The larva of C. robusta has a central nervous system (CNS) composed of 177 neurons, the notochord, the neural tube, and the brain vesicle (Ryan et al., 2016). Whereas the notochord develops into the vertebral column in vertebrates, in C. robusta it is known to disappear during metamorphosis together with the regression of the larval organs. The CNS of C. robusta contains the otolith (OT), the ocellus (OC), and photoreceptors (PR). These structures are called the brain vesicle (BV). Also, it has the peripheral interneurons (PNIN), motor neurons (MN), the Ascending motor ganglion neurons (AMG) and the motor ganglion interneurons (MGIN) which are involved in movement of the tail. These are called the motor ganglion (MG). It has peripheral sensory neurons (PSN) which detect stimuli from the papillae and convey them to BV. (Olivo et al., 2021; Ryan et al., 2016).

For imaging neural activity in C. robusta larvae, constructs employing the prohormone convertase 2 (PC2) promoter has been used. However, this construct carries limitations. Although PC2 has been widely used as a pan-neuronal marker (Sekiguchi et al., 2007), its expression is not restricted to neurons and extends to a diverse range of peripheral peptidergic cells outside the nervous system (Osugi et al., 2020). This calcium signals thereby obtained inevitably contain extensive background arising from non-neural cells in addition to neural origin. The βtubulin>GCaMP6s construct generated in the present study overcomes the limits of above construct in principle ( Fig. 1A ). The Ci-βtubulin promoter was originally discovered as a marker of neural cell differentiation in C. robusta larvae (Miya and Satoh, 1997), and it has been used as a pan-neuronal marker for various purposes (Hamada et al., 2011; Horie et al., 2011; Hozumi et al., 2015; Kusakabe et al., 2004). However, few studies have used it for the purpose of imaging neural activity itself. Thus, βtubulin>GCaMP6s is superior to those having been used in that they enable comprehensive measurement of activity throughout the nervous system while excluding non-neural background signals.

As the experiment progressed, although βtubulin>GCaMP6s reports calcium signals in axons and neuropil in addition to cell bodies, we found it was difficult to isolate signals from individual somata in densely packed regions, and larval movement during observation occurred fluorescence fluctuations unrelated to neural activity. To deal these issues, βtubulin>H2B-GCaMP6s-T2A-mCherry was constructed ( Fig. 1B ) in which GCaMP6s is confined to the nucleus and mCherry is co-expressed in the somata by the self-cleaving peptide T2A (de Felipe et al., 2006; Donnelly et al., 2001; Liu et al., 2017). T2A is a self-cleaving peptide that has been used as a means of co-expressing multiple proteins of interest from a single construct (Totsuka and Hotta, 2026). Concentrating GCaMP6s in the nucleus eliminates overlapping axonal signals and allows individual somata to be identified clearly as ROIs. Because mCherry fluorescence is independent of calcium concentration, its fluctuations can be seen as motion artifacts or defocusing. Thus, we can use mCherry fluorescence as a reference signal for ratiometric correction of the GCaMP6s signal. This approach has been used to reduce motion artifacts in calcium imaging (Nguyen et al., 2016; Venkatachalam et al., 2016).

In this study, βtubulin>GCaMP6s and βtubulin>H2B-GCaMP6s-T2A-mCherry, based on the Ci-βtubulin promoter, were newly generated. As described in before section, these are considered to overcome in principle the limits of the constructs having been used. Indeed, βtubulin>GCaMP6s's fluorescence was observed at a broad range of the nervous system (Figs. 1C, 1D, 1F, 1G, 1I, 1J, 1L, 1M, Extended data). This construct can therefore be regarded as a tool suitable for pan-neuronal imaging in C. robusta larvae that surpasses existing constructs in simultaneously satisfying neural specificity by excluding non-neural signals and comprehensiveness by visualizing the entire nervous system. Its novelty is achieving with a single construct the two requirements of specificity and comprehensiveness, which are often difficult to reconcile. However, some nervous systems were rarely observed while they should be more observed, and room for further improvement of the constructs remains. In βtubulin>H2B-GCaMP6s-T2A-mCherry, localizing GCaMP6s to the nucleus by fusing with H2B eliminated the overlapping signals of axonal origin and enable us to identify individual neuronal somata obviously as ROI. Because GCaMP6s's fluorescence was localized in the nucleus, the definition of ROIs is facilitated, which suited the aim of this study of tracking multiple cells at the same time. You can also choose between the two types of constructs depending on the subject you want to observe (e.g. if you want to see the axon in addition to the somata, you can use βtubulin>GCaMP6s and if you want to see specific nervous system clearly, you can use βtubulin>H2B-GCaMP6s-T2A-mCherry).

The principal feature of βtubulin>H2B-GCaMP6s-T2A-mCherry lies in its design, in which nuclear-localized GCaMP6s and cytoplasm-specific mCherry are co-expressed. Because larvae of C. robusta move actively and the specimen shift out of the focal plane, the fluorescence of GCaMP6s contains artifacts from defocusing in addition to changes in calcium concentration. Since the fluorescence of mCherry does not depend on calcium concentration, its fluctuations can be regarded as components attributable to defocusing (Figs. 1E, 1H, 1K, 1N, Extended data). By using mCherry, which is always co-expressed at a constant ratio in the same cell, these artifacts can be corrected in the GCaMP6s signal. This procedure is expected to separate genuine excitation of neural structures by increasing calcium concentration from motion artifacts and thereby to enhance the reliability of the analysis. Using a dual-color design makes a core advantage of this construct in that it not only facilitates ROI identification but also provides a basis for quantitative noise correction. In the network analysis using CNER, some communities were detected from the correlation structure of the fluorescence time series of each ROI. Also, it demonstrated that the community structure differed according to the presence or absence of stimulation (Figs. 1C, 1E, 1F, 1H, 1I, 1K, 1L, 1N). Sample of Figs 1D, 1G, 1J and 1M reacts to mechanical stimulation with a glass needle, and Samples of figs 1C, 1E, 1F, 1H, 1I, 1K, 1L and 1N had fluorescence even without mechanical stimulation. it is thought that this is because that these samples react to laser light used for observation. These results suggest that the larva of C. robusta's CNS may have a functional modular structure superimposed upon the known connectome (Ryan et al., 2016). In particular, the finding that MG1 showed negative correlations with multiple ROIs (Figs. 1C, 1F, 1I, 1L) may reflect inhibitory connectivity. Taken together, this study demonstrates that this approach captures not only the visualization of neurons but also the fact that different neuronal populations are excited depending on the type of stimulus at the level of functional connectivity. However, since the ROIs where GCaMP6 and mCherry are expressed differ in this construct, it may not be suitable for completely eliminating noise due to the differences in concentration of fluorescence between the nucleus and the cytoplasm. As a future challenge, to achieve more accurate measurements, we could develop a construct that localizes non-calcium-derived signals within the nucleus while keeping the construct length short.

Construction of βtubulin>GCaMP6s

Genomic DNA was extracted from the eggs and sperm of C. robusta using NucleoBond HMW DNA (Takara Bio). The nucleotide sequence of the Ci-βtubulin promoter was identified on the basis of the Ghost Database compiled by the Satoh group, Laboratory of Developmental Biology, Department of Zoology, Graduate School of Science, Kyoto University. The C. robusta used in the present study were obtained from the National BioResource Project (NBRP) through the Shimoda Marine Research Center, University of Tsukuba; the Graduate School of Science, Kyoto University; and the Misaki Marine Biological Station, Graduate School of Science, The University of Tokyo.

The βtubulin>GCaMP6s construct was generated by nested PCR. In the first step (KOD One PCR), the βtubulin promoter region was amplified from genomic DNA (primers: forward 5'-GGAGTCCAGATAAACTTGGA-3', reverse 5'-TAGCTCCGATTTGGTTGCCG-3'). In the second step (HiFi PCR), the insert was amplified from the product of the first step (primers: forward 5'-ATGGGTTCTCATCAT-3', reverse 5'-GACTCTAGAGGATCC-3'), and the vector was amplified from a PC2>GCaMP6s plasmid previously prepared in our laboratory (primers: forward 5'-GACTCTAGAGGATCCGATCAAGCACTGAGG-3', reverse 5'-CATCTGCAAGCTGGCCATGGGTTCTCATC-3'). The products of each step were verified by agarose gel electrophoresis and then purified from the gel using the Wizard SV Gel and PCR Clean-Up System (Promega). The purified insert and vector fragments were joined by homologous recombination using In-Fusion HD Enzyme Premix (Takara Bio) to construct the βtubulin>GCaMP6s plasmid. The sequence of the purified construct was verified by Sanger sequencing (Azenta Life Sciences) and by ClustalW alignment (Thompson et al., 1994).

Construction of βtubulin>H2B-GCaMP6s-T2A-mCherry

In order to express GCaMP6s localized to the nucleus of neurons together with mCherry localized to the cytoplasm, the βtubulin>H2B-GCaMP6s-T2A-mCherry construct was generated. The insert was amplified from PC2>H2B-GCaMP6s-T2A-mCherry, which had been prepared in our laboratory (forward: 5'-ATCTGCAAGCTGGCCATGGTTGCATCCAAA-3', reverse: 5'-GCTCAGCTGGAATTCTTACTTGTACAGCTC-3'), and the vector was amplified from the βtubulin>GCaMP6s plasmid described in the before section (forward: 5'-GAATTCCAGCTGAGC-3', reverse: 5'-GGCCAGCTTGCAGAT-3'), each serving as template for KOD One PCR. All subsequent procedures, from purification onward, were performed as described in the before section.

Electroporation into Ciona robusta eggs

DNA constructs were introduced into fertilized eggs by electroporation using a Gene Pulser Xcell system (Bio-Rad). After introduction, the fertilized eggs were distributed into gelatin-coated dishes and incubated at 20°C.

Fluorescence imaging

Swimming larvae were immobilized by adhering them to the bottom of Poly-L-Lysine (PLL)-coated dishes in order to restrict their movement. For imaging of neural activity, a confocal laser scanning microscope (Olympus FLUOVIEW FV4000) was used with a 20× objective lens. Mechanical stimulation was delivered to the papillae of larvae immobilized on PLL-coated dishes using a glass needle fabricated with a PC-10 puller (NARISHIGE) and mounted on a motorized micromanipulator (NARISHIGE MM-89). All observations were performed at a room temperature of 20°C.

Image processing with Fiji

The acquired fluorescence images were processed using the open-source image analysis platform Fiji (Schindelin et al., 2012), and time-series data of fluorescence intensity were extracted for each neuronal ROI. Data subjected to ratiometric correction were used for the analysis, and the following expression was used for the correction: F = GCaMP6s / mCherry

Analysis with CalciumNetExploreR (CNER)

Network diagrams and heatmaps representing functional connectivity between neurons were generated using CNER (Lenci and Sieger, 2025), an R package dedicated to network analysis of calcium imaging data. The extracted fluorescence time series of each ROI were normalized to the interval [0, 1] by min–max normalization, and a correlation matrix based on Pearson's product-moment correlation coefficient was calculated in order to evaluate the similarity of activity patterns between cells. Community detection was then performed on the resulting data using the Louvain method (Blondel et al., 2008). Pairs with a correlation coefficient of 0.5 or greater were subsequently regarded as edges, and a neural network diagram was constructed (Extended data). In drawing the network diagram, nodes were color-coded according to cell type and the line width of each edge was made proportional to the strength of the connection. Communities were further enclosed together so as to emphasize the subgraph structure visually. The correlation matrix was additionally visualized as a heatmap (Figs. 1F,1G and 1H). A red–blue gradient was used for the color scale, and a dendrogram was appended by hierarchical clustering using the group average method. Community information was also displayed in a sidebar as row and column annotations so that the cluster structure could be grasped intuitively.