To identify peripheral non-neuronal cell types that contribute to nociception, we conducted an optogenetic screen for light-evoked nocifensive behavior. First, as a benchmark for comparison we used the light-activated cation channel CsChrimson (Klapoetke et al., 2014) to optogenetically activate nociceptive C4da neurons.

Consistent with prior reports (Hwang et al., 2007; Hu et al., 2017), C4da activation triggered nocifensive behaviors including c-bending and rolling in 100% of larvae (Fig. 1A, Fig. 1S1A, Movie S1). Next, we selectively expressed CsChrimson using GAL4 drivers in combination with elav-GAL80, which effectively silences GAL4 expression in larval sensory neurons (Fig. 1S2), to target the six principle non-neuronal cell types within the larval body wall: epidermis, trachea, muscle, hemocytes, oenocytes, and glia (Fig. 1S3, Key Resources Table). We then monitored light-evoked behavioral outputs associated with stimulation of each cell type. We found that optogenetic stimulation of epidermal cells, like C4da neurons, elicited nocifensive c-bending and/or rolling behaviors in 73% of larvae (Fig. 1A, 1S1B), without significantly altering nociceptor morphogenesis (Fig. 1S4). In contrast, stimulation of other body wall cell types elicited a variety of non-nociceptive behavior outputs: for example, muscle stimulation triggered hunching behavior followed by prolonged freezing, whereas glia stimulation reproducibly induced only hunching behavior (Fig. 1S1C-I) (Zimmermann et al., 2009). Thus, epidermal cells are the only non-neuronal body wall cell type that triggers robust nocifensive behavioral responses.

To further validate the selective ability of body wall epidermal cells to drive nocifensive behaviors, we examined eight other epidermal drivers in addition to R38F11-GAL4, which displays no expression in sensory neurons and limited non- epidermal cell expression overall (Fig. 1S5). We found that optogenetic stimulation evoked nocifensive behaviors with each of the eight epidermal driver lines we tested: seven of the lines displayed rolling behavior while all eight displayed c-bending (Fig. 1B, Fig. 1S6). Although the previously described pan-epidermal A58-GAL4 driver (Galko and Krasnow, 2004) drove robust nocifensive rolling responses (Fig. 1B, 1S6), A58- GAL4 is expressed broadly in the larval CNS (Fig. 1S7) and stochastically expressed in sensory neurons (Jiang et al 2014). In contrast, the remaining seven drivers including R38F11-GAL4 exhibited limited expression aside from epidermal cells, with no detectable expression in nociceptors, other larval SSNs, or peripheral glia, and highly restricted or undetectable expression in the CNS (Fig. 1S7). Further underscoring the connection between epidermal stimulation and nocifensive responses, the nocifensive behavioral response with these epidermal drivers correlated with the proportion of epidermal expression (Fig 1B).

We next used thermogenetic stimulation with the warmth-activated TRP channel dTRPA1 (Hamada et al., 2008) as an independent method of probing nociceptive responses triggered by epidermal cell activation. On its own, the thermal stimulus (35° C) rarely induced rolling behavior in control larvae bearing UAS-TRPA1 alone. In contrast, we found that >75% of larvae expressing TRPA1 in all nociceptors exhibited rolling behavior in response to a thermal stimulus (Fig. 1C). Likewise, thermogenetic activation of epidermal cells induced robust rolling responses in >75% of larvae, and addition of GAL80 transgenes (tsh-GAL80 elav-GAL80) that silenced the sparse R38F11-GAL4 VNC expression (Fig. 1S5) had no effect on the rolling frequency (Fig. 1C, Fig. 1S3). Altogether, these results demonstrate that epidermal stimulation evokes nocifensive responses in Drosophila. Of note, prior studies demonstrated that sparse thermogenetic activation of nociceptors (<5 cells) yielded no significant increase in nocifensive rolling whereas activation of >10 cells was required to elicit rolling responses in a majority of larvae (Robertson et al., 2013). Hence, epidermal stimulation likely engages numerous C4da neurons to elicit these behavioral responses.

In addition to C4da nociceptors, the epidermis is innervated by a variety of other SSNs including mechanosensory C3da and chordotonal (Cho) neurons and proprioceptive C1da neurons. Whereas direct stimulation of C4da nociceptors principally elicited nocifensive behavioral outputs, epidermal stimulation elicited an array of behaviors in addition to nocifensive responses, including freezing and hunching (Fig. 2A, 2B, Fig. 2 video 1 and 2), behaviors associated with stimulation of C3da and Cho neurons (Zhang et al., 2013; Turner et al., 2016). These data suggest that epidermal cells may broadly modulate SSN activity in Drosophila.

To examine whether different epidermis-evoked behaviors were associated with activation of distinct classes of SSNs, we compared epidermis-evoked and SSN-evoked behaviors. Stimulation of C4da, C3da and Cho neurons elicited distinct behavioral motifs: only C4da neurons elicited rolling behavior; stimulation of C3da and Cho neurons together elicited hunching, C-bending, and backing; stimulation of Cho neurons alone principally elicited hunching and freezing responses (Fig. 2A-2C, 2F). In contrast, optogenetic epidermal stimulation elicited all of these behaviors, with nocifensive behaviors (c-bending, rolling) predominating initially, followed by non-nociceptive behaviors (backing, freezing) (Fig. 2D, 2F, 2S1). We note that neither the behavioral motifs induced by epidermal or SSN stimulation nor the behavioral sequence induced by epidermal stimulation was recapitulated in effector-only controls (UAS-CsChrimson ATR+; Fig. 2S1E), demonstrating that the observed responses were driven by activation of the respective cell types.

We observed three striking differences in behavior evoked by stimulation of epidermal cells versus individual SSNs. First, although rapid, latency to rolling was significantly longer following epidermal stimulation compared to stimulation of C4da (Fig. 2F). Second, the duration of rolling, bending, and backing responses was significantly longer for epidermis versus SSN stimulation (Fig. 2G). Third, backing and freezing behaviors persisted beyond the duration of the light stimulus for epidermis but not SSN stimulation (Fig. 2H). In summary, we find that epidermal stimulation triggers more robust, varied and prolonged behaviors compared to responses from direct stimulation of discrete SSN subtypes.

We next asked whether epidermal stimulation activates larval SSNs including C4da, C3da, C1da, and Cho neurons. To test this possibility, we developed a semi-intact larval preparation in which we optogenetically stimulated epidermal cells while simultaneously monitoring calcium responses in axon terminals of SSNs (Fig. 3A). We found that epidermal stimulation triggered rapid and robust calcium transients in nociceptive C4da neurons, responses that were not observed in the absence of ATR or in effector-only controls (Fig. 3B). Epidermal stimulation likewise evoked calcium transients in mechanosensory C3da and Cho neurons, and in proprioceptive C1da neurons (Fig. 3C- 3E, 3S1). Hence, epidermal stimulation can broadly modulate activity of larval SSNs.

We next tested the requirement for SSN synaptic transmission in epidermis- evoked behaviors. We stimulated epidermal cells with CsChrimson while blocking SSN neurotransmitter release using tetanus toxin light chain (TnT) (Sweeney et al., 1995). We found that inhibiting C4da or C3da + Cho neurotransmission significantly reduced the frequency and duration of epidermal-evoked rolling and backing behaviors, respectively (Fig. 3F, 3G, 3S2). These data suggest that C4da and C3da/Cho neurons act downstream of epidermal cells to drive behaviors. We note that TnT expression in C4da neurons did not completely block epidermis-evoked nocifensive behaviors, and this likely reflects both incomplete C4da neuron silencing and epidermal activation of other SSNs that promote nociceptive outputs including C3da neurons, C2da neurons, and Cho neurons (Ohyama et al., 2015; Hu et al., 2017; Burgos et al., 2018). Further, silencing C4da or C3da/Cho neurons while stimulating epidermal cells led to an increase in the non-nocifensive behaviors hunching and freezing (Fig. 3F, 3G). These results, along with the observation that rolling behaviors predominate the early behavioral responses to epidermal stimulation (Fig. 2B), suggest that the nervous system prioritizes nocifensive behavioral outputs following epidermal stimulation. These data support a model in which epidermal cells and SSNs are functionally coupled.

What is the physiological relevance of this functional coupling of epidermal cells and SSNs? To address this question, we compared calcium responses in C4da neurons to either simultaneous epidermal and C4da stimulation or C4da stimulation alone. Simultaneous stimulation significantly enhanced the magnitude and duration of calcium responses in C4da axons (Fig. 4A-4D, 4S1). Based on this prolonged calcium response, we hypothesized that simultaneous epidermis and C4da neuron stimulation would yield enhanced nocifensive behavior output. To test this, we optogenetically stimulated C4da neurons and epidermal cells individually or in combination using low intensity CsChrimson activation and monitored larval behavior responses. In this stimulation paradigm, simultaneous epidermal cell and C4da neuron stimulation resulted in rolling in 100% of larvae whereas selective stimulation of C4da neurons or epidermal cells induced rolling in only 63% or 18% of larvae, respectively (Fig. 4E, 4F). Furthermore, simultaneous stimulation elicited a significantly higher number of rolls among responders than stimulation of nociceptors or epidermal cells alone (26.9 rolls for C4da + Epi, 4.9 for C4da, and 5.3 for Epi stimulation; Fig 4G, 4H). Likewise, simultaneous stimulation significantly reduced the latency to the first roll (Fig. 4I) and increased the duration of rolling behaviors (Fig. 4J). We next tested whether this functional coupling extends to mechanical stimuli. We simultaneously presented larvae with a noxious mechanical stimulus and a low intensity optogenetic epidermal stimulus that was insufficient to trigger rolling on its own (0% response rate, n = 200). This concurrent epidermal stimulation significantly increased touch-evoked nocifensive responses, yielding a 91% or 49% increase in rolling responses to 20 mN or 50 mN Von Frey stimulus, respectively (Fig. 4K). We next probed the kinetics of this epidermis- induced mechanical sensitization.

When Drosophila larvae are presented with two nociceptive mechanical stimuli in succession, they exhibit enhanced behavioral responses to the second stimulus (Hu et al., 2017). We hypothesized that selective epidermal stimulation would sensitize larvae to subsequent nociceptive mechanical stimuli. To test this hypothesis, larvae expressing the warmth-activated calcium-permeable channel dTRPA1 in epidermal cells were presented with a thermal stimulus, 32° C to activate dTRPA1, followed by a 40 mN mechanical stimulus 10 s later (Fig. 4L). Indeed, we found that dTRPA1-mediated epidermal stimulation significantly sensitized larvae to a subsequent mechanical stimulus, increasing the roll probability more than two-fold. In contrast, dTRPA1- mediated stimulation of C4da neurons did not induce mechanical sensitization, and we confirmed this result with two independent C4da neuron drivers (Fig. 4L). Thus, activation of epidermal cells but not C4da nociceptors alone induces prolonged sensitization to noxious mechanical stimuli. We next assessed the duration of sensitization following transient epidermal activation. Thermogenetic epidermal stimulation yielded persistent sensitization that recovered over a timescale of minutes (τ = 337 sec, Fig. 4M, 4N). The magnitude and duration of mechanical sensitization by thermogenetic epidermal stimulation was remarkably similar to sensitization evoked by a prior mechanical stimulus (63% roll probability in response to a second stimulus, τ = 334 sec, Fig. 4N, 4S1B). Altogether our data support a model whereby epidermal cells are mechanosensitive cells that signal to SSNs to drive acute nocifensive behaviors and prolong mechanical sensitization.

Prior studies have shown that vertebrate epidermal cells directly respond to mechanical stimuli (Koizumi et al., 2004; Haeberle et al., 2008; Tsutsumi et al., 2009; Ranade et al., 2014; Woo et al., 2014; Moehring et al., 2018). Therefore, we next assessed whether Drosophila epidermal cells are intrinsically mechanosensitive. We developed a protocol to acutely dissociate epidermal cells and measure the responses of individual GCaMP6s-expressing epidermal cells to mechanical stimuli (Fig. 5A). We found that radial stretch elicits calcium responses in epidermal cells in a dose-dependent manner. For example, a low 0.5% stretch activated 18% of cells and a subsequent 1% stretch recruited an additional 10% of stretch-responding cells (Fig. 5B-5D). Overall, 51% of epidermal cells displayed stretch sensitivity (Fig. 5C, 5D). We also found that 43% of epidermal cells responded to hypoosmotic challenge and 35% responded to laminar flow; 19% of epidermal cells responded to both hypoosmotic challenge and laminar flow (Fig. 5S1). Given that dissociated epidermal cells were intrinsically mechanosensitive, we next assessed mechanically evoked responses in a semi-intact body wall preparation (Fig. 5E). We found that 50% of epidermal cells exhibited a robust calcium transient in response to a 25 μm membrane displacement using a glass probe (Fig. 5E- 5G, 5S1G). Altogether, these results indicate that Drosophila larval epidermal cells are intrinsically mechanosensitive.

Our studies demonstrate that, like vertebrate keratinocytes, Drosophila epidermal cells exhibit mechanically evoked calcium transients. What is the mechanism of mechanotransduction in these cells? RNA-seq analysis of acutely dissociated epidermal cells revealed expression of more than 20 cation channels, including the mechanosensitive ion channels Piezo, TMEM63, and TMCO (Fig. 6S1). We assessed the epidermal requirements of these channels in mechanical nociception using available RNAi transgenes (Fig. 6A). Our behavioral screen identified one channel, Orai, the sole Drosophila pore-forming subunit of the Ca²⁺ release-activated Ca²⁺ (CRAC) channel (Feske et al., 2006), that blocked mechanically-evoked nociceptive sensitization without impacting behavioral responses to the first stimulus (Fig. 6A, 6B, 6S2A) or altering nociceptor morphogenesis (Fig. 1S4). Interestingly, our screen uncovered an epidermal role for Task6, an orthologue of stretch-sensitive 2-pore potassium channels (Fink et al., 1996), in mechanonociception, as Task6 RNAi increased nocifensive rolling responses to the initial mechanical stimulus (Fig. 6S2B). Finally, although our RNAi studies did not reveal an epidermal requirement for other known mechanosensitive cation channels in mechanonociceptive behaviors, it is possible that multiple channels function redundantly, or that RNAi knockdown was incomplete.

To gain insight into mechanically evoked nociceptive sensitization, we focused on probing the role of Orai in epidermal mechanosensory responses. We first asked whether Orai is functional in Drosophila epidermal cells. Orai is a store-operated calcium (SOC) channel that is activated by the calcium-sensitive, endoplasmic reticulum (ER) molecule Stim, upon calcium release from ER calcium stores. Thapsigargin (TG) induces calcium release from intracellular stores and thus triggers Stim-dependent activation of Orai channels. Indeed, Drosophila epidermal cells displayed TG-induced calcium release from stores in the absence of extracellular calcium, followed by calcium influx upon re-addition of extracellular calcium (Fig. 6C). Calcium influx was significantly inhibited by the addition of low nanomolar lanthanum, consistent with the high sensitivity of Orai channels to lanthanides (Fig. 6S2C). This characteristic store operated calcium entry (SOCE) response was significantly reduced by epidermis-specific Stim or Orai RNAi knockdown (Fig. 6S2D-6S2F). Consistent with a key role for SOCE in mechanotransduction, we found that radial stretch in the absence of extracellular calcium induced calcium release from intracellular stores, as well as calcium influx upon re-addition of extracellular calcium. These data show that mechanically evoked responses in epidermal cells involve both ER calcium release and store-operated calcium entry. While both store release and calcium influx constitute the calcium response to stretch, in 69% of cells, calcium due to store release exceeded that of calcium re-entry (Fig. 6E). Consistent with this observation, depletion of intracellular stores and inhibition of calcium influx reduced the number of stretch sensitve cells by 61% (stretch non-responsive cells in WT = 49% vs. store depleted = 80%) and 30% (stretch non-responsive cells in WT = 49% vs. La³⁺ = 64%; Fig. 6F-G), respectively.

Given that Stim and Orai mediate SOCE, we investigated requirements for epidermal Stim and Orai in mechanically evoked calcium responses. RNAi knockdown of either Stim or Orai significantly reduced the fraction of stretch-responsive epidermal cells (RNAi control = 48%, Stim RNAi = 22%, Orai RNAi = 24%; Fig. 6H-I), with Stim or Orai RNAi preferentially attenuating stretch evoked responses to larger magnitude stretch stimuli. We also found that human keratinocytes display dose-dependent stretch evoked calcium responses, though they respond to higher magnitudes of stretch than Drosophila epidermal cells (Fig. 6J). Like Drosophila epidermal cells, both ER calcium release and store-operated calcium entry constitute the mechanically-evoked calcium responses in human keratinocytes (Fig. 6K).

Two hallmarks of Orai channels are steep inward rectification, with larger currents at hyperpolarizing potentials, and highly cooperative Orai activation by Stim (Hoover and Lewis, 2011). Since Stim and Orai mediate mechanical responses of epidermal cells in vitro, we predicted that increasing the calcium driving force through Orai activity by either hyperpolarizing Drosophila epidermal cells or by activating additional Orai channels via Stim overexpression would enhance behavioral responses to mechanical stimuli. Indeed, we found that hyperpolarizing epidermal cells with the light-activated anion channelrhodopsin GtACR1 (Mohammad et al., 2017) increased behavioral responses to mechanical stimuli (Fig. 6L). In addition, overexpressing Stim in epidermal cells significantly enhanced nocifensive behavioral responses to mechanical stimuli (Fig. 6M). Altogether, these results demonstrate that mechanically evoked responses of epidermal cells and the resulting nocifensive behavior outputs require store-operated calcium entry.

How might mechanically evoked calcium entry in epidermal cells drive nociceptor activation and behavior? Stim/Orai-mediated calcium entry contributes to exocytosis in a variety of cell types, including neurons and immune cells (Pores-Fernando and Zweifach, 2009; Ashmole et al., 2012; Maneshi et al., 2020; Chanaday et al., 2021; Ramesh et al., 2021). Therefore, we investigated the contribution of epidermal exocytosis in nociceptive sensitization with the temperature-sensitive dynamin mutant shibire^ts (shi^ts) to inducibly block vesicle recycling, as this treatment rapidly and potently blocks neurotransmitter release (Koenig et al., 1983) and we found that acute epidermal dynamin inactivation using UAS-shi^ts had no discernable effect on nociceptor morphogenesis (Fig. 1S4). In this paradigm, larvae expressing shi^ts in epidermal cells, but not control larvae, exhibited significant attenuation of mechanically evoked nociceptive sensitization following pre-incubation at the non-permissive temperature (Fig. 6N). In contrast, both genotypes exhibited comparable responses to a mechanical stimulus at the permissive temperature (25° C) and to the first mechanical stimulus following pre-incubation at the non-permissive temperature (30° C). Taken together, these results are consistent with a model in which mechanical stimuli induce calcium influx and vesicular release from epidermal cells, which in turn activates nociceptors to induce acute nocifensive behaviors and prolonged sensitization (Fig. 6O). Although our RNA-seq analysis of epidermal cells did not reveal expression of neurotransmitter biosynthesis genes, epidermal cells express a large repertoire of genes involved in vesicular release as well as several neuropeptide genes, providing an entry point to defining the molecules involved in epidermis-SSN communication (Fig. 6S3).

Raw sequencing reads and gene expression estimates are available in the NCBI Sequence Read Archive (SRA) and in the Gene Expression Omnibus (GEO) under accession number GSE284380. Raw data used for analyses in this study is presented in the supplementary materials as Source Data and details of statistical analyses are presented in Supplementary File 1. ICMJE guidelines were used to define authorship roles and the ARRIVE essential 10 guidelines were used for the reporting of our in vivo studies.

Flies were maintained on standard cornmeal-molasses-agar media and reared at 25° C under 12 h alternating light-dark cycles. For all experiments involving optogenetic manipulations, larvae were raised in the constant dark at 25 °C on Nutri-Fly Instant Food (Genesee Scientific #66-117), supplemented with 1 mM all-trans retinal (ATR; Sigma #R2500). A complete list of alleles used in this study is provided in the Key Resources Table. Experimental genotypes are listed in figure legends.

A human keratinocyte cell line (HaCaT) was used in this study. HaCaT cells were obtained from Cytion (Sioux Falls, SD), who performed STR authentication and mycoplasma-free certification.

RNA-Seq libraries were prepared from the picked cells following the Smart-Seq2 protocol for full length transcriptomes (Picelli et al., 2014). To minimize batch effects, primers, enzymes, and buffers were all used from the same lots for all libraries.

Libraries were multiplexed, pooled, and purified using AMPure XP beads, quality was checked on an Agilent TapeStation, and libraries were sequenced as 51-bp single end reads on a HiSeq4000 at the UCSF Center for Advanced Technology.

For each experimental assay control populations were sampled to estimate appropriate sample numbers to allow detection of ∼33% differences in means with 80% power over a 95% confidence interval. Details of statistical tests including treatment groups, sample numbers (which correspond to independent biological replicates), statistical tests, p- values and q-values are provided in Supplementary File 1.