Understanding the Fluorescence Change in Red Genetically Encoded Calcium Ion Indicators

Abstract

For over 20 years, genetically encoded Ca²⁺ indicators have illuminated dynamic Ca²⁺ signaling activity in living cells and, more recently, whole organisms. We are just now beginning to understand how they work. Various fluorescence colors of these indicators have been developed, including red. Red ones are promising because longer wavelengths of light scatter less in tissue, making it possible to image deeper. They are engineered from a red fluorescent protein that is circularly permuted and fused to a Ca²⁺-sensing domain. When Ca²⁺ binds, a conformational change in the sensing domain causes a change in fluorescence. Three factors can contribute to this fluorescence change: 1) a shift in the protonation equilibrium of the chromophore, 2) a change in fluorescence quantum yield, and 3) a change in the extinction coefficient or the two-photon cross section, depending on if it is excited with one or two photons. Here, we conduct a systematic study of the photophysical properties of a range of red Ca²⁺ indicators to determine which factors are the most important. In total, we analyzed nine indicators, including jRGECO1a, K-GECO1, jRCaMP1a, R-GECO1, R-GECO1.2, CAR-GECO1, O-GECO1, REX-GECO1, and a new variant termed jREX-GECO1. We find that these could be separated into three classes that each rely on a particular set of factors. Furthermore, in some cases, the magnitude of the change in fluorescence was larger with two-photon excitation compared to one-photon because of a change in the two-photon cross section, by up to a factor of two.

Introduction

The advent of fluorescent genetically encoded Ca²⁺ indicators (GECIs) has been revolutionary in understanding key aspects of biology, ranging from immunology (1, 2) to neuroscience (3, 4). These engineered proteins make it possible to visualize dynamic Ca²⁺ levels in genetically specific live cells and tissues under both one-photon or two-photon excitation. There are many different GECI designs (5, 6). The most popular one is based on a single fluorescent protein (FP) that is circularly permuted in the middle of its β-barrel and then fused to calmodulin (CaM) at the new C-terminus and a CaM-binding peptide at the new N-terminus. Initially, only green FP-based indicators of this type were made (7, 8, 9). Their continuous development through several generations to increase Ca²⁺ sensitivity resulted in the widely used GCaMP6 series (10). Later, protein engineers broadened the color palette to include blue and red (11). Red GECIs are especially desirable because red light penetrates deeper in living tissue than green or blue, enabling deeper imaging (12). In addition, red-shifted excitation wavelengths are associated with less autofluorescence and photodamage. Red GECIs also provide another spectral channel for multicolor imaging and optogenetic manipulation.

Current red GECIs originate from three different circularly permuted monomeric red FPs, which in turn come from three different wild-type FPs (Fig. 1). The first red GECI, R-GECO1 (11), was based on the directed evolution of circularly permuted mApple (13) fused to the same Ca²⁺-sensing domain as the green GECIs. From the R-GECO1 template, protein engineering efforts through random mutagenesis as well as site-specific mutations yielded the improved R-GECO1.2, a red-shifted variant CAR-GECO1, a long Stokes shift variant REX-GECO1, and a blue-shifted variant O-GECO1 (14, 15). Other R-GECO1 variants include R-CaMP1.07 and R-CaMP2 (16, 17). One of the latest is jRGECO1a (18), which was generated through directed evolution with a neuron screening platform to better detect Ca²⁺ transients produced by neuronal action potentials. Another line of red GECIs based on circularly permuted mRuby (19) was created and designated the RCaMPs (RCaMP1h was the most popular variant) (20). These were also further improved with the neuron screening platform, resulting in jRCaMP1a and jRCaMP1b (18). Most recently, K-GECO1 was developed from a third circularly permuted FP, mKate-derived FusionRed (21, 22), whose parent wild-type FP originates from the same organism as mRuby’s.

Figure 1.

Genealogy of red GECIs. The ones under study are in color. Adapted from (21).

In this wide range of red GECIs, Ca²⁺ binding causes an increase in fluorescence intensity. Three factors can contribute to a Ca²⁺-dependent fluorescence change on a molecular level. The first is a shift in the protonation equilibrium of the chromophore because the protonated and deprotonated forms behave quite differently. The second is a change in the fluorescence quantum yield. The third is a change in the molecular absorption coefficient at the excitation wavelength. For one-photon excitation, this is the extinction coefficient; for two-photon excitation, it is the two-photon cross section. This last factor could depend on shifts of the excitation spectrum, and how it changes could differ between one-photon and two-photon excitation.

In a detailed study of the popular green GECI GCaMP6m, Barnett et al. found that the main mechanism of its Ca²⁺-dependent fluorescence increase was a shift in the protonation-deprotonation equilibrium of the chromophore (23). When Ca²⁺ binds, the conformational change of the protein translates into a change in the electrostatic potential around the chromophore. A consequent shift in the chromophore pK_a pushes the equilibrium from the protonated, neutral form toward the deprotonated, anionic form. For GCaMP6m, only the anionic form fluoresces when excited at the wavelengths used to detect Ca²⁺, so this results in an increase in fluorescence.

Although some of the red GECIs may depend on a similar mechanism, others appear to exhibit more of a change in the quantum yield (20, 21). How the molecular absorption coefficient plays a role is unclear. Additionally, any contrasts between the fluorescence change under one-photon and two-photon excitation have not been highlighted or explained. Our goal was to find the main factors responsible for the Ca²⁺-dependent change in fluorescence in a cohort of red GECIs.

Background and analytical considerations

Fig. 2 A presents a typical red GECI design, using K-GECO1 in its Ca²⁺-saturated state as an example (21). It consists of a red FP that is circularly permuted so that the original N- and C-termini are linked together, and new ones are introduced into the seventh β-strand of the 11 β-strand barrel adjacent to the phenolic oxygen of the chromophore. The new termini are attached via short peptide linkers to a two-part Ca²⁺-sensing domain: CaM at the C-terminus and a CaM-binding peptide at the N-terminus. Depending on the red GECI, the N-terminus arm is either the rat CaM-dependent kinase peptide ckkap (as in K-GECO1 and R-CaMP2) or the chicken myosin light-chain kinase peptide RS20 (as in the other red GECIs mentioned). In the Ca²⁺-free state, CaM and its binding peptide are apart, and the FP is dim. In the Ca²⁺-saturated state, four Ca²⁺ ions bind to CaM, and it undergoes a conformational change to wrap around the CaM-binding peptide. This causes the FP to increase in fluorescence. Because the entire construct is genetically encoded, it is possible to follow Ca²⁺ dynamics by expressing it in living systems such as neurons and imaging fluorescence over time (Fig. 2 B).

Figure 2.

Red GECIs structure, function, and chromophore. (A) Crystal structure of K-GECO1 in the Ca²⁺-saturated state (PDB: 5UKG (21)) is shown. The FP is colored red, CaM is cyan, the CaM-binding peptide (ckkap in K-GECO1) is yellow, and the Ca²⁺ ions are purple. (B) (Adapted from (21)). Shown is the imaging of spontaneous Ca²⁺ oscillations in dissociated rat hippocampal neurons with K-GECO1. Graph shows the change in fluorescence normalized to baseline fluorescence over time in three different regions of interest. Inset: widefield image of K-GECO1 expression in neurons (Scale bars, 30 μm). (C) Shown is the molecular structure of the typical red FP chromophore in its neutral and anionic forms. (D) Absorption spectrum illustrates the presence of both forms of the chromophore in a red GECI. Shown is R-GECO1 in the Ca²⁺-free state.

How can binding Ca²⁺ modulate the fluorescence of the FP? As the chromophore is the molecule responsible for the fluorescence, we begin by examining its two possible protonation states, the protonated neutral form and the deprotonated anionic form (Fig. 2 C).

Nestled inside the FP β-barrel, the red FP chromophore exists in an equilibrium between its neutral and anionic forms that can shift in a Ca²⁺-dependent manner. The neutral (protonated) form maximally absorbs light at ∼450 nm, whereas the anionic (deprotonated) form absorbs at ∼570 nm (Fig. 2 D). Each form has a distinct fluorescence quantum yield (φ), maximal extinction coefficient (ε), and maximal two-photon cross section (σ₂). These properties can depend on whether the protein is bound to Ca²⁺. Because of this, the neutral and anionic forms have their own Ca²⁺-dependent one-photon brightness (φ × ε) and two-photon brightness (φ × σ₂). These, weighted by their relative fractions (ρ), combine to make the total brightness of a GECI in the Ca²⁺-saturated and Ca²⁺-free states. Equations 1 and 2 show the total one-photon brightness (F₁) and two-photon brightness (F₂) as a function of excitation wavelength (λ), where the subscripts A and N denote the anionic or neutral form, respectively. The relative fraction ρ_X is defined by the concentration of form X divided by the total concentration of chromophore.

$$F_1\left(\lambda \right)={\rho }_A{\phi }_A{\epsilon }_A\left(\lambda \right)+{\rho }_N{\phi }_N{\epsilon }_N\left(\lambda \right),$$ (1)

$$F_2\left(\lambda \right)={\rho }_A{\phi }_A{\sigma }_{2,A}\left(\lambda \right)+{\rho }_N{\phi }_N{\sigma }_{2,N}\left(\lambda \right)\text{.}$$ (2)

The fluorescence ratio between the Ca²⁺-saturated and Ca²⁺-free states can be expressed as F_n,sat/F_n,free, where the subscript n is 1 or 2. Conveniently, at the excitation wavelengths that are relevant to sensing Ca²⁺ (that is, the F_n ratio is at or near its maximum), only one form of the chromophore per Ca²⁺ state has to be considered because the brightness of the other form is close to zero. Here, the fluorescence ratios simplify to Eqs. 3 and 4, where the subscript e denotes the “excitable” form of the chromophore: the form that yields appreciable fluorescence when excited at the wavelengths where the F_n ratio is maximal. We call it excitable because it dominates the excitation spectrum of the red fluorescence at these wavelengths. The excitable form could be neutral or anionic, and it could also be different depending on whether the protein is bound to Ca²⁺.

$${\left(\frac{F_{1,sat}}{F_{1,free}}\right)}_{max}=\frac{{\left({\rho }_e{\phi }_e{\epsilon }_e\left(\lambda \right)\right)}_{sat}}{{\left({\rho }_e{\phi }_e{\epsilon }_e\left(\lambda \right)\right)}_{free}},$$ (3)

$${\left(\frac{F_{2,sat}}{F_{2,free}}\right)}_{max}=\frac{{\left({\rho }_e{\phi }_e{\sigma }_{2,e}\left(\lambda \right)\right)}_{sat}}{{\left({\rho }_e{\phi }_e{\sigma }_{2,e}\left(\lambda \right)\right)}_{free}}\text{.}$$ (4)

Because there are two forms of the chromophore, measuring the quantum yield, or φ, corresponding to just the excitable form can require a more involved approach. Specifically, this can occur when the excitable form of the chromophore is anionic, but the concentration of the neutral form is significant. In this case, the red tail of the neutral absorption band can overlap considerably with the blue tail of the anionic absorption band (as in Ca²⁺-free R-GECO1, see Fig. 2 D). Therefore, both forms could contribute to the optical density (OD) at a typical excitation wavelength (that does not overlap with emission) used to excite the anionic form and measure its full fluorescence spectrum. We developed a new method that accounts for this and produces the correct φ of the anionic form (see Materials and Methods).

Finding the extinction coefficient, ε, of the excitable form is also a more involved process because of the presence of two forms of the chromophore. With the alkaline denaturation method, both forms of the chromophore give rise to the denatured form with the known ε. Because of this, the extinction coefficients listed in the original red GECIs articles (11, 14, 15, 18, 20, 21) are effective values and technically equal to a product of ε_e and ρ_e. We measured the individual ε values for both the neutral and the anionic form of each protein in both Ca²⁺-free and -bound states (see Materials and Methods).

The first two factors, ρ_e and φ_e, are the same for both one-photon and two-photon excitation. However, the ratios ε_e,sat/ε_e,free and σ_2,e,sat/σ_2,e,free might not be equal since ε and σ₂ can depend differently on the local environment of the chromophore (24, 25, 26). Therefore, the Ca²⁺-dependent change in fluorescence could depend on the mode of excitation.

Materials and Methods

Fluorescence quantum yields

Fluorescence quantum yields (φ) were measured with an integrating sphere fluorometer (Quantaurus-QY; Hamamatsu Photonics, Hamamatsu City, Japan). The solvent-only reference measurement was done in the same cuvette as the sample (1 cm pathlength quartz cuvette). The peak OD of the sample was less than 0.1. Values presented in Table 1 are from one independent measurement that includes averaging three to four reference measurements and 3–20 sample measurements from the same software acquisition file.

TABLE 1.

Photophysical Properties of the Nine Red GECIs

Class	Protein	Ca2+	ea	ρe ± 11%b	φe ± 5%	εA,max ± 5% mM−1cm−1 (λ nm)	εN,max ± 5% mM−1cm−1 (λ nm)	σ2,e,max ± 17% GM (λ nm)	F1,max ± 13% mM−1cm−1 (λ nm)	F2,max ± 21% GM (λ nm)	F1 max. ratio ± 9% (λRmax nm)c	F2 max. ratio ± 9% (2λRmax nm)c
I	R-GECO1	−	A	0.06	0.15	89 (577)	33 (447)	24 (1072)	0.8 (577)	0.21 (1072)	23 (516)	38 (1032)
		+	A	0.82	0.21	69 (563)	30 (451)	31 (1056)	12 (563)	5 (1056)
	R-GECO1.2	−	A	0.033	0.21	71 (573)	27 (443)	21 (1072)	0.49 (573)	0.14 (1072)	45 (514)	73 (1028)
		+	A	0.8	0.31	56 (558)	26 (445)	31 (1044)	14 (558)	8 (1044)
	jRGECO1a	−	A	0.054	0.18	88 (577)	32 (447)	18 (1076)	0.8 (577)	0.17 (1076)	23 (514)	44 (1028)
		+	A	0.86	0.21	74 (563)	39 (443)	31 (1056)	13 (563)	5 (1056)
	O-GECO1	−	A	0.0076	0.22	80 (557)	30 (433)	19 (1044)	0.14 (557)	0.032 (1044)	140 (502)	220 (1004)
		+	A	0.9	0.23	59 (544)	23 (422)	24 (1028)	12 (544)	5 (1028)
	CAR-GECO1	−	A	0.039	0.17	92 (575)	38 (447)	25 (1072)	0.59 (575)	0.16 (1072)	42 (516)	62 (1032)
		+	A	0.9	0.26	71 (563)	40 (425)	36 (1060)	17 (563)	8 (1060)
II	K-GECO1	−	A	0.25	0.19	75 (569)	45 (451)	16 (1068)	3.6 (569)	0.8 (1068)	11 (530)	13 (1060)
		+	A	0.75	0.48	84 (566)	34 (440)	23 (1068)	30 (566)	8 (1068)
	jRCaMP1a	−	A	0.55	0.29	78 (573)	37 (451)	25 (1076)	13 (573)	4.0 (1076)	3.4 (518)	4.3 (1036)
		+	A	0.8	0.54	77 (572)	24 (450)	23 (1076)	34 (572)	10 (1076)
III	REX-GECO1	−	A	0.43	0.05	89 (583)	34 (449)	24 (1076)	2.1 (583)	0.6 (1076)	32 (450)	31 (900)
		+	N	1	0.21	83 (577)	33 (489)	35 (920)	6.9 (489)	7 (920)
	jREX-GECO1	−	A	0.37	0.06	86 (579)	32 (447)	30 (1080)	1.9 (579)	0.7 (1080)	28 (450)	26 (900)
		+	N	1	0.21	78 (570)	34 (488)	39 (920)	7.0 (488)	8 (920)

^aExcitable form of the chromophore, anionic (A) or neutral (N).

^bEstimated relative SD.

^cThe F_n maximal ratios were measured directly by taking the ratio of total fluorescence signals normalized to the known relative protein concentration between the Ca²⁺-free and Ca²⁺-saturated samples.

It is trivial to measure the φ of a sample with only one chromophore. However, in these Ca²⁺ indicators, there are two forms of the red chromophore that maximally absorb at distinct wavelengths: the neutral form at ∼450 nm and the anionic form at ∼570 nm. We measured the φ of the excitable form of the chromophore: the form that dominates the fluorescence signal when excited at the wavelengths where the Ca²⁺-induced fluorescence change is largest. If the excitable form is anionic and the relative fraction of the neutral form is significant, then because of spectral overlap, both forms could contribute to the absorbance at a typical excitation wavelength used to excite the anionic form and measure its full fluorescence spectrum. We developed the following method to correct for this.

The method requires measurements at two different wavelengths. One is at the typical excitation wavelength to get the full emission spectrum of the anionic form, called λ_blue, where the neutral form still absorbs. The other is at the peak of the anionic form absorption, λ_red, where the neutral form absorption becomes insignificant in comparison. With information from these two measurements and the following equations, we can get the correct φ of the anionic form.

In an integrating sphere, the φ is calculated as the integral of the emission spectrum of the sample, P, divided by the difference of the integrals of the excitation light intensity transmitted by the reference (solvent only), R, and the sample, S: φ = P/(R − S). The integral P is proportional to the total number of photons emitted by the sample, and (R − S) is proportional to the number of photons absorbed. Each integral is written explicitly in Eqs. 5, 6, and 7, where λ_ex is the excitation wavelength, Δλ is the excitation half-bandwidth (∼10 nm), f(λ) is the fluorescence spectrum, and R(λ) and S(λ) are the spectral distributions of excitation light intensity reaching the detector from the reference and sample, respectively.

$$P=\underset{{\lambda }_{ex}+\mathit{\Delta }\lambda }{\overset{\infty }{\int }}f\left(\lambda \right)d\lambda ,$$ (5)

$$R=\underset{{\lambda }_{ex}-\mathit{\Delta }\lambda }{\overset{{\lambda }_{ex}+\mathit{\Delta }\lambda }{\int }}R\left(\lambda \right)d\lambda ,$$ (6)

$$S=\underset{{\lambda }_{ex}-\mathit{\Delta }\lambda }{\overset{{\lambda }_{ex}+\mathit{\Delta }\lambda }{\int }}S\left(\lambda \right)d\lambda \text{.}$$ (7)

When exciting the sample at λ_red, the absorbance (R_red − S_red) should belong solely to the anionic form. However, the measured effective φ can be underestimated or overestimated because of two counteracting factors. First, only part of the emission spectrum can be integrated because the excitation light overlaps with it. We define this “short” fluorescence integral as H_red (Eq. 8). Second, the part of the fluorescence resonant with the excitation light contributes to the integral of the excitation light intensity of the sample S_red, resulting in the underestimation of absorbed photons. We call this the “very short” fluorescence integral E_red, which is the sample emission integrated over the bandwidth of excitation (Eq. 9). The correct φ, denoted C, requires the “full” integral, or the integral of the full fluorescence spectrum, F_red. It is described by Eq. 10.

$$H_{red}=\underset{{\lambda }_{red}+\mathrm{\Delta }\lambda }{\overset{\infty }{\int }}f\left(\lambda \right)d\lambda ,$$ (8)

$$E_{red}=\underset{{\lambda }_{red}-\mathrm{\Delta }\lambda }{\overset{{\lambda }_{red}+\mathrm{\Delta }\lambda }{\int }}f\left(\lambda \right)d\lambda ,$$ (9)

$$C=\frac{F_{red}}{R_{red}-\left(S_{red}-E_{red}\right)}\text{.}$$ (10)

To find C, we make an additional measurement with the excitation wavelength λ_blue, which is out of the range of emission. With this measurement, we can get the relative values of the full, short, and very short fluorescence integrals as F_blue, H_blue, and E_blue, respectively. Now we can use the relations F_red/H_red = F_blue/H_blue and E_red/H_red = E_blue/H_blue to find C (Eq. 11).

$$C=\frac{\frac{H_{red}}{H_{blue}}{F}_{blue}}{R_{red}-\left(S_{red}-\frac{H_{red}}{H_{blue}}{E}_{blue}\right)}\text{.}$$ (11)

This method can be applied to find the correct quantum yield of the redder chromophore in a mixture of two chromophores as long as the emission of the bluer chromophore is negligible within the range of emission of the redder chromophore when exciting at λ_blue. This method was checked with a mixture of 6.5 μM 3,3′-dioctadecyloxacarbocyanine perchlorate (absorption peak 487 nm) and 1.2 μM Cresyl violet (CV) (absorption peak 601 nm) in ethanol (Fig. S1 A). The mixture was measured with λ_blue = 520 nm (where the absorption of the two overlap) and λ_red = 600 nm. The correct quantum yield C of CV as calculated with Eq. 11 was identical (<1% difference) to the quantum yield obtained with an excitation wavelength in which 3,3′-dioctadecyloxacarbocyanine perchlorate does not absorb (565 nm). The short and very short fluorescence integrals of CV in this case are illustrated in Fig. S1 B.

We applied this method to measure the correct quantum yield C of the anionic form for the Ca²⁺-free states of jRGECO1a, K-GECO1, jRCaMP1a, CAR-GECO1, O-GECO1, R-GECO1, and R-GECO1.2 (values are displayed in Table 1). For the Ca²⁺-saturated states of all proteins, and the Ca²⁺-free states of REX-GECO1 and jREX-GECO1, no corrections were necessary. In these cases, the correct quantum yields in Table 1 were simply measured with λ_blue excitation.

Chromophore pK_a values and extinction coefficients

The Ca²⁺-free and Ca²⁺-saturated states of each protein were subjected to stepwise alkaline and acid titrations (Figs. 4, S7, and S8) by gradually adding 2–10 μL of 0.1–1 M NaOH or HCl, respectively, to a sample of 500–1200 μL. Absorption was measured at each step with a LAMBDA 950 Spectrophotometer (PerkinElmer, Waltham, MA), and pH was measured with an Orion PerpHecT ROSS combination pH microelectrode (Thermo Fisher Scientific, Waltham, MA). For the alkaline titrations of the Ca²⁺-saturated state, the protein was first diluted 1:100 into a pH 7.2 buffer with 30 mM 3-(N-morpholino)propane sulfonic acid (MOPS), 100 mM KCl, 50 μM CaCl₂, and no EGTA because at a higher pH the K_d of EGTA for Ca²⁺ decreases dramatically (27). To determine the chromophore pK_a values, the OD at the peak of the anionic absorbance (even if it shifted throughout the titration) was plotted versus pH in a range in which it was not denaturing, and the protein was not precipitating (which happened in some instances at the end of the acid titrations). The plot was fitted with 1, 2, or 3 pK_a values using OriginPro 2017 (OriginLab, Northampton, MA) and Eq. 12 in which ODmax_n is the maximal OD of specie n for pK_a,n (and was a free parameter):

$$OD=\frac{ODma{x}_1·{10}^{pH-p{K}_{a,1}}}{1+{10}^{pH-p{K}_{a,1}}}+\frac{ODma{x}_2·{10}^{pH-p{K}_{a,2}}}{1+{10}^{pH-p{K}_{a,2}}}+\frac{ODma{x}_3·{10}^{pH-p{K}_{a,3}}}{1+{10}^{pH-p{K}_{a,3}}}.$$ (12)

Figure 4.

Absorbance pH titrations and shifts in pK_a are starkly different between classes and between their Ca²⁺-free and Ca²⁺-saturated states. Titrations from neutral to acidic pH and neutral to alkaline pH were done separately and then combined. The figures are normalized to the observed maximal absorbance of the anionic form. (A–F) Representative spectra from absorbance pH titrations are shown, measured in either Ca²⁺-free or Ca²⁺-saturated buffer as noted. Each spectrum was measured at the pH value designated in the legends. The final spectrum (red dashed line) belongs to the denatured chromophore. (G–I) Apparent pK_a curves are shown, showing the OD of the anionic form as a function of pH. The pK_a values are indicated on the fitted curve. (H) Only part of the K-GECO1 titrations could be fitted because of protein precipitation at pH values below those displayed and early denaturation of the anionic form at higher pH values.

The anionic and neutral forms of the chromophore have distinct extinction coefficients (ε). To get the ε values of the anionic form for R-GECO1, R-GECO1.2, O-GECO1, jRGECO1a, REX-GECO1, and jREX-GECO1, we took the OD of the anionic peak at the pH in which it reached its maximum during the alkaline titration, divided it by the maximal OD of the alkaline denatured form with a known ε (44,100 M⁻¹cm⁻¹ (28)), and multiplied it by that known ε. The presence of the immature green chromophore was considered to be insignificant in these proteins. For R-GECO1, R-GECO1.2, O-GECO1, and jRGECO1a, in which the anionic peak shifted red at pH ≳ 8.5 in the Ca²⁺-saturated state, we took the maximal OD before the shift. The shift is probably due to a titration of residue(s) near the chromophore, and our method provided the ε of the chromophore in the environment corresponding to pH 7.2. To get the ε of the neutral form using the now-known ε of the anionic form, we plotted the anionic peak OD versus the neutral peak OD across the titration. Where the dependence was linear, we used the slope to calculate the neutral ε. The OD at the neutral and denatured peaks was corrected for any contribution of the blue tail of the anionic absorption, using the spectrum at the pH when the anionic form is maximal as a reference. This slope method is further detailed in (23).

For CAR-GECO1, K-GECO1, and jRCaMP1a, the slope method was used to find both the ε value of the anionic form and the neutral form because of the significant presence of the immature green chromophore in CAR-GECO1 and the appearance of a 380 nm peak in K-GECO1 and jRCaMP1a during denaturation. To obtain the anionic ε for these proteins, we plotted the OD of the anionic form versus the OD of the ∼450 nm denatured form and fit the slope at the end of the titration where it was linear and, in the case of K-GECO1 and jRCaMP1a, where the 380 nm peak had already reached its maximum. Linear plots with a fitted slope are displayed in Fig. S2 using CAR-GECO1 as an example. For special corrections to the pH titrations of jREX-GECO1 and REX-GECO1, see Supporting Materials and Methods.

Determination of the relative fraction of the excitable form of the chromophore

The relative fraction of the excitable form of the chromophore in the Ca²⁺-free and Ca²⁺-saturated states for each protein, ρ_e, was calculated with Eq. 13 using the absorption spectra collected at pH 7.2.

$${\rho }_e=\frac{\frac{O{D}_e}{{\epsilon }_e}}{\frac{O{D}_e}{{\epsilon }_e}+\frac{O{D}_{ne}}{{\epsilon }_{ne}}},$$ (13)

where ε is the extinction coefficient; the subscripts e and ne stand for the excitable and nonexcitable form, respectively. For special corrections to determining the relative fraction for K-GECO1 and CAR-GECO1, see Supporting Materials and Methods.

Two-photon characterization

To collect the two-photon absorption spectra, the tunable femtosecond laser InSight DeepSee (Spectra-Physics, Santa Clara, CA) was used to excite the fluorescence of the sample contained in a 3 mm glass cuvette (Starna Cells, Atascadero, CA) within a PC1 Spectrofluorimeter (ISS, Champaign, IL). The laser was automatically stepped to each wavelength over the spectral range with a custom LabVIEW program (National Instruments, Austin, TX), with at least 30 s at each wavelength to stabilize. We measured two samples per laser scan by using both the sample and reference holders and switching between them with the autoswitching mechanism on the PC1. The laser was focused on the sample through a 45-mm near-infrared achromatic lens with antireflection coating from 750–1550 nm (Edmund Optics, Barrington, NJ). Fluorescence was collected from the first 0.7 mm of the sample at 90° with the standard PC1 collection optics through a combination of either 680/SP, 633/SP, or 694/SP filters together with a 745/SP filter (Semrock) to remove all laser-scattered light and a 570/longpass filter (Edmund Optics) to remove the residual green fluorescence of the immature chromophore. To correct for wavelength-to-wavelength variations of laser parameters, we used LDS798 (Exciton, Lockbourne, OH) in 1:2 CHCl₃:CDCl₃ as a reference standard between 900 and 1240 nm (29) and Coumarin 540A (Exciton) in 1:10 dimethyl sulfoxide:deuterated dimethyl sulfoxide between 700 and 952 nm (30). Adding the deuterated solvents (MilliporeSigma, Darmstadt, Germany) was necessary to decrease near-infrared solvent absorption. The standards were measured at least twice a week in both the sample and the reference holder. All the dyes were magnetically stirred throughout the measurements. Quadratic power dependence of fluorescence in the proteins and standards was checked at several wavelengths across the spectrum.

For the two-photon cross sections, we measured each protein versus Rhodamine 6G in MeOH at 1052 nm (two-photon cross section, σ₂ = 9.9 GM (31)). Rhodamine 6G was chosen because the value for σ₂ at this wavelength agrees across the literature (30, 31). (Rhodamine B, which was the standard for previous two-photon measurements of some red GECIs, is pH-sensitive with variable σ₂ values.) Power dependence of fluorescence intensity was recorded with the PC1 monochromator at a wavelength in which the OD was ≳ 0.05 and fitted to a parabola with the curvature coefficient proportional to σ₂. These coefficients were normalized for the differential fluorescence quantum yield at the registration wavelength and the concentrations. The differential quantum yields of the standard and the sample were obtained using the data from the fluorescence quantum yield measurements and selecting an integral range from ∼5 nm to the left and right of the registration wavelength. Concentrations were determined by Beer’s Law. For Rhodamine 6G in MeOH, ε = 100,500 M⁻¹cm⁻¹ (31). For most of the proteins (except for the Ca²⁺-saturated state of REX-GECO1 and jREX-GECO1), the σ₂ at 1052 nm corresponds to the anionic form of the chromophore, and the concentration was determined with the maximal OD of the anionic form and its ε. For the Ca²⁺-saturated state of REX-GECO1 and jREX-GECO1, the OD at the peak of the absorption spectrum and the ε of the neutral chromophore was used to find the concentration.

Direct measurements of Ca²⁺-saturated/free F₁ and F₂ ratios

We measured the Ca²⁺-saturated/free F₁ and F₂ ratios for each protein directly by comparing the Ca²⁺-saturated and Ca²⁺-free states with known concentrations of protein relative to each other. To make the samples, a portion of Ca²⁺-free protein was diluted 10 times to a total volume of 150 μL into a Ca²⁺-saturated buffer (30 mM MOPS, 100 mM KCl, and 10 mM CaCl₂) and then the pH was adjusted back to 7.2 with 3–5 μL of 0.1 M NaOH. An equivalent volume of 18.2 MΩ water was added to the Ca²⁺-free sample to maintain a known concentration ratio between the two samples.

For the F₂ ratio spectra, the shape was calculated by dividing the Ca²⁺-saturated two-photon excitation spectrum by the Ca²⁺-free two-photon excitation spectrum. Then, the ratio of the fluorescence signals was measured at three laser wavelengths around the maximal ratio. The setup from the two-photon characterization section was employed, and fluorescence was collected through 745/shortpass (SP) and 770/SP filters (Semrock, Rochester, NY). After applying the known concentration ratio, the F₂ ratio spectrum was normalized to the three fluorescence signal ratios, and the results were averaged (Fig. S3).

For the F₁ ratios, the samples were diluted further into 1 cm glass cuvettes for a total volume of 2 mL. The excitation spectrum for each sample was scanned with the LS 55 Fluorescence Spectrometer (PerkinElmer) with excitation slits set to 10 nm, and the entire fluorescence was collected by setting the emission grating to the 0-th diffraction order. The background fluorescence spectrum of water in a 1 cm glass cuvette was subtracted from each spectrum, scaled so that the fluorescence at 650 nm became approximately zero. The ratio between the Ca²⁺-saturated and Ca²⁺-free spectra was calculated and adjusted to account for the concentration difference between the two.

Results

The red GECIs studied here are highlighted in Fig. 1: R-GECO1, R-GECO1.2, CAR-GECO1, O-GECO1, REX-GECO1, jRGECO1a, jRCaMP1a, and K-GECO1. Although the chromophores in O-GECO1 and jRCaMP1a have slightly different structures than the typical red FP chromophore displayed in Fig. 2 C (Fig. S4), the same analysis still applies because they also undergo the protonation-deprotonation reaction. We additionally present a new red GECI, jREX-GECO1, which is derived from REX-GECO1 and incorporates mutations present in jRGECO1a. A protein sequence alignment of these nine GECIs can be found in Fig. S5.

The first clue to the mechanism behind the Ca²⁺-dependent change in fluorescence comes from the one-photon absorption spectra of the Ca²⁺-free and saturated states (Fig. 3). Any changes in spectra between the two states could reveal changes in the neutral/anionic chromophore equilibrium and thus the relative fraction of the excitable form, ρ_e. In all but two of the indicators, the concentration of the anionic form appears to increase when Ca²⁺ binds. The two exceptions, jREX-GECO1 (Fig. 3 G) and REX-GECO1 (Fig. 3 H), display an opposite effect: the chromophore distribution shifts from predominantly anionic in the Ca²⁺-free state to neutral in the Ca²⁺-saturated state. Among the other seven, there is a consistent blue shift of the anionic form absorption peak upon Ca²⁺ binding. K-GECO1 (Fig. 3 D) and jRCaMP1a (Fig. 3 E) stand out as having both a smaller increase and a shorter blue shift of the anionic form compared to the other five (≤2 nm vs. ≥13 nm). The spectral shifts are also apparent in the one-photon excitation and emission spectra (Fig. S6). Based on these observations, these nine red GECIs can be separated into three classes: class I, which includes jRGECO1a, CAR-GECO1, R-GECO1, R-GECO1.2, and O-GECO1 (names colored blue in the figures); class II, which includes K-GECO1 and jRCaMP1a (names colored gold); and class III, which includes jREX-GECO1 and REX-GECO1 (names colored magenta).

Figure 3.

The red GECIs under study can be separated into three classes based on their one-photon absorption spectra. Plotted are the respective spectra of the nine indicators under study (A–I). In each plot, the Ca²⁺-free state (green) and Ca²⁺-saturated state (purple) are shown at the same protein concentration and normalized to the peak absorbance in the Ca²⁺-saturated state. The names are colored by class: blue for class I, gold for class II, and magenta for class III.

Again, the excitable form of the chromophore is the one that dominates the excitation spectrum at the wavelengths where the Ca²⁺-dependent fluorescence change is maximal. In classes I and II, we find that the excitable form of the chromophore is anionic regardless of the presence of Ca²⁺. In class III, the excitable form is anionic in the Ca²⁺-free state and neutral in the Ca²⁺-saturated state (Fig. S6). Quantitative values of ρ_e are displayed in Table 1.

We can gain insight into why ρ_e changes with Ca²⁺ binding by looking at the one-photon absorption spectra as a function of pH and finding the apparent pK_a values for each Ca²⁺ state (Figs. 4, S7, and S8). Starting from Fig. 4, we chose to highlight jRGECO1a, K-GECO1, and jREX-GECO1 as examples of their respective classes. Figures for the other six red GECIs can be found in the Supporting Materials and Methods. The absorbance pH titrations generally show the neutral form titrating to the anionic form as the pH increases, and they display different behaviors depending on both the Ca²⁺ state and the class.

In class I, when no Ca²⁺ is present, the titration displays a clean isosbestic point between the neutral and anionic peaks belonging to the red chromophore and no spectral shifts (Figs. 4 A, S7, A–C, and S8 A). On the contrary, in the Ca²⁺-saturated state, the anionic form undergoes clear spectral shifts (Figs. 4 D, S7, D–F, and S8 D). The Ca²⁺-free state has one wave of titration and one apparent pK_a, whereas the Ca²⁺-saturated state exhibits two to three waves of titration, each corresponding to a different pK_a value (Figs. 4 G, S7, G–I, and S8 G). This is likely due to titratable residues near the chromophore in the Ca²⁺-saturated state that affect its pK_a as well as the absorption of the anionic form (32, 33, 34). The titration of jRGECO1a exemplifies these observations. In the Ca²⁺-free titration (Fig. 4 A), there is just one pK_a, and the peak position of the anionic form stays at 577 nm (Fig. 4 G). For Ca²⁺-saturated jRGECO1a, there are three apparent pK_a values (Fig. 4 D), and the absorption peak of the anionic form shifts to three distinct wavelengths throughout the titration (Fig. 4 A): 577 nm at pH ≤ ∼5.5, 563 nm from pH ∼6.5–9, and 575 nm at pH ≥ ∼9.

The titrations in class II were best fit with a single pK_a (Figs. 4 H and S8 H), except for K-GECO1 in the Ca²⁺-free state, which displayed a broad titration curve that could only be satisfactorily fit with two pK_a values. There are subtle spectral shifts of the anionic peak in both Ca²⁺ states of class II; it shifts blue as it increases and shifts red right before it denatures (Figs. 4, B and E and S8, B and E). The pH titrations in class II are also characterized by a peak at ∼380 nm that irreversibly appears at alkaline pH, which may be due to the denatured form of the cleaved protein chromophore shown in Fig. S4 C (20, 22, 35, 36, 37).

Interestingly, in class III, when titrating the Ca²⁺-saturated state from neutral to acidic pH, the spectrum unexpectedly morphs from the broad ∼490 nm peak into two peaks that presumably belong to the neutral and anionic forms of the chromophore (Figs. 4 F and S8 F). This could be due to protonation of the Glu49 residue (numbered as in Fig. S5) near the phenolic oxygen of the chromophore that somewhat stabilizes the anionic form. The clean isosbestic point when titrating from neutral to alkaline pH between the 490 nm form and the 570 nm form suggests that the 490 nm form is singularly neutral. Assuming it is, the alkaline titration can be fit with one very high pK_a (11.63 ± 0.02 for jREX-GECO1). The Ca²⁺-free titrations of class III are best fit with three pK_a values (Figs. 4 I and S8 I), and spectral shifts of the anionic form are present.

For classes I and II in which the excitable form is always anionic, the Ca²⁺-induced changes in ρ_e qualitatively follow from the shifts in apparent pK_a (Table S1). In class I, the Ca²⁺-saturated pK_a with the largest amplitude change of the curve is always the first (most acidic) one (Figs. 4 G, S7, G–I, and S8 G). It is usually ∼3.5 pH units smaller than the Ca²⁺-free pK_a. In class II, the pK_a changes less between Ca²⁺ states. For K-GECO1, the shift is ∼0.4 pH units if we consider the first wave of titration in the Ca²⁺-free state, and jRCaMP1a has a shift of only 0.3 pH units. The Ca²⁺-saturated/free ρ_e ratios of class I and class II reflect their differences in pK_a shifts. In class I, the ratio ranges between 14 and 121, whereas in class II, the ρ_e ratio is 3 for K-GECO1 and 1.5 for jRCaMP1a. In class III, there is a large pK_a shift in the opposite direction when Ca²⁺ binds, showing how the neutral form is favored in the Ca²⁺-saturated state (Figs. 4 I and S8 I).

After investigating the change in ρ_e, the next clue is the fluorescence quantum yield of the excitable form, φ_e (Table 1). In class I, φ_e ranges from 0.15 to 0.31, and φ_e increases when Ca²⁺ is bound, from an insignificant change in O-GECO1 to 1.5 times in CAR-GECO1. In class II, the φ_e values in the Ca²⁺-saturated state are higher than in class I (0.48–0.54). Also, the Ca²⁺-dependent increase of φ_e is more prominent: 1.85 times for jRCaMP1a and 2.5 times for K-GECO1. Class III has the largest increases of φ_e (2.5–3.5 times) because whereas the Ca²⁺-saturated φ values of the neutral form equal 0.21, the φ values of the Ca²⁺-free anionic form are much lower (0.05–0.06). In every class the φ_e increases when Ca²⁺ binds, although the degree to which it does depends on the class.

The final factor contributing to the Ca²⁺-saturated/free ratio of the fluorescence signal is the molecular absorption coefficient of the excitable form and how it changes because of Ca²⁺ at the excitation wavelength. This is the factor that can be quite different in one-photon versus two-photon excitation. Directly measuring the Ca²⁺-saturated/free F₁ and F₂ ratios as a function of excitation wavelength can reveal these differences. To do this, we took the ratio of the integrated fluorescence signal across the excitation spectrum for samples with and without Ca²⁺, normalized to known relative concentrations of protein. When the Ca²⁺-saturated/free F₁ and F₂ ratios are plotted together versus transition wavelength (corresponding to half the laser wavelength in the two-photon case), the shapes of the ratios look similar for both modes of excitation (Figs. 5, A–C, S9, A–C, and S10, A–C). However, the absolute values of the ratios are not always the same. This is especially apparent at the transition wavelength where the ratios are maximal, marked with a vertical dashed line, and called λ_Rmax. In classes I and II, the ratio spectra have a double-humped nature because of the spectral shift of the anionic form between Ca²⁺ states, and we selected the shorter wavelength hump as λ_Rmax for consistency. For all members of class I, the F₂ ratio is significantly larger than the F₁ ratio by 1.5–1.9 times as exemplified by jRGECO1a (Figs. 5 A, S9, A–C, and S10 A). For classes II and III, represented by K-GECO1 and jREX-GECO1, respectively, the F₁ and F₂ ratios are not significantly different (Figs. 5, B and C and S10, B and C).

Figure 5.

Spectral analysis of jRGECO1a (row 1), K-GECO1 (row 2), and jREX-GECO1 (row 3), representing classes I, II, and III of red GECIs, respectively. The vertical dashed line on each plot indicates the transition wavelength of excitation (λ_Rmax) where the Ca²⁺-saturated/free F₁ and F₂ ratios are approximately maximal. (A–C) Ca²⁺-saturated/free (Sat/Free) F₁ (dotted line) and F₂ (solid line) ratios as a function of excitation wavelength were measured directly by taking the ratio of the integrated fluorescence signal normalized to the known relative protein concentration between the Ca²⁺-free and Ca²⁺-saturated samples. (D–F) Shown are the spectra of the one-photon brightness (F₁, dotted lines, left and bottom axes) and two-photon brightness (F₂, solid lines, right and top axes) of the Ca²⁺-free and Ca²⁺-saturated states. (G–I) Shown are the one-photon absorption (1PA) and two-photon absorption (2PA) spectra of the excitable form of the chromophore, shown in values of ε_e (left and bottom axes) and σ_2,e (right and top axes), respectively, for the Ca²⁺-free and Ca²⁺-saturated states.

To get the molecular brightness spectra for the Ca²⁺-free and saturated states separately, we normalized their one- and two-photon absorption spectra to F₁ and F₂, respectively (Figs. 5, D–F, S9, D–F, S10, D–F, and S11). F₁ and F₂ values were calculated from the independently measured parameters ρ_e, φ_e, ε_e, and σ_2,e. Next, to explain the differences between the directly measured F₁ and F₂ ratios for class I versus the other classes, we normalized the one- and two-photon absorption spectra of the excitable form to ε_e and σ_2,e, respectively (Figs. 5, G–I, S9, G–I, and S10, G–I). At λ_Rmax (again marked with a vertical dashed line), the σ_2,e of jRGECO1a increases by 2.4 times in the Ca²⁺-saturated state. In comparison, the increase of ε_e at λ_Rmax is much smaller (∼1.2 times) (Fig. 5 D). A similar phenomenon occurs for the rest of class I. Curiously, ε_e at λ_Rmax only increases because of the spectral shift, whereas ε_e at the peak actually decreases slightly for class I when Ca²⁺ binds (Table 1). In class II, there is no significant change of σ_2,e or ε_e (Figs. 5 H and S10 H). Although the σ_2,e of K-GECO1 appears to increase, the uncertainty of the measurement is too large to consider it statistically significant (Table 1). Finally, class III, represented by jREX-GECO1, exhibits equivalent Ca²⁺-induced increases of ε_e and σ_2,e (Figs. 5 I and S10 I). They both increase ∼3.6 times on average at λ_Rmax. This is due to the switch of the excitable form from the anionic to the neutral, and the major shift between their absorption spectra under both one- and two-photon excitation.

The final question remains: of the three different factors, which one matters the most for the Ca²⁺-saturated/free fluorescence ratios? We have found that the answer to this question depends on the indicator class. The Ca²⁺-saturated/free ratios for each parameter are visualized in Figs. 6 and S12. Representing class I, jRGECO1a (Fig. 6 A) has a ρ_e,sat/ρ_e,free ratio that far exceeds the other ratios. An increase in σ_2,e increases the F₂ ratio by two times. From class II, K-GECO1 (Fig. 6 B) has comparable contributions to the overall F_n ratios from changes in both ρ_e and φ_e. Last, in class III, all three factors play a significant role as illustrated by the example of jREX-GECO1 (Fig. 6 C).

Figure 6.

Each class depends on a different set of factors for the Ca²⁺-induced fluorescence change. Shown are Ca²⁺-saturated/free ratios (Sat/Free) of ρ_e, φ_e, ε_e(λ_Rmax), and σ_2,e(λ_Rmax) of (A) jRGECO1a, (B) K-GECO1, and (C) jREX-GECO1. Insets: Ca²⁺-saturated/free F₁ and F₂ ratios were both calculated from the independent measurements of ρ_e, φ_e, ε_e, and σ_2,e (tan) and measured directly (green) as described for Fig. 5. The horizontal dashed line marks a Ca²⁺-saturated/free (Sat/Free) ratio of 1 (no change) for each plot. Error bars represent estimated SDs.

Discussion

We set out to determine the main factors that contribute to the Ca²⁺-dependent fluorescence change in nine red GECIs under both one-photon and two-photon excitation. The key parameters that could change because of Ca²⁺ are the following: 1) the relative fraction of the chromophore in the excitable protonation state (ρ_e), 2) the quantum yield of the excitable state (φ_e), and 3) the extinction coefficient (ε_e) or two-photon cross section (σ_2,e) of the excitable state at the excitation wavelength.

We found that the nine red Ca²⁺ indicators could be separated into three different classes, each of which relied on different factors. Class I consists of jRGECO1a, CAR-GECO1, R-GECO1, R-GECO1.2, and O-GECO1. Class II includes K-GECO1 and jRCaMP1a. In these two classes, the excitable form of the chromophore is anionic for both the Ca²⁺-free and saturated states. Class III contains jREX-GECO1 and REX-GECO1. It is unique because the excitable form switches from anionic in the Ca²⁺-free state to neutral in the Ca²⁺-saturated state. The fluorescence change in class I is mostly due to a dramatic increase in ρ_e because of a large Ca²⁺-dependent shift in pK_a. This is much like GCaMP6m (23). Additionally, an increase in σ_2,e when Ca²⁺ binds contributes to the F₂ ratio by approximately twofold on average for class I. The indicators in class II depend on both a Ca²⁺-induced increase in ρ_e and φ_e. In class III, all the factors are comparably important to the change in fluorescence. Classes I and II were previously alluded to in (20), where the importance of the change in the chromophore pK_a between Ca²⁺ states was highlighted for R-GECO1, and the change in φ was seen to be more important for the RCaMP series (which jRCaMP1a was based on).

The spectral shapes of the Ca²⁺-saturated/free fluorescence excitation ratios are very similar under one-photon excitation and two-photon excitation. Interestingly, for class I, the value of the fluorescence ratio is greater under two-photon excitation compared to one-photon excitation at λ_Rmax. This is because the σ₂ of the anionic form increases upon Ca²⁺ binding, whereas the ε of the anionic form changes little or not at all. The increase in σ₂ for the anionic chromophore can be explained by its sensitivity to the local electric field created by the protein surrounding (24, 25, 26). This sensitivity is due to a factor describing the change of the permanent dipole moment of the chromophore upon excitation that enters the expression for σ₂, and the electric field affects this factor through polarizability.

Notably, the increase in σ₂ of the anionic chromophore is always paralleled by a shift of the one-photon absorption peak to a shorter wavelength. We previously developed a physical model (24, 25, 26) that predicts that both of these changes will happen if the electric field increases, directed from the center of the phenolate ring to the center of the imidazole ring. In red GECIs, the chromophore is oriented such that its phenolate oxygen points toward the site of circular permutation where the Ca²⁺-sensing domain is fused to the FP (as observed for RCaMP, Protein Data Bank (PDB): 3U0K (20); R-GECO1, PDB: 4I2Y (20); and K-GECO1, PDB: 5UKG (21); see Fig. 1 A). Because of this geometry, the conformational change of the protein when it binds to Ca²⁺ would likely modulate the electric field along the phenolate-imidazole axis of the chromophore. We propose that the increase in σ₂ results from an increase in the electric field in this direction, in which positive charges move closer to the phenolate oxygen and/or negative charges move farther away.

With this in mind, we can compare potential Ca²⁺-induced electric field changes for both class I and class II because the excitable form of the chromophore is always anionic. The indicators in class I show the greatest increases in σ₂ and the largest blue shifts (13–15 nm), whereas those in class II show insignificant increases in σ₂ and small blue shifts (1–2 nm). These differences might be because of what is hydrogen bonded to the phenolate oxygen of the chromophore in the Ca²⁺-saturated state. In class I, it is a positive residue (Lys49, numbered according to Fig. S5), based on the crystal structure of R-GECO1 (20). In class II, only neutral residues (Asn32 and Thr278 for K-GECO1 and jRCaMP1a, respectively) and water molecules are bound to that oxygen (20, 21). If those hydrogen bonds are disrupted in the Ca²⁺-free state, then this could explain the largest changes in σ₂ for class I and insignificant changes for class II. There is likely a Ca²⁺-induced increase in the electric field along the phenolate-imidazole axis that is much stronger for class I.

The changes in ρ_e for class I and class II qualitatively follow the same trend as the changes in σ₂, which can be explained on similar electrostatics grounds. For class I, the major Ca²⁺-induced shift of the acid-base equilibrium of the chromophore toward its deprotonated, anionic form is probably a result of the positive charge on Lys49 coming closer to the phenolic oxygen. In class II, this pK_a shift is not as prominent because there are no charges to redistribute.

In support of these observations, the indicators in class III were initially developed by mutating the positive Lys49 in R-GECO1 to the negative Glu49, stabilizing the protonated, neutral form in the Ca²⁺-saturated state (15). Importantly, the proximity of the Glu49 carboxylate in the Ca²⁺-saturated state also makes excited-state proton transfer possible for class III proteins (15, 38). The neutral form of the chromophore undergoes excited-state proton transfer upon excitation by giving its proton to Glu, and then it emits from the anionic state, resulting in the long Stokes shift. This pathway is disrupted in the Ca²⁺-free state because the Glu49 carboxylate is not in close proximity. This is why the excitable form is anionic in the Ca²⁺-free state and neutral in the Ca²⁺-saturated state, even though, technically, it is the anionic form of the chromophore which emits in both cases.

We also observe that the φ of the anionic form increases in the Ca²⁺-saturated state in all cases except O-GECO1 (where it did not change) and class III indicators (where no anionic form is present in the Ca²⁺-saturated state). This effect can also be explained by the Ca²⁺-induced changes in the electric field. Quantum chemistry calculations predict that in a vacuum, the nonradiative decay of the excited state proceeds through substantial twisting around the phenolate exocyclic C-C bond (39). This is coupled with electron density transfer from the phenolate to the imidazolinone and acylimine groups (39). If the field increases in the same direction of electron density transfer, the activation energy for the transfer will increase. As a result, the nonradiative decay rate will slow, and the fluorescence quantum yield will increase. The independence of quantum yield on conformational changes in GCaMP6m (23) and O-GECO1 (Table 1) may be due to no acylimine group in their chromophore structures. In fact, it was shown that the electronegativity of acylimine makes the described mechanism of nonradiative relaxation highly probable specifically for the red FP chromophore (39).

In summary, red GECIs can be divided into three classes, and the factors that mainly contribute to the Ca²⁺-dependent change in fluorescence vary based on the class. It is interesting to note that the three factors, ρ_e, φ_e, and σ_2,e, synergistically increase in almost all the red GECIs because of a favorable geometric arrangement of the chromophore with respect to the fusion site and to the local electrostatic/hydrogen bonding interactions. In class I, the change under one-photon excitation was smaller than the change under two-photon excitation. We suggest that this should be carefully considered when further developing these red GECIs.

We characterized the Ca²⁺-free and Ca²⁺-saturated states for each red GECI. These measurements can only give us information about the protein in these limiting states. They do not tell us about the dynamics of the conformational change between these states or the indicators’ performance in live cells. Some of these aspects were recently reviewed for red GECIs in other work (40, 41). Our results do reveal that there is no single perfect red GECI. For maximal changes in fluorescence, the indicators in classes I and III appear to be the best. For fluorescence lifetime imaging, the proteins with the largest changes in quantum yield are promising (classes II and III). This follows from the fact that the fluorescence lifetime τ is proportional to the quantum yield: τ = τ_Rφ, where τ_R is the radiative lifetime and can be considered virtually constant (24). For two-photon imaging simultaneously with a green FP-based probe, jREX-GECO1 is an excellent candidate because it shows the maximal change in fluorescence at ∼900 nm. This has been demonstrated with its predecessor REX-GECO1 (15). Finally, to inform the optimum optical setup for these red GECIs, it is important to consider the Ca²⁺-dependent spectral shifts in excitation (one-photon and two-photon) and emission.

The red fluorescent Ca²⁺ indicator is a man-made molecular machine found nowhere in nature. Just as understanding how microscopes work can lead to better microscopes and better imaging, so it is important to understand how these biological tools work, both to improve them and make the best use out of them.

Author Contributions

R.S.M. and M.D. designed and conducted research. J.W. and Y.Q. developed jREX-GECO1. Y.S. provided plasmids and critical advice along the way. R.S.M. analyzed data and wrote the manuscript. R.E.C., T.E.H., and M.D. supervised research. R.S.M., Y.Q., Y.S., R.E.C., T.E.H., and M.D. revised the manuscript.

Editor: Amy Palmer.