Ventral motor thalamic input to prelimbic cortex mediates cost-benefit decision-making in rats

Corticostriatal neurons in prelimbic cortex contribute to decisions that require a trade-off between cost and benefit. The ventral motor thalamus sends dense projections to many cortical areas, including the prelimbic cortex. We investigated whether this input from the ventral motor thalamus to prelimbic cortex contributes to cost-benefit decision-making. Optogenetic inhibition of ventral motor thalamic axon terminals in prelimbic cortex biased rats towards a high cost-high benefit option and, in anesthetized rats, decreased neuronal activity in deep layers of prelimbic cortex. Stimulation of ventral motor thalamic nuclei induced a neuronal response in deep layers of prelimbic cortex and simultaneous optogenetic inhibition of layer 1 inhibitory interneurons similarly decreased neuronal activity. Our results indicate that ventral motor thalamic input to prelimbic cortex mediates cost-benefit decision-making. Significance Statement Our results indicate that ventral motor thalamic input to prelimbic cortex plays a critical role in decisions that require a trade-off between two conflicting reward values. Traditionally, ventral motor thalamic nuclei were primarily associated with motor control, but more recently these thalamic nuclei have been implicated in tasks that require animals to choose between two alternatives. Our results highlight the need to reevaluate the role of the ventral motor thalamic nuclei in cognition. Furthermore, prelimbic cortex and, more generally, prefrontal cortex have been associated with chronic stress and major depressive disorder, highlighting the possibility that ventral motor thalamic nuclei might be involved in these disorders.


Introduction 31
Prefrontal cortical areas are involved in a variety of decisions that require a trade-off: anterior 32 cingulate cortex regulates the willingness to expend physical (Walton et al., 2003) or mental 33 effort to receive a larger reward (Hosking et al., 2014); orbitofrontal cortex is necessary in 34 risk-and delay-based decision-making (Mobini et al., 2002); dorsomedial prefrontal 35 corticostriatal neurons encode approach-avoidance behavior (Loewke et al., 2021); and 36 prelimbic corticostriatal neurons mediate the trade-off between a more costly, more 37 beneficial and a less costly, less beneficial option (Friedman et al., 2015). Recent  2021). In rodents, MT sends dense projections to prelimbic cortex (Herkenham, 1979;42 -11-stereotaxic tilted at a 30° angle; Paxinos and Watson, 2004). Experiments were performed 251 12-16 days after the virus was injected and rats were perfused after the experiment. been injected with the archaerhodopsin-expressing virus (ArchT rats) and 12 rats that had 275 -12-been injected with the control virus (controls). Rats were between 4 to 5 weeks of age, when 276 we started habituation to the sweetened condensed milk and behavioral training. Rats were 9 277 weeks of age at the time of surgery and 11 weeks of age, when we started behavioral testing. 278 The required number of animals for behavioral experiments was predicted a priori based on a 279 similar study conducted in the past (Friedman et al., 2015). 280 281 Assignment of rats to groups for behavioral training was pseudo-randomized. For the initial 282 lever pressing training we presented the left lever on the first day and the right lever on the 283 second day for one group of rats, while for a second group of rats the right lever was 284 presented on the first day and the second lever was presented on the second day. For 285 behavioral training and testing for one group of rats the left and for a second group of rats the 286 right lever was associated with the high benefit, high cost and high cost-high benefit option. 287 The lever on the opposite side was associated with the other choice option. Assignment of 288 rats to the ArchT or control groups was also pseudo-randomized. 289 290 Rats had a 12-to 16-day-long recovery period after surgery before their choice behavior on 291 each of the three decision-making tasks was assessed over a period of 9 days. Rats were 292 presented with 40 trials per day over 3 consecutive days on each decision-making task. The 293 first 20 trials were presented without delivery of the 590 nm light (light OFF) and the last 20 294 trials were paired with delivery of the 590 nm light (light ON). In ArchT rats, light delivery 295 induced optogenetic inhibition in MT axon terminals located in prelimbic cortical layer 1. In 296 total, we presented rats with 60 light OFF and 60 light ON trials on each decision-making 297 task. The 590 nm light was turned on at the same time as the tone stimulus and remained on 298 until rats made a choice or until the levers were retracted, i.e. a maximum of 20 secs. 299 300 -13-The implanted LED fiber optics were controlled by wireless receivers. The maximum light 301 intensity that was reached using the wireless receivers and LED fiber optics was between 0.7-302 1.2 mW. Given that LED fiber optics had a diameter 250 μm, the emitted 590 nm light 303 reached approximately 3-6 mW/mm 2 . 304

305
We decided on a block design for presentation of light OFF and ON trials, since a previous 306 study suggested that sustained optogenetic inhibition of axon terminals may cause 307 unexpected long-term effects such as an increase in spontaneous neurotransmitter release 308 (Mahn et al., 2016). Hence, light ON trials were always presented after light OFF trials. All 309 rats were first tested on the benefit-benefit, then on the cost-cost and last on the cost-benefit 310 decision-making task. We presented the decision-making tasks in this specific order, to 311 increase rats' retention of the tasks. A timeline for behavioral testing is provided in Figure 1. Behavioral responses in each session were stored in an automatically generated text 318 document. We extracted any variables that were analyzed from these text documents using 319 custom-written Python 3.7 scripts. The reaction times stored in each text document 320 represented a measurement from the presentation of the tone stimulus marking the beginning 321 of each trial to the time rats indicated their choice by pressing one of the two levers. 322 However, reaction times were corrected to represent times from presentation of the two 323 levers to the time rats indicated their choice. 324 325

Statistical Analyses -Choice Behavior on the Three Decision Making Paradigms 326
To quantify and compare decision-making behavior of rats on the three decision-making 327 tasks, we primarily assessed the percentage of trials, in which rats chose 1) the high benefit 328 option on the benefit-benefit decision-making task, 2) the high cost option on the cost-cost-329 decision-making task or 3) the high cost-high benefit option on the cost-benefit decision-330 making task out of the total number of non-omitted trials. 331

332
In addition, to test whether motor function was disrupted by optogenetic inhibition of MT 333 axon terminals in prelimbic cortical layer 1, we assessed 1) the percentage of omitted trials; 334 and 2) the average reaction time across all non-omitted trials. 335

336
To analyze the effect of the injected virus and of light ON versus OFF, we performed 337 separate ANOVAs for each decision-making task and each variable using the injected virus 338 as between and treatment as within animal variable. When we observed a main or interaction 339 effect, we performed two post-hoc paired t-tests to compare the treatment and non-treatment 340 condition within each group of rats. To confirm that effects were not caused by pre-existing 341 differences in behavior between the two groups of rats, we further performed two Student's t-342 -15-tests to compare behavior on the first 20 trials of the last day of training as well as behavior in 343 the non-treatment condition between the two groups of rats. To confirm that effects were not 344 caused by surgery, we also compared the behavior pre-and post-surgery within each group of 345 rats. We usually performed 6 t-tests to analyze each behavioral variable. We applied a 346 Bonferroni correction to the significance level resulting in a significance level of p=0.0083. ArchT rats, to analyze whether the difference in the percentage of high cost-high benefit trials 359 was significantly different between the two groups of rats. 360

361
We adjusted individual dilutions of sweetened condensed milk for each animal, so that each 362 animal would choose the high cost-high benefit option as well as the low cost-low benefit 363 option in about 50% of non-omitted trials. We used the first 3 days of behavioral training on 364 the cost-benefit decision-making task to adjust individual dilutions. We systematically varied 365 the dilution of sweetened condensed milk on every 20 trials. Hence, data on the percentage of 366 high cost-high benefit choices, which was collected across these 3 days, does not represent 367 -16-the actual choice behavior of rats on days after we determined and used their individual 368 dilutions. Only a few rats did not reach criterion on the cost-benefit decision-making task 369 within the first 3 days, resulting in few rats being trained for additional days. When running 370 statistical tests that involved data from the last day of behavioral training on the cost-benefit 371 decision-making task, we only used the data from the first 20 trials on the last day of 372 behavioral training for these few rats. For all other rats, when the determined individual 373 dilution corresponded to a dilution that was used during the adjustment process, we used data 374 from those 20 trials. Otherwise, when the individual dilution was between dilutions used 375 during the adjustment process, we averaged data from the 20 trials with the next higher and 376 the 20 trials with the next lower dilution. Given the limitations of this approach, we compared 377 a second metric between groups of rats to determine whether pseudo-random assignment of 378 rats to either the control or ArchT group influenced the percentage of high cost-high benefit 379 choices within each group. We analyzed for each group of rats whether the individual 380 dilution of sweetened condensed milk correlated with the percentage of high cost-high 381 benefit choices in light OFF trials during behavioral testing. We assumed that if these two 382 metrics did not correlate, pseudo-random assignment of rats would not have influenced the 383 percentage of high cost-high benefit choices in each group, even if the determined individual 384 dilutions of sweetened condensed milk would have varied vastly. We calculated Kendall's 385 Tau to determine the correlation between the two metrics. We chose Kendall's Tau since 386 individual dilutions of sweetened condensed milk were rounded to the closest whole number 387 and, hence, on an ordinal scale. We chose Kendall's Tau over Spearman's rho due to the small 388 sample size and higher robustness of Kendall's Tau for small sample sizes. 389 390 -17-Statistical analyses were performed in R (version 3.6.3). We tested the assumption of the data 391 being normally distributed, which is a prerequisite for parametric tests, with Levene's test 392 and the assumption of homogeneity of variances with Shapiro's test. 393 394

Experimental Design -In-vivo Electrophysiology 395
We acquired extracellular recordings from deep-layer pyramidal neurons in anesthetized, 396 male Sprague-Dawley rats (Charles River Laboratories, Japan). We stimulated MT and, in 397 four rats, simultaneously inhibited MT axon terminals in prelimbic cortical layer 1 using 398 optogenetics. Surgeries were performed when these rats were between 9 to 11 weeks of age, 399 and data was collected when rats were between 11 to 13 weeks of age. We recorded from a 400 total of 39 cells. In three rats, we stimulated MT and simultaneously inhibited prelimbic layer 401 1 inhibitory interneurons using optogenetics. Surgeries were performed when these rats were 402 between 9 to 10 weeks of age, and data was collected when rats were between 11 to 12 weeks 403 of age. We recorded from a total of 37 cells. To test whether rebound spiking was observed upon turning the 590 nm light off, which 495 deactivated the archaerhodopsin and terminated optogenetic inhibition of MT axon terminals, 496 we extracted segments from light ON trials starting 100 ms before and ending 100 ms after 497 the light was turned off. Data was normalized by subtracting the mean activity across each 498 segment from each data point in the segment. We again identified extracellular spikes by 499 running the Matlab 'findpeaks'-algorithm and using the previously determined value for 500 minimum spike prominence. We constructed a peri-stimulus time histogram from the data 501 using a bin size of 1 ms. In addition, the average number of spikes per second for the 100 ms 502 before and the 100 ms after the light was turned off was calculated and indicated in the peri- Raw behavioral data, raw electrophysiology data and analysis scripts used to generate figure  574 1 through 4 are available on Github (https://github.com/bsieveri/sieveritz-2022-mt-575 prelimbic). Microscopy data will be made available upon request. 576 577

Results 578
Optogenetic inhibition of MT input to prelimbic cortex biased rats towards a high cost-high 579 benefit option 580 We trained twenty-two 4-to 7-week-old male Sprague-Dawley rats on a benefit-benefit, cost-581 cost and cost-benefit decision-making task (Figure 2A; see methods section for details). At 9 582 weeks of age, twenty-two rats that had learned all three decision-making tasks received 583 unilateral virus injections into MT. We injected 50-70 nl of either an adeno-associated virus 584 expressing archaerhodopsin (AAV5-CAG-ArchT-GFP, n=10; ArchT rats) or a control virus 585 (AAV5-CAG-GFP, n=12; controls). In addition, we implanted a short optical fiber attached 586 to a 590 nm light emitting diode (LED fiber optic) through the contralateral hemisphere into 587 ipsilateral prelimbic cortical layer 1 ( Figure 2B). We confirmed that virus injections were 588 primarily confined to MT ( Figure 2C). Tips of LED fiber optics were located in prelimbic 589 cortical layer 1 in close proximity to virus-expressing MT axon terminals ( Figure 2C). 590 591 After a 12-to 16-day-long recovery period, we assessed rats' choice behavior. For each of the 592 three decision-making tasks, rats were presented with 40 trials per day on 3 consecutive days. 593 We compared rats' choice behavior without (light OFF) and with delivery of the 590 nm light 594 -26-(light ON). In ArchT rats, delivery of the 590 nm light induced optogenetic inhibition of 595 virus-expressing MT axon terminals in prelimbic cortical layer 1. On each day, the first 20 596 trials were light OFF trials, while the last 20 trials were light ON trials. Optogenetic 597 inhibition of MT axon terminals in prelimbic cortical layer 1 on the cost-benefit decision-598 making task biased rats towards the high cost-high benefit option ( Figure 2D). In contrast, 599 choice behavior on the benefit-benefit and cost-cost decision-making tasks was not affected 600 ( Figure 2E and 2F). To compare the percentage of high cost-high benefit choices on the cost-601 benefit decision-making task, we performed a mixed-design ANOVA with the injected virus 602 as between animal factor and light ON/OFF as within animal factor. We observed a 603 significant interaction effect between the injected virus and light ON/OFF (p=0.031, Cohen's 604 F=0.517, df=1). and a main effect for light ON/OFF (p=0.005, Cohen's F=0.707, df=1). We 605 performed post-hoc testing using multiple t-tests and applied a Bonferroni correction to 606 account for multiple comparisons (adjusted significance level=0.0028). In light ON trials, we 607 observed a significant increase in the percentage of high cost-high benefit choices for ArchT 608 rats (p=0.00004, r=0.926, df=9), but not for controls (p=0.552, r=0.182, df=11). This 609 indicates that optogenetic inhibition of MT axon terminals in prelimbic layer 1 biased rats' 610 choices towards the high cost-high benefit option ( Figure 2D). To determine whether the 611 increase in the percentage of high cost-high benefit choices from light OFF to light ON trials 612 differed significantly between controls and ArchT rats, we used an uncorrected Student's t-613 test. The increase was significantly larger for ArchT rats than for controls (p=0.026, r=0.536, 614 df=20), further confirming that optogenetic inhibition of MT input to prelimbic cortex biased 615 choices towards the high cost-high benefit option. 616

617
We observed a similar trend when data was split by day (Figure 2 -figure supplement 1). 618 Given that only 20 light ON and OFF trials were administered on each individual day of 619 -27-behavioral testing, it is not surprising that none of the effects reached significance on any 620 individual day. However, we did observe an increase in high-cost high-benefit choices in 621 ArchT rats in light ON trials on the second day of behavioral testing that approached 622 significance (p=0.0037, r=0.791, df=9; adjusted significance level=0.0028).