The anterior insula bidirectionally modulates cost-benefit decision-making on a rodent gambling task

Abstract

Deficits in cost-benefit decision-making, as assessed in the Iowa Gambling Task (IGT), are commonly observed in neuropsychiatric disorders such as addiction. There is considerable variation in the maximization of rewards on such tasks, both in the general population and in rodent models, suggesting individual differences in decision-making may represent a key endophenotype for vulnerability to neuropsychiatric disorders. Increasing evidence suggests that the insular cortex, which is involved in interoception and emotional processes in humans, may be a key neural locus in the control of decision-making processes. However, the extent to which the insula contributes to individual differences in cost-benefit decision-making remains unknown. Using male Sprague Dawley rats, we first assessed individual differences in the performance over the course of a single session on a rodent analogue of the IGT (rGT). Rats were matched for their ability to maximize reward and received bilateral excitotoxic or sham lesions of the anterior insula cortex (AIC). Animals were subsequently challenged on a second rGT session with altered contingencies. Finally, animals were also assessed for instrumental conditioning and reversal learning. AIC lesions produced bidirectional alterations on rGT performance; rats that had performed optimally prior to surgery subsequently showed impairments, and animals that had performed poorly showed improvements in comparison with sham-operated controls. These bidirectional effects were not attributable to alterations in behavioural flexibility or in motivation. These data suggest that the recruitment of the AIC during decision-making may be state-dependent and help guide response selection towards subjectively favourable options.

Keywords: decision-making; gambling task; individual differences; insular cortex.

Figure 1

General experimental design and rGT choice structure. (A) The behavioural procedure before surgery consisted of nine daily sessions (see details in Materials and methods). Animals initially responded in any of the four active holes during free choice sessions to obtain one food pellet. Forced choice sessions subsequently ensured that animals explored every active hole. Two further free choice sessions were then administered, during which animals received two‐sugar pellets for half of each session. Rats were then tested on a single pre‐surgery rGT session from which good (n = 16) and poor (n = 16) decision‐makers were identified according to their performance. Rats underwent either bilateral AI‐ or sham lesion (respectively, n = 9 and n = 7). After recovery, animals were exposed to similar training sessions, but over 4 days and finally a post‐surgery rGT session was conducted. The second rGT was based on the same principle of disadvantageous and advantageous two‐hole options but utilised a new combination of conditions to avoid learning effects. To further explore the effects of AIC lesions, the acquisition of instrumental responding for food reward in a two‐lever operant conditioning chamber under increased FR response requirements (FR1, FR3 and FR5) was measured. Subsequently a PR schedule during which the cost of the reward is progressively increased within a single session measured animals’ motivation for food reward. Lastly, rats’ ability to update or alter their behaviour was measured via five reversal learning sessions according to a FR5 schedule under reversed values of the levers (FR5r). (B) Diagram showing the utility of the four options in each of the two rGT sessions. Rats are offered four options, two of which were associated with small reward (one pellet) and the other two with higher reward (two pellets, represented here pictorially). The higher rewards were associated with higher punishment in the form of time–out periods, the duration of which is shown in seconds along with the relative probability (P). Timeouts were delivered concomitant with reward. The options delivering single pellets were considered advantageous due to the lower potential timeout punishments. Thus, rats had to sample the various options and learn to move away from high reward/high cumulative loss choices to low reward/low cumulative loss choices. Advantageous options were counterbalanced against the animals preferred side during training. During the second rGT, the contingencies were reshuffled and the duration of penalties was altered so animals had to sample each hole again in order to re‐learn the optimal strategy. [Colour figure can be viewed at wileyonlinelibrary.com].

Figure 2

Bilateral AIC lesions impair maximization of reward in rats performing optimally in a rat gambling task. (A) Schematic representation of the extent of the bilateral excitotoxic lesions of the AIC from rats identified as good (n = 16) and poor (n = 16) decision‐makers. Areas shaded in grey represent the extent of neuronal damage. Coronal sections are 4.20 mm anterior through 0.12 mm anterior to the bregma. (B and C) Behavioural performance in the rGT of GDM and PDM rats was not similarly affected by AI lesion. Thus, sham GDM rats, akin to sham PDM, displayed similar performance in rGT prior to, and after surgery. In contrast, GDM lesioned animals showed an impairment in rGT performance as compared to their pre‐lesion test, and PDM lesioned animals exhibited an improvement in their performance, so that the performances of these two groups are similar across the course of the second rGT session. (D) During the exploration phase of the rGT, all good decision‐maker (GDM) rats tended to randomly explore all available options; the decision‐making score defined by the ratio of advantageous choices relative to disadvantageous choices for the same period was around 1 in both lesioned (n = 9) and sham (n = 7) groups before and after surgery. During the last minutes of the test, in this subpopulation the aim of maximizing rewards defined an exploitation phase; it was reflected by a marked preference for advantageous choices. There were around 40 advantageous choices for one disadvantageous choice in lesioned rats before surgery, and in sham rats before and after surgery. But on the other hand, good decision‐makers after AIC lesion showed a ~70% decrease in their decision‐making score. (E) In poor decision‐makers, the exploration phase was also similar through the different groups before and after surgery. Before surgery, in the last minutes of the test, when exploitation of gathered information about options would have been possible, both groups failed to identify the advantageous options as preferable and pursued risky choices; however, lesioned (n = 9) and sham (n = 7) PDM rats increased their exploitation performance by, respectively, 5.5 and 2.5 times during the second rGT despite recombined but matching conditions as compared to the previous one. Data are group means and SEM. Each dot represents an individual. [Colour figure can be viewed at wileyonlinelibrary.com].

Figure 3

Preserved acquisition of instrumental response in all rats and impaired behavioural flexibility only in PDM. (A and C) Neither decision‐making trait nor AIC lesion affected the acquisition of instrumental conditioning. The total number of lever #1 presses, that is current active lever presses, increased in parallel with the increased behavioural requirement from FR1 to FR5 in both lesioned and sham operated good and poor decision‐makers. Lever #2 presses, that is current inactive lever presses, remained low in all groups. (B) After two additional FR5 sessions in which only lever #1 is active, the contingencies were reversed for five sessions such that lever #2 is now active (rFR5). In both lesioned and sham good decision‐makers, after switching, the total number of lever #1 presses decreased in parallel with the increased total number of lever #2 presses. Active lever presses summed for the five FR5 sessions of the acquisition phase and for the five rFR5 sessions of the reversal phase did not differ in these groups. (D) PDM under the same reversal conditions decreased their responses on the now inactive lever #1, but exhibited lower lever presses for the lever #2, that is the new active lever, in comparison to GDM. PDM did not attain the same mean level of responsing as represented in grey following reversal. Their sums of active lever presses through FR5 sessions and through rFR5 sessions differed, reflecting also the impaired reallocation of the instrumental response. Data are group means and SEM. [Colour figure can be viewed at wileyonlinelibrary.com].

Figure 4

Unaltered general motivation after AIC lesion. (A and B) The breakpoint on a PR schedule did not differ between good and poor decision‐makers and was not influenced by AIC lesions. Data are group means and SEM. Each dot represents an individual. [Colour figure can be viewed at wileyonlinelibrary.com].