Dehydrogenase Binding Sites Abolish the “Dark” Fraction of NADH: Implication for Metabolic Sensing via FLIM

Abstract

The fluorescence of dinucleotide NADH has been exploited for decades to determine the redox state of cells and tissues in vivo and in vitro. Particularly, nanosecond (ns) fluorescence lifetime imaging microscopy (FLIM) of NADH (in free vs bound forms) has recently offered a label-free readout of mitochondrial function and allowed the different “pools” of NADH to be distinguished in living cells. In this study, the ultrafast fluorescence dynamics of NADH–dehydrogenase (MDH/LDH) complexes have been investigated by using both a femtosecond (fs) upconversion spectrophotofluorometer and a picosecond (ps) time-correlated single photon counting (TCSPC) apparatus. With these enhanced time-resolved tools, a few-picosecond decay process with a signatory spectrum was indeed found for bound NADH, and it can best be ascribed to the solvent relaxation originating in “bulk water”. However, it is quite unlike our previously discovered ultrafast “dark” component (~26 ps) that is prominent in free NADH (Chemical Physics Letters 2019, 726, 18–21). For these two critical protein-bound NADH exemplars, the decay transients lack the ultrafast quenching that creates the “dark” subpopulation of free NADH. Therefore, we infer that the apparent ratio of free to bound NADH recovered by ordinary (>50 ps) FLIM methods may be low, since the “dark” molecule subpopulation (lifetime too short for conventional FLIM), which effectively hides about a quarter of free molecules, is not present in the dehydrogenase-bound state.

Graphical Abstract

1. INTRODUCTION

In the past decades, the reduction–oxidation pair NADH/NAD⁺ has been gradually recognized as a key indicator of the cellular metabolic state associated with health and disease.¹ Modern applications are frequently based on the metabolic differences between normal and malignant cells. In normal cells, NADH is produced in cytosol and then converted to NAD⁺ during the mitochondrial oxidative phosphorylation activity. However, there is a significant reprogramming of metabolism in cancer cells. The mitochondrial oxidative phosphorylation is largely diminished and more NADH is produced through the anaerobic glycolysis (Warburg effect),² resulting in an obvious increase of the NADH/NAD ratio.

Using the most prominent endogenous fluorophore in cells, Chance and co-workers proved decades ago that the autofluorescence of NADH could report the oxidation–reduction levels.^3,4 In their pioneering work, they described how the spectral differences between free and protein-bound NADH could be used to track the intracellular metabolic state. Moreover, in 2005, Bird and co-workers demonstrated that the NADH/NAD⁺ ratio was related to the changes of the NADH free/bound ratio,⁵ which could be obtained from ns lifetime data (e.g., FLIM). The NADH free/bound ratio has become an important and frequently used indicator for tracking the intracellular metabolism.^6–12

The optical absorption of NADH in the ultraviolet (UV) is from its adenine and nicotinamide, and NADH has an absorption band at 340 nm which has long been exploited to measure the kinetics of many pyridine nucleotide redox enzymes. Fluorescence originates from the reduced nicotine amide with the absorption at 340 nm, and the emission maximum is near 460 nm for the free coenzyme or variably blue-shifted when the coenzyme is bound to proteins. These changes of fluorescence on protein binding have been used to study the thermodynamics of protein/coenzyme interaction and the fraction of reduced coenzyme bound to proteins in the living cells.

It has previously been shown via nanosecond fluorescence decay measurements that the free NADH molecules have a set of sub-nanosecond lifetimes with an average lifetime of ~0.4 ns. Visser and colleagues¹³ were the first to resolve those lifetimes further into ~0.25 and 0.70 ns components by using a mode-locked Ar⁺ laser. Much speculation about the origins of those components in the stacked adenine—pyridine complexes has occurred,^14–17 but it is difficult to make definitive assignments without an extensive quantum mechanical-molecular mechanics (QM-MM) simulation. The bound forms of NADH (principally to dehydrogenases found in the metabolic cycles of cells) were extensively studied by Gafni and Brand^18–20 along with other groups^21–25 to learn more about the quantum yield (QY) and lifetime enhancement characteristics of the rigid binding in cognate sites. Lifetimes were obtained in the 2—12 ns range with a typical 4 ns bound lifetime, which enhanced an order of magnitude over the free and self-quenched molecules. The emission spectrum of bound NADH was also considerably blue-shifted from the free form. This feature actually led to a long-time misunderstanding about NADH: the emission spectrum of NADH in mitochondria was blue-shifted as well, so the bound state was presumed to be dominant. Actually, since the contributions to a spectrum were lifetime weighted (i.e., the amplitude proportional to concentration must be multiplied by lifetime when calculating the steady-state spectral weights), the order-of-magnitude lifetime difference made the “bound” state dominate the steady-state spectra, even though the bound state was eventually found to represent less than a quarter of mitochondrial NADH concentration.²⁶

These contribution calculations all assumed that the free and bound species shared any other excited state deactivating mechanisms equally (similar radiative rate, intersystem crossing, etc.). The possibility that ultrafast deactivation could further alter quantum yield per molecule was introduced by R.F. Chen and colleagues for tryptophan in peptides and proteins. He deduced the presence of dark subpopulations created by QSSQ (quasi static self quenching). This, in turn, was derived from a disparity between the quantum yield and ns lifetimes via “apparent radiative lifetime” differences (lifetime/QY).²⁷ More recently, direct measurement of ultrafast decay revealed QSSQ in peptides in the form of rapid decay terms (tens of picoseconds) ascribable to charge transfer kinetics.^28–30 We have recently reported similar QSSQ phenomena in free NADH.^31,32 We found (in the sub-500 nm observation window) that 25–30% of free NADH was “dark”. Thus, free NADH values gleaned from TCSPC or phase lifetime (especially FLIM imaging of cells) should be corrected upward ~25% (depending somewhat on the detection band).

To further explore the consequences, in this paper, dehydrogenases were chosen as exemplars of the bound state host. We present direct evidence, using the ultrafast fluorescence upconversion technique, that the “dark” component in free NADH essentially disappeared when bound to active sites in cognate proteins. This disparity in the number of molecules generating rapid decay components must be taken into account when estimating the fraction of coenzyme bound to protein in the living cells.

2. EXPERIMENTAL SECTION

Upconversion Spectrophotofluorometer.

The detailed information about the experimental setup has been described elsewhere.³³ Briefly, the fundamental pulse centered at 800 nm was generated by a seeded ultrafast Ti:sapphire regenerative amplifier (Astrella, Coherent Inc.). The output infrared beam typically had an autocorrelation pulse width of ~90 fs and a repetition rate of 1 kHz with an average power of 7 W. The NADH excitation pulse centered at 340 nm was generated by an optical parameter amplifier (OPerA Solo, Coherent Inc.). A simple spectrometer made up of a pair of UV prisms was used to further purify the excitation pulse. The power of the excitation pulse was carefully attenuated to around 0.1 mW to avoid photodegradation or hole burning. The sample was held in a UV quartz disk-shaped cuvette with a diameter of 80 mm and continuously spun with a tangential velocity of ~5 m/s to avoid overexposure at the same sample spot. After the emission was collected by a pair of off-axis parabolic mirrors, the fluorescence was focused into a BBO frequency mixing crystal (with a thickness of 0.2 mm), together with the delayed fundamental infrared beam (800 nm) acting as a timing “gate” pulse. The upconverted signal was obtained via type I sum frequency generation with the gate pulse. In order to reduce various background signals such as unmixed fluorescence, remnant UV, or transmitted gate pulse, the fluorescence of samples and the gate pulse were arranged with non-collinear geometry. The polarization of excitation was adjusted to the magic angle (54.7°) relative to the crystal acceptance axis (perpendicular) by a zero-order half-wavelength plate to prevent any kinetic contribution from fluorescence anisotropy. The upconverted signal (275–330 nm) was directed into a monochromator (Omnik500, Zolix), and then detected by a slow PMT (CR317, Hamamatsu). The impulse instrument response function was ~350 fs by measuring the crosscorrelation between Raman scattering (near 384 nm) of water and the gate pulse.

TCSPC Apparatus.

In this system, the excitation pulse centered at 340 nm (with 10 MHz repetition) was generated by a picosecond pulsed diode laser (PDL 800-B, PicoQuant). The fluorescence was recorded using a time-correlated single photon counting module (PicoHarp 300, PicoQuant) and a single photon counting PMT (PMA165A-N-M, PicoQuant). The instrument response function was found to be ~440 ps in width by measuring the Rayleigh scattering of excitation pulses in a suspension of 0.34 wt% SiO₂ nanoparticles in water. More details about the system can be found elsewhere.³⁴ For constructing decay associated spectra (DAS), the fluorescence was scanned (in the range 420–560 nm for NADH–MDH and 420–500 nm for NADH–LDH) with 10 nm bandwidth and analyzed by a global fitting technique³⁵ using a biexponential model. The fits in this work usually yielded x² values of 1.01–1.15.

Steady-state absorption spectra were measured with a UV–visible spectrophotometer (TU1901, Beijing Purkinje General Instrument Co. Ltd.). Steady-state fluorescence spectra were recorded using a commercial spectrofluorometer (FluoroMax-4, Horiba). The decay associated spectra were normalized to the steady-state spectra of NADH–MDH/LDH in Tris–HCl buffers.

Samples.

β-NADH disodium salt (>98%) was purchased from Aladdin (Shanghai, China). Malate dehydrogenase (MDH, ammonium sulfate suspension, 600 units/mg protein) from porcine heart and lactate dehydrogenase (LDH, type XI, lyophilized powder, 600–1200 units/mg protein) from rabbit muscle were bought from Sigma. All samples were prepared in Tris–HCl buffer (50 mM Tris, 150 mM NaCl, pH 7.35).

MDH, ammonium sulfate suspension, was first centrifuged for 30 min (9000 rpm); then, the precipitate was redissolved in Tris-HCl buffer. After that, the solution was centrifugated for 30 min again to remove any insoluble particles. The supernatant was then washed via ultrafiltration centrifugation (Millipore, 15 mL, 30 KD) for four times to remove any ammonium sulfate. Then, the MDH was eluted from the tube by Tris–HCl buffer. The solution was centrifuged for another 30 min to remove any insoluble particles. LDH was first dissolved in Tris–HCl buffer, and then, the solution was passed through a 0.22 m sterile syringe filter (PES membrane) to remove insoluble particles from the solution.

The concentrations of MDH and LDH in solutions were determined to be about 130 and 40 M using the extinction coefficients of ε₂₈₀,_MDH = 20160 M⁻¹ cm⁻¹ and ε₂₈₀,_LDH = 200000 M⁻¹ cm⁻¹, respectively.^36,37 The NADH concentration was checked with the absorbance at 340 nm using an extinction coefficient of ε₃₄₀,_NADH = 6200 M⁻¹ cm−¹.³⁸

MDH in solution is a dimer, and each subunit is capable of binding one NADH molecule.³⁹ LDH in solution, on the other hand, is a tetramer with four independent subunits.^40,41 In our experiments, the concentration of NADH was 75% of that of binding sites of proteins so that most of NADH existed in protein-bound form. The fraction of free NADH in mixed solution was calculated to be only 3–6% according to eq 1⁴²

$$K_{\text{d}}=\left(1-\alpha \right)\left(\frac{L_0}{\alpha }-E_0\right)$$ (1)

where L₀ and E₀ are the concentration of NADH and binding sites of proteins, respectively. K_d represents the dissociation constant (~6.2 × 10⁻⁶ M for NADH–MDH²⁴ and (4.2 00B1 0.2) × 10⁻⁶ M for NADH–LDH,²⁵ respectively).

All samples were freshly prepared for each time-resolved measurement, and the absorption spectra were carefully checked after each experiment to make sure there was no obvious photodamage or photodegradation for the samples.

3. RESULTS AND DISCUSSION

The fluorescence of a low quantum yield, photolabile ultraviolet compound would make NAD(P)H seem a poor target, but the biological abundance of the metabolic cofactor NAD(P)H has proven valuable for decades in the analysis of physiology of cells and tissues. More recently, the popularity of FLIM (fluorescence lifetime imaging microscopy) has grown with its application to NADH, since the apparent free/bound ratio allows one to derive factors, such as fluorescence lifetime redox ratio,^43–45 that are instrumental in defining cell status and even malignant tumor behavior.

The free/bound ratio of NADH has been widely sought; it is gleaned from lifetime data⁴⁶ in living cells and related to the metabolic potential and mode. That useful feature is largely unchanged by what follows, but when one calculates the actual concentration of free NADH for compartmental models, the amplitude of “free” will need to be adjusted.^31,32 In the key dehydrogenase sites we tested here (as proxies for the bound state in general), little or no amplitude (population) correction was needed.

Figure 1a shows the steady-state emission spectra (340 nm excitation) of NADH–MDH complex and free NADH in solution. The fluorescence of NADH–MDH complex has its maximum at 457 nm, which is blue-shifted by ~10 nm relative to that of NADH alone in solution (467 nm). Notably, the steady-state fluorescence intensity of NADH–MDH complex is much higher than that of free NADH. The fluorescence titration experiments we used to confirm complete binding setpoints also show an ~2.2 times higher quantum yield of NADH–MDH complex compared with free NADH (data not shown). Figure 1b (inset) shows the representative nanosecond-resolved fluorescence decay of NADH–MDH complex at 460 nm collected at the magic angle in TCSPC. On the time scale from 500 ps to 15 ns, the decay curves of NADH–MDH complex are apparently multiexponential. Global fitting results also show two lifetime components (0.28 and 0.96 ns) with similar DAS (decay associated spectra, see below). The intensity-weighted mean lifetime²⁷ is around 0.8 ns in agreement with previous reports.²¹ For comparison, free NADH in solution yields a biexponential with lifetimes of ~0.25 and 0.70 ns, and the average lifetime is ~0.4 ns.¹³

Figure 1.

Steady-state and nanosecond spectral properties of free/bound NADH. (a) Peak normalized steady-state spectra of free NADH and NADH–MDH complex. (b) Decay associated spectra of NADH–MDH complex extracted from TCSPC data (0.28 ns for blue line and 0.96 ns for red line). The inset is the typical TCSPC-collected magic angle decay of the NADH–MDH complex. Emission: 460 nm. The concentrations of NADH and MDH are 119 μM and 131 μM, respectively.

Decay associated spectra (DAS) have been widely used to evaluate the relative importance of ground-state heterogeneity vs solvent relaxation in tryptophan emission. A DAS with the “positive blue, negative red” characteristics is a typical hallmark for a two-state excited state reaction (e.g., solvent relaxation).³⁰ Figure 1b shows the decay associated spectra of NADH–MDH complex in solution. Both ns DAS show positive amplitudes at all detected emission wavelengths, which means that heterogeneity is dominant here. Interestingly, the two lifetime components nearly share the same shape and width, reflecting the similarity of dielectric environments for the apparent two different conformers.

The upconversion decay surfaces of NADH–MDH complex in solution are presented as a compendium of individual decay profiles at different emission wavelengths in Figure 2. The decay behavior in the earliest ~12 ps is clearly wavelength-dependent, with a fast decay at bluer emission wavelengths and a corresponding fast rise at redder emission wavelengths. To contrast, the peak normalized transients (40 ps full range) at two representative wavelengths for free (blue line) and protein-bound (red line) NADH in solution are shown in Figure 3a. Clearly, free NADH displays a fast decay process not present in the protein-bound form. At 480 nm, the fluorescence of NADH in both free and protein-bound forms shows similar decay behavior in the first ~40 ps. However, at 435 nm, free NADH shows a fast decay population component with a lifetime of ~26 ps, which has been described in our previous work.^31,32 Protein-bound NADH, in contrast, does not exhibit the same ultrafast component. Note that at 480 nm+, the fast decay component is much less prevalent; this provides one strategy for collecting NADH fluorescence in a band with less worry about the precise correction factor.³¹

Figure 2.

Raw upconversion decay surfaces for the NADH–MDH complex: (a, b) 0–12 ps; (c, d) 0–50 ps. The solid line is the fitting curve.

Figure 3.

Femtosecond spectral properties of free/bound NADH. (a) Peak normalized transients of free NADH and NADH–MDH complex. (b) Ultrafast decay associated spectra of NADH–MDH complex extracted from upconversion data. (c) Peak normalized transients of free NADH and NADH–LDH complex. (d) Ultrafast decay associated spectra of NADH–LDH complex extracted from upconversion data. The missing “QSSQ’ component (seen only in the free state) is also added to the graph (gray dashed line, arbitrarily normalized to show shape differences) for the reader’s reference.

By properly normalizing the upconversion data to the TCSPC derived (ns, conventional) DAS, we extract the fs-resolution DAS of NADH–MDH complexes, as shown in Figure 3b. Note that the small amplitude, few ps duration decay (green line) has a positive amplitude to the left of the fluorescence maximum at 460 nm but a negative amplitude at redder emission wavelengths. We ascribe this to a bulk water (~2 ps) response, possibly tainted by some slower relaxation in the binding pocket. For comparison, the gray dashed line shows our prior result in free NADH (arbitrarily normalized) to indicate what a “dark state” positive-dominant QSSQ term should have looked like in this context.

Ultrafast DAS seen in the bound state are small and cross zero near the peak of emission and are therefore relaxation dominated, though the water term expected at ~2 ps rates is likely mixed with some slightly slower relaxation and/or quenching. We did not further resolve this term.

The dark fraction of free NADH below 500 nm was seen to be ~25–33% (depending on instrument response vs wavelength characteristics), while, in the NADH binding site, at most 3–5% of net population was lost to such early relaxation and quenching. (Note the arbitrary normalization of the dashed gray result for free NADH was not scaled in proportion to reflect this disparity in Figure 3b.)

The LDH results shown in Figure 3c are remarkably similar. Again, the dark component from free NADH expected at 26 ps is absent in the bound state and the fast dark decay term is smaller (hence the disparity is smaller) near 480 nm. Figure 3d shows the fs-resolution DAS of NADH–LDH complex, which further confirms the MDH result: only a small (water relaxation dominated, green line) term is seen below 200 ps (the dashed gray is again a free NADH DAS comparison, to guide the reader). It appears that the binding sites of LDH and MDH are quite similar in their ability to restrict QSSQ

4. CONCLUSIONS

In this work, we show that the metabolically relevant, apparent ratios of free to bound recovered by FLIM will underrepresent the free cofactor, since the fraction of “dark” (QSSQ lifetime too short for FLIM) molecules is small in the dehydrogenase-bound state. The real-world consequences in normal FLIM metabolic imaging are (if these sites are representative of most “bound” sites) a 25% underestimation of the free species. Coupled enzymatic models of oxidative phosphorylation and glycolysis, which require actual concentration rather than a relative redox ratio along a metabolic trajectory, will need to take this dark species effect into account.

The QSSQ phenomenon in tryptophan, like here, contained both rapid charge transfer quenching and relaxation of the environment; extensive QM-MM and analog studies were needed to tease those phenomena apart. We anticipate that the same will be true for NADH. More importantly, for biological relevance, it will be useful to learn if any of the other significant NADH hosts in either oxidative phosphorylation or glycolysis can create a significant cryptic reserve of dark states. This seems unlikely, but quantum yield and lifetime studies (“apparent radiative rate”) of all main cycle enzymes and complex I–V should be carried out to search for any quantum yield defects that might motivate fuller upconversion studies.

More widespread studies of other metabolic enzymes may be needed to confirm this disparity. We suspect that the hemes of ETC (electron transport chain) enzymes might, for example, provide some pathways for the fast FRET (fluorescence resonance energy transfer) deactivation for some of their NADH sites. The relative contribution of ETC vs other bound sites remains to be established. Further, there are known “initial brightness”-derived dark contributions to QSSQ in tryptophan that even 300 fs resolution does not extract;²⁸ hence, we may yet find further dark populations in NADH as well. Although NADH is small compared to a typical dehydrogenase protein, it is large enough to have segmental flexibility, and numerous potential quenching mechanisms are available. Studies of modified forms of the flexible coenzyme will be of value in this regard. These include acetylpyridine NADH, pyridine aldehyde NADH, deamino NADH, alpha NADH, and N-methyl nicotinamide. Future steady-state and time-resolved fluorescence studies of these and other NADH derivatives should help to elucidate the actual quenching mechanisms.

In summary, the typical Rossman-fold dehydrogenase binding site abolishes a key self-quenching of NADH fluorescence that is relevant to the FLIM analysis of metabolism.