Beneficial and adverse effects of THC on cognition in the HIV-1 transgenic rat model: Importance of exploring task- and sex-dependent outcomes

Abstract

HIV-associated neurocognitive impairment (NCI) is an untreated concern among people living with HIV (PLWH). Cannabis use in PLWH may complicate outcomes on cognition, with evidence to suggest function-dependent effects that are modulated by several factors including use patterns (e.g., frequency of use) and demographic influences (e.g., age). Animal studies can control for these factors.

Here, we characterized the impact of the primary psychoactive ingredient in cannabis (delta-9-tetrahydrocannabinol; THC), on function-dependent cognitive outcomes in HIV-1 transgenic (Tg) rats using cross-species translatable assays. Female and male HIV-1Tg rats and their controls were tested in the rat Iowa Gambling Task (IGT; to measure risk-based decision-making), and the Probabilistic Reversal Learning Task (PRLT; to measure learning and cognitive flexibility). Rats were tested at baseline, then retested following acute and chronic exposures to THC (0, 0.3, 3 mg/kg, intraperitoneal injection).

At baseline, HIV-1Tg rats took longer to make decisions, but exhibited intact cognition across tasks, suggestive of a speed-accuracy trade-off and early cognitive deficits. Both acute and chronic THC exposures produced selective effects on primary performance measures in HIV-1Tg rats, including enhanced learning performance but worsened risk-based decision-making, not observed in controls.

This work confirms function-dependent effects of THC on cognitive function in an animal model of HIV using cross-species translatable tasks used in the clinic. Findings are consistent with evidence for function-dependent cannabis effects observed in HIV, and suggest THC may drive cannabis-induced changes observed on cognitive performance in PLWH. These data may serve as guidance for clinicians prescribing cannabis to patients with HIV, and for further research exploring the interactive effects of HIV and cannabinoid treatment on cognitive function.

1. Introduction

Although advancements in the treatment of human immunodeficiency virus (HIV) have reduced mortality rates among affected individuals, HIV-associated neurocognitive impairment (NCI) remains a persistent and untreated concern (Heaton et al., 2010). People living with HIV (PLWH) frequently report cognitive deficits (Knippels et al., 2002; Thames et al., 2011), and nearly half receive formal diagnoses of HIV-associated NCI (Wang et al., 2020; Wei et al., 2020). The complexity of HIV-associated NCI is compounded by its heterogeneous nature, with distinct functional domains disturbed across individual diagnoses. Indeed, the cognitive domains most commonly affected are verbal fluency, attention/working memory, executive function, learning and memory, information-processing speed, and motor skills, though patients are diagnosed based on observed deficits in any two or more of these cognitive processes (Heaton et al., 2011). There is an urgent need for treatments specifically targeting HIV-associated NCI, however, the development of effective therapeutics will likely necessitate an individualized treatment approach informed by research exploring domain-specific outcomes (Ayoub et al., 2024).

Cannabis use is more prominent among PLWH compared to the general population (Shiau et al., 2017; Costiniuk et al., 2019). This elevated use pattern may relate to self-medication tendencies as PLWH report using cannabis to alleviate HIV-related symptoms including anxiety, anorexia, and pain (Costiniuk et al., 2019; Harris et al., 2014). While cannabis can adversely affect cognition in healthy individuals (Shrivastava et al., 2011), its impact on cognition in PLWH remain unclear. A recent review of clinical and preclinical literature indicated that cannabis use among PLWH is rarely linked to worsened cognitive outcomes and, in fact, may alleviate HIV-associated NCI (Ayoub et al., 2024). Notably, the interactive effects of HIV and cannabis use appear to be highly dependent on the specific cognitive function assessed, as well as key moderating factors such as use patterns and age. For instance, studies have reported that cannabis use is linked to superior learning, executive function, and verbal fluency in PLWH (Watson et al., 2023; Watson et al., 2020; Watson et al., 2021; Flannery et al., 2022). For example, daily, but not moderate, cannabis use was associated with lower CNS proinflammatory chemokines in PLWH, which were linked to enhanced learning performance (Watson et al., 2021). Conversely, aspects of executive function, such as risk-based decision-making, and memory performance may be negatively impacted by cannabis use in PLWH, particularly at higher doses (Gomez et al., 2017; Cristiani et al., 2004; Skalski et al., 2018). While these association studies provide valuable guidelines for assessment, they remain correlational, as they are based on people choosing to use cannabis, with limited controlled studies involving cannabis administration to PLWH. Nonetheless these data underscore the need for directionality and domain-specific approaches to defining NCI in PLWH to ultimately advance treatment options.

The variability across clinical studies may also stem from differences in studied cohorts’ exposure to commercially available cannabis products which widely range in their composition of cannabinoids, including varying concentrations of the primary psychoactive component delta-9-tetrahydrocannabinol (THC). This variability is an important factor to consider since THC, when compared to other cannabinoids such as cannabidiol (CBD), produces differential, or in some cases even opposing effects, on function-dependent cognitive outcomes (Woelfl et al., 2020; Fusar-Poli et al., 2009; Fusar-Poli et al., 2010; Curran et al., 2020; Bhattacharyya et al., 2010). Some experimental clinical studies have orally administered the synthetic version of THC, dronabinol, to PLWH and measured cognitive outcomes on attention, learning, information-processing and memory domains. The findings revealed a negative impact of acute higher doses of dronabinol (30 mg) on memory and attentional abilities, that were not observed at lower doses (≥20 mg) whether administered acutely (Haney et al., 2005) or subchronically (4 days QUID) in a separate cohort (Haney et al., 2007). Further, when one of these doses (10 mg) was administered chronically for 16 days, it produced minimal albeit adverse effects on measures of attention and processing-speed (Bedi et al., 2010). These findings largely mimic the function-dependent and dose/frequency-dependent impact of cannabis use on HIV-associated neurocognitive disorder, suggesting THC may drive the observed effects. Further investigations defining the impact of THC on HIV-associated neurocognitive impairment are warranted to both probe its utility as a treatment option, but also to better understand its role in the function- and dose-dependent effects of cannabis use observed in PLWH (Ayoub et al., 2024).

The independent and interactive effects of cannabinoid use and HIV seem dependent on factors that are not always easy to control for in clinical populations (e.g., gender, age, age of HIV onset, cannabis use, etc.). Hence, preclinical animal models of neuroHIV are essential for elucidating discrepancies in the literature. Preclinical studies also help to define more casual mechanisms through which cannabis exposure and HIV influence cognitive performance. The HIV-1 transgenic (Tg) rat model expresses 7 of the 9 HIV-1 viral proteins (Reid et al., 2001); mirroring several aspects of HIV as observed in PLWH including immune deficiency (Reid et al., 2001; Yadav et al., 2006; Li et al., 2013), neuroinflammation (Royal et al., 2012; Cho et al., 2017; Repunte-Canonigo et al., 2014), and cognitive impairment (Repunte-Canonigo et al., 2014; Moran et al., 2019; McLaurin et al., 2018; Vigorito et al., 2007). Importantly, the HIV-1Tg rat lacks the gag and pol proteins, rendering the viral genome non-replicable and non-infectious. Hence, this model is effectively simulates the context of modern HIV infection wherein PLWH have ready access to effective therapies that suppress viral load and subsequent infection. To our knowledge only one study has tested the impact of THC on function-dependent cognitive outcomes in an animal model of HIV. Mice exposed to the viral HIV-Tat protein exhibited impaired recognition memory and acute THC did not worsen these deficits, while chronic effects were not explored (Yadav-Samudrala et al., 2024).

In this study, we utilized the HIV-1Tg rat model to investigate the independent and interactive effects of THC and HIV on cognitive outcomes across discrete functional domains. We first assessed baseline performance before retesting the animals following exposure to both acute and chronic THC at low to medium doses (0, 0.3, 3 mg/kg). These doses were chosen because they attenuated cognitive decline in rodent models of aging (0.002 mg/kg-3 mg/kg) (Bilkei-Gorzo et al., 2017; Suliman et al., 2018; Wang et al., 2022; Nidadavolu et al., 2021; Sarne et al., 2018), with the higher dose also affording an approximate THC exposure level comparable to the average daily cannabis user (Borodovsky et al., 2025; Larsen et al., 2023; Kerr and Ye, 2022). Importantly, the tasks used are cross-species compatible (i.e., have versions that are administered to humans) to enhance translational relevance of our findings to the clinic. To evaluate risk-based decision-making, the Iowa Gambling Task (IGT) was utilized; a task in which NCI-diagnosed PLWH perform more poorly than both non-NCI diagnosed PLWH and healthy participants (Hardy et al., 2006; Martin et al., 2004; Iudicello et al., 2013; Nakao et al., 2020; Nigro et al., 2021). To assess both learning and cognitive flexibility (reversal learning), cognitive functions known to be impaired in PLWH (Kesby et al., 2015; Kanmogne et al., 2018), the Probabilistic Reversal Learning Task (PRLT) was utilized. Based on the existing literature on the function-dependent effects of cannabis use and HIV on cognition (Ayoub et al., 2024), we hypothesized that THC exposure would: 1) impair risk-based decision making in HIV-1Tg rats; and 2) improve initial and reversal learning in HIV-1Tg rats.

2. Materials and methods

2.1. Animals

A total of 142 adult male and female HIV-1Tg (n = 58) and control (n = 84) rats (49 % female) were utilized in this study. The HIV-1Tg rats were bred in-house onto the inbred Fischer F344 rat strain and control animals included both HIV-1Tg wildtype littermates (n = 48) and Fischer F344 rats (n = 36), commonly used in the literature (Reid et al., 2001; Li et al., 2013; Royal et al., 2012; Vigorito et al., 2007). Animals were pair-housed in ventilated shoebox cages with standard environmental enrichment (plastic tube housing and nesting material; LWH: 15.5″ x 11.5″ x 8″) in a temperature-controlled room (21 ± 1°C) on a reversed light–dark cycle (7:00/19:00). Food and water was provided ad libitum, except during operant training and testing when animals were food restricted to 85 % of their free-feeding body weight. All animals began experimental procedures at 3 months of age (adult). All behavioral testing began at least 2 h into the animal’s dark (active) phase and only occurred during their active period. Rats were maintained in a University of California San Diego (UCSD) animal facility which meets all federal and state requirements for animal care, and all procedures were approved by the Institutional Animal Care and Use Committee (IACUC) at UCSD.

2.2. Drugs

THC dissolved in ethanol was obtained from the National Institute on Drug Abuse (NIDA). The ethanol was evaporated under a stream of dry nitrogen, and the residue was dissolved to final concentrations of 0.3 and 3 mg/mL in a vehicle consisting of 7.5 % Tween-80 and 7.5 % propylene glycol (Sigma-Aldrich, St. Louis, MO, USA) in saline. Animals were treated with 0, 0.3 or 3 mg/kg THC 30 min prior to cognitive testing, or at the same time of day on non-test days (i.e., during chronic exposure). This range of doses were chosen because they attenuated cognitive decline in rodent models of aging (0.002 mg/kg-3 mg/kg) (Bilkei-Gorzo et al., 2017; Suliman et al., 2018; Wang et al., 2022; Nidadavolu et al., 2021; Sarne et al., 2018), with the higher dose also affording an approximate THC exposure level to the average daily cannabis user (Borodovsky et al., 2025; Larsen et al., 2023; Kerr and Ye, 2022). When delivered via intraperitoneal injection at a similar doses (2.5–––5 mg/kg) to our higher dose in the current study (3 mg/kg), the half-life of THC was 2 h in plasma, with minimal detection of THC levels in plasma or brain by 4 h (Baglot et al., 2021; Ruiz et al., 2021). Thus, we are confident that the pharmacological effects of THC were onboard throughout all tests (1 h duration), and that the 48 h between each acute THC test was sufficient to wash out any previous administrations of THC (see experimental Timeline below).

2.3. Apparatus

Cognitive behavior was measured using 5-choice nosepoke operant chambers (Med Associates Inc., St. Albans, VT and Lafayette Instrument Company, Lafayette, IN; LWH: 10.5–12″ x 10.5 “ x 12”), located in larger sound-attenuating cabinets (LWH: 23” x 18” x 18.5), with built-in fans to mask noise and provide ventilation. Nosepoke apertures consisted of an LED light for stimulus presentation and an infrared beam to detect responding. Reward (strawberry milkshake; Nesquik in nonfat milk), was delivered into a magazine on the opposite wall of the chamber that contained an LED light to signal reward delivery and an infrared beam to detect reward collection. Stimulus outputs and response inputs were managed by a Smart Ctrl Package (8-In/16-Out) with additional interfacing by MED-PC for Windows (Med Associates Inc., St. Albans, VT) using custom programming.

2.4. Operant behavioral training

Rats were first conditioned to associate an illuminated magazine with reward delivery during a 20 min session wherein 50 µl strawberry milkshake was dispensed on a 15 sec fixed interval schedule. Rats were maintained on this program until at least 60 reward collections were made across two consecutive days. Next, animals were trained to make an operant response (nosepoke) into one of five illuminated choice apertures to earn 50 µl of reward on a fixed-ratio 1 (FR1) schedule of reinforcement. Importantly, after 5 continuous responses in a single aperture the aperture was disabled to prevent the development of side and/or aperture biases. Disabled apertures were reactivated following 2 responses in other apertures. Rats were maintained on this program until at least 70 reward collections were made across 5 consecutive days. To avoid over-training, animals that reached these training criterion faster than their peers were maintained on this program twice weekly.

2.5. IOWA Gambling task (IGT)

The IGT is a 60 min risk-based decision-making task designed to mimic the human IGT by providing four choice options that vary in reward and punishment contingencies and magnitudes (see Fig. 1A). A nosepoke in an illuminated reward magazine was required to start each trial. Following a 5 sec inter-trial interval (ITI) from trial initiation four evenly spaced apertures were illuminated for 10-sec. Two choice apertures are designated “risky” since selection resulted in larger rewards (80 µl), but also greater punishments (66–132 sec). Conversely, two “safe” options delivered smaller rewards (40 µl), but lesser punishment (6–12 sec) upon selection. Failure to select a choice aperture during the response-window resulted in a new trial and were recorded as an omission. Responses to choice apertures during the ITI also resulted in a new trial and were recorded as a premature response. Omissions and premature responses were not included in total choice trial counts. Table 1 defines all primary and secondary variables in the IGT. The primary outcome measures included a difference score calculated by subtracting the number of risky from safe choices, and decision-making metrics: percentage [%] of trials with a safe-win stay, safe-lose shift, risky-win stay, risky-lose shift. Secondary outcome variables included choice trials, %premature trials, %omission trials, response latency and reward latency.

Fig. 1.

Cross species translatable tasks. The Iowa Gambling Task (A) measures risk-based decision-making by assessing choice patterns across options that yield differing probabilities and magnitudes of wins and losses. In the rat version (right) animals choose from two risky options that yield higher reward values but lengthier timeout punishments, and safe options that yield less reward but are associated with lower probabilities of experiences less-lengthy timeout punishments. The Probabilistic Reversal Learning Task (B) measures initial reinforcement learning and cognitive flexibility. In this task correct selection of the target location results in a 80% chance of earning reward and 20% chance of experiencing a timeout punishment. The non-target location has the opposite reward/punishment contingencies. Once the target location is selected for 8 consecutive trials, the target and non-target switch locations (i.e., a reversal is achieved).

Table 1.

Description of primary and secondary IGT outcome measures. Consistent with the human IGT, the main outcome variable of the rat IGT is the difference score which provides a measured of risk-based decision-making. Other primary variables include decision-making metrics (i.e., safe-win stay, risky-lose shift, etc.) which can inform how an animal updates risk-based decision-making choices following task feedback. Secondary variables are used to gauge task performance, less related to cognition per se.

Dimension	IGT Variable	Description
Risk-based decision making	Difference Score	No. of safe choices – risky choices
Low-reward-focused decision making metric	%safe-win stay	No. of safe-win stay choices/(no. of safe-win stay + no. of safe-win shift)
Low-punishment-focused decision making metric	%safe-lose shift	No. of safe-lose shift choices/(no. of safe-lose shift + no. of safe-lose stay)
High-reward-focused decision making metric	%risky-win stay	No. of risky-win stay choices/(no. of risky-win stay + no. of risky-win shift)
High-punishment-focused decision making metric	%risky-lose shift	No. of risky-lose shift choices/(no. of risky-lose shift + no. of risky-lose stay)
Secondary Variables:	Choice Trials	No. of trials – premature + omission trials
	%prematures	No. of trials where a non-illuminated choice aperture response was made during the 5-sec ITI period
	%omissions	No. of trials where no response was made during the 10-sec response window
	Response Latency	Time (sec) to respond following the 5-sec ITI period
	Reward Collection Latency	Time (sec) to collect reward from magazine following on a rewarded trial

2.6. Probabilistic Reversal-Learning task (PRLT)

The PRLT is a 60 min within-sessions reversal learning paradigm that measures both initial and reversal learning about probabilistic reinforcement (see Fig. 1B). This task utilized 2 of the 5 choice apertures, evenly spaced apart. One choice aperture represented a target location where 80 % of responses were rewarded with 50 µl of strawberry milkshake and 20 % of responses were punished with a 2 sec flash of the houselight and no reward. The second choice aperture represented a non-target location that had reversed reward/punishment contingencies (i.e., 20 % reward, 80 % punishment). Trials did not time out and therefore required choice behavior prior to trial succession. Table 2 defines all primary and secondary variables in the IGT. The primary outcome measures included trials to first reversal, and the number of reversals achieved. Secondary outcome variables included choice trials, %premature trials, target response latency, and non-target response latency.

Table 2.

Description of primary and secondary PRLT outcome measures. There are two main outcome variables of the rat PRLT. The number trials to first reversal switch is used as a measure of initial reinforcement learning, whereas the number of reversal achieved is used as a measure of reversal learning indicative of cognitive flexibility. Secondary variables are used to gauge task performance, less related to cognition per se.

Dimension	PRLT Variable	Description
Reinforcement Learning	Trials to first reversal switch	No. of trials required to reach the 8 consecutive target responses criterion and cause the first reversal switch
Reversal Learning	Reversals	No. of reversal switches achieved
Secondary Variables:	Choice Trials	No. of trials – premature trials
	%premature	No. of trials where a non-illuminated choice aperture response was made during the 5-sec ITI period
	Target Response Latency	Time (sec) to respond to the target location following the 5 sec ITI period
	Non-target Response Latency	Time (sec) to respond to the non-target location following the 5 sec ITI period

2.7. Experimental timeline

The full experimental timeline is depicted in Fig. 2. All animals completed training and were then tested in the IGT followed by the PRLT. Each test was separated by a FR1 schedule of reinforcement training day which utilized the number of choice apertures required for the respective task (IGT n = 4; PRLT n = 2). After PRLT testing all animals were then tested in the progressive-ratio breakpoint task (data not presented here). Animals were assessed in both tasks at 1) baseline, 2) following acute THC, 3) following chronic THC. Baseline behavior was used to assign animals to one of three THC doses matched on the primary outcomes measures of risk-based decision-making (IGT) and initial and reversal learning (PRLT). Treatments were administered 30 min prior to cognitive testing, and at the same time of day on non-testing days. All animals received 5 consecutive days of FR1 training prior to each bout of cognitive testing (i.e., baseline, acute, chronic) to re-stabilize operant responding.

Fig. 2.

Experimental Timeline. Animals were trained on a fixed-ratio 1 (FR1) schedule of reinforcement then tested for baseline performance in the Iowa Gambling Task (IGT) and the Probabilistic Reversal Learning Task (PRLT). Between each test “habituation” (HAB) training sessions occurred wherein rats were exposed to the apertures used in the subsequent test to familiarize and restabilize FR1 activity. Following baseline, rats were retested in the IGT and PRLT following acute exposure to THC (0, 0.3 or 3 mg/kg; intraperitoneal) 30 min prior to being placed in operant chambers. Rats were then exposed to the same dose of THC for 11 days, then retested in the IGT and PRLT during chronic steadystate THC exposure. Between each testing bout animals received at least 5 days of FR1 training to re-stabalize operant behavior.

2.8. Statistical analyses

Data were first analyzed using separate univariate analyses of variance (ANOVA) tests with gene (Control, HIV-1Tg), drug (0, 0.3, 3 mg/kg THC), and sex (male, female) as factors where appropriate. These initial analyses determined some independent effects of sex, but no interactions of sex and pretreatment on any main outcome variable across all tests and timepoints, with the sole exception of sex-dependent effects of chronic THC on IGT behavior. Based on these findings, and given a priori hypotheses of THC-specific effects on HIV-1 Tg rats, data was collapsed across sex and reanalyzed to increase statistical power when there were no observed sex interactions on main outcome variables. Nonetheless, independent and interactive effects of sex are reported within the main text and/or Supplemental Materials. Where no sex interactions were observed, baseline data was analyzed with two-sided independent samples t-tests to determine effects of gene on cognitive performance. For within-sessions time-binned data, each animals’ total trial count was used to split session data into 3 equal time bins for analysis. Repeated-measures ANOVAs were utilized to assess time-binned data, with Greenhouse-Geisser corrections applied to analyses with violations of sphericity. Bonferroni adjustments were used to correct for multiple comparisons and LSD comparisons were used to explore a priori predictions. To avoid skewed data from response biases, animals that did not select each choice option (safe vs. risky for IGT; target vs. non-target for PRLT) at least twice during testing were removed from analyses. Data from animals that did not complete the minimum number of trials for each task (40 choices for IGT, 20 choices for PRLT) were also removed from analyses. Supplemental Fig. 1 outlines the number of animals by group that were removed for each analysis. Statistical analyses were conducted using SPSS v 27 (Chicago, IL), with alpha set at p < 0.05, though trend effects (p < 0.1) were reported where observed.

3. Results

3.1. HIV-1 transgenic rats exhibited normal initial and reversal learning at baseline

Baseline PRLT performance is illustrated in Fig. 3. A total of 3 animals were excluded from these analyses based on response biases (n = 1) or low trial counts (n = 2). On main outcome measures, a trending effect of sex [F_(1,135) = 3.55, p = 0.062] revealed females tended to have achieve a higher number of reversal switches than males. However, sex did not interact with drug or genotype to impact reversal switches, and no main or interactive effects of sex were observed on the other main outcome variable of trials to first reversal switch (Fs < 2, ps > 0.1). Data were therefore collapsed across male and female rats. At baseline HIV-1Tg and control rats produced comparable number of trials to first reversal [t₍₁₂₄₎ = 1.641, p = 0.103; Fig. 3A] and reversal switches [t₍₁₃₇₎ = 0.48, p = 0.962; Fig. 3B] as controls.

Fig. 3.

HIV-1 transgenic animals exhibited normal baseline initial and reversal learning, but altered choice latencies. Compared to controls, HIV-1Tg rats exhibited similar initial learning, as measured by trials to first reversal switch (A) and similar reversal learning, as measured by number of reversal switches achieved (B). HIV-1Tg rats completed less choice trials (C) but similar %premature trials (D) as controls. Choice latencies for both the target location (E) and the non-target location (F) were slower in HIV-1Tg rats. Data presented as individual data-points, plus mean ± S.E.M. Circles and triangles represent female and male data points, respectively. *p < 0.05, #p < 0.1, as indicated.

On secondary measures, some sex effects were observed that were consistent with IGT test performance. For example, compared to females, males had lower choice trials [F_(1,135) = 22.456, p < 0.001] and % prematures [F_(1,135) = 5.79, p = 0.017]. Males also had slower choice latencies for both target [F_(1,135) = 6.96, p = 0.009] and non-target locations [F_(1,135) = 5.04, p = 0.026]. However, sex did not interact with genotype or drug on any measure (Fs < 2, ps > 0.1), hence as with primary variables, data was collapsed across male and female animals. HIV-1Tg rats completed less choice trials [t₍₁₃₇₎ = 2.368, p = 0.019; Fig. 3C], but a similar number of %premature trials [t₍₁₃₇₎ = 0.595, p = 0.553; Fig. 3D] when compared to controls. HIV-1Tg rats also tended to have slower choice latencies for the target location [t₍₁₃₇₎ = 1.838, p = 0.068; Fig. 3E] and significantly slower latencies for the non-target [t₍₁₃₇₎ = 2.146, p = 0.034; Fig. 3F] location.

3.2. Acute THC enhanced initial learning, selectively in HIV-1Tg rats

The impact of acute THC on PRLT performance is illustrated in Fig. 4. A total of 7 animals were excluded from these analyses based on response biases (n = 3) or low trial counts (n = 4). A trending sex*genotype interaction on trials to first reversal [F_(1,105) = 3.22, p = 0.076] revealed HIV-1Tg rats took less trials to achieve their first reversal switch compared to controls (p = 0.032), selectively in male rats. Despite this faster initial learning, male HIV-1Tg rats tended to achieve fewer reversals than male controls [p = 0.081; sex*genotype: F_(1,123) = 3.6, p = 0.060]. No other main or interactive effects of sex were observed on main outcome variables (Fs < 2, ps > 0.1), therefore data was collapsed across male and female rats. There was a trend effect of drug [F_(1,111) = 2.687, p = 0.072] on the main outcome variable of trials to first reversal but no significant group comparisons withstood Bonferroni corrections. Given a priori predictions that THC would enhance learning in HIV-1Tg rats, the genotype*drug interaction [F_(2,111) = 2.049, p = 0.134] was explored. Group comparisons revealed that HIV-1Tg rats had lower trials to first reversal than controls at the 3 mg/kg dose (p = 0.02). Further, 3 mg/kg THC-treated HIV-1 Tg rats had lower trials to first reversal switch compared to 0.3 mg/kg (p = 0.013) and VEH (p = 0.026) rats (Fig. 4A). No effect of genotype or drug were observed on the main outcome variable of number of reversal switches (Fig. 4B). Given a priori predictions that THC would enhance cognitive flexibility in HIV-1Tg rats the genotype*drug interaction [F_(2,129) = 0.751, p = 0.474] was explored, but no significant differences between groups were observed.

Fig. 4.

Acute THC enhanced initial learning, selectively in HIV-1Tg rats. Exploratory analyses based on a priori hypotheses revealed high-dose THC enhanced reversal learning, selectively in HIV-1Tg rats, as indicated by decreased trials to first reversal switch (A). There were no effect of gene or drug on reversal learning (B). HIV-1Tg rats tended to complete less choice trials (C), and high-dose THC tended to reduce choice trials across genes. There were no effects of gene or drug on % premature trials (D). HIV-1Tg rats had normal choice latencies for both the target (E) and non-target (F) locations, with high-dose THC slowing target choice latencies across all rats. Data presented as individual data-points, plus mean ± S.E.M. Circles and triangles represent female and male data points, respectively. **p < 0.01, *p < 0.05, #p < 0.1, as indicated.

On secondary measures, females had higher choice trials [F_(1,123) = 47.781, p < 0.001], %prematures [F_(1,123) = 20.95, p < 0.001] and faster choice latencies for both the target [F_(1,123) = 4.05, p = 0.046] and non-target [F_(1,123) = 4.09, p = 0.045] options. However, sex did not interact with drug or genotype for any of these measures, therefore male and female data were collapsed. HIV-1Tg rats tended to have lower choice trials [F_(1,129) = 3.585, p = 0.061; Fig. 4C] and %prematures [F_(1,129) = 3.01, p = 0.085; Fig. 4D] than controls. A main effect of drug [F_(2,129) = 4.02, p = 0.02] revealed 3 mg/kg THC reduced choice trials compared to 0.3 mg/kg THC (p = 0.016). HIV-1Tg rats had similar choice latencies as controls (Fig. 4E and F), but 3 mg/kg THC slowed target choice latency [F_(2,129) = 6.01, p = 0.003] compared to VEH (p = 0.007) and 0.3 mg/kg (p = 0.013) THC, across all animals.

3.3. Chronic THC enhanced learning selectively in HIV-1Tg rats, and tended to impair reversal learning in controls

The impact of chronic THC on PRLT performance is illustrated in Fig. 5. A total of 13 animals were excluded from these analyses based on response biases (n = 6) or low trial counts (n = 7). Additionally, one male transgenic animal (0.3 mg/kg THC group) fell sick and was removed from to study prior to testing. Sex produced no main or interactive effects (Fs < 2, ps > 0.1) on the main outcome variables, therefore data were collapsed across males and females rats. There were no main or interactive effects observed on trials to first reversal (Fig. 5A). Given a priori predictions that THC would enhance learning in HIV-1Tg rats, the genotype*drug interaction [F_(2,115) = 1.56, p = 0.214] was explored. Similar to the acute test, chronic 3 mg/kg THC tended to reduce trials to first reversal relative to VEH (p = 0.053), selectively in HIV-1Tg rats. Further, HIV-1Tg rats tended to have reduced trials to first reversal compared to controls only following 3 mg/kg THC treatment (p = 0.093). There were no main or interactive effects observed on the main outcome measure of number of reversal switches achieved (Fig. 5B). Given a priori predictions that THC would enhance cognitive flexibility in HIV-1Tg rats, the genotype*drug interaction [F_(2,122) = 1.74, p = 0.180] was explored. Chronic 0.3 (p = 0.058) and 3 (p = 0.092) mg/kg THC tended to reduce reversals achieved compared to VEH in control, but not HIV-1Tg, rats.

Fig. 5.

Chronic THC enhanced learning, selectively in HIV-1Tg rats and tended to lower reversal learning performance in control animals. Exploratory analyses based on a priori hypotheses revealed high-dose THC enhanced learning, selectively in HIV-1Tg rats, as indicated by decreased trials to first reversal switch (A). There were no effects of gene or drug on reversal learning, though exploratory analyses revealed THC decreased this cognitive flexibility measure in controls (B). HIV-1Tg rats completed less choice trials (C), which tended to be lowered by high-dose THC across all rats. Within high-dose THC treated animals, HIV-1Tg rats tended to have reduced %premature trials compared to controls (D). HIV-1Tg rats had slowed choice latencies for the target (E) and non-target (F) locations. Circles and triangles represent female and male data points, respectively. ***p < 0.01 **p < 0.01, *p < 0.05, #p < 0.1, as indicated.

On secondary outcomes, females completed less choice trials than males [F_(1,117) = 14.79, p < 0.001]. Sex did not produce any other main or interactive effects on secondary measures, therefore data were collapsed across male and female animals. HIV-1Tg rats had lower choice trials [F_(1,122) = 4.37, p = 0.039] than controls and 3 mg/kg THC lowered choice trials relative to 0.3 mg/kg THC (p < 0.001) and VEH (p = 0.003; [drug; F_(2,122) = 9.45, p < 0.001; Fig. 5C]. A trending gene*-dose interaction [F_(1,122) = 2.62, p = 0.077] revealed lower %prematures in HIV-1Tg rats only following 3 mg/kg THC exposure (p = 0.036; Fig. 5D). Trending and significant main effects of genotype on choice latencies for the target [F_(1,122) = 3.76, p = 0.055; Fig. 5E], and non-target [F_(1,122) = 6.85, p = 0.01; Fig. 5F] locations revealed HIV-1Tg rats took longer than controls to select options during PRLT testing.

3.4. HIV-1 transgenic rats exhibited normal risk-based decision-making at baseline

Baseline IGT performance is illustrated in Figs. 6 and 7. A total of 9 animals were excluded from these analyses based on response biases (n = 6) or low trial counts (n = 3). Sex did not produce main or interactive effects (Fs < 2, ps > 0.1) on the main outcome variables of difference score or decision-making metrics, therefore data was collapsed across male and female rats. Difference scores were comparable between HIV-1Tg and control rats, [t₍₁₃₁₎ = 0.63, p = 0.53; Fig. 6A]. Consistently, HIV-1Tg rats did not differ from controls on secondary measures of decision-making metrics including %safe-win stay [t₍₁₂₉₎ = 0.23, p = 0.82; Fig. 6B], %safe-lose shift [t₍₁₂₅₎ = 0.69, p = 0.49; Fig. 6C], %risky-win stay [t₍₁₂₅₎ = 0.98, p = 0.33; Fig. 6D], or %risky-lose shift [t₍₁₂₉₎ = 0.27, p = 0.79; Fig. 6E]. Difference scores were also analyzed across time-binned data to test whether learning to choose safe options occurred throughout the IGT testing session (Supplemental Fig. 2), consistent with the human literature. A main effect of time [F_(1.2,262) = 6.649p = 0.007], wherein difference scores were lower at time-bin 1 compared to time-bin 2 (p = 0.024) and time-bin 3 (p = 0.028). While there were no effects of, or interactions with, genotype these data validate the preclinical IGT task utilized herein.

Fig. 6.

HIV-1Tg rats exhibited normal risk-based decision-making at baseline. Risk-based decision-making was comparable between HIV-1Tg and controls as measured by difference score (A), and decision-making metrics of %safe-win stay trials (B), %safe-lose shift trials (C), %risky-win stay trials (D), %risky-lose shift trials (E). Circles and triangles represent female and male data points, respectively. Data presented as individual data-points, plus mean ± S.E.M.

Fig. 7.

HIV-1Tg exhibited altered choice and reward collection latencies. HIV-1Tg performed similarly to controls on secondary measures of choice trials (A) %omission trials (B), and %premature trials (C). HIV-1Tg rats tended to take longer to choose a response option (D) but took less time to collect rewards (E). Data presented as individual data-points, plus mean ± S.E.M. Circles and triangles represent female and male data points, respectively. # p < 0.1, ***p < 0.001 as indicated.

On secondary outcome measures, some sex effects were observed. For example, compared to males, female rats had higher %premature responses [F_(1,129) = 4.512, p = 0.036], less %omissions [F_(1,129) = 18.474, p < 0.001], quicker choice latencies [F_(1,129) = 22.580, p < 0.001] but slower reward collection latencies [F_(1,129) = 37.692, p < 0.001]. Sex and gene did not interact for any of these measures (Fs < 1, ps > 0.1), other than %omissions [F_(1,129) = 4.718, p = 0.032] whereby females exhibited fewer %omissions than males only in control animals (p < 0.001). Hence, as with primary outcome variables, data were collapsed and examined by genotype. HIV-1Tg rats completed a similar number of choice trials [t₍₁₃₁₎ = 0.86, p = 0.39; Fig. 7A], %omissions [t₍₁₃₁₎ = 0.32, p = 0.75; Fig. 7B] and %prematures [t₍₁₃₁₎ = 1.27, p = 0.21; Fig. 7C] as controls. HIV-1Tg rats, however, tended to have slower choice latencies [t₍₁₃₁₎ = 1.69, p = 0.095; Fig. 7D] and displayed faster reward collections [t₍₁₃₁₎ = 3.65, p < 0.001; Fig. 7E]. Hence, there were minimal IGT performance differences between HIV-1Tg and control animals at baseline.

3.5. Acute THC administration increased risky decision-making selectively in HIV-1 transgenic rats

The impact of acute THC on IGT performance is illustrated in Figs. 8 and 9. A total of 29 animals were excluded from these analyses based on response biases (n = 10) or low trial counts (n = 19). Females had higher %risky-win stay trials than males [F_(1,97) = 5.81, p = 0.018]. No other main or interactive effects of sex (Fs < 2, ps > 0.1) were observed on the main outcome variables of difference score or decision-making metrics, therefore data were collapsed across males and females animals. A main effect of genotype [F_(1,107) = 5.307, p = 0.023] revealed HIV-1Tg animals exhibited reduced difference scores compared to controls, indicative of increased risk-based decision-making (Fig. 8A). Given a priori predictions that THC would increase risk-based decision-making in HIV-1Tg rats the genotype*drug interaction [F_(2,107) = 0.908, p = 0.407] was explored despite lack of significance. HIV-1Tg rats had lower difference scores compared to controls only when administered 3 mg/kg THC (p = 0.022). Further, 3 mg/kg THC tended to reduce difference scores (increase risk preference) compared to VEH treatment, selectively within HIV-1Tg animals (p = 0.085).

Fig. 8.

Acute THC increased risk-based decision-making selectively in HIV-1Tg rats, likely by reducing safe-win stay decisions. HIV-1Tg exhibited poorer risk-based decision-making compared to controls, as indicated by lower difference scores. Exploratory analyses based on a priori hypotheses revealed high-dose THC increased risk-based decision-making, only in HIV-1Tg rats (A). High-dose THC also dose-dependently lowered %safe-win stay trials, only in HIV-1Tg rats (B). Compared to controls, HIV-1Tg rats has similar %safe-lose shift trials (C) and %risky-win stay trials (D), but lower %risky-lose shift trials. Data presented as individual data-points, plus mean ± S.E.M. Circles and triangles represent female and male data points, respectively. **p < 0.01, *p < 0.05, #p < 0.1 as indicated.

Fig. 9.

Acute THC effects on secondary performance outcomes in the IGT were consistent across HIV-1Tg rats and controls. Compared to controls, HIV-1Tg rats completed less choice trials (A), had higher %omission trials (B) but showed no differences in %premature trials (C). High-dose THC decreased choice trials across all animals (A). HIV-1Tg rats had slower choice latencies compared to controls, with high-dose THC dose-dependently increasing choice latencies across all animals (D). HIV-1Tg rats had faster reward collection latencies than controls, with high-dose THC decreasing, and low-dose THC increasing, reward collection times across all animals (E). Data presented as individual data-points, plus mean ± S.E.M. Circles and triangles represent female and male data points, respectively. ***p < 0.001, **p < 0.01, *p < 0.05, as indicated.

The increased risk-based decision-making by 3 mg/kg THC in HIV-1Tg rats was likely driven in part by a trending interaction of drug*genotype [F_(2,105) = 2.92, p = 0.059] and trending main effect of drug [F_(2,105) = 2.98, p = 0.055] on %safe-win stay trials (Fig. 8B). The interaction revealed that HIV-1Tg rats had lower %safe-win stay trials compared to controls only when administered 3 mg/kg THC (p = 0.013) and that 3 mg/kg THC reduced %safe-win stay trials compared to VEH (p = 0.01) and 0.3 mg/kg THC (p < 0.1), selectively in HIV-1Tg rats. HIV-1Tg rats also tended to have lower %risky-lose shift trials (Fig. 8E) than controls [gene: F_(1,104) = 2.99, p = 0.086], suggestive of reduced sensitivity to high punishment, and possibly a contributing factor to lower differences scores observed. No other main effects of drug or genotype, nor their interactions were observed on %safe-lose shift (Fig. 8C) or %risky-win stay (Fig. 8D) trials.

Secondary outcome measures of IGT performance during the acute THC testing period are illustrated in Fig. 9. Consistent with baseline testing, females continued to have higher %premature responses [F_(1,101) = 5.24, p = 0.024], less %omissions [F_(1,101) = 7.15, p = 0.009], and faster choice [F_(1,101) = 5.99, p = 0.017] but slower reward [F_(1,101) = 19.02, p < 0.001] latencies. Sex did not interact with gene nor drug for any of these measures (Fs < 1, ps > 0.1), hence, as with primary outcome variables, data were collapsed and examined by genotype. HIV-1Tg rats completed less choice trials [F_(1,107) = 10.11, p = 0.002; Fig. 9A] and had higher %omission trials [F_(1,107) = 5.13, p = 0.026; Fig. 9B] than controls, but no differences in %premature trials (Fig. 9C). Consistent with baseline responding, HIV-1Tg rats had slower choice [F_(1,107) = 7.40, p = 0.008; Fig. 9D] but quicker reward collection [F_(1,107) = 19.99, p < 0.001; Fig. 9E] latencies.

3 mg/kg THC reduced trial counts [F_(2,107) = 3.88, p = 0.024; Fig. 9A] compared to VEH administration (p = 0.029) and dose-dependently slowed choice latencies [F_(2,107) = 8.21, p < 0.001; Fig. 9D] compared to VEH (p < 0.001) and 0.3 mg/kg THC (p = 0.002). Despite 3 mg/kg THC-induced slowed choice latencies, this dose sped reward collection latencies compared to VEH (p = 0.007) and 0.3 mg/kg (p < 0.001), while 0.3 mg/kg THC instead slowed reward latencies compared to VEH (p = 0.004; Fig. 9E). A trending effect of drug on % premature responses [F_(2,107) = 2.52, p = 0.085] was observed but no group differences withstood Bonferroni corrections. (Fig. 9C).

3.6. Chronic THC administration increased risk-based decision-making selectively in HIV-1 transgenic rats

The impact of chronic THC on IGT performance is illustrated in Figs. 10 and 11. A total of 10 animals were excluded from these analyses based on response biases (n = 8) or low trial counts (n = 2). Additionally, one male transgenic animal (0.3 mg/kg THC group) fell sick and was removed from the study prior to testing. Sex produced main and interactive effects on the main outcome variables (difference score and decision-making metrics) therefore data was not collapsed across males and females animals except when exploring the a priori hypothesis on difference score. A main effect of genotype [F_(1,119) = 4.44, p = 0.037] revealed HIV-1Tg animals had reduced difference scores compared to controls, indicative of increased risk-based decision-making (Fig. 10A). Given a priori predictions that THC would negatively impact risk-based decision-making in HIV-1Tg rats, the genotype*drug interaction [F_(2,119) = 1.95, p = 0.147] was explored. HIV-1Tgs had lower difference scores compared to controls only after chronic administration of 3 mg/kg THC (p = 0.04), and a similar trend effect observed with 0.3 mg/kg THC (p = 0.089). Chronic 3 mg/kg THC administration also decreased difference scores compared to VEH in HIV-1Tg rats (p = 0.034), but not in controls (p = 0.785). These effects were driven by females, as revealed by sex*drug*genotype [F_(2,119) = 4.96, p = 0.008] and sex*drug [F_(2,119) = 4.61, p = 0.012] interactions. However, group comparisons revealed sex-differences were only observed in VEH treated HIV-1Tg rats, with females having higher difference scores compared to males (Supplementary Fig. 3A). Since 1) no sex*drug*gene effects were observed on any other measure at any other test, and 2) to avoid over-interpretation of sex-dependent effects of treatment (Pape et al., 2024), further reporting and illustration of the influence of sex can be found in Supplemental Fig. 3. Subsequent analyses continued with data collapsed across the sexes.

Fig. 10.

Chronic THC increased risk-based decision-making selectively in HIV-1Tg rats. HIV-1Tg exhibited poorer risk-based decision-making compared to controls, as indicated by lower difference scores (A). Exploratory analyses based on a priori hypotheses revealed high-dose THC worsened risk-based decision-making, only in HIV-1Tg rats. Compared to controls, HIV-1Tg rats had similar %safe-win stay trials (B) and %safe-lose shift trials (C), but higher %risky-win stay trials (D) and a tendency for reduced %risky-lose shift trials (E). Data presented as individual data-points, plus mean ± S.E.M. Circles and triangles represent female and male data points, respectively. **p < 0.01, *p < 0.05, #p < 0.1 as indicated.

Fig. 11.

Chronic THC effects on secondary performance outcomes in the IGT were consistent across HIV-1Tg and controls. Compared to controls, HIV-1Tg rats completed less choice trials (A), and high-dose THC reduced trials relative to VEH treatment across genes. HIV-1Tg rats completed a similar number of %omission trials (B) and %premature trials (C) as controls. HIV-1Tg rats had slower choice latencies compared to controls (D), and high-dose THC slowed choice latencies relative to low-dose THC and VEH treatment across genes. HIV-1Tg rats had faster reward collection latencies compared to controls, with low-dose THC increasing reward collection times compared to high-dose THC across genes (E). Data presented as individual data-points, plus mean ± S.E.M. Circles and triangles represent female and male data points, respectively. ***p < 0.001, *p < 0.05, #p < 0.1, as indicated.

Decision-making metrics are illustrated in Fig. 10B–E. HIV-1Tg rats had lower %safe-win stay trials [F_(1,119) = 4.79, p = 0.031; Fig. 10B] but similar %safe-lose shift trials (ns; Fig. 10C) as controls. Instead, HIV-1Tg rats had higher %risky-win stay trials [F_(1,120) = 4.57, p = 0.035; Fig. 10D] and also tended to have higher %risky-lose shift trials [F_(1,122) = 3.809, p = 0.053; Fig. 10E] when compared to controls. Sex interacted with drug and gene to alter %safe-win stay and %risky-lose shift, but not other decision-making metrics, which are further discussed and illustrated in Supplementary Fig. 3.

On secondary measures HIV-1Tg rats had lower choice trials [F_(1,125) = 4.216, p = 0.042; Fig. 11A] than controls and 3 mg/kg THC reduced trial count [F_(2,125) = 3.373, p = 0.037] compared to VEH treatment (p = 0.033) across both genes. There were no effects or interactions on % omission (Fig. 11B) or %premature (Fig. 11C) trials. HIV-1Tg rats had slower choice latencies [F_(1,125) = 8.487, p = 0.004; Fig. 11D] than controls and 3 mg/kg THC reduced choice latencies [F_(2,125) = 9.817, p < 0.001] compared to 0.3 mg/kg THC (p < 0.01) and VEH treatment (p < 0.001) across genes. HIV-1Tg rats had quicker reward collection latencies [F_(1,125) = 19.106, p < 0.001; Fig. 11E] than controls and a trending drug effect [F_(1,125) = 2.37, p = 0.098] revealed 3 mg/kg THC tended to speed reward collections relative to 0.3 mg/kg THC (p = 0.094) across genes. Sex interacted with drug and/or gene on several secondary outcome measures which are discussed in Supplemental Fig. 4.

4. Discussion

Here, HIV-1Tg rats were tested in two cross-species translatable tasks – the Iowa Gambling Task and the Probabilistic Reversal Learning Task - at baseline and following acute and chronic THC exposures. At baseline, HIV-1Tg rats exhibited notable differences in choice and reward collection latencies, suggestive of possible altered information- and reward-processes. Aligning with the existing literature on the function-dependent cognitive impacts of cannabis use in PLWH (Ayoub et al., 2024), THC enhanced learning performance in the HIV-1Tg rat, while simultaneously resulting in worsened risk-based decision making in the IGT. In most cases, THC effects specific to the HIV-1Tg rat were only observed on main, rather than secondary, outcome measures. Hence, the selective effects of THC in HIV-1Tg rats suggest a complex interaction with the specific cognitive domains targeted by the IGT and PRLT, rather than reflecting differential drug-induced processing or motivational shifts among the genetic groups. Unexpectedly, THC did not appear to affect cognitive flexibility in HIV-1Tg rats, diverging from clinical studies that have demonstrated beneficial associations between cannabis use and executive function in PLWH (Flannery et al., 2022; Kallianpur et al., 2020; Chang et al., 2006; Dastgheyb et al., 2021; Rogers et al., 2023; Wang et al., 2020). Instead, chronic THC non-significantly reduced cognitive flexibility among control animals, a phenomenon not observed in HIV-1Tg rats. Together, these data provide the first preliminary evidence that THC produces function-dependent effects on cognition in the HIV-1Tg rat model using cross-species translatable tasks poorly performed by PLWH in the clinic. These data are consistent with the function-dependent effects of cannabis use on cognition noted in PLWH (Ayoub et al., 2024), thereby suggesting THC may drive cannabis effects on cognition in PLWH, resulting in improved or deficient cognitive ability, dependent on the function measured.

This study revealed HIV-1Tg rats did not exhibit baseline performance deficits in primary outcome measures of the IGT or PRLT relative to control animals. While this finding was somewhat surprising, HIV-associated NCI is heterogeneous, and cognitive performance deficits are not always observed within every population of PLWH (Iudicello et al., 2013; Nigro et al., 2021; Vassileva et al., 2013). Nonetheless, in this study HIV-1Tg rats demonstrated explicit differences in secondary performance measures across tasks relevant to early-stages of NCI. For example, HIV-1Tg rats consistently took longer to choose an option in the IGT and the PRLT. While this slowed choice could indicate potential motoric differences, HIV-1Tg rats also demonstrated quicker reward collection latencies across all three IGT tests. Further, previous reports from our laboratory using this model have determined intact gross locomotor responding (Roberts et al., 2021). The longer choice latencies may instead indicate a speed-accuracy trade-off in the HIV-1Tg model, necessary for their ‘normal’ baseline cognitive performances, as observed in PLWH performing cognitive tasks (Karlsen et al., 1992) and in HIV-1Tg rats performing the PRLT (Roberts et al., 2021). At the neural level, when compared to healthy participants, increased cognition-dependent brain activation has been observed in PLWH alongside similar cognitive performance levels (Chang et al., 2001; Ernst et al., 2002), suggesting the increased activation of these systems may compensate for their reduced efficacy. Interestingly, this pattern of activation changes as cognitive deficits advance, as PLWH diagnosed with more progressive NCI have instead exhibited reduced cognition-dependent brain activation during cognitive task performance (Tucker et al., 2004). Taken together, our data suggest a speed-accuracy trade off in HIV-1Tg rats is necessary to maintain intact cognitive performance in the cross-species IGT and PRLT. The quicker reward collection in HIV-1Tg rats may instead indicate altered reward processing, consistent with previous reports (Bertrand et al., 2018; McIntosh et al., 2015), warranting further future research with use of available cross-species compatible tools for measuring positive valence systems (Dexter et al., 2025). Future work should also prioritize testing HIV-1Tg rats at an older age, following further exposure to viral proteins, to assess if this may render baseline IGT and PRLT deficits.

In the clinic, PLWH often demonstrate learning and executive function deficits across a variety of tasks (Kanmogne et al., 2018; Walker and Brown, 2018; Kanmogne et al., 2020; Wang et al., 2015; Martin et al., 2019; Hardy et al., 2021; Jiang et al., 2016) and several studies have reported beneficial associations between cannabis use and learning performance (Watson et al., 2023; Watson et al., 2020; Watson et al., 2021) and executive function (Flannery et al., 2022; Kallianpur et al., 2020; Rogers et al., 2023; Wang et al., 2020) in this population. Based on this literature, we predicted that THC exposure would enhance learning performance and executive function in the HIV-1Tg rat as measured by the PRLT. We have previously tested HIV-1Tg rats in the PRLT and observed enhanced initial learning and normal reversal learning performance (Roberts et al., 2021). Here, in a larger sample size, both initial and reversal learning in HIV-1Tg rats were comparable to controls at baseline assessment. As predicted however, acute and chronic THC exposures dose-dependently improved initial probabilistic reversal learning in HIV-1Tg rats, without altering task performance in controls. No effects of THC were observed on cognitive flexibility (reversal learning) in HIV-1Tg rats following either acute or chronic THC exposure. Interestingly, chronic THC tended to impaire reversal learning in controls, but not HIV-1Tg rats, suggesting that the HIV-1 transgene may provide protective effects that render rats immune to the harmful impact of THC on executive function, as we previously reported in PPI (Roberts et al., 2021). These findings also align with studies linking cannabis use to negative cognitive outcomes and related brain activity in healthy individuals, but not in PLWH (Hall et al., 2021; Flannery et al., 2021; Okafor et al., 2019).

The beneficial effects of THC on learning function in the HIV-1Tg rat may be mediated by the neuroprotective properties of cannabinoids, including their ability to inhibit excitotoxicity and reduce inflammatory processes. Research consistently shows that cannabis use is associated with lower levels of proinflammatory markers in PLWH (Watson et al., 2021; Kallianpur et al., 2020; Yin et al., 2022; Manuzak et al., 2018) and unlike other biomarkers of HIV progression (e.g., CD4 counts), proinflammatory markers are often linked to cognitive deficits (Watson et al., 2021; Burlacu et al., 2020; Portilla et al., 2019). Additionally, lower levels of neuroinflammatory markers, including MCP-1 and IP-10, have been connected to improved learning performance in PLWH, and within the same population cannabis use was related to enhanced learning abilities (Watson et al., 2021). Indeed, THC is capable of reducing inflammation in vivo and in vitro (Kinsey and Cole, 2013; Fitzpatrick et al., 2020), and the endocannabinoid (eCB) system neurotransmitters anandamide (AEA) and 2-arachidonyl glycerol (2-AG), which bind to the same receptors, reduce HIV-induced neuroinflammation and associated damage (Esposito et al., 2002; Xu et al., 2017). Hence, the THC-induced learning enhancements observed in HIV-1Tg rats may be caused, in-part, by the known anti-inflammatory properties of cannabinoids. The reduction of MCP-1 in cannabis using PLWH is particularly noteworthy since this molecule facilitates HIV-infected monocytes migration into the central nervous system (Eugenin et al., 2006), and is associated with NCI in HIV, aging, and Alzheimer’s populations HIV (Anderson et al., 2020; Sanchez-Sanchez et al., 2022; Lee et al., 2018). It would be interesting for future studies to test the impact of THC, and other anti-inflammatory compounds, on markers of neuroinflammation in the HIV-1Tg rat model. Such studies could be conducted longitudinally using the same PET ligand used to measure neuroinflammation in PLWH (Brody et al.) and rodents (Young et al., 2022).

Compared to healthy control participants, PLWH perform worse on the IGT in clinical studies, indicative of worse risk-based decision making processes (Hardy et al., 2006; Martin et al., 2004; Iudicello et al., 2013; Nakao et al., 2020; Nigro et al., 2021). The poorer IGT performance of PLWH has been observed alongside NCI diagnoses (Iudicello et al., 2013; Nakao et al., 2020) and worse performance on tests of cognitive flexibility (Iudicello et al., 2013), inhibitory control, and verbal memory (Hardy et al., 2006). Hence, IGT deficits in PLWH may indicate NCI more broadly, while also providing a useful diagnostic tool for detecting risk-based decision-making deficits in NCI-diagnosed PLWH. To our knowledge we are the first group to investigate the performance of any rodent neuroHIV model in the cross-species IGT. We confirmed similar learning profiles in our rodent IGT as seen in humans, with performance improving across time, thereby validating the translational value of the IGT. While HIV-1Tg rats had similar baseline IGT performance as controls broader performance deficits were observed following acute or chronic THC exposure. More specifically, acute 3 mg/kg THC increased risky decision-making, while also reducing safe-winstay behavior. In contrast however, chronic administration of either the lower (0.3 mg/kg) or higher (0.3 mg/kg) dose of THC increased risky decision-making. These data are aligned with cannabis use correlating with gambling disorder and self-reported impulsivity (Langan et al., 2019), while also predicting risky decision-making deficits in PLWH (Gomez et al., 2017). Hence, our findings further support the potential adverse impact of cannabis use on risk-based decision-making in HIV, and implicate THC in underlying these effects. Further, this work suggests the HIV-1Tg rat model can be used to further probe the underlying neurobiology that may contribute to altered risk-based decision making in cannabis using PLWH.

It is possible that the differential THC-induced effects observed across HIV-1Tg rats and their controls are a result of differences in eCB system functioning observed in HIV. For example, eCB receptors and the ligands that bind to them, including the eCBs AEA and 2-AG, appear altered in the brains of PLWH and neuroHIV rodent models (Yadav-Samudrala et al., 2024; Swinton et al., 2021; Cosenza-Nashat et al., 2011). More specifically, cannabinoid receptor expression levels are elevated in HIV encephalitis (Cosenza-Nashat et al., 2011) in NCI-diagnosed PLWH (Swinton et al., 2021), and in transgenic female mice expressing the HIV-1 Tat protein (Yadav-Samudrala et al., 2024). Importantly, these studies implicate that HIV-related eCB system changes occur in a region-dependent fashion, which may help to explain the function-dependent effects of cannabinoids on HIV-associated NCI. Indeed, cannabinoid-1 receptor expression was increased in white and gray matter of NCI diagnosed PLWH, whereas cannabinoid-2 receptor expression was decreased solely in gray matter regions. Limited data on eCB changes in HIV exist however, and further clarification on these effects may lead to a more mechanistic understanding of how cannabinoids interact with HIV to produce both beneficial and adverse effects on cognition. We have preliminary findings that THC induced similar physiological effects across HIV-1Tg rats and controls in the cannabinoid tetrad test (Ayoub et al., 2023), a commonly used screening tool for cannabinoid-receptor functionality. Hence, there is reason to believe that eCB receptor function, and THC-induced physiological responses to THC, are preserved despite transgene expression.

Here, sex-specific effects were not predicted given the relative lack of support for such in the literature of cannabinoid and HIV effects on cognition (Ayoub et al., 2024). Consistently, sex-specific effects of drug and gene were only observed at one of six testing time-points. More specifically, the adverse impact of chronic THC on IGT performance in HIV-1Tg rats appeared to be driven by females. These findings suggest sex may moderate the impact of cannabinoid effects in HIV. Somewhat similarly, we previously observed heightened risk-taking, and exacerbated responses to dopamine-2 receptor activation on risk-taking in female rats using a different (non-cross species) task (Ayoub et al., 2024). Hence, females may be more susceptible to drug-induced effects on risk-taking behavior and should continue to be explored further in clinical and preclinical studies. Further, some sex-differences have been observed in both the magnitude of NCI among PLWH, and in cannabis-induced effects (Dreyer et al., 2022; Cuttler et al., 2016; Matheson et al., 2020), therefore the inclusion of sex as a factor in subsequent studies remains vital to understanding its possible interactive effects with cannabinoid exposure, and HIV. Gender may also modulate the experience of cannabinoids exposure (Matheson and Le Foll, 2023), however gender cannot be measured in animals, limiting their use in exploring its impact on cannabinoid-induced effects in the context of HIV-associated NCI.

This study has provided the first preliminary evidence for function-dependent effects of THC in an animal model of neuroHIV¸ consistent with clinical data that details the impact of cannabis use on cognition in PLWH. Indeed, within the same cohort of PLWH, recent cannabis use was associated with weaker resting-state functional connectivity and more physical HIV symptoms, while concomitantly being linked to larger regional brain volumes, lower proinflammatory markers, and better executive functioning (Kallianpur et al., 2020). Similarly, we demonstrate here, that within the same cohort of HIV-1Tg rats, THC worsened risky decision-making, while enhancing learning outcomes. These data are critical for understanding the impact of high cannabis use in PLWH, as well as the potential therapeutic value of THC for HIV-associated NCI.

While our findings are promising, they must be interpreted alongside potential limitations of the study. For example, THC-specific administration is a strength, knowing whether effects are THC-specific, but they only partially inform cannabis-induced effects as seen in humans. The cannabis plant contains over 500 individual compounds, including other phytocannabinoids (e.g., cannabidiol; CBD) and terpenes (e.g., myrcene; (Rock and Parker, 2021)), with differences in pharmacological profiles observed between many of these compounds. Most of our THC-induced function-dependent findings in HIV-1Tg rats were consistent with the cannabis-induced function-dependent findings in PLWH (Ayoub et al., 2024), some however, were not (e.g., the tendency for cannabis to improve cognitive flexibility in PLWH). Therefore, these data can only be used to interpret the contributions of THC to cannabis-induced effects in PLWH and explore the potential of THC to alter cognitive performance in PLWH. Future work should explore of the effects produced by other independent cannabis ingredients (e.g., CBD), or a combination of cannabis compounds (e.g., THC + CBD), on function-dependent outcomes in the HIV-1Tg rat model to better interpret cannabis use findings in PLWH, and determine therapeutic avenues for HIV-associated NCI.

Another limitation of the current study that requires exploration in future studies is the route of administration used for THC exposure, which is not relevant to methods used by human cannabis users (i.e., inhalation, oral consumption). Indeed, these routes produce independent pharmacokinetic profiles including differences in peak brain and plasma concentrations, as well as differences in drug clearance (Baglot et al., 2021; Hlozek et al., 2017). Our approach was a necessary first-step, however, as the use of injection protocols enables precise dosing consistent across all animals, also important for interpretation of drug effects. Nonetheless, future studies should replicate our observed effects using more translationally-relevant THC delivery methods.

Finally, while HIV-1Tg rats are positioned as a model of modern HIV-infection undergoing cART, this strain does not express 2 of the 9 key viral proteins that have still been found in brain reservoirs of cART-suppressed PLWH (Tang et al., 2023). As HIV is inherently a human disease, each animal model attempting to re-create its expression is limited in its direct translational value to the clinic (for a comprehensive overview of the advantages and disadvantages of each model see (Ayoub et al., 2024). While humanized mouse models exist for HIV in immunodeficient mice that express the full HIV genome, these models are met with poor longevity, transmission, and testing concerns, which limit their use in lengthier longitudinal cross-species studies such as the current report. HIV-1Tg rats instead enable this longevity testing and by lacking the viral components required for HIV transmission and replication, this genetic strain is conceptualized as an appropriate model for studying modern HIV infection wherein PLWH have suppressed viral loads. Further, the HIV-1Tg rat recreated cannabinoid-induced function-specific effects, consistent with those observed in PLWH who use cannabis, suggesting them as an appropriate model for further probing drug-induced effects on cognitive outcomes.

5. Conclusions

Here, THC produced beneficial effects on learning (PRLT), but adverse effects on risk-based decision making (IGT) in HIV-1Tg rats using cross-species tasks with direct relevance to testing in clinical populations. These findings further support function-dependent effects of cannabis use on cognitive function in HIV, and suggest that THC may be a key contributing cannabinoid to such effects. Our study also suggests that the HIV-1Tg rat provides a suitable model for exploring function-dependent cannabinoid effects in the context of HIV, and can be further used to probe underlying neural mechanisms. These findings may serve as guidance for clinicians prescribing cannabis to patients with HIV, and for further research exploring the interactive effects of HIV and cannabis on cognitive function.

Supplementary Material

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.bbi.2025.04.030.

Data availability

Data will be made available on request.