically varied, were thoroughly studied using fluorescence spectroscopy, molecular dynamics simulations and mixed quantum mechanical/molecular mechanics calculations, yielding a comprehensive understanding of its photophysics in DNA. In matched duplexes, ^thG’s hypochromism was larger for flanking G/C residues but its fluorescence quantum yield (QY) and lifetime values were almost independent of the flanking bases. This was attributed to high duplex stability, which maintains a steady orientation and distance between nucleobases, so that a similar charge transfer (CT) mechanism governs the photophysics of ^thG independently of its flanking nucleobases. ^thG can therefore replace any G residue in matched duplexes, while always maintaining similar photophysical features. In contrast, the local destabilization induced by a mismatch or an abasic site restores a strong dependence of ^thG’s QY and lifetime values on its environmental context, depending on the CT route efficiency and solvent exposure of ^thG. Due to this exquisite sensitivity, ^thG appears ideal for monitoring local structural changes and single nucleotide polymorphism. Moreover, ^thG’s dominant fluorescence lifetime in DNA is unusually long (9–29 ns), facilitating its selective measurement in complex media using a lifetime-based or a time-gated detection scheme. Taken together, our data highlights ^thG as an outstanding emissive substitute for G with good QY, long fluorescence lifetimes and exquisite sensitivity to local structural changes.

Graphical Abstract

1. Introduction

DNA molecules are fundamental actors in living organisms as they encode genetic information needed for development, function and reproduction. Such processes rely on interactions with a large set of proteins that can impact the structure, conformation, bending or dynamics of the DNA double helix. These effects can be localized, as in base-flipping and bending events, or affect the DNA molecules globally, when compacted by histones or unwound by helicases or nucleic acid chaperones. To interrogate the molecular mechanisms that govern such protein-mediated variations, fluorescence techniques are a tool of choice due to their high sensitivity and exquisite spatio-temporal resolution. However, as the natural nucleobases are essentially non-fluorescent, there is a need for suitable emissive nucleobase analogues, which are able to seamlessly substitute the natural nucleobases while displaying bright and environmentally sensitive emission. The recently developed thienoguanosine (^thG), an isomorphic guanosine (G) surrogate, appears particularly promising in this respect, almost perfectly replacing the natural G in both matched and mismatched duplexes.^1,2 Moreover, unlike many other fluorescent nucleobase analogues, ^thG is not strongly quenched when incorporated into DNA oligonucleotides and can thus be used to selectively and faithfully monitor the conformation and dynamics of a given G residue in a DNA sequence³. The only anticipated limitation of ^thG may be its imperfect ability to form higher DNA structures (i.e., triplexes, quadruplexes) due to its lack of basic nitrogen corresponding to N7 in the purine skeleton.

Importantly, ^thG exhibits rich information content, as it can exist in two tautomeric forms (i.e. H1 and H3 keto-amino tautomers) at neutral pH, exhibiting considerably shifted absorption and emission spectra⁴ and different fluorescent lifetimes^5,6. The equilibrium between these tautomers informs on the local environment of ^thG. Due to its unique properties, ^thG was successfully used to monitor the local conformational changes associated to the binding of the HIV-1 nucleocapsid protein NCp7 to the (−)Primer Binding Site (PBS) and to decipher the mechanism of the NCp7-promoted (+)/(−)PBS annealing reaction⁷. The fluorescent nucleoside analogue in its deoxyribo or in its ribo form has also been successfully used to monitor B to Z transitions of DNA², single nucleotide polymorphism⁷, distance measurements in DNA^8,9, base flipping reactions¹⁰, ribozyme-mediated cleavage¹¹, and cellular activity of siRNAs¹².

Most of the applications of ^thG reported so far rely on empirical changes in fluorescence intensity and thus, do not exploit the full potential of this information-rich probe. The main reason is that a full understanding of ^thG photophysical properties, especially when ^thG is included in DNA, is still missing. Such an understanding would be useful for extracting additional information about any conformational changes or protein binding events by interpreting the observed changes in the photophysics on the basis of structural and electronic effects. For instance, this knowledge would be useful to explain the previously observed 2.5-fold difference in the quantum yield (QY) and mean fluorescence lifetime between corresponding matched and mismatched ^thG-labeled duplexes³. This large change in the fluorescence properties highlights the exquisite sensitivity of ^thG to its environment. Its understanding would be critical for gaining insight into the influencing factors. Such knowledge would also be very useful for laying the foundations for future applications of this probe and, in particular, for selecting the appropriate sequences for optimal spectroscopic changes in the presence of a given protein or other external effectors and ligands.

To augment the understanding of the photophysics of ^thG-labeled duplexes and, notably, their dependence on the DNA sequence context, we investigate herein the steady-state and time-resolved fluorescence properties of ^thG-labeled DNA duplexes, where we systematically vary the composition of neighboring and opposite bases of ^thG. As a model DNA duplex, we use the 18-mer (+)/(−)PBS duplex in which the G7 residue of (−)PBS was replaced by ^thG (Figure 1). The experimental data were complemented by molecular dynamics (MD) simulations to explore the structural behaviour of the investigated duplexes, as well as by mixed quantum mechanical/molecular mechanics (QM/MM) calculations to map the main photophysical processes in each duplex. Taken together, our integrated analysis provides a rationale for the observed spectroscopic properties and their dependence on ^thG’s flanking or opposite nucleobases, disentangling the different factors modulating the photophysical features of ^thG.

Figure 1:

Sequences of the ^thG-labeled (−)/(+)PBS duplexes used in this study. In all duplexes, the G7 residue of (−)PBS was replaced by the fluorescent analogue ^thG. To investigate the sequence context dependence of the fluorescence properties of ^thG, its flanking and opposite bases were systematically varied. Ab designates an abasic site, which corresponds to a site where the base is missing, but the sugar is present. The natural flanking T residues were systematically changed by two similar residues, and the opposite bases were concomitantly replaced by their matched Watson-Crick residues to minimize duplex destabilization. The duplexes were named using an X^thGX/YZY triplet code comprising ^thG and its flanking residues and their opposite nucleobases. For example, the wild-type (−)/(+)PBS duplex is encoded as the T^thGT/ACA duplex.

2. Materials and Methods

(−)PBS DNA sequences labelled by thienoguanosine (^thG) at 7^th position (Figure 1) were purchased from TriLink Biotechnologies (USA). The complementary matched and mismatched (+)PBS DNA sequences were purchased from IBA Nucleic Acids Product Supply (Germany). Note, a stable tetrahydrofuran residue was used as a substitute for a “true” apurinic site. This commonly used substitution has been demonstrated to be a structurally and functionally viable surrogate^13–17. ^thG-labeled (−)/(+)PBS duplexes were obtained by heating 10 μM ^thG7(−)PBS and 30 μM (+)PBS at 85 °C for 3 min and then, slowly cooling them down to room temperature. All experiments were done in 25 mM TRIS-HCl buffer pH=7.5, 30 mM NaCl and 0.2 mM MgCl₂.

Absorption and fluorescence measurements

Absorption spectra were collected on a Cary 4000 UV-visible spectrophotometer (Varian). Fluorescence spectra were recorded at 20°C on a FluoroLog spectrofluorometer (Jobin Yvon) equipped with a thermostated cell compartment. Matched and mismatched ^thG-labeled duplexes were excited at 315 nm and 360 nm wavelengths for steady-state fluorescence studies. The spectra were corrected for buffer fluorescence, lamp fluctuations, and detector spectral sensitivity. Quantum yields (QYs) of the ^thG-labeled duplexes were determined using quinine sulphate (QY=0.546 in 0.5 M H₂SO₄) as a reference¹⁸. Time-resolved fluorescence measurements were performed with the time-correlated single-photon counting technique¹⁹. Excitation wavelength at 360 nm was generated by a supercontinuum laser (NKT Photonics SuperK Extreme) with 10 MHz repetition rate. The fluorescence emission was collected at different wavelengths across the spectrum through a polarizer set at the magic angle and a 16 mm bandpass monochromator (Jobin Yvon). The single-photon events were detected with a micro-channel plate photomultiplier R3809U Hamamatsu coupled to a pulse pre-amplifier HFAC (Becker-Hickl GmbH) and recorded on a time-correlated single photon counting board SPC-130 (Becker-Hickl GmbH). The instrumental response function (IRF) recorded with a polished aluminium reflector was fitted with a Gaussian function (FWHM=50 ps). The measured fluorescence decays were first deconvoluted with the instrumental response function and fitted using the maximum entropy method^20,21 to assess the number of fluorescence lifetimes in the decays. Then, time-resolved exponential decays were fitted by a global analysis algorithm. The fitting function was a multi-exponential decay convolved with a normalized Gaussian curve of standard deviation σ standing for the temporal IRF and a Heavyside function. The fitting function was built in Igor Pro (Wavemetrics). All emission decays were fitted using a weighting that corresponds to the standard deviation of the photon number squared root. The lifetimes were shared for all emission wavelengths while the amplitudes were kept as free parameters during the fitting (Levenberg–Marquardt algorithm). A very good match was observed for the lifetime values obtained by the maximum entropy method and the global analysis. The results of the global fitting were used to construct the decay associated spectra (DAS) using: I_i(λ)=I(λ)α_i(λ)τ_i/Σα_i(λ)τ_i, where I(λ) is the steady-state emission spectrum and α_i (λ) are the wavelength-dependent amplitudes. The DAS provide the spectra associated to each lifetime component, and thus indicate the contribution in intensity of each species to the whole emission of the duplex. The real amplitudes associated to each lifetime component were calculated from: $A_i=\frac{\int I_i(\lambda )d\lambda }{\int I(\lambda )d\lambda }$, where the numerator is the area under the DAS associated to the species i and the denominator is the area under the steady-state emission spectrum.

Melting temperatures

Melting curves were recorded by monitoring the changes in the absorbance at 260 nm of solutions containing final concentrations of 2 μM (−)PBS and 2.2 μM (+)PBS in in 25 mM TRIS-HCl buffer pH=7.5, 30 mM NaCl and 0.2 mM MgCl₂. The two sequences were annealed by a heating/cooling cycle prior to the measurement of the thermal denaturation. The melting curves were recorded in a 40–90°C range on a Cary 4000 spectrophotometer (Varian/Agilent) equipped with a Peltier thermostatted cell holder. The heating speed was 1°C/min. 200 μL of mineral oil were added on the top of the DNA solutions to avoid evaporation. Melting curves were converted into a plot of α versus temperature, where α represents the fraction of single strands. The melting temperatures were extracted from these curves by calculating their first derivative.

The melting temperatures of the different duplexes in their unlabelled form were also predicted by using the DINAMelt web server (http://unafold.rna.albany.edu/). The concentrations of PBS sequences and salt concentrations used for the predictions corresponded to the experimental ones.

Molecular dynamics simulations

DNA duplexes were built by the nucleic acid builder (NAB) molecular manipulation language²², while parameters for ^thG and other MD settings were retrieved from a previous work⁴. Briefly, MD simulations were performed with AMBER 16²³, using the OL15 force field for DNA and the GAFF2 force field for ^thG. A box of TIP3P type water molecules buffering 8 Å from the solute’s surface was built, while Na⁺ ions were added to achieve the system neutrality. Analysis of MD trajectories was carried out by cpptraj²⁴, while the calculation of the delta energy of binding between the two strands in the duplex was performed with the MMPBSA.py script²⁵. To check the conformational and thermodynamic stability of matched and mismatched (+)/(−)PBS duplexes, MD simulations were carried out on the most representative systems. All the systems were submitted to 200 ns of unrestrained MD production in explicit solvent, while analysis was carried out at the conformational (cluster analysis) and thermodynamic (calculation of the delta energy of binding) levels. Two replicas of MD simulations were carried out for all matched and mismatched duplexes. For systems showing flipped out ^thG residues, a third replica was carried out as well. The main atoms forming the nucleobase cycles were considered in the definition of the center of mass, which was used in the measurements of base-base distances along MD trajectories.

QM/MM calculations

Computational details:

The ground- and excited-state minima of all the duplexes have been optimized by QM/MM calculations, using the ONIOM²⁶ interface as implemented in Gaussian09²⁷. Density Functional Theory (DFT) and its time-dependent version (TD-DFT), adopting M052X^28,29 functional and 6–31G(d) basis set, was selected as QM method. This functional was shown to well reproduce the tautomeric equilibrium in ^thG and the related absorption spectra, with the exception of a small blue shift (0.2–0.3 eV)⁵. Furthermore, M052X can reliably treat dispersion interactions and electronic transitions with Charge Transfer (CT) character²⁹, making it suitable for the study of a strongly coupled multi-chromophore system as DNA, as shown by many previous studies³⁰. We chose the Amber parm96.dat Force Field for MM calculations³¹. Bulk solvent (water) effect has been included (in both QM and MM parts) by the Polarizable Continuum Model (PCM)^32,33. Ground- and excited-state minima have been optimized by using the Linear Response version of the PCM, for which analytic gradients are available. We instead resorted to State Specific (SS-PCM)/TD-DFT calculations to provide a more accurate estimate of electronic transitions with significant CT character^34,35. Further analysis concerning excited-state densities, molecular charges, and energetics were done considering a simplified model including only the four bases which were considered at the QM level and including environmental effects by PCM. In this latter case, however, we have used the dielectric constant of CHCl₃, 4.8, which can better mimic that experienced by a base within a duplex^36,37.

Computational models:

Our computational analysis has been performed on a model including 8 bases (Figure S1), with the phospho-deoxyribose backbone and Na⁺ counterions, namely bases 5, 6, 7 (^thG), 8 and the corresponding bases 5’−8’ of the complementary strand. In our computations, four bases have been treated at the QM level, whereas the remaining ones, together with the phospho-deoxyribose backbone and Na⁺ counterions, were described at the MM level. For both matched and mismatched systems, the QM region includes ^thG at position 7 and its Watson-Crick partner as well as the stacked base pair closest to ^thG, according to MD simulations. A stacked dimer appears the minimal model that, though being still computationally tractable, is able to provide a qualitative correct understanding of DNA photoactivated dynamics. Actually, there are convergent experimental and computational indications that, though the spectroscopic states in absorption can be delocalized over multiple bases, the main photophysical and photochemical processes in DNA involve one base or two stacked bases^38,39. Nevertheless, the choice of the ‘stacked’ partner of ^thG to be included in the QM part is an additional source of ambiguity. To address this question, two models were considered for the matched T^thGT/ACA duplex where ^thG is flanked by T residues. In model 1, T6 and A6’ (in addition to ^thG7 and C7’) are included in the QM region whereas T8 and A8’ are in the MM region. On the opposite, in model 2, T8 and A8’ are instead included in the QM region whereas T6 and A6’ are in the MM region. In this last case, the model has been extended to include also bases 9 and 9’. Since T6 and T8 are not symmetrically stacked to ^thG, in this way, we can get insights on the effect of conformational heterogeneity on the spectral properties of ^thG.

Starting from the most representative clusters issuing from MD simulations, we have optimized the ground state QM/MM minima for all models. In a second step, we have characterized the lower energy excited states in the Franck-Condon (FC) region and their Potential Energy Surfaces at the PCM/TD-M052X/6–31G(d)/MM level. Our re-optimization procedure, necessary in order to avoid any artefact due to the use of cluster average geometry and the MM field in the MD simulations, is also prone to predict a too close stacking geometry between ^thG and the adjacent stacked residue treated at the QM level. Two limitations are the main responsible of this result. First, the other base stacked to ^thG is treated at the MM level, favouring a closer approach to that described at the QM level. Second, the effects of conformational fluctuations of the DNA are underestimated (a single minimum is analysed).

3. Results

To explore the photophysics of ^thG in DNA oligonucleotides and its dependence on sequence context, the 18-mer (+)/(−)PBS duplex was used, where the G7 residue in (−)PBS was replaced with the isomorphic emissive nucleoside ^thG, and the flanking T6 and T8 residues were systematically replaced by A, C or G residues (Figure 1). The complementary residues on (+)PBS were changed accordingly to maintain Watson-Crick pairing. We also explored the impact of the base opposite to ^thG, by replacing its complementary C residue in (+)PBS by A, T, G or an abasic residue, thus intentionally generating a mismatch. In total, the absorption and emission properties of 20 (+)/(−)PBS duplexes, including 4 fully matched duplexes and 16 mismatched duplexes, were analysed.

3.1. Matched Sequences

3.1.1. Spectral characterization

The absorption spectra of all matched duplexes (with C opposite to ^thG) were recorded (Figure 2A). As the main absorption band of ^thG is red-shifted compared to the natural bases, it could be compared to the spectrum of ^thG H1 tautomer, which has previously been shown to be the only tautomer present in matched (+)/(−)PBS duplexes⁴. A slight red-shift and a large decrease in the extinction coefficient was observed for ^thG in matched duplexes, as compared to the free ^thG H1 tautomer (Figure 2A, green spectrum). This strong hypochromic effect (29–41%, Table 1) clearly suggests that Watson-Crick pairing with a complementary C residue enables optimal stacking of ^thG with the flanking nucleobases⁴⁰, especially when the ^thG–C base pair was flanked with another G–C base pairs (Figure 2A, blue and orange spectra).

Figure 2:

Absorption spectra of ^thG-labelled matched (A) and mismatched (B-E) (+)/(−)PBS duplexes. In the mismatched duplexes, ^thG was opposite G (B), A (C), T (D) or an abasic site (E). In each panel, the neighboring residues of ^thG were systematically varied. The concentration of all duplexes was 10 μM. The absorption spectrum of the H1 tautomer of free ^thG at the same concentration in water is included for comparison (green spectra). All experiments were done in 25 mM TRIS-HCl buffer (pH=7.5), 30 mM NaCl and 0.2 mM MgCl₂.

Table 1:

Photophysical data of ^thG labelled (−)/(+) PBS matched duplexes

(−)PBS	(+)PBS	Hypochromism (%)a	QYb	τ1 (ns)c	α1d	f1e	τ2 (ns)c	α2d	f2e	<τ> (ns)	kr (107×S−1)	knr (107×S−1)
AthGA	TCT	32	0.18	1.7	0.17	0.03	11.9	0.83	0.97	10.1	1.8	8.1
TthGT	ACA	29	0.15	2.4	0.14	0.04	11.1	0.86	0.96	9.9	1.5	8.6
CthGC	GCG	41	0.16	3.4	0.31	0.13	10.6	0.69	0.87	8.4	1.9	10.0
GthGG	CCC	41	0.15	4.4	0.39	0.19	12.3	0.61	0.81	9.2	1.6	9.2
thG H1 in waterf			0.51				20.5				2.49	2.39
thG H1 in MeOHf			0.42				14.5				2.9	4.0

Standard Deviation (SD)= ^a ± 2%; ^b ± 0.02; ^c ± 0.1 ― 0.3 ns; ^d ± 0.01 ― 0.05. ^eData from Martinez-Fernandez et al, 2019⁵. Excitation wavelength was 360 nm. The amplitudes, α_i, are calculated from the integrated areas under the DAS of each lifetime component normalized with respect to the total emitted intensity. The fractional intensities were calculated by f_i = α_iτ_i/<τ>. The radiative and non-radiative rate constants were calculated by: k_r = QY/<τ> and k_nr = 1/<τ>−k_r, respectively.

Emission spectra were recorded upon excitation at 360 nm (where only the H1 tautomer absorbs) and 315 nm (where both the H1 and H3 tautomers absorb) (Figure S2). When normalized, the emission spectra of all matched duplexes perfectly overlapped at the two excitation wavelengths, confirming that the H3 tautomer was essentially absent in these duplexes⁴. The emission maxima of the duplexes are blue-shifted by ~10 nm (457–463 nm, Table S1) as compared to the free ^thG H1 tautomer (470 nm) in water⁴ and are thus comparable to that the free ^thG H1 in methanol (458 nm). Interestingly, the duplexes exhibited rather low QYs (0.15–0.18, Table 1) almost independent on the nature of the flanking residues, in agreement with previous data on a 12 bp duplex, where ^thG at different positions (and thus, with different neighbors) was found to exhibit similar QY values¹⁰. The emission spectra of the matched duplexes were also checked to be independent on the NaCl concentration in the 30 – 150 mM range (Fig. S3A).

Next, time-resolved fluorescence decays were recorded for the ^thG-labelled matched duplexes over the entire emission spectrum of ^thG (420–560 nm) to recover the decay associated spectra (DAS) (Figure 3). For all matched duplexes, the ^thG fluorescence decay was well fitted with two lifetime components (Figure S4A). The long-lived lifetime τ₂ (10.6–12.3 ns) was almost independent of the neighboring nucleobases’ identity (Table 1), in line with the observed QYs. The short-lived component τ₁, displayed greater variations, with lower values being associated with neighboring A and T (1.7–2.4 ns), as compared to C and G (3.4–4.4 ns). With respect to amplitudes, the long-lived lifetime, τ₂, was the dominant species in all duplexes, but its contribution was more pronounced with A and T (83–86%) than with C and G (61–69%) as neighboring bases. As a result, the DAS associated with the long-lived lifetime nearly overlapped the steady-state spectrum when ^thG was flanked with A or T (Figure 3A and B), while the short-lived lifetime significantly contributed to the emission of the duplexes when ^thG was flanked with G or C (Figure 3C and D). From the comparison of the QY and mean excited-state lifetimes, the radiative, k_r, and non-radiative, k_nr, rate constants could be calculated. The k_r values (1.5–1.9×10⁷ s⁻¹) of ^thG in the matched duplexes were significantly lower than those of the ^thG H1 monomer in water (2.5×10⁷ s⁻¹) or methanol (2.9×10⁷ s⁻¹). As the k_r values are proportional to the extinction coefficients of the chromophores⁴¹, the decreased k_r values are in line with the hypochromism observed in the absorption spectra (Figure 2A and Table 1). Moreover, our data further revealed that the lower QY values of ^thG in the duplexes as compared to free ^thG H1 in water were mainly due to an about four-fold increase in the k_nr value, suggesting the presence of additional non radiative pathways as compared to the ^thG H1 monomeric nucleoside.

Figure 3:

Deconvolution of the fluorescence spectra of ^thG-labelled matched (+)/(−)PBS duplexes into two spectra associated with the long-lived lifetime component (red) and short-lived lifetime component (blue). The black lines represent the steady-state fluorescence spectra. Excitation wavelength was 360 nm.

3.1.2. MD Analysis

MD simulations on matched duplexes indicated that replacement of the T bases flanking ^thG with A, C, or G had a minimal structural effect, since the global conformation of the duplex as well as the ^thG-C Watson-Crick base pair was preserved (Figures 4 and S5). From the cluster analysis of MD simulations, a single representative cluster containing more than 99% of MD frames was extracted for each duplex.

Figure 4:

Conformational stability of the matched ^thG-labelled (+)/(−)PBS duplexes in MD simulations. In the duplexes, ^thG was flanked either by T (black), A (red), C (blue) or G (green) residues. A) Structural superimposition of the most representative MD frames of (+)/(−)PBS matched duplexes. Only the cartoon representation of the phosphate backbone is shown, ^thG is shown as sticks. B) Magnification of the interaction of the ^thG-C base pair with its flanking base pairs. ^thG is shown as sticks, other nucleotides are shown as lines while the phosphate backbone is shown as a cartoon.

The thermodynamics stability of the matched ^thG-labelled duplexes was estimated by the molecular mechanics Poisson-Boltzmann surface area (MM-PBSA) approach by calculating the delta energy of binding, ΔEb (Table 2), which corresponds to the difference in total energy between the duplexes and their constitutive single-stranded sequences and is thus related to the hybridization enthalpy.⁴² The duplex stability was also characterized by its melting temperature T_m, either determined experimentally with the ^thG-labelled duplexes from the temperature dependence of their absorbance changes at 260 nm (Figure S6) or predicted by DINAMelt⁴³ on the corresponding unlabelled duplexes (Table 2). Interestingly, an excellent match was found between the experimental and the DINAMelt-predicted Tm values, confirming that substitution of G by ^thG does not impact the melting temperature in DNA duplexes.^2,10 The only deviations between experimental and calculated Tm were observed for the duplexes with abasic sites, probably due to the fact that DINAMelt is not optimized for abasic sites.

Table 2.

Thermodynamic stability of ^thG-labeled (−)/(+)PBS duplexes.

	(−)PBS	(+)PBS	ΔEb (kcal/mol) ± SEMa	Tme
				Calc.e	Exp.f
Matched	GthGG	CCC	−75.0 ± 0.6	74.2	75.5
	CthGC	GCG	−71.2 ± 0.7	74.3
	AthGA	TCT	−63.1 ± 0.6	68.2	70
	TthGT	ACA	−71.3 ± 0.8b	69	68.5
Mismatched	GthGG	CTC	−66.8 ± 0.7	68.6
		CAC	−73.1 ± 0.8	69.3	68
		CGC	−61.5 ± 0.7	67.5
		CAbC	−64.3 ± 0.8c	71.1	65.5
	CthGC	GTG	−65.6 ± 0.6	67.6	68
		GAG	−65.7 ± 0.8	64.9	66
		GGG	−59.4 ± 1.0	67.3
		GAbG	−65.2 ± 1c	70.9	64
	AthGA	TTT	−56.2 ± 0.7	60.4
		TAT	−48.7 ± 0.9	62.8
		TGT	−51.1 ± 0.7	62.7	62
		TAbT	−55.8 ± 0.7c	68.7	60
	TthGT	ATA	−59.2 ± 1.0d	62.2	63
		AAA	−46.5 ± 0.7	62.4
		AGA	−48.7 ± 1.0	63.1
		AAbA	−40.2 ± 1.2c	68.9

^aDelta energy of binding, ΔEb, of ^thG-labeled (−)PBS to (+)PBS was calculated by the MM-PBSA approach along MD trajectories.

^bTaken from our previous work⁴.

^cΔEb was calculated by considering explicitly the 15 water molecules closest to ^thG in MD trajectories.

^dThe value refers to the most populated cluster (first group, around 70% of MD frames), while the second group (around 30% of MD frames) was estimated with a theoretical affinity of −51 ± 1 kcal/mol.

^eThe melting temperatures of the unlabeled duplexes were predicted by using the DINAMelt web server (http://unafold.rna.albany.edu/). They correspond to the inflection point of the predicted melting absorbance curves, using the experimental PBS (2 μM) and salt (30 mM NaCl and 0.2 mM MgCl₂) concentrations.

^fExperimental melting temperatures in brackets were determined on the ^thG-labelled duplexes from the first derivative of the melting curves monitored from the absorbance changes at 260 nm (Fig. S6). The experimental Tm values were given for n = 2 as means +/−0.5 °C.

As expected from the higher stability of G-C base pairs as compared to A-T base pairs, the G^thGG/CCC and C^thGC/GCG duplexes were more stable by 5–6°C as compared to the A^thGA/TCT and T^thGT/ACA duplexes (Table 2).

3.1.3. QM/MM calculations

After having optimized the minima for the matched sequences (Figures 5A and S1) at the QM(PCM/M052X/6–31G(d))/MM level, we have computed the 12 lowest energy transitions at the TD-M052X levels (Tables 3 and S2), in order to be sure to include all the main electronic transitions involving ^thG, including the CT ones.

Figure 5: Ground-state geometries (A) and excited-state electronic densities (B) of the matched duplexes.

A) Most relevant geometries for the optimized minima at the QM(PCM/M052X/6–31G(d))/MM level of theory. Nucleobases treated at the QM level are in ball and sticks, whereas the MM system is depicted in lines. Numbering of residues is as in Figure 1. B) Excited-state density difference (S_n-S₀), green means negative difference (i.e from where the electron moves) while purple indicates more density (i.e where the electron moves to).

Table 3: Absorption energies.

Vertical Absorption Energy (VAE) and oscillator strength f of the lowest energy excited-state (S₁), corresponding to a π^thG)→ π*(^thG) transition in the FC region. Values were obtained from LR-PCM/TD-M052X/6–31G(d) excited-state calculations on the structures extracted from QM/MM geometry optimizations. Hypochromism is estimated as the relative difference in % between the oscillator strength computed for free ^thG H1 tautomer and the one for the duplex.

(−)PBS	(+)PBS	VAE (eV)	VAE (nm)	f	Hypochromism
AthGA	TCT	4.30	288	0.11	27%
TthGT	ACA	4.31	287	0.11	27%
CthGC	GCG	4.24	292	0.11	27%
GthGG	CCC	4.37	283	0.11	27%
TthGT	ATA	g14.43	279	0.10	33%
		g24.59	270	0.12	20%
GthGA	CTC	4.24	292	0.07	53%
AthGA	TAT	4.63	268	0.16	0%
TthGT	AAA	g14.44	279	0.07	53%
		g24.56	272	0.10	33%
		g34.61	269	0.14	<1%
TthGT	CGC	4.56	272	0.16	0%
Free thG watera		4.55	273	0.15
thG+2H2Ob		4.44	279	0.14

^aBulk solvent effects are included by PCM.

^bIn addition to bulk solvent, two explicit water molecules were added.

When compared to the free ^thG H1 in water, the predicted moderate red shift of the absorption maximum and strong hypochromicity (~30%) of the transition (Table 3) are in line with the experimental data (Figure 2A and Table 1). Inspection of the lowest energy transitions (Table S2) shows that in addition to the spectroscopic ππ* state, essentially localized on ^thG, a relatively close-lying CT transition (<1 eV less stable than S₁), involving ^thG and a flanking base is always present. However, its direction depends on the nature of the neighbour base, since ^thG serves as electron donor when stacked with C but an electron acceptor when stacked with G (Figure 5B).

In the second step, we have optimized both S₁ and the lowest energy CT transition at the QM/MM level for all matched systems. The main features of the resulting minima are summarized in Table 4.

Table 4. Emission Energies of the matched and mismatched duplexes.

Vertical Emission Energies (VEE) and oscillator strengths (f) at the QM(PCM/M052X/6–31G(d) Linear Response (LR) level of theory are compared to those computed at the State Specific (SS)-PCM. Charge Transfer character (CT in a.u), radiative lifetime (τ_R in ns) and computed fluorescence lifetimes (^c<τ> in ns) of the different minima for the different duplexes.

	(−)PBS	(+)PBS	Description.	LRVEE (eV)	LRf	LRCT	SSVEE (eV)	SSf	SSτR	c<τ>
Match	AthGA	TCT	Min-ππ*	3.28	0.07	0.03	3.20	0.05	74	13.4
			Min-CT (thG→A)	3.77	0.01	0.58	2.53	0.00
	TthGT	ACA	Min-ππ*modell	3.43	0.08	0.01	3.32	0.05	64	9.7
			Min-CTmodel2 (thG→T)	3.34	0.00	0.86	1.68	0.00
	CthGC	GCG	Min-ππ*	3.42	0.08	0.01	3.33	0.06	52	8.3
			Min-CT (thG→C)	a3.88	0.03	0.72	2.10	0.01
	GthGG	CCC	Min-ππ*	3.47	0.08	0.03	3.33	0.05	64	9.6
			Min-CT (G→thG)	3.15	0.02	−0.22	2.80	0.01
Mismatched	GthGG	CTC	Min-ππ*	b
			Min-CT (G→thG)	4.24	0.07	−0.22	2.91	0.01	7000
	AthGA	TAT	Min-ππ*	3.67	0.11	0.03	3.39	0.08	29	14.0
			Min-CT	c
	TthGT	ATA	g1Min-ππ*	3.46	0.07	0.04	3.32	0.05	68	23.7
			g1Min-ct (thG→T)	3.36	0.06	0.06	3.24	0.04
	TthGT	AAA	g3Min-ππ*	3.50	0.08	0.01	3.29	0.06	62	22.8
			g1Min-ct (thG→T)	2.93	0.01	0.16	2.80	0.01
	TthGT	AGA	Min-ππ*	3.63	0.11	0.00	3.33	0.08	28	2.8
			Min-CT	c
thG+water			Min-ππ*	3.60	0.11	-	3.22	0.08	31	15.8
thG+2H2O			Min-ππ*	3.45	0.11	-	3.15	0.08	31	15. 8

Notes: a) No real minimum but a low energy gradient region was found, b) no clear ππ * minimum was found, c) no clear CT minimum was located. The radiative lifetime (τ_R) was estimated by using a simplified version of the Strickler-Berg formula: ${}^{\text{Theor}}{\tau }_{\text{R}}=\frac{1}{{\text{k}}_{\text{R}}}$ with ${\text{k}}_{\text{R}}=\frac{4\Delta {\text{E}}^3}{3{\text{c}}^3}{\mu }^2$, where ΔE is the vertical emission energy and the dipole moment of the transition. The computed lifetime is calculated by: ^c<τ> = k_R×QY, where QY is the experimental quantum yield. Further details on the calculations are given in section 4 of SI.

The high stability of the CT minima strongly suggests that an additional, non-emissive, deactivation channel is available for ^thG within duplexes, in line with the multi-exponential fluorescence decay and the decrease of both fluorescence QY and lifetimes observed experimentally when ^thG is included in DNA. The comparison between the different duplexes indicates that the CT minimum is relatively less stable for the (^thG→A) CT state (Table S3), in line with the slightly higher experimental QY for the matched A^thGA/TCT duplex (Table 1). Only for the G^thGG/CCC duplex, ^thG acts as an electron acceptor in CT, whereas ^thG serves as electron donor for the other systems (Table 4). Our calculations also show that the minimum of the π(^thG)→ π*(^thG) excited state acquires a small CT character (0.01≤^LRCT≤0.03), since some contribution of the stacked bases is found. This leads to a decrease of the oscillator strength and thus, of the emission intensity as well as an increase of the radiative lifetime with respect to ^thG H1 tautomer in water, in line with the experiments (Table 1). Moreover, using the experimental QY values, the mean lifetime values (last column of Table 4) recalculated by the product of QY with the calculated k_r value match very well with the experimental values (Table 1). Comparison of the results obtained for the two models used to study the T^thGT/ACA duplex provides interesting indications on the effect of the variation of the stacking geometry (due for example to conformational fluctuations of the duplex) on the ^thG optical properties. As explained in the Materials and Methods section, the two models differ by the T base (5’ or 3’ to ^thG) included in the QM region. Model 2 monitors the effect of T8 at the 3’ end with respect to ^thG. MD simulations and QM/MM calculations show that T8 is strongly stacked to ^thG (with a closer stacking distance and a larger overlap between the rings), leading to a minimum with strong ^thG⁺→T⁻ CT character. In Model 1, where the focus is on T6, our calculations show that T6 is more loosely stacked with ^thG, leading to an excited-state minimum localized on ^thG. This clearly indicates that the photophysics of ^thG is primarily governed by the interaction with its nearest flanking base which most strongly interacts with ^thG.

3.2. Mismatched duplexes

For all mismatched duplexes, comparison of the spectra at 315 nm and 360 nm excitation wavelengths revealed an only marginal contribution of the H3 tautomer (Figure S7). Its contribution was most prominent with opposite A, but represented less than 10%. Due to its marginal contribution, no accurate analysis of the H3 tautomer could be performed. Therefore, all the measurements in the remaining part of this study were performed at 360 nm, where only the H1 tautomer absorbs. As for the matched duplexes, the spectra of the mismatched duplexes were found to be independent on the NaCl concentration in the 30–150 mM range (Figure S3B and C). For all duplexes, the fluorescence decays were biexponential over the whole spectral range (Table 5, Figure S4B and C). The short-lived lifetimes τ₁ were rather homogeneous (2–5.5 ns) and contributed to less than 10% of the total emission of ^thG (as shown by their f₁ values) in 12 out of the 16 tested duplexes. As a consequence, the emission of the mismatched duplexes was largely dominated by the long-lived lifetimes τ₂, which varied in a large range (9.3 – 28.6 ns) and were therefore mainly discussed in the following. Interestingly, the k_r values were distributed in a narrow range (1.4–2.1×10⁷ s⁻¹) indicating that the QY and <τ> values change in a highly correlated manner and thus, that dark species, with lifetimes under the 20 ps resolution of our set-up, are likely marginal in all duplexes.

Table 5:

Photophysical properties of ^thG-labelled (−)/(+) PBS mismatched duplexes

(−)PBS	(+)PBS	Hypochr omism (%)	QY	τ1 (ns)	α1	f1	τ2 (ns)	α2	f2	<τ> (ns)	kr (107×S−1)	knr (107×S−1)
GthGG	CTC	37	0.10	2.4	0.43	0.16	9.3	0.57	0.84	6.3	1.6	14.2
GthGG	CAC	13	0.17	4.4	0.39	0.18	12.9	0.61	0.82	9.6	1.8	8.7
GthGG	CGC	32	0.23	2.5	0.29	0.05	19.8	0.71	0.95	14.8	1.5	5.2
GthGG	CAbC	28	0.15	4.0	0.49	0.25	11.6	0.51	0.75	7.9	1.9	10.8
CthGC	GTG	24	0.15	2.1	0.42	0.10	13.5	0.58	0.90	8.7	1.7	9.8
CthGC	GAG	15	0.34	2.7	0.19	0.03	18.8	0.81	0.97	15.7	2.2	4.2
CthGC	GGG	30	0.21	3.5	0.26	0.08	14.8	0.74	0.92	11.8	1.8	6.7
CthGC	GAbG	26	0.12	4.7	0.68	0.42	14.5	0.32	0.58	7.8	1.5	11.2
AthGA	TTT	33	0.37	2.3	0.31	0.04	27.3	0.69	0.96	19.6	1.9	3.2
AthGA	TAT	12	0.48	2.8	0.22	0.03	28.6	0.78	0.97	22.9	2.1	2.3
AthGA	TGT	28	0.42	2.3	0.17	0.02	26.9	0.83	0.98	22.7	1.8	2.6
AthGA	TAbT	28	0.47	2.4	0.18	0.02	25.0	0.82	0.98	20.9	2.2	2.5
TthGT	ATA	25	0.35	3.0	0.27	0.04	26.8	0.73	0.96	20.4	1.7	3.1
TthGT	AAA	25	0.42	3.0	0.24	0.04	23.8	0.76	0.96	18.8	2.2	3.1
TthGT	AGA	31	0.33	5.5	0.25	0.09	19.4	0.75	0.91	15.9	2.1	4.2
TthGT	AAbA	23	0.25	3.4	0.28	0.07	16.9	0.72	0.93	13.1	1.9	5.7

All reported values are the means for two to four experiments. The standard errors of the mean of the reported values are 8% for the QY, 10% for hypochromism, ± 0.1 ― 0.3 ns for τ1, ± 0.2 ― 0.8 ns for τ2, < 0.05 for the amplitudes (α_i) and fractional intensities (f_i). The radiative and non-radiative rate constants were calculated as described in Table 1.

3.2.1. Mismatched duplexes with flanking G and C bases

In mismatched duplexes where ^thG is flanked by G residues, the absorption spectra show a strong hypochromism (up to 37%), except for the G^thGG/CAC duplex where ^thG is mismatched with A (13%) (Table 5). Among the mismatched duplexes, those with flanking G residues exhibit the lowest QY, τ₂ and <τ> values, being comparable to the values of the corresponding matched duplex (Table 1). The only exception is for the G^thGG/CGC duplex, showing an only two-fold decrease of its QY as compared to free ^thG H1 in water, and a τ₂ value comparable to the ^thG H1 lifetime in water. Another striking feature is the high amplitude (~ 50%) associated to the short-lived lifetime τ₁ of the G^thGG/CAbC duplex with an abasic site, so that this lifetime strongly contributes to the emission spectra (Figure S8).

The spectroscopic properties of mismatched duplexes with ^thG flanked by C bases are not too different from those with flanking G bases. The observed hypochromism is fairly high (24–30%), with again the exception of the C^thGC/GAG duplex with ^thG opposite A (15%). The QY, τ₂ and <τ> values are comparable to those observed when ^thG is flanked by G, with the exception of the ^thG-A mismatch, where the QY, τ₂ and <τ> values are 1.5 to 2-fold higher for the duplex with flanking C residues. Finally, the short-lived fluorescence lifetime of the duplex with an abasic base was found to exhibit a very high amplitude (68%) and thus, strongly contributes to the overall emission (Figure S9).

To further understand the spectroscopic data of the mismatched duplexes with ^thG flanked by G or C bases, we performed MD and QM/MM calculations. The G^thGG/CTC and C^thGC/GTG duplexes with a ^thG-T mismatch show rather high Tm values (68–69 °C) and stable ΔEb values (−66 kcal/mol), as a result of the Wobble-like conformation adopted by the ^thG-T mismatch (Figure 6A–B). This conformation is stable over time as shown by the RMSD analysis (Figure S10A) and the observation of a single cluster of MD frames with a population over 99%, which is structurally close to the matched duplex (Figure S10B). The similarity in structure and stability of these mismatched duplexes with the matched duplexes is well consistent with the similarity of their spectroscopic properties.

Figure 6. Conformational analysis of mismatched (+)/(−)PBS duplexes where ^thG is flanked by G and C bases.

A) Magnification of the interaction of the ^thG-T base pair with flanking G residues in the G^thGG/CTC duplex. B) Magnification of the interaction of the ^thG-T base pair with flanking C residues in the C^thGC/GTG duplex. C) Magnification of the interaction of the ^thG-G base pair with flanking G residues in the G^thGG/CGC duplex. D) Magnification of the interaction of the ^thG-G base pair with flanking C residues in the C^thGC/GGG duplex. E) Magnification of the interaction of the ^thG-A base pair with flanking G residues in the G^thGG/CAC duplex. F) Magnification of the interaction of the ^thG-A base pair with flanking C residues in the C^thGC/GAG duplex. ^thG is shown as sticks, other nucleotides are shown as lines while the phosphate backbone is shown as cartoon. H-bonds are shown as dashed lines. Mismatched duplexes where ^thG is flanked by G are coloured blue, while systems bearing ^thG flanked by C are coloured green.

The G^thGG/CGC and C^thGC/GGG duplexes with ^thG-G mismatch show lower stability, with Tm ~ 67°C and ΔEb ~ −60 kcal/mol. In contrast to the ^thG-T mismatch, ^thG is unable to establish a Wobble-like pairing with an opposite G. The amino group of ^thG forms an H-bond with the carbonyl group of G, while the remaining portion of ^thG is exposed towards the solvent (Figure 6C–D). By monitoring the distance between the amino group of ^thG and the carbonyl group of opposite G, ^thG is partially flipped out from the duplex for a short period that corresponds to 13% of MD trajectory (Figure S11A). Both the solvent exposure of ^thG and its ability to flip out are well consistent with the τ₂ lifetime value close to that of the free nucleobase in water.

The G^thGG/CAC duplexes show a high stability (Tm = 68–69 °C and ΔEb = −73 kcal/mol), likely due to the ability of ^thG to establish H-bonds with the opposite A (Figure 6E–F) and properly stack with the flanking G bases. As a result, ^thG remains within the duplex for the majority of MD simulation time and its photophysical properties do not deviate too much from those of the corresponding matched duplex. In contrast, there is an additional destabilization of 7.4 kcal/mol and a 4 °C decrease in T_m for the C^thGC/GAG duplex. This additional destabilization likely reduces the base stacking of ^thG with its flanking C bases, explaining the substantial increase in the QY, <τ> and τ₂ values.

Simulation of the duplexes with an abasic site opposite to ^thG using the parameters previously reported by Chen et al.⁴⁴ indicates that the abasic site does not dramatically affect the conformation of ^thG, which is stacked within the duplex by flanking bases and eventually connected to the opposite sugar by water-bridged H-bond interactions (Figure S12A and B). For the G^thGG/CAbC duplex, the distance between the centers of mass of the main rings of ^thG and G6, the closest flanking nucleobase along the MD trajectory was found remarkably constant and similar to the corresponding distance in the matched G^thGG/CCC duplex (Figure S13A), in agreement with the highly similar photophysical properties of ^thG in the two duplexes. Noteworthy, the distance between ^thG and G8 has larger fluctuations and is on the average longer in the mismatched duplex than in the matched duplex. Nevertheless, in a few fluctuations, G8 came very close to ^thG in the mismatched duplex. Interestingly, the mismatched C^thGC/GAbG duplex with flanking C residues was less stable by 1.5 °C than the G^thGG/CAbC duplex with flanking G residues (Table 2). This lower stability is also perceived in the highly fluctuating distances of ^thG with its C6 and C8 neighbors (Figure S13B). These fluctuations, notably with the closest C6 residue, induce an overall increase in the distance with ^thG as compared to the C^thGC/GCG matched duplex, in line with the increase in the τ₂ value observed in the mismatched duplex (compare Tables 5 and 1). Interestingly, the distance fluctuations between C8 and ^thG were very large, occasionally leading to very close proximity.

QM/MM calculations on the G^thGG/CTC duplexes (Figure S1) confirm a strong interaction of ^thG with the close-lying G6 base, leading to a strong hypochromism (Table 3). Excited-state geometry optimizations of S₁ lead to a minimum with a very strong G→^thG CT character (Table 4), suggesting an efficient decay channel through this CT, in line with the low QY (0.10) of these duplexes. The coupling between the bright ^thG ππ* state and the CT state is so large that it has not been possible to locate a ππ* minimum on ^thG. However, as explained in Materials and Methods, our procedure likely overestimates the interaction between the stacked bases considered at the QM level and underestimates the conformational flexibility of the duplexes. This limitation is expected to be more important for mismatched than for matched duplexes. As a consequence, there are probably structures for which the decay to an excited-state localized on ^thG is favoured.

3.2.2. Mismatched duplexes with flanking A bases

In mismatched duplexes, where ^thG is flanked by A nucleobases, the hypochromism with respect to free ^thG was substantial (28–33%), with the exception again of the A^thGA/TAT duplex with a ^thG-A mismatch (12%). All QY values (0.37–0.48) were close to that (0.51) of free ^thG H1 tautomer, indicating that flanking A residues had the lowest quenching potency on ^thG in mismatched duplexes (Table 5). The values of both τ₁ and τ₂ fluorescence lifetimes were rather uniform for the different duplexes with flanking As. Strikingly, the τ₂ values (25–28.6 ns) were substantially higher than the lifetime of free ^thG H1 in water (20.5 ns). This may be mainly attributed to the decrease of k_r from 2.5×10⁷s^–1 for free ^thG H1 in water to ~2 ×10⁷s⁻¹ for ^thG in the duplexes, while the k_nr values were comparable to those of ^thG H1 in water. The τ₁ lifetimes (2.3–2.6 ns) exhibit small amplitudes and thus, contribute only to a very small extent (≤ 4%) to the overall emission of these duplexes (Figure S14). Thus, the lifetime data suggest that only a small ^thG population is quenched by its A neighbors in these mismatched duplexes.

Among the mismatched duplexes, the lowest stability (Tm = 60–62.8 °C and ΔEb between −49 and −56 kcal/mol) was observed when ^thG is flanked by A nucleobases (Table 2). For the A^thGA/TGT and A^thGA/TAT duplexes, MD simulations show that ^thG mostly occupies local single-stranded regions, being paired with the opposite +1 base or temporarily exposed to the solvent. In both cases, stacking interactions of ^thG with flanking base pairs are completely lost (Figure 7A–B), in full line with the high QY, <τ> and τ₂ values observed with these duplexes. When ^thG was mismatched with T in the A^thGA/TTT duplex, it establishes a Wobble-like base pair (Figure 7C), which marginally stabilizes the duplex (T_m = 60.4 °C). In line with this low stability, the distance between the centers of mass of the main rings of ^thG and the closest flanking adenine A6 along the MD trajectory is highly fluctuating (Figure S15A) leading to conformations where A6 is highly shifted in respect to ^thG (Figure S16B). This is in sharp contrast to the A^thGA/TCT matched duplex, where this distance is consistent (Figure S15A), keeping ^thG and A6 in close proximity (Figure S16A). These simulations are in agreement with the high QY, <τ> and τ₂ values observed with the A^thGA/TTT duplex. Low stability (T_m = 60°C) was observed also for the A^thGA/TAbT duplex with abasic site (Figure S12C), although the relative position of ^thG with respect to the flanking bases does not significantly deviate from that measured in the matched duplex (Figure S13C). In this particular case, the explicit water molecules included in the MD simulation may artificially stabilize the duplex structure, which can obscure the fluctuations in distances between ^thG and its neighbors occurring in this less stable duplex.

Figure 7.

Conformational behaviour of ^thG in the A^thGA/TAT (A), A^thGA/TGT (B) and A^thGA/TTT (C) (−)/(+)PBS duplexes. Representative structures from MD simulations are shown. ^thG is shown as sticks, surrounding nucleotides as lines and the phosphate backbone is shown as cartoon. H-bonds are highlighted by black dashed lines. ^thG is shown as sticks, other nucleotides are shown as lines.

QM/MM calculations on the A^thGA/TAT duplex where ^thG is opposite A (Figure S1) confirmed that ^thG is highly exposed to the solvent and does not strongly interact with the close-lying A nucleobases. As a consequence, the computed hypochromism is very small (Table 3) and the S₁ geometry optimization provides a picture similar to that for free ^thG H1, i.e. only a ππ * minimum is optimized, as it can be noted by the estimated radiative lifetime (Table 4).

3.2.3. Mismatched duplexes with flanking T bases

All mismatched sequences with flanking T bases exhibit a substantial hypochromism (23–31%, Table 5). The QY, τ₂ and <τ> values are in between the values observed for duplexes with flanking A residues and those with flanking G or C residues. The effect of flanking T bases on ^thG photophysics is therefore smaller than that of G or C residues and larger than that of A residues (Table 5). The τ₁ values are rather high (3.0–5.5 ns) but contribute only little (4–9%) to the overall emission of these duplexes (Figure S17 and Table 5). Of note, the data of the T^thGT/ATA duplex with opposite T in Table 5 should be considered as a refinement of our data on the same duplex in a previous paper³, which were very similar, except for the presence of small population of dark species. This small population was not found anymore and probably resulted from small errors in the determination of the QY.

As for the mismatched duplexes where ^thG is flanked by A nucleobases, low stability (T_m ≤ 63 °C and ΔEb between −40 and −59 kcal/mol) was observed for the mismatched duplexes with flanking T nucleobases. For the T^thGT/ATA duplex with a ^thG-T mismatch, MD simulation show that ^thG interacts with the mismatched T mostly by a Wobble-like pairing that is persistent in MD trajectories for around 70% of the simulation time (Figure 8A). In this major conformation, ^thG loses partially its stacking interaction with one of the neighbouring bases. The remaining fraction of MD simulation time is represented by a destabilized conformation (Figure 8B) in which ^thG does not interact with the opposite and flanking bases. Similarly, when ^thG is mismatched with A in the T^thGT/AAA duplex it remains within the duplex for around 22.6% of MD simulation time (Figure S11B), while it moves towards the solvent accessible space for 68.5% of MD time (Figure 8C) and flips out from the duplex for the remaining 8.9% of MD simulation time (Figure S11B). In the T^thGT/AGA duplex with a ^thG-G mismatch, ^thG is unable to establish direct interactions with the opposite nucleobase, but remains in the solvent-accessible space within the duplex for most of the MD simulation time (Figure 8D). Finally, for the T^thGT/AAbA duplex, we also observed a significant conformational instability of the duplex within the abasic site, such as underlined by the representative structure (Figure S12D) and the analysis of the distance between ^thG and the flanking T nucleobases with respect to the T^thGT/ACA matched duplex (Figure S13D). Both the loss of stacking interaction as well as the partial or full exposure of ^thG to solvent are well consistent with the high QY, <τ> and τ₂ values observed with these duplexes

Figure 8.

Conformations of ^thG flanked with T residues, as obtained by MD simulations. A) The canonical Wobble-like ^thG-T base pair accounts for around 70% of MD frames (group 1). B) Destabilized conformation of the ^thG-T base pair, which accounts for around 30% of MD frames (group 2). C) Destabilized conformation of the ^thG-A mismatch; D) Destabilized conformation of the ^thG-G mismatch. ^thG is shown as sticks, nucleotides are shown as lines. H-bonds within the ^thG-T base pair are highlighted by black dashed lines.

QM/MM calculations were then performed with the T^thGT/ATA and T^thGT/AAA duplexes. Based on the MD simulation data, we have analysed two families (groups 1 and 2) for the duplexes where ^thG is opposite T and three different conformational families (groups 1, 2, 3) for the duplexes where ^thG is opposite A (Figure S1). These structures differ on the degree of ^thG stacking (Figure S1). As detailed in the SI, both the mismatches with T and A present geometries (groups 1 and 2) where the S₁ excitation is delocalized over the closest base (A/T). The higher the delocalization, the larger is the charge transfer (CT) and the lower is the radiative constant. In the mismatch with A, there is also a monomer-like group 3, where the excitation is localized on ^thG without any coupling to other bases. Finally, for the T^thGT/AGA duplex where ^thG is mismatched with G, ^thG is flipped out into the solvent and does not interact with the surrounding bases (Figure S1) in agreement with MD results. Although no clear CT minimum was located in this case (Table 4), its population cannot be ruled out, since an additional T base (T8, not included in this model) is close enough to provide such kind of minimum, as we found for the T^thGT/ACA matched duplex.

4. Discussion

Being able to faithfully report on the structure and dynamics of G residues in nucleic acids (NAs), ^thG, an environmentally sensitive isomorphic fluorescent analogue of guanosine, has emerged as a major tool for characterizing NA structural changes and interactions with proteins. To further evaluate its potency and better interpret the changes in its spectroscopic properties upon external perturbations, it is essential to fully understand the mechanisms underlying the photophysics of ^thG when incorporated into NAs. Toward this end, we combined experiment and theory to study (−)/(+)PBS DNA duplexes, where the flanking and opposite bases of a central ^thG in the (−)PBS sequence were systematically varied.

A strong hypochromism in ^thG’s absorption spectra was observed for both matched and mismatched sequences (up to 41% and 37%, respectively; Figure 1, Table 1 and 5). As for the natural bases in DNA⁴⁰, this hypochromism is likely due to exciton coupling between ^thG and its flanking bases. The strongest hypochromism (41%) was observed in G^thGG/CCC and C^thGC/GCG matched duplexes, where ^thG was flanked by G-C base pairs. It was substantially lower (~30%) for the A^thGA/TCT and T^thGT/ACA matched duplexes with flanking A-T base pairs. This difference likely results from the stiffness brought about by the three consecutive G-C like pairs stabilized by three hydrogen bonds each. This rigidity likely governs the strength of the stacking interaction of ^thG with the nearby bases, and thus, the extent of hypochromicity. In line with this conclusion, a strong hypochromism (24–37%) was also observed with G^thGG/CTC and C^thGC/GTG duplexes where ^thG forms a wobble base pair with T, stabilized by two hydrogen bonds (Figures 6 and 8A). In contrast, the lowest hypochromicity was observed when ^thG was opposite A, as a result of a steric clash between the two residues, which shifts ^thG toward the solvent (Figure 7A). The absorption spectra computed at the QM/MM calculations are fully consistent with the experimental results, confirming the general reliability of our conclusions.

The QY values in DNA were found to be lower than the QY of free ^thG H1 in water (0.51) and to vary in a large range (0.1–0.48), indicating that the quenching of ^thG in the duplexes is highly context dependent. However, a very different picture emerged for matched and mismatched duplexes. In matched duplexes, the QY values as well as the τ₂ and <τ> values were rather low, and minimally dependent on the nature of the bases flanking ^thG. This minimal dependence of the photophysical features of ^thG on its flanking nucleobases also applies when it is flanked by nucleobases of different nature, as shown by our ab initio calculations (Figure S18A–C) and by the independence of ^thG’s QY when placed in different positions and therefore with different neighbors (G^thGG, C^thGC, G^thGC or C^thGT) in a matched duplex.¹⁰ QM/MM calculations indicated that the matched duplexes were characterized by a relatively close-lying CT transition (<1 eV less stable than S₁) with a strong CT character, involving ^thG and the adjacent base. ^thG in the excited state serves as an electron acceptor when it is flanked by G, whereas it serves as an electron donor for the other bases (Table 4). This CT transition is likely responsible of an additional, non-emissive, deactivation channel in the matched duplexes that is consistent with the measured increase in the k_nr value, responsible of the decrease in QY and lifetime values. This efficient CT transition with all neighboring bases in matched duplexes is correlated with the high stability of these duplexes (Table 2) that places ^thG in the proper orientation and distance to the flanking bases. These conditions of relatively high stability that facilitate efficient CT transition are also met with the G^thGG/CTC and C^thGC/GTG mismatched duplexes where ^thG forms a Wobble base pair with T and is flanked by G-C base pairs (Table 2). In these duplexes, these three base pairs are stabilized by a set of eight H bonds, which probably preserve the same orientation and distance of ^thG with the flanking bases as in the matched duplexes, and thus, yield QY, τ₂ and <τ> values (Table 5) close to those of the matched duplexes (Table 1). Calculations also revealed that the minimum of the π(^thG)→ *(^thG) excited state in matched duplexes acquires a low CT character (Table 4), which reduces its oscillator strength and radiative constant with respect to free ^thG H1, and thus leads to a further decrease in the QY and lifetime values.

In contrast to matched sequences, the values of QY, τ₂ and <τ> in mismatched duplexes are highly dependent on the nature of ^thG’s flanking bases and to a lesser extent on its opposite base. Our experiments indicate that the impact of the flanking bases on ^thG fluorescence varies in the following order: G>C>T>A. The strong quenching potency of G residues is not surprising, as G is generally the strongest quencher among the natural bases for a large number of fluorescent nucleobases^45–48 and fluorophores⁴⁹, due to its high reduction potential and thus, high electron donating properties^49,50. In line with our experiments, computational analysis suggests that both electronic and structural effects modulate the fluorescence of ^thG in DNA. QM calculations indicate that the strongest quenching effects are observed when ^thG is stacked with another base and involves a CT state (e.g. ^thG⁺→Pyrimidine⁻ or G⁺→^thG⁻), leading to an effective non-radiative decay channel (Table 4). Moreover, inspection of the FC region (e.g. energy difference between the lowest energy CT state and the bright state) and analysis of our excited-state geometry optimizations suggest that the coupling between the bright spectroscopic state localized on ^thG with the CT state is larger when the flanking bases are G or C, and much smaller for A. This stronger coupling suggests a more effective decay to CT states for the former two bases. The availability of this additional decay channel obviously depends on the proximity of ^thG to its flanking base, as shown by the excellent correlation between the photophysical properties of ^thG and the distance fluctuations between ^thG and its closest flanking nucleobase over MD time (Figures S13 and S15). Thus, the local stiffness of the duplex and, more generally, its conformational behaviour, also modulate ^thG spectroscopic properties. On this ground, flanking G-C base pairs increase the local stiffness of the duplex, due to their three hydrogen bonds, explaining why flanking C residues are more effective quenchers than flanking T residues.

Interestingly, the τ₂ values of all mismatched duplexes with flanking A residues and two out of the four mismatched duplexes with flanking T residues (23.8–28.6 ns) were much higher than the fluorescence lifetime of ^thG H1 in water (20.5 ns). As fluorescence lifetimes are dictated by 1/(k_r+k_nr), the large τ₂ values can be explained in part by their k_r values, which are lower (1.7–2.2×10⁷ s⁻¹) than that (2.5×10⁷ s⁻¹) of ^thG H1 in water. The real k_r values associated to τ₂ may even be lower, since the k_r values reported in Table 5 were calculated from the mean lifetime, assuming that the k_r values are the same for both τ₁ and τ₂ lifetimes. As the k_r values are proportional to the chromophore’s extinction coefficients,⁵¹ the decreased k_r values can be related to the hypochromism observed for ^thG in all duplexes (Table 1). For all other duplexes, the observed hypochromism also reduces the k_r value in comparison to that of the ^thG H1 monomer, but its effect on the lifetime is masked by the large increase in k_nr due to the CT pathway in these duplexes.

Short-lived τ₁ lifetimes are observed for all the matched and mismatched duplexes, but their amplitudes are systematically higher when ^thG is flanked by G-C as compared to A-T base pairs. The amplitude and thus, the contribution of the τ₁ component to the overall emission is especially high in the highly dynamic context provided by an abasic site, where ^thG can be efficiently quenched by its G or C neighbors. This view is supported by MD simulations on mismatched duplexes with abasic site, showing that the distance between ^thG and its flanking G8 or C8 residues is highly fluctuating, occasionally leading to very short distances favourable for quenching (Figure S13A and B). These data suggest that the τ₁ lifetime might be related to a dynamic quenching of ^thG by CT to its neighbors, modulated by the conformational fluctuations of DNA.

From the complementary base perspective, the poor quenching of ^thG by its flanking bases in mismatches with A or G (Table 5) can be easily explained by the steric clash of ^thG with its opposite purine base, which dislodges the ^thG toward the solvent and prevents its stacking with the flanking bases (Figure 7). In contrast, a close contact and thus a strong quenching of ^thG by flanking C-G residues is observed when ^thG is opposite to a pyrimidine or an abasic site. This can be illustrated by the C^thGC/GTG mismatched duplex where ^thG faces T and is flanked by C residues (Figure 6A and B). In this case, the most representative cluster shows that ^thG is close to both C and G residues on the opposite strand, in line with the observed low QY (0.15).

Overall, our experimental and computational results can be rationalized by a simple model, where the key interactions are those established between ^thG and its most closely stacked neighbours, which depending on the global conformational behaviour of the particular duplex examined, can be one of its flanking residues or less frequently, a base on the opposite strand. Increasing numbers of experimental and computational contributions indicate that the ‘stacked dimer’ is the most meaningful minima model to understand the photophysics and photochemistry of nucleic acids.^30,52,53 On this ground, we can explain why, as discussed in detail in section 6 of the SI, other effects, as those related to the presence of two different flanking residues for ^thG (Figure S18A–C), play a minor role in its photophysics (Table S4 and Figure S19). Analogously, the nature of the next closest neighbour, even in the extreme case of a GG step (Figure S18D), does not have a significant direct effect on the CT excited states in which ^thG is involved (Table S5). On the other hand, it is clear that any change in the sequence, especially for mismatched duplexes can have a substantial and not a priori predictable effect on the photophysics of ^thG, but this study highlights the major chemical physical effects that have to be taken into consideration.

5. Conclusions

By combining steady-state and time-resolved fluorescence spectroscopy with molecular modelling and quantum mechanical calculations, the photophysical properties of ^thG in matched and mismatched DNA duplexes were collected, interpreted and rationalized. In all twenty sequences examined, ^thG was significantly emissive (QY 0.10–0.48) and exhibited a simple biexponential fluorescence decay, confirming its utility as a highly sensitive and robust probe in DNA duplexes. Due to its high QY, the brightness (ε×QY) of ^thG in DNA duplexes is between 500 and 2250 M⁻¹ cm⁻¹ and thus, not too distinct from that of the brightest fluorescent nucleoside ever reported in nucleic acids (2600 M⁻¹ cm⁻¹).⁵⁴ Moreover, the long-lived excited-state lifetime τ₂, which contributes to more than 75% of the total emission for 19 out of 20 of the tested duplexes (Table 5), shows unusually high values (9–29 ns). Such long-lived excited states are rarely observed with common small fluorophores, potentially facilitating selective measurements even in complex mixtures of fluorescently labelled molecules or in autofluorescent media by using a fluorescence lifetime-based or time-gated detection scheme.

For matched duplexes, the emission quantum yields (0.15–0.18), as well as τ₂ (10–12 ns) and <τ> (8–10 ns) were found to narrowly vary. The minimal dependence of the QY and excited-state lifetime values on ^thG’s neighbors in matched duplexes appears to be advantageous, because ^thG can be used to replace any G residue in a DNA duplex, while maintaining similar photophysical properties. In this respect, the values given here can be considered representative for referencing in future studies.

In sharp contrast, the QY, τ₂ and <τ> values of ^thG in mismatched duplexes strongly depend on the nature of its flanking and paired residues. An obvious application for ^thG-labelled sequences emerging from our observations is the detection of single nucleotide polymorphism. This application was already suggested for ^thG on the basis of the analysis of the limited but context dependent contribution of the ^thG-H3 tautomer, when the duplexes were excited at 310 nm.⁴ Herein, comparison of Table 1 and 5 shows clear differences in the QY, τ₂ and <τ> values when the matched C is replaced by any other nucleotide or an abasic site. Thus, the QY, τ₂ and <τ> values appear to reflect a simpler and more sensitive approach to detect single nucleotide polymorphism and identify the nature of the mismatched nucleotide. Moreover, comparison of the matched with mismatched duplexes further indicates that the context insensitive QY, τ₂ and <τ> values strictly depend on the local stability and stiffness of the matched duplexes, which ensure the strength of the stacking interactions of ^thG with the nearby bases. The local destabilization and structural changes in DNA induced by a mismatch or an abasic site opposite to ^thG are sufficient to restore a strong dependence of ^thG photophysics on its context, and notably on the nature of its flanking bases. In contrast to matched duplexes, the photophysical properties of ^thG flanked by nucleobases of different nature are more difficult to predict in mismatched duplexes, as they obviously depend on the exact geometry adopted by ^thG and its flanking nucleobases as a result of the local destabilization induced by the mismatch. To our knowledge, ^thG’s exquisite photophysical sensitivity to its context, while preserving good QYs, is unique among fluorescent nucleoside analogues^46,55 and is thus ideal for monitoring local structural changes induced by DNA ligands acting as helix destabilizers, NA chaperones, DNA bending proteins or base flipping proteins. This conclusion is fully in line with previous observations and helps explaining the strong changes in ^thG fluorescence observed when ^thG-labelled DNA duplexes or stem-loops interacted with the base flipping domain of UHRF1¹⁰ or the HIV-1 nucleocapsid protein NCp7, acting as a NA chaperone.⁷ In both cases, the protein-induced base flipping prevents the stacking between ^thG and its flanking nucleobases, which impedes the CT mechanism and thus, explains the observed fluorescence increase. Similarly, the fluorescence increase observed for a ribozyme-mediated cleavage that releases a short duplex with ^thG at its 5’ terminus¹¹ could be easily explained by a dramatic decrease in CT efficiency as a result of the loss of one of ^thG’s flanking residues and the poor stability of the duplex termini.

Taken together, our integrated analysis provides a rationale for the observed spectroscopic properties of ^thG in DNA duplexes and enables disentangling the different effects modulating its photophysical features. This analysis validates ^thG as a very unique fluorescent nucleoside surrogate that combines almost perfect substitution of G residues, preservation of good QY, unique long-lived excited-state lifetimes and exquisite sensitivity to solvent exposure and DNA context. This study opens new possibilities for the rational use and proper data interpretation of ^thG-labelled oligonucleotides in applications such as single nucleotide polymorphism and DNA/protein interactions. Applications of ^thG photoexcited dynamics, whose assessment is fundamental for the study of photoinduced oxidative damage, can also be envisioned.