A Cre-Dependent GCaMP3 Reporter Mouse for Neuronal Imaging In Vivo

Fluorescent calcium indicator proteins, such as GCaMP3, allow imaging of activity in genetically defined neuronal populations. GCaMP3 can be expressed using various gene delivery methods, such as viral infection or electroporation. However, these methods are invasive and provide inhomogeneous and nonstationary expression. Here, we developed a genetic reporter mouse, Ai38, which expresses GCaMP3 in a Cre-dependent manner from the ROSA26 locus, driven by a strong CAG promoter. Crossing Ai38 with appropriate Cre mice produced robust GCaMP3 expression in defined cell populations in the retina, cortex, and cerebellum. In the primary visual cortex, visually evoked GCaMP3 signals showed normal orientation and direction selectivity. GCaMP3 signals were rapid, compared with virally expressed GCaMP3 and synthetic calcium indicators. In the retina, Ai38 allowed imaging spontaneous calcium waves in starburst amacrine cells during development, and light-evoked responses in ganglion cells in adult tissue. Our results show that the Ai38 reporter mouse provides a flexible method for targeted expression of GCaMP3.

GECIs have to be introduced using gene transfer methods. Mouse experiments have mostly relied on in utero electroporation (IUE) Mao et al., 2008) or viral infection (Tian et al., 2009;Lütcke et al., 2010;Borghuis et al., 2011). These methods require surgery on each animal, produce inhomogeneous expression patterns, and lead to undesirable expression timelines. For example, IUE produces high expression early during development, which can cause permanent cytotoxicity (Tian et al., 2009). Viral transduction ramps up over time, causing nonstationary expression levels. Transgenic methods of expression could potentially solve these issues.
Several GECI-expressing transgenic mouse lines have been reported (Hasan et al., 2004;Díez-García et al., 2005;Tallini et al., 2006Tallini et al., , 2007Heim et al., 2007;Atkin et al., 2009), but each of these transgenic mouse lines produced robust expression only in selected tissues, limiting their utility. Among GECIs, the recently developed indicator GCaMP3 has superior sensitivity, SNR, and photostability (Tian et al., 2009). Here, we report a GCaMP3 reporter mouse (Ai38) based on the flexible Cre/lox system (Sauer and Henderson, 1988;Orban et al., 1992). Upon crossing with different Cre lines, Ai38 reporter mice allow flexible and reproducible GCaMP3 expression in a wide variety of neuronal classes, including ganglion and starburst amacrine cells in the retina, pyramidal cells in the cortex, and Purkinje cells in the cerebellum. We validate the performance of GCaMP3 in these neuronal populations under physiological conditions.

Materials and Methods
All surgeries and experimental procedures were conducted under guidelines of the Janelia Farm Research Campus Institutional Animal Care and Use and Biosafety Committees. Adult mice (P56 -P120) of either sex were used in most experiments, unless otherwise noted.
Transgenic mice. The Ai38 floxed GCaMP3 reporter mouse line was generated using a knock-in strategy into the ROSA26 locus that was previously demonstrated to have robust and ubiquitous expression of the transgene by using the CAG (cytomegalovirus early enhancer/chicken ␤-actin) promoter and the woodchuck hepatitis virus posttranscriptional regulatory element (WPRE), made Cre-dependent with a loxstop-lox (LSL) cassette (Madisen et al., 2010). The GCaMP3 insert was cloned into a ROSA26-pCAG-LSL-WPRE-bGHpA targeting vector (Madisen et al., 2010), in between the LSL and the WPRE sequences. The LSL sequence contained LoxP-Stop codons-3ϫ SV40 poly(A)-LoxP. The final targeting vector ( Fig. 1 A) also contained 5Ј and 3Ј homology arms of 1.1 and 4.3 kb, as well as an AttB-pPGK-FRT-Neo-PGK poly(A)-AttP cassette for positive selection, but it did not contain a pPGK-DTA cassette for negative selection. The targeting vector was linearized and transfected into the 129/B6 F1 hybrid embryonic stem (ES) cell line, G4. G418 (Geneticin)-resistant (thus containing the neomycin resistance marker aminoglycoside 3Ј-phosphotransferase, Neo) ES clones were screened by PCR using primers spanning the 1.1 kb 5Ј genomic arm (forward primer, 5Ј-gggctccggctcctcagaga-3Ј; reverse primer, 5Ј-atgccaggcgggccatttac-3Ј).
Correctly targeted ES clones were injected into C57BL/6J blastocysts to obtain chimeric mice following standard procedures. Chimeric mice were bred with C57BL/6J mice to obtain germline transmitted F1 mice. Progeny mice from a single correctly targeted ES clone were maintained and used for studies. Southern blot analysis (see Fig. 1 A) was performed on these mice to confirm the correct targeting into the ROSA26 locus by probing HindIII-digested genomic DNA with a 1.1 kb genomic fragment from immediately upstream of the 5Ј arm, as well as to confirm the single-copy integration of the transgene by probing SulI/SapI-or StuI/ BclI-digested genomic DNA with an 798 bp SalI/MluI fragment from GCaMP3. Genotyping of the Ai38 mice was performed by PCR (forward primer, 5Ј-cttcaagatccgccacaacatcg-3Ј; reverse primer, 5Ј-ttgaagaagatggtgcgctcctg-3Ј), which amplifies a 546 bp fragment of GCaMP3.
The Ai38 mice used in the current studies still contain the PGK-Neo marker cassette downstream of the GCaMP3 expression cassette. If desired, the PGK-Neo cassette can be deleted by breeding the Ai38 mice with ROSA26-PhiC31 mice (JAX stock no. 007743). Our previous experience did not indicate any expression difference between the Ai-series of reporter lines with or without the PGK-Neo cassette (e.g., between Ai9 and Ai14 floxed tdTomato reporters) (Madisen et al., 2010).
Mouse and DNA availability. We have deposited the Ai38 mice to The Jackson Laboratory for distribution (JAX stock no. 014538; http://jaxmice.jax.org/strain/014538.html). We have also deposited the Ai38 targeting vector DNA plasmid to Addgene (http://www.addgene.org/). Viral infection. Adenoassociated virus AAV2/1 carrying the construct syn-GCaMP3 (GCaMP3 under the neural-specific human synapsin-1 promoter) (Kügler et al., 2001) (2.25 ϫ 10 13 genome copies/l) was injected at a depth of ϳ250 m to the primary visual cortex (two sites, 2.5 and 2.9 mm lateral from the lambda suture) of adult (ϳ2 months) C57BL/6J mice. Small volumes (ϳ30 nl) of viral suspension were delivered over a period of 5 min per site using a custom-built volumetric injector. The AAV was allowed to infect and express GCaMP3 for 3 weeks after injection.
Mouse preparation for in vivo imaging. Mice were anesthetized using isoflurane (3% for induction, 1.5-2% during surgery). In some cases, 2 mg/kg dexamethasone was administered intraperitoneally to reduce brain edema. A circular craniotomy (2-3 mm diameter) was placed above V1 (centered 2.7 mm lateral from lambda suture). OGB-1-AM (Invitrogen) was injected as previously described (Stosiek et al., 2003;Komiyama et al., 2010;Zariwala et al., 2011). The craniotomy was covered with agarose (1.2-1.5%) and a round glass coverslip (Warner Instruments; 5 mm diameter; no. 1 thickness) was cemented to the skull to reduce motion of the exposed brain. A custom titanium head post was fixed to the skull using black dental cement (Contemporary Ortho-Jet). The animal was then transferred to the imaging setup, where it was placed on a warm blanket (37°C) and kept anesthetized using 0.5% isoflurane and sedated with chlorprothixene (20 -40 l at 0.33 mg/ml, i.m.) (Niell and Stryker, 2008).
Visual stimuli. The moving grating stimuli were generated using the Psychophysics Toolbox (Brainard, 1997;Pelli, 1997) in MATLAB (Math-Works). Each stimulus trial consisted of a 4 s blank period (uniform gray at mean luminance) followed by a 4 s drifting sinusoidal grating (0.05 cycles per degree; 1 Hz temporal frequency). The visual stimuli were synchronized to individual image frames using frame-start pulses provided by ScanImage 3.7. The gratings were presented through an LCD monitor (30 ϫ 40 cm), placed 25 cm in front of the center of the right eye of the mouse. The monitor subtended an angle of Ϯ38°horizontally and Ϫ20 to ϩ38°vertically around the eye of the mouse.
Data analysis. Slow drifts in brain position in the X and Y directions were corrected using the Turboreg plug-in in ImageJ (Thévenaz et al., 1998). All remaining analyses were performed in MATLAB. Regions of interest (ROIs) corresponding to visually identifiable cell bodies were selected using a semiautomated algorithm. For GCaMP3, ring-shaped ROIs were placed at the cytosolic regions of the cells [excluding the nucleus; GCaMP3 expression is typically restricted to the cytoplasm (Tian et al., 2009)]. For OGB-1, circular ROIs covering the whole soma were used. The fluorescence time course of each cell was measured by averaging all pixels within the ROI. The neuropil contamination was corrected using published methods ; in short, the neuropil signal F neuropil (t) surrounding each cell was measured by averaging the signal of all pixels within a 20 m circular region from the cell center (excluding all selected cell bodies). The true fluorescence signal of a cell body was estimated as follows: with r ϭ 0.3 throughout the study. After neuropil correction, the ⌬F/F 0 of each trial was calculated as (F Ϫ F 0 )/F 0 , where F 0 is the baseline fluorescence signal averaged over a 2 s period immediately before the start of grating stimulation. Visually responsive neurons were defined using ANOVA across blank and eight direction periods ( p Ͻ 0.01) (Ohki et al., 2005). Of the responsive cells, orientation-selective cells were defined by ANOVA across eight direction periods ( p Ͻ 0.01).
The orientation selectivity index (OSI), tuning width, and direction selectivity index (DSI) were calculated for visually responsive cells. First, the preferred orientation ( pref ) of the cell was determined as the angle of the grating that produced the strongest response. The orientation tuning curve was constructed by measuring the mean ⌬F/F 0 over the 4 s stimulus period for each orientation. We then fitted the tuning curve as the sum of two Gaussian functions centered on pref and pref ϩ with equal width , different amplitudes A 1 and A 2 , and a constant baseline B (Niell and Stryker, 2008). The value of was required to be Ͼ15°to reflect the limit of our stimulus set (45°separation) in resolving sharper tuning. The OSI was defined as follows: where R pref and R ortho are the response amplitudes at the preferred ( pref ) and the orthogonal orientation ( pref ϩ /2). Tuning width (TW) was defined as the half-width at half-maximum of the fitted Gaussian (TW ϭ ⅐ ͱ2 ⅐ ln 2). Finally, DSI was calculated as follows: where R opposite is the response in the opposite direction ( pref ϩ ). Estimation of intracellular GCaMP3 concentration. We measured intracellular GCaMP3 concentrations in acute coronal brain slice of Emx1-Cre:Ai38 mice. The cytosolic GCaMP3 fluorescence in multiple individual neurons was measured using a two-photon microscope. The slice was incubated in artificial CSF with 20 mM extracellular [Ca 2ϩ ] and Figure 1. Expression of GCaMP3 in various cell populations in the Ai38 mouse. A, Schematic of the gene targeting strategy to generate the Ai38 reporter mice. GCaMP3-containing Cre reporter cassette (shown in between the dashed red lines) was targeted to the ROSA26 locus in the intron between endogenous exons 1 and 2. Locations of the 5Ј and 3Ј arms (1.1 and 4.3 kb, respectively) for homologous recombination are indicated by the black bars. Genotyping primers are located in the GCaMP3 sequence, as indicated by the open arrows (the F and R primers). Correct gene targeting was confirmed by Southern blot (top right, lanes 1-2) using the ROSA26 locus-specific probe (the R probe) immediately upstream of the 5Ј arm on HindIII (H)-digested genomic DNA, which gives rise to a 4.5 kb band for the wild-type (WT) locus and a 5.6 kb band for the gene-targeted locus. Lane 1, HindIII on WT ES cell DNA. Lane 2, HindIII on Ai38 mouse tail DNA. The single-copy integration of the transgene was confirmed by the appearance of a single band in Southern blot (top right panel, lanes 3-6) using the GCaMP3-specific probe (the G probe) on StuI/BclI-or StuI/SapI-digested genomic DNA, which gives rise to a 5.9 kb or a 5.8 kb band, respectively. Lane 3, StuI/BclI on WT ES cell DNA. Lane 4, StuI/SapI on WT ES cell DNA. Lane 5, StuI/BclI on Ai38 mouse tail DNA. Lane 6, StuI/SapI on Ai38 mouse tail DNA. B, Native fluorescence of GCaMP3 in a coronal cortical section from a Wfs1-Tg2-CreERT2:Ai38 mouse. C, GCaMP3 fluorescence in a cortical section of a Emx1-Cre:Ai38 mouse. D, Pcp2-Cre:Ai38 mouse expressing GCaMP3 in cerebellar Purkinje cells, native GCaMP3 fluorescence. E, GCaMP3 fluorescence near the AAV-syn-Cre injection site in the cortex of an Ai38 mouse. F, Pvalb-2A-Cre:Ai38 mouse expressing GCaMP3 in retinal ganglion cells. The retina was radially incised to permit flattening on a microscope glass slide for confocal microscopy. Green, GCaMP3; blue, DAPI. Radial blood vessels (black) are visible, converging on the optic disk (center). Right, Z-projection of confocal stack through the fiber and ganglion cell layers (higher magnification of the boxed area). The image shows GCaMP3-expressing ganglion cells and bundles of GCaMP3-positive axons (arrows), oriented toward the optic disk. G, Ai38 crossed with Chat-Cre targets GCaMP3 expression to SACs. Left, Maximum intensity projection of a confocal stack through the ganglion cell layer (GCL), inner plexiform layer (IPL), and inner nuclear layer (INL) of a whole-mount retina (age P8). Inset, Single z-plane through somata of SACs in the INL. Right, Confocal image of a side-on view of the same retinal slice.
1 mM ionomycin to produce saturating intracellular calcium levels. Neuronal brightness was compared with calibrated solutions of GCaMP3 in a thin cuvette.
Histology. Coronal sections (50 m) of the visual cortex were imaged using a confocal microscope (Zeiss LSM 510). Special care was taken to avoid saturation of the image pixels. Identical laser power and image settings were used for the intensity comparisons shown in Figure 2E. The brightness of individual cells was measured by averaging pixels within ROIs covering the cytosolic regions. Background (measured in nearby cell-free regions) was subtracted.
Retinal imaging in vitro. Confocal fluorescence images of GCaMP3 expression patterns in the retina (Fig. 1 F, G) were obtained with a Zeiss LSM 510 laser-scanning microscope. For functional imaging in vitro, retinas were prepared using standard methods (Borghuis et al., 2011). A retina was removed from the sclera and mounted on nitrocellulose filter paper (HAWP01300; Millipore), photoreceptor side down. Apertures (1 mm diameter) in the filter paper permitted light stimulation of the photoreceptors through the condenser lens. Retinas were continuously perfused with oxygenated (95% O 2 /5% CO 2 ) Ames medium (Sigma-Aldrich) at 37°C.
Two-photon fluorescence measurements were obtained with a modified Olympus BX microscope with an Olympus 60ϫ, 0.9 NA, LUMPlanFl/IR objective, an ultrafast pulsed laser (Chameleon Vision II; Coherent) tuned to 910 nm, and controlled with ScanImage. Images (512 ϫ 128 pixels) were acquired at 16 frames per second. Electrical field stimulation (1 kHz, 20 nA peak current) was generated with the capacitance compensation circuit on a Neurodata intracellular amplifier and delivered through the patch pipette (ϳ3 M⍀). All data were analyzed with custom algorithms in MATLAB as described previously (Borghuis et al., 2011).

Results
The Ai38 line was crossed to Cre lines to express GCaMP3 in selected cell populations in the retina, cerebral cortex, and cerebellum. We labeled retinal ganglion cells by crossing the Ai38 mouse to Pvalb-2A-Cre (Madisen et al., 2010) and Chat-Cre (Ivanova et al., 2010) mice. For imaging primary visual cortex (V1), Ai38 was crossed with three different Cre lines, Wfs1-Tg2-CreERT2, Pvalb-2A-Cre, and Emx1-Cre, which label different but overlapping cortical cell populations (Madisen et al., 2010). For imaging Purkinje cells, Ai38 was crossed with Pcp2-Cre mice (Barski et al., 2000). Results from the Ai38 mouse were compared with published data collected with small molecule dyes and virally transduced GECIs. For V1 imaging, we also performed a side-by-side comparison of Ai38-GCaMP3 with OGB-1 and AAV-syn-GCaMP3, under identical experimental conditions.

Characterization of the Ai38 line
The Ai38 reporter mouse carries the GCaMP3 gene under control of the CAG promoter to drive high levels of transgene expression. Expression was made Cre-dependent by introduction of a lox-stop-lox cassette (Fig. 1 A). The construct was targeted to the permissive ROSA26 locus (Soriano, 1999). Cell-specific GCaMP3 expression was achieved by crossing the Ai38 reporter mice with Wfs1-Tg2-CreERT2, Emx1-Cre, Pcp2-Cre, Pvalb-2A-Cre, and Chat-Cre mice. Expression patterns in the cortex (Fig. 1 B, C), cerebellum ( Fig. 1 D), and the retina (Fig. 1 F, G) are consistent with the Cre expression patterns of the driver lines (Barski et al., 2000;Gorski et al., 2002;Ivanova et al., 2010;Madisen et al., 2010). GCaMP3 expression in the Ai38 mouse was also confirmed with Cre delivery through AAV infection in the cortex. This led to homogeneous GCaMP3 expression in cells around the injection site (Fig. 1 E).
Previous GECI delivery using IUE or viral infection produced a subpopulation of cells (typically at the site of injection) with strong nuclear fluorescence (Tian et al., 2009). These filled cells displayed abnormal physiology and attenuated stimulus-evoked responses (Tian et al., 2009). Furthermore, the number of filled neurons increased with time, which fundamentally limits the duration and stability of chronic GECI imaging. We examined GCaMP3 fluorescence in various Ai38 crosses in brain areas including the cortex (Figs. 1 B, C, 2A), retina (Fig. 1 F, G), and cerebellum (Fig. 1 D). In all cases, we detected virtually no cells with filled nuclei. We further examined the effect of long-term trans- genic GCaMP3 expression in the Wfs1-Tg2-CreERT2:Ai38 cross. GCaMP3 expression was induced by tamoxifen at P7, and cortical tissue was examined 5 and 10 months after induction. In both cases, we found virtually no cells with filled nuclei (Fig. 2 A, B), and the brightness of cytosolic GCaMP3 was comparable between 5 and 10 months (Fig. 2 E). In contrast, filled neurons started to appear 3-4 weeks after AAV mediated GCaMP3 expression (Fig. 2C, arrows), and 8 months after injection most cells were brightly labeled (ϳ5.5-fold brighter than at 4 weeks; Fig. 2 E) with strong nuclear fluorescence (Fig. 2 D). The baseline GCaMP3 fluorescence was ϳ14-fold brighter with AAV than with the Ai38 reporter mouse (Fig. 2 E). We further determined the intracellular GCaMP3 concentration in Emx1-Cre:Ai38 cortical slices (see Materials and Methods). The intracellular GCaMP3 concentration was low (5.4 Ϯ 1.0 M; n ϭ 9). Thus, control of GCaMP3 expression might permit long-term expression with minimal adverse effects.
The Wfs1-Tg2-CreERT2 line labels ϳ40% of the excitatory cells in cortical layer 2/3 (Fig. 1 B). This is unlike the Emx1-Cre line, which labels almost all excitatory neurons in the cortex (Fig.  1C), including deeper layers (Madisen et al., 2010). Because of these differences in labeling, the Wfs1-Tg2-CreERT2 line shows reduced nonspecific signal due to the densely labeled neuropil (Fig. 2 A), and we therefore used this mouse for in vivo measurements in the primary visual cortex.

Wfs1 ؉ neurons show highly tuned GCaMP3 responses
Adult Wfs1-Tg2-CreERT2:Ai38 mice were anesthetized and implanted with a cranial window above primary visual cortex (V1) immediately before in vivo imaging. Mice were presented with oriented gratings moving in eight different directions.
Two-photon imaging revealed subsets of GCaMP3-positive neurons showing highly selective responses to oriented gratings ( Fig. 3 A, B,G). The orientation tuning of these cells was consistent with previous experiments where Wfs1 ϩ cells were identified based on tdTomato expression in the Ai9 reporter mouse line (Madisen et al., 2010) and loaded with OGB-1 (Zariwala et al., 2011). A majority of the responsive neurons were modulated at the temporal frequency of the moving gratings (1 Hz) (Fig.  3 B, C,E), consistent with a predominance of simple cells in layer 2/3 of mouse V1 (Mangini and Pearlman, 1980;Niell and Stryker, 2008;Liu et al., 2010;Smith and Häusser, 2010). Significant GCaMP3 responses (ANOVA, p Ͻ 0.01) were observed in ϳ10% of GCaMP3-positive cells (77 of 773 cells in 3 mice), with a range of 1-25% across 29 fields of view at depths ranging from 110 to 250 m (Fig. 3D). The percentage of visually responsive Wfs1-GCaMP3 ϩ cells was lower than that reported in previous studies using OGB-1 (Zariwala et al., 2011). This likely reflects a lower sensitivity of GCaMP3 to weak activity (one to three action potentials per 0.5 s bin) (Tian et al., 2009) than OGB-1. The responding cells, however, often showed strong fluorescence changes (up to 400%; Fig. 3B, cell 2). Overall, the average ⌬F/F of the cells at the preferred orientation showed a skewed distribution ( Fig. 4 A, all 773 cells).
The decay time of GCaMP3 fluorescence was determined as the time to half-decay of the transients after stimulus offset. To ensure robust SNR, this analysis was confined to trials with ⌬F/F at stimulus offset (⌬F/F stim_end Ͼ25%). The median T 1/2 of these calcium responses was 230 Ϯ 70 ms (n ϭ 320 trials in 65 cells, SD; Fig. 3F ). These decays are faster than those measured in previous studies using OGB-1 in V1 Smith and Häusser, 2010) and Fluo-4 in the somatosensory cortex (Sato et al., 2007). The GCaMP3 signal in the Ai38 mouse permitted us to resolve the phasic excitation at the drifting frequency during individual stimulus presentations (Fig. 3 B, C). Spectral analysis of the response waveform showed a clear peak at the drifting frequency (1 Hz) (Fig. 3E). Such modulations are hardly resolved in OGB-1 experiments (Fig. 4 F) (Ohki et al., 2005;Mrsic-Flogel et al., 2007;Kerlin et al., 2010;Zariwala et al., 2011).
We quantified the tuning properties of cells that showed significant visually driven responses (77 cells). For each responsive neuron, we measured the average ⌬F/F response at each orientation and fitted the response with a sum of two Gaussian functions (Fig. 3G) (see Materials and Methods) (Niell and Stryker, 2008). The OSI, tuning width, and DSI were calculated from the fitting parameters (Fig. 3H ). Visually responsive cells showed high orientation selectivity (mean Ϯ SD, OSI, 0.90 Ϯ 0.14), sharp tuning (mean Ϯ SD, tuning width, 21.8 Ϯ 9.6°) and a broad distribution of direction selectivity (mean Ϯ SD, DSI, 0.4 Ϯ 0.28). These properties are consistent with previously reported tuning properties in identified excitatory neurons in V1 using electrophysiology (Niell and Stryker, 2008;Liu et al., 2009) and OGB-1 imaging Zariwala et al., 2011).

Comparison of GCaMP3 and OGB-1 in V1
On average, a lower percentage of Wfs1 ϩ neurons showed significant responses to visual stimuli compared with studies using bulk loading of OGB-1-AM. For a direct comparison, we performed OGB-1 imaging in wild-type mice under identical stimulus and imaging conditions. Similar experiments were also performed with AAV-mediated GCaMP3 expression (see Materials and Methods).
The percentage of cells that were significantly visually responsive was ϳ10% for Wfs1-Tg2-CreERT2:Ai38 (77 of 773 cells; three animals), ϳ46% for OGB-1 (1673 of 3623 cells; three animals), and ϳ16% for AAV-syn-GCaMP3 (277 of 1762 cells; four animals). Of the responsive cells, the percentage of cells with significant orientation selectivity (ANOVA across eight orientations, p Ͻ 0.01) was ϳ78% for Wfs1-Tg2-CreERT2:Ai38 (60 of 77), ϳ48% for OGB-1 (807 of 1673), and ϳ45% for AAV-syn-GCaMP3 (126 of 227). The median fluorescence change at the preferred orientation was higher in OGB-1-labeled neurons than for GCaMP3 (Fig. 4 A). However, the maximum ⌬F/F achieved in GCaMP3-expressing neurons was substantially higher (Fig. 4 A). Whereas OGB-1 provided a higher average signal for low responders (cells that fall within the 50th to 80th percentile of the ⌬F/F distribution), among the top responders (Ͼ97th percentile) GCaMP3 did considerably better (Fig. 4 B). The percentage of GCaMP3-responsive cells was slightly higher with AAV than with the Ai38 reporter mouse. This is likely due to significantly higher expression levels and brightness of GCaMP3-expressing neurons in the case of viral expression (Fig. 4C), providing greater SNR.
The OSI, tuning width, and DSI of the visually responsive cells across the two GCaMP3 expression systems was then compared with those obtained in OGB-1 imaging (Fig. 4G-I ). The mean OSI for Wfs1-Tg2-CreERT2:Ai38 cells was markedly higher (0.90 Ϯ 0.14) compared with AAV-syn-GCaMP3 (0.70 Ϯ 0.27) and OGB-1 (0.53 Ϯ 0.27) (Fig. 4G). A possible explanation is that the Wfs1-Tg2-CreERT2:Ai38 selectively labels excitatory neurons, whereas in OGB-1 and AAV-syn-GCaMP3 the labeling includes inhibitory neurons known to show weaker or no orientation selectivity (Sohya et al., 2007;Niell and Stryker, 2008;Liu et al., 2009;Kerlin et al., 2010;Zariwala et al., 2011). Furthermore, the contamination of signal due to nonspecific neuropil responses (Kerr et al., 2005) can lower the orientation selectivity of neurons. Wfs1-Tg2-CreERT2:Ai38 animals showed sparser labeling (Fig. 1C) compared with AAV-syn-GCaMP3 and OGB-1 and therefore could be less susceptible to neuropil contamination. Indeed, compensating nonspecific neuropil signals reduced the percentage of responsive neurons for AAV-syn-GCaMP3 and OGB-1, but had minimal effect on Wfs1-Tg2-CreERT2:Ai38 (Fig.  4 J). Finally, GCaMP3 may fail to detect some weak responses at nonpreferred orientations. This can lead to an overestimation of the orientation selectivity in GCaMP3 experiments. The inclusion of poorly selective neurons in OGB-1 and AAV-syn-GCaMP3 imaging is also reflected in the comparison of both tuning width and DSI of neurons across these experiments (Fig.  4 H, I ). The highly selective responses of GCaMP3-expressing cells in Ai38 mice demonstrate that long-term expression of GCaMP3 at low, steady-state levels likely does not impact the functional tuning properties of the neurons imaged in this study.
Together, these results indicate that GCaMP3-responsive cells show orientation and direction preference typical for visual cortical neurons. The Ai38 reporter mouse gave lower signal and faster kinetics, but the functional properties of the neurons imaged appear normal compared with those following AAV infection or OGB-1 bulk loading.

OGB-1 responses in neurons expressing GCaMP3
The lower percentage of visually responsive neurons in GCaMP3 compared with OGB-1 experiments might be due to adverse effects of GCaMP3 expression. To test this possibility, we performed OGB-1 imaging in a mouse where GCaMP3 was expressed in the majority of cortical neurons throughout development. Emx1 is a marker for cortical excitatory neurons (Chan et al., 2001), and this expression pattern is recapitulated in the Emx1-Cre transgenic mouse line (Gorski et al., 2002). The Emx1-Cre mouse was crossed with the Ai38 line to produce expression of GCaMP3 in all cortical excitatory neurons. Adult Emx1-Cre:Ai38 mice were prepared for OGB-1 imaging in an identical manner to the OGB-1 recordings in wild-type mice. Imaging was performed at 800 nm to excite OGB-1, while minimizing GCaMP3 fluorescence.

Functional imaging in the developing and adult retina
Ai38 crossed with Chat-Cre showed retinal GCaMP3 expression exclusively in starburst amacrine cells (SACs) (Fig. 1G), consistent with known expression patterns (Ivanova et al., 2010). Synchronized, periodic calcium transients in starburst amacrine cells are a well known but incompletely understood feature of the developing vertebrate retina (Feller et al., 1996;Zhou, 1998;Wong, 1999;Zheng et al., 2006). Two-photon fluorescence imaging from SAC populations in a P8 Chat-Cre:Ai38 retina (Fig. 6A) showed large, periodic calcium transients on either side of the inner plexiform layer (Fig. 6B). Temporal delays in the activation of individual cells within the field of view were consistent with a traveling wave (Fig. 6A,B; arrowhead in Fig. 6A indicates the direction of wave propagation).

GCaMP3 responses in the cerebellum of Pcp2-Cre:Ai38 mice
In Pcp2-Cre:Ai38 mice, fluorescence microscopy confirmed high levels of expression in Purkinje cells (PCs) of the cerebellum (Fig.  1D). GCaMP3 expression appeared to be localized exclusively to PCs with nearly all PCs labeled. PC dendritic arbors, somata, and proximal axons could be clearly discerned, suggesting expression levels sufficient for functional imaging (Fig. 8A-C).
Climbing fibers generate all-or-none complex spikes and dendritic calcium transients in PCs with a frequency of Ͻ1 Hz in lightly anesthetized mice (Ozden et al., 2009;Schultz et al., 2009), whereas cerebellar granule cells can also induce subdendritic signals that span only part of the PC arbor (Wang et al., 2000). Under light anesthesia (0.75-1.0% isoflurane), we observed sparse but clearly visible fluorescence transients in one or a few isolated PC arbors at a rate of Ͻ1 Hz (Fig. 8 D, F ). We also observed large amplitude (Ͼ100% ⌬F/F) subdendritic signals that spanned a few PC branchlets (Fig. 8 E). Mean amplitudes for all transients (n ϭ 611 in 3 mice) were 30 Ϯ 11% ⌬F/F (mean Ϯ SEM; range, 9 -260% ⌬F/F). The low frequency of events compared with previous studies using OGB-1 (Ozden et al., 2009) could result from the high levels of the endogenous calcium binding proteins calbindin D28K and parvalbumin in PCs (Celio, 1990;Baimbridge et al., 1992), and slower calcium binding to GCaMP3 compared with OGB-1. Calbindin limits calcium concentration increases in response to single complex spikes to a few hundred nanomolar (Schmidt et al., 2003). This falls in the lower range of what GCaMP3 can detect (Tian et al., 2009).

Discussion
The sensitivity and kinetics of GCaMP3 make it a promising GECI for in vivo imaging of neuronal activity in a variety of different brain regions and organisms (Tian et al., 2009;Dombeck et al., 2010;O'Connor et al., 2010;Seelig et al., 2010;Borghuis et al., 2011). The Ai38 mouse now facilitates labeling of genetically defined populations of neurons with GCaMP3. The possibilities are mainly limited by the available Cre driver lines, the number of which is rapidly expanding (http://nagy.mshri.on.ca/cre_new/). Ai38 yielded homogeneous and stable expression levels over many months. Cell morphology, physiology, and response properties appeared unchanged during this time. This stable expression will greatly enable chronic imaging experiments, for which the time window provided by AAV expression is insufficient. The low (undetectable) cytomorbidity indicates that usable GCaMP3 concentrations can be tolerated longterm. Head-to-head comparison of Ai38 with AAV-syn-GCaMP3 and OGB-1 showed similar fluorescent response properties in vivo, with the former having superior temporal resolution.
The homogeneous and repeatable expression will permit easier pooling of imaging results across multiple experiments. For the crosses evaluated here, the expression levels in the retina ap-   (Hyvärinen, 1999). F, Various signals recorded from PC dendrites (ROIs selected with ICA are sorted left-to-right, and signals bottom-to-top).
peared to match the average expression levels obtained with viral transduction (Borghuis et al., 2011), and no increase in laser power was required to image those cells. However, high laser powers were required for cortical imaging in the Ai38 mouse. Higher expression levels might be required for some applications. Breeding mice homozygous for the GCaMP3 allele should increase the expression level. Further optimization of transgenic strategies should improve the expression level while preserving low cytomorbidity.
GCaMP3 is not the perfect calcium indicator. Further engineering of the GCaMP scaffold, or of other GECI scaffolds, will be required to achieve the sensitivity, kinetics, and SNR levels required to robustly detect very sparse neural activity in vivo. The expression and targeting cassette used here, as well as the standardized imaging assays established, will be useful for creating reporter mice from the next generation of GECIs as well. As GECI sensitivity to single spikes improves, we expect that the performance of the resulting transgenic mice in standardized assays such as the V1 experiment presented here will approach that of the best small molecule indicators.