Hydrogen peroxide assisted photorelease of an anthraquinone-based ligand from [Ru(2,2'-bipyridine)2(9,10-dioxo-9,10-dihydroanthracen-1-olate)]Cl in aqueous solution.

A new class of light-activated ruthenium(ii) complex was designed as a potential blocker of biological functioning, especially for targeting redox reactions within mitochondria under light activation. Based on our concepts the complex [Ru(bipy)2(1-hydroxyanthra-9,10 quinone)]Cl (RU1) was prepared and studied to understand the preliminary reaction mechanisms and its excited state behaviour through a series of stability tests, electrochemistry, UV-Visible kinetics and femtosecond transient absorption spectroscopy experiments. Under white light in the presence of H2O2 two different reactions (fast and slow) appear to take place. The complex loses the quinone-based ligand and a resulting Ru(iii) or Ru(v) species is produced. The complex RU1 shows potential to consume H2O2 from the one carbon metabolism in mitochondria, and hence may cut the energy cycle pathway of tumor cells.


Introduction
The cellular uptake of metallodrugs at the nucleus is much less than that on other subcellular organelles such as mitochondria. 1,2 In classical biochemistry textbooks mitochondria is known to be the major place where cells can make energy from aerobic respiration processes, and it is also called the "power house" of a cell. Mitochondria also contain their own genetic materials and systems through its genome, and can function as semi-autonomous organelles. Because of this, and besides simply providing energy to cells, they also serve as a checkpoint in cellular metabolism such as cell differentiation, activation of signaling pathways and programmed cell apoptosis. [2][3][4] Considering these important features, metabolism in mitochondria is believed to be a potential target site for the treatment of cancer, because if they disfunction at any point in the metabolism cycle cells are unlikely to survive. [5][6][7][8][9][10][11][12] Many different metabolism pathways have been investigated as a possible targeting site for cancer treatment. Here regulation of intracellular reactive oxygen species (ROS) levels is our research direction. The electron transport chain (ETC) on the mitochondria, which is also called the respiratory chain, is the main site of intracellular ROS production; 95% of ROS (including the oxygen anion (O 2 − ), hydrogen peroxide (H 2 O 2 ), hydroxyl radical (HO • ), oxygen (O 2 ) and hypochlorous acid (HOCl)) in most mammalian cells are produced from here. 13 In a normal cell oxidation and antioxidant systems are usually maintained in a relatively balanced level, and the increase of oxidation level or the decrease of anti-oxidation ability will lead to the increase of ROS in the body, which will be followed by a series of changes. 14,15 Compared with normal cells, ROS level in tumor cells is higher than that in normal cells due to both environmental and internal mechanisms, so tumor cells are usually under oxidative stress, and the sensitivity of tumor cells to ROS is higher than that of normal cells. Nevertheless, at the same time a tumor cell has its own unique antioxidant capacity, which can control the level of ROS within a range suitable for their biological function. So ROS will not accumulate to a high level that promote cell death and acute cell damage. [16][17][18] Hydrogen peroxide is a common intermediate product of cellular oxygen metabolism in aerobic organisms. The earliest idea was that H 2 O 2 was derived from the super-oxide anion (O 2 •− ) which is generated by the ETC and then converted into H 2 O 2 by superoxide dismutase 2 (SOD2) or voltage-dependent anion channels (VDACs). 19 In the process of cancerization normal cells will produce a large amount of H 2 O 2 which will accumulate in tumor cells to counterbalance the increased antioxidant capacity of cancer cells. In order not to build up large amounts of ROS, a high concentration of H 2 O 2 will also promote cell division to prevent oxidative stress (OS). Tumor cells have a strong dependence on H 2 O 2 and are very sensitive to its changes, which means it cannot tolerate excessive increase or decrease of its amount. 12,20 Based on this observation the concept was to use a ruthenium complex to consume H 2 O 2 based on the following reasons: one is to accumulate other ROS by prohibiting cell division due to the lack of H 2 O 2 , and secondly to affect its redox balance to achieve the purpose of treatment.
An archetypical photoreactive complex is [Ru(bipy) 3 ] 2+ , where bipy = 2,2′-bipyridine, which can behave as a photooxidant (Ru 3+ ) or a photoreductant (bipy •− ) in the excited state. 21 Quenching by either an oxidant or reductant is possible and has been extensively studied in the context of artificial photosynthesis (i.e., water splitting). 22 As far as we are aware from the literature there are no reports of excited state quenching of [Ru(bipy) 3 ] 2+ by peroxide; in fact, reports highlighted H 2 O 2 formation from the reduction of O 2 . 23 Hence, modification of the complex appeared to be required, and so we rationalized that incorporation of an anthraquinone-based ligand would fulfill several beneficial requirements (Scheme 1). The low reduction potential often associated with an anthraquinone would selectively localize the excited electron at the site. 24 Also, the single electron reduced CvO group in the excited state would afford a protonation site, and thus a potential pH dependency to any redox reactions. 25 The hard donor oxygen atoms would in addition facilitate stabilization of the Ru 3+ oxidation state, hence perturbing the redox potential at the metal centre. 26 Release of the anthraquinone ligand could also introduce an additional species to perturb the complicated redox chemistry within mitochondria. 27 To these ends, the chirally pure complex RU1 was targeted containing the deprotonated form of 1-hydroxyanthra-9,10-quinone (HAQ). Even though chirality at the metal centre was not critical for the current studies there is an interest in these type of complexes for binding to DNA. 28 The method employed, as developed by von Zelewsky, 29 is as straight forward as the classic approach using racemic [Ru(bipy 2 )Cl 2 ].

Characterization
The optically pure new compounds Λand Δ-RU1 were synthesized in a single step by treating the resolved ruthenium complexes 29 Λ-PR or Δ-PR with HAQ (Scheme 1). The reaction was easily monitored due to the colour change from redorange to deep purple. The base Et 3 N was used to deprotonate the 1-hydroxyanthra-9,10-quinone. The two enantiomeric complexes RU1 were firstly isolated as their PF 6 − salts. While column chromatography purification is often required to obtain pure complexes this procedure was not necessary for RU1. Both complexes were isolated as black solids with yields around 50-60%. To obtain water-soluble versions the PF 6 − anion was exchanged for Cl − using a Dowex resin. The complexes Λ, Δ-RU1 were characterized by 700 MHz high-field NMR including 1-D 1 H and 13 C NMR spectroscopy (see ESI †). Also, 2D COSY and HSQC spectra were recorded to investigate the proton-proton correlations and proton-carbon correlations. The NMR spectroscopic data are consistent with the proton and carbon numbering of RU1. The low symmetry of RU1 is a consequence of the HAQ ligand. For the bipyridine segment the low symmetry gives rise to protons in different environments. The below assignments agree with those previously reported for cis-[Ru(bpy) 2 L] n+ (L = other ligand) complexes in the literature, taking into account the close proximity of certain protons to the aromatic rings. For example, H 6a and H 6c are more directed to the ring current of the anthraquinone-based ligand while H 6b and H 6d are more directed to the ring current of a bipyridine segment. Because the anthraquinone-based ligand is more electron rich than the bipyridine fragments, the ligand therefore has a larger ring current resulting in a greater upfield shift for H 6a and H 6c compared to H 6b and H 6d . [30][31][32][33][34][35][36][37] The proton H 8 has an upfield chemical shift since it is directed towards the bipyridine group's ring current. In the free ligand, H 3 has an upfield chemical shift relative to H 4 . The numbering scheme for RU1 and assignment for the signals are in Table 1.
The ESI-MS mass spectrum for RU1 displayed a cluster of intense molecular peaks centered at m/z = 637.0813, which are consistent with the structure for the singly-charged species [RU1-PF 6 ] + . The observed and expected isotopic pattern corres- ponding to the loss of one hexafluorophosphate anion were in good agreement.
Single-crystals of RU1 were grown using the slow diffusion method by layering a solution of RU1 in CH 3 CN on top of an aqueous solution of KPF 6 in an NMR tube. The CH 3 CN layer and aqueous layer slowly mixed together over time and black, needle-like crystals formed at the interface of the two layers. These crystals were analysed by single-crystal X-ray diffraction.
The molecular structure of RU1 is labelled and illustrated in Fig. 1. Selected bond lengths and angles along with DFT modelling results are shown in Table 2. The asymmetric unit consists of one RU1 cation and one PF 6 − anion. The deprotonated 1-hydroxyanthra-9,10-quinone binds to the ruthenium ion in a bidentate coordination mode. The two bonds between the ruthenium and oxygen atoms Ru-O1 and Ru-O2 have average bond lengths of 2.038(4) Å and 2.052(4) Å. There is very little variation in the Ru-N bond lengths which are observed to be statistically similar. The structure is best described as exhibiting a distorted octahedral coordination geometry. The deprotonated 1-hydroxyanthra-9,10-quinone ligand is slightly bowed with a fold angle of 8.617(2)°between the two benzene rings. The experimental results were compared with a structure obtained by DFT computational studies (Gaussian 09 (B3LYP/3-21G*/6-311G). 38 The calculated bond lengths and bond angles are similar with those observed in the crystal structure. The packing diagram reveals that the RU1 cations form columns along the crystallographic [100] direction with each cation related to the next by inversion symmetry. Given the orientation of adjacent groups in this direction these columns are most likely formed of parallel offset π-π interactions. The closest Ru⋯Ru distances in the structure (7.610(2) Å) are also observed along the [100] direction.

Electrochemistry
The redox behaviour of the complex RU1 was performed by CV experiments in dry and deoxygenated CH 3 CN using a platinum working electrode with TBAPF 6 (0.1 M) as the background supporting electrolyte versus an Ag/AgNO 3 reference (Fig. 2). The complex exhibits one dominant reversible wave at E 1/2 = +0.45 V. The ΔE = 70 mV is typical for a one-electron diffusion controlled process and represents the Ru(III)/Ru(II) redox couple. Upon reductive scanning at a scan rate of 0.1 V s −1 , HAQ = 1-hydroxyanthra-9,10-quinone ligand; coupling constants ( J) in Hz are given in the brackets. d = doublet, t = triplet, m = multiplet.  a Value is the experimental value representing average bond lengths and angles, the number in the bracket is standard deviation of the bond lengths. b Value is calculated bond lengths and angles from DFT calculation in the gas phase using Gaussian 09 (B3LYP) and a split basis set: Ru 3-21G*; C, H, O, N 6-311G.

Dalton Transactions Paper
This journal is © The Royal Society of Chemistry 2020 there is one irreversible reduction peak at E 1 = −1.10 V which is associated with two electron reduction of the anthraquinone-based ligand. The lack of reversibility of the wave would suggest that after two-electron reduction there is a chemical reaction. This is likely protonation by any water that is often found even in dried CH 3  The broad and much weaker absorption tail around 700 nm is possibly due to the spin-forbidden singlet to triplet MLCT transition (Fig. 3), [40][41][42] or the LLCT transition associated with electron migration from the quinone-based ligand to a bipy group. 43 The absolute configurations of Λ-RU1 and Δ-RU1 were investigated by circular dichroism (CD) in both CH 3 CN and H 2 O. Within experimental error, the spectra of both compounds are the mirror images of each other with strong oppo-site but equal Cotton Effects as expected for enantiomeric pairs. Seven clear signals are observable in both CH 3 CN and H 2 O (see ESI †). [44][45][46][47] Transient absorption spectroscopy Since no apparent room temperature luminescence was observed for RU1 in solution, femtosecond pump-probe transient absorption spectroscopy was employed to investigate its excited state properties. The complex RU1 in H 2 O ( pH 7) was excited at 500 nm with an ultrashort 100 fs laser pulse, and 100 time-resolved spectra were recorded in the 0.1 to 300 ps time frame (Fig. 4). At a very early time delay, there is a clear bleach centered around 600 nm consistent with depletion of the ground state complex. Well within 1 ps a new positive transient feature is unmistakably observed towards the red-end of the spectrum and this is assigned to the MLCT excited state. The kinetics for recovery of the ground-state bleach and the decay of the transient profile are in good agreement, supporting the direct restoration of the ground state ( Fig. 4

inserts).
Multi-wavelength global analysis of the decay kinetics required three exponentials resulting in time constants of 0.14 ps, 0.70 ps and 5.3 ps. The first point of note is the ultrashort lifetime of the excited state as opposed to the normal long-lived 3 MLCT state observed for [Ru(bipy) 3 ] 2+ . 48 It is well established that intersystem crossing of the 1 MLCT to the 3 MLCT in [Ru (bipy) 3 ] 2+ is ultrafast (<30 fs). 49 It is reasonable to assume the same process would occur in RU1, especially since at the excitation wavelength of 500 nm (Fig. 3) a proportion of photons go into forming a similar looking 1 MLCT state (i.e., [Ru(III) (bipy •− )(bipy)AQ] + ). It is also likely that the preferentially formed lower in energy 1 MLCT state (i.e., [Ru(III)(bipy) 2 AQ •− ] + would also undergo rapid intersystem crossing to the 3 MLCT state. The 0.14 ps time constant is outside the limit of representing intersystem crossing and is more likely a relaxation timescale as the complex vibrationally cools. The simplest explanation for the 0.70 ps time constant is it exemplifies the relaxation of the upper-lying 3 MLCT state (bipy-based) to the   3.0 ps (blue) and 11.4 ps (green) after excitation of RU1 in water with a 70 fs laser pulse delivered at 500 nm. Insert shows decay curves and fits in red at two different wavelengths to a three exponential model. lower-lying 3 MLCT state (quinone-based); this latter state then returns to the ground state in 5.3 ps. The ultrashort lifetime is likely linked to the fact that the 3 MLCT state (quinone-based) is so low in energy that it is strongly coupled to the ground state (energy-gap law). 50 The simplest analogues of RU1 are the catechol-based complexes as described by Lever and coworkers, 51 for which there are no reported excited state lifetimes. Other complexes containing directly metal attached anthraquinone units (e.g., emodin) are known but no lifetime data are given. 52 Similar pump-probe experiments performed in the pH 2 and 4 did not reveal any tangible changes in the transient profiles or decay kinetics. Likewise, no effect on the excited-state lifetime was observed for RU1 in the presence of H 2 O 2 as discussed later. This finding would appear to rule out any major complex formation between the two species in the ground state prior to excitation.

Light activated ligand disassociation
Given the ultrafast excited state decay of RU1, and the lack of any effect on its lifetime by the presence of peroxide, it was unexpected that clear alterations were observed under steadystate light activation conditions (Fig. 5). As a typical example the reaction of RU1 in an aqueous solution treated with H 2 O 2 at pH 2 is used to highlight certain features. The UV-Vis spectra of the solution were recorded every 5 min till 70 min, when the colour of the solution was almost clear. The spectral change at 580 nm versus time is the most distinctive feature (Fig. 5). The disappearance of the absorption band can be due to complete disassociation of the quinone-based ligand, or the oxidation of Ru 2+ to Ru 3+ which would essentially switch off the MLCT transition. A very subtle effect is observed at around 400 nm where over some 20 min the absorbance increased slightly and then decreased steadily over time. The spectrum recorded after 70 min showed a clear broad band centered at 752 nm. The spectrum is remarkably similar to that recorded for the in situ generated complex [Ru(III)(bipy) 2 AQ] 2+ (see ESI †). 53 It does appear that the first part of the photoreaction is formation of the oxidised metal centre. However, there is likely an additional very slow photodriven reaction that results in loss of the quinone-based ligand. A sample of RU1 in D 2 O irradiated for several hours showed clear alterations in the collected 1 H NMR spectra, and resulted in the precipitation of a white solid (see ESI †). Large chemical shifts were also observed for several resonances supporting the generation of paramagnetic species such as Ru(III)-d 5 or even Ru(V)-d 3 . Existence of a higher oxidation ruthenium species was confirmed by collection of an electrospray mass spectrum of the irradiated mixture. A cluster of peaks at m/z = 445/447 was observed which can be assigned to species such as [Ru(V)(bipy) 2 O 2 ] + (445) or [Ru(III)(bipy) 2 (OH) 2 ] + (447) (see ESI †). These species presumably originate from an aqua complex initially formed from water molecules occupying vacant coordination sites.
A similar photo-driven degradation of RU1 in the presence of peroxide was observed in the pH range 2-7, albeit that the reaction did slow down as the solution became neutral. Under the basic conditions of pH 8-10 the reaction was extremely slow, but at pH >11 decomposition of the complex again became clear. At high pH it is likely that reactions involving hydroxide ion become important, since it is recognized that [Ru(bipy) 3 ] 2+ is subject to decomposition by hydroxide attack at the bipy ligand.
There are several additional points of note regarding the decomposition of the complex. The solution when left in the dark showed very little change supporting that light is required at some point to initiate the reaction. In the absence of both hydrogen peroxide and oxygen the decomposition reaction is again non-existent under light illumination. However, there is a very clear reaction when both oxygen and hydrogen peroxide are present both in the dark and under light stimulation (see ESI †). Control experiments performed using [Ru(bipy) 3 ] 2+ under conditions discussed above proved that peroxide assisted decomposition of RU1 must arise because of the presence of the anthraquinone ligand. The decomposition of the complex is the same for both isomers and likely results in racemization.

Degradation kinetics and model
The kinetics for the degradation reaction as monitored at the main absorption band 580 nm versus time is zero order (Fig. 5). This situation is not unprecedented for photoactivated reactions. 54 Analysis of the data at pH 2 by conversion of absorbance to concentration and using eqn (1) afforded k obs = 3 × 10 −8 M s −1 , where [C] is the concentration of RU1 at time t and its initial concentration is [C] 0 . The corresponding halflife (t 1/2 = [C] 0 /2k obs ) is 22 min.
The kinetics of degradation were also analysed from pH 3-7 in a similar manner to above (see ESI †) and the results are col- lected in Table 3. The abrupt change in k obs between pH 6-7 would suggest that this point represents the pK a for the excited state of RU1 located at the C-O unit. The pK a (eqn (2)) for hydrogen peroxide is 11.8, 55 signifying that under the conditions described the ratio falls within the range 1.6 × 10 −10 to 1.6 × 10 −5 . Consequently, the probability of any ground state equilibrium complex with the cation RU1, because of ion-pair formation, would be extremely low. Certainly, the concentration of any ion-pair complex would have been far too small to show up as lifetime changes in the pump-probe experiments discussed previously.
The major pathway for RU1 following excitation is assigned to a four-cycle system as illustrated in Fig. 6. The pK a of the conjugate acid of anthraquinone (ANQ) is −8.2 in H 2 O. 56 Given this pK a value and the lack of evidence for protonation of the ground state complex it is proposed that following excitation the C-O •− radical is protonated; 57 the pK a being around 6.5 which is reasonable since the pK a of the conjugate acid of ANQ •− is ca. 5.3 in DMF. 58 The protonated version of RU1 following charge recombination then rapidly loses a proton to complete the cycle. To account for the slow decomposition of RU1 a very minor pathway must exist that removes an electron from the quinone-based ligand to generate the ruthenium(III) complex (cf. Fig. 5). 59 The one electron reduction of hydrogen peroxide in the presence of acid (eqn (3)) has a standard redox potential E°= +0.80 V (vs. NHE). 60 In addition, the hydroxyl radical formed is an extremely strong oxidizing agent (eqn (4)) with E°= +2.73 V (vs. NHE). 60 One electron oxidation of photoexcited RU1 by H 2 O 2 appears to be a feasible reaction 61 and as a result would produce a highly oxidising radical, which may go on to oxidise a further ground state complex. Hence, the input of one photon could in principle result in the formation of two ruthenium(III) complexes. Over a much longer timescale the ruthenium(III) complex (Fig. 6) decomposes further, possibly promoted by light, to release HAQ. There are likely further redox reactions involving the [Ru(bipy) 2 (OH 2 ) 2 ] 3+ ion, because of the excess peroxide present in solution.

Conclusion
The incorporation of an anthraquinone-based ligand onto the "Ru(bipy) 2 " fragment is a facile way to produce a complex (i.e., RU1) that absorbs further to the red. This positive feature is unfortunately accompanied by the loss of any discernable phosphorescence. The efficient non-radiative excited state deactivation is doubtless the result of strong excited to groundstate coupling. The poor emission does rule out the use of RU1 as a luminescent probe. However, it is a photoactivatable complex that can consume hydrogen peroxide, albeit the reaction is slow under the conditions employed and operates best at low pH. The pH of the inner membrane of normal mitochondria is around 7, and so it has the potential to operate as a peroxide concentration disruptor. 62 Clearly, there are many other redox active species (e.g., NAD + /NADH, FAD + /FADH) within mitochondria that could interact with the excited state of RU1. We expect to test these concepts in follow up work in the future, exploring the localization of RU1 in mitochondria and its toxicity under light illumination.

Materials
All the chemicals were purchased from commercial sources and used as received without further purification. For the studies undertaken in water, ruthenium hexafluorophosphate salts were converted to chloride salts by using Amberlite® IRA-410 chloride form. Aqueous solutions were prepared using a PURELAB® Option-Q water purification system. High-field nuclear magnetic resonance (NMR) spectra including 1 H NMR, 13 C NMR, COSY, HSQC, HMBC were collected on Bruker Advance III HD 700 MHz spectrometer. Chemical shifts for all complexes were reported relative to acetonitrile-d 3 (Cambridge Isotope Laboratories, CD 3 CN, 99.8%) at δ = 1.940 ppm. Highresolution electron spin ionization (ESI) mass spectrometry was collected by the National Mass Spectrometry Facility (NMSF) in Swansea.   1, 30 mL) was added 1-hydroxyanthra-9,10-quinone (22 mg, 0.1 mmol, 1 e.q.) and trimethylamine (0.015 mL, ∼0.1 mmol, 1 e.q.). The reaction mixture was shielded from light, stirred and heated at 120°C under a nitrogen atmosphere for 4 h until all the starting material was consumed during which time the solution changed from orange to dark-red. The solution was cooled down to R.T. and Λ-RU1 was precipitated by addition of excess KPF 6 from water. The black compound was collected by filtration and rinsed with distilled water and allowed to air dry overnight (

X-Ray crystallography
Crystal structure data for RU1 were collected at 100 K on beamline I19 at Diamond Light Source using synchrotron radiation (λ = 0.68890 Å). 63,64 These data were processed using the software APEX3 (Bruker, 2015). The structure of RU1 was solved using XT 65 and refined using XL 66 through the Olex2 interface. 67 All non-hydrogen atoms were refined anisotropically. Hydrogen atoms were positioned with idealized geometry and their displacement parameters were restrained using a riding model with U(H) set to be 1.2 times the U eq value of the parent atom.

Computer calculations
Computational calculations were performed using a 32-bit version of Gaussian09 68 on a quadruple-core Intel Xeon system with 4GB RAM. The calculations were run in parallel, fully utilising the multi-core processor. Becke's three paramater exchange correlation functional and the Lee, Yang and Parr correlation functional that recovers dynamic electron correlation (B3LYP) coupled to a split basis set was used in all calculations. The ground-state structure for RU1 was calculated using a 3-21G* basis set for the ruthenium ion and 6-311G for all other atoms. Energy minimisation calculations were monitored using Molden and run in parallel with frequency calculations to ensure optimised geometries represented local minima.
Circular dichroism (CD) spectra CD spectra were measured by a JASCO 810 spectropolarimeter (JASCO International Co., Ltd). The spectra were obtained in acetonitrile at 20°C using a 1 cm path length quartz cuvette (Starna® Scientific Ltd). Each spectrum was accumulated from 3 scans between the range of 200-950 nm, at a scan rate of 100 nm min −1 with a standard sensitivity and continuous scan mode. The data were recorded every 0.2 nm. Concentrations of Λ or Δ-RU1 were prepared in a 1 × 10 −3 M stock solution and adjusted to an appropriate concentration where the HT [V] was lower than 600 V.

Electrochemistry
Cyclic voltammetry experiments were used to determine the redox potentials of the individual components of RU1. Measurements were performed using a fully automated CH Instruments Electrochemical Analyzer and a three electrode set-up on BASi® C-3 Cell Stand at room temperature. The three electrode set-up consisted of a platinum working electrode, a platinum wire counter electrode and an Ag/AgNO 3 reference electrode. All studies were carried out in deoxygenated and anhydrous CH 3 CN containing TBAPF 6 (0.1 M) as the background electrolyte. Solution concentrations were 0.1 mM.

Photochemical kinetic studies
UV-visible absorption properties under different conditions of RU1 were collected on Shimadzu Spectrophotometer UV-1800 at a resolution of 0.5 nm with medium scan speed from range

Transient absorption spectroscopy
Excited state dynamic processes in femtoseconds to nanoseconds time scales were studied using the pump-probe method as described in detail elsewhere. 69 The fundamental light source at 800 nm was generated by a Ti:sapphire laser system (Libra F laser system, Coherent Inc.) which produced 1 mJ pulses with duration 100 fs at a 1 kHz repetition rate. Majority (∼90%) of the fundamental beam power was delivered to Optical Parametric Amplifier (OPA) (TOPAS C, Light Conversion Ltd) and was tuned to generate pump pulses at 410, 470 or 500 nm for RU1. The pump beam diameter at the sample was around 0.5 mm and the power was attenuated by optical filters to excitation density at the sample <0.1 mJ cm −2 .
A small amount (∼10%) of the Libra-output light was delivered to the measurement system (ExiPro, CDP Inc.) which utilized a sapphire plate as a White Light Continuum (WLC) generator to produce the probe beam. The probe beam was then split into two parts: the probe signal and probe reference beams (∼0.1 mm diameter). The delay time limit for the machine is ∼6 ns. The signal and reference beams were passed through the sample and their spectra were measured by monochromator with a pair of array detectors. The gap between the wavelength ranges is due to strong WLC distortions caused by the fundamental (800 nm) pulse of the Libra laser. The transient absorption spectra were obtained by comparing responses with and without excitation using a chopper synchronized with the fundamental laser pulses and blocking every second pump pulse. ExiPro Program (CDP Corp.) was used to control the experiment and calculate the spectra. The sample solutions for pump-probe measurements were prepared to have absorbance between 0.2-0.8 optical densities (OD) in 2 mm cuvette used for the measurements and equipped with magnetic stirred bar. The optical scheme was aligned for maximum overlap of the pump and prove signal through the whole cuvette thickness. Zero reference time was found by looking for the start of signal build-up, with the delay line first set for the pump pulse hitting the sample after the probe pulse and moving the delay line so that pump started to overlap the probe in time. Steady state absorption spectra were measured before and after the pump-probe measurement to confirm that the sample did not change during the measurement.

Conflicts of interest
There are no conflicts of interest.