Flavinium Catalysed Photooxidation: Detection and Characterization of Elusive Peroxyflavinium Intermediates

Abstract Flavin‐based catalysts are photoactive in the visible range which makes them useful in biology and chemistry. Herein, we present electrospray‐ionization mass‐spectrometry detection of short‐lived intermediates in photooxidation of toluene catalysed by flavinium ions (Fl+). Previous studies have shown that photoexcited flavins react with aromates by proton‐coupled electron transfer (PCET) on the microsecond time scale. For Fl+, PCET leads to FlH.+ with the H‐atom bound to the N5 position. We show that the reaction continues by coupling between FlH.+ and hydroperoxy or benzylperoxy radicals at the C4a position of FlH.+. These results demonstrate that the N5‐blocking effect reported for alkylated flavins is also active after PCET in these photocatalytic reactions. Structures of all intermediates were fully characterised by isotopic labelling and by photodissociation spectroscopy. These tools provide a new way to study reaction intermediates in the sub‐second time range.


Synthesis of flavinium compounds
Flavinium salt 1a, precursor S1 and analogue of 1a with methyl in N3 position instead of octyl (1d) were synthetized by the previously published procedures. [1] Synthesis of the charge-tag derivatives 1b and 1c is outlined in figure S1. N Figure S1. Synthesis of the charge-tagged flavinium derivates 1b and 1c.
NOTE: Attempts to precipitate the compound from Acetonitrile / Et2O have led to a crystalline mass which is 1:1 mixture of an unknown impurity (m/z: M-2Br + OH) and the desired compound. No procedure to isolate and precipitate the desired compound as pure crystalline mass has been found. During MS experiments the desired ion was isolated by mass selection. From the 19F-NMR it can be established that the impurity and the target compound are in 1:3 ratio in the case of the NMR sample.

Instrumentation
If not stated otherwise, reported MS and MS 2 spectra were measured on Finnigan TSQ-7000 machine with an ESI ion source. For sample irradiation, 1W laser diode (445nm, blue, e-bay) with Thorlabs ballast set at 1A current was used. In all cases fused silica capillary (internal diameter 100 µm, outer diameter 190 µm, polyimide coating PostNova, part no. Z-FSS-100190 was used).

2.1.2
Irradiation layout In our experiments, we have irradiated the sample at three different positions: at a syringe/vial (1), at a capillary (2) and at the tip of a capillary (3). (1) For the syringe irradiation, we placed Hamilton syringe to a syringe pump and we focused a laser beam through the end of a glass part of the syringe as illustrated on Figure S2a. We placed alumina foil in opposite to the laser beam to reflect the laser beam back through the solution. (2) For the capillary irradiation ( Figure S2b) we removed coating of the fused capillary by published procedure. [ 2] First we burnt the capillary coating quickly by a lighter. Then we removed the carbon deposit by a tissue wetted with acetonitrile. We removed the coating approximately 10-15 cm away from the capillary tip (uncoated capillary often broke and replaced by a new one where the window was at slightly different position).
(3) When we irradiated the capillary tip ( Figure S2c, S2d) we removed coating in same way as previously described. We were forced to bypass the microswitch of the ESI head so we could leave it opened in the course of spraying. Although this system resembles the previously published nano-ESI setup, [ 3] we applied much higher flow-rate (0.20-0.40 ml/h compared to nano-ESI flow) to improve our chance to detect reactive intermediates. Figure S2. Irradiation at the syringe (a), the capillary (b) and the tip of a capillary (c), (d)

2.1.3.1
Syringe ionization When noted that syringe ionization has been applied, the current for ESI-MS has been delivered directly to the sample by custom-made syringe piston for Hamilton syringe.
Piston description: Stainless steel screw has been used as an electrode. The sealing and cap comes from original Hamilton piston. The custom piston has been made from 3D printed PET. We have quickly tested the piston and did not observe any ions coming from this syringe when sprayed under conditions of experiments mentioned below. This piston has been connected to ESI-MS source through 20 MΩ resistor. The .stl 3D model ready for printing (optimized for 0.20 mm layer-height) is downloadable on our group website, or accessible through the corresponding author.

"
Overpressure unit" When noted that "overpressure unit" ionization technique has been applied, [ 4] high voltage was applied by stainless steel wire to the diluted mixture in 4ml vial. The wire, fused silica capillary connected to MS and needle connected to tube for pressurizing the vial were induced through the septum in the vial cap. Pressure has been regulated manually based on the sample viscosity and sample flow to optimize the signal intensity.

2.1.4
Preparation of solution for the ESI ionization and ionization conditions Syringe was covered by alumina foil if irradiation was applied to the capillary or capillary tip. In some capillary-tip irradiation experiments syringe was replaced by overpressure unit described above. Acetonitrile was used as an exclusive solvent in the whole study as it is the solvent used in the original work. [1] If not stated otherwise, solvent was not degassed and was shaken with air in vial or syringe to ensure that substantial amount of oxygen is dissolved in the analysed mixture.

2.2.1
MS spectra acquired without irradiation Except the total ion current change, no other change of spectrum was observed if small amount of water, hydrogen peroxide, toluene, 1,3-dimethoxybenzene, chlorobenzyl-alcohol or 4-chlorobenzyaldehyde was added to the solution of flavinium salt 1a. Spectrum of pure flavinium salt is shown in Figure S5.
Sample: 0.7 mg of 1a was dissolved in 4 ml of acetonitrile. 100 µl of this concentrated solution was then diluted by 1 ml of acetonitrile, transferred into the syringe and sprayed by syringe ionization technique. Conditions: flow rate 0.20 ml/h, spray voltage 4.5 kV, capillary temperature 150°C, Capillary voltage 20 V, Tube lens voltage 100 V. Figure S5. Spectrum of non-irradiated flavinium salt 1a.
Conversely, in the case of charge-tagged derivatives 1b and 1c, in some cases pseudobase formed through a reaction with an additive is detected. The most important pseudobase is the one formed after hydrogen peroxide addition to the solution of flavinium salt as seen in the Figure S6 because it represents the charge-tagged hydroperoxyflavin 2.
Samples and conditions used in Figure S6: (a) Sample: 0.5 mg of 1b was dissolved in 2 ml of acetonitrile. 100 µl of the concentrated solution was then diluted by 1 ml of acetonitrile, transferred into the syringe and sprayed by syringe ionization technique. Conditions: flow rate 0.20 ml/h, spray voltage 5 kV, emult 1100 V, sheath gas flow ~20, Capillary voltage 30 V, Tube lens voltage 70 V.
(b) Sample: sample used in (a) after addition of 100 µl of 4-chlorobenzyl alcohol solution (1.4 mg dissolved in 4 ml of acetonitrile (the spectrum still looked like in the case (a)) and 50 µl of concentrated (30%) H2O2 ([1b+H,2O] + formed). It was sprayed using the same spraying conditions as (a), except emult 1200 V.
(c) Sample: 2.6 mg of 1c was partially dissolved in 0.6 ml of acetonitrile. 50 µl of the concentrated solution was then diluted by 2 ml of acetonitrile, transferred into the "overpressure unit" and sprayed. Conditions: spray voltage 4.5 kV, emult 1200 V, sheath gas flow ~25, capillary temperature 150°C, Capillary voltage 0 V, Tube lens voltage 60 V. The ion of m/z corresponding to the the precursor S3b described in Figure 1 is present as an impurity in the 1c, from which we prepared the sample (confirmed by NMR). Figure S6. a) Spectrum of non-irradiated charge-tagged flavinium 1b and b) the spectrum after addition of H2O2 to the sample used in a). c) Spectrum of the trimethylammonium charge tag 1c + H2O2.

2.2.2
Experimental details of MS spectra acquired with sample irradiation under oxygen-free conditions We have prepared an air-free sample by Schlenk-line technique and irradiated it by a home-made diode light-source (Luxeon STAR/0, 4 x 1030 mW@700 mA, 448 nm). In all oxygen-free conditions experiments, Schlenk tubes were dried by a heat-gun under vacuum and then cooled under Argon atmosphere to room temperature prior to use.
(a) Sample: 0.6 mg of 1a and 8 µl of toluene-d8 were dissolved in 1 ml of acetonitrile (dry). This solution was then degassed by 5 freeze-thaw cycles. Then the solution was irradiated for 5 minutes under Argon atmosphere. After the reaction the sample was diluted by 10 ml of acetonitrile (dry). To ensure minimal contact with external atmosphere, MS capillary, wire for ionization and N2 source for the overpressure injection of the sample were introduced through the septum. Conditions: spray voltage 4.5 kV, capillary temperature 275°C, Capillary voltage 0 V, Tube lens voltage 100 V, emult 1400 V, sheath gas 20, auxiliary gas 35.
(b) Sample: Sample from a) + 10 µl D2O. Conditions as in (a) (c) Sample: 0.9 mg of 1a and 8 µl of toluene-d5 were dissolved in 2 ml of acetonitrile (dry). This solution was then degassed by 5 freeze-thaw cycles. Then the solution was irradiated for 5 minutes under Argon atmosphere. After the reaction the sample was diluted by 10 ml of acetonitrile (dry). To ensure minimal contact with external atmosphere, MS capillary, wire for ionization and N2 source for the overpressure injection of the sample were introduced through the septum. Conditions: spray voltage 4.5 kV, capillary temperature 275°C, Capillary voltage 0 V, Tube lens voltage 80 V, emult 1400 V, sheath gas 8, auxiliary gas 0.
(d) Sample: Sample from c) + 10 µl D2O in 100 µl of acetonitrile (dry). Conditions as in (c) (e) Sample: 0.8 mg of 1a, 8 µl of toluene-d8 and 8 µl of toluene-d0 were dissolved in 1 ml of acetonitrile (dry). This solution was then degassed by 5 freeze-thaw cycles. Then the solution was irradiated for 5 minutes under Argon atmosphere. After the reaction the 100 µl of the reaction mixture was diluted by 2 ml of acetonitrile (dry). MS experiment has been carried out in overpressure unit. Conditions: spray voltage 4.5 kV, capillary temperature 275°C, Capillary voltage 0 V, Tube lens voltage 100 V, emult 1200 V, sheath gas 21, auxiliary gas 35 Figure S8. a) Spectrum of 1a + and toluene-d8 irradiated under anaerobic conditions b) (a) after addition of D2O c) spectrum of 1a + and toluene-d5 irradiated under anaerobic conditions d) (b) after addition of D2O e) spectrum of 1a + , toluene-d8 and toluene-d0 irradiated under anaerobic conditions.

MS spectra acquired with syringe irradiation method
When we irradiated the syringe, we have observed mainly formal adduct with oxygen. We observed this adduct in case that solution of sole flavinium salt or flavinium salt and water was irradiated (b). One hydrogen atom can be exchange for deuterium in this adduct. This exchange has been proved by addition of deuterium-labelled water (c). In case that 4-chlorobenzyl alcohol is added instead of water formal adduct of flavinium salt and water is observed (d).
In case of spectra (a) and (d) decrease of the total ion current is observed during irradiation. It can be explained by formation of water which then forms neutral flavin pseudobase. This effect is not observable in cases (b) and (c) as water is present from the beginning and total ion current remains relatively low, but constant over the time of the irradiation.
All spectra shown in Figure S9 were measured on the Finnigan LCQ Deca XP mass spectrometer. Blue laser diode current was set to 0.4 A. All spectra in this figure represent an average of the acquisition between 7-20 minutes after the start of the irradiation.
Samples: a) Sample: 0.8 mg of 1a was dissolved in 4 ml of acetonitrile. 100 µl of the concentrated solution was then diluted by 1 ml of acetonitrile, transferred into the syringe and sprayed by syringe ionization technique.

S14
As mentioned in the article, formal adduct with oxygen ([1a+O] + ) and water ([1a+H2O] + ) are present even after the irradiation is stopped as shown in Figure S10.
Sample and conditions used in Figure S10: Sample: 0.5 mg of 1a was dissolved in 4 ml of acetonitrile. 100 µl of the concentrated solution was then diluted by 1 ml of acetonitrile and 100 µl of 4-chlorobenzyl alcohol solution (1.4 mg dissolved in 4 ml of acetonitrile) was added.
Conditions: flow rate 0.40 ml/h, spray voltage 5 kV, capillary temperature 80°C, Capillary voltage 30 V, Tube lens voltage 110 V, emult was changed from 1200 to 1350 V during acquisition due to decreasing signal intensity. For the latter spectra when irradiation was stopped flow rate was decreased to 0.18 ml/h Figure S10. Spectrum (a) is an average of the acquisition between 3-6 minutes after the start of the irradiation. (b) Spectrum of the same sample after prolonged irradiation time and turning the light source off (emult was changed to 1800-2000 V). The impurities on the baseline are caused by relatively low intensity of the target ions and are most probably product of sidereactions during a photocatalysis. We speculate that [1a-2H] + is most probably a partially decomposed catalyst 1.

2.2.4
MS spectra acquired with capillary irradiation method We have tried a slight variation of the recently published capillary irradiation technique 2 where capillary is irradiated in front of the ESI head as illustrated in Figure S2b. However, it was not effective for our system. Figure  Figure S11. Spectrum acquired with irradiation applied to a capillary just before ESI-head entrance. Ion m/z 105 has fragmentation which fits to [Na+2CH3CN] + cluster.

2.2.5
MS spectra acquired with capillary tip irradiation method To observe reaction intermediates we developed our own system which was inspired by recently published nano-esi tip irradiation technique. [3] For our system liquid-ionization is crucial because when capillary is pulled out of the source as in our case (Fig S2c and d) classical tip irradiation is ineffective. By this system we have been finally able to see elusive reaction intermediates as illustrated in Figure S12.
During tip-irradiation we observed rapid drop of total ion current in virtually all cases. Subsequently relative intensity of trace impurities raise. To rule out possibility of misinterpretation of such impurity for emerging intermediate we have employed technique where for each spectrum of acquisition two scans were acquired. First scan was irradiated and the second was not. Then we have normalized our spectra to the total ion current of the average of non-irradiated averaged spectra and overlaid averaged irradiated spectrum with the non-irradiated one (pink dash-dotted lines in Figure S12). Thus minimize a chance of misinterpretation caused by total ion current shift induced by irradiation.
Samples and conditions used in Figure   S12. Spectrum acquired with irradiation applied to the tip of a capillary a) and spectra after addition either of water b), deuterated water c), hydrogen peroxide d), 4-chlorobenzyl alcohol e), toluene f), or dimethoxybenzene g) to the sole flavinium salt 1a solution.

S17
We have employed the same technique to the charge-tagged analogues as illustrated in Figure S13.
Samples and conditions used in Figure

2.2.6
Time evolution of the spectra To estimate the timescale on which we are operating we have sequentially irradiated two distinguished points on the capillary tip. The distance between points was approximately 2-3 mm, which corresponds to time difference of ~300-450 milliseconds under the flow which we have used. The irradiation points are described in Figure S15, the spectra in Figure S16.

2.2.7
Fragmentation of the selected ions To further describe ions observed by us we performed their fragmentation by collisions with xenon gas (CID = collision induced dissociation). In all cases, the structures of parent and daughter ions are merely hypothetical and should be regarded as possible interpretation of the m/z. We have started by identification of oxo-intermediates present after irradiation of the original mixture or addition of water ( Figure S17). If not stated otherwise all observed parent ions were generated by capillarytip irradiation. It is apparent that original flavinium 1a + and its adduct [1a+O] + both undergoes McLafferty rearrangement with neutral loss of octene with similar intensity. This leads us to hypothesis that oxidation occurred on the aromatic moiety or at the ethylene bridge. Conversely the adduct [1a+O2] + has completely different and complex fragmentation pattern which has not been assigned.

S22
We have also performed CID of formal water and hydrogen peroxide adducts summed in Figure S19 and fragmentation of adducts with aromatics summed in Figure S20. Figure

2.2.8
Fragmentation in the source When we hardened the source conditions we were able to generate fragments of OOH • (m=33) radical loss directly in the source which was used for IRPD experiments. This is illustrated on [1b+OOH] + in-source fragmentation to 1b •+ in figure  S21 Samples and conditions used in Figure

2.2.9
Reactivity of the selected ions

Reactivity of 1a+H2O2
We have measured the gas-phase reactivity of 1a+H2O2. We have observed a reaction with dimethyl sulphide. To document the reactivity, we have recorded multiple spectra with varying pressure and collision energy. For comparison we also recorded spectra using unreactive xenon gas instead of dimethyl sulphide. Results of these experiments ( Figure S24) indicate that the OOH (m=33) neutral loss is a fragmentation channel and that O (m=16) neutral loss is a reactivity channel.    Figure S23. Scheme of the main reactivity and fragmentation channels of the 1a+H2O2 ion. Figure S24. Relative intensity of fragmentation channels based on the pressure and collision offset variation. The relative intensity has been obtained as intensity of fragmentation channel divided by total ion current. Noise has been subtracted from both values prior to the calculation. a) Collision offset dependence of OOH (m=33) neutral loss; b) Collision offset dependence of O (m=16) neutral loss; c) pressure dependence of OOH (m=33) neutral loss; d) pressure dependence of O (m=16) neutral loss. Xenon (blue dash-dotted lines) or dimethyl sulphide (full red lines) has been used as a collision gas.

2.2.9.2
Reactivity of 1b+H2O2 We have also recorded gas-phase reactivity of charge-tagged analogue 1b+H2O2 with dimethyl sulphide in analogy to abovementioned experiments with 1a+H2O2. The fragmentation/reactivity pattern is the same. Only difference is that in the charge-tagged species we observed visible a single-charged fragments. These fragments are most probably experimental artefacts as they are present even in collisions with unreactive xenon. We hypothesize that these artefacts are caused by an unknown persistent impurity present in the collision cell. In analogy to 1a+H2O2, the results of the experiments with 1b+H2O2 ( Figure S27) indicate that the OOH (m=16.5) neutral loss is a fragmentation channel and that O (m=8) neutral loss is a reactivity channel.  . Relative intensity of fragmentation channels based on the pressure and collision offset variation. The relative intensity has been obtained as intensity of fragmentation channel divided by total ion current. Noise has been subtracted from both values prior to the calculation. a) Collision offset dependence of OOH (m=33) neutral loss; b) Collision offset dependence of O (m=16) neutral loss; c) pressure dependence of OOH (m=33) neutral loss; d) pressure dependence of O (m=16) neutral loss. Xenon (blue dash-dotted lines) or dimethyl sulphide (full red lines) has been used as a collision gas. Figure S28. Relative intensity of artefact fragmentation channels based on the pressure and collision offset variation. The relative intensity has been obtained as intensity of fragmentation channel divided by total ion current. Noise has been subtracted from both values prior to the calculation. a) Collision offset dependence of H2O2 + (m=34) loss; b) Collision offset dependence of H + (m=1) loss; c) pressure dependence of H2O2 + (m=34) loss; d) pressure dependence of H + (m=1) loss. Xenon (blue dash-dotted lines) or dimethyl sulphide (full red lines) has been used as a collision gas. Figure S29. Relative intensity of artefact fragmentation channels based on the pressure and collision offset variation. The relative intensity has been obtained as intensity of fragmentation channel divided by total ion current. Noise has been subtracted from both values prior to the calculation. a) Collision offset dependence of OH + (m=17) loss; c) Collision offset dependence of OH + (m=17) loss. Xenon (blue dash-dotted lines) or dimethyl sulphide (full red lines) has been used as a collision gas.

2.2.9.3
Reactivity of 1b+OOH We did not observed any reactivity with dimethyl sulphide in the case of 1b+OOH ion ( Figure S30), which is in a strict contrast to abovementioned cases. However, both the fragmentation (neutral loss of OOH) and the IRPD spectra of this ion suggest that this ion is truly a flavin-hydroperoxide. We speculate, that the absence of the reactivity is caused by a steric hindrance of the active site by bulky imidazolium group, or by the fact, that nucleophilic dimethyl sulphide is attracted by the charge tag, thus cannot react with the hydroperoxy moiety.
Sample and conditions used in Figure S30:

Determination of the KIE from the MS experiments
The KIE was determined from repeated experiments illustrated in Figure 2c,d of the main article and Figure S31. First, we have evaluated the toluene d0/d5 experiments ( Figure S31a) In cases of pattern overlap, intensities of deuterated channels have been renormalized.

Experimental Details
Ion spectroscopy (IRPD) technique has been described elsewhere in detail (see citations in the main article). For the generation of ions, the capillary tip irradiation technique described in the mass spectrometry section was used. Samples were prepared in a way as described in previous mass spectrometry sections. Ionization conditions have been almost the same to the one described in mass spectrometry section. Special attention has been paid to keeping the voltage of transfer quadrupole of TSQ-7000 part of our IRPD machine at higher voltage than the first lens to ensure that the ions do not pre-trap in transfer quadrupole as otherwise it impairs the spectrum. When possible, the background check has been carried out by observing a sudden drop of signal of the desired photo-generated ion, when LED laser irradiating the silica capillary tip has been shut down.
The general procedure for IRPD, VIS and two-colour experiments is described in Figure S32. The conditions varies slightly at each case, mainly in a change of helium pulses or time how long the quadrupole bender guides ions to the trap. These changes are done exclusively to ensure the stability of a signal to give a reasonable signal to noise ratio. In most cases the cycle time (N) has been 1 second. shutter open shutter open Figure S32. Pulse sequence of IRPD experiments. Second shutter was opened only in case of two-colour experiments, otherwise the laser and the shutter was not part of the setup. General Experimental procedure for one-color experiments: Ions were mass selected by a first quadrupole and guided into wired 4-pole trap. There they were cooled to 3K by helium buffer gas. At this temperature and pressure, the ions forms clusters with helium. These clusters are then irradiated at various wavelengths by IR or visible range tunable laser(s). Usually energy of one absorbed photon is enough to dissociate the helium from the ion and thus reporting absorption at specific wavelength. We measure the whole spectrum by varying the wavelength of the irradiation, reporting the dissociation yield (1-Ni/Ni0) as a function of absorption coefficient at specific wavelength. The Ni0 cycle serves as a reference value for the number of cluster formed. In case of two-colour experiments used in this article, we use second laser to deplete population of specific isobaric impurity by targeting a wavelength where the impurity absorbs and the target ion does not.

Computational Details
If not stated otherwise the calculations were done by the B3Lyp DFT functional with the D3 version of Grimme´s dispersion with Becke-Johnson damping on the 6-311+G** basis in Gaussian G09. For spectra presented in the main article in figures 4 and 5, carbon atom in the CF3 group was described by an additional PC-3 basis set for a better prediction of the electron dispersion. Structures pre-optimizations were carried out by AM1 and further refined by the B97 DFT functional on the 6-31+G basis set. If not stated otherwise, the methylated analogue of 1a has been calculated instead of the original octylated one. Scaling factor for vibrational spectra stays the same as in the main article.

Computational study of different conformers of ([1b+OOH] + )
To determine most stable conformer of [1b+OOH] + we have performed series of DFT calculation. We quickly discovered that the most stable conformers are ones in which the imidazolium charge-tag is located above or below flavinium moiety ( Figure S34). We have then performed variety of dihedral angle studies to estimate the amount of conformers with similar energies. The big amount of conformers which are close in the energy corresponds well with the broadened peak observed during IRPD analysis of this ion.

4.2.2.1
Dihedral angle studies of conformers of ([1b+OOH] + ) To map potential energy surface of found minima we have performed studies of dihedral angles of the OOH group and the chain between imidazolium charge-tag and flavinium moiety ( Figures S35 and S36).  Notice that the spectra a) and d) has been recorded at a lower attenuation compared to spectrum b), which is apparent from the attenuation of the C-H stretch bands present in range 2850-3000 cm -1 .
All spectra contain a signal between 3450 cm -1 and 3560 cm -1 that is slightly above the level of noise. For the ions [1a+2H,2O] + (b) and its methyl-analogue shown in (c), the band could be ascribed to hydrogen bonding of the hydroxyl group. However, it seems that the presence of the octyl group also affects the spectra in this range. Possible contribution of some combination bands reflecting vibrations of the octyl group is also partly supported by a weaker dependence of intensity of this band on the irradiation power ( Figure S41).

HRMS-ESI+ spectra of the synthetized compounds
HRMS spectra were acquired on the JEOL JMS-T100CS AccuTOF-CS mass spectrometer using polypropylene glycol (PPG) as an internal standard for mass drift compensation. The allowed difference limit below m/z 1000 is 3 mmu.