Molecular Framework Modification in Mass Spectrometry: Atom Exchange, Insertion, and Deletion

Abstract

The ionization and activation of molecules, as performed using mass spectrometry (MS), has long been known to induce unimolecular rearrangements in gas-phase ions, and in some cases, reorganization of the underlying skeletal framework. Here, we explore skeletal rearrangements occurring from the electrospray ionization process, including atom exchange, insertion, and deletion, within pre-existing ring systems. The Baeyer–Villiger oxidation, the Beckmann rearrangement, and the Favorskii rearrangement are longstanding organic reactions that are here explored in MS, alongside a unique sulfur-for-oxygen exchange reaction within a 2,4,6-triphenylpyrylium scaffold. MSⁿ and high-resolution MS are used to examine four representative transformations of molecular frameworks. Reactions across the four types emphasize the fact that ESI-MS can be intrinsically reactive. Product assignments in each case are supported by exact-mass measurements, isotopic distribution analysis, and MS/MS comparison with authentic standards or by characteristic fragmentation behavior. Beyond demonstrating framework modifications in electrospray droplets, these results point to an expanded suite of reactions for reactive MS derivatization, as well as the development of future strategies for late-stage functionalization of biomolecules.

Graphical Abstract

Introduction:

The process of recording mass spectra by classical electron impact ionization (EI) has long been known to induce molecules to undergo unexpected rearrangements.^1,2 In some cases, most notably in the work of Djerassi and colleagues,³ these modifications extend past simple hydrogen rearrangements to involve skeletal reorganization of the ionized molecules.^3–6 Beyond the classical unimolecular rearrangement processes studied by EI-MS, discharge ionization methods can also be accompanied by modifications of molecular frameworks, for example, through the insertion of nitrogen^7–9 or oxygen atoms¹⁰ into carbon-carbon bonds^7,9,10 due to the generation of reactive species (N₃⁺, NO⁺, O₂^+., O₃^+.) in the high-energy ionization environment.^7–10

These observations connect the ion chemistry in the mass spectrometer to the broader field of core scaffold modification reactions. The term ‘skeletal editing’ is relatively recent, but the underlying concept has long been significant to organic chemistry.^11–13 Rather than constructing a new scaffold de novo, these reactions can be used to alter the elemental composition, size, or connectivity of an existing core through the selective exchange, insertion, or deletion of individual atoms. In synthetic organic and medicinal chemistry, this strategy is attractive because it allows complex drug or natural-product scaffolds to be modified both generally and selectively, avoiding the need to redesign and resynthesize each analogue from earlier reaction steps.^13,14 As a result, skeletal editing provides a route to late-stage diversification and focused structure–activity relationship studies by generating bio-isosteres or shape-divergent analogues from an existing molecular framework.^15–18 The recent emergence of accelerated reactions from electrospray devices^19–23 and their application to late-stage functionalization of molecules in drug discovery^24–28 raises the question of whether skeletal transformations in this reaction format are possible.

In the present work, we examine the possibility of skeletal editing in droplets generated by electrospray ionization. All reactions are run without external catalysts or oxidants, preheating, and occur within the short time frame it takes to mix reagents and be analyzed, in contrast to the traditional bulk conditions.

The reactions examined herein span several modes of framework remodeling (Scheme 1) and were selected to demonstrate the range of transformations accessible. The Baeyer–Villiger oxidation,²⁹ the Beckmann rearrangement,³⁰ and the Favorskii rearrangement³¹ are longstanding organic reactions that are here explored with MS, alongside a sulfur-for-oxygen exchange reaction within a triphenylpyrylium scaffold, a reaction closely related to the pyridinium synthesis from pyrylium salts.³²

Scheme 1.

Overview of the skeletal editing reactions performed in this study.

The conversion of pyrylium to pyridinium using amines is a well-established chemical transformation.^33–35 This reaction is of immense analytical value and has been used in conjunction with mass spectrometry for quantitation of water content in organic liquids,³⁶ charge labeling of peptides,^37,38 and the enhanced detection of amino acid neurotransmitters in mouse brains.³⁹ This type of O-for-N swap reaction is accessible in a variety of formats, including online reactions in microdroplets generated using ESI.^40,41 The corresponding O-for-S transformation shown in Scheme 1 would generate the sulfur-containing thiapyrylium product when using a sulfide salt in the presence of water. If such a reaction is possible through droplets, this could enable rapid derivatization based online analysis of the bisulfide anion (HS⁻). At biological pH the bisulfide anion is the dominant species of aqueous hydrogen sulfide equilibria and an important biological signaling molecule of analytical interest.^42–44

More broadly, such transformations are of analytical interest because they could expand reactive MS workflows while also revealing an unexpected and fundamentally important capability for electrospray to mediate skeletal remodeling reactions. Knowledge of the possibility of skeletal rearrangement may be significant in structural studies in the omics areas, not least in work like that of Livia Eberlin that is being celebrated in this issue of the journal.^45–47

Materials and Methods

Materials:

Liquid chromatography-MS–grade water (Fisher Scientific) was utilized for all experiments containing water. Cyclohexanone, 2-chlorocyclohexanone, hydroxylamine, cyclohexanone oxime, caprolactam, caprolactone, formic acid, graphite foil (0.5 mm thickness, flexible graphite, 99.8%), and potassium sulfide were purchased from Sigma-Aldrich. 2,4,6-triphenylpyrylium tetrafluoroborate (98+%) and acetonitrile (ACN; ACS 99.5+%) were purchased from Thermo-Fisher Scientific. Acetonitrile-D₃ (D, 99.8%) was purchased from Cambridge Isotope Laboratories, Inc. High-purity compressed N₂ (≥ 99.95%) was procured from Indiana Oxygen. Reactions were run using concentrations of 10 mM unless otherwise denoted. All reagents and solvents were used as received.

Methods:

Nano-electrospray ionization (nESI) tips were made from borosilicate glass capillaries (Sutter Instrument Co.; with an outer diameter of 1.5 mm, and an inner diameter of 0.86 mm) pulled to an approximate 5 μm O.D. using a Flaming/Brown micropipette puller (P-97, Sutter Instrument Company). Solutions were ionized by applying a positive or negative 1.7 kV DC spray potential (supplied from the mass spectrometer) to an electrode placed inside the nESI capillary via a copper alligator clip, unless otherwise noted. The electrosonic spray deposition apparatus used was similar to that of previous reports.⁴⁸ Briefly, this apparatus was constructed using fused silica lines with 250-μm inner diameter. and 360-μm outer diameter (PolyMicro), one tee assembly, one union assembly, two NanoTight sleeves, and a silica capillary (IDEX Health and Science) at an applied potential of 4 kV. Solutions were sprayed at a graphite surface that was held at ground. To maintain the flow of reagent solution at a rate of 10 μL/min, syringe pumps (Standard Infusion PHD 22/2000, Harvard Apparatus) were utilized with gastight syringes (Hamilton Robotics). N₂ was used as the nebulizing gas at 120 psi, as read on the regulator.

All ion trap full scan (also referred to as MS¹, full scan, or single stage mass spectrometry) and ion trap collision-induced dissociation (IT-CID) tandem mass spectrometry (MS/MS) experiments were performed using an LTQ XL ion trap mass spectrometer (Thermo Scientific) with a helium bath gas and nESI as the ionization method. High-resolution mass spectrometry (HRMS) measurements on the products and standards were made using a Waters AQUITY Q-TOF MS with nESI as the ionization method. More information regarding instrument conditions and relevant parameters can be found in the Supplementary Information. A Bruker AV-III-400-HD with a 5mm BBFO Z-gradient SmartProbe was used for ¹HNMR data.

The effects of spray distance were explored for the synthesis of caprolactam. nESI emitters were placed 64 mm away and 10 mm away from the inlet to study the effect of distance, while standards were studied 10 mm away from the inlet (i.e., non-accelerating conditions²¹).

No additional reagents were used to promote the ionization of each compound apart from the formation of caprolactone. This reaction was run in a solvent of 99:1 acetonitrile/H₂O (% v/v) with 0.1% v/v formic acid. The formic acid was added here to give a clear reproducible signal, given that both species of interest have low ionization efficiency in positive mode and hence small aberrations in spray conditions, tip geometry, etc. can skew observed peak intensities.

Results and Discussion:

Atom Exchange:

The bulk reaction for the simplest thiapyrylium (also named thiopyrylium) is known to occur when pyrylium salts are treated with sodium sulfide in acetone, followed by reaction with acid, a replacement of the heterocyclic oxygen atom with sulfur yields the thiapyrylium salts.^49,50 The speculated interfacial acidity of microdroplets generated from electrospray ionization (using positive mode ESI)^20,51 suggests that acid-catalyzed pyrylium conversion to thiapyrylium may occur spontaneously (Figure 1A). This reaction was performed by spraying from a mixture of 2,4,6-triphenylpyrylium tetrafluoroborate and potassium sulfide in acetonitrile with 1% v/v water, onto a grounded graphite surface (Figure 1B). The reaction mixture turned blue upon mixing, indicating the formation of a highly conjugated open-chain sulfur-containing intermediate (Figure 1C, Figure S1).^49,50 Figure 1A summarizes the sulfur-for-oxygen atom-exchange reaction of 2,4,6-triphenylpyrylium under external acid-free electrospray ionization conditions. Shown in Figure 1B, the reaction mixture was introduced by electrospray deposition onto a grounded graphite surface, after which the deposited material was recovered and examined by nESI-MS using non-accelerating conditions. The resulting full scan mass spectrum (Figure 1C) contains the starting pyrylium ion signal at m/z 309 and the putative product at m/z 325, the latter matching the expected nominal mass of the sulfur-exchanged thiapyrylium species. High-resolution mass measurement of the nominal m/z 325 ion, experimentally observed at m/z 325.1046, agrees with the calculated theoretical value of m/z 325.1045 within a 5-ppm mass tolerance for an ion with formula C₂₃H₁₇S⁺. A minor ion observed at m/z 365 is tentatively assigned as the sodium-adducted intermediate consistent with the proposed pathway.

Figure 1.

A) Atom exchange reaction of sulfur-for-oxygen occurs in 2,4,6-triphenylpyrylium upon addition of potassium sulfide. B) Electrosonic spray deposition set-up with solutions containing the reaction mixture being spray-deposited onto a grounded graphite surface. The surface of the graphite was washed with acetonitrile which provided the single-stage mass spectrum shown in C) Ions are seen corresponding to the starting material, as well as the protonated thiapyrylium product and a putative sodium-bound intermediate. Separate exact mass measurements of these ions matched the expected molecular formulas with less than 5 ppm error. D) MS/MS spectrum of precursor ion m/z 325 showing characteristic fragmentation by loss of H₂S, benzene (C₆H₆), and loss of both these neutral molecules.

Collision-induced dissociation of 2,4,6-triphenylpyrylium reagent (Figure S2) gave characteristic product ions at m/z 294, 291, 231, 203, and 105, with m/z 231 being observed as the most abundant ion. The ions at m/z 294 correspond to a loss of •CH₃, m/z 291 to loss of H₂O, m/z 231 corresponds to loss of C₆H₆, and m/z 203 corresponds to loss of benzaldehyde (C₇H₆O), while m/z 105 is assigned to the formation of the benzoyl cation from a total neutral loss of C₁₆H₁₂. These fragment ions match previously described fragmentation channels.⁵²

Under similar IT-CID conditions, the putative 2,4,6-triphenylthiapyrylium ion (Figure 1D) gave product ions at m/z 310, 292, 291, 247, 215, 203, 191, and 121, with m/z 247 giving the base peak. The ions correspond to losses, respectively, of H₃C•, HS•, H₂S, C₆H₆, C₈H₁₄, thiobenzaldehyde (C₇H₆S), C₈H₆S, and C₁₆H₁₂. Tandem MS of the m/z 325 ion supports product assignment, as does the exact mass determination. Comparison of the MS/MS spectra of the two compounds shows several distinct and parallel fragmentation channels, including heteroatom loss as H₂O for the pyrylium ion and H₂S/HS• for the thiapyrylium ion, C₆H₆ loss as the most abundant pathway in both cases, as well as benzaldehyde (C₇H₆O) or the parallel thioaldehyde loss (C₇H₆S), and C₁₆H₁₂ loss with retention of the cationic heteroatom-containing core. Given the limited precedent for this reaction and the absence of a commercially available standard for comparison, formation of the thiapyrylium product was further corroborated by product collection and subsequent ¹H NMR (Figure S3).

The proposed intermediate implicated in the literature as leading to the ring-closing condensation associated with product formation^49,50,53 observed here as the sodium-bound intermediate (m/z 365) was investigated with IT-CID (Figure S4). This species yielded substantial fragmentation at a lower normalized collision energy than both the starting material and the substitution product and yielded product ions at m/z 347, 337, 332, and 231 corresponding to neutral losses of H₂O, C₂H₄, HS•, and C₈H₆S, respectively. The presence of these fragment ions and neutral losses indicates that this species contains both a sulfur and oxygen heteroatom and has fragmentation channels that overlap with both the starting material and the product.

Despite the sulfur reagent introduced being a potassium salt a sodium is shown to adduct to the intermediate. It is worth noting that sodium is an abundant and common ambient ionization contaminant.⁵⁴ It is well known that sodium ions can often have a higher affinity to analytes than potassium ions; thus, the main adducted intermediate interrogated here is the sodium-bound intermediate despite the addition of potassium as a starting material.^55,56 However, as expected, under different instrument settings, both the potassium-bound and sodium-bound adducts can be observed (Figure S5).

Oxygen Insertion:

In the traditional Baeyer–Villiger reaction, oxidation of a ketone with a peroxyacid reagent effects formal oxygen insertion adjacent to the carbonyl, converting acyclic ketones to esters and cyclic ketones to ring-expanded lactones.²⁹ Prior work has indicated that Baeyer–Villiger oxidation may occur spontaneously in microdroplets during negative mode ionization, giving oxidation products from ketones without externally added peroxide.⁵⁷ However, a recent theory advanced by the authors suggests that applying a positive potential results in the eventual generation of a reactive oxygen species.⁵⁸

Shown in Figure 2A is oxygen insertion into a carbon framework during positive ionization, converting cyclohexanone into ε-caprolactone through a ring-expansion consistent with Baeyer–Villiger-type reactivity. Although no external H₂O₂ was added, the observed product profile is consistent with prior studies showing that oxidizing species, including H₂O_2, may be generated under reactive conditions^58–60 like those shown in Figure 2B. However, it is worth noting that the mechanism of formation of H₂O₂ remains uncertain with reports in the literature stating the production of reactive species can arise under various conditions, including the electrochemical reactions of dissolved oxygen.⁵⁹ Moreover, it remains unclear if the oxidizing species generated in the positive mode are identical to those generated in the negative mode. The corresponding single-stage mass spectrum in Figure 2C shows the starting ketone as protonated cyclohexanone, at m/z 99, and a lower-abundance peak at m/z 115 assigned to protonated ε-caprolactone. High-resolution mass measurement of the nominal m/z 115 ion, experimentally observed at m/z 115.0749, agrees with the calculated theoretical value of m/z 115.0753 within 5 ppm for an ion with formula C₆H₁₁O₂⁺. Tandem MS of precursor ion m/z 115 (Figure 2D) provides additional support for the structural assignment, producing a dominant fragment at m/z 97 via a neutral loss of H₂O and additional product ions at m/z 87, 73, and 69 arising from neutral losses of CO, C₃H₆, and C₂H₆O, respectively. Agreement of this MS/MS spectrum with that of an authentic ε-caprolactone standard (Figure S6) supports the conclusion that oxygen insertion occurs during the analysis process to furnish the ring-expanded lactone product. Formation of caprolactone during the charged spray process was further corroborated by product collection made using a deposition set-up analogous to that described in Fig. 2B, followed by ¹H NMR characterization (Figure S7).

Figure 2.

A) Oxygen insertion reaction of Baeyer-Villiger type to synthesize caprolactone. H₂O₂ is believed to be generated in-situ during the ionization process, none was added. B) Online nano-electrospray produces reactive droplets. C) Single-stage mass spectrum showing protonated reagent ion (m/z 99) as well as the formation of protonated ε-caprolactone (m/z 115) shown by separate exact mass measurements to match the expected mass with an error of less than 5 ppm. D) MS/MS spectrum of the protonated product (m/z 115) matches the fragmentation of the authentic product (shown in Figure S6) with neutral losses of H₂O, CO, C₃H_6, and C₂H₆O.

Nitrogen Insertion:

Figure 3A expands the observed reaction scope to include nitrogen insertion by another nESI reaction, showing that cyclohexanone can be converted to ε-caprolactam through initial oxime formation followed by a Beckmann rearrangement. In this case, the reaction proceeds by initial condensation of cyclohexanone with hydroxylamine to give cyclohexanone oxime, after which activation with acetic anhydride promotes rearrangement to the ring-expanded lactam. Note that condensation reactions are commonly reported in microdroplet literature generated from positively charged sprays.^19,58,61 Given that the oxime intermediate and ε-caprolactam share the same molecular formula, the positive-mode MS¹ spectrum in Figure 3B is most appropriately interpreted as evidence for the formation of an ion at m/z 114 consistent with either isomer. High-resolution mass measurement of the m/z 114 ion, experimentally observed at m/z 114.0914, agrees with the calculated theoretical value of m/z 114.09134 for an ion with formula C₁₆H₁₂NO⁺. Thus, exact mass measurements support the elemental composition, but they do not distinguish between the oxime and lactam structures.

Figure 3.

A) Nitrogen insertion reaction performed through initial condensation of cyclohexanone with hydroxylamine, followed by a Beckmann rearrangement to yield caprolactam. B) Using a nESI set-up identical to the one shown for the previous insertion reaction the single-stage mass spectrum showing almost exclusively the ions (m/z 114) associated with the caprolactam product, with exact mass measurements confirming the match with less than 5 ppm mass error. C) MS/MS product ion spectrum showing formation of the cyclohexanone oxime intermediate before the addition of acetic anhydride at a sprayer distance of 64 mm from the inlet. D) MS/MS spectrum illustrating the near-complete conversion of the cyclohexanone oxime to the caprolactam with the one pot reaction including acetic anhydride at a sprayer distance of 64 mm from the inlet.

Structural assignment instead comes from the tandem mass spectra in Figures 3C and 3D. Without the inclusion of acetic anhydride (Figure 3C), IT-CID of the m/z 114 ion produced a fragmentation pattern consistent with cyclohexanone oxime indicating that. The spectrum showed product ions at m/z 97, 96, 94, 86, 81, 79, and 69, corresponding to neutral losses of NH₃, H₂O, H₂O + H₂, CO, NH₂OH, H₂O + NH₃, and CH₃NO, respectively, from the m/z 114 precursor. Although IT-CID-induced isomerization has been reported for other compounds, it does not appear to induce isomerization to the caprolactam, as the fragmentation profile matches that of the commercial cyclohexanone oxime standard (Figure S8).

With the inclusion of acetic anhydride (Figure 3D), the product ion spectrum from precursor m/z 114 changes substantially and is assigned to ε-caprolactam, with dominant fragmentation to m/z 96 by loss of H₂O and comparatively higher relative abundance of product ions at m/z 79, 69, 97, and 86 arising from neutral losses of H₂O + NH₃, CH₃NO, NH₃, and CO, respectively. Unexpectedly, the synthesized caprolactam fragmentation yields the neutral loss of both heteroatoms as ammonia and water to give the second most abundant product ion at m/z 79. The formation of this product ion would most likely have to be formed through the formation of a skeletal rearrangement upon fragmentation. The product ion at m/z 79 is believed to be protonated benzene, and indeed, upon isolation and fragmentation of m/z 79 (MS³) both the standard and product show the characteristic fragment, the phenyl cation^62,63 at m/z 77, formed by neutral loss of H₂ (Figure S9). The resulting spectra from both MS² and MS³ are nearly identical to that of the authentic caprolactam standard (Figure S10). Between the obtained product and the standard MS² spectrum, however, there are small residual peaks, most noticeably one at m/z 81. The presence of this peak in the tandem MS of the product likely reflects a conversion of less than 100%, with the m/z 81 fragment ion being due to the unreacted oxime intermediate.

The extent of oxime-to-lactam conversion is expected to depend on the distance between the ionization source and the mass spectrometer inlet because this distance changes the lifetime of the sprayed droplets before sampling. At shorter distances, droplets are sampled more rapidly, limiting evaporation, desolvation and reaction time, and therefore preserving a larger population of the oxime intermediate. At longer distances, the droplets undergo greater solvent loss and have a longer reaction residence time, increasing the effective concentration of reagents within the droplet and favoring further conversion to caprolactam. Consistent with this interpretation, placing the ionization source close to the inlet (10 mm) gives a spectrum with higher relative intensities of ions associated with the cyclohexanone oxime, such as m/z 81 and m/z 94, while other ions remain at relative intensities similar to those observed for the caprolactam standard (Figure S11). In contrast, placing the nESI emitter farther from the inlet (64 mm) yields the tandem spectrum shown in Figure 3D, with noticeably lower m/z 81 and 94 abundances, indicating greater conversion to the lactam product.

Ring Contraction via Atom Deletion

Figure 4A completes the series of reactions by illustrating ring contraction through atom deletion from 2-chlorocyclohexanone to give the cyclopentane carboxylic acid under nESI aqueous conditions. In this case, the reaction is consistent with a Favorskii rearrangement, in which the six-membered α-chloroketone framework is converted into a five-membered scaffold with an exo-cyclic carboxylic moiety; no basic catalyst is added, unlike the traditional bulk reaction conditions.³¹ The formation of this product would then likely be driven by the enhanced autoionization of water in the droplet.^20,58 The full-scan positive-mode spectrum in Figure 4B shows the starting material as protonated 2-chlorocyclohexanone at m/z 133 together with its expected chlorine isotope peak at m/z 135, while another feature at m/z 115 is assigned to protonated product, cyclopentane carboxylic acid. Exact mass measurement of the m/z 115 ion, experimentally observed as m/z 115.0750, agrees with the theoretical value of m/z 115.0754 for an ion with formula C₆H₁₁O₂⁺. Further structural support was obtained from MS/MS spectra recorded in both positive and negative polarities. In positive mode (Figure 4C), IT-CID of the m/z 115 precursor ion gives product ions at m/z 97, 79, and 69, corresponding to neutral losses assigned in the figure as H₂O, two consecutive H₂O losses, and H₂O + CO, respectively, from protonated cyclopentane carboxylic acid. In the negative mode (Figure 4D), IT-CID of the m/z 113 ion, assigned as deprotonated cyclopentane carboxylic acid, yields fragments at m/z 85, 83, and 69, corresponding to neutral losses of C₂H₄, C₂H₆, and C₃H₈, respectively. Both positive and negative mode fragmentation patterns are consistent with those of the authentic compound (Figure S12–S13).

Figure 4.

A) Cyclopentane carboxylic acid is produced through nESI spray of 2-chlorocyclohexanone solution with equal parts acetonitrile and water. B) Using a nESI set-up identical to previous iterations positive mode mass spectrum shows the protonated reagent ion (m/z 133) and chlorine-37 isotope (m/z 135) as well as the product, the protonated cyclopentane carboxylic acid (m/z 115). C) MS/MS of m/z 115 in positive mode matches the spectrum of the authentic compound (Figure S11) with neutral losses of H₂O and CO. D) MS/MS fragmentation of m/z 113 in negative mode, which matches that of the authentic compound (Figure S12), showing neutral losses of C₂H₄, C₂H₆, and C₃H₈.

Conclusion:

It is noteworthy that mass spectrometry serves not only to reveal, classify, and compare changes in core molecular architecture, but that the process of ionization itself can induce these changes in nESI-generated sprays; here, bimolecular reactions are responsible, whereas in the case of EI, unimolecular reactions are involved. These examples show that classic modes of skeletal framework remodeling can be produced in the nESI-MS and this may inform future biomedical workflows. The resulting data is analytically informative with unexpected fragmentation and also of fundamental interest in that previously studied skeletal rearrangements in EI mass spectrometry often have relatively low activation energies, which makes them competitive with simple cleavages.² The energy deposited in nESI is low enough that fragmentation is not typically observed.⁶⁴ Despite this, unsuspected rearrangements may still occur, and their occurrence will be reflected in the MS/MS spectra of the products, even in cases where only isomerization has occurred.

This work provides significant motivation to explore reactive derivatization, as well as droplet-based late-stage functionalization strategies based on the reactions shown. Analogies to known solution-phase chemistry deserve further detailed exploration to establish the degree of parallelism between the techniques reported here and bulk organic chemistry with respect to these transformations.

Supplementary Material

The Supporting Information is available free of charge at XXXX

• ¹HNMR, Tandem MS of standards (PDF)