Prior Cocaine Exposure Increases Firing to Immediate Reward While Attenuating Cue and Context Signals Related to Reward Value in the Insula

The insula contributes to behavioral control and is disrupted by substance abuse, yet we know little about the neural signals underlying these functions or how they are disrupted after chronic drug self-administration. Here, male and female rats self-administered either cocaine (experimental group) or sucrose (control) for 12 consecutive days. After a 1 month withdrawal period, we recorded from insula while rats performed a previously learned reward-guided decision-making task. Cocaine-exposed rats were more sensitive to value manipulations and were faster to respond. These behavioral changes were accompanied by elevated counts of neurons in the insula that increased firing to reward. These neurons also fired more strongly at the start of long-delay trials, when a more immediate reward would be expected, and fired less strongly in anticipation of the actual delivery of delayed rewards. Although reward-related firing to immediate reward was enhanced after cocaine self-administration, reward-predicting cue and context signals were attenuated. In addition to revealing novel firing patterns unique to insula, our data suggest changes in such neural activity likely contribute to impaired decision making observed after drug use. SIGNIFICANCE STATEMENT The insula plays a clear role in drug addiction and drug-induced impairments of decision making, yet there is little understanding of its underlying neural signals. We found that chronic cocaine self-administration reduces cue and context encoding in insula while enhancing signals related to immediate reward. These changes in neural activity likely contribute to impaired decision making and impulsivity observed after drug use.


INTRODUCTION
The insula has recently gained traction as a key contributor to relapse and drug-seeking 58 behaviors, and as a potential therapeutic target for addiction. Work in humans has shown that 59 insula is activated during the presentation of drug cues and recollection of past drug use (Wang 60 et al., 1999;Bonson et al., 2002). Further, it has been shown that damage to insula can promote 61 drug abstinence (Naqvi et al., 2007). Rodent work has supported these ideas by showing that 62 disruption of insula function reduces the ability of drug-associated cues and contexts to drive During this task (Fig. 1A,B) rats use reward-predicting cues to guide choice behavior 79 between two options, the values of which are manipulated across two contexts: in one, rats 80 choose between an immediate and delayed reward; in the other, rats choose between a large and 81 small reward with delays to reward held constant. In both contexts, rats learn which option yields 82 the preferred outcome, and maintain those expectations during delays to reward and across trial 83 blocks while following forced-choice rules. We have previously shown that rats prefer 84 immediate over delayed reward, and large over small reward, and that rats are more motivated  Given insula's role in addiction and its dysregulation following drug use, we 90 hypothesized that functional signals related to reward processing would be disrupted by chronic 91 cocaine self-administration. We found reward delivery-related signals in cocaine-exposed rats 92 were enhanced, while context-and cue-related firing was reduced. Further, encoding of 93 immediate delivery of reward was more prominent in cocaine-exposed rats. These results suggest 94 heightened reward responding and lower cue and context selectivity in insula may contribute to 95 increased impulsivity and poor decision-making observed after chronic drug use (Roesch et al.,    During each recording session, reward value was independently manipulated across four 116 blocks of 60 correct trials (Trial summary in Fig. 1A; Block summary in Fig. 1B). During the 117 first two blocks, one well was randomly designated to deliver reward immediately (500ms delay; 118 0.05ml) while the other well delivered sucrose with delays that would gradually increase (1000-119 7000ms). Delay contingencies would then be switched at the start of the second block, so that the 120 well previously containing higher-valued reward (short delay) now carried the longer delay. 121 During the final two trial blocks, delays on both wells were held constant at 500ms, and reward 122 value was instead manipulated by size. At the start of the third block, the well that previously 123 carried the long delay condition now delivered two boli of sucrose solution. Value contingencies 124 were again switched at the fourth block (i.e, well previously containing large reward now 125 contained small reward, and vice versa). Rats were water deprived to increase motivation to 126 complete the task. Water bottles were removed by noon the day before training began. After a 127 session, rats were given water for 20 minutes. Because testing took place Monday through 128 Friday, rats were given ad lib access to water after Friday's session and then had water bottles 129 removed Sundays at noon. This deprivation schedule has been used in our previous studies, and 130 has been shown to increase motivation in the behavioral task without detriment to the rats' health  All subjects were implanted with catheters in the jugular vein for self-administration and 135 drivable electrodes (+1.5AP, +/-5.0ML, -5.0DV from brain surface) for single-unit recordings   Associates operant chambers (St. Albans, VT). Animals in the cocaine group could press a lever 145 for an infusion of cocaine, while control animals pressed for sucrose pellets. Cocaine dosage was 146 calculated by weight, with separate doses calculated for males and females to account for weight 147 differences. On days 1-6, rats self-administered 1 mg/kg dosage of cocaine (experimental group) 148 or 2 sucrose pellets (control) per lever press for a maximum of 30 infusions or for a 3 h time 149 limit. For the final 6 days, the dosage of cocaine was halved to 0.5mg/kg, and only 1 sucrose 150 pellet was delivered per lever press, for a maximum of 60 presses. This procedure allowed us to 151 assess increases in drug-seeking when doses are cut in half to maintain the desired level of drug 152 intake.

153
Importantly, rats were trained on the Delay/Size task before cocaine self-administration 154 occurred. No behavioral or recording data were collected during cocaine exposure, or during the 155 1-month withdrawal period. Self-administration procedures expose rats to cocaine prior to 156 recording during the Delay/Size task in order to determine how drug exposure impacts brain and 157 behavior in the long term. These procedures are consistent with work establishing that 158 continuous access to high cocaine doses evokes drug-taking and drug-seeking behaviors that are  Single-unit recording 164 Procedures were the same as described previously (Bryden and Roesch, 2015). Wires 165 were screened for activity each day; if no activity was detected, the rat was removed and the 166 electrode assembly was advanced 40 or 80µm. Otherwise, a session was conducted, and the 167 electrode was advanced at the end of the session. Neural activity was recorded using four 168 identical Plexon Multichannel Acquisition Processor systems (Dallas, TX). Signals from 169 electrode wires were amplified 20x by an op-amp headstage located on the electrode array.

170
Immediately outside the training chamber, the signals were passed through a differential pre-171 amplifier (Plexon Inc, PBX2/16sp-r-G50/16fp-G50) where single unit signals were amplified 172 50x and filtered at 150-9000Hz. The single unit signals were then sent to the Multichannel 173 Acquisition Processor box, where they were further filtered at 250-8000Hz, digitized at 40kHz 174 and amplified at 1-32x. Waveforms (>2.5:1 signal-to-noise) were extracted from active channels 175 and recorded to the disk by an associated workstation with event timestamps from the behavior 176 computer.   was no overlap between epochs, even during short delay (500ms) conditions. To analyze neural 203 activity during these epochs, we normalized firing rates so that high and low-firing cells could be 204 analyzed together. We determined significant changes in firing rate by taking difference scores  Cressyl violet stain to confirm electrode placement (Fig. 1D).    Finally, cocaine-exposed rats showed a stronger response bias for high vs low value rewards 237 (Fig. 2B, inset; significant interaction between group and value F(1,5841) = 4.7, p = 0.030, 238 ANOVA; t(543) = 4.856, p < 0.001, unpaired t-test). Thus, cocaine-exposed rats exhibited an 239 exaggerated response bias toward high-valued rewards, as well as faster reaction times. we found that cocaine-exposed rats were overall more accurate on forced-choice trials (Fig.2D).

251
There were significant main effects of block and value, with rats being more accurate during size  Cue-related value-encoding was attenuated after cocaine exposure 259 We first asked if cocaine exposure altered the counts of insula neurons that were 260 responsive during odor sampling (odor epoch: 100 ms after odor onset to port exit) compared to 261 baseline (1s before odor onset; p < 0.05, Wilcoxon). In controls, 28% (n = 88, out of total 315 262 cells) of neurons were responsive during odor sampling (Fig. 3A), whereas only 12% (n =51, out 263 of total 416 cells) were responsive in rats that self-administered cocaine (Fig. 3B). Both counts 264 were significantly greater than expected by chance alone (control: χ 2 = 617.371, p < 0.001, chi-265 square; cocaine: χ 2 = 135.51, p < 0.001, chi-square), and the frequency of cells that increased 266 firing in controls was significantly higher compared to cells from cocaine-exposed rats (χ 2 = 267 18.395, p < 0.001, chi-square). Thus, rats that had self-administered cocaine had fewer cue-268 responsive neurons compared to controls. An example of a cue-responsive neuron is illustrated in 269 Figure 3C. This neuron fired more strongly for odor cues that predicted higher value reward for 270 behavioral responses to be made into the cell's response field (i.e., the cell's preferred response 271 direction, or direction that elicited the strongest firing; left panels in Fig. 3C).

272
To quantify value selectivity during the odor epoch, we computed the difference between  (Fig. 3A vs B), and only slightly attenuating outcome selectivity in those that 286 remained (Fig. 3D, E vs Fig. 3F, G).  Figure 4A and B.

297
Many neurons fired in anticipation of reward across long delays, exhibiting sustained firing from 298 well-entry until reward delivery (Fig. 4A). However, other neurons seemed to fail in representing 299 reward across a long delay (Fig. 4B). These cells exhibited increases in firing after well-entry, at 300 the time when reward would have been delivered on the majority of trials (i.e., after 500 ms), 301 followed by a decrease in firing during the remainder of the delay. To determine response 302 patterns across the entire population of reward responsive neurons, we plotted average firing 303 aligned to both well-entry (Fig. 4E,F) and reward delivery (Fig. 4G,H) for actions made into the 304 response field.

305
Aligning activity to well-entry, we found that insula neurons from control rats exhibited 306 sustained elevated firing during long delays, starting at well-entry ( Fig. 4E; red). In controls, 307 anticipatory firing after well-entry for longer delays (red) rose less rapidly than for rewards 308 delivered after a short delay (blue dashed). This was not true after cocaine exposure (Fig. 4F). 309 Instead, early activity during long delays tracked firing as if on a short delay trial, peaking 310 around the time when the more immediate reward would have been delivered before dropping to 311 sustained levels. Thus, qualitatively it appeared that firing after cocaine exposure held on to the 312 expectation that reward would be delivered after a shorter delay (which occurs on 3 (short; big; 313 small) of the 4 trial-types).

314
To quantify this observation, we computed the difference in activity during the first and 315 last 500ms of the delay period on long-delay trials (Late -Early / Late + Early; early = first 316 500ms after well-entry; late = last 500ms before reward). These two time points encompass non-317 overlapping activity early and late in the delay period, when rats stay in the fluid well before 318 reward delivery. In controls, we observed no significant shift in the distribution, demonstrating 319 that both early and late expectancy-related firing were similarly represented across longer delays 320 (p = 0.194, µ = 0.034, Wilcoxon, Fig. 4I). However, for rats that self-administered cocaine, this that cocaine-exposed rats had a higher frequency of neurons firing more strongly at the start 324 compared to the end of long delays. 325 These results indicate that in cocaine-exposed rats, neural correlates of immediate reward 326 expectancy were maintained during performance of long-delay trials. These findings also suggest 327 that cocaine exposure impairs the ability of insula neurons to maintain expectancy-related firing 328 from early to late in delay trials. Such a lack of firing might lead to a reduced or discounted 329 neural representation of delayed rewards compared to those delivered after 500ms (i.e., short-330 delay trials). Overall, this result suggests that immediate rewards are better represented in insula 331 after cocaine exposure. Indeed, examining average firing aligned to reward delivery (Fig. 4G,H) 332 suggests that anticipatory firing for immediate rewards (i.e., short-delay trials; blue) was stronger 333 compared to firing for delayed rewards (Fig. 4H, red). To quantify this effect, we computed the  Context-related value signals were attenuated by cocaine exposure 344 Upon constructing the population histograms for the analysis above (Fig. 4G,H), we 345 unexpectedly found that neurons in insula appeared to encode block context in control rats.

346
Specifically, firing was higher after reward delivery during size blocks (green and orange vs blue 347 and red). Remarkably this was true even for 'small' rewards (orange) that were physically the 348 same delay and size as reward on 'short' delay trials (blue).

349
To quantify this effect, we created a 'context' index comparing average activity from size 350 and delay blocks (Size Block -Delay Block / Size Block + Delay Block) 1-2s after reward that 351 indexed the global value differences between blocks (Fig.4M,N). This analysis epoch occurs  Fig. 4M). Interestingly, this effect was not observed in cocaine-exposed rats. Instead, the context 357 index distribution was significantly shifted below zero, indicating that cells tended to fire more 358 strongly for delay blocks compared to size blocks (p = 0.001, µ = -0.022; Wilcoxon, Fig. 4N. 359 Cocaine vs control: z = -4.287, p < 0.001, Wilcoxon). 360 Remarkably, we also observed that neurons that ramped up firing in anticipation of house 361 light onset (i.e., the stimulus that signaled the start of each trial) fired more strongly during 'size' 362 compared to 'delay' trial blocks. This finding is illustrated in Figure 5A, which plots the average 363 firing of 69 control neurons (22% of total). To quantify this effect, we again computed the 364 'context' index during the 4s prior to light onset (the period of time immediately preceding the 365 onset of the trial; Fig. 5B). Consistent with observations in the population histogram, we found a 366 significant positive shift in firing rate distributions, demonstrating that insula neurons tended to 367 fire more strongly during size compared to delay blocks prior to trial onset (p < 0.001, µ = 0.062, 368 Wilcoxon). 369 Interestingly, cells that carried this signal were nearly nonexistent in rats that had self-370 administered cocaine (Fig. 5C). Only 7% (n = 28) of neurons exhibited such pre-trial ramping

384
In control rats, there was a significant positive relationship between size vs delay block 385 accuracy and context-encoding (p < 0.001, r 2 = 0.095, Regression). Thus, increases in firing 386 were correlated with better performance. Interestingly, after cocaine exposure there was no 387 correlation between firing and percent correct (p = 0.808, r 2 < 0.001, Regression, Fig. 5E).

388
During free-choice trials there was also a significant, positive relationship between firing rate 389 and reaction time in control rats (p = 0.032, r 2 = 0.032, Regression). This result suggests that 390 stronger firing rates were actually related to slower reaction times during size blocks. This 391 relationship was also disrupted by cocaine exposure (p = 0.991, r 2 < 0.001, Regression, Fig. 5F). 392 Taken together, these novel firing patterns in insula are related to accuracy and reaction time 393 during our Delay/Size task. These relationships are additionally disrupted by cocaine exposure.

407
Our work demonstrates signals in insula related to cues, context, and differently-valued rewards, 408 and how chronic cocaine profoundly affects insula activity in reward-guided decision-making.  Specifically, our study found that the insula encoded predicted outcomes during 414 presentation of reward-predictive odor cues, and fired in anticipation of reward after completion 415 of an instrumental response. Further, we showed firing was higher for cues that predicted high- Interestingly, we also saw that average firing was stronger during size blocks, and 425 correlated with changes in reaction time and accuracy. When designing our task, our goal was to 426 completely counter-balance value manipulation and direction within each session. We 427 accomplished this objective by running size blocks last, countering satiation by promising rats 428 more and faster access to reward later in the task. Indeed, rats exhibited faster reaction times and 429 better percent correct scores during size blocks due to the increased benefit of completing trials 430 (i.e., no long delays and potential for a larger reward). These distinct differences in behavior 431 indicate changes in overall value between delay and size blocks, which may trigger increases in 432 contextual firing observed in the present study. This interpretation is further supported when 433 considering that higher firing during size blocks reflects more than just receiving a larger reward 434 within trials, as activity was stronger before trial initiation, and after delivery of 'small' rewards 435 that were physically identical to 'short-delay' rewards previously experienced during delay 436 blocks (i.e., both trial types delivered 1 bolus after 500ms delay). However, presenting these size 437 blocks last raises the question of whether or not this context encoding may be a reflection of 438 satiety, or a recency-weighted response to blocks in the second half of the task. Directly testing 439 this issue has proven difficult in the past, as we have found that rats are highly unmotivated to 440 work for delayed rewards when presented with size blocks first. While beyond the scope of the 441 current study, readdressing this questionperhaps through shorter recording sessions with 442 reversed delay and size blocksmay add interesting insight to this data.

443
It may therefore be more accurate to hypothesize that these increases in activity track  We also determined how these correlates are disrupted after chronic cocaine self- representing the anticipation that reward should have been delivered sooner. Consistent with this 464 hypothesis, not only did insula neurons fire more strongly in anticipation of reward on short-465 delay trials, but they also fired strongly at the start of long-delay trials at the time when reward 466 would have been delivered on trials with shorter delays (i.e., 500 ms after well entry).

467
Remarkably, insula neurons in cocaine-exposed rats also failed to increase firing at the 468 end of recording sessions during size trial blocks. Above, we suggest this signal is important for 469 motivating satiated animals when they encounter global increases in reward value in the context 470 of size blocks. The explanation for its absence in cocaine-exposed rats may be that the signals 471 (related to satiety, motivation or absolute value) necessary to drive behavior during size blocks 472 are disrupted after cocaine exposure. Alternatively, these size-encoding signals may be absent 473 because they are not necessary to drive motivation in cocaine-exposed rats, as seen by their faster 474 reaction times and higher accuracy in size blocks. However, the absence of these global signals 475 may in turn disrupt processing of long vs short-delay rewards and subsequently bias behavior for 476 immediate rewards.

477
Relatedly, we found cocaine exposure disrupted relationships between context encoding 478 and behavior-specifically, greater accuracy and slower reaction time were both associated with 479 higher rates of context encoding. The implications of these relationships appear counterintuitive 480 at face value, especially given that control and cocaine-exposed rats exhibited faster reaction   Finally, we must note that insula spans a large portion of the brain and -across humans,  Thus, our findings demonstrate that cocaine exposure enhances signals related to reward 504 delivery, while attenuating signals related to cue and context. Our novel context encoding results 505 provide physiological evidence of the insula's role in interoception and decision-making (Naqvi 506 and Bechara, 2010). Moreover, disruptions of these signals and related changes in behavior due 507 to cocaine exposure suggest an important role for insula in impaired decision-making observed 508 after drug abuse. Further study of insula using different behavioral assays will be important for 509 understanding how this region is involved in both optimal decision-making, and its drug-induced   Block 4. Each block shift was unsignaled so that rats needed to learn through behavior that 750 reward values had changed. C. Average lever-press rates for Control (black) and Cocaine-751 exposed (gray) rats for each day of self-administration. Data points represent each day. Bars on 752 the cocaine-exposed group represent standard error of the mean (SEM). D. Electrode placement 753 for each rat, verified by histology. neurons with firing that significantly differed between trial types (p < 0.05, Wilcoxon). C, D. 808 Same as A and B for cocaine-exposed rats (n = 28). E, F. Scatter plots of context index (size 809 blockdelay block/ size block + delay block) distribution data from neurons that increased