Spectroscopic Studies on Photoinduced Reactions of the Anticancer Prodrug, trans,trans,trans‐[Pt(N3)2(OH)2(py)2]

Abstract The photodecomposition mechanism of trans,trans,trans‐[Pt(N3)2(OH)2(py)2] (1, py=pyridine), an anticancer prodrug candidate, was probed using complementary Attenuated Total Reflection Fourier Transform Infrared (ATR‐FTIR), transient electronic absorption, and UV/Vis spectroscopy. Data fitting using Principal Component Analysis (PCA) and Multi‐Curve Resolution Alternating Least Squares, suggests the formation of a trans‐[Pt(N3)(py)2(OH/H2O)] intermediate and trans‐[Pt(py)2(OH/H2O)2] as the final product upon 420 nm irradiation of 1 in water. Rapid disappearance of the hydroxido ligand stretching vibration upon irradiation is correlated with a −10 cm−1 shift to the antisymmetric azido vibration, suggesting a possible second intermediate. Experimental proof of subsequent dissociation of azido ligands from platinum is presented, in which at least one hydroxyl radical is formed in the reduction of PtIV to PtII. Additionally, the photoinduced reaction of 1 with the nucleotide 5′‐guanosine monophosphate (5′‐GMP) was comprehensively studied, and the identity of key photoproducts was assigned with the help of ATR‐FTIR spectroscopy, mass spectrometry, and density functional theory calculations. The identification of marker bands for some of these photoproducts (e.g., trans‐[Pt(N3)(py)2(5′‐GMP)] and trans‐[Pt(py)2(5′‐GMP)2]) will aid elucidation of the chemical and biological mechanism of anticancer action of 1. In general, these studies demonstrate the potential of vibrational spectroscopic techniques as promising tools for studying such metal complexes.


Introduction
Platinum therapy is standard practice for approximately 50 % of patients undergoing anticancer treatment, but suffers from major drawbacks. [1] These include acquired and inherentr esistance as well as off-target effectsl eading to short-and longterm strain on the patient. [2] Similar shortcomings foro ther chemotherapeutic agents in clinical use have stimulatedt he searchf or new anticancer drugs exhibiting novel mechanism(s) of action.
Photoactivatable metal-basedp rodrugs can release their reactive metal-based species locally upon activation by light, also referred to as photoactivated chemotherapy (PACT). It has been shown in the past decade that such metal-based PACT candidates can provide localised toxicity throught heir metal core and/or their releasedligands. [3] Trans,trans,trans-[Pt(N 3 ) 2 (OH) 2 (py) 2 ]( 1,p y= pyridine, Figure 1) is one such potent PACT prodrug candidate capable of provid- ing reactiveP t(IV/II)-, azido-andh ydroxyl-(radical) based species. [4] Complex 1 can be specifically activated locally in cancer cells by UVA, blue or green light, affording am ultitargeted biological activity,w hilst being inert andn on-toxic in the dark. [4,5] Recently,w er eportedacomprehensive vibrational spectroscopic study of 1,i dentifying its highly defined mid-andf ar-IR vibrational fingerprints. [6] In order to provide new insighti nto our understandingo ft he mechanism of action of 1,w en ow study the photodecomposition of 1 using Attenuated To tal Re-flectionF ourier TransformI nfrared( ATR-FTIR)s pectroscopy. [6,7] Additionally,t he photoinduced reactiono f1 with guanosine 5'-monophosphate (5'-GMP) was investigated. Transient Electronic Absorption Spectroscopy (TEAS) has been employedt o investigate the ultrafast photodynamicso f1,w ith UV/Vis spectroscopya nd mass spectrometry being used to study the photoinducedr eactions and for identification of the potential photoproducts. [4,8] Data analysis has been performed using Principal Component Analysis (PCA) and Multi-Curve ResolutionA lternating Least Squares (MCR-ALS), which allow for correlation of data from the various techniques. Moreover,i nterpretation of the captured spectral changes has been supported with Density Functional Theory( DFT) calculations, aiding the assignment of the observed dynamics.

Results
Photodecomposition of 1b yA TR-FTIR Studieso ft he photodecomposition of 1 under 420 nm irradiation, monitored by ATR-FTIR, were carried out as described in the Experimental section. The evolution of the ATR-FTIR spectra of the full spectral range between 3700-640 cm À1 is shown in Figure 2A,w hichi ncludes selected characteristic peak assignmentsf or 1 reported previously. [6] The detection of changes in Figure 2A is aided by the corresponding second derivatives pectra divided into three spectral windows, 3650-2950 cm À1 ,2 100-1980 cm À1 ,a nd 1650-640 cm À1 (Figure 2B-D).
The first spectralw indow between 3650 and 2950 cm À1 ,F igure 2B,s hows as ingles harp hydroxido stretching vibration, n(OH Pt-OH ), at 3550 cm À1 ,t hree azido overtones (3373 cm À1 , 3330 cm À1 ,a nd 3302 cm À1 ), pyridine CH stretching vibrations n(CH py )b etween 3126 cm À1 and 3008cm À1 ,a nd a n(OH) band centred at approximately 3300 cm À1 ,w hich is very broad due to hydrogen bonding. Upon 420 nm irradiation, ar apid decrease in n(OH Pt-OH )i so bserved in the first 10 min ( Figure 2B) in which complete disappearance is reached after 20 min (Figure S2 in the Supporting Information, intensity vs. time plot), whilst the broad n(OH) feature increases from the start to the Figure 2. ATR-FTIR spectra of the photodecomposition of 1 (2 mm in water) under 420 nm irradiation.A )Full spectral range, 3700-640cm À1 .B)Second derivative spectra,3 650-2950 cm À1 ,highlightingr apid decrease in n(OH Pt-OH )a t3 350 cm À1 in the first few minutes. C) Second derivative spectra, 2100-1980 cm À1 , highlightingthe decrease and shifts in n sym (N 3 )at2048/2033 cm À1 to 2023 cm À1 and 2044 cm À1 sequentially prior to complete disappearance. D) Second derivative spectra, 1650-640 cm À1 ,highlighting decrease in n sym (N 3 )at1 271 cm À1 and increase in d(OH Pt-OH )at1 381 cm À1 and 1334 cm À1 .N otations used: n = stretch, d = in-plane angle bending, g = out-of-plane anglebending, sym = symmetric and asym = antisymmetric. The labelling of pyridine modesi sg iven in Wilson's notations,asd escribed previously. [9] end ( Figure 2A). Concurrently,t he n(CH py )v ibrations broaden and shift slightly (< 5cm À1 )o ver time. Furthermore,s low removal of the azido overtones is observed.
The second spectral window between 2100 cm À1 and 1980 cm À1 coverst he antisymmetric azido stretching vibration, n asym (N 3 ). Prior to irradiation, n asym (N 3 )c onsists of one main vibrationa t2 033 cm À1 and as houlder at 2048 cm À1 as indicated in the second derivative spectra in Figure 2C.D uring the first 5min of irradiation, ad ecreasei ni ntensity of both of these bands is observeda nd additionally the main vibration at 2033 cm À1 shifts to 2023 cm À1 .C ontinuedi rradiation shifts the main vibration to 2044 cm À1 prior to complete removal (Figure 2C).
As shown in Figure 2A and D, there is little effect on the pyridine vibrations throughout the irradiation process. Minor shifts do occur as shown by the gradual shift of the pyridine bending mode, d(py,19b), from 1460 cm À1 to 1456 cm À1 (Figure 2D). The removal of the azido vibrations can be observed by the changes to n sym (N 3 )a nd d(N 3 ), but is lessd ramatic compared to the antisymmetric stretching vibration. Both these vibrations overlap with pyridine modes, d(py, 14) and g(py,11), respectively,w hich remainp resenta t1 250 cm À1 and 690 cm À1 after completion of the photodecompositiona ss hown in Figure 2D. Interestingly,abroad band gradually appearsu pon irradiation underneath the two previously assigned weak pyridine bending modes at 1381cm À1 and 1360 cm À1 (d(py,3 ), in-and out-of-phase) (Figure2A,D), with the peak at 1360 cm À1 shifting to 1334 cm À1 .T his broad band can be reasonably assigned to the OH bending mode of bound water to platinum (d(OH Pt-OH=OH 2 )) and is considered in more detail in the Discussion section, below.A tthes ame rate, ab road water bending mode, d(OH H2O ), appears at % 1650 cm À1 (Figure 2A  The presence of water in the resulting product matrix is clearly indicated by the growth of the broad n(OH) and d(OH) bands of water centred at % 3300 cm À1 and % 1650 cm À1 (Figure 2A). Furthermore, an ew OH bending mode is observed underlying the pyridine bending modes at 1381 cm À1 and 1334 cm À1 (Figure 2A,D).
Part of the experimental procedure involves the evaporation of solvent by aN 2 flow directed onto the ATRc rystal prior to measurement by ATR-FTIR (see Experimental Section). When comparing spectra taken with the N 2 flow left on during measurement,t he intensity and band shapes of the n(OH) and d(OH) bands are reduced ( Figure S4). The new band that ap-pears underlying the pyridine bending modes at 1381 cm À1 and 1334 cm À1 is also influenced by the nitrogen flow,h owever,its intensity is increased,whereas all other bands remainunchanged. Additionally,w hen 1s can ( % 1s)A TR-FTIR measurementsa re recorded continuously, whilst turning the N 2 flow ON and OFF with two second intervals, the spectra interconvert without intermediates visible ( Figure S4).

ATR-FTIR:PCA
Principalc omponent analysis (PCA) was carried out in order to capturet he major trends in the data and correlate the observed spectral changes.P CA is an unsupervised analytical data method to assessl arge data sets and is widely used in a variety of sciences and industries. [10] In PCA, ad ata set is reduced in dimensionality by finding directions through the data set where the variation is greatest. The first principal component (PC) accountsf or the majority of variation in the sample spectra and each successive orthogonal PC accounts for ad ecreasingp roportion of the variance. Each spectrum can ultimately be represented as ap oint in multivariate space plotted on the PCs. In the case of the ATR-FTIR data set, each PC represents an ew axis that the projected objects (scores) can be plotteda gainst each other in order to find patterns or clusters, that is, how samples are correlated or anti-correlated. The PC loadings are the cosine angles that the scores make with the centroid and can be analysed to determine what variables (wavenumber values) are responsible forthe clustering.
The full spectralr egion (3700-650 cm À1 )w as used to carry out the PCA analysis on the photodecomposition of 1 under 420 nm irradiationb yA TR-FTIR and three PCs were selected, capturing 93 %o ft he variance. The resulting PCA scores and loading plotsa re shown in Figure3;i ndividual loading plots and PC vs. time plotsa re in Figure S5 in the Supporting Information. In this case, positive scores are associatedwith positive loadings and vice versa. The PCA loading vibrations and irradiation times with positive PC scores are tabulated in Table 1, and are described below.
Both score plots PC1 vs. PC2 and PC1 vs. PC3, follow at rajectory over time as indicated by the black arrow in Figure 3B and C. Positive scoresf or PC1 are observed mostly from 0t o 20 min ( Figure 3B,C), with highest scores in the first 4min. After 5min of irradiation, ac ombination of positive scores is obtained for PC1 and PC3, reducing towards 15 min (Figure 3C,b lue and yellow ellipse). It continues to transition into PC2, which remains dominantu ntil 120 min of irradiation (Figure 3B). Subsequently, the scores transition back into PC3, which continues to become more prominent until the end of irradiation ( Figure 3C). Due to the positive scores on both PC1 and PC3 from 5t o1 5min ( Figure 3C,b lue/yellow circle), it can be deduced confidently,w itht he related loadings, that this is due to an initial shift in the n asym (N 3 )o fÀ10 cm À1 (2033 to 2023 cm À1 )c ombined with the reduction of n(OH Pt-OH )a t 3550 cm À1 ( Figure 3A,b lue and yellow). Thisf eature is strongly present after 5min and reduces towards 15 min. Further transition of the n asym (N 3 )i st hen capturedi nP C2, in whicht he main peak is located at 2044 cm À1 ( Figure 3A,B). Thus, as econd shift of n asym (N 3 )i sc aptured by the PCA of + 21 cm À1 after removal of n(OH Pt-OH )( 2023 to 2044cm À1 ). Positive scores on PC2 remain present until 120 min of irradiation, hereafter PC3 becomes dominant until the completiono ft he photodecomposition of 1 ( Figure 3C). This captures the growth of the new broad d(OH Pt-OH=OH 2 )b and (1381 and 1334 cm À1 )f rom 60 min onwards. The PCA analysis further captures more gradual changes to the spectra,s uch as the shift of d(py,19b) from 1460 cm À1 to 1456 cm À1 ,a sw ell as the À45 cm À1 shift of the g(py,10a)vibration from 874 to 829 cm À1 .   Figure S6. Starting from complex 1 at time zero, two new products appear after 5min with different retention times (P1 and P2) prior to the appearance of another absorption P3, which remains as the final product. Products P1 and P2 evolve at the same ratio based on the integrated HPLC peak areas (ratio P2/P1 = 0.43 AE 0.09, between 1-60 min), which indicate that one of these speciesc ould be the result of solvent exchange ( Figure S7).

Photodecompositionof1b yU V/Vis
The evolution of photoproducts formed during the photodecomposition of 1 was further investigated by UV/Vis spectroscopy.T he UV/Vis spectrum of 1 (50 mm,w ater) in the dark prior to irradiation was in line with ap reviousr eport ( Figure S8). [4] It containst wo characteristic absorption bands at 294 nm and 260 nm, assigneda sL MCT (N 3 ,O H !Pt;L MCT = ligand-tometal charge-transfer) and mixed 1 LMCT/ 1 IL (OH!Pt,N 3 ;I L = interligand) transitions. [4] The decrease in intensity of the absorptionb and at 294 nm upon irradiation was monitored to determine the pseudo quantum yields. [4] Here, the trends and speciesf ormed upon irradiation have been investigated, which has not been carriedo ut previously.F igure S8 shows the photodecomposition of 1 (50 mm)u nder 420 nm irradiationb yU V/ Vis, where the absorption band at 294 nm disappears whilst the band at 260 nm increasesa nd shifts to 256 nm in the first 10 min before decreasing with no furtherc hanges observed after 150 min. The photodecomposition of 1 andi ts synthesised precursor, trans-[Pt(N 3 ) 2 (py) 2 ]( 2,F igure 1), under 310 nm irradiation in acetonitrile were additionally examined by UV/Vis. The same evolution for 1 was observed regardless of the solvent or excitation wavelength ( Figure S13). The UV/Vis spectrumo f2 (50 mm,a cetonitrile) in the dark prior to irradiation contains two characteristic absorptionb ands at 330 nm and 263 nm, with another weakb and at 403 nm observable at higher concentrations ( Figure S12). Upon 310 nm irradiation, both the 330 nm and 263 nm absorption bands decrease over time, including ab lue shifto f1 1nmi nt he band at 263 nm to 252 nm at completion ( Figure S14). Overlaying the photodecomposition time points 10 and 4min of 1 and 2 in acetonitrile reveal ah igh degree of similarity ( Figure S15).

ATR-FTIR and UV/Vis:m ulti-curve fitting
Multi-Curve Resolution Alternating Least Squares (MCR-ALS) was carried out to analyse the photodecomposition of 1 and 2,a sd escribed in the Experimental Section. The UV/Vis data were fitted using at wo-step kinetic model, capturing the observed changes in the spectra ( Figure S8, S13, and S14 in the Supporting Information). The resulting fitted spectraa nd con-centration profiles, including rate constants are shown in Figure S16. Fitted spectra of the photoproducts are labelledP 1+ P2 and P3 for 1 based on the analytical HPLC traces and P1 and P2 for 2.I na ll three instances, the first trace is alike to the correspondingdark spectra of 1 in water as well as acetonitrile, and 2 in acetonitrile (Figure S16 A,C,E, respectively). The first intermediate product of 1 in water under 420 nm irradiation has two absorption bands at 315 nm and 256 nm (P1 + P2), with the final product (P3) containingasingle absorption band at 256 nm. The photodecompositions of 1 and 2 in acetonitrile under 310 nm irradiation both show identicals pectra for their first intermediate species( 1:P 1 + P2 and 2:P 1) with absorption bands at 297 nm and 254 nm. Theirc orresponding final products (1:P 3a nd 2:P 2) contain as inglea bsorption band at 253 nm and 252 nm, respectively.
Using the same kinetic model, the ATR-FTIR spectrab etween 2100-640 cm À1 of the photodecomposition of 1 in water under 420 nm irradiation weref itted by MCR-ALS and photoproduct assignments can be proposed based on the deconstructed ATR-FTIR spectra ( Figure S17 in the Supporting Information). The deconstructed spectrum of the photodecomposition of 1 at 0min is in line with the dark spectrum ( Figure 2A) with the main n asym (N 3 )v ibration at 2031 cm À1 .T he P1 + P2 speciesh as as ingle n asym (N 3 )v ibration centred a2 042 cm À1 , the d(OH Pt-OH=OH 2 )b ands at 1379 cm À1 and 1334 cm À1 and all corresponding trans pyridine modesa so utlined in Figure 2A.

TEAS:p hotodecomposition of 1
TEAS studies of 1 in acetonitrile using 310 nm UVAe xcitation pulsesw ere carriedout to gain insight into the ultrafastphotodynamics occurring prior to the formationo ft he observed species by steady state ATR-FTIR, HPLC, and UV/Vis. Attempts using blue 420 nm excitation in both water and acetonitrile resulted in no observablet ransient species.
The transient absorption spectra (TAS) are presented as a false colour-map, shown in Figure 4. After initial excitation, the TASa re dominated by ab road excited state absorption (ESA) feature spanning 350-690 nm. As the pump-probe time delay (Dt)i ncreases ( % 1ps),t he ESA begins to undergo two spectral changes,c onsisting of ad ecay in the ESA at wavelengths longer than 550 nm and ag rowth in absorption at wavelengths shorter than 350 nm. This evolutionl eads to ab road sloped ESA spanning the entire probe window.A sDt further increases, three (positive) features begin to emerge, around 345 nm, 420 nm and 500 nm. At Dt > 5pst he peak at 345 nm begins to decay away to reveal an egative spectral feature by % 20 ps, leaving the ESA peaks at 420 nm and 500 nm, which extends to the maximum available Dt (2 ns). The resultingn egative feature is assigned to ag round state bleach (GSB) as it overlaps with the ground state absorption of 1.
To extract the dynamical information in the TAS, as equential global fitting analysis was carried out using the Glotaran software package. [11] To fully model the TAS, four time-constants (t n = k n À1 )w ere used and the fit wasc onvoluted with the instrumentr esponse function ( % 80 fs). The resulting evolutionassociated difference spectra (EADS) and corresponding timeconstants are shown in Figure 5. We will discuss the significance of these EADS in more detail later (see Discussion). We note that the value given for t 4 was significantly greater than the maximum available experimental Dt,t hus quoted as @ 2ns. The TASr esidual plot, which reports on the goodness of fit, can be found in the Supporting Information ( Figure S18).

ATR-FTIR:photoinduced reaction of 1w ith 5'-GMP
The photoinduced reaction of 1 with 5'-GMP monitoredb y ATR-FTIR under 420 nm irradiation was carried out as described in the Experimental Section. Results are divided into the ATR-FTIR spectra of compounds in the dark and the photoinduced reactions pectra between 3800-1950 cm À1 ,1 800-1200 cm À1 and 1190-640cm À1 .
Photoinduced reaction of 1with 5'-GMP (1150-640 cm À À1 ) The last spectral window between 1190-640 cm À1 in Figure 6D,shows the changes to the phosphate-related vibrations throughout irradiation. The broad antisymmetricp hosphate stretching vibration, n asym (PO 3 2À ), peaking at 1076 cm À1 (d(py,9a)) increases slightly during irradiation whilst the 1111 cm À1 and the new peak at 1095 cm À1 increase moren otably.T he symmetric phosphate stretching vibration, n sym (PO 3 2À ), at 976 cm À1 shows am inor increase in intensity during irradiation. The n(PO) vibrations at 802 cm À1 and 777 cm À1 change throughout irradiation, in which the 777 cm À1 peak decreases and two new peaks at 783 cm À1 and 768 cm À1 develop and increaseinintensity.Lastly,aminor shift in the g(py,11) + d(N 3 )vibrationo f1 from 689 cm À1 to 692 cm À1 is observed, which is in line with the photodecomposition results of 1.

Discussion
Photochemotherapy is attractive for the treatment of cancer because it uses non-toxic prodrugs, which are activated only in the region of tumoursw ith spatially directedl ight, so minimizing toxic side-effects on healthy tissue. Photoactive Pt IV complexes such as 1 show promise, being active against cancer cells at low micromolar doses and active in vivo using visible light. Moreover,t hey kill cancerc ells by unusual mechanisms. It is importantt oe xplore methods to elucidate the photochemistry and photo-biochemistry of these complexes.A l-thoughN MR spectroscopy is often informativef or solution studies, in the present case, it is difficult to monitort he azido and hydroxidol igandsb ecause resonances for quadrupolar 14 N are broad and 15 Nr esonances cannot be enhanced by polarization transfer because there are no coupled protons. [14] The presents tudies show that vibrational spectroscopy can make a significant contributioni nt his area.
The photodecomposition of 1 under 420 nm irradiation is likely to involve binding and/orinteraction with water (solvent) when no other competing binding sites are present. ATR-FTIR revealed the formationo fanew OH bendingm ode underlying the peaks 1381 cm À1 and 1360 cm À1 (Figure 2A and D). This broad band is reasonably assigned to the OH bending mode of water bound to platinum (d(OH Pt-OH=OH 2 )), behaving similarly to the n(OH) and d(OH) bands of water when exposed to aN 2 flow during spectrala cquisition ( Figure S4). This is further supported by previousr eports on platinum group metal hydroxides, where broad metal-OH bending modes were observed between 1600-1000 cm À1 . [15] Additionally,w hen 1scan ( % 1 sec) ATR-FTIR measurements are taken continuously whilst turning the N 2 flow ON and OFF,w ith two second intervals, the spectra change from one to another without intermediates visible ( Figure S4), that is, the photodecomposed product of 1 rapidly takes up water vapour from air.T hisc oulds uggest that free water is presenti nt he final product matrix undergoing hydrogen bonding with Pt-OH/OH 2 .
The PCA on the full ATR-FTIR spectral region of the photodecomposition of 1 under 420 nm irradiation capturesa nd correlates the effect of the removal of n(OH Pt-OH )a t3550 cm À1 with ad ecreaseo fÀ10 cm À1 to n asym (N 3 )p rior to an increase to n asym (N 3 )o f2 1cm À1 when little-to-no n(OH Pt-OH )i sp resent. Furthermore, the appearance of the broad d(OH Pt-OH=OH 2 )b ands (1381 and 1334 cm À1 )c orrelates to the n asym (N 3 )a t2 044 cm À1 , where d(OH Pt-OH=OH 2 )i ncreases in intensity until the single n asym (N 3 )v ibration at 2044 cm À1 is completely removed as shown in Figure 2A.F urther deconstruction of the ATR-FTIR spectra and investigation of the same reactionof1 and its synthetic precursor 2 by UV/Vis and analytical HPLC were performed in order to connect these observed changes to possible formed intermediates.
Analytical HPLC revealed two possible intermediates with different retention times, which evolve at the same rate (P1 and P2) until one final product (P3) remains ( Figure S6, S7 in the SupportingI nformation). Further,t he steady state UV/Vis resultsa nd their MCR-ALS spectra of 1 and 2 suggest that both compounds undergo photodecomposition through a similar intermediate before obtaining the final photoproduct, via at wo-stepp rocess ( Figure S15, S16).
The deconstruction of the ATR-FTIR spectra by MCR-ALS, using the same kinetic model, revealed an intermediate (P1 + P2) containing as ingle n asym (N 3 )v ibration at 2042 cm À1 ,t he d(OH Pt-OH=OH 2 )b ands at 1379 cm À1 and 1334 cm À1 and all corresponding trans pyridine modes (Figure 2A,S 16), closely matching the PCA results.
This allows us to make ar easonable assignment for P1 + P2 as trans-[Pt II (N 3 )(py) 2 (H 2 O/OH)].T he MCR-ALS spectrum of P3 is ac lose match to the final time points of the photodecomposition with no presence of azido vibrationso bserved, and can be reasonablya ssigned as trans-[Pt II (py) 2 (H 2 O/OH) 2 ] ( Figure S17). Although there are somer eports of stable Pt III complexes, the general observed trend for Pt IV reduction is the two-photon reduction of Pt IV to Pt II ,w hich supports our findings. [3c, 16] The formation of an intermediate species with as ingle azido ligand, trans-[Pt(N 3 )(py) 2 (H 2 O/OH)] implies that reduction of 1 from Pt IV to Pt II occurs via release of at least one hydroxyl radical and one azidyl radical or two hydroxyl radicals. This is in line with the previously reported detection of azidyl radicals released from 1 upon irradiationb yu sing spin traps. [5b] Recombination of two hydroxyl radicals results in formation of H 2 O 2 , which furtherd ecomposes to generate O 2 .T he evolution of oxygen from the reduction of Pt IV via hydroxyl radicalr elease rathert han from the solvent, has been reported to occur for a similars tructure to 1. [17] Furthermore, the PCA of the ATR-FTIR data suggest ap ossible second intermediate product during the first few minuteso fi rradiation, in which a À10 cm À1 reduction in wavenumber for n asym (N 3 )i so bserved which correlates to the changing OH environment( removalo fn(OH Pt-OH ), Figure 3.
From the TEAS measurements and resulting global analysis, insight into the ultrafast photodynamics involved in photodecomposition of 1 is obtained.Apossible photodecomposition mechanism is shown in Figure9.H owever,f urtherw ork is re- Figure 9. Proposed mechanism for the ultrafast photochemical processes of 1 and the extracted time-constants (t n )from the TEAS studiesa sd iscussed in the main text. IRF = instrument response function. quired to validate this scheme (e.g.,t ransient vibrational spectroscopy) to assign intermediate states and high-level electronic structure calculations (both beyond the scope of the present study). The initial excitation populates an array of singlet excited states, which undergo rapid internal conversion (IC), populating the lowest 1 MLCT within our instrument response ( % 80 fs). The excess vibrational energy imparted to this state undergoes intramolecular vibrational redistribution( IVR) and occurs with at ime-constanto ft 1 ,a st he EADS (t 1 )i s evidently broader and red-shifted compared to EADS (t 2 ). This proceeds via intersystem crossing (ISC) to av ibrationally hot tripletexcited state (t 2 , 3 MLCT). This conclusion is drawn from the evident change in spectral profile between EADS (t 2 )a nd EADS (t 3 ), a sign of as tate change. The timescale for this process also sensibly compares with that reportedf or other platinum complexes. [18] The population of the vibrationally hot 3 MLCT is likely capable of accessing an additional near degenerate state 3 MC state. This near degeneracy in the two states might allow the population to flow freely between these states. Upon population of the 3 MC state, the molecule can undergo rapid loss of one or more ligands, leadingt ot he formation of the photoproduct. Our TEAS setup is blind to the formation of the photoproduct, as it absorbsb eyond of the spectral window of our white light (< 345 nm, see Figure S13). As the excessv ibrational energy is shed to the solventb ath, access to the 3 MC state is no longer available,s witching off the formation of the photoproduct, with the remaining population becomingt rapped in the vibrationally-cold 3 MLCT state. Furthermore,t he emission spectrumo f1 reveals ab and at % 550 nm in acetonitrile upon 310 nm excitation ( Figure S19). This large red-shift points to either fluorescence, following al arge geometry change in the exciteds tate or,m orel ikely, phosphorescence. Thel atters upportsthe idea of population becoming trappedi natriplets tate. This vibrationalc ooling processa nd thust he loss of access to the 3 MC state, occurs with at ime-constant of t 3 ,a st he spectral features in EADS (t 4 )a re spectrally sharper and blue-shifted compared to EADS (t 3 ). We note that our fitting procedure is unidirectional in population flow and does not include branched kinetics, therefore the fit convolutes the processes mentioned above into the time-constant t 3 .T he vibrationallycold 3 MLCT state persists beyondo ur maximum available Dt.
Prior to this work, onlyt heoretical studies had been carried out in order to investigate the ultrafastp hotodynamics of 1. Analysis of Mulliken and NBO charges of the S 0 and 3 MLCT geometries of 1 revealed al oss in negative charge from the azido ligands and positive charge of platinum, including the elongation of platinum to nitrogen (azide) bond lengths, suggesting reductive elimination of the azides. [4] Earlier computational work on as imilarc omplex, cis,trans,cis-[Pt(N 3 ) 2 (OH) 2 (NH 3 ) 2 ], presented the accessibility of several photorelease pathways via low-lying singlet states resulting in the releaseoft he azide anion and low-lyingt riplet states for which dissociation of ammonia and an azide anion was suggested withoutt he reduction of platinum. for their ground and excited state properties. [8b] The further inspection of their three lowest singlet exciteds tates using the time-dependent DFT method revealed the potentiald issociative behaviour of the two azido ligands and photoreduction of platinum, from Pt IV to Pt II .More recently,Shushakov et al. investigated the photophysical andp hotochemical processes of cis,trans,cis-[Pt(N 3 ) 2 (OH) 2 (NH 3 ) 2 ]a nd trans,trans,trans-[Pt(N 3 ) 2 (OH) 2 (NH 3 ) 2 ]o np sa nd mst imescales, capturing multistage photolysis for both complexes including initial substitution of one azide by as olventm olecule. [19] Overall,t he present findings by TEAS open up an ew path for more comprehensive ultrafastp hotodynamic and high-level theory studies of 1, which are outside the scope of the present study.
The photoinduced reaction of 1 with 5'-GMP was closely followed by ATR-FTIR as shown in Figure 6. Prior to irradiation, splitting of the n(OH) into four peaks is observed. This is indicative of increased hydrogen bondingb etween the hydroxyl groups of 1 and 5'-GMPa nd/or water.W hilst 20 min of irradiation is required to completely remove the n(OH) of 1 in water, the presence of 2mol equiv of 5'-GMP increases the rate of removal. The change in the antisymmetrica zido vibration of 1 is in line with the photodecomposition of 1 in which after 2hof irradiation, as ingle peak at 2046 cm À1 is present. Interestingly, am inor band emerges at 2164 cm À1 upon irradiation, which matches the infrared frequency of sodium azide. This suggests that the mechanismi nvolves the releaseo ft he azide anion, N 3 À from 1 in the presence of 5'-GMP. Large changes are observed to the purine-ring-related CO, CC, and CN stretching vibrations and NH/NH 2 bending modes of 5'-GMP in the 1800-1485 cm À1 ATR-FTIR regiont hroughout irradiation. Firstly,areduction in intensity of the n(CO) + n(C5C6) vibration at 1696 cm À1 is observed. Similar results were obtainedf or several M 2 + -5'-GMP complexes by Theophanides and co-workers, in which as imilarr eduction in intensity of this band is observed but only small changes to the wavenumber itself ( % 5cm À1 ). [20] This was attributed to the maintained hydrogen bondingb etween C6O of 5'-GMP and water boundt ot he metals as observed in the crystal structures. In this case ac lear shift is observed for the n(CO) + n(C5C6)t o 1670 cm À1 .T his could be due to the pyridines of 1 shielding the C6O from undergoing hydrogen bonding. The DFT optimized S 0 structures of potential photoproducts, trans-[Pt(N 3 )(py) 2 (5'-GMP)] 1À and trans-[Pt(py) 2 (5'-GMP) 2 ] 2À ,s how a 2.333 ,2 .250 and 2.293 distance betweent he C6O oxygen and ap yridine proton( Figure S20). Furthermore,t he DFT-predicted frequenciesf or this vibration shiftf rom 1705 cm À1 to 1681 cm À1 ,g oing from free 5'-GMP to 1 bound to the N7 position of 5'-GMP (trans-[Pt(N 3 )(py) 2 (5'-GMP)] 1À ) (Table S1). In the case of trans-[Pt(py) 2 (5'-GMP) 2 ] 2À ,t wo n(CO) + n(C5C6) peaks are predicted at 1687 cm À1 and 1682 cm À1 . These predicted frequenciesd ecreasing by 25/18/23 cm À1 are well matched by the experimental value of 27 cm À1 .S tronger infrared absorption is observed for the NH/NH 2 bending modesa t1 610 cm À1 ,w hich indicates the decreasei nh ydrogen bonding. Continued increase of the new n(C4C5) + n(N1C2) vibrations at 1599 cm À1 and 1570 cm À1 are further indications of binding of 1 to 5'-GMP.T he changes occurring to the breath-ing modes of 5'-GMP (from 1377/1360 cm À1 to 1358/ 1336 cm À1 )a re likely to be influenced by the binding of platinum to the N7 position of 5'-GMP,l owering the infrared wavenumber value for this vibration. This can be attributed to changes to the hydrogen bondingn etwork upon platination of the N7 position of the pyrimidine ring system. [20][21] Theearlier assigned d(OH Pt-OH=OH 2 )b ands during the photodecomposition of 1 in water at 1381 cm À1 and1 334 cm À1 are in close proximity to the breathing modes of 5'-GMP, thus the presence of water bound platinum vibrations cannot be ruled out. Interestingly, the decrease in intensity of the puriner ing breathing mode at 1178 cm À1 throughout irradiation is predictedb yt he DFT calculations, in which a % fivefold decrease in intensity is observedw hen platinum binds to N7 of 5'-GMP.P erturbation of the electronic structure upon binding to the purine ring could be the cause of the observed decrease in intensity.E arlier reported direct bindingo fp hosphate to magnesium resulted in as hifto f+ 18 cm À1 to the n sym (PO 3 2À )v ibration, whereas indirecti nteraction of the water bridging between phosphate and copper or magnesium resulted in as plit in this vibration into two peaks. [20,21] Reported platinum 5'-GMP complexes showedl ess effect on the phosphate-related stretching vibrations, in line with our results, which were reported to be primarily due to conformational changes around the phosphate and sugar moiety. [20,21] Thus, binding of the phosphate to 1 might be ruled out.I ng eneral,b roadening of vibrations at and near 3300 cm À1 and 1650cm À1 is observed throughout the photoinduced reactiono f1 and 5'-GMP,w hich indicate the presence of water in the resulting product matrix with n(OH) and d(OH)u nderlying other vibrations, respectively ( Figure 6).
The ATR-FTIR spectra of separatedp hotoproducts correlate well to the observedc hanges throughout irradiation of 1 with 5'-GMP,e specially in the 1800-1300 cm À1 region.L arger changes between the ATR-FTIR spectra of the mixture and separated photoproductsw ere seen in the phosphate region.T he separation by HPLC and lyophilisation is expected to change the chemical environmenta nd molecular geometry around PO 3 2À ,t he charge depends on pH where counter ions might be exchanged. This could account for the observed increasei n wavenumber values where fewers hort distance interactions are present compared to the reactionm ixture containing 2mol equiv of the disodium 5'-GMP salt. Photoproduct 1b decomposed after lyophilisation according to MS and no ATR-FTIR spectra could be obtained. HRMS results prior to lyophilisation match aP t IV species (trans-[Pt IV (N 3 )(OH)(py) 2 (5'-GMP-2Na)] + H + ). Initial studies of 1 revealed the formation of am inor Pt IV species upon irradiation in presence of 5'-GMP. [4] Further studies of derivatives of 1 revealed the oxidizing capabilities of such complexes,w here oxidation on the C8 positiono f5 '-GMP was postulated. [17] The m/z of aP t II species bound to an oxidized 5'-GMP (trans-[Pt II (N 3 )(py) 2 (5'-GMP-2Na-H + OH)] 1À + 2H + )i si dentical to the proposed Pt IV species and therefore cannot be ruled out.
The evolution of species 1a to 1e in the dark after 30 min of irradiation captured changes to absorption intensity of each photoproduct during the course of two weeks. The decrease in absorption intensity of 1b could be due to the reduction of the proposedP t IV species to yield the Pt II species 1c, which increasesina bsorption intensity over time.

Conclusions
This work has demonstrated thepotential for ATR-FTIR to elucidate the photodecompositiono ft he photoactivatable diazido Pt IV anticancer prodrug candidate (1)i ncluding the photoinduced reactiono f1 with 5'-GMP.P rincipal Component Analysis (PCA) captured and correlated the observed changes of the disappearance of n(OH Pt-OH )a t3 550 cm À1 with ad ecrease of À10 cm À1 to n asym (N 3 )p rior an increase to n asym (N 3 )o f2 1cm À1 when little to no n(OH Pt-OH )ispresent. Furthermore, the appearance of an ew broad d(OH Pt-OH=OH 2 )b and (1381 cm À1 and 1334 cm À1 )c orrelates to the single n asym (N 3 )v ibration at 2044 cm À1 ,w here d(OH Pt-OH=OH 2 )g radually increased until the single n asym (N 3 )a t2 044 cm À1 completelyd isappeared. Transient Electronic Absorption Spectroscopy (TEAS) provide ap robable vibrational switching mechanism for 1 where the excited state complex has access to the dissociative 3 MC state only when it is vibrationally hot, after which it remains trapped in av ibrationallyc ool state with al ong-lived absorption (@ 2ns). This paves the way for future comprehensive ultrafast photodynamic (transient vibrational absorption spectroscopy) and high-level theory studies of 1,w hich were outside the scope of this study.
The ATR-FTIR of the photoinduced reaction of 1 with 5'-GMP exposed the faster removal of n(OH)c ompared to the photodecomposition ( 5min vs. 20 min), whilst showingi ncreased hydrogen bonding between the hydroxyl groups of 1 and 5'-GMP and/or water prior to irradiation. Changes to the antisymmetrica zido vibration are in line with the photodecomposition resultsw ith only as ingle vibrationp resent after two hours of irradiation and additionally an ew vibration matching sodium azide, that is, releaseo fa na zide anion.C omprehensiveA TR-FTIR assignmentso f5 '-GMP and photoproducts formed upon irradiation of 1 and 5'-GMP were achieved by DFT calculations. Separation of photoproducts by HPLC, followed by MS and ATR-FTIR allowed ac omprehensive ATR-FTIR assignment. These furthera ided the detailed elucidation of the observed changes in the ATR-FTIR of the reaction mixturet hroughout irradiation. For instance, trans-[Pt(N 3 )(py) 2 (5'-GMP)] is the major product in the reaction mixture, however am arker band for the trans-[Pt(py) 2 (5'-GMP) 2 ]p roduct at 1772 cm À1 allows the trackingo f this product.
These studies show that vibrational spectroscopy can make am ajor contribution to understanding the chemical basis for the mechanism of action of photoactivatable diazidoP t IV prodrugs including their interactions with DNA bases.

Experimental Section Materials
All reactions were performed under nitrogen atmospheres using standard Schlenk techniques. K 2 PtCl 4 (99 %) was purchased from Precious Metals Online. Guanosine 5'-monophosphate disodium salt hydrate ! 99 %( 5 '-GMP) and all other chemicals and solvents were purchased from Sigma-Aldrich and used as received.

Instrumentation and methods
UV/Vis spectra were acquired on aV arian Cary 300 BIO UV/Vis spectrophotometer equipped with aV arian Cary temperature controller (298 K, unless otherwise stated). LCMS spectra were acquired with an Agilent 6100 Series Single Quadrupole LC/MS incorporating ap hotodiode array detector (214/254 nm) coupled directly to an electrospray ionization source.
HRMS was conducted using an Agilent 6224 TOFL C/MS Mass Spectrometer coupled to an Agilent 1290 Infinity.A ll data were acquired and reference-mass corrected via adual-ESI source.
Fluorescence spectra were recorded on aH oriba Fluorolog 3s pectrophotometer using an excitation wavelength of 310 nm with a bandwidth of 8nm. The data obtained for 1 were only usable between 363 and 580 nm due to scatter and 2nd order effects.

Transient electronicabsorption spectroscopy (TEAS)
The experimental procedures pertaining to TEAS have been described in detail elsewhere and ab rief overview is presented herein. [22] The femtosecond pump pulses were generated using a commercially available optical parametric amplifier,( TOPAS-C, Spectra-Physics). The white light continuum (330-690 nm) utilized as the probe pulse was produced through supercontinuum generation from the 800 nm fundamental in a2mm thick CaF 2 window; translated vertically.T he pump pulse wavelength was set to 310 nm (4.00 eV). The fluence of the pump beam was set between 1-2 mJ cm À2 .T he difference between the pump and probe polarizations was held at magic angle (54.78)t on egate rotational effects using ah alf-wave plate in the probe beam path. The pump-probe time delay (Dt)w as varied by adjusting the optical delay of the probe pulse, the maximum obtainable Dt was 2ns. Changes in the optical density (DOD) of the samples were calculated from probe intensities, collected using as pectrometer (Avantes, AvaSpec-ULS1650F). Samples of 1 were made to ac oncentration of 2mm in acetonitrile (99.9 %, VWR). The delivery system for the samples was af low-through cell (Demountable Liquid Cell by Harrick Scientific Products Inc.) with a1 00 mmp ath. The sample was circulated using ap eristaltic pump (Masterflex) recirculating sample from a reservoir to provide each pulse with fresh sample.

Theoretical calculations
DFT calculations were carried out using the Gaussian 9p ackage [23] with the hybrid PBE0 functional with 25 %H Fe xchange. [24] cc-pVDZ was used for carbon, hydrogen, nitrogen, oxygen and phosphorus, whereas for platinum an augmented cc-pVDZ-PP with effective core potential (ECP) was used. [25] The conductor-like polarizable continuum model (CPCM) was applied (water) to better describe the electrostatic interactions between metal to ligand bonds. [26] Calculations converged to optimised geometries by allowing all parameters to relax, corresponding to true energy minima as confirmed by the lack of imaginary frequencies. The ground-state geometries reported previously for 1 were used as a starting guess and 5'-GMP starting geometries were obtained using universal force field optimizations, prior to af ull conformational screening by systematic variation of all the dihedral angles until corresponding minima were attained. [6] Sodium ions did not influence the geometry optimization of 5'-GMP and were therefore omitted. Starting geometry estimates for photoproducts of 1 and 5'-GMP were based on the S 0 geometries of each, respectively, with conformational screening carried out by systematically rotating along the platinum-N7 axis and varying the dihedral angles between platinum and the N7 position of 5'-GMP until consequent minima were reached. TwoS 0 geometries for trans-[Pt(py) 2 (5'-GMP) 2 ] 2À and one for trans-[Pt(N 3 )(py) 2 (5'-GMP)] 1À were obtained ( Figure S20). Calculated infrared spectra were exported from Gauss-View 5w ith ahalf-width of 16 cm À1 .

Data processing
PLS_Toolbox 8.2 (Eigenvector Research) was used to carry out the Principal Component Analysis (PCA). ATR-FTIR spectra of the photodecomposition of 1 were cut (3700-640 cm À1 ), Standard Normal Variate (SNV) scaled, mean centred and cross-validated (venetian blinds, 10 splits and 1s ample per split). Three principal components (PC) were selected, quantifying the majority of changes in the spectra over time.

Photodecomposition of 1a nd 2i nacetonitrile
The photodecomposition of 1 and 2 in acetonitrile (50 mm)u nder 310 nm irradiation monitored by UV/Vis (3 mL) in 1cmp athlength quartz cuvettes was carried out in triplicates until completion.
Photoinduced reaction of 1w ith 5'-GMP Solutions (2 mL) of 1 (2 mm)a nd 5'-GMP (4 mm)i nM ili-Q water were irradiated a1cm pathlength quartz fluorescence cuvette for 2h.5 0mLa liquots were taken from the solution during irradiation at selected time points (0, 5, 10, 15, 30, 45, 60, 90 and 120 min). Samples were protected from light and stored on ice prior to measurement by ATR-FTIR. The photoinduced reaction was carried out in triplicate, with each time point measured at least twice by ATR-FTIR. A2mL solution of 5'-GMP (4 mm)w as exposed to 24 ho fi rradiation and measured by ATR-FTIR. The stability of 1 (2 mm)a nd 5'-GMP (2 mol equiv) was confirmed by ATR-FTIR and HPLC for 35 days ( Figure S22 and S23) and 22 ho fi rradiation of 5'-GMP (4 mm) showed no changes by ATR-FTIR ( Figure S21). Separation of photoproducts formed by irradiation of 1w ith 5'-GMP Solutions (3 mL) of 1 (3.58-3.86 mm)a nd 5'-GMP (1:2) in Mili-Q water were irradiated for one hour in a1cm pathlength quartz fluorescence cuvette prior to separation by semi-preparative HPLC. Fractions were lyophilised and protected from exposure to light. Analytical HPLC was carried out before and after irradiation, after separation and after lyophilisation. Pure fractions (> 95 %according to HPLC) were stored in the freezer and protected from light. Samples were dissolved in Mili-Q water prior to ATR-FTIR, LCMS and HRMS analysis.
Evolution of photoproducts after irradiation of 1w ith 5'-GMP A1 .5 mL solution of 1 and 5'-GMP (50 mm 1,1 00 mm 5'-GMP) in a glass LCMS vial was irradiated for 30 min. Analytical HPLC measurements were taken before irradiation and after irradiation, t: 0, 1, 2, 3, 4, 5, 6, 9, 10, 12 and 14 days. Analytical HPLC measurements were carried out in triplicate and peaks were auto integrated (slope sensitivity:5mAU s À1 ,p eak width:0 .02 min, area reject:5 mAU*s, height reject:2mAU, shoulders:o ff).A nalytical HPLC samples were directly taken from the vial and punctured stoppers were replaced after each run. The vial was stored at RT and covered from light in between measurements. (1) and trans,trans,trans-[Pt(N 3 ) 2 (py) 2 ]( 2)w ere synthesized as reported previously [6] with characterization data in agreement with those reported originally. [4] Caution!H eavy metal azides are known shock sensitive detonators. Whilst no problems were encountered during this work, it is crucial not to apply excessive pressure to platinum azido compounds. Exposure of 1 and 2 to light was strictly limited during synthesis, sample preparation and measurements.