Mechanistic Insights into Reversible Photoactivation in Proteins of the GFP Family

Abstract

Light-controlled modification of the fluorescence emission properties of proteins of the GFP family is of crucial importance for many imaging applications including superresolution microscopy. Here, we have studied the reversibly photoswitchable fluorescent protein mIrisGFP using optical spectroscopy. By analyzing the pH dependence of isomerization and protonation equilibria and the isomerization kinetics, we have obtained insight into the coupling of the chromophore to the surrounding protein moiety and a better understanding of the photoswitching mechanism. A different acid-base environment of the chromophore’s protonating group in its two isomeric forms, which can be inferred from the x-ray structures of IrisFP, is key to the photoswitching function and ensures that isomerization and protonation are correlated. Amino acids near the chromophore, especially Glu212, rearrange upon isomerization, and Glu212 protonation modulates the chromophore pK_a. In mIrisGFP, the cis chromophore protonates in two steps, with pK_cis of 5.3 and 6, which is much lower than pK_trans (>10). Based on these results, we have put forward a mechanistic scheme that explains how the combination of isomeric and acid-base properties of the chromophore in its protein environment can produce negative and positive photoswitching modes.

Introduction

Fluorescent proteins (FPs) of the green fluorescent protein (GFP) family have become key research tools in a wide range of life science applications. These small proteins of ∼250 amino acids autocatalytically form a fluorescent chromophore in the interior of their 11-stranded β-barrel fold (1–4). When fused to a protein of interest on the DNA level, FPs can serve as genetically encoded, highly specific fluorescence markers in cellular imaging (5–13). In recent years, photoactivatable or optical-highlighter FPs have emerged as especially powerful marker tools, notably for pulse-chase imaging and a variety of superresolution microscopy techniques (7,14–17). Their fluorescence properties can be modified by light irradiation at specific wavelengths. Two different modes of photoactivation are presently known. Irreversible photoactivation, also known as photoconversion, involves a permanent photochemical modification of the FP chromophore. As a consequence, a nonfluorescent (dark) state may become permanently activated to a fluorescent (bright) state, or a bright state may be turned into another bright state with a different emission wavelength (18–23).

Reversible photoactivation, also known as photoswitching, typically arises from chromophore isomerization between two configurations, only one of which emits fluorescence with high yield (24–30). We note in passing that Dreiklang, a recently introduced photoswitching GFP variant, is a notable exception to this rule, featuring a reversible hydration/dehydration reaction of the chromophore (31). Chromophore isomerization is intimately coupled to a chromophore protonation reaction (20,24,32). In most reversibly photoswitchable FPs, the cis configuration is the thermodynamically stable form; its anionic form constitutes the bright fluorophore (24,33,34). Electronic excitation triggers a cis-trans isomerization accompanied by a change of protonation state, resulting in an essentially nonfluorescent chromophore. In a few photoswitchers, including KFP1 (22) and asFP595 (25,35,36), the nonfluorescent trans state rather than the cis state is the thermodynamically stable form, and light irradiation induces a trans-cis isomerization to the fluorescent cis state. In the dark, these FPs spontaneously relax to their thermodynamically stable isomeric state on timescales ranging from minutes to hours. This process can be markedly accelerated by irradiation with light of a suitable wavelength. In the so-called positive photoswitchers, fluorescence excitation enhances the emission intensity; in negative switchers, the emission is quenched by the excitation light (37). The body of acquired knowledge on photoconversion and photoswitching has been presented in several reviews (14,16,38). There is general agreement on the basic principles of photoactivation; the mechanistic details, however, are still controversial. For Dronpa, for example, it was proposed that on-switching is initiated by excited-state proton transfer (ESPT) from the neutral trans chromophore (39), and that off-switching involves an intersystem crossing from the excited anionic cis form to the triplet state and its subsequent protonation (40,41). For asFP595, results from molecular-dynamics simulations led to the suggestion that trans-cis isomerization of the chromophore occurs via a hula-twist mechanism, with a concerted rotation of both bonds of the methine bridge connecting the phenoxy and the imidazolinone rings, possibly accompanied by proton transfer (25). Because the various photoswitching FPs that exist all have monomethine chromophores, Olsen et al. (42) proposed that photoswitching is an intrinsic property of the chromophore. For efficient photoswitching, the protein environment should 1), restrict torsion around the phenoxy bond (P-bond) to enhance the radiative decay channel, and 2), provide a suitable acid-base chemistry so that protonation-state change and isomerization by torsion around the imidazolinone bond (I-bond) are correlated.

We recently introduced FP variants that combine reversible and irreversible photoactivation, tetrameric IrisFP (20) and its monomeric variant, mIrisFP (43). In their green fluorescent form, they display reversible photoswitching between a fluorescent and a nonfluorescent state. In addition, they can be photoconverted irreversibly to red-emitting FPs by ∼400 nm light. In their red form, they can again be toggled between a bright fluorescent and a dark state. Due to their multiple photoactivation modes, these FPs enable entirely novel experiments, e.g., pulse-chase imaging with superresolution in living cells (43).

To understand the competition between photoswitching and photoconversion, we have, as a first step, performed a detailed study on the Phe¹⁷³Ser variant of the advanced EosFP variant mEosFPthermo (43,44). This protein can be photoswitched efficiently and reversibly between a green fluorescent and a dark state; its irreversible photoconversion from a green fluorescent to a red fluorescent form, however, is largely suppressed and only observable under extremely high laser powers. We refer to this green variant as mIrisGFP because it shares the crucial Phe¹⁷³Ser mutation of IrisFP, rendering it photoswitchable.

Here, we present an in-depth spectroscopic characterization of the cis and the trans chromophore species of mIrisGFP and analyze the thermally and light-activated chromophore isomerization reactions over a wide pH range. We provide evidence that both chromophore isomerization and protonation occur during reversible photoswitching. Therefore, even a simple model of this process comprises a minimum of four chromophore species in the ground state (cis neutral, cis anionic, trans neutral, trans anionic) and, upon light exposure, the corresponding excited-state species (Fig. 1). In the following, C and T will denote protein subpopulations with cis and trans chromophore isomers, respectively, with superscript minus signs for anionic and superscript H for neutral.

Figure 1.

General scheme capturing reversible photoactivation of mIrisGFP and other photoswitchable FPs. It contains eight chromophore species that are potentially involved. The states are denoted by C and T for cis and trans chromophore isomers, respectively, with superscript minus sign for anionic and superscript H for neutral; an asterisk denotes the electronically excited states.

Materials and Methods

Protein mutagenesis, expression, and purification

Taking the mEosFPthermo vector as a template, we introduced an additional mutation, Phe¹⁷³Ser, by site-directed mutagenesis, yielding mIrisGFP (QuikChange II Mutagenesis Kit, Stratagene, La Jolla, CA). The variant Glu²¹²Gln was produced by further site-directed mutagenesis of mIrisGFP. DNA primers were purchased from Biomers (Ulm, Germany); the resulting clones were sequenced by a commercial provider (GATC, Konstanz, Germany). Proteins were expressed in E. coli strain XL1-blue (Stratagene) and purified as described (43).

Spectroscopic characterization

Absorption spectra of mIrisGFP were recorded on a Cary 50 Bio UV/Vis spectrophotometer (Varian, Darmstadt, Germany). Fluorescence spectra were measured with a SPEX Fluorolog II spectrofluorometer (Spex Industries, Edison, NJ). The excitation line width was set to 0.85 nm; emission spectra were taken at 2.2-nm resolution and corrected for the wavelength dependence of the detector efficiency. Circular dichroism (CD) spectra were recorded on a J-810 spectropolarimeter (Jasco Europe, Cremella, Italy).

For spectroscopic experiments, concentrated stock solutions of mIrisGFP were diluted to protein concentrations of ∼10 μM (absorption), ∼1 μM (fluorescence), and ∼30 μM (CD) in 20 mM buffer solutions (sodium citrate/phosphate, sodium phosphate, sodium carbonate, and glycine buffer systems, depending on the desired pH in the range 3–11). The ionic strength was adjusted to 150 mM with sodium chloride.

Photoswitching quantum yields

To measure the kinetics of photoswitching, we illuminated mIrisGFP solutions with 473-nm light at 5–50 mW cm⁻² (off-switching) and 405-nm light at 5–25 mW cm⁻² (on-switching) from solid-state lasers and recorded the resulting light-induced changes of the fluorescence emission intensity at 520 nm as a function of time. The power density of the activating light was adjusted as appropriate for the specific experiment. The measured kinetic traces were fitted with exponential decays, yielding the sum, k_on + k_off, as an apparent rate coefficient. Consequently, a single, microscopic rate coefficient results directly from the kinetics only if one of the two predominates. The rate coefficient, k, can be converted to a quantum yield of the different photo-induced processes, ϕ, by using the equation

$$\varphi =\frac{k}{\epsilon }\times \frac{1}{\lambda P}\times \left(N_{A}hc\right),$$ (1)

with wavelength, λ, power density, P, Avogadro’s number, N_A, Planck’s constant, h, and the velocity of light, c, and the extinction coefficient, ε, of the reacting species at the wavelength chosen for triggering the process (20).

Thermal relaxation to the fluorescent form

mIrisGFP samples were illuminated with 473-nm light (25 mW cm⁻²) until the absorption band at ∼490 nm was suppressed to <5% of its original area. Subsequently, thermal relaxation was monitored as a function of time via the absorption increase at 490 nm. The absorption at 490 nm was plotted as a function of time and fitted with a single exponential. Because of the rather slow relaxation kinetics, these experiments were carried out at 310 K (37°C) unless stated otherwise.

Action spectrum of light-induced off-switching

mIrisGFP samples were illuminated for 300 s with light at 20 different wavelengths, λ_exc, between 330 and 530 nm. The wavelengths were selected from the 450-W xenon lamp of the Fluorolog II spectrometer by using its grating monochromator. The relative fluorescence decrease upon light exposure was measured at 520 nm (λ_exc = 330–490 nm) and 540 nm (λ_exc = 490–530 nm). The data were adjusted for intensity variations of the source spectrum and photon energy and plotted against the excitation wavelength to yield the action spectrum.

Results

Spectroscopic characterization of mIrisGFP in the cis isomeric form

In its dark-adapted form, the green chromophore of IrisFP adopts the cis configuration, as shown in the structural model in Fig. 2 A (20). The very similar optical spectra of IrisFP and mIrisGFP suggest that the chromophore of dark-adapted mIrisGFP also assumes the cis configuration. Its absorption spectra show a strong pH dependence (Fig. 3 A), which is often—but not always (45,46)—observed in GFPs and can be described by an exchange between an A (∼400 nm) and a B (∼500 nm) band associated with a neutral and an anionic chromophore, respectively. To assign an absorption band to a protein with a particular chromophore isomer, we use either C (cis) or T (trans) as a subscript. At low pH, the spectrum is dominated by the A_C band, peaking at 389 nm (ε_max = 24,000 ± 2000 M⁻¹ cm⁻¹). With increasing pH, the B_C band at ∼488 nm (ε_max = 47,000 ± 1700 M⁻¹ cm⁻¹) gains intensity at the expense of A_C (Fig. 3 A). The peak amplitudes of the A_C and B_C bands are plotted as a function of pH in Fig. 3 B.

Figure 2.

Structural environment of the green chromophore in IrisFP. Hydrogen bonds are shown as dotted lines. (A) Cis conformer (PDB accession code 2VVH). (B) Trans conformer (PDB accession code 2VVI).

Figure 3.

pH variation of the optical absorption spectra of mIrisGFP in the dark-adapted state (cis chromophore). (A) Absorption spectra in the pH range 3.6–10.4. Arrows point in the direction of increasing pH. (B) Peak absorption of the A_C (open circles) and B_C (solid circles) bands. Solid lines show the fit of the model depicted in panel c in the pH range 3.6–8. An additional protonation step described by the Henderson-Hasselbach equation was added to fit the B_C band amplitudes at pH >8. Dotted line shows one-site protonation with pK = 5.3. Open triangles indicate pH dependence of the reaction rate coefficient, k_off(473), for off-switching of the fluorescence with 473-nm light. (C) Four-state model describing the protonation equilibrium of the chromophore, C, interacting with a titratable amino acid, X. (D) pH dependence of the peak positions of the A_C (open circles) and B_C (solid circles) absorption bands of mIrisGFP, as well as the peak position of B_C of the Glu²¹²Gln mutant (stars).

For a simple protonation equilibrium between a neutral C^H and an anionic C⁻ chromophore species, the fraction of C⁻ varies with pH according to

$$f_{C^{-}}\left(\text{pH}\right)=\frac{1}{1+{10}^{n\left(\text{p}{K}_{\text{a}}-\text{pH}\right)}},$$ (2)

with a Hill coefficient n = 1; n ≠ 1 indicates that a simple one-site protonation is inadequate. This function, plotted in Fig. 3 B (dotted line) for pK_a = 5.3 and n = 1, fails to describe the pH dependence of the C⁻ population. Apparently, a more complex model is needed to fit the observed two-step behavior, which includes (at least) one other protonating group, which we shall denote by X in the following. Located near the chromophore, X modulates the chromophore’s proton affinity via electrostatic interactions. Such a model thus contains four discrete species, C^H/X^H, C^H/X⁻, C⁻/X^H and C⁻/X⁻, as shown in the scheme in Fig. 3 C (47–51). The equations governing the fractional populations of all four species are provided in the Supporting Material. However, those are not directly accessible via the pH-dependent absorption spectra of mIrisGFP, because we can only distinguish between the overall A_C and B_C bands and, thus, the fractional populations of neutral and anionic chromophores,

$$\begin{array}{l}{f}_{{\text{C}}^{-}}\left(\text{pH}\right)=f_{{\text{C}}^{-}\text{XH}}\left(\text{pH}\right)+f_{{\text{C}}^{-}{\text{X}}^{-}}\left(\text{pH}\right)\hfill \\ f_{\text{CH}}\left(\text{pH}\right)=f_{\text{CHXH}}\left(\text{pH}\right)+f_{{\text{CHX}}^{-}}\left(\text{pH}\right).\hfill \end{array}$$ (3)

Although the protonation state of X is not observable from the relative weights of the A_C and B_C bands, the presence or absence of a charge in the close vicinity of the neutral or anionic chromophore may affect its delocalized π-electron system. Therefore, the protonation state of X may be inferred from Stark shifts of the A_C or B_C bands, as discussed below.

A global fit to the data (Fig. 3 B) in the range 3.5 < pH < 7.5, based on the four-state model (Eq. 3 and Eq. S3 in the Supporting Material), yields pK_a1 = 5.3 ± 0.1 and pK_a4 = 6.0 ± 0.1 for the two-step protonation reaction of the cis chromophore. The amplitude of the B_C band at 489 nm displays yet another step with pK_a = 8.9 ± 0.2 (Fig. 3 B) that is accompanied by a marked red shift. This effect is likely associated with the appearance of a chromophore species with a more red-shifted absorption band at 496 nm, for which the negative charge of the hydroxybenzyl anion is no longer stabilized by a hydrogen bond to Ser¹⁴² (46).

We have also examined the spectra for band shifts that may indicate a change in the protonation state of X. The B_C band indeed shows a pronounced red shift by 6 nm, with increasing pH that can be described by pK_a2 = 6.6 ± 0.1 (Fig. 3 D). This shift is absent in the mIrisGFP mutant Glu²¹²Gln (compare Fig. 3 D and Fig. S1), strongly suggesting that the protonation state of Glu²¹² affects the proton affinity of the cis chromophore. Likewise, we have observed a shift of the A_C band that signals protonation of X in the presence of a neutral chromophore, C^H (Fig. 3 D, dotted line). At pH >5, the shift can be described by pK_a3 = 5.9 ± 0.1; the strong deviation below pH 4.5 is due to the onset of acid denaturation (Fig. 3 D). We note that the value of pK_a3 agrees quantitatively with that required by the model (Eq. S2) and thus lends strong credence to our two-site protonation model. The coupling free energy between the two sites (Eq. S5) amounts to ∼4 kJ/mol.

The excitation spectra of mIrisGFP, collected with the emission monochromator set to λ_em = 540 nm, display a single band at 488 nm, showing that C⁻ is the fluorescent species (Fig. 4 A, dotted lines); the emission band is centered on 515 nm (Fig. 4 A, dashed lines). The fluorescence intensity shows the same pH dependence as the B_C band absorption at pH <8 and consequently can be modeled with the same pK_a values (see Fig. 4 C, triangles). At pH >8, the emission intensity remains constant, whereas the absorption still increases. This discrepancy suggests that the high-pH chromophore species represented by the absorption band at 496 nm is nonfluorescent, as would be expected for a less stabilized chromophore. The fluorescence quantum yield of C⁻ at pH 7 was determined as Φ = 0.63 ± 0.06. The neutral chromophore, C^H, is only poorly fluorescent (Φ < 0.03). Its emission peaks at ∼460 nm (λ_exc = 390 nm); concurrent fluorescence emission at ∼515 nm is not observed, thus excluding excited-state proton transfer as a cause of chromophore deprotonation in the excited state and the subsequent red-shifted emission by the anionic fluorophore seen in a variety of FPs (52–54).

Figure 4.

pH dependence of optical spectra of mIrisGFP in the dark-adapted state (cis chromophore) (A–C) and immediately after 473-nm illumination (predominantly trans chromophore) (D–F). (A) Absorption, excitation, and emission spectra (solid, dotted, and dashed lines, respectively). Excitation and emission spectra were measured with emission at 540 nm and excitation at 390 nm. (B) CD spectra. (C) pH dependence of the (normalized) peak amplitudes of the A_C (open circles) and B_C (solid circles) absorption bands and of the emission intensity (triangles). (D and E) Absorption spectra and CD spectra, respectively. Arrows indicate increasing pH. (F) Normalized peak amplitudes of the A (open circles) and B (solid circles) bands and normalized peak shift of the A band in D (solid diamonds) and E (open diamonds) and of mutant Glu²¹²Gln (stars).

The CD spectra of mIrisGFP display two bands with negative ellipticity corresponding to the A_C and B_C absorption bands (compare Fig. 4, A and B). Within the experimental error, the CD bands display the same pH-dependent population exchange as the absorption bands. The optical parameters of mIrisGFP (pH 7) with a cis chromophore are compiled in Table 1.

Table 1.

Optical properties of mIrisGFP at pH 7

Cis chromophore	Anionic (Bc)	Neutral (Ac)
Wavelength, λmax (Ex/Em)/nm	488/516	391/450
Extinction coefficient, ε (λmax) (M−1 cm−1)	47,000 ± 1,700	24,000 ± 2,000
Fluorescence quantum yield, Φ	0.63 ± 0.06	<0.03
pKa	5.3 ± 0.1
	6.0 ± 0.1
trans chromophore	Anionic (Bt)	Neutral (At)
Wavelength, λmax (Abs) (nm)	ND	390
Extinction coefficient, ε (λmax) (M−1 cm−1)	ND	18,000 ± 2,300
Quantum yield, Φ	—	—
pKa	>10

Spectroscopic characterization of mIrisGFP in the trans state

The protein with the chromophore in trans is a light-induced, metastable conformation that spontaneously reverts back to cis. To measure its absorption spectra as a function of pH, we prepared the trans state by illuminating mIrisGFP solutions with 473-nm light until the color faded completely. Then, the 473-nm light was switched off and spectra were taken. Because of thermal trans-cis relaxation during data acquisition (∼1 min), it was unavoidable that the spectra of the trans form contain minor contributions of the cis species. The absorption spectra (Fig. 4 D) are dominated by an A band at ∼390 nm representing a neutral chromophore species. With increasing pH, a second band emerges at ∼490 nm at the expense of A, which we assign to an anionic chromophore species. Its pH-dependent absorbance change can be modeled with a pK_a of 7.8 ± 0.1 (see Fig. 4 F, solid circles). However, even at pH 10, only a minority of proteins are involved in this pH-dependent spectral change, which indicates that it cannot originate from deprotonation of the trans chromophore. Comparison of the spectra in Fig. 4, A and D, rather suggests that the 490-nm band represents anionic cis chromophore species that have reformed during data collection. The pH-dependent variation of the 490-nm band with pK_a = 7.8 thus describes the pH dependence of the trans-cis relaxation kinetics rather than a protonation equilibrium. Additional spectral features that would indicate the presence of anionic trans chromophores are absent in the spectra. Consequently, the trans isomer does not deprotonate in the observed pH range, so that pK_a > 10 for this isomer.

We have calculated pure absorption spectra of the trans species by removing the cis fraction from the spectra. To this end, the cis spectra in Fig. 4 A (at the respective pH) were scaled to match the 490-nm band of the spectra in Fig. 4 D and subtracted. We note that the apparent pK_a of 7.8 is significantly above the pK_a values of 5.3 ± 0.1 and 6.0 ± 0.1, which characterize cis chromophore protonation, and that contributions of neutral cis in the spectral region of the A band are therefore small. The pure A_T band of the neutral trans species shows a pH-dependent shift from 396.8 to 384.3 nm that can be modeled by proton binding with pK_a = 7.8 ± 0.1. The observation that the spectral shift has the same pH dependence as the thermal trans-cis recovery kinetics, shown by plotting the data in a normalized fashion in Fig. 4 F, suggests that the protonation state of the amino acid responsible for this Stark shift also influences the kinetics of isomerization.

Pronounced changes are visible in the CD spectra upon cis-trans isomerization. The bands with negative ellipticity associated with the cis chromophore (see Fig. 4 B) are essentially absent. Instead, there is a strong band with positive ellipticity at ∼390 nm (Fig. 4 E) associated with the nonfluorescent, neutral trans species, T^H. Like the A_T absorption band, this CD band shifts markedly to the blue, and its pH dependence can also be modeled with pK_a = 7.8 ± 0.1 (Fig. 4 F, open diamonds). Because the blue shift is absent for mutant Glu²¹²Gln, it is apparently associated with a change in Glu²¹² protonation. The 390-nm band dominates the CD spectrum even at pH 10; a band associated with an anionic trans chromophore, T⁻, could not be discerned (Fig. 4 E). This observation lends further credence to our hypothesis that the trans chromophore protonates with pK_a > 10. The optical properties of mIrisGFP with a trans chromophore are also included in Table 1.

Thermally activated chromophore isomerization

The neutral trans chromophore of mIrisGFP, T^H, relaxes back to cis over time, with a half-life of ∼90 and ∼20 min at 290 and 310 K, respectively (at pH 7). We have confirmed that thermal relaxation goes to completion at all pH values examined here (pH 5–10.4). The recovery kinetics of the cis isomer is pH-dependent (Fig. 5 A). Between pH 5.5 and 10, its rate coefficient, k_th, increases ∼40-fold, from 0.167 × 10⁻³ s⁻¹ to 6.45 × 10⁻³ s⁻¹, as determined from the increase of the B_C band at 488 nm (T = 310 K). The acceleration occurs in two steps. Using a sum of two Henderson-Hasselbalch relations (Eq. 2), we obtain pK_th1 = 7.8 ± 0.4 and pK_th2 = 9.5 ± 0.6, indicating a destabilizing influence on the trans chromophore due to successive deprotonation of two moieties. The step at lower pH is absent for mutant Glu²¹²Gln (Fig. 5 A).

Figure 5.

pH dependence of the on-switching kinetics (trans-cis isomerization). (A) Thermally activated recovery of the anionic cis chromophore with rate coefficient, k_th, measured at T = 290 K (open circles) and T = 310 K (solid circles). Stars, mIrisGFP Glu²¹²Gln at T = 310 K; solid line, two-site protonation with pK_a1 = 7.8 and pK_a2 = 9.5; dotted line, one-site protonation with pK = 7.8; dashed line, one-site protonation with pK = 10.3. (B) Photoactivated trans-cis isomerization: k_on(405) for P₄₀₅ = 13 mW cm⁻². Solid line shows one-site protonation with pK = 7.8.

Light-activated chromophore isomerization

To study the kinetics of photoactivated cis-trans isomerization of mIrisGFP, we prepared a sample at pH 5.3 with roughly equal fractions of C^H and C⁻ chromophores. We illuminated the sample for 300 s with light between 330 and 550 nm in wavelength and measured the relative loss in fluorescence at 520 nm (540 nm for excitation wavelengths >480 nm). These data were normalized for differences in light intensity and photon energy to reveal the wavelength dependence of the photoisomerization efficiency (i.e., the product of extinction coefficient and quantum yield of photoisomerization), which represents the probability of photoisomerization by light irradiation at a particular wavelength.

In Fig. 6, the action spectrum (Fig. 6, symbols) is plotted against the excitation wavelength, together with the absorption spectrum (Fig. 6, solid line). Both spectra were scaled to equal amplitudes at the peak of A_C. Remarkably, the action spectrum has a maximum in the A_C band and is significantly lower in the region of the B_C band. Consequently, light-induced cis-trans isomerization is more efficient when exciting the nonfluorescent C^H than the it is for the fluorescent C⁻ species.

Figure 6.

Action spectrum of cis-trans photoisomerization of mIrisGFP. The fluorescence decrease (symbols) at 520/540 nm due to photo-induced off-switching is plotted as a function of the wavelength. Also shown are the absorption (solid line), excitation (dotted line), and emission (dashed line) spectra at pH 5.3. Action spectrum and absorption amplitudes were normalized to 1 at the peak of the A_C absorption band at 390 nm; the fluorescence spectra were normalized to match the peak of the B_C absorption band.

With our 473-nm laser source (P₄₇₃ = 24 mW cm⁻²), light-induced off-switching via excitation of C⁻ occurs within tens of milliseconds; it is about three times faster at pH 9.5 (k_off(473) = 0.060 ± 0.001 s⁻¹) than at pH 5 (k_off(473) = 0.024 ± 0.004 s⁻¹). Its pH dependence follows that of the C⁻ population (Fig. 3 B). The off-switching quantum yield, ϕ_off(473) ≈ k_off(473)/ε_c(473) (Eq. 1), increases from (2.3 ± 0.5) × 10⁻³ at pH 5 to (6.9 ± 0.3) × 10⁻³ at pH 9.5.

At the laser powers employed here, off-switching via excitation of the anionic cis chromophore at 473 nm is a simple one-photon process, as shown by the linear dependence of the rate coefficient, k_off(473), on the intensity (Fig. 7 A). However, even at the highest intensities (70 mW cm⁻²), we could not switch off all chromophores: for a sample at pH 7, 4.3 ± 0.1% of the initial fluorescence remained (Fig. 7 A). The residual fluorescence depended on the wavelength used for photoactivation (Fig. 7 B), suggesting that a second light-driven process opposes off-switching. Apparently, the activating light can also excite the neutral trans species, T^H, and trigger the back reaction, T^H → C⁻, although with a lower yield. Because of the high pK_a of the trans chromophore, light activation of T⁻ is negligible at pH 7.

Figure 7.

Effect of illumination power and wavelength on photo-induced off-switching of mIrisGFP (cis-trans isomerization). (A) Dependence of the rate coefficient, k_off(473), on the 473-nm excitation laser power (open circles, solid line: linear fit) and residual fluorescence upon extended 473-nm laser illumination (solid circles, dotted line: fit) as a function of laser power. (B) Time dependence of the fluorescence decay upon illumination with light at three different wavelengths.

To single out the effect of light on mIrisGFP molecules with a T^H chromophore, samples were first illuminated with 473-nm light to populate this state to the largest possible extent. Afterward, the temporal increase of the fluorescence at 520 nm, which is proportional to the amount of C⁻, was recorded under 405-nm illumination. At pH 7.8, the on-switching rate coefficient, k_on(405), was 0.17 s⁻¹ (P₄₀₅ = 13.3 mW cm⁻²), so that Eq. 1 yields a quantum yield of on-switching of ϕ_on(405) = 0.15 ± 0.01. Fig. 5 B shows that k_on(405) decreases with increasing pH, with pK_a = 7.8 ± 0.1. As the absorption at 405 nm also decreases with pK_a = 7.8 ± 0.1 due to the shift of the A_T band (Fig. 4 F), ϕ_on(405) is constant between pH 4.5 and 9.

Because of the large spectral overlap of the A absorption bands of the neutral species, λ_max (C^H) = 389 nm and λ_max (T^H) = 391 nm at pH 7, 405-nm light irradiation excites both isomeric forms. Therefore, a light-controlled equilibrium, ${\text{T}}^{\text{H}}\rightleftarrows {\text{C}}^{\text{H}}$, is established, with an equilibrium coefficient of

$$K_{405}=\frac{\left[{\text{T}}^{\text{H}}\right]}{\left[{\text{C}}^{\text{H}}\right]}=\frac{k_{\text{off}}\left(405\right)}{k_{\text{on}}\left(405\right)}=\frac{{\varphi }_{o\text{ff}}\left(405\right)\times {\epsilon }_C}{{\varphi }_{\text{on}}\left(405\right)\times {\epsilon }_T}.$$ (4)

K₄₀₅ is pH-dependent because k_off(405) depends on the protonation equilibrium of the cis chromophore. With 405-nm illumination at pH 5.3 (P = 20 mW/cm²), the fluorescence was reduced to 70% of its initial value, K₄₀₅ = 0.41 ± 0.05. With the extinction coefficients of the neutral forms, T^H and C^H, at 405 nm, ε_T = (12,000 ± 1600) M⁻¹cm⁻¹ and ε_C = (20,700 ± 2000) M⁻¹ cm⁻¹, respectively, and ϕ_on(405) = 0.15 (see above), we finally obtain ϕ_off(405) = 0.036 ± 0.008. The photophysical parameters of mIrisGFP are compiled in Table 2 together with those of its cousin, mIrisFP (43), for comparison.

Table 2.

Photophysical parameters of mIrisFP and mIrisGFP at pH 7

	mIrisFP (green)	mIrisGFP
ϕoff(473)	0.0069 ± 0.0001	0.0045 ± 0.0004
ϕoff(405)	ND	0.036 ± 0.008
ϕon(405)	0.36 ± 0.02	0.15 ± 0.01
Residual fluorescence (%)∗	4.8 ± 0.7	5.0 ± 0.6
Thermal recovery half-life t1/2 (min)†	60 ± 4.6	64 ± 2

All data were taken on samples in potassium phosphate buffer. ND, not determined.

^∗Determined with 473-nm illumination at 25 mW cm⁻².

^†Measured at 290 K.

Discussion

Chromophore protein interactions in mIrisGFP

The photophysical properties of (photoactivatable) FPs strongly depend on the coupling between the chromophore and the surrounding protein moiety. In the cis and trans isomeric forms of mIrisGFP, the chromophore’s protonating hydroxyphenyl group resides in two very different local environments (Fig. 2), as revealed by the x-ray structures of IrisFP (20). In the cis configuration, the hydroxyphenyl group forms a hydrogen bond to the Ser¹⁴² hydroxyl group and, therefore, is presumably deprotonated (Fig. 2 A). The anionic state is further stabilized by a hydrogen bond between Arg⁶⁶ and the imidazolinone carbonyl stabilizing the electron density on the imidazolinone ring. In contrast, the hydroxyphenyl group of the mIrisGFP trans chromophore points toward the Glu¹⁴⁴ side chain (Fig. 2 B). As in IrisFP (20), Glu¹⁴⁴ is most likely deprotonated. Consequently, the trans chromophore has a much higher proton affinity (pK_a > 10) than the cis chromophore (pK_a = 5.3/6.0) and is thus protonated up to pH values of ∼9.5.

Chromophore protonation and the kinetics of trans-cis isomerization are coupled to the protonation of a neighboring group (Fig. 3), which we assign to the side chain of Glu²¹² based on several observations. In the Glu²¹²Gln mutant, the shifts of the absorption bands are absent and, moreover, the kinetics of photoisomerization are altered, strongly suggesting that it is the protonation of Glu²¹² that affects the chromophore properties. Glu²¹² is part of a tightly coupled network of residues that also includes His¹⁹⁴, Arg⁶⁶, and Glu¹⁴⁴. If, instead, Glu¹⁴⁴ were the residue in question, its deprotonation at pH >7.8 would cause a stabilization of the proton on the neutral trans chromophore and, concomitantly, a destabilization of the anionic cis form. Consequently, we would expect a slower thermal trans-cis isomerization but the contrary is observed (Fig. 5 A).

In the presence of anionic and neutral cis chromophores, Glu²¹² protonation occurs with pK_a values of 6.6 and 5.9, respectively. This pK_a difference of 0.7 pH units is identical to that between the chromophore’s pK_a values of 6.0 and 5.3, and it reflects their mutual coupling free energy of ∼4 kJ/mol. We note that an earlier report assumed a much stronger coupling for GFP (55). With a neutral trans chromophore, Glu²¹² is more basic (pK_a = 7.8). This pronounced pK_a difference likely originates from the changed orientation of the Glu²¹² side chain (Fig. 2). Next to the trans chromophore, it points at the free electron pair on the imidazolinone nitrogen, so it should have a rather high proton affinity. With the chromophore in cis, Glu²¹² is part of an extended, planar network parallel to the chromophore plane, in which its acidity is governed by its direct-interaction partners, His¹⁹⁴ and Arg⁶⁶.

Thermal versus photoactivated trans-cis isomerization

The metastable, trans form of mIrisGFP is produced by light irradiation and spontaneously relaxes back to cis in the dark. This process is pH-dependent; its rate coefficient, k_th, increases in two steps toward higher pH (Fig. 5 A). The first step has a pK_a of 7.8 and is, therefore, presumably connected to a change of the Glu²¹² protonation state. Upon deprotonation, the stabilizing hydrogen bond with the imidazolinone nitrogen is relinquished and, instead, the charged Glu²¹² carboxylate side chain causes electrostatic repulsion. This structural change is expected to enhance chromophore flexibility and thereby may facilitate twisting of the chromophore and isomerization to cis. Concomitantly, the Glu²¹² carboxylate reorients to form a salt bridge to the presumably positively charged His¹⁹⁴ side chain (56,57) and a hydrogen bond to a water molecule. Furthermore, an extensive network of salt bridges and hydrogen-bonding interactions is established adjacent to the chromophore, which also involves Arg⁶⁶, Glu¹⁴⁴, and Tyr¹⁷⁷ (Fig. 2 A). Such a network has also been found for the cis species of Dronpa (58) and mTFP0.7 (59). The trans-cis relaxation kinetics is further accelerated upon deprotonation of the chromophore hydroxyphenyl (pK_a > 10, Fig. 5 A). Removal of a stabilizing hydrogen bond to the Glu¹⁴⁴ side chain and introduction of electrostatic repulsion between the charged phenolate moiety and the deprotonated Glu¹⁴⁴ side chain may cause an enhanced mobility of the chromophore. Thermally activated on-switching upon chromophore deprotonation is markedly faster for mutant Glu²¹²Gln than for mIrisGFP (Fig. 5 A), supporting the idea of a more flexible chromophore. We note that the fluorescence quantum yield of the mutant in decreased, Φ = 0.26 ± 0.01, as expected for a more mobile chromophore.

Photoactivated trans-cis isomerization is also a pH-dependent process (pK_a = 7.8). Its rate coefficient, k_on, decreases with increasing pH (Fig. 5 B), in contrast to that of thermal relaxation (Fig. 5 A). Apparently, protonation of Glu²¹² has a different effect on the light-induced transition in the electronically excited state than on thermal recovery. Trans-cis photoisomerization is initiated by excitation of T^H with 405-nm light, so its kinetics depends on the probability of absorbing a photon, i.e., on the absorption at 405 nm and on the probability to subsequently isomerize. The latter is an intrinsic property of the particular chromophore species and, therefore, depends only weakly on pH due to coupling to other, more remote groups. However, the absorption of T^H at 405 nm decreases with increasing pH because of the A_T band shift, with the pH dependence described by pK_a = 7.8 (Fig. 4 F). Note that the overall concentration of T^H species and, therefore, the concentration of reactants, is essentially constant for 5 < pH < 9 because of the high pK_trans.

In general, one should expect the isomerization probability to depend on the stabilization of the chromophore in its particular isomeric state, which is substantially determined by the numbers and strengths of hydrogen-bonding interactions between the chromophore hydroxyphenyl and the protein moiety. In the C⁻ conformation of mIrisFP, the cis chromophore is engaged in hydrogen-bonding with Ser¹⁴² (2.7 Å) and two water molecules (2.7 Å), and by π-stacking with His¹⁹⁴ (20). In the T^H species, the chromophore hydrogen-bonds to Glu¹⁴⁴ (2.9 Å) and a water molecule (20) and, therefore, appears less stabilized. Indeed, the trans-cis photoisomerization (on-switching) quantum yield is much larger than the cis-trans photoisomerization (off-switching) quantum yields via either the C^H or the C⁻ states (Table 2).

Photoisomerization mechanism

Reversible photoswitching of FPs involves a light-activated change of both isomerization and protonation states of the chromophore (24,25,57,60), but the mechanistic details are not fully understood (41,57,61,62). Especially the order in which isomerization and protonation occur is still under debate (63,64). The results obtained here provide several pieces of evidence in support of the model of Olsen et al. (42). Trans-cis isomerization can be induced by 405-nm excitation of the neutral trans chromophore (Fig. 5 B) with a quantum yield of ϕ_on(405) = 0.15 ± 0.01. According to the photoswitching model (42), the excited neutral trans chromophore twists around the I-bond of the methylene bridge, deprotonates, and isomerizes to the anionic cis state (Fig. 1). The molecular structure (Fig. 2 B) suggests that Glu¹⁴⁴ may accept the proton from the neutral trans chromophore. Irradiation of the anionic cis chromophore with 473-nm light causes the opposite chromophore transition, i.e., cis-trans isomerization (off-switching), with ϕ_off(473) = 0.0045 ± 0.0004. Rather unexpectedly, mIrisGFP can be switched off much more efficiently by exciting the neutral, nonfluorescent cis chromophore (ϕ_off(405) = 0.036 ± 0.008) (Fig. 6). This observation also agrees with the result of the quantum-chemical calculations by Olsen et al. (42), who found a barrierless energy surface for excited-state isomerization of the neutral chromophore, whereas a substantial barrier was found for the anionic chromophore. In addition, one may argue that the C⁻ species should be especially stabilized by hydrogen-bonding so as to ensure a high fluorescence quantum yield, which makes a decay via isomerization less probable.

Finally, we note that off-switching by exciting the neutral cis chromophore occurs with a lower quantum yield than on-switching via the neutral trans chromophore (Table 2). This difference likely originates from the different proton-accepting properties of the moieties near the neutral hydroxyphenyl group in the cis and trans forms, which contribute to stabilization of the chromophore by hydrogen bonding or electrostatic interactions.

Conclusions for the photoswitching behavior

Reversible photoswitching of FPs may proceed along several pathways (Fig. 1). Depending on the illumination wavelength and the chromophore species present, several pathways may even be activated in parallel. So a whole range of FP properties may arise depending on a number of determinants, including the relative free energies of the protein conformations with a cis or a trans chromophore in the ground and excited states, the isomer-dependent proton affinities of the chromophore’s hydroxyphenyl group in its local environment in the ground and excited states, and the shapes of the energy surfaces connecting the different states, which control the yields of fluorescence and isomerization transitions. An important issue is also that the availability of protons at the internal donor and acceptor sites must be assured. Taking a GFP as an example, blue light (∼400 nm) excites both protonated species, C^H and T^H, and green light (∼500 nm) excites both anionic species, C⁻ and T⁻, owing to their spectral overlap. Therefore, the net result of irradiating into the blue and green bands is governed by the relative populations and isomerization yields of the cis and trans species.

If the proton affinities are such that pK_cis < pH < pK_trans, only the C⁻ and T^H species will be appreciably populated. The low pK_cis results from a strong hydrogen bond formed between a serine and the anionic cis chromophore. This bond contributes significantly to the overall stabilization of C⁻, so it is typically the thermodynamically stable and thus predominant species at neutral pH in the absence of light. Green light excites the fluorescent species, C⁻, and also switches it to the nonfluorescent T^H form; blue light induces the reverse transition from T^H to C⁻ and switches the fluorescence back on. Because irradiation into the excitation band of the fluorescent species causes a reduction of fluorescence emission, this scheme has been termed a negative switching mode (37). Dronpa (pK_cis = 5 (65)) and mTFP0.7 (pK_cis = 4.0 (32,59)) are two examples of such negative photoswitchers. If pK_cis is closer to neutral, which is relevant for marker applications, e.g., in mIrisGFP (pK_cis = 5.3/6), fastLime (27), the mGeos variants (66), rsTagRFP (pK_cis = 6.6 (29)), and rsEGFP (pK_cis = 6.6 (26)), a third state (C^H) comes into play, resulting in a lower dynamic range between the on and off states, as explained in this work in connection with the data in Fig. 7. In a similar way, for pH ≈ pK_trans, T⁻ as the third state gives rise to more complicated behavior.

For a reverse order of pKs, i.e., pK_cis > pH > pK_trans, only the nonfluorescent C^H and T⁻ species are appreciably populated. In this case, the low pK_trans results from a strong hydrogen bond formed between the amino acid at position 158 (residue numbering according to mIrisGFP), which is often a serine, and the anionic trans chromophore, so T⁻ is typically the thermodynamically stable species at neutral pH. Photoisomerization can be induced but remains unnoticed in the fluorescence emission due to the lack thereof. If, however, pK_cis ≈ pH ≫ pK_trans, the fluorescent C⁻ species is present in addition to C^H and T⁻. This scenario is characteristic of the so-called positive photoswitchers, for which fluorescence excitation enhances the emission (37). One example is Padron, with pK_cis = 6 and pK_trans = 4.5 (27,64). Excitation of the ground state T⁻ chromophore with green light generates C^H. Subsequently, the protonation equilibrium ${\text{C}}^{\text{H}}\rightleftarrows {\text{C}}^{-}$ readjusts to maintain the ratio of the two species, so the net effect is an increase in the fluorescent species, C⁻. Green-light irradiation excites C⁻ fluorescence, and moreover, photoisomerization may occur as well. However, due to preferential T⁻ excitation and, possibly, different isomerization quantum yields of T⁻ and C⁻, the net effect is an enhanced C⁻ species. Off-switching is achieved by blue-light excitation of C^H, which photoisomerizes to generate nonfluorescent T⁻. Subsequently, the pH equilibrium between C^H and C⁻ readjusts, thereby reducing the fraction of the fluorescent C⁻ species. We note that the fluorescent C⁻ is a minority species in this positive photoswitching mode, so the emission will, in principle, be significantly reduced in comparison to that of negative photoswitchers.

Red FPs can exhibit similar photoswitching properties because the chromophore extension leading to red-shifted optical spectra is practically decoupled from the protonation/isomerization properties of the hydroxyphenyl group (67). The red asFP595 also belongs to the class of positive photoswitchers (25,68). In the dark, the chromophore adopts the anionic trans state. Illumination of T⁻ induces isomerization to the C^H state (see Fig. 1 B in Chudakov et al. (69). and Schüttrigkeit et al. (70).). Because of the high pK_cis, most cis chromophores are neutral and thus nonfluorescent. The reported fluorescence quantum yield, Φ = 0.001 (71), is extremely low, presumably because it was referenced to the total amount of protein rather than the tiny fraction of C⁻ species responsible for fluorescence.

To conclude, we have studied the coupling of the chromophore to its protein environment in the reversibly photoswitchable mIrisGFP by using optical spectroscopy. By analyzing pH-dependent equilibria and kinetic properties in the native protein and the mutant Glu²¹²Gln, chromophore-protein interactions could be interpreted in structural terms based on the x-ray structure of mIrisFP (20). Our measured quantum yields of photoisomerization are in qualitative agreement with a recently published photoswitching model (42). From these results, we have put forward a general scheme as to how isomeric and acid-base properties of the chromophore in its protein environment can give rise to negative and positive photoswitching modes. This knowledge is required to further analyze the competition between photoconversion and photoswitching in mIrisFP and beneficial for the design of advanced FPs.