Photo-physical characterization of high triplet yield brominated fluoresceins by transient state (TRAST) spectroscopy

Photo-induced dark transient states of fluorophores can pose a problem in fluorescence spectroscopy. However, their typically long lifetimes also make them highly environment sensitive, suggesting fluorophores with prominent dark-state formation yields to be used as microenvironmental sensors in bio-molecular spectroscopy and imaging. In this work, we analyzed the singlet–triplet transitions of fluorescein and three synthesized carboxy-fluorescein derivatives, with one, two or four bromines linked to the anthracence backbone. Using transient state (TRAST) spectroscopy, we found a prominent internal heavy atom (IHA) enhancement of the intersystem crossing (ISC) rates upon bromination, inferred by density functional theory calculations to take place via a higher triplet state, followed by relaxation to the lowest triplet state. A corresponding external heavy atom (EHA) enhancement was found upon adding potassium iodide (KI). Notably, increased KI concentrations still resulted in lowered triplet state buildup in the brominated fluorophores, due to relatively lower enhancements in ISC, than in the triplet decay. Together with an antioxidative effect on the fluorophores, adding KI thus generated a fluorescence enhancement of the brominated fluorophores. By TRAST measurements, analyzing the average fluorescence intensity of fluorescent molecules subject to a systematically varied excitation modulation, dark state transitions within very high triplet yield (>90%) fluorophores can be directly analyzed under biologically relevant conditions. These measurements, not possible by other techniques such as fluorescence correlation spectroscopy, opens for bio-sensing applications based on high triplet yield fluorophores, and for characterization of high triplet yield photodynamic therapy agents, and how they are influenced by IHA and EHA effects.


Introduction
Photo-induced dark transient states of fluorophores have been extensively studied for the limitations they set for single molecule fluorescence applications, as well as for their role in enhancing resolution in fluorescence-based super-resolution microscopy [1][2][3].Moreover, population dynamics of long-lived, non-fluorescent or weakly fluorescent transient states of fluorophores arising from intersystem crossing [4], trans-cis isomerization [5] or photo-induced charge transfer processes [6] are often highly environment sensitive.This opens for using these states as microenvironmental read-out parameters in bio-molecular research, and to use fluorophores with prominent formation yields of such states.Fluorescent probes with high intersystem crossing (ISC) enable efficient transfer to their lowest triplet state, with possible long lifetime phosphorescence emission [7].Such fluorophores have found use in several biomedical applications, including photodynamic therapy (PDT) [8,9], as well as sensing [10] and imaging [11] based on phosphorescence emission.Compared to fluorescence, the phosphorescence signal is however typically weak and strongly compromised by oxygen quenching.ISC can be enhanced by covalent addition of heavy atoms, such as bromine or iodide, yielding an internal heavy atom (IHA) effect by spin-orbit coupling [12,13].ISC enhancement can also be effectuated by collisional encounters between heavy atoms and the fluorophores, an external heavy atom (EHA) effect, which is commonly used in fluorescence spectroscopy to get insights about molecular interactions [13].IHA and EHA effects have been largely studied by transient absorption spectroscopy (or flash photolysis, FP) [14,15], with transient states monitored via their absorption by a separate probing beam, following an excitation pulse.However, FP is technically relatively complicated, lacks the sensitivity for measurements at low (<μM) concentrations, and absorption spectra of transient states can overlap and thus be difficult to spectrally separate from each other.Moreover, while the decay of the transient state absorption can be followed, it is more complicated to determine the absolute rates of formation of the transient states.
Transient dark state transitions in fluorophores can also be followed via fluorescence-based single molecule detection (SMD) methods, in which fluorophore blinking behavior can be monitored on the fly, without need of synchronization.However, SMD requires high brightness fluorophores, which typically makes fluorophores with prominent triplet quantum yields too dim to be properly analyzed.Moreover, their singlet-triplet state transitions typically take place at sub-microsecond time scales.This is typically too fast and generates too few photons within the event to be properly analyzed on a single molecule level.Compared to SMD methods, fluorescence correlation spectroscopy (FCS), analyzing fluorescence intensity fluctuations of larger numbers of single fluorophores on a cumulative, ensemble level, offers extended possibilities to analyze transient dark states of fluorophores, such as singlet-triplet [16], isomerization [17,18] and photo-induced charge transfers [19].On these premises, FCS was used to investigate EHA effects of potassium iodide (KI) under typical excitation conditions for SMD, and where it was found that KI can function both as a fluorescence quencher and promoter [3,20].However, as a method relying on SMD conditions, also FCS measurements ultimately depend on the molecular brightness of fluorophores, and the higher their triplet quantum yield, the lower their fluorescence brightness, and the more challenging it is to analyze their singlet-triplet transitions.Increasing the excitation intensity can lead to higher fluorophore brightness, but fluorescence saturation due to triplet state buildup sets a limit for this gain.With the relaxation time for this buildup then also moved into shorter time ranges, it may even overlap with the antibunching relaxation of the fluorophores [21], and signal-to-noise conditions in the recorded FCS curves are further compromised [22].Moreover, the contribution of a fluorescent species to a recorded FCS curve scales with its molecular brightness squared.With a typically non-uniform excitation intensity within the detection volume used in FCS experiments, the contribution from different parts of the detection volume can strongly vary depending on the degree of dark state build-up/saturation [16], which further complicates analyses of high triplet yield fluorophores.Thus, both SMD methods and FCS possess limitations in resolving the higher ISC rates of heavy atom-containing fluorophores.Transient state (TRAST) monitoring offers a means to circumvent many of these limitations and allows transient state dynamics of fluorophores to be followed in a widely applicable manner [3,23].
TRAST analyses the time-averaged fluorescence intensity, á ñ F w , exc ( ) detected from fluorophores sub- ject to a modulated excitation.With the modulation systematically varied over the time range of the dark state transitions of the fluorophores, these transitions can be determined form how á ñ F w exc ( ) varies with the modulation characteristics [6,[24][25][26][27].The TRAST method does not rely on SMD conditions nor on a high time resolution in the detection, as needed for FCS.This makes it possible to extend dark state transition studies in fluorophores to more demanding samples with limited signal-to-background conditions, to fluorophores with lower brightness, including autofluorescent compounds [27][28][29], and allow investigations under far lower excitation intensities than used in FCS experiments.Heavy atom substitution by e.g.bromines offers a means to enhance triplet state formation by an IHA effect.Thereby, prominent triplet state populations can be reached also under moderate excitation intensities, which offers an advantage in live cell TRAST and phosphorescence studies.Yet, brominated fluoresceins, such as e.g.Eosin, are also broadly used as fluorophores in cellular imaging, in which case triplet state formation rather constitutes a limiting factor.In this work, we applied wide-field TRAST spectroscopy to study the overall effects of heavy ions on the photo-physics of carboxy-Fluorescein (CFl) and three synthesized CFl derivatives, with different numbers of bromine atoms.We investigated how triplet state transition and photo-oxidation rate parameters varied with the number of bromines attached on the common CFl body, and how EHA effects mediated by potassium iodide (KI) added into the fluorophore solutions were manifested.We found that the intrinsic ISC rate dramatically increased by the number of covalently added bromines, and with increasing KI concentrations.Interestingly, at the same time addition of KI leads to an overall reduction in the triplet state populations of the brominated fluorophores.By enhancing not only the ISC rates, but also the triplet decay rates, iodine ions can act as a net fluorescence promoter for these fluorophores, as opposed to bromine-free CFl.The TRAST experiments were complemented with density functional theory (DFT) calculations, suggesting that for the observed, high ISC rates to be generated in the fluorophores, ISC may have to take place to a higher triplet state, from which relaxation to the lowest triplet state then takes place.Fluorescein fluorophores, and brominated variants such as Eosin, are broadly used as labels of molecules, cells and tissues.Apart from a typical pH sensitivity, rendering them suitable as pH sensors [13], modified fluoresceins have also found use for lifetime-based sensing [30].In lifetime-based sensing, quenching needs to be effectuated in the time range (ns) of the excited singlet state (S 1 ) lifetime to be properly detected [13].If instead the quenching of more long-lived photoinduced states, such as triplet (T) and photooxidized ( + R  ) states, can be monitored then also more rare and lower frequency interactions can be detected [6,31].Taken together, the TRAST experiments show that dark state transitions within high triplet yield fluorophores can be readily analyzed, and that IHA and EHA effects can be followed, also in e.g.biological samples not amenable to SMD-, FCS-or FP-based analyses.By use of high-triplet yield fluorophores, prominent triplet state populations can be generated also at moderate excitation intensities, minimizing possible photo-toxic effects.This offers a promising basis for bio-imaging of high triplet yield fluorophores, as well as for characterization of high triplet yield PDT agents, for local monitoring of their therapeutic effects via their ISC rates.

Materials and methods
2.1.Fluorophore synthesis and sample preparation Four carboxy-Fluoresceins were synthesized without bromination, or with a mono-, di-and tetra-bromination (figures 1(a)-(d)).The synthesis of CFl and the corresponding brominated fluorophores (CFl-Br, CFl-2Br and CFl-4Br) and the subsequent verification by nucleic magnetic resonance (NMR) and mass spectrometry (MS) are further described in the Supplementary part, sections S4 and S5.Fluorescein-5isothiocyanate (FITC, Thermo Fisher Invitrogen, Cat.No: 143)) was used without further purification.Aliquoted fluorophore stock solutions (100 μM) in phosphate-buffered saline (PBS, pH = 7.4) were prepared from powder samples, stored in freezer at −20 °C, then thawed and diluted in PBS to a final concentration of 100 nM before experiments.Potassium iodide (KI, Sigma Aldrich) was added in concentrations up to 50 mM in the diluted samples.The pH of all measured samples were set at 9.3, to eliminate any pH dependent effects and to keep the fluorophores in a fluorescent deprotonated state (pK a of ∼6.5 [32]).

TRAST spectroscopy 2.2.1. Theoretical concept and photophysical model
In TRAST spectroscopy, photo-induced blinking kinetics of fluorescent molecules are studied via the time-averaged fluorescence intensity from samples under a modulated excitation scheme, varied on the time scales of the dark state transitions underlying the blinking [3,25].Blinking kinetics occuring in a m s-ms time range can then be quantified without the need for time-resolved detection.To calculate the fluorescence intensity in the TRAST experiments, we used a common photo-physical model for all fluorophores, including a ground singlet state, S , 0 an emissive, excited singlet state, S , 1 a dark triplet state, T, and a dark photo-oxidized radical state, + R ,  (figure 2).This model corresponds to previously used models for similar xanthene fluorophores [20,24], but due to the lower excitation rates applied in this work, population of higher excited singlet and triplet states can be disregarded.For a fluorescent molecule subject to a rectangular excitation pulse initiated at = t 0, the resulting detected fluorescence intensity can then be given by where c is the fluorophore concentration, q f is the fluorescence quantum yield, q D denotes the overall detection quantum yield of the microscope, CEF r (¯) represents the collection efficiency function and S r t , (¯) the probability that a fluorophore, residing at position r ¯in the detection volume, is in either its ground (S 0 ) or excited / denotes the decay rate within the fluorophores from S 1 to S 0 , represents the excitation rate from S 0 to S , 1 where s exc is the excitation cross section of the fluorophore, F r exc (¯) the local excitation flux, I r exc (¯) the excitation intensity and hv is the excitation photon energy [27].With a constant excitation starting at t = 0, the probability for a fluorophore to be in a singlet state (S 1 or S 0 ) at time t can in a general form be described by: k 10 representing the excitation and de-excitation rates, respectively.k isc denotes the intersystem crossing rate from S 1 to the triplet state, T, and k T the triplet decay rate back to S 0 .Photo-oxidation of fluorophores to + R  state is assumed to occur from both S 1 and T, with a rate denoted k .
ox From the long-lived + R  state, the fluorophores can relax back to S 0 through a photo-reduction rate, k .red (b) Simplified electronic three-state model, describing state transitions taking place on a time scale much longer than the equilibration of the S 0 and S 1 states, where S represents both these singlet states and Since the singlet-triplet state transitions are much faster, the S and T populations are equilibrated on the timescale of the transitions to and from + R .
 We can then for simplicity assume photo-oxidation to take place only from T.
Here, λ i are the eigenvalues, i.e. the rates of relaxation modes of S(t) upon onset of constant excitation, and A i the related amplitudes, reflecting the population build-up of the different photo-induced non-fluorescent states at steady-state With a suitable initial condition, typically S(0 ) = 1, λ i and A i can be described analytically.At onset of excitation, equilibration between the S 1 and S 0 states takes place within the fluorescence lifetime (ns), with the so-called antibunching relaxation time ) [21].This relaxation is typically averaged out on the time scales of the dark state transitions monitored by TRAST, and we can thus reduce our photophysical model of figure 2 R  states, and with characteristic relaxations in timescales of m s to ms of S t ( ) and F t , ( ) attributed to build-up of T and + R  state populations (corresp- onding rate equations and their analytical solution are given in section S1).This relaxation process is also reflected in the time-averaged fluorescence signal generated by a rectangular pulse with a duration of w: where the change of á ñ F w exc ( ) with w allows the population kinetics of long-lived, dark transient states, such as T and + R ,  to be determined, which is the general basis for TRAST method.So-called TRAST curves are then generated by collecting á ñ F w exc ( ) over multiple pulses, M, for different pulse durations, w, normalized by á ñ F w exc ( ) recorded with a pulse train with short pulse duration, w : Here, w 0 is selected to be short enough to avoid the build-up of dark transient states yet longer than the anti-bunching relaxation time ( t w ab 0  ).By this normalization, the terms, c, q D and q f (equation ( 1)) cancel out.Moreover, at low duty cycles of the excitation pulse trains, allowing the dark transient states of the fluorescent molecules to completely relax back to the singlet ground state, S , 0 before the onset of next excitation pulse, all generated pulses in the pulse train can be treated as identical.We can then write:
The beam was then reflected by a dichroic mirror (FF506-Di03, Semrock) and focused to the back aperture of a water-immersion microscope objective (Olympus, UPLSAPO 60x/NA1.20) to generate a wide-field illumination with a beam waist of ω 0 = 18 μm (1/e 2 radius) in a sample plane.The fluorescence signal from the sample was collected by the same objective, transmitted through the dichroic mirror and an emission filter (BrightLine 530/55, Semrock), and was then detected by an sCMOS camera (Hamamatsu, ORCA-Flash4.0v2).The duty cycle of the excitation pulse trains was kept low, h = 0.01, to allow the fluorophores in the sample to fully recover back to S 0 before the onset of the next pulse.To obtain sufficient photon counts, even for short w, the number of identical pulse repetitions in each excitation pulse train, M (equation ( 4)), was adjusted to maintain a constant laser illumination time, = ⋅ t M w, ill c for all w, and with t ill kept at 10 ms.The AOM operation and the TRAST measurements were controlled by a custom software developed in Matlab environment and by use of a digital I/O card (from National Instruments, PCI-6602) [26].

Wide-field TRAST data analysis
The experimental TRAST curves (equation ( 4)) were obtained from a stack of 30 fluorescence images, where each image was recorded over an entire excitation pulse train with a certain w.Pulse trains with different pulse durations were applied, with w distributed logarithmically between 100 ns and 1 ms, and measured in a randomized order to avoid bias due to time effects.An additional 10 reference frames, all using 100 ns pulse duration to avoid dark state build-up, were inserted at regular intervals between the 30 main images to track any permanent bleaching of the sample.The overall observed bleaching was typically 5%-10% of the total detected intensity in the measurements.Prior to analysis, the recorded TRAST data was first pre-processed by subtraction of the static ambient background.In all measurements, TRAST curves were produced by recording á ñ F w exc norm ( ) within a region of interest (ROI) corresponding to a 13 mm radius in the sample plane, centered on the excitation beam.Fitting of photophysical rate parameters was then performed following the same general procedure as previously described [27], here specifically simulating theoretical TRAST curves using equations (1)-( 5) and based on the photophysical model of figure 2(b), and then comparing these curves to the experimental data.The set of rate parameter values best describing the experimental data was then found using non-linear least squares optimization.In the fitting of the TRAST curves, the excited state lifetime for each fluorophore, t , f was fixed to a value determined by time-correlated single photon counting (TCSPC) measurements (described in section 2.3 below), and with t 1 f / comprising all deactivation (including ISC) rates from S 1 .Moreover, an average singlet excitation rate, k , 01 ŵas calculated for each ROI using equation (S5) (see [27] or Supplementary section S2 for details).

Fluorescence lifetime measurements
TCSPC measurements were performed with a pulsed excitation beam of a diode laser (485 nm, Picoquant GmbH, LDH-D-C-485) fed into an inverted, epiilluminated confocal microscope (Olympus FV1200) with a water immersion objective (60x, NA1.2, Olympus, UPlanSApo).The fluorescence was collected by the same objective, passed through a dichroic mirror (ZT405/473-491/NIRrpc-UF2, Chroma) and an emission filter (HQ535/70, Chroma) and then recorded by single photon counting avalanche photodiodes (Perkin & Elmer, SPCM-AQR-14).The signals were fed into a data acquisition card (Hydraharp, Picoquant GmbH), deconvoluted and then fit to an exponential decay based on non-linear least squares minimization (Symphotime, Picoquant GmbH).Deconvolution was based on an instrument response function (IRFs) determined from the back-reflected light from the laser excitation pulses.

Absorption and fluorescence emission spectra measurements
Absorption spectra of the fluorophores (1 μM, PBS, pH = 9.3) were recorded in cuvettes by a spectrophotometer (UV5, Mettler Toledo), corrected for the spectral response of the instrument, and with background contributions subtracted using blank PBS solutions.Emission spectra were recorded by a spectrofluorometer (FluoroMax-3, HORIBA), using an excitation wavelength of 488 nm and a 1 nm bandpass slit.

Density functional theory (DFT) calculations
To further investigate the IHA effects for the different fluorophores (CFl, CFl-Br, CFl-2Br and CFl-4Br), quantum-chemical estimations of the main photophysical rate parameters (the radiative rate from the S 1 state, k r , and the ISC rate, k isc ) were performed at different levels of density functional theory (DFT), and employing a polarizable continuum model (PCM) to take solvent effects into account.Computational details for these quantum chemical estimations are given in the Supplementary part S3.

Fluorescence lifetime and spectrometer measurements
CFl and mono-, di-and tetra-brominated forms of CFl (CFl-Br, CFl-2Br and CFl-4Br) were synthesized and characterized by NMR and MS (described in Supplementary, section S4), dissolved in PBS (pH 9.3, 100 nM), and then characterized by TCSPC and spectrometer measurements, as described in Methods and Materials.Both the absorption and fluorescence emission spectra displayed a clear red shift with higher degree of bromination of the fluorophores (figures 4(a) and (b)).Such red shift has been previously reported and has been explained by quantum chemical calculations to be due to charge redistributions upon halogenation [34] and can also be inferred from our DFT calculations (see section 3.3 below).The fluorescence intensities for the different fluorophores were also found to decrease with higher degree of bromination (figure 4(c), table 1).Shorter fluorescence lifetimes were also recorded with higher degree of bromination of the fluorophores (figure 4(d)).This together indicates that

Transient state (TRAST) spectroscopy measurements
Wide-field TRAST measurements were performed on the fluorophores (CFl, CFl-Br, CFl-2Br and CFl-4Br, 100 nM) in PBS (pH 9.3), as described in Materials and Methods.In the recorded TRAST curves, a prominent relaxation was observed in the microsecond time range, which increased amplitude with higher degree of bromination of the fluorophores (figure 5(a)).This is consistent with triplet state formation and an IHA effect, promoting the k isc rates of the fluorophores.This promotion was found to be significant also for the mono-brominated fluorophore (CFl-Br).On a slower, sub-millisecond, time scale a second relaxation process could be observed in the TRAST curves, most prominent for the non-brominated fluorophore (CFl).We attribute this relaxation to photo-oxidation, as has been previously observed for e.g.rhodamine fluorophores in FCS [20] and TRAST [37] experiments, and consistent with the photophysical model of figure 2.
We then recorded TRAST curves from CFl at different excitation intensities (figure 5(b)), which were globally fitted to the model of figure 2 (procedure described in Methods and Materials).In the fit, s exc and k 10 were fixed to ⋅ - Next, we studied effects upon adding potassium iodide (KI), known to act as a fluorescence quencher via an EHA effect, by dynamic quenching upon collisional encounters of the I − ions with the fluorophores [13].Depending on the conditions, however, iodide can act both as a fluorescence quencher and a promoter.In FCS experiments on rhodamine fluorophores with similar excitation and emission spectra as the fluorophores studied here, KI has been found not only to enhance k isc by an EHA effect, but also to enhance their k T rates by a charge-coupled deactivation, as well as to act as an antioxidant, promoting the recovery of photo-oxidized fluorophores [20].However, in contrast to the trend observed for the rhodamine fluorophores in [20] (Rhodamine green, RhGr), the triplet relaxation amplitude for CFl was found to increase with higher KI concentrations, [KI] (figure 4(c)).For the brominated, CFl-Br, CFl-2Br and CFl-4Br fluorophores, however, a reduction in the triplet state amplitude (or build-up) was observed with higher [KI] (figures 5(d)-(f)).The recorded TRAST curves in figures 5(c)-(f) were fitted in a similar way as the curves in figure 5(b), with s exc and k 10 fixed to the values determined from the spectrofluorometer and TCSPC measurements (table 1) and using the same model (figure 2).In the fits, we also introduced a linear dependence on [KI] for the k isc , k T and k red rates, as found previously for rhodamine fluorophores [20]: In the fits the rates at [KI] = 0 mM, k , KI has previously been found to generate two major effects on the triplet state kinetics of xanthene fluorophores [20].On the one hand, the k isc rates are significantly enhanced by an EHA effect.This EHA effect, given by k Qisc in equation (6A), was found to be quite similar for CFl, CFl-Br, CFl-2Br and CFl-4Br, compared to previously reported k Qisc [20].Addition- ally, for fluorophores with an excitation maximum in the blue spectral range, a KI-mediated charge-coupled deactivation of their triplet states can also take place, indicating that the triplet state energies of these fluorophores lie above a threshold level, above which deactivation by an electron exchange reaction with KI is possible [20].Consequently, addition of KI can have opposite effects on different fluorophores, on the one hand strongly promoting triplet state formation in xanthene fluorophores in the green-red spectral range, while even diminishing triplet state populations in corresponding fluorophores in the blue spectral range.
CFl deviates from this 'spectral rule', showing an increased triplet state population with higher [KI] (figure 5(c)).A likely reason is that CFl is negatively charged, making close collisional encounters and triplet state deactivation by charge transfer between CFl and I − less probable than in zwitterionic or positively charged (RhGr) fluorophores.Indeed, the k QT quenching coefficient for CFl, CFl-Br, CFl-2Br and CFl-4Br was found to be about two orders of magnitude lower than for RhGr [20], although the fluorophores are in the same blue spectral range.Similarly, k Qred was found to be orders of magnitude lower for CFl, CFl-Br, CFl-2Br and CFl-4Br than in RhGr (table 2), which reflects that these, all negatively charged fluorophores are far less likely than RhGr to accept an electron from I − upon collisional encounter.In contrast to CFl, the brominated fluorophores showed a significant decrease in their triplet state populations with increasing [KI] (figures 5(d)-(f)), although very similar k Qisc and k QT coefficients were found for these fluorophores.This is, however, to be expected since the measured triplet state populations depend on the ratio k isc /k .
T For the brominated fluorophores, but not for CFl, the relative increase of k isc is lower than the relative increase of k , T and the k isc /k T ratio then decreases with higher [KI].For reference, we also studied the photodynamics of Fluorescein 5-Isothiocyanate (FITC) and its dependence on F exc Table 2. Globally fitted rates from the TRAST curves shown in figures 5(c)-(f).The uncertainties refer to the 95% confidence intervals obtained from the fitting of the TRAST curves.
and [KI] under the same conditions as for the other fluorophores (experimental and fitted TRAST curves presented in SI, section S6, figure S1).As for CFl, we also observed an increase in the triplet state buildup for FITC with higher [KI], which can be explained by the negative charge of FITC (di-anion at pH 9.3).Also, the singlet-triplet and redox kinetics were very similar to those of CFl (table 2), which indicates that alterations outside of the anthracene body of the fluorophore have less effects on these kinetics.Irrespective of which of the methodologies was employed, the fluorescence redshift with higher degree of bromination (figures 4(a)-(c)) was well reproduced in the calculations, together with a slight decrease of the S 1 -S 0 oscillator strength for CFl-4Br, further contributing to the lower fluorescence intensities observed for this fluorophore.However, using either of the range-separated wB97X functional with COSMO model or the B3LYP functional with the SM12 model expectedly overestimated the energy of the S 1 state, E(S 1 ), compared to the results of the B3LYP functional with the COSMO model, which perfectly matched with the experimental data (table 1).At the same time, all methods predicted the position of the T 1 energy level near 1.7 eV (with some exceptions for 4Br, table 3).This, together with the small spinorbit coupling matrix elements (SOCMEs), S H T, | ˆ| resulted in very small ISC rates, at a level of 10 2 -10 4 s −1 .This is three to six orders of magnitude smaller than what was observed in the experiments (figure 5).Moreover, the TDA/B3LYP/TZP (COSMO) methodology couldn´t reproduce the IHA trend for k isc (for CFl-2Br it is smaller than for  | ˆ| reaches 1.92 cm −1 for the CFl-4Br tri-anion, which due to the large S 1 -T 1 energy gap, however, still does not significantly improve the matching with the experimental data (k isc (S 1 -T 1 ) theor = 5.5 × 10 5 s −1 versus k isc (S 1 -T 1 ) exp = 8.4 × 10 8 s −1 for CFl-4Br).
As a possible explanation to the experimental trends in k isc one may then consider the second lowest triplet state of the fluorophores, T 2 , as a possible acceptor state for S 1 deactivation.Both the wB97X and B3LYP functionals with the COSMO solvent model predict the T 2 state to be higher in energy than the S 1 state.This would make S 1 -T 2 ISC thermodynamically forbidden.However, TDA/B3LYP/TZP calculations based on electron density distributions within the SM12 model predict a stabilization of the T 2 state that makes it lower in energy than the S 1 state (table 3).This suggests that an additional S 1 -T 2 channel for the ISC can be considered.By taking this S 1 -T 2 ISC channel into account, we obtain k isc rates which are in reasonable agreement with the experimental results, with deviations between the calculated and experimental data not exceeding one order of magnitude (table 4).The reason for the difference between the COSMO and the SM12 solvent models in this case might originate from the better parametrization of density for the description of multicharged anions by the SM12 model.By orbital nature, the T 2 state corresponds to a ππ * symmetry, as well as the T 1 and S 1 states.
Thus, the IHA effect, increasing from CFl, over CFl-1Br, CFl-2Br to CFl-4Br, and as directly observed via the measured ISC rates (figure 5), can be correctly reproduced by accounting for the additional channel of S 1 -T 2 ISC.In contrast, different computational methodologies only accounting for ISC via an S 1 -T 1 channel give considerably underestimated ISC rates compared to the experimental data, largely due to the large S 1 -T 1 splitting values.It can be noted that a ISC via a S 1 -to-T 2 transition is still consistent with FP studies, in which T 1 state absorption was identified following fluorophore excitation [38], given that a fast T 2 -to-T 1 relaxation likely occurs after the S 1 -to-T 2 state transition, before recovery to S 0 .Such fast T 2 -to-T 1 relaxation is also fully consistent with the TRAST measurements, and it would take place within the T state (figure 2(a)), well before its relaxation to S 0 or + R .

Conclusions
By TRAST experiments, we could directly measure excitation-induced population build-up of triplet and photo-oxidized states in carboxy-fluorescein and in high triplet yield, brominated variants thereof.Bromination was found to generate a prominent internal heavy atom (IHA) enhancement of the intersystem crossing (ISC) rates, and our DFT calculations suggest this ISC to take place via a higher triplet state, followed by relaxation to the lowest triplet state.A corresponding external heavy atom (EHA) enhancement was found upon adding potassium iodide (KI).In contrast to the common view of KI as a fluorescence quencher, addition of KI resulted in lowered triplet state populations in the brominated fluorophores.This can be attributed to much lower relative enhancements of the (intrinsically very prominent) ISC rates, than of the (orders of magnitude lower) triplet decay rates.
Together with an antioxidative effect on the fluorophores, KI can thus act as a fluorescence enhancing compound on the brominated fluorophores.
In contrast to SMD and FCS techniques, TRAST experiments do not rely on high signal-to-background conditions, nor on fluorophores with high brightness or a high time resolution.The experiments are therefore broadly applicable, also on biological samples and living cells, and can be performed under more moderate excitation intensities.However, TRAST experiments rely on excitation-induced build-up of fluorophore dark transient states, such as triplet states, in the samples.Therefore, excitation intensities and doses may sometimes reach a level where photo-toxic effects need to be considered.For the very high triplet yield (>90%) fluorophores used here however, much lower excitation intensities are needed in the TRAST measurements to drive these fluorophores into their triplet states.The photophysical characterization of this work thus suggests TRAST measurements based on high triplet yield fluorophores as a useful approach for biomolecular interaction and sensing studies, with long-lived, highly sensitive dark states of these fluorophores followed under more biologically relevant conditions, for sensing and quenching studies in living cells and biological tissues.

Figure 2 .
Figure 2. (a) Electronic state model used in this work for CFl and brominated CFl derivatives.S 0 and S 1 denote the ground and first excited singlet states , with k 01andk 10 representing the excitation and de-excitation rates, respectively.k isc denotes the intersystem crossing rate from S 1 to the triplet state, T, and k T the triplet decay rate back to S 0 .Photo-oxidation of fluorophores to + R  state is assumed to occur from both S 1 and T, with a rate denoted k .oxFrom the long-lived + R  state, the fluorophores can relax back to S 0 through a photo-reduction rate, k .red (b) Simplified electronic three-state model, describing state transitions taking place on a time scale much longer than the equilibration of the S 0 and S 1 states, where S represents both these singlet states and ¢ = + k k .
(a), to three states (figure 2(b)); the fluorescent S (S 1 and S 0 ), and the non-fluorescent T and +

Figure 3 .
Figure 3. Visual representation of the wide-field TRAST setup used in this study (see main text for details).

isc 0 k T 0 and k , red 0
as well as the KI enhancement pre-factors, k , Qisc k QT and k , Qred were all fitted as global parameters for the set of eight TRAST curves recorded (for [KI] between 0 and 50 mM) for each fluorophore.The resulting sets of fitted curves could well reproduce the experimental TRAST curves for all the fluorophores (figures 5(c)-(f)).The fitted parameter values are plotted in figures 6(a)-(d) and summarized in table 2.

Figure 5 .
Figure 5. Fitted (lines) and experimental (dots) TRAST curves recorded from 100 nM fluorophore solutions (1.2 mM PBS, pH = 9.3).Fitting residuals plotted below.TRAST curves measured with I exc = 1.5 kW cm −2 , if not otherwise stated.(a) Comparison of TRAST curves recorded from CFl, CFl-Br, CFl-2Br and CFl-4Br solutions.(b) TRAST curves recorded from a CFl solution with different I exc applied (inset).Global fitting of the curves, with k , isc k , T k red and k ox fitted as global parameters, yielded: m = k s 9.14 , isc 1

3. 3 .
Quantum chemical estimations of radiative and ISC rates from S 1In order to investigate possible pathways for the IHA effect in the brominated fluorophores, to what extent they can contribute to the ISC rates and relate to the fluorescence rates we performed quantum-chemical calculations.The fluorescence rate (k r ) and the ISC rate (k isc ) were calculated for all four dyes, at different levels of density functional theory (DFT) and employing a polarizable continuum model (PCM) to account for solvent effects.Three different DFT/PCM methodologies were employed (TDA/B3LYP/TZP (COSMO), TDA/wB97X/TZP (COSMO) and TDA/ B3LYP/TZP (SM12), further described in Supplementary part S3).

Figure 6 .
Figure 6.Photo-physical rate parameters obtained from the fitting of the experimental TRAST curves in figure 5, plotted versus [KI].Fitted parameter values listed in table 2.

Table 1 .
Fluorescence lifetimes, t , Br and CFl-2Br assumed to have the same peak s exc as CFl-4Br.The resulting estimated s exc values at 488 nm excitation are listed in table 1, together with the peak absorption and emission wavelengths, and the relative fluorescence intensities of the different fluorophores.
f estimated s exc at 488 nm, relative fluorescence intensities and peak absorption and emission wavelengths for CFl and its brominated derivatives.

Table 3 .
Energies of lowest lying electronic states (in eV) for CFl, CFl-Br, CFl-2Br, CFl-4Br and FITC, oscillator strengths for the corresponding S 0 -S 1 transitions and SOCMEs S H T n -1Br).The main reason for the small values of k isc for the S 1 -T 1 channel, irrespective of the level of theory used, is the very large S 1 -T 1 energy gap.A second major reason is that both the S 1 and T 1 states correspond to ππ * electronic states, without any significant contribution from natural transition orbitals (NTOs) of the Br atoms (See SI, tableS1).Thus, the resulting | ˆ| between S 1 state and T n states (in cm −1 ).a Experimental emission maxima.CFl