Negative Electron Transfer Collision-Induced Dissociation of G-Quadruplexes: Uncovering the Guanine Radical Anion Loss Pathway

Abstract

G-quadruplex (G4) DNA can form highly stable secondary structures in the presence of metal cations, and research has shown its potential as a transcriptional regulator for oncogenes in the human genome. In order to explore the interactions of DNA with metal cations using mass spectrometry, employing complementary fragmentation methods can enhance structural information. This study explores the use of ion–ion reactions for sequential negative electron transfer collision-induced dissociation (nET-CID) as a complement to traditional ion-trap CID (IT-CID). The resulting nET-CID data for G4 anions with and without metal cations show an increase in fragment ion type diversity and yield of structurally informative ions relative to IT-CID. The nET-CID yields greater sequence coverage by virtue of fragmentation at the 3′-side of thymine residues, which is lacking with IT-CID. Potassium adductions to backbone fragments in IT-CID and nET-CID spectra were nearly identical. Of note is a prominent fragment resulting from a loss of a 149 Da anion seen in nET-CID of large, G-rich sequences, proposed to be radical anion guanine loss. Neutral loss of neutral guanine (151 Da) and deprotonated nucleobase loss (150 Da) have been previously reported, but this is the first report of radical anion guanine loss (149 Da). Confirmation of the identity of the 149 Da anion results from the examination of the homonucleobase sequence 5′-GGGGGGGG-3′. Loss of a charged adenine radical anion at much lower relative abundance was also noted for the sequence 5′-AAAAAAAA-3′. DFT modeling indicates that the loss of a nucleobase as a radical anion from odd-electron nucleic acid anions is a thermodynamically favorable fragmentation pathway for G.

Graphical Abstract

INTRODUCTION

Guanine-rich sequences of DNA can form four-stranded G-quadruplexes (G4) in the presence of metal cations, usually K⁺ or Na⁺ under physiological conditions.¹ Four guanine bases can form a guanine tetrad held together by Hoogsteen-hydrogen bonds (Figure 1A), the stacking of which is stabilized through the coordination of central cations between tetrads by the guanine carbonyl groups (Figure 1B). Potassium is the most common cation in G4 structures because it is favorable for coordination in the quadruplex core and has the highest cellular abundance.² DNA G4s can be found throughout the human genome, enriched in specific regions such as the telomeres and oncogene promoter regions.³ Moreover, they can be involved in critical processes including transcription, translation, and replication.⁴ Thus, DNA G4s are of functional importance and have emerged as promising drug targets.^3,5 For example, G4 DNA in oncogene promoters has been shown to be targetable by small molecule ligands to kill cancer cells.^6,7

Figure 1.

(A) Arrangement of a guanine tetrad in the presence of a potassium cation. (B) Multiple guanine tetrads stacked on top of each other result in a G-quadruplex structure, with the MycG4 sequence as an example. Guanine bases involved in the quadruplex structure are represented as orange parallelograms, and potassium ions are represented as blue circles. (C) MS fragmentation nomenclature of nucleic acids based on phosphate backbone cleavage.⁸ (D) Structures of the four common DNA bases guanine (G), adenine (A), cytosine (C), and thymine (T).

In the field of mass spectrometry, the study of G4 DNA can determine folded and unfolded G4 states through collisional cross section measurements using ion mobility MS,^9–11 and the stoichiometry of ligand binding to G4 DNA,^11–14 as it has been shown that the G4 structure is largely preserved from the solution phase to the gas phase when the structure is rigid.¹⁵ Additionally, tandem MS of nucleic acids can also reveal primary sequence information,¹⁶ modification locations in oligomers,^17,18 and structural characteristics of G4-folded DNA using low energy collision-induced dissociation kinetics.¹⁹

Tandem MS of nucleic acids can provide sequence information, but due to the limited number of bases compared to peptide residues (4 compared to 20, respectively), the probability of generating isobaric product ions increases.²⁰ As with many bioion structure characterization applications, the use of different precursor ion-types and complementary dissociation methods can overcome sequencing challenges associated with a single ion-type and dissociation method.^21,22 Due to the high acidities of nucleic acids, precursor ions are usually generated in the negative ion mode. The most commonly used nomenclature for the fragmentation products of nucleic acids,^8,16 based on fragmentation along the phosphate backbone, is provided in Figure 1C.

Aside from backbone fragmentation, the losses of nucleobases are commonly observed from DNA anions. The charge state of the sequence determines whether the nucleobase is lost as a neutral moiety or as an anion, as prominent charge states formed from native-like solution conditions favor neutral base loss, and Coulombic repulsion in relatively high charge states favors anionic base loss.^16,23 The mechanisms and products of neutral and anion base loss are well understood.^16,23 For G-quadruplex DNA anions of moderate-to-low charge, a neutral guanine base loss peak is the primary fragmentation pathway as the sequences are guanine-rich.^19,24

One of the most common fragmentation methods for nucleic acids is ion-trap collision-induced dissociation (IT-CID). For DNA anions, the loss of a nucleobase (B) followed by cleavage to yield complementary a-B and w ions is the dominant dissociation pathway for even-electron ions. When the base is cleaved from the sugar ring, this forms a double bond on the sugar ring, which, in combination with a phosphate backbone “a-type” fragmentation, forms a stable furan ring.¹⁶ One of the drawbacks of IT-CID of DNA is the absence of backbone fragmentation on the 3′-end of thymine bases. Due to thymine’s low proton affinity,^16,25 proton transfer from a backbone phosphate, which is proposed to be the first step in nucleobase loss, is less likely.^26–28 Another drawback of IT-CID is the low fragment diversity of backbone fragmentation (only a-B and w ions), increasing the chance for isobaric backbone fragments.

Ultraviolet photodissociation (UVPD) is another method of fragmentation applied to DNA anions, leading to detachment of an electron followed by extensive fragmentation, resulting in a diversity of fragment types and modification preservation.²⁰ Fragmentation can occur spontaneously following electron detachment, but subsequent activation of the odd-electron product is usually desirable to generate abundant sequence-informative product ions.²⁰ Such an approach using electron photodetachment (ePD) followed by IT-CID, referred to herein as ePD-CID, has been applied to anions of G-quadruplex DNA and G-quadruplex-ligand complexes.²⁹ An alternative to ePD is electron transfer from a multiply charged anion to a cation in the gas phase,³⁰ referred to as negative electron transfer (nET). An attractive feature of nET is that the efficiency of the ion/ion reaction can approach 100% under readily achievable conditions. As in the photodetachment experiment, dissociation can occur spontaneously, but additional activation is usually desirable. For example, activated-ion negative ETD (AI-nETD), incorporating IR photoactivation with nETD, has been recently used to sequence RNA with complete sequence coverage for RNA up to 21 nucleotides.³¹ Electron detachment dissociation (EDD) is another method for radical fragmentation of acidic species, where high energy electrons interact with an anionic analyte, removing an electron from the analyte.³² This method has been used in multiple oligonucleotide studies to determine sequence information, increase fragment diversity and minimize secondary fragmentation.^33–35 Many studies^36–38 have explored common radical cation fragments that occur from these various radical fragmentation methods, along with radical anion products^29,39 that form due to oligonucleotides’ negatively charged backbone.

The ion–ion analog to the ePD-CID experiment is nET-CID. This approach has been used in previous studies for single stranded oligonucleotide sequences⁴⁰ and modified DNA and RNA¹⁷ but has not been applied to systems with particularly stable secondary structures, like G-quadruplexes or G-quadruplex-ligand complexes. In this work, we have applied IT-CID to even-electron and odd-electron G-quadruplex DNA anions (i.e., IT-CID versus nET-CID) with and without stabilizing K⁺ ions. This work indicates that nET-CID provides more abundant sequence-informative ions and more complete primary sequence information compared to IT-CID. Furthermore, a reaction channel leading to loss of a guanine radical anion, which is unique to radical anions, is also observed to be prominent and appears to be sensitive to G-quadruplex formation. The model G-quadruplex system used here, MycG4,^41,42 represents the major G-quadruplex found in the MYC gene promoter sequence (Figure 1B), where it acts as transcriptional silencer.^5,43 It is a parallel G-quadruplex for which the parallel oriented G-runs are connected by propeller loops of 1:2:1 nt length.⁴² The MYC oncoprotein is overexpressed in many cancers and plays an important role in the control of cell growth and apoptosis.^44,45 Furthermore, the stabilization of the MycG4 by small molecules downregulates c-Myc expression and induces cancer cell death.^3,43,46

METHODS

Materials.

The MycG4 sequence 5′-TGAGGGTGGGTAGGGTGGGTAA-3′ (MW: 7013 Da) was synthesized and purified as described in detail previously.⁴⁷ Solid-phase DNA synthesis was performed using phosphoramidites (Glen Research, USA) on an Expedite 8909 nucleic acid synthesis system (Applied Biosystems, USA). MycG4 was purified with reverse-phase Micropure II columns (BioSearch Technologies, U.K.) and several rounds of dialysis. MycG4 samples were reconstituted in Optima LC/MS grade water (Fisher Scientific, USA), creating a stock solution of 100 μM. The sample underwent desalting three times via centrifugation with Optima LC/MS grade water using a 3 kDa molecular weight cutoff (MWCO) Amicon Ultra 0.5 mL filter (Millipore Sigma, USA). The recovered sample was then diluted back to 100 μM in water. For MS analysis, solutions comprised 10 μM MycG4, 37.5 mM trimethylammonium acetate (TMAA) buffer (Santa Cruz Biotechnology, USA), and 10% (v/v) Optima LC/MS grade methanol (Fisher Scientific, USA).⁹ To create the G-quadruplex folded structures, 0.25 mM KCl (Fisher Scientific, USA) was used.

DNA sequences 5′-GGGGGGGG-3′ (MW = 2570.5 Da), referred to as G8, 5′-AAAAAAAA-3′ (MW = 2442.5 Da), referred to as A8, 5′-TTTTTTTT-3′ (MW = 2370.8 Da), referred to as T8, and 5′-CCCCCCCC-3′ (MW = 2250.4 Da), referred to as C8 were ordered from Integrated DNA Technologies (Coralville, IA, USA). These samples were reconstituted to 100 μM in water, then diluted further for MS analysis, comprised of 10 μM DNA, 37.5 mM TMAA buffer, 10% (v/v) dimethyl sulfoxide (DMSO), and 10% (v/v) Optima LC/MS grade methanol (Fisher Scientific). DMSO was added as a supercharging reagent for the DNA. A rubrene (Millipore Sigma, USA) stock solution of 1 mg/mL in dichloromethane (Fisher Scientific, USA) was created and then diluted to 0.3 mg/mL in 3:7 dichloromethane:acetonitrile (Fisher Scientific, USA) for MS analysis.

Density Functional Theory.

Density functional theory (DFT)^48–51 was used at the B3LYP/6–311++G(d,p) level using Gaussian 16⁵² for molecule modeling. The total electronic (ε₀) and thermal enthalpies (H_corr) of products and reactants were calculated to solve for the change of enthalpies of base loss reactions (ΔH_rxn). The equation for reaction enthalpies used is shown in eq 1. Multiple variations for each structure were made with varying charge and radical positions, and the variation resulting in the lowest ΔH_rxn for each reaction was used.

$$\Delta H_{\text{rxn}}=\sum {\left({\epsilon }_0+H_{\text{corr}}\right)}_{\text{product}}-\sum {\left({\epsilon }_0+H_{\text{corr}}\right)}_{\text{reactants}}$$ (1)

Mass Spectrometry.

This study uses a TripleTOF 5600 quadrupole time-of-flight mass spectrometer (SCIEX, Canada) that has been modified for ion–ion reactions.^53,54 Anions and cations formed alternately for sequential injection via pulsed dual nanoelectrospray ionization.⁵⁴ For negative electron transfer CID (nET-CID), an ion–ion reaction consists of the anionic DNA [M − nH]ⁿ⁻ reacting with the rubrene radical cation [R]^+•. The analyte anions and the reagent radical cations are generated from nESI emitters, mass selected in Q1, and transferred to the q2 linear ion trap for mutual storage (applying AC on both entrance and exit lenses of q2). The ion–ion reaction for nET results in electron transfer from the anion to the cation, yielding charge-reduced radical anion products. After the ion–ion reaction, sequential resonance ejection ramps in q2 were used to mass-select the resulting product ions for MSⁿ experiments.⁵⁵ The radical anion DNA then undergoes IT-CID (nitrogen gas) in q2, resulting in more diverse fragmentation types compared to IT-CID.^17,40 Each scan for nET-CID experiments takes less than a second. For IT-CID experiments, the anionic DNA were mass selected in Q1 and fragmented in q2 using ion trap CID at amplitudes that drive the dissociation reacts at rates of 6–12 s⁻¹. Nucleic acid fragmentation masses of the sequence MycG4 were determined using the Mongo Oligo Mass Calculator v2.06 (http://rna.rega.kuleuven.be/masspec/mongo.htm).⁵⁶

RESULTS AND DISCUSSION

IT-CID of Even- and Odd-Electron G4 Anions Devoid of K⁺.

Product ion spectra from IT-CID of even- and odd-electron 5⁻ precursor anions of MycG4 that are devoid of metal ions were collected to serve as benchmarks for interpreting the data derived from metal-containing ions (see Figure 2). Both IT-CID (Figure 2A) and nET-CID (Figure 2B) spectra of the DNA anions show abundant neutral loss (NL) of a single guanine base (NL = 151 Da) along with loss of a single adenine base (NL = 135 Da).

Figure 2.

(A) IT-CID of [MycG4 − 5H]⁵⁻ and (B) nET-CID of [MycG4 − 6H]^5−• for the potassium-deficient MycG4 (lightning bolt symbol) precursor. The inset of each spectrum details the neutral base loss peaks. Y-axis zoom-in of (C) IT-CID (33×) and (D) nET-CID (3×) fragmentation of the unbound MycG4 sequence (lightning bolt symbol) precursor. Emergence of charged guanine loss and more diverse fragment types are seen in nET-CID fragmentation.

When the y-axis is expanded (Figure 2C,D), many fragment ions of low abundance are apparent. In the case of the even-electron precursor ion (Figure 2C, 33-fold y-scale expansion), the main product ions aside from base loss arise from sequential cleavage of the backbone to yield a-B ions and w ions, as mentioned above. The fragmentation map in the Figure 2C inset shows the identified fragment ions from the even-electron MycG4 anions, in which IT-CID generated fragments between every residue on the phosphate backbone except for the 3′ side of the five thymine residues. Figure 2D provides the roughly 3-fold y-scale expansion of the nET-CID spectrum of the odd-electron MycG4 anions, which shows mixtures of first- and second-generation product ions. For example, complementary a-B and w ions that are formed after base loss are apparent, as are z, a, and d ions, which can be formed directly from the precursor in a radical-induced process. The sequence inset shows that nET-CID yields 100% sequence coverage by producing a-ions and some d-ions at the 3′-sides of the internal thymine residues, in addition to cleavages between all of the other residues.

Aside from the difference in sequence coverage and backbone fragment ion types generated from each of the precursor ion types (i.e., [M − 5H]⁵⁻ versus [M − 6H]^5−•), the precursor ion generated via ion/ion electron transfer (i.e., [M − 6H]^5−•) shows a relatively abundant loss of a (G – 2H)^−• anion (149 Da) (see Figure S1B). It is not uncommon to observe losses of deprotonated bases upon IT-CID of relatively highly charged even-electron oligonucleotide anions (e.g., loss of (G − H)⁻, 150 Da).⁵⁷ However, the loss of a radical anion guanine base of 149 Da has, to our knowledge, not been previously noted from either even-electron or odd-electron DNA precursor ions. The mass loss of 149 Da indicates that the DNA strand gains two hydrogens from the guanine base loss and is specific to the presence of the radical on the DNA.

IT-CID of Even- and Odd-Electron G4 Anions Containing Two K⁺ Ions.

As indicated above, it is known that at least some elements of the G4 structure can be preserved in the gas phase, and previous studies have shown that the addition of two potassium ions to MycG4 indicates a G4 structure.¹⁹ Figure 3A,C shows that IT-CID of the G4 structure [M + 2K − 7H]⁵⁻ yields similar fragmentation results to those of the K⁺-devoid DNA (i.e., [M − 5H]⁵⁻) at the CID amplitudes used here, where neutral loss of a single guanine is the base peak (Figure 3A,C). There appears to be fragmentation at the 3′-end of the thymines, but the products are of minor abundance. With nET-CID of the G-quadruplex structure (Figure 3B,D), fragmentation appears slightly different from that of its K⁺-devoid counterpart. From the K⁺-devoid DNA results, nET-CID increases the sequence coverage, but with the [M + 2K − 7H]⁵⁻ G-quadruplex structure, the nET-CID results show a slight fragmentation gap near the middle of the sequence. When comparing the distribution of potassium adducts along the sequence for IT-CID vs nET-CID (Figure S2), there is not a significant difference between potassium location.

Figure 3.

(A) IT-CID of [MycG4 + 2K − 7H]⁵⁻ and (B) nET-CID of [MycG4 + 2K − 8H]^5−• for the G4-folded MycG4 + 2K (lightning bolt symbol) precursor. Inset of each spectrum details the neutral base loss peaks. Y-axis zoom-in of (C) IT-CID (10×) and (D) nET-CID (1.8×) fragmentation of the G4 + 2K structure of MycG4 (lightning bolt symbol) precursor. Emergence of charged guanine loss and more diverse fragment types seen in nET-CID fragmentation. Green dotted lines represent nET-CID fragments that are identical to the IT-CID fragments. In the nET-CID spectrum, the peak directly to the right of the precursor is a preexisting sodium adduction.

Similar to its K⁺-devoid counterpart, [M − 6H]^5−•, the nET-CID spectrum of the [M + 2K − 8H]^5−• ion shows a prominent signal corresponding to a loss of a radical anion guanine base (149 Da) leading to the charge reduced peak [M + 2K − 8H − (G − 2H)^−•]⁴⁻. The 149 Da anion loss is not as prominent in the potassium-bound G4 structure compared to the potassium-devoid MycG4, which could indicate that G-base stabilization tends to inhibit the radical anion guanine pathway (compare Figure 2B with Figure 3B).

Abundances and Diversity of Structurally Informative Fragment Ions: nET-CID versus IT-CID.

As indicated in the comparisons of Figures 2 and 3, the relative abundances of the structurally informative product ions are greater in the nET-CID data compared with the IT-CID data, as reflected by the 5 to 10-fold greater y-axis expansions required for the IT-CID data. This is due, at least in part, to the fact that structurally informative fragments from the even-electron ions requires sequential fragmentation. Figure 4 provides summed abundances of the types of fragmentation generated via IT-CID and nET-CID for potassium-devoid and -adducted MycG4 anions. For IT-CID (Figure 4A,C), by far the dominant product ion signal arises from neutral guanine base loss, with all backbone fragment abundances appearing minor in comparison. When comparing Figure 4A–C, it appears that there is a relative decrease in neutral guanine loss in the MycG4 + 2K anions versus MycG4, which could arise from G-base stabilization in G4 structures, but this difference is minor at the relatively high IT-CID amplitudes used in this work. Figure 4B,D shows that for nET-CID, neutral G loss significantly decreases compared to all other fragmentation types in favor of backbone fragmentation and the anion radical guanine loss. Because of this, backbone fragmentation products are relatively more prominent, thereby facilitating sequence determination. Of interest is the prominence of the anion radical guanine loss fragment peak, which by itself constitutes 15% of the total fragment signal for potassium devoid MycG4. This decreases in relative abundance for the MycG4 + 2K anions to 3.7% of the total fragment signal but remains a prominent peak on its own. The lower abundance of the anion radical loss peak from the MycG4 + 2K ion is consistent with a degree of G stabilization via the G4 structure. This peak is not unique to G4-forming DNA, as Figure S3 shows this 149 Da loss with a non-G4-forming DNA sequence. Additionally, Figure S4 indicates that it is present regardless of the dissociation rate. Because this product appears to be a general phenomenon, further exploration is necessary.

Figure 4.

Percentage fragmentation intensity for major fragmentation types in potassium-devoid MycG4 (A) IT-CID and (B) nET-CID, and potassium adducted MycG4 (C) IT-CID and (D) nET-CID. Percentage intensity (y-axis) based on identified fragments only.

nET-CID of Small Poly-G or Poly-A DNA Sequences.

Due to the prominence of the 149 Da anion loss (i.e., [G − 2H]^−•) signal in the nET-CID of G-rich DNA sequences, small model DNA sequences were examined to study this process further. Monoisotopic peaks and isotope distributions for the precursor radical anion DNA from MycG4 (i.e., [MycG4 − 6H]^5−•) and the 149 Da loss fragments can be seen in Figure S1. These spectra show that the monoisotopic peaks appear to line up with a loss of 149 Da, and the isotope distributions appear to reasonably follow the predicted distribution based on the chemical formula (calculated by the program IsoPro), showing it is unlikely that there is significant isobar overlap. However, due to the relatively large m/z ratio of the ions, the presence of overlapping isobars could be difficult to detect. In addition, due to the large m/z of the ions, the low mass cutoff used for the precursor ion activation was too high to see the complementary m/z 149 anion fragment. To confirm the existence of the 149 Da loss, nET-CID was conducted on smaller homonucleobase DNA.

In order to confirm the existence of the prominent 149 Da loss fragment found in the G-rich MycG4 sequence and the possibility for analogous anion losses from the other nucleobases, sequences 5′-GGGGGGGG-3′ (MW = 2570.5 Da), referred to as G8, and 5′-AAAAAAAA-3′ (MW = 2442.5 Da), referred to as A8, were examined. Figure 5 shows the nET-CID steps for G8 (Figures 5A–C) and A8 (Figure 5D–F), showing the initial ion–ion reaction with anion [DNA − 4H]⁴⁻ and radical cation rubrene (Figure 5A,D), isolation of ion–ion reaction product radical anion [DNA − 4H]^3−• (Figure 5B,E), and IT-CID of [DNA − 4H]^3−• (Figure 5C,F). When comparing the initial ion–ion reactions of G8 vs A8, the only difference is the appearance of a rubrene adduct (R) to the second-generation ion–ion reaction product. This indicates that both guanine and adenine are readily able to release a single electron to the first rubrene that reacts with the product, forming [DNA − 4H]^3−•. However, if a second rubrene reacts with this product, guanine is able to donate a second electron, but adenine is less likely to do so, leading to the formation of the rubrene adduct. Ion–ion reactions with radical cation rubrene were also conducted with sequences C8 (5′-CCCCCCCC-3′, MW = 2250.4 Da) and T8 (5′-TTTTTTTT-3′, MW = 2370.8 Da), seen in Figure S5, but these reactions showed either very little radical anion product (in the case of C8) or no radical anion product at all (in the case of T8). Rather, rubrene adduction is observed to be the prominent ion–ion reaction pathway. These findings are consistent with ePD results^39,58 showing that photodetachment efficiency is highest for the purines (guanine having the lowest PD threshold⁵⁹) and much lower for the pyrimidines.⁶⁰ Consistent with the various ePD studies, the ion–ion electron transfer results involving G-containing multiply charged anions suggest that the electron transferred to the rubrene cation originates from a guanine nucleobase.

Figure 5.

nET-CID spectra for DNA sequence G8 (A–C) and A8 (D–F), starting with the initial ion–ion reaction between the [DNA − 4H]⁴⁻ anion and the rubrene radical cation (A, D). Isolation of the radical anion DNA product (B, E) via resonance ejection ramp isolation is observed. Due to the radical product’s fragility, fragmentation can already be observed. IT-CID of the radical anion DNA product (C, F) shows extensive backbone fragmentation. Due to the sequence base homogeny, most backbone fragments are isobars. The peaks of interest representing anion base loss (blue inset) and their complementary fragments (orange inset) can be observed. The prominent unlabeled peak in (F) is an ion–ion reaction product that was not fully removed during the isolation step.

Isolation of the [DNA − 4H]^3−• products (Figure 5B,E) suggests that the radical product is relatively fragile, as low abundance fragments are formed via off-resonance power absorption from the resonance ejection ramps.⁵⁵ nET-CID of [DNA − 4H]^3−• (Figure 5C,F) shows significant backbone fragmentation. Due to the use of homonucleobase DNA, almost all backbone fragments are isobaric with different fragment types, such as w/d ions and a/z ions. Of interest is the doubly charged fragment highlighted in orange (m/z 1208.7) in Figure 5C, which is the fragment analogous to the 149 Da loss fragment found in the MycG4 nET-CID spectrum. This fragment is much smaller in relative abundance than the fragment found in MycG4, possibly due more extensive secondary fragmentation in the DNA ions with fewer degrees of freedom. Comparing the monoisotopic peaks of the precursor [G8 − 4H]^3−• and this fragment reveals a loss of 149 Da. Furthermore, highlighted in blue in Figure 5C, the complementary m/z 149 fragment is observed, providing confirmatory evidence for 149 Da anion loss.

To probe for an analogous adenine radical anion loss from odd-electron oligonucleotide anions, the nET-CID of A8 (Figure 5F) was performed. In this case, even-electron anion loss (134 Da loss leading to m/z 1152.2), highlighted in orange, gives rise to most of the observed base anion loss. Even-electron adenine anion loss is commonly seen with the IT-CID of highly charged DNA oligomer anions containing adenine, and the complementary fragment at m/z 134.1 confirms this. Interestingly, there is evidence for a low abundance product ion at m/z 133.1, which would be analogous to guanine’s 149 Da loss. Nevertheless, the much greater relative abundance of the m/z 134.1 product ion indicates that this pathway is much less favorable. Due to the low abundance of radical anion product formed for C8 and T8 from the initial ion–ion reaction, CID of these peaks was not conducted.

Density Functional Theory (DFT) Modeling of Base Loss Reaction Enthalpies.

Density functional theory (DFT) calculations can be used to estimate the reaction enthalpies for base loss. Experiment shows that radical anion guanine is a major pathway in nET-CID of G-rich DNA. In order to determine the most stable structure of the 149 Da anion and to compare neutral, anion, and radical anion loss of guanine, DFT modeling was used.

Anions of the 5′-mononucleotides, GMP and AMP, were used as model structures for the precursor species (Figure 6) for guanine- and adenine-containing precursors, respectively. Because fragmentation from even-electron sites can compete with fragmentation from radical sites in the oligonucleotide ions studied here, DFT calculations were performed for base loss channels from [GMP − H]¹⁻ and [AMP − H]¹⁻, to represent base loss channels from even-electron sites, and [GMP − 2H]^1−• and [AMP − 2H]^1−•, to represent odd-electron sites. Structures used to determine the reaction enthalpies for base loss can be seen in Figures S6−S8. Multiple possible structures for the different base loss types were examined to determine which structure was most stable, and the base structures shown in Figure 6 reflect the most stable guanine loss structures for neutral G loss, G⁰, even-electron anion G loss, [G − H]¹⁻, and radical anion G loss, [G − 2H]^1−•. Table 1 summarizes the results for five relevant base loss reactions for the model ions. The first two rows of Table 1 summarize the enthalpies for [B − H]⁻ versus B⁰ loss from the even-electron [GMP − H]¹⁻ and [AMP − H]¹⁻ ions. The bottom three rows summarize the enthalpies associated with the losses of B⁰, [B − H]¹⁻, and [B − 2H]^1−• from the corresponding odd electron precursors. We note that it is difficult to determine how much neutral base loss contributes to the spectra of Figure 5 because most of the sequence-informative ions can result from either neutral or charged base loss. Comparisons of the relative propensities for [B − H]¹⁻ versus [B − 2H]^1−• loss, however, can be made directly from the relative abundances of the [B − H]¹⁻ versus [B − 2H]^1−• ions (see expanded regions outlined in blue in Figure 5). It is noteworthy that although G⁰ loss is the most favorable pathway (ΔH = 43.0 kcal/mol), the loss of [G − 2H]^1−• (ΔH = 56.3 kcal/mol) is significantly more favored than [G − H]¹⁻ loss (ΔH = 68.8 kcal/mol). These theoretical results are in agreement with our experimental results, as Figure 2D and Figure 3D show a combination of products that originate from G⁰ loss and [G − 2H]^1−• loss. In contrast, while A⁰ loss is also the most favorable pathway (ΔH = 34.3 kcal/mol) for radical anion adenine, [A − 2H]^1−• loss (ΔH = 58.6 kcal/mol) and [G − H]¹⁻ loss (ΔH = 58.4 kcal/mol) are comparable in favorability, unlike with guanine. Additionally, the difference in ΔH between B⁰ and [B − 2H]^1−• for guanine is 13.3 kcal/mol, and for adenine this difference is 24.3 kcal/mol, indicating that [G − 2H]^1−• is more likely to occur than [A − 2H]^1−•. The loss of neutral adenine, is highly favored from both AMP ion types, which may well be related to the reported tendency for the preferred loss of adenine either as a neutral or charged species from multiply charged mixed base oligonucleotides.⁶¹

Figure 6.

Proposed final structures for the loss of neutral G⁰, anion [G − H]¹⁻, or radical anion [G − 2H]^1−• guanine from the radical anion of GMP [GMP − 2H]^1−• based on the lowest ΔH values calculated from DFT modeling at the B3LYP/6–311++G(d,p) level. Adenine equivalents can be seen in Figure S8.

Table 1.

Reaction Enthalpies (ΔH) for Different Base Loss Pathways Are Based on DFT Modeling at the B3LYP/6–311++G(d,p) Level^a

guanine loss reaction	ΔH (kcal/mol)	adenine loss reaction	ΔH (kcal/mol)
[GMP − H]1− → MP1− + G0	38.9	[AMP − H]1− → MP1− + A0	44.3
[GMP − H]1− → MP0 + [G − H]1−	63.7	[AMP − H]1− → MP0 + [A − H]1−	67.4
[GMP − 2H]1−• → MP1−• + G0	43.0	[AMP − 2H]1−• → MP1−• + A0	34.3
[GMP − 2H]1−• → MP0• + [G − H]1−	68.8	[AMP − 2H]1−• → MP0• + [A − H]1−	58.4
[GMP − 2H]1−• → MP0 + [G − 2H]1−•	56.3	[AMP − 2H]1−• → MP0 + [A − 2H]1−•	58.6

^aNeutral B⁰ and anion [B − H]¹⁻ base loss from the even-electron anion nucleotide-like starting structure [BMP − H]¹⁻ is shown in the first two rows. Neutral B⁰, anion [B − H]¹⁻, and radical anion [B − 2H]^1−• loss from the odd-electron anion nucleotide-like starting structure [BMP−2H]^1−• are shown in the last three rows.

The mononucleotide anions modeled here obviously do not account for the possible role of multiple charging on the propensities for charged versus neutral base loss, higher order structural effects that may be present in oligonucleotides, or the fact that the dissociation reactions are under kinetic control. Nevertheless, the DFT results indicate that the generation of a [G − 2H]^1−• ion from a G-containing oligonucleotide radical anion is a favorable process.

To our knowledge, this is the first report of radical anion guanine base loss as an odd-electron oligonucleotide anion fragmentation pathway. A recent study⁶² involving MSⁿ using ion trap CID of deprotonated mononucleosides with methylated nucleobases reported a 149 Da fragment anion from CID of a deprotonated O⁶-methylated guanine base. The structure of the product ion was proposed to arise from loss of NH from the deprotonated nucleobase. A 2023 review⁶³ pointed out that the loss of CH₃^• is much more likely but the mass accuracy of the measurement could not distinguish the two. The structure proposed in the review is almost identical to the structure indicated in this work. The difference is that the charge and radical locations are reversed. In the current work, the negative charge is localized on oxygen with the radical site on ⁹N whereas the structure proposed in the review localizes the charge on ⁹N and the radical on the oxygen. While the two structures are similar in energy, we assign the charge and radical site locations based on the Mulliken charge distribution for the lowest energy structure determined via DFT modeling of [G − 2H]^1−• seen in Figure S9.

CONCLUSIONS

In this work, the IT-CID and nET-CID behavior of anions derived from a DNA sequence with the potential for G-quadruplex formation was explored. Specifically, 5⁻ even- and odd-electron single stranded M = MycG4 (5′-TGAGGGTGGGTAGGGTGGGTAA-3′) anions were subjected to study using CID amplitudes that give rise to dissociation rates of 6−16 s⁻¹. Precursor ion-types included [MycG4 − 5H]⁵⁻, [MycG4 − 6H]^5−•, [MycG4 + 2K-7H]⁵⁻, and [MycG4 + 2K-8H]^5−•, and all precursor ion-types showed the loss of a neutral guanine to be a major dissociation channel upon IT-CID. However, the odd-electron precursor ions generated via electron transfer yielded both a greater diversity of backbone fragments and a greater relative abundance of sequence-informative product ions. More complete sequence coverage was derived from the odd-electron precursors by virtue of cleavages on the 3′-side of thymine residues, which are known to be inhibited with even-electron precursors. For the [MycG4 − 5H]⁵⁻ and [MycG4 − 6H]^5−• precursors, 100% sequence coverage was obtained for the latter ion, whereas the lack of cleavages at thymine residues resulted in lower sequence coverage for the former ion. Both the IT-CID and nET-CID of the [M + 2K − 8H]^5−• ions provided extensive sequence coverage, albeit with lower contributions from fragmentation toward the middle of the sequence. The odd-electron [MycG4 − 6H]^5−• and [MycG4 + 2K − 8H]^5−• precursor ions also showed a prominent loss of a 149 Da anion, determined to be radical anion guanine base loss [G − 2H]^1−•, which was absent for the analogous even-electron ions. nET-CID experiments conducted on anions of homonucleobase sequences G8 and A8 provided confirmatory evidence of the origin of the 149 Da anion loss. DFT modeling indicates that radical anion guanine base loss from a radical anion precursor is a thermodynamically favorable process for G-containing sequences. The ion–ion reaction results for homonucleobase anions (G8, A8, C8, and T8) (as well as the MycG4 anions) are consistent with photodetachment studies that have demonstrated that electron loss from nucleic acid anions composed of guanidine residues originates from guanine nucleobases. While the guanine radical anion base loss process was noted here using an ion/ion reaction to remove an electron from an even-electron anion, other approaches to remove an electron, such as via photodetachment, electron irradiation, high energy collision excitation, etc., might also be expected to result in the same process.