Excited-state structural dynamics of a dual-emission calmodulin-green fluorescent protein sensor for calcium ion imaging

Significance

Fluorescent proteins (FPs) started their incredible, colorful journey in bioimaging and biomedicine with the extraction and purification of GFP from the Pacific jellyfish Aequorea victoria more than 50 years ago. Recently, an expanded palette of genetically encodable Ca²⁺-sensing FPs have paved the way to image neural activities and important biological processes where Ca²⁺ is the ubiquitous messenger. To unravel the molecular choreography of FPs engineered for visualizing Ca²⁺ movement, we study the embedded chromophore upon photoexcitation and monitor its subsequent excited-state structural evolution with femtosecond Raman spectroscopy. The vivid insights on H-bonding network and functional roles played by strategic mutations provide a deep understanding of excited-state processes in biology and will guide future bioengineering efforts toward better biosensors.

Abstract

Fluorescent proteins (FPs) have played a pivotal role in bioimaging and advancing biomedicine. The versatile fluorescence from engineered, genetically encodable FP variants greatly enhances cellular imaging capabilities, which are dictated by excited-state structural dynamics of the embedded chromophore inside the protein pocket. Visualization of the molecular choreography of the photoexcited chromophore requires a spectroscopic technique capable of resolving atomic motions on the intrinsic timescale of femtosecond to picosecond. We use femtosecond stimulated Raman spectroscopy to study the excited-state conformational dynamics of a recently developed FP-calmodulin biosensor, GEM-GECO1, for calcium ion (Ca²⁺) sensing. This study reveals that, in the absence of Ca²⁺, the dominant skeletal motion is a ∼170 cm⁻¹ phenol-ring in-plane rocking that facilitates excited-state proton transfer (ESPT) with a time constant of ∼30 ps (6 times slower than wild-type GFP) to reach the green fluorescent state. The functional relevance of the motion is corroborated by molecular dynamics simulations. Upon Ca²⁺ binding, this in-plane rocking motion diminishes, and blue emission from a trapped photoexcited neutral chromophore dominates because ESPT is inhibited. Fluorescence properties of site-specific protein mutants lend further support to functional roles of key residues including proline 377 in modulating the H-bonding network and fluorescence outcome. These crucial structural dynamics insights will aid rational design in bioengineering to generate versatile, robust, and more sensitive optical sensors to detect Ca²⁺ in physiologically relevant environments.

Green fluorescent protein (GFP) first emerged as a revolutionary tool for bioimaging and molecular and cellular biology about 20 years ago (1–3), and the quest to discover and engineer biosensors with improved and expanded functionality has yielded exciting advances. Recently, the color palette of genetically encoded Ca²⁺ sensors for optical imaging (the GECO series) has been expanded to include blue, improved green, red intensiometric, and emission ratiometric sensors (4–7). The GECO proteins belong to the GCaMP family of Ca²⁺ sensors that are chimeras of a circularly permutated (cp)GFP, calmodulin (CaM), and a peptide derived from myosin light chain kinase (M13) (8). The CaM unit undergoes large-scale structural changes upon Ca²⁺ binding as it wraps around M13. These changes, especially at the interfacial region where CaM interacts with cpGFP, allosterically alter the local environment of the tyrosine-derived chromophore and lead to dramatic fluorescence change in the presence of Ca²⁺ (9, 10). Because GCaMP and GECO proteins are genetically encodable, show sensitivity to physiologically relevant Ca²⁺ concentrations, and respond to Ca²⁺ concentration changes rapidly, they have gained increasing popularity for in vivo imaging of Ca²⁺ in neural and olfactory cells (11–13).

Among the engineered GECO proteins, GEM-GECO1 is an intriguing case with the serine–tyrosine–glycine (SYG) derived chromophore (4). Upon ultraviolet (UV) excitation it fluoresces green in the absence of Ca²⁺ but blue upon Ca²⁺ binding with a K_d of 340 nM. This dual-emission behavior is unique among ratiometric Ca²⁺-sensing FPs as well as the few reported pH-dependent dual-emission GFP variants, which typically require a threonine–tyrosine–glycine (TYG) chromophore to keep the nearby glutamate largely protonated (14, 15). Dual emission is particularly useful for imaging in vivo because the signal color change is a direct consequence of analyte concentration. There remains room to further improve GEM-GECO1 because it has a low quantum yield and decreased dynamic range in vivo (5).

The advanced imaging capabilities of the GECO series have been explored (4, 5, 7), but few spectroscopic studies exist for these unique Ca²⁺ sensors. In contrast, spectroscopy on wild-type (wt)GFP included infrared pump probe (16), time-resolved fluorescence (17–19), transient infrared (20, 21), and femtosecond Raman spectroscopy (22), as well as computational studies (23–25), providing a fairly complete picture of the photophysical and photochemical steps leading to green fluorescence (26, 27). In the electronic ground state (GS), wtGFP exists as a mixture of neutral chromophore (A, ∼400 nm peak absorbance) and a small population of anionic chromophore (B, ∼475 nm peak absorbance; Fig. 1B). Emission from the excited state of either form (A* at ∼460 nm and B* at ∼500 nm, respectively) is possible. The main emission pathway upon 400-nm excitation involves excited-state proton transfer (ESPT) from A* on a picosecond timescale to form the intermediate green fluorescent state (I*). We hypothesize that ESPT occurs in the Ca²⁺-free state of GEM-GECO1, but upon Ca²⁺ binding ESPT is disrupted and blue fluorescence occurs from A* (28). In this work, we aim to elucidate the ESPT mechanism of GEM-GECO1 as a function of the chromophore environment, using time-resolved femtosecond stimulated Raman spectroscopy (FSRS). Previous results (22) identified a low-frequency skeletal motion facilitating ESPT in wtGFP that provides guidance to unravel the chromophore dynamics in a flexible CaM–GFP complex. In the absence of a GEM-GECO1 crystal structure, we will compare spectroscopic signatures of the Ca²⁺-free/bound proteins at equilibrium in conjunction with site-specific mutagenesis and molecular dynamics simulations. We then infer how the local environment influences the chromophore structural evolution on the electronically excited state and leads to distinct fluorescence hues.

Fig. 1.

Schematic structure and electronic spectroscopy of the M13-cpGFP-CaM chimera, GEM-GECO1. (A) The Ca²⁺-bound GCaMP2 structure (PDB ID 3EVR) for illustration purposes is shown with the GFP β-barrel in green, CaM in orange, the M13 peptide in magenta, and the Ca²⁺ bound to CaM in yellow. The black arrow indicates the β-barrel opening. The autocyclized SYG chromophore is highlighted by a black box with an asterisk and shown in Inset. (B) Normalized absorption and relative emission spectra of Ca²⁺-free (green solid and dashed lines) and Ca²⁺-bound GEM-GECO1 (blue solid and dashed lines, respectively). The corresponding spectra in wtGFP are shown in black. Blue fluorescence from Ca²⁺-bound protein (peak at 462 nm) is broad and asymmetric relative to the sharp green fluorescence of the Ca²⁺-free protein (peak at 511 nm).

Results

Electronic Spectroscopy Shows Distinctive Features of GEM-GECO1 ±Ca²⁺.

The UV/visible spectrum of Ca²⁺-free protein resembles wtGFP with ∼398 nm absorption maximum and a 511-nm emission peak (Fig. 1B). Upon Ca²⁺ binding the protein absorption/emission maximum blueshifts to 392/462 nm, indicating that the excited-state (ES; S₁) potential energy surface (PES) of the chromophore is modified even though Ca²⁺ binding occurs in the remote CaM domain. In contrast to wtGFP, no significant B state absorbance near 476 nm exists for GEM-GECO1, so the protein pocket strongly favors the neutral chromophore (A state) in GS (S₀). The small shoulder at ∼508 nm for the Ca²⁺-bound protein (4, 14) is consistent with the ensemble measurement and the intrinsic inhomogeneity of the protein complex.

Comparison of GS and ES Spectra Infers the Ca²⁺-Binding Effect.

The GS Raman spectra of GEM-GECO1 both with and without Ca²⁺ are similar, indicative of a largely conserved protein pocket before photoexcitation (Fig. 2 and Fig. S1). Raman bands at 1,603 and 1,565 cm⁻¹ are markers for the neutral chromophore (29). Upon photoexcitation, GEM-GECO1 without/with Ca²⁺ displays key changes starting at time zero: the 1,247/1,250 cm⁻¹ GS C–O stretching mode blueshifts to 1,265 cm⁻¹; the 1,155/1,157 cm⁻¹ mode redshifts to 1,138/1,147 cm⁻¹; and the 1,174 cm⁻¹ mode blueshifts to 1,180 cm⁻¹ (Table S1). The 1,155/1,157 cm⁻¹ GS shoulder is significantly enhanced in A*, indicative of the mode assignment to a largely protonated GS chromophore that experiences electronic redistribution and enhanced polarizability in ES. Moreover, the mode redshift magnitude matches ES calculation (TD-DFT, time-dependent density functional theory) results (30) on a localized phenol H-rocking mode, reflecting a larger extent of S₀→S₁ electronic redistribution in the Ca²⁺-free protein than the Ca²⁺-bound one across the conjugated ring system. In contrast, the 1,174 cm⁻¹ mode shows a small blueshift, consistent with a more delocalized ring H-rocking motion of a partially deprotonated chromophore in GS that has stronger Raman intensity. Also, neither the Ca²⁺-free or -bound protein has a ∼1,300 cm⁻¹ peak in GS (Table S2), but this mode emerges in A* and manifests quite different dynamics in the two species (see below). The mode enhancement is attributed to the increased polarizability of the ES chromophore (31).

Fig. 2.

Time-resolved excited-state FSRS spectra of (A) Ca²⁺-free and (B) Ca²⁺-bound GEM-GECO1 from −5 to 650 ps following 400-nm photoexcitation. The Raman pump is at ∼800 nm. The buffer-subtracted protein spectrum is plotted at the bottom for comparison with the GS-subtracted ES spectra from ca. 750 to 1,700 cm⁻¹. The double-headed arrow shows the Raman gain magnitude of 0.1%. Vibrational modes discussed are indicated with vertical dashed lines and frequency labels. On a much longer timescale, the fluorescence lifetime of the Ca²⁺-free (-bound) GEM-GECO1 is ca. 4 (<1) ns, and the reported lifetime for wtGFP green fluorescence is ca. 3–4 ns (17, 18, 27).

Time-resolved FSRS spectra in Fig. 2 offer unprecedented insights into ES structural dynamics of Ca²⁺-free and -bound GEM-GECO1 across a spectral window spanning over 1,200 cm⁻¹ on the intrinsic timescale for relevant photophysics and photochemistry. It is immediately apparent that the Ca²⁺-free protein resembles wtGFP (22), showing a rapid decay of A* modes and a concomitant increase of I* modes at 1,305 and 1,540 cm⁻¹, with the latter mode redshifted from the 1,570 cm⁻¹ mode. In the Ca²⁺-bound protein, the A* modes persist throughout the entire 650-ps sampling window without significant frequency change, confirming that blue fluorescence (Fig. 1B) is directly emitted from a trapped A* state.

Transient Dynamics of Vibrational Marker Bands in A* and I* Report on PES.

Fitting results and mode assignments for ES peaks (Table S1) and relevant GS modes are tabulated to show the effect of S₀→S₁ transition on the normal modes (Table S2). The kinetic plot of the strongest A* mode intensity at 1,180 cm⁻¹ (Fig. 3A) shows an ultrafast rising component within the 140-fs cross-correlation time and two decay components of 730 (630) fs and 27 (592) ps for the Ca²⁺-free (bound) protein. The first decay time constant is similar to wtGFP in H₂O or D₂O at ∼700 or 600 fs (22), consistent with the initial Franck–Condon (FC) dynamics seeing minor proton motions. Instead, vibronic relaxation occurs in A* despite the fate of the wavepacket when it reaches the lower portion of PES. Therein, the Ca²⁺-free protein modes exhibit a time constant of 30–40 ps attributed to ESPT, much longer than the 5–9 ps counterpart for wtGFP in water (22, 26). In contrast, the Ca²⁺-bound protein shows characteristic mode decay on the 500–900 ps timescale, revealing that ESPT is essentially blocked and blue emission from A* dominates (17, 19).

Fig. 3.

Time evolution of two excited-state vibrational modes for both Ca²⁺-free (green) and bound (blue) GEM-GECO1 up to 650 ps following 400-nm photoexcitation. (A) The 1,180 cm⁻¹ A* mode shows biexponential relaxation but with much lengthened long-time decay with Ca²⁺. The enlarged dynamics plot up to 10 ps exhibits a similar initial decay. The −Ca²⁺ peak is normalized to the stronger +Ca²⁺ peak with a scaling factor of 2. (B) The 1,305/1,300 cm⁻¹ mode in Ca²⁺-free/bound protein shows distinct dynamics. There are overlapping A* and I* modes in the Ca²⁺-free protein and the dominant ESPT timescale is ∼30 ps. The −Ca²⁺ peak is normalized to the stronger +Ca²⁺ peak with a scaling factor of 3. Error bars, SD of the Gaussian-fitted peak area (n = 3). Solid lines are fits to data points, and the peak integrated gain is plotted to better represent the mode intensity due to small variations of the broad ES peak width.

Kinetic analysis of the nascent 1,305 cm⁻¹ ES mode (Fig. 3B) reports on structural dynamics changes induced by Ca²⁺. It involves bridge-H rocking and is particularly sensitive to electronic distribution across the two rings of the chromophore (22). Without Ca²⁺, this mode is much weaker than the nearby 1,265 cm⁻¹ mode during the first 4 ps and is from A* (Fig. 2A); then the mode starts to grow with I* formation. The peak fit to three exponentials yields an initial decay with ∼540 fs time constant, a rising component of ∼31 ps, and a second decay constant of ∼4.1 ns. The intermediate time constant matches ESPT timescale (Table S1), suggesting a direct A*→I* transition. In contrast, although the Ca²⁺-bound protein also exhibits the 1,300 cm⁻¹ mode relaxation, the dynamics are characterized by a biexponential decay with time constants of ca. 700 fs and 472 ps, indicative of vibronic relaxation within a confined A* state.

Remarkably, several modes manifest spectral oscillations (i.e., quantum beating) promptly following photoexcitation. The 819; 885; 967; 1,180; 1,265; 1,400; 1,446; and 1,570 cm⁻¹ modes in the Ca²⁺-free protein all exhibit a reproducible ∼200 fs period oscillation in the peak intensities and frequencies (Fig. S2), wherein the 1,265 cm⁻¹ C–O stretch is antiphase with the 1,570 cm⁻¹ C = N stretch (Fig. 4). Discrete Fourier transform (DiFT) of the quantum beats after removing the broad incoherent component reveals a dominant ∼170 cm⁻¹ peak (Fig. 4, Inset, and Fig. S2, Inset) that corresponds to an in-plane phenol-ring rocking motion about the bridge carbon. The collective nature of the skeletal motion leads to similar mode frequency in S₀ and S₁ (22, 32). In contrast, the Ca²⁺-bound protein exhibits some oscillations within 2 ps, but without a dominant modulating component; several low-frequency modes between 100 and 300 cm⁻¹ of roughly equal intensity are present (Fig. S3). It is conceivable that multiple skeletal modes project onto the PES and contribute to initial energy dissipation but provide no effective driving force for ESPT.

Fig. 4.

Temporal evolution and quantum beats of the 1,265 and 1,570 cm⁻¹ excited-state vibrational modes of the Ca²⁺-free GEM-GECO1 following 400-nm photoexcitation. The two marker band intensities during the first 2 ps manifest an oscillatory pattern with a 200-fs period, whereas the 1,265 cm⁻¹ phenolic C–O stretch (red) is anti-phase with the imidazolinone C = N stretch (black) at 1,570 cm⁻¹. Inset shows normalized DiFT spectra of the coherent residuals after removing the incoherent exponential fits (dashed lines), which manifests a prominent ∼170 cm⁻¹ mode.

Molecular Dynamics Simulation and Site-Directed Mutagenesis of GEM-GECO1 Pinpoint Specific Roles of Residues.

Although the GEM-GECO1 crystal structure is unavailable, some essential insights can be drawn from parent GCaMP structures that show the conserved serine residue along the ESPT path largely in the same plane of the chromophore (10, 33). We made all of the relevant GEM-GECO1 mutations (4) to the Ca²⁺-free GCaMP2 crystal structure 3EKJ (10) and attached the missing part of CaM and M13 units, followed by a 10-ns molecular dynamics simulation to find the lowest-energy structure (SI Text, Fig. S4). To obtain additional insight, we used site-directed mutagenesis to identify potential roles of single residues in modulating the fluorescence response. Specifically, we introduced the individual mutations P60L, E61H, V116T, S118G, and P377R and measured the absorption and emission spectra of the resulting proteins (Figs. S5 and S6). In the crystal structure of Ca²⁺-bound GCaMP (3EVR) (9) that is considered to closely resemble the Ca²⁺-bound GEM-GECO1 (Table S3), residues E61 (M13 to FP linker), T116 and S118 (FP domain), and R377 (CaM domain) all participate in an extensive H₂O-bridged H-bonding network that includes the phenolate oxygen of the chromophore. All five mutations (Table S4) showed a decrease in the ratio of blue to green fluorescence, ranging from slight to severe, in the Ca²⁺-bound form (Figs. S5 and S6A). P377R shows the most profound effect on GEM-GECO1 function and essentially eliminates its dual-emission imaging capability (SI Text).

Discussion

The emerging group of FPs capable of Ca²⁺ sensing paves the way to a myriad of important experiments including imaging neural activities (6) and cancer metastasis (34). However, to realize the full potential of these FPs, we need to unravel the molecular choreography of ESPT that is at the core of the fluorescence change upon Ca²⁺ binding. Structural differences can be inferred from crystallography, but the dynamics insights are pivotal to understand chemical reactivity. The FSRS results unveil important details of the ES structural evolution of the photoexcited GEM-GECO1 chromophore, and paint a clear portrait of its specific local environment affecting transient atomic motions and determining the fluorescence outcome.

Initial Structural Motion of the Chromophore Gates ESPT.

The spectral oscillations in the first few picoseconds of both Ca²⁺-free and -bound proteins indicate that characteristic low-frequency modes project strongly onto the initial reaction coordinate out of the FC region. This pre-ESPT timescale sees dominant collective skeletal motions of the chromophore that tend to dissipate photoexcitation energy (31) or push its phenolic proton away (22, 35). In the Ca²⁺-free protein, the ∼200 fs oscillatory period reveals a dominant ∼170 cm⁻¹ in-plane rock of the phenol ring (Fig. 5). This motion is structurally relevant as it is capable of modulating the phenolic proton position in reference to nearby water molecules and protein residues, leading to frequency and intensity oscillations of the phenolic C–O stretching mode and other high-frequency modes. In the lowest-energy structure for Ca²⁺-free GEM-GECO1 (Fig. S4), a side-to-side in-plane phenol ring rocking motion is more efficient for proton transport than an out-of-plane ring wag, like the 120 cm⁻¹ mode observed in wtGFP (22). Meanwhile, the partially open β-barrel hosts multiple labile H₂O molecules (33, 36), so the lack of a conserved H₂O molecule along the 170 cm⁻¹ rocking motion trajectory toward S118 also explains the lengthened ESPT in Ca²⁺-free GEM-GECO1 (∼30 ps) vs. wtGFP (∼5 ps). The involvement of bridging H₂O is essential for ESPT because the distance between the phenolic proton and S118 is larger (∼4.9 Å) than a typical H-bond length. Our molecular dynamics simulations show such bridging H₂O molecules in the lowest-energy configuration (Fig. S4B). Ultrafast structural motions of H₂O and S118, E135 sidechains are expected in S₁ to facilitate ESPT via an optimized H-bonding chain (Fig. 5).

Fig. 5.

Illustration of initial structural evolution gating ESPT in the Ca²⁺-free GEM-GECO1. Geometric constraints from the Ca²⁺-free GCaMP2 crystal structure 3EKJ are used. An optimized ESPT pathway is delineated with an intervening H₂O molecule added (Fig. S4B) due to limited resolution of the crystal structure. To obtain an SYG chromophore, the methyl group of T223 is replaced by an H atom without further structural modification. The 170 cm⁻¹ mode facilitates ESPT with three positions depicted (DFT-calculated mode displacements are ±2° in equilibrated S₀). The bridge C—C single bond shifts from the inherent structure (gray, 125°, the bridge C = C–C angle) back (orange, 132°) and forth (cyan, 118°) to modulate H-bonding geometry in S₁ with the adjacent H₂O molecule. C, O, N, and H atoms are shown in gray, red, blue, and white, respectively. The surrounding protein environment is shown in green. Rendered using visual molecular dynamics (VMD) (49).

Notably, one conserved low-frequency mode does not dictate ESPT in all systems. The local environment of the embedded chromophore plays a key role in shaping the PES and electronic redistribution, and the coherent low-frequency motions along the ESPT reaction coordinate (22, 37). Therefore, the local geometry and nuclear motion in tandem are responsible for the ESPT efficiency as the chromophore searches phase space to convert from A* to I* before fluorescence. The dominant 120 cm⁻¹ ring wag in wtGFP arises from a conserved H₂O molecule and S205 above the chromophore ring plane (22), with nearby ESPT-prone residues H148 and T203 situated out of the plane to stabilize the deprotonated chromophore. Also, the wtGFP crystal structure shows T203 (Table S3) in two distinct positions (38), indicating that its sidechain can rotate to H bond and stabilize the anionic chromophore, and thus facilitate ESPT. In contrast, the T116V mutation and the largely in-plane S118 (Fig. S4) diminish the functional relevance of a phenolic ring wag in the Ca²⁺-free GEM-GECO1, which instead adopts a 170 cm⁻¹ in-plane rocking motion to drive ESPT (SI Text) but less efficiently. This is in accord with previous reports of longer ESPT time in wtGFP upon T203V mutation (17, 18).

Structural Dynamics in the Chromophore Local Environment for Ca²⁺ Sensing.

The FSRS data reveal the dependence of chromophore dynamics on allosteric Ca²⁺ binding to the CaM domain of GEM-GECO1. The presence of competing skeletal modes in the Ca²⁺-bound protein (Fig. S3) reveals that initial energy dissipation takes on multiple yet less directional pathways. The coherent low-frequency modes originate from vibronic coupling and are likely ubiquitous in photoexcited environments (22, 39–41), similar to the case of photoacid pyranine in various solvents (31, 42–45). Pyranine undergoes ESPT in water following photoexcitation (45), but ESPT is blocked in methanol and multiple low-frequency modes are observed (31). These coherent low-frequency modes are anharmonically coupled to high-frequency modes and are functionally requisite to search local phase space to dissipate energy in lieu of ESPT.

In canonical GCaMPs, the Ca²⁺-bound protein is brightly fluorescent and the Ca²⁺-free protein emits dim fluorescence (4–13). Based on crystal structure 3EVR, it was conjectured that chromophore deprotonation occurs when Ca²⁺ binds to CaM because a conserved R377 coordinates an H₂O molecule near the chromophore phenolic end and participates in an extensive H-bonding network. The Ca²⁺-free protein is dimly fluorescent due to both a protonated chromophore and solvent access to the chromophore pocket that results in fluorescence quenching. In contrast, GEM-GECO1 fluoresces green without Ca²⁺ but blue with Ca²⁺, so the GCaMP fluorescence modulation cannot be the dominant mechanism for this sensor, which possesses several key point mutations relative to its GCaMP progenitor. Our reversal mutagenesis results manifest that the interfacial R377P mutation plays an essential role in removing the aforementioned R377 interaction and effectively inhibits ESPT in the Ca²⁺-bound protein. As position 377 is in a helical segment of CaM, proline mutation will affect the secondary and tertiary protein structure.

In Ca²⁺-bound GEM-GECO1, it is thermodynamically unlikely for water to be completely excluded from the protein interior because the β-barrel opening is large (10, 46), but the availability of water molecules alone cannot enable ESPT. Previous work explored pH-dependent dual-emission GFP variants (15, 47) with mutations at positions 65, 148, and 203: blue fluorescence was attributed to decreased H-bond interactions and increased hydrophobicity of the protein pocket. As an analogy, although P377 in the Ca²⁺-bound protein cannot directly participate in the extensive H-bond network due to lack of an ionizable group and restricted sidechain conformation, this mutation likely repositions other CaM residues (e.g., M375 or M379; SI Text) such that they increase the hydrophobicity of chromophore environment, promote the neutral form, reduce the available H bonds, and impede ESPT pathways by increasing the reaction barrier. Generation and analysis of the Ca²⁺-bound GEM-GECO1 crystal structure should reveal which repositioned residues are involved in these critical interactions at the FP–CaM interface.

In addition, the wtGFP mutant S65T/H148E exhibited dual emission but with strong pH dependence (47), hinting the role played by a repositioned glutamate to the chromophore. Our mutagenesis results show that the Ca²⁺-bound E61H variant has reduced blue but enhanced green fluorescence, indicative of less ESPT disruption and reduced dual-emission capability (Fig. S6). Therefore, the closing in of some negatively charged residues could destabilize the deprotonated chromophore and contribute to A* trapping in the Ca²⁺-bound GEM-GECO1.

Conclusion

Using FSRS, we studied the time-resolved excited-state structural dynamics of dual-emission GEM-GECO1 and inferred its fluorescence modulation mechanism. Distinct vibrational dynamics are associated with Ca²⁺-free and -bound states due to their different chromophore pocket environments. The Ca²⁺-free protein A* modes decay with a dominant 30–40 ps time constant, and vibrational modes associated with the intermediate deprotonated I* state rise on a similar timescale. Remarkably, most of the A* modes have a well-defined 200-fs spectral oscillation period that corresponds to a 170 cm⁻¹ phenol ring-rocking motion about the bridge carbon. This impulsively excited coherent motion affords directionality of the wavepacket moving out of the FC region toward ESPT barrier crossing, facilitating phenolic proton transfer to a largely in-plane adjacent serine residue via labile bridging water molecules.

Upon Ca²⁺ binding, the GEM-GECO1 chromophore gets trapped in A* after photoexcitation and emits blue fluorescence. Analysis of spectral oscillations therein reveals multiple competing low-frequency modes, dissipating energy in lieu of ESPT. The key ingredients for Ca²⁺-induced ESPT inhibition are identified to include (i) structural rearrangement as CaM compresses and forms a new interface with GFP, increasing the hydrophobicity around the chromophore; (ii) disruption of the pro-ESPT H-bonding network; and (iii) negative residues moving closer to the phenolic end to hinder chromophore deprotonation. The elucidation of elementary steps in ES structural evolution of a 3-residue chromophore inside a 450+ residue protein complex offers crucial insights into the dual-emission Ca²⁺-sensing GEM-GECO1, which not only paves the way to mechanistically study photosensitive biomolecules in physiological conditions, but also enables targeted protein design to develop more powerful biosensors.

We anticipate that the allosteric mechanism of transient conformational dynamics induced by Ca²⁺ can be further studied with a photolabile molecular cage that releases Ca²⁺ with femtosecond time resolution (48). The nonequilibrium structural evolution of the chromophore may be tracked using a common 400-nm photoexcitation pulse.

Materials and Methods

A full description of methods is given in SI Materials and Methods. To prepare GEM-GECO1 proteins for in vitro spectroscopic characterization, Escherichia coli cells expressing 6-histidine tagged GEM-GECO1 were grown in 1 L of a modified terrific broth at 30 °C for two days, collected by centrifugation and lysed by French press. Proteins from the clarified supernatant were then purified by nickel-nitrilotriacetic acid (Ni-NTA) affinity chromatography and finally exchanged to 3-(N-morpholino)propanesulfonic acid (MOPS) buffer (OD > 1/mm at 400 nm) with either 10 mM EGTA (Ca²⁺-free sample) or 10 mM CaEGTA (Ca²⁺-bound sample). Time-resolved excited-state spectra of photoexcited GEM-GECO1 proteins with and without Ca²⁺ were collected by an FSRS setup that was previously described (44, 45), using similar photoexcitation power on a 1-mm-thick protein sample solution (OD ∼ 1/mm at 400 nm) in a flow cell to avoid thermal effects or photodegradation (22).