Gated Coupling of Dopamine and Neuropeptide Signaling Underlies Perceptual Processing of Appetitive Odors in Drosophila

How olfactory stimuli are perceived as meaningful cues for specific appetitive drives remains poorly understood. Here we show that despite their enormous diversity, Drosophila larvae can discriminatively respond to food-related odor stimuli based on both qualitative and quantitative properties. Perceptual processing of food scents takes place in a neural circuit comprising four dopamine (DA) neurons and a neuropeptide F (NPF) neuron per brain hemisphere. Furthermore, these DA neurons integrate and compress inputs from second-order olfactory neurons into one-dimensional DA signals, while the downstream NPF neuron assigns appetitive significance to limited DA outputs via a D1-type DA receptor (DoplR1)-mediated gating mechanism. Finally, Dop1R, along with a Gβ13F/Irk2-mediated inhibitory and a Gαs-mediated excitatory pathway, underlie a binary precision tuning apparatus that restricts the excitatory response of NPF neurons to DA inputs that fall within an optimum range. Our findings provide fresh molecular and cellular insights into cognitive processing of olfactory cues.

irreversibly turns Ca 2+ -bound CaMPARI protein from green to red fluorescence, thereby 92 capturing the excitatory state of the DL2 neurons in the aroused larvae. We observed that 93 stimulation by an appetizing dose of PA (7µl) induced excitatory responses from the four DA 94 neurons in each cluster, as evidenced by increased red fluorescence signals in the DL2 neurons 95 over a 5-min test period ( Figure 2B). Similarly, an appetizing dose of balsamic vinegar (20 µl) 96 also triggered a similar rise of intracellular Ca 2+ levels in the DL2 neurons during its stimulation. 97

98
In parallel, we also examined how the DL2 neurons respond to an ineffective dose of PA in fed 99 larvae under the same experimental conditions. In response to 20µl PA stimulation, DL2 100 neurons showed a rapid surge of intracellular Ca 2+ within the first minute, followed by a gradual 101 appetitive arousal. However, such a binary mixture became effective if the testing larvae have 138 reduced TH-activity or pre-fed with 3IY ( Figure 3D). Together, these findings suggest that the 139 minimal strength of an odor stimulus required to arouse appetite is inversely correlated with the 140 baseline level of DA in fed larvae. 141 Our previous study showed that the Drosophila neuropeptide Y-like NPF system contributes to 160 the odor-aroused feeding response of fed larvae by modulating the excitatory responses of the 161 DL2 neurons to olfactory stimuli 7 . Anatomical analyses have revealed that a npf-GAL4 driver 162 predominantly labels six NPF neurons in the larval central nervous system (CNS; Figure 4A, B; 163 16,17 . Another npf-LexA driver, which predominantly labels two dorsomedial NPF neurons in the 164 larval brain, shows that their dendrites are enriched in the lateral horn region (Figure4-- Figure  165 supplement 3). To determine which subset(s) of NPF neurons is necessary to induce appetitive 166 arousal, we performed targeted laser lesioning of NPF neurons. We found that odor-stimulated 167 fed larvae missing the two-lesioned dorsomedial NPF neurons failed to display appetitive 168 response, while those larvae with two lesioned dorsolateral NPF neurons showed normal 169 appetitive behavior ( Figure 4C). 170 171 Given that NPF signaling was previously implicated in modulation of DL2 neuronal responses to 172 the outputs from second-order projection neurons, we decided to determine the observed activity 173 of the dorsomedial NPF neuron could be explained by its positive regulation of the upstream DA 174 neurons. To test this possibility, we used a fluorescent Ca 2+ indicator, GCaMP6 to analyze the 175 excitatory responses of four DL2 neurons to PA stimulation over a 60-sec period in the presence 176 or absence of two dorsomedial NPF neurons. We found that in both cases, the odor responses of 177 DL2 neurons were largely similar. We also performed a reciprocal imaging test under the same 178 odor treatment. In the presence of the DL2 neurons, the NPF neurons were excited by the PA 179 stimulation, but their odor responses were significantly attenuated when the two clusters of DL2 180 neurons were lesioned ( Figure 4D). We also observed that the DL2 neurons responded to the PA 181 stimulation within a few seconds. However, the response of the NPF neurons showed a 182 significant time lag of up to 25 seconds. These results raise the possibility that the dorsomedial 183 NPF neurons may contribute to a yet uncharacterized circuit mechanism by acting downstream 184 from odor-responsive DL2 neurons. 185

186
To test this hypothesis, we functionally knocked down Dop1R1 activity in npf-187 GAL4/Dop1R1 RNAi fed larvae. Similar to Dop1R1/+ fed larvae mutant larvae, these Dop1R1- at any of the doses tested, revealing an essential role of NPF in arousing larval appetite. 192 Furthermore, we genetically activated NPF neurons using a fly TRP family channel, dTrpA1 18 . 193 The npf-GAL4/UAS-dTrpA1 fed larvae, heat shocked at 31 o C for 15 or 30 min, showed in larval and adult flies 20-24 . We postulate that the dorsomedial NPF neuron may contribute to a 212 food odor-aroused motivational state in fed larvae by selectively assigning appetitive 213 significance to DA signals that fall within a limited range. To test this hypothesis, we examined 214 how the dorsomedial NPF neurons respond to appetizing and non-appetizing odor stimuli. We 215 first examined the potential impact of PA stimulation on the membrane potentials of the 216 dorsomedial NPF neurons in a larval preparation using a fluorescent indicator of membrane 217 potential (Arclight) 25 . The dorsomedial NPF neurons showed a gradual increase in excitatory 218 response during a 10-minute continuous exposure to the vapor of PA at an effective dose. In 219 contrast, when a higher ineffective dose of PA was applied, no excitatory responses were 220 observed, except for a transient depolarization immediately following the odor application 221 ( Figure 5A and B). The responses of the dorsomedial NPF neurons to appetizing and non-222 appetizing odor stimuli were also analyzed in behaving larvae using CaMPARI-based imaging 223 ( Figure 5C). Similar to the Arclight imaging results, these neurons showed an excitatory 224 response to an appetizing dose of PA (7µl), but failed to respond to non-appetizing PA doses that 225 are either too high or too low (e.g., 20 or 3.5µl). Therefore, these results indicate that the 226 inverted-U effects of odor stimuli can also be observed at the level of NPF neuronal excitation. 227 On the other hand, in the Dop1R1-defieinct NPF neurons, the effective dose of PA required to 228 activate the dorsomedial NPF neurons is upshifted from 7µl to 20µl PA. 229 230 A binary genetic mechanism for Dop1R1-mediated precision tuning of NPF signaling 231 A key remaining question is how Dop1R1 precisely tunes the two dorsomedial NPF neurons to 232 assign appetitive salience to otherwise behaviorally meaningless DA signals. D1-type DA 233 receptor is associated with the heterotrimeric G protein complex consisting of Gαs, Gβ, Gγ 234 subunits 26 . Upon its activation by DA, the dissociated Gαs subunit and Gβγ complex each 235 defines a separate effector pathway. We found that RNAi-mediated knockdown of Gαs activity 236   that when the Gβ13F/IRK2 pathway is deficient, the effective range of odor-induced DA signals 254 is greatly expanded, converting the inverted U dose-response curve to the sigmoidal shape in 255 these larvae. Together, our findings have revealed a binary genetic mechanism in the NPF 256 neurons that underlies Dop1R1-gated coupling of odor-evoked DA and NPF signaling. 257 Furthermore, Dop1R1 gating of the NPF neurons appears to employ a two-layered precision 258 tuning strategy that ultimately shapes the inverted-U dose response of fed larvae: the Gαs-259 mediated excitatory pathway is used to set up a minimal threshold level of odor-evoked DA 260 signals required for NPF neuronal excitation; the Gβ13F/IRK2-mediated inhibitory pathway is 261 specialized for preventing NPF neuronal response to any odor-evoked DA signals that are 262 excessively strong ( Figure 6). 263

DISCUSSION 265
To date, how sensory inputs are perceptually processed in deep brain centers to arouse specific 266 behavioral drives remains poorly understood 32 . We have shown that Drosophila larvae display 267 aroused appetite for selected macronutrients in response to stimulation by food-related odors. 268 Using this behavioral paradigm, we have identified a pair of higher-order neural circuits for The dose-response analysis shows that the appetizing effects of both monomolecular and mixed 290 odor stimuli follow an inverted U function. We have also provided evidence that the amount of 291 DA released from the DL2 neurons is positively correlated with the strength of odor stimulation, 292 and induction of appetitive arousal requires an optimum level of odor-evoked DA released from 293 the DL2 neurons. For example, a higher non-appetizing dose of PA (e.g., 20µl PA) triggered a 294 rapid and intense excitatory response from the DA neurons. As the result, a 30-sec stimulation 295 by 20µl PA triggered appetitive arousal while its stimulation for 5 minutes failed to do so. 296 Another key insight from this study is that odor-evoked DA signals encode no intrinsic appetitive 297 values. Instead, larval perceptual recognition of appetitive odor cues requires a Dop1R gating 298 mechanism, which is preset genetically, to selectively assign appetitive significance to a narrow 299 spectrum of the DA signals. For example, a 50% reduction of Dop1R1 activity in Dop1R1/+ 300 larvae led to a right-shift of the dose-response curve regardless whether the odor stimulus is 301 monomolecular or chemically complex. Therefore, to be perceived as appetizing cues, odor-302 evoked DA signals must fall within a narrow range that matches the pre-existing level of 303 Dop1R1 activity. 304 305

The central role of NPF neurons in regulation of appetitive arousal 306
We have shown that two NPF neurons, located in the dorsomedial region of each brain lobe, are 307 necessary and sufficient to elicit appetitive arousal in fed larvae. Anatomical and functional 308 analyses also suggest that these NPF neurons likely form synaptic connections to the upstream 309 clustered DL2 neurons ipsilaterally. In freely behaving fed larvae, the NPF neurons display 310 excitatory responses to stimulation by appetizing doses of PA but not the non-appetizing doses, 311 and such differential responses are gated by the Dop1R1 activity in the dorsomedial NPF 312 neurons. Since dTRPA1-mediated genetic activation of the two dorsomedial NPF neurons, even 313 under prolonged heat treatment, was sufficient to trigger increased feeding response, these results 314 suggest that elevated NPF neuronal signaling plays an acutely role in the induction and 315 maintenance of appetitive odor-aroused motivational state. It remains to be determined whether 316 the NPF neurons are responsive to sugar stimulation. If so, this would provide evidence for an 317 important role of these neurons in integrating DA-coded olfactory and sugar-evoked gustatory 318

signals. 319
Precision tuning of Dop1R1-gated NPF neuronal response to odor-evoked DA signals 320 We have provided molecular and cellular evidence for how the Dop1R1 activity in the NPF 321 neurons shapes the inverted-U effects of appetitive odor stimuli. Using both behavioral and 322 functional imaging assays, we show that Dop1R1 determines which range of odor-evoked DA 323 signals may acquire appetitive significance through precision tuning of NPF neurons. Two 324 separate genetic pathways have been identified for this process ( Figure 6B). One of them, 325 involving a Gβ13F/IRK2-mediated pathway, sets an upper limit of the optimum effective range 326 of odor-induced DA signals by silencing the NPF neurons when Dop1R1is hyper-activated. 327 Attenuation of the Gβ13F or IRK2 activity greatly broadens the effective dose range of DA 328 signals, as evidenced by the change of the dose-response curve from an inverted U shape to a 329 sigmoidal shape. Again, this result also strongly supports the notion that the ineffectiveness of 330 stronger odor stimuli is caused by an excessive release of odor-evoked DA, which leads to a high 331 level of dissociated G complex that activates IRK2 channels and subsequently silences the NPF 332 neurons. The second Dop1R1/Gαs-mediated pathway provides a default excitatory mechanism 333 that mediates NPF neuronal response to any odor stimuli that are at or above the minimal 334 threshold strength. Therefore, they together define a two-layered precision tuning strategy. 335

336
The inverted-U effects of DA on cognitive functions have been widely observed in humans and 337 other animals 33,34 . Malfunctioned dopamine systems also underlie many psychiatric disorders 338 such as schizophrenia 35,36 . In the prefrontal cortex of mammals, an optimum level of D1-type 339 DA receptor activity is crucial for spatial working memory, while its signaling at levels that are 340 too low or too high leads to impaired working memory 37-39 . Therefore, our findings raise the 341 possibility that a homologous DA receptor-mediated tuning strategy may be used to mediate the 342 inverted-U effects of DA in flies and mammals. 343

Fly Stocks and Larval Growth
All flies are in the w 1118 background. Larvae were reared at 25°C, and early third instars (~74 hr after egg laying, AEL) were fed before behavioral experiments as previously described 7

Behavioral Experiments
Quantification of mouth hook contraction rate in liquid food was performed as previously were tested for their feeding responses. Feeding media include agar paste (US Biological, A0940) containing 10% glucose, 0.5% Tryptone (Becton-Dickinson, 628200) or 3% oleic acid (Sigma-Aldrich, 112-80-1). UAS-dTrpA1 was expressed by allowing larvae to feed in prewarmed yeast paste in a 31°C incubator for defined periods, followed by rinsing with 31°C water prior to feeding assays.

Molecular Cloning and Immunostaining
To construct the npf-LexA driver, a DNA fragment of ~1-kb containing a region spanning from the 5' regulatory sequence to the beginning of the second axon was amplified by genomic PCR.
This fragment was subsequently cloned into the KpnI site in the pBPnlsLexA::GADflUw vector. Forward Primer: cagggagagagaacggagac; Reverse primer: gtgtcacaatgcaattgttcg. Tissue dissection and fixation, antibodies used and dilution conditions were described previously 7 .

Targeted Laser Lesion
Protocols for calibration of 337 nm nitrogen laser unit and laser lesion experiments have been described 42 .

CaMPARI Imaging
The conditions of larval feeding and odor treatment for CaMPARI imaging are identical to those for odor-aroused feeding behavioral assays. After odor stimulation, larvae were irradiated with PC light 405 nm LED array (200 mW/cm 2 , Loctite) for 5s. The treated larval CNS was dissected and individually scanned using a Zeiss LSM 510 confocal microscope.

GCaMP imaging
The neural tissues of larvae were processed as previously described 7 . A Ca2+ indicator, GCaMP6s , was used for imaging odor responses by DA and NPF neurons. The odor delivery system involves a sealed 1.25L glass chamber fumigated with 7.5μl PA for 5 minutes. Odor was continuously delivered to larval head region by pumping at the rate of 0.28 L/min. The treated larval preparation was imaged using a Zeiss LSM 510 confocal microscope.

Arclight imaging and data processing
The method for making larval CNS preparation is the same as previously described for calcium imaging preparation (7). The preparation was incubated in Drosophila PBS. Effective and ineffective odor vapors were prepared by fumigating a sealed 24L foam box with 150 or 800µl PA for 2 hours, respectively. Odor was continuously delivered to larval head region by pumping at the rate of 0.36L/min.
The protocol for ArcLight Imaging 25 was followed with minor modifications. Briefly, larval CNS was imaged under 40X water immersion lens using Zeiss Axio Examiner. NeuroCCD-SM camera and Turbo-SM software (RedShirt Imaging) were used for recording and data processing.
Images were captured at a frame rate of 100 Hz, and exposure time is 10ms. 2000 frames were collected for each of the seven 20s periods.
All the time series curves were low pass filtered with a Kaiser-Bessel 30 filter (200 Hz cut off).
Then, each curve was fitted with a single exponential equation, I=Ae(-at). Bleaching of each curve was corrected based on the formula: It,corrected=It+(A----Ae(-at)). Normalization of each trace was achieved by dividing It,corrected with the average value. Standard deviation was obtained using normalized data.

Statistical Analysis
Statistical analyses for behavioral and CaMPARI imaging assays were performed using One-way ANOVA followed by Dunnett's or Sidak's multiple comparisons test.        to appetitive response to a normally appetitive binary mixture containing 2.5µl PA and 12.5µl

Supplemental information includes Supplemental Experimental
Hep, but showed appetitive response to the normally non-appetitive binary mixture (n>16). *p<0.05;**P < 0.001. The presumptive synaptic connections are labeled by split GFP (green) between DA neurons (red) and the NPF neuron in the lateral horn (DL2-LH, see Figure 4H).