Introduction

2′-Deoxyribonucleic acids (DNA) and ribonucleic acids (RNA) are the molecular materials responsible for the coding and decoding of the genetic information of life [1]. These biopolymers consist of long chains of nucleosides attached sequentially via 5′–3′ phosphodiester bonds. Nucleosides consist of a planar, nitrogenous, and aromatic nucleobase and a functionalized five-membered ring sugar. The common purine nucleobases are adenine (Ade) and guanine (Gua), whereas the common pyrimidine nucleobases are cytosine (Cyt), uracil (Ura), and thymine (Thy). The purine and pyrimidine nucleobases are connected to their pentose sugar constituents via C1′–N9 and C1′–N1 glycosidic bonds, respectively. The stability of these N-glycosidic bonds is crucial. In some regards, the N-glycosidic bond stability must maintain a balance such that it is stable enough to maintain the proper sequence fidelity of DNA and RNA strands, but must also be available for cleavage via various base-excision repair pathways to correct undesirably mutated nucleobases [2,3,4,5,6,7,8].

The 2′-deoxyribose sugar constituents of DNA nucleosides (dNuo) have a 2′-hydrogen, whereas the ribose sugar constituents of RNA nucleosides (Nuo) have a 2′-hydroxy group. This difference in the 2′-substituent functionality is the primary characteristic that defines the chemistry and classification of DNA and RNA molecules. Synthetic replacement of the 2′-substituent with fluorine produces the unnatural nucleoside analogues, 2′-fluoro-2′-deoxynucleosides (Nuofl). The pyrimidine Nuofl include 2′-fluoro-2′-deoxycytidine (Cydfl) and 2′-fluoro-2′-deoxyuridine (Urdfl) [9] and the purine Nuofl include 2′-fluoro-2′-deoxyadenosine (Adofl) and 2′-fluoro-2′-deoxyguanosine (Guofl). Fluorine substitutions are among the most commonly pursued and medically useful synthetic modifications to natural products [10,11,12,13,14,15,16]. Syntheses of Nuofl and other fluorinated nucleoside derivatives have been continually pursued, improved, and screened for pharmaceutical activity, especially anticancer and antiviral effects, for decades [11, 17,18,19,20,21,22]. The extensive pharmaceutical applications of fluorine arise from its unique ability to impart significant electronic effects with minimal structural changes to molecules. The very high electronegativity of fluorine [23] affects many critical aspects of potentially therapeutic molecules, including lipophilicity, bond strengths, dipole moments, polarizabilities, and the acidity or basicity of nearby functional groups [12, 24, 25]. 2′-Fluoro modifications to nucleosides are of further fundamental interest due to the unique ability of fluorine to mimic some properties of both hydrogen and hydroxyl groups. For example, fluorine is intermediate in size to hydrogen and hydroxy groups [26]. Additionally, fluorine substituents can accept but not donate hydrogen bonds, whereas hydroxy substituents can both accept and donate hydrogen bonds, and hydrogen substituents cannot do either. The small steric bulk and hydrogen bond accepting ability of fluorine has implications in enzyme binding, where noncovalent interactions are used to guide and stabilize substrates into specific spaces and conformations for subsequent catalyzed chemical reactions [1, 27]. Some fluorinated substrates, such as the 2′,2′-difluorinated nucleoside drug gemcitabine [28], display substantially increased enzyme–substrate binding affinities and competitively inhibit the unmodified substrates from entering the enzyme active sites, whereas others, such as 5-fluorouracil [29], display mechanism-based inhibition through the irreversible formation of enzyme–substrate complexes [13, 27, 30]. The occurrence and stability of H∙∙∙F interactions in the condensed phase remains a subject of interest and debate [31,32,33,34,35,36,37,38]. Although F has higher electronegativity, H∙∙∙F hydrogen bonds tend to be approximately half as strong as parallel H∙∙∙O interactions [25].

The various structures and functions of DNA and RNA strands are ultimately emergent features arising from the makeup of their constituents. The study of the isolated constituents of complex biological systems provides insight into their biochemistry from a “bottom-up” perspective. The solvent-free and controllable chemical environment of tandem mass spectrometry (MS/MS) experiments has proven an invaluable asset in the study of various aspects of the canonical and noncanonical DNA and RNA constituents. For example, energy-resolved collision-induced dissociation (ER-CID) experiments performed under multiple collision conditions in trap** mass spectrometers have been used to study the relative stability of various nucleosides [9, 39,40,41,42,43,44,45]. Additionally, threshold analysis of ER-CID experiments performed under nominally single collision conditions on guided ion beam mass spectrometers (GIBMS) have provided thermodynamically accurate dissociation energetics for various noncovalent [46,47,48,49,50,51,52] and covalent [53,54,55,56] interactions of nucleobases and nucleosides. Infrared multiple photon dissociation (IRMPD) action spectroscopy has been used to determine the experimentally populated structures of many canonical and noncanonical forms of nucleobases [57,58,59,60,61,62,63,64,65,66], nucleosides [9, 39,40,41,42,43,44, 67,68,69,70,71,72,73,74], and nucleotides [75,76,77,78,79,80].

In this work, two types of MS/MS are utilized complementarily with electronic structure calculations to study the protonated forms of the two purine Nuofl, [Adofl+H]+ and [Guofl+H]+. The neutral chemical structures of Adofl and Guofl are shown in Figure 1. ER-CID experiments with survival yield analyses provide relative N-glycosidic bond stabilities. IRMPD action spectroscopy experiments compared with theoretical IR spectra of low-energy stable structures of these species allows determination of the three-dimensional gas-phase ion conformations populated in the experiments. Comparisons are primarily made between this work and previously reported ER-CID, IRMPD, and computational studies of the protonated 2′-fluoro pyrimidine nucleosides [Nuofl+H]+ [9], the protonated canonical RNA nucleosides, [Nuo+H]+, and the protonated canonical DNA nucleosides, [dNuo+H]+ [42, 43, 55, 56, 70, 71].

Figure 1
figure 1

Chemical structures of neutral 2′-fluoro-2′-deoxyadenosine (Adofl) and 2′-fluoro-2′-deoxyguanosine (Guofl). The numbering of the atoms of the nucleobase and sugar moieties are shown

Full size image

Experimental and Theoretical Methods

Materials and Locations

Adofl and Guofl (see Figure 1 for neutral chemical structures) were purchased from Hongene Biotechnology Limited (Morrisville, NC, USA) and used as supplied for ER-CID and IRMPD experiments. ER-CID experiments were performed at Wayne State University (WSU; Detroit, MI, USA). In these experiments, methanol (HPLC grade) and glacial acetic acid were purchased from Fisher Scientific (Waltham, MA, USA), and water (HPLC grade) was purchased from Sigma-Aldrich Corporation (St. Louis, MO, USA). IRMPD action spectroscopy experiments were performed at the FELIX Facility at Radboud University (Nijmegen, The Netherlands). For these experiments, methanol (HPLC grade), water (HPLC grade), and fuming hydrochloric acid were purchased from Sigma-Aldrich Corporation (St. Gallen, Switzerland).

Energy-Resolved Collision-Induced Dissociation

Separate working solutions of the nucleosides Adofl and Guofl were prepared at concentrations of ~ 10 μM in 50/50 methanol/water with ~ 0.1% acetic acid. These solutions were introduced via mechanical syringe pump at a flow rate of ~ 3 μL/min to the Apollo II electrospray ionization (ESI) source of an amaZon ETD quadrupole ion trap mass spectrometer (QIT MS; Bruker Corporation, Billerica, MA; Bruker Daltonics, Bremen Germany). The precursor ions, [Adofl+H]+ and [Guofl+H]+, were transferred via rf ion guides and static lenses to the QIT MS where they were mass isolated. Helium (and background residual gases) present in the trap chamber at a stagnation pressure of ~ 1 mTorr was used as the CID collision gas. Mass spectra were collected as a function of the rf excitation amplitude (rfEA) used to excite the ion kinetic energies during CID. The rfEA was ramped from 0.00 V to a value beyond that required to achieve 100% precursor ion dissociation (~ 0.6 V maximum for the systems studied here) in steps of 0.01 V per 30 s segments. The rfEA was applied for 40 ms during each MS/MS sequence. With the experimental sequence employed, ~ 50 spectra were collected and averaged at each rfEA during each ER-CID scan. Experiments were performed in triplicate. Compass Data Analysis 4.0 software (Bruker Daltonics, Bremen, Germany) was used to acquire, extract, and export the mass spectral data. Custom software developed in our laboratory was used to extract time-resolved raw intensity data from the Bruker data files and calculate rfEA-dependent survival yields according to Eq. (1):

$$ \mathrm{Survival}\ \mathrm{yield}={I}_{\mathrm{p}}/\left({I}_{\mathrm{f}}+{I}_{\mathrm{p}}\right) $$
(1)

where Ip and If are the precursor and fragment ion intensities, respectively. The survival yield data was fit with a four-parameter logistic dynamic curve as shown in Eq. (2):

$$ \mathrm{Survival}\ \mathrm{yield}=\min +\left\{\left(\max -\min \right)/{\left(1+\frac{{\mathrm{rf}}_{\mathrm{EA}}}{{\mathrm{CID}}_{50\%}}\right)}^{{\mathrm{CID}}_{\mathrm{slope}}}\right\} $$
(2)

where max is the maximum possible value, set to 1 here and corresponding to precursor isolation without fragmentation; min is the minimum possible value, set to 0 here and corresponding to complete precursor ion dissociation; CIDslope is the slope of the declining region of the survival yield curve; and the CID50% is the rfEA required to give a survival yield of 50%. The CID50% value is extracted from this fit and used as a relative measurement of N-glycosidic bond stability. SigmaPlot Version 10.0 (Systat Software, Inc., San Jose, CA, USA) was used to fit, analyze, and plot the data.

Infrared Multiple Photon Dissociation Action Spectroscopy

Separate working solutions of the nucleosides Adofl and Guofl were prepared at concentrations of ~ 1 mM in 50/50 methanol/water with ~ 0.1% hydrochloric acid. These solutions were introduced via mechanical syringe pump at a flow rate of ~ 5 μL/min to the “Z-Spray” ESI source (Micromass, Milford, MA, USA) of a homebuilt 4.7 T Fourier transform ion cyclotron resonance mass spectrometer (FT-ICR MS). This FT-ICR MS is coupled with two sources of tunable infrared (IR) laser light: the free electron laser for infrared experiments (FELIX) [81, 82] and a benchtop optical parametric oscillator/optical parametric amplifier (OPO) laser system (LaserVision, Bellevue, WA, USA States) [74, 83] pumped by an InnoLas SpitLight 600 Nd:YAG (Krailling, Germany). The precursor ions, [Adofl+H]+ and [Guofl+H]+, were accumulated and thermalized in a hexapole ion trap and then pulsed through a quadrupole bender and injected into the ICR cell where they were trapped and mass isolated using stored waveform inverse Fourier transform (SWIFT) techniques. The isolated precursor ions were then irradiated with IR light from either FELIX or the OPO. With absorption of a sufficient number of resonant photons, the ion internal energies may reach or exceed their dissociation energy thresholds and fragment. IRMPD yields were calculated as a function of laser wavelength using Eq. (3):

$$ \mathrm{IRMPD}\ \mathrm{yield}={I}_{\mathrm{f}}/\left({I}_{\mathrm{f}}+{I}_{\mathrm{p}}\right) $$
(3)

where If and Ip are defined as in Eq. (1). FELIX was operated at 5 Hz, and the [Adofl+H]+ and [Guofl+H]+ ions were irradiated for 3.0 and 3.5 s, respectively, over the vibrational frequency range of ~ 600–1850 cm−1. The OPO was operated at 10 Hz, and the [Adofl+H]+ and [Guofl+H]+ ions were irradiated for 7.5 s over the vibrational frequency range of ~ 3300–3800 cm−1. IRMPD yields were normalized linearly based on the irradiation time length and a least squares fit of the laser power as a function of wavelength.

Computational Approaches

The molecular structures and atom numbering for neutral Adofl and Guofl are shown in Figure 1. Potentially favorable protonation sites of these nucleosides were investigated. For Adofl, this included protonation at the N1, F2′, N3, and N7 positions. For Guofl, this included protonation at the F2′, N3, O6, and N7 positions. Candidate structures were generated for each protonated nucleoside form via a molecular dynamics simulated annealing procedure using the HyperChem software (Hypercube, Inc., Gainesville, FL, USA) and the Amber 3 force field. During simulated annealing, the ion was heated from 0 to 1000 K over 0.3 ps, conformational space was sampled at 1000 K for 0.2 ps, and the ion was cooled from 1000 to 0 K over 0.3 ps. A snapshot of the resulting structure was saved and used to initiate the next simulated annealing cycle. This process was repeated 300 times for each of the initial protonated forms of the nucleosides, and ~ 30 of the most stable conformers from each simulated annealing run were subjected to density functional theory (DFT) calculations. The optimized structures previously reported for [Nuo+H]+ and [dNuo+H]+ along with chemical intuition were used to generate additional conformers to complement those found via simulated annealing in an attempt to comprehensively sample all possible low-energy structures.

The Gaussian 09 suite of programs [84] was used to perform geometry optimizations, frequency analyses, and single-point energy determinations. Geometry optimizations and frequency analyses were performed at the B3LYP/6-311+G(d,p) level of theory at standard ambient temperature and pressure and with a frequency scaling factor of 0.9887 [85]. Theoretical linear IR spectra were generated by the software and scaled by factors of 0.980 and 0.957 in the ~ 600–1850 cm−1 fingerprint and ~ 3300–3800 cm−1 hydrogen-stretching regions, respectively, for both [Adofl+H]+ and [Guofl+H]+. The predicted linear IR spectra were broadened using 30 cm−1 and 15 cm−1 full width at half maximum (FWHM) Gaussian line shapes for the fingerprint and hydrogen-stretching regions, respectively. Single-point energy determinations were performed at the B3LYP/6-311+G(2d,2p) and MP2(full)/6-311+G(2d,2p) levels of theory for each optimized structure. Additional single point energy calculations at the B3LYP/6-311+G(2d,2p) and MP2(full)/6-311+G(2d,2p) levels of theory were performed for selected conformers using the polarizable continuum model (PCM) with an implicit water solvent. Zero-point energy (ZPE) and thermal corrections at 298 K were included in the relative stabilities computed.

Conformer Designations and Structural Characterizations

Comparisons between the experimental IRMPD and predicted linear IR spectra allow for the inclusion or exclusion of calculated conformers as being present as gas-phase ions in the MS experiments. Calculated conformers are named based on their site of protonation followed by a letter incremented alphabetically based on their relative B3LYP/6-311+G(2d,2p) Gibbs energy at 298 K. GaussView 5.0 (Gaussian, Inc., Wallingford, CT, USA) was used to extract bond lengths and angles. Structural descriptions are made according to methods described in detail in previous work [39,40,41,42,43, 86,87,88] and described briefly here. Purine nucleobase orientations are classified as anti or syn as defined by ∠C4N9C1′O4′ dihedral angle ranges of − 90° to 90° and 90° to 270°, respectively. Sugar puckers are described in two ways. In the first classification scheme, descriptors are given to the sugar ring atom that deviates the greatest distance from the plane formed by the other four sugar ring atoms. Endo configurations deviate toward the 5′-hydroxyl, whereas exo configurations deviate away. In the second classification scheme, more descriptive designations based on the pseudorotation angle (P°) of the sugar are used. P° is calculated using the five dihedral angles of the sugar moiety. Envelope (E) structures have only one atom deviating from the plane of the sugar, whereas twist (T) structures have two atoms deviating in opposite directions. Superscripted numbers indicate endo configurations, whereas subscripts represent exo. The number prior to the letter designation (E or T) indicates the atom number of the major deviator from the sugar plane, whereas the number after the letter indicates the atom number of the minor deviator.

Results and Discussion

CID and IRMPD Dissociation Pathways

Both CID, when performed in a QIT MS, and IRMPD are multi-step, slow-heating processes [89,90,91,92]. Activation of [Adofl+H]+ by multiple collisions, Eq. (4), and multiple IR photons, Eq. (5), produces the same reaction products.

$$ {\left[\mathrm{Adofl}+\mathrm{H}\right]}^{+}\overset{n\ \mathrm{He}}{\to }{\left[\mathrm{Ade}+\mathrm{H}\right]}^{+}+\left(\mathrm{Adofl}-\mathrm{Ade}\right) $$
(4)
$$ {\left[\mathrm{Adofl}+\mathrm{H}\right]}^{+}\overset{n\ \mathrm{h}\upnu}{\to }{\left[\mathrm{Ade}+\mathrm{H}\right]}^{+}+\left(\mathrm{Adofl}-\mathrm{Ade}\right) $$
(5)

In both reactions, [Adofl+H]+ solely dissociates via N-glycosidic bond cleavage to produce the protonated adenine nucleobase [Ade+H]+ and the corresponding neutral sugar moiety (Adofl–Ade). Similarly, activation of [Guofl+H]+ by multiple collisions, Eq. (6), and multiple IR photons, Eq. (7), also produces parallel reaction products.

$$ {\left[\mathrm{Guofl}+\mathrm{H}\right]}^{+}\overset{n\ \mathrm{He}}{\to }{\left[\mathrm{Gua}+\mathrm{H}\right]}^{+}+\left(\mathrm{Guofl}-\mathrm{Gua}\right) $$
(6)
$$ {\left[\mathrm{Guofl}+\mathrm{H}\right]}^{+}\overset{n\ \mathrm{h}\upnu}{\to }{\left[\mathrm{Gua}+\mathrm{H}\right]}^{+}+\left(\mathrm{Guofl}-\mathrm{Gua}\right) $$
(7)

In both of these reactions, [Guofl+H]+ solely dissociates via N-glycosidic bond cleavage to produce the protonated guanine nucleobase [Gua+H]+ via neutral loss of the corresponding neutral 2′-fluorinated sugar moiety (Guofl–Gua).

CID MS/MS spectra for [Adofl+H]+ and [Guofl+H]+ acquired at an rfEA near their CID50% value are shown in Figure 2. Previously published [39, 40] MS/MS spectra for [Ado+H]+, [dAdo+H]+, [Guo+H]+, and [dAdo+H]+ are also included in Figure 2 for comparison. The N-glycosidic bond cleavage pathways observed for [Adofl+H]+ and [Guofl+H]+ exactly parallel those found for their canonical DNA and RNA forms: [dAdo+H]+, [dGuo+H]+, [Ado+H]+, and [Guo+H]+ [39, 40, 53, 54, 67, 68], as well as for the protonated noncanonical purine 2′-O-methylnucleosides, [Adom+H]+ and [Guom+H]+ [44], and other protonated and sodium cationized noncanonical methylguanosines [45]. The pyrimidine [Nuofl+H]+, [Nuo+H]+, and [dNuo+H]+ also primarily undergo N-glycosidic bond cleavage [41,42,43, 69,70,71]; however, several competing neutral loss pathways are also observed for the protonated uracil and thymine nucleoside analogues.

Figure 2
figure 2

CID MS/MS spectra acquired at rf excitation amplitudes (rfEA) that produced ~ 50% dissociation of [Adofl+H]+, [Ado+H]+, and [dAdo+H]+ (top panel) and [Guofl+H]+, [Guo+H]+, and [dGuo+H]+ (bottom panel). Results for the protonated canonical RNA and DNA nucleosides were taken from previous work [39, 40]

Full size image

Relative Stability Comparisons

With judicious employment of QIT MS ER-CID and survival yield analyses, the relative stability of ions can be determined [9, 44, 45, 92,93,94,95,96,97,98,99]. Figure 3 compares survival yield curves for [Adofl+H]+, [Ado+H]+, [dAdo+H]+, [Guofl+H]+, [Guo+H]+, and [dGuo+H]+. The [Adofl+H]+ and [Guofl+H]+ results were acquired in this work, whereas data for [Ado+H]+, [dAdo+H]+, [Guo+H]+, and [dGuo+H]+ are reproduced from previous work [40]. With the comparable conditions and reaction pathways observed in the experiments, the CID50% values of these protonated nucleosides can be understood and compared as a measure of the relative stability of their N-glycosidic bonds.

Figure 3
figure 3

Survival yield curves for [Adofl+H]+, [Ado+H]+, and [dAdo+H]+ (top panel) and [Guofl+H]+, [Guo+H]+, and [dGuo+H]+ (bottom panel). Error bars represent one standard deviation of the data measured in triplicate. Fits to the survival yield measurements using a dynamic four-parameter logistic curve, the extracted CID50% values, and the standard error associated with the fit (in parentheses) are shown. Survival yield data for the protonated canonical RNA and DNA nucleosides were taken from previous work [39, 40]

Full size image

Two important trends are observed in the data. First, the N-glycosidic bond stability is observed to vary with the 2′-substituent. The CID50% values of [dAdo+H]+, [Ado+H]+, and [Adofl+H]+ increase from 0.252 ± 0.001 to 0.290 ± 0.001 to 0.305 ± 0.001 V, respectively. Similar behavior is seen in the CID50% values of [dGuo+H]+, [Guo+H]+, and [Guofl+H]+, which increase from 0.186 ± 0.001 to 0.220 ± 0.001 to 0.244 ± 0.001 V, respectively. Parallel trends regarding the influence of the 2′-substituent on the glycosidic bond stability of the protonated pyrimidine nucleosides were also found [9]. These results can be explained by examining the electronegativities of the 2′-substituents, which follow that same order: 2′-H (2.20) < 2′-OH (3.44) < 2′-F (3.98) [23, 100]. Increasing electronegativity of the 2′-substituent draws more electron density from the 2′-C, giving it greater partial positive character. This partial positive character destabilizes the proposed oxocarbenium ion intermediate [53,54,55,56, 101, 102] and increases the energy necessary to activate the glycosidic bond. The second trend relates to the identity of the nucleobases. For the purine nucleobases, the adenine nucleoside analogues exhibit greater glycosidic bond strength than the guanine nucleoside analogues. Figure S1 compares the survival yield curves of [Adofl+H]+ and [Guofl+H]+ with previously published [9] results for [Cydfl+H]+ and [Urdfl+H]+. From this data, the relative glycosidic bond stabilities of the [Nuofl+H]+ increase with the trend: [Urdfl+H]+ < [Cydfl+H]+ < [Guofl+H]+ < [Adofl+H]+ (respective CID50% values, 0.213 ± 0.001 V < 0.224 ± 0.001 V < 0.244 ± 0.001 V < 0.305 ± 0.001 V). Parallel trends as a function of nucleobase were observed in the previously studied [Nuo+H]+ and [dNuo+H]+ systems [39,40,41,42,43]. These results indicate that the purine nucleoside analogues have a greater glycosidic bond stability than the pyrimidine nucleoside analogues and that the glycosidic bond stabilities increase from the uracil to the cytosine to the guanine to the adenine nucleoside analogues.

IRMPD Action Spectra

Figure 4 displays the IRMPD action spectrum of [Adofl+H]+ compared with the previously published [Ado+H]+ and [dAdo+H]+ spectra [67]. The spectral features of the three IRMPD spectra are overall highly parallel, suggesting similar structural conformations. Table 1 shows the vibrational band assignments for the [Adofl+H]+ spectrum based on the experimentally populated conformers determined in this work. In the fingerprint region, from ~ 550 to 1850 cm−1, the primary change between the [Adofl+H]+ spectrum and the canonical [Ado+H]+ and [dAdo+H]+ spectra is an increase in the relative intensities of the absorption bands, especially between ~ 1150–1650 cm−1 and ~ 600–1000 cm−1. Modest peak broadening and absorption intensity changes are observed in the ~ 1575–1725 cm−1, ~ 1350–1550 cm−1, and ~ 1000–1150 cm−1 regions. In the hydrogen-stretching region, from ~ 3300–3750 cm−1, the primary spectral difference is the splitting of the ~ 3668 cm−1 band corresponding to the O3′–H and O5′–H stretches in the [Ado+H]+ and [dAdo+H]+ spectra into two distinct O–H stretches at 3644 and 3668 cm−1 in the [Adofl+H]+ spectrum. The additional peak at 3644 cm−1 is observed due to red shifting of the O3′–H stretch resulting from a hydrogen-bonding interaction with the highly electronegative 2′-F substituent. An additional band at ~ 3580 cm−1 is observed in the [Ado+H]+ spectrum that arises from the 2′-O–H stretch, a feature not present in the [Adofl+H]+ and [dAdo+H]+ spectra as this substituent is absent in these nucleosides. The absorption bands corresponding to N–H stretches are highly parallel between the [Adofl+H]+, [Ado+H]+, and [dAdo+H]+ spectra.

Figure 4
figure 4

Experimental IRMPD action spectra of [Adofl+H]+, [Ado+H]+, and [dAdo+H]+. Data for the protonated canonical DNA and RNA nucleosides were taken from previous work [67]

Full size image
Table 1 Vibrational Band Assignments for [Adofl+H]+
Full size table

Figure 5 displays the IRMPD action spectrum of [Guofl+H]+ compared with the previously published spectra of [Guo+H]+ and [dGuo+H]+ [68]. Table 2 lists the vibrational band assignments for the [Guofl+H]+ spectrum. Again, the spectral features of the three experimental IRMPD spectra are overall highly parallel, suggesting similar structures. In the fingerprint region, the primary change between the [Guofl+H]+ spectrum and the canonical [Guo+H]+ and [dGuo+H]+ spectra is an increase in the relative intensities of absorption bands, especially in the ranges ~ 1725–1825 cm−1, ~ 1150–1550 cm−1, and ~ 600–1050 cm−1. Modest peak broadening is found in the [Guofl+H]+ spectrum, as particularly exemplified in the most intense spectral features in the ~ 1725–1825 cm−1, ~ 1550–1700 cm−1, and ~ 1050–1300 cm−1 regions. In the hydrogen-stretching region, [Guofl+H]+ again displays a greater number of absorption bands than its canonical RNA and DNA counterparts. In the [Guofl+H]+ spectrum, two red-shifted bands arise at 3644 and 3624 cm−1 from O3′–H stretches associated with two distinct O3′H∙∙∙F geometries among the conformers populated, whereas in the [Guo+H]+ and [dGuo+H]+ spectra, the O3′–H stretch coalesces with the O5′–H stretch at ~ 3665 cm−1. The number and position of the N–H, NH2 symmetric, and NH2 antisymmetric stretches are consistent among [Guofl+H]+, [Guo+H]+, and [dGuo+H]+, but the relative intensity of the bands at 3564 and 3536 cm−1 is reduced for [Guofl+H]+ compared to its canonical counterparts.

Figure 5
figure 5

Experimental IRMPD action spectra of [Guofl+H]+, [Guo+H]+, and [dGuo+H]+. Data for the protonated canonical DNA and RNA nucleosides were taken from previous work [68]

Full size image
Table 2 Vibrational Band Assignments for [Guofl+H]+
Full size table

The wavelength-dependent IRMPD yields of [Adofl+H]+ and [Guofl+H]+ and their canonical RNA and DNA forms are overlaid for more direct comparison in Figure S2. Despite their greater glycosidic bond stability, the [Nuofl+H]+ nucleosides exhibit greater IRMPD yields than their canonical forms. The large IRMPD yields of [Nuofl+H]+ likely arise from increased IR photon absorption cross sections due to the increased interatom dipole moments in the sugar moiety (see Figure S3) and the lone pairs of the highly electronegative 2′-F substituent. The slightly broader peaks in the [Nuofl+H]+ IRMPD spectra are also likely due to the greater absorption cross sections, and therefore the increased absorption band intensities, in the protonated 2′-fluoro-substituted nucleosides compared to the canonical systems. However, minor differences in laser beam profiles from day-to-day, i.e., the shape of the beam about the central frequency, could also lead to differences in absorption peak widths.

B3LYP vs. MP2(full) Stability Comparisons

Previous experimental and theoretical works, especially those reported by Wu et al. and Zhu et al., have determined that single-point energy calculations at the B3LYP/6-311+G(2d,2p) are robust and satisfactory for describing canonical and modified nucleosides [39,40,41,42,43, 53,54,55,56, 67,68,69,70,71]. The MP2(full)/6-311+G(2d,2p) level of theory has also proven useful in providing ancillary insight into the conformer energetics for some systems. However, the MP2(full) results consistently mirror the B3LYP results (see Table 3 and Figures S4 and S5) for [Adofl+H]+ and [Guofl+H]+. Additionally, the [Nuofl+H]+ energetics are highly parallel to those calculated for analogous [Nuo+H]+ and [dNuo+H]+ structures. Therefore, throughout this work, the B3LYP/6-311+G(2d,2p) calculated values will be used in all discussions, and the MP2(full)/6-311+G(2d,2p) values are included in tables and figures for additional reference.

Table 3 Relative 0 and 298 K Enthalpies and 298 K Gibbs Energies of Selected Low-Energy Conformers of Protonated Purine Nucleoside Analogues
Full size table

Calculated [Adofl+H]+ and [Guofl+H]+ Conformers

Unique conformers calculated for [Adofl+H]+ are shown in Figure S4. Energetic and structural information for selected conformations (N3A, N3B, N1A, and N7A) is provided in Tables 3 and 4. The ground-state conformation, N3A, has a syn nucleobase orientation and C2′-endo (2T1) sugar puckering. The N3 proton aids in stabilization of the syn nucleobase orientation by participating in an N3H+∙∙∙O5′ ionic hydrogen-bonding interaction. The C2′-endo (2T1) sugar pucker is stabilized by an O3′H∙∙∙F interaction. The N3B conformer lies 2.5 kJ/mol higher in B3LYP Gibbs energy, and is highly parallel in structure to N3A, except with a C4′-exo (4T3) sugar pucker. Disrupting the N3H+∙∙∙O5′ hydrogen bond of N3B produces the N3C conformer after optimization at 8.0 kJ/mol B3LYP Gibbs energy, indicating this particular hydrogen–fluorine interaction stabilizes the system by ~ 5.5 kJ/mol. After N3 protonation, the next most favorable protonation site is at the N1 position. N1A, at 20.3 kJ/mol, has an anti nucleobase orientation stabilized by a noncanonical C8H∙∙∙O5′ hydrogen bond and C2′-endo (2T3) sugar puckering stabilized by an O3′H∙∙∙F hydrogen bond. N7A, the most stable N7-protonated conformer, lies 35.2 kJ/mol above the ground-state conformation. N7A has structural features that are highly parallel to N1A, except that rotation of the 3′-hydroxy substituent and thus reorientation of the O3′H∙∙∙F hydrogen bond lead to C3′-endo (3T2) sugar puckering.

Table 4 Geometric Parameters of Selected Low-Energy Conformers Computed for Protonated Purine Nucleoside Analogues
Full size table

Unique calculated conformers of [Guofl+H]+ are shown in Figure S5. Energetic and structural details for selected conformations (N7A, N7B, O6A, and N3A) are provided in Tables 3 and 4. The ground-state conformation, N7A, has an anti nucleobase orientation and C3′-endo (3T2) sugar puckering. N7B, the C2′-endo (2T3) analogue of N7A, lies 1.4 kJ/mol above the ground-state conformation. O6A is the O6 protonated analogue of N7B with an anti nucleobase orientation and C2′-endo (2T3) sugar puckering. The relative Gibbs energy of O6A, 33.9 kJ/mol above the ground-state conformation, results primarily from the differences in the proton affinities of these sites. Whereas the ground [Adofl+H]+ conformer is protonated at the N3 position, the N3A conformer of [Guofl+H]+ is 50.4 kJ/mol higher in B3LYP Gibbs energy than the ground-state conformation, N7A. N3A has a syn nucleobase orientation and C2′-endo (2T1) sugar puckering. As with the [Adofl+H]+ conformers, the nucleobase orientations and sugar puckerings are stabilized by hydrogen bonds. The anti nucleobase orientations are preferentially stabilized by a noncanonical C8H∙∙∙O5′ hydrogen bond, whereas the syn nucleobase orientation of N3-protonated conformers possess an N3H+∙∙∙O5′ hydrogen bond. The sugar puckers are preferentially stabilized by an O3′H∙∙∙F hydrogen-bonding interaction.

Experimentally Populated [Adofl+H]+ Conformers

The experimental [Adofl+H]+ IRMPD action spectrum is compared with the predicted linear IR spectra for the N3A, N3B, N1A, and N7A conformers in Figure 6. Similar comparisons for all unique calculated conformers are shown in Figure S6. Based on spectral comparisons, the conformers that were predominantly populated in the MS/MS experiments are N3A and N1A. Spectral comparisons indicate that N3B and N7A were not significantly populated in the MS/MS experiments.

Figure 6
figure 6

Comparisons of the experimental IRMPD action spectrum of [Adofl+H]+ (shown in black in the top panel and overlaid in gray and scaled to match the most intense feature in each region in all other panels) with theoretical linear IR spectra predicted for selected conformations of [Adofl+H]+ (shown in blue for experimentally populated conformers and in red for nonpopulated conformers in the bottom four panels). The optimized structure, conformer designation, nucleobase orientation, sugar puckering, and B3LYP/6-311+G(2d,2p) (shown in black) and MP2(full)/6-311+G(2d,2p) (shown in red) relative Gibbs energy at 298 K are given. The predicted IR spectrum for a 10:7 mixture of the N3A and N1A conformers, respectively, is overlaid in green in the top panel

Full size image

The vibrations predicted for the N3A and N1A conformers nicely replicate all experimental absorption bands. However, calculated linear IR spectra often underestimate the intensities of the absorption bands in nonlinear IRMPD action spectra [103], as is particularly observed below ~ 1625 cm−1 in the [Adofl+H]+ spectrum. The relative IRMPD yields of these vibrations are particularly amplified in the [Adofl+H]+ spectrum compared to those of [Ado+H]+ and [dAdo+H]+ (see Figure 4). Although N1A lies 23.6 kJ/mol higher in Gibbs energy than N3A in the gas-phase calculations, many of its spectral features match the experimental IRMPD spectrum exceptionally well. N1A lies only 4.9 kJ/mol higher in Gibbs energy than N3A in the PCM calculations with an implicit water solvent, suggesting some sampling of the solution-phase structures in the experimental measurements. Similar results were observed in the [Ado+H]+ and [dAdo+H]+ studies, where implicit solvation calculations were also found to lower the Gibbs energy difference between the N3A and N1A conformers to 2.4 and 4.5 kJ/mol, respectively, and both protonation sites were accessed in the experiments [67]. Some [Adofl+H]+ experimental bands, especially at ~ 3420, ~ 1600, ~ 1550, ~ 1500, and ~ 1275 cm−1, are more accurately replicated by N1A than by N3A. Other experimental bands, especially at ~ 3675, ~ 1660, ~ 1380, and ~ 1100 cm−1 are more accurately replicated by N3A than by N1A. A 10:7 combination of the N3A and N1A spectra, shown with a green line in the top plot, fits most of the experimental data well. This ratio reproduces the wide and angled shape of the feature at ~ 1625–1725 cm−1.

The hydrogen-stretching region of the experimental [Adofl+H]+ spectrum (see the 3300–3750 cm−1 region of Table 1 and Figure 6) has four distinct and highly informative absorption bands. The two highest frequency absorption bands at 3668 cm−1 and 3644 cm−1 correspond to the O5′–H and O3′–H stretches, respectively, of the N3A and N1A conformers. The O5′–H stretch of the N3A conformer matches the position of the observed experimental peak more accurately than that of the N1A conformer. The C2′-endo sugar puckering of N3A and N1A induce ∠O3′H∙∙∙F hydrogen bond angles of ~ 97.1° and lengths of ~ 2.423 Å (see Table 4), resulting in red shifting of the O3′–H stretch away from the O5′–H stretch by 24 cm−1 (see Table 1). The C4′-exo and C3′-endo sugar puckers of the N3B and N7A conformers produce ∠O3′H∙∙∙F hydrogen bond angles of ~ 107.2° and length of ~ 2.283 Å. The shorter and more on-axis O3′H∙∙∙F interaction of N3B and N7A result in further red shifting of the O3′–H band to ~ 3625 cm−1, where no appreciable resonant photon absorption is observed in the experimental IRMPD spectrum of [Adofl+H]+. The N–H stretches of N3A, N3B, and N1A all align well with the experimental spectrum. Further evidence for the exclusion of N7A from the experimental population is the mismatch of the predicted versus experimentally measured N–H vibrational frequencies in the region from ~ 3440 to ~ 3565 cm−1.

The fingerprint region of the IRMPD spectrum of [Adofl+H]+ (see the 550–1850 cm−1 region of Table 1 and Figure 6) displays an abundance of spectral features, all of which are replicated by the N3A and N1A conformers. The many features in the ~ 1150–1625 cm−1 and ~ 600–1050 cm−1 ranges are best replicated by the N1A conformer, whereas the more intense features at ~ 1625–1725 cm−1 and 1050–1150 cm−1 are better represented by the N3A conformer. The ~ 1625–1725 cm−1 region is probably most diagnostic for the presence of both the N3A and N1A conformers to explain its distinct shape and width, with a slight (~ 10:7) estimated preference for N3A over N1A. The N3B conformer displays parallel fingerprint absorptions to the N3A conformer, but with slight mismatching in the ~ 850–925 cm−1 and ~ 1150–1225 cm−1 regions and therefore could not be excluded from the experimental population based on the fingerprint region alone. The N7A conformer exhibits obvious spectral mismatches in the ~ 1425–1525 cm−1 and ~ 1625–1725 cm−1 regions.

Some conformers of [Adofl+H]+ not shown in Figure 6 have theoretical IR spectra that match the experimental IRMPD spectrum at least reasonably well (see Figure S6) and therefore cannot be conclusively excluded from the experimental population. However, the overall better quality match of the N3A and N1A predicted spectra with experiment over these conformers, as well as their higher Gibbs energies, suggest these additional conformers if present to be populated to a much lesser degree. These additional conformers with predicted spectra that reasonably match the experimental spectrum include N3C, N3D, N3G, N1C, N1G, N1H, and N1I. N3C, at 8.0 kJ/mol relative Gibbs energy, and N3D, at 13.9 kJ/mol, also have a syn nucleobase orientation, with N3D having a smaller ∠O4′C1′N9C4 dihedral angle. N3C has C4′-exo sugar puckering and does not contain an O3′H∙∙∙F interaction, whereas N3D has C2′-endo sugar puckering and is stabilized by an O3′H∙∙∙F interaction. N3G, at 29.8 kJ/mol, has a syn nucleobase orientation, C3′-endo sugar puckering, and an N3H+∙∙∙O5′ hydrogen bond. N1C, at 31.3 kJ/mol, has an anti nucleobase orientation, C3′-endo sugar puckering, and lacks an O3′H∙∙∙F interaction. N1G, at 43.1 kJ/mol, has an anti nucleobase orientation and C3′-endo sugar puckering. N1H, at 44.3 kJ/mol, has an anti nucleobase orientation and C4′-exo sugar puckering. N1I, at 46.1 kJ/mol, has a syn nucleobase orientation and C4′-exo sugar puckering.

Experimentally Populated [Guofl+H]+ Conformers

The experimental IRMPD action spectrum of [Guofl+H]+ is compared with the predicted linear IR spectra for the N7A, N7B, O6A, and N3A conformers in Figure 7. Similar comparisons for all unique calculated conformers are shown in Figure S7. Based on spectral comparisons, the conformers that were predominantly populated in the MS/MS experiments are N7A and N7B. The data also suggest the O6A conformer may have been present in minor population. Spectral comparisons indicate that N3A was not populated in the MS/MS experiments.

Figure 7
figure 7

Comparisons of the experimental IRMPD action spectrum of [Guofl+H]+ (shown in black in the top panel and overlaid in gray and scaled to match the most intense feature in each region in all other panels) with theoretical linear IR spectra predicted for selected conformations of [Guofl+H]+ (shown in blue for the dominant experimentally populated conformers, in purple for minor experimentally populated conformers, and in red for nonpopulated conformers in the bottom four panels). The optimized structure, conformer designation, nucleobase orientation, sugar puckering, and B3LYP/6-311+G(2d,2p) (shown in black) and MP2(full)/6-311+G(2d,2p) (shown in red) relative Gibbs energy at 298 K are given. The predicted IR spectrum for a 5:10:1 mixture of the N7A, N7B, and O6A conformers, respectively, is overlaid in green in the top panel

Full size image

The vibrations predicted for the N7A, N7B, and O6A conformers replicate all experimental absorption bands. Similar to that seen for [Adofl+H]+, the predicted linear IR spectra for [Guofl+H]+ generally underestimate absorption band intensities below ~ 1625 cm−1. The IRMPD yield for [Guofl+H]+ in this region are higher than the parallel spectral features observed in the IRMPD spectra of [Guo+H]+ and [dGuo+H]+ (see Figure 5). The predicted IR spectrum of N7A matches the experimental spectrum better than N7B at ~ 675, ~ 880, ~ 950, ~ 975, and ~ 1400 cm−1, whereas the predicted spectrum for N7B matches the experimental spectrum better than N7A in the region from ~ 1050–1375 cm−1. A 5:10:1 linear combination of the N7A, N7B, and O6A spectra, respectively, fits most of the experimental data quite well (see Figure 7). Because the presence of O6A cannot be conclusively proven or ruled out based on the data, a linear combination of the predicted spectra for a 5:10 mixture of N7A and N7B is also compared with the experimental spectrum in Figure S8 for comparison. The fit degrades only very slightly when the O6A conformer is not included.

The hydrogen-stretching region of [Guofl+H]+, the ~ 3300–3750 cm−1 region of Table 2 and Figure 7, displays seven distinct and highly informative resonant vibrational modes. The O5′–H stretch is the highest frequency vibration at 3668 cm−1. The 5′-hydroxyl hydrogen atoms of most low-energy stable conformers do not participate in a canonical hydrogen-bonding interaction and therefore the IR spectra of most conformers reproduce this experimental band. The other two O–H stretches, at 3644 and 3624 cm−1, arise from O3′–H stretches located in two distinct chemical environments. These two environments, readily replicated by N7A and N7B, are defined by the orientation and positioning of the O3′H∙∙∙F hydrogen-bonding interactions. The O3′H∙∙∙F geometry is primarily determined by the sugar puckering adopted by the conformer. N7B, with C2′-endo sugar puckering, has an ∠O3′H∙∙∙F angle of 97.7° and O3′H∙∙∙F distance of 2.410 Å and produces the O3′–H stretch at 3644 cm−1. N7A, with C3′-endo sugar puckering, has an ∠O3H′∙∙∙F angle of 106.1° and O3′H∙∙∙F distance of 2.310 Å and produces the O3′–H stretch at 3624 cm−1. The 3644 cm−1 band arising from N7B is more intense than the 3624 cm−1 band from N7A, suggesting that N7B is present in greater population than N7A in the experiments despite B3LYP predicting it to be 1.4 kJ/mol less stable in Gibbs energy. However, the shorter and more on-axis O3′H∙∙∙F interaction of N7B could potentially dampen the absorption of the O3′–H stretch and reduce the IRMPD yield in comparison to the longer and less on-axis O3H′∙∙∙F interaction of N7A; therefore, the suggested relative populations of the N7A and N7B conformers cannot be considered conclusive. The band at 3564 cm−1 arises from N2H2 antisymmetric stretching. This vibrational mode is replicated by conformers with an unbound N2H2 group, a feature of all conformers with an anti nucleobase orientation. N7A, N7B, and O6A all have anti nucleobase orientations and, therefore, all reproduce this absorption band, whereas N3A, with a syn nucleobase orientation in which the N2H2 substituent engages in an O5′∙∙∙HN2 hydrogen-bonding interaction, does not. The smallest feature in the hydrogen-stretching region is found at 3536 cm−1 and is the most challenging band to explain. This band probably arises from the highly anharmonic O6–H+ stretch predicted at 3582 cm−1 for the most stable O6 protonated conformer, O6A. Red shifting of this type of O–H+ stretch has been previously observed in other IRMPD studies, particularly in the IRMPD spectra of [dCyd+H]+, [Cyd+H]+, and [Cydfl+H]+ [9, 70]. However, the experimental O2–H+ stretch in these Cyd systems was red shifted from its predicted position by ~ 20 cm−1, whereas here the proposed assignment for the 3536 cm−1 peak corresponds to a red shift from theory by an even larger difference of 46 cm−1. However, the O6A conformer also readily replicates other necessary features of the [Guofl+H]+ IRMPD spectrum, and therefore the interpretation of its minor presence seems justified. The most intense hydrogen stretch absorption occurs at 3456 cm−1 and primarily arises from a coalescence of N2H2 symmetric and N7–H+ stretches from the N7A and N7B conformers. The slight shoulder to the red of this peak is readily replicated by the N2H2 symmetric stretch of O6A, as this band is predicted to be at a slightly lower vibrational frequency than the corresponding stretches of the N7 protonated conformers. The lowest frequency peak in the hydrogen-stretching region is observed at 3420 cm−1 and corresponds to N1–H stretching. This vibrational mode is readily reproduced by most conformers independent of their specific structural motifs and is therefore generally undiagnostic.

The fingerprint region displays an abundance of spectral features. The highest intensity and largest vibrational frequency band observed at 1767 cm−1 corresponds to the C6=O carbonyl stretch of N7 protonated conformers such as N7A and N7B. This band is not present in O6 protonated conformations such as O6A due to shifting of the resonant vibrational frequency by the bonded proton. Although the C6=O carbonyl is not directly perturbed by protonation in the N3 protonated conformers, the change in the chemical environment is sufficiently different to shift the predicted C6=O stretch to ~ 1825 cm−1; this peak is not reproduced in the experimental spectrum, and therefore N3 protonated conformers such as N3A are not populated in the experiments. The next major absorption band occurs at 1631 cm−1 and corresponds to the unbound N2H2 scissoring primarily from the anti nucleobase orientation of the N7A, N7B, and O6A conformers. The syn orientation of N3A aids the formation of an N2H∙∙∙O5′ interaction that red shifts this vibration, adding further justification for elimination of N3A and other syn conformers from the experimental population. The shoulder to the blue of this peak arises from C5=C6 and C2=N3 stretching and O6–H+ rocking of O6A.

Some conformers of [Guofl+H]+ not shown in Figure 7 have predicted linear IR spectra that match the experimental IRMPD spectrum at least reasonably well (see Figure S7) and therefore cannot be conclusively excluded from the experimental population but are likely much less important than N7A, N7B, and O6A. These conformers include N7C, N7E, N7F, N7H, N7I, O6B, and O6D. N7C, at 6.3 kJ/mol, is structurally parallel to N7A and has an anti nucleobase orientation and C3′-endo sugar puckering. N7C does not have an O3′H∙∙∙F interaction, and thus no red shifting of its O3′–H stretch is observed. N7E, at 19.6 kJ/mol, also has an anti nucleobase orientation and C3′-exo sugar puckering, but has a wider ∠O4′C1′N9C4 dihedral angle than N7C. N7F, at 19.7 kJ/mol, has a syn nucleobase orientation and C2′-endo sugar puckering stabilized by an O3′H∙∙∙F interaction. N7H, at 29.7 kJ/mol, is parallel to N7A, except with a different O5′ orientation that disrupts the noncanonical C8H∙∙∙O5′ interaction. N7I, at 30.1 kJ/mol, has an anti nucleobase orientation and C1′-endo sugar puckering. O6B, at 50.0 kJ/mol, has a syn nucleobase orientation and C2′-endo sugar puckering. O6D, at 89.6 kJ/mol, has an anti nucleobase orientation and C3′-exo sugar puckering.

Effects of the 2′-Substituent on the Protonated Adenine and Guanine Nucleosides

Visual comparisons of selected comparable conformers of [Adofl+H]+, [Ado+H]+, and [dAdo+H]+ are given in Figure S9 and of conformers of [Guofl+H]+, [Guo+H]+, and [dGuo+H]+ in Figure S10. Comparisons of conformer energetics and structures are provided in Table 3 and Table 4. The purine [Nuofl+H]+ conformers are highly parallel to their counterpart [Nuo+H]+ and [dNuo+H]+ conformers. The 2′- and 3′-hydroxy substituents of [Ado+H]+ and [Guo+H]+ allow for stabilization of a wider variety of hydrogen-bonding interactions in their sugar moieties than for [Adofl+H]+ and [Guofl+H]+. In [Adofl+H]+, [Ado+H]+, and [dAdo+H]+, N3 protonation is preferred over N1 protonation by ~ 22 kJ/mol, which is favored over N7 protonation by ~ 13 kJ/mol or ~ 35 kJ/mol relative to N3 protonation. Implicit solvation lowers the Gibbs energies predicted for N1 protonation of [Adofl+H]+, [Ado+H]+, and [dAdo+H]+ to ~ 5 kJ/mol above the corresponding ground-state conformation [67]. N3 protonation preferentially favors the syn nucleobase orientation, whereas N1 and N7 protonation favor an anti nucleobase orientation. A slight ~ 3 kJ/mol preference for C2′-endo sugar puckering is computed for the N3 and N1 protonated conformers, whereas N7 protonated conformers prefer C3′-endo sugar puckering by ~ 2 kJ/mol. In [Guofl+H]+, [Guo+H]+, and [dGuo+H]+, N7 protonation is preferred over O6 protonation by ~ 36 kJ/mol, and over N3 protonation by ~ 48 kJ/mol. N7 and O6 protonation favor an anti nucleobase orientation, whereas N3 protonation favors a syn nucleobase orientation. C3′-endo sugar puckering is predicted to be preferred over C2′-endo by ~ 2 kJ/mol in N7 protonated conformers, whereas O6 protonation prefers C2′-endo sugar puckering, and the N3 protonated conformers do not exhibit a consistent trend.

The C1′–N9 glycosidic bond lengths decrease slightly from [dNuo+H]+ to [Nuo+H]+ to [Nuofl+H]+ for the purine nucleosides. The C1′–N9 glycosidic bond lengths decrease by ~ 0.008 Å from [dAdo+H]+ to [Ado+H]+ and by ~ 0.001 Å from [Ado+H]+ to [Adofl+H]+. Similarly, the C1′–N9 bond lengths shrink by ~ 0.013 Å from [dGuo+H]+ to [Guo+H]+ and by ~ 0.001 Å from [Guo+H]+ to [Guofl+H]+. These trends are conceptually consistent with the ER-CID survival yield results that suggest increasing glycosidic bond stability from [dNuo+H]+ to [Nuo+H]+ to [Nuofl+H]+. However, the experimental survival yields show a larger increase in CID50% from [Nuo+H]+ to [Nuofl+H]+ than is suggested by the calculated glycosidic bond lengths alone. Parallel trends were observed in the pyrimidine [Nuofl+H]+, [Nuo+H]+, and [dNuo+H]+ survival yield analyses and C1′–N1 glycosidic bond lengths [9].

Shorter and more linear, i.e., nearer to 180°, interactions are indicative of stronger hydrogen bonds [33, 35, 104,105,106]. In nucleoside sugar rings, sugar puckering dictates the geometry, and thus the relative strength, of their intramolecular hydrogen bonds. Purine [Nuofl+H]+ conformers with C2′-endo character have ∠O3′H∙∙∙F angles of ~ 97.5° and lengths of ~ 2.416 Å, whereas conformers with C3′-endo or C4′-exo puckers have ∠O3′H∙∙∙F angles of ~ 106.8° and lengths of ~ 2.292 Å (see Table 4). Red shifting in both the purine [Nuofl+H]+ O3′–H stretches occur due to these hydrogen-bonding interactions with fluorine. The longer and less direct O3′H∙∙∙F interactions of C2′-endo conformers red shift the O3′–H stretches ~ 24 cm−1 away from the O5′–H stretches, whereas the shorter and more direct O3′H∙∙∙F interactions of C3′-endo and C4′-exo conformers red shift the O3′–H stretches ~ 44 cm−1 away from the O5′–H stretches. Parallel intramolecular OH∙∙∙O hydrogen-bonding interactions are observed in the sugar moiety of the experimentally populated [Nuo+H]+ conformers [54, 67]. Conformers with O2′H∙∙∙O interactions were generally calculated as lower in Gibbs energy than conformers with O3′H∙∙∙O interactions. These O2′H∙∙∙O interactions have angles of ~ 114.5° and lengths of ~ 2.123 Å regardless of their sugar puckering (see Table 4). Therefore, these [Nuo+H]+ OH∙∙∙O hydrogen bonds are shorter and more linear than those of the [Nuofl+H]+ OH∙∙∙F hydrogen bonds and thus provide greater stabilization. These calculations are consistent with condensed phase studies, which have generally demonstrated OH∙∙∙O interactions to be shorter and more stable than parallel OH∙∙∙F interactions [32,33,34,35, 38]. Regardless, no O2′–H stretch red shifting dependence on the sugar puckering or ∠OH∙∙∙O angle is observed in the experimental and predicted [Nuo+H]+ IRMPD spectra [54, 67]. Parallel observations were found for the pyrimidine nucleosides, where geometric variations in the O3′H∙∙∙F interactions red shift the O3′–H stretches to different degrees, whereas OH∙∙∙O interactions display no such variation [9, 69,70,71]. Thus, the red shifting of the O3′–H stretches observed in the [Nuofl+H]+ spectra must result both from the strength of the hydrogen bonds and the highly electronegative character of fluorine.

Conclusions

The structural and energetic effects of 2′-fluoro modification on the purine nucleosides were investigated using IRMPD action spectroscopy, ER-CID with survival yield analysis, and electronic structure calculations. Results for the [Adofl+H]+ and [Guofl+H]+ are reported here and compared with previous work on their counterpart canonical purine systems [Ado+H]+, [Guo+H]+, [dAdo+H]+, and [dGuo+H]+ [67, 68], as well as with [Cydfl+H]+ and [Urdfl+H]+ [9]. Unimolecular dissociation of [Adofl+H]+ and [Guofl+H]+ via IRMPD and CID activation results in cleavage of their C1′–N9 glycosidic bonds with retention of the excess proton by the nucleobase; no other fragmentation pathways were observed. Based on survival yield analysis of the ER-CID experiments, 2′-fluoro modification of the protonated DNA and RNA purine nucleosides strengthens their C1′–N9 glycosidic bonds. These experimental results are reinforced by the shrinking of the C1′–N9 glycosidic bonds predicted across the [dNuo+H]+, [Nuo+H]+, and [Nuofl+H]+ series. This behavior can be further explained based on increasing destabilization of the proposed oxocarbenium ion intermediates [53, 54] with increasing electronegativity of the 2′-substituent: H < OH < F. In general, the absorption band identities and resonant vibrational frequencies of the experimental IRMPD spectra of [Adofl+H]+ and [Guofl+H]+ are parallel to their analogous DNA and RNA forms. However, important differences are also apparent. The [Nuofl+H]+ systems have higher IR absorption efficiency, resulting in greater IRMPD yields compared with the [Nuo+H]+ and [dNuo+H]+ systems. Additionally, the relative peak intensities in most of the fingerprint region are much greater for [Nuofl+H]+ than [Nuo+H]+ and [dNuo+H]+. Evidence for intramolecular gas-phase O3′H∙∙∙F hydrogen bonds is found in the hydrogen-stretching region of the IRMPD spectra of [Adofl+H]+ and [Guofl+H]+ and allows for reliable sugar puckering determinations, which are not possible for the canonical analogues. Recent studies on the pyrimidine [Nuofl+H]+ also displayed clear spectroscopic evidence for gas-phase O3′H∙∙∙F interactions [9]. For [Adofl+H]+, the N3A and N1A conformers are populated experimentally, with a slight preference for N3A over N1A suggested by the spectral comparisons. N3A, the ground-state conformation, has a syn nucleobase orientation and C2′-endo sugar puckering, whereas N1A, at 20.3 kJ/mol, has an anti nucleobase orientation and C2′-endo sugar puckering. C3′-endo and C4′-exo sugar puckerings, such as in N3B and N1B, were not present in the experimental population. For [Guofl+H]+, the N7A and N7B conformers are dominant in the experimental population with a slight preference for N7B. O6A is also suggested to be a minor contributor to the experimental population. N7A, the ground-state conformation, has an anti nucleobase orientation and C3′-endo sugar puckering, whereas N7B, at 1.4 kJ/mol, has anti nucleobase orientation and C2′-endo sugar puckering. O6A, at 33.9 kJ/mol, has an anti nucleobase orientation and C2′-endo sugar puckering. Overall, 2′-fluoro modifications to the canonical DNA and RNA purine nucleosides impart minimal structural changes and result in overall highly parallel conformers among the analogous [Nuofl+H]+, [Nuo+H]+, and [dNuo+H]+ systems [67, 68]. In contrast, substantial electronic effects are produced by the highly electronegative 2′-fluoro modification and are readily apparent in the IRMPD, ER-CID, and theoretical results.