Characterization and Subcellular Targeting of GCaMP- Type Genetically-Encoded Calcium Indicators Tianyi Mao1,2., Daniel H. O���Connor1,2., Volker Scheuss1,2��, Junichi Nakai3, Karel Svoboda1,2* 1 Cold Spring Harbor Laboratory, Cold Spring Harbor, New York, United States of America, 2 Howard Hughes Medical Institute, Janelia Farm Research Campus, Ashburn, Virginia, United States of America, 3 Laboratory for Memory and Learning, RIKEN Brain Science Institute, Wako-shi, Saitama, Japan Abstract Genetically-encoded calcium indicators (GECIs) hold the promise of monitoring [Ca2+] in selected populations of neurons and in specific cellular compartments. Relating GECI fluorescence to neuronal activity requires quantitative characterization. We have characterized a promising new genetically-encoded calcium indicator���GCaMP2���in mammalian pyramidal neurons. Fluorescence changes in response to single action potentials (17610% DF/F [mean6SD]) could be detected in some, but not all, neurons. Trains of high-frequency action potentials yielded robust responses (302650% for trains of 40 action potentials at 83 Hz). Responses were similar in acute brain slices from in utero electroporated mice, indicating that long-term expression did not interfere with GCaMP2 function. Membrane-targeted versions of GCaMP2 did not yield larger signals than their non-targeted counterparts. We further targeted GCaMP2 to dendritic spines to monitor Ca2+ accumulations evoked by activation of synaptic NMDA receptors. We observed robust DF/F responses (range: 37%���264%) to single spine uncaging stimuli that were correlated with NMDA receptor currents measured through a somatic patch pipette. One major drawback of GCaMP2 was its low baseline fluorescence. Our results show that GCaMP2 is improved from the previous versions of GCaMP and may be suited to detect bursts of high-frequency action potentials and synaptic currents in vivo. Citation: Mao T, O���Connor DH, Scheuss V, Nakai J, Svoboda K (2008) Characterization and Subcellular Targeting of GCaMP-Type Genetically-Encoded Calcium Indicators. PLoS ONE 3(3): e1796. doi:10.1371/journal.pone.0001796 Editor: Ernest Greene, University of Southern California, United States of America Received December 19, 2007 Accepted February 13, 2008 Published March 19, 2008 Copyright: �� 2008 Mao et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Funding: HHMI, NIH. Funders had no role in the design and conduct, analysis or reporting of the study. Competing Interests: The authors have declared that no competing interests exist. * E-mail: svobodak@janelia.hhmi.org �� Current address: Department of Cellular and Systems Neurobiology, Max Planck Institute for Neurobiology, Martinsried, Germany . These authors contributed equally to this work. Introduction Understanding the function of neural networks will require the ability to monitor action potentials and synaptic activity in populations of identified neurons. In mammalian pyramidal neurons, action potentials trigger a transient calcium influx though voltage-gated calcium channels that can occur both at the soma and in the dendrites following backpropagation of the action potential [1���4]. Action potential (AP)-evoked calcium transients have been used extensively to measure neuronal spiking activity in vitro and in vivo [5���9]. In addition, NMDA receptor-dependent calcium accumulation in dendritic spines has been used to monitor the activity of individual synapses [10���16]. The vast majority of calcium imaging experiments have employed synthetic calcium indicators, which permit measure- ments of AP- and synaptically-evoked calcium transients. Howev- er, genetically-encoded calcium indicators (GECIs) provide advantages over synthetic indicators [17]. They allow: (1) monitoring activity among genetically-defined subsets of cells, (2) measuring calcium dynamics in specific subcellular compartments, and (3) long-term calcium imaging in vivo. GECIs are engineered based on either changes in the florescence intensity of a single fluorophore, or changes in fluorescence resonance energy transfer (FRET) efficiency. For example, the GCaMP family of GECIs is composed of a single circularly permuted GFP with calmodulin (CaM) and its binding peptide myosin light-chain kinase M13 linked to its C- and N- termini, respectively. Upon calcium binding, conformational changes in the CaM/M13 complex cause a fluorescence change in the circularly permuted GFP-based fluorophore [18]. FRET- based GECIs are based on two designs. In the cameleon family [19], a calcium-dependent increase in FRET between a CFP and YFP FRET pair is coupled by the binding of calmodulin to the M13 peptide. The troponin family of sensors utilizes the skeletal muscle calcium sensor troponin C (TnC). Binding of calcium to troponin causes a conformational change that increases FRET between CFP and YFP [20]. Since endogenous TnC, unlike calmodulin, is not expressed in neurons, TnC-based sensors may show reduced interference with endogenous signal transduction processes in neurons [21]. Recently developed GECIs have provided improved brightness, dynamic range, speed, pH- and Mg2+- sensitivity, thermal stability and folding efficiency [17,21��� 25]. Several lines of mammalian GECI transgenic animals have been engineered [23���26], but the small signal levels in these mice [25���27] have so far not permitted widespread use for in vivo physiology. Better results have been achieved in invertebrate systems [28���30]. To understand the advantages and limitations of each GECI for measuring neuronal activity a quantitative comparison of GECI signals under identical experimental conditions is required. In pilot PLoS ONE | www.plosone.org 1 2008 | Volume 3 | Issue 3 | e1796

studies we screened through several members of the latest generation of GECIs and identified GCaMP2 [24], the latest member of the GCaMP family, as particularly promising. We evaluated several versions of GCaMP2 (Figure 1), focusing on its suitability for monitoring action potentials and NMDA-R activation in single spines in mammalian pyramidal neurons. We found that GCaMP2, compared to its predecessors, displayed improved fluorescence change in response to action potential trains and in addition showed robust responses to two-photon glutamate uncaging stimuli in dendritic spines. However, our studies also reveal significant limitations of GCaMP2 for monitoring neural activity in vivo. Results Responses of GCaMP-type GECIs to action potential trains We made whole���cell recordings from GCaMP-expressing cultured hippocampal pyramidal neurons [31,32] and in acute cortical brain slices at room temperature. Under baseline conditions GCaMP fluorescence was very low. For example, it was often difficult to image small dendritic branches and to detect dendritic spines. Action potentials were evoked by short current injections (3���5.5 nA, 2 ms). Our basic experiment comprised measuring GECI responses to high-frequency (83 Hz) action potential trains (Figure 2). Under our experimental conditions the peak Ca2+ accumulations are approximately proportional to action potential frequency [5,32,33]. We acquired linescans from the proximal apical dendrite (within 50 mm of the soma) (Figure 2). In cultured hippocampal neurons transfected with GCaMP2 and the cytoplasmic red protein mCherry [34] single action potentials caused clear fluorescence increases in some, but not all, neurons (Figure 3A). The average response to single action potentials was small (17610% [mean6SD] DF/F across n = 13 cells). A train of 40 actions potentials (APs) at 83 Hz gave a robust response of 302650% DF/F (n = 12 cells), close to GCaMP2���s dynamic range measured in cuvettes [24]. GCaMP2 responses in layer 2/3 pyramidal cells in acute cortical brain slices (postnatal day 14���21, see Materials and Methods) were similar (1 AP response, 13617% DF/F, n = 8 40 AP response, 248651% DF/F, n = 8) (Figure 3B) to the responses measured in cultured neurons. The recorded cells had healthy input resistances and resting potentials (see Materials and Methods) and apparently normal morphology. Thus, even though GCaMP2 was expressed at high concentrations for up to 4 weeks, the similar DF/F responses suggest that endogenous calmodulin did not interfere with the function of the calmodulin-based GCaMP2. Furthermore, GCaMP2 did not appear to degrade the health of the transfected neurons. We next measured GCaMP2 responses near physiological temperature (34.5���35.5u). Consistent with faster calcium extrusion [35] and a narrower action potential, GCaMP2 responses were smaller (1 AP response, 668% DF/F, n = 10 40 AP response, 134648% DF/F, n = 10) (Figure 3C). GCaMP2 responses were also much faster (Figure 3C room temperature: rise T1/2: 95615 ms decay T1/2: 4836127 ms, n = 13 cells near-physio- logical temperature: rise T1/2: 73615 ms decay T1/2: 134639 ms, n = 10 cells all measurements for the 10 AP stimulus). The decay time of the GCaMP2 fluorescence transient is ,2 fold slower than the decay time of [Ca2+] accumulations [35]. These values are in general agreement with GCaMP2 response kinetics measured in cerebellar granule cells in vivo following electrical stimulation [36]. We also tested GCaMP1.6 [37] (see also [32]) and GCaMP1.6- CaM(E140K) . The E140K mutation is located in a calcium binding site and has been shown to increase the brightness of the sensor and decrease the affinity of the sensor for calcium [37]. GCaMP1.6 (Figure 3D) gave much smaller response amplitudes than GCaMP2 (1 AP, 464% DF/F, n = 5 40 AP, 155628% DF/ F, n = 5) at room temperature. Single action potentials did not Figure 1. Domain structures of the GCaMP-family of genetically encoded calcium indicators (GECIs) and fusion constructs. A, Domain comparisons of GCaMP2 and GCaMP1.6 red labels indicate the differences. B, Constructs for subcellular targeting of the GECIs. doi:10.1371/journal.pone.0001796.g001 Targeted Calcium Indicators PLoS ONE | www.plosone.org 2 2008 | Volume 3 | Issue 3 | e1796