Distinct and common dynamics of central brain spontaneous neuronal activity across cell types during Drosophila pupal development

Abstract

The frequency or degree of spiking activity and its relative timing, in pre- and post- synaptic neurons is a major determinant of the nature of synaptic plasticity. Such activity driven plasticity based on stimuli forms the basis of learning and formation of memory. However, in the absence of any sensory stimuli particularly during development, neurons are active and often in a synchronous manner. Such spontaneous activity driven plasticity is thought to underlie formation and development of circuits. Drosophila has become the species of choice for many neural developmental studies due to ease of handling and genetics. However, little is known in terms of spontaneous activity in the central brain of the Drosophila pupa and how such activity is distinct in different cell types during development. In the current work, using two-photon Ca²⁺ imaging, we obtain spontaneous activity of Drosophila pupae at different stages of development of 3 different cell types, namely, dopaminergic, cholinergic and GABAergic neurons. We show that all cell types follow a similar pattern over development with periodic synchronous activity separated by durations of silence. We find that the central brain spontaneous activity is different from what has been previously observed in the optic lobe, which is true across the cell types. However, the three cell types have different periodicities, spontaneous event timing precision and strength of spontaneous activity. Together our results can form the basis of understanding cell type specific circuit development and future investigations of alterations in spontaneous activity in different cell types with neurodevelopmental disorders.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1038/s41598-026-56092-6.

Introduction

Patterned spontaneous neuronal activity (SNA) is a fundamental firing property of developing as well as mature neurons^1–3. This firing is not induced by external sensory stimulus or synaptic transmission, instead it has been proposed that SNA originates from internal ionic oscillation within neurons due to random opening of certain types of voltage-gated and hyperpolarization-activated cyclic nucleotide-gated (HCN) ion channels^4,5. Spontaneous activity in developing neurons orchestrates diverse aspects of neural development, like differentiation, maturation and migration of neurons⁶, wiring specificity^7,8, formation of synapses⁹, synaptic refinement^10,11 pruning¹² and also neuronal survival³. To regulate different developmental and physiological properties of neurons, patterned SNA modulates expression of genes coding for transcription factors like CREB, NFAT as well as various other proteins^13,14. Disruptions in SNA patterns, whether through excessive or diminished activity, can lead to miswiring of neural circuits^{15–18,18–20}. Alterations in SNA has been reported as the key pathophysiological mechanism implicated in various psychiatric disorders like Schizophrenia, depression, bipolar disorder, neurological disorders like epilepsy, neurodevelopmental disorder like autism spectrum disorder and also in neurodegenerative disorders like Alzheimer’s and Parkinson’s Disease²¹.

In Drosophila embryo, first electrical activity of neurons is observed during 13–14 h AEL (hours after egg laying) as small amplitude currents. Synaptic transmissions start at 16 h AEL leading to EPSCs and first action potentials generated by 17 h AEL⁴. Myogenic developmental spontaneous activity was also reported in the form of spontaneous embryonic twitching of the body wall segments; initial low frequency and low amplitude twitches mature into a complex, patterned, rhythmic bursts of motion interspersed with episodes of motor inactivity²². This motion, though myogenic in origin, comes under motor neuronal modulation by the 17 h AEL²³. The SNA pattern in the embryonic motor neurons was found to be shaped by mechanosensory inputs as well as via regulatory feedback from myogenic activity. The SNA was found to be vital for the precise development of motor circuits and the emergence of coordinated larval locomotion^24,25.

Patterned spontaneous neuronal activity has also been reported in the mid and late pupal optic lobe neurons which become electrically active at around 50 h APF (after puparium formation). From 50 to 70 h APF, a clear ‘periodic stage’ is observed which comprises of 12 to15 minute cycles of coordinated active and silent phases. Each active phase consists of multiple bouts of neural activity, termed sweeps. The periodicity progressively exhibits greater irregularity at around 70 h APF onwards and is termed as ‘turbulent stage’²⁶. The oscillatory pattern of the brain is regulated by a small population of neurons which express Trpγ and is required for synapse and circuit formation²⁷.

Although the SNA patterns of the fly optic lobe have been detailed, the activity signature within the central brain neurons of the fly is yet to be elucidated. Here, we characterize the dynamics of SNA patterns in the developing central brain of the fly in the pupal stage and compare the same among different neurotransmitter systems, namely, the Dopaminergic and Cholinergic (excitatory) and the GABAergic (inhibitory) neurons. We have performed live calcium imaging from three windows of pupal development using 2-photon microscopy. We have considered the SNA patterns in three temporal windows of development during the second half of pupal development. We found nearly periodic occurrence of phases of activity across cell types at different developmental stages, with the silence period between active phases generally reducing from the ‘early stage’ (50–65 h APF) to ‘middle stage’ (65–80 h APF) and then the ‘late stage’ (80–95 h APF). We characterized the activity through detection of events in the Ca²⁺ fluorescence and then by quantifying the strength and pattern or timing of events. While a number of aspects of SNA were similar across cell types, we found distinct difference in amplitude of events, the duration of periods and the precision of events. These results can be the benchmarks for understanding altered spontaneous activity in disease and to guide understanding of cell type specific circuit development.

Materials and methods

In vivo calcium imaging:

Fly lines and their sources:

Genotype	Description	Source/Stock ID	References
Canton S	Wild type fly line	–
TH-GAL4	Dopaminergic driver line	BDSC # 8848	28
ChAT-GAL4 (7.4)	Cholinergic driver line	BDSC # 607181	29
Gad1-Gal4/CyO	GABAergic driver line	BDSC # 51630	30
UAS-GCaMP6s/CyO	Reporter line for driving the GECI GCaMP6s in desired cell type	BDSC # 42756	31

Experimental model and subject details:

Flies were reared at 25 °C on standard cornmeal medium in bottles or vials. Pupal development was staged with respect to white pre-pupa formation (0 h APF) or head eversion (12 h APF). The specific developmental age studied in each experiment is indicated in the figures and figure legends Using a two-photon microscope, imaging was done from the central brain region of Drosophila melanogaster between 50 and 95 h after puparium formation (h APF). We used the terms, ‘early window’ for pupae between 50 and 65 h APF, ‘middle window’ for pupae whose age is between 65 and 80 h APF and ‘late window’ for 80–95 h APF pupae which has completed most of the brain circuit development and are almost ready to eclose.

Preparation of pupal Drosophila for 2-photon based calcium imaging

The methods for the preparation of pupae, glue and mounting are adapted and modified from^26,27.

Materials required:

See Table 1.

Table 1.

List of essentials required for live calcium imaging using 2-photon microscope.

Items	Company
Jewelers forceps: Dumont No. 5, L 4 1/4 in., Inox alloy	Sigma-Aldrich
Scotch double-sided tape	Locally purchased
Cover glass (22 mm × 22 mm, No.1)	Blue Star
Glass slide (75 mm x 25 mm, Thickness: 1.45 mm)	Blue Star
O-ring or rubber band (Diameter ~ 1.5 cm)	(Locally purchased)
Heptane	Sigma-Aldrich
Paint brush	Locally purchased

Preparation of tape glue

30 ml of heptane is taken in a 50 ml Falcon tube. 2–3 feet long double-sided Scotch tape is cut into small (~ 1 cm) pieces. The pieces were added to the heptane containing tube and the lid was sealed with parafilm. The heptane and the pieces of the tapes were mixed overnight by continuously rotating it in a shaker, to allow the solvent (heptane) to dissolve the adhesive from double-sided tape. The glue was then filtered to remove any remaining tape. The prepared tape glue was covered with aluminium foil and stored inside a 4 °C fridge. The lid of the tube was sealed with parafilm every time after use to avoid evaporation of volatile heptane.

Preparation of water reservoir on cover glass

In order to perform two-photon microscopy-based hour-long imaging of live pupal prep, provision was required to be made on the cover glass so that sufficient water can be held between the glass and the water-immersion objective lens which would sustain for the entire imaging event. For this purpose, an O-ring or simply a small round rubber band (diameter ~ 1.5 cm) was affixed onto the cleaned cover glass using adhesive (Fig. 1A).

Fig. 1.

Imaging of spontaneous Ca²⁺ activity and its patterns in the developing central brain of the Drosophila pupae. (A) Schematic diagram showing the imaging set up of spontaneous neuronal activity patterns from the live pupal brain. (B1-B2) Brightfield view of pupal head in 4X (B1) and water-immersion 20X (B2). (C) A schematic diagram of the pupal brain showing the landmark synaptic structures, as viewed in conventional Confocal microscopy. The corresponding imaging plane for the 2-photon-based imaging is indicated by a grey rectangle. (D–F) Confocal image showing pupal neurons labelled by mCD8::GFP driven by TH-GAL4 (D), GAD-GAL4 (E) and ChAT-GAL4 (F). (G) Schematic diagram of the pupal brain showing the anterior view, as observed in the ‘hanging fly’ prep. The imaging area is indicated by a grey rectangle. (H, I and J) Representative fluorescence from two-photon microscopy imaging of pupal brains expressing genetically encoded calcium indicator GCaMP6s driven by the three GAL4 driver lines. (E, F and G) Representative ΔF/F traces (in %, scale bars shown to the right of each trace) showing patterns of Ca²⁺ activity across three windows of pupal development in the dopaminergic neurons (E), GABAergic neurons (F) and Cholinergic neurons (G). Windows indicate age groups of the pupae: window I represents 50 to 65 h APF, window II pupae belong to 66–80 h APF and window III stands for pupae of 81–95 h APF age group. The arrows and the dashed arrows indicate active phases and silent phases respectively. Red ticks mark location of peaks with each tick 2% in height.

Application of tape glue on cover glass

On the day of imaging, the opposite surface of the ‘reservoir cover glass’ was cleaned thoroughly with alcohol and allowed to dry. 2–3 drops of the tape glue were applied on the centre of the cover glass (on the opposite surface of the rubber band) and were allowed to dry for at least ~ 30 min.

Preparation of pupa (the ‘hanging fly’ prep)

Pupae of appropriate age were selected and taken out from their growth vial using a soft brush onto a glass slide. Using a pair of fine forceps, the pupal case was removed from the head region of the fly (Figure S1A-C). This exposes the head and provides optical access into the central brain region of the pupae. Next, the exposed head of the pupa is attached to the glued central part of the cover slip (Figure S1D). In order to image multiple pupae in a single imaging session, up to 2–3 ‘hanging fly’ pupal preps were affixed on a single cover glass. Glass-bound pupae were then placed on the 2-photon microscope stage (Fig. 1A and Figure S1D). The stage was specially customized to contain a depression in the middle to hold water, on top of which the cover glass containing the ‘hanging fly pupae’ was placed in such a way that the rubber band on top, pupa at the bottom, below the cover glass. (NOTE: The pupal bottom tip lies a few millimetres above the water. The pupal body should not touch the water surface). This second water reservoir minimizes dehydration of the pupae and provides passive humidification for the animals; otherwise, the pupae may dry out over hours-long imaging.

Imaging details

Two-photon calcium imaging was conducted using a two-photon microscopy system from Prairie Technologies, controlled by Prairie View software. The excitation beam, with a wavelength of 920 nm, was generated by an IR Insight Ti-Sap laser (Spectra-Physics) and directed over the head region for imaging. A 4X objective was used with white light to locate the pupae in the middle of the stage. A 20x/0.8 NA water immersion objective (2 mm working distance, Olympus) was used for imaging. Depending on the clarity of images through the head cuticle, the laser power was adjusted between 100 and 150 mW. Imaging was performed with 2X optical zoom by choosing rectangular regions of interest (ROIs) with the median of the x-axis side and y axis side being 297 µm (Range: 177–429 µm) and 224 µm (Range: 159–302 µm). The frame durations varied between 0.56 s up to 3 s with most being around the median frame period of 1.26 s. Imaging of spontaneous activity was performed from a single ROI for ~ 1–1.5 h.

Image processing and analysis

Images were analyzed using custom written codes in MATLAB as done previously^1,32–34. From the average image obtained with all frames collected, first the absence of x–y movements was ascertained. In case of movements an x–y movement correction was performed with correlation analysis. The method has been detailed in earlier work, in short, successive frames were moved relative to each other in + x and − x and in + y and − y in a grid pattern. Correlation coefficient for each relative shift of (x,y) was calculated between frames. If the maximum correlation was found to be different from (0,0) shift, the next image was corrected by the appropriate shift. Next with the average image ROIs in the field of view were manually marked. An ROI which covered all the neural structures observed was analyzed further after calculation of ΔF/F. The sum of fluorescence over all pixels within an ROI was converted into a single discrete time series with a sampling period the same as the frame duration. Since we perform spontaneous activity imaging there are no reference time points to define boundaries for obtaining baseline fluorescence. Thus, the baseline fluorescence was taken as the median fluorescence across all the frames of the above time series, and that was used to calculate ΔF/F as a function of frame number or time ( , where, F_k is the fluorescence in the k^th frame and F₀ is the baseline fluorescence or median fluorescence, as majority of the fluorescence hovered around baseline. The ΔF/F was high pass filtered to emphasize the peaks using a first difference filter and then smoothed with a moving average filter with a span of seven frames. The resulting time series was used for further analysis to detect peaks.

In the next stage of analysis, we primarily dealt with events or activity sweeps obtained from the above trace of . Events were detected using MATLAB inbuilt function findpeaks using a minimum peak prominence of 3σ, where σ is the standard deviation of the entire time series. This allowed us to identify events that were relatively large as the chance of false positive was 0.015 and we found that the shapes of the events were consistent across the dataset of the same cell type.

Results

Imaging of SNA from the Drosophila pupal central brains neurons

Three subsets of Drosophila central brain neurons, namely dopaminergic, GABAergic and cholinergic neurons, were labelled by endogenous expression of a genetically encoded calcium indicator (GECI) GCaMP6s using TH-GAL4, Gad1-GAL4 and ChAT-GAL4 respectively. Pupae of specific ages were prepared (see methods, Fig. 1A, Figure S1 and Supplementary video V1) and imaging was performed in a 2-photon microscope (Fig. 1A–C). For each genotype, 6–7 pupae were used for SNA recording from each time Window (described below).

All cell types show periodic active phases of SNA in the central brain at the early stages

The central brain was imaged in live Drosophila pupae (Fig. 1A–C) of 3 different genotypes, TH-GAL4 (n = 19, Fig. 1D), Gad1-GAL4 (n = 22) and ChAT-GAL4 (n = 23), with GCaMP6s expressed in Dopaminergic, GABAergic and Cholinergic cell types respectively. The neuronal subsets labelled by each of these GAL4 driver lines were first identified by crossing them with UAS-mCD8::GFP and pupal brains were dissected at 65 h APF and imaged in Confocal microscope (Fig. D, E and F). The orientation and the imaging planes of the dissected brains is indicated in Fig. 1C; for live 2-photon calcium imaging of the brains, the imaging plane is shifted by about 90˚ due to constraints posed by pupal orientation in the live prep (compare the schematics in Fig. 1C and G). Generally, the target imaging region was 30–50 microns deep from the cuticle surrounding the central complex, and mushroom body regions (grey rectangles in Fig. 1C and G) and often some neural structures, including cell bodies, neurites and synaptic regions, could be visualized (Fig. 1H,I and J). The time course of imaged fluorescence changes in manually constructed ROIs encompassing neural structures (see Methods) was converted into ΔF/F for the entire 1–1.5 h of imaging (Fig. 1K–M) based on the median fluorescence as the reference fluorescence (see Methods). All the imaged subsets of pupal neurons labelled by the three GAL4 lines (Fig. 1H–J) display pulsating higher and lower activities, evident from changes in fluorescence intensities, throughout the imaging period (Supplementary Fig. 2). Each of the genotypes were imaged across pupal ages (in hours APF, Fig. 1F–G, each row) which was grouped into three windows, Window I (50 to 65 h APF), Window II (66–80 h APF) and Window III (81–95 h APF). Typical ΔF/F traces of each strain at each Window (Fig. 1E–G, each row) show distinctive features which are not contaminated by z-movement (Fig. S3) which was minimal in our preparation. At the earliest stage, Window I, all strains are characterized by intermittent periods of high activity followed by long periods of silence (Fig. 1E–G, first row, solid and dashed arrows respectively). Such synchronized activity is common across the cell types and is similar to previously reported PSINA (patterned stimulus-independent neuronal activity), which was observed in the pupal optic lobe of Drosophila²⁶ independent of cell type (pan-neuronal GAL4 driving GCaMP6s in all cells). Thus, like in the optic lobe, the central brain of the Drosophila pupae is also characterized by developmental periodic synchronized neural activity reflected in Ca²⁺ buildup in the neural structures, cell bodies and neurites.

A remarkable departure in terms of spontaneous activity dynamics is observed in the developing central brain as compared to that in the optic lobe^26,27,35. The turbulent stage, observed in the optic lobe, characterized by long phases of activity with relatively short phases of silence, seems largely absent across all 3 cell types of the central brain. The silent periods in the optic lobe’s turbulent stage tend to be at 2.5 min, which is not the case in any of the cell types in the central brain in Window II and in Window III (Fig. 1E–G, bottom two rows).

Event magnitudes follow distinct patterns across age in the different cell types

Active phases of the PSINA are marked by Ca²⁺ transient events (red ticks, Fig. 1D–F), which could be bursts of synchronized spikes across the neural structures. The amplitude and width of the individual events is indicative of the number of synchronized spikes. Ca²⁺ transient events were detected in the ΔF/F traces for each time series recorded (see Methods) and were grouped based on cell type and stage of development. Traces of the mean event waveform with standard error for each of the above groups (Fig. 2A–C) show a stereotypical shape and time course for GCaMP6s based activity. The mean event peaks suggest distinct patterns of developmental modulation of activity peaks across the cell types. Cholinergic neurons show the least sized events of the three cell types across all three pupal developmental time windows indicating lesser synchronized spiking in this population of neurons. Moreover, size of the events of cholinergic neurons decreases over developmental progression (Fig. 2C). The magnitudes of events in dopaminergic neurons are larger than that in GABAergic neurons except in Window III, when they are the same (Fig. 2A and B). GABAergic neurons have a decreased event size in Window II compared to that in Window I while dopaminergic neurons have an increase in event size during the same transition. Comparison of peak heights of events between every pair showed significant difference (ANOVA with Tukey post hoc analysis, p < 0.0001) except TH and GAD in Window III (Fig. 2D).

Fig. 2.

Event magnitude variation across cell types and developmental stages. (A–C) Traces of calcium responses (ΔF/F) depicting the mean event waveforms across developmental stages in different neuronal cell types; (A) dopaminergic (TH + , blue), (B) GABAergic (GAD + , red), and (C) cholinergic (ChAT + , green). Data are shown as mean ± SEM. Scale bars: 30 s (horizontal) and 0.5–2% ΔF/F (vertical, indicated on each panel) (D) Comparison of peak heights of events across cell types and windows are represented by Box and Whisker plot with outliers shown as individual points. Cholinergic neurons showed significantly smaller ΔF/F peaks compared to the other two neuronal cell types (ANOVA with Tukey post hoc analysis, p < 0.0001) and gradual decrease in peak height with increase in pupal age.

Duration and rate of active phases and silence durations are similar across cell types

By detecting events in the Ca²⁺ activity-based fluorescence, the presence of active phases and silent phases in the SNA was observed at all stages of development. To clearly demarcate active phases, the inter-event-interval (IEI) distributions (Fig. 3) were considered. Since Window I of all groups (Fig. 3, blue cdf in all panels) showed the clearest active and silent phases, the cumulative distribution functions (CDFs) of IEIs of each cell type in that phase was considered to obtain a partition in IEIs for short IEIs (within active phase silences: gap between events) and long IEIs (between active phase silences: silent phases). Bimodality in IEIs emerges based on the knee in the Window I IEI CDFs (Fig. 3, blue, all panels) with 120 s as the boundary (Fig. 3, black vertical line). The 120 s boundary is also consistent with exponentially decaying ΔF/F of each event (Fig. 2) that lasted ~ 90 s. The 120 s number matches with the knee in the CDFs as it is the boundary separating the 2 peaks in the bimodal distribution of IEIs. Combined distribution of IEIs of the 3 cell types in Window I showed clear 2 peaks with the lower peak falling to below 50% of the maximum value (0.22 to ~ 0.05) first at 124 s. We used 120 s as the boundary which is the nearest to the above in 10 s multiples. Thus, we considered 3 or more events with IEIs < 120 s to be an active phase and silences of > 120 s (up to 1200 s) were considered as silent phases between active phases. Interestingly the IEI distributions were statistically identical for GAD at the 3 Windows (KS-test), while they were all different from each other (p < 0.05, KS-test) for TH neurons as well as for ChAT neurons (p < 0.05, KS-test).

Fig. 3.

Inter-event-interval distributions. (A–C) Cumulative distribution functions (CDFs) of inter-event intervals (IEIs) of each cell type in over three Windows showing bimodal distribution pattern of IEIs. CDFs of Window I are in blue, Window II in red and Window III in green. 120 s is considered as a cutoff (black vertical line parallel to Y-axis, arrowhead) for distinguishing ‘active phase events’ from ‘sweeps within active phases’; such that three or more events occuring within 120 s is defined as an active phase while silences lasting 120 s to 1200 s were considered as silent phases. IEI distributions of all 3 Windows were statistically identical (KS-test) in case of GAD neurons, but differed significantly from each other for TH neurons (p < 0.05, KS-test) as well as for ChAT neurons (p < 0.05, KS-test).

Having obtained the active phases and silent phases in each neuronal type and pupal age group, differences among the cell types in terms of the activity dynamics were investigated. Unlike that observed in the optic lobe²⁶, all cell types in the central brain showed similar active phase durations across developmental Windows, barring ChAT neurons, which showed larger active phase durations in Window III compared to Window I (ANOVA with Tukey post hoc analysis, p < 0.01, Fig. 4A). Further, within each Window, active phase durations were same except ChAT neurons showing shorter active phases than that of GAD neurons in Window I (ANOVA with Tukey post hoc analysis, p < 0.05, Fig. 4A). Moreover, the silent phase durations were compared among the three age-groups and it was observed that they were significantly longer in Window I than that of the subsequent Windows (ANOVA with Tukey post hoc analysis, p < 0.05, Fig. 4B) in all three cell types, the only feature showing similarity with optic lobe PSINA²⁶. This single resemblance to peripheral SNA is insufficient to align central brain SNA dynamics with those of the optic lobe, as active phase durations clearly do not increase with progress in pupal age.

Fig. 4.

Similarity of active phase and silent phase durations across cell types. (A) Plots showing Window-wise comparison of the active phase durations of each of the three neuronal sub-types. In all Windows, dopaminergic and GABAergic neurons show similar active phase durations, but ChAT shows significantly longer active phases in Window III compared to Window I (ANOVA with Tukey post hoc analysis p < 0.01) and shorter active phases compared to the GAD neurons within the same Window (ANOVA with Tukey post hoc analysis, p < 0.05). (B) Comparison of the silent phase durations among the three age-groups show significantly longer silent phases in Window I than that of the other two later Windows (ANOVA with Tukey post hoc analysis, p < 0.05) which is consistent across all three cell types.

The rates at which active phases occurred across the different age-groups among the different cell types were also not significantly different from each other. The rate varied between 1.9 and 3.7 active phases per hour (1.9, 2.4 and 2.5 in the 3 Windows of dopaminergic neurons type, 3.1, 3.6 and 3.7 in case of GABAergic neurons and 2.3, 3.3 and 2.8 in cholinergic neurons). The lack of differences among the rates of occurrence of the active phases also underline lack of conversion from periodic to turbulent stage in the central brain during the considered period of development.

Rate of events within active phases are mostly similar across cell types and age

The number of events or sweeps occurring in each active phase is also known to vary with developmental progression in the optic lobe neurons. Even though there are largely no differences in the duration of active phases and silent phases in the different cell types and age-groups, the rate of events happening in the active phases could be very different among the groups leading to alterations in SNA dynamics prompting different long-term outcomes in network formation. Thus, the rate of events within active phases were compared (Fig. 5) between cell types and age-groups. In general, all the event rates in active phases were between 1 and 1.5 events per minute in active phases (Fig. 5), in all types of cells and all age-groups. Most pairs of mean event rates were not significantly different (ANOVA with Tukey post hoc analysis, Fig. 5), except only in the cholinergic neurons with Window I rates lower than Window II rates (ANOVA with Tukey post hoc analysis, p < 0.05, Fig. 5), which may be expected during the transition from a periodic stage to a more turbulent one. However, on the contrary, TH neurons showed significantly reduced rates in Window III compared to that in Window II (ANOVA with Tukey post hoc analysis, p < 0.05, Fig. 5), which goes against the above possibility.

Fig. 5.

Event rates in active phases. Plots showing number of events, or sweeps, within the active phases for different neuronal types across developmental stages of the pupae. In cholinergic neurons, rate of sweeps in Window I is lower than Window II rates (ANOVA with Tukey post hoc analysis, p < 0.05), whereas in case of the dopaminergic neurons, sweep-rates per active phase are significantly reduced in Window III compared to that in Window II (ANOVA with Tukey post hoc analysis, p < 0.05). Sweep-rates in dopaminergic neurons were significantly lower than that in GABAergic neurons in Window III (ANOVA with Tukey post hoc analysis, p < 0.05).

Between the different neuronal types, only the active phase-event rates of the dopaminergic neurons were significantly lower than that in GABAergic neurons in Window III and not in any of the other cases and types (ANOVA with Tukey post hoc analysis, p < 0.05, Fig. 5). Thus, again largely, along with active phase duration, active phase occurrence rates, silent phase durations and event rates in active phase, all remained almost the same across cell types and Windows except for a few cases. The above observations only reiterate the major point of difference between peripheral SNA and SNA observed in the central brain with the 3 cell types in the current study.

Variability in event timing lowers with age with cholinergic neurons showing the least variability

Timing of spikes and bursts of spikes play a profound role in plasticity, particularly in developmental plasticity, leading to establishment and refinement of network circuitry^36–38. Since the Ca²⁺ events in the developing Drosophila pupae likely play a role in circuit formation and refinement, there may be subtle differences in the timing of the events of the different types of neurons as average rate was found to be similar. Investigating the event timing through the IEI distributions showed differences in the types and phases of activity (Fig. 3).

Variability in event timing was quantified through coefficient of variation (CV) in the IEIs. Due to the large silence intervals in the IEIs the variability was found to be > 1 in all cases (Fig. 6), with a decay in variability from the initial stage (Window I) to the latter stages (Window III) in all the 3 cell types. While dopaminergic and GABAergic neuronal types showed similar CV (90% confidence interval with bootstraps), the cholinergic neurons showed significantly lower variability or CV in their IEIs compared to the others. The differences in CV with developmental stage and cell types are likely a reflection of the differences in the silent phase duration and numbers. To consider the above further, we took only the IEIs within active phases. These are the intervals between peaks within each of the active phases as defined earlier (Fig. 4). This allowed us to investigate the variability in even timing in only the active phases. In the 9 groups of data, 3 cell types and 3 time Windows each, we found that the calculated CVs showed no difference between the 9 groups, with CV being ~ 0.5 in all cases (not shown). A CV of 0.5 indicates a random pattern of occurrence of peak events in the activity within each active phase. Thus the peaks within the active phase generally did not follow any pattern or sequence.

Fig. 6.

Developmental refinement of spike timing precision across neuronal populations. Cell type-specific trajectories (line graph) highlighting changes in Coefficient of Variation (CV) in three different neuronal populations with respect to progress in developmental stages. The dopaminergic neurons show a gradual decline in variability across all Windows, GABAergic neurons stabilized by Window II, while cholinergic neurons exhibited a marked decrease only during Window III. Solid lines represent mean CV values, with dashed lines indicating 90% confidence intervals obtained by bootstrapping.

Distinct SNA dynamics among cell types revealed with autocorrelation functions

Precise linear temporal structure in the dynamics of a time series can be assessed using autocorrelation functions. Thus, to investigate possible differences in temporal dynamics in the SNA of different cell types the weighted activity (see Methods) over time for every SNA time course were obtained at a resolution of 30 s, by summing activity in overlapping moving windows of 60 s (Fig. 7). Moving from Window I to Window III in the case of each cell type (Fig. 7) changes appeared in the temporal structure. Next the average autocorrelation (using MATLAB ‘xcorr’ function) was calculated as a function of lag, from 0-lag to lag of 40 min for each time series of SNA from each animal. Finally, the average autocorrelation in each Window (Fig. 8, each row) for each cell type (Fig. 8, each column) was obtained along with 95% confidence intervals based on bootstrap resampling with replacement³⁹. The autocorrelation profiles displayed distinct dynamics in each cell type and Window. Peaks in autocorrelation function indicate repetitive or periodic structure embedded in other forms of temporal structure that may be present. The presence and degree of periodicity is measured by the peak away from 0 and its size. The same is obtained by using a Fourier analysis as the Fourier Transform of the autocorrelation function gives the power in different frequencies. Clearly in Window I all three cell types showed periodicity as expected, with GAD neurons showing the longest and ChAT neurons showing the shortest (21 min and 14 min respectively, TH neurons 19 min; Inset Fig. 8, top right, dashed arrows) periodicities. Thus, the different types have distinct periodicity dynamics and hence the three types of neurons would sometimes become synchronized and for the rest of the time remain desynchronized in their combined activities. The above difference can have a profound impact on plasticity and maturation of synapses that connect pairs of the 3 cell types considered.

Fig. 7.

Patterns of activity of different cell types at different Windows. Heatmaps showing spontaneous network activity (SNA) in dopaminergic (TH, top), GABAergic (GAD, middle), and cholinergic (ChAT, bottom) neuronal populations during three pupal developmental Windows (I–III). Each row within a panel corresponds to the recording from an individual pupal brain. The activity levels are represented in arbitrary units (a.u.) from minimum (blue) to maximum (yellow). Time is shown on the x-axis (0 to > 60 min).

Fig. 8.

Autocorrelation of activity profiles of spontaneous network activity across neuronal populations and developmental stages. Average autocorrelation functions of spontaneous network activity (SNA) were computed using the MATLAB xcorr function for dopaminergic (TH, left column, blue), GABAergic (GAD, middle column, red), and cholinergic (ChA, right column, green) neurons across three developmental windows (rows I–III). Autocorrelation was calculated for time lags ranging from 0 to 40 min, and the population averages are shown as thick colored traces. Shaded dashed lines represent 95% confidence intervals. Distinct temporal structures were observed in different cell types and Windows, with peaks in the autocorrelation indicating the presence of periodic or repetitive patterns within the overall activity dynamics. Inset (top right) shows an overlay of autocorrelation functions for all three neuronal populations, comparing periodicities in Window I; dashed arrows indicate the average periodicities of each celltype in each color. Black lines under the peaks (Row 3, Window III, GAD and ChA, indicate the presence of peaks with 95% confidence as the CI at peak do not intersect the black line connecting minima beside peaks.

With the progression of time into Window II, all three cell types show reduction in the periodicity marked by a decrease in the peak height, while maintaining the same location of peak indicated by the time of maximum value of peak in the autocorrelation function away from 0. Thus, with time, although there is an underlying periodicity in the activity, it is smeared by many other events that are not completely repetitive in nature. Finally at the third Window, TH neurons lose their periodicity while GAD and ChAT neurons retain it (Fig. 8, peak over black lines). Thus, if absence of periodicity is in anyway a marker of progress towards maturation, then dopaminergic circuits likely mature the earliest. It is yet to be elucidated how such periodicities in SNA progress in the other cell types beyond Window III.

Discussion

Developing neural circuits exhibit patterned spontaneous neuronal activity (SNA) that plays instructive roles in wiring and maturation of neural circuits across species¹. In our study, we systematically characterized the patterns of SNA in three crucial neuronal populations, namely dopaminergic, GABAergic, and cholinergic neurons in the developing central brain of Drosophila.

Despite the diversity of neurotransmitter identities, dopaminergic, GABAergic, and cholinergic neurons all displayed periodic active phases of SNA interspersed with silent phases. The durations of active and silent phases, as well as the rate of phase occurrence, were strikingly similar across cell types. Similarly, the rate of events within active phases (sweeps) showed little variation across cell types and pupal stages. These shared dynamics suggest that central brain SNA may be governed by a global regulatory mechanism. Our observations correlate with the patterns of SNA observed in developing vertebrate circuits, where activity waves coordinate refinement across distributed networks¹. It remains to be determined whether similar roles are played by the patterned SNA in pupal neurons in assigning circuit accuracy culminating into behavioural precision in Drosophila.

Some features of the developmental SNA diverged across neuronal types in the pupal brain. Firstly, the event magnitudes followed distinct developmental trajectories, with cholinergic neurons showing the greatest temporal precision by late pupal stage which may indicate a broader role of excitatory cholinergic transmission in shaping the dopaminergic and GABAergic signals. Secondly, variability in spike/event timing decreased with age at different rates in different populations, which could be due to the differential maturation rates of distinct neuronal populations. Overall, all three neuronal populations showed reduced variability in spike timing with maturation, but the trajectory differs by cell type, dopaminergic and GABAergic neurons reached their minimum variability at Window II, whereas cholinergic neurons continued to refine till Window III. Thirdly, sweep rate per active phase of cholinergic neurons increase to maximum in Window III, possibly indicating shift from periodic to more ‘turbulent-like stage’, although dopaminergic neurons showed the reverse pattern, i.e., a decrease in sweep rate in Window III, indicating that the age-dependent maturity of SNA has cell-type specific patterns which may have specific implications in circuit maturation. The coexistence of general and cell type-specific features suggests that central brain SNA arises from a combination of global pattern generators²⁷ and local, cell type–specific modulators.

The cell type-specific SNA patterns in excitatory, inhibitory and modulatory neurotransmitter systems prompt us to ask relevant questions which might be addressed in future studies. It would be valuable to determine the interrelation between the three neuronal types and whether SNA features are influenced by other neuronal types through synaptic transmission during development. One approach would be to measure SNA simultaneously across multiple neuronal populations and finding whether their firing happen in coherent or in asynchronous fashion. Moreover, the power of Drosophila genetics might be employed to find the interdependence of different cell types by blocking SNA firing in one subset and simultaneously recording from another subset.

During metamorphosis, the nervous system undergoes a large-scale reconfiguration including circuit composition, intrinsic excitability, neurite pruning, branch formation, and synaptic architecture. During mid-to-late pupal stages, there is robust formation of new synapses accompanied by activity-dependent refinement. Changes in neurotransmitter receptor composition, synaptic strength, and connectivity patterns may contribute to the maturation of coordinated spontaneous activity that are observed in this study. Mammalian neurons are known to undergo developmental changes in neurotransmitter phenotype or receptor expression profiles, leading to shifts in cholinergic, GABAergic, or dopaminergic signaling which results in altered excitation-inhibition balance. It remains to be seen whether similar changes in pupal neurons determine the gradual changes in SNA patterns. Moreover, glial cells actively participate in pruning, regulate extracellular ion homeostasis and also perform neurotransmitter clearance. Developmental changes in neuro-glial would also influence neuronal excitability and spontaneous firing patterns.

Spontaneous activity has long been recognized as a driver of developmental plasticity, influencing synaptic refinement and connectivity prior to sensory experience⁴⁰. The observed reduction in timing variability with age indicates a progressive stabilization and refinement of SNA driving the synchronized firing which contributes to the maturation of the circuit. This maturation process mimics the spontaneous activity seen in mammalian cortex and retina, where it shifts from noisy to highly organized pattern¹. The fly system provides a unique model to assess the contribution of the precisely patterned SNA in activity regulated developmental gene expression, fostering neuronal development and synaptic plasticity. As neurodevelopmental disorders like Autism are associated with atypical SNA pattern, the Drosophila SNA will be a perfect system to check whether altering SNA pattern affects behavioral accuracy.

Previous work on Drosophila developmental SNA has emphasized in the optic lobe and neuromuscular systems, where stereotyped waves of activity guide topographic mapping^26,41. Our results reveal that central brain SNA shares certain global features with the optic lobe but also displays cell type-specific temporal organization. One key component of the optic lobe SNA was the gradual transition from periodic to turbulent firing pattern; the central brain SNA begins with very similar periodic firing, the frequency of which increases in later stages, but it does not transition to turbulent form till the very late pupal stage. This distinction may highlight the asynchronous rate of maturation in optic lobe and central brain, potentially reflecting important differences in their contributions to circuit development. It is tempting to examine if SNA goes on to full turbulent form later during early adult stage, the critical period of experience dependent plasticity.

One major limitation in our study is that the entire 2-photon-based recording of SNA was restricted to the neurons near the dorsal surface of the central brain, which may not reflect the SNA patterns of various other brain regions. Also, the three GAL4 driver lines used in this study label large subsets of neurons; it would be valuable to detect SNA patterns arising in specific brain regions and find their variabilities using GAL4 lines labelling fewer neurons.

Conclusion

Together, our findings reveal a unique pattern of developmental spontaneous neuronal activity in the Drosophila central brain neurons which is broadly invariant across neuronal subtypes, with specific points of divergence with progression of development. The organizational principle comprises invariable periodic active phases interspersed by gradually widening silent phases. Within the active phases, the firing events display a minor trend of increase in all neuronal types. Age dependent refinements in magnitude, variability, and temporal motifs occur in a cell type-specific manner. Such an organized SNA pattern may be predicted to orchestrate both coordinated network maturation and tailored circuit-specific development. Overall, our observations emphasize the importance of SNA not merely as a hallmark of early development but also as a prime factor for sculpting refinement of neuronal circuits during critical windows of brain maturation.

Author contributions

AD, SB and SD conceived the project, SD performed all experiments, SD, AD and SB analyzed data. AD, SB supervised the work. SB, AD and SD wrote the manuscript.

Funding

Open access funding provided by Indian Institute of Technology Kharagpur. AD laboratory received research grants from Govt. of India: Science and Technology Research Board- DST (ECR/2017/002963), Department of Biotechnology Ramalingaswami fellowship (BT/RLF/Re-entry/11/2016) and Department of Biotechnology (BT/PR53805/BMS/85/101/2024); from Govt. of West Bengal: WBDSTBT Core Research Grant (2297(SANC.)/STBT-13015/10/2024); and intramural research grant from IIT Kharagpur, India. SB laboratory work is supported by Challenge Grant from IIT Kharagpur. This work was also supported by the Department of Science and Technology, India through DST/CSRI/2021/340 awarded to SB. SD received Senior Research Fellowship for PhD from Council for Scientific and Industrial Research (CSIR, Govt. of India). We acknowledge the Confocal Microscope and the 2-photon microscope purchased under DST-FIST grant conferred on the School of Bioscience, File no. SR/FST/LS-I/2019/595(C) and intramural funds to the Department of Biotechnology respectively.

Data availability

All data in the study will be made available by corresponding authors on reasonable request.

Declarations

Competing interests

The authors declare no competing interests.