Replicability in Measures of Attentional Set-Shifting Task Performance Predicting Chronic Heavy Drinking in Rhesus Monkeys

Abstract

This study was designed to replicate and extend a previous report that the increase in performance of an attentional set-shifting task (ASST) in rhesus monkeys predicted their future alcohol drinking status as a heavy drinker (HD) or non-heavy drinker (NHD). A cohort of 6 young adult male monkeys was trained and tested under the same ASST and then underwent a alcohol self-administration protocol that maintained open-access (22 hours/day) choice of alcohol or water 7 days/week for approximately 6 months. The average improvement in performance in the ASST, as measured by a performance index, was replicated in the cohort of 6 monkeys when compared to the increase in the task performance in a previous cohort of 9 male monkeys. The alcohol self-administration protocol was then used to determine the drinking status (HD: n = 4 or NHD: n = 2) of the replicate cohort, which was accurately predicted by the performance on the ASST. Finally, individuals from both cohorts could be combined based on future drinking status of HD (n = 8) or NHD (n = 7), and the association with pre-alcohol ASST performance remained. Specifically, monkeys that had lower rates of PI improvement were more likely to become HDs. To our knowledge, this is the first study to replicate that deficits in the set-shifting performance can predict chronic heavy alcohol drinking in primates.

Introduction

The Attentional Set-Shifting Task (ASST) is a variant of the Wisconsin Card Sorting Task that reflects sensory-motor and associative processes and has been used to assess behavioral flexibility (Brown & Tait, 2016; Miyake et al., 2000; Nyhus & Barceló, 2009). The wide use of this procedure across many species, (including rodents, non-human primates, and humans) provides an opportunity for identifying fundamental mechanisms of control over repetitive behavioral patterns that are integral to many public health problems, such as addiction and obsessive-compulsive disorders (for example, Kim et al., 2017).

A novel ASST was developed to identify individual differences in behavioral flexibility of rhesus monkeys (Shnitko, Allen, Gonzales, Walter, & Grant, 2017). This ASST task is based on a visual discrimination of one of two geometrical objects presented on a touchscreen. Subjects advance at their own pace through a sequence of up to eight sets, with sets differentiated on intra-dimensional, extra-dimensional, and reversal shifts in the correct choice (Shnitko et al., 2017). Uniquely, all monkeys housed in the same room (i.e., a cohort) are presented with the task in their housing cages at the same time. Each session began with a simple discrimination regardless of the performance during the previous session. This approach minimizes differential training across individuals but increases task-extraneous stimuli primarily from viewing and hearing other monkeys engaging in the task. The application of the ASST in rhesus monkeys revealed rates of improvement in the performance over 30 consecutive sessions and predicted future chronic alcohol drinking status (Shnitko, Gonzales, & Grant, 2019; cohort 14). Specifically, low rates of performance improvement (slopes <2.4) were significantly associated with future status as a heavy drinker.

Chronic heavy drinking is defined as >20% of days with 22-hour access to alcohol or water, consuming more than 3.0 g/kg of alcohol per day (Baker, Farro, Gonzales, Helms, & Grant, 2014). This definition was derived using a mathematical modeling approach and applied to the pattern of alcohol drinking in daily 22-hour sessions, over 12 consecutive months of daily sessions from 31 monkeys from 4 cohorts (6–12 (Baker et al., 2014) and confirmed with subsequent cohorts (Baker et al., 2017; Moore et al., 2019). Ultimately, the ability to use the ASST and alcohol self-administration protocols to identify brain circuity underlying the risk for heavy drinking with in vivo brain imaging or genomic measures of brain function requires a large population of subjects characterized under standard conditions. A key aspect of adding subjects is showing the reliability of the measures within and across cohorts over time. Therefore, the object of this study was to assess the reliability of performance on the novel ASST itself and in predicting future heavy drinking in a new cohort of monkeys.

Methods

Animals

Two cohorts of male rhesus macaques were obtained 97 months apart from the Oregon National Primate Research Center (ONPRC) breeding colony, ranging in age from 3.6 to 5.7 years (“cohort 14”, n = 9, and “cohort 16”, n = 6; cohort details at www.matrr.com). Their weight ranged between 4 and 10 kg at the beginning of the study (6.2 ± 0.4 kg in cohort 14 and 9.6 ± 1.2 kg in cohort 16). Monkeys from cohort 14 were the subjects of previous publications (Shnitko et al., 2017; Shnitko, Gonzales and Grant, 2019; Shnitko, Gonzales, Newman, & Grant, 2020) and those in cohort 16 were experimentally naive subjects. The monkeys were housed in individual cages (0.8 × 0.8 × 0.9 m) in a housing room under temperature (20–22 °C) and humidity (65%) control with an 11-hour light:13-hour dark cycle. Monkeys had constant audio-visual-olfactory access to all animals in the room and were daily placed in a social pair with a compatible partner for 1–2 hours/day. They were fed a diet of nutritionally complete 1-g banana-flavored pellets (TestDiet, United States) and were given fresh produce daily. All monkeys were weighed weekly and ethanol intakes were calculated on contemporary weights (g/kg). All procedures were conducted according to the Guide for the Care and Use of Laboratory Animals (Committee for the Guide, 2010) and approved by the ONPRC Animal Care and Use Committee in an AALAC-accredited facility.

Equipment

A panel with a centrally located computer-operated touchscreen was embedded in a side wall of each housing cage as previously described (Grant, Stafford, et al., 2008; Shnitko et al., 2017). The panel included a dowel below the screen, two drinking spouts on the left and right sides of the screen, and an infrared finger-poke located to the right side of the screen (Med Associates Inc., United States) that operated a food pellet dispenser located outside the cage that delivered 1-g banana-flavored food pellets. Each drinking spout was connected to a carboy filled with tap water or 4% ethanol (w/v diluted in water) and located outside the cage on two scales (Ohaus Adventurer, United States). All panel-connected computers were linked to a network and with a main computer controlling the alcohol self-administration and set-shifting program procedures (National Instruments interface and LabView software; LabView 2011, SP1, National Instruments, Texas, United States).

Procedures and experimental design

Set-shifting task

All animals underwent set-shift testing that was followed by the induction to alcohol self-administration and approximately 6 months of free access to 4% ethanol and water. The set-shifting procedure has been described for cohort 14 (Shnitko et al., 2017) and was replicated for the new cohort of 6 monkeys reported here (cohort 16). Briefly, all animals were given the set-shift sessions within their housing cage each morning (from 8:30 to 9:30 AM). Unique to this procedure is the concurrent training and testing of all monkeys in the cohort within their housing environment. Training to respond on the screen was accomplished with minimal experimenter-delivered differential reinforcement (i.e., shaping) using the contingent presentation of a highly preferred photograph (Shnitko et al., 2017). After the monkeys reliably touched the screen, training on the set-shift procedure began. In training, a series of trials with random left/right presentation of two geometrical shapes were presented on the touch screen. A trial ended after a monkey touched one of the shapes or 30 seconds had elapsed. A monkey learned to discriminate which shape was predetermined as ‘correct’ or ‘incorrect’ based on the contingent presentation of a preferred photograph and a banana-flavored pellet or a blank screen for 10 seconds, respectively. The schedule of reinforcement following a correct response was gradually increased to a second order of FR1(FR3), where every correct response resulted in a preferred photo and every third correct response resulted in a pellet (details in Shnitko et al., 2017). The only criteria set for training the second order schedule was that responding was maintained throughout the 45-minute session in all animals. There were seven training sessions for cohort 14 (Shnitko et al., 2017), and the number of training sessions for cohort 16 was limited to seven. The second order schedule was effective in maintaining responding, as every third correct response was reinforced with a food pellet, and this was independent of the criteria for advancing through the discrimination sets.

After training under the second order schedule, the next 30 consecutive sessions were used to assess set-shifting performance. For each session, there was a maximum of four original (simple and compound discriminations, intradimensional and extradimensional shifts) and four reversal sets. A set was a pair of shapes, filled with either of two colors presented side by side, with the side and the color of the shape randomly chosen in each trial of the set (i.e., a black or white circle and a black or white triangle presented either on the left or on the right of center). Each set retained the same two shapes and colors for all trials. Reversal trails again presented the same set of stimuli but with the previous contingency for a correct touch reversed. Criteria for acquiring each set or reversal set was a running 12 correct out of 15 consecutive trials. The next set was the presentation of two new shapes and two new colors once the criterion was met for the reversal trials. Each session began with a pair of black or white shapes (same color for both shapes) displayed on the screen, and the discrimination was based only on stimulus shape (i.e., a simple discrimination). The second set introduced unique colors to two new geometric shapes (i.e., a compound stimulus), although the discrimination remained based on new shapes (i.e., an intradimensional shift). The third set was again a different combination of unique shapes, again with the discrimination based on shape (a second intradimensional shift). The final (fourth) set was again a different combination of unique shapes, but with the discrimination based on color (an extradimensional shift). The progression through the four possible original and reversal sets in the daily sessions was self-paced. The session ended after 45 minutes had elapsed or when a monkey met the criteria for all eight sets (i.e., under 45 minutes).

Alcohol self-administration

Alcohol self-administration was established using a standard protocol that induced set amounts of alcohol intakes (4% w/v) with a schedule-induced polydipsia (SIP) procedure prior to daily ‘open-access’ sessions (22 hours/day) of alcohol or water availability (Baker et al., 2014; Grant, Stafford, et al., 2008). During the daily SIP sessions, 1-gram banana-flavored pellets were delivered under a fixed-time of 300 seconds until the monkey consumed a preset volume of either water or ethanol. Monkeys were first induced to drink water for 30 sessions, and then water was replaced with 4% ethanol and the monkeys were induced to drink a volume equivalent to 0.5 g/kg, 1.0 g/kg, and 1.5 g/kg, over 30 consecutive sessions at each dose. After induction, the open-access phase began and continued for seven days a week and lasted approximately 6 months. The total daily food ration for each monkey was divided equally into three “meals”, with the first meal occurring at the beginning of each open-access session and subsequent meals available at 2-hour intervals. The pellets were contingent on a finger poke under a fixed ratio 1 schedule. The total number of the open-access self-administration sessions was 152 in cohort 14 and 172 in cohort 16. During open access, blood ethanol concentrations (BECs) were taken every 5–7 days, 7 hours after the onset of the session and just before the lights were shut off in the housing room.

Data analysis

The main set-shift dependent variable used in analyses was the performance index (PI) that was derived from three orthogonal variables: the session duration, a ratio of errors to trials, and the maximum ‘set’ that a monkey reached during a session (see Shnitko et al., 2017). All three variables were scaled from 0 to 100 and were summed to give a session PI, and then the cohort average over 30 consecutive sessions was calculated. The PI was first derived in cohort 14 and used to predict future drinking status (Shnitko, Gonzales, et al., 2019) as well as the consequences of drinking alcohol (Shnitko et al., 2020).

The main dependent variable of the alcohol self-administration procedure used in analyses was the daily average ethanol intake (g/kg/day). Individual monkeys were categorized as low, binge, heavy, and very heavy drinkers based on percentage of sessions where intakes were over 4.0 g/kg, 3.0 g/kg, and 2.0 g/kg ethanol during open access as previously described (Baker et al., 2014). In this cohort design, individuals could fall into any of the four categories (due to the voluntary nature of self-administration). Therefore, to investigate group differences in the limited number of monkeys in each cohort (here n = 9 and n = 6), monkeys categorized as heavy and very heavy drinkers were also combined into a “heavy” drinking group (HD: a mean ethanol intake of ≥3 g/kg for more than 20% of drinking sessions), and monkeys categorized as low and binge drinkers were combined into a “non-heavy” drinking (NHD) group (Allen, Gonzales, & Grant, 2018; Cervera-Juanes et al., 2016; Grant, Leng, et al., 2008; Helms, Park, & Grant, 2014; Helms, Rau, et al., 2014). The HD and NHD group assignments were then used for analysis of set-shift performance predicting future heavy (fHD) and future non-heavy (fNHD) drinkers.

A mixed-model linear regression was used to compare slopes and intercepts of set-shift PI over sessions analyzed for cohort effect or for predicting future drinking status (fHD, fNHD) with a monkey as a random factor. An unpaired t test was used to compare ethanol intake and the BECs. All analysis was performed using SPSS, and results were graphed using GraphPad Prizm 9.

Results

Figure 1 shows that the average daily improvement in PI on the set-shift task prior to alcohol self-administration was similar in the replicate cohort of monkeys studied here (cohort 16) to the original cohort (cohort 14; Shnitko et al., 2017). Cohort average PI improved as a factor of sessions in both cohort 14 (slope [β], PI/session) = 3.1 ± 0.3, R² = 0.23, p < 0.001, 95% CI [2.4–3.7]) and cohort 16 (β = 2.6 ± 0.6 R² = 0.15, p < 0.001, 95%CI [1.7–3.5]). These measures did not differ between cohorts (slope, t = 0.89, p = 0.37; intercept, t = 0.065, p = 0.94).

Figure 1.

Pre-ethanol average set-shift performance as a function of daily consecutive sessions in cohort 16 (present data) and cohort 14 (previously published). ***significant effect of session on PI (linear regression analysis (p < 0.0001).

Cohort 16 monkeys then entered the ethanol self-administration protocol. The average ethanol intake during open access ranged between 2.0 and 3.9 g/kg/day in this cohort. Intakes greater than 3.0 g/kg on >20% of days were the defining limit between heavy drinking (HD) and non-heavy drinking (NHD) status. By this definition, there were 2 HD (52 and 68% of sessions >3 g/kg/day) and 4 NHD (0–14% of sessions >3 g/kg/day) monkeys in cohort 16. The stability of daily alcohol intakes can be seen in the hive plot of Figure 2A, where the average daily intake of each monkey during the 180 sessions of open access (top node) is compared to consecutive 60-session averages in the clockwise direction. Only one monkey escalated its ethanol intake and switched categories from NHD to HD, and ultimately ranked as the heaviest drinker by the last 60 days of access and as the second heaviest drinker in the 6-month average daily intake. Such steep escalation after 2 months of drinking is a relatively rare phenomenon (seen here in 1/15 monkeys) and was not due to weight loss or decrease in daily food allotment. The average daily ethanol intake of monkeys in the HD group (n = 8) was 3.3 ± 0.2 g/kg/day (Figure 2B), compared to 2.1 ± 0.1 g/kg/day in NHD monkeys (t = 5.2, df = 3, p < 0.001). HD monkeys also had higher average BECs 7 hours after the start of the session in the HD group (89 ± 11.6 mg/dL) compared to the NHD groups (50 ± 8.8 mg/dL, t = 2.6, df = 13, p < 0.05, Figure 2C).

Figure 2.

Stability of the average daily ethanol intakes across approximately 6 months of daily self-administration, and drinking characteristics of HD and NHD designation. A. The top node of the graph represents average ethanol intake of each monkey from both cohorts (14 and 16, n = 15) across the entire 6-month period. The self-administration was divided into three sub-periods of 2 months each, and average ethanol intake per subsequent sub-periods (1 of 3, 2 of 3, and 3 of 3) is plotted at the 2nd, 3^rd, and 4th nodes. B. Average daily ethanol intake of HD and NHD groups (combined cohorts 14 and 16) during 6 months of open-access self-administration (22 hours/day). Data are means (± SEM). C. Average blood ethanol concentration (BEC) of HD and NHD groups (± SEM.) *p < 0.05 (unpaired t test).

Figure 3A shows the improvement in PI over sessions for cohort 16 differentiated future classification as a Heavy Drinker (fHD, n = 2) or a Non-Heavy Drinker (fNHD, n = 4), replicating published results from cohort 14 (Shnitko, Gonzales, et al., 2019). The group average PI slopes over sessions was 3.17 ± 0.57 in fNHD monkeys and 1.42 ± 0.45 in fHD monkeys. There was a significant improvement in the group-average performance (fNHD: R² = 0.21, p < 0.0001, 95%CI [2.0–4.3] and fHD: R² = 0.14, p < 0.01, 95%CI [0.5–2.3]), indicating that monkeys were learning to perform more efficiently (fewer errors in reaching higher sets in less time). However, the rate of improvement over the 30 sessions was significantly different between the two groups (slope, t = 2.0, p < 0.05; intercept, t = 1.3, p = 0.17). Similar results were observed when data were collapsed across cohorts 16 and 14 (fHD, n = 8; fNHD, n = 7, Figure 3B). While all animals improved performance over time (fNHD: β = 3.7 ± 0.38, R² = 0.32, p < 0.001, 95%CI [2.9–4.5] and fHD: β = 2.2 ± 0.35, R² = 0.13, p < 0.001, 95%CI [1.5–2.8]), the average increase in PI/session was higher in the fNHD compared to the fHD (slope, t = 3.0, p < 0.01; intercept, t = 1.1, p = 0.27). By the last ASST (session 30), 6 out of 7 fNHD monkeys (~86%) were reaching the reversal for extradimensional shift (set 8, Figure 3C, blue circle). In contrast, only 3 out of 8 fHD monkeys (38%) were reaching the reversal for extradimensional shift (Figure 3C, red circle). Three fHD monkeys (38%) did not advance beyond the simple discrimination set or its reversal. The ability of the animals to learn a simple discrimination was compared between fHD and fNHD groups of monkeys. Specifically, the number of trials to criterion (TTC, 12 correct out of 15 consecutive trials) for learning the simple discrimination set was examined for all monkeys across five initial sessions. The discrimination was considered learned if a monkey reached the criterion during 4 out of 5 consecutive sessions and the total number of TTC recorded during these sessions were summed for each animal. The group-average number of TTC was not different (p = 0.8, t = 0.2, df = 13, unpaired t test) between fNHD (370 ± 142) and fHD (386 ± 149), suggesting the ability to learn a simple discrimination could not predict future drinking status.

Figure 3.

Average daily pre-ethanol set-shift performance as a function of future drinking status of heavy (fHD) and non-heavy (fNHD) drinkers. A. Group average pre-ethanol PI by session in the present (replicate) cohort 16 across 30 daily sessions. Shading shows 95% confidence intervals. B. Group average pre-ethanol PI by session combining individuals from both cohort 14 and cohort 16 based on future drinking status as fHD or fNHD. *** indicates significant effect of session (p < 0.0001) and * indicates significant effect of group (p < 0.05) identified with linear regression analysis. C. Proportion of terminal ASST set reached for fNHD (blue outline) and fHD (red outline) monkeys in the final set-shift session (session 30).

Discussion

The novel task was self-pacing and was performed in the presence of many other monkeys performing the task at the same time. Therefore, showing it was replicable was an important aim of the present study (Figure 1). This replication suggests the task is engaging even though the animals were not isolated and the rate of acquiring proficiency in the task is a stable feature of this population of rhesus monkeys. This finding also reflects the utility of a composite measure of performance (errors/trials, set reached and time to finish), captured here as a performance index (PI) (Shnitko et al., 2017). Given the possibility of different outcomes using an index composed of three variables, the replication of PI improvement over 30 sessions shows a high degree of task robustness. This is notable because in this approach to assessing set-shift performance no monkey was removed from the data analyses due to preset criteria for performance and all monkeys acquired the simple discrimination (first set). A comparison to other attentional set-shifting studies in rhesus monkeys is not possible because the number of training and testing sessions in other ASST tasks reported in monkeys do not include this information (Moore, Killiany, Herndon, Rosene, & Moss, 2005; Weed, Bryant, & Perry, 2008; Weiss, White, Richardson, & Bachevalier, 2019).

Acquiring a strategy to advance through the sets in the ASST appears to be a trait-like characteristic that is associated with the probability of acquiring a heavy drinking phenotype. Specifically, the daily self-administration protocol provided an opportunity for replicating ASST performance across cohorts as well as replicating ASST performance as a risk for chronic heavy alcohol consumption. Further, neither cohort in the current study contributed data to the original data modeling that defined categorical levels of drinking (i.e., low, binge, heavy, and very heavy: Baker et al., 2014), showing that these defined levels of drinking using this self-administration procedure are informative when adding individuals from different cohorts, as has been done when analyzing the effect of age or sex on synaptic function following chronic drinking (Cuzon Carlson, Grant, & Lovinger, 2018). Finally, collapsing data from animals in the “low and binge” categories into “NHD” and animals from the “heavy and very heavy” categories into “HD” within a cohort, as was done in this study, allows enough statistical power to explore group comparisons of chronic alcohol drinking threshold effects. For example, reliable differences in drinking patterns, synaptic function, genomic changes, and relapse are seen when data from NHD are compared to HD data (Allen et al., 2018; Cervera-Juanes et al., 2016; Cuzon Carlson et al., 2018; Grant, Stafford, et al., 2008; Helms, Park, et al., 2014; Helms, Rau, et al., 2014; Moore et al., 2019; Shnitko, Gonzales, et al., 2019; Shnitko, Liu, Wang, Grant, & Kroenke, 2019). Here we examined the grouping of NHD and HD with the 15 monkeys from both cohorts 14 and 16 and found clear differences between the groups in their daily average ethanol intakes (Figure 2). Overall, the robust self-administration protocol provides the addition of monkeys over time to the drinking categories for genomic and imaging data sets that require large numbers for the multiple comparisons made with these approaches (Iancu et al., 2018; Shnitko, Liu, et al., 2019).

The predictive relationship of ASST and future drinking status replicated the original finding from cohort 14 (Shnitko, Gonzales, et al., 2019) in the current cohort of monkeys (cohort 16; Figure 3A), as well as extended the findings when the cohorts were combined for each drinking outcome (HD or NHD; Figure 3B). By the end of the procedure (session 30), a majority of fNHD monkeys routinely completed all eight sets of the ASST (Figure 3C), which is seen in the high average PI of this group (Figure 3C). In contrast, the fHD group had a variety of deficits while performing the task (Figure 3C). Specifically, a subset of these monkeys was significantly challenged by the reversal of original sets (25% not advancing in SD reversal [set 2] and 25% not advancing in EDS), indicating potential deficits in functioning of the orbitofrontal cortex and amygdala (Rudebeck & Murray, 2008). A smaller group of fHD monkeys (12.5%) was unable to perform the interdimensional shift, indicating deficits in the ventrolateral prefrontal cortex (Oh, Vidal, Taylor, & Pang, 2014). The phenomenon that low flexibility in ASST is associated with heavier alcohol intake corresponds to a growing animal and human subject literature on prefrontal control of perseveration and inhibitory mechanisms involved in excessive alcohol exposure (Moorman, 2018). Additionally, data showing pre-existing sub-optimal functioning within the frontal lobe and parietal cortex are essential for flexibility and executive control over behavior in general (Barbas & Zikopoulos, 2007; Collins, Roberts, Dias, Everitt, & Robbins, 1998; Moorman & Aston-Jones, 2015; Rougier, Noell, Cohen, & O’Reilly, 2005). A number of studies, including our previous study on cohort 14, have documented aspects of behavior that become more habitual (automatic) with chronic heavy drinking or chronic ethanol exposure (Renteria, Baltz, & Gremel, 2018; Sey, Gómez, Madayag, Boettiger, & Robinson, 2019; Trantham-Davidson et al., 2014). However, this is the first to characterize and replicate that a predisposition toward habitual behavior can predict chronic heavy drinking in primates.

A main limitation of the study is the inclusion of only male subjects. It will be important to document whether the association between the behavioral flexibility as measured in this set-shifting task is also seen with female rhesus monkeys, particularly in light of recent reviews that AUD produces sex-specific effects on women’s cognition (Fama, Le Berre, & Sullivan, 2020) and response to treatment (McCrady, Epstein, & Fokas, 2020).

Highlights.

• The average improvement in performance on a self-pacing attentional set-shifting task by rhesus monkeys was replicated in a separate cohort of 6 rhesus monkeys, suggesting that the task is engaging and that the average rate of acquiring proficiency in the task is a stable feature of this population of rhesus monkeys.

• Categorical drinking levels, defined across cohorts with deep data modeling from the ethanol self-administration protocol, applied here as “future drinking category”, were robustly predicted by improvement on the set-shifting task 16 months prior to ethanol access in a replicate cohort.

• When combined with a previous cohort, low rates of performance improvement continued to predict future status as a chronic heavy alcohol drinker, demonstrating a robust risk factor for developing a heavy drinking phenotype.