Quantitative comparison of novel GCaMP-type genetically encoded Ca2+ indicators in mammalian neurons

New variants of GCaMP-type genetically encoded Ca2+ indicators (GECIs) have been continuously developed and heavily used in many areas of biology including neuroscience. The latest subfamily called “GECOs” were developed with in vitro high-throughput screening, and shown to have novel spectral properties and/or improved fluorescent responses over their ancestor GCaMP3. The most critical parameter in evaluating performance in neurons, however, remains uncharacterized: the relationship between the GECI responses and the number of action potentials (APs). Here we analyzed the GECI responses to APs in cortical pyramidal cells of mouse acute brain slices. Unexpectedly, we found that none of the GECOs exhibited any improved performance over GCaMP3. Our results imply that careful validation is required for the accurate prediction of the actual performance of GECIs in mammalian neurons. We propose that appropriate guidelines for evaluating their efficacy should be established for the benefit of research community, given the rapidly expanding use of GECIs in neuroscience.


INTRODUCTION
Genetically encoded Ca 2+ indicators (GECIs; for review: Mank and Griesbeck, 2008;Tian et al., 2012), or Ca 2+ -sensitive fluorescent proteins, are regarded as promising tools for many areas of biology including neuroscience. Since GECIs can in principle be stably expressed in a targeted type of cells, they emerged as critically important tools for analyzing long-term changes of in vivo multi-neuronal activity during learning, development and disease Huber et al., 2012).
Among several types of GECIs, GCaMP family (Nakai et al., 2001) has attracted intense attention in the field. It consists of circularly-permutated GFP, calmodulin (CaM), and M13 (Ca 2+ /CaM-binding peptide derived from skeletal muscle myosin light chain kinase), and changes its fluorescence intensity in response to [Ca 2+ ] i changes. The first prototypes suffered from poor expression at mammalian physiological temperature (Nakai et al., 2001;Ohkura et al., 2005), but this problem was overcome by subsequent mutagenesis, resulting in GCaMP2 (Tallini et al., 2006). GCaMP2 was successfully used to monitor activation of cerebellar parallel fibers (Díez-García et al., 2005 and vomeronasal neurons (He et al., 2008), yet it turned out to be mostly insensitive to [Ca 2+ ] i changes caused by single or small number of action potentials (APs) (Mao et al., 2008). Taking advantage of the crystal structure information described (Wang et al., 2008;Akerboom et al., 2009), GCaMP2 was mutagenized into GCaMP3 to show higher baseline fluorescence and larger dynamic range . Although GCaMP3 has been widely used for recording in vivo activity of mammalian neurons (Dombeck et al., 2010;Huber et al., 2012;Keller et al., 2012), its detection reliability of single APs is still relatively low under physiological conditions Yamada et al., 2011). This issue makes it difficult to relate the fluorescent responses with neuronal activity, and thus should be overcome in the next generation of GCaMP.
The most recent subfamily of GCaMP was developed with random mutagenesis of GCaMP3 accompanied by high-throughput in vitro screening (Zhao et al., 2011). The best variants with improved performance and/or novel spectral properties were efficiently selected from >10 5 clones, and termed genetically encoded Ca 2+ indicators for optical imaging (GECOs). The superior functionality of GECOs was confirmed by Ca 2+ titration with purified proteins and live imaging of cancer cell line. All of the three variants of green GECOs (G-GECOs) with different dissociation constants (K d ) for Ca 2+ (G-GECO1.0, 750 nM; G-GECO1.1, 620 nM; and G-GECO1.2, 1150 nM) showed twice as large dynamic range as GCaMP3 (K d : 540 nM). Red-shifted GECO (R-GECO1; K d : 480 nM) showed dynamic range similar to GCaMP3. Blue-green emission ratiometric GECO (GEM-GECO1; K d : 340 nM) showed the largest dynamic range, 6-9-fold larger than GCaMP3. Furthermore, G-GECOs were shown to be superior to, and R-GECO1 comparable to, GCaMP3 in detecting spontaneous activity of cultured rat hippocampal neurons, and GEM-GECO1 was shown to be functional in sensory neurons of C. elegans.
Nevertheless, the most critical parameter of GECOs for evaluating their performance in neurons remains poorly characterized: the relationship between fluorescent responses and the number of APs. This leaves open the possibility that the GECO-expressing neurons happened to be more actively firing than GCaMP3-expressing neurons, resulting in apparently improved performance. In addition, most data in the original study were acquired in a culture system, where GECIs sometimes show larger responses compared to non-culture or in vivo systems . It therefore remains unclear whether GECOs can indeed show improved performance in mammalian neurons.
To address these issues, we analyzed the GECI responses to APs in cortical pyramidal cells of mouse acute brain slices by simultaneous two-photon imaging and patch-clamp recording, and investigated whether GECOs would indeed show better responses than their ancestor GCaMP3.

MATERIALS AND METHODS
All experimental procedures were performed in accordance with the guidelines of the Animal Experiment Committee of the RIKEN Brain Science Institute.
For experiments with GECIs, image acquisition began typically after 2 min of break-in and terminated within 30 min given washout of GECIs Mao et al., 2008). For experiments with OGB-1, image acquisition began after 15 min of break-in for equilibration of the dyes.
The electrophysiological recording and 2-photon imaging were synchronized by a trigger pulse generated upon laser scanning.

DATA ANALYSIS
After subtraction of dark noise on the photomultipliers, the mean baseline fluorescence (F 0 , GCaMP3, G-GECOs, R-GECO1, and OGB-1) or the mean baseline ratio of blue to green fluorescence (R 0 , GEM-GECO1) was calculated as the mean fluorescence or the mean ratio, respectively, of the approximately 1 s window immediately before stimulus onset (baseline period). Subsequently, the fractional change of the fluorescence ( F/F 0 , GCaMP3, G-GECOs, R-GECO1, and OGB-1) or the fractional change of the ratio ( R/R 0 , GEM-GECO1) was calculated. To facilitate comparison across GECIs with different baseline noise level, the signal-to-noise ratio (SNR) was calculated as F/F 0 or R/R 0 divided by the baseline standard deviation. Peak SNR was calculated from SNR trace filtered with a 35 ms moving window and defined as the maximum value between the stimulus onset and 500 ms after the stimulus

THE PERFORMANCE OF GCaMP3 AND GECOs IN CORTICAL LAYER 2/3 PYRAMIDAL CELLS
We characterized the performance of GECIs expressed in the cortical layer 2/3 pyramidal cells by simultaneous 2-photon imaging and whole-cell patch-clamp recording in acute brain slice preparations. Overall, the expression of GECIs did not have significant effects on the electrophysiological properties of pyramidal cells, except G-GECO1.0 and GEM-GECO1, the expression of which seemed to result in higher threshold and broader half width of APs (Table 2).

COMPARISON WITH OGB-1
In order to clarify the factors to be improved in the next generation of GCaMP, we quantified the performance of OGB-1, one of the most commonly used synthetic dyes for in vivo imaging. We loaded OGB-1 through recording patch pipettes at 20 μM, which is close to the concentration obtained by the bolus loading technique (Stosiek et al., 2003). Consistent with previous studies (Waters et al., 2003;Kerr et al., 2005), OGB-1 reliably detected single APs (Figures 2, 3 and Table 4). Responses of OGB-1 to 1, 2, and 5 APs at 20 Hz were significantly larger than those of Table 2

DISCUSSION
In the present study, we found that the latest variants of GCaMP or GECOs did not exhibit any improved responses over their ancestor GCaMP3: G-GECOs had lower baseline fluorescence, similar or smaller dynamic range, and slower rise and decay kinetics; R-GECO1 showed expression invading the nucleus and punctate patterns in the cytosol, and had much smaller dynamic range; GEM-GECO1 also had much smaller dynamic range.

VALIDITY OF THE TECHNIQUES USED IN THE PRESENT STUDY
The present study was designed to be optimal for efficient comparison across many GECI constructs as well as to be extendable to future in vivo experiments.  Ideally, the characterization of all the GECIs should be performed in vivo, yet this is not practical for comparison across many constructs. We believe that characterization of GECIs in acute brain slices at physiological temperature should be a reasonable compromise, as it not only gives a good yield of data but also seems to predict in vivo GECI performance more reliably than other in vitro preparations ; also see the next section).
Compared to other techniques applicable for acute brain slice preparation (virus and transgenic animals), in utero electroporation should be preferable for early screening of many constructs, as it is faster and less laborious, only requiring a purified plasmid and pregnant mice for each new construct. One concern may be that Ca 2+ buffering by GECI expression during the development might perturb the properties of neurons, but the electrophysiological parameters of GECI-expressing neurons were overall not significantly different from those of wild-type cells. It would be interesting to test in the future whether (1) the functionality of GECIs remain comparable when they are expressed for a longer period of time Tian et al., 2009;Yamada et al., 2011) and (2) different transfection methods (virus and transgenic animals) can result in different GECI performance, especially when more promising GECIs are developed and validated.
Imaging in this study was performed exclusively at the proximal apical dendritic segments as previously described Mao et al., 2008;Tian et al., 2009;Yamada et al., 2011), yet the future experiments should be preferably performed at the soma, where most in vivo multi-cell imaging are performed. Since AP-associated Ca 2+ transients are generally smaller in the soma than in the proximal apical dendrites (Schiller et al., 1995), the apparent performance of GECIs would be expected to be lower, presumably making the criteria for GECI selection even more stringent.

COMPARISON WITH PREVIOUS STUDIES USING THE SAME GECIs
Like many other variants of GCaMP generated by other groups (Souslova et al., 2007;Muto et al., 2011), the relationship between fluorescent responses of GECOs and the number of APs was poorly characterized in the original study, where it was claimed that GECOs showed larger dynamic range compared to GCaMP3 (Zhao et al., 2011). The source of inconsistencies between the original results and ours is currently unknown, yet they might be attributed to the difference in preparation (culture in the original study vs. acute slice in ours) as well as whether or not imaging was accompanied with electrophysiology. Given that GCaMP3 showed much larger single AP responses in culture ( F/F 0 : 46 ± 4.2%) than in acute slice preparation (14 ± 2.7%) or in vivo (7.9 ± 2.8%) , we believe that the experimental design used here should be better suited to assess the actual performance of GECIs in mammalian neurons.

IMPLICATIONS FOR FUTURE IMPROVEMENT OF GCaMP AND OTHER GECIs
As indicated from the comparison with OGB-1, one of the obvious goals for the next generation of GCaMP is reliable detection of single APs. For successful improvement, appropriate guidelines for evaluating the efficacy of GECIs should be established and accepted in the research community (Hires et al., 2008). We propose that the following implications drawn from our study should be taken into consideration: (1) Properties of GECIs (dynamic range, affinity, kinetics, etc.) measured with purified protein often show limited correlation with the performance in neurons (e.g., Purified protein of GEM-GECO1 has a larger dynamic range and a higher affinity compared to GCaMP3, but performs far worse in neurons; see similar reports in Hendel et al., 2008); (2) Screening of optimal GECIs for mammalian neurons should be finalized with combined imaging and electrophysiology in a non-culture system; and (3) GECI performance can be strikingly different from one species to another (e.g., GEM-GECO1 showed limited responsiveness in mouse cortical neurons, but good functionality in C. elegans sensory neurons).
We also propose that factors responsible for the reduced GECI performance in mammalian neurons in physiological preparations should be identified and overcome; these might include interaction of GECIs with endogenous CaM (Miyawaki et al., 1999;Palmer et al., 2006) or with other as yet unknown binding proteins, or direct modifications of GECI protein such as phosphorylation. Some useful clues might be obtained by biochemical comparison of cell lysates containing GECI protein from different cell types (e.g., HeLa cells, where GECOs show high performance vs. cortical pyramidal cells, where GECOs show reduced performance) or from the same type of neurons in different preparations.
We hope that our findings will alert the research community to the limitations of current GECIs, stimulate future development and screening of a "holy-grail" GECI with high sensitivity and fast kinetics, and facilitate appropriate selection of optimal GECIs for different experimental requirements.