Introduction

The nervous system is composed of a vast number of neurons that perform diverse functions. How are these neurons generated and how do they communicate with each other to form functional networks? To better understand the function and development of the nervous system, it is crucial to examine the dynamic behavior of neuronal populations and how these dynamics evolve over time. However, such investigations are often hampered by the huge number of neurons involved and the invasiveness of conventional electrophysiological recording techniques.

Optical methods provide a powerful alternative, as light facilitates non-invasive access to large populations of neurons, which is difficult with electrophysiological recordings that insert electrodes into target tissues as single-point detectors. Voltage imaging enables fast, direct, and simultaneous detection of membrane potentials across multiple cells, providing an excellent system for understanding the dynamic behavior of neuronal populations [1–3]. Its high speed and directness allow the detection of action potentials, hyperpolarization, and subthreshold responses, which are difficult to examine using calcium imaging, partly because of the slow kinetics of the indicators and the dynamics of calcium [4, 5]. Various voltage-sensitive dyes have been developed to improve brightness, signal-to-noise ratio, and spectral range, thereby advancing our understanding of network dynamics in the brain [6–10]. More recently, genetically encoded voltage indicators (GEVIs) have been developed, allowing for cell-type-specific voltage imaging [11–13]. These GEVIs include the ones based on voltage-sensitive phosphatase domains [14–17], as well as microbial rhodopsin-based sensors [18–21]. Although GEVIs have been successfully applied to investigate brain and cardiac function [22–29], their application to developmental studies of the nervous system remains limited. This is largely due to the technical challenges of achieving long-term, non-invasive voltage imaging in developing tissues.

One of the earliest neural structures to develop is the spinal cord, which serve as the core of rhythmic motor behaviors through central pattern generators [30]. Neurons in the spinal cord, including motor neurons, exhibit prominent population activity even at early in development. This activity drives fetal movement and is suggested to play a critical role for circuit formation [31, 32]. While electrophysiological studies have characterized the firing properties of these neurons, the spatiotemporal organization and developmental emergence of their coordinated voltage dynamics, the core of the remain to be understood [33–35].

In this study, to examine the spatiotemporal dynamics of neuronal populations and their emergence during development, we applied GEVI ArcLight to zebrafish embryos, taking advantage of their optical transparency, genetic tools, and rapid development [36, 37]. Although recent studies have employed GEVIs in zebrafish larvae primarily to explore circuit function [20, 21, 24, 25, 28], long-term developmental imaging has not yet been achieved. Here, we establish a long-term voltage imaging approach that enables the observation of voltage dynamics in neuronal populations over the course of development. This approach allowed us to reveal how coordinated voltage activity emerge during early neural development.

Materials and methods

Fish lines

Wild-type zebrafish (Danio rerio) with RW genetic background or nacre line were used [38]. Tg(elavl3:GAL4-VP16), Tg(UAS:lynRFP), and Tg(elavl3:GCaMP6f) fish lines have been described previously and obtained from the National BioResource Project [24, 39]. The Tg lines were crossed with the nacre line. All procedures were performed in accordance with a protocol approved by the Saitama University Committee on Animal Research.

DNA Construction and establishment of ArcLight transgenic line

To generate pTol2-UAS-ArcLight plasmid, ArcLight-A242 sequence was excited from the pCS2-ArcLight-A242 plasmid (Addgene) and inserted into pTol2-UAS-MCSF-polyA, in which UAS and polyA sequences were integrated into the pTol2-MCS plasmid [40]. To establish pTol2-UAS-ArcLight line, Tol2 plasmid DNA and transposase mRNA were co-injected into one-cell-stage embryos. Genotyping was performed using the primers for ArcLight-A242 (5′-TCGACGGTATCGATAAGCTTGC-3′,5′-CACCTCCCCCTGAACCTGAAACATA-3′).

Transient expression of ArcLight in the spinal cord and cerebellum

For the spinal cord, 25 ng/μl pTol2-UAS-ArcLight-A242-polyA plasmid DNA was co-injected with 25 ng/μl Tol2 transposase mRNA into one-cell-stage embryos of Tg(elavl3:GAL4-VP16) embryos, which express Gal4 in neurons [39]. After the injection, the embryos were incubated at 23°C and then transferred to 28°C by 18 hour post fertilization (hpf). ArcLight-positive embryos were screened at the 18.5 to 19 hpf, followed by voltage imaging.

Confocal imaging

Zebrafish embryos were paralyzed in 0.02% tricaine and mounted on 2% low melting-point agarose (Sigma-Aldrich). Optical sectioning was performed using an Olympus FV1000 or a Nikon A1R confocal microscope. ImageJ, NIS-Elements (Nikon), and Imaris (Bitplane) programs were used to perform the image analysis.

Voltage imaging

Embryos from 16 to 23 hpf were paralyzed with tubocurarine (0.5 mM, Sigma-Aldrich), with their tail tips cut for better penetration of tubocurarine [41, 42]. Then, embryos were mounted on 1.8% low melting-point agarose containing 0.1-0.2 mM tubocurarine. After embedding the embryos onto 1% agarose, an extracellular solution (134 mM NaCl, 2.9 mM KCl, 2.1 mM CaCl₂, 1.2 mM MgCl₂, 10 mM HEPES, and 10 mM glucose, adjusted to pH 7.8 with NaOH) containing 150 mM tubocurarine was added. The agarose around the tail was removed using a surgical blade to allow efficient transfer of the drug solution. For high-speed voltage imaging, a fluorescence microscope (FN-1, Nikon) equipped with a CMOS camera (ORCA-Flash4.0, Hamamatsu Photonics) was used with a 40×/0.8 NA water-immersion lens. For most recordings, a 0.7× relay lens (NIKON) was also used. Fluorescence images were acquired using NIS-Elements (Nikon) software at approximately 16-200 Hz. High-speed confocal scanning was performed using a confocal microscope (Nikon A1R) with a 25×/1.0 NA water-immersion lens. For drug treatment, tetrodotoxin (1 μM, TOCRIS) was added to the extracellular solution using a peristaltic pump (World Precision Instruments). A thermo plate was used to keep the recording chamber at 28°C (Tokai hit).

Image analysis

Images were analyzed using the NIS-Elements (Nikon), Fiji, and OriginPro (Lightstone) programs as described previously [41]. To extract changes in fluorescence intensity, regions of interest (ROIs) were manually set using NIS-element software (Nikon). Normalized changes in fluorescence intensity in the target regions (ROIs) were quantitatively analyzed by calculating dF/F in the ROIs (dF/F₀=[F_(t)–F₀]/F₀, where F_(t) and F₀ are the fluorescence intensities at a certain time point and initial time points, respectively). If the larvae/embryos moved during the voltage imaging, the images were aligned using a Fiji program (TurboReg plugin). Denoising and bleach corrections were also applied to some data using OriginPro (Lightstone). Cross-correlation and hierarchical clustering analyses were performed using OriginPro (Lightstone). The coupling rate was calculated as the ratio of the frequency of target cell activity to that of mature cells. Among depolarizing signals with small voltage changes, those characterized by a longer time to peak and a shorter decay time were defined as "immature firing" and were distinguished from other small depolarizations.

Electrophysiological recordings

Patch clamp recordings were conducted as described previously, using borosilicate glass pipettes at 28 ℃ (World Precision Instrument, IB150F-4WPI) [24]. Electrical responses were acquired via a Multiclamp 700B (Molecular devices) and DAQ interface (National Instruments, USB-6212 BNC) with WinWCP software (Univ. of Strathclyde), which also recorded the timing of image acquisition.

Results

Establishment of voltage imaging using ArcLight zebrafish embryos

To investigate voltage dynamics of neuronal populations in neuronal populations during development, we established a transgenic zebrafish line that expresses ArcLight specifically in neurons. We first established the transgenic line Tg(UAS:ArcLight) using the GAL4-UAS and Tol2 transposon systems [37] (Fig. 1a). To confirm ArcLight localization to cellular membranes in Tg(UAS:ArcLight) fish, we examined Tg(elavl3:GAL4-VP16;UAS:ArcLight;UAS:lyn-RFP) embryos, in which GAL4 is expressed in neurons and cellular membranes are labeled with Lyn-RFP.

Figure 1

At 1 dpf, ArcLight was widely observed in the neural tube and co-localized with Lyn-RFP, indicating proper membrane localization of ArcLight in neurons (Fig. 1b, c).

Spinal cord neurons exhibit spontaneous firing early in development, with coordinated activity of the neuronal populations beginning around 17 hpf [34, 43]. Consistent with this, Tg(elavl3:GAL4-VP16;UAS:ArcLight) embryos displayed oscillatory changes in ArcLight fluorescence in the ventral spinal cord (Fig. 1d–g, 0.14±0.10 Hz, 6 fish). Fluorescence fluctuations alternated between the left and right sides of the spinal cord, consistent with previous reports of periodic depolarization (PD) in spinal neurons [43]. This activity was largely abolished following treatment with the sodium channel blocker tetrodotoxin (TTX) (Fig. 1e-g). These results demonstrate that ArcLight can reliably detect spontaneous activity in developing spinal neurons.

We next examined whether ArcLight can resolve spinal voltage dynamics at both single-cell and population levels. As shown in Figure 2, oscillatory activity was clearly detected in individual neurons, with distinct amplitudes and temporal patterns among neighboring neurons (Fig. 2a-c). Depolarizations typically propagated sequentially along the rostrocaudal axis. In addition to large depolarizations that closely resemble PD, small membrane potential fluctuations were also observed between the PDs (Fig. 2d arrowheads). Based on their size and timing, these may be synaptic bursts or subthreshold events, which are generally only captured by electrophysiological recordings [43]. Cross-correlation analysis revealed strong positive correlations among ipsilateral neurons and negative correlations between contralateral pairs, indicating high synchrony within each side of the spinal cord (Fig. 2e). Hierarchical clustering analysis also supported this synchrony with partially disordered activity along rostrocaudal axis (Fig. 2f). The depolarization was confirmed by electrophysiological recordings (Supplementary Fig S1). To evaluate the strength of ArcLight voltage imaging, we performed spinal calcium imaging using Tg(elavl3:GCaMP6f) and compared its temporal resolution with ArcLight. Calcium imaging exhibited substantially lower temporal resolution and reduced variability in activity patterns (Supplementary Fig S1), highlighting the superior precision of the GEVI ArcLight in capturing neural population dynamics.

Figure 2

Subcellular voltage imaging of neural populations

Measuring membrane potential dynamics in subcellular compartments such as axons and dendrites is crucial for understanding neuronal information processing [44, 45]. As shown in Figure 3, ArcLight voltage imaging successfully captured depolarizations in somata as well as in proximal and distal axonal regions. Although somata and axons exhibited similar depolarization patterns, the amplitude of voltage changes was slightly smaller in axons than in somata (Fig. 3b). Axonal depolarizations were also confirmed by confocal microscopy (Supplementary Fig. S2). Furthermore, high-resolution recordings revealed voltage fluctuations and morphological dynamics in fine structures such as filopodia (Supplementary Fig. S3).

Figure 3

Single-cell voltage imaging reveals simultaneous voltage and morphological dynamics

Spinal circuits comprise diverse neuronal types with distinct distributions, morphologies, and physiological properties [46]. To understand how these circuits develop, it is important to simultaneously monitor voltage dynamics and morphological changes. To achieve this, we used mosaic expression of ArcLight. By injecting UAS-ArcLight pDNA into Tg(elavl3:GAL4-VP16) fish, we sparsely labeled neurons, enabling high-resolution tracking of membrane potential changes and cellular morphology (Fig. 4a–c, 20 hpf, PMNs). We also captured voltage dynamics from interneurons that are otherwise difficult to identify (Fig. 4d, e). In addition to PD and synaptic burst (SB)-like activity, some neurons showed hyperpolarizing responses (Fig. 4e). Over the course of one hour, we observed notable morphological changes in these neurons, suggesting active developmental remodeling (Supplementary Fig. S4).

Figure 4

Developmental changes in voltage dynamics of neuronal populations

Finally, we investigated the emergence and maturation of voltage dynamics during development. Periodic firing in the spinal cord drives coiling behavior, analogous to fetal movement in humans, which are thought to contribute to spinal circuit formation [32]. This characteristic activity appears around 17 hpf and rapidly increases in frequency and amplitude until 20 hpf, followed by a gradual decline [34]. However, how population-level coordinated activity arises during this process remains to be elucidated.

To address this, we performed long-term voltage imaging of the spinal cord along the course of development (Fig. 5). At 18 hpf, primary motor neurons (PMNs) possesses an axonal protrusion but seldom exhibited firing, instead showing small voltage fluctuations (Fig. 5c, arrowheads, cell 1, located at somite 8). These fluctuations were observed in both soma and axons (Supplementary Fig. S5) and displayed slower kinetics than mature firing (Supplementary Fig. S6). Over the next 30 minutes, PMNs gradually extended axons and increased firing frequency. By 19.5 hpf, the firing frequency reached ∼0.3 Hz (0.29±0.05 Hz). In addition to cell 1, PMN cell 2, located more caudally, showed slower axonal extension and began firing after 18.5 hpf, ultimately synchronizing with other neurons by 19.5 hpf (Fig. 5). Similar temporal maturation patterns were also observed in other PMNs with some variability (Supplementary Fig. S7). Quantitative analyses revealed that firing frequency and coordination increased in tandem with axonal extension (Fig. 6a, b). In deede, the emergence of small depolarization and matured firing strongly correlated with morphological maturation (Fig. 6c, d). Moreover, coordination among PMNs also increased alongside the maturation of their individual voltage dynamics (Fig. 6e, f). These data suggest a sequence in the emergence of coordinated activity in spinal neurons: (1) absence of depolarization, (2) small depolarizations, (3) irregularly coordinated activity, and (4) fully synchronized firing.

Figure 5

Figure 6

To further assess how spontaneous activity evolves later in development, we extended voltage imaging to 23 hpf. By this stage, the previously regular and rapid firing observed at 20 hpf had decreased (Fig. 7, Supplementary Fig. S8). Depolarization patterns also changed—bursts became longer and left–right alternation diminished (Fig. 7b). Cross-correlation analysis revealed that at 20 hpf, neurons on both sides of the spinal cord were highly synchronized and exhibited strong periodicity (Fig. 7c). By 23 hpf, this coordination was largely diminished, although synchrony among ipsilateral neurons persisted. Notably, left–right alternating activity, which was prominent at 20–21 hpf, was markedly reduced at 23–24 hpf (Fig. 7d). These results suggest that while coordinated PMN activity declines during later stages of development, intra-hemispheric synchrony is maintained.

Figure 7

Discussion

Understanding how neuronal populations generate coordinated activity during development is a central challenge in neuroscience. Electrophysiological recordings have historically provided valuable insights into the membrane dynamics of spinal neurons. However, these techniques are invasive and largely limited to short-duration and single-cell level [43, 46]. Calcium imaging has been used to examine developing circuits but lacks the temporal resolution and direct voltage sensitivity required to resolve fine membrane potential dynamics [42, 47, 48]. In this context, genetically encoded voltage indicators (GEVIs) offer a promising alternative, providing direct, fast, and non-invasive readouts of membrane potential. While GEVIs have been applied in mature brain tissues and cardiac tissues, their application to vertebrate embryonic development has remained rare due to technical limitations. To the best of our knowledge, this is the first study to achieve long-term voltage imaging of genetically defined neuronal populations in the spinal cord during development in a vertebrate model.

High-resolution imaging of voltage and morphological dynamics

Subcellular compartments—such as dendrites axons, and filopodia—play essential roles in neuronal function, shaping synaptic integration and signal propagation [44, 45]. In this study, ArcLight enabled voltage imaging not only at the soma, but also at proximal and distal axonal regions of primary motor neurons (PMNs). Notably, voltage fluctuations were also observed in filopodia, suggesting that even fine structures can exhibit membrane potential dynamics. Such fine-scale resolution opens opportunities to investigate the bioelectrical properties of neurite outgrowth, synaptogenesis, and structural remodeling. Combining voltage imaging with ratiometric or dual-color imaging approaches may allow simultaneous measurement of voltage and morphological parameters at high spatiotemporal precision [49].

Long-term population-level voltage dynamics in the developing spinal cord

The zebrafish spinal cord possesses one of the earliest neuronal circuits to form during development, and its transparency makes it an ideal model for in vivo imaging. Using a newly generated ArcLight transgenic line, we visualized voltage dynamics in spinal neuron populations from early developmental stages. At 20 hpf, we detected left-right alternating depolarizations, consistent with periodic depolarizations (PDs) previously observed via electrophysiology [34]. Compared to calcium imaging or earlier voltage sensors like ASAP1, ArcLight offered superior temporal resolution and signal-to-noise ratio, respectively [13, 50]. This enabled detection of not only suprathreshold PDs, but also subthreshold events and hyperpolarizations, capturing a wider dynamic range of neuronal activity. Importantly, ArcLight could resolve activity patterns at both the single-cell and population levels, providing a comprehensive picture of how spinal circuits operate and evolve during development.

Emergence of voltage dynamics in the spinal neuron populations

One of the significant findings of this study is the stepwise emergence of coordinated voltage dynamics in motor neurons, closely tied to their morphological maturation. By performing long-term voltage imaging from, we found that PMNs followed a developmental sequence: (1) absence of depolarization, (2) emergence of small voltage fluctuations, (3) irregular and weakly coordinated activity, and (4) synchronized, robust left-right firing. This progression was strongly correlated with axonal extension, suggesting a tight coupling between structure and function.

It has been implicated that spinal cord neurons do not fire before axonal outgrowth, but detailed observation was missing [46]. Interestingly, here early voltage fluctuations preceded axon outgrowth, possibly driven by intrinsic dynamics of ion channels or electrical coupling via gap junctions, which predominate at these early stages. These small fluctuations may represent primitive forms of population synchrony, preceding the establishment of mature synaptic connections. This idea is consistent with other model systems where early electrical coupling plays an instructive role in circuit patterning [51].

Our approach provides a first-in-class demonstration of long-term, non-invasive voltage imaging in a developing vertebrate nervous system. Beyond the spinal cord, this approach can be extended to other regions, enabling a detailed dissection of diverse population dynamics during development. Furthermore, applying this imaging system to genetic models of neurodevelopmental disorders— such as autism spectrum disorders (ASD), epilepsy, or motor neuron disease—may allow the identification of early disruptions in coordinated neuronal activity. Detecting such early electrophysiological phenotypes could contribute to a better understanding of disease mechanisms.

In conclusion, we have established GEVI-based long-term voltage imaging platform for long-term, high-resolution monitoring of developing neuronal populations in vivo. This system revealed the temporal emergence and maturation of population-level voltage dynamics in the spinal cord, offering new insights into how coordinated neural activity is built during development. The combination of spatial, temporal, and functional resolution achieved here represents a critical advance for the field of developmental neuroscience. By integrating this approach with optogenetics, cell-type-specific reporters, or functional perturbation, future studies can address how intrinsic and extrinsic factors shape circuit assembly. Moreover, our findings highlight the value of voltage imaging as a general tool for probing bioelectric mechanisms in morphogenesis, circuit development, and disease [52, 53].

Data availability

Data will be deposited in a public repository (SSBD:repository) during the review process. Key supporting data are available from the corresponding author upon request.

Acknowledgements

We thank Drs Koichi Kawakami, Shinichi Higashijima, Hitoshi Okamoto, Hisashi Kakinuma, and Ryunosuke Amo for providing fish lines; Kyo Yamasu for plasmid DNA; Junichi Nakai, Masamichi Ohkura, and Keiko Ando for technical support; and Lawrence Cohen and Shinichi Higashijima for helpful discussions. This work was supported in part by JSPS KAKENHI (Grants-in-Aid for Scientific Research), JST Tenure Track program at Saitama University (SUTT), Strategic Research Program for Brain Science from Ministry of Education, Culture, Sports, JST A-STEP Grant Number JPMJTM20C05, New Energy and Industrial Technology Development Organization (NEDO), Kao Foundation grant, Asahi Glass Foundation, and Kato-bio Foundation.

Additional information

Author Contributions

A.S, A.H., N.F., M.H., K.M., and S.T. performed the experiments. S.T. conceived and designed the experiments. S.T., A.S., A.H., and N.F. wrote the manuscript. All authors reviewed the manuscript.

Funding

JSPS KAKENHI (19K06756)

• Sachiko Tsuda

JST-ASTEP

https://doi.org/10.52926/jpmjtm20c5

• Sachiko Tsuda

Additional files

Supplementary figures