Jorge E.
Ramos-Sanchez‡§
a,
Yasser
Gidi‡¶
a,
Terri C.
Lovell§
a,
Liam
Payne||
a,
Aya
Sakaya
ab and
Gonzalo
Cosa
*a
aDepartment of Chemistry and Quebec Center for Advanced Materials (QCAM), McGill University, 801 Sherbrooke Street West, Montreal, QC H3A 0B8, Canada. E-mail: gonzalo.cosa@mcgill.ca
bDepartment of Chemistry, University of Toronto, Toronto, ON M5S 3H6, Canada
First published on 18th March 2025
Fluorophore signal and photostability are desirable when conducting single molecule fluorescence (SMF) studies. For cyanine dyes, this is typically achieved by quenching their excited triplet states via photoinduced electron transfer (PeT). However, to date, only a handful of electron donors or acceptors have been – empirically – identified, most of which do not show diffusion controlled PeT quenching rate constants (kPeT). Here we report a screening tool to rapidly explore the potential of thiol-based electron donors to serve as photostabilizers of Cy3, Cy5, and their bridged congeners Cy3B and Cy5B. Based on density functional theory (DFT) and utilizing Marcus theory of electron transfer, our method estimates in silico the activation free energy of PeT. A correlation function is then established between theoretical estimates and empirical values of kPeT acquired through transient absorption spectroscopy. The correlation function then enables screening of the photostabilizing potential of untested reagents from a kinetic standpoint. A new compound, glyceryl monothioglycolate (GMTG), was thus identified. SMF studies show the effectiveness of GMTG as a domestic use, non-toxic, non-volatile alternative to the widely utilized β-mercaptoethanol. Altogether, a predictive model is proposed toward discovering photostabilizers to achieve single molecule fluorescence experiments with enhanced photo- and signal stability.
Quenching of the excited triplet state enhances the signal intensity and typically prevents photodamage, resulting in increased photostabilities in SMF. Triplet state quenching may also yield a desired photoproduct, e.g. a non-absorbing adduct that can be subsequently photo-uncaged, of interest in SMLM (Fig. 1a).7–9 Carefully tuning the lifetime of the excited triplet state of the fluorophore has been instrumental to the many advances brought about by SMF and SMLM in biology, chemistry, physics, and materials science.10–13
Photoinduced electron transfer (PeT) has become the method of choice to quench fluorophores in their excited triplet state.13,14 Water soluble electron donors can be utilized at concentrations high enough to quench excited triplet states, while sparing the excited singlet state.15 While typically a counter oxidant may be needed in conjunction with the electron donor, when using thiol-based electron donors, such counter oxidants are not required. This is because newly formed thiyl radicals undergo rapid spin flip enabling back electron transfer (BeT) to occur in a spin allowed process within the solvent cage, before escape of newly formed radicals takes place (Fig. 1a).16–18 Notably, while the chemical space offers limitless possibilities for thiol-based electron donors, only a handful of these types of reagents have been identified to date. This is largely explained by the fact that the screening of these compounds is a highly laborious task, requiring bulk time resolved studies to determine the rate constants of PeT and single molecule studies to then validate the performance of such compounds.
To accelerate the discovery of novel and improved triplet state quenchers, here we report a density functional theory (DFT)-based method to estimate PeT activation energy and rate constant (kPeT) values for the quenching of the excited triplet state of the single-molecule fluorophores Cy3, Cy5 and their bridged congeners Cy3B and Cy5B.19 Our method relies on obtaining via DFT – utilizing Marcus theory of electron transfer – the parameters to calculate the activation energy (
, Fig. 1b), namely the electronic energy change
and the reorganization energy (λ) of PeT. We next correlate the theoretically estimated values of
with empirical values for the PeT rate constant (kobsPeT) obtained through bulk laser flash photolysis (LFP) studies (Fig. 1c). The correlation then enabled us to utilize predicted activation energy parameters to estimate rate constants of PeT for unknown compounds. In our work, we utilized popular electron donating triplet state quenchers (TSQs): β-mercaptoethanol (β-ME),16,20 ascorbic acid (AA),21n-propyl gallate (n-PG),15,22 and the vitamin E analogue Trolox (TX)11,23,24 in combination with four cyanine dyes (Cy3, Cy3B, Cy5 and Cy5B; Table 1).
a Values taken from ref. 25. b Determined with only 3 points. | |||||
---|---|---|---|---|---|
![]() |
k obsPeT | 3.54 ± 0.23 × 108 | 1.94 ± 0.18 × 107 | 9.44 ± 0.43 × 107 | 1.30 ± 0.39 × 106 |
![]() |
−0.46 | −0.11 | −0.22 | 0.04 | |
![]() |
0.40 | 0.55 | 0.49 | 0.60 | |
![]() |
k obsPeT | 4.41 ± 0.14 × 108 | 6.93 ± 0.23 × 107 | 2.54 ± 0.16 × 108 | 1.19 ± 0.06 × 107 |
![]() |
−0.51 | −0.16 | −0.27 | −0.004 | |
![]() |
0.35 | 0.50 | 0.45 | 0.55 | |
![]() |
k obsPeT | 1.17 ± 0.04 × 109 | 2.91 ± 0.97 × 108 | 1.19 ± 0.12 × 108 | 5.49 ± 2.19 × 107 |
![]() |
−0.42 | −0.07 | −0.18 | 0.08 | |
![]() |
0.32 | 0.47 | 0.41 | 0.52 | |
![]() |
k obsPeT | 7.64 ± 0.65 × 108 | 1.14 ± 0.06 × 109 | 8.07 ± 0.72 × 108 | 5.99 ± 1.00 × 108 |
![]() |
−0.82 | −0.47 | −0.58 | −0.32 | |
![]() |
0.18 | 0.30 | 0.26 | 0.34 | |
![]() |
k obsPeT | 1.37 ± 0.13 × 109 | 5.42 ± 0.16 × 108 | 9.15 ± 0.30 × 108 | 1.22 ± 0.04 × 108 |
![]() |
−0.55 | −0.20 | −0.31 | −0.05 | |
![]() |
0.30 | 0.44 | 0.39 | 0.49 | |
![]() |
k obsPeT | 3.77 ± 0.77 × 108 | 7.78 ± 0.27 × 108 | 1.11 ± 0.06 × 109 | 3.52 ± 0.21 × 108 |
![]() |
−0.93 | −0.33 | −0.44 | −0.18 | |
![]() |
0.15 | 0.38 | 0.33 | 0.42 | |
![]() |
k obsPeT | 2.35 ± 0.08 × 108 | 1.73 ± 0.12 × 106 | 2.19 ± 0.10 × 107 | 1.61 ± 0.43 × 105 |
![]() |
−0.33 | 0.02 | −0.09 | 0.15 | |
![]() |
0.44 | 0.60 | 0.54 | 0.65 |
The excellent correlation between empirical kobsPeT values and theoretical values (while not
) positions our predictive model as a suitable tool to rapidly evaluate the potential of existing and newly designed compounds to serve as photostabilizers of cyanine dyes. As a case example, our method screened a range of compounds, primarily based on the values of kobsPeT, but also introducing solubility and pKa as additional selection parameters. Improved photostability is predicted and shown with GMTG, a low vapor pressure alkyl thiol-based compound – of ample use in the cosmetic industry for perms26 – that surpasses the performance of β-ME, providing an odorless, inexpensive, and safe-to-operate PeT-based photostability and photoswitching agent.
The predictive model proposed to select photostabilizers, while centered on cyanine dyes, should be of general applicability to other fluorophores utilized in SMF experiments, toward conducting experiments with enhanced photo- and signal stability.
![]() | (1) |
To obtain and λ, we next performed DFT calculations using the B3LYP functional with the polarizable continuum model (PCM, water) for solvation. Here, we parameterized the reactants separately using the Gaussian package v.16,29 as previously described by Buda30 (see Section S1.2, ESI,† for more information). We first optimized the cyanines in the excited triplet state at their minimum (Fig. 1b, P1) using time dependent DFT (TD-DFT). We “froze” the molecule with the solvent charges using the “NONEQ” keyword formalism in Gaussian and added an electron to the cyanine to perform a vertical transition (Fig. 1b, P2). The molecular geometry was then optimized to accommodate for the new electron (Fig. 1b, P3). A similar treatment was performed on the TSQ, where the ground state was optimized (Fig. 1b, P1), an electron was removed from the molecule (Fig. 1b, P2) and then it was optimized to accommodate for the loss of the electron (Fig. 1b, P3). Energies were calculated from the three different points for each cyanine and TSQ independently (see Tables S1 and S2, ESI,† for values).
was obtained from the optimized structures (eqn (2)) for each donor–acceptor pair, while reorganization energies (λ) were calculated from the vertical excitations and the optimized product (eqn (3)). Finally, the activation energy
for each combination was obtained according to eqn (4) where both free energy and reorganization energy were combined from both the fluorophore and the quencher. Results are summarized for each dye–TSQ combination in Table 1.
![]() | (2) |
λ = P3 − P2 | (3) |
![]() | (4) |
![]() | (5) |
![]() | (6) |
![]() | (7) |
![]() | (8) |
![]() | (9) |
The correlation between kobsPeT and the activation energies calculated for each of our 16 different combinations is shown in Fig. 2a. Consistent with the reaction evolving from activation to diffusion controlled as the value of drops, kobsPeT is seen to increase in a parabolic manner and ultimately plateau as it reaches the value for the rate constant of diffusion (kd). Here, a break point around 0.4 eV signals the transition from an activation- to a diffusion-controlled reaction. Eqn (8) provides a single parameter (Z, eqn (9)) to fit the functional relationship between experimental kobsPeT and theoretical
. Compounds serving as TSQs are thus sought where the
values – calculated for their electron transfer to photoexcited cyanine dyes in the excited triplet state – ensure diffusion controlled quenching rate constants.
![]() | ||
Fig. 2 Empirical PeT rate constant (kobsPeT) as a function of different energy parameters: (a) the theoretical activation energy of the PeT process ![]() ![]() |
Analysis of Fig. 2a additionally shines light on the behaviour recorded for TSQs with different cyanines. It thus illustrates that β-ME, a well-known photostabilizer used with Cy3 and Cy5, is ineffective when used in conjunction with the bridged cyanine forms, Cy3B or Cy5B, as there is a significant activation barrier to PeT. While the values of kobsPeT recorded with Cy3–βME and Cy5–βME are on the order of 108 M−1 s−1, the values for Cy3B–βME and Cy5B–βME drop by one and two orders of magnitude to ∼107 and 106 M−1 s−1, respectively. Likewise, n-PG and AA show the same trend, where kobsPeT follows the order Cy3 > Cy5 > Cy3B > Cy5B. In the case of TX, Cy5 and Cy3B invert the order, rendering Cy3 > Cy3B > Cy5 > Cy5B. For TX and AA, the diffusion-controlled regime is reached with Cy3, although most of the pairs are in the activation energy-controlled regime. Although all dyes exhibit a diffusion-limited rate with n-PG, its low solubility, and the high likelihood of radicals in the triplet manifold escaping and reacting elsewhere, poses a significant drawback for its universal applicability.
Importantly, while the correlation of kobsPeT with provided satisfactory results, correlations with calculated values for the free energy change of the reaction (
, see the values in Table 1) or with experimental values of
calculated through experimental values for their excited triplet state energy35 and redox potentials (see Section S1.6, ESI,† for the experimental procedure and Table S4, ESI,† for values) yielded poorer outcomes (see Fig. 2b and c).
![]() | ||
Fig. 3 Predicted kobsPeT for thiolates with cyanine dyes. Molecules are parametrized as presented in their thiolate form. Rate constants are calculated after computing the DFT obtained energies, by using the fitting parameters obtained in the correlation model. Numerical data are presented in Table S5 (ESI†). |
Three targets were ultimately chosen utilizing our predictive model and selection criteria: ergothioneine (Erg), 2-thiol histidine (2-TH) and glyceryl monothioglycolate (GMTG). Erg and 2-TH were recently reported by our group as cyanine photostabilizers.25 Consistent with our expectations, Erg exhibited diffusion-controlled values for kobsPeT with Cy3, Cy3B, Cy5, and Cy5B and provided superior photoprotection for these cyanine dyes.
Turning the attention next to GMTG, this compound was selected based on its low pKa (7.8),26 low toxicity and its wide range of predicted kobsPeT, facilitating the assessment of the model for a situation where a range of rate constants were predicted. The kobsPeT values for GMTG with all four cyanines were determined experimentally using LFP. The experimentally obtained and predicted kobsPeT values are summarized in Fig. 4a. The accurate prediction of kobsPeT is apparent from the inspection in Fig. 4b, showing the linear correlation between the calculated (Calc.) and experimental (Exp.) kobsPeT values obtained with GMTG, Erg, or 2-TH. Importantly, consistent with the previously reported rapid intersystem crossing observed within the geminate radical pair in the presence of newly formed thiyl radicals and leading to BeT before radical escape, transient absorption studies of Cy3B and Cy5B with GMTG showed negligible radical escape following PeT (Fig. S7, ESI†). The simultaneous decay of the excited triplet state and recovery of the ground state demonstrated efficient back electron transfer in the presence of GMTG as an electron donor, consistent with previous reports for β-ME17 and ergothioneine.25
![]() | ||
Fig. 4 (a) Calculated or experimentally determined kobsPeT for GMTG. (b) Calculated versus experimental log(kobsPeT) values for GMTG, Erg and 2-TH to show the accuracy of the reported computational model. (c) Single molecule image of 7.8 mM GMTG with Cy3 (scale bar is 5 μm) and (d) the corresponding histogram for survival time (a marker of photostability). (e) Average number of total photons obtained in single molecule experiments of Cy3 with 143 mM β-ME and 5.2 mM GMTG yielding equal thiolate concentration as estimated by their respective pKa values; at 7.8 mM GMTG ensuring equal quenching efficiency as that observed with 143 mM β-ME, and at an optimal GMTG concentration (20 mM); see Table S6 (ESI†). (f) Single molecule trajectories of Cy5 with 143 mM β-ME and 22.9 mM GMTG at 54 mW laser excitation power. |
GMTG performance was then tested in single molecule fluorescence studies with Cy3 and Cy5 (Fig. 4c–f). While the predicted kobsPeT rate constants for GMTG are smaller than those of β-ME, its pKa is more than two units lower, indicating that GMTG should achieve a similar triplet quenching to that of β-ME with significantly lower total thiol concentrations. We thus tested GMTG in experiments conducted at three different concentrations: (1) a GMTG concentration that ensures the thiolate concentration is similar to that present when working with 143 mM β-ME, as estimated by their respective pKa values; (2) a GMTG concentration that renders the same decay rate constant for the fluorophore excited triplet state (kT) as that obtained with 143 mM of β-ME, and (3) a GMTG concentration that ensures a two-fold reduction of the triplet lifetime as compared to β-ME. Cy3 and Cy5 were chosen based on their larger kobsPeT. Additionally, the former is expected to exhibit enhanced photostability, while the latter should display photoinduced blinking.
Single-molecule experiments were conducted with 18-mer single stranded DNA decorated with a single fluorophore dye and immobilized on the surface of a polyethylene glycol passivated glass coverslip.36 The fluorophores were excited under standard total internal reflection microscopy conditions where images were captured with an EMCCD camera (Fig. 4c, see also Section S1.7, ESI,† for more information). The fluorescence response was recorded over time for the different GMTG concentrations. We extracted the total number of photons, survival time, and intensity per frame for each single molecule from the intensity-time trajectory. Data from all molecules were combined to create a probability distribution histogram (Fig. 4d). The obtained data exhibit an exponential decay, enabling determination of the average number of photons and survival time through fitting. These data were then compared with those from β-ME.
The average number of total photons obtained per fluorophore with 7.8 mM total GMTG (∼3.8 × 105) was similar to that recorded with 143 mM β-ME (Fig. 4e). Thus, a ∼20-fold lower total GMTG concentration was required to achieve a comparable level of Cy3 photostabilization to that of β-ME, highlighting the potential of GMTG as an alternative to the latter. Improved results were obtained when increasing GMTG concentration by 2.56× (20 mM, see Fig. 4e). Here we note that with increasing GMTG concentration the intensity distribution shifted to smaller values (Fig. S8, ESI†). This shift can be attributed to partial quenching of the excited singlet states with increasing thiolate concentrations.
Single molecule studies with Cy5 in turn showed that GMTG leads to photoinduced blinking, as it has been shown for β-ME (Fig. 4f). As expected, given the precursor nature of the thiolate in forming the Cy5–thiol adduct,12,17 both the duration of the dwell-on time and accordingly the number of photons per blink got reduced as we increase the GMTG concentration (Table S7, ESI†). Importantly the number of photons per blink (∼2200 photons) and the intensity per unit time during a blink were similar regardless of the thiol source at comparable quenching rates (143 mM β-ME vs. 22.9 mM GMTG), highlighting the importance (an advantage) of a lower pKa for thiol-based switching agents. Notably, the lifetime of the dark adduct under our experimental conditions was one order of magnitude larger for GMTG than for β-ME, going from ∼11 s to ∼100 s for β-ME vs. GMTG, respectively. This discrepancy may reflect either a decrease in the extinction coefficient for the adduct, or in the acid catalyzed uncaging rate constant of the Cy5–GMTG adduct.17 The increased lifetime for the dark adduct would enable better duty cycles and enhanced resolution in SMLM for GMTG when compared with β-ME.
Our model is a valuable tool for rapid computational testing of potentially effective triplet quenchers. This strategy can avoid a lengthy synthesis and further screening through single-molecule methods for promising molecules, a resource- and labor-intensive task. Since both the dye and the quencher are modeled separately, parameters described herein can be directly applied to combinations with new triplet state quenchers after calculations for the latter are performed. Our method should be directly extended to structurally related cyanine fluorophores. We also anticipate that this strategy, with proper calibration, can also be extended to other dyes utilized in single-molecule studies, e.g. rhodamines.
A series of thiol-based compounds were selected though our screening method, where a versatile thiol of ample use by the general public was singled out, due to its low vapor pressure and toxicity, and desirable solubility and acidity. While this work highlights the opportunities brought by this compound toward single molecule fluorescence imaging and super-resolution, most importantly, it underscores the potential of the model to rapidly screen for optimal photostabilizers using the vast number of available compounds. It is also relevant to mention that the model is certainly applicable to all scenarios where PeT is a desired outcome in the encounter between an excited organic chromophore and an electron donor/acceptor quencher. While this work emphasizes the need for back electron transfer toward photostabilization and photoswitching, and thus the choice of thiol-based compounds as PeT electron donors, certainly similar ideas could be extended in, e.g., photoredox catalysis, where the desired outcome would be escape of newly formed transient species from a solvent cage, rather than back electron transfer.
Footnotes |
† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d4cp04794j |
‡ These authors contributed equally. |
§ Department of Chemistry and Biomolecular Science, University of Ottawa, Ottawa, ON K1N 6N5, Canada. |
¶ Department of Radiology, School of medicine, Stanford University, Stanford, California 94305, United States. |
|| Discipline of Internal Medicine, Faculty of Medicine, Memorial University of Newfoundland, St. John’s, NL A1B 3V6, Canada. |
This journal is © the Owner Societies 2025 |