Functional expression of Δ12 fatty acid desaturase modulates thermoregulatory behaviour in Drosophila

Polyunsaturated fatty acids (PUFAs) play crucial roles in adaptation to cold environments in a wide variety of animals and plants. However, the mechanisms by which PUFAs affect thermoregulatory behaviour remain elusive. Thus, we investigated the roles of PUFAs in thermoregulatory behaviour of Drosophila melanogaster. To this end, we generated transgenic flies expressing Caenorhabditis elegans Δ12 fatty acid desaturase (FAT-2), which converts mono-unsaturated fatty acids to PUFAs such as linoleic acid [C18:2 (n-6)] and linolenic acid [C18:3 (n-3)]. Neuron-specific expression of FAT-2 using the GAL4/UAS expression system led to increased contents of C18:2 (n-6)-containing phospholipids in central nerve system (CNS) and caused significant decreases in preferred temperature of third instar larvae. In genetic screening and calcium imaging analyses of thermoreceptor-expressing neurons, we demonstrated that ectopic expression of FAT-2 in TRPA1-expressing neurons led to decreases in preferred temperature by modulating neuronal activity. We conclude that functional expression of FAT-2 in a subset of neurons changes the thermoregulatory behaviour of D. melanogaster, likely by modulating quantities of PUFA-containing phospholipids in neuronal cell membranes.

Drosophila melanogaster has been used as a model to evaluate the role of PUFAs in cold acclimation. As shown in other species, acclimation to low temperatures increased the proportions of PUFA-containing phospholipids in the cell membranes [26][27][28][29] , and these have been shown to contribute to the modulation of membrane fluidity and improve their development and survival in cold environments 30 . Recently, it was reported that breeding of flies at 12 °C switches dietary preference from yeast to PUFA-containing plant foods 30 . Because feeding behaviour is altered to modulate PUFA contents in response to temperature changes, we hypothesised that other temperaturerelated behaviour is affected by PUFA contents.
Thermoregulation is accomplished by temperature-sensing followed by effector reactions 31,32 . Behavioral thermoregulation (migrating to environments with more comfortable temperatures) is an essential part of the effector reactions in D. melanogaster 33 , because the body temperature of ectotherms is strongly affected by environmental temperature. Therefore, we focused on the behavioral thermoregulation in Drosophila. Previously, we reported that the Dystroglycan mutant fly atsugari showed cold-seeking behaviour with enhanced mitochondrial oxidative energy metabolism 34 ; however, the factors that influence behavioral thermoregulation are not completely understood.
In this study, we investigated the roles of PUFAs in thermoregulatory behaviour and characterised the underlying molecular mechanisms. Although PUFAs have been shown to play important roles in the visual system 35 , synaptic functions 36 and follicle maturation 37 in D. melanogaster, the role of PUFAs in thermoregulatory behaviour has not been reported. Because D. melanogaster cannot synthesise PUFAs de novo, we established a transgenic D. melanogaster strain that expresses C. elegans FAT-2 under the control of the galactose-responsive transcription factor (GAL4)/upstream activating sequence (UAS) system. Then, we examined the effects of tissue-and cell type-specific expression of FAT-2 on thermoregulatory behaviour. From these data, we describe mechanisms how PUFAs affect thermoregulatory behaviour in D. melanogaster.

Results cold acclimation alters fatty acid compositions and thermoregulatory behaviour in D. melanogaster.
To define relationships between cold acclimation and thermoregulatory behaviour in Drosophila, temperature preference in third instar larvae of D. melanogaster was evaluated using a thermal gradient plate after incubation at 25 °C or 18 °C for 1 day. Cold exposure significantly altered average preferred temperature from 22.5 °C ± 0.2 °C to 20.7 °C ± 0.3 °C (Fig. 1a,b and Supplementary Table S1), indicating that thermoregulatory behaviour is affected by environmental temperature. In subsequent experiments, we assessed the effects of cold exposure on fatty acid compositions of phospholipids using liquid chromatography-tandem mass spectrometry (LC-MS/MS) (Fig. 1c,d). 3)] were present at significantly increased in cold-exposed larvae. To determine whether PC and PE molecules with two or three double bonds includes PUFAs in their acyl chains, we performed product ion scan analyses of PC (32:2), PE (34:2) and PC (34:3) using LC-MS/MS ( Supplementary Fig. S1). PC (32:2) yielded a product ion that coincided with C16:1 (n-7) ( Supplementary Fig. S1a,b), but PE (34:2) and PC (34:3) yielded a product ion of C18:2 (n-6) (Supplementary Fig. S1c-f). The detected C18:2 (n-6) is presumed to be derived from their diet, because flies cannot synthesize C18:2 (n-6). In contrast, the product ion scan analysis of PC (32:1), PC (34:1) and PE (34:1), whose proportions were decreased in cold-exposed larvae, yielded mainly product ions of C16:1 (n-7), C18:1 (n-9) and C16:0 (Supplementary Fig. S2a-f). Hence, we hypothesised that cold exposure increases the presence of PUFA-containing phospholipids that promote thermoregulatory cold preference in D. melanogaster because C18:2 (n-6)-containing phospholipids were accumulated following cold exposure.

Establishment of transgenic flies expressing Δ12 fatty acid desaturase.
To determine relationships between tissue-specific changes in lipid profiles and thermoregulatory behaviour, we employed the Drosophila GAL4/UAS system 38 to modulate PUFA contents in a tissue-specific manner. Wild-type D. melanogaster only expresses Δ9 fatty acid desaturase and hence cannot produce PUFAs. Unlike Drosophila, C. elegans expresses Δ12 fatty acid desaturase (fat-2), which converts mono-unsaturated fatty acids to PUFAs such as C18:2 (n-6) and C18:3 (n-3) 14,25 (Fig. 2a). In the present transgenic D. melanogaster, the UAS-fat-2 construct was introduced using P-element-mediated transformation. The transgene was integrated between the two protein-encoded genes, namely, tolloid (tld) and abnormal spindle (asp), in chromosome 3 (Fig. 2b). To confirm functional expression of FAT-2 from the introduced GAL4/UAS system, phospholipids were extracted from third instar larvae ectopically expressing FAT-2 and fatty acid compositions were measured. The amount of the total phospholipids was not affected by the expression of FAT-2 (Fig. 2c). In the control strain (w 1118 > FAT-2), C16:0, C16:1 (n-7) and C18:1 (n-9) were the most abundant fatty acids, and C18:2 (n-6) was only present in 6.3% of all phospholipid acyl chains. Ubiquitous expression of FAT-2 using the tub-GAL4 driver increased the proportion of C18:2 (n-6) to 36.4% and hexadecadienoic acid [C16:2 (n-6)] to 9.0% of all acyl chains of phospholipids. The proportions of C18:3 (n-3) was also increased in FAT-2-expressing larvae (Fig. 2d). This confirms that the introduced FAT-2 enzyme has Δ12/Δ15 desaturase activity.
Effects of tissue-specific FAT-2 expression on lipid composition. To evaluate the effects of FAT-2 expression in a tissue-specific manner, we induced its expression by a variety of tissue-specific GAL4 drivers. The larvae expressing FAT-2 driven by r4-GAL4 (fat body) and Myo31DF-GAL4 (gut) showed that the proportions of C18:2 (n-6) in acyl chains of whole-body phospholipids were increased by up to 23 Fig. S3). In contrast, the FAT-2 expression driven by elav-GAL4 (neuron) did not significantly affect the proportion of C18:2 (n-6) in acyl chains of whole-body phospholipids. To determine why the pro- Figure 1. Cold exposure-induced changes in temperature preference and lipid compositions of third instar D. melanogaster larvae. Comparison of temperature preference of the w 1118 wandering third instar larvae continuously cultured at 25 °C (n = 6) (a) and exposed at 18 °C for 1 day (n = 6) (b). The histogram shows distributions of the third instar larvae on the thermal gradient plate. The distribution curve is denoted by a black solid line and the average temperature preference is shown as a blue vertical line. The dotted curve represents the distribution of w 1118 control larvae cultured at 25 °C. The numerical analyses of data are also shown in Supplementary Table S1. The proportions of phosphatidylcholine (PC) (c) and phosphatidylethanolamine (PE) (d) molecules in w 1118 larvae continuously cultured at 25 °C (white bar, n = 3) and w 1118 larvae exposed to 18 °C for 1 day (black bar, n = 3) were analysed using LC-MS/MS. Phospholipid molecules are shown in the format PC (X:Y) or PE (X:Y), where X denotes the total number of acyl chain carbons and Y denotes the total number of double bonds in acyl chains. Data are presented as means ± standard errors (SE); * p < 0.05; ** p < 0.01; *** p < 0.001, Student's t-test. www.nature.com/scientificreports/ portion of C18:2 (n-6) in whole-body phospholipids was not affected by elav > FAT-2, we analysed lipids from central nervous system (CNS; brain and ventral nerve cord) using LC-MS/MS (Fig. 3a,b). In larvae expressing  Fig. S5g,h) were not significantly increased in CNS of elav > FAT-2 larvae (Fig. 3a,b), the proportions of C18:2 (n-6)-containing species were increased in these phospholipids.   Table S2), respectively. Given that the greatest change in temperature preference was induced by neuron-specific expression of FAT-2, we further analysed the role of neuronal PUFAs in the regulation of thermoregulatory behaviour.
To determine whether the thermoregulatory behaviour is also affected by the expression of the desaturase with distinct substrate preference, we overexpressed the native D. melanogaster Δ9 fatty acid desaturase DESAT1 that produces mono-unsaturated fatty acids. We established a UAS-DESAT1 transgenic strain using the same methods as that of the generation of a FAT-2 transgenic strain (see "Methods"). In contrast to FAT-2 overexpression, no significant difference was observed in preferred temperature between control (elav > w 1118 )  Table S3). These results suggest that the production of PUFAs, rather than mono-unsaturated fatty acids, in neurons contribute to the regulation of thermoregulatory behaviour in Drosophila.

Energy metabolism is unchanged by neuronal expression of FAT-2. In a previous study, we
showed that the cryophilic mutant fly, atsugari, had increased energy metabolism 34 . Thus, to clarify the roles of energy metabolism in the phenotypes associated with FAT-2 expression, we measured adenosine triphosphate (ATP) concentrations and metabolic rates in FAT-2-expressing larvae. ATP concentrations were significantly  The distribution curve is denoted by a black solid line and the average temperature preference is shown as a blue vertical line. The dotted curve represents the distribution of the control (w 1118 > FAT-2). The temperature preference of third instar larvae expressing FAT-2 under the control of tissue-specific GAL4 drivers (tub-GAL4, r4-GAL4, Myo31DF-GAL4 and elav-GAL4) (n = 6) (d, f, h and j, respectively). As control, the temperature preference of third instar larvae in the w 1118 strain crossed with the w 1118 (n = 7) (a), tissue-specific GAL4 drivers (tub-GAL4, r4-GAL4, Myo31DF-GAL4 and elav-GAL4) (n = 6) (c, e, g and i, respectively) and UAS-FAT-2 strain (n = 6) (b) was analysed. The numerical analyses of data are also shown in Supplementary Table S2. www.nature.com/scientificreports/ decreased in larvae expressing FAT-2 in fat bodies, but were not affected by ubiquitous and neuron-or gutspecific expression of FAT-2 (Fig. 5a). Similarly, metabolic rates of neuronal or ubiquitous FAT-2-expressing larvae were comparable to those of control larvae (Fig. 5b). These results suggest that changes in the temperature preference of elav > FAT-2 larvae is likely not related to alterations of energy metabolism.

FAT-2 expression affected TRPA1-expressing neuron-mediated thermoregulatory behav-
iour. Because thermoregulatory behaviour in D. melanogaster is closely associated with thermosensation 39 , neuron-specific expression of FAT-2 may affect thermosensory neuron activities and influence temperature preference. To identify thermosensory neurons that are responsible for FAT-2-induced changes in temperature preference, FAT-2 was expressed in various thermoreceptor-expressing neurons, which reportedly determine temperature preference at ambient temperatures. We used transient receptor potential A1 (TRPA1)-GAL4 as a driver of warm sensory neurons 40 and iav-GAL4 41 , R11F02-GAL4 42,43 , TRP-GAL4 44 and TRPL-GAL4 44 as drivers of cold sensory neurons. Compared with control larvae (w 1118 > FAT-2) ( Fig. 6a and Supplementary Table S4) or thermoreceptor-specific GAL4 drivers crossed with w 1118 (Fig. 6b,d,f,h,j and Supplementary Table S4), the preferred temperature was significantly decreased in larvae expressing FAT-2 induced by all thermoreceptorspecific GAL4 drivers (Fig. 6c,e,g,i,k and Supplementary Table S4). Although decreases in preferred temperature was observed with all GAL4 constructs (Fig. 6a-k and Supplementary Table S4), the most drastic shift to low temperature was induced by FAT-2 expression in TRPA1-expressing neurons ( Fig. 6c and Supplementary  Table S4). TRPA1 has at least four isoforms and is transcribed from two distinct promoters 45 . TRPA1-A and TRPA1-B were expressed in the brain using the first promoter, whereas TRPA1-C and TRPA1-D were expressed mainly in multidendritic class IV neurons using the second promoter. FAT-2 expression under the control of either TRPA1-AB-GAL4 or TRPA1-CD-GAL4 caused significant decrease in preferred temperature (Fig. 7a
These results indicate that increases in C18:2 (n-6)-containing phospholipids in TRPA1-expressing neurons enhance their sensitivity to warm temperature, resulting in changed thermoregulatory behaviour in D. melanogaster.

Discussion
In this study, we established a transgenic Drosophila line of the C. elegans Δ12 fatty acid desaturase FAT-2, which was functionally expressed in Drosophila (Fig. 2d). The desaturation of fatty acids is performed by a multienzyme complex comprising cytochrome b5, cytochrome b5 reductase and fatty acid desaturase. FAT-2 carries the putative cytochrome b5 domain but requires the specific cytochrome b5 reductase HPO-19 and T05H4.4 for activation in C. elegans 51 . In D. melanogaster, the putative cytochrome 5b reductase CG5946 (NP_729751.1) has 59% and 55% amino acid sequence identity with HPO-19 (NP_504638.1) and T05H4.4 (NP_504639.1), respectively, and likely donates electrons for FAT-2 activity.
Dietary PUFAs are known to affect thermoregulation in animals, such as lizards 52,53 and birds 54 . In addition, the dietary supplementation with C18:2 (n-6) has been reported to affect the visual system 36  www.nature.com/scientificreports/ cold temperatures in Drosophila 30 . However, the significant roles being played by the changes in lipid profiles of specific tissues remain unknown. In the present study, we employed the GAL4/UAS system to express FAT-2 in a tissue-specific manner and found that neuronal expression of FAT-2 confers the most significant decrease in preferred temperature in the third instar larvae of Drosophila. Since significant increases in the proportion of PUFA-containing phospholipids were observed in the CNS of the larvae expressing FAT-2, it is apparent that the modulation of lipid molecules in the neurons was sufficient to affect behavioral thermoregulation in Drosophila.  www.nature.com/scientificreports/ In D. melanogaster, various thermosensitive receptors have been identified using a combination of genetic approaches, electrophysiological techniques and analyses of temperature preference 55,56 . The TRPV family member inactive (IAV) is expressed in chordotonal organs of flies and is required for cold sensation 41 . Moreover, the TRPC family members TRP and TRPL are reportedly required for cold avoidance 44 . Channel proteins of other protein families have also been associated with thermoregulatory behaviours in D. melanogaster larvae. Among these, the ionotropic receptors IR21a and IR25a are expressed in cold-sensitive dorsal organ neurons of larvae and are required for cold avoidance 43 . TRP channel family members are the most extensively studied thermoreceptor members and function as sensors for various environmental cues, including chemical and physical stimuli [57][58][59] . Among them, the TRPA1 channel is well known for its roles in the detection of noxious chemicals and unfavorable temperatures in various animals 60 . Although the thermosensitivity of TRPA1 remains controversial in mammals, the Drosophila TRPA1 channel is demonstrably activated by elevated temperature and by noxious chemicals 60,61 . In Drosophila, TRPA1 was identified as the receptor that controls thermoregulatory behaviour 40 . Moreover, TRPA1 is expressed in anterior cell neurons of adult D. melanogaster, and TRPA1deficient D. melanogaster show decreased aversion to warm temperatures 62 . TRPA1 is also important for the temperature preference 47,63 or warm temperature-induced rolling behaviour 49 of D. melanogaster larvae. We demonstrated that responses of TRPA1-expressing neurons to warm temperatures were significantly increased by FAT-2 expression (Fig. 9c), suggesting that PUFAs, such as C18:2 (n-6), affects thermoregulatory behaviour Figure 7. Temperature preference of third instar larvae expressing FAT-2 in distinct TRPA1-expressing neurons. The histogram shows distributions of third instar larvae on the thermal gradient plate. The distribution curve is denoted by a black solid line and the average temperature preference is shown as a blue vertical line. The dotted curve represents the distribution of control (w 1118 > FAT-2) flies. The temperature preference of third instar larvae expressing FAT-2 under the control of TRPA1-expressing neuron-specific GAL4 drivers (TRPA1-AB-GAL4 and TRPA1-CD-GAL4) (n = 6) (c and e, respectively). As control, the temperature preference of third instar larvae in w 1118 strain crossed with TRPA1-expressing neuron-specific GAL4 drivers (TRPA1-AB-GAL4 and TRPA1-CD-GAL4) (n = 3) (b and d, respectively) and UAS-FAT-2 strain (n = 9) (a) was analysed. The numerical analyses of data are also shown in Supplementary Table S5 39 . Although the process from temperature reception to thermoregulatory behaviour has not been completely revealed, it is apparent that the modulation of the activity of TRPA1-expressing neurons is a crucial mechanism for the modulation of thermoregulatory behaviour in Drosophila in different conditions. FAT-2-mediated decreases in preferred temperature were also observed in larvae expressing FAT-2 under control of other GAL4-thermoreceptor constructs that are associated with cold sensation (Fig. 6 and Supplementary Table S4). Previous studies showed that the TRP channel activities were modulated by lipid molecules 64 . In particular, C18:2 (n-6) reportedly activated the TRPL in Drosophila 35,65 , and inhibit the response to menthol in mammalian TRPM8 and response to capsaicin in mammalian TRPV1 66 . Although the effects of PUFAs on the cold sensation by iav and ionotropic receptors have not been reported, the C18:2 (n-6) production by FAT-2 in cold-sensitive neurons may desensitize or inhibit thermoreceptors that are involved in the cold sensations. The evaluation of the effects of C18:2 (n-6) on each predicted cold channel using the cell-based assay that was used in this study for TRPA1 analysis (Supplementary Fig. S7) may reveal the actual regulatory functions of C18:2 (n-6) with respect to cold receptors.
Motter et al. expressed rat TRPA1 in HEK293T cells and showed strong activation following treatments with arachidonic acid [C20:4 (n-6)], eicosapentaenoic acid [C20:5 (n-3)] and docosahexaenoic acid [C22:6 (n-3)] but only weak activation by C18:2 (n-6) 67 . Yet in their experiments, D. melanogaster TRPA1 was not activated by C22:6 (n-3). These investigators also showed that activation of TRPA1 by PUFAs does not involve the known ligand-binding domains, suggesting that other transmembrane or intracellular domains mediate direct interactions with PUFAs. In our study, non-esterified C18:2 (n-6) did not affect the function of D. melanogaster TRPA1 as an activator ( Supplementary Fig. S8), indicating that C18:2 (n-6) is not likely to be a ligand of Drosophila TRPA1. We also showed that the responses to AITC were not increased in FAT-2-expressing neurons (Fig. 9e). This suggests that the increased temperature-dependent activities in FAT-2-expressing neurons were not caused by temperature-independent events, such as increased expression of TRPA1. Several TRP channels carry phosphatidylinositol-binding domains that are required for regulation of channel activity 68 . Structural analysis also demonstrated the presence of a phospholipid-binding motif in TRPA1, suggesting that TRPA1 is regulated by direct interactions with phospholipids 69 . In another study, TRP channels activities were highly sensitive to changes in the physicochemical properties of bilayer membranes, such as membrane tension 70 . Moreover, GCaMP6m (n = 11) (white dot) and TRPA1-AB > FAT-2; GCaMP6m (n = 12) (black dot) neurons. AITC activations were measured in the same neurons after temperature-induced activation. Activity was calculated as ∆F max /F 20°C or ∆F max /F 0 , where F 20°C and F 0 were the fluorescent intensities at 20 °C and at the initial timepoint, respectively, and ∆F max was calculated by subtracting F 20°C or F 0 from maximum fluorescent intensities. Bars indicate medium values. P-values were calculated using Student's t-tests (*p < 0.05).
Scientific RepoRtS | (2020) 10:11798 | https://doi.org/10.1038/s41598-020-68601-2 www.nature.com/scientificreports/ the domains required for thermal activation of TRPA1 were found in pore regions 71 or in N-terminal regions upstream of ankyrin repeats 45,72 . These domains were distinct from those related to activation by AITC 73 . Thus, PUFA-containing phospholipids may affect the thermosensitive domains of TRPA1 through direct binding or the modulation of the physicochemical properties of the lipid membrane ( Supplementary Fig. S9); however, further studies are required to elucidate the mechanisms involved at the molecular level. We recently observed an accumulation of PUFA-containing phospholipids in the CNS of D. melanogaster 74 . PUFA-containing phospholipids are transported to the CNS through the receptor-mediated endocytosis of lipophorin 74 . Accordingly, we detected remarkably high expression levels of lipophorin receptors LpR1 and LpR2 in the CNS. From these observations, it is apparent that PUFAs are selectively transported to the CNS and might have specific functions in neurons. In the present study, we revealed that PUFAs in the phospholipid acyl chains in the thermosensor-expressing neurons affect the activity of thermosensory neurons and thermoregulatory behaviour. Further detailed analysis of lipophorin receptor expression in thermosensor-expressing neurons will elucidate the mechanisms of PUFA-mediated control of thermoregulatory behaviour at the molecular level. Moreover, novel physiological meanings and molecular mechanisms underlying the regulations of the contents and distribution of PUFA-containing lipid molecules will be revealed by tissue-specific expression of FAT-2 using a transgenic fly strain established in this study. Fly stocks were raised on a corn/yeast/glucose medium containing 80 g brewer's yeast powder, 100 g glucose, 40 g cornmeal, and 7 g agar, 8.8 mL propionic acid, and 0.88 g butyl parahydroxybenzoate per 1 L water. Fly stocks were cultured at 25 °C with a 12-h light/12-h dark cycle. The strain w 1118 was used as a control. In cold acclimation assays, vials were placed in an incubator at 18 °C for 1 day (20-23 h).
For generating UAS-DESAT1 transgenic Drosophila, the coding sequence of D. melanogaster DESAT1 was isolated from cDNA library of the fly strain of Canton-S as previously described 50 . The DESAT1 cDNA was amplified by PCR using primers 5′-CTG AAG TAA AAC AGT TGT TGC AAC ATGC -3′ and 5′-CAT GAT TGG CCC TAC GCT CAA CCT GCCT -3′, and the resulting PCR product was cloned into a vector using the TA-cloning method. The cDNA insert was then digested with EcoRI and BglII at the 5′ and 3′ terminals, respectively.
The digested cDNA inserts were cloned into the pUAST vector 38 at the corresponding cloning sites. The resulting constructs were subjected to P element-mediated transformation of the w 1118 strain 75 . The integrated position of the transgene was determined through a sequence analysis of the franking genome. temperature preference assays. Temperature preference of wandering third instar larvae was assayed as described previously 34 . Briefly, a temperature gradient was generated on an aluminium plate with a peltier device at both ends of the plate (DIA Medical System Co.). A glass plate covered with 2.2% agarose was placed on the aluminium plate. Subsequently, 30-40 larvae were placed on the position of the glass plate at 28 °C. Distributions of larvae were measured after 20 min. Lipid analysis. Lipids were extracted from homogenised samples using the Bligh and Dyer method 76 and were then dissolved in chloroform. Phospholipids were fractionated from total lipid extracts using thin layer chromatography, using hexane/diethyl ether/acetic acid (60:40:1, v/v/v) as the solvent.
The amount of the total phospholipids was determined by inorganic phosphate quantification as previously described 77 and normalised to protein contents that were measured using Pierce BCA Protein Assay Kits (Thermo Fisher, USA).
Lipid analyses using gas chromatography (GC) were performed as described previously 5,50 . The extracted phospholipids were incubated in a 5% hydrogen chloride/methanol solution (Nacalai Tesque, Japan) at 100 °C for 3 h. Fatty acid methyl esters were then analysed using GC-14A (Shimadzu, Kyoto, Japan) with a flame ionisation detector and Supelco Omegawax Capillary GC columns (0.25 μm, 30 m × 0.25 mm; Sigma-Aldrich, USA). The column temperature was held at 180 °C for 5 min, was ramped to 220 °C at 3 °C/min and then held for 7 min and was finally ramped to 240 °C at 3 °C/min and held for 15 min. Fatty acid peaks were identified using GC-MS.
LC-MS/MS analyses were performed using a high-performance liquid chromatography system LC-30AD (Shimadzu, Kyoto, Japan) coupled to a triple quadrupole mass spectrometer LC-MS-8040 (Shimadzu, Kyoto, Japan) that was equipped with an electrospray source as described previously 5,74 . The extracted phospholipids were separated using a Kinetex C8 column (2.6 μm, 2.1 × 150 mm) (Phenomenex, USA) with mobile phases comprising 10 mM ammonium formate in water (mobile phase A) and 10 mM ammonium formate in 2-propanol/acetonitrile/water ( Atp measurements. ATP concentrations were measured as described previously 34 . Briefly, ten third instar larvae were homogenised in lysis buffer of an ATP Bioluminescence Assay Kit HS II (Roche, Switzerland) on ice and were then incubated at 72 °C for 15 min. Subsequently, homogenates were centrifuged at 15,000 rpm for 5 min and supernatants were centrifuged again. Supernatants were mixed with luciferase reagent and luciferase activity was quantified using an Infinite F200 pro (TECAN, Switzerland) instrument. ATP concentrations were normalised to protein concentrations that were measured using Pierce BCA Protein Assay Kits (Thermo Fisher, USA).
Measurements of metabolic rates. Metabolic rates of D. melanogaster were measured according to CO 2 production as described previously 34 . Briefly, ten third instar larvae were placed in a 1-mL plastic syringe containing a small piece of soda lime (Wako, Japan). A glass capillary was connected to one end of the syringe and a small amount of ink was placed at the end of the capillary. The syringe was then preincubated at 25 °C for 15 min and movements of ink were measured over 1 h. Volumes of CO 2 produced were normalised to the body weights of ten larvae. ca 2+ imaging. To measure responses of TRPA1-expressing neurons, UAS-GCaMP6m was expressed using TRPA1-AB-GAL4. Brains were dissected from third instar larvae and were collected in recording buffer containing 5 mM TES, 10 mM HEPES, 120 mM NaCl, 3 mM KCl, 4 mM MgCl 2 , 2 mM CaCl 2 , 10 mM NaHCO 3 , 10 mM trehalose, 10 mM glucose and 10 mM sucrose (pH 7.25). Brain samples were placed on glass-bottomed dishes coated with poly-L-lysine, and the dishes were then filled with recording buffer and placed on a DTC-300C temperature controller (DIA Medical System Co., Tokyo, Japan). The setting temperature was increased from 19 to 40 °C, and the temperature near sample was recorded using a TA-29 thermistor (Warner instrument, USA). Ca 2+ imaging was acquired using an inverted Zeiss LSM 800 confocal microscope equipped with a 10 × /0.25 objective and ZEN2.3 software (https ://www.zeiss .com/micro scopy /int/produ cts/micro scope -softw are/zen.html#modul es). Increases in Ca 2+ concentrations in BLP neurons were measured and activities were calculated as ∆F/F 20 °C , where F 20 °C is the fluorescent intensity at 20 °C and ∆F is the change in fluorescence from F 20 °C . For cellular Ca 2+ imaging, Drosophila S2 cells were maintained in Schneider's Drosophila medium supplemented with 10% fetal bovine serum (FBS), 50-units/ml penicillin, and 50 μg/ml streptomycin at 25 °C. The TRPA1 (isoform A) expression vector was generated by cloning into the pAc5.1 plasmid and was subsequently transfected with the pCoBlast plasmid using TransFectin lipid reagent (Bio-Rad). Stable transformants were selected using 20 μg/ml blasticidin. Cells were incubated with 100 μΜ fatty acid [C18:1 (n-9) or C18:2 (n-6)]-BSA complex for 6 h as described previously 50 . Prior to Ca 2+ measurements, cells were seeded on glass coverslips and were loaded with Fura2-AM (10 μM) (Dojindo, Japan) and probenecid (500 μM) in Schneider's Drosophila medium for 60 min at 25 °C. After washing with recording buffer, coverslips were placed and slides were mounted in a perfusion chamber on a microscope (Axio-observer Z1, Carl-Zeiss, Germany). The recording buffer used in TRPA1-expressing neurons was supplemented with 500 μM probenecid and 1 mg/ml BSA. The temperature of the perfusion solution was controlled using a CL-100 bipolar temperature controller equipped with an SC-20 inline solution heater/cooler (Warner Instruments, USA). Time-lapse images were recorded every 3 s. Ratiometric images (F 340 /F 380 ) were analysed using the Physiology module of AxioVision (Axiovs40 V 4.8.2.0, https ://www. zeiss .de/axiov ision , Carl-Zeiss, Germany).

Statistics.
All experiments were performed at least three times. Distribution analyses of larvae in temperature preference assays were performed using nonparametric Mann-Whitney U tests to compare two samples and Kruskal-Wallis tests followed by Steel-Dwass tests for multiple comparison. In the analyses of parametric data, Student's t-tests were used to compare two samples, and Dunnett's tests or Tukey-HSD tests were used to compare three or more samples. Statistical analyses were performed using JMP software or R programmes 78