Recent work on competitive interactions and behavioral transitions in Drosophila larvae has reported CIII md and Ch connectomes in the context of larval mechanosensation (Masson et al., 2020). We performed a comparative circuit analysis of somatosensory (Ch, CIII md, and CIV md) neurons using an online repository hosted by LMB Cambridge, which contains a ssTEM volume of Drosophila melanogaster first instar larval central nervous system and a repository of neural reconstructions (https://l1em.catmaid.virtualflybrain.org). Here, we report and contextualize relevant CIII md neuron downstream neurons, whose functional roles in cold nociception were evaluated. The meta-analyses of previously published literature on Drosophila larval somatosensory connectomes revealed several common post-synaptic partners that are shared amongst Ch, CIII md, and CIV md neurons (Figure 1A). CIII md neurons are upstream of multisensory integration neurons, pre-motor neurons, and projection neurons. Class III md neurons are synaptically connected to multisensory integration neurons (Basin-2, –3, and –4). The downstream pathway from Basins includes connectivity to A00c neurons and a polysynaptic pathway to the rolling command neuron (Goro) via A05q neurons. CIII md neurons are also connected to various pre-motor neurons, which are ventral nerve cord localized neurons providing synaptic input to motor neurons, including DnB, A02n, Chair-1, A01d3, A02f, A02h, and A27k, and projection neurons including A05q, A09e, TePn05, A08n, A09o, A02o, and dILP7. The CIII md neuron connectome analyses reveal broad second-order connectivity and complex interconnectivity amongst second-order neurons (Figure 1A, Figure 1—figure supplement 1). Functional and behavioral roles of CIII md neuron second-order interneurons in cold nociception remain unexplored and below we address this knowledge gap.

Due to shared post-synaptic neural connectivity of Ch, CIII md, and CIV md neurons (Figure 1A, Figure 1—figure supplement 1), we first assessed functional roles of these sensory neurons in cold nociception by investigating: (1) stimulus-evoked calcium responses in the Drosophila larval ventral nerve cord; (2) cold-evoked calcium responses of these sensory neurons; (3) necessity and sufficiency of these sensory neurons in cold-evoked behaviors; and (4) examining whether co-activating multiple sensory neurons facilitates CIII-mediated behavioral response in Drosophila larvae.

Basin interneurons function as multisensory integrators receiving convergent inputs from mechano-, chemo- and thermo-sensitive peripheral sensory neurons (Figure 2A, Figure 2—figure supplement 1A). Previous behavioral and functional studies have revealed that Basin interneurons are required for nociceptive escape responses mediated by Ch and CIV md neurons (Ohyama et al., 2015). Both Ch and CIV md neurons play roles in CIII md-mediated behavioral responses either in cold nociception and/or can detect cold stimuli. We hypothesized that Basin interneuron function is required downstream of cold nociceptive somatosensory neurons. We evaluated whether Basin interneurons are necessary and sufficient for cold nociception, exhibit cold-evoked increases in calcium response and function downstream of CIII md neurons.

Basin neurons innervate a set of projection neurons (A05q and A00c), which are upstream of a command neuron (Goro) that is responsible for initiating CIV md neuron-mediated nociceptive escape behaviors (Figure 4A, Figure 4—figure supplement 1A; Ohyama et al., 2015). We set out to test whether cold-evoked behavioral responses mediated by CIII md and Basin neurons function through A00c, A05q, and/or Goro neurons. We first assessed whether these neurons are required for cold-evoked behavioral responses (Figure 4B–E). Neural silencing of A05q neurons via tetanus toxin led to mild, yet significant reductions in cold-evoked responses, when compared to w¹¹¹⁸ genetic control (Figure 4B–E). Neural silencing of A00c neurons resulted in significantly lower cold-evoked cumulative CT response compared to w¹¹¹⁸ (Figure 4C). There were also significant reductions in CT duration and CT magnitude when A00c neurons were silenced compared to both controls (Figure 4D and E). In contrast, the Goro command neuron for nociceptive rolling behavior is not required for cold-evoked CT responses (Figure 4B–E). Next, we assessed whether neural activation of these neurons led to evoked CT responses. Like Basin neurons, single cell-type activation of A00c, A05q, or Goro neurons did not lead to CT behavior (Figure 2—figure supplement 2). Similarly, optogenetic co-activation of CIII md neurons and A05q or Goro neurons did not lead to significant facilitation of the CT response (Figure 4F–I). However, simultaneously co-activating A00c and CIII md neurons led to significant increases in CT duration and CT magnitude (Figure 4H-I). There were no changes in Drosophila larval immobility upon co-activation of CIII md plus A00c, A05q, or Goro neurons (Figure 4—figure supplement 1B–D). Only co-activation of A00c neurons leads to notable enhancement of CIII md-mediated CT response, whereas neural silencing of A00c or A05q neurons led to greater than 25% impairment in cold-evoked behavioral responses (Figure 4J). In conclusion, the Basin to Goro polysynaptic pathway does not significantly contribute to the cold-evoked CT response.

Select Drosophila larval premotor neurons were previously implicated in CIV-mediated nociceptive escape responses. Specifically, DnB premotor neurons are involved in noxious thermal stimulus-evoked c-bending and rolling behavior (Burgos et al., 2018; Lopez-Bellido et al., 2019). Drosophila larvae also roll in response to activation of mCSI premotor neurons, which are synaptically connected to CIV md neurons and predicted to be A02m/n neurons from EM connectomes (Lopez-Bellido et al., 2019; Yoshino et al., 2017). Additionally, Chair-1 (A10a) premotor neurons have been implicated in anemotaxis (Jovanic et al., 2019). Collectively, these nocifensive premotor neurons primarily receive inputs from both primary sensory neurons (CIII and CIV md) and multisensory integrators (Basin-2 and –4) (Figure 5A, Figure 5—figure supplement 1A). CIII md neurons have additional synaptic connectivity with intersegmental feedback circuitry involved in locomotion, where Ifb-fwd (A01d3) is active during forward locomotion and Ifb-bwd (A27k) is active during backwards locomotion (Kohsaka et al., 2019). The intersegmental feedback circuit neurons (A01d3 and A27k) are interconnected via A02e and A02g (Kohsaka et al., 2019). GABAergic A31k premotor neuron is downstream of A01d3 and is responsible for inhibiting motor activity causing relaxation of the larval body. Lastly, A02f has been implicated in delaying nocifensive behavioral responses (Garner, 2020). We predicted that select premotor neurons are required for cold-evoked responses and function downstream of CIII md neurons in a stimulus-specific manner.

Inhibition of neural transmission via cell-type-specific expression of tetanus toxin in individual premotor neurons led to reduced cold-evoked responses in Drosophila larvae (Figure 5B–E). Silencing Chair-1 neurons resulted in the strongest reduction of cold-evoked CT responses, where instantaneous %CT was the lowest of all premotor neuron subtypes tested (Figure 5B). CT duration, magnitude, and cumulative percent response were all significantly reduced compared to controls (Figure 5B–E). Impairment in Chair-1 neuronal signaling leads to 75% reduction from controls in cold-evoked behaviors (Figure 5J). We silenced DnB neurons using two independent cell-type-specific driver lines (DnB’ (IT4051^GAL4) & DnB’’(IT412^GAL4)), where both resulted in reduced instantaneous %CT response, along with significant reductions in CT duration and magnitude compared to controls (Figure 5B–E). Silencing mCSI (R94B10^GAL4) neurons also resulted in significantly reduced cold-evoked CT responses compared to controls (Figure 5B–E). DnB or mCSI inhibition of neurotransmitter release leads to approximately 50% reduction in cold-evoked CT responses (Figure 5J). Furthermore, we silenced premotor neurons previously implicated in locomotion (A01d3, A27k, A02e, A02g, A31k) or delaying nocifensive responses (A02f); this did not lead to significant changes in cold-evoked CT responses (Figure 5—figure supplement 2). Collectively, the previously implicated nocifensive premotor neurons are required for cold-evoked behavioral responses.

Based on EM connectivity and neural silencing experiments, we predicted that activation of these premotor neurons would be sufficient for Drosophila larval CT response. First, we found that activation of premotor neurons alone did not evoke CT response (Figure 2—figure supplement 2). Next, we performed optogenetic co-activation of CIII md and premotor neuron subtypes. Chair-1 neuron co-activation with CIII md neuron did not have an effect on CT responses (Figure 5F–I); however, there was a reduction in larval immobility, where average mobility was increased and duration of immobility was reduced compared to when only CIII md neurons are activated (Figure 5—figure supplement 1B–D). Simultaneous activation of CIII md neurons and DnB or mCSI both led to significant increases in CT responses, as measured by peak CT response, CT duration or magnitude, compared to CIII md neuron activation alone (Figure 5F–I). DnB and CIII md neuron co-activation led to short lived, subtle but insignificant increases in larval immobility (Figure 5—figure supplement 1B–D). Co-activation of mCSI and CIII md neurons resulted in mild reductions in larval immobility (Figure 5—figure supplement 1B–D). These data reveal DnB and mCSI neuronal activity enhances CIII md neuron-mediated CT responses.

Since these premotor neurons are postsynaptic to CIII md neurons and their activity is required for proper cold-evoked behaviors, we predicted that either activation of CIII md neurons or exposure to cold would cause increases in Ca²⁺. To test this prediction, we selectively expressed the Ca²⁺ integrator CaMPARI2 in premotor neurons. Unexpectedly, there was no change in cold-evoked Ca²⁺ levels in mCSI or Chair-1 neurons (Figure 6A and B). Noxious cold exposure did, however, lead to significant Ca²⁺ increases in DnB neurons (Figure 6C and D). To further assess how CIII md neuronal activity affects DnB function, we optogenetically activated CIII md neurons and assessed evoked Ca²⁺ levels of DnB using GCaMP6. Interestingly, Drosophila larvae raised without all trans-retinal, a requisite light-sensitive cofactor for optogenetic experiments, also had a mild light-evoked increase in Ca²⁺ (Figure 6E and F). CIII md neuron optogenetic activation led to significant increases in DnB Ca²⁺ levels that slowly returned to baseline levels (Figure 6E and F). Upon repeated CIII md neuron activations, DnB neurons exhibit a blunted Ca²⁺ response relative to initial stimulation (Figure 6E and F). Taken together, we find that premotor neuronal function is required for cold nociception and premotor neuron activity can facilitate CIII md neuron-mediated CT responses (Figure 6—figure supplement 1).

Sensory, second order multisensory integration neurons (Basins) and a premotor neuron (DnB) have further direct synaptic connectivity to a set of projection neurons including A09e, A08n, A02o, dILP7, and R61A01^GAL4 labeled neurons (labels: A10j, A09o, TePn04, TePn05) that have previously been implicated in anemotaxis, mechanosensory, or chemosensory evoked behavioral responses (Hu et al., 2017; Jovanic et al., 2019; Kaneko et al., 2017; Masson et al., 2020; Takagi et al., 2017; Vogelstein et al., 2014). Briefly, A09e receives synaptic inputs from CIII md, CIV md, A08n, DnB, and TePn05 (Figure 7A, Figure 7—figure supplement 1A). A09e and CIII md neurons are both required for Drosophila larval anemotaxis (Jovanic et al., 2019; Masson et al., 2020). A08n primarily receives inputs from CIV md neurons and very few inputs from CIII md neurons (Figure 7A, Figure 7—figure supplement 1A). However, Drosophila third instar larval synaptic connectivity visualization using GFP reconstituted across synaptic partners (GRASP) revealed that A08n are not synaptic partners of CIII md neurons (Kaneko et al., 2017). A08n neurons function downstream of CIV md neurons for noxious chemical and mechanical nociception (Hu et al., 2017; Kaneko et al., 2017), and neural activation of Ch and CIII md neurons does not lead to activation of A08n neurons (Tenedini et al., 2019). R61A01^GAL4-labeled neurons receive inputs from Ch, CIII md, CIV md, Basins, DnB, and A09e (Figure 7A, Figure 7—figure supplement 1A). Neurons labeled by R61A01^GAL4 have been implicated in mechanosensation (Masson et al., 2020). CIII md neurons synapse onto two additional neurons that function in processing mechanosensitive information A02o and dILP7. A02o neurons are command-like neurons required for directional avoidant responses to noxious touch (Takagi et al., 2017), whereas dILP7 neurons are required for noxious mechanical and/or UV/blue light avoidance using short neuropeptide F or Ilp7, respectively (Hu et al., 2017; Imambocus et al., 2022). CIII md to dILP7 synaptic connectivity is observed in the first instar larval EM connectome, however, assessing synapse connectivity using activity-dependent GRASP in third instar larvae, there was no detectable connectivity (Hu et al., 2017). We assessed whether the following projection neurons A09e, A08n, neurons in the expression pattern of the GAL4 line R61A01 (A10j, A09o, TePn04, TePn05), A02o and dILP7 neurons are required for cold-evoked behaviors and function in conjunction with CIII md neurons for generating CT behavioral response.

Neural silencing of projection neurons using cell-specific expression of tetanus toxin led to impairments in cold-evoked behaviors. Neurotransmission inhibition in projection neurons (A09e and R61A01) that receive strong connectivity from CIII md neurons led to significant reductions in all the CT behavioral metrics that we analyzed (Figure 7B–E). As expected, A08n neurons, which primarily receive inputs form CIV md neurons, are not required for cold-evoked CT response (Figure 7B–E). Whereas neural silencing of A09e or R61A01 neurons led to greater than 50% reduction in cold-evoked responses from controls (Figure 7J). Furthermore, neural silencing of dILP7 led to a mild but significant reduction in cold-evoked CT responses, whereas A02o neurons were dispensable for cold nociception (Figure 7A, Figure 7—figure supplement 2).

Next, we assessed whether these projection neurons were able to elicit a CT response upon neural stimulation. Optogenetic stimulation of projection neurons alone did not lead to any CT response (Figure 2—figure supplement 2). Simultaneous co-activation of CIII md and R61A01^GAL4 or A08n neurons led to mild but statistically insignificant increases in CT responses (Figure 7F–I and J). Co-activation of A09e and CIII md neurons led to facilitation of CT behavior, where compared to controls there were significant increases across all behavioral metrics that we analyzed (Figure 7F–). There were mild, statistically insignificant, reductions in larval immobility when these projection neurons were co-activated with CIII md neurons; however, they were significant for A08n (Figure 7—figure supplement 1B–D). A09e, which receives substantial inputs from CIII md neuron, was sufficient to facilitate CIII md neuron-evoked behavioral responses (Figure 7J).

Our neural silencing analyses suggest that A09e and R61A01^GAL4 (A10j, A09o, TePn04, TePn05) neurons are required for cold-evoked responses. We hypothesized that these neurons are cold sensitive and function downstream of CIII md neurons. We predicted that these neurons would have increases in Ca²⁺ upon cold stimulation or optogenetic activation of CIII md neurons. A09e neurons present indeed significant increases in Ca²⁺ levels upon cold stimulation, as measured by cell-type-specific expression of CaMPARI2 (Figure 8A). Optogenetic activation of CIII md neurons led to significant increases in evoked Ca²⁺ levels of A09e neurons (Figure 8B and C). Multiple stimulations of CIII md neurons did lead to slightly lower levels of evoked Ca²⁺ but overall Ca²⁺ response was largely similar between stimulations (Figure 8B and C). For R61A01^GAL4 Ca²⁺ imaging experiments, we restricted our analyses to TePn04 and TePn05 neurons, which were reliably identifiable in an intact Drosophila larval preparation. There were significant increases in CaMPARI2 response of TePn04/05 neurons upon cold stimulation (Figure 8D). Optogenetic stimulation of CIII md neurons led to strong increases in Ca²⁺ levels of TePn04/05 neurons; however, Ca²⁺ levels rapidly returned to baseline levels (Figure 8E and F). Repeated stimulations of CIII md neurons led to a blunted Ca²⁺ response in TePn04/05 neurons following the initial stimulation (Figure 8E and F). Combined, these data indicate A09e and R61A01^GAL4 neurons function downstream of CIII md neurons for cold nociception (Figure 8—figure supplement 1), indicating that ascending neurons relay cold somatosensation to the brain.

Thus far, data were presented in logical groups based on their previously known cell types and functions; however, these neurons function in an interconnected network and behavioral and functional imaging data must be assessed collectively to study circuit function. As discussed in previous sections, of all the cell types we tested, only optogenetic activation of CIII md neurons alone is sufficient to elicit CT responses (Figure 2—figure supplement 2). Neural co-activation of CIII md neurons plus additional cell types resulted in marked increases in CT responses compared to only CIII md neuron activation (Figure 9A). Neural silencing of cell types downstream of CIII md neurons led to significant reductions in cold-evoked CT (Figure 9B). To understand how these behavioral phenotypes are interrelated, we performed t-distributed stochastic neighbor embedding (t-SNE) analysis on both neural silencing and co-activation behavioral datasets. We identified five different clusters that exhibit varying impacts on behavioral phenotypes upon either neural co-activation or silencing (Figure 9C and D). The first group clusters together with the control genotype and includes CIV, A05q, Goro, and A08n neurons. This group on average had less than 25% change in behavioral phenotypes in either neural co-activation or silencing experiments (Figure 9C and D). Basin-1 and DnB neurons clustered together, where on average they had 49% enhancement of CT response upon neural co-activation and 45% reduction in cold-evoked CT response upon neural silencing (Figure 9C and D). R61A01 and CIII md neurons formed a cluster, where neural silencing resulted in nearly 60% reduction cold-evoked behavior and 20% enhancement of CIII md neuron-evoked behaviors upon neural co-activation (Figure 9C and D). A group of neurons including Chair-1, Basin-4, and Ch were required for cold-evoked behavioral responses; however, these neurons did not facilitate CIII-evoked CT responses (Figure 9C and D). In the last group, neural silencing of all Basins, Basin-2, mCSI, A00c, or A09e led to an overall 69% reduction in cold-evoked behaviors and 34% enhancement in CIII evoked CT response (Figure 9C and D). Additionally, we assessed how CIII md neuron synaptic connectivity to post-synaptic neurons could inform whether CIII md post-synaptic neurons are required for cold nociception (Figure 9E). Generally, neural silencing of highly connected CIII md post-synaptic neurons exhibited a greater requirement in cold nociception; however, there are a few notable exceptions (A02f and A02o), where greater connectivity did not lead to significant impairment in cold sensitivity (Figure 9E).

Analyses of Drosophila larval behavioral phenotypes revealed distinct roles for downstream neurons in cold nociceptive circuitry. Mapping synaptic connectivity of a particular circuit is only the first step in understanding how individual behaviors arise. Here, we draw attention to select CIII md neuron first-order neurons (Basins (–2 and –4), DnB, TePns, and A09e) that showed robust requirements for cold nociceptive behaviors and had functional connectivity to CIII md neurons (Figure 9F). TePns synapse onto Basin-4 and form reciprocal connections to Basin-2, DnB, and A09e (Figure 9F). DnB further synapses onto Basin-4 and A09e. Within this circuit motif one might predict that A09e neurons function as the master integrators, where they might be computing synaptic information from various sources (Figure 9F). Consistent with EM connectivity, A09e neurons have the highest CIII md neuron-evoked Ca²⁺ responses of the tested cell types (Figure 9G and H). Compared to Basin-4 neurons, Basin-2 neurons receive greater synaptic input from CIII md neurons (Figure 9F). However, CIII md neuron activation leads to nearly twice as much Ca²⁺ response in Basin-4 neurons than Basin-2 (Figure 9G and H). DnB neurons receive the second highest synaptic input from CIII md neuron and have the second highest CIII md neuron-evoked Ca²⁺ response (Figure 9G and H). TePns receive significantly lower synaptic input compared to Basin-2 neurons; however, they both exhibit similar levels of CIII md neuron-evoked Ca²⁺ response (Figure 9G and H). Comparative analyses reveal that synaptic connectivity is informative about functional neural activity, however, additional molecular and functional studies are required to fully understand roles of these neurons in cold nociception. Collectively, our findings on neural substrates of cold nociception indicate that second-order multisensory integration by Basin neurons, select pre-motor neurons, and projection neurons are preferentially activated in a stimulus-specific manner to elicit appropriate behavioral responses.

Sensory perception is the first node for interfacing with the external environment, which is processed through various pathways in the brain eventually leading to motor commands that generate stimulus-specific behavioral responses. Thus far, we have implicated CIII md neurons as the primary cold-sensitive neurons, which transduce thermal cues to select multisensory integration neurons, premotor neurons, and ascending neurons. However, it remains unexplored how the sensory and central processing of cold leads to CT responses. To this end, assessing CIII md neuron’s premotor neuron network revealed that collectively CIII md premotor neurons have connectivity to five of the six major muscle groups in a Drosophila larva hemi-segment (Figure 10A). Next, we assessed how individual Drosophila larva muscular segments respond to noxious cold. Analysis of muscle localized Ca²⁺ responses revealed that all larval segments have cold-evoked increases in Ca²⁺ (Figure 10B–D). Interestingly, the strongest cold-evoked response is present in abdominal segments (A2-A5), whereas anterior (T1-A1) and posterior (A6-A8) segments have marginally smaller cold-evoked Ca²⁺ responses (Figure 10B and C). Similar to cold-evoked CT behavior, where the head and tail contract towards the center of the animal, muscle segments in anterior and posterior reach peak activity sooner than the central segments by an average of 4s (Figure 10B’). Evaluating individual larval cold-evoked Ca²⁺ levels revealed segmental responses that can be attributed to previously identified cold-evoked behavioral responses (i.e. CT and head and/or tail raises) (Figure 10D; Turner et al., 2016; Turner et al., 2018).

Furthermore, we assessed whether impaired circuit function would lead to reduced cold-evoked muscle responses. Firstly, while maintaining sensory perception, silencing motor neuron activity using GtACR1 resulted in significant reductions in cold-evoked Ca²⁺ levels compared to controls (Figure 10—figure supplement 1A–E). Secondly, application of ethyl ether also resulted in significant reductions in cold-evoked Ca²⁺ activity in postsynaptic densities (PSDs) in larval muscles (Figure 10—figure supplement 1F–H).

To achieve greater granularity of how muscles respond to noxious cold, we quantitatively assessed how individual muscles respond to changes in temperature. Based on electrical stimulations of abdominal motor nerves, larval body wall muscles have equivalent force generation capabilities for dorsal, lateral, or ventral muscles (Ormerod et al., 2022). Dorsal longitudinal and oblique (DL and DO) muscles respond acutely and maintain high cold-evoked Ca²⁺ levels throughout the stimulus period (Figure 10E–I). Meanwhile, lateral transverse (LT) muscles exhibit a gradually increasing cold-evoked Ca²⁺ levels reaching a peak response at the end of the cold stimulus (Figure 10E–I). Lastly, ventral lateral and oblique (VL and VO) muscles respond most robustly to the onset of cold stimulation which gradually declines over the course of the stimulation, in contrast to DL and DO muscles, which maintain their levels of Ca²⁺ activation throughout the stimulus (Figure 10E–I). Both dorsal (DO and DL) and ventral (VO and VL) muscles reach peak response ~10s before lateral muscles (Figure 10I). Thus, Drosophila larvae exhibit stimulus-specific muscle activation patterns that are consistent with cold-evoked behavioral responses arising from sensory processing of temperature changes.

Accurately responding to potentially harmful stimuli by appropriately executing energetically expensive escape behaviors is critical for animal survival. Incorrectly performing escape responses in absence of dangerous stimuli can have long-term detrimental consequences. Therefore, the need for neural mechanisms responsible for integrating various noxious and innocuous sensory stimuli for accurately executing stimulus-appropriate escape responses. Here, we focus on the role of Basin neurons, known to mediate various escape responses (Jovanic et al., 2016; Masson et al., 2020; Ohyama et al., 2015).

EM connectome analyses revealed that somatosensory (Ch, CIII md, and CIV md) neurons all synapse onto multisensory integrating Basin neurons, and that amongst the Basins subtypes, Basin-1 neurons are presynaptic to Basin-2, –3, and –4 neurons (Masson et al., 2020; Ohyama et al., 2015). On the basis of the known connectome, we analyzed the function of Basin neurons in cold-evoked escape behaviors. Our findings indicate that optogenetic activation of Basin-1 is not sufficient to evoke noxious cold-evoked behavioral response (Figure 2—figure supplement 2), and it was known that thermogenetic activation of Basin-1 is also not sufficient for evoked rolling escape (Ohyama et al., 2015). Therefore, Basin-1 neuron activity alone is not sufficient to elicit a noxious stimuli-evoked behavioral response. However, due to intricate connectivity amongst the Basins, the total integrated output of multiple Basin neurons may lead to threshold activation of noxious stimulus-evoked escape responses. And indeed, thermogenetic co-activation assays found that Basin-1 neural activity could facilitate a Basin-4-mediated rolling response (Ohyama et al., 2015), and also the activation of Basin-1 neurons leads to the activation of Basin-2 neurons through lateral disinhibition (Jovanic et al., 2016). Here, we found that co-activation of Basin-1 plus CIII md neurons led to strong-evoked CT responses (Figure 2). Taken together, these studies concluded that Basin-1 facilitates the activation of other Basin neurons, and our results also provide additional support for the role of Basin-1 neurons in fine-tuning behavioral outcomes and enhancing cumulative output of multisensory integration Basin neurons leading stimulus-specific escape responses.

In order to evaluate the functional and behavioral roles of neurons within a circuit, a comprehensive analysis of both upstream and downstream synaptic connectivity is necessary. Basin-2 and –4 neurons integrate different amounts of synaptic input from somatosensory neurons, other Basin neurons, feedforward/back inhibitory neurons, and projection neurons. Basin-4 neurons receive greater excitatory synaptic input compared to Basin-2 neurons, which have greater connectivity with local inhibitory neurons (Jovanic et al., 2016; Masson et al., 2020; Ohyama et al., 2015). Larger inhibitory connectivity of Basin-2 neurons likely arises from much broader downstream connectivity including several projection/descending neurons (A00c, A02o, A05q, A10j, TePn04), premotor neurons (A02m, A02n, Chair-1), descending feedback inhibitory neuron (SeIN128), local feedback neurons (Handle-a/-b), whereas Basin-4 downstream connectivity is restricted to limited set of neurons (A00c, A02o, Basin-3, Chair-1) (Zhu et al., 2024). These synaptic level differences in multisensory integration neurons underlie noxious stimuli-specific evoked responses and suggest that CIII md neuron-mediated responses primarily function through Basin-2 neurons and CIV md neuron-mediated environmental cues are processed via Basin-4 neurons.

Chemical nociception in Drosophila larvae primarily functions via CIV md neurons. Among the md neurons, CIII md neurons are the least sensitive to noxious chemical stimulus (Lopez-Bellido et al., 2019). Analyses of Basin neuron requirement in chemical nociception revealed that all Basins are necessary for chemical-evoked rolling, where comparative analyses of Basin subtypes revealed that chemical nociception is primarily mediated through Basin-4 with weaker behavioral phenotypes of Basin-1 or –2 (Lopez-Bellido et al., 2019). Our work shows that inhibition of neurotransmission in Basin-2 or Basin-4 neurons leads to significant reductions in cold nociceptive responses, where across the analyzed behavioral metrics Basin-2 had stronger deficits in cold-evoked responses (Figure 2). Furthermore, neural co-activation analyses revealed that Basin-2 plus CIII md led to enhanced CT response, whereas Basin-4 plus CIII md co-activation did not facilitate but rather suppressed the CT response (Figure 2). Basin-2 neural activity likely has cascading effects on functions of downstream neurons (i.e. TePns, A02m/n, Chair-1, and A00c), which we also found to be required for cold nociception. Behavioral assessment of multisensory integration neurons supports the notion that CIII md neuron sensory input is processed primarily by Basin-2 and CIV md neuron sensory cues are processed via Basin-4 neurons.

Despite the accumulating evidence that cold-evoked responses are primarily mediated by Basin-2, our Ca²⁺ analyses revealed that compared to Basin-2, Basin-4 neurons have much greater cold- or CIII md neuron-evoked Ca²⁺ responses. There are a few possible explanations for greater Basin-4 activity levels. Firstly, greater Basin-4 Ca⁺² activity could be due to greater synaptic input from projection neurons (TePn04/05) and premotor neurons (DnB), which are both required for cold nociception and exhibit cold- or CIII md neuron-evoked Ca²⁺ responses (Figure 7, Figure 8—figure supplement 1; Masson et al., 2020). Secondly, reduced Basin-2 Ca²⁺ responses could be due to greater connectivity with feedback/feedforward inhibitory local neurons (Jovanic et al., 2016; Masson et al., 2020) or descending feedback neuron (SeIN128) (Zhu et al., 2024). Descending inhibitory, SeIN128, neuron functions in transitioning from rolling to fast crawling response in Drosophila larvae (Zhu et al., 2024). SeIN128 receives synaptic input from Basin-2 but not Basin-4, while providing feedback inhibition to Basins (−1,–2, –3), premotor neurons (A02m/n, Chair-1, and DnB) and projection neurons (A00c, A02o, and A05q) (Ohyama et al., 2015; Zhu et al., 2024). Thus, lower cold-/CIII md neuron-evoked Basin-2 neural activity could be due to greater feedback inhibitory input compared to Basin-4. Comparative Ca²⁺ imaging or electrophysiological analyses can further delineate the precise interplay between Basin-2 and –4 neurons in integrating sensory information across somatosensory modalities leading to distinct behavioral responses. However, to date, synaptic connectivity and behavioral analyses collectively indicate an initial sensory discrimination node amongst the Basin neurons, where downstream processing of CIII-mediated sensory input is conveyed through Basin-2 and its downstream connectivity. Meanwhile, CIV somatosensory information is transduced through the Basin-4 neuronal pathway.

Animal locomotive behavioral responses are implemented through neurosecretory systems and muscle contractions evoked by motor neuron activity. Motor neurons integrate a conglomerate of neural impulses from premotor neurons, leading to appropriate muscle activation patterns to implement specific behavioral responses (Bellardita and Kiehn, 2015; Berkowitz et al., 2010; Carreira-Rosario et al., 2018; Green and Soffe, 1996; Huang and Zarin, 2022; Liao and Fetcho, 2008; Talpalar et al., 2013; Zarin et al., 2019). Recent work in premotor and motor neuron connectomics has provided evidence for both labeled line and combinatorial connectivity between promotor and motor neurons that give rise to co-active motor neuron states leading to selective muscle group activation (Cooney et al., 2023; Huang and Zarin, 2022; Zarin et al., 2019). Various Drosophila larval premotor neuron subtypes and/or motor pools have been implicated in locomotion, nociception, and innocuous mechanosensation (Burgos et al., 2018; Cooney et al., 2023; Huang and Zarin, 2022; Kohsaka et al., 2017; Kohsaka et al., 2014; Kohsaka et al., 2019; Yoshino et al., 2017; Zarin et al., 2019). Neural reconstruction efforts in Drosophila larvae have revealed that somatosensory (CIII md and CIV md) neurons are not directly connected to motor neurons, but rather feed into polysynaptic pathways leading to motor neurons via premotor neurons (such as the DnB, mCSI, and Chair-1 neurons studied here) (Cooney et al., 2023; Jovanic et al., 2019; Ohyama et al., 2015; Winding et al., 2023; Yoshino et al., 2017). Premotor neurons (DnB, mCSI, and Chair-1) that receive synaptic input from CIII md neurons are required for cold-evoked responses. However, only premotor neurons previously implicated in nociceptive escape responses, DnB and mCSI, can facilitate CIII-evoked CT responses (Burgos et al., 2018; Lopez-Bellido et al., 2019; Yoshino et al., 2017). Meanwhile, Chair-1 neurons, which are required for innocuous mechanosensation (Jovanic et al., 2019), do not facilitate CIII md neuron-mediated CT responses. We showed, with Ca²⁺ imaging, that Chair-1 neurons are not activated upon cold exposure. It remains unclear whether Chair-1 neuronal function is sufficient to drive specific sensory-evoked behaviors or requires a combinatorial pre-motor neuronal activity for generating evoked behavioral responses. Given the data, we speculate that Chair-1 neurons likely gate innocuous-to-noxious stimulus-evoked behaviors; however, additional functional studies are required to tease out how Chair-1 neurons function in behavioral selection.

Animals detect changes in temperature through a variety of sensors that occupy and function in different spatiotemporal timescales. Thermosensory systems function to detect changes at the epidermal, brain and visceral levels, and these broad thermosensory inputs must be integrated to generate appropriate behavioral responses (Nakamura, 2018). The underlying neural circuitry for thermosensation differs based on body plan and type of thermosensation. In mice, cool and warm sensory signals, which are detected by TRP channels, are transmitted to the spinal dorsal horn, where distinct cool/warm pathways relay epidermal changes in temperature to thalamocortical regions for perception and discrimination (Nakamura, 2018). In Drosophila larvae and adults, thermotaxis occurs through temperature detection via dorsal organ ganglion (DOG) and antennal thermoreceptors, respectively (Frank et al., 2015; Gallio et al., 2011; Hernandez-Nunez et al., 2021; Klein et al., 2015; Liu et al., 2003). Investigations of circuit bases of thermotaxis have revealed different strategies for larval and adult systems. Adult antennal thermoreceptors have distinct hot or cold sensors, which project to the proximal antennal lobe (PAL), where hot and cold information is segregated into discrete regions (Gallio et al., 2011; Macpherson et al., 2015). Thermal coding properties of thermosensory projections neurons (tPNs), whose dendrites are located in PAL, include both slow/fast adapting tPNs and broadly/narrowly tuned tPNs (Frank et al., 2017). Drosophila larval thermosensory DOG contains both warm- and cool-sensing cells that are connected to individual cool or warm projection neurons (Hernandez-Nunez et al., 2021). DOG thermosensory information is further integrated by integration projection neurons (iPNs), which receive inputs from both warm and cool projection neurons (Hernandez-Nunez et al., 2021). The Drosophila larval noxious thermosensory system has distinct hot (CIV md) or cold (CIII md) sensing neurons (Babcock et al., 2009; Tracey et al., 2003; Turner et al., 2016). Larval somatosensory thermal inputs mediated by CIII md neurons are heavily integrated by second order neurons (Basins, DnB, A09e, etc.), where population coding leads to cold-evoked behavioral response. Noxious cold stimulus is encoded downstream of CIII md neurons by ascending neurons (A09e and TePns) that allow for systems-level integration of multiple thermal inputs including those from cool/warm sensitive larval DOG neural circuits and cold/heat information from CIII/CIV md neurons, respectively. Thermal information from the larval DOG and somatosensory neurons converges onto the mushroom body, where cool or warm projection neurons downstream of DOG and feed-forward projection neurons downstream of ascending neurons (A09e) synapse onto the same mushroom body output neuron (Eichler et al., 2017; Hernandez-Nunez et al., 2021; Winding et al., 2023). Thus far, ascending pathways into the brain have been characterized for larval cold nociception; however, descending output neurons responsible for cold-evoked behaviors remains unidentified. Diversity in processing thermosensory stimuli reflects heterogeneity at molecular, cellular, and circuit levels that collectively function to maintain proper thermal homeostasis and ultimately behavioral action selection. The Drosophila larval thermosensory system is well-suited to dissect mechanisms underlying central nervous system integration of varying spatial and functional thermosensory information by utilizing various molecules, sensory systems, and interconnected circuits for appropriate behavioral responses.

Animals interact with the external environment by detecting external cues via sensory systems that are transduced to the CNS, where motor commands are generated to elicit stimulus-relevant muscle activity patterns leading to behavioral responses. Drosophila larval body wall muscle have unique behavior relevant muscle activity patterns for locomotion and nociceptive responses (rolling and CT;) (Cooney et al., 2023; Heckscher et al., 2012; Liu et al., 2023; Zarin et al., 2019, This work). In forward locomotion, posterior segments contract before anterior segments generating a forward contraction wave, where the longitudinal (dorsal and ventral) muscles contract first during the wave phase and lateral muscles contract during the interwave interval (Heckscher et al., 2012; Liu et al., 2023; Zarin et al., 2019). Exposure to nociceptive heat, mechanical or chemical cues larvae execute a corkscrew rolling response (Babcock et al., 2009; Hu et al., 2017; Hu et al., 2020; Hwang et al., 2012; Im et al., 2018; Im et al., 2015; Kaneko et al., 2017; Lopez-Bellido et al., 2019; Ohyama et al., 2013; Ohyama et al., 2015; Robertson et al., 2013; Tracey et al., 2003). In rolling response, multiple abdominal segments respond synchronously for executing a roll (Cooney et al., 2023). Furthermore, within a hemi-segment, the muscle activation pattern moves circumferentially during the roll durations, where DL, DO, and VL muscles are active during the bend phase and LT, VO, & VA muscles transition the larval hemi-segment into a stretch phase completing a roll (Cooney et al., 2023). In cold-evoked CT responses, anterior and posterior segments reach peak activation before central segments. Within a hemi-segment individual muscles groups have unique activation profiles, where upon cold exposure both dorsal and ventral muscles reach peak response before lateral muscles. Thus, we can conclude the following. For both cold-evoked CT and locomotion, longitudinal (dorsal and ventral) muscles contract before lateral muscles, whereas rolling response has a mixed activation pattern. The larval segments respond uniquely and are dependent on the behavior being generated. In forward locomotion, there is a propagating wave of activation from posterior to anterior; in rolling, there is a synchronous activation of segments; and in CT response, anterior/posterior segments contract before central. The underlying circuitry responsible for generating these three unique behaviors differs based on sensory, central, and motor processing modules.

All Drosophila melanogaster strains used in this study are listed in (Table 1). All Drosophila reagents were maintained on standard cornmeal-molasses-agar diet in 12:12 hr light-dark cycle at ~22°C. All experimental crosses were raised in 12:12 hr light-dark cycle at 29°C, unless otherwise stated. We used two separate CIII md neuron driver lines in this study:19–12^GAL4, R83B04^GAL4 and R83B04^lexA.

GMR83B04^GAL4 was characterized as CIII md neuron driver using both optogenetics and visualization using membrane markers (Galindo et al., 2023; Himmel et al., 2023; Patel et al., 2022). The R83B04 enhancer containing entry vector was a gift from the FlyLight team at Janelia Research Campus, Ashburn, VA. We performed Gateway cloning to insert the R83B04 enhancer upstream of lexA. LR reaction was performed using R83B04 containing entry vector and lexA containing (pBPLexA::p65Uw) destination vector (Addgene: 26231). Transgenic fly generation was conducted by GenetiVision. R83B04^lexA was inserted in the VK20 docking site using PhiC31 integrase-mediated transformation.

All data for synaptic connectivity, circuit diagrams and wire frame projections of neural cell types were extracted from Neurophyla LMB Cambridge (https://neurophyla.mrc-lmb.cam.ac.uk/). Relevant primary literature sources are cited within the main text. Sensory neurons from abdominal segments 1–4 were analyzed and for the remaining cell types, all reconstructed neurons were included (date accessed: March 04, 2022).

We assessed requirements of neurons downstream from peripheral sensory neurons in cold-evoked responses using the cold plate assay (Patel and Cox, 2017; Patel et al., 2022; Turner et al., 2016). GAL4 drivers for select neuronal subtypes were used to express the light chain of tetanus toxin (UAS-TNT). We used two sets of controls: w1118, genetic background control, and Empty^GAL4, which contains a GAL4 construct but no regulatory promoter (GAL4 only control). Genetic crosses were raised on standard cornmeal-molasses-agar diet and at 29°C. To assess cold-evoked behavioral responses of age-matched Drosophila third instar larvae, we exposed the ventral surface to noxious cold (10°C) temperatures. Briefly, using a brush we remove third instar larvae from food and place them on wet Kimwipe. Food debris is removed passively by allowing the larvae to freely locomote on wet Kimwipe. We place 6–8 larvae on a thin black metal plate that is subsequently placed on a pre-chilled (10°C) Peltier plate TE technologies Peltier plate (CP-031, TC-48–20, RS-100–12). Larval responses are recorded from above using Nikon DSLR (D5300). Changes in larval surface area were extracted using FIJI and Noldus Ethovision XT (GitHub, copy archived at Patel and CoxLabGSU, 2025; Patel et al., 2022) (https://github.com/CoxLabGSU/CaMPARI-intensity-and-cold-plate-assay-analysis/tree/AtitAPatel-cold_plate_assay). Next, we isolated individual larva from each video, removed the background, and used Ethovision to measure larval surface area. Using custom built r scripts, we compiled data from each larva and each genotype. Utilizing r, we calculated percent change in larval surface area (Area change = (Area_N – Average_Area_baseline)/ Average_Area_baseline*100). From the percent change in area dataset, we report three behavioral metrics: Average change in area, which is average percent change in area for the stimulus duration. We defined cold-evoked CT response as a change in area of –10% or less for at least 0.5 consecutive seconds. CT duration, time spent at or below –10% change in area. Lastly, percent CT response, which is cumulative percent of animals that CT for at least 0.5 consecutive seconds.

Statistical analysis: We performed the following statistical tests for all cold plate assay data analysis. %CT response: Fisher’s exact with Benjamini-Hochberg for multiple comparison. We used r for performing comparisons of percent behavior response between genotypes, we used Benjamini-Hochberg multiple comparison correction. CT duration: Kruskal-Wallis with Benjamini, Krieger and Yekutieli for multiple comparisons. CT duration data are not normal. CT magnitude: One-way ANOVA with Holm-Šídák’s for multiple comparisons.

We performed two types of neural activations to assess sufficiency for CT response: single downstream neural activation and co-activation, where we simultaneously activated CIII md neurons and individual downstream neurons. For optogenetic experiments in Drosophila, a light-sensitive co-factor all trans-retinal (ATR) is required. For all conditions, all adult animals in the genetic cross were placed in ATR (1500 µM) supplemented food and subsequently F1 progeny were also raised in food containing ATR and raised in dark. For control condition, we used an Empty^GAL4 containing GAL4 construct but no regulatory promoter. Optogenetic experiments were conducted using a similar setup as previously described (Patel et al., 2022). Briefly, we used principles of dark field microscopy to enhance signal to noise ratio and capture high resolution larval videos. We created a custom dark field stage, where a Canon DSLR T3i camera captures video from above. Neural activation is performed by two blue led lights that are controlled remotely using the Noldus control box (Thorlabs: DC4100, DC4100-hub, and two M470L3-C4 led light. Noldus: mini-IO box). Larval behaviors are directly captured using Noldus Ethovision XT software, which also controls blue led activation. All optogenetic experiments were conducted in a dimly lit room. Third instar Drosophila larva were removed from food plug and placed onto wet Kimwipe, where larval locomotion allowed for passive removal of food debris. We lightly sprayed water onto a thin glass plate, then placed a single larva for optogenetic stimulation. The glass plate was manually moved on XY-axis to keep larva in the field of view. The following stimulus paradigm was used: 5s of baseline (light off) and 5s of neural activation (blue led lights on).

We performed video processing and behavioral analysis using FIJI and data compilation and analysis using r (GitHub, copy archived at CoxLabGSU, 2025). In order to analyze Drosophila larval behavioral responses, we first automatically stabilized (XY axis) and then measured changes in larval surface area and mobility (described below).

Raw videos from Noldus Ethovision XT were uncompressed using video-to-video convertor (https://www.videotovideo.org/). The following steps were scripted in FIJI macro language for automatic video processing and data acquisition. For increasing processing speed, the uncompressed videos were automatically cropped to dimensions to contain all of the larva’s movement. Next, using a pre-determined threshold, we created a mask followed by background removal using erode, remove outlier and dilate functions. We then used the ‘Analyze Particles’ function to obtain XY coordinates of the larva in each frame. Larval movements were stabilized using XY coordinates, and the ‘Translate’ function was used to create a highly stabilized video. Next, we measured larval surface area using automatic thresholding ‘Huang method’ and ‘Analyze Particles’ to obtain area. We define Drosophila larval mobility as changes in occupied pixels between two frames. We used our stabilized larval video to measure larval mobility, where larval peristaltic movements (XY displacement) are not captured. However, changes in occupied pixels resulting from turning and head sweeps are captured. Specifically, larval mobility was measured by subtracting thresholded larva in each frame from the previous frame (Raw mobility = Thresholded larva_{Frame N} - Thresholded larva_{Frame N-1}).

Data compilation and analysis were performed in r using custom scripts. Optogenetically evoked changes in behavior were analyzed independently for mobility and changes in area. For each larva, we calculated percent change in area (Area change = (Area_N – Average_Area_baseline)/ Average_Area_baseline*100). We measured CT duration as the amount of time the larva has –10% or lower change in area. We also measure CT magnitude by analyzing average change area for stimulus duration. We report percent instantaneous CT over time as percent of animals that are at or below –10% change in area. Additionally, we report peak %CT from percent instantaneous CT dataset for each genotype. For analysis of mobility, we calculated percent change in mobility (Mobility change = (Mobility_N – Average_ Mobility_baseline)/ Average_ Mobility_baseline*100). We report average percent change in mobility during stimulus for each genotype. We also plot percent immobility, which is calculated by the percent of animals with –25% or more reduction in mobility. Immobility duration is calculated based on amount of time individual animals spend at or below –25% mobility.

Statistical analysis: We performed the following statistical tests for all behavioral optogenetic data analysis. %CT peak response: Fisher’s exact with Benjamini-Hochberg for multiple comparison. We used r for performing comparisons of percent behavior response between genotypes, we used Benjamini-Hochberg multiple comparison correction. CT/immobility duration: Kruskal-Wallis with Benjamini, Krieger, and Yekutieli for multiple comparisons. CT duration data are not normal. CT/immobility magnitude: One-way ANOVA with Holm-Šídák’s for multiple comparisons.

For assessing cold-evoked larval behavior across life stages, we developed a specialized behavior apparatus that improved to signal-to-noise ratios, allowed for similar stimulus magnitude between trial and across larval life stages, and an improved assay throughput. The apparatus consists of VWR circulating water chiller (model: MX7LR20), two aquarium mini-pumps (model: AD20P-1230D), custom 3D-printed water chamber with glass 1 inch glass walls, a remote control to activate the aquarium pumps (Syantek model: BHZ0320U-RF), three-way water valve (Malida model: 607569660968), iPad screen protector (Ailun model: B0BJQ24MK7), Canon R6 with RF50mm F1.8 and RF85mm F2 macro lens and led lights for illumination from all sides (Daybetter model: FLSL-DB-501215RGBUS). The larvae were removed and cleaned as described above, then placed on a thin transparent glass plate (screen protector). The glass plate bearing larvae is placed on a 3D-printed water chamber, then pre-chilled water is piped into the chamber using mini-aquarium pumps. After the water touches the bottom of the glass plate a three-way water valve is turned off and the chilled water remains in the 3D-printed water chamber providing cold stimulation to the larvae through the glass surface. At the end of the trial, the water valve is opened to drain the water into the VWR water chiller. For first and second instar larval assays, an 85 mm RF macro lens was used at different distances to increase magnification. A 50 mm RF lens was used to image third instar larval cold-evoked responses.

To quantitatively analyze the larval responses, we used slightly modified versions of FIJI and r macros as described above. In the FIJI macro, individual larvae were extracted into their individual video file and the background was removed. A technical challenge arose in the new assay, where each larva gets the cold stimulus at slightly different time points as the water fills up in the water chamber. There is a minor but consistent change in brightness/contrast for each larva, when the water below the glass plate touches the bottom of the glass plate. Using Noldus Ethovision, changes in brightness/contrast are quantified and then using r, the time of stimulus delivery is obtained. After determining the first frame automatically in r, the analysis for various cold-evoked CT response metric are obtained using the same r code as described above.

Statistical analysis: We performed the following statistical tests for all behavioral data analysis. CT magnitude and duration: One-way ANOVA with Holm-Šídák’s for multiple comparisons.

To evaluate CIII-evoked CT responses across life stages, we utilized optogenetic neural activation of CIII md neurons using ChR2-H134R±ATR. In order to maintain consistent blue-light-based neural activation, we used Canon R6 camera mounted on AxioScope V16 Zeiss microscope for visualizing Drosophila larvae at life stage specific magnification while applying oblique blue light exposure (Thorlabs: DC4100, DC4100-hub, and one M470L3-C4 led light). We recorded larva responses without blue light (pre-stimulus) for 5s and 5s of blue light exposure. To quantitatively analyze larval responses, we used a similar method as described above for neural activation.

Statistical analysis: We performed the following statistical tests for all behavioral optogenetic data analysis. CT magnitude and duration: One-way ANOVA with Holm-Šídák’s for multiple comparisons.

For visualizing cold-evoked Ca²⁺ responses of larval muscles in each segment, we expressed jRCaMP1a in Drosophila larval muscles using mef2^GAL4 and motor neuron silencing using GtACR1 with muscle Ca²⁺ imaging. Cold stimulus was delivered using the CherryTemp temperature controller (Cherry Biotech: Cherry Temp dual channel temperature control), which allows for rapid changes in temperature. To analyze Drosophila larval cold-evoked segmental response, we used AxioScope V16 Zeiss microscope with a Canon R6 camera mounted. In vivo jRCaMP1a responses were observed in third instar larva, where the larva was mounted with dorsal midline centered on the top. Using FIJI, changes in jRCaMP1a fluorescence intensities in each segment were quantified with manually drawn regions of interest (ROI). We report changes in jRCaMP1a fluorescence as ΔF/F = (F-F_prestimulus)/ F_prestimulus*100 and average ΔF/F for each larval segment.

Drosophila larva’s individual muscles cold-evoked Ca²⁺ responses were visualized using mef2^GAL4>jRCaMP1 a and Zeiss LSM 780 confocal microscope with EC Plan-Neofluar 10 x objective. Time-lapse acquisition was acquired at 500.11µm x 500.11 µm x 307.2ms using 561 nm laser wavelength. Three different imaging planes (dorsal, lateral, and ventral) were used to capture the majority of the muscles of a hemi-segment. Using FIJI, changes in jRCaMP1a fluorescence intensities in each muscle were quantified with manually drawn ROIs. We report changes in jRCaMP1a fluorescence as ΔF/F = (F-F_prestimulus)/ F_prestimulus*100 and maximum ΔF/F for each muscle. No statistical analyses were performed.

The following temperature exposure paradigm was used for both segmental and individual muscle cold-evoked Ca²⁺ response assays: Pre-stimulus (25°C for 45s), temperature ramp down, steady state cold (5°C for 15s), temperature ramp up, and post-stimulus (25°C for 40s). Temperature ramp duration 10s.

We used the above-mentioned Cherry Temp. dual channel temperature control system, Canon R6 and Zeiss V16 to visualize cold-evoked Ca²⁺ responses for assessing delay in mouth hook retraction (10°C stimulus) and ethyl ether pharmacological application and muscle post-synaptic density Ca²⁺ imaging (5°C stimulus). We used the following temperature exposure paradigm: pre-stimulus (20s), 10s ramp down, steady state cold (10s), temperature ramp up, baseline (10s), temperature ramp down, steady state cold (10s), temperature ramp up, baseline (10s), temperature ramp down, steady state cold (10s), temperature ramp up, and baseline (20s). The temperature ramp duration was10 seconds. For muscle Ca²⁺ fluorescence analyses, we used ΔF/F = (F-F_prestimulus)/ F_prestimulus*100. For assessing, CIII md Ca²⁺ increase and delay in mouth hook retraction, we used ΔF/F = (F-F_prestimulus)/ F_prestimulus*100 for CIII md Ca²⁺ levels and visually determined the time at which the mouth retracted.

Statistical analysis: Delay in mouth hook retraction - Mann-Whitney test. For motor neuron silencing assay - Two-way ANOVA with Holm-Šídák’s for multiple comparisons. For pharmacology assay - Mann-Whitney test for non-parametric data and Welch’s t-test for parametric data.

Post-synaptic neuron CaMPARI2 imaging: For assessing cold-evoked Ca²⁺ responses of sensory neurons and CIII md neuron downstream neurons, we utilized CaMPARI2, which upon photoconverting light and high intracellular Ca²⁺ stably photoconverts fluorescence from green to red. We performed the cold plate assay as described above. For the stimulus condition, the Peltier plate was set to noxious cold (6°C) temperature, and for the control condition, the Peltier plate was turned off (room temperature). We placed individual third instar larvae onto the Peltier plate and simultaneously exposed the animal to photoconverting light for 20 seconds. CaMPARI2 fluorescence was imaged in live, intact larvae via confocal microscopy. Zeiss LSM 780 Axio examiner microscope, Plan-Apochromat 20x objective, and excitation wavelengths of 561nm and 488nm were used to image larval ventral nerve cord or peripheral sensory neurons. We mounted live intact larva onto microscope slide and immobilized the larva by placing a coverslip, as previously described (Im et al., 2018; Patel and Cox, 2017; Patel et al., 2022; Turner et al., 2016). Three-dimensional CaMPARI2 fluorescence in ventral nerve cord localized downstream neurons was imaged at 607.28µm x 607.28µm (XY resolution) and 2µm z-slices. ROIs were identified and area normalized fluorescence intensity for red and green signals were obtained via Imaris 9.5 software. CaMPARI responses are reported as a ratio of F_red/F_green. Statistical comparisons: Parametric data – Welch’s t-test and non-parametric data – Mann-Whitney test.

PNS CAMPARI2 imaging: Sensory neuron CaMPARI2 responses were analyzed using custom FIJI macros that automatically detected cell bodies and sholl intensity analyses were performed using semi-automated custom FIJI macros (GitHub; Patel and CoxLabGSU, 2025; Patel et al., 2022) (https://github.com/CoxLabGSU/CaMPARI-intensity-and-cold-plate-assay-analysis/tree/AtitAPatel-CaMPARI_Analysis).

Cell body analysis – We created a set of three sequential macros that draw ROIs around the cell body, user verification of the ROIs and lastly quantification of fluorescence intensities. The first custom FIJI script generates maximum intensity projections of z-stacks. Image masks were created by thresholding (Moments method) GFP signal, and next, background and dendritic branches were removed using erode and dilate functions. At the end of background clearing only cell bodies remain, where Analyze particle function is used to draw ROIs around soma. The second FIJI script is used for manual verification of each ROI and manually redrawing any incorrect ROIs. Lastly, upon ROI verification, area normalized F_red and F_green intensities are quantified. As previously described (Fosque et al., 2015; Im et al., 2018; Patel and Cox, 2017; Patel et al., 2022; Turner et al., 2016), we report evoked photoconverted CaMPARI signal as F_red/F_green ratio. Statistical comparisons: Non-parametric data – Mann-Whitney test.

Sholl intensity analysis – Sholl intensity analysis was performed using a set of two custom FIJI scripts as previously described (Patel et al., 2022). Briefly, we first perform background clearing by manually thresholding the GFP signal and select all branches and soma of neuron of interest using the ‘Wand Tool’ in FIJI and then a mask of neuron of interest is created. The second FIJI script draws five-pixel wide radial ROIs at a single pixel interval, here only the dendrites and soma from the neuron of interest are selected at each radial interval. After Sholl ROIs are drawn, area and area-normalized F_red and F_green fluorescence intensities are extracted for each radial step away from the soma. Similar to CaMPARI cell body analysis, we measured CaMPARI2 signal as F_red/F_green ratios away from the soma.

Ventral nerve cord CaMPARI imaging: We utilized Pan-neural (R57C10^GAL4>CaMPARI) driver to visualize ventral nerve cord Ca²⁺ responses to various stimuli including innocuous touch, noxious heat (45°C) and noxious cold (6°C). Stimulus and photoconverting light were delivered for 20s. Whole ventral nerve cord was imaged at 607.28µm x 607.28 µm (XY resolution) and 2 µm z-slices. The rest of the stimulus delivery and imaging was similar to previously described CaMPARI experiments. No statistical analyses were performed.

CIII md neuron-evoked responses in downstream neurons were evaluated by using optogenetics and GCaMP. We expressed lexAop-CsCrimson in CIII md neurons using R83B04^lexA and used downstream neuron-specific GAL4 to drive expression of GCaMP6m. For the experimental condition, all adult animals in the genetic cross were reared in ATR (1500 µM) supplemented food, and subsequently, F1 progeny were also raised in food containing ATR. For control condition, adult flies and F1 progeny were raised in standard cornmeal-molasses-agar diet. Both control and experimental crosses were reared in 24 hr dark. We mounted live intact third instar larvae in between microscope slide and a coverslip. Larval ventral nerve cord and cells of interest were located using epifluorescence on Zeiss LSM 780 confocal microscope. Larval GCaMP6 responses were allowed to return to baseline for at least 2 minutes. Time-lapse acquisition was imaged at 250.06µm x 250.06 µm x 307.2ms using 488 nm laser wavelength. CIII md neural activation was performed using two oblique 617 nm leds (Thorlabs M617F2 and M79L01) that were manually operated using Thorlabs led controller (Thorlabs DC4100 and DC4100-hub). We performed three sequential neural activations using the following paradigm: Baseline light off (30s)->neural activation (617 nm for 15s)->light off (30s)->neural activation (617 nm for 15s)->light off (30s)->neural activation (617 nm for 15s)->light off (30s). Time-lapse videos were stabilized using Stack reg – Rigid transformation in FIJI (Linkert et al., 2010; Schindelin et al., 2012; Thévenaz et al., 1998). ROI were manually drawn and area normalized GCaMP fluorescence over time was exported. We report changes in GCaMP6m fluorescence as ΔF/F = (F-F_prestimulus)/ F_prestimulus*100 and max ΔF/F for each of the three neural activation epochs. Statistical comparisons: Parametric data – Welch’s t-test and non-parametric data – Mann-Whitney test.

Statistical analyses were performed using r (Fisher’s exact test) and GraphPad Prism. All graphical visualization of the data were created using Prism GraphPad. Details on specific statistical tests are listed in the respective methods section.