Ultrafast excited-state proton transfer dynamics in dihalogenated non-fluorescent and fluorescent GFP chromophores

The green fluorescent protein (GFP) discovered in the jellyfish Aequorea victoria has been widely used as a fluorescence marker in bioimaging applications.^1–3 The pertinent mechanistic research has facilitated the development of a broad library of fluorescent proteins (FPs) with versatile functions for various imaging and sensing applications.^4,5 Despite their diverse properties, many FPs share the same chromophore motif, p-hydroxybenzylidene-imidazolinone [p-HBDI, Fig. 1(a)], which highlights the importance of sidechain modifications and protein environment-chromophore interactions in governing the FP functions. Notably, in contrast to an intact protein chromophore undergoing efficient excited-state proton transfer (ESPT) with a high fluorescence quantum yield (FQY),^4,6,7 the synthetic chromophore p-HBDI exhibits distinctly different excited-state dynamics in solution, leading to no fluorescence. It has been established that the FQY loss in solution is caused by facile torsional motions around the methine bridge of the chromophore; however, the exact mechanism is still unclear. Two mechanisms centered on the C=C isomerization coordinate have been proposed: one-bond-flip (C=C torsion) and hula-twist (concerted C=C and C—C torsion).^8–11 This nonradiative decay is rather fast (up to a few picoseconds), and the excited-state dynamics are essentially dominated by the S₁/S₀ internal conversion instead of an ESPT reaction followed by fluorescence as prevalent in the protein matrix.

In essence, the competition between ESPT and other decay pathways determines the photophysics/photochemistry of the protonated p-HBDI in solution and GFP. The pK_a^* (photoacidity or excited-state acidity) of p-HBDI in aqueous solution is predicted to be ∼2.1 by the Förster equation, hence classified as a weak photoacid.¹² It corresponds to a 2–3 orders of magnitude slower ESPT process with respect to torsional motions.^10,13 Baranov and co-workers reported that the conformationally locked GFP chromophores resemble GFP in terms of ESPT, electronic transitions, and FQY.¹⁴ The borylation-locked chromophore, p-HOBDI-BF₂ [see Fig. 1(a) for related compounds], exhibits ESPT to water with a rate constant of 0.45 ns⁻¹. Observing such a slow ESPT process is enabled by conformational locking that extends the excited-state lifetime. The internal locking greatly suppresses both free C—C rotation and C=C isomerization [see Fig. 1(a)] and hence leads to a high FQY in solution comparable to GFP. This case differs from another modified p-HBDI analog, p-LHBDI, where only the C—C bond is exolocked such that C—C rotational motion is restricted. As a result, p-LHBDI does not experience a significant FQY boost and remains nonfluorescent (FQY ∼ 10⁻⁴).¹⁵ These studies support the double-bond isomerization as a major decay pathway. We note that the pK_a value of p-HOBDI-BF₂ (∼6.4) (Ref. 14) with structural restraints is lower than those of p-HBDI (∼8.5) (Ref. 16) and p-LHBDI (∼8.2) in aqueous solution. This could be ascribed to the torsional degrees of freedom of non-BF₂-locked compounds that lead to an enhanced entropic contribution (ΔS⁰ < 0, hence larger |TΔS⁰|) to the free energy change (ΔG⁰) in the ground-state PT reaction (deprotonation in this case).^17,18

Recently, we reported the fluorinated and internally locked p-HBDI analogs that are highly fluorescent superphotoacids.^18,19 They undergo ESPT not only in water but also in nonaqueous solvents such as alcohols. In methanol, the elucidated inhomogeneous dynamics consist of three parallel ESPT pathways: direct, solvation-controlled, and rotational diffusion-controlled across a broad time regime from subpicosecond (sub-ps) to hundreds of picoseconds. The nanosecond (ns) excited-state lifetimes of these locked molecules play an important role in the observation of slower ESPT processes, since otherwise a slower ESPT may be truncated or overwhelmed by faster nonradiative decay pathways. However, little light has been shed on ESPT dynamics of these superphotoacids in aqueous solution, and several intriguing questions arise that need to be addressed. First, how does dihalogenation near the proton dissociation site affect the ESPT rate? Second, how would the conformational locking or reduced nonradiative decay modulate the ESPT pathways?

To address these questions, we modified p-HBDI into nonfluorescent and fluorescent superphotoacids to enable and disable the ultrafast isomerization-induced decay, respectively. The superphotoacidity is achieved by difluorinating and dibrominating the phenolic ring at the sites adjacent to the phenolic hydroxyl group. Fluorescence is turned on by conformationally locking p-HBDI through borylation [Fig. 1(a)]. The four molecules under study are labeled 2F, 2Br, L2F, and L2Br, where “L” denotes the locked version. Halogenation weakens the O—H bond and decreases the pK_a by several units from 8.5 in p-HBDI to 4–6 in these photoacids (Table I, also see titration curves in Fig. S1 of the supplementary material). The dibrominated compounds exhibit lower pK_a than the difluorinated counterparts, and their pK_a^* values have the same ordering on the basis of the Förster equation.^14,19 The locked photoacids show slightly lower pK_a than the unlocked counterparts due to the entropic contributions (ΔS⁰), and the small difference is likely because the dihalogenation-induced enthalpy difference (ΔH⁰) dominates the free energy change (ΔG⁰) over the entropy term (−TΔS⁰), which is also observed in our previous work.¹⁸ Furthermore, the locked compounds (L2F and L2Br) are highly fluorescent (FQY ∼ 0.8), while the unlocked compounds (2F and 2Br) are essentially nonfluorescent (FQY ∼ 10⁻⁴).^18,19

Upon varying pH, two forms are present in L2F and L2Br, attributed to the neutral and anionic forms [Figs. 1(d) and 1(e)], while 2F and 2Br show a cationic form due to N-protonation at the imidazolinone ring at very low pH in addition to the neutral/anionic forms at the phenolic moiety [Figs. 1(b) and 1(c)]. In the absorption spectra, the cation peak is red-shifted with respect to the neutral form. This result is rationalizable by the photobase nature of the imidazolinone moiety at the π-conjugated N site for which its protonated state lies lower in energy than its unprotonated state.²⁰ The emission spectra show that the excitation of neutral and/or cationic states results in ESPT for all four chromophores. A similar ESPT process is also observed in acetonitrile, which confirms the superphotoacidity of these compounds (Fig. S3).^14,19 In particular, the emission spectrum after excitation of the cationic form has a dual-band profile, with the higher-energy band clearly attributable to the cationic emission that is redder than the neutral emission [Figs. 1(b) and 1(c)]. The lower-energy band is likely due to the zwitterion rather than the anion because it is to the red side of the anionic emission band [see Table I and Figs. S2(a) and S2(b)]. More importantly, the photoinduced intramolecular charge transfer (ICT) from the phenolic to imidazolinone moiety of the p-HBDI backbone^18,21 would lead to an increase in photobasicity at the imidazolinone N site, rendering the protonation even more robust with a further stabilized zwitterionic emission state.

To elucidate the ESPT dynamics of these fluorescent and nonfluorescent chromophores in aqueous solution, time-resolved femtosecond transient absorption (fs-TA) spectroscopy²² was employed by the excitation of the cationic form of 2F and 2Br and the neutral form of L2F and L2Br. Details about our ultrafast spectroscopic setups and fs-TA experiments using the femtosecond 400 nm pump and supercontinuum white light probe can be found in some previous reports.^19,23–25 The neutral form excitation was not chosen for 2F and 2Br due to their low solubility at the corresponding pH conditions while the resultant weak absorption band between 350 and 400 nm does not support an effective excitation by a 400 nm pulse (i.e., from second harmonic generation of the 800 nm laser fundamental output).^7,26 In comparison, the cationic form with better solubility in water absorbs redder than the neutral form, therefore making the 400 nm excitation viable. Besides, Meech and co-workers reported time-resolved fluorescence dynamics for the difluorinated p-HBDI (termed 2F in this work) and showed that the cationic and neutral excitations yield very similar ESPT dynamics, reflective of their similar photoacidity.²⁷ 

Notably, fs-TA spectra of 2F, 2Br, L2F, and L2Br in Figs. 2(a), 2(b), 3(a), and 3(b) all feature a stimulated emission (SE) band conversion from blue (∼450–480 nm) to red (∼520–530 nm), characteristic of ESPT. This can be verified by comparing the SE band positions to steady-state emission spectra [Figs. 1(b)–1(e)]. It appears that starting from the reactant species (i.e., the photoexcited acid form or A^*), the locked compounds (L2F and L2Br) have slower ESPT rates than the unlocked compounds (2F and 2Br). The SE band of the photoproduct (i.e., the conjugate base or B^*) reaches its maximum intensity within the first picosecond for the unlocked superphotoacids (Fig. 2), while the locked ones reach the maximum at a few tens of picoseconds (Fig. 3). The subsequent B^* decay dynamics in 2F and 2Br are dominated by the torsional coordinates that lead to picosecond excited-state lifetimes [see Figs. 2(a) and 2(b)]. Meanwhile, this SE band decay is accompanied by the rise of a positive absorption band at ∼480 nm that decays afterward on the tens of picoseconds timescale. The origin of this transient band cannot be the trans isomer as the C=C bond isomerization usually leads to a long-lived trans photoproduct that needs to overcome a high barrier of reverse isomerization in the ground state (S₀).^11,28 Due to the passage through a peaked S₁/S₀ conical intersection with branching into two conformational states,^25,29 the hot ground state (HGS)²⁴ of the zwitterionic cis form is supported by the red-shifted peak at ∼480 nm with respect to the equilibrated ground-state absorption peak of the chromophore anion [∼420 nm, Figs. 1(b) and 1(c), and Table I] and zwitterion [not available from the experimental spectra but predicted to be 436 nm for 2F and 442.5 nm for 2Br by our time-dependent density functional theory (DFT) calculations at the B3LYP level with 6-311G+(d,p) basis sets].³⁰ In contrast, L2F and L2Br exhibit much lengthened B^* decay dynamics on the nanosecond time scale due to the suppression of facile torsional motions [Figs. 3(a) and 3(b)].

Global analysis of the fs-TA spectra dissects the ESPT dynamics via the evolution-associated difference spectra (EADS) shown in Figs. 2(c), 2(d), 3(c), and 3(d), wherein three and five components under a sequential scheme were found to produce the best fit for the unlocked and locked superphotoacids, respectively. For 2F and 2Br, the first EADS exhibits dual SE peaks at ∼460 and 520 nm, characteristic of A^* (protonated at —OH) and B^* species, respectively [black trace, Figs. 2(c) and 2(d)]. This indicates the ultrafast formation of B^* within the cross-correlation time (∼140 fs) following A^* excitation.^18,19 The A^* SE band diminishes within 200 fs and converts to a B^* SE band. Since a dominant species could be tracked at a key probe wavelength,^22,31 the “local” TA dynamics at ∼520 nm reveal the B^* population rise component of 90 fs for 2F [Fig. 2(e)] and 150 fs for 2Br [Fig. 2(f)], comparable to the decay at 460 nm (A^*) but faster than the early-time decay at ∼480 nm. As the B^* rise is a better indicator for ESPT, the larger time constant retrieved from global analysis (i.e., 140 fs for 2F and 200 fs for 2Br) suggests that it goes beyond a directional two-state model and other nonradiative decay pathways depopulate A^* in parallel with ESPT.

Further corroborating pieces of evidence are as follows: (1) The presence of A^* and B^* fluorescence upon excitation of A indicates that some A^* population is trapped and decays through radiative emission despite the highly efficient ESPT [Figs. 1(b) and 1(c)]. (2) The decay-associated difference spectra (DADS, Fig. S4) show that the ensemble-averaged A^* lifetime is ∼140 fs for 2F and 200 fs for 2Br, longer than their fastest ESPT time constant obtained by the B^* rise [Figs. 2(e) and 2(f)], in accord with various decay pathways from multiple A^* subpopulations with different solute-solvent environments (e.g., H-bonding configurations).^18,19 Notably, the difference between A^* lifetime and ESPT time constant may be reduced by our experimental cross-correlation time. The fluorescence up-conversion method used by Meech and co-workers yielded a larger difference between A^* decay (∼200 fs) and B^* rise (∼50 fs) dynamics of the same compound (2F here) with a sub-50 fs time resolution.²⁷ Therefore, the ESPT reaction of 2F and 2Br in aqueous solution is ultrafast with a sub-100 fs time constant and can be considered kinetically barrierless.^13,19,27,32

Following ESPT, the zwitterionic photoproduct or B^* decays via isomerization likely through a conical intersection to the HGS,^25,33 characterized by the red → blue EADS conversion with a lifetime of ∼1 ps (840 fs for 2F and 1.2 ps for 2Br).²⁵ The zwitterionic HGS could then relax to the ground vibrational state (in accord with an apparent blueshift of the HGS absorption band)²⁴ and undergo ground-state back proton transfer to regenerate the cationic form in water. These two processes may correspond to the ∼480 nm HGS band biexponential decay time constants of 4.3–4.9 ps and 55–59 ps, respectively [see the long-time decay of black traces in Figs. 2(e) and 2(f)].

In contrast, L2F and L2Br show more complex dynamics involving solvent activities. The first EADS [black trace, Figs. 3(c) and 3(d)] with a lifetime of 70–100 fs features an SE band at ∼460 nm, assigned to the A^* Franck-Condon (FC) region. The second EADS (red) exhibits a multiband profile with two dominant peaks at 485–490 and 505–510 nm. The former band is likely associated with the FC-relaxed A^* inferred from Kamlet-Taft analysis in our previous work,²¹ and hence, the vibrational cooling time within A^* is in the ∼100 fs range. The latter, however, is debatable as its peak wavelength lies between the relaxed A^* (∼485 nm) and B^* (∼530 nm). The departure from the A^* emission peak indicates that it undergoes certain degrees of O—H bond elongation or breaking from A^* species, which remains solvated by the largely unoptimized H-bonding matrix due to the ultrashort reaction time of ∼100 fs (lifetime of the initial EADS)^13,19,32 and can be denoted as I^*.

Previous photoacid studies invoked a contact ion-pair (CIP) as an intermediate to account for the multistep dynamics, although the terminology has continued to be discussed in the literature.^18,34,35 Superphotoacids provide valuable insights into the intermediate state because they are typically rate-limited by solvent reorientation as a result of the photoinduced dipole moment change from S₀ to S₁.^13,36 In the case of L2F and L2Br, the red to blue EADS conversion is characteristic of local solvent reorientation or proton hopping in water since the time constant of 680–890 fs largely matches the literature value of 0.5–1.5 ps.^37–39 However, the notion that the aforementioned I^* state characterized by the SE band at 505–510 nm is directly followed by further solvent reorientation, as the sequential model suggests, requires caution because global analysis with a sequential scheme may not represent true kinetics.¹⁸ In this case, the probe-wavelength-dependent dynamics can provide complementary insights into ESPT pathways, especially when the retrieved time constants for ultrafast electronic dynamics match to a large extent.

In Figs. 3(e) and 3(f), the multiexponential fits at 485, 510, and 530 nm represent ultrafast dynamics of the FC-relaxed A^*, I^*, and the equilibrated B^* states, respectively. After an initial ∼100 fs rise, the A^* SE band (at 485 nm) decay shows a solvent reorientation stage (2.2 ps for L2F and 1.6 ps for L2Br) and a longer time constant of 3.5–10 ps owing to the increase in an adjacent excited-state absorption band of B^* (Fig. 3). Interestingly, the I^* SE band (at 510 nm) features an initial ∼400 fs rise and a slower decay (7.2 ps for L2F and 3.1 ps for L2Br) that matches one component of the B^* SE band rise dynamics [6.1 ps for L2F and 3.1 ps for L2Br, Figs. 3(e) and 3(f)], while missing the local solvent reorientation component on a shorter timescale (670–920 fs).^13,37–39 This result is corroborated by the DADS analysis (see Fig. S4 of the supplementary material) showing a more defined SE band ascribed to I^*, which has characteristic lifetimes of 7 and 3.2 ps for L2F and L2Br, respectively, and can be further validated by the target analysis (see below).

These results suggest that two distinct ESPT pathways are present for L2F and L2Br in water. One is the local solvent reorientation-controlled ESPT pathway, which can be considered as the “standard” dynamics for superphotoacids.^13,19 Note that the CIP is not invoked here due to the absence of a structurally distinct state from the relaxed A^* (emission at 485–490 nm). The other pathway is mediated by an intermediate (I^*) state that energetically lies between A^* and B^* states and is formed within the first ∼400 fs [see Figs. 3(a)–3(d)], which then undergoes a process longer than local solvent reorientation time to convert to B^*. Due to the redder wavelength of the I^* SE band with respect to the relaxed A^* SE band, I^* species has acquired B^* or anionic character to some extent, and its formation is favored by a relatively optimized local H-bonding configuration such that only small-scale solvent motions are required. The I^* decay time constant (3–7 ps) is consistent with the Debye relaxation of water (∼8 ps),³⁷ which usually involves long-range solvent reorganization or H-bond switching between water molecules beyond the first solvation shell. This process further stabilizes the solute, manifested as a Stokes shift of the SE band that was observed in L2F and L2Br [see curved arrows in Figs. 3(a) and 3(b)]. Due to the lengthened excited-state lifetimes in these locked compounds, the ESPT processes can be effectively completed and no discernible A^* fluorescence is observed [see Figs. 1(d) and 1(e), Table I]. The last two EADS characterize the B^* decay dynamics with time constants of 86 ps and 3.1–5.4 ns, which are associated with major nonradiative and radiative pathways, respectively [green and magenta traces in Figs. 3(c) and 3(d)]. In accord with this kinetic scheme, target analysis was performed (see Fig. S5), where the branching pathways occur from the Franck-Condon region. Reasonable species spectra are yielded when the I^*-mediated ESPT pathway is greater than ∼70% and 75% weight for L2F and L2Br, respectively.

In aggregate, the unlocked superphotoacids (2F and 2Br) exhibit direct and ultrafast ESPT, whereas their locked counterparts (L2F and L2Br) are more rate-limited by solvent motions and exhibit an overall slower ESPT reaction. Meanwhile, thermodynamics predicts L2F and L2Br to be more exergonic than 2F and 2Br as reflected by their pK_a^* values (Table I). This counterintuitive phenomenon is similar to the Marcus inverted region where the electron-transfer rates slow down as the reaction exergonicity increases.⁴⁰ It essentially correlates with an increased activation energy when the driving force is greater than the solvent reorganization energy. A similar theory could be applied to proton transfer as an appreciable amount of work has been performed through quantum treatment of a proton.^41–44 However, theoretical efforts on the photoinduced proton transfer remain scarce with respect to the thermally driven proton transfer.⁴⁵ We expect that future advances in computational chemistry could provide more insights into the ESPT dynamics for these contrasting GFP chromophore derivatives in this work. Moreover, isotopic labeling of these superphotoacids could shed more light.¹⁹ From a classical perspective, we speculate that the structural difference between unlocked and locked compounds may result in intrinsically different H-bonding geometries near the chromophore, wherein the solvent reorientation barrier is smaller for 2F and 2Br than for L2F and L2Br upon photoexcitation; hence, ultrafast ESPT on the ∼100 fs time scale is only observed in 2F and 2Br. More importantly, structural restraints in the locked compounds effectively limit their phase space and conformational flexibilities. As a result, so after A^* relaxation or I^* formation within ∼400 fs (see Fig. 4), two distinct solvent reorientation phases are resolved on the ∼1.5 ps and 7 ps time scales involving the first solvation shell and beyond, which substantiate a multistaged dynamic portrait of locked superphotoacids (with A^* subpopulations) undergoing ESPT surrounded by labile water molecules.

Furthermore, the difference between difluorination and dibromination is manifold in the context of ESPT and excited-state energy dissipation. First, bromination leads to lower pK_a and pK_a^* values than fluorination (Table I), opposite to the ordering of their electron-withdrawing abilities (i.e., —F > —Br). This is usually considered as a consequence of the interplay between the electron-withdrawing effect and intramolecular H-bonding that connects —OH to halogens in the ortho-substituted phenols.^46,47 Regarding ESPT dynamics, the unlocked compounds exhibit “inverted” kinetic behavior (2F is faster than 2Br), while the locked compounds show “normal” behavior (L2Br is faster than L2F), which could be a result of the interplay between the electronic distribution and nuclear motions. Second, bulkier halogens exert opposite effects on energy-dissipating pathways in the unlocked and locked compounds. In the unlocked case, bromination slows down the isomerization process (840 fs in 2F vs 1.2 ps in 2Br, Fig. 2) and hence leads to a higher FQY than fluorination (Table I). In contrast, bromination causes a tiny FQY drop in reference to fluorination in the locked case, which could be due to an increase in the size-dependent nonradiative decay pathways such as collision or thermal cooling.^23,48,49

Deeper mechanistic insights into the ESPT and energy-dissipation pathways due to structural modifications can be afforded by time-resolved vibrational spectroscopy as it directly reports on the nuclear motions involved in these processes.^26,31,32 For example, distinct structural dynamic features of the transient I^* state in L2F and L2Br could delineate the bifurcated ESPT pathways. We plan to employ tunable femtosecond stimulated Raman spectroscopy (FSRS)^31,50 to further study the superphotoacids of this work with a focus to capture key vibrational motions that play functional roles during the photocycle (Fig. 4).^7,19,23,25 An experimental advantage of tunable FSRS is that both the equilibrium⁵¹ and nonequilibrium^19,23,25 vibrational signals of interest can be enhanced by properly tuning the resonance conditions. As a preview, we present the ground-state FSRS spectra of four superphotoacids (Fig. 1) in the cationic, neutral, and anionic forms, in comparison to the parent molecule p-HBDI (Fig. S6), laying the foundation for future nonequilibrium structural characterization of these nonfluorescent and fluorescent compounds.^18,19,25

In conclusion, we investigated the ESPT dynamics of superphotoacids engineered from the GFP chromophore p-HBDI in aqueous solution. By strategic modification and systematic ultrafast characterization of the excited-state decay pathways via chemical synthesis and fs-TA spectroscopy, our results show that (1) the ESPT reaction rates display an inverted relationship between thermodynamics and kinetics and (2) the conformationally locked fluorescent superphotoacids undergo two ESPT pathways with active solvent engagement, whereas the unlocked nonfluorescent superphotoacids exhibit essentially barrierless and more homogeneous ESPT dynamics with little solvent participation. Further details could be provided by advanced modeling and excited-state FSRS, which are currently underway in our laboratory. Since these rationally designed and engineered^31,52 dihalogenated GFP chromophore derivatives retain a small size and are hydrophilic, they can be used in living systems. Moreover, their photoacidity and ESPT properties elucidated in this work could enable their optogenetic applications such as opening acid-sensitive ion channels (ASICs)⁵³ in live cells under light irradiation.

See the supplementary material for chemical synthesis methods of the unlocked and locked dihalogenated compounds with representative NMR spectra, pK_a titration curves for 2F, 2Br, L2F, and L2Br, second-derivative analysis of the fluorescence spectra in aqueous solution and acetonitrile, DADS and target (SADS) analysis of the fs-TA data, and the ground-state FSRS data of p-HBDI and its four derivatives in solution.

This work was supported by the U.S. NSF CAREER grant (No. CHE-1455353 to C.F.) and Russian Science Foundation (Grant No. 18-73-10105 to M.S.B.). The Wei Family Private Foundation Scholarship to C.C. is appreciated. We thank Dr. Longteng Tang and Taylor Krueger for helpful discussions.