The cellular and circuit basis of temperature sensation in Drosophila

Abstract

The thermosensory system enables animals to detect and respond to changes in external temperature and is therefore essential for survival yet remains significantly understudied. This review summarizes current knowledge of its organization in adult Drosophila melanogaster: from peripheral cellular receptors and molecular detection mechanisms to the brain circuits that process thermal information, beginning with second-order thermosensory projection neurons and their targets. The powerful tools available in Drosophila have driven significant advances, revealing the organization of this system at the periphery, the reach of thermosensory pathways within the brain, and the range of behaviors directly influenced by external temperature. These findings also open new avenues to examine how the thermosensory system is reshaped under changing thermal conditions as insects evolve to colonize diverse thermal environments.

INTRODUCTION

Most animal species thrive in a relatively narrow thermal range and possess sensory systems that allow them to avoid sub-optimal temperatures as well as potentially dangerous thermal extremes. In both vertebrates and invertebrates, the sense of temperature — thermosensation— is separate from touch, having dedicated receptor channels and primary sensory neurons^1,2. However, despite its critical role in survival, thermosensation has remained a “Cinderella” amongst the senses, receiving far less attention than modalities such as sight, hearing, smell, and taste. As a result, we know little about the mechanisms that allow us to distinguish hot, cold, and noxious temperatures, or about how different thermal ranges become innately associated with either positive (“warm/cool”) or negative (“hot/cold”) hedonic valence in the brain. Moreover, because temperature sensing is an ancient and fundamental sensory modality, one may expect broad conservation at least at the level of the molecular mechanisms that allow sensory neurons to respond to changes in external temperature. Yet, for example, temperature sensing mechanisms do not appear especially conserved even between flies and mosquitoes^3–5. These perplexing results may be explained in part by the fact that the thermosensory system is subject to continuous evolutionary pressure under constantly changing climatic conditions and may therefore be comparably less evolutionarily stable. But this may also suggest that we do not yet understand its organizing principles in any species, so that the conserved logic is still difficult to see. Precisely in an area like this, work in relatively simple animal systems like the fruit fly Drosophila can prove essential.

Matching the Goldilocks principle, Drosophila flies are not too simple so that their behavior is fundamentally different from our own, and not too complex as to become intractable in their own right⁶. In fact, flies are proving especially well-suited to study the processing of temperature stimuli both at the periphery and in the brain: as small poikilotherms, flies display complex but highly stereotypical temperature responses, ranging from species-specific temperature preference, to robust escape responses to noxious heat, to temperature modulation of sleep and activity rhythms. What is more, the recent release of fly brain connectomes^7,8 has dramatically enhanced our ability to study the central processing of sensory stimuli, allowing us to easily identify brain regions targeted by sensory pathways and to produce hypotheses on how sensory cues may modulate behavioral and physiological responses.

The goal of this review is to highlight recent progress on our understanding of the cellular and circuit mechanisms of temperature sensation in adult Drosophila melanogaster flies, with the hope that this may serve as a useful resource to advance studies on fly thermosensation in flies and other insect species, as well as advancing work that includes thermal stimuli to address more complex aspects of sensory processing such as learning and memory, multi-sensory integration, decision making, and more.

The Drosophila peripheral thermosensory receptor neurons or TRNs

Much like humans, Drosophila flies possess thermosensory neurons that are distinct from mechanosensory neurons that respond to stretch and physical pressure, as well as from multimodal nociceptors that detect noxious thermal stimuli, chemical irritants, and mechanical injury. In adult Drosophila, the best characterized thermosensory neurons reside in (or in close association with) the antenna, a true hub for the senses in the insects (Figure 1a). Six thermosensory receptor neurons (TRNs) are housed in the last antennal segment, the arista, arranged into three thermosensory sensilla⁹, each containing a pair of one TRN activated by heating stimuli (Arista “Hot” Cell, ArHC) and one TRN activated by cooling stimuli (Arista “Cold” Cell, ArCC; Figure 1a,b;¹⁰). Sensilla are insect sensory organs consisting of a cuticular structure (often a hair or a peg) innervated by sensory neurons. Within the arista, each hot-cold TRN pair is tightly apposed and ensheathed by a number of support cells (forming an individual sensillum) yet, unlike for example insect olfactory and taste ones, thermosensory sensilla of the arista are not intimately associated with the cuticle⁹ (Figure 1b). Instead, ultrastructural studies have demonstrated that, while both hot- and cold-activated neurons of the arista have long slender dendrites with the usual subdivision into an inner and an outer segment, the cold cell’s outer segment membrane appears quite distinctive (is highly lamellated and bears unusual particulate protrusions termed BOSS-structures⁹; Figure 1c) potentially reflecting distinct molecular sensing mechanisms¹¹. Each hot-cold TRN pair also appears closely coordinated at the functional level: for example, hot cells respond with a calcium increase to heating, but with a calcium decrease when challenged with the opposite stimulus (cooling), and the opposite is true for cold cells¹⁰ (Figure 1c). Due to their intimate association, these opposing responses may be mediated by some form to ephaptic communication within the pair¹². Arista TRNs are essential for responses to temperature such as rapid thermal preference and thermotaxis^10,13,14, which is sensible as this structure is the first part of the animal to come in contact with new conditions in the surrounding environment as the fly moves. Moreover, the two aristae are symmetrically positioned on the fly head so that small temperature differentials between them can be used to chart efficient navigational trajectories¹⁴. Finally, since the arista is covered by a thin 1–2 μm layer of chitin⁹, a material with good thermal conductivity¹⁵, the aristae match the external temperature nearly instantaneously (as modeled in¹⁴).

Figure 1: Organization of the Drosophila peripheral thermosensory system.

a) The Drosophila antenna houses specialized hot- (red) and cold-activated (blu) thermosensory receptor neurons (TRNs). b) The Arista contains 3 thermosensory sensilla (one is drawn here) each composed of one hot- and one cold-activated TRN, their outer segments appear very different at the ultra-structural level. c) Calcium imaging shows that the arista TRNs have opposing responses to heating and cooling stimuli (4 TRNs are visible here, red dot = heating stimulus; cold dot = cooling stimulus). d) The Sacculus consists of three chambers lined by sensilla. e) Cold activated cells of the sacculus innervate No-pore sensilla coeloconica in chamber 1 and basiconica in chamber 2, including contributing to a “hygrosensory triad” in Chamber 2. f) The Antennal nerve thermosensory ganglion is composed of two heat-activated “Anterior cells” and one cold-activated “anterior cold cell”. Abbreviations: TRNs=thermosensory recepto neuron, ArHC=arista hot cell, ArCC=arista cold cell, sCC=sacculus cold cell, AC=anterior cell, ACC=anterior cold cell, ANTG=antennal nerve thermosensory ganglion, np-SC=no-pore sensilla coeloconica, np-SB=no-pore sensilla coeloconica. Scalebar in c: 20μm.

Panel b was adapted with permission from Foelix et. al. 1989

Panel c was adapted from Gallio et. al. 2011

Panels d and e were adapted with permission from Shanbhag et. al. 1995

A second structure within the antenna housing thermosensory neurons is the sacculus, a small invagination of the third antennal segment (or funiculus, Figure 1a). In D. melanogaster, the sacculus consists of three chambers, lined with olfactory, hygrosensory, and thermosensory sensilla each with distinct structure and properties¹⁶ (Figure 1d). The TRNs described in the sacculus are cold-activated cells that innervate specialized sensilla of chambers 1 and 2^17–21 (Figure 1e). These sensilla lack the openings that allow chemical cues to reach the sensory neurons in the more common olfactory sensilla (for example) and are therefore termed “no-pore” sensilla. Interestingly, the cold activated TRNs of chamber 2 of the sacculus are part of a “triad” of neurons that, together, define a unique humidity sensing mechanism also found in other insects²². The triad consists of a cold-activated, a humid-air activated and a dry-air activated cell all innervating the same sensillum^17–20 (Figure 1e). Each cell type is thought to contribute to the detection of changes in air humidity, either by sensing hygroscopic swelling/shrinking of the cuticular structure encapsulating the hygrosensory sensillum or by monitoring evaporative cooling²².

Finally, a small but important group of thermosensory neurons has been described within the head capsule, forming a small cluster at the base of the antennal nerve (Figure 1b,f). Because of its distinctive position and organization, we term this structure the Antennal Nerve Thermosensory Ganglion (ANTG; Figure 1b,f). The first cell type identified in this ganglion was a hot-activated neuron called the “Anterior Cell” (AC;²³) an unusual cell type that combines properties of a sensory neuron (i.e. the responses to temperature stimuli) with those of a central neuron (i.e. long axonal projections that target directly higher-brain centers). A second cell type that is activated by cold temperature was discovered more recently and called “Anterior Cold Cell” for symmetry (ACC;²¹)(Figure 1b,f). The projection pattern of these TRNs within the brain suggest that they may be involved in a set of partially overlapping behaviors (see below).

When it comes to the dynamic response properties of TRNs (adaptation, potential thresholds of activation etc.), beyond the simple characterization of the TRNs as hot- or cold-activated offered above, their properties have been rather difficult to describe in detail. Fly sensory neurons are tiny (the cell body is on the order of 2–6 μm) and often surrounded by hard cuticle, which makes them very difficult to reach for patch clampelectrophysiology (but see²⁴). Unlike chemosensory structures (for example), thermosensory sensilla are well shielded from the outside, making extracellular recordings more challenging (but see¹¹). Calcium imaging using genetically encoded indicators has been amply utilized to characterize the responses of Drosophila sensory neurons²⁵ (including those of TRNs^10,21), and yet calcium levels do not allow one to estimate the firing rate of a neuron at steady state (i.e. at a particular temperature). As a result of these limitations, it has been easier to evaluate the dynamic response properties of the more accessible second-order or even third-order neurons of the system than those of the TRNs (^{21,24,26–28}, and see below). With these caveats in mind, if we define cold/hot as anything below/above the species-specific favorite thermal range (as determined experimentally, ~25°C for D. melanogaster), current evidence suggests that most TRNs display a combination of phasic responses (bursts of activity in response to temperature change) and tonic responses (persistent activity in either the hot or cold range;^10,11,21); a profile rather similar to that reported for their mammalian counterparts²⁹. Notable exceptions include cold-activated neurons of chamber I of the sacculus, which do not produce any calcium transients unless the temperature dips below ~25°C (cold range;²¹), and the AC neurons, which only produce robust firing responses above ~27°C (hot range;²⁴).

An introduction to the Drosophila molecular temperature sensors

Any neurons’ firing activity is likely to be modulated by abrupt changes in temperature, but the TRNs have the unique ability to detect and respond robustly to even the most minute changes in temperature (as small as 0.1°C, see¹⁴). The exquisite temperature sensitivity and specific tuning of TRNs for hot or cold is determined by the selective expression of molecular sensors for temperature. In mammals, these molecules are members of the Transient Receptor Potential (TRP) family of ion channels which are directly gated by temperature². The first heat sensors identified in Drosophila were indeed also TRP channels: painless³⁰ (which responds to heat in the noxious range³¹ and that curiously can function both as a nociceptor^30,32 and as an anti-nociceptor³³) and Drosophila TRPA1 (dTRPA1^34,35). While much remains to be discovered in this area, all other molecular temperature sensors identified so far in Drosophila belong to different, invertebrate-specific ion channel lineages: arista hot cells express the “Gustatory” ion channel Gr28b.d³⁶, while cold cells depend on a specific complex of Ionotropic receptor (Ir) family members, consisting of Ir93a, Ir25a and Ir21a³⁷ (Figure 2).

Figure 2: Molecular receptors for temperature detection.

a) Organization of the Antennal nerve thermosensory ganglion, showing dTRPA1 expression in AC. b) Organization of the dTRPA1 genomic locus and relevant transcripts. c) TRPA1 tetramere structure (based on pdb_00007ykr). d) In vitro, dTRPA1 is activated by heat with a threshold of ~28°C (Gallio Lab, unpublished). e) The arista TRNs. f) Organization of the Gr28b.d genomic locus and relevant transcripts. g) Gr28b.d tetramere structure (predicted by Alphafold). h) In vitro, Gr28b.d is activated by heat with a threshold of ~26°C (see Capek et al., 2025). i) In a rapid 2-choice assay for temperature preference, flies avoid temperatures higher or lower than 25°C (BT=base temperature, 25°C, TT=test temperature). j) Silencing of ArHC or removing Gr28b (in a null mutant) abolishes avoidance of 25°C and reduces avoidance of 35°C. k) Silencing ACs has no effect on preference in this assay. l) In a dTRPA1, Gr28b double mutant heat avoidance is completely eliminated but cold avoidance remains normal (red median=different from controls, black box=not different from zero. see Capek et al., 2025 for controls, experimental and statistics details).

Panel d is unpublished data from Matthew Capek, Gallio Lab

Panels f and h are adapted from Capek et., al. 2025

Drosophila TRPA1 was first cloned based on homology with mammalian TRPA1 and shown to respond to heat (rather than cold, as did its mammalian counterpart) in vitro³⁵; a fly mutant was reported soon after to have phenotypes in temperature preference behavior³⁴. A version of this channel has even become a neurogenetic staple, allowing researchers to artificially activate neurons in response to temperature by transgenic mis-expression of the channel (a technique termed “thermogenetics”³⁸). Despite a promising start, the full spectrum of functions of dTRPA1 in vivo has proven harder to nail down, in part due to the fact that its locus comprises at least four splice variants (dTRPA1A-D) encoding proteins with different sequences, activation profiles, and expression patterns^39,40 (Figure 2b,c). It is clear that variants A and D are bona fide heat receptors and that dTRPA1 is expressed in AC neurons, where it is responsible for their activation by heat²³. In line with this, the in vitro threshold for activation of dTRPA1-A (~28°C, Figure 2d; and see³⁹) matches well the in vivo threshold of activation of ACs (~27°C;²⁴). Variants B and C instead do not show significant responses to temperature in vitro^39,40 (and when overexpressed in vivo), which is consistent with roles in the taste and olfactory systems (e.g.^41,42, and with the broad expression of dTRPA1 in the brain⁴³). More complicated are cases in which dTRPA1 mutants show phenotypes in temperature responsive behavior which can be nevertheless rescued by the expression of non-heat-sensitive variants^40,44,45. This is the case of noxious heat^40,44: flies display increasingly robust avoidance of hot temperatures, spanning the innocuous (i.e. not harmful, 25–35°C) and noxious range (i.e. potentially harmful or even lethal, >35°C). Adult D. melanogaster dTRPA1 mutants have striking defects in noxious heat avoidance, in that their specific avoidance of 40°C is reduced from almost complete to levels comparable to the avoidance of mild, innocuous heat (see Figure 5 of⁴⁴). Remarkably, this phenotype can be rescued by expression of dTRPA1-C, a Drosophila variant that does not respond to heat, and even by expression in Drosophila of a flatworm TRPA1 (insensitive to temperature in vitro) or of a human TRPA1 homolog (cold-activated;⁴⁴). This puzzle was resolved when it was discovered that noxious heat stimuli in this range rapidly lead to the production of reactive oxygen species (ROS), well known TRPA1 ligands –so that ROS function as chemical mediators of dTRPA1 activation by noxious heat⁴⁴.

When it comes to milder temperatures, in the adult fly the main molecular sensor for innocuous heat is the ion channel Gr28b.d³⁶, which is expressed in the hot TRNs of the arista⁴⁶ and is responsible for their activation by heat (^14,36; Figure 2e–h). This protein belongs to a large family of ion channels that is otherwise populated by gustatory receptors^47,48 (hence the prefix Gr). The Gr28b locus encodes five variants (a-e) based on the alternative usage of specific promoters which in turn determine the inclusion of an alternative first exon (Figure 2f,g). In D. melanogaster, only the d variant (Gr28b.d) encodes a molecular heat sensor⁴⁹. The other four isoforms (Gr28b.a,b,c, and e) code instead for putative chemoreceptors^50–52. In vitro, the D. melanogaster Gr28b.d is activated by heat with a distinct threshold of ~26°C (⁵³; Figure 2h). Excitingly, sequence differences in the Gr28b.d of different Drosophila species determine differences in the threshold of heat activation of the channel, which in turn matches remarkably well the onset of behavioral heat responses recorded in the laboratory (heat avoidance or attraction;⁵³). As these thresholds also correlate with thermal conditions in the native habitat of each species, Gr28b.d and the neurons that express it may have played an important role in the evolution of species-specific thermal preference in Drosophila⁵³. In contrast, Gr28b.d-like sequences do not appear to have a conserved thermosensory function in mosquitoes⁴, nor in other insects outside of the Diptera, suggesting the intriguing idea that a thermosensitive Gr28b may be a quite recent evolutionary innovation within the fly lineage that contains Drosophila.

In contrast to the heat sensors, the molecular receptors that mediate the responses to cold temperature in Drosophila have been much harder to express (and therefore validate) in vitro. A number of ion channels have been proposed as candidate molecular cold receptors over the years based on behavioral phenotypes in cold avoidance^10,54. The current best candidate cold receptor is a specific complex of Ionotropic receptor (Ir) family members, consisting of the broadly expressed co-receptors Ir93a and Ir25a and the more narrowly expressed Ir21a, which is considered as the subunit that confers cold sensitivity³⁷. The picture is unfortunately complicated by the fact that these Irs also have a developmental function so that Ir21a and Ir25a/93a mutants have defects in cilia morphology¹¹. The best evidence that Ir21a may directly mediate cold responses (when co-expressed with Ir25a and Ir93a) comes from in vivo mis-expression in hot cells of the arista, where it is sufficient to produce partial cold responses in otherwise heat-activated TRNs (and partial transformation of cilia morphology;¹¹).

The coding of temperature at the periphery

As detailed above, work in the past years has led to a catalog of peripheral temperature sensing neurons in adult Drosophila with different tuning properties and anatomical locations. One initial observation is that, like in mammals, Drosophila thermosensory neurons seem to specialize in responding to either heat/heating or cold/cooling. Moreover, while we have found “cold”²¹ and “hot”²⁴ thermometers (neurons that encode absolute temperature in the hot or cold range as spike frequency), we have not found neurons that function as true thermometers across the range of temperature that spans the preferred 25°C. These observations may be explained by a simple argument: rapid changes in stimulus space are best encoded by neurons that are silent at steady state and that increase their firing rapidly when a relevant change in stimulus space occurs⁵⁵. Moreover, in this configuration, the fly thermosensory system is expected to remain quiet when thermal conditions are permissive, which is energetically advantageous. An important question is to which extent cold and hot temperature can indeed be processed independently. Examples of largely independent processing can be found in the regulation of sleep-wake cycles by temperature, where silencing the relevant hot or cold-sensing circuits has different and non-overlapping effects on behavior^21,24. An example of this can also be found in rapid thermal preference: genetic silencing of arista hot cells or removal of Gr28b.d (in a null mutant) abolishes the avoidance of mild innocuous heat (30°C) and reduces the avoidance of warmer, still largely innocuous temperature (35°C; Figure 2i–j), but does not affect the avoidance of noxious heat (40°C;¹⁴) which instead depends on dTRPA1⁴⁴ (but, interestingly, not on the AC neurons, which instead help regulate sleep-wake cycles in hot conditions –see below; Figure 2k and see¹⁴). Consistent with this, removal of both Gr28b.d and dTRPA1 (in a double mutant) completely abolishes heat avoidance in the full heat range tested (30°C-40°C). Importantly, none of the manipulations affecting heat avoidance significantly impacted the avoidance of cold temperature (Figure 2i–l see⁴⁴ for details; and see^10,56 for similar results from the silencing of cold circuits). It has been much debated if peripheral sensory systems are organized as “labelled lines”, where sensory neurons with different tuning act as independent conduits for information, or if any meaningful responses necessarily involve “across fiber” interactions between different sensors. The argument has often been muddled by stretching the idea of “labelled lines” past the receptor layer and to include central sensory pathways, where of course cross-modal interactions are the rule. The data from the Drosophila thermosensory system highlighted above seems to suggest that, at least for some behavioral outcomes, hot and cold information from the TRNs can be processed rather independently, which is sensible as it may allow additional evolutionary flexibility in shaping the responses to different habitat conditions.

Temperature representation in the brain: the second order thermosensory projection neurons or TPNs

Next in line from the TRNs of the antenna, temperature information is conveyed to higher brain centers by second-order neurons –the thermosensory projection neurons or TPNs. With the exception of the ACs which, as mentioned, directly target higher brain centers, all antennal TRNs connect to the cognate TPNs in a region of the brain termed the Posterior Antennal Lobe (PAL⁵⁶; Figure 3). Here, the axon termini of hot and cold TRNs are segregated into different “hot” and “cold” synaptic domains (the “hot” and “cold” glomeruli) which are adjacent, but in turn distinct from glomeruli formed by the hygrosensory neurons responding to dry and humid air (the “arm” and “bean” glomeruli, respectively –so named based on their characteristic shape,^17,18; Figure 3a). As a result of this organization, changes in external temperature and/or humidity are represented by a simple spatial map of neural activity within the PAL^18,56, a “sensory map” that resembles those found in the mammalian brain for the internal representation of more complex stimuli. The TPNs receive TRN synaptic input within the PAL, and their role is that of deciphering this map of activity and to relay information about external temperature to higher brain centers for further processing. The PAL is also innervated by an abundance of local interneurons which also participate in sensory processing²⁷. The major TPN pathways were first described using neurogenetic methods such as focal photo-activation of GFP at the hot or cold glomerulus (⁵⁶; Figure 3b,c), as well as by anatomical and functional studies using selective drivers^27,56, and have been more recently reconstructed in the fly brain EM connectome^7,57.

Figure 3: Second-order projection neurons, anatomy, connectivity, and functional properties.

a) Schematic of cell types and structures (top), and detail of the PAL glomeruli annotated by functional properties (first cirle), using anatomical nomenclature (second circle), and based on color-coding of axonal projections on a 3D reconstruction based on the EM hemibrain connectome (third circle). b,c) Focal labeling using photo-activated GFP in hot or cold glomeruli highlights the major TPN pathways. d) EM reconstruction of TPNs based on the hemibrain connectome showing TPNs that innervate the hot glomerulus (left), the cold glomerulus (right), or both (bottom). The full ensemble is also shown. e) TPNs are also distinguished by the dynamics of their responses as fast adapting (a patch-clamp trace from TPN-III is shown at the top) or slow/non-adapting (a trace from TPN-II is shown at the bottom). f) TPN nomenclature and properties. g) TPN synaptic output by brain region, based on the hemibrain connectome and MB targets within the lACa (green=KC-gamma s2, grey=KC-alpha’Beta’). h) Detail of the lACa; h’) is a two-photon stack of the Calyx of a TPN-II:RFP expressing fly. Abbreviations: TRN=thermosensory receptor neuron, TPN=thermosensory projection neuron, LH=lateral horn, MB=mushroom body, PAL=posterior antennal lobe, ArCC/HC=arista hot cell/cold cell, ACC=anterior cold cell, sCC=sacculus cold cell, SMP=superior medial protocerebrum, SLP superior lateral protecerebrum, Wed=wedge, lACa=lateral accessory calyx; Scalebars in b,c: 30μm, in h’:10μm.

Panels b and c are reproduced from Frank et al., 2015

Panel h’ is adapted from Alpert et al., 2020

A first conclusion that can be drawn from the data is that there is a whole ensemble of TPNs receiving direct synaptic input from TRNs in the PAL (at least ~22 cells belonging to 7 major cell types, Figure 3d–f). TPNs cell types can be categorized based on a combination of anatomical, physiological and functional characteristics. TPNs can be categorized at the anatomical level based on their pattern of innervation of the PAL glomeruli (e.g. innervating either hot or cold PAL glomeruli); in many cases we also know a TPN’s own functional tuning (as in hot- or cold-activated, e.g., from recordings) and this is generally consistent with their innervation pattern. For example, “narrowly tuned” cold-activated (e.g., TPN-II, TPN-VI) and hot-activated (e.g., TPN-IV, TPN-V) cell types have been described which innervate exclusively either the cold or hot glomerulus (respectively). There are also TPN cell types in which an individual neuron receives input in both the hot and cold glomeruli and responds to both heating and cooling (“broadly tuned” e.g., TPN-III^26,27,56). In addition, in at least one case, a TPN cell type includes distinct cells with opposite tuning^27,56. Figure 3d–f). TPN cell types have also been found that receive mixed input from thermosensory and hygrosensory receptor neurons, directly acting as substrates for multisensory integration¹⁸.

TPNs have been easier to access than TRNs for electrophysiological recordings so that their dynamic responses to temperature can be further used to subdivide them in fast-adapting versus slow- or non-adapting cells, based on the persistence of their response to a defined temperature step (e.g., 25°C to 20°C; Figure 3e)^21,26,27,56. Moreover, at least in some cases, a TPN’s firing rate was found to scale with the rate of temperature change, suggesting the existence of specific “rate of change” detectors^26,27. Finally, as will be detailed in the following paragraph, TPNs can be characterized by their functional roles, either indirectly, based on the known functions of the higher brain regions they target, or directly, through the use of selective drivers for genetic silencing in vivo. Indeed, the diversity of anatomical, electrophysiological, and tuning properties of the TPNs described above is beginning to make sense in light of recent studies that have elucidated their specific involvement in a variety of temperature-responsive behaviors, often occurring on distinct temporal scales (see below).

Third order pathways and beyond

The availability of EM-based connectomes of the fly brain has revolutionized circuit mapping in this species, enabling one to generate comprehensive catalogs of the potential downstream targets of a neuronal cell type of interest (at least through chemical synapses) in hours or days on a computer, rather than over weeks or months in the laboratory. When it comes to the TPN higher brain targets, a broad theme is that, by far, the main target of TPNs in the higher brain is the Lateral Horn (LH), a region long implicated in the processing of innate behavior (41% of the total TPN ensemble synaptic output;^56,57; Figure 3g). The Mushroom Body (MB), the main associative center of the fly brain, also receives significant TPN input (15% of total TPN output) but curiously this is concentrated on the lateral Accessory Calyx (lACa) and onto a few Kenyon cells⁵⁷ that receive strong convergent input from two excitatory, heat-activated TPN cell types (TPN-IV and -V) and one cold-activated, inhibitory cell type (TPN-II; Figure 3g,h). The function of this specialized MB circuit is not known.

As alluded above, recent work has started to delineate entire thermosensory sub-circuits comprising sensory neurons, cognate TPNs, third- and in some cases higher-order neurons based on their involvement in specific temperature-responsive behavior. For example, two distinct and asymmetrical thermosensory circuits have been found to directly target neurons of the clock network that participate in the regulation of sleep^21,24. Cold-activated TPN-II receives input from a variety of TRNs within the cold glomerulus and targets the clock neuron DN1a (Dorsal Neuron 1a) within the lACa²¹ (Figure 4a,b). TPN-II has been described as a “cold thermometer” as its steady-state firing rate scales with absolute temperature, but only when the temperature dips below the D. melanogaster species-specific preferred ~25°C (a neuronal correlate for what it must mean for a fly to be cold). Interestingly, TPN-II expresses the inhibitory transmitter GABA, mediating powerful cold inhibition of DN1a²¹. In contrast, temperatures above ~25°C persistently activate AC neurons, which function as a “hot thermometer” and in turn directly target a second clock network component known as the LPNs (for Lateral Posterior Neurons), this time for activation²⁴ (and note that LPNs are also independently targeted by TPN-IV and -V²⁴, Figure 4b,c). As alluded to, both these hot and cold “thermometer” circuits and their targets ultimately converge on the regulation of daytime sleep by temperature (^21,24 and see recent work which expanded on these circuits^58–60). Flies are characteristically crepuscular and, in standard laboratory conditions, display a prominent “siesta” sleep in the middle of the day. Cold temperature-mediated inhibition of DN1a increases morning sleep and advances the peak of siesta sleep to earlier in the day (indeed cold conditions often accompany the seasonal shortening of days;²¹). In contrast, hot temperature-mediated LPN activation expands siesta sleep, a phenomenon that has been suggested to help keep files away from high noon heat (²⁴ and see Figure 1). This circuit also impacts night-time sleep regulation⁵⁸.

Figure 4: Third-order neurons, properties and functions.

a) Reconstruction of a TPN circuit that carries information about persistent cold to the clock neuron DN1a in the lACa. b) Summary of direct hot and cold TPN input on clock neurons and on Kenyon cells in the lACa, and effect on daytime sleep; representative electrophysiological traces are also shown (reproduced from Alpert et al, 2020 and Alpert et al, 2022). c) Reconstruction of a circuit comprised of AC and TPN-IV that carries information about heat to the LPN clock neurons. d) Reconstruction of a TPN circuit comprising TPN-III and TPN-V targeting LHPV2a and TLHON-I neurons of the ventral lateral horn (top) and electrophysiological responses of TPN-III showing adapting responses to heating and cooling (“ON” responses). e) Summary of electrophysiological responses of TPN-III and TLHON-I to heating and cooling stimuli of different rate, showing that both cells’ responses are proportional to the rate of temperature change, but TLHON-I only responds to rapid change. f) Schematic representing phenotypic consequences of TLHON-I silencing. In a two-choice assay, at the boundary between hot and cold quadrants, flies encounter very steep thermal gradients and frequently produce “escape turns” see Jouandet et al, 2023 for details). TLHON-I silenced flies more frequently ignore the boundary and cross over into the hot quadrant or turn later within the boundary (traces and speeds are real data, gradient is from hi-resolution thermal simulation (see Jouandet et al, 2023). Reconstructions are based on the EM hemibrain connectome (Scheffer et al 2020).

Panel b is adapted with permission from Alpert et al, 2022

Panels d-f are adapted from Jouandet et al, 2023

While “thermometer” circuits characterized by persistent firing have a prominent role in regulating behavior occurring on longer temporal scales (minutes to hours, to days and potentially seasons –such as the timing and duration of sleep and wake cycles, see above), rapid responses such as navigation, escape, and preference seem to involve a number of fast-adapting pathways that instead operate on the timescales of hundreds of milliseconds to seconds. Case in point is the circuit defined by TPN-III, a cell type with unusual properties^26,56. TPN-III innervates both the hot and cold glomerulus of the PAL and fires a burst of action potentials in response to any change in temperature, irrespective of the direction of change (i.e. either cooling or heating) or of the absolute temperature (e.g. cooling from a hot or cold baseline)²⁶. Rather, TPN-III’s firing scales with the rate of temperature change: the faster the change in temperature the more spikes a TPN-III burst contains. Connectomic analysis shows that TPN-III innervates target neurons in the ventral LH, including an output neuron of the circuit named TLHON-I (for thermosensory lateral horn output neuron -I; Figure 4d). Indeed, TLHON-I’s responses also depend to the rate of thermal change, but their rate responses appear high-pass filtered: TLHON-I only fires in response to fast (>0.2°C/second), but not slow, thermal change²⁶. Genetic silencing of this circuit suggest that such “fast change” responses represent an alarm system which alerts the fly that a rapid temperature change is going to bring about dangerous thermal conditions, priming locomotor behavior for escape²⁶ (i.e., a substrate for the proverbial frog’s successful escape from a rapidly heating pot).

Interestingly, the ventral LH is densely innervated by TPNs axons, including by projections that differ between species in a way that correlates with the nature of their behavioral responses to temperature. As discussed above, D. melanogaster’s temperature preference behavior emerges as a result of hot and cold avoidance. Perhaps as a result of extreme conditions in its native habitat, the desert endemic Drosophila mojavensis seems instead to have evolved active attraction to its favorite thermal range of ~30–35°C, so that Gr28b.d-mediated activation of hot arista TRNs in this fly species signals attraction, rather than avoidance⁵³. TPN-V is the main (albeit not only) synaptic output of hot-activated TRNs of the arista and, excitingly, targeted electroporation of TPN-V demonstrated that its pattern of projections within the ventral LH in D. mojavensis is unique and different as compared to that of heat-avoiding species⁵³. This result suggests that the LH valence circuit may have re-programmed in this species to accompany the adaptation to the extreme desert environment. Together with the fact that shift in the activation threshold of Gr28b.d can shift behavioral thresholds of heat response (see above and⁵³), the data demonstrate that, in Drosophila, changes in species-specific thermal preference can be accomplished with mechanisms ranging from shifts in the activation threshold of molecular temperature sensors to a re-wiring of the brain pathways that process temperature valence⁵³.

CONCLUSION

A tractable model system such as the fruit fly affords the possibility of studying a neurobiological process at multiple levels of organization and from multiple angles, in most cases using relatively affordable reagents, and with the added benefit of an ever-growing experimental toolkit which is freely shared within the community. The advantages are evident: with comparably little investment and with the dedicated work of only a dozen groups around the world, rapid progress can paint a remarkably complete picture, bringing together discoveries that range from the molecular and structural level, to the level of cells and circuits, to that of systems and behavior, and all the way to the scale of evolutionary processes. The goal of this review was to highlight the recent progress in the study of Drosophila thermosensation as a good example of this. Since the pioneering work of Seymour Benzer nearly 30 years ago^13,30, discoveries from a number of groups around the world has contributed a remarkably nuanced and layered portrait of the fly thermosensory system. A portrait that offers a framework to understand how sensory stimuli are processed in the brain to produce innate, species-specific behavioral responses and one that advances our understanding of how animals interact with the thermal environment.

This review is by no means complete. For consistency, the work covered focused in large part on the adult fly and ignored the larva, a fantastic model for the study of the senses⁶¹ but also anatomically and functionally almost a distinct animal species. Likewise, the main focus was kept on research on innocuous, rather than noxious, temperature responses –a very active area of research in Drosophila, especially in the larva⁶². Systems-level work that incorporates computational models for the study of thermal navigation and thermotaxis was not covered in detail (e.g.,^14,63–65), but this too is an exciting area of research where a combination of modeling and connectomics is expected to truly test our understanding of the functional organization of the thermosensory system. Finally, this review did not touch on a body of literature on the thermal ecology and natural behavior of Drosophila species, an important inspiration for any work on thermosensation and one that helps to contextualize the relevance of any discovery in this area, especially given the immediate impact of climate change on habitats⁶⁶.

The advances outlined above and the reagents that have just recently become available for work in Drosophila (i.e. the EM connectomes) are poised to propel further exploration in this area and further integration and synergy with other related fields of research. If the funding “climate” for research on model systems remains favorable, we can expect many new exciting discoveries and likely a few surprises. This is especially important now, when changes in the actual climate (i.e. a warming planet) are demanding we pay attention to the effects of temperature on our own behavior as well as on that of insects.

TERMS AND DEFINITION LIST:

Fly brain connectome

A wiring diagram of the fruit fly central nervous system reconstructed from an electron microscopy volume of the fly brain or ventral nerve cord, detailing all the neurons and the synaptic connections between them.

Sensillum

A common insect sensory organ, consisting of a cuticular structure (a hair or a peg) innervated by one or more sensory neurons. Insect antennae, legs and even wings and other body parts are often covered by sensilla, permitting the detection of chemical and physical stimuli such as odor cues or mechanical deflection.

TRN

thermosensory receptor neuron. A sensory neuron specialized in the detection of thermal stimuli.

TPN

thermosensory projection neuron. A central neuron that receives direct synaptic input from thermosensory receptor neurons and that extends axons targeting higher brain centers.

ANTG

antennal nerve thermosensory ganglion. A small group of thermosensory receptor neurons located at the based of the antennal nerve.

PAL

posterior antennal lobe. The posterior aspect of the antennal lobe, containing specialized clusters of synapses called glomeruli. The main antennal lobe (the anterior part) contains ~50 distinct glomeruli each innervated by olfactory receptor neurons with distinct profiles of odor responses, while the posterior antennal lobe contains only a few glomeruli, most of which are targeted by thermosensory and hygrosensory neurons.

MB

mushroom body. The associative center of the insect brain, so named because of its fungal shape.

LH

lateral horn. Like the mushroom body, the lateral horn is a crucial target for sensory pathways and a center for the processing of innate behavioral responses.

TLHON

thermosensory lateral horn output neuron.