Determination of quantum yield of NADH and FAD in alcohol-water solutions: the analysis of radiative and nonradiative relaxation pathways.

Reduced nicotinamide adenine dinucleotide (phosphate) (NAD(P)H) and oxidized flavin adenine dinucleotide (FAD) are essential biological coenzymes involved in regulation of living cell metabolism that are actively used nowadays as intrinsic fluorescent biomarkers for monitoring the respiratory chain activity and cell metabolism [1, 2, 3]. A high potential of NAD(P)H and FAD autofluorescence as biomarkers was highlighted in a series of seminal works by Chance et al. [4, 5]. Chance et al. have demonstrated that NADH in its reduced form exhibits autofluorescence at 400–500 nm, whereas its oxidized form, NAD⁺ does not fluoresce and developed spectroscopic ways to measure a redox ratio NAD⁺/NADH for real-time monitoring of the metabolic state of a cell during pathophysiological changes.

Reduced FAD (FADH₂) and FAD are another redox pair largely associated with respiration in mitochondria of eukaryotic cells. As known, the reduced form FADH₂ does not fluoresce while the oxidized form FAD can fluoresce under excitation in the spectral region shorter than 420 nm [6]. Zhang et al. [7] suggested a prominent way of determination the mitochondria redox state by measuring the redox ratio HADH/FAD and shown that the measurement of the ratio of fluorescence has important advantages over the absolute values.

Chance et al. [4, 8] suggested that the changes occurring in redox state in cancer cells can be effectively studied by monitoring NADH fluorescence. This approach was developed and widely used by many groups (see, e.g. [9, 10]) however further experiments elucidated a number of biological and technical problems dealing with light scattering in tissue, absorption of the laser light at 360 nm by other intracellular species, and the influence of a number of factors on fluorescence intensity.

To overcome these problems Lakowicz et al. [11] employed the measurements of NADH fluorescence lifetimes that are independent of light intensity and can be used for separation of free and protein-bound NADH quantitatively. Moreover, Bird et al. [12] reported that the ratio of free/protein-bound NADH is related to changes in the NADH/NAD⁺ ratio. Therefore, quantifying NAD(P)H and FAD fluorescence through fluorescence lifetime imaging allows in general to determine the optical redox ratio with high sensitivity and therefore to monitor small changes in metabolic activity [3]. Fluorescence lifetime imaging microscopy (FLIM) is now widely used for monitoring of metabolic pathways activated in cancer cells [2, 13, 14, 15].

As demonstrated in numerous studies mentioned above, FLIM in NAD(P)H and FAD can be used for non-destructive determination of changes in cellular metabolism. However, till now these changes are difficult to interpret [16]. The isotropic (polarization insensitive) part of the fluorescence signal $I_{tot}(t)$ is usually presented as a sum of exponentials [17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27]:

where $I_{0}$ is a time-independent initial fluorescence intensity, $\tau_{i}$ are decay times, and $a_{i}$ are corresponding weighting coefficients that are normalized to unity: $\sum_{i}a_{i}=1$.

Nowadays using several supplementary fluorescence detection experimentalist can measure a great number of fluorescence decay times $\tau_{i}$ describing various aspects of the fluorescence dynamics. One of them is time correlated single photon counting (TCSPC) technique allowing for determination of decay times in the nanosecond and near sub-nanosecond time domain. The others are femtosecond transient monitoring and up-conversion techniques that allow for determination of the decay times in picosecond and sub-picosecond time domain.

In the nanosecond and near sub-nanosecond time domain free NADH in aqueous solution is known to exhibit biexponential fluorescence decay with the lifetimes of about 0.3 ns and 0.7 ns [28, 29, 18, 30] and in other solvents the lifetimes and corresponding weighting coefficients depend strongly on temperature and solvent type [31, 29, 28, 17]. The explanation of the nature of the two observed fluorescence decay times in NADH for long time remained controversial. In particular it was unclear if the two fluorescence lifetimes either correspond to different geometric conformations of the whole molecule, or they are intrinsic features of the NA ring [28, 32, 31, 33, 34, 17, 19]. Gorbunova et al. [35] suggested that the heterogeneity in the measured decay times in NADH is due to different charge distributions in the cis and trans configurations of the nicotinamide ring that results in different electrostatic field distributions and different non-radiative decay rates $(\tau^{\,nrad}_{l})^{-1}$ (see eq. (8)). This model was also shown to be useful for elucidation of the single-exponential and relatively slow fluorescence decay in NADH-alcoholedegidrogenase (ADH) complex [21] because as known from the X-ray structure spectroscopy and NMR experiments, NAD(P)H imbedded into the ADH binding site in a single unfolded anti ”trans-out-of-plane” nuclear conformation [36, 37, 38]. Within the same model a significant enhancement of the decay time value in the ADH-bounded NADH compare with the free NADH in solution can be attributed [21] to a significant decrease of charges separations in the nicotinamide (NA) ring in the conditions of a typical apolar binding site environment in the ADH-NADH complex [39, 40]. Similar reasons are likely responsible for significantly different lifetime values observed in NADH bounded with various enzymes [16].

In the nanosecond time domain FAD in solutions and living cells demonstrates a multiexponential fluorescence decay dynamics. At physiological pH and temperature conditions in solution it is most common to fit FAD fluorescence decay after excitation of the isoalloxazine moiety by two exponentials with the lifetimes of about 2.3 ns and 3.5 to 4.5 ns [41, 42, 43, 44]. However, in other solvents the number of lifetimes, their values and corresponding weighting coefficients depend strongly on temperature, pH and solvent type [45, 6, 22, 42, 41]. Despite numerous studies carried out the origin of the two fluorescence nanosecond decay times in FAD still remains poorly understood.

Multiexponential fluorescence decay was also observed in the picosecond time domain by many groups using ultrafast fluorescence up-conversion and transient monitoring methods. A very short decay time of about $\tau\simeq$ 1 ps that was observed by several groups in NADH [26, 27, 46], FAD [25, 42, 22, 23, 47, 24], and many other molecules (see, e.g. [48, 49, 50]). It is commonly attributed to water relaxation around the corresponding excited chromophore moiety. A signature of this vibrational relaxation is Decay Associated Spectra (DAS) that is characterised by the positive value of the corresponding pre-exponential weighting coefficient $a_{i}$ in eq. (1) in the blue wing of the fluorescence band and negative, or near zero coefficient $a_{i}$ in the red wing of the fluorescence band [26, 46, 24, 50]. However, the understanding of DAS still remained controversial [49].

Another typical picosecond decay times of about $\tau\simeq$ 26 ps observed in NADH [26, 27] and of about $\tau_{d}\simeq$ 5-9 ps observed in FAD [23, 24, 42, 25] were associated with picosecond relaxation occurring between the excited $S_{1}$ and the ground $S_{0}$ electronic states and resulting in fast fluorescence quenching. These fluorescence decay is known as quasi-static-self-quenching (QSSQ) [50] and results in decrease of the observed quantum yield. In the case of NADH the routes of QSSQ are still not known exactly [26, 27, 46]. In particular, these can be either interactions of excited NADH molecules with surrounding solvent molecules, or intramolecular internal conversion to the ground electronic state, or intersystem crossing to the triplet manifold [51]. In the case of FAD it is now well adopted that the mechanism of the fluorescence quenching is an electron transfer reaction in the $\pi$-stacked conformation [6, 52, 53, 22, 54].

A common way for determination of the fractional population of free and bound species from the multiexponential expansion in eq. (1) is to assume that the radiation decay rate $\gamma_{rad}=\tau_{rad}^{-1}$ remains constant and then suggest that the weighting coefficients $a_{i}$ in eq. (1) are equal to the fractional concentrations of the species associated with it and to use the fraction of the total fluorescence signal generated by each species $i$ [40, 55, 16]:

However, the origin of the fluorescence decay times numbers and their values in NADH and FAD bound to different enzymes still remains controversial. Therefore, the relative number of coenzyme molecules bound to different enzymes cannot be obtained without further characterization [55, 56, 21]. As pointed out by Niesner et al. [55] in general $a_{i}$ and $f$ values are influenced by the concentration, absorption cross-section, and fluorescence quantum yield. Consequently, it is necessary to know the photophysical properties of the fluorescing species in order to calculate their relative concentration.

Fluorescence quantum yield is one of the most important parameter needed for determination of the redox ratio and the ratio of free and protein-bound NADH in biological solutions and living cells from fluorescence lifetime measurement [3, 57] as it is known to be highly dependent on solution/intracellular viscosity and polarity [58, 59, 32, 28, 60, 61, 17]. Scott et al. [59] studied fluorescence lifetime and quantum yield of NADH and its derivatives in water and propylene glycol (PG) and revealed that the excited state dynamics in NADH are mostly characterized by non-radiative processes. This suggestion was supported in the following studies [32, 17, 28, 61, 60] where the dependence of fluorescence quantum yield and non-radiative decay rates in NADH on temperature or solvent type was extensively studied. Gafni et al [32], Fischer et al. [60], and Visser et al. [28] reported an increase in the NADH quantum yield by an order of magnitude upon binding with enzymes. Nakabayashi et al. [62] studied the effect of the local electric field on NADH fluorescence in solutions and demonstrated an increase in average fluorescence decay time due to the decrease in solvent polarity. Blacker et al. [17] inquired time-resolved fluorescence in NADH in water-glycerol solution and calculated non-radiative decay rate using Kramers-Hubbard equation.

However, the origin of non-radiative pathways accounting for excited-state dynamic in FAD and especially NADH still remain controversial. Also it is unclear ether intermolecular quenching processes or intramolecular interaction of nicotinamide and adenine can be account for the very low fluorescence quantum yield in NADH [17, 59, 60].

Fluorescence quantum yield in FAD in solutions and proteins was intensively studied during latest decades [6, 42, 54, 63]. Nakabayashi et al. [42] observed rapid growth of fluorescence excited at 410 nm depending on alcohol concentration increase in alcohol-water solutions and suggested that the addition of a less polar solvent breaks the $\pi$-$\pi$ interaction in the stacked conformation and produces the open FAD conformation which results in the rise of fluorescence intensity. Radoszkowicz et al. [54] performed molecular dynamics simulations of FAD in water-methanol mixtures and showed that methanol affected the dynamics of FAD by enhancing the frequency of unfolding events without any effect on the lifetime of unfolded states.

Despite a number of studies carried out the detailed physical interpretation of experimental data in the form of multiexponential expansion in eq. (1) and in the form of fluorescence quantum yields is still a challenging problem.

This paper aims to address this problem, at least partly. We carried combined experimental studies including the determination of fluorescence quantum yields in NADH and FAD dissolved in water mixtures with several alcohols: methanol (MeOH), ethanol (EtOH), and propylen glycol (PG), and the time-resolved fluorescence studies of the same molecules using TCSPC method under excitation at 450 and 355 nm. The experimental results were analysed using a new model taking into consideration radiative excitation and fluorescence channels as well as nonradiative excited state relaxation pathways and based on the quantum mechanical theory developed recently by the authors [64].

The main results obtained are as follows. A significant difference was observed in the behavior of NADH and FAD quantum yields in alcohol solution. In particular, the dependence of quantum yield in NADH on alcohol concentration was similar for monohydric alcohols (methanol and ethanol) and significantly different for polyhydric alcohol (propylene-glycol). In the case of FAD, the behavior of quantum yield was almost independent of the type of alcohol and exhibited a dramatic increase of 5–6 times with alcohol concentration.

The analysis of general theoretical expression describing fluorescence decay in a polyatomic molecule excited by a short laser pulse led to better understanding of the nature of the DAS phenomenon. A generalised expression for the fluorescence quantum yield that allowed to separate contributions from the fast and slow nonratiative processes in molecular excited states was derived and used for clarification of the role of nanosecond and picosecond relaxation channels in NADH and FAD in various water-alcohol mixtures.

The contribution of the fast picosecond nonradiative decay in NADH to the quantum yield in water-MeOH and water EtOH mixtures was found to be practically independent of the alcohol concentration and hence of the NADH conformation distribution. This result suggests that the picosecond decay in NADH does not likely occur through electron transfer in the stacking configuration of the nicotinamide and adenine moieties but through other mechanisms. It also suggests that the known growth of NADH quantum yield at high alcohol concentration occurs mostly due to relatively slow nanosecond nonradiative mechanisms related with conformation distributions. At the same time the contribution of the fast picosecond nonradiative decay in NADH to the quantum yield in water-PG solutions was found to depend on PG concentration and the sharp nonlinear increase of the measured quantum yield with PG concentration was associated with decrease of both fast picosecond and relatively slow nanosecong quenching nonradiative rates.

In FAD, the contribution of the fast picosecond quenching to the quantum yield was found to depend significantly on the alcohol concentration in all MeOH, EtOH. This finding suggests that the fast quenching gives a profound contribution to the rise of the measured quantum yield with alcohol concentration and supports the established mechanism of the fluorescence quenching in FAD through electron transfer reaction in the $\pi$-stacked conformation between isoalloxazine and adenine moieties.

Fluorescence quantum yield in FAD determined upon excitation at 450 nm and 355 nm in three different water-alcohol solutions of various concentrations are presented in Figs. 1a–c as a function of the alcohol concentration.

According to Figs. 1a–c the absolute quantum yield in FAD in pure water under excitation at 450 nm and 355 nm was found to be $Q$=3.33$\pm$% and $Q$=1.87$\pm$%, respectively. The quantum yield in pure water at 450 nm in Figs. 1a–c agrees perfectly with that reported earlier by Islam et al. [6] and Sengupta et al. [63]. To the best of our knowledge the quantum yield in FAD under excitation at 355 nm was determined in this paper for the first time.

As can be seen in Fig. 1 the fluorescence quantum yield in FAD rose dramatically with alcohol concentration in all solutions under study at each excitation wavelength. In water-propylene glycol solutions fluorescence quantum yield in Fig. 1a increased practically linearly with propylene glycol concentration and in pure propylene glycol reached the values of $Q$=17% and $Q$=9% under excitation at 450 and 355 nm, respectively. In water-methanol solutions the quantum yield in Fig. 1b also increased linearly up to the methanol concentration of about 80% and reached the values of $Q$=15% and $Q$=10% under excitation at 450 and 355 nm, respectively. At higher methanol concentrations the quantum yield decreased somehow by 2–3% at both excitation wavelengths. In water-ethanol solutions the quantum yield in Fig. 1c increased almost linearly with alcohol concentration up to the concentration of about 70% and then increased more slow at higher concentrations reaching the maxima of 19% and 11% under excitation at 450 and 355 nm, respectively.

Quantum yields in NADH determined upon excitation at 355 nm in water–propylene glycol, water-methanol, and water-ethanol solutions as a function of alcohol concentrations are shown in Figs. 2a–c.

The fluorescence quantum yield in aqueous solution was determined to be $Q$=2.1$\pm$0.2 that agrees perfectly with the value reported earlier by Scott et al. [59] and Gafni et al. [32]. As can be seen in Fig. 2a in water-propylene glycol solutions, fluorescence quantum yield in NADH exhibited a nonlinear behavior with propylene glycol concentration and reached the maximum of $Q$=8% in pure propylene glycol. This result is also in agreement with the earlier result reported by Scott et al. [59]. The quantum yield dependence in Fig. 2a almost perfectly follows the propylene glycol viscosity dependence, also the quantum yield value in pure propylene glycol was about 4 times larger than in water. As can be seen in Figs. 2b,c in the case of water-methanol and water-ethanol solutions fluorescence quantum yield in NADH increased practically linearly with alcohol concentration and reached the maxima of about 3.2 % in the pure alcoholes. As can be seen in Figs. 2b,c fluorescence quantum yield in NADH was about 1.5 times larger in pure alcohol than in water.

The dependence of FAD and NADH quantum yields on ethanol and methanol concentrations was studied earlier by several groups [42, 54, 60, 59]. Our results presented in Figs. 1 and 2 are at least in qualitative agreement with them. The increase of fluorescence intensity in FAD in water-ethanol and water-glycerol solutions was observed early by Nakabayashi et al [42] who observed an almost 5 times increase of FAD fluorescence intensity upon excitation at 445 nm in 50% ethanol solution compared with pure water. The dependence of FAD quantum yield on methanol concentration shown in Fig. 1b is consistent with the result reported earlier by Radoszkowicz et al [54]. In the case of NADH, an 1.5 times increase of fluorescence intensity upon excitation at 360 nm in methanol solution compared with water was reported by Freed et al. [58]. Moreover, the same increase in fluorescence intensity and relative quantum yield in dihydronicotinamide in ethanol and methanol was observed by Fischer et al [60]. At the same time, the complete quantitative data on fluorescence quantum yields in NADH and FAD in Figs. 1 and 2 are presented in this paper at the first time.

The experimental results shown in Figs. 1 and 2 can be briefly summarized as follows. A significant difference was observed in the behavior of NADH and FAD quantum yields in alcohol solution. In particular, the dependence of quantum yield in NADH on alcohol concentration was similar for monohydric alcohols (methanol and ethanol) and significantly different for polyhydric alcohol (propylene-glycol). In the case of FAD, the behavior of quantum yield was almost independent of the type of alcohol and exhibited a dramatic increase of 5–6 times with alcohol concentration.

The fluorescence decay times and corresponding weighting coefficients are given in Tables 1 and 2 for NADH and FAD in water-alcohol solutions, respectively. The data in Table 1 were obtained in our recent publication [68] and the data in Table 2 were obtained in this paper experiments. As can be seen in Tables 1 and 2 the fluorescence decay in both molecules was double exponential with decay times $\tau_{1}$ and $\tau_{2}$.

Experimental data in Table 1 for aqueous solution are in agreement with our recent results [30, 35] and with those those reported earlier in NADH under one-[61, 31] and two-photon [19] excitation within experimental error bars. Also, both decay times shown in Table 1 in pure methanol agree roughly with those reported earlier [31, 33] where a three-exponential model was used.

The existence of two excited state lifetimes in FAD obtained in this paper is in agreement with the results reported by Sengupta et al. [63] and Esposito et al. [69], however contradicts to the conclusions given by Nakabayashi et al. [42], where the existence of four lifetimes have been reported under similar experimental conditions. The decay time $\tau_{1}$ and $\tau_{2}$ values in Table 2 obtained in aqueous solution are in perfect agreement with those reported earlier [63, 69, 44].

The studies of the quantum yield and the dynamics of fluorescence decay in NADH and FAD excited at 355 nm and 450 nm in water- methanol, ethanol, and propylen glycole mixtures have been carried out. The global fit procedure was used for determination of the fluorescence parameters from experimental data. A significant difference was observed in the behavior of NADH and FAD quantum yields in alcohol solution. In particular, the dependence of quantum yield in NADH on alcohol concentration was similar for monohydric alcohols (methanol and ethanol) and significantly different for polyhydric alcohol (propylene-glycol). In the case of FAD, the behavior of quantum yield was almost independent of the type of alcohol and exhibited a dramatic increase of 5–6 times with alcohol concentration. The contribution of the fast picosecond nonradiative decay in NADH to the quantum yield in water-MeOH and water EtOH mixtures was found to be almost independent of the alcohol concentration. The contribution of the fast picosecond nonradiative decay in NADH to the quantum yield in water-PG solutions was found to depend on PG concentration. The contribution of the fast picosecond quenching to the quantum yield in FAD was found to depend significantly on the alcohol concentration in all MeOH, EtOH. The experimental results were analysed using a new model based on the quantum mechanical theory developed recently by the authors. The analysis of general theoretical expression describing fluorescence decay in polyatomic molecules excited by a short laser pulse gave a new insight of the nature of DAS phenomenon. An expression for the fluorescence quantum yield that allowed to separate the contributions from the fast and slow nonratiative processes in molecular excited states was derived and used for clarification of the role of nanosecond and picosecond relaxation channels in NADH and FAD in various water-alcohol mixtures. A sharp increase of the measured quantum yield with PG concentration was associated with decrease of both fast picosecond and relatively slow nanosecong quenching nonradiative rates. As was shown the fast quenching gives a profound contribution to the rise of the measured quantum yield with alcohol concentration. The results obtained were compared with the results reported by other authors.

The experiments were carried out under the support of Russian Science Foundation, grant No 18-53-34001. Calculations and models developed were performed under the financial support of BASIS Foundation, grant no. 19-1-1-13-1. The authors are grateful to the Ioffe Institute for providing the equipment used in experiments.