Genetic visualization with an improved GCaMP calciumindicator reveals spatiotemporal activation of thespinal motor neurons in zebrafishAkira Mutoa, Masamichi Ohkurab, Tomoya Kotania,1, Shin-ichi Higashijimac, Junichi Nakaib,2, and Koichi Kawakamia,d,2

aDivision of Molecular and Developmental Biology, National Institute of Genetics, and dDepartment of Genetics, Graduate University for Advanced Studies(SOKENDAI), Mishima, Shizuoka 411-8540, Japan; bBrain Science Institute, Saitama University, Sakura-ku, Saitama 338-8570, Japan; and cNational Institutes ofNatural Sciences, Okazaki Institute for Integrative Bioscience, Okazaki, Aichi 444-8787, Japan

Edited by Thomas M. Jessell, Columbia University College of Physicians and Surgeons, New York, NY, and approved February 8, 2011 (received for reviewJanuary 23, 2010)

Animal behaviors are generated by well-coordinated activation ofneural circuits. In zebrafish, embryos start to show spontaneousmuscle contractions at 17 to 19 h postfertilization. To visualize howmotor circuits in the spinal cord are activated during this behavior,we developed GCaMP-HS (GCaMP-hyper sensitive), an improvedversion of the genetically encoded calcium indicator GCaMP, andcreated transgenic zebrafish carrying the GCaMP-HS gene down-stream of the Gal4-recognition sequence, UAS (upstream activationsequence). Thenweperformed a gene-trap screen and identified theSAIGFF213A transgenic fish that expressed Gal4FF, a modified ver-sion of Gal4, in a subset of spinal neurons including the caudal pri-mary (CaP) motor neurons.We conducted calcium imaging using theSAIGFF213A; UAS:GCaMP-HS double transgenic embryos during thespontaneous contractions. We demonstrated periodic and synchro-nized activation of a set of ipsilateral motor neurons located on theright and left trunk in accordance with actual muscle movements.The synchronized activation of contralateral motor neurons oc-curred alternately with a regular interval. Furthermore, a detailedanalysis revealed rostral-to-caudal propagation of activation of theipsilateral motor neuron, which is similar to but much slower thanthe rostrocaudal delay observed during swimming in later stages.Our study thus demonstrated coordinated activities of the motorneurons during the first behavior in a vertebrate. We propose theGCaMP technology combinedwith the Gal4FF-UAS system is a pow-erful tool to study functional neural circuits in zebrafish.

neuronal activity | calcium transient | transgenesis | Tol2 | gene trapping

Vertebrate behaviors are generated by coordinated activationof neural networks. The zebrafish provides an excellent sys-

tem to study the neural activities of vertebrates because of itstransparency and external development. It has been reported thatzebrafish embryos show a first locomotive activity as early as 17 hpostfertilization (hpf), which is called spontaneous contractionor coiling; namely, they spontaneously contract the left and righttrunk muscles alternatively without any external stimulus. Thisbehavior precedes touch-evoked escape response and swimming,which emerge in later developmental stages (1). The neuronalactivities during the spontaneous contractions were analyzed byelectrophysiological approaches, and periodic and synchronizeddepolarization of the spinal neurons has been observed (2, 3).However, it is still challenging to analyze the activity of a neuralnetwork by electrophysiology because it requires recording frommultiple neurons by using multiple electrodes. To understandhow functional neural circuits are formed and operated, a newtechnology to monitor the activity of multiple neurons is desired.An alternative approach to record activities from multiple

neurons is to detect Ca2+ influx associated with generation of ac-tion potentials. For this purpose, fluorescent calcium-indicatordyes, which should be introduced into cells by microinjection or bybulk-loading using their acetoxymethyl ester forms (4, 5), or ge-netically encoded calcium indicators (GECIs) have been used

(6, 7). Previously we created a fusion of the calmodulin-bindingdomain from themyosin light-chain kinase (which is called anM13peptide), permutated EGFP, and the calmodulin, and named itGCaMP (7). GCaMP increases its intensity of green fluorescenceupon binding to calcium ions. Since the first version was created,efforts have been made to generate brighter and more sensitiveversions of GCaMP by introducing amino acid substitutions (8–10).In this study, we aim to visualize a motor network during the

first behavior in zebrafish. For this purpose, we used the GCaMPtechnology and the Gal4FF-UAS (upstream activator sequences)system, which also was described recently in zebrafish (11). First,we constructed an improved version of GCaMP, GCaMP-HS(GCaMP-hyper sensitive), which is brighter at the resting leveland more sensitive to the change of cellular Ca2+ concentrationsthan the previous version. Second, to express GCaMP-HS underthe control of the Gal4FF transcription activator, we constructedan effector fish line that carried the GCaMP-HS gene down-stream of the Gal4 recognition sequence UAS. Third, we per-formed gene-trap screens and isolated a driver line that expressedGal4FF in a specific subset of motor neurons. Here we demon-strate successful imaging of the activity of the motor neuronsduring the spontaneous contractions in the vertebrate embryos.

ResultsDevelopment of GCaMP-HS. Previously, it was shown that amino acidsubstitutions in GFP could enhance the folding activity and therebyincrease itsfluorescence. Thismore recent versionofGFPwas called“superfolder GFP” (12). To develop a better GECI, we tested if thesubstitutions described in the superfolderGFP can improveGCaMPin terms of the brightness and the signal amplitude. We introducedfour of the six superfolder mutations, S30R, Y39N, N105T in theN-terminus half of GFP, and A206V in the C-terminus half of GFP,which were shown to have more effects in folding (12), into theprevious version of GCaMP, GCaMP2 (9), and named the newversion of GCaMP GCaMP-HS (Fig. 1A).First, we purified the GCaMP-HS protein produced in bac-

teria and characterized it in vitro. Spectrum analysis of GCaMP-HS showed maximal excitation at 488 nm and maximal emissionat 509 nm in the presence of 300 nM Ca2+, which were similar to

Author contributions: A.M., M.O., J.N., and K.K. designed research; A.M., M.O., T.K.,S.-i.H., J.N., and K.K. performed research; A.M., M.O., T.K., J.N., and K.K. contributednew reagents/analytic tools; A.M., M.O., S.-i.H., J.N., and K.K. analyzed data; and A.M.,M.O., J.N., and K.K. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1Present address: Laboratory of Reproductive and Developmental Biology, Faculty ofAdvanced Life Science, Hokkaido University, Sapporo 060-0810, Japan.

2To whom correspondence may be addressed. E-mail: kokawaka@lab.nig.ac.jp or jnakai@mail.saitama-u.ac.jp.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1000887108/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1000887108 PNAS | March 29, 2011 | vol. 108 | no. 13 | 5425–5430

NEU

ROSC

IENCE

those of GCaMP2 (Fig. 1B). To test the effect of the superfoldermutations on the folding activity, we analyzed restoration of thefluorescence after heat-denaturation of the GCaMP proteins.The GCaMP-HS showed higher efficiencies of restoration (Fig.1C), indicating that the substitutions introduced in GCaMP-HSincreased the refolding activity. Then we measured their fluo-rescence intensities at various Ca2+ concentrations. From thisCa2+ titration assay, we calculated that Kd and Hill coefficient ofGCaMP-HS were 102 nM and 5.0, respectively (Fig. 1D). Kd andHill coefficient of GCaMP2 were 146 nM and 3.8, respectively,indicating that GCaMP-HS has a higher affinity to Ca2+ ions anda higher cooperativity in binding to Ca2+. The actual values of(Fmax − Fmin)/Fmin (maximal fluorescence minus minimal fluo-rescence per minimal fluorescence) were comparable: 4.1 inGCaMP-HS and 3.9 in GCaMP2. These findings suggested thatGCaMP-HS may be used to detect Ca2+ transient where Ca2+

concentrations are low.Second, we tested the cellular performance of GCaMP-HS.

The expression plasmids pN1-GCaMP-HS and pN1-GCaMP2were used to transfect HEK293 cells. The cells transfected withpN1-GCaMP-HS showed ∼1.6-fold higher fluorescence at theresting levels (Fig. 1 E and F), and 1.4-fold intensity changes whenactivated with 100 μM carbachol (Fig. 1G). We analyzed thecellular protein levels by Western blotting and found that thelevels of the GCaMP-HS protein were constantly 1.7-fold higherthan those of the GCaMP2 protein (Fig. 1H), suggesting that thehigher folding activity increased the cellular expression level, andhence increased the basal fluorescence of GCaMP-HS.

Construction of the UAS:GCaMP-HS Transgenic Zebrafish. We aimedto apply the new GCaMP-HS to in vivo imaging of neuronal ac-tivities in zebrafish. To take advantage of the Gal4FF-UAS systemwe described previously (11), we constructed the T2RUASG-CaMP-HS plasmid that carried the GCaMP-HS gene downstreamof the upstream activation sequence (UAS). The T2RUASG-CaMP-HS plasmid was injected with the transposase mRNAto fertilized eggs, and the injected fish were crossed with thehspGGFF27A fish that expressed the Gal4FF-GFP fusion proteinin the central nervous system or with the SAGFF73A fish thatexpressed Gal4FF ubiquitously (11, 13). All of four injected fishused for the mating produced transgenic F1 embryos that carriedthe T2RUASGCaMP-HS transgene in their offspring. We ana-lyzed theF1fish by Southern blot hybridization and determined thenumber of insertions carried by these fish. Then we crossed the F1fish further with wild-type fish, and finally identified transgenic fish

that carried a single transposon insertion and showed the brightestfluorescence when crossed with the Gal4FF-expressing fish lines.We named the fish line and the insertionUAS:GCaMPHS4A. TheSAGFF73A;UAS:GCaMPHS4A double-transgenic embryo wasembedded in agarose at 5 d postfertilization (dpf) and observedunder a fluorescence microscope. We could detect the fluores-cence changes in the trunk muscles in accordance with twitch-ing (Movie S1). Therefore, the UAS:GCaMPHS4A transgenicfish was used for further studies.

Identification of a Driver That Expressed Gal4FF in a Subset of thePrimary Motor Neurons in the Spinal Cord. Spontaneous contractionis the first behavior of a zebrafish embryo. It can be observed asearly as 17 hpf, reaches the maximum frequency at 19 hpf, andgradually declines by 26 hpf (1). We aimed to visualize the activityof themotor circuit during the spontaneous contraction. To isolatea driver line that expresses Gal4FF in the primary motor neurons,we performed genetic screens using the Tol2-based gene-trapconstructs T2KSAGFF and T2TSAIGFF. We observed offspringfrom ∼500 fish injected with these constructs and identified the

A

B C D

E

F G HGCaMP2

GCaMP2 GCaMPHS

GCaMPHS

Fig. 1. Construction and analysis of GCaMP-HS. (A) The structure of GCaMP-HS. The positions of the amino acid substitutions are indicated. (B) Excitation(blue) and emission (green) spectra of GCaMP-HS in Ca2+-chelated (20 mM BAPTA; dotted lines) and saturated (300 nM Ca2+; solid lines) solutions. Excitationand emission are maximum at 488 nm and 509 nm, respectively. (C) Restoration of the fluorescence of GCaMP-HS (solid line) and GCaMP2 (dotted line) afterheat-denaturation. (D) Titration assay with defined Ca2+concentrations. Comparison of GCaMP-HS (solid line) and GCaMP2 (dotted line). Means and SD areshown (n = 3). (E) GFP fluorescence in HEK cells transfected with pN1-GCaMP2 (Left) and pN1-GCaMP-HS (Right). (Scale bar, 30 μm.) (F) The basal fluorescenceintensity in HEK cells transfected with the GCaMP2 and GCaMP-HS plasmids. (G) The fluorescence changes upon addition of 100 μM carbachol to the GCaMP2-and GCaMP-HS–transfected cells. Means and SD from three independent transfection experiments. (H) Western blot analysis using an anticalmodulin anti-body in three independent transfection experiments. Protein levels are increased in the GCaMP-HS–transfected cells.

A

*

C

B

Fig. 2. GFP expression in the SAIGFF213A;UAS:GFP double transgenic em-bryos. (A) A side view at 24 hpf. GFP is expressed in the central nervoussystem. (Scale bar, 100 μm.) (B) A side view of the spinal cord at 24 hpf ob-served with a confocal microscope. Arrowheads indicate CaP motor neurons.In some cases, two CaP neurons are found in one spinal segment (asterisks).(Scale bar, 100 μm.) (C) The soma (arrowhead) and axon of the GFP-expressing single CaP neuron at 24 hpf. The CaP neurons project their axonsto the ventral muscle (arrow). Dotted line shows the boundary betweendorsal and ventral muscles. (Scale bar, 50 μm.)

5426 | www.pnas.org/cgi/doi/10.1073/pnas.1000887108 Muto et al.

SAIGFF213A fish. In the SAIGFF213A;UAS:GFP double-transgenic embryos, GFP was expressed rather ubiquitously in thecentral nervous system at 24 hpf (Fig. 2A). In the spinal cord,several different types of neurons, includingRohon-Beard neurons,subpopulations of interneurons, and a set of neurons that were lo-cated in the ventral spinal cord, one per each spinal segment, thatextended their axons to the ventral muscles were identifiable (Fig. 2B and C). The zebrafish embryos have three types of the primarymotor neurons in the spinal cord, middle primary neurons (MiP),rostral primary neurons (RoP), and caudal primary neurons (CaP).Among them, only theCaPmotor neurons extend their axons to theventral muscles (14). From the location and the morphology, weconcluded that the SAIGFF213A;UAS:GFP expressed GFP in theCaP motor neurons, and MiP and RoP, which should be detectedadjacent to CaP, did not express GFP at the detectable levels. Insome cases, two CaP neurons were observed in one segment, asreported previously (15) (Fig. 2B). The GFP expressing CaPmotorneurons were detectable as early as 17 hpf (the 16-somites stage)and observable at 2 dpf, and then the expression was gradually de-creased and hardly detected at 7 dpf, although GFP expression inthe other neurons were still observable (Fig. S1). We analyzed theTol2 insertion in the SAIGFF213A line and found that the insertionwas mapped close to the prdm14 gene on the chromosome 24.

Spontaneous Contractions and Detection of the Activity of the CaPPrimary Motor Neurons.Weobserved the spontaneous contractionsin our laboratory conditions (∼28 °C). The spontaneous con-tractions started around 19 hpf, showed the maximum frequency(0.6 Hz) at 21 hpf, and gradually ceased by 27 hpf (Fig. 3A andMovie S2). The 2-h delay of the start time in comparison with theprevious work (1) may be a result of slightly lower temperatures inour laboratory condition or to a difference in the fish line we used.To analyze how motor circuits are activated during this behavior,we crossed the UAS:GCaMPHS4A fish with the SAIGFF213Afish and created double-transgenic embryos. Then we carried outimaging of the Ca2+ transient in the soma of the CaP neuronsbetween 21 and 27 hpf. We found that the maximal changes of thefluorescence increased from 53% (21 hpf) to 400% (27 hpf) as themotor neurons matured; the basal fluorescence was nearly con-stant (Fig. 3B).First, we analyzed the activity of the CaP neurons at early de-

velopmental stages. At 18.5 hpf, an embryo contracted its trunkmuscles at a frequency less than 0. 1 Hz (Fig. 3A). We performedimaging of the SAIGFF213A;UAS:GCaMPHS4A double trans-genic embryo embedded in agarose at 18.5 hpf (Fig. 3C andMovieS3). We used axons of the CaP neurons as ROIs (regions of in-terest) because the fluorescence changes were obvious in theaxons but were hardly detected in the soma at this stage. Periodicfluorescence changes at a frequency of 0.03 Hz were detected inabout half of single CaP neurons observed (ROI-2 and ROI-4 inFig. 3C andD). The fluorescence changes in the other half of CaPneurons were rarely detected (ROI-1 and ROI-3 in Fig. 3 C andD). We found that the periodic fluorescence changes in differentipsilateral CaP neurons were synchronized at the temporal reso-lution of 10 fps (frames per second) (ROI-2 and ROI-4 in Fig.3D). Moreover, nonperiodic activation of a CaP neuron also oc-curred simultaneously with activation of the other synchronizedCaP neurons (ROI-1 in Fig. 3D), suggesting these neurons arephysically connected or receiving common inputs at this stage. Itshould be noted that actual muscle contractions were not asso-ciated with 46% (18 of 39) of the cases where the fluorescenceincreases were detected in 10 different CaP neurons. Thesefindings indicate that the motor network to generate periodic andsynchronized activation starts to form before the muscle con-tractions are fully coupled with the activity of the motor network.Second, we carried out imaging at 10 fps using the same

SAIGFF213A;UAS:GCaMPHS4A double-transgenic embryo at24 hpf. The periodic synchronized fluorescence changes in the

ipsilateral CaP neurons were detected in the embryos embedded inagarose with a frequency of 0.41± 0.24Hz (n=20 embryos) (ROI-1 and -2 in Fig. 3 E and F, and Movie S4). It should be noted thatthe muscle contractions occurred concomitantly with the fluores-cence increases of GCaMP-HS at this stage. Occasionally, twoconsecutive contractions occurred at the same side. In such cases,we observed two consecutive peaks in the ipsilateral CaP neurons

F

1

2

12 20%

2 sec

A B

* ** * *

4

D

12

3

1

2

3

4

C

E

Fig. 3. Imaging of the activity of CaP motor neurons in the SAIGFF213A;UAS:GCaMPHS double-transgenic embryo during spontaneous contractions.(A) Time course of the frequency of spontaneous contractions between 17and 27 hpf. (B) The increase of the maximal fluorescence changes (Upper)and the fluorescence at the resting state (Lower) of GCaMP-HS in the somaof the CaP motor neuron between 21 and 27 hpf. Means and SD are shown(n = 5). (C) A side view of the double-transgenic embryo at 18.5 hpf. Anteriorto the left. GCaMP fluorescence was detected in the CaP neurons and someother neurons. The axons of the CaP neurons were used as ROI (1–4). (Scalebar, 50 μm.) (D) The fluorescence changes in the ROI-1, -2, -3, and -4 areshown as graphs. Downward peaks at the left ends of ROI-1, -2, and -3 areartifacts caused by the movement of the embryo. (E) A dorsal view of theSAIGFF213A;UAS:GCaMPHS4A embryo at 24 hpf. Anterior to the left. TheCaP neurons that are circled and numbered were used as ROIs. (Scale bar,200 μm.) (F) The fluorescence changes in the ROI-1 (dotted line) and ROI-2(solid line). Two consecutive peaks detected in ROI-2 are marked withasterisks.

Muto et al. PNAS | March 29, 2011 | vol. 108 | no. 13 | 5427

NEU

ROSC

IENCE

(asterisks in Fig. 3F). The synchronized activation of contralateralmotor neurons occurred alternately with approximately regularintervals (Fig. 3F).

Spatiotemporal Activation of the Motor Network: Synchronization,Alternation, and Propagation. As it was difficult to calculate thebrightness of ROIs precisely in embryos that moved in accordancewith muscle contractions, we paralyzed the embryo with a neuro-muscular junction blocker, D-tubocurarine, and performed imagingat 10 fps. Because it is not cell-permeable, the embryos were soakedin 100 μM D-tubocurarine and their tail tips were cut. The periodicand synchronized fluorescence changes with a frequency of 0.67 ±0.11 Hz (n= 5 embryos) were detected in the tubocurarine-treatedembryos, which was comparable but slightly faster than the fre-quencies detected in the untreated embryos (ROI-1, -2, -3, and -4 inFig. 4A and B andMovie S5).We also imaged embryos whose tailswere cut but not treated with D-tubocurarine as a control, and

detected thefluorescence changewith a frequency of 0.39± 0.10Hz(n = 6) (0.41 ± 0.24 Hz in uncut embryos), suggesting that thesurgery had little effect on the frequency.We applied the cross-correlogram, which is commonly used to

analyze correlation of timing of spike trains (16), to the analysisof the calcium signals (Fig. 4 C and D). The cross-correlogram oftwo ipsilateral neurons created a peak at zero, indicating that theactivation of ipsilateral neurons signals were synchronized (Fig.4C). The synchronized activation of the right and left side oc-curred alternately in both tubocurarine-untreated and treatedembryos. The cross-correlogram of pairs of contralateral neuronsshowed the peak was shifted by a 0.46 to 0.53 cycle (n = 6 pairsfrom 6 embryos), indicating relatively precise alternate activationof the contralateral neurons (Fig. 4D).Although the imaging at 10 fps demonstrated synchronized ac-

tivation of ipsilateral CaP neurons, one could imagine activation ofmotor neurons in a rostral-to-caudal sequence, which was detectedduring a swimming behavior in later developmental stages (17). Totest if propagation of the CaP activation occurs during the spon-taneous contractions and if GCaMP-HS can detect such rapidpropagation, we performed imaging at 20 fps. Although time res-olution of 20 fps was still too low to directly observe the propaga-tion during a single activation event, the averaged timing of thepeaks of each CaP neuron during a total of 7-min imaging revealeddelayed activation of CaP neurons located caudally (Fig. 4E).Withlinear regression [the coefficient of determination (R2) was in therangeof 0.8–0.85], the velocity of the propagation ofCaPactivationwas calculated as 5.6 μm/ms ± 0.8 (mean ± SE, n= 4 embryos).Finally, to get insight into correlation between Ca2+ imaging

using GCaMP-HS and spikes of the CaP neurons, we performedCa2+ imaging at 8 fps and electrophysiological recording simul-taneously using∼24-hpf embryos. On average, 53.5± 9.0 peaks perburst (n = 20 bursts from two embryos) were detected (Fig. 4F).With linear regression, we calculated a linear relationship (R2 >0.99) between the fluorescence change and the number of spikesthat had occurred to give rise to the change in the rising phase (Fig.S2). The increment of the fluorescence changes per spike wascalculated as 0.29% ± 0.019 (mean ± SE, n= 21). Because, in ourimaging, a range of the noise or a fluctuation of the base line weretypically within 1%, it is reasonable to assume that neuronal acti-vation composed of at least four spikes could be detected by usingthe current GCaMP technology.

DiscussionIn this study, we analyzed spontaneous contractions, the first lo-comotive behavior in zebrafish, by combination of an improvedGCaMP technology and the Gal4FF-UAS system. We success-fully imaged Ca2+ signals in the motor circuit during the naturalvertebrate behavior and demonstrate how the motor neurons arecoordinately activated.

Improvement of GCaMP.GCaMP-HS was generated by introducingsubstitutions described in the GFP superfolder mutant (12) intothe previous version of GCaMP, GCaMP2. As expected, the sub-stitutions conferred the higher refolding activity after heat-denaturation toGCaMP-HS. Consistent with this, theGCaMP-HSprotein became more stable and showed stronger fluorescencethanGCaMP2 in transfected cells, suggesting that the substitutionsindeed affected in vivo activities. Moreover, GCaMP-HS showedthe increased sensitivity to Ca2+ [i.e., the affinity for Ca2+ (Kd =102 nM)was higher than that of GCaMP2 (Kd = 146 nM) (9)], andHill coefficient became 5.0 although it is 3.8 in GCaMP2. How isthe sensitivity to Ca2+ affected by the “superfolder” substitutionsin the GFP domain? It has been shown by X-ray crystallographythat Arg-377 (position 74 of the calmodulin domain in Fig. 1A),which is located close to the second EF hand of the calmodulin, isinteracting with the GFP chromophore (18). Therefore, we inferthat the structural changes in the GFP chromophore affected the

1 2

3 4

1

2

3

4

20%2 sec

Time (sec)

Volta

ge (x

0.5m

V)

Norm

alized fluorescence

E

A

B

C

D

F

Distance (um)

Del

ay (s

ec)

Fig. 4. Spatiotemporal patterns of activation of the CaP motor neurons.(A) A dorsal view of the SAIGFF213A;UAS:GCaMPHS4A embryo at 24 hpftreated with 100 μM D-tubocurarine. (Scale bar, 200 μm.) (B) The fluorescencechanges in ROIs in A. (C) Cross-correlogram of the ipsilateral CaP neurons. (D)Cross-correlogram of the contralateral CaP neurons. (E) The averaged timingof the peaks of the ipsilateral CaP neurons located in a rostral (Left)-to-caudal(Right) sequence. Data from four embryos are shown. (F) A whole cell-patchrecording and simultaneous calcium imaging of the CaP neurons duringspontaneous contractions. Action potentials (gray) and calcium signals (black)are shown.

5428 | www.pnas.org/cgi/doi/10.1073/pnas.1000887108 Muto et al.

Ca2+ interacting part of the calmodulin domain, causing thechange in the Ca2+ sensitivity. This Kd value is closer to the restinglevel of estimated intracellular Ca2+ concentrations (∼100 nM)(19, 20), suggesting thatGCaMP-HS is better suited to detect Ca2+

transients in vivo.Recently, Tian et al. (10) reported construction of GCaMP3 by

introducing amino acid substitutions into GCaMP2. None of thesesubstitutions is overlapping with the substitutions introduced inGCaMP-HS. Therefore, it will be interesting to combine thesesubstitutions and test whether they show additive effects in de-tecting cellular Ca2+ transients.

Alternating and Synchronized Periodic Activation of the MotorNetwork. The activity of the primary neurons during the sponta-neous contraction has been analyzed by electrophysiology. Periodicdepolarization of single primary motor neurons and interneurons,synchronized depolarization of a pair of ipisilateral neurons of oneside, and alternating activation of a pair of contralateral neurons at19 to24hpf havebeenobserved (2, 3).However, it has beendifficultto analyze multiple neuronal activities by the electrophysiologicalapproach. In this article, we demonstrated the periodic and syn-chronizedactivationof the right and left sets of the ipsilateralmotorneurons and their alternating activation.Our present system enabled analysis of the onset of the sponta-

neous contraction in early stages. We found periodic and synchro-nized activation of ipsilateral CaP neurons in the very beginning ofthe spontaneous contractions. Although the soma of CaP neuronscan be fluorescently labeled by GCaMP-HS even before 18 hpf inthedouble-transgenic embryo, thefluorescence changes in the somawere hardly detected at this stage. In contrast, axonal Ca2+ tran-sients were observable at 18.5 hpf. Larger Ca2+ transients in theaxons than in the soma were detected in developing Purkinje neu-rons in the rat by using a calcium-indicator dye (21), and it wasshown thatCa2+channels are functioning in theaxonsofdevelopingchick motor neurons (22). Although it remains to be proven, thesefindings suggest that voltage-gated Ca2+ channels may be moreabundant in the axonal membranes in developing neurons. Wefound that the Ca2+ transients in the CaP neurons were not alwaysassociatedwithmuscle contractionsat this stage,demonstrating thatperiodic and synchronized activation of the CaP neurons precedesits functional coupling to muscle contractions. The clustering ofacetylcholine receptors and formation of neuromuscular junctionsproceed as the formation of the myotomes proceeds in a rostral-to-caudal direction between 16 and 20 hpf (23) (Fig. S3). Therefore, inthese stages, the connection between the motor circuits and musclecellsmaynot be complete or the activity of themotor circuitmaynotbe strong enough to reliably contractmuscles. It has not been knownhow the rhythmicity emerges in the early motor circuit. Although itstill remains to be elucidated, our present study demonstrated thatthe periodic and synchronized activity is generated by a networkinvolving a small number of CaP neurons before synchronization ofall of the CaP neurons.

Propagation of Activation of CaP Motor Neurons. We found prop-agation of activation of the CaP neurons along a rostrocaudal axisin the spinal cord during spontaneous contractions. The calcu-lated velocity was 5.6 μm/ms. A similar activity, the rostrocaudaldelay, was detected during a swimming behavior in later stages(17). However, its velocity was much faster (about 20-fold faster)than those in the spontaneous contractions. It will be interestingto disclose what makes this difference. Interestingly, a similar slowrostrocaudal wave has been reported in the neonatal mouse spinalcord (24). The calculated velocity was 13 to 18 μm/ms, which isfaster but closer to that we observed for the CaP motor neurons.These findings may suggest the existence of a commonmechanismfor motor circuits in the vertebrate developing spinal cord.

GCaMP Technology and the Gal4FF-UAS System. In vivo calciumimaging of neuronal activities has been carried out in zebrafishlarvae. Bulk loading of the fluorescent calcium-indicator dyes hasbeen used to record the activity of a neuronal ensemble in thespinal cord (4). However, with this method, neuronal subtypescould not be specified. To overcome this problem, Cameleon,a FRET-based GECI, was used in zebrafish (25). However, untilnow it has not been widely used probably because it is not easy toanalyze the calcium transient with a poor signal-to-noise ratio.GCaMP1.6 and GCaMP2 were successfully used to detect cal-cium signals in the neuropil of the zebrafish tectum, which con-sisted of axonal arbors of retinal ganglion cells and dendriticarbors of the tectal cells (26, 27). In contrast to these studies, ourpresent study clearly demonstrated that GCaMP-HS can detectcalcium signals in both the cell bodies and neurites. This findingis important because axonal imaging does not immediately allowus to identify neurons that are being activated, especially whenmultiple neurons are imaged.It has been shown that GCaMP has faster kinetics in calcium

ion binding than other GECIs (28). In our recording of actionpotentials and simultaneous calcium imaging (at 8 fps), we coulddetect the fluorescence change even on the same frame as the firstspike was generated (Fig. S2). Furthermore, the half-life of thedecay of the maximal fluorescence change was 0.92 ± 0.13 second(mean and SD, n=20 events) (Fig. S2), which was about 0.8 s witha calcium-indicator dye Oregon Green BAPTA (26). From thesefindings, we think that time resolution of GCaMP-HS in vivo issimilar to that of calcium-indicator dyes.We are currently constructing transgenic fish that express Gal4FF

invarious typesof specificneural circuits (11, 29).By combining thesewith the present GCaMP technology, we expect that neuronal ac-tivities of various neural circuits will be imaged and studies of func-tional neural circuits in a living vertebrate will be facilitated.

Materials and MethodsConstruction and Analysis of GCaMP-HS.WeusedpN1-GCaMP2 (9) asa templatefor mutagenesis. S30R (TCC to CGC), Y39N (TAC to AAC), N105T (AAC to AAC),and A206V (AAA to GTA) were introduced by using QuikChange Multi Site-directed Mutagenesis Kit (Agilent Technologies), giving rise to pN1-GCaMP-HS.The SalI-MluI fragment from pN1-GCaMP-HS was cloned into pRSETb, givingrise to pRSETb-GCaMPHS that was used to produce the His-tagged GCaMP-HSprotein in Escherichia coli. For calcium titration assay, the His-taggedGCaMP-HSprotein was purified by metal affinity chromatography, and fluorescence (F) ofGCaMP-HS was measured in different Ca2+ concentrations. Fmin and Fmax weremeasured in Ca2+-chelated buffer (20 mM BAPTA) and Ca2+-saturated buffer(300 nM Ca2+), respectively. All of the measurements were carried out in trip-licate. (F − Fmin)/(Fmax − Fmin) was plotted against Ca2+ concentrations (X). Kd

(dissociation constant) and n (Hill coefficient) were calculated by applying thereal data to theHill equation,Y =Xn/(Kd

n+Xn). To test the refolding activity, thepurified proteins were denatured at 95 °C for 5 min, cooled down to 25 °C, andmeasured for restoration of the fluorescence every 17 s in KM buffer (100 mMKCl, 20 mMMops pH 7.5) in the presence of 1 mM EGTA. The fluorescence wasnormalized by the initial fluorescence before heat-denaturation. The pN1-GCaMP2 and pN1-GCaMP-HS plasmids were transfected to HEK cells, as de-scribed previously (7). In three independent transfection experiments, thefluorescence intensity was calculated by measuring those of 150 cells per dish.ForWestern blot analysis, anticalmodulin antibody (Upstate #05–173) was usedto detect both GCaMP2 and GCaMP-HS.

Construction of the UAS:GCaMPHS4A and SAIGFF213A Transgenic Zebrafish.The GCaMP-HS gene was cloned into the EcoRI site of pT2MUASMCS (11),giving rise to pT2RUASGCaMP-HS. The DNA construct was injected withtransposase mRNA into zebrafish embryos at one-cell stage (30). The injec-ted fish were crossed with the hspGGFF27A fish or with the SAGFF73A;UAS:RFP fish (11). Among the offspring, UASGCaMP-HS;hspGGFF27A double-transgenic embryos were identifiable because they showed much brightergreen fluorescence than the Gal4FF-GFP transgene alone, and UASGCaMP-HS;SAGFF73A double-transgenic embryos were distinguishable because ofthe different color of the fluorescence.

For gene trapping, the Tol2 constructs T2KSAGFF (11) and T2TSAIGFF wereused. T2TSAIGFF was constructed by cloning a splice acceptor, the Gtx IRES

Muto et al. PNAS | March 29, 2011 | vol. 108 | no. 13 | 5429

NEU

ROSC

IENCE

sequence (31) and the Ga4FF gene into a minimal Tol2 vector (32). The gene-trap screens were performed essentially as described previously (33). TheSouthern blot hybridization, inverse PCR and adaptor-ligation PCR werecarried out as described previously (34).

Calcium Imaging and Microscopy. A Zeiss Axiovert upright microscope witha 10×/NA0.3 lens and a water-immersion 40×/NA0.8 lens, equipped with ahigh-sensitivity cooled CCD camera (Hamamatsu Photonics, ORCA-R2), wasused for calcium imaging. Each frame was taken under 100-ms exposure at10 fps or 50-ms exposure at 20 fps using the manufacturer’s image-acquisitionsoftware AquaCosmos. Images were analyzed off-line with custom-madesoftwares developed on LabVIEW (National Instruments) and Matlab(Mathworks). Zebrafish embryos were embedded in 5 mL of 2% low-meltingagarose in a lid of a 6-cm dish covered with 5 mL E3 water with or without100 μM D-tubocurarine chloride hydrate (Sigma; T2379). Because tubocrarineis not cell-permeable, their tail tips were cut for tubocurarine to permeate

throughout the body. The GFP-expressing CaP neurons were observed witha confocal laser microscope Zeiss LSM510Meta.

Electrophysiological Recording. The SAIG213A;UAS:GCaMPHS embryos at 24hpf were immobilized by treatment of Tricaine and D-tubocrarine and whole-cell recording of the activity of CaP neurons was performed essentially asdescribed previously (25). Simultaneously, the image of the CaP neurons weretaken under 100-ms exposure at 8 fps by a cooled CCD camera (ORCA-R2) witha software Metamorph (Molecular Devices). The TTL pulses from the camerafor each frame were recorded in the electrophysiology measurements.

ACKNOWLEDGMENTS. We thank M. Suster for helpful discussion; N. Mouri,M. Mizushina, M. Suzuki, Y. Kanebako, and A. Ito for fish room works; andM. Itoh-Hiratani and H. Takakubo for technical assistance. This work wassupported in part by the Mitsubishi Foundation (A.M.) and the NationalBioResource Project, and grants from the Ministry of Education, Culture,Sports, Science, and Technology of Japan.

1. Saint-Amant L, Drapeau P (1998) Time course of the development of motor behaviorsin the zebrafish embryo. J Neurobiol 37:622–632.

2. Saint-Amant L, Drapeau P (2001) Synchronization of an embryonic network ofidentified spinal interneurons solely by electrical coupling. Neuron 31:1035–1046.

3. Saint-Amant L, Drapeau P (2000) Motoneuron activity patterns related to the earliestbehavior of the zebrafish embryo. J Neurosci 20:3964–3972.

4. Fetcho JR, O’Malley DM (1995) Visualization of active neural circuitry in the spinalcord of intact zebrafish. J Neurophysiol 73:399–406.

5. Smetters D, Majewska A, Yuste R (1999) Detecting action potentials in neuronalpopulations with calcium imaging. Methods 18:215–221.

6. Miyawaki A, et al. (1997) Fluorescent indicators for Ca2+ based on green fluorescentproteins and calmodulin. Nature 388:882–887.

7. Nakai J, Ohkura M, Imoto K (2001) A high signal-to-noise Ca(2+) probe composed ofa single green fluorescent protein. Nat Biotechnol 19(2):137–141.

8. Ohkura M, Matsuzaki M, Kasai H, Imoto K, Nakai J (2005) Genetically encoded brightCa2+ probe applicable for dynamic Ca2+ imaging of dendritic spines. Anal Chem 77:5861–5869.

9. Tallini YN, et al. (2006) Imaging cellular signals in the heart in vivo: Cardiac expressionof the high-signal Ca2+ indicator GCaMP2. Proc Natl Acad Sci USA 103:4753–4758.

10. Tian L, et al. (2009) Imaging neural activity in worms, flies and mice with improvedGCaMP calcium indicators. Nat Methods 6:875–881.

11. Asakawa K, et al. (2008) Genetic dissection of neural circuits by Tol2 transposon-mediated Gal4 gene and enhancer trapping in zebrafish. Proc Natl Acad Sci USA 105:1255–1260.

12. Pédelacq JD, Cabantous S, Tran T, Terwilliger TC, Waldo GS (2006) Engineering andcharacterization of a superfolder green fluorescent protein. Nat Biotechnol 24(1):79–88.

13. Asakawa K, Kawakami K (2010) A transgenic zebrafish for monitoring in vivomicrotubule structures. Dev Dyn 239:2695–2699.

14. Myers PZ, Eisen JS, Westerfield M (1986) Development and axonal outgrowth ofidentified motoneurons in the zebrafish. J Neurosci 6:2278–2289.

15. Sato-Maeda M, Obinata M, Shoji W (2008) Position fine-tuning of caudal primarymotoneurons in the zebrafish spinal cord. Development 135:323–332.

16. Perkel DH, Gerstein GL, Moore GP (1967) Neuronal spike trains and stochastic pointprocesses. II. Simultaneous spike trains. Biophys J 7:419–440.

17. Masino MA, Fetcho JR (2005) Fictive swimming motor patterns in wild type andmutant larval zebrafish. J Neurophysiol 93:3177–3188.

18. Akerboom J, et al. (2009) Crystal structures of the GCaMP calcium sensor reveal themechanism of fluorescence signal change and aid rational design. J Biol Chem 284:6455–6464.

19. Nohmi M, Hua SY, Kuba K (1992) Intracellular calcium dynamics in response to actionpotentials in bullfrog sympathetic ganglion cells. J Physiol 458(1):171–190.

20. Wanaverbecq N, Marsh SJ, Al-Qatari M, Brown DA (2003) The plasma membranecalcium-ATPase as a major mechanism for intracellular calcium regulation in neuronesfrom the rat superior cervical ganglion. J Physiol 550(1):83–101.

21. Callewaert G, Eilers J, Konnerth A (1996) Axonal calcium entry during fast ‘sodium’

action potentials in rat cerebellar Purkinje neurones. J Physiol 495:641–647.22. Hata K, Polo-Parada L, Landmesser LT (2007) Selective targeting of different neural

cell adhesion molecule isoforms during motoneuron myotube synapse formation inculture and the switch from an immature to mature form of synaptic vesicle cycling. JNeurosci 27:14481–14493.

23. Flanagan-Steet H, Fox MA, Meyer D, Sanes JR (2005) Neuromuscular synapses canform in vivo by incorporation of initially aneural postsynaptic specializations.Development 132:4471–4481.

24. Bonnot A, Whelan PJ, Mentis GZ, O’Donovan MJ (2002) Spatiotemporal pattern ofmotoneuron activation in the rostral lumbar and the sacral segments duringlocomotor-like activity in the neonatal mouse spinal cord. J Neurosci 22:RC203.

25. Higashijima S, Masino MA, Mandel G, Fetcho JR (2003) Imaging neuronal activityduring zebrafish behavior with a genetically encoded calcium indicator. J Neurophysiol90:3986–3997.

26. Sumbre G, Muto A, Baier H, Poo MM (2008) Entrained rhythmic activities of neuronalensembles as perceptual memory of time interval. Nature 456(7218):102–106.

27. Dreosti E, Odermatt B, Dorostkar MM, Lagnado L (2009) A genetically encodedreporter of synaptic activity in vivo. Nat Methods 6:883–889.

28. Mank M, et al. (2006) A FRET-based calcium biosensor with fast signal kinetics andhigh fluorescence change. Biophys J 90:1790–1796.

29. Koide T, et al. (2009) Olfactory neural circuitry for attraction to amino acids revealedby transposon-mediated gene trap approach in zebrafish. Proc Natl Acad Sci USA 106:9884–9889.

30. Kawakami K, et al. (2004) A transposon-mediated gene trap approach identifiesdevelopmentally regulated genes in zebrafish. Dev Cell 7(1):133–144.

31. Chappell SA, Edelman GM, Mauro VP (2004) Biochemical and functional analysis ofa 9-nt RNA sequence that affects translation efficiency in eukaryotic cells. Proc NatlAcad Sci USA 101:9590–9594.

32. Urasaki A, Morvan G, Kawakami K (2006) Functional dissection of the Tol2transposable element identified the minimal cis-sequence and a highly repetitivesequence in the subterminal region essential for transposition. Genetics 174:639–649.

33. Asakawa K, Kawakami K (2009) The Tol2-mediated Gal4-UAS method for gene andenhancer trapping in zebrafish. Methods 49:275–281.

34. Urasaki A, Kawakami K (2009) Analysis of genes and genome by the tol2-mediatedgene and enhancer trap methods. Methods Mol Biol 546(2):85–102.

5430 | www.pnas.org/cgi/doi/10.1073/pnas.1000887108 Muto et al.