g 

t

1/2off

values were determined from the dissociation kinetics curves in the presence of 1 mM MgCl

2

.

t

1/2off

for GCaMP6s is 1.01 ± 0.06 s; t

1/2off

for GCaMP6f is 0.37 ± 0.04 s.

FGCaMP5 and FGCaMP7 had 19 amino acid substitutions relative to the original FGCaMP, and FGCaMP6 had 20 amino acid substitutions (Figure S4 and S5). Among these mutations, 13 or 12 were located in the A. niger/A. fumigatus-derived M13-peptide/CaM of FGCaMP5 and FGCaMP6 or FGCaMP7, respectively; four or five mutations were located in the cpEGFP-based fluorescent domain of FGCaMP5 or FGCaMP6 and FGCaMP7, respectively; and two mutations were located in linkers between the Ca²⁺-binding and fluorescent domains (Figure S4 and S5).

Figure S4. Alignment of the amino acid sequences for FGCaMP, FGCaMP5, FGCaMP6, FGCaMP7, and GCaMP6f calcium indicators. Alignment number follows that of FGCaMP. Residues of fluorescent part buried in β-barrel are highlighted with green. Residues that are forming chromophore are marked with asterisks. Calcium-binding EF1–EF4 hands are highlighted with cyan. Linkers between fluorescent and calcium-binding parts are highlighted with yellow. Mutations in FGCaMP variants relative to the original FGCaMP are highlighted with red. Secondary structure of FGCaMP is presented on top.

All mutations located in fluorescent domain were external to the β-barrel of GFP and were not likely to affect chromophore properties (Figure S5). All three chosen indicators had five mutations in the Ca²⁺-binding EF-hands: three mutations in EF2, and two in EF3. N327D, N364D, and S368D substitutions were previously introduced by site-directed mutagenesis; these mutations have been shown to decrease the K_d value of high-affinity components for the FGCaMP indicator ^[1]. FGCaMP5 had one unique N15I mutation in the M13-peptide, FGCaMP6 had two unique T162A and T272A substitutions, and FGCaMP7 had one unique K88Q mutation in the fluorescent domain (Table S3 and Figure S4). The location of the FGCaMP7 mutations along the gene of FGCaMP are presented schematically in Figure 1a (bottom).

Figure S5. Cartoon representation of FGCaMP calcium indicator with mutations from FGCaMP5, FGCaMP6, and FGCaMP7 variants. Orthogonal views of the structure of FGCaMP indicator in Ca²⁺-bound state (PDB 6XU4). β-barrel of cpEGFP is shown as a green cylinder, Ca²⁺ ions are shown as grey spheres, M13-peptide and CaM are shown in purple and light blue, respectively. Positions of FGCaMP5, FGCaMP6, and FGCaMP7 mutations relative to FGCaMP structure are depicted as blue sticks.

The key in vitro characteristics of the improved variants of the FGCaMP indicator were determined and are summarized in Table 1 and Table S4. Spectroscopic data showed that FGCaMP5, FGCaMP6, and FGCaMP7 exhibited absorbance peaks at 400 nm in Ca²⁺-free state and 496–498 nm in Ca²⁺-saturated state (Table S4 and Figure 1b,c). When excited at 400 and 496–498 nm, the FGCaMP variants fluoresced at one emission peak at 516 nm (Table S4and Figure 1b,c). All three FGCaMP variants had slightly (3–5 nm) smaller Stokes shift, compared to the original FGCaMP. The 400 and 496–498 nm absorption maxima can be attributed to the protonated and anionic forms of GFP chromophore, respectively, similar to the original FGCaMP indicator ^[1][3]. In Ca²⁺-saturated states, the brightness of FGCaMP5, FGCaMP6, and FGCaMP7 were 1.37-, 1.13-, and 1.16-fold higher than the brightness of the FGCaMP anionic form, respectively (Table S4). In Ca²⁺-free states, the brightness of FGCaMP5 and FGCaMP7 were 1.66- and 1.23-fold higher than that of FGCaMP, respectively; while the 400-nm absorbing form of FGCaMP6 had only 65% brightness of the original FGCaMP indicator (Table S4).

Upon calcium ion binding, the maximal fluorescence of FGCaMP7 and FGCaMP6 in the presence of 1 mM MgCl₂ (equivalent to 0.58–1 mM free Mg²⁺ ion concentration, resembling that of 0.5–1.5 mM in the cytosol of mammalian cells ^[14][15]) showed 10.1- and 9.0-fold decreases and 32.7- and 33.0-fold increases at 400 and 496–498 nm excitation, respectively (Table 1 and Table S4). These contrasts were 1.4- and 1.3-fold and 2.1- and 2.2-fold higher than the fluorescence contrasts of the respective forms of the original FGCaMP indicator. The fluorescence contrasts for the FGCaMP5 mutant (8.0 ± 0.4- and 17.6 ± 1.1-fold at 400 and 498 nm, respectively) were comparable to those of FGCaMP (7.1 ± 0.2- and 15.3 ± 0.5-fold at 402 and 493 nm, respectively). Thus, the selected FGCaMP variants are excitation-ratiometric calcium indicators with 140–330-fold maximal fluorescence contrast ratio change upon calcium ion binding. The maximal fluorescence contrast ratio for FGCaMP7 (330-fold) was 3.0-fold higher than that for FGCaMP (109-fold) in the presence of 1 mM MgCl₂. Recently, another ratiometric calcium indicator series, Y-GECO2s, has been published with maximal contrasts ratio 1.1–2.3-fold higher than the contrast ratio for the FGCaMP7 indicator, which demonstrated the largest contrast among the studied FGCaMPs ^[16]. However, Y-GECO2s still suffer from low extinction coefficients and its molecular brightness at 412 and 522 nm excitation were 2.8–3.3-fold and 3.1–3.5-fold lower than the brightness of the FGCaMP indicators at 400 and 498 nm excitation, respectively ^[16]. Hence, enhanced FGCaMPs are the GECIs of choice among the currently available ratiometric indicators, in terms of high dynamic range and molecular brightness.

We next assessed the affinity of FGCaMP variants to Ca²⁺ ions. According to equilibrium binding titration curves, FGCaMP demonstrates bi-phasic Ca²⁺ binding in the range of 0–39 μM free Ca²⁺ at 493 nm excitation. Such bi-phasic Ca²⁺ binding of FGCaMP is a limitation, as it decreases the sensitivity of the indicator to low calcium ion concentration transients and violates its linear response at high calcium ion concentrations when expressed in neurons. In the presence of 1 mM MgCl₂, FGCaMP5, FGCaMP6, and FGCaMP7 demonstrated mono-phasic calcium titration curves with K_d values of 98 ± 3 nM, 350 ± 13 nM, and 240 ± 6 nM, respectively (Figure 1d and Figure S6a,b and Table 1 and Table S4). These values were 4.7-, 1.3-, and 1.9-fold lower than the K_d value for FGCaMP under the same conditions and were optimal for the monitoring of calcium concentration changes during neuronal activity, which typically varies from 50–100 nM at rest to 250–10,000 nM during activation in the cytosol of mammalian cells ^[17][18]. The high calcium affinity of the FGCaMP5 indicator is probably a result of the N15I substitution in the M13-peptide (Figure S4), which may strengthen its interactions with the Ca²⁺-binding domains of CaM. Indeed, N15I was a unique mutation among the FGCaMP5, FGCaMP6, and FGCaMP7 indicators; it has been previously shown that mutations in the M13-peptide remarkably affected the indicator’s affinity to calcium ions ^[1]. According to the Hill coefficient values calculated by fitting the equilibrium titration curves to the Hill equation, the FGCaMP5, FGCaMP6, and FGCaMP7 indicators bound calcium ions with slightly less co-operativity, compared to the control FGCaMP and GCaMP6s indicators (Hill coefficients: 1.6–2.3 vs. 2.3–4.0; Table 1 and Table S4) ^[1][19]. Hill coefficients in the range of 3–4 are common for many GECIs, which implies linear dependence of the ΔF/F response to changes in calcium concentration in a narrow calcium ion concentration range but with higher ΔF/F response. When the Hill coefficient is close to 1 or less (which is characteristic of most synthetic dyes), the ΔF/F dependence vs. [Ca²⁺] is linear in a wider range of calcium ion concentrations ^[20][21]; however, this is at the expense of lower sensitivity. Hence, the enhanced versions of FGCaMP demonstrated advantageous mono-phasic fluorescence responses to calcium ions and had affinities to calcium ions appropriate for the monitoring of neuronal activity.

Figure S6. Calcium titration curves and pH dependences for FGCaMP5 and FGCaMP6. (a) Ca²⁺ titration curves for FGCaMP5 and FGCaMP6 at 365 nm excitation. (b) Ca²⁺ titration curves for FGCaMP5 and FGCaMP6 at 490 nm excitation. (c) Fluorescence of FGCaMP5 as a function of pH at 365 nm excitation. (d) Fluorescence of FGCaMP5 as a function of pH at 490 nm excitation. (e) Fluorescence of FGCaMP6 as a function of pH at 365 nm excitation. (f) Fluorescence of FGCaMP6 as a function of pH at 490 nm excitation.

Next, we characterized another important property of the improved FGCaMPs GECIs: their pH stability. According to fluorescence changes, the 400 nm absorbing form of FGCaMP7 exhibited a slight shift in pK_a value (from 6.63 to 6.00) after the addition of calcium ions (Figure 1e and Table 1). Upon Ca²⁺ binding, the pK_a values for the 498 absorbing form shifted from 5.28 and 7.81 to 6.87 (Figure 1f and Table 1). In the range of pH values of 7–8, which is typical in the cytosol of neurons ^[22], the changes of fluorescence intensity for the protonated form in Ca²⁺-free state and anionic form in Ca²⁺-saturated state were 25% and 36%, respectively, while the change of the dynamic ranges for both forms in the pH range 7–8 was minor—which is an advantage of FGCaMP7 (Figure 1e,f) the protonated forms of FGCaMP5 and FGCaMP6 showed larger shifts in pK_as, from 7.01 to 5.75 and from 6.84 to 5.57, respectively (Table S4), upon calcium-binding. The pK_a shifts of 496–497 nm absorbing forms for FGCaMP5 and FGCaMP6 were negligible (Table S4), considering that their anionic forms are more pH-stable than the respective form of FGCaMP7. In contrast to the FGCaMP7 indicator, the dynamic ranges for the protonated forms of FGCaMP5 and FGCaMP6 changed by about 25%–30% in the range of pH values 7–8 (Figure S6c,e). At the same time, changes in the dynamic ranges of anionic forms of FGCaMP5 and FGCaMP6 were negligible under the same conditions (Figure S6d,f). Hence, the fluorescence response of these FGCaMP variants may be affected by variations in pH.

Next, we assessed the photobleaching stability for the FGCaMP7 indicator, as the indicator combining high contrast and brightness for both forms and showing the best performance in neuronal cultures (see results below). Both protonated and anionic forms of FGCaMP7 were extremely photostable in vitro. Under a wide-field microscope equipped with a metal halide lamp and 470/40BP excitation filter, FGCaMP7 in Ca²⁺-saturated state photobleached 2.7- and 1.8-fold slower than a control mEGFP protein and FGCaMP indicator in Ca²⁺-saturated state. In the absence of Ca²⁺ ions, FGCaMP7 photobleached 8.7- and 8.6-fold slower than mTagBFP2 and FGCaMP as a result of illumination with 355 ± 25 nm light, respectively (Figure 1g and Table 1). Hence, both 400- and 498-forms of the FGCaMP7 indicator demonstrated photostability in vitro higher than the photostabilities of FGCaMP, mTagBFP2, and mEGFP proteins.

Both in the absence of Ca²⁺ and the presence of 1 mM Ca²⁺, purified FGCaMP7 sensor eluted in size-exclusion chromatography as a monomer (Figure S7); note that monomeric proteins are usually preferable in terms of reduced cytotoxicity ^[23].

Figure S7. Size-exclusion chromatography for FGCaMP7 protein. (a) Fast protein liquid chromatography of FGCaMP7 in 20 mM Tris-HCl (pH 7.5), 100 mM NaCl buffer supplemented with 1 mM CaCl₂ or (b) in 20 mM Tris-HCl (pH 7.5), 200 mM NaCl buffer supplemented with 2 mM EDTA.

As fast Ca²⁺ association and dissociation kinetics are crucial for the resolution of neuronal dynamics, these kinetics were studied in the FGCaMP variants using stopped-flow fluorimetry. To investigate the association kinetics, we calculated the observed association rate constants (k_obs) at [Ca²⁺] 300–1000 nM in the presence of 1 mM MgCl₂ by fitting the kinetic curves of FGCaMP5, FGCaMP6, and FGCaMP7 to double exponentials at 400 nm excitation, to a single exponential at 498 nm excitation, and the kinetic curves of the original FGCaMP to a single exponential at both excitation wavelengths (Figure S8a,b and Figure 1h). The contribution of exponent 1 at 400 nm excitation varied from 10% to 29% for FGCaMP5, from 34% to 49% for FGCaMP6, and from 0% to 35% for FGCaMP7 (Figure S8c and Figure 1i). According to the k_obs values obtained from the stopped-flow data, in the range of free Ca²⁺ concentrations 300–1000 nM, the FGCaMP5 indicator was the fastest among the FGCaMP mutants, which associated with calcium ions 1.1–1.9-fold faster than FGCaMP6 and FGCaMP7 at both 400 and 498 nm excitation (Figure S8b). At 300 nM calcium concentration, as observed in neurons at one action potential (AP), FGCaMP5 associated with calcium ions 2.1/1.7-fold or 2.7/2.3-fold faster than control GCaMP6s or progenitor FGCaMP at 400/498 nm excitation, respectively. At the same time, at 300 nM [Ca²⁺] at 400 nm excitation, the FGCaMP7 indicator associated with calcium 2.1-fold slower than the fastest GCaMP6f indicator and 1.2- and 1.6-fold faster than GCaMP6s and FGCaMP (Figure 1h and Table S4). At 498 nm excitation, FGCaMP7 associated with calcium ions was 3.0-fold slower than GCaMP6f and had an association rate constant comparable to the respective constants for GCaMP6s and FGCaMP (Figure 1h and Table S4).

Figure S8. Calcium-association kinetics for the FGCaMP indicator series characterized by stopped-flow fluorimetry. (a) Calcium-association kinetics curves for FGCaMP, FGCaMP5, FGCaMP6, and FGCaMP7 GECIs at Ca²⁺ concentrations of 300 nM, 700 nM, and 1000 nM. Calcium-association curves are shown for both forms with maximum excitation at 400 and 498 nm. (b) Observed Ca²⁺-association rate constants determined from association curves for FGCaMP, FGCaMP5, FGCaMP6, and FGCaMP7 GECIs. For the form with excitation maximum at 400 nm fast (exp. 1) and slow (exp. 2) exponents are shown. The data was fitted to the equation k_obs = k_on × [Ca²⁺]ⁿ + k_off. (c) Relative contribution of monoexponents A1/(A1 + A2) and A2/(A1 + A2) for FGCaMP5, FGCaMP6, and FGCaMP7 GECIs at 400 nm excitation, where A1 and A2 are the pre-exponential factors in the association curve equation ΔFlu(t) = A1 × exp(-k^on_obs1 × t) + A2 × exp(-k^on_obs2  × t).

The dissociation half-time (t_1/2^off) of FGCaMP5 was 1.5–2.1-fold longer than the respective half-times for FGCaMP6 and FGCaMP7 (Table S4). The FGCaMP7 indicator at 400 nm excitation released calcium ions 1.5- and 1.3-fold slower than GCaMP6s and FGCaMP and 4.1-fold slower than GCaMP6f, respectively (Figure 1j and Table 1). At 498 nm excitation, the dissociation half-time of FGCaMP7 was similar to the dissociation half-time of FGCaMP and 1.3- or 3.6-fold longer than those obtained with GCaMP6s or GCaMP6f, respectively (Figure 1j and Table 1). Hence, enhanced FGCaMPs have fast association–dissociation kinetics for calcium ions, comparable to the kinetics of the standard GCaMP6s GECI, which demonstrates their suitability for the visualization of fast neuronal activity.

Thus, using directed molecular evolution, we substantially improved the new class of FGCaMP calcium indicators, based on Ca²⁺-binding parts from fungi. As a result, for further characterization in HeLa cells and neuronal cultures, we selected three enhanced FGCaMPs variants which demonstrated a mono-phasic response to calcium ions, had different calcium affinities and dynamic ranges superior to that of FGCaMP, and which preserved the ratiometric phenotype of the original FGCaMP indicator.

To evaluate the behaviors of the enhanced FGCaMP indicators in live mammalian cells, we expressed FGCaMP5, FGCaMP6, and FGCaMP7, as well as control FGCaMP and GCaMP6s, in HeLa Kyoto cells together with a red reference R-GECO1 calcium indicator, and studied their response to ionomycin-induced elevation of calcium concentration in the cell cytosol ^[24]. After the addition of 2.5 μM ionomycin, we observed increases in red and green fluorescence when excited at 561 and 488 nm, respectively; green fluorescence at 405 nm excitation decreased (Figure 2 and Figure S9). The ΔF/F₀ value for the fast calcium-associated FGCaMP5 normalized to ΔF/F₀ of the R-GECO1 indicator was only 87 ± 25% at 488 nm excitation (Table 2). The mean normalized ΔF/F₀ response for FGCaMP7 was 237 ± 81% from the R-GECO1 response, which was 2.1- and 1.5-fold higher than normalized responses for GCaMP6s (p ˂ 0.0001) and FGCaMP (p = 0.0045), respectively. The ΔF/F₀ response for FGCaMP6 was similar (p = 0.9313) to that of the original FGCaMP and 1.4-fold higher (p = 0.0297) than the response of the control GCaMP6s indicator. At 405 nm excitation, the ΔF/F₀ values for all three indicators (FGCaMP5, FGCaMP6, and FGCaMP7) were similar and 1.4–1.6-fold less than that of FGCaMP (p = 0.0243, p = 0.0230, and p = 0.0493, respectively; Table 2). Hence, the FGCaMP7 indicator demonstrated the highest ionomycin-induced ΔF/F₀ response in cultured mammalian cells among the three FGCaMP variants tested and GCaMP6s, while FGCaMP5 had the lowest ΔF/F₀ response; this correlated with its lowest contrast and highest calcium affinity in vitro (Table 2).

We next compared the enhanced FGCaMP variants in their ability to monitor calcium dynamics during the spontaneous activity of live neurons. We co-transduced dissociated neuronal cultures isolated from mice pups with recombinant adeno-associated viral particles (rAAV) carrying FGCaMP5, FGCaMP6, FGCaMP7, FGCaMP, and GCaMP6s GECIs together with control red R-GECO1. We observed calcium oscillations in all cultures by day 11–17 in vitro (Figure 3 and Figure S10). In the case of the ratiometric FGCaMP series, calcium oscillations were detected with opposite fluorescence changes at 405 and 488 nm excitation, respectively (Figure 3 and Figure S10). The mean ΔF/F₀ response (normalized to ΔF/F₀ of R-GECO1) was similar for FGCaMP, FGCaMP6, and FGCaMP7 at 405 nm excitation. However, ΔF/F₀ for FGCaMP5 was 2.6-fold lower (p = 0.0014) than that of FGCaMP, perhaps because the protein was partially bound to Ca²⁺ in resting neurons due to its lower K_d value of 98 nM, as compared to the K_d value of 460 nM for FGCaMP. At 488 nm excitation, the mean normalized ΔF/F₀ responses for FGCaMP7 and FGCaMP6 were comparable to that of FGCaMP (p = 0.8026 and p = 0.8895, respectively) and GCaMP6s (p = 0.4966 and p = 0.3932, respectively; Table 2). The mean normalized ΔF/F₀ for FGCaMP5 at 488 nm excitation was 2.3–2.6-fold lower (p < 0.0001) than ΔF/F₀ for FGCaMP, FGCaMP6, FGCaMP7, and GCaMP6s. Finally, we estimated ratiometric ΔR/R₀ responses for the FGCaMP5, FGCaMP6, and FGCaMP7 indicators normalized to ΔF/F₀ of R-GECO1 and compared them to ΔR/R₀ for the original FGCaMP and ΔF/F₀ for GCaMP6s. The mean normalized ΔR/R₀ for FGCaMP7 was 1.6-fold higher (p = 0.0014) than ΔF/F₀ of GCaMP6s, comparable to ΔR/R₀ of FGCaMP (p = 0.3054) and FGCaMP6 (p = 0.6803), and 2.1-fold higher (p = 0.0002) than ΔR/R₀ of FGCaMP5 (Table 2). Hence, during the spontaneous activity of neuronal cultures, FGCaMP7 revealed ΔF/F₀ or ΔR/R₀ responses similar or superior to those for other FGCaMP variants, FGCaMP, and GCaMP6s.

When comparing mean rise half-times at 488 nm excitation, the rise of FGCaMP7 fluorescence was slightly faster (1.36 ± 0.47 s) than the rise of the FGCaMP fluorescence (2.08 ± 0.57 s) and the other FGCaMP variants (FGCaMP5: 2.01 ± 1.23 s; FGCaMP6: 1.65 ± 0.71 s) and slightly slower than the rise of GCaMP6s fluorescence (1.14 ± 0.61 s). FGCaMP5 was the slowest indicator, with mean decay half-time values of 6.51 ± 2.50 and 5.78 ± 2.56 s at 488 and 405 nm excitation, respectively. Decay half-times for the FGCaMP7 (3.84 ± 0.88 and 3.40 ± 0.72 s at 488 and 405 nm excitation, respectively), FGCaMP (4.01 ± 1.65 and 3.56 ± 1.37 s at 488 and 405 nm excitation, respectively), and GCaMP6s (3.77 ± 1.27 s) GECIs were similar. Mean rise and decay half-times for the control co-expressed R-GECO1 indicator were similar to the respective half-times of FGCaMP7 (t_1/2^rise R-GECO1 = 1.73 ± 0.63 s, t_1/2^decay R-GECO1 = 4.17 ± 1.45 s).

Figure S10. Response of FGCaMP (a), FGCaMP5 (b) and FGCaMP6 (c) GECIs to variations in the cytoplasmic calcium concentration in cultured neurons. (Upper panels) Confocal images of neurons co-expressing respective indicator from FGCaMP series at 488 nm (left panels) and 405 nm (middle panels) excitations and red indicator R-GECO1 at 561 nm excitation (right panels). (Lower panels) The graphs illustrate changes in green fluorescence of the respective indicator from FGCaMP series at 488 nm (green lines) and 405 nm (cyan lines) excitations and red fluorescence of co-expressed R-GECO1 (red lines) GECI as a result of spontaneous activity in neuronal cultures. Changes in fluorescence of each FGCaMP indicator are shown for three neurons. The changes shown in graphs correspond to the areas indicated with white circles in the images in upper panels. Scale bars: 20 μm.

Both FGCaMP7 and GCaMP6s indicators demonstrated even distribution in the cytosol of neurons, however R-GECO1 indicator revealed puncta-like aggregates during prolonged expression in neuronal cultures (Figure S11). This difference may be attributed to different fluorescent domains utilized in FGCaMP7 and GCaMP6s (EGFP fluorescent domain) as compared to R-GECO1 (mApple fluorescent domain). Puncta formation was attributed to lysosomes ^[25], therefore autophagy should adjust the number of these puncta ^[26].

Figure S11. Localization of FGCaMP7 (a) and GCaMP6s (b) GECIs co-expressed with R-GECO1 in the cultured neurons. Confocal images of neurons co-expressing respective green indicator at 488 nm (green) and/or 405 nm (LSS-green) excitations with even distribution and red indicator R-GECO1 at 561 nm excitation (red) with uneven puncta-like distribution. Cultures were imaged on DIV 13 (a) and 16 (b). Scale bar: 150 μm.

Hence, all three enhanced FGCaMP variants were able to visualize spontaneous calcium activity in neuronal culture. According to the rise and decay half-times values—as well as to the ΔF/F

₀

and ΔR/R

₀

response values—during spontaneous neuronal activity, FGCaMP7 was chosen as the best variant among them. Its normalized ratiometric ΔR/R

₀

response was 1.6-fold higher (

p

= 0.0014) than the normalized ΔF/F

₀

response for GCaMP6s.

We next compared the responses of the enhanced FGCaMP variants, control FGCaMP, GCaMP6s, and R-GECO1 GECIs during external electric field stimulation of neuronal cultures. FGCaMP variants were co-expressed in neuronal cultures together with reference red R-GECO1 ^[24]. FGCaMP and GCaMP6s GECIs were used as controls. The number of APs induced by 150–300 external electric field pulses were evaluated according to the ΔF/F₀ changes of co-expressed R-GECO1, assuming linearity of its response and 4% ΔF/F₀ value at 1 AP stimuli ^[27].

All indicators responded to trains of 1–40 APs; except for FGCaMP and FGCaMP5, which started to sense trains from 10 APs and up (Figure 4a,b) and FGCaMP6, which responded to trains starting from 4 APs and up (Figure 4c). The green fluorescence in GCaMP6s-expressing neurons increased in response to electric field stimulation, while green fluorescence in neurons expressing ratiometric indicators FGCaMP, FGCaMP5, FGCaMP6, and FGCaMP7 increased at 488 nm excitation and decreased at 405 nm excitation. To compare the responses of enhanced FGCaMP variants, we estimated ΔF/F₀ or ΔR/R₀ values per 1AP, calculated as a slope of linear dependence of ΔF/F₀ or ΔR/R₀ vs. trains of 1–40 APs in the case of FGCaMP7 and GCaMP6s, 4–40 APs in the case of FGCaMP6, and 10–40 APs in the case of FGCaMP5 and FGCaMP (Figure 4). The ΔF/F₀ response per 1 AP for FGCaMP7 at 405 nm excitation was 1.9–3.5-fold higher (p < 0.0001) than responses of FGCaMP5, FGCaMP6, and control FGCaMP (Table 2). At 488 nm excitation, ΔF/F₀ response per 1AP for FGCaMP7 was 3.7- (p < 0.0001), 4.3- (p < 0.0001), and 2.1-fold (p < 0.0001) higher than the responses of FGCaMP, FGCaMP5, and FGCaMP6, respectively (Table 2). At 488 nm excitation, ΔF/F₀ response per 1 AP for FGCaMP7 was similar (p = 0.8805) to that of GCaMP6s (Table 2). We also compared ratiometric ΔR/R₀ responses per 1AP for enhanced FGCaMP variants with the respective ΔF/F₀ value for GCaMP6s (Figure 4f and Table 2). The ΔR/R₀ response per 1 AP for the FGCaMP7 variant with highest contrast in vitro was 3.8- (p < 0.0001), 4.0- (p < 0.0001), and 2.2-fold (p < 0.0001) higher than respective values for FGCaMP, FGCaMP5, and FGCaMP6, respectively. The ΔR/R₀ response per 1 AP for FGCaMP7 was 1.4-fold higher (p = 0.0002) than the respective ΔF/F₀ response per 1 AP for GCaMP6s (Table 2).

Thus, according to the external electric field stimulation of neuronal cultures, the enhanced FGCaMP variants, similarly to control FGCaMP and GCaMP6s GECIs, demonstrated linear dependence of ΔF/F response in the range of 0–40 APs. The FGCaMP7 indicator outperformed the FGCaMP progenitor and FGCaMP5/FGCaMP6 indicators, in terms of larger ΔF/F₀ and ΔR/R₀ responses. The ΔF/F₀ response per 1 AP for FGCaMP7 at 488 nm excitation was similar to that of the GCaMP6s control, while ratiometric ΔR/R₀ response per 1 AP for FGCaMP7 was 1.4-fold higher (p = 0.0002) than ΔF/F₀ per 1 AP for GCaMP6s. The FGCaMP7 variant was considered as best and chosen for further characterization in vitro, in mammalian cells, and in vivo.

To suggest the mechanism of fluorescence changes of the FGCaMP7 and GCaMP6s calcium indicators caused by calcium ion binding and to assess the possible interactions between these indicators and the intracellular environment, we constructed truncated versions of FGCaMP7 and GCaMP6s with deleted M13-like peptide (called FGCaM7 and GCaM6s, respectively) and studied their responses to calcium ions in vitro as purified proteins and in the cytosol of mammalian cells. These truncated versions, as purified proteins, practically did not respond to the addition of calcium ions in the range of 39–2000 µM free-calcium ion concentrations (Table S5). Hence, the CaM domain binding to calcium ions does not evoke fluorescent changes in the FGCaMP7 and GCaMP6s indicators, per se. It is likely that only subsequent M13-like peptide binding with the calcium-saturated CaM domain is translated into the fluorescence response of these indicators.

Table S5.

In vitro ΔF/F₀ response of truncated versions (with deleted M13-like peptide) of the purified FGCaM7 and GCaM6s indicators to the saturating calcium ion concentrations.

Indicator

ΔF/F0, %

39 µM a

820 µM a

2000 µM a

FGCaM7

-4 ± 10

To validate the in vivo application of the FGCaMP7 indicator using one-photon non-ratiometric imaging, we performed in vivo calcium imaging with an NVista head-mounted microscope (Figure 6a) and compared the amplitudes and kinetics of spikes detected during specific and non-specific activities of neurons in the CA1 field of the hippocampus while mice explored an O-shaped track with landmarks. The FGCaMP7 indicator was delivered into CA1 hippocampal neurons of 1–2 month-old mice using rAAVs carrying NES-FGCaMP under CAG promoter. Several weeks after lens probe implantation, we recorded neuronal activity in the hippocampus of the mice while they explored an O-shaped track with landmarks. Using the MIN1PIPE procedure pipeline ^[28] and manual inspection, we successfully identified active cells and extracted respective calcium activity ΔF/F₀ traces. An example of typical neuronal activity for eight neurons is shown in Figure 6b. Averaged peak ΔF/F₀ for FGCaMP7 was 1.0 ± 1.2, which was 2.1-fold lower (p < 0.0001) than averaged ΔF/F₀ for GCaMP6s (Figure 6c and Table S6). Averaged rise and decay half-times for spikes detected for FGCaMP7 (0.97 ± 0.60 and 3.07 ± 0.80 s, respectively) were 1.3- (p = 0.0002) and 1.2-fold (p < 0.0001) longer than the respective times of the GCaMP6s indicator (0.76 ± 0.49 and 2.54 ± 0.94 s, respectively; Table S6). Therefore, the FGCaMP7 indicator detected hippocampal neuronal activity with slightly less sensitivity and resolution than the control GCaMP6s.

a 39, 820, and 2000 µM free calcium concentration corresponds to the buffer supplemented 

with 10 mM CaEGTA, 10 mM NTA or 2 mM CaCl2, respectively. Data were averaged across 3–6 repeats. SD is shown.

To assess the possible interactions of the FGCaMP7 and GCaMP6s proteins with the intracellular environment, we analyzed the responses of the FGCaM7 and GCaM6s truncated versions to ionomycin-induced calcium ion elevations in HeLa cells. The FGCaM7 and GCaM6s truncated versions transiently expressed in the HeLa cells had an even distribution in the cytosol of the cells (Figure 5a,b, respectively). FGCaM7 practically did not respond to the elevation of calcium ions induced by ionomycin addition, with ΔF/F₀ values of 6.5 ± 2.4% (Figure 5a,c,d). However, GCaM6s sensed the ionomycin-induced elevation of calcium ions, with an ΔF/F₀ response of 139 ± 62% (Figure 5b–d). Although this response was 4.3-fold lower (p < 0.0001), as compared to the ΔF/F₀ response for the full-length GCaMP6s of 600 ± 333% under the same conditions, it was 21-fold larger (p < 0.0001) than that of FGCaM7. Taking into account that purified FGCaM7 and GCaM6s proteins practically did not respond to saturating calcium ion concentrations, we concluded that the CaM part of the GCaM6s protein interacted with intracellular proteins, as opposed to that in the FGCaM7 indicator. Earlier, the fluorescence recovery after photobleaching (FRAP) experiments demonstrated that the progenitor of the FGCaMP7 indicator, FGCaMP, was freely diffusible in the cytoplasm of HeLa cells at physiological Ca²⁺ concentrations, but that about 30% of GCaMP6s and G-GECO1.2 were not freely diffusible, suggesting that they may become bound to CaM or other cellular proteins ^[1]. Therefore, the GCaM6s protein is prone to interaction with intracellular proteins, in contrast to the FGCaM7 protein. We speculated that this difference may be attributed to the fungal origin of the CaM part in the FGCaM7 indicator, which has a different amino acid sequence than metazoan CaM. Overall, the M13-like peptide complex formation with calcium-bound CaM, rather than calcium ion binding by CaM domain, was translated into the fluorescence response of the GCaMP6s and FGCaMP7 indicators; in mammalian cells, the CaM domain of the FGCaMP7 indicator practically did not interact with intracellular proteins, in contrast to the CaM domain of the GCaMP6s indicator.