A genetically encoded Ca2+ indicator based on circularly permutated sea anemone red fluorescent protein

Genetically-encoded calcium ion (Ca2+) indicators (GECIs) are indispensable tools for measuring Ca2+ dynamics and neuronal activities in vitro and in vivo. Red fluorescent protein (RFP)-based GECIs enable multicolor visualization with blue or cyan-excitable fluorophores and combined use with blue or cyan-excitable optogenetic actuators. Here we report the development, structure, and validation of a new red fluorescent Ca2+ indicator, K-GECO1, based on a circularly permutated RFP derived from the sea anemone Entacmaea quadricolor. We characterized the performance of K-GECO1 in cultured HeLa cells, dissociated neurons, stem cell derived cardiomyocytes, organotypic brain slices, zebrafish spinal cord in vivo, and mouse brain in vivo.


Introduction
Protein engineering efforts have yielded three major lineages of monomeric red fluorescent proteins (RFPs) derived from their naturally oligomeric precursors (Fig.   1a). One lineage comes from the Discosoma sp. mushroom coral RFP, DsRed, and includes the first monomeric RFP, mRFP1 (Ref. 1), and the mRFP1-derived "mFruit" variants such as mCherry, mCherry2, mOrange, and mApple [2][3][4] . The second and third lineages stem from the sea anemone Entacmaea quadricolor RFPs eqFP578 (Ref. 5) and eqFP611 (Ref. 6), respectively. EqFP578 is the progenitor of the bright monomeric proteins TagRFP, TagRFP-T, mKate, mKate2, and the low cytotoxicity variant FusionRed 5,7-9 . Engineering of eqFP611 produced mRuby, mRuby2, and mRuby3, a line of RFPs with relatively large stokes-shift and bright red fluorescence [10][11][12] . Together, these three lineages of monomeric RFPs are commonly used in a variety of fluorescence imaging applications and have served as templates for developing red fluorescent indicators of various biochemical activities 13 .
Among the many FP-based indicators of biochemical activity, genetically-encoded calcium ion (Ca 2+ ) indicators (GECIs) are particularly versatile tools. Most notably, they enable imaging of neuronal activity in contexts ranging from dissociated neurons in vitro to brain activity in behaving animals 14 . Green fluorescent GCaMPs, in particular, have proven extremely useful for imaging Ca 2+ activities in various neural systems [15][16][17] . The development of the first single RFP-based Ca 2+ indicators, the DsRed-derived R-GECO1 (Ref. 18) and eqFP611-derived RCaMP1h 19  K-GECO1 shows a strong 2-photon excitation peak at approximately 1100 nm ( Fig.   2c) in the Ca 2+ -bound state. A ~25-fold maximal increase of fluorescence signal, using 2-photon excitation in the excitation region from 1050 to 1150 nm, occurs upon binding Ca 2+ (Fig. 2c). The peak 2-photon molecular brightness of K-GECO1 was compared with R-GECO1, using mCherry as a standard with 1060 nm excitation.
The peak 2-photon molecular brightness, defined as the maximum detected fluorescence count rate per emitting molecule 35 , was obtained from the average fluorescence count rate and the average number of emitting molecules in the beam as determined by fluorescence correlation spectroscopy (FCS). Using this approach, K-GECO1 was found to be approximately 1.5-fold brighter than mCherry and over 2fold brighter than R-GECO1 (Fig. 2d), which is consistent with the comparison of one-photon brightness for the Ca 2+ -bound state (Supplementary Table 1). 9 Crystal structure of K-GECO1. To gain insight into the molecular mechanism of K-GECO1 Ca 2+ sensitivity and to assist future protein engineering efforts, we determined the x-ray crystal structure of K-GECO1 in the Ca 2+ -bound form. The structure was determined to 2.36 Å resolution by molecular replacement (Fig. 2e, Supplementary Table 2). The crystal structure reveals the distinctive features of the ckkap/CaM complex in K-GECO1 (and presumably in other ckkap-based GECIs) relative to other RS20/CaM-based GECIs including R-GECO1 (Fig. 2f), RCaMP ( Fig. 2g), and GCaMP6 (Fig. 2h). The major difference is that the binding orientation of the ckkap peptide to the CaM domain is opposite to that of RS20 to CaM 36,37 .
Another difference is that the RS20 peptide consists entirely of an α-helix in the CaM-binding region, whereas the CaM-binding region of ckkap consists of both an αhelical segment as well as a hairpin-like loop structure at its C-terminus 34 .
Examination of the molecular interactions between the protein and the chromophore at the circular permutation site provides insights into the mechanism of Ca 2+dependent fluorescence modulation. The side chain of Asn32 of linker1 is in direct hydrogen bonding with the phenolate oxygen of the chromophore (Fig. 2i), and is positioned similarly to Ser143 of FusionRed, which engages in a similar interaction with the chromophore 9 . We reason that Asn32 plays a critical role in communicating the Ca 2+ -dependent conformational change in the ckkap/CaM domain to the chromophore in the cpRFP domain. Lys79 of R-GECO1 (Fig. 2j), Thr243 of RCaMP1h (Fig. 2k), and Arg376 of GCaMP6 (Fig. 2l), are likely to have similar roles in their respective mechanisms of fluorescence modulation. Saturation mutagenesis of Asn32 of K-GECO1 resulted in a library of variants that all had dimmer 10 fluorescence and/or smaller Ca 2+ -induced fluorescence intensity fold change. This results indicates that Asn is the optimal residue in this position.
Performance of K-GECO1 in cultured cells. To demonstrate the utility of K-GECO1 for imaging of Ca 2+ dynamics, we expressed it in cultured human cells, dissociated rat neurons, organotypic rat brain slices, zebrafish sensory neurons, and mouse primary visual cortex. We first recorded the response of K-GECO1 to changes in the cytoplasmic Ca 2+ concentration in HeLa cells using established protocols ( Fig. 3a) 38 . HeLa cells expressing K-GECO1 had maximum fluorescence intensity changes of 5.2 ± 1.1-fold (n = 44) on treatment with histamine, which is similar with to the 4.9 ± 1.9-fold (n = 22) response previously reported for R-GECO1 expressing HeLa cells 18 .
Next, we tested K-GECO1 in dissociated rat hippocampal neurons. The relatively low Ca 2+ Kd of 165 nM for K-GECO1 is comparable to that of current best green GECI, GCaMP6s 17 , which has been highly optimized for detection of neuronal Ca 2+ transients. Cultured dissociated neurons expressing K-GECO1 had fluorescence distributed throughout the cytosol and nucleus, and exhibited close to 2-fold maximum increases for spontaneous Ca 2+ changes (Fig. 3b). We did not observe intracellular fluorescent punctate structures, as have been observed for R-GECO1 and its variants 22,27 , in the cell bodies of dissociated neurons expressing K-GECO1 ( Supplementary Fig. 3a,b). We also did not observe noticeable photoactivation of K-GECO1 in neurons when illuminated with 0.5 W/cm 2 405 nm laser light. Under the same illumination conditions, R-GECO1 exhibited substantial photoactivation ( Supplementary Fig. 3c,d). The absence of photoactivation for K-GECO1 under 11 these conditions might be due to the relative low laser intensity (0.5 W/cm 2 ) compared with the intensity (1.76 W/cm 2 ) used for in vitro characterization.
To compare the performance of K-GECO1 with other red GECIs in dissociated neurons, we performed an automated imaging assay with field stimulation as previously described 17,24 . For a single action potential, K-GECO1 exhibited a similar response to jRGECO1a (Fig. 3c) and GCaMP6s 17 , two of the most sensitive indicators currently available. The peak ΔF/F0 amplitude of K-GECO1 with 3 or more action potentials wassmaller than that of jRGECO1a, yet better than other red GECIs ( Fig. 3d, e). In terms of signal-to-noise ratio (SNR), K-GECO1 had similar performance to jRGECO1a, but less than that of jRCaMPa/b (Fig. 3f). K-GECO1 exhibits fast kinetics, with has a half decay time that is faster than jRGECO1a and jRCaMP1a/b (Fig. 3g), and a half rise time that similar to jRGECO1a but faster than jRCaMP1a/b (Fig. 3h).
As our in vitro characterization indicated that K-GECO1 has less blue-light photoactivation than R-GECO1, we tested its performance in human induced pluripotent stem cell-derived cardiomyocytes (iPSC-CMs) in combination with Channelrhodopsin-2 (ChR2). As expected, transfected iPSC-CMs expressing K-GECO1 exhibited spontaneous Ca 2+ oscillations (Fig. 4a). To compare photoactivation of K-GECO1 and R-GECO1 in iPSC-CMs, we illuminated transfected cells (GECI only, no ChR2) with 0.19 W/cm 2 470 nm LED light (Fig. 4b, c). Under these conditions, R-GECO1 exhibited a substantial photoactivation effect with a transient 200% increase in red fluorescence. Under the same illumination conditions, K-GECO1 had negligible change in red fluorescence. When we co-transfected iPSC- 12 CMs with both K-GECO1 and ChR2, blue light stimulation reliably induced Ca 2+ transients (Fig. 4d), demonstrating that the combination of K-GECO1 and ChR2 are suitable for all-optical excitation and imaging of iPSC-CMs.
Performance of K-GECO1 in organotypic brain slices. We further tested the performance of K-GECO1 by expressing it in organotypic slices of the newborn rat ventromedial nucleus (VMN) of the hypothalamus. Expression of K-GECO1 enabled visualization of both neuronal cell bodies and processes (Fig. 5a). We investigated the performance of K-GECO1 under pharmacological stimulation by ATP (100 μM), which activates suramin-sensitive ATP-receptors and induces an influx of extracellular Ca 2+ , thus increasing cytosolic Ca 2+ concentration 39 . Upon treatment with ATP, neurons expressing K-GECO1 underwent a mean increase in fluorescence intensity of 3.26 + 0.18-fold (n = 21) (Fig. 5b).
To compare the performance of K-GECO1 with the small molecule-based green cytosolic Ca 2+ indicator, Fluo-4AM, we loaded the dye into VMN neurons that were expressing K-GECO1 (Fig. 5c). When treated with ATP, these neurons (n=3) exhibited a 3.01 + 0.86-fold increase in K-GECO1 fluorescence, but only a 0.70 + 0.12-fold increase in Fluo-4 fluorescence (Fig. 5d). In non-transfected cells stained with Fluo-4AM, we did not observe any crosstalk from Fluo-4AM into the red channel. Overall, K-GECO1 unravels robust responses to cytosolic Ca 2+ concentration changes in neurons in organotypic brain slices. 13 In vivo Ca 2+ imaging with K-GECO1. To test K-GECO1 in zebrafish spinal cord sensory neurons in vivo, we transiently expressed K-GECO1 in Rohon Beard cells.
Zebrafish Rohon-Beard (RB) cells have previously been used for in vivo GECI imaging and shown to fire a single spike in response to each electrical pulses to the skin 40 . Electrical stimulations were applied to trigger Ca 2+ transients at 3 days post fertilization. 2-photon imaging with excitation at 1140 nm ( Fig. 6a) revealed that K-GECO1 filled both the cytoplasm and nucleus in vivo in zebrafish RB neurons (Fig.   6b). Cytoplasmic K-GECO1 exhibited ~40% fluorescence intensity increase to Ca 2+ transients triggered by a single pulse stimulus (Fig. 6c). When the RB neurons were stimulated with 5 to 20 repetitive stimuli, 50-100% increases in K-GECO1 fluorescence were observed (Fig. 6d). As expected, the fluorescence response in the nucleus was diminished with respect to the response in the cytosol, and exhibited a slower recovery to baseline (Fig. 6c, d). Compared to the optimized red fluorescent indicator jRGECO1a, K-GECO1 showed decreased sensitivity in zebrafish in terms of stimulus-evoked fluorescence change (Fig. 6e, f), whereas the half decay time was comparable (Fig. 6g, h). Consistent with the results from dissociated neurons, even distribution of the K-GECO1 red fluorescence in RB cells were observed in zebrafish neurons in vivo ( Supplementary Fig. 4a,b), while jRGECO1 exhibited fluorescence accumulations ( Supplementary Fig. 4c).
To evaluate K-GECO1 in the mouse primary visual cortex (V1) in vivo, V1 neurons were infected with adeno-associated virus (AAV) expressing nuclear export signal (NES) tagged K-GECO1 under the human synapsin1 promoter (AAV-SYN1-NES-K-GECO1). The majority of V1 neurons can be driven to fire action potentials in response to drifting gratings. Eight-direction moving grating visual stimuli were 14 presented to the contralateral eye (Fig. 7a). K-GECO1 expressing L2/3 neurons exhibited cytoplasmic red fluorescence (Fig. 7b), and 2-photon imaging revealed visual stimulus-evoked fluorescence transients in subsets of neurons (Fig. 7c). We compared the performance of K-GECO1 with other red GECIs using previously established metrics 17,24 . In terms of the fraction of neurons detected as responsive in the visual cortex, K-GECO1 is higher than RCaMP1h, but lower than R-GECO1 and other optimized red indicators (Fig. 7d). The mean ΔF/F0 at the preferred visual stimulus is reflective of indicator sensitivity. By this metric, K-GECO1 has sensitivity that is comparable to R-GECO1 and jRCaMP1a, but less than jRGECO1a (Fig. 7e).
Lysosomal accumulation was previously observed in mouse V1 neurons labeled with jRGECO1a, but not in the ones with jRCaMP1a/b 24 . Fixed brain tissue sections, prepared as previously reported for jRGECO1a and jRCaMP1a/b 24 , revealed no signs of lysosomal accumulation in K-GECO1 expressing V1 neurons ( Supplementary Fig. 5a). As with both jRGECO1a and jRCaMP1a/b, in vivo functional imaging of K-GECO1 did exhibit fluorescent clump-like structures ( Supplementary Fig. 5b), yet these structures were not observed in fixed sections of the same tissue. We are currently unable to explain this discrepancy. Overall, the results demonstrate that K-GECO1 can be used to report physiological Ca 2+ changes in neurons in vivo with performance that matches or surpasses that of other firstgeneration red fluorescent Ca 2+ indicators. 15

Discussion
Although green fluorescent GECIs are currently the most highly effective tools for in vivo visualization of neuronal signaling, we anticipate that they will one day be made redundant by red fluorescent GECIs due to the inherent advantages associated with longer wavelength fluorescence. The transmittance of tissue increases as wavelength increases, so red fluorescent GECIs will enable imaging of neuronal activity deeper into brain tissue than is possible with green fluorescent GECIs, assuming all other properties are equivalent 24,41 . In addition, red fluorescent GECIs enable multiparameter imaging in conjunction with green fluorescent indicators, and facilitate simultaneous imaging and optical activation when used in conjunction blue light activatable optogenetic actuators such as channelrhodopsin-2 (Ref. 42).
These limitations include decreased sensitivity for RCaMP variants and complicated photophysics and lysosomal accumulation for R-GECO variants. As both green and red GECIs have analogous designs and contain identical Ca 2+ -binding domains, these undesirable characteristics are related to the RFP scaffold used to generate red GECIs.
In an effort to overcome the limitations associated with current RFP scaffolds, we turned our attention to the eqFP578-derived lineage of monomeric RFPs (i.e., mKate and its derivatives) [7][8][9] , which tend to give bright and evenly-distributed fluorescence when expressed in the neurons in transgenic mice 30 . Using semi-rational design and directed evolution we developed a new red fluorescent Ca 2+ indicator, K-GECO1, based on a mKate variant FusionRed 9 . We anticipated that K-GECO1 would retain 16 the favorable traits associated with its starting template RFP. We have found this expectation to be generally true, as we did not observe lysosomal aggregation in dissociated rat neurons, or zebrafish neurons, or fixed mouse brain tissue expressing K-GECO1. Some fluorescent punctate-like structures were observed during in vivo functional imaging.
The other distinctive feature of K-GECO1 is the use of the ckkap peptide as the CaM binding partner for the Ca 2+ -binding motif. Consistent with previous reports 23, 34 , the ckkap/CaM motif yielded a lower apparent Kd for Ca 2+ and faster kinetics (relative to RS20/CaM), and an apparent Hill coefficient close to 1. These characteristics should enable more sensitive detection of Ca 2+ dynamic at physiological ranges, as is evident from K-GECO1's large single action potential fluorescence response amplitude. With a Hill coefficient close to 1, K-GECO1 should provide a more linear Ca 2+ response following multiple stimuli. The x-ray crystal structure of K-GECO1 suggests that the indicator has a "selfcontained" fluorescence modulation mechanism, similar to that proposed for R-GECO1 (Ref. 22,29). Unlike GCaMP, in which the fluorescence modulation mechanism is dependent on the interactions with a residue of CaM 43 (Fig. 2l), the K-GECO1 Ca 2+ -bound state is likely stabilized by the hydrogen bonding between the phenolate group of chromophore and linker1 residue Asn32 (Fig. 2i). This makes the cpFusionRed protein in K-GECO1 a potentially useful template as a signal transduction domain to be combined with other binding domains for development of new types of red fluorescent indicators. The crystal structure also reveals that the ckkap/CaM motif in K-GECO1 has a reversed binding orientation for CaM when compared with the RS20/CaM binding patterns in R-GECO1, RCaMP, and GCaMP6 ( Fig. 2e-h). These results indicate that the GCaMP-like design is versatile enough to tolerate different peptide conformations and CaM orientations, and that exploring a wider range of CaM binding partners is likely to lead to GECIs with new and improved properties.
First generation red GECIs, including mApple-based R-GECO1 and mRuby-based RCaMP1h, have been optimized using a neuron screening platform 24,44 , resulting in jRGECO1a and jRCaMP1a/b with greatly improved in vivo performance for detection of action potentials. Although K-GECO1 is a first generation red GECI, it already provides performance that, by some criteria, is comparable to second generation red GECIs. Specifically, K-GECO1 has a fluorescent response to single action potentials that is similar to that of jRGECO1a (and superior to jRCaMP1a/b) and faster dissociation kinetics than either jRGECO1a or jRCaMP1a/b. However, by other criteria, K-GECO1 will require further optimization in order to match the performance of second generation red GECIs. For example, K-GECO1 does not provide the same level of in vivo sensitivity as the highly optimized jRGECO1a. In addition, K-GECO1 showed some blue light-dependent photoactivation during in vitro characterization, though less than R-GECO1. The photoactivation of K-GECO1 was not detectable in our characterization in cultured dissociated neurons ( Supplementary Fig. 3c) or in iPSC-CMs (Fig.4c), suggesting that it is more suitable than R-GECO1 for use with blue/cyan excitable optogenetic actuators.
In summary, we have demonstrated the utility of K-GECO1 in various cell types including HeLa cells, dissociated neurons, iPSC-CMs, neurons in organotypic rat 18 brain slices, zebrafish RB cells, and mouse V1 neurons in vivo. Though not yet ideal by all criteria, K-GECO1 represents a step forward in the development of red GECIs.
As with R-GECO1 and RCaMP1h, further optimization using a neuron-based screening approach is likely to yield K-GECO variants with much improved sensitivity and performance in vivo. 19

In vitro characterization
To purify K-GECO variants for in vitro characterization, pBAD/His B plasmid encoding the variant of interest was used to transform electrocompetent E. coli DH10B cells and then plated on LB agar plate with ampicillin (400 µg/mL). Single colonies were picked and inoculated into 5 mL LB medium supplemented with 100 g/mL ampicillin. Bacterial subcultures were incubated overnight at 37°C. 5 mL of bacterial subculture was added into 500 mL of LB medium with 100 µg/mL ampicillin.
The cultures were incubated at 37°C to an OD of 0.6. Following induction with Larabinose to a final concentration of 0.02% (wt/vol), the cultures were then incubated 21 at 20°C overnight. Bacteria were harvested by centrifugation at 4,000 g for 10 min, resuspended in 30 mM Tris-HCl buffer (pH 7.4), lysed using French press, and then clarified by centrifugation at 13,000 g for 30 mins. Proteins were purified from the cell-free extract by Ni-NTA affinity chromatography (MCLAB). The buffer of purified proteins was exchanged into 10 mM MOPS, 100 mM KCl, pH 7.2. Absorption spectra were recorded on a DU-800 UV-visible spectrophotometer (Beckman) and fluorescence spectra were recorded on a Safire2 fluorescence plate reader (Tecan).
For quantum yield (QY) determination, the fluorescent protein mCherry was used as a standard. The detailed protocol has been described previously 18 . Briefly, the fluorescence emission spectra of each dilution of mCherry and K-GECO variants protein solution were recorded. The total fluorescence intensities were obtained by integration. Integrated fluorescence intensity versus absorbance was plotted for both mCherry and K-GECOs. QY was determined from the slopes of mCherry and K-GECOs. Extinction coefficient (EC) was determined by first measuring the absorption spectrum of K-GECO variants in Ca 2+ -free buffer and Ca 2+ -buffer. Following alkaline denaturation, the absorption was measured. With the assumption that the denatured chromophore has an EC of 44,000 M -1 cm -1 at 446 nm, the protein concentration was determined. EC of K-GECO variants were calculated by dividing the peak absorbance maximum by the concentration of protein.
For the Ca 2+ Kd determination, purified protein solution was diluted into a series of buffers which were prepared by mixing Ca 2+ -buffer and Ca 2+ -free buffer with free Ca 2+ concentration ranges from 0 nM to 3,900 nM. The fluorescence intensity of K-GECO variants in each solution was measured and subsequently plotted as a 22 function of Ca 2+ concentration. Data was fit to the Hill equation to obtain Kd and the apparent Hill coefficient.
Two-photon excitation spectra and cross sections were measured as previously reported 49 , with the following adjustments. For the two-photon excited (2PE) spectra, fluorescence was collected through a 694/SP filter for K-GECO1 (Semrock). To correct for wavelength-to-wavelength variations in the laser parameters, a correction function using Rhodamine B in MeOH and its known 2PE spectrum was applied 50 . Two-photon cross sections were measured at 1100 nm for K-GECO1, with Rhodamine B in MeOH as a reference standard. The fluorescence for cross sections were collected through a narrow bandpass filter, 589/15 (Semrock), and differential quantum efficiencies were obtained at 582 nm with a PC1 ISS spectrofluorimeter (this wavelength corresponded to the bandpass center of the above filter when used in the MOM Sutter Instruments microscope due to its tilted position). Since the filter (694/SP) used for the 2PE spectra measurements covers the fluorescence of both the neutral and anionic forms of the chromophore, the spectrum of a particular Ca 2+ state of a protein represents a combination of the unique 2PE spectra of the neutral and anionic forms, weighted to their relative concentrations (ρ, concentration of one form divided by the total chromophore concentration) and quantum yields. The y-axis of the total 2PE spectrum is defined by F2(λ) = σ2,N(λ) φN ρN + σ2,A(λ) φA ρA , where σ2(λ) is the wavelength-dependent 2-photon cross section and φ is the fluorescence quantum yield of the corresponding form (N for neutral or A for anionic in the subscript). At the wavelengths used to measure the cross sections (1060 and 1100 nm), σ2,N is assumed to be zero, and φA and ρA were independently measured to give a value for F2 (GM). The relative concentrations of the neutral and anionic forms 23 were found by measuring the absolute extinction coefficients of each respective form in the Ca 2+ -free and the Ca 2+ -bound states. These differ from the effective extinction coefficients reported in Supplementary Table 1 For comparison of K-GECO1 and Red GECIs in stimulated cultured neuron cells, the procedure was done as previously reported 24 . Briefly, red GECIs were expressed after electroporation into rat primary hippocampal neurons (P0) using Nucleofector system (Lonza). For stimulation, action potentials were evoked by field stimulation.
TxRed (540-580 nm excitation, 593-668 nm emission, and 585 nm long pass dichroic mirror) filter set was used for illumination. Responses were quantified for each cell as change in fluorescence divided by baseline fluorescence before stimulation. Signal-to-noise ratio (SNR) was quantified as the peak fluorescence signal over baseline, divided by the standard deviation of fluorescence signal before the stimulation. 28 iPSC-CMs were purchased from Axol Bioscience. Cells were plated in two wells of a 6-well plate and cultured for 4 days in Cardiomyocyte Maintenance Medium (Axol Bioscience) to 60-80% confluency. Cells then were then transferred to Fibronectincoated (1%) coverslips and imaged in Tyrode's buffer. Cells were transfected using Organotypic hypothalamic rat brain slice imaging For transfection of organotypic slices, after 8-10 days of organotypic slice culturing, the VMN areas were transfected with an electroporation technique previously described 47 . Specifically, the insert with the slice was placed on a platinum plate petri dish electrode (Bex Co Ltd) and electroporation buffer (HBSS with 1.5 mM MgCl2 and 10 mM D-glucose) was filled between the electrode and the membrane.
Plasmids of pcDNA3.1-K-GECO1 were dissolved in electroporation buffer at a concentration of 1 μg/ml and 10 µl of this solution was added to just cover the slice.
Then, a square platinum electrode (Bex Co Ltd) was placed directly above the slice.
Five 25 V pulses (5 ms duration, interval 1 s) were applied twice (the second time with reversed polarity) using a pulse stimulator (Sequim) and an amplifier (Agilent).
The electroporation buffer was replaced with supplemented NbActiv4 medium and slices were returned to the incubator. 30 For imaging of cytosolic Ca 2+ dynamics using K-GECO1, an upright FV1000 confocal microscope equipped with FluoView software and a 20× XLUMPlanF1 water immersion objective (NA 1.0) was used (Olympus). The millipore insert containing the transfected brain slice was placed in a custom-made chamber and mechanically fixed with a platinum harp. The slices were then perfused at 31°C with artificial cerebrospinal fluid (ACSF) containing (in mM) 120 NaCl, 3 KCl, 1 CaCl2, 1.

Mouse V1 imaging
For in vivo mouse V1 imaging, the procedure was done as previously reported 24 .
Briefly, AAV injection was used for expression of K-GECO1 in Mouse V1 neurons. 32 After the virus injection, cranial window was implanted. The animal was then placed under a microscope at 37°C and anesthetized during imaging. A custom-built 2photon microscope was used for imaging with a pulse laser as light source and a 16× 0.8 NA water immersion lens as objective. Laser power was 100-150 mW at the front aperture of the objective lens. Moving grating stimulus trial consisted of a blank period followed by a drifting sinusoidal grating with eight drifting directions with 45° separation. The gratings were presented with an LCD screen placed in front of the center of the right eye of the mouse. For fixed tissue analysis, mice were anesthetized and transcardially perfused. The brains were then removed and postfixed. Sections of the brains were coverslipped and imaged using confocal microscopy (LSM 710, Zeiss).

Statistical analysis
All data are expressed as means ± s.d. Sample sizes (n) are listed for each experiment. For V1 functional imaging, ANOVA test (p=0.01) was used to identify responsive cells for each of the grating stimuli. 33    The number in parentheses is for the highest resolution shell.