Genetically Encoded Green Fluorescent Ca2+ Indicators with Improved Detectability for Neuronal Ca2+ Signals

Imaging the activities of individual neurons with genetically encoded Ca2+ indicators (GECIs) is a promising method for understanding neuronal network functions. Here, we report GECIs with improved neuronal Ca2+ signal detectability, termed G-CaMP6 and G-CaMP8. Compared to a series of existing G-CaMPs, G-CaMP6 showed fairly high sensitivity and rapid kinetics, both of which are suitable properties for detecting subtle and fast neuronal activities. G-CaMP8 showed a greater signal (F max/F min = 38) than G-CaMP6 and demonstrated kinetics similar to those of G-CaMP6. Both GECIs could detect individual spikes from pyramidal neurons of cultured hippocampal slices or acute cortical slices with 100% detection rates, demonstrating their superior performance to existing GECIs. Because G-CaMP6 showed a higher sensitivity and brighter baseline fluorescence than G-CaMP8 in a cellular environment, we applied G-CaMP6 for Ca2+ imaging of dendritic spines, the putative postsynaptic sites. By expressing a G-CaMP6-actin fusion protein for the spines in hippocampal CA3 pyramidal neurons and electrically stimulating the granule cells of the dentate gyrus, which innervate CA3 pyramidal neurons, we found that sub-threshold stimulation triggered small Ca2+ responses in a limited number of spines with a low response rate in active spines, whereas supra-threshold stimulation triggered large fluorescence responses in virtually all of the spines with a 100% activity rate.


Introduction
Understanding brain function requires techniques for monitoring the spatio-temporal activity patterns of individual neurons and synapses. A promising approach for this purpose is Ca 2+ imaging that can detect neuronal events as a change in Ca 2+ fluorescence intensity. Recently, Ca 2+ imaging using green fluorescent protein (GFP)-based genetically encoded Ca 2+ indicators (GECIs) has been introduced as an alternative to using chemically synthesized fluorescent Ca 2+ indicators [1][2][3][4][5][6]. GECIs offer two remarkable advantages over synthesized indicators: (i) GECIs can be targeted to specific cell types and specific subcellular compartments [7][8][9][10], and (ii) GECIs are applicable to long-term expression (over months) [4,[11][12][13]. Although GECIs have improved, there remains a need for GECIs with greater signals and more rapid kinetics to allow the reliable detection of individual neuronal spikes.
In this study, we developed high-sensitivity and fast-responsivity GECIs, termed G-CaMP6 and G-CaMP8, by mutagenizing existing G-CaMPs. These novel indicators allow us to reliably monitor neural spikes with larger fluorescence signals and higher temporal resolution than G-CaMP3, a recently reported variant of G-CaMP2 [4]. We also demonstrate that G-CaMP6-actin, a fusion protein of G-CaMP6 and actin, can be used to image spinespecific Ca 2+ signals in response to presynaptic single spikes at the single-synapse level.

Development of Improved G-CaMPs by Site-directed and Random Mutagenesis
In an effort to create a superior GECI, we first introduced mutations from ''superfast GFP'' [14], which was recently reported to enhance the folding activity of GFP, into a prototype GECI, G-CaMP2 [15], because some known folding mutations improve the functionality of GECIs [16,17]. Through screening, we found that a G-CaMP2 variant with two mutations (N105Y and E124V) introduced into the circularly permutated enhanced GFP (EGFP) domain, termed sfG-CaMP2 (Fig. 1A), showed a greater dynamic range (F max /F min = 9.0360.06, n = 3) than G-CaMP2 [15] (F max /F min = 4.8) (Fig. 1B). For further improvement, mutations known to stabilize the chromophore [i.e., T203V in the circularly permutated EGFP domain and D78Y in the calmodulin (CaM) domain] were introduced into sfG-CaMP2 [18], and this variant was termed sfG-CaMP2.02 (Fig. 1A). sfG-CaMP2.02 showed a greater signal increase (F max /F min = 14.860.28, n = 3) than sfG-CaMP2 (Fig. 1B). Subsequently, mutations from G-CaMP3 [4] were introduced into sfG-CaMP2.02 to examine whether certain mutations could further improve the functionality of GECIs. Among the variants of sfG-CaMP2.02, we identified a superior variant termed G-CaMP5.09 (Fig. 1A), which showed not only a high signal (F max /F min = 18.464.94, n = 3) but also a high affinity for Ca 2+ (K d = 20067.0 nM, n = 3) ( Fig. 1B and C). G-CaMP5.09 differs from G-CaMP2.02 by M153K in the circularly permutated EGFP domain and N60D in the CaM domain (Fig. 1A). Next, we attempted to enhance the Ca 2+ sensitivity of G-CaMP5.09 to improve the detection of weak Ca 2+ signals. For this purpose, we introduced mutations known to modify the affinity of CaM for myosin light chain kinase (MLCK) [19] because G-CaMP Ca 2+ sensitivity is based on the intramolecular interaction between the CaM domain and the M13 domain of MLCK. As expected, G-CaMP6, a variant of G-CaMP5.09 bearing an M36L substitution in the CaM domain (Fig. 1A), showed a higher Ca 2+ affinity (K d = 15864.0 nM, n = 3) than G-CaMP5.09 or the previously reported G-CaMP2 variants G-CaMP-HS [17] and G-CaMP3 [4] (Fig. 1B and C). The dynamic range of G-CaMP6 (F max /F min = 11.460.11, n = 3) was not significantly different from that of G-CaMP3 (F max / F min = 11.161.59, n = 3) (Fig. 1B). The spectra of G-CaMP6 were similar to those of G-CaMP2, with the exception that the Ca 2+ absorbance peak (498 nm) was red-shifted by ,10 nm (Fig. 1D). Next we performed random mutagenesis on G-CaMP6 by using an error-prone PCR [15] and were able to screen a highly responsive variant termed G-CaMP7, which differs from G-CaMP6 by a deletion of histidine (DH) in the RSET domain and an S205N mutation in the circularly permutated EGFP domain (Fig. 1A). The dynamic range of G-CaMP7 (F max / F min = 36.664.10, n = 3) was ,3-fold greater than that of G-CaMP6, even though this variant showed a lower Ca 2+ affinity (K d = 243614 nM, n = 3) than G-CaMP6 ( Fig. 1B and C). By performing further random mutagenesis on G-CaMP7, we obtained a more sensitive variant of G-CaMP7 termed G-CaMP8, bearing a deletion of arginine (DR2) in the RSET domain and an I47F mutation in the circularly permutated EGFP domain (Fig. 1A). It was intriguing that the DR2 mutation reported in G-CaMP3 [4] was incidentally incorporated into G-CaMP8. The Ca 2+ affinity of G-CaMP8 (K d = 20061.1 nM, n = 3) was between those of G-CaMP6 and G-CaMP7 and similar to that of G-CaMP3 (K d = 20565.0 nM, n = 3) ( Fig. 1B and C). The spectra of G-CaMP8 were similar to those of G-CaMP6 (Fig. 1D). To assess the functionality of the GECIs in a cellular environment, we next expressed the G-CaMPs in HeLa cells. The baseline fluorescence (Fig. 1E) and the response to ATP stimulation (DF/F) (Fig. 1F) of each G-CaMP variant are summarized in Figure 1G. As expected from their dynamic ranges and Ca 2+ affinities ( Fig. 1B and C), the signal amplitude (DF/F) of G-CaMP6, G-CaMP7 and G-CaMP8 was greater than that of G-CaMP3. The baseline fluorescence of G-CaMP7 and G-CaMP8, unlike G-CaMP6 and other variants, was lower than that of G-CaMP3.

Comparison of G-CaMPs in Pyramidal Neurons
We next evaluated the performance of G-CaMPs in pyramidal neurons in cultured rat hippocampal slices. In the cultured slices, the expression of G-CaMPs and mCherry was driven by the CMV promoter following transfection of the cells with the construct via targeted single-cell electroporation [20]. Simultaneous patchclamp recording and confocal Ca 2+ imaging were performed on G-CaMP-expressing neurons 24-48 h after electroporation. The baseline fluorescence of the neurons expressing G-CaMP6 ( Fig. 2A) was similar to that of neurons expressing G-CaMP3, whereas G-CaMP8-expressing neurons exhibited lower fluorescence intensity than those expressing the other G-CaMPs (Fig. 2B). To monitor spike-induced Ca 2+ responses, the neurons were current-injected to evoke 1-6 spikes at a frequency of 50 Hz. All experiments were carried out at room temperature (25-28uC), unless otherwise specified. G-CaMP6 and G-CaMP8 responded to single spikes with 100% probability. The DF/F amplitudes of Ca 2+ transients evoked by single spikes were 17.463.5%, 27.964.5% and 37.865.2%, and the signal-to-noise ratios (SNRs) were 8.061.5, 18.361.5 and 16.463.5 for G-CaMP3, G-CaMP6 and G-CaMP8, respectively ( Fig. 2C and D; n = 7 each). The signal amplitudes grew almost linearly as the spike number increased ( Fig. 2C and D). Over the entire stimulus range, the amplitudes of the Ca 2+ transients and the SNRs of G-CaMP6 and G-CaMP8 were consistently higher than those of G-CaMP3. The rise time of the spike-induced Ca 2+ transients did not differ among G-CaMP3, G-CaMP6 and G-CaMP8 (P.0.05, Tukey's test). On the other hand, the signal decay of G-CaMP6 and G-CaMP8 was significantly faster than that of G-CaMP3 (G-CaMP3, decay t 1/ 2 = 638638 ms; G-CaMP6, decay t 1/2 = 457620 ms; G-CaMP8, decay t 1/2 = 428611 ms; Tukey's test; n = 7 each) (Fig. 2E). The rapid kinetics and the fairly high Ca 2+ sensitivity ( Fig. 1B and C) of G-CaMP6 contributed to an increased temporal resolution of the signals within spike trains up to 15-20 Hz (Fig. 2F).
The detectability of G-CaMPs was also evaluated in pyramidal neurons in acute cortical slices. The expression of G-CaMPs in the mouse brain was driven by in utero electroporation, as previously described [21]. Consistent with the results presented in Figure 2C and D, G-CaMP6 performed better than G-CaMP3 in acute cortical slices prepared from mice at postnatal day 10-16 ( Fig. 3). This result also implies that G-CaMP6 can be stably expressed in neurons for at least 4 weeks.
By contrast, it was confirmed that the expression of G-CaMP6 does not affect the electrophysiological properties [i.e., input resistance, membrane capacitance, resting potential, excitatory postsynaptic current (EPSC) amplitude, and EPSC frequency] of hippocampal neurons (Fig. 5).

Imaging of Spine Ca 2+ Activity with G-CaMP6-actin
Next, we targeted G-CaMP6 to dendritic spines, the putative synaptic sites, to reveal the dynamics of individual spine activities. For this purpose, G-CaMP6 was fused with actin, a major cytoskeletal protein within spines, to yield G-CaMP6-actin (Fig. 7A). G-CaMP6-actin was effectively localized to the spines in rat hippocampal CA3 pyramidal neurons ( Fig. 7B and C), as has been reported for EGFP-actin and G-CaMP2-actin [8]. We then electrically stimulated the granule cells of the dentate gyrus, which We also tested G-CaMP6 in innervate synapses in the striatum lucidum of CA3 region, with signals of two different strengths (Fig. 7B). Intriguingly, the subthreshold stimulations (DV m = 18.564.8 mV) triggered small fluorescence responses (DF/F = 337686%, n = 256 responses of 63 spines from 5 slices) in a limited number of spines (48.666.3%) in the striatum lucidum, with a low response rate in the active spines (57.6613.8%) ( Fig. 7D and E). In contrast, the suprathreshold stimulations triggered large fluorescence responses (DF/ F = 4436182%, n = 222 responses of 131 spines from 5 slices) in virtually all of the spines in the imaged region including the striatum lucidum and the striatum radiatum, with a 100% activity rate ( Fig. 7D and E).
One of the significant advantages of GECIs over chemically synthesized fluorescent indicators is that once indicator genes have been introduced into neurons, the stable expression of the indicator proteins allows long-term recording of the neurons [4,[11][12][13]. To test whether G-CaMP6-actin is applicable to longterm monitoring, Ca 2+ activity was imaged in spines in slices cultured for 8 and 29 days. After 29 days in vitro (Div), the amplitudes of spine Ca 2+ transients in response to supra-threshold stimulation were not significantly different from those at 8 Div (253630.5% and 201646.6% at 8 Div and 29 Div, respectively; n = 25 spines, P.0.05, Student's t-test). These results confirmed that the expression of G-CaMP6-actin in spines remained stable after at least 4 weeks of culture (Fig. 8).

Discussion
In this study, we developed high-sensitivity and fast-responsivity GECIs, termed G-CaMP6 and G-CaMP8, by introducing sitedirected and random mutations into a prototype GECI, G-CaMP2. Both indicators showed superior performance for reliable detection of neuronal activity with larger fluorescence signals and higher temporal resolution than G-CaMP3. In addition, G-CaMP6-actin captured spine Ca 2+ dynamics in response to the stimulation of presynaptic afferent fibers.
In the course of developing these superior G-CaMPs, we found three novel mutations for improving the GECI functionality [i.e., DH mutation in the RSET domain (in G-CaMP7 and G-CaMP8) and S205N (in G-CaMP7 and G-CaMP8) and I47F (in G-CaMP8) mutations in the circularly permutated EGFP domain]. Based on the G-CaMP2 structure, the residue Ser-205 is located in the bstrand of the circularly permuted EGFP domain (corresponding to the tenth b-strand in EGFP) and facing the inside of the chromophore [18]. This residue has been shown to interact with the chromophore in Ca 2+ -saturated G-CaMP2 [18]. By contrast, the residue Ile-47 is located in the b-strand of the circularly permutated EGFP domain (corresponding to the third b-strand in EGFP) and facing the outside of the chromophore [18]. In addition, this residue is apart from the M13 domain and the CaM domain. Topology of the DH position in the RSET domain is unknown, because the available structural analyses of G-CaMPs based on crystallography have been performed using G-CaMP2 without the RSET domain [23] or with another tag [18]. The DR2 mutation has been known to enhance the G-CaMP fluorescence in cells by stabilizing the protein [4], but G-CaMP8  bearing this mutation did not show brighter fluorescence than G-CaMP7 in a cellular environment (Fig. 1E and G).
Recently, Akerboom et al. have reported new series of GECIs termed G-CaMP5s [6]. Among these indicators, they have demonstrated that G-CaMP5A, 5G and 5K outperform G-CaMP3 in a wide variety of neuronal preparations. G-CaMP5G, which shows ,3-fold greater dynamic range (F max / F min = 32.761.5) than G-CaMP3 (F max /F min = 12.360.4), is the variant which responds with the greatest signals among G-CaMP5s to maximal stimulation when expressed in cultured neurons. Indeed G-CaMP5G is reported to show ,70% greater signals (DF/F) than G-CaMP3 in response to 1-5 spike trains, but its SNR is not improved with respect to that of G-CaMP3 [6]. Besides, the decay kinetics of G-CaMP5G seems to be almost the same as that of G-CaMP3, judging from the shape of trial-averaged responses of G-CaMP5G and G-CaMP3 (Fig. 2B of [6]). In contrast, G-CaMP8, of which dynamic range (F max / F min = 37.563.6) is similar to that of G-CaMP5G, shows ,100% greater signals than G-CaMP3 in terms of both DF/F and SNR (Fig. 2C and D) and ,2-fold more rapid decay kinetics than G-CaMP3 (Fig. 2E). On the other hand, a drawback of G-CaMP8 is its dim baseline fluorescence in neurons, which needs to be improved in the future. G-CaMP5K is the most sensitive G-CaMP5 variant (K d = 18965.0 nM) [6] and is likely to be useful for detecting small neuronal Ca 2+ signals, similar to G-CaMP6 (K d = 15864.0 nM)( Fig. 1B and C). G-CaMP5K is reported to show ,2-fold greater signals (DF/F and SNR) than G-CaMP3 in response to 1-5 spike trains [6]. G-CaMP5A is the variant with intermediate sensitivity (K d = 307612 nM) and signal amplitudes (F max /F min = 17.461.2) among G-CaMP5s, but is reported as the  preferred variant over G-CaMP5G and G-CaMP5K for use in worm and zebrafish [6]. It is good for researchers to have the option to select the ideal GECI depending on their own applications. Because new G-CaMPs (G-CaMP6 and G-CaMP8) and G-CaMP5s have been optimized by different strategies, it may be possible to combine the mutations in the different sets of G-CaMPs to further improve them.
The detection of neuronal activity patterns with single-spike resolution is required to elucidate neural network dynamics. We demonstrated that G-CaMP6 and G-CaMP8 faithfully detected Ca 2+ transients in response to single spikes in pyramidal neurons in hippocampal slices at 25-28uC. However, it is still unknown whether these G-CaMPs exhibit similar performance in vivo. As shown in Fig. 4, both the dynamics of intracellular Ca 2+ and the sensitivity of Ca 2+ indicators are temperature dependent. Indeed, it has been reported that GECI fluorescence is less intense in vivo compared to in vitro [4,24]. Another point to note is that the detectability of indicators might be affected by the expression levels of indicator proteins. Therefore, further studies are needed to determine whether similar results can be obtained in the other gene expression systems, such as transgenic mouse lines or viruses. The decay time constant of spike-induced Ca 2+ transients of the newly-developed G-CaMPs ranged between 400 and 450 ms, which is shorter than that of G-CaMP3 [4]. We demonstrated that the rapid kinetics of Ca 2+ indicators contribute to discrete fast individual spikes in burst-spike trains with a temporal resolution of up to 15 Hz. To our knowledge, G-CaMP6 is the most suitable GECI currently available for detecting and isolating fast individual spikes in spike trains.
Excitatory synaptic activity induces a transient Ca 2+ increase in individual spines through the activation of voltage-sensitive Ca 2+ channels and/or NMDA receptors. In previous studies, spine Ca 2+ activity was imaged with synthetic indicators, such as Oregon Green BAPTA-1 [25,26]. In fly neuromuscular junctions, postsynaptically targeted G-CaMP2 (SynapG-CaMP2) has been reported to respond to excitatory postsynaptic currents [9]. In mammalian cells, Mao et al. [8] developed G-CaMP2actin to record Ca 2+ signals within spines but failed to detect synaptically evoked Ca 2+ activity, presumably because of the low Ca 2+ sensitivity of G-CaMP2. In this study, we demonstrated that G-CaMP6-actin is the first GECI that allows the visualization of Ca 2+ signals in response to synaptic stimulation at the single-spine level. Although the exact mechanisms of spine Ca 2+ signals remain unknown, it seems likely that subthreshold stimulation triggers Ca 2+ transients through postsynaptic NMDA receptors [27], while supra-threshold stimulation triggers Ca 2+ transients at 100% of the spines by opening of voltage-gated Ca 2+ channels through backpropagation of action potentials. In principle, we should be able to visualize spine responses to evaluate long-term plasticity, which is thought to be an elementary component of learning and memory. We expect that these novel G-CaMP technologies, together with advanced imaging systems [26,28], will facilitate our under-standing of neuronal network dynamics in the brain at the single-synapse level.

Plasmid Construction
Complementary DNAs (cDNAs) encoding sfG-CaMP2, sfG-CaMP2.02, G-CaMP5.09 and G-CaMP6 were synthesized by mutagenizing the cDNA encoding the prototype GECI, G-CaMP2 [15], using the QuikChange Lightning Multi Site-Directed Mutagenesis Kit (Agilent). cDNAs encoding G-CaMP7 and G-CaMP8 were synthesized by randomly mutagenizing the cDNAs encoding G-CaMP6 and G-CaMP7, respectively, as previously described [15]. The cDNA encoding G-CaMP3 was constructed by introducing mutations [4] into the G-CaMP2 cDNA. These cDNAs were subcloned into a pRSET B vector (Invitrogen) containing a T7 promoter, as described [15] for bacterial expression, or into a pEGFP-N1 vector (Clontech) with a CMV promoter, as described [3] for expression in HeLa cells and cultured rat hippocampal neurons. For in utero electroporation, cDNAs encoding G-CaMPs and mCherry (Clontech) were subcloned into a pCAGGS vector containing a CAG promoter (CMV enhancer, b-actin promoter and woodchuck hepatitis virus regulatory element [WPRE]) [4]. To target G-CaMP6 to dendritic spines in neurons, a G-CaMP6-actin indicator was generated by fusing a cDNA encoding human b-actin (derived from pAcGFP1acin, Clontech) to the 39 end of a cDNA encoding G-CaMP6 via encoding the amino-acid sequence GGGTGGSRSRARGTVDCRIRSLSSRSRA (in one-letter code). To generate plasmids to express G-CaMPs in the DA motoneurons in C. elegans, cDNAs encoding G-CaMPs were subcloned into a pFX_EGFPT vector containing the unc-4 promoter [29]. All of the constructs were verified by sequencing.

Bacterial Protein Expression and in vitro Characterization
E. coli KRX (Promega) transformed with pRSET B -G-CaMP was grown at 37uC, and protein expression was induced by adding 0.1% rhamnose and incubating for an additional 5 h at 20uC. The indicator proteins with N-terminal histidine tags were purified, dialyzed against KM buffer containing (in mM) 100 KCl and 20 MOPS (pH 7.5) and used for in vitro characterization [15]. Spectral analyses were performed as previously described [16,17]. The term ''dynamic range'' was defined as F max /F min , where F max is the fluorescence intensity at saturating [Ca 2+ ], and F min is the fluorescence intensity at nominally zero [Ca 2+ ] with 1 mM EGTA. The Ca 2+ titration experiments were performed at pH 7.2 with 10 mM solutions of K 2 H 2 EGTA and Ca 2 EGTA from the Ca 2+ Calibration Kit #1 (Invitrogen), as previously reported [30].

Ca 2+ Imaging in HeLa Cells
HeLa cells were cultured in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum and transfected with plasmids using Lipofectamine 2000 (Invitrogen) according to the manufacturer's manual. Fluorescence images of cells expressing G-CaMPs were acquired with a fluorescence microscope (IX71, Olympus) equipped with a CCD camera (ORCA-ER, Hamamatsu), as previously described [16,17]. The cells were perfused with HEPES-buffered saline (HBS) containing (in mM) 135 NaCl, 5.4 KCl, 2 CaCl 2 , 1 MgCl 2 , 10 glucose and 5 HEPES (pH 7.4), and after reading the baseline fluorescence, 100 mM ATP was bath-applied for 1 min. The images were analyzed using AquaCosmos version 2.0 software (Hamamatsu). The transient increase in fluorescence (DF/F) was calculated after subtracting the background fluorescence.

Ca 2+ Imaging in C. elegans
The expression plasmid carrying G-CaMP6 (Punc-4::G-CaMP6) or G-CaMP3 (Punc-4::G-CaMP3) was co-injected with the plasmid carrying DsRed-Express-1 (Punc-4::DsRed-Express-1) [29] into wildtype N2 worms using a standard protocol [31]. The jqEx97 (G-CaMP6) strain and the jqEx216 (G-CaMP3) strain were used in this study. Ca 2+ imaging was performed in worms on a 1.5% agar pad placed on a glass slide (76626 mm, 1.0-to 1.2-mm thickness, Matsunami). L1 animals were placed in M9 buffer [32] and dropped onto the agar pad, and the glass slide was covered by a cover glass (24624 mm, 0.12-to 0.17-mm thickness, Matsunami). The worms were then subjected to imaging analyses using an A1R laser confocal microscope (Nikon) and NIS-Elements AR 3.2 image acquisition software (Nikon). The images were captured with manual movement of the X and Y positions of the stage to Figure 8. Long-term imaging of Ca 2+ activity in spines in a cultured hippocampal pyramidal neuron. A, Z-projection of a representative CA3 pyramidal neuron expressing G-CaMP6-actin at 8 (upper) and 29 (lower) days in vitro (Div). After 7 days in vitro, the G-CaMP6-actin plasmid was introduced into the neuron via single-cell electroporation. Two spines of interest (S1, S2) are indicated by yellow circles. B, Changes in fluorescence at S1 and S2 upon supra-threshold electrical stimulation (Stim). The average spine DF/F ratios in response to supra-threshold stimulation were 253630.5% and 201646.6% at 8 Div and 29 Div, respectively (n = 25 spines, P.0.05, Student's t-test). doi:10.1371/journal.pone.0051286.g008