Universal quenching of common fluorescent probes by water and alcohols

Although biological imaging is mostly performed in aqueous media, it is hardly ever considered that water acts as a classic fluorescence quencher for organic fluorophores. By investigating the fluorescence properties of 42 common organic fluorophores recommended for biological labelling, we demonstrate that H2O reduces their fluorescence quantum yield and lifetime by up to threefold and uncover the underlying fluorescence quenching mechanism. We show that the quenching efficiency is significantly larger for red-emitting probes and follows an energy gap law. The fluorescence quenching finds its origin in high-energy vibrations of the solvent (OH groups), as methanol and other linear alcohols are also found to quench the emission, whereas it is restored in deuterated solvents. Our observations are consistent with a mechanism by which the electronic excitation of the fluorophore is resonantly transferred to overtones and combination transitions of high-frequency vibrational stretching modes of the solvent through space and not through hydrogen bonds. Insight into this solvent-assisted quenching mechanism opens the door to the rational design of brighter fluorescent probes by offering a justification for protecting organic fluorophores from the solvent via encapsulation.

solvents were carried out using the relative method by integration of the fluorescence spectra and using the given fluorophore in H2O as a relative standard. Reference dyes were used as received without further purification, dissolved at millimolar concentration in 0.1 M NaOH (F27), absolute ethanol (R6G, R101, Ox1), H2O (R6G), or DMSO (HITCI), and stored at 4 °C.
Stock solutions of the synthetic target dyes at 0.5-10 mM concentration were prepared in anhydrous DMF or DMSO and stored at -20 °C until used. Target fluorophores and reference dyes were freshly diluted into the solvent of interest directly before the measurement.
Concentrations were adjusted to similar absorbance values at the excitation wavelength, and the absorbance values of the absorption band corresponding to the S0-S1 transition were kept below 0.1. All measurements were carried out in air-saturated solutions.
Fluorescence quantum yields (Ffl) of the target fluorophores were calculated according to the relationship S1 (S1) In this equation, the subscripts x and s indicate the target fluorophore and the reference dye, respectively, F(l) denotes the wavelength-dependent intensity of the corrected fluorescence spectrum which is integrated over the whole emission range, and n is the refractive index of the solvent. The term B(lex) represents the fraction of incident light absorbed by the sample and is given by (S2) where A(lex) is the absorbance of the sample at the excitation wavelength.
Fluorescence lifetime measurements. Fluorescence lifetime measurements were performed using the time-correlated single-photon counting (TCSPC) technique under magic angle conditions. S2 Lifetimes in H2O and D2O were measured using excitation at 470 nm (PicoQuant LDH-P-C-470 pulsed laser diode) or 635 nm (PicoQuant LDH-P-C-635B). Excitation light in the fluorescence pathway was removed using a 500 nm (Semrock BLP01-488R-25) or a 650 nm (Semrock BLP01-635R-25) long-pass filter. The instrument response function (IRF) was around 210 ps with 470 nm excitation and 660 ps with 635 nm excitation. Alternatively, fluorescence lifetime measurements were performed on a setup using a tuneable pulsed picosecond excitation source (NKT Photonics SuperK EXTREME). The fluorescence emission from the sample was collected into an optical fibre, spectrally filtered using a spectrograph (Horiba Triax 190) and detected on a photomultiplier tube (PicoQuant PMA 192C). The wavelength-dependent time resolution of this instrument was 170-200 ps. Both setups agree on the measured excited-state lifetimes within the experimental uncertainty.
All fluorescence lifetimes were extracted from the measured traces by iterative reconvolution of a trial function (single exponential or sum of exponential functions when required) with the measured IRF. In the case of multiexponential decays (Cy dyes), S3 the average lifetime was calculated from the amplitude average of the exponential components. S4 The uncertainty on fluorescence lifetimes is estimated to ± 0.05 ns.

Photophysical data analysis
Standard photophysical relationships were used to determine the radiative rate (krad) and the sum of non-radiative (knr) and solvent-assisted quenching rate (ks) from the fluorescence quantum yield (Φfl) and the excited-state lifetime (τS1). The rate constant for the depopulation of the excited state of a standard organic dye (kS1) in a pure solution may be written as the sum of radiative and nonradiative deactivation processes: (S3) where krad = Φfl·kS1 is the radiative rate constant, knr a first-order nonradiative decay rate constant taking into account all intrinsic nonradiative decay processes (internal conversion, intersystem crossing, etc.), and ks a rate constant representing the nonradiative decay selectively induced by the solvent. In order to determine the solvent quenching rate constant ks for ATTO655 in different solvents, we first calculated krad in all solvents from the fluorescence quantum yield and excited-state lifetime, and postulated that ks = 0 in acetonitrile-d3, the solvent in which the longest excited-state lifetime was measured. With this we obtained a value of knr = 6.8(8)·10 7 s -1 which we assumed constant in all solvents in order to calculate ks.
For all other dyes, ks was determined from the excited-state lifetimes as follows. Under the reasonable assumptions that (1) krad is independent of the solvent for these well-behaved dyes, and that (2) knr does not significantly change between a given protonated solvent (SH,e.g. H2O) and its deuterated analogue (SD, e.g. D2O), the difference in excited-state decay rate constants k S1 = τ S1 ( ) −1 = k rad + k nr + k s between a protonated and a deuterated solvent directly yields the quenching rate constant for this solvent, within the limit that quenching by the deuterated solvent is negligible: (S4) Instead of measuring the photophysical properties of the dye solely in pure solvents, one can also treat the protonated solvent as a quencher which can be added in different amounts to a solution of fluorophore in a deuterated solvent. In the case of purely dynamic quenching, the excited-state decay rate constant of a fluorophore F depends on the concentration of protonated solvent (the quencher) and can be expressed as (S5) where kq is the bimolecular quenching rate constant for the reaction (S6) In the absence of quencher, the observed decay rate is simply (S7) Dividing eq. S5 by eq. S7 yields the Stern-Volmer expression for dynamic quenching: S5 By comparing equations S3 and S5, one concludes that This relationship has been experimentally verified in the present work for the dyes for which a Stern-Volmer titration was performed.

Decomposition of the H2O absorption spectrum
Because of interference patterns arising in the signal when using an empty reference cuvette, the absorption spectrum of pure H2O and pure MeOH was obtained using D2O as a reference.
Neither D2O nor MeOH-d4 absorb in any measurable manner over a path length of 1 cm in the investigated spectral region. S6, 7 We further made sure that the measured value for the molar absorption coefficient matched known literature values (e700 » 4·10 -5 cm -1 ·M -1 for H2O and e750 » 3·10 -4 cm -1 ·M -1 for MeOH). S7, 8 The spectrum of H2O was then decomposed on the wavenumber scale using a sum of 6 Gaussian functions and of a baseline component (l -4 dependence) using a custom written Matlab script. Inclusion of all 7 components was required in order to reproduce the data. The centres of the Gaussian functions were fixed at or close to known overtones and/or combination modes of the fundamental vibrational frequencies n1 (symmetric stretching), n2 (bending), and n3 (asymmetric stretching). The central frequencies of the Gaussians and their assignment are listed in Table S6.

FRET analysis
Transition dipole moment magnitudes were estimated from the relationship (S10) where h is Planck's constant, e the charge of the electron, me the mass of the electron, and n0 the central transition frequency in s -1 . In this equation, f is the oscillator strength which is obtained by integrating over the absorption band of interest on the wavenumber ( ̅ ) scale (in cm -1 ) using the following relationship: S9 (S11) The overlap integrals (Q) between the fluorophore emission and the solvent absorption were computed on the wavenumber scale using area-normalized spectra as required by eq. 5 of the main text. An integral Qi was calculated for every overlap between the emission spectrum of the dye and the i th component of the water absorption spectrum. In parallel, the coupling, Vi, between a fluorophore and the i th absorption band of one water molecule was estimated using the transition dipole magnitude for the i th absorption band: (S12) where V is expressed in cm -1 , µ in D, and d in nm.
From the coupling and the overlap integral for the i th absorption of water, the rate constant for excitation energy transfer by dipolar coupling between a fluorophore and the i th absorption band of one water molecule, kdip,i, could be determined ( Figure S16): where kdip is expressed in ps -1 , V in cm -1 , and Q in cm.
The total energy transfer rate constant between a fluorophore and one water molecule was obtained by summing over all water absorption bands: The total dipole-dipole resonance energy transfer rate constant, kFRET, was obtained by multiplying kdip by the total number of solvent molecules N (main text, eq. 7). For the sake of this computation the result of which is represented on a semi-log scale, N was kept constant (N = 86 in water, N = 50 in methanol, see main text).
The total coupling and total overlap integral ( Figure S17) were estimated by summing over all i components of the water absorption spectrum: The energy transfer rate constant, kFRET, is also defined in Förster's theory as: S17) where kD = 1/t 0 D is the decay rate constant of the excited donor fluorophore in the absence of energy transfer, d is the distance, and R0 is the Förster radius or critical quenching radius defined as the distance at which the energy transfer rate constant, kFRET, and the decay rate constant of the excited donor in the absence of energy transfer, kD, are equal. S10 R0 can be evaluated from the relationship S10 (S18) where FD is the quantum yield of the fluorophore in the absence of energy transfer and Jl is the spectral overlap integral expressed in units of M -1 ·cm -1 ·nm 4 , with In equation S19, FD is the area-normalized fluorescence spectrum of the donor fluorophore.
The values of R0 in Table S7 and Figure S13 were obtained by using FD = Ffl(D2O), k 2 = 2/3, and n = 1.332 as the index of refraction for H2O.
The energy transfer efficiency, FFRET, is directly related to kFRET and given by (S20) In the present case, kFRET was experimentally determined in water as ks = kS1(H2O) -kS1(D2O), and under the same assumption kD = kS1(D2O). Therefore, equation S20 can be rewritten as: (S21) By analogy, the following relationship holds true in methanol: Equations S21 and S22 were used to evaluate the energy transfer efficiency in Table S7 and Figure S18.

MD simulations
MD simulations of fluorophores in water were performed using GROMACS 2018.1 S11 and the AMBER03 forcefield. Point charges for atoms in the solute were determined from RESP fitting of the electrostatic potential from quantum calculations performed using Gaussian 09 S12 with the B3LYP functional and a 6-311G** basis set. The antechamber S13 and acpype S14 programs were used to generate topologies for the probes. The TIP4P model was used to simulate water, S15 and standard AMBER03 parameters were used for MeOH and EtOH.
Each solute-solvent system was simulated using 1 chromophore molecule in a cubic box with dimensions of 6 × 6 × 6 nm. The simulation boxes were generated by placing the chromophore at the centre of the box and adding water using the gmx solvate command or gmx insertmolecules for the other solvents. There were ∼7100 water molecules, 3305 MeOH, and 2273 EtOH molecules in each solvent system.
Integration was carried out using the Verlet leap-frog algorithm with a step size of 2 fs. Nonbonded interactions were calculated using a Verlet neighbour list with a cutoff radius of 14 Å.
Long-range electrostatics were calculated using the particle-mesh Ewald method. S16 Hydrogencontaining bonds were constrained using the LINCS algorithm. S17 The temperature was 293.15 K and was controlled using the modified Berendsen thermostat S18 with a relaxation time of 0.5 ps.
Equilibration was accomplished using the following procedure. First, an energy minimization procedure was carried out using the steepest-descent method with maximum force stopping point of 500 kJ·mol −1 ·nm −1 . Next, three consecutive 500 ps simulations were carried out in the NVT, NPT, and NVT ensembles. The pressure in the NPT simulation was set to 1.013 bar and regulated using the Berendsen barostat with a relaxation time of 5 ps. Following equilibration, a 5 ns production simulation was carried out in the NVT ensemble and coordinates were saved at 1 ps intervals, giving 5000 configurations for later analysis.
Minimum-distance distribution functions, g(rmin), were then calculated according to standard literature procedures. S19, 20 These distribution functions are calculated with respect to the distance between the solvent centre of mass and the closest atom of the solute. Normalization was accomplished using a Monte Carlo procedure that generated randomly distributed centres of mass from which minimum-distance distributions were calculated. Solute structures for the normalization procedure were taken from the production simulations, and the number of randomly generated centres of mass per frame was the same as the number of solvent molecules in the simulation.
The average FRET orientation factor, ⟨k 2 ⟩, for energy transfer between ATTO655 and water was estimated by averaging individual k 2 values for all water molecules within 0.45 nm, the radius of the first solvent shell as determined by g(rmin). The orientation factor for the i th solvent molecule and the average are calculated according to: In these equations, N is the total number of first solvent shell waters in the 5000 saved simulation frames, and the vectors ' '⃗ and ⃗ are unit vectors lying along the transition dipole moments for the FRET donor (ATTO655) and acceptor (water) molecules. For ease of computation, ' '⃗ was taken to be a vector connecting the ATTO655 central N and O atoms of the oxazine core, and ⃗ connects the two hydrogens of water. Finally, '⃗ is a unit vector connecting the centres of mass of ATTO655 and the water molecule in question. These vectors are illustrated in Figure S15. We find that ⟨k 2 ⟩ = 0.66 with a standard deviation of 0.70. This average corresponds well with the isotropic prediction, which is not surprising given the large number of water molecules and the lack of strong specific interactions between ATTO655 and water. We performed the same evaluation for MeOH following the same procedure and using the O and H atoms of the OH group to define the vector ⃗ . Tables   Table S1. Steady-state photophysical properties of the investigated fluorophores in water and dye class to which they belong: absorption (labs) and emission (lem) maxima, S0-S1 energy gap in eV (DE00) and in nm (l00), molar decadic absorption coefficient at the absorption maximum provided by the manufacturer (e), and calculated S0-S1 transition dipole moment magnitude µF.