A subset of sweet-sensing neurons identified by IR56d are necessaryand sufficient for fatty acid taste

Fat represents a calorically potent food source that yields approximately twice the amount of energy as carbohydrates or proteins per unit of mass. The highly palatable taste of FAs promotes food consumption, activates reward circuitry, and is thought to contribute to hedonic feeding underlying many metabolism-related disorders. Despite a role in the etiology of metabolic diseases, little is known about how dietary fats are detected by the gustatory system to promote feeding. Previously, we showed that a broad population of sugar-sensitive taste neurons expressing Gustatory Receptor 64f (Gr64f), is required for reflexive feeding responses to both FAs and sugars. Here, we report a genetic silencing screen to identify specific populations of taste neurons that mediate fatty acid taste. We find neurons identified by expression of Ionotropic Receptor 56d (IR56d) are necessary and sufficient for reflexive feeding response to FAs. Functional imaging reveals that IR56d-expressing neurons are responsive to short- and medium-chain FAs. Silencing IR56d neurons selectively abolishes fatty acid taste, while their activation is sufficient to drive feeding responses. Analysis of co-expression with Gr64f identifies two subpopulations of IR56d-expressing neurons. While physiological imaging reveals that both populations are responsive to FAs, IR56d/Gr64f neurons are activated by medium-chain FAs and are sufficient for reflexive feeding responses to FA. Moreover, flies can discriminate between sugar and FAs in an aversive taste memory assay, indicating that FA taste is a unique modality in Drosophila. Taken together, these findings localize fatty acid taste within the Drosophila gustatory center and provide an opportunity to investigate discrimination between different categories of appetitive tastants. Authors Summary Fat represents a calorically potent food source that yields approximately twice the amount of energy as carbohydrates or proteins per unit of mass. Dietary lipids are comprised of both triacylglycerides and FAs, and growing evidence suggests that it is the free FAs that are detected by the gustatory system. The highly palatable taste of FAs promotes food consumption, activates reward centers in mammals, and is thought to contribute to hedonic feeding that underlies many metabolism-related disorders. Despite a role in the etiology of metabolic diseases, little is known about how dietary fats are detected by the gustatory system to promote feeding. We have identified a subset of sugar-sensing neurons in the fly that also respond to medium-chain FAs and are necessary and sufficient for behavioral responses to FAs. Further, we find that despite being sensed by shared neuron populations, flies can differentiate between the taste of sugar and FAs, fortifying the notion that FAs and sugar represent distinct taste modalities in flies.


Authors Summary
(FAs), and growing evidence suggests that it is the free FAs that are detected by the 7 gustatory system [5][6][7]. Sensing of FAs promotes food consumption, activates reward 8 circuitry, and is thought to contribute to hedonic feeding that underlies many metabolism-9 related disorders [8,9]. Despite a role in the etiology of metabolic diseases, little is known 1 0 about how dietary fats are detected by the gustatory system to promote feeding. taste modalities such as sweet, bitter, salty, sour or umami, and project to higher order 1 5 brain structures for processing [10,13,14]. While these taste modalities have been housed in gustatory sensilla located on the tarsi (feet), proboscis (mouth) and wings. include one group that senses sweet tastants and promotes feeding, and a second, non-2 2 overlapping group that senses bitter tastants and promotes food avoidance [17,18]. We have previously shown that Drosophila is attracted to medium-chain FAs [19].

4
Consumption of FAs relies on taste, rather than smell, as it is not impaired by surgical information to the SEZ, and how this information is represented in higher order brain 1 5 centers, is central to understanding the neural basis for taste processing and feeding 1 6 behavior. Identifying the neural principles underlying FA taste processing requires neurons that are responsive to diverse modalities including salt, sugar, amino acids, 2 0 water, carbonation, bitter, polyamines, and electrophilic tastants [17,[26][27][28][29][30][31][32]], yet little is 2 1 known about the populations underlying FA taste. Here, we show that the IR56d 2 2 population of gustatory neurons, which partially overlaps with Gr64f neurons, is 2 3 necessary and sufficient to produce a feeding response to FAs. The IR56d-Gr64f 2 4 population of neurons is selectively responsive to medium-short-chain FAs. Our results 2 5 reveal a defined channel that senses FAs to promote food consumption, providing a 1 mechanism for differentiation between attractive tastants of different modalities. water alone (Fig. 1C-F). The temporal dynamics of Ca 2+ activity differed between the two 1 2 tastants, with HxA eliciting a broader response (Fig. 1E, F), yet both elicited comparable 1 3 peak changes in fluorescence ( Fig 1C). Therefore, both sugars and FAs activate Gr64f 1 4 sweet-sensing GRNs, fortifying the notion that this neuronal class is generally 1 5 responsive to attractive tastants.

6
To localize FA-sensitive neurons within the broad Gr64f population, we selectively with Gr64f (Table S1) [20,30,35,36]. Of the 10 GAL4 lines tested, silencing with GAL4 2 2 drivers for Gr61a, IR56b, and Gr64f resulted in PER defects to sucrose and HxA, By contrast, silencing Gr64e neurons significantly reduced response to sucrose without 2 4 affecting response to HxA, and silencing Gr5a, Gr43a, or IR56d neurons significantly 1 reduced PER to HxA without affecting sucrose response. 2 We chose to further investigate the role of IR56d neurons in FA sensing, since IRs have 3 been found to be involved in detection of non-sweet appetitive tastants, including salt 4 and amino acids [28,37,38]. IR56d neurons have previously been reported to project to 5 two distinct regions in the SEZ. One region overlaps with sweet-sensing projections, and neurons did not affect PER in response to sucrose or HxA, while expression of TNT 2 0 selectively inhibited HxA response (Fig 2D). Therefore, IR56d neurons are required for 2 1 behavioral response to HxA, but dispensable for response to sucrose.

9
IR56d neurons project to both taste peg and sweet-sensing regions of the SEZ, and 1 0 each region can be distinguished anatomically (Fig 2A-C; [36]). To determine whether 1 1 sugars and FAs can differentially activate the two regions, we expressed GCaMP5 in which overlaps with Gr64f neurons, responded to both HxA and sucrose with similar 1 5 magnitude (Fig 3A-B). The projections from the taste pegs, however, showed a strong 1 6 response to HxA, but did not respond to sucrose (Fig 3C-D). Therefore, while IR56d 1 7 neurons that project posteriorly are responsive to both sucrose and HxA, those 1 8 originating in taste pegs are selectively responsive to HxA, revealing two distinct 1 9 functional classes of IR56d taste neurons. to short-chain pentanoic acid (5-carbon), medium-chain octanoic acid (8 carbon), and 1 long chain oleic acid (18-carbon). As compared to control IR56d>imp-TNT flies, IR56d-2 silenced flies exhibited weaker responses to octanoic acid, but retained robust PER to 3 pentanoic acid (Fig 4A), suggesting that PER to short chain FAs is not dependent on 4 IR56d neurons. Oleic acid did not elicit strong PER in either. We directly examined 5 activity induced by different FAs in IR56d neurons with in vivo Ca2+ imaging in IR56d-6 GAL4>GCaMP5 flies. Octanoic acid robustly activated posterior IR56d projections, while 7 both pentanoic and octanoic acid activated anterior IR56d projections (Fig 4B,C). Oleic 8 acid, which did not induce PER, also did not activate IR56d projections in either posterior 9 or anterior regions (Fig 4B,C). Together, these findings show that IR56d neurons silencing IR56d neurons that do not overlap with Gr64f did not affect PER to HxA, or as 1 expected, to sugar, and no difference was observed between flies expressing imp-TNT 2 and TNT, suggesting the taste peg neurons are dispensable for the reflexive feeding 3 response to FAs (Fig 5B). Flies lacking Gr64f-LexA, but still harboring a copy of 4 LexAop-GAL80 (IR56D-GAL4-TNT; LexAop-GAL80/+), showed reduced PER as 5 expected (Fig S1). 6 Although both sugars and FAs induce feeding behavior, it is unclear whether flies can 7 qualitatively differentiate between these two classes of appetitive tastants. We have 8 developed an assay in which an appetitive tastant is applied to the tarsi, paired with 9 bitter quinine application to the proboscis, and the suppression of PER in subsequent 1 0 response to the appetitive tastant is measured [40,45]. To determine whether flies can 1 1 differentiate between sugars and FAs, we applied sucrose (conditioned stimulus) to the 1 2 tarsi followed immediately with quinine application (unconditioned stimulus) to the 1 3 proboscis. Following three training trials, memory was tested by application of sucrose 1 4 (control) or HxA to the tarsi, in the absence of quinine, and measuring PER (Fig 6B). We reduction that was not generalized to sucrose (Fig. 6C). Quantification of the percentage 2 4 reduction in PER reveals that the 'matched' groups, where quinine is paired with an 2 5 attractive tastant suppressed PER by 79-94%, while there was little PER suppression 1 when the tested tastant was unmatched from the one previously paired with quinine (Fig   2   6D). This reciprocal discrimination between sucrose and HxA is different from the 3 unilateral discrimination between two sugars reported previously [45] and indicates that 4 flies can discriminate between sucrose and HxA based on their identity. Thus, sugars 5 and FAs act as independent taste modalities in flies. Here, we identify a small population of neurons within the taste system that is responsive 9 to FAs. Other taste modalities, such as sweet and bitter, have been well studied in flies 1 0 and mammals, but much less is known about FA taste. We previously identified a appetitive tastants such as amino acids [17,28,48]. Our findings indicate that IR56d We identify a subpopulation of sweet sensing neurons that are responsive to FAs. We tastants including a Gr64e population that is sensitive to glycerol [48] and Gr5a subset 1 that is sensitive to trehalose [17]. To localize the Gr64f neurons responsible for FA taste 2 we conducted a targeted screen and silenced neurons that are believed to overlap with 3 Gr64f neurons, which led us to study the IR56d population of neurons. ISilencing Ir56d 4 neurons appears to selectively disrupt HxA taste without affecting response to sucrose, 5 supporting the notion that independent channels within the Gr64f population mediate robust PER in response to pentanoic acid, HxA, and octanoic acid, but not oleic acid.

4
Therefore, it seems likely that flies are selectively responsive to short/medium chain FAs, 1 5 but less responsive to long-chain and/or unsaturated FAs. It is possible that a population 1 6 of taste neurons independent from IR56d are activated by pentanoic acid because it 1 7 elicited a robust behavioral response that was independent of Ir56d neurons. Further, 1 8 IR56d neurons may be activated by long-chain FAs that were not tested, and these 1 9 could modulate feeding response and induce PER. Nevertheless, the findings presented 2 0 here reveal specificity for short/medium chain FAs within a defined population of taste 2 1 neurons, and further specificity for medium chain FAs in a subset of those neurons. The robust PER response induced by two different medium-chain FAs, hexanoic and is sufficient for survival [19]. Therefore, it is possible that FA attraction evolved to 7 promote consumption of calorically rich fruits consumed by Drosophila. by HxA appears more restricted, and therefore it is possible that differences in activation 1 2 allow for differentiation [54]. Alternatively, we find that HxA also activates IR56d neurons 1 3 that emanate from the taste pegs and do not co-express Gr64f, so it is possible that the window of cuticle exposed, then sealed using dental glue (Tetric Flow -Ivoclar 1 2 Vivadent). The proboscis was extended and a small amount of dental glue was used to 1 3 secure it in place, ensuring the same position throughout the experiment.

4
The cuticle and connective tissue was dissected to expose the SEZ, which was bathed  interval and the number of full proboscis extensions was recorded. During training, same 5 stimulation as before was followed by a droplet of 10 mM quinine was placed on 6 extended proboscis and flies were allowed to drink it for up to 2 sec or until they 7 retracted their proboscis. After each session, tarsi and proboscis were washed with 8 water and flies were allowed to drink to satiation. After training, flies were tested with 9 either that same substance without quinine or with the untrained substance (matched or 1 0 opposite trained groups). Another group of flies was measured as described above but control of Drosophila using a red-shifted channelrhodopsin reveals experience-  Trained Naïve