Important Requirements for Desorption/Ionization Mass Spectrometric Measurements of Temozolomide-Induced 2′-Deoxyguanosine Methylations in DNA

Simple Summary Monitoring chemical action of drugs directly at their molecular target would be particularly valuable for personalized medicine. In temozolomide-exposed biological samples, 2′-deoxyguanosines and O6-methylated species are compounds of interest to monitor chemical effects. However, their analysis can be hampered by molecular interferences. Hereby, desorption/ionization mass spectrometry was evaluated for such investigation. Here, we illustrate that, without following specific requirements in terms of sample preparation and mass spectrometric instrumentation, these analyses are prone to important artefacts. Abstract In clinical pharmacology, drug quantification is mainly performed from the circulation for pharmacokinetic purposes. Finely monitoring the chemical effect of drugs at their chemical sites of action for pharmacodynamics would have a major impact in several contexts of personalized medicine. Monitoring appropriate drug exposure is particularly challenging for alkylating drugs such as temozolomide (TMZ) because there is no flow equilibrium that would allow reliable conclusions to be drawn about the alkylation of the target site from plasma concentrations. During the treatment of glioblastoma, it appears, therefore, promising to directly monitor the alkylating effect of TMZ rather than plasma exposure, ideally at the site of action. Mass spectrometry (MS) is a method of choice for the quantification of methylated guanines and, more specifically, of O6-methylguanines as a marker of TMZ exposure at the site of action. Depending on the chosen strategy to analyze modified purines and 2′-deoxynucleosides, the analysis of methylated guanines and 2′-deoxyguanosines is prone to important artefacts due to the overlap between masses of (i) guanines from DNA and RNA, and (ii) different methylated species of guanines. Therefore, the specific analysis of O6-methyl-2′deoxyguanosine, which is the product of the TMZ effect, is highly challenging. In this work, we report observations from matrix-assisted laser desorption/ionization (MALDI), and desorption electrospray ionization (DESI) MS analyses. These allow for the construction of a decision tree to initiate studies using desorption/ionization MS for the analysis of 2′-deoxyguanosine methylations induced by TMZ.


Introduction
In clinical pharmacology, approaching the anatomical, cellular, and molecular sites of action (SOAs) of drugs allows to monitor the ultimate chemical and biological actions

Synthesis of DHPT
The synthesis is a slightly modified version of a previously published procedure [21] that could not directly be reproduced.
The NMR properties of the product compound were the following: 1  To a solution of 2,3,4,5-tetrakis(3,4-dimethoxyphenyl)thiophene (1.115 g, 1.773 mmol, 1.0 eq) in anhydrous CH 2 Cl 2 (30 mL), was added a solution of BBr 3 (32 mL, 1.0 M in CH 2 Cl 2 , 18 eq) at -78 • C under argon. The resulting wine-colored solution was stirred at this temperature for 1 h and at 0 • C for 3 h. After cooling to -78 • C, the reaction mixture was quenched with MeOH (10 mL) and concentrated to an oil. The mixture was suspended in H 2 O (30 mL) and heated in a 70 • C oil bath for 1 h, during which time a white precipitate formed. The mixture was cooled to 0 • C, filtered, and the solid was washed with H 2 O (4 × 30 mL), and air dried. The resulting powder was dissolved in acetone, concentrated, and then dried on a lyophilizer to give DHPT as a gray powder (650 mg, 71%). Spectroscopic data were consistent with the literature [22].

Mass Spectrometric Analyses
Analyses were performed on a Synapt G2-Si instrument ( Figure 1, Waters Corp, Milford, MA, USA) consisting of orthogonal acceleration (oa)-quadrupole (Q)-ion mobility (IM)-time-of-flight (TOF) MS equipped with either a MALDI source ( Figure 1A) or a DESI source ( Figure 1B) and controlled under MassLynx v4.1 (Waters Corp), as fully described previously [20]. The instrument was used in positive and resolution mode ("W" mode, Figure 1H). We previously described the IM-MS parameters for the analysis of native compounds (no selection of the native compounds or selection of a specific native ion in Q ( Figure 1D), followed by IM separation ( Figure 1F) before MS detection ( Figure 1H), referred as Methods 2 and 3, respectively in [20] and detailed in Table 1) and IM-MS/MS parameters for the analysis of fragments (selection of specific native ions in Q followed by collision-induced dissociation at 20-25 V ( Figure 1E) and subsequent IM separation of the fragments before MS detection, referred as Method 4 in [20,23] and detailed in Table 1). Related acquisition methods applied in this study are further detailed in Table 1. 71%). Spectroscopic data were consistent with the literature [22].

Mass Spectrometric Analyses
Analyses were performed on a Synapt G2-Si instrument ( Figure 1, Waters Corp, Milford, MA, USA) consisting of orthogonal acceleration (oa)-quadrupole (Q)-ion mobility (IM)-time-of-flight (TOF) MS equipped with either a MALDI source ( Figure 1A) or a DESI source ( Figure 1B) and controlled under MassLynx v4.1 (Waters Corp), as fully described previously [20]. The instrument was used in positive and resolution mode ("W" mode, Figure 1H). We previously described the IM-MS parameters for the analysis of native compounds (no selection of the native compounds or selection of a specific native ion in Q ( Figure 1D), followed by IM separation ( Figure 1F) before MS detection ( Figure 1H), referred as Methods 2 and 3, respectively in [20] and detailed in Table 1) and IM-MS/MS parameters for the analysis of fragments (selection of specific native ions in Q followed by collision-induced dissociation at 20-25 V ( Figure 1E) and subsequent IM separation of the fragments before MS detection, referred as Method 4 in [20,23] and detailed in Table 1). Related acquisition methods applied in this study are further detailed in Table 1.  The MALDI-MS analyses were performed as described before [23] using a spiral pattern, and DESI-MS analyses were performed using parameters previously described for the high-performance DESI sprayer [24].

Data Processing
MS spectra and mobilograms were extracted from MassLynx v4.1. Recommendations for reporting results of IM-MS measurements [25] were followed. Because IM is used here as a separation method and not for structural analyses, the drift times (DT) are reported as IM spectrometry (IMS) data. The previously described MobA method [20] was used for data extraction; the mobility peaks specific to the compounds of interest (extracted ion mobilograms (XIM)) were extracted from their specific m/z peak in the combined mass spectra. The obtained XIM were then automatically integrated to retrieve the peak areas of the targeted compounds using MassLynx software [20,23].

MALDI-MS Analyses of Guanine and 2 -Deoxyguanosine
For initial ionization tests with MALDI-MS, non-methylated guanine ( Figure 2) and 2dGO ( Figure 3) were tested using five different MALDI matrices: (i) Graphene, a MALDI matrix allowing for interference-free analysis of small molecules and demonstrated to be efficient for the analyses of guanines and 2dGO [26]; (ii) DHPT, developed for the analyses of low-mass amines including guanines [21]; (iii) 3-HPA, reported for the ionization of oligonucleotides [27]; (iv) 2,5-DHB [20] and (v) CHCA [28], two MALDI matrices universally used for small compounds. Initial analyses of non-methylated guanine and 2dGO were performed using methods Q-152 and Q-268 (Table 1).
Guanine (parent peak at m/z 152.06) gave a clear signal with four out of five MALDI matrices (i.e., intensities > 1.10 5 arbitrary units (a.u.)) ( Figure 2B-E), and only a weak signal could be detected with graphene (i.e., 1.56 × 10 3 a.u., Figure 2A). For 2dGO, the parent peak at m/z 268.11 was only weakly detected with intensities close to those of interfering signals in the control analyses for graphene (9.62 × 10 2 a.u. vs. 1.78 × 10 3 a.u., Figure 3A), DHPT (9.55 × 10 3 a.u. vs. about 3.0 × 10 3 a.u., Figure 3B), and 2,5-DHB (3.68 × 10 4 a.u. vs. 2.08 × 10 4 a.u., Figure 3D). Using 3-HPA, the m/z 268.11 peak was detected for both the control sample and the 2dGO sample ( Figure 3C), with a higher intensity in the 2dGO sample (i.e., 1.10 × 10 5 a.u., twice the intensity of the 3-HPA peak at m/z 269.10) compared to the intensity in the control sample (i.e., 3.52 × 10 4 a.u., half the intensity of 3-HPA peak at m/z 269.10). As for guanine samples, the highest signal for 2dGO was obtained with the CHCA MALDI matrix (1.72 × 10 6 a.u. vs. 8.10 × 10 4 a.u. in control, Figure 3E). The analysis of 2dGO also revealed the presence of another peak at m/z 152 corresponding to the guanine moiety of the compound, thus revealing fragmentation processes during the analyses. Because the quadrupole was set to the parent mass for the analysis (i.e., focus on m/z 268 for the analysis of 2dGO), it seems that the fragmentation event took place in or after the quadrupole ( Figure 1D), and could, thus, be described as PSD. The fragment signal was visible for all MALDI matrices, even when the parent compound was detected with a weak signal ( Figure 3) but PSD events had different impact depending on the MALDI matrix. Using graphene, the intensity of the fragment at m/z 152 was only 10% of the parent intensity ( Figure 3A). In DHPT, 2,5-DHB, and CHCA, the fragment ion represented 28%, 37%, and 29% of the parent peak, respectively ( Figure 3B,D,E). The PSD event was  In order to investigate whether fragmentations could also happen before the quadrupole selection, further analyses were performed using the two best MALDI matrices (i.e., 3-HPA and CHCA) with the full scan and the Q-152 methods (Table 1, Figure 4). With method Q-152, 3.1-fold and 8.6-fold higher intensities could be detected for the fragment m/z 152 using 3-HPA ( Figure 4A) and CHCA ( Figure 4C), respectively, compared to method Q-268, thus suggesting that fragmentation events mainly occur before the quadrupole. This process could take place during the ionization (i.e., ISD, Figure 1A) or after, in the hexapole ( Figure 1A) or the ion guide ( Figure 1C). This was confirmed by analyses with the full-scan method showing predominant fragment peaks: (i) Compared to the fragment peaks obtained with method Q-268 (i.e., 4.2-fold and 7.5-fold higher intensities using 3-HPA ( Figure 4B) and CHCA ( Figure 4D), respectively); (ii) compared to the parent peak at m/z 268.12 on the same mass spectrum (i.e., 7.5-fold and 1.7-fold higher intensities for the fragment using 3-HPA ( Figure 4B) and CHCA ( Figure 4D), respectively). Overall, these results indicated that fragmentation events were dominant as compared to the stable compounds and were mainly taking place early in the ion path (i.e., mainly in the source, Figure 1A,C).

DESI Analyses of 2′-Deoxyguanosine, O6-Methyl-2′-Deoxyguanosine
ISD and PSD events are well known in MALDI-MS, where the fluence of the laser plays a major role [29,30]. In order to test whether these events could be avoided using another DI-MS method, 2dGO was also analyzed by DESI-MS, as well as O6-m2dGO, the critical compound to monitor TMZ action.
Using the parent-focused methods Q-268 and Q-282 (Table 1), DESI-MS data demonstrated that 2dGO ( Figure 5A) and O6-m2dGO ( Figure 5B) were also instable using DI under ambient conditions, with dominant ISD/PSD fragments compared to the native compounds (10.8-fold higher intensity of the m/z 152 peak compared to the m/z

DESI Analyses of 2 -Deoxyguanosine, O6-Methyl-2 -Deoxyguanosine
ISD and PSD events are well known in MALDI-MS, where the fluence of the laser plays a major role [29,30]. In order to test whether these events could be avoided using another DI-MS method, 2dGO was also analyzed by DESI-MS, as well as O6-m2dGO, the critical compound to monitor TMZ action.
Using the parent-focused methods Q-268 and Q-282 (Table 1), DESI-MS data demonstrated that 2dGO ( Figure 5A) and O6-m2dGO ( Figure 5B) were also instable using DI under ambient conditions, with dominant ISD/PSD fragments compared to the native compounds (10.8-fold higher intensity of the m/z 152 peak compared to the m/z 268 peak for 2dGO, and 1.1-fold higher m/z 166 peaks compared to m/z 282 peaks for O6-m2dGO). The analysis of the compounds with methods Q-152 and Q-166 (Table 1) resulted in a 4.1-fold higher signal for the 2dGO fragment ( Figure 5C) and a 10.2-fold higher signal for the O6-m2dGO fragment ( Figure 5D) compared to methods focused on the parent ion, indicating an early fragmentation occurring during ionization or shortly thereafter. Similarly, the analyses in full scan indicated a 19-fold higher intensity for the 2dGO fragment compared to the parent (Figure 5E), and a 3.6-fold higher intensity for O6-m2dGO fragment compared to the parent ( Figure 5F).
Using DI, it appeared that the quantification of 2dGO and O6-m2dGO would be hampered by the intrinsic decay of these compounds in the gas phase. When the decay is very pronounced and the stability of the parent compound is low, the analysis of fragments (purines) may be mandatory in order to reach the highest sensitivity in quantitative assays. Because purines of DNA and RNA bear the same mass, it will be important to consider this information for the sample preparation step, in order to specifically digest DNA instead of both DNA and RNA (e.g., by using DNAse instead of exonucleases for digestion).

Ion Mobility Separation of O6-Methylguanine and N7-Methylguanine
An additional parameter to consider while measuring guanine modifications induced by TMZ is the presence of the N7-mG species, which is the most abundant modified species (80 %) with the same mass as the O6-mG species. Any method for the quantification of O6-mG would necessitate to separate these two species to enable specific analyses. Since DI-MS cannot be coupled to liquid chromatography for separation before ionization, an alternative would consist of post-ionization IM separation. In the present context, we investigated whether N7-mG and O6-mG fragments could be separated using the present instrumentation. Since N7-m2dGO is not a commercially available compound, we performed the analyses of N7-mG, which has a parent peak at m/z 166 ( Figure 6A, N7-mG), and compared the mobility profiles with the m/z 166 fragment of O6-m2dGO corresponding to O6-mG. The XIM of the m/z 166 peak indicated a similar drift time of 0.79 ms for both N7-mG and O6-mG ( Figure 6A); therefore, IM cannot be used for the differentiation of these two species. The last option to differentiate O6 from N7 species would then be the generation of fragments with different m/z and/or DT between guanine, N7-mG, and O6-mG. We, therefore, further fragmented the m/z 166 fragments and two signals appeared as possible signatures of O6-mG that would differentiate it from N7-mG and guanine, m/z 134 and m/z 121 ( Figure 6B-D). Both MS/MS and mobility peaks of m/z 134 from O6-mG ( Figure 6C) interfered with a peak from both guanine ( Figure 6D) and N7-mG ( Figure 6B). The peak at m/z 121 and the related DT peak from O6-mG ( Figure 6C) did not interfere with any signal from guanine ( Figure 6D) but one from N7-mG ( Figure 6B). Although the intensity of the peak was 10-fold lower in N7-mG, it is important to note that up to 16-fold higher concentrations of N7-mG can be found in DNA exposed with TMZ. Therefore, minor signal interferences might still represent important hurdles for quantification assays. However, the IM data indicated a slight shift in the maximum DT (∆DT of 0.08 ms). Although the present IM instrumentation did not provide a clear separation of these two compounds, this is a promising approach for developing quantification assay using instrumentation equipped with ion mobility with higher resolving power. resulted in a 4.1-fold higher signal for the 2dGO fragment ( Figure 5C) and a 10.2-fo higher signal for the O6-m2dGO fragment ( Figure 5D) compared to methods focused o the parent ion, indicating an early fragmentation occurring during ionization or short thereafter. Similarly, the analyses in full scan indicated a 19-fold higher intensity for th 2dGO fragment compared to the parent (Figure 5E), and a 3.6-fold higher intensity f O6-m2dGO fragment compared to the parent ( Figure 5F). Using DI, it appeared that the quantification of 2dGO and O6-m2dGO would hampered by the intrinsic decay of these compounds in the gas phase. When the decay very pronounced and the stability of the parent compound is low, the analysis of fra ments (purines) may be mandatory in order to reach the highest sensitivity in quantit tive assays. Because purines of DNA and RNA bear the same mass, it will be important consider this information for the sample preparation step, in order to specifically dige

Discussion
The development of assays for the quantification of DNA methylations responsible of the biological effect of TMZ (i.e., methylation of guanines at the O6 position) is challenged by the presence of possible sources of interferences from RNA and competitive

Discussion
The development of assays for the quantification of DNA methylations responsible of the biological effect of TMZ (i.e., methylation of guanines at the O6 position) is challenged by the presence of possible sources of interferences from RNA and competitive methylated species induced by TMZ. The development of DI-MS-based methods further adds a layer of difficulty due to the impossibility to separate compounds from the samples before their ionization, e.g., by liquid chromatography. The present characterization of compounds using different DI-MS sources demonstrates that the additional fragmentation, likely due to ISD or early PSD events, further limits the ability to generate quantification assays. However, these observations provide valuable guidance for decisions in initiating developments. First, post-ionization separation of compounds is an instrumental prerequisite to enable the separation of O6-mG from N7-mG species (e.g., using IM instrumentations). In our present context, the IM resolving power of our setup (travelling wave IM spectrometry (TWIMS)) was not sufficient to permit a separation of O6-mG and N7-mG. In addition, the occurrence of fragmentation events should be investigated. In such cases, it should be determined whether it is a quantitatively relevant event and whether it would result in an increase in the lower limit of quantification (LLOQ) of O6-m2dGO above the expected concentrations in the samples. In this case, it should be determined when the fragmentation mainly occurs during the MS analyses, i.e., before or after the quadrupole, in order to set it accordingly. The resolution of quadrupoles is also a critical issue for the development of such quantification assays. Quadrupoles that allow for high-resolution mass selectivity without loss of sensitivity (e.g., in a triple quadrupole) allow for the secure selection of 2dGO from DNA and the exclusion of guanosines (GO) originating from RNA. In a situation where increasing the resolution of quadrupoles reduces analytical sensitivity, the 17-Da molecular weight difference between 2dGO and GO and between O6-m2dGO and O6-mGO may be critical for their analysis with high selectivity. In this case, the sample preparation must be adapted to target digestion only to DNA and not to RNA (e.g., with DNAases).
Based on this knowledge, several options would be available. This is illustrated in Figure 7, which can be used as a guide for future decisions to initiate the development of assays for the monitoring of DNA modifications induced by TMZ. Although in our present context, the IM resolution of TWIMS was not sufficient for the separation of O6-mG and N7-mG, other IM equipment, such as trapped IM spectrometry (TIMS), might provide higher mobility resolving power for a better separation. Preliminary results in our context indicated that IM separation might be more efficient on pseudo-MS3 fragments of O6 and N7-methylated purines than on ISD/PSD fragments of methylated purines. The IM separation of pseudo-MS3 fragments will, however, not be possible with instrumentations having the mobility cell before the quadrupole and collision cells, e.g., with the timsTOFflex (Bruker Daltonics, Billerica, USA) [31]. The success of the separation of O6 and N7-methylated species would rely on the IM separation of methylated guanosine species.
It is noteworthy that Kimura and co-workers observed ISD effects in MALDI on oligonucleotides with nucleobase derivatives, which may be attributed to the structural proximity and close contact between the matrix and the analyte [32]. It could be hypothesized that ISD/PSD events could be due to a concentration effect of the compounds in the deposits with a close proximity of the compounds to one another. This effect might be reduced in biological matrices, when nucleosides are present at low concentrations and mixed with endogenous compounds. Preliminary results of electrospray ionization (ESI)-MS and the development of assays of cell culture models currently indicate that (i) fragmentation events also take place with ESI but can be attenuated with specific source-tuning, such as lower cone voltage, which seems to prove the predominance of ISD events in the analyte fragmentation and that (ii) exposure to TMZ can lead to concentrations of O6-mG in the pg/µg tissue range, which represent low concentrations for DI-MS-based analyses. These assumptions and observations may lead to additional recommendations for workflows before investigating O6-mG concentrations by DI-MS analyses: It is noteworthy that Kimura and co-workers observed ISD effects in MALDI on oligonucleotides with nucleobase derivatives, which may be attributed to the structural proximity and close contact between the matrix and the analyte [32]. It could be hypothesized that ISD/PSD events could be due to a concentration effect of the compounds in the deposits with a close proximity of the compounds to one another. This effect might be reduced in biological matrices, when nucleosides are present at low concentrations and mixed with endogenous compounds. Preliminary results of electrospray ionization (ESI)-MS and the development of assays of cell culture models currently indicate that i) fragmentation events also take place with ESI but can be attenuated with specific source-tuning, such as lower cone voltage, which seems to prove the predominance of ISD events in the analyte fragmentation and that ii) exposure to TMZ can lead to concentrations of O6-mG in the pg/µg tissue range, which represent low concentrations for DI-MS-based analyses. These assumptions and observations may lead to additional recommendations for workflows before investigating O6-mG concentrations by DI-MS analyses:

Conclusions
The analysis of the pharmacodynamic methylation effects of TMZ is a complex task, especially using DI-MS, which is hampered by several possible artefacts. Here, we provide important information to avoid pitfalls in DI-based analyses of O6-mG induced by DNA exposure to TMZ.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.