Aging alters the role of basolateral amygdala in intertemporal choice

Aging is associated with an increased ability to delay gratification. Moreover, even when matched for performance, young and aged subjects recruit distinct brain circuitry to complete complex cognitive tasks. Experiments herein used an optogenetic approach to test whether altered recruitment of the basolateral amygdala (BLA), a brain region implicated in valuation of reward-cost contingencies, contributes to age-dependent changes in intertemporal decision making. BLA inactivation while rats deliberated prior to choices between options that yielded either small, immediate or large, delayed rewards rendered both young and aged rats less impulsive. In contrast, BLA inactivation after choices were made (during evaluation of choice outcomes) rendered young rats more impulsive but had no effect in aged rats. These data define multiple, temporally-discrete roles for BLA circuits in intertemporal decision making and implicate altered recruitment of BLA in the elevated preference for delayed rewards that is characteristic of advanced age. Impact Statement Basolateral amygdala (BLA) performs multiple, temporally-discrete functions during intertemporal decision making. Differential engagement of BLA contributes to the enhanced ability to delay gratification that is characteristic of advanced ages.


51
Intertemporal choice refers to decisions between rewards that differ with respect to both their magnitude 52 and how far in the future they will arrive. Biases in intertemporal choice, whether manifesting as extreme 53 impulsivity or patience, strongly associate with psychiatric disease. For example, greater preference for smaller, 54 immediate rewards (greater impulsive choice) is a hallmark of attention deficit hyperactivity disorder and 55 substance use disorders (Bickel et  pattern of choice behavior is sometimes characterized as "wisdom", increased preference for delayed over 64 immediate rewards may also be maladaptive. For example, biases toward delayed gratification in older adults 65 could contribute to inappropriately conservative financial strategies that forgo expenditures necessary to 66 maintain quality of life. 67 The neural circuits underlying age-associated changes in intertemporal choice remain poorly understood. 68 Relevant to elucidating this circuitry is the fact that intertemporal choice is a multicomponent process that 69 involves a series of temporally distinct cognitive operations (Rangel et al., 2008;Fobbs and Mizumori, 2017). 70 Specifically, most decisions begin with representations of past choices, as well as some idea of the outcomes 71 associated with each choice option. These representations are weighted by one's motivation to obtain the choice 72 outcomes at the time of the decision. A second phase of decision making occurs after a choice is made and 73 involves evaluating the outcome to determine the degree to which it matches its predicted value. Feedback from 74 this evaluative process can be used to adjust representations of the options to guide future choices. Both 75 deliberation before a choice and outcome evaluation after a choice are supported by a network of brain structures 76 that mediate reward processing, prospection, planning, prediction error, and value computations (Peters and 77 Behavioral Testing Procedures 155 Apparatus. Testing was conducted in 4 identical standard rat behavioral test chambers (Coulbourn Instruments) 156 with metal front and back walls, transparent Plexiglas side walls, and a floor composed of steel rods (0.4 cm in 157 diameter) spaced 1.1 cm apart. Each test chamber was housed in a sound-attenuating cubicle and was equipped 158 with a custom food pellet delivery trough fitted with a photobeam head entry detector (TAMIC Instruments) 159 located 2 cm above the floor and extending 3 cm into the chamber in the center of the front wall. A nosepoke 160 hole equipped with a 1.12 W lamp for illumination was located directly above the food trough. Instruments) was used to control the behavioral apparatus and laser delivery, and to collect data. 170 Behavioral shaping and initial training. The intertemporal choice task was based on a design by Evenden 171 and Ryan (1996) and was used previously to demonstrate age-related alterations in decision making in both 172 lever press to initiate delivery of a food pellet into the food trough and were then trained to nosepoke to initiate 174 lever extension. Each nosepoke initiated extension of either the left or right lever (randomized across pairs of 175 trials), a press on which yielded a single food pellet. After two consecutive days of reaching criterion performance 176 (45 presses on each lever), rats began testing on the intertemporal choice task. 177 Intertemporal choice task. Each 60-min session consisted of 3 blocks of 20 trials each. The trials were 60 s in 178 duration and began with a 10 s illumination of both the nosepoke port and house light. A nosepoke into the port 179 during this time extinguished the nosepoke light and triggered lever extension. Any trials on which rats failed to 180 nosepoke during this 10 s window were scored as omissions. Each 20-trial block began with 2 forced choice 181 trials, in which either the right or left lever was extended, in order to remind rats of the delay contingencies in 182 effect for that block. These forced choice trials were followed by 18 free choice trials, in which both levers were 183 extended. For all trials, one lever (either left or right, counterbalanced across age groups) was always associated 184 with immediate delivery of one food pellet (the small reward), and the other lever was associated with 3 food 185 pellets (the large reward) delivered after a variable delay. Lever assignment (small or large reward) remained 186 constant throughout testing. Within a session, the duration of the delay preceding large reward delivery increased 187 across the three blocks of trials. The actual delay durations were adjusted individually for each rat on a daily 188 basis, such that the percent choice of the large reward for each rat corresponded to approximately 100% in block 189 1, 66% in block 2, and 33% in block 3. On all trials, rats were given 10 s to press a lever, after which the levers 190 were retracted, and food was delivered into the food through. If rats failed to press a lever within 10 s, the levers 191 were retracted, lights were extinguished, and the trial was scored as an omission. An inter-trial interval (ITI) 192 followed either food delivery or an omitted trial, after which the next trial began. 193 Rats were initially trained for 15 sessions on the intertemporal choice task. They were then lightly 194 anesthetized and optic fibers (Thor Labs) were inserted into the guide cannulae such that they extended 1 mm 195 past the end of the guide cannulae, and then were cemented in place. After recovery, rats resumed training but 196 were now tethered to the rotary joint. 197 Effects of optogenetic inhibition during specific task epochs. The effects of temporally-discrete optogenetic 198 inhibition of BLA were tested in both young and aged rats using a within-subjects design. Data from sessions 199 occurring just prior to inactivation sessions (in which rats did not receive light delivery) were used as the baseline 200 against which to compare the effects of BLA inhibition. Task  dispensed and continued for 4 s]. Finally, the that the order in which the BLA was inactivated during different 208 task epochs was randomized and counterbalanced across the two age groups. 209

Vector Expression and Cannula Placement Histology 210
After completion of behavioral testing, rats were administered a lethal dose of Euthasol (sodium placements and mCherry expression were mapped using a rat brain atlas (Paxinos and Watson, 2005). 225 Decisions regarding inclusion/exclusion of rats based on cannula placements and mCherry expression within 226 the BLA were conducted by an observer for whom rats' behavioral performance was masked. 227 228

Experimental Design and Statistical Analysis 229
Evaluation of age differences in intertemporal choice under baseline conditions. Raw data files were 230 extracted using a Graphic State 4.0 analysis template that was custom-designed to extract the number of lever 231 presses on each lever (large or small rewards) during forced and free choice trials in each trial block. First, age 232 differences in intertemporal choice performance were tested by analyzing the actual delays used to achieve the 233 target 100%, 66% and 33% choice of the large reward in blocks 1, 2 and 3, respectively. Actual delays were 234 compared using a mixed-factor ANOVA, with age (2 levels: young and aged) as the between-subjects factor and 235 block (3 levels: blocks 1-3) as the within-subjects factor. Second, the choice indifference point, or the delay at 236 which a rat showed equivalent choice of the small and large reward, was calculated and compared between 237 young and aged rats. Choice indifference points were calculated by fitting a trendline to each rat's percent choice 238 of the large reward at each delay block. The slope-intercept formula, y=mx+b (where "y" is percent choice or the 239 large reward, and "x" is delay), was then used to solve for the number of seconds (x) at which y=50% choice of 240 the large reward (the delay at which the rat was equally likely to choose the large or small reward). Choice 241 indifference points were compared between young and aged rats using an independent samples t-test. For all 242 statistical analyses and reported results, alpha was set to 0.05, η 2 and Cohen's d were used to report the effect 243 size for mixed-factor ANOVAs and t-tests, respectively, and 1-β was used to report the observed power. 244

Evaluation of BLA inactivation on intertemporal choice. Power analyses based on data from an initial cohort 245
of rats (n=3) were used to determine sample sizes necessary to evaluate the effects of BLA inactivation on 246 choice behavior. These analyses indicated the presence of large effect sizes (greater than 1.0), and that n=6 247 rats should be sufficient to detect effects of BLA inactivation, with a power to detect significant differences of 248 0.95. The effects of light delivery were tested separately for each task epoch (deliberation, small reward delivery, 249 large reward delivery, delay, delay + large reward delivery, and ITI). For each of these task epochs, comparisons 250 were made using a mixed factor ANOVA (age  delay block  inactivation condition), with age as the between-251 subjects factor (2 levels: young and aged), and delay block (3 levels: delay blocks 1-3) and inactivation condition 252 (2 levels: laser on or off) as within-subjects factors. To better understand significant main effects or interactions, 253 post hoc analyses were conducted in each age group separately using a repeated-measures ANOVA (block  254 inactivation condition). Note that for those epochs in which effects of BLA inhibition during the delay were tested, 255 data analyses were confined to blocks 2 and 3 as no delay was used in block 1. 256

Evaluation of choice strategy resulting from BLA inactivation. Additional analyses were conducted to better 257
elucidate the shifts in young rat choice performance following BLA inactivation during deliberation and small 258 reward outcome. Graphic State 4.0 templates were created to assess trial-by-trial choices during baseline and 259 BLA inactivation sessions for the deliberation and small reward epochs. Trials were categorized based on 260 choices made on the previous trial. For the deliberation epoch, trials were categorized as "small-shift-to-large" 261 or "large-stay-on-large". For the small reward delivery epoch, trials were categorized as "large-shift-to-small" or 262 "small-stay-on-small". The number of trials in each category was divided by the total number of trials in that 263 session and expressed as a percentage. For each task epoch, percentages were compared using paired-264 samples t-tests comparing baseline and inactivation condition. 265 Effects on other task performance measures resulting from BLA inactivation. Other task measures were 266 compared between BLA inactivation and baseline conditions in task epochs in which BLA inactivation produced 267 significant changes in choice behavior. Specifically, on free choice trials, response latency (the time between 268 lever extension and a lever press) was compared. Previous work shows that response latencies can differ for 269 large and small reward levers  and hence analyses were conducted separately for each 270 lever using data from delay block 2, during which rats made roughly equivalent numbers of choices on each 271 reward lever. Response latency and total number of trials completed were compared using a mixed factor 272 ANOVA (age  inactivation condition).

Fiber placement and AAV transduction 285
Expression of mCherry was used to confirm viral transduction in the BLA of rats used in behavioral studies 286 that were injected with either AAV5-CamKIIα-eNpHR3.0-mCherry (AAV-eNpHR3.0, black circles in Figure 3) or 287 AA5-CamKIIα-mCherry alone (AAV-control, white circles in Figure 3). Cannula placements were centered in the 288 BLA, and the brain volumes transduced by AAV-eNpHR3.0 and AAV-control (calculated from the atlas of Paxinos 289 & Watson, 2005) were comparable in young and aged rats. 290

Altered choice strategy resulting from BLA inactivation during the deliberation and small reward epochs 335
The data above show that BLA inactivation in young rats during the deliberation and small reward epochs 336 altered choice behavior in opposite directions (i.e., BLA inactivation during deliberation increased whereas BLA 337 inactivation during small reward outcome decreased choice of the large, delayed reward). A trial-by-trial analysis 338 was conducted on these data to determine the effects of BLA inactivation on two distinct behavioral strategies 339 that could mediate these shifts in choice preference. Specifically, during the deliberation epoch, this analysis 340 determined the degree to which BLA inactivation influenced rats to "shift" to the large reward option following a 341 choice of the small reward on the previous trial, versus "stay" with the large reward option following a choice of 342 large reward on the previous trial. In the small reward outcome epoch, the analysis assessed the degree to which 343 BLA inactivation influenced rats to "shift" to the small reward option following a choice of the large reward on a 344 previous trial, versus "stay" with the small reward option following a choice of the small reward on the previous 345

trial. 346
As shown in Figure 5C, the percentage of trials during deliberation epoch inactivation on which a large 347 Other task-specific measures were compared between BLA inactivation and baseline conditions in both 361 deliberation and small reward epochs. The number of trials completed did not differ as a function of laser 362 condition or age in either epoch (see Table 1). Similarly, no differences in response latency were observed as a 363 function BLA inactivation, age, or lever type in these epochs (See Table 2). 364

Effects of BLA inactivation during epochs associated with the large reward outcome 365
Choosing the large reward lever resulted in a variable delay period that was followed by large (3 food 366 pellets) reward delivery. The effects of BLA inactivation during the delay and large reward delivery epochs were 367 initially tested in separate sessions (n=6 young and n=6 aged). Subsequently, the effects of BLA inactivation 368 across both the delay and large reward epochs were tested in a subset of these rats (n=3 young and n=3 aged). 369

Effects of BLA inactivation during the delay epoch. The effects of BLA inactivation during the delay epoch 370
were tested in delay blocks 2 and 3 using a three-factor ANOVA (laser condition  age  block). As expected, 371 there was a main effect of delay block (F (2,20) =146.811, p<0.001, η 2 =0.936, 1-β=1.000) such that both young and 372 aged rats decreased their choice of the large reward as the delay prior to the large reward increased (Figure  373 6A). Compared to baseline, however, no reliable differences in choice behavior resulted from BLA inactivation 374 during the delay epoch (F (1,10

Effects of BLA inactivation during both delay and large reward epochs. One possible explanation for the 391
null effects of BLA inactivation during either the delay or large reward epochs is that, given the role of the BLA 392 in integration of rewards and costs, inactivation may only be effective when conducted during both of these 393 epochs. To evaluate this possibility, rats were tested while the BLA was inactivated during both the delay and 394 large reward epochs. Continuous inactivation across both epochs yielded no effects on choice performance. As 395 shown in Figure 6C, a three-factor ANOVA (laser condition  age  delay block) revealed the expected main 396 effect of delay block (F (2,8) =193.743, p<0.001, η 2 =0.980, 1-β=1.000) but no main effects or interactions with laser 397 condition or age (main effect of laser condition: F (1,4) =0. 757 Figure 7). 411

Effects of light delivery into BLA in rats with control virus (AAV5-CamkIIα-mCherry) 412
To control for non-specific effects of light delivery (e.g., changes in tissue temperature), the effects of 413 light delivery in rats transduced with a control virus that did not contain the eNpHR3.0 gene were tested during 414 behavioral epochs in which BLA inactivation influenced choice behavior [i.e, deliberation (n=4 young and n=4 415 aged rats) and small reward (n=4 young rats)]. 416

Effects of light delivery during the deliberation epoch. Light delivery during the deliberation epoch in rats 417
transduced with a control virus had no effects on choice performance ( Figure 8A). A three factor ANOVA (laser 418 condition  age  delay block) indicated the expected main effect of delay block (F (2,12) =100.272; p<0.001, 419 η 2 =0.944, 1-β=1.000) but no main effects or interactions involving laser condition or age (main effect of laser 420 condition: F (1,6) =0.128; p=0.733, η 2 =0.021, 1-β=0.061; main effect of age: F (1,6) =0.055; p=0.823, η 2 =0.009, 1-421 β=0.055; laser condition  age: F (1,6) =0.028; p=0.874, η 2 =0.005, 1-β=0.052; laser condition  delay block: 422 F (2,12) =0.121; p=0.887, η 2 =0.020, 1-β=0.065; age  delay block: F (2,12) =0.105; p=0.902; , η 2 =0.017, 1-β=0.063 423 laser condition  age  delay block: F (2,12) =0.434; p=0.658, η 2 =0.067, 1-β=0.105). relative to young rats, aged rats display greater preference for large, delayed over small, immediate rewards in 436 a "fixed delays, block design" intertemporal choice task. This difference is not readily attributable to age-related 437 deficits in cognitive flexibility, working memory, or food motivation, nor is it attributable to impairments in reward 438 or temporal discrimination (Simon et al., 2010;Hernandez et al., 2017). The present study replicated these prior 439 findings using a task variant in which the fixed delays/block design used in our previous work was maintained, 440 but the delays to large reward delivery were adjusted on an individual basis to obtain equivalent levels of choice 441 preference in young and aged rats. Under these conditions, aged rats required longer delays to achieve levels 442 of choice preference comparable to young, suggesting that delays are less effective at discounting reward value 443 in aged compared to young rats. These data are consistent with findings in human subjects (Green et al., 1994(Green et al., , 444 1999Jimura et al., 2011;Eppinger et al., 2012) and indicate that an enhanced ability to delay gratification is a 445 consistent feature of aging across species. 446 Data from the current study leveraged optogenetic approaches in young and aged rats to elucidate the 447 contributions of BLA to intertemporal choice in young adult rats and to age-associated changes in this aspect of 448 decision making. Temporally-discrete inactivation revealed distinct roles for BLA in intertemporal choice during 449 the periods immediately before and after a choice was made. Specifically, BLA inactivation during the period 450 prior to a choice (the deliberation epoch) increased choice of large, delayed over small, immediate rewards in 451 both young and aged rats. In contrast, BLA inactivation during receipt of the small, immediate reward (after the 452 choice) decreased choice of the large, delayed reward in young rats, whereas the same inactivation in aged rats 453 had no effect on choice behavior. It should be noted that these results were not likely due to non-specific effects 454 on behavior. Light delivery in rats expressing the control vector did not affect choice behavior, demonstrating 455 that effects of BLA inactivation are due specifically to optogenetic rather than non-specific mechanisms. The fact 456 that inactivation during the delay, large reward delivery, and ITI epochs had no effect on choice performance 457 lends further specificity to the effects of BLA inactivation. Moreover, there were no effects of BLA inactivation 458 during either the deliberation or small reward epochs on the number of trials completed or response latencies. 459 Finally, effects on choice performance were not driven by alterations in reward magnitude discrimination, as BLA 460 inactivation during either the deliberation or small reward epochs did not alter choice performance under no 461 delay conditions (Block 1). The fact that effects were observed only under conditions in which a cost 462 accompanied the large reward (Blocks 2 and 3) suggests a role for the BLA in assigning value to rewards based 463 on cost parameters. Considered together, these data indicate that BLA activity is engaged in different roles at 464 distinct points in the decision process, and that these roles change across the lifespan. analysis. This pattern of results, which is similar to that induced by less temporally-specific approaches to 470 inhibiting BLA activity, suggests that BLA inactivation causes a failure to acquire or integrate information about 471 the negative properties of the small reward (i.e., that it is smaller than the large reward), rendering rats more The finding that BLA inactivation during the small reward epoch (which decreased choice of large, 480 delayed rewards in young rats) had no effect in aged rats indicates that aging is associated with a reduced role 481 for BLA in reward outcome evaluation. Importantly, this lack of effect in aged rats was likely not due to age-482 related impairments in viral transduction or optogenetic efficacy, as BLA inactivation during the deliberation 483 epoch in these same aged rats produced effects on behavior that were as robust as in young rats. In addition, 484 because delays were adjusted to equate baseline performance in young and aged rats, the lack of effect in aged 485 rats cannot be attributed to insufficient parametric space. Instead, the body of data in the current study support 486 the idea of multiple BLA circuits that play unique roles in intertemporal decision making, and indicate that these 487 circuits are differentially recruited in aged subjects. BLA neurons respond differentially to reward anticipation and 488 specific outcomes (Schoenbaum et al., 1998;Belova et al., 2007Belova et al., , 2008 outcome evaluation are not engaged in aged subjects thus may "unmask" the influence of a putative "BLA-521 deliberation" circuit on choice behavior. This pattern of results observed following BLA inactivation during 522 deliberation in intertemporal choice is similar to those in a prior study from our labs which showed that BLA 523 inactivation during the deliberation epoch in a risky decision making task decreased preference for risky rewards 524 . One interpretation of the finding that BLA activity during deliberation contributes to both 525 impulsive and risky choices is that this activity is important for incentive motivation, driving individuals to seek 526 more immediate rewards during intertemporal choices and larger, more salient rewards despite potential 527 punishment during risky choices. This interpretation agrees with evidence from other behavioral contexts. For 528 example, an intact BLA is necessary for the potentiating influence of reward-predictive cues on instrumental 529 responding for reward (Everitt et al., 2003), as well as for maintenance of effortful choices of preferred options 530 (Hart and Izquierdo, 2017). In addition, the trial-by-trial analysis of the current data indicates that BLA inactivation 531 during deliberation increases the frequency with which rats shift from choosing the small, immediate to the large, 532 delayed reward. 533 In addition to its projections to the NAc described above, the BLA projects to many other output structures 534 is reduced in aged rats (Roesch et al., 2012). These data, together with those in the current study, suggest that 548 BLA circuits that normally encode the incentive value of rewards may be hypoactive in aging. In combination 549 with a failure to engage BLA circuits during outcome evaluation, these effects of aging on BLA neurons may 550 contribute to the attenuated impulsive choice observed in aging. Experiments focusing on specific molecular 551 mechanisms underlying age differences in neural activity within discrete BLA circuits will be useful for elucidating 552 the neural substrates that account for the increased ability of aged subjects to delay gratification. An increased 553 appreciation of such mechanisms within the context of the nuanced roles of BLA across multiple stages of the 554 decision process could reveal therapeutic targets for optimizing decision making in both older and younger 555 adults.  intertemporal choice task illustrating the choices and trial blocks across which the duration of the delay to the 560 large reward increased. On each trial, rats were presented with two response levers that differed with respect to 561 the magnitude and timing of associated reward delivery. Presses on one lever delivered a small (one food pellet), 562 immediate reward, whereas presses on the other lever delivered a large (three food pellets), delayed reward. 563 Trials were presented in a blocked design, such that the delay to the large reward increased across successive  open circles) and aged (n=7, closed circles) rats in order to place all rats in the same parametric space. B: Mean 587 actual delays required to achieve the comparable young and aged rat choice performance shown in panel A. 588 Aged rats required longer delays in Blocks 2 and 3 to achieve choice performance comparable to young rats. C: 589 The mean indifference point (the delay at which rats showed equivalent preference for the small and large 590 rewards) was significantly greater in aged rats compared to young. In all panels, error bars represent the 591 standard error of the mean (SEM). *p<0.05, main effect of age;  p<0.01, age  delay block interaction.