ABSTRACT
Animals commonly use thermosensation, the detection of temperature and its variation, for defensive purposes: to maintain appropriate body temperature and to avoid tissue damage. However, some animals also use thermosensation to go on the offensive: to hunt for food. The emergence of heat-dependent foraging behavior has been accompanied by the evolution of diverse thermosensory organs of often exquisite thermosensitivity. These organs detect the heat energy emitted from food sources that range from nearby humans to trees burning in a forest kilometers away. Here, we examine the biophysical considerations, anatomical specializations and molecular mechanisms that underlie heat-driven foraging. We focus on three groups of animals that each meet the challenge of detecting heat from potential food sources in different ways: (1) disease-spreading vector mosquitoes, which seek blood meals from warm-bodied hosts at close range, using warming-inhibited thermosensory neurons responsive to conductive and convective heat flow; (2) snakes (vipers, pythons and boas), which seek warm-blooded prey from ten or more centimeters away, using warmth-activated thermosensory neurons housed in an organ specialized to harvest infrared radiation; and (3) fire beetles, which maximize their offspring's feeding opportunities by seeking forest fires from kilometers away, using mechanosensory neurons housed in an organ specialized to convert infrared radiation into mechanosensory stimuli. These examples highlight the diverse ways in which animals exploit the heat emanating from potential food sources, whether this heat reflects ongoing metabolic activity or a recent lightning strike, to secure a nutritious meal for themselves or for their offspring.
Introduction
On a moonless night, a mosquito buzzes near a sleeping camper, searching for the warmth of exposed skin. Outside the tent, a hungry rattlesnake detects the presence of a mouse moving quietly on the cool ground a short distance away. Tens of kilometers away, a forest fire rages – most here are oblivious, but several beetles take flight, racing to be the first to lay their eggs on the freshly burned wood. Although driven by sources near and far, from humans to burning trees, all these behaviors share a common reliance on thermosensation, the detection of temperature and temperature variation. Although animals commonly rely on their thermosensory systems to maintain optimal body temperatures and avoid tissue damage, some animals also use these systems to forage. In this Review, we focus on some of the diverse specializations that animals have evolved to locate nutrition based on the heat that its sources emit.
Physical properties of temperature, heat transfer and thermosensation
In discussion of thermosensory-driven foraging, it is important to consider the nature of temperature and heat (see Glossary). Temperature [often reported in kelvin (K) or degrees centigrade (°C)] measures molecular motion. It is proportional to the average kinetic energy of the molecules in a system: higher temperatures reflect higher energies. In contrast, heat [often reported in joules (J)] refers to the quantity of thermal energy transferred from regions of higher to lower temperature. In biology, temperature affects all physiological processes. Raising the temperature increases chemical reaction rates by increasing the frequency and energy of molecular collisions. It also alters macromolecular structure and function by increasing internal molecular motion. Finally, it changes the energetic driving forces of biological processes, and hence equilibrium constants. Together, these effects explain the widespread impact of temperature on animal physiology (Huey and Kingsolver, 2011). They also underlie the mechanisms by which animals detect heat emission. Throughout this Review, the term ‘heat’ is used when the relevant parameter is heat energy being released or absorbed, and the term ‘temperature’ is used when the relevant parameter is the average kinetic energy of a given piece of matter. For example, ‘heat seeking’ refers to an animal's attraction to a target based on the heat energy the target emits, whereas ‘temperature sensing’ refers to an animal's ability to detect the temperature of its own tissue.
As detailed below, animals from mosquitoes to snakes to beetles locate potential food sources based on heat transferred to them from their targets (Fig. 1). Heat transfer can occur in three ways: conduction, convection and radiation. Conduction is energy exchange through direct interaction of molecules without bulk motion of matter, such as when feet contact hot ground. Convection is heat transfer by movement of a fluid, such as when warm air currents signal the presence of a warm object nearby. Finally, all objects with temperatures above absolute zero emit thermal radiation as electromagnetic waves. At animal body temperatures, thermal radiation is maximal in the infrared (emission from 35°C human skin, for example, peaks at 9.4 μm), providing a far-traveling signal that some specialized animals have harnessed to sense targets at a distance.
Heat transfer can be detected by thermosensory neurons in many ways, both directly and indirectly. At the molecular level, temperature changes induced by conduction or convection can initiate temperature-sensitive signaling events; for example, by activating highly thermosensitive ion channels (Xiao and Xu, 2021). For heat energy transferred by infrared radiation, one could imagine a detection system like that for visible light, with photon absorption by a visual pigment initiating a phototransduction signaling cascade (Arshavsky et al., 2002). To date, however, no visual pigments have been discovered with peak-absorption wavelengths in the infrared range; this is potentially because sensitivity to longer-wavelength (less-energetic) photons comes at the cost of greater noise (Luo et al., 2011). Instead, infrared detection is thought to be mediated by tissue absorption of infrared photons, leading to local temperature increases. These increases either directly activate temperature-sensitive neurons or indirectly activate mechanosensory neurons, the latter responding to forces exerted by temperature-induced expansion or structural changes in the surrounding tissue (Gracheva et al., 2010; Moiseenkova et al., 2003; Schmitz et al., 2007).
Bolometer
A temperature sensor that operates through temperature-dependent changes in electrical resistance.
Golay cell
A temperature sensor that operates by monitoring the temperature-dependent expansion of a structure or chamber.
Heat
Thermal energy transferred between systems or between a system and its surroundings.
Poikilotherm
An animal whose body temperature tracks ambient temperature.
Schwann cell
A glial cell that supports neuronal function.
Spatiotopic map
The spatially ordered representation of external spatial coordinates within the brain.
Temperature
The average kinetic energy of molecules in a system. Animals sense heat emission by absorbing heat energy and responding to the resulting changes in tissue temperature.
Topographic map
A set of spatially ordered neuronal projections in which neurons representing adjacent sensory surfaces project to adjacent locations in the brain.
Benefits and challenges of using heat emission as a feeding cue
The utility of heat emission as a feeding cue is evident from that fact that many independent lineages have evolved such behaviors. Among insects alone, hematophagy (blood feeding) is believed to have arisen independently ≥12 times, and temperature is an important host-seeking cue in most, if not all, of these lineages (Grimaldi and Engel, 2005; Lazzari, 2019; Wiegmann et al., 2011). Such convergence could, in part, reflect the fact that thermosensation is an ancient sensory modality. Rather than requiring de novo development of novel sensors, existing temperature-detection mechanisms can be repurposed to support heat seeking (Greppi et al., 2020). An additional benefit of using heat emission as a feeding cue is that, at least for animals feeding on homeotherms, heat can be used to locate many different types of potential hosts, in contrast to other cues, such as odors, which may be species specific. Finally, local heat emission correlates with vascularization (Bagavathiappan et al., 2009), so it could also denote particularly nutritious areas of the host body on which to feed (Ferreira et al., 2007).
Using heat emission as a feeding cue also presents challenges. Temperature gradients generated by conduction and convection dissipate rapidly, e.g. air temperature around a human body usually drops to ambient within <10 cm (Craven and Settles, 2006; Lewis et al., 1969). Thus, these cues are mostly useful at short range. In addition, because temperature is ubiquitous, organisms must have the means to filter background thermal noise and limit their attraction to inappropriate targets (Corfas and Vosshall, 2015). For poikilotherms (see Glossary), such as insects, a large temperature differential with the host makes host detection easier, but can result in dramatic body temperature changes upon ingestion of host blood. Mosquito body temperature can rise from 22°C to 32°C within a minute of drinking blood. If not properly accounted for, this rapid body temperature change could be detrimental to the mosquito. In addition to gene expression changes, which act over longer time scales (Benoit et al., 2011), mosquitoes may protect themselves by lowering their body temperature by evaporative cooling using fluid excreted during blood feeding (Lahondere and Lazzari, 2012).
Direct versus indirect effects of temperature on feeding behavior
Because temperature affects all aspects of physiology, it can be challenging to differentiate between heat seeking as a means to orient towards a food source versus temperature-induced foraging secondary to thermoregulation. For instance, honeybees prefer feeding on warmed nectar at cool temperatures, but this appears to reflect a thermoregulatory strategy rather than attraction to heat emitted by the target (Dyer et al., 2006; Nicolson et al., 2013; Norgate et al., 2010). In this Review, we focus instead on how animals exploit heat emission to locate a source of food. We highlight three animal groups that each use distinct mechanisms to detect heat emitted by targets at different distances: mosquitoes, snakes and pyrophilous (‘fire-loving’) beetles. By comparing and contrasting the thermosensory systems of groups that have independently evolved the capacity for specialized temperature-driven foraging, we can deepen our understanding of the principles and mechanisms underlying thermosensation more generally.
Convective heat detection as a short-range cue for mosquito blood feeding
Heat-seeking behavior drives mosquito blood feeding from warm-bodied hosts
In many mosquito species, females must ingest blood from an appropriate host to nourish their developing eggs (Clements, 2012a). These species include major vectors of human disease, such as the malaria vector Anopheles gambiae and the dengue vector Aedes aegypti, and it is during blood feeding that mosquitoes contract and transmit disease-relevant pathogens (Clements, 2012b). Because blood-feeding targets are often warm bodied, heat emission is useful for locating hosts. The ability of temperature to drive mosquito host seeking was first reported in 1910, when Frank Howlett, a British entomologist, observed: ‘On more than one occasion when mosquitoes have been troublesome at tea-time I have noticed that they seemed to be fond of hovering in the neighbourhood of the tea-pot, being attracted apparently by the heat’ (Howlett, 1910). He noted that female mosquitoes were not only attracted to heat, they also stabbed at heat sources with their proboscises, ‘displaying the utmost eagerness in their fruitless efforts’ to feed from inanimate warmed objects.
Subsequent work has confirmed that heat emission is a powerful mosquito attractant in the laboratory and in the field (Brown, 1951; Corfas and Vosshall, 2015; Marchand, 1918; Peterson and Brown, 1951; Zermoglio et al., 2017). It is now appreciated that mosquitoes exploit additional host cues, including carbon dioxide (CO2), visual stimuli, humidity and host-specific odors (Carde, 2015; van Breugel et al., 2015). This multiplicity of cues permits host location and, in some cases, blood feeding even when thermal cues are absent or thermosensation is compromised (Greppi et al., 2020; Laursen et al., 2023; McMeniman et al., 2014; Reinhold et al., 2022). Nonetheless, heat emission strongly promotes attraction and blood feeding, and robustly synergizes with other host-associated cues in driving these behaviors (Greppi et al., 2020; Laursen et al., 2023; Liu and Vosshall, 2019; McMeniman et al., 2014; Raji et al., 2019). For Ae. aegypti mosquitoes, attraction of CO2-exposed mosquitoes to warm objects is detected as soon as target temperatures rise above ambient, peaking near ∼40°C, and declining above ∼45°C, likely due to the engagement of high-temperature avoidance pathways (Corfas and Vosshall, 2015). Heat-driven attraction also diminishes as ambient temperature rises, likely as a result of the decreased thermal contrast between target and environment (Corfas and Vosshall, 2015).
Heat-driven attraction of mosquitoes could, in principle, be mediated through a combination of conduction, convection and/or radiation. To date, studies in Ae. aegypti suggest that conduction and convection are the primary mediators, at least at close range (Peterson and Brown, 1951; Zermoglio et al., 2017). For example, an infrared-transparent material that blocks convection, but not infrared radiation, eliminates mosquitoes' preference for a 34°C versus 25°C source (Zermoglio et al., 2017). Furthermore, all faces of a 33°C ‘Leslie cube’ (whose faces all transfer similar heat by conduction and convection, but which differ by up to ∼7 fold in infrared radiation emitted) are similarly attractive to mosquitoes (Peterson and Brown, 1951). However, because all heated objects emit infrared radiation, definitively excluding a role for infrared is challenging – the detected photons could ultimately come from the mosquito's own tissue warmed by heat from the source. Irrespective of the contribution of radiation, it is clear that temperature gradients and convection currents robustly attract female mosquitoes (Carde, 2015; Eiras and Jepson, 1994; Howlett, 1910; Lehane, 2005; van Breugel et al., 2015; Zermoglio et al., 2017).
Mosquito antennae contain highly thermosensitive warming- and cooling-activated neurons
Ablation experiments demonstrated that the antennae are required for heat seeking in mosquitoes (Ismail, 1962; Roth, 1951). Further investigation localized the thermosensors to coeloconic (peg-in-pit) sensilla at the antenna's distal tip, as well as several more proximal segments (Davis and Sokolove, 1975; Gingl et al., 2005; Greppi et al., 2020; McIver, 1973). In these sensilla, the pit wall extends above the peg and only leaves a small opening at the top (Gingl et al., 2005; McIver, 1973; Fig. 1). Three bipolar neurons innervate each sensillum, two of which extend unbranched dendrites into the lumen of the sensilla, while a third neuron terminates with a highly lamellated dendrite beneath the peg base (McIver, 1982).
Physiological studies have confirmed the presence of thermosensitive neurons in the coeloconic sensilla at the distal tip of the antenna in Ae. aegypti (Davis and Sokolove, 1975; Gingl et al., 2005; Laursen et al., 2023) and An. gambiae (Greppi et al., 2020; Wang et al., 2009). These neurons appear to act as an opponent pair of warming- and cooling-activated thermosensors (Davis and Sokolove, 1975; Gingl et al., 2005). The warming-sensitive neuron (warming cell) transiently increases its spike rate as temperature increases and transiently decreases its rate of spiking as temperature decreases. The cooling-sensitive neuron (cooling cell) exhibits the opposite responses. In Ae. aegypti, both neurons are highly thermosensitive, responding to step changes of ∼0.05°C (Davis and Sokolove, 1975). In contrast, they appear to be relatively insensitive to infrared radiation, requiring steps of ≥40 mW cm−2 of infrared light to elicit responses (Davis and Sokolove, 1975; Gingl et al., 2005). Note that infrared emission from a human hand is estimated to be only ∼4.1 mW cm−2 at 10 cm (Zopf et al., 2014). The location of these neurons in pegs recessed within thick-walled pits has been suggested to limit their direct exposure to radiation (Gingl et al., 2005).
Mosquito heat seeking involves cooling-activated receptors
Work on mosquito olfaction indicates that rather than evolving new host-focused chemosensory systems de novo, mosquitoes and other hematophagous arthropods have largely co-opted existing chemosensory systems and leveraged them for host seeking (DeGennaro et al., 2013; McMeniman et al., 2014; Raji et al., 2019). Initial studies indicate this is also the case for thermosensing (Corfas and Vosshall, 2015; Greppi et al., 2020; Laursen et al., 2023). Current work on the molecular basis of heat seeking was inspired by discoveries initially made in the vinegar fly Drosophila melanogaster. The Drosophila antenna contains opponent cooling- and warming-sensitive thermoreceptors, reminiscent of thermoreceptors found in the coeloconic sensilla of mosquitoes (Budelli et al., 2019; Davis and Sokolove, 1975; Gallio et al., 2011; Ni et al., 2013). Interestingly, Drosophila cooling cells drive not only avoidance of cool temperatures but also avoidance of innocuous warm temperatures (Budelli et al., 2019). Such bidirectional control suggested an alternative mechanism by which heat seeking could be controlled: not by a warming-activated cell, but rather by a warming-inhibited cell. Although the behavioral outcome would be the same in both cases (movement towards a warm object), in the latter scenario, heat seeking would be more akin to avoidance of the cooler surroundings.
At the molecular level, thermosensitivity of the Drosophila cooling cells is mediated by a combination of ionotropic receptors, an invertebrate-specific family of channels related to ionotropic glutamate receptors (iGluRs). The structure and function of ionotropic receptors is not well understood, but they are proposed to form heteromeric channels composed of one or more broadly expressed co-receptor ionotropic receptors (IR25a and IR93a in the cooling cell) along with stimulus-specific ionotropic receptor subunit(s) (IR21a for cooling; Benton et al., 2009; Budelli et al., 2019; Ni et al., 2016). These ionotropic receptors also mediate cooling cell function in mosquitoes, with disruption of either Ir21a or Ir93a eliminating cooling cell thermosensitivity in An. gambiae (Greppi et al., 2020; Laursen et al., 2023; Fig. 2A). Similar effects were observed when Ir93a was eliminated in Ae. aegypti (Laursen et al., 2023). Consistent with a central role for cooling cell activity, both Ir21a mutant and Ir93a mutant animals show strong deficits in CO2-stimulated heat seeking (Greppi et al., 2020; Laursen et al., 2023). This suggests that mosquitoes have repurposed an ancestral cooling-activated receptor to drive an opposing behavioral response to that seen in non-hematophagous dipterans – that is, to drive them towards a warm host (Fig. 2B). It should be noted that, similar to other mutations that affect a single sensory modality, when provided with sufficient additional sensory cues, Ir21a mutants are nevertheless able to successfully find a host (Greppi et al., 2020). Furthermore, although heat seeking is dramatically reduced in Ir21a and Ir93a mutants, it is not completely eliminated, suggesting that additional thermosensory mechanisms remain to be uncovered (Greppi et al., 2020; Laursen et al., 2023).
Infrared sensitivity for longer-range detection
Although a mosquito primarily depends on conduction and convection to detect the heat energy emitted by a nearby host, the effectiveness of these modes of heat transfer decreases rapidly with distance. To detect heat emission over longer distances (tens of centimeters to kilometers), animals instead rely on heat transferred in the form of infrared radiation. In the sections below, we discuss the systems that have evolved independently in snakes and in beetles to enable them to locate distant sources of infrared radiation.
Snakes: temperature-based infrared detectors
Snakes use radiant heat to locate warm-bodied prey
Although many carnivores rely on acute vision to detect prey movement, three groups of snakes – pit vipers (subfamily Crotalinae), boas (family Boidae) and pythons (family Pythonidae) – have specialized thermosensory organs on their faces that extend their ability to detect electromagnetic radiation into the infrared (Goris, 2011; Figs 1 and 3A). This infrared sensitivity allows them to ‘see’ thermal emissions from objects in their environment (Bullock and Cowles, 1952; Bullock and Diecke, 1956; Moiseenkova et al., 2003). Experiments with blindfolded or congenitally blind snakes show that infrared detection can be used to effectively find prey when visual cues are limited or absent, such as when hunting at night or underground (Kardong and Mackessy, 1991; Newman and Hartline, 1982).
In addition to facilitating hunting in the dark, infrared detection has a significant advantage compared with relying on conductive and convective heating alone: range. The Amazon tree boa (Corallus hortulana) accurately strikes at a moving, cloth-covered hot (69°C) lightbulb 48.5 cm away, and the copperhead (Agkistrodon mokasen) preferentially strikes at a freshly killed warm rat versus a freshly killed chilled rat 10 cm away (Noble and Schmidt, 1937). In the latter study, the temperature differential between the two targets at the strike initiation position was <0.2°C, barely perceptible by the thermometers used. Although not discussed in Noble and Schmidt (1937), the minute temperature differential between target and ambient conditions at the snake's position suggests that they are detecting radiant heat, an idea that was subsequently validated using electrophysiology (Bullock and Cowles, 1952; Bullock and Diecke, 1956; Goris and Nomoto, 1967) as well as behavior; for example, a Crotalus atrox rattlesnake can strike a mouse-sized 34°C target at 80 cm (Ebert and Westhoff, 2006). Although their heat-sensing abilities are generally discussed with regard to prey seeking, there is evidence that facial pits are also used for behavioral thermoregulation in snakes, which are poikilotherms (Bakken and Krochmal, 2007; Krochmal and Bakken, 2003; Krochmal et al., 2004).
Snake pit organs are anatomically specialized to act as detectors of infrared radiation
Conspicuous pits are present on the faces of pit vipers, pythons and some boas (although boas lacking pits also possess infrared receptors; Fig. 3A). Infrared detection appears to have evolved independently at least three times in snakes, leading to noticeable interspecies variation in pit organ anatomy, morphology and sensitivity (Goris, 2011). However, the specialized thermosensory role of the pit organ is shared across snakes (Goris, 2011; Noble and Schmidt, 1937).
The function of snake pit organs has been a topic of scientific interest since at least the 1680s (Goris, 2011). One prominent theory held that the facial pits were ears or responded to mechanical stimuli, including air vibrations. In the 1930s, Margarete Ros demonstrated that blocking the facial pits with petroleum jelly was sufficient to disrupt the attraction of her pet African rock python (Python sebae) to a warm object, leading her to conclude that the main function of the pit organ is heat detection to aid in the hunt for warm-blooded prey (Ros, 1935). Subsequent investigations demonstrated that pit organs are exquisitely sensitive thermosensors, responsive to temperature changes of the pit membrane as low as ∼0.003°C (Bullock and Diecke, 1956; De Cock Buning, 1983; Noble and Schmidt, 1937). For a snake at ∼300 K, this corresponds to a 1×10−5-fold change in average kinetic energy! It should be noted, however, that such thermosensitivity is not unique to snakes. For example, cave beetles, D. melanogaster and Caenorhabditis elegans all possess thermosensors with similar sensitivity (Goodman and Sengupta, 2018; Klein et al., 2015; Loftus and Corbière-Tichané, 1987).
Pit vipers have a single pair of pit organs referred to as ‘facial’ or ‘loreal’ pits because of their location in the region between the eye and nostril (Lynn, 1931). The loreal pits are more anatomically specialized (and sensitive) than the labial pits of boas and pythons (labial pits are discussed below). In pit vipers, infrared receptors are located in a thin pit membrane (∼8–16 µm thick) that is loosely suspended across the pit, separating the pit into air-filled inner and outer chambers (Amemiya et al., 1995; Bleichmar and de Robertis, 1962; Terashima et al., 1968; Figs 1 and 3A). Being surrounded by air on both sides insulates the membrane from conductive heat loss to surrounding tissue, and the membrane's location within a recessed pit prevents heat loss through wind convection (Bakken and Krochmal, 2007). A small pore connects the inner chamber to the outside environment and helps to equalize pressure and air temperature (Amemiya et al., 1995, 1996; Fig. 3A). These architectural features place the pit membrane in an excellent position to reliably respond to small changes in temperature.
The pit membrane itself exhibits additional anatomical features that support its exceptional thermosensitivity. It is highly vascularized and densely innervated by three branches of the trigeminal nerve (Amemiya et al., 1999; Bleichmar and de Robertis, 1962; Bullock and Fox, 1957; Lynn, 1931). The nerve fibers extend and branch towards the membrane surface, losing their myelin sheath before ultimately terminating in mitochondria-filled free nerve endings, referred to as terminal nerve masses (TNMs; Bleichmar and de Robertis, 1962; Bullock and Fox, 1957; Terashima et al., 1970; Fig. 3B). Schwann cells (see Glossary) are associated with these TNMs but, rather than providing myelination, they are proposed to maintain proper TNM morphology. The nerve endings are located within ∼2 µm of the membrane's outer surface, placing them at the site of infrared photon absorption. Compared with mammalian thermoreceptors, which are often ∼300 µm below the skin surface, it is estimated that ∼20 times less incident energy is required to heat loreal membrane receptors (Campbell et al., 2002).
Running beneath and between the TNMs, the pit membrane vasculature forms loops around groups of nerve terminals, such that each TNM contacts a capillary on at least one side (Fig. 3B). The capillary density is positively correlated with the activity level of the membrane region – regions of the membrane that detect infrared signals from the front of the animal are more densely vascularized than areas that receive infrared signals from the sides (Amemiya et al., 1999; Goris et al., 2007, 2003; Goris et al., 2000). Together, this pattern suggests the circulatory system supporting the pit organ has a dual function: it not only provides sufficient oxygen to support the metabolic needs of the mitochondria-rich nerve endings but also serves as a heat exchange mechanism to cool the pit membrane, preventing infrared afterimages (Amemiya et al., 1999; Goris et al., 2007, 2003; Goris et al., 2000).
In contrast to the loreal pits of vipers, pythons and boas possess labial pits (Goris, 2011). In pythons, these pits are depressions within labial scales that contain receptors in the pit receptive membrane region (the fundus) at the pit's base (Amemiya et al., 1998, 1995; Fig. 3A). In boas, the pit invaginations are located between adjacent labial scales. However, the infrared receptors are located not in the pits, but rather on the edges of the scales adjacent to the pits (Amemiya et al., 1995; Goris, 2011; von During, 1974; Fig. 3A). As in the loreal pits, the nerve masses associated with labial pits are filled with mitochondria and supplied by a vascular network so dense as to give the pits a red appearance (Amemiya et al., 1998; Goris et al., 2003).
The scales of infrared-detecting snakes also contain specialized architectural features to control detection of electromagnetic radiation. Tiny pores cover the surface of the infrared receptor-containing areas of the labial pits as well as both sides of the loreal pit membrane and the wall of the inner chamber (Amemiya et al., 1995; Campbell et al., 1999). The pores are hypothesized to function as a filter, allowing longer-wavelength infrared light to pass through while reflecting shorter-wavelength visible light (Goris, 2011). In this way, the receptors could be protected from heating by visual light. The back wall of the inner chamber of the loreal pit is lined with large and small domed structures that are proposed to function as a light trap, preventing light from being reflected back onto the membrane (Amemiya et al., 1995).
Mechanistically, the ∼1–3 mm diameter pit openings of pit vipers and pythons function like the aperture of a pinhole camera. Electromagnetic waves traveling through the small opening illuminate the pit's receptive surface and create an inverted image of the target object on the pit receptive membrane. The size of the pit opening reflects a trade-off between sensitivity and resolution. The pit opening must be sufficiently large to allow enough infrared radiation to enter in order to stimulate the receptors, but the relatively large pinhole means the resulting image is of poor resolution (Bakken and Krochmal, 2007; Clark et al., 2022; De Cock Buning, 1984; Sichert et al., 2006). The ability of infrared-sensitive snakes to accurately find and strike prey, even when vision is occluded (De Cock Buning, 1983; Ebert and Westhoff, 2006; Kardong and Mackessy, 1991), suggests that downstream neuronal processing refines these infrared images.
Snake infrared detection relies on highly sensitive thermoreceptors
The primary afferent thermoreceptor neurons that terminate as TNMs are composed of individual fibers from the trigeminal nerve (Bullock and Fox, 1957; Kishida et al., 1982; Lynn, 1931). Electrophysiologically, at constant temperature, these neurons display a baseline firing rate that is transiently suppressed by cooling and transiently increased by warming by as little as ∼0.003°C (Bullock and Diecke, 1956; Ebert and Westhoff, 2006; Goris and Nomoto, 1967). Thus, they can encode bidirectional thermal information (much like the insect cooling cells, noted above). Thermoreceptor responses were not affected when an infrared-transparent filter was placed between the heat source and the pit membrane, but disappeared when the filter was swapped for an infrared-opaque version or a glass of water, indicating that the pit membrane detects radiant heat (Bullock and Diecke, 1956).
An early theory proposed that pit organ thermosensors detected radiation through a mechanism similar to that of a ‘Golay cell’ (see Glossary). In this model (discussed in more detail below for pyrophilous beetles), the absorption of infrared radiation would cause local increases in temperature, expanding the gas in the inner chamber and ultimately activating mechanosensitive neurons in the pit membrane (Block, 1950). However, tearing the pit membrane to eliminate potential pressure differentials does not result in the loss of radiation responsiveness (Bullock and Cowles, 1952). This lends support to the currently accepted mechanism, in which the pit neurons function as a ‘bolometer’ (see Glossary), responding directly to changes in temperature caused by infrared photon absorption (Gracheva et al., 2010; Moiseenkova et al., 2003; Schmitz and Trenner, 2003) (Fig. 4A).
Information from snake infrared detectors is integrated with visual input in the brain
The descending trigeminal tract of infrared-detecting snakes carries the sensory information from the pits, and projects ipsilaterally to a specialized medullar nucleus in the hindbrain known as the lateral descending tract and nucleus of the trigeminal nerve (LTTD) (Meszler et al., 1981; Molenaar, 1974; Schroeder and Loop, 1976; Fig. 3C). In boas and pythons, output from the LTTD is sent directly to the contralateral optic tectum, which is analogous to the superior colliculus of mammals (Kishida et al., 1980; Newman et al., 1980). In crotalines, infrared information is first routed from the LTTD to another ipsilateral medullary processing center, the reticularis caloris (RC), before also being sent to the contralateral tectum (Bothe et al., 2019; Gruberg et al., 1979; Kishida et al., 1980; Newman et al., 1980; Fig. 3C).
The neural circuitry of the infrared detection system has parallels with that of the visual system (Goris, 2011). Similar to the retina, incoming electromagnetic radiation, albeit in the infrared as opposed to visual range, activates receptors on the pit membrane. Electrophysiological recordings from the LTTD show that these inputs form a topographic map (see Glossary; Bothe et al., 2018; Meszler et al., 1981; Schroeder and Loop, 1976). LTTD neurons exhibit center-surround antagonism, in which they show burst responses to a thermal object entering their receptive field, followed by delayed inhibition, which is likely to originate from inputs from adjacent fields. This type of activity is reminiscent of the lateral inhibition that occurs in the visual system – as well as the somatosensory system from which the infrared receptors evolved – and serves to increase contrast (Bothe et al., 2018, 2019; Stanford and Hartline, 1984). As a thermal object moves across the pit's receptive field, the spatiotemporal activation sequence of LTTD neurons can likely be used to compute the direction of movement. Although recordings from LTTD neurons show weak directional sensitivity, much stronger directional motion sensitivity is observed for RC neurons (Bothe et al., 2019).
In the current view of ‘heat vision’, low-resolution infrared images from pit organ receptors are processed in two sequential steps before being integrated with visual information in the optic tectum (Bothe et al., 2019). First, inputs are fed into the LTTD, where their spatiotemporal activation sequence provides weak directional motion encoding, while lateral inhibition from adjacent receptive fields enhances signal contrast and effectively reduces receptive field size (Bothe et al., 2018, 2019; Stanford and Hartline, 1984). Outputs from the LTTD then converge on directionally tuned RC neurons to further improve motion information, before the information is routed to the contralateral optic tectum (Goris and Terashima, 1973; Hartline et al., 1978). In the tectum, infrared information is integrated with inputs from the visual system (Newman and Hartline, 1981). Spatiotopic maps (see Glossary) of the two sensory modalities are roughly overlaid. However, the overlap is not perfect; this has been attributed to the fact that the overall field of the pit is smaller than that of the eye, whereas individual infrared fields are larger than visual fields (Hartline et al., 1978). From the tectum, signals are sent to forebrain regions for further processing (Berson and Hartline, 1988).
Snake pit organs express thermosensitive TRP channels
Although the extreme thermosensitivity of the snake pit organ – and the morphological and circuit adaptations that support it – has been known for decades, the molecular basis for how heat is transduced to activate thermoreceptor neurons is only beginning to be understood. Gracheva et al. (2010) compared the transcriptomes of the trigeminal and the dorsal root ganglia (which innervate the face/pits and the body, respectively) to search for molecules specifically required for the function of the pit organ thermosensors. This study found that transient receptor potential A1 (TRPA1) was highly expressed in the trigeminal compared with the dorsal root ganglia in rattlesnake, boa and python species. By contrast, no differential expression was detected in two non-pit-bearing species. TRPA1 is a cationic ion channel known colloquially as the ‘wasabi receptor’, as it serves as a molecular receptor for electrophilic chemical ligands, including allyl isothiocyanate (AITC) from wasabi and horseradish (Jordt et al., 2004; Laursen et al., 2014). Notably, some TRPA1 orthologs are also activated by temperature, an intrinsic property of the channel that persists when they are expressed in artificial lipid bilayers (Laursen et al., 2014).
Several lines of functional evidence point to a role for TRPA1 in snake infrared detection. Cultured neurons from trigeminal ganglia of pit-bearing pythons show a higher proportion of warmth-activated neurons with lower thermal activation thresholds than in non-pit-bearing ratsnakes (29.5±1.7°C versus 35.6±1.2°C, respectively), and these neurons respond to the TRPA1 agonist AITC (Gracheva et al., 2010). Moreover, a TRPA1-specific blocker significantly reduces the thermosensitivity of cultured python trigeminal neurons (Gracheva et al., 2010). Finally, when expressed in heterologous systems, the thermal activation profiles of snake TRPA1 orthologs roughly mimic those seen for cultured neurons: activation thresholds are lowest for rattlesnake TRPA1, intermediate for boa and python TRPA1s and highest for ratsnake TRPA1 (Gracheva et al., 2010).
The identification of heat-sensitive TRPA1 channels in pit organ thermoreceptors provides additional support for the prevailing theory that the snake pit responds to infrared-induced temperature increases rather than by expressing an infrared-sensitive photoconvertible molecule analogous to those found in visible light photoreceptors. The exceptional thermosensitivity of TRPA1 orthologs from rattlesnakes, pythons and boas also suggests that the enhanced thermosensitivity of infrared-sensitive snakes is not restricted to anatomical and morphological adaptations, but extends to heat receptor molecules as well (Du and Kang, 2020; Gracheva et al., 2010; Kang, 2016; Panzano et al., 2010).
Pyrophilous insects: photo-mechanical infrared detection
Pyrophilous insects use infrared to sense distant forest fires as sites for laying eggs
Unlike the previous examples, pyrophilous insects use thermal cues to find food, not for themselves, but for their offspring. Two groups of buprestid jewel beetles, the Australian fire-beetle Merimna atrata, and species of the genus Melanophila, are attracted to forest fires, where they mate and lay their eggs under the bark of the freshly burned trees that serve as a food source for their larvae (Schmitz, 2004). Other pyrophilous insects, including the little ash beetle (Acanthocnemus nigricans) and some Aradus flat bugs, display similar attraction to forest fires (Kreiss et al., 2007; Schmitz et al., 2008a). Pyrophilous behavior and infrared detection probably evolved independently in different beetle groups, offering an opportunity to explore the different (or in some cases, apparently convergent) mechanisms that have evolved for extremely long-distance heat detection (Schmitz, 2004).
As they exploit forest fires to reproduce, pyrophilous insects have evolved highly sensitive heat-detection mechanisms. Particularly for Melanophila, their exceptional thermosensitivity occasionally leads them astray, causing them to mistakenly congregate at smelting plants, lumber kilns, sugar refineries, oil fires and even football stadiums filled with smoking fans (Linsley, 1943; Schmitz and Bousack, 2012). A notable historical incident – the beetles' appearance at a 1924 oil fire in Coalinga, CA, USA – provides a striking demonstration of this phenomenon (Van Dyke, 1926). The oil fire was over 50 miles from the nearest coniferous forest, from which the beetles are likely to have originated, suggesting that the beetles were able to detect and navigate to the site from a great distance (Linsley, 1943; Schmitz and Bousack, 2012).
How can Melanophila detect heat from such a distance? The temperature of a typical forest fire is 500–1000°C, with maximum emission of infrared radiation in the 2.2–4 μm range (Schmitz and Bleckmann, 1998). Temperature differentials from convective/conductive heating decrease rapidly over comparatively small spatial scales, making it unlikely that the insects could locate fires at a distance by detecting elevations in temperature. By contrast, radiation in the range emitted by forest fires travels through the atmosphere with minimal absorbance. Therefore, an appropriate receptor could potentially detect radiant heat from miles away (Evans, 1966b). Behavioral studies demonstrate that a brief pulse of infrared radiation elicits a twitching of the antennae in Melanophila, with a threshold intensity of 60 μW cm−2. The beetles are most sensitive to 2.4 to 4 μm wavelength light, which corresponds to the emission spectrum of forest fires, suggesting that infrared sensitivity could guide them to freshly burned trees (Evans, 1964, 1966b).
Melanophila infrared sensing organs appear to act as radiation-responsive mechanoreceptors
In Melanophila, infrared receptors are located in two elliptical sensory pits situated on the metathorax, adjacent to the coxae of the mesothoracic legs (Evans, 1966a; Vondran et al., 1995) (Fig. 1). During search flights, the beetles lift their middle legs to fully expose the sense organs to potential radiation (Schmitz and Bleckmann, 1998). At the bottom of each pit, there are 50–100 dome-shaped sensilla, each accompanied by a wax gland (Evans, 1966a; Vondran et al., 1995; Fig. 1). The filamentous wax secreted by the glands fills the pit, yet the function of the wax remains relatively unknown. It has been hypothesized to prevent water loss, to protect the sensilla from dust particles, and/or to insulate receptors from background noise (Evans, 1966a; Hammer et al., 2001). Electron microscopy studies revealed that the interior of the sensillum is filled by a multilayer cuticular sphere consisting of a fluidic core surrounded by an outer lamellated zone comprising layers of chitinous exocuticle (Schmitz and Bleckmann, 1998; Vondran et al., 1995). A single neuron innervates each sensillum with its unbranched, tubular body-containing dendritic tip anchored to the base of the sphere (Fig. 4B).
In contrast to infrared detection in snakes, which relies on thermosensitive neurons, that in Melanophila is thought to involve mechanosensitive neurons, with the absorption of infrared radiation generating pressure within the sensilla that activates mechanosensitive neurons (Schmitz and Bleckmann, 1998; Schmitz et al., 2010; Fig. 4B). Supporting this view, the neurons innervating the sensilla morphologically resemble mechanoreceptors and respond to weak mechanical stimulation as well as infrared radiation (Schmitz and Bleckmann, 1997, 1998; Vondran et al., 1995). The cuticular spherules in which the neurons are embedded also appear to be adapted for this task – they are composed of chitin, which has a stretch resonance of ∼3 μm, allowing radiation of this wavelength to be maximally absorbed. This wavelength matches both the forest fire emission maxima and the maximum behavioral sensitivity of the beetles (Schmitz and Bleckmann, 1998; Vondran et al., 1995).
The sensitivity of Melanophila sensors to infrared radiation measured with electrophysiology is at least 5 mW cm−2 and potentially as low as 500 μW cm−2 (Schmitz and Bleckmann, 1998). It has been calculated that a 500 μW cm−2 threshold would allow detection of a 10 hectare (100,000 m2) fire from 12 km (Schmitz and Bleckmann, 1998), although this would be less sensitive than expected from the >80 mile detection range of Melanophila beetles in historical accounts or the 60 μW cm−2 threshold determined from behavioral experiments (Evans, 1966b; Linsley, 1943; Schmitz and Bousack, 2012). This discrepancy may result from underestimation of the organ's sensitivity in electrophysiology experiments – it could reflect the use of broader wavelength ranges than in behavioral experiments or the insertion of a potentially heat-wicking recording electrode (Evans, 1966b; Schmitz and Bleckmann, 1998).
An alternative explanation for the exceptional sensitivity of Melanophila to forest fires in the field could involve synergism between infrared and other cues. For example, Melanophila antenna detect guaiacol derivatives – major components of wood smoke –with a sensitivity estimated to enable them to detect a single burning tree from >1 km (Schütz et al., 1999). Thus, beetles could use a combination of chemical and infrared cues to achieve greater sensitivity than with either cue alone (Linsley, 1943; Paczkowski et al., 2013; Schmitz, 2004; Schmitz et al., 2008b). However, another study reported that smoke failed to arouse resting beetles and drove no attraction (Evans, 1964; Schmitz and Bousack, 2012). Finally, it has been suggested that beetles might implement active amplification, active sensing and stochastic resonance mechanisms to improve sensitivity (Schneider et al., 2015). In this scenario, Melanophila could tune their wingbeat intensity during flight such that the energy produced by flight muscles brings the mechanosensors just below activation threshold, allowing even a small infrared stimulus to generate action potentials (Schneider et al., 2015).
Other pyrophilous insects may rely on thermosensor-based infrared detection
While not as extensively studied as in Melanophila, the infrared-detecting organs of the pyrophilous flat bug Aradus albicornis display a similar structure to those of Melanophila (Schmitz et al., 2008a), suggesting convergent evolution between these two groups. By contrast, the infrared receptors of M. atrata and A. nigricans differ from those of Melanophila in structure (Schmitz et al., 2000, 2002). In M. atrata, two sets of round, slightly sunken infrared-detecting areas are located at the second and third abdominal sternite. Each area is innervated by a single multipolar neuron (Schmitz et al., 2000). Multipolar neurons of M. atrata have extensively branched dendrites and a high concentration of mitochondria (Kreiss et al., 2005; Schmitz et al., 2000, 2001, 2002), reminiscent of the TNMs of infrared-sensitive snakes (Terashima et al., 1970). Therefore, they are believed to be thermoreceptive neurons that directly sense the temperature change caused by infrared absorption (Schmitz et al., 2000, 2002). In the third species, A. nigricans, there is a pair of disc-shaped organs anterior to the prothoracic legs (Schmitz et al., 2002), each of which is innervated by about 90 ciliary neurons which each terminate in a peg sensillum reminiscent of thermoreceptors (Kreiss et al., 2007, 2005; Schmitz et al., 2002). In another parallel with snakes, specifically pit vipers, A. nigricans has an air-filled inner chamber under the infrared-detecting disc, which is proposed to insulate and lower the thermal mass of the disc, thereby functioning similarly to the membrane and inner chamber of the loreal pit in vipers (Schmitz et al., 2002).
Conclusions and future directions
As we have discussed in this Review, the exploitation of heat emission as a cue to locate food has emerged repeatedly in distantly related animals, with different animal groups implementing distinct approaches for detecting thermal cues at different distances. For space reasons, here we have focused on a few species to detail the ways in which animals have evolved to engage in heat seeking, but such behavior is highly developed in many other animals, including ticks (Carr and Salgado, 2019), kissing bugs (Flores and Lazzari, 1996; Lazzari, 2009), bedbugs (DeVries et al., 2016) and vampire bats (Gracheva et al., 2011; Kurtin and Schmidt, 1982). It is likely that genomic and gene editing approaches will enable rapid progress in identifying the molecules involved in heat seeking in many of these non-traditional animals in the near future.
Although few of the thermosensory systems used for foraging have been investigated in molecular detail to date, these initial studies have pointed to the repurposing of ancestral thermosensors implicated in thermoregulation as a common evolutionary theme (Gracheva et al., 2011, 2010; Greppi et al., 2020; Laursen et al., 2023). Even in these cases, however, the underlying mechanisms of thermotransduction are not well understood. For thermosensitive ion channels, the biophysical basis of temperature sensitivity is increasingly pursued at the structural level (Cao, 2020; Diver et al., 2022; Nadezhdin et al., 2021). Examining related ion channels with differing thermal activation profiles, such the TRPA1s of different snake species, may offer an opportunity to exploit naturally occurring variation to understand how thermal sensitivity is encoded. In addition, although the molecular receptors initiating the sensory neuron's electrical response are critical, these receptors are just one of multiple features that endow these systems with their exceptional thermosensitivity. As highlighted above, the specialized morphology and material composition of the sensory organs that house these receptors also make essential contributions to the sensitivity of the transduction process. Little is understood about how these structures are built or operate at a molecular level.
Finally, foraging ultimately involves the integration of multiple sensory modalities. It will therefore be important to further map the neuronal circuits downstream of these thermosensory organs. We need to identify where and how the neural pathways that control thermoregulation and foraging diverge, as well as to examine how heat seeking is integrated with other external and internal sensory information to allow an animal to successfully find the food sources it requires. Together, the exceptional thermosensory organs and abilities of these varied animals provide promising areas for future investigations of sensory system evolution as well as for addressing fundamental questions in animal physiology and neuroscience.
Acknowledgements
We thank Rachel Busby for drawing the snakes shown in Fig. 2. P.A.G. thanks Jane Kondev, Stan Lazopulo, David Zimmerman and the late Richard Goris for helpful discussions.
Footnotes
Funding
Supported by funding from the National Institute of General Medical Sciences (R01 GM130842) and the National Institute of Allergy and Infectious Diseases (R01 AI157194) to P.A.G.
References
Competing interests
The authors declare no competing or financial interests.