Quantitative Comparison of Genetically Encoded Ca2+ Indicators in Cortical Pyramidal Cells and Cerebellar Purkinje Cells

Genetically encoded Ca2+ indicators (GECIs) are promising tools for cell type-specific and chronic recording of neuronal activity. In the mammalian central nervous system, however, GECIs have been tested almost exclusively in cortical and hippocampal pyramidal cells, and the usefulness of recently developed GECIs has not been systematically examined in other cell types. Here we expressed the latest series of GECIs, yellow cameleon (YC) 2.60, YC3.60, YC-Nano15, and GCaMP3, in mouse cortical pyramidal cells as well as cerebellar Purkinje cells using in utero injection of recombinant adenoviral vectors. We characterized the performance of the GECIs by simultaneous two-photon imaging and whole-cell patch-clamp recording in acute brain slices at 33 ± 2°C. The fluorescent responses of GECIs to action potentials (APs) evoked by somatic current injection or to synaptic stimulation were examined using rapid dendritic imaging. In cortical pyramidal cells, YC2.60 showed the largest responses to single APs, but its decay kinetics were slower than YC3.60 and GCaMP3, while GCaMP3 showed the largest responses to 20 APs evoked at 20 Hz. In cerebellar Purkinje cells, only YC2.60 and YC-Nano15 could reliably report single complex spikes (CSs), and neither showed signal saturation over the entire stimulus range tested (1–10 CSs at 10 Hz). The expression and response of YC2.60 in Purkinje cells remained detectable and comparable for at least over 100 days. These results provide useful information for selecting an optimal GECI depending on the experimental requirements: in cortical pyramidal cells, YC2.60 is suitable for detecting sparse firing of APs, whereas GCaMP3 is suitable for detecting burst firing of APs; in cerebellar Purkinje cells, YC2.60 as well as YC-Nano15 is suitable for detecting CSs.


INTRODUCTION
Understanding the spatio-temporal patterns of neuronal activity underlying brain function is one of the fundamental goals in neuroscience research, and requires techniques for the large-scale recording in living animals. The repertoire of in vivo multi-cell recording techniques has been enriched by the recent establishment of in vivo Ca 2+ imaging, a combination of multi-photon imaging and bolus loading of synthetic Ca 2+ dyes (Stosiek et al., 2003). In vivo Ca 2+ imaging allows not only multi-cell recording based on fast Ca 2+ transients generated by action potentials (APs; Markram et al., 1995;Schiller et al., 1995;Helmchen et al., 1996), but also the precise localization of recorded cells. It has thus contributed to unveiling the functional micro-architecture of many brain regions (Ohki et al., 2005(Ohki et al., , 2006Kerr et al., 2005Kerr et al., , 2007Sullivan et al., 2005;Rothschild et al., 2010;Smith and Häusser, 2010), which was difficult to achieve using classical electrode-based techniques. However, the lack of cell type specificity, the unrepeatability, and the short-lived nature (typically less than 1 day) of imaging using synthetic Ca 2+ dyes has remained an obstacle for further applications.
Genetically encoded Ca 2+ indicators (GECIs; for review, Miyawaki, 2005;, which are Ca 2+sensitive fluorescent proteins (FPs), can in principle offer an excellent solution to these problems, since they can be stably and specifically expressed in a targeted cell type by the use of appropriate promoters and transfection methods. [Ca 2+ ] i changes cause structural changes of the Ca 2+ -sensing domains in GECIs, which further cause changes in either (1) fluorescence resonance energy transfer (FRET) efficiency between two FPs or (2) the fluorescent intensity of a single circularly permutated (cp) FP, depending on the design of GECIs. GECIs have been successfully applied in many model organisms including Caenorhabditis elegans (Kerr et al., 2000), Drosophila melanogaster (Fiala et al., 2002), and Danio rerio (Higashijima et al., 2003), where electrode penetration and exogenous dye application are technically challenging. In the mammalian central nervous system (CNS), initial attempts using prototypical GECIs were somewhat disappointing (Hasan et al., 2004;Pologruto et al., 2004), but recently developed GECIs have been shown to display improved performance (Heim et al., 2007;Wallace et al., 2008;Tian et al., 2009;Horikawa et al., 2010;Lütcke et al., 2010) and have been used to address biologically relevant questions (e.g., Dombeck et al., 2010). Nevertheless, most applications of GECIs in the mammalian CNS have been limited to cortical and hippocampal pyramidal cells, and how GECIs perform in other cell types has remained largely unknown. There are a few exceptions where GCaMP2 has been tested in cerebellar granule cells as well as Purkinje cells (Díez-García et al., 2005, 2007Akemann et al., 2009), but the relationship between fluorescent changes and intracellular electrical responses of imaged cells was not investigated, nor was the performance of multiple GECIs compared under the same experimental conditions. The application of novel GECIs to broader contexts should be facilitated by quantitative comparison of their performance in reference to intracellular electrical signals.
In the present study, we selected the latest series of FRETbased GECIs ( Figure 1A): yellow cameleon (YC) 2.60, YC3.60 , YC-Nano15 (Horikawa et al., 2010), and the latest cpGFP-based GECI, GCaMP3 (Tian et al., 2009). We expressed each of them in mouse cortical pyramidal cells as well as in cerebellar Purkinje cells by in utero injection of recombinant adenoviral vectors Mikoshiba, 2003, 2004). All the YCs above utilize calmodulin (CaM) and M13 (Ca 2+ /CaMbinding peptide derived from skeletal muscle myosin light chain kinase) as Ca 2+ -sensing domain, but their in vitro affinities are modified by molecular engineering: YC3.60 carries a mutation in EF-hand motif of CaM (E104Q) resulting in a larger dissociation constant (K d ) value (∼250 nM) than that of YC2.60 (∼95 nM), while YC-Nano15 has an elongated linker between CaM and M13 (GGGGS) than that used in YC2.60 and YC3.60 (GGS), resulting in an extremely smaller K d value (∼15 nM). GCaMP3, which also utilizes CaM and M13 as Ca 2+ -sensing domain, was constructed by mutagenesis of GCaMP2, resulting in slightly lower K d value (660 nM; GCaMP2, 840 nM), improved baseline brightness, and expanded dynamic range. Using simultaneous patch-clamp recording and two-photon imaging in acute brain slices at physiologically relevant temperatures (33 ± 2˚C), we characterized the performance of these GECIs, and investigated which GECIs could be optimal for applications in each cell type.

MATERIALS AND METHODS
All experimental procedures were performed in accordance with the guidelines of the Animal Experiment Committee of the

EXPRESSION OF GECIs
cDNAs encoding YC2.60, YC3.60, YC-Nano15, and GCaMP3 ( Figure 1A) were subcloned into a cosmid vector carrying the cytomegalovirus enhancer and β-actin (CAG) promoter, woodchuck hepatitis virus post-transcriptional regulatory element Frontiers in Cellular Neuroscience www.frontiersin.org (WPRE), and bovine growth hormone (BGH) polyadenylation signal ( Figure 1B). Recombinant adenovirus was generated either by full-length DNA transfer method (Takara) or by COS-TPC method (Miyake et al., 1996) using HEK293 cells (kindly provided by the Cell Resource Center for Biomedical Research, Institute of Development, Aging and Cancer, Tohoku University). Viral clones were screened by restriction mapping of their genomic DNA, and appropriate clones were amplified and then purified by double cesium chloride step gradient ultracentrifugation (Kanegae et al., 1994). The titer of purified virus was measured by plaque forming assay with HEK293 cells. Purified viral solution (total 0.2 to 6 × 10 7 plaque forming units) was pressure-injected (IM-300, Narishige) into the lateral ventricle of ICR mice on embryonic day (E) 11 or E12 for expression in cerebellar Purkinje cells or E14 for cortical layer 2/3 pyramidal cells, respectively. In the following, a mouse virus-injected at E14 and sacrificed at P20, for instance, is described as "E14:P20."
The electrophysiological recording and two-photon imaging were synchronized by a trigger pulse generated upon laserscanning.

DATA ANALYSIS
After subtraction of dark noise on the photomultipliers, the baseline ratio of yellow to cyan fluorescence (R 0 , YCs) or baseline fluorescence (F 0 , GCaMP3) was calculated as the mean ratio or mean fluorescence, respectively, of the approximately 1-s window immediately before stimulus onset (baseline period). Subsequently, the fractional change of the ratio (ΔR/R 0 , YCs) or the fractional change of the fluorescence (ΔF/F 0 , GCaMP3) was calculated. Peak amplitude was calculated from ΔR/R 0 or ΔF/F 0 trace filtered with a 35-ms moving window and defined as the maximum value between the stimulus onset and 500 ms after the stimulus cessation. The signal-to-noise ratio (SNR) was calculated as the peak Frontiers in Cellular Neuroscience www.frontiersin.org amplitude divided by the baseline noise (the SD of the raw trace during the baseline period). Both peak amplitude and SNR were calculated from individual trials and averaged over three trials for each stimulus condition ( Table 2, 4, and 5). Responses were judged to be suprathreshold when SNR exceeds 2. Half rise time and half decay time were calculated from the 3-trial-averaged and filtered traces for 10 APs (pyramidal cells) or 5 complex spikes (CSs; Purkinje cells) only when responses were suprathreshold. Statistical difference was assessed using one-way ANOVA (p = 0.05) followed by Tukey's post-hoc tests unless otherwise noted. Data analysis was performed with AxoGraphX, Igor Pro 6 (WaveMetrics), NeuroMatic (http://www.neuromatic.thinkrandom.com/), Fluoview (Olympus), ImageJ (US National Institutes of Health), Excel (Microsoft), and GraphPad Prism4 (GraphPad software). All values are presented as mean ± SD unless otherwise noted.

ADENOVIRUS-MEDIATED EXPRESSION OF GECIs IN CORTICAL PYRAMIDAL CELLS AND CEREBELLAR PURKINJE CELLS
In order to express GECIs in the mouse brain, we performed in utero injection of adenoviral vectors Mikoshiba, 2003, 2004). This allows "neuronal birthday-specific" introduction of a foreign gene, since adenoviral infection is temporarily short (up to 4 h) and the adenoviral gene is transferred exclusively to the neuronally committed-daughter cells divided from stem cells on the ventricular surface of embryonic brain. We previously used LacZ-carrying adenovirus and demonstrated that injection at E14 led to expression in cortical superficial layer and that injection at E11 or E12 led to expression in cortical deep layer as well as cerebellar Purkinje cells. We performed immunohistochemical analysis to test if this specific expression pattern is reproducible with adenoviral vectors carrying GECIs (Figure 1). The injection of YC2.60-carrying adenovirus at E14 resulted in expression in the superficial layer of neocortex ( Figure 1C). The majority (97%; n = 119 of 123 cells) of expressing cells were immunopositive for CaMKII ( Figure 1D), indicating that they were pyramidal cells. The injection of YC2.60-carrying adenovirus at E11 or E12 resulted in expression in cortical deep layer (data not shown) as well as cerebellum ( Figure 1E). All (n = 76 of 76 cells) of the expressing cells in cerebellum were immunopositive for IP 3 R1, indicating that they were Purkinje cells. These results show that in utero injection of recombinant adenoviral vectors carrying GECIs could lead to their specific expression in cortical pyramidal cells as well as cerebellar Purkinje cells, successfully reproducing our previous results.

THE STABILITY OF GECI EXPRESSION AND RESPONSES
One of the major advantages of GECIs over synthetic Ca 2+ dyes should be its stable expression over time, allowing its application to tracking long-term plasticity of neuronal activity. It was previously demonstrated that the performance of GECIs could remain stable for weeks to months after expression in cortical pyramidal cells Tian et al., 2009;Horikawa et al., 2010). To test if this is also the case in Purkinje cells, we prepared acute cerebellar slices from YC2.60-expressing mice older than P100. The expression of YC2.60 was stable (Figure 4A), and its responses to a train of CSs were not significantly different from those in younger animals (P22-56; Figures 4B,C; Table 5; unpaired Student's t -test). These results confirm the idea that GECIs can be promising tools for chronic recording of neuronal activity.

DISCUSSION
We tested the performance of YC2.60, YC3.60, YC-Nano15, and GCaMP3 in mouse cortical pyramidal cells and cerebellar Purkinje cells. Our results suggest that (1) YC2.60 would be suitable for reliable detection of sparse firing of APs in cortical pyramidal cells; (2) GCaMP3 would be suitable for detecting burst firing of APs in pyramidal cells; and (3) YC2.60 as well as YC-Nano15 would be suitable for detecting CSs in cerebellar Purkinje cells. To our knowledge, this is the first study that quantitatively compares the performance of multiple GECIs in Purkinje cells, and thus should provide useful implications for the broader application of GECIs in mammalian CNS.

COMPARISON WITH PREVIOUS STUDIES USING THE SAME GECIs
In the present study, we performed the first quantitative characterization of the performance of YC2.60 in multiple mammalian neurons, and found that it exhibits good performance with little sign of signal saturation both in cortical pyramidal cells and cerebellar Purkinje cells.
In our previous work, YC-Nano15 showed much higher affinity than YC2.60 in Ca 2+ titration experiment using purified proteins (K d of YC-Nano15, 15 nM; K d of YC2.60, 95 nM; Horikawa et al., 2010). In the present study, we found (1) in pyramidal cells YC2.60 was as sensitive to single APs as YC-Nano15 and showed better responses to larger number of APs without signal saturation and (2) in Purkinje cells they showed comparable responses, YC-Nano15 showing slightly (but not significantly) better SNR over the entire stimulus range tested (1-10 CSs at 10 Hz). The fact that YC-Nano15 was still responsive to stimulation without being saturated by the resting [Ca 2+ ] i (∼53 nM in cortical pyramidal cells (Schiller et al., 1995) and ∼67 nM in Purkinje cells (Konnerth et al., 1992) implies that its affinity may decrease when expressed in neurons (see Hendel et al., 2008). YC-Nano15 still seems to have a higher affinity than YC2.60 in cortical pyramidal cells, whereas there seems to be little difference between YC-Nano15 and YC2.60 in Purkinje cells. This inconsistency between the two cell types will be discussed in the next section.
The relatively low reliability of single AP detection with YC3.60 and GCaMP3 in cortical L2/3 pyramidal cells contrasts with previous in vitro results (Tian et al., 2009;Lütcke et al., 2010), but is reminiscent of in vivo results in the same studies. This could be explained in part by the difference in the recording temperatures: ∼33˚C in our study vs. room temperature (22-24˚C) for their in vitro experiments. Higher temperature should make Ca 2+ transients smaller and faster, probably due to more active Ca 2+ extrusion mechanisms and narrower APs (Markram et al., 1995). These factors should in turn decrease responses of GECIs, as predicted and demonstrated by the same group (Hires et al., 2008;Mao et al., 2008). Indeed, we also found that SNR of GCaMP3 in

Frontiers in Cellular Neuroscience
www.frontiersin.org  response to single APs was larger at room temperature (2.2 ± 0.4, n = 4) than at 31-35˚C (1.3 ± 0.6, n = 7; p < 0.05, unpaired twotailed Student's t -test). Taken together, our results underscore the importance of appropriately designed in vitro experiments for accurate estimation of GECI performance in vivo.

DIFFERENCE IN GECI PERFORMANCE BETWEEN PYRAMIDAL CELLS AND PURKINJE CELLS
Previous studies using synthetic Ca 2+ dyes show that Ca 2+ transients generated by single APs in cortical pyramidal cells (262 ± 25 nM, Helmchen et al., 1996) and those by single CSs in Purkinje cells (∼150 nM, Wang et al., 2000) are comparable in vitro, and that both are detectable in vivo with comparably high fidelity (up to ∼97% for L2/3 pyramidal cells (Kerr et al., 2005) and ∼95% for Purkinje cells (Ozden et al., 2009). However, all the tested GECIs had a tendency to show remarkably smaller responses in Purkinje cells than in pyramidal cells, as is evident from the smaller SNR (for instance, SNR of YC2.60 in response to single pulse of stimulation were 4.3 ± 1.7 in pyramidal cells and 2.3 ± 0.8 in Purkinje cells, respectively) and the smaller percentage of cells with suprathreshold responses (for instance, YC2.60 showed suprathreshold responses to single pulse of stimulation in 89% of pyramidal cells and 59% of Purkinje cells, respectively). The mechanism responsible for this strikingly different performance of GECIs in these two cell types is unclear, but it may be accounted for, at least in part, by the much higher endogenous Ca 2+ buffering capacity in Purkinje cells (Fierro and Llano, 1996;Helmchen et al., 1996;Maeda et al., 1999), which presumably reflects the total activity of Ca 2+ -binding proteins. We speculate that a larger amount of Ca 2+ -binding proteins expressed in Purkinje cells might have decreased the performance of all the GECIs tested and also somehow masked the difference between YC2.60 and YC-Nano15. In line with the notion above that GECI performance could be dramatically altered in different cell types expressing different amount of Ca 2+ -binding proteins, it was previously reported that the performance of YCs containing wild type CaM could be interfered with CaM in a concentration-dependent manner (Miyawaki et al., 1999;Palmer et al., 2004Palmer et al., , 2006.

IMPLICATIONS FOR FUTURE IMPROVEMENT AND APPLICATION OF GECIs
YC2.60 and YC-Nano15 reliably responded to single APs and CSs, but their decay kinetics were slow, which could be a disadvantage for detecting individual events at frequency higher than 1 Hz (Horikawa et al., 2010). In contrast, YC3.60 and GCaMP3 showed faster signal decay, but they did not reliably detect small numbers of spikes. Resolving this tradeoff between sensitivity and on/off kinetics should be the focus of future improvement or development of GECIs. Nevertheless, YC2.60 and YC-Nano15 might be fast enough for detecting spontaneous AP firing of cortical L2/3 pyramidal cells as well as spontaneous CS firing of Purkinje cells in vivo, which are both known to occur at relatively low frequency, typically 1 Hz or less (Thach, 1968;Margrie et al., 2002). It would thus be interesting to apply these GECIs to monitoring long-term plasticity of spontaneous activity in the context of learning, development and disease.