An opposing self-reinforced odor pre-exposure memory produces latent inhibition in Drosophila

Prior experience of a stimulus can inhibit subsequent acquisition or expression of a learned association of that stimulus. However, the neuronal manifestations of this learning effect, named latent inhibition (LI), are poorly understood. Here we show that odor pre-exposure produces LI of appetitive olfactory memory performance in Drosophila. Behavioral expression of LI requires that the context during memory testing resembles that during the odor pre-exposures. Odor pre-exposure forms an aversive memory that requires dopaminergic neurons that innervate the γ2α′1 and α3 mushroom body compartments - those to α3 exhibit increasing odor-driven activity with successive pre-exposures. In contrast, odor-specific responses of the corresponding mushroom body output neurons are suppressed. Odor pre-exposure therefore recruits specific dopaminergic neurons that provide teaching signals that attach negative valence to the odor itself. LI of Drosophila appetitive memory consequently results from a temporary and context-dependent retrieval deficit imposed by competition with this short-lived aversive memory.


Introduction
Many studies have shown that LI is sensitive to properties (frequency, amount and duration) 161 of the conditioned stimulus [40][41][42] . We therefore also tested whether two exposures of a lower 162 odor concentration (10 -6 rather than 10 -3 ) with 15 min ITI induced LI of appetitive memory. 163 Although training with this odor concentration produced robust appetitive memory, no LI effect 164 was observed following odor pre-exposure (Fig. 1H). 165 166 Odor pre-exposure can produce facilitation of aversive memory 167 We also tested whether odor pre-exposure altered performance measured after aversive 168 training, pairing odor with electric shock. Starved flies were given two 2 min odor X 169 presentations with a 15 min ITI before being trained by presenting odor X for 1 min, paired 170 with electric shocks, then 45 s later presenting odor Y. Surprisingly, flies pre-exposed to the 171 CS+ exhibited faciliated aversive memory performance as compared to flies pre-exposed to 172 mineral oil (Fig. 1I). Finding that the same odor pre-exposure schedule can inhibit appetitive 173 memory but faciliate aversive memory performance led us to hypothesize that pre-exposure 174 might form an avoidance memory for the CS+, that in the former case is competing and the 175 latter complementary. 176 177

Pre-exposure forms a short-lived mushroom body-dependent aversive odor memory 178
Prior work has shown that odor pre-exposure can enhance subsequent odor avoidance 179 behavior 14,15,33 , consistent with our hypothesized aversive learning model. We therefore tested 180 whether such an effect resulted following our LI odor exposure regimen. As in the prior 181 experiments starved flies were exposed twice to 2 min of odor X (10 -3 dilution in mineral oil) 182 with a 15 min ITI. They were then immediately tested for preference between the pre-exposed 183 odor X and another odor Y, without any training. Flies pre-exposed twice to odor X showed a 184 selective avoidance of odor X, consistent with pre-exposure forming an aversive odor memory 185 ( Fig. 2A). In contrast, a single odor pre-exposure, of either 2 or 4 min, did not alter odor 186 preference, suggesting that repeated exposure is required to form the avoidance memory (Fig.  187 2A). Measuring odor preference at different times after pre-exposure revealed that the 188 avoidance memory is labile and slowly decays between 15 min and 2 h (Fig. 2B). We also 189 tested whether expression of pre-exposure memory was sensitive to context by pre-exposing 190 flies to odor in copper grid tubes and testing them for odor preference in clear tubes. Performance was unaffected by this change of context (Fig. 2C), demonstrating that context 192 is uniquely important for the LI effect of pre-exposure memory. We last tested whether the 193 explicit absence of food acts as aversive reinforcement for hungry flies by pre-exposing 194 satiated flies. However, pre-exposure induced similar odor avoidance in satiated flies as 7 compared to starved flies (Fig. 2D). Since our LI effect was assayed in hungry flies, all 196 subsequent experiments in this study were performed in hungry flies. 197 198 Olfactory memories typically depend on the neuronal circuitry of the MB 39,43-47 . We therefore 199 tested the consequence of blocking output from αβ and γ subsets of MB KCs, on the effect of 200 odor pre-exposure. We expressed the dominant temperature-sensitive UAS-Shibire ts1 (UAS-201 Shi ts1 ) transgene 48 with MB247-GAL4 and blocked KC output throughout the experiment by 202 raising the temperature to restrictive 33˚C. This manipulation abolished the development of 203 odor avoidance performance (Fig. 2E). Moreover, restricting inhibition of KC activity to only 204 the periods of odor pre-exposure, using expression of the green light-sensitive anion-selective 205 GtACR1 channel 49 revealed that KC activity is necessary during pre-exposure (Fig. 2F). 206 Finding a role for the MB suggests the pre-exposure effect is a form of associative learning. 207

208
Since we observed LI when flies were twice pre-exposed to 10 -3 odor concentration but not to 209 10 -6 , we also tested whether this lower odor concentration produced an aversive pre-exposure 210 memory. Consistent with the lack of LI, no enhanced avoidance was observed following 10 -6 211 odor exposures (Fig. 2G). In addition, when naïve flies were given the choice between a 10 -3 212 odor stream and air they exhibited avoidance of the odor (Fig. 2H). In contrast, they either 213 showed no preference, or odor approach when tested with a 10 -6 odor stream and air. We 214 therefore reasoned that lower odor concentrations, which are less repellent 50 , do not act as 215 an aversive reinforcer during odor pre-exposure. 216 217 Pre-exposure memory requires γ2α′1 and α3 DANs 218 Aversive olfactory learning reinforced by electric shock, bitter taste or heat depends on 219 punishment coding DANs from the PPL1 cluster 35,37,51-54 . In contrast, DANs in the PAM cluster 220 mostly code for reward 29,30,34 . We therefore first used TH-GAL4 and R58E02-GAL4 to express 221 UAS-Shi ts1 and test the respective roles of PPL1 and PAM DANs in pre-exposure learning. 222 Blocking TH-GAL4 neurons switched the effect of pre-exposure from generating aversion to 223 approach (Fig. 3A). This reversal of valence implies that removing aversive signalling may 224 either release (mutually exclusive) or unmask (in parallel) positive reinforcement induced by 225 odor pre-exposure. Blocking the rewarding DANs with R58E02-GAL4; with UAS-Shi ts1 226 increased the aversive effect of pre-exposure ( learning 54 , are dispensable for pre-exposure learning. In contrast, blocking PPL1-γ2α′1 DANs 234 (MB-MV1) abolished pre-exposure learning, while blocking the PPL1-α3 DANs converted pre-235 exposure-induced aversion to approach (Fig. 3C). 236

237
Since PPL1-DANs are thought to provide an aversive teaching signal, common to electric 238 shock, heat and bitter taste we also tested for a role of PPL1-α3 DANs in shock-reinforced 239 aversive memory. Although blocking the four PPL1-DANs impaired shock learning, blocking 240 only the PPL1-α3 DANs left immediate shock memory intact (Fig. 3E). In addition, replacing 241 shock with optogenetic stimulation of the four PPL1-DANs, could artificially implant an aversive 242 memory whereas a single pairing of odor presentation with PPL1-α3 DAN activation did not 243 form aversive memory (Fig. 3F). These data are consistent with a prior study that showed 244 learning requires multiple trials of PPL1-α3 DAN activation 55 . These results demonstrate that 245 two-trial pre-exposure learning and single-trial electric shock learning involve different PPL1-246 DANs and emphasize the importance of PPL1-α3 DANs for pre-exposure learning. 247

248
The activity of PPL1-γ2α′1 and PPL1-α3 DANs is altered following odor presentation 249 Since pre-exposure learning depends on repeated trials and requires the PPL1-γ2α′1 and 250 PPL1-α3 DANs we imaged DAN activity during and after odor presentations under the 251 microscope (Fig. 4). Flies were constructed that expressed the fluorescent UAS-GCaMP6m 252 calcium sensor in PPL1-γ2α′1 or PPL1-α3 or DANs. As before in the behavioral 253 experiments, flies were given two odor presentations with 15 min ITI. They were then 254 immediately given another 5 s exposure of the trained odor followed by 5 s of a novel odor, 255 to mimic the behavioral test situation under the microscope. Odor-evoked responses of 256 PPL1-α3 DANs increased from the first to the second trial ( Fig. 4A), whereas PPL1-γ2α′1 257 DAN responses did not change (Fig. 4C). However, PPL1-γ2α′1 DANs exhibited a reduced 258 response to the trained compared to the novel odor in the 5s test, whereas PPL1-α3 DANs 259 showed no difference between novel and test odor responses ( Fig. 4B and 4D).

261
In addition, consistent with a lack of LI (Fig. 1H) and pre-exposure memory (Fig. 2F), lower 262 10 -6 odor concentration did not increase the odor-evoked responses of PPL1-α3 DANs (Fig.  263   4E). Together these results suggest that the increased odor-driven activity of the PPL1-α3 264 DANs, specifically in the 2 nd pre-exposure, is crucial for the formation of an aversive pre-265 exposure memory. The predominant model for Drosophila learning is dopamine-driven depression of synapses 269 between odor-specific KCs and MBONs 27,56-60 . We therefore used MBON expression of 270 GCaMP6m to test whether KC-MBON connections underlying the PPL1-α3 and PPL1-γ2α′1 271 DANs were changed following repeated odor exposure (Fig. 5). Odor responses of both 272 MBON-α3 and MBON-γ2α′1 were both reduced compared to their responses to a novel odor 273 ( Fig. 5A and 5B). In addition, blocking either the α3 or γ2α′1 MBONs throughout an odor 274 exposure experiment with UAS-Shi ts1 , or specifically during the test phase with UAS-GtACR1 275 abolished exposure-induced avoidance behavior in the T-maze (Fig. 5C-5E). Taken together 276 these data indicate that spaced odor exposure forms aversive memory, that manifests as 277 reduced odor-evoked activity of the α3 and MBON-γ2α′1 MBONs resulting from increased 278 odor-evoked activation of the corresponding DANs. 279

α3 MBON output is required for expression of latent inhibition of appetitive memory 280
We last tested whether the activity of the γ2α′1 and α3 MBONs was required for the 281 expression of LI. Flies expressing UAS-GtACR1 in γ2α′1 and α3 MBONs were subjected to 282 the standard pre-exposure regimen followed by appetitive conditioning (Fig. 5F). Blocking 283 output from γ2α′1 MBONs impaired the expression of appetitive memory, so their 284 contribution to LI could not be further tested (data not shown). However, silencing the α3 In this study we demonstrate latent inhibition (LI) in Drosophila and identify an underlying 290 neuronal mechanism. We find that repeated odor presentation forms a labile self-reinforced 291 aversive memory for that odor, which can temporarily compete with the expression of a 292 newly acquired appetitive memory for that same odor. During memory testing the 293 conditioned odor should therefore activate both the memory of the pre-exposure (odor-self) 294 and that of the appetitive conditioning (odor-sugar). Importantly, the aversive pre-exposure 295 memory is labile which means the LI effect is transient. As a result, the appetitive memory 296 performance exhibits 'spontaneous recovery'. These results demonstrate that a retrieval (R) 297 model underlies LI in the fly. An acquisition (A) model is not supported because flies acquire 298 an associative reward memory for the odor after pre-exposures of that odor. Instead, the 299 expression of the learned approach performance is impeded by the co-expression of a 300 competing aversive pre-exposure memory. In further support of this R model, the same pre-301 exposure regimen caused facilitation of a subsequently acquired aversive olfactory memory. 302 In this instance the pre-exposure memory adds to the new aversive associative memory, 303 rather than competes with an appetitive memory. 304 A defining feature of LI is a sensitivity to the consistency of the context in which the pre-305 exposure, learning and testing are carried out 61 . Changing between the clear and paper-306 lined tubes did not impair LI, suggesting that the flies likely consider these to be a similar 307 context. However, if odor pre-exposure, learning and testing were performed in different 308 contexts (ie. a copper grid-lined versus a paper-lined, or clear tube) LI was abolished. Most 309 strikingly, LI could be restored if copper grid tubes were used to provide the same context 310 during pre-exposure and testing. In line with prior theories and studies of LI in other animals 311 62,63 , these results suggest that flies learn an association between the odor and the context in 312 which it is experienced during the non-reinforced pre-exposure. As a result, the pre-313 exposure memory gains context-dependence, and our experiments show it is not retrieved if 314 the context is different when memory is tested. The failure to retrieve the pre-exposure 315 memory in a different context manifests as a loss of LI -the appetitive memory is fully 316 expressed. Our study therefore reveals that the context-dependency of LI results from the 317 ability (correct context, LI evident), or inability (wrong context, no LI), to retrieve the pre-318 exposure memory. In addition, context only plays a role in the expression of the pre-319 exposure memory when it is in conflict with a subsequently acquired appetitive memory. 320 Further work will be required to define what the flies recognise as a 'change of context'. 321 There are many possibilities including, background odors, tube/paper texture, relative 322 luminance, and other flies in the group. 323 We found that the odor-driven activity of γ2α′1 and α3 DANs increased with repeated odor 324 pre-exposure and that they were required for the formation of the odor pre-exposure 325 memory. In addition, the odor-specific responses of the corresponding MBONs were 326 depressed following pre-exposure. We therefore conclude that ramping odor-driven DAN 327 activity assigns negative value to the odor itself by depressing odor-specific KC connections 328 onto the γ2α′1 and α3 MBONs. In support of this model, repeated pre-exposure of flies to a 329 lower and less innately aversive odor concentration did not increase the activity of the α3 330 DANs, or form an aversive pre-exposure memory. Importantly, reduced odor-activation of 331 the approach-directing γ2α′1 and α3 MBONs MBONs is sufficient to account for the aversive 332 nature of pre-exposure memory. Moreover, both expression of pre-exposure memory and LI 333 are abolished if the α3 MBONs are blocked during testing, confirming the model that LI is 334 produced by the expression of the aversive pre-exposure memory competing with that of the 335 associative reward memory. 336 Prior work has described odor-driven activity of the PPL1-α′3 DANs, and subsequent 337 depression of odor-specific responses of the α′3 MBONs to underlie how flies become 338 familiar with an odor following repeated short exposures 17 . In contrast, we show that two 339 longer and spaced odor exposures produce an aversive memory that manifests as plasticity 340 of γ2α′1 and α3 DANs and MBONs. Moreover, whereas we show a retrieval defect underlies 341 LI, a reduced attention/familiarity to the odor following pre-exposures would be expected to 342 result in a subsequent acquisition defect. 343 LI has often been compared to memory extinction 64 and our work in the fly shows that very 344 similar neuronal mechanisms account for both of these phenomena. Pre-exposure learning 345 in Drosophila appears to follow similar rules to extinction learning following aversive olfactory 346 conditioning; 2 spaced trials with 15 min ITI is more efficient than massed training with 1 min 347 ITI 33 and in both cases a resulting parallel opposing odor-nothing memory inhibits the 348 retrieval/expression of the odor-punishment or odor-reward memory 33,65 . The obvious 349 difference is that the interfering non-reinforced odor memory is formed before learning for LI, 350 and after learning for extinction. 351 Our studies of learning, extinction and LI suggest that flies acquire and store all of their When assessing LI, odor pre-exposure was followed by appetitive olfactory conditioning, 396 performed essentially according to Krashes and Waddell 39 : flies were exposed for 2 min to 397 odor Y without reinforcement, in a tube with dry filter paper (the conditioned stimulus -, CS-), 398 30 s of clean air, then 2 min with odor X with saturated 5.8M sucrose, dried on a filter paper 399 (the conditioned stimulus+, CS+). For assessing facilation of aversive memory, odor pre- KCl, 5mM N-Tris, 10 mM trehalose, 10 mM glucose, 7mM sucrose, 26 mM NaHCO3, 1mM 435 NaH2PO4, 1.5 mM CaCl2, 4mM MgCl2, osmolarity 275 mOsm, pH 7.3) and the fly, in the 436 recording chamber, was placed under the Two-Photon microscope (Scientifica). 437 Flies were exposed to odors under the microscope using essentially the same regimens and 438 odor concentrations as those in the behavioral experiments. Flies were subjected to a 439 constant air stream, carrying vapor from mineral oil solvent (air). For the odor pre-exposures, 440 an odor stream was added to the air for 2 min. Flies in the custom chamber were then 441 removed from the microscope and rested for 15 min until being returned to the microscope 442 and given the 2 nd odor exposure. The carbogenated buffer was changed before each re-443 exposure. To emulate the testing phase, after the 2 nd exposure, the flies were sequentially 444 exposed to the re-exposed odor and a novel odor, each for 5 s, interspersed by 30 sec of air. 445 As in the behavior experiments the odors were MCH and OCT, and they were used 446 reciprocally. GCaMP responses were measured in the relevant DANs and MBONs during 447 pre-exposure and test phases. 448 One hemisphere of the brain was randomly selected to image the dendritic field of each 449 MBON and the presynaptic terminals of each DAN. Flies that did not respond to one of the 450 two presented odors were excluded from the analyses in this study. Each n corresponds to a 451 recording from a single fly. 452 Fluorescence was excited using ~140 fs pulses, 80 MHz repetition rate, centered on 910 nm 453 generated by a Ti-Sapphire laser (Chameleon Ultra II, Coherent). Images of 256 x 256 pixels 454 were acquired at 5.92 Hz, controlled by ScanImage 3.8 software 80 . Odors were delivered 455 using a custom-designed system 81 . 456 For analysis, two-photon fluorescence images were manually segmented using Fiji 82 . 457 Movement of the animals was small enough such that images did not require registration.

Statistical Analysis 465
Statistical analyses were performed in GraphPad Prism. All behavioral data were analyzed 466 with an unpaired t-test or a one-way ANOVA followed by a posthoc Tukey's or Bonferroni's 467 multiple comparisons test. No statistical methods were used to predetermine sample size. 468 For the imaging experiments odor-evoked responses were compared by a paired t-test for 469 normally distributed data, otherwise a Wilcoxon matched-pairs signed rank test was used for 470 non-Gaussian distributed data. Normality was tested using the Shapiro-Wilk normality test. 471 For imaging data, a method for outlier identification was run for each dataset (ROUT 472 method), which is based on the False Discovery Rate (FDR). The FDR was set to the 473 highest Q value possible (10%). In datasets in which potential outliers were identified, 474 statistical analyses were performed by removing all odor-evoked responses for those flies. 475 The analyses with or without the outliers were not different, so we decided to maintain and 476 present the complete datasets, which may contain potential outliers.  Psychol. 4, 113-134 (1929).

7.
Wagner, A. R. Priming in STM: An information-processing mechanism for self-492 generated or retrieval-generated depression in performance.  Sci. 80, 1482-1486 (1983).  (80-. ). 288, 672-675 (2000). (A) Flies were twice pre-exposed to odor X or mineral oil (MO) in clear tubes (context A), with 672 a 15 min inter-trial interval. Immediately following the last exposure, they were trained by 673 presenting odor Y for 2 min, then odor X for 2 min with sugar (CS+; context A*). Flies pre-674 exposed to odor X exhibited appetitive memory performance to odor X (one-sample t-test: 675 t(7)=2.419, p=0.0461) that was significantly reduced (LI) in comparison to flies exposed to MO 676 [t(14)=4.719, p=0.0003]. (B) When pre-exposure and training context were more closely 677 matched, using filter paper in the pre-exposure tube (context A*A*A), flies pre-exposed to odor 678 X did not exhibit appetitive memory performance (LI) (one-sample t-test: t(7)=0.9866, 679 p=0.3567). Performance was also significantly reduced compared to flies pre-exposed to MO 680 Matching context between pre-exposure and testing (both in copper grid tube; context BA*B) 684 restored LI. Performance of flies pre-exposed to odor X was significantly impaired compared 685 to flies pre-exposed to MO [t(14)=5.449, p<0.0001]. (E) The inhibitory effect on appetitive 686 learning is evident 2 h after odor pre-exposure. Flies were pre-exposed to Odor X or MO then 687 2 h later were appetitively trained and tested immediately for memory. Performance of flies 688 pre-exposed to odor X was significantly reduced compared to flies pre-exposed to MO 689 immediately trained with 1:10 -6 odor X paired with sugar (CS+) and tested for memory. Flies 698 pre-exposed to 1:10 -6 odor X showed similar memory performance to those pre-exposed to 699 MO [t(14)=0.6741, p=0.5112]. (I) Odor pre-exposure facilitates memory performance after 700 aversive conditioning. Flies were pre-exposed to odor X or mineral oil (MO) in a copper grid-701 lined tube (context BBA), then immediately trained by pairing odor X for 1 min with twelve 30V 702 electric shocks (CS+). Flies pre-exposed to odor X exhibited increased aversive memory