Introduction

During nervous system development, critical periods (CPs) are transient windows of heightened plasticity, which are fundamental to the emergence of appropriate network function. Because of this, networks are particularly vulnerable to activity perturbations during a CP, which can result in lasting structural and/or functional changes, while comparable perturbations outside the CP are less impactful (Coulson et al., 2022; Reh et al., 2020). The functions and concepts of CPs have traditionally been investigated in the visual system of mammals (Hensch, 2005; Hensch and Quinlan, 2018; Hubel and Wiesel, 1970) and imprinting of birds (Hess, 1975; Horn, 2004). More recently, CPs have also been identified and investigated across a range of species and neuronal ensembles, from networks underlying the development of speech (Harrison et al., 2005) to motor behaviour of rats (Soiza-Reilly and Azcurra, 2009), fish (Hageter et al., 2023) and insects (Dombrovski and Condron, 2021; Giachello and Baines, 2015). The presence of a CP in developing networks is therefore increasingly considered a universal phenomenon. The precise changes within networks and associated mechanisms, as they occur during normal development, as well as those resulting from perturbations during a CP, are not well understood. This is in part due to the complexity of many of the experimental model systems, notably mammalian cortical systems that have been the predominant model for studying CPs.

To overcome this complexity, we have utilized the comparatively simpler nervous system of the fruit fly larva, Drosophila melanogaster, with which to study CPs. Though an insect, Drosophila displays CPs with features that suggest commonalities to mammalian CPs (for review see: Coulson et al., 2022). Currently, the best characterised CP during Drosophila development is one that occurs during embryogenesis, linked to the emergence of normal locomotor network output (Carreira-Rosario et al., 2021; Corke et al., 2025; Coulson et al., 2022; Crisp et al., 2008, 2011; Ackerman et al., 2021; Fushiki et al., 2013; Giachello et al., 2021; Giachello and Baines, 2015; Hunter et al., 2024; Sobrido-Cameán et al., 2025; Williamson et al., 2021; Zeng et al., 2021). This experimental system benefits from the comprehensively annotated connectome of the Drosophila larval nervous system (Giachello et al., 2020; Heckman and Doe, 2022; Valdes-Aleman et al., 2021; Winding et al., 2023) as well as an unparalleled set of genetic tools for manipulating identified neurons with both spatial and temporal precision (Simpson, 2009). Combined with imaging and electrophysiological techniques, this model system enables investigation of cellular responses to CP manipulations at the level of identified cells and synapses (Carreira-Rosario et al., 2021; Coulson et al., 2022; Crisp et al., 2008, 2011; Fushiki et al., 2013; Giachello et al., 2021; Giachello and Baines, 2015; Hunter et al., 2024; Sobrido-Cameán et al., 2025; Williamson et al., 2021; Zeng et al., 2021).

Neuronal activity manipulations during the CP of Drosophila embryos have been shown to cause decreased network stability. This manifests in significantly longer times required for larvae to recover from a strong activity challenge, such as electric shock-induced seizures (Coulson et al., 2022; Giachello et al., 2021; Giachello and Baines, 2015; Hunter et al., 2024, 2021). The premise that balanced network activity during the CP is necessary for stable, resilient networks to form has been explicitly supported by findings from our recent studies (Giachello et al., 2021; Giachello and Baines, 2015; Hunter et al., 2024). By extension, rectifying activity manipulations during identified CPs could have therapeutic potential (Nardou et al., 2023).

While experimental perturbations of neuronal activity during development have been a useful methodological approach for studying CPs as a phenomenon and the associated biological mechanisms, we tested if temperature change, in the form of 32°C heat stress, might be an alternative perturbation stimulus, with ecological relevance. For example, most animals are not warm-blooded and in poikilothermic animals, such as Drosophila, changes in ambient temperature, daily as well as seasonal, lead to changes in behavioural activity, likely caused by changes in neuronal and network activity (Lee and Montell, 2013; Powsner, 1935). Studies from the crustacean stomatogastric nervous system showed that these minimal networks produce a characteristic, rhythmic firing pattern, which is robustly maintained across a range of external temperatures. Importantly, neuronal network properties were adapted relative to the long-term temperatures that crabs had experienced, resulting in significantly different upper “crash temperatures” at which networks ceased to function (Marder, 2023; Marder and Rue, 2021). In Drosophila, the temperature experienced by late embryos determines larval locomotor output, including crawling speed and social feeding behaviour (Williamson et al., 2021). Similarly, studies on the Drosophila adult visual and olfactory systems showed that neuronal wiring properties and associated behaviours of adults are determined by the temperature experienced during a second CP of Drosophila, which takes place during pupariation (Chodankar et al., 2020; Kiral et al., 2021; Züfle et al., 2025), a phase of extensive network remodelling (Baumann et al., 2024; Leier et al., 2025; Lowe et al., 2023; Nelson et al., 2024).

In this study, we used heat stress of 32°C as an ecologically relevant stimulus with which to probe the embryonic CP. To gain insights at the level of single, identified connected cells, we focused on the neuromuscular system of the Drosophila larva, and found that the components of this network respond differently to heat stress experienced during embryonic development. Specifically, we find that elevated temperature increases neuronal activity and, when experienced during the locomotor CP, causes long term reduced network stability, as seen for neuronal activity manipulations (Giachello et al., 2021; Giachello and Baines, 2015; Hunter et al., 2024). At the neuromuscular junction (NMJ), embryonic heat stress experience causes presynaptic terminal overgrowth and changes in postsynaptic neurotransmitter receptor composition, though synaptic transmission remains normal. In contrast, within the central circuitry, we find that embryonic heat stress leads to enhanced synaptic drive from premotor interneurons and, potentially as a response, reduced excitability of motoneurons. Overall, this study presents a first comprehensive analysis of how several interconnected elements of a defined locomotor network develop in response to CP perturbation by heat stress, an ecologically relevant stimulus.

Results

Temperature manipulation during the locomotor CP leads to an unstable crawling network

We had previously identified a CP of nervous system development in late embryogenesis, from 17 – 19 hrs after egg laying (AEL), equivalent to 80-90% of embryonic development (Giachello and Baines, 2015). Perturbation of network activity in late embryogenesis, during this CP, pharmacologically or optogenetically, leads to the formation of an unstable, seizure prone network, as revealed by acute activity challenges. For example, an electric shock administered to late larvae, which develop from CP-manipulated embryos, causes seizures in those larvae, with recovery times that are significantly longer compared to unmanipulated controls. These pharmacological and optogenetic large-scale activity manipulations, while useful to study the embryonic CP, are artificial. We therefore asked to what extent changes in environmental conditions, that embryos might normally encounter, could also impact on network development. As poikilothermic animals, Drosophila development strongly depends on ambient temperature (Powsner, 1935). Several studies have shown that temperature experiences during development alter neuronal wiring as well as behaviours in larvae and in adult flies (Kiral et al., 2021; Williamson et al., 2021; Züfle et al., 2025). Increases in ambient temperature can lead to increased activity in the locomotor network (Oswald et al., 2018; Sigrist et al., 2003; Zhong and Wu, 2004). We verified this by functional imaging in isolated CNS, measuring spontaneous calcium transients from motoneurons, via GCaMP8f selectively expressed in aCC motoneurons using R94G06-GAL4. Stepwise increases in temperature led to increases in the frequency of rhythmic bursts. In contrast, CNS kept at the control temperature of 25°C for the duration of functional imaging (600 seconds) tended to show a decrease in frequency over time (Fig. 1A). These experiments suggest that increases in ambient temperature, above 25°C, lead to concomitant increases in locomotor network activity.

Figure 1:

Next, we tested the effect that a transient experience of 7°C above the evolved temperature preference of 25°C (Ashburner et al., 1983), namely 32°C, might have on nervous system development. We chose 32°C as this is known to be experienced by Drosophila melanogaster (Williamson et al., 2021). We transiently exposed embryos for two hours to 32°C heat stress either before, during, or after the previously characterised embryonic CP for locomotor network development (17 – 19 hrs AEL) (Giachello and Baines, 2015), then returned them to the control temperature of 25°C. After four days, at the end of larval development, we measured as a proxy for network stability the speed of recovery from electroshock-induced seizures. Seizure recovery was unchanged, at control levels, following transient heat stress before or after the embryonic CP. In contrast, animals that had experienced 32°C heat stress during the embryonic CP (i.e. between 17 – 19 hrs) had significantly prolonged recovery times, indicative of reduced network stability (Fig. 1B). This shows that when Drosophila embryos experience a 32°C heat stress during the embryonic CP of the locomotor network it results in altered developmental outcomes, notably the formation of sub-optimal, less stable networks, similar to the effect of network-wide activity perturbations during this developmental window (Hunter et al., 2024; Giachello et al., 2021; Giachello and Baines, 2015).

To investigate structural and functional changes that result from a CP heat stress perturbation we first focused on the well characterised, and easy to access, peripheral neuromuscular junction (NMJ). Specifically, we focused on the NMJ of dorsal acute muscle 1 (DA1), which is innervated by the aCC motoneuron, a cell that has been the focus of our previous studies (Fig. 2A) (Baines et al., 2001, 1999; Fujioka et al., 2003; Hunter et al., 2024; Landgraf et al., 2003; Giachello et al., 2021; Giachello and Baines, 2015 Sobrido-Cameán et al., 2025). Because temperature manipulations outside the CP did not lead to an unstable network (Fig. 1A), we adopted a simplified experimental paradigm of putting embryos from a 6h egg collection to 32°C for 24h. This produces outcomes comparable to heat stress exposure centred around the CP, while ensuring that all embryos go through their CP during heat stress irrespective of minor variations in developmental timing or due to retention of fertilised eggs. We find that heat stress during embryogenesis leads to a significant overgrowth of the aCC motoneuron terminal (composed of type Ib boutons) by the late larval stage. This increase in the number of boutons is accompanied by a corresponding increase in active zones, while overall active zone density is not altered (Fig. 2B). On the postsynaptic side, embryonic heat stress causes an altered receptor composition: a selective reduction of ion channels containing the larger conductance GluRIIA subunit. Across the aCC NMJ, this leads to a reduced ratio of the larger conductance GluRIIA versus the lower conductance GluRIIB subunit-containing glutamate receptors (Fig. 2C).

Figure 2:

In recognition of the potential caveat that prolonged heat stress during most of embryogenesis could lead to compound effects, we additionally carried out 32°C heat stress exposure of embryos during shorter time windows, followed by rearing at the control temperature of 25°C. We then measured postsynaptic GluRIIA vs GluRIIB levels at the aCC NMJ on muscle DA1 of late wandering third instar larvae. We found that the GluRIIA reductions caused by prolonged heat stress throughout most of embryogenesis is recapitulated by a shorter, 3h exposure to 32°C, though during a different developmental window, from 13-16 hrs AEL. Heat stress before or after this 13-16 hrs AEL developmental window had no effect on the composition of the postsynaptic glutamate receptors (Fig. 3). GluRIIB subunit levels, in contrast, remained unaffected by embryonic exposure to 32°C heat stress. The period of 13-16 hrs AEL is the developmental window during which the embryonic body wall muscles become electrically active and start to exhibit myogenic contractions (Broadie and Bate, 1993; Crisp et al., 2008). Thus, the maturation of the embryonic body wall muscles precedes the CP of the central locomotor circuitry (17-19 hrs AEL), when central neurons develop their electrical properties (Baines and Bate, 1998). To our surprise, we found that NMJ size is impacted by heat stress during an even earlier window of muscle development (9-12 hrs AEL), during which myoblast fusion occurs (Bate et al., 1993) (Fig. 3). Overall, we found that transient embryonic heat stress of 32°C leads to reproducible changes in NMJ development during subsequent larval stages. Our observations show that postsynaptic glutamate receptor composition is specified during the developmental window that is characterised by electrical maturation of the body wall muscles, preceding by several hours the equivalent maturation of the innervating motoneurons.

Figure 3:

Despite morphological changes, larval NMJ physiology is not altered

In view of the prominent changes in NMJ growth, morphology and postsynaptic glutamate receptor composition that result when heat stress (32°C) is experienced during embryonic development, we next assessed NMJ synaptic physiology. Because the NMJ is a composite structure that develops through interactions between the postsynaptic body wall muscle with presynaptic terminals from motoneurons, whose cell bodies are located within the CNS, we proceeded using an experimental paradigm of prolonged heat stress (32°C) during most of embryogenesis, so as to include CPs of both body wall muscles and the central locomotor circuitry. We used the two-electrode voltage clamp (TEVC) technique to assess muscle synaptic physiology, initially measuring single pulse transmission across a gradient of external calcium concentrations. Despite the marked structural changes in heat stress manipulated animals, single pulse transmission at the NMJ was not significantly different between embryonically manipulated (32°C experience during embryogenesis) and controls that had been reared at 25°C throughout (Fig. 4 A&B). Quantification of the number of release sites utilised per action potential and quantal size (Jusyte et al., 2023) also showed no statistically significant difference between manipulated and control specimens (Fig. 4C). We quantified the size of the readily releasable pool (RRP) of synaptic vesicles by recording responses to a high frequency train of action potentials (Fig. 4D), then plotting the cumulative amplitude of responses to the first 60 pulses and linearly back extrapolating through the last 20 pulses to y=0 (Fig. 4E). This also did not show a significant difference between manipulated and control animals (Fig. 4F). Though we noticed that during the 60Hz pulse train, NMJs in embryonic heat stress-manipulated specimens exhibited an increased facilitation during the first 10 pulses, indicating a slight reduction in release probability of the presynaptic terminal (Fig. 4D). Last, we tested if the NMJ remained able to adjust homeostatically to acute changes in synaptic transmission. We utilized the well-established method of acutely blocking glutamate receptors with Philanthotoxin (PhTx), a GluRIIA blocker, to induce presynaptic homeostatic potentiation (Frank et al., 2006; Genç and Davis, 2019; Müller and Davis, 2012). Acute GluRIIA blockade by a 10 min incubation with PhTx caused a significant reduction mEPSC amplitude while evoked EPSC amplitude remained unchanged in both controls and heat stress-manipulated specimens. Calculating quantal content shows that in response to acute GluRIIA blockade, both control and manipulated animals are indeed able to induce presynaptic homeostatic potentiation by increasing quantal content (Fig. 4G-Giii). While establishing baseline conditions, we found mEPSC amplitudes and quantal content to be comparable between controls and experimentals Fig. 4G-Giii). This was contrary to expectation considering the overall reduction in the GluRIIA/B ratio that results from an embryonic heat stress manipulation. Overall, we conclude that despite the significant structural changes that emerge by the late larval stage following an embryonic 32°C heat stress experience, synaptic physiology at the late larval NMJ remains normal. Moreover, when considering the unstable network phenotype, as manifest by increased recovery times from electroshock-induced seizures, our finding that synaptic transmission at the NMJ remained normal suggests that the network instability likely originates from changes within the central locomotor circuitry.

Figure 4:

Motoneurons display reduced excitability and receive increased synaptic input following embryonic perturbation by heat stress

We noted that following embryonic exposure to 32°C heat stress, motor axons in late third instar larvae required an increased stimulation voltage to reliably induce action potentials. This hinted at motoneurons in these larvae having acquired a state of reduced excitability. To directly assay motoneuron excitability, we used whole cell patch clamp to measure action potential firing of aCC, evoked by injection of current steps (4pA). These measures were made in the presence of mecamylamine (200 µM in external saline) to block endogenous cholinergic excitatory synaptic drive. We found that in late third instar larvae, the aCC motoneuron had significantly decreased excitability following an embryonic heat stress experience of 32°C (Fig. 5A). These results are comparable to, and validate, results obtained after optogenetic locomotor network activity manipulation during the embryonic CP of the locomotor network (17-19 hrs AEL) (Giachello and Baines, 2015). In previous studies, we had shown with recordings from aCC that its synaptic drive is enhanced following transient network activity manipulations during the embryonic CP, using optogenetic or pharmacological means to induce network over-excitation or genetic persistent overactivation of neurons, as in para^bss seizure mutants (Giachello and Baines, 2015; Hunter et al., 2024). This manifests as an increase in the duration (though not amplitude) of endogenous excitatory cholinergic synaptic currents, termed spontaneous rhythmic currents (SRCs) (Giachello et al., 2019; Giachello and Baines, 2015). We asked if heat stress experienced during embryonic development led to similar changes. Indeed, in late larvae that had experienced heat stress during embryonic development, SRCs recorded from aCC motoneurons were significantly increased in duration, though with amplitude unchanged (Fig. 5Bi), i.e. phenocopying the lasting effects of network activity manipulations during the embryonic CP. Note these SRC recordings were undertaken in the absence of mecamylamine and suggest an increase in premotor network activity.

Figure 5:

To further explore changes in the premotor network, we recorded synaptic drive to aCC under whole cell patch clamp configuration. We focused on one of the most the strongly connected excitatory premotor interneurons, termed “A27h”. Because motoneurons are quiescent without synaptic drive, a small amount of current was injected into aCC to maintain a modest and consistent mean firing frequency of ∼10-20 Hz. We then assayed how this mean firing frequency changed upon optogenetic activation of the A27h premotor interneuron (Hunter et al., 2024). As might be expected, and as per our previous study (Hunter et al., 2024), acute optogenetic activation of the A27h premotor excitor resulted in increased aCC firing frequency. Significantly, this increase was greater in larvae that had been exposed to heat stress during their embryonic development, as compared to controls (Fig 5.Bii). This observation demonstrates that heat stress during embryonic development results in increased excitatory drive from central premotor interneurons, and it suggests that this might then countered by the postsynaptic motoneurons through reduction in their excitability.

Adjustments of the pre-motor synaptic drive onto the aCC motoneuron result in normal endogenous firing patterns

If the reduction in excitability displayed by the aCC motoneuron was a homeostatic response to the increased premotor synaptic drive, then one might predict this to be a mechanism to maintain the endogenous level of motoneuron activity. To test this, we used cell-attached patch recordings from aCC motoneurons, in isolated nerve cords of late third instar larvae.

This extracellular recording mode allows quantification of endogenous action potential firing without disturbance of the intracellular composition of a neuron. As hypothesized, our recordings show that no significant differences in endogenous motoneuron activity result from an embryonic experience of heat stress, as measured by burst rate, burst duration, or action potentials per burst (Fig. 6A). Next, to test if this translates to normal muscle output, we recorded spontaneous synaptic drive from muscles in a semi-intact preparation with an intact nervous system attached. Current clamp recordings from the DA1 muscle, while supressing muscle contractions using nifedipine (50µM) (Dyson et al., 2022; Kratschmer et al., 2021), showed no significant differences in spontaneous burst frequency or duration, nor the overall time spent bursting (Fig. 6B), between controls and embryonic heat stress-manipulated specimens. Therefore, we conclude that, at least in this locomotor system, CP manipulations might act primarily on the cells of the premotor circuitry, namely excitatory premotor interneurons as tested in this study. Changes in the motoneurons could thus be seen to constitute secondary homeostatic adjustments in response to those premotor network changes.

Figure 6:

Propagation speed of neuronal activation waves is reduced following embryonic heat stress perturbation, mirroring reduced crawling speed

To establish how the excitatory premotor network is affected by a transient embryonic 32°C heat stress manipulation we conducted functional imaging of the segmentally repeated A27h premotor interneurons. Specifically, we targeted expression of the calcium indicator GCaMP8f to A27h premotor interneurons. Then, in isolated third instar nerve cords, we recorded the propagation of neuronal activation waves, indicative of fictive locomotion, across abdominal segments (Fig. 7A) (Streit et al., 2016; Pulver et al., 2015). These experiments show that embryonic 32°C exposure leads to a significant reduction in the speed with which these neuronal activity waves travel from posterior to anterior abdominal segments, as compared to controls, and mirrors the decrease in larval crawling speed caused by the same embryonic manipulation (Fig. 7B). Because our electrophysiological analyses showed that motoneuron firing patterns and output to the muscle remain normal after an embryonic 32°C manipulation (Fig. 4-6), these results further support the hypothesis that it is the lasting changes to the central premotor circuitry, rather than to the motoneurons, which are responsible for the behavioural changes, in crawling behaviour, and likely also for network instability.

Figure 7:

The CP of the locomotor network determines behavioural outcomes

Having identified consecutive phases of embryonic development that represent CPs for the body wall muscles and NMJ development, followed by a CP for the central locomotor circuitry (Fig. 3), we asked which of these CPs shaped larval crawling behaviour. To this end, we subjected embryos to 32°C heat stress at different windows of their development, separating the periods that principally affects NMJ development (prior to 17 hours AEL) from the subsequent 2-hour window during which the development of the locomotor circuitry is vulnerable to neuronal activity and heat stress manipulations (17-19 hours AEL) (Giachello and Baines, 2015). Assaying crawling behaviour of third instar larvae showed that crawling speed is reduced only when the embryonic CP of the locomotor network (17-19 hours AEL) had been subjected to heat stress. This demonstrates that it is the impact on the developing central circuitry, rather than NMJ formation, which causes behavioural changes in terms of crawling speed (Fig. 8A) and of network stability, as assessed by seizure recovery times (Fig. 1).

Figure 8:

We wondered whether larvae that emerge from heat stressed embryos crawl more slowly simply because they are no longer able to crawl at the normal speed. To test this, we subjected embryonic development to 18, 25 or 32°C and, after hatching, reared larvae at 25°C until the mid-third instar stage (72 hrs ALH). We then assayed how these larvae adjusted their crawling speed in response to an acute change in ambient temperature, when placed into an arena equilibrated to 18, 25, 29, 32 or 35°C (Fig. 8B). This clearly showed that while embryonic heat stress leads to a significant reduction in the default crawling speed, larvae retained the ability to accelerate in response to an acute increase in ambient temperature, as they normally do. Moreover, animals that experienced a higher temperature during embryogenesis retained the ability to increase their crawling speed at higher temperatures (e.g. 32 or 35°C) than those with cooler embryonic experiences, whose maximum speed plateaued at 29°C (Fig. 8B). These data suggest that while an embryonic experience of heat stress reduces the default crawling speed, it preserves a response range at higher temperatures, which might therefore potentially be adaptive to larval life at higher temperatures.

Discussion

An experimental model system with which to study CP biology

In this study we utilised the Drosophila embryo and larva as an experimental model with which to study the role of CP-regulated plasticity across interconnected cells within a defined circuit, from a premotor excitatory interneuron, to motoneuron to muscle. We have identified 32°C heat stress as a CP stimulus that has ecological relevance, as a temperature experienced in the wild (Williamson et al., 2021). Transient exposure of embryos to 32°C during the previously defined CP for the developing CNS (17-19 hours after egg laying) leads to lasting changes in subsequent nervous system development, resulting in decreased crawling speed of third instar larvae and in reduced network stability, reflected by an increased susceptibility to electroshock-induced seizures. As is characteristic for CPs, the same manipulation before or after this defined developmental window has no lasting effect on nervous system properties or animal behaviour. A well-known way by which increases in temperature impact on neurons is their activity. Using calcium imaging of motoneurons we showed that stepwise increases in ambient temperature cause acute increases of fictive rhythmic locomotor network activity in isolated nerve cords. This agrees with past studies, which have demonstrated the well-established link between temperature and neuronal activity in this same experimental system (Zhong and Wu, 2004). It is therefore not surprising that 32°C heat stress during the embryonic CP of nervous system development causes outcomes that are comparable to those created by neuronal activity manipulations (pharmacological or genetic) during this developmental window (Giachello et al., 2021; Giachello and Baines, 2015; Hunter et al., 2024).

Different cell types respond differently to CP manipulations

A particular strength of the Drosophila larval motor system is the accessibility of interconnected network elements. We focused on a small part of the locomotor network: an excitatory premotor interneuron (A27h) connected to a motoneuron (aCC), and its postsynaptic body wall muscle (DA1). Following transient application of heat stress during embryonic development, we could identify that within this section of the locomotor network each of these three connected nodes responded differently. We observed increased excitatory drive from A27h interneurons to aCC motoneurons, which might reflect changes in A27h neuronal excitability and/or its synaptic transmission onto motoneurons. Potentially in response to these changes in the premotor network, aCC motoneurons displayed a decrease in excitability, which facilitated normal motoneuron firing patterns to be maintained. In contrast to these changes within the central circuitry, at the peripheral NMJ terminal, we documented normal synaptic transmission properties, despite a significant presynaptic terminal overgrowth. The latter might be a secondary adjustment to a decrease in transmitter sensitivity by the muscle, which showed reduced levels of the high conductance subunit GluRIIA.

The first observation to make is that these responses to embryonic heat stress, which manifest during subsequent larval stages, are reproducible. This suggests that there might be direct cause-effect relationships at the cellular level, which can now be investigated in detail using this experimental model. Building on these findings, we thus identified mitochondrial reactive oxygen species (ROS) and Hypoxia-inducible factor-1 alpha (HIF-1α) as heat stress-activated signals required for the long-lasting changes of cellular properties that result from embryonic CP heat stress (Sobrido-Cameán et al., 2025). An alternative, not mutually exclusive interpretation is that activity patterns in this part of the network might be robustly canalised so as to lead to reproducible outcomes. Second, the changes in neuronal excitability and synaptic transmission that we have documented are compatible with a temporal sequence, or a functional hierarchy, during network maturation. For example, following an embryonic heat stress, premotor interneurons, such as A27h, potentially part of the central pattern generator (Carreira-Rosario et al., 2021; Zeng et al., 2021), increase their collective synaptic drive to postsynaptic motoneurons, which in turn reduce their excitability (Fig. 5). Along these lines we interpret the changes in motoneuron properties that we have characterised as homeostatic responses by motoneurons to the increase in presynaptic excitatory drive from premotor interneurons. This perspective implies a hierarchy, whereby the specification of premotor interneuron properties dominates that of their downstream integrators, the motoneurons (Fig. 9). This is supported by previous findings of motoneurons flexibly adjusting their electrical properties in response to changes in presynaptic excitatory drive (Baines et al., 2001). However, the precise temporal and causal relationships of cell type-specific adjustments during the CP of this network remain to be established. In the developing mouse neocortex, sequential cell type-specific maturation windows have recently been shown. There, somatostatin-expressing inhibitory interneurons differentiate early, and promote the maturation of network activity patterns through inhibition of the later differentiating parvalbumin-expressing inhibitory interneurons (Mòdol et al., 2024).

Figure 9:

NMJ structural development is affected by CP manipulations, but not synaptic transmission

Despite embryonic heat stress leading to significant NMJ structural changes, in terms of presynaptic terminal overgrowth (an average increase of active zone number by 30%) and overall reduction of GluRIIA (by 50%), evoked transmission resulting from the firing of single action potentials in the motor axon remained normal. Distal boutons had previously been shown to house release sites that are preferentially recruited under low frequency stimulation regimes, such as used for making NMJ recordings (Peled and Isacoff, 2011). Speculatively, preferential maintenance of GluRIIA in distal boutons, might have explained maintenance of normal synaptic transmission at low frequency stimulation regimes. However, we found no evidence for this. On the contrary, in embryonic heat stress-manipulated specimens, GluRIIA levels were comparably reduced in both proximal and distal boutons (data not shown). An alternative explanation for maintained synaptic transmission in the face of increased active zone number is a reduction in release probability. Although we did not conduct functional imaging to investigate this directly, an indirect indication for this is the increased facilitation in animals that had experienced embryonic heat stress, which indicates a reduction in presynaptic release probability across the whole NMJ (Fig. 4D).

On reflection, to find evoked synaptic transmission at the NMJ not significantly affected by CP manipulations, is perhaps not unexpected. A well-established feature of this particular synapse is that its transmission is tightly regulated so as to maintain constancy through the deployment of multiple NMJ plasticity mechanisms, including presynaptic homeostatic potentiation (PHP), which we have shown to function effectively even after an embryonic heat stress manipulation (Fig. 4G) (Delvendahl and Müller, 2019; Frank et al., 2020, 2006; Goel and Dickman, 2021).

Distinct developmental windows with sensitivity to heat stress suggest sequential network maturation

It is unlikely that structural changes at the NMJ contribute to making the nervous system less stable, or that they effect CP-regulated changes to larval behaviour, such as crawling speed or social feeding (Williamson et al., 2021). Based on previous CP manipulations targeted to various neuronal subsets (Giachello et al., 2021), it seemed more likely that behavioural phenotypes, including reduced crawling speed and increased recovery times from electroshock-induced seizures, would be the consequence of changes in the central circuitry. Indeed, we found that the developing body wall muscles have an earlier CP that is sensitive to heat stress (13-16 hrs AEL) and leads to changes in postsynaptic GluR composition (Fig. 3), than the central locomotor circuitry (17-19 hrs AEL) (Figs, 1 & 8A). By applying heat stress manipulations to these distinct developmental phases we were able to demonstrate that behavioural phenotypes on both larval crawling speed and network stability are linked to CP-induced changes to the central circuitry, not the peripheral NMJ (Figs. 1, 3 & 8A). These CPs correlate with the developmental phases when the body wall muscles and central neurons become electrically active, respectively (Broadie and Bate, 1993; Baines and Bate, 1998). Thus, they reflect the sequence of earlier myogenic and subsequent neuronal maturation (Crisp et al., 2008). This is comparable with observations from mammalian systems, where different parts of the developing brain mature in sequence (Hensch, 2005; Hensch and Quinlan, 2018; Reh et al., 2020).

Mechanistically, heat stress causes increased neuronal activity (Oswald et al., 2018; Sigrist et al., 2003; Zhong and Wu, 2004). Functional imaging of motoneurons showed that in larvae acute temperature elevations cause corresponding increases in both fictive locomotor activity waves (Fig. 1B) as well as larval crawling (Fig. 8B). Interestingly, this manipulation, when carried out during the embryonic CP, leads to a locomotor network with significantly slower intersegmental propagation of activity waves and, similarly, reduced crawling speed (Fig. 7). One possibility is that heat stress during the CP might cause a lasting malfunctioning of pacemaker activity. Alternatively, the slower central pattern of locomotor network activity in larval stages could be the consequence of a network-wide homeostatic response to an abnormally fast rate provoked by transient heat stress during the embryonic CP. The latter explanation is compatible with our recent findings of the CP as an integration window of excitatory vs inhibitory network activity (Hunter et al., 2024) and the identification of nitric oxide signalling pathways downstream of CP network activity (Giachello et al., 2021). To identify additional genetic pathways associated with CP plasticity, we propose that the NMJ, as a large, readily accessible and morphometrically quantifiable peripheral synapse, could serve as a convenient and effective assay.

Materials and Methods

Fly husbandry

Wild-type and transgenic strains were maintained on standard yeast–agar–cornmeal medium at 25°C. For CP manipulations, animals were kept in laying pots with apple juice agar plates and yeast paste for 6h. Collected eggs were either kept at 25°C for controls or at 32°C for 24h, before being moved back to 25°C until animals developed to the L3 larval stage.

Dissection and Immunostainings

Wandering L3 larvae were dissected in Sørensen’s sodium phosphate buffered saline (0.1 M, pH 7.2) and fixed for 10min at room temperature with Bouińs fixative (Sigma-Aldrich). After fixing, specimens were washed in washing solution (Sørensen’s buffer + 0.25% BSA + 0.3% TritonX-100 from Sigma-Aldrich) and incubated with mouse nc82 (anti-Bruchpilot to reveal active zones) antibody (1:100, DSHB) overnight, visualized with goat α-mouse StarRed (1:1000, Abberior, cat.no.STRED-1001-500UG, Lot. 2042PK-1) and AlexaFluor488 conjugated goat α-HRP (1:1000, Jackson ImmunoResearch, cat.no. 123-545-021, Lot. 152767) to visualise presynaptic, neuronal structures. For stainings of the postsynaptic glutamate receptor subunits GluRIIA and GluRIIB we used monoclonal mouse-α-DGluRIIA antibody (1:100, DSHB, product ID: 8B4D2 (MH2B)) and polyclonal rabbit-α-DGluRIIB antiserum (1:1000, gift from Mihaela Serpe (Sulkowski et al., 2016)) in conjunction with goat α-mouse StarRed (1:1000) and goat α-rabbit Atto594 secondary antibodies (1:1000, Hypermol, cat.no. 2306, Lot. 0356RG22), respectively, in combinations with AlexaFluor488 conjugated goat α-HRP (1:1000, Jackson ImmunoResearch, cat.no. 123-545-021, Lot. 152767) to visualise presynaptic terminals.

Fly stocks and other Key Resources

Image acquisition

Confocal images were acquired, using a Leica Stellaris point scanning confocal microscope, with an 40x/1.25N.A. oil immersion lens (Leica) and a 1.25 digital zoom. Images were acquired with a voxel size 0.16x 0.16x 0.3 µm. Images were processed using FIJI and Affinity Designer.

Two electrode voltage clamp

Two Electrode Voltage Clamp (TEVC) recordings were performed in HL3.1 saline (Feng et al., 2004) composition in mM: NaCl 70, KCl 2.5, MGCl₂ 4, CaCl₂ 2 (if not other mentioned), NaHCO₃ 10, trehalose 5, sucrose 115, HEPES 5 (pH was adjusted to 7.24-7.25 with 1M NaOH), using an Olympus BX50WI compound microscope with 60X dipping (Olympus 60x/0.9 N.A.) objective lenses. Experiments were performed on third-instar larval NMJs of the dorsal acute muscle 1 (DA1) in segments A3 and A4. The respective nerve was sucked in with a self-built suction electrode (Johnson et al., 2007) which was filled with HL3.1 saline. Subsequently, the nerve was stimulated for 0.1ms to trigger action potential firing. Briefly, voltage was stepwise increased until the first synaptic transmission could be recorded. From that voltage, 1 V was added to reliably evoke APs in control conditions, in 32°C CP manipulated animals 2-3 V had to be added to reliably recruit aCC and RP2. For PhTx experiments, samples were incubated in 4 µM PhTx final concentration (Frank et al., 2006) for 10 min. While dissecting, great care was taken to not stretch animals too much. Successful PhTx application was determined by dividing miniature EPSC amplitude of PhTx treated animals by the average Amplitude of control mEPSCs. If PhTx treated mEPSCs were at least 40% reduced compared to control, PhTx application was considered successful (Frank et al., 2006). Quantal content was determined by dividing EPSC by mEPSC for each recording. Recordings were only further considered from cells with an initial Vm below -50 mV and input resistance between 8-10 MΩ, using intracellular electrodes filled with 3M KCl with resistances of ∼20 MΩ (Electrodes for current injection and the membrane voltage). Current passing und recording electrodes were pulled from Borosilicate glass microelectrodes with filament (OD1 mm, ID 0.58mm; Harvard Apparatus LTD, GC100F-10 Part No. 30-0019), suction electrodes from Borosilicate glass microelectrodes without filament (OD1 mm, ID 0.58mm; Harvard Apparatus LTD, GC100-10 Part No. 30-0016) with a Flaming/Brown Puller Model P-97 (Sutter Instruments USA). Excitatory postsynaptic currents (EPSCs) were evoked by an Isolated Pulse Stimulator (Model 2100 A-M Systems USA), recorded at a clamped voltage of -60 mV with a sampling rate of 50 kHz and filtered with the anti-alias filter at 25 kHz using the AxoClamp 2B amplifier (Molecular Devices Axon Instruments USA). Data analysis was done with Clampfit 10.7. Traces were first filtered with a 360 Hz Gaussian low pass filter, then template searches were run over the trace for single pulses and spontaneous miniature excitatory postsynaptic currents (mEPSCs or minis). The resulting files were used to analyse mEPSC and EPSC properties. Rise and decay tau were measured with an one-term exponential product fit (Levenberg-Marquadt method with a max of 5000 iterations)

Current clamp muscle recordings

Current clamp recordings from muscle DA1 were performed in HL3.1 saline (see above), in semi intact, fillet dissected wandering L3 larvae with intact nervous system. After finishing the dissection, only animals performing crawling-like movements were considered for further experiments. To block muscle contractions, a 500 µM stock solution of nifedipine in HL3.1 and 1% DMSO was added to the bath for a final concentration of 50 µM nifedipine and 0.1% DMSO for 15 min. Recordings were only further considered from cells with an initial Vm below -60 mV and input resistance between 8-10 MΩ, using intracellular electrodes filled with 3M KCl with resistances of ∼20 MΩ. Spontaneous activity was recorded for 5min. Burst analysis was performed using the Clampfit Burst Analysis, using the Poisson Surprise model with a minimum of 5 events per burst.

Electroshock assay

A handheld stimulator consisting of two tungsten wires (0.1 mm diameter) fixed to a plastic rod was used. Wires were shaped to be ∼1-2 mm apart, attached to a DS2A Mk.II stimulator (Digitimer Ltd), which was set to provide a 6 V, 2 s current pulse. Larvae were placed on a plastic plate lid, with tissue paper used to dry residual moisture from the larvae. Once larvae had returned to normal crawling, the tungsten wires were brought into contact with the larvae, positioned on the larval cuticle, perpendicular to and over the approximate position of the CNS, with gentle pressure. Current stimulation was then delivered, which usually induced a sustained contraction and paralysis representing a seizure (videos can be viewed in Huertas Radi et al., 2025 and Marley and Baines 2011). The time to resumption of normal behaviour (regular, whole body forward propagating peristaltic waves) was recorded and defined as the duration of recovery from seizure.

Dissection and CNS mounting for calcium imaging and electrophysiology

Larval CNSs were dissected and mounted for electrophysiological recordings and for GCaMP imaging. Third instar (L3) larvae (of either sex) were dissected in a dish in standard saline (135 mM NaCl (Fisher Scientific), 5 mM KCl (Fisher Scientific), 4 mM MgCl₂·6H₂O (Sigma-Aldrich), 2mM CaCl₂·2H₂O (Fisher Scientific), 5 mM TES (Sigma-Aldrich), 36 mM sucrose (Fisher Scientific), pH 7.15), to remove the CNS (ventral nerve cord and brain lobes). The isolated CNS was then transferred to a droplet of external saline, within which it was laid flat (dorsal side up) and glued (GLUture Topical Tissue Adhesive; World Precision Instruments USA) to a Sylgard-coated cover slip (1 to 2mm depth of cured SYLGARD Elastomer (Dow-Corning USA) on a 22 x 22mm square coverslip). The preparation was then placed on a glass slide and viewed under a microscope (Olympus BX51-WI).

Calcium imaging

Dissected CNSs were imaged (QImaging EXi-Aqua; photometrics, Arizona, US) with an acquisition frame rate of 10Hz and frame duration of 100 ms using a x20 immersion lens. GCaMP8f was excited using a 470 nm collimated LED (Thorlabs, New Jersey, US). Time-course of neuronal fluorescence was determined in WinFluor (V.4.1.5; University of Strathclyde, UK) by adding regions of interest (ROIs) to axons, close to the cell body (Streit et al, 2016). In Clampfit (10.3.1.5; Molecular Devices, California, US) fluorescence traces were smoothed using a lowpass BoxCar filter using 11 smoothing points prior to analysis.

For motoneuron calcium transients (Fig. 1), recordings were conducted for 600s, whilst temperature of the explanted CNS was increased steadily from ∼23°C to ∼34°C using a digital heating unit (CO 102; Linkam Scientific, Surrey, UK). For sham recordings, temperature was maintained at a constant 25°C. Temperature was measured in Clampex (10.3.1.5; Molecular Devices, California, US) using a temperature sensor placed in the external saline surrounding the brain, via a temperature controller (TC-10; npi electronic GmbH, Tamm, Germany). The frequency of calcium fluorescence peaks was measured for each segment (A8/9 – A1) in three sequential 200 s time periods and averaged per larva. Average temperature was calculated for each recording during these same 200 s time periods.

For calcium imaging of interneurons after CP manipulation (Fig. 5), 180 s recordings were taken. Time lag was determined by manually assessing the time between peak fluorescence corresponding to ipsilaterally adjacent interneurons (segments A6 – A1. For each recording the first complete forward waves of activity were used (up to three repeats per trace) to generate an average intersegment duration for each pair of interneurons.

Whole cell patch clamp electrophysiology

CNSs were dissected, mounted (see calcium imaging methods) and viewed under a 60x water-immersion lens. To access soma, 1% protease (Streptomyces griseus, Type XIV, Sigma-Aldrich, in external saline) contained within a wide-bore glass pipette (GC100TF-10; Harvard Apparatus UK, approx. 10 µm opening) was applied to abdominal segments, roughly between A5-A2 (Baines and Bate, 1998). This was done to remove overlaying glia to facilitate access to underlying nerve cell soma (e.g., aCC). Motoneurons were identified by anatomical position and relative cell size, with aCC being positioned close to the midline and containing both an ipsilateral and contralateral projection. Whole-cell patch recordings were made using borosilicate glass pipettes (GC100F-10, Harvard Apparatus) that were fire polished to resistances of 10 - 15MΩ when filled with intracellular saline (140 mM potassium-D-gluconate (Sigma-Aldrich), 2 mM MgCl₂·6H₂O (Sigma-Aldrich), 2 mM EGTA (Sigma-Aldrich), 5 mM KCl (Fisher Scientific), and 20 mM HEPES (Sigma-Aldrich), (pH 7.4). Input resistance was measured in the ‘whole cell’ configuration, and only cells that had an input resistance ≥ 0.5 GΩ were used for experiments. Cell capacitance and break-in resting membrane potential was also measured for each cell recorded. Data for current steps recordings was captured using a Multiclamp 700B amplifier controlled by pCLAMP (version 10.7.0.3), via an analogue-to-digital converter (Digidata 1440A, Molecular Devices). Trace/s were sampled at 20 kHz and filtered online at 10 kHz.

Current step recordings

Mecamylamine (200 µM in saline) was used to block postsynaptic nACh receptors to synaptically isolate neurons for experiments designed to measure intrinsic motoneuron excitability. Once patched, neurons were brought to a membrane potential of -60 mV using current injection. Each recording consisted of 20 x 4 pA (500 ms) current steps, including an initial negative step, giving a range of -4 to +72 pA. Number of spikes were counted and plotted against injected current. Both cell capacitance and input resistance were compared between conditions to ensure that any observed differences in excitability were not due to differences in either cell size or resistance.

Synaptic drive to motoneurons

To measure synaptic drive between A27h premotor interneurons and aCC, A27h neurons were identified via Chronos-mVenus reporter (as above), allowing corresponding aCC motoneurons to be patched. Recordings were conducted in current-clamp, with current injected into aCC sufficient to evoke a mean spike frequency of ∼10-20 Hz, prior to optogenetic stimulation. Recordings consisted of: individual sweeps, beginning with 1 s LED off, followed by 1 s LED on, and 1 s LED off (repeated 5-8 times per cell). Change in AP frequency (%) was calculated for each recording (per sweep) comparing AP frequency before and during stimulation, which was then averaged for each cell.

Spontaneous rhythmic currents (SRCs) were recorded for 180 s with aCC voltage clamped at -60 mV. Amplitude, duration, and frequency were measured for the first 15 events of each trace using Clampfit’s event detection, threshold search, function. The first 15 events with a stable baseline, an amplitude of >300 pA and no double peaks were accepted for analysis. Amplitude was measured as the change from baseline immediately prior to event initiation to the peak current and normalised for cell capacitance, providing a measure of current density. SRC duration was measured as the width of an event at 100 pA amplitude.

Spontaneous motoneuron recordings

Cell-attached recordings, measuring motoneuron activity, were conducted using unpolished patch electrodes. Motoneuron spontaneous activity was recorded using an adapted cell-attached current clamp protocol (Multiclamp 700B amplifier with I=0) to measure passive action potential bursting properties. In Clampfit (version 10.3.1.5), 180 s recordings, were analysed to detect action potential ‘spike times’. These were imported into MATLAB and analysed using a custom script (developed by Mituzaite et al., 2021). An action potential burst was considered as an event consisting of at least 3 spikes occurring within 100 ms and ending when no spike was detected for 100 ms.

Crawling Assay

Crawling behaviour was recorded at the mid-third instar stage, 72 hours after larval hatching (ALH). Larvae were rinsed in water and placed inside a 24cm x 24cm crawling arena with a base of 5mm thick 0.8% agarose gel (Bacto Agar), situated inside an incubator. Temperature was maintained at 25±0.5°C, reported via a temperature probe in the agar medium. Humidity was kept constant. Larval crawling was recorded using a frustrated total internal reflection-based imaging method (FIM) in conjunction with the tracking software FIMTrack (Risse et al., 2017, 2013) using a Basler acA2040-180km CMOS camera, fitted with a 16mm KOWA-IJM3sHC.SW-VIS-NIR lens, acquisition controlled by Pylon software (Basler) and Streampix (v.6) software (NorPix). Larvae were recorded for 20 minutes at five frames per second.

Recordings were split into four 5-minute sections with the first five minutes of acclimatisation discarded. The remaining three 5-minute sections were used to analyse crawling speed, choosing crawling periods of uninterrupted forward crawling, devoid of pauses, turning or collision events. Each larva was sampled up to once per 5-minute section, i.e. up to three data points were derived per larva from up to three consecutive 5-minute segments. Crawling speed was calculated using FIMTrack software.

Statistics

Statistical analyses were done using GraphPad Prism Software (Version 10.1.2). Datasets were tested for normal distribution with the Shapiro-Wilk Test. Normally distributed data were then tested with students t-test (for pairwise comparison). Normally distributed analysis for more than two groups was done with a one-way ANOVA and post hoc tested with a Tukey multiple comparison test. Non-normally distributed data sets of two groups were tested with Mann-Whitney U Test (pairwise comparison) and datasets with more than two groups were tested with a Kruskal Wallis ANOVA and post hoc tested with a Dunns multiple post hoc comparison test. For whole cell electrophysiology experiments in isolated neurons, linear regression analysis was used to compare the intercepts between conditions. This analysis was restricted to linear sections of data, represented by inset graphs. For all datasets mean and standard error of mean (SEM) are shown. Significance levels were * p<0.05; ** p<0.01; *** p<0.001; **** p<0.0001. Sample sizes are stated in figures or legends, each data point derived from a unique specimen, unless otherwise stated; e.g. Figure 3B, NMJ sizes were sampled at up to four data points per larva; for larval crawling speeds up to three data points were derived per larva, one for each unique 5-minute segment of recorded crawling. All experiments indiscriminately included larvae of both sexes.

Data availability

Figures 1-8 source data contain the numerical data used to generate the figures.

Acknowledgements

This work was made possible through support by a Walter Benjamin Programme Fellowship to NK by the Deutsche Forschungsgemeinschaft (DFG) KR5597/1-1, and funding from the Biotechnology and Biological Sciences Research Council (BBSRC) to ML (BB/V014943/1) and the Wellcome Trust, through a Joint Wellcome Trust Investigator Award to RAB and ML (217099/Z/19/Z). Research reported in this publication was supported by an institutional startup fund from Texas A&M University (AAZ). DSC was supported by the European Molecular Biology Organization (EMBO) with a long-term EMBO fellowship (ALTF 62-2021). The work benefited from the Imaging Facility, Department of Zoology, supported by Matt Wayland and Tom Pettini, and funds from a Wellcome Trust Equipment Grant (WT079204) with contributions by the Sir Isaac Newton Trust in Cambridge, including Research Grant [18.07ii(c)]. Work on this project further benefited from the Manchester Fly Facility, established through funds from the University and the Wellcome Trust (grant 087742/Z/08/Z).

The authors would like to thank members of the Baines and Landgraf research teams for feedback on the manuscript. The authors are grateful to Ela Serpe for Rabbit polyclonal anti-GluRIIB antiserum.

Stocks obtained from the Bloomington Drosophila Stock Center (NIH P40OD018537) were used in this study.

The 8B4D2 (MH2B) monoclonal antibody to visualise GluRIIA, developed C. Goodman, was obtained from the Developmental Studies Hybridoma Bank, created by the NICHD of the NIH and maintained at The University of Iowa, Department of Biology, Iowa City, IA 52242.

Additional information

Funding

Deutsche Forschungsgemeinschaft (DFG) (KR5597/1-1)

• Matthias Landgraf

UKRI | Biotechnology and Biological Sciences Research Council (AFRC) (BB/V014943/1)

• Matthias Landgraf

Wellcome Trust (WT)

https://doi.org/10.35802/217099

• Matthias Landgraf

European Molecular Biology Organization (EMBO) (ALTF 62-2021)

• Matthias Landgraf

Wellcome Trust (WT) (WT079204)

• Matthias Landgraf

Wellcome Trust (WT)

https://doi.org/10.35802/087742

• Matthias Landgraf

Isaac Newton Trust (18.07ii(c))

• Matthias Landgraf