Cornichon Homolog-3 (Cnih3) deletion impairs spatial memory, operant learning, and fentanyl self-administration behavior

Abstract

Opioid misuse remains rampant as new synthetic opioids reach the market. Large-scale genetic tools like the GWAS identify previously unrecognized targets and biomarkers in opioid misuse with the hopes of combating the opioid epidemic. One such target is the α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor (AMPAR) auxiliary protein Cornichon Homolog-3 (human analog: CNIH3, mouse analog: Cnih3), which determines AMPAR subunit composition and kinetics: important factors in opioid use. Though CNIH3 was identified as a gene of interest in OUD, its role in opioid use and related behavior has not been studied. Using mice with Cnih3 deletion, we characterize the role of Cnih3 in a battery of behaviors that encompass well-being, affect, spatial and social memory, operant learning, reversal learning, and opioid use itself. We find that Cnih3 deletion moderately impairs spatial memory in the novel object recognition task, as well as operant learning in a sucrose self-administration paradigm. Cnih3 deletion also delays acquisition of fentanyl IVSA in females and blunts fentanyl intake during IVSA in both sexes. We use principal component analysis to pinpoint the dimensions in which Cnih3 deletion impacts the behaviors tested in an unbiased manner. These findings, combined with initial GWAS findings, identify Cnih3 as a behaviorally relevant protein in opioid-related plasticity.

Introduction

In 2024, 2.7% of people 12 or older reported misusing opioids in the last year [1]. Frequent use or misuse of opioids can lead to opioid use disorder (OUD), which is characterized by repeated drug-taking, drug craving during attempted abstinence, and lastly, relapse [2]. While a decade ago, prescription opioids were the major contributors to opioid misuse, highly potent and readily available synthetic opioids, such as fentanyl, have exacerbated overdose deaths in the US [3]. Uncovering why certain populations are predisposed to or protected against opioid misuse and OUD is critical for overdose prevention as more potent opioids continue to flood the market.

OUD is heritable, with 50–70% of variation due to segregating loci [4]. The genome-wide association study (GWAS) is a valuable tool that can identify common variants that contribute to OUD heritability. There have been numerous large-scale GWAS of OUD (e.g., [5–8]), and more recently, these GWAS have utilized control individuals with a history of some opioid exposure. Comparing OUD cases to such exposed controls ensures that identified variants specifically relate to opioid addiction rather than behavioral attributes that might contribute to using opioids initially. However, in a majority of these GWAS, exposure was minimally defined. One prior GWAS utilized a novel design in which a small group of individuals with opioid dependence who injected daily (n = 1167) was contrasted with neighborhood controls, as well as a small cohort of opioid-dependent individuals who never progressed to daily injection use. Further controlling for environmental factors, controls were recruited from the vicinity of the methadone clinics where cases were recruited. This GWAS identified variants in the Cornichon homolog 3 (human analog: CNIH3; mouse analog: Cnih3) gene [9]. CNIH3 is detected in all brain regions, with high expression in the cerebral cortex, hippocampal formation, and hypothalamus [10, 11], and functions as an α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor (AMPAR) auxiliary protein important for glutamatergic plasticity – a critical player in maladaptive drug use [12–14]. CNIH3 regulates synaptic AMPAR expression and subunit composition necessary for long-term potentiation (LTP) by increasing surface expression of GluA1 and enhancing AMPAR kinetics and glutamatergic transmission [15–17]. This form of plasticity is critical for several processes contributing to maladaptive opioid use, including learning, memory, stress-coping, nociception, and affect [12, 18–20] with alterations found in human and animal models of major depressive disorder [21–24], anxiety [25, 26], stress [19, 27], and pain [28, 29]. Therefore, SNPs in CNIH3 may protect against OUD by disrupting plasticity that contributes to any of the above-mentioned underlying processes. That these variants arose when case status involved a severe form of OUD, and controls were individuals engaged in problematic but not severe OUD, suggests the specificity of CNIH3, and given that CNIH3 determines AMPAR composition and trafficking, CNIH3 may influence maladaptive opioid use directly and/or indirectly.

Glutamate transmission, which is supported by CNIH3, is also necessary for the development, maintenance, and recall of opioid-associated memories [12, 30, 31]. Alterations in CNIH3 may facilitate maladaptive opioid use by impacting glutamatergic plasticity that underlies the acquisition, expression, or longevity of memory, including opioid-associated memory. Increased GluA1-containing AMPARs facilitated opioid-associated contextual memory [32–34], and altering AMPAR subunit composition reduced opioid seeking [35–39]. Furthermore, in previous studies, we overexpressed Cnih3 in the hippocampus and found that this enhanced the performance of female mice in the Barnes Maze spatial memory task [10]. Our previous work also shows that female mice with a Cnih3 deletion exhibited deficits in performance on the Barnes Maze spatial memory task, as well as decreased synaptic connectivity, altered AMPAR subunit composition, and impaired LTP maintenance [10]. Based on this evidence, CNIH3 may influence the learning and memory required for the development and progression of OUD.

Given the GWAS identification of SNPs in CNIH3 protecting against OUD, we hypothesize that CNIH3 plays an essential role in the AMPAR trafficking necessary for opioid-associated learning. To test this, we use a Cnih3 knockout (KO) mouse model to characterize the role of Cnih3 in behavior (well-being, anxiety- and depression-like behavior, spatial memory, sociability, nociception), natural reward-seeking and cognitive flexibility, and last, in instrumental opioid self-administration. We found that Cnih3 deletion impairs spatial memory, operant learning, and fentanyl self-administration and reinstatement in a sex specific manner. These experiments are the first to assess the role of Cnih3 in opioid use and behavior, which seems to act through facilitating reward-associated memory formation. These results lay the foundation of CNIH3 as a behaviorally relevant component of opioid-related plasticity.

Results

Cnih3 deletion does not significantly impact well-being, motor coordination, or baseline nociception

To fully assess the potential role of CNIH3 in opioid use and relevant behaviors, we employ a preclinical mouse model using a global Cnih3 deletion [10]. First, we conducted a battery of behavioral tests to assess the role of Cnih3 in overall well-being, motor coordination, and nociception, which may impact the ability to perform operant tasks or impact opioid use. First, we measured well-being using the nest-building assay, motor coordination using the accelerating Rotarod, and thermal nociception using the hot plate assay (Fig. 1A). There were no differences between sexes or genotype in untorn nestlet (Fig. 1B; 2-way ANOVA: genotype x sex, F (1, 49) = 0.1816, p = 0.6719), latency to fall from the Rotarod on the first (S1; 2-way ANOVA: genotype x sex, F(1, 49) = 0.7284, p = 0.3976) or last trial (21 C; 2-way ANOVA: genotype x sex, F(1, 49) = 0.005529, p = 0.941), or paw withdrawal latency on the hot plate test (Fig. 1D; 2-way ANOVA: genotype x sex, F(1, 49) = 0.6475, p = 0.4249), indicating that Cnih3 deletion does not have significant baseline effects on well-being, motor coordination, or thermal nociception. Together, these findings demonstrate that Cnih3 does not significantly contribute to innate nesting behavior or sensorimotor function.

Fig. 1. Cnih3 deletion does not significantly impact well-being, motor coordination, or baseline nociception.

A Schematic of nest building, rotarod, and hot plate experiments. B Well-being as measured by percentage of nestlet torn during the 24-hr testing period in male (♂) and female (♀) wildtype (WT) and Cnih3 knockout (KO) mice. C Motor coordination as measured by latency to fall (s) on the last trial of the accelerating rotarod. D Thermal nociception as measured by latency (s) to paw withdrawal on the hot plate set to 54° C. Data are shown as mean ± SEM. Data analyzed with two-way ANOVAs (sex x genotype). Full statistical reporting is available in Table S1. Additional data related to this dataset are available in Figure S1.

Cnih3 deletion impairs spatial memory but does not affect socialization or social memory

Next, given the importance of associative memory in OUD, and prior work showing spatial memory deficits in Cnih3 KO mice [10], we assessed whether Cnih3 KO led to memory impairments. First, we assessed spatial memory using the Novel Object Recognition Test (NORT; Fig. 2A) in which mice were exposed to an open field under different conditions for 10 min/day. The first day (D1), mice were placed in an empty open field to habituate them to the apparatus and examine exploratory behavior (Fig. 2B). On day two (D2), mice returned to the field and were allowed to explore two identical objects. On day 3 (D3), one of the objects was replaced with a novel object, and the latency and total time investigating the novel object were measured as indices of spatial memory (Fig. 2C, D). Interestingly, independent of sex, Cnih3 deletion decreased total investigation time (Fig. 2B; 2-way ANOVA: genotype, F(1, 49) = 4.191, p = 0.046), increased latency to investigate the novel object (Fig. 2C; 2-way ANOVA: genotype, F(1, 49) = 5.232, p = 0.0265), and reduced time investigating the novel object (Fig. 2D; 2-way ANOVA: genotype, F(1, 49) = 9.246, p = 0.0038), suggesting that Cnih3 may be important for spatial memory, consistent with our prior work [10]. Next, we assessed sociability and social memory using the social interaction test (SIT; Fig. 2E), in which mice were exposed to a three-chambered apparatus under different conditions for 10 min/day. On the first day (D1), mice were placed in the empty apparatus (including an empty cup in the left- and right-most chambers) to habituate them to the apparatus and assess exploratory behavior (Fig. 2F). On day two (D2), mice returned to the apparatus and were allowed to explore a cup containing an unknown mouse, M1, or an empty cup to assess socialization (Fig. 2G). On day 3 (D3), the previously empty cup contained a novel mouse, M2, while the familiar mouse (M1) remained on the opposite side. Time investigating M2 was measured as an indication of social memory (Fig. 2H). Cnih3 deletion increased total investigation time in both sexes on D1 (Fig. 2F; 2-way ANOVA: genotype, F(1, 49) = 12.11, p = 0.0011), but there were no group differences in time investigating M1 on D2 (Fig. 2G; 2-way ANOVA: genotype x sex, F(1, 49) = 2.823, p = 0.0993) or time investigating M2 on D3 (Fig. 2H; 2-way ANOVA: genotype x sex, F(1, 49) = 1.159, p = 0.287), indicating that Cnih3 does not significantly impact social behavior or memory. Together, these findings show that Cnih3 deletion impairs spatial memory in the context of external cues, but not social factors.

Fig. 2. Cnih3 deletion impairs spatial memory but does not affect socialization or social memory.

A Schematic of the Novel Object Recognition Test (NORT). D1: Habituation/OFT, D2: Object exposure, D3: NORT. B Exploratory behavior as measured by total investigation time (s) during NORT D1. C Spatial memory as measured by latency (s) to the first investigation of the novel object on NORT D3. D Spatial memory as measured by time investigating (s) the novel object on NORT D3. E Schematic of the Social Interaction Test (SIT). D1: Habituation, D2: Sociability, D3: Social novelty, M1: known mouse, M2: novel mouse. F Exploratory behavior as measured by total distance traveled (m) during SIT D1. G Socialization as measured by time investigating (%) M1 on SIT D2. H Social memory as measured by time investigating (%) M2 on SIT D3. Data are shown as mean ± SEM. Data analyzed with two-way ANOVAs (sex x genotype). Symbols denote ^#main effect of genotype (^#p < 0.05, ^##p < 0.01). Full statistical reporting is available in Table S1.

Cnih3 deletion does not significantly impact affective behaviors

We next assessed how Cnih3 deletion impacted affective behaviors, which are common risk factors for OUD [40–42]. We assessed anxiety-like behavior using the Elevated Plus Maze (EPM; Fig. 3A) in which mice were placed in the center of the 4-armed apparatus (2 open, anxiogenic arms; 2 enclosed, anxiolytic arms) and allowed to explore for 10 min. We observed no differences between groups in exploratory behavior measured by total distance traveled in the maze (Fig. 3B; 2-way ANOVA: genotype x sex, F(1, 49) = 0.5360, p = 0.4676) or anxiety-like behavior measured by time spent (Fig. 3C; 2-way ANOVA: genotype x sex, F(1, 49) = 1.417, p = 0.2396) and distance traveled (Fig. 3D; 2-way ANOVA: genotype x sex, F (1, 49) = 1.480, p = 0.2296) in the open arms. We also probed anxiety-like behavior and exploration using the Open Field Test (OFT; Fig. 3E) in which mice freely explored a square apparatus for 10 min. Similar to the EPM, we observed no differences between groups in total distance traveled in the apparatus (Fig. 3F; 2-way ANOVA: genotype x sex, F (1, 41) = 2.308, p = 0.1364) or anxiety-like behavior indicated by time spent (Fig. 3G; 2-way ANOVA: genotype x sex, F (1, 41) = 1.156, p = 0.2887) and distance traveled (Fig. 3H; 2-way ANOVA: genotype x sex, F (1, 41) = 2.069, p = 0.1579) in the anxiogenic, center zone. Last, we measured learned helplessness behavior, a model of depression-like behavior, using the Tail Suspension Test (TST; Fig. 3I) in which mice were suspended by their tail for 6 min and latency to the first immobile episode (Fig. 3J), time spent immobile (Fig. 3K), and the number of immobile episodes (Fig. 3L) were measured. We observed sex differences in latency to the first immobile episode, with females taking longer to reach immobility (Fig. 3J; 2-way ANOVA: sex, F (1, 29) = 7.538, p = 0.0103), but no genotype differences. We saw no sex or genotype effects on time spent immobile (Fig. 3K; 2-way ANOVA: genotype x sex, F (1, 29) = 0.6842, p = 0.4149) or number of immobile episodes (Fig. 3L; 2-way ANOVA: genotype x sex, F (1, 29) = 0.4734, p = 0.4969). Together, these results indicate that Cnih3 deletion does not impact anxiety-like or depression-like behavior, suggesting that Cnih3 does not significantly modulate these affective behaviors.

Fig. 3. Cnih3 deletion does not significantly impact affective behaviors.

A Schematic of the Elevated Plus Maze (EPM) experiment. B Exploratory behavior during EPM as measured by total distance traveled (m) during EPM. C and D Anxiety-like behavior as measured by C. time spent in the open arm (%) and D. distance traveled in open arm (%) during EPM. E Schematic of the Open Field Test (OFT) experiment. F Exploration as measured by total distance traveled (m) during OFT. G and H Anxiety-like behavior as measured by G. time in the center zone (%) and H. distance traveled in the center zone (%) of OFT. I Schematic of the Tail Suspension Test (TST). J, K and L Depression-like behavior as measured by (J). latency to the first immobile episode, (K). total time immobile (s) and (L). number of immobile episodes during the TST. Data are shown as mean ± SEM. Data analyzed with two-way ANOVAs (sex x genotype). Symbols denote ^&main effects of sex in J (^&p < 0.05). Full statistical reporting is available in Table S1. Additional data related to this dataset are available in Figure S2.

Cnih3 deletion impairs acquisition in females but not spatial reversal learning during sucrose self-administration

We demonstrated that Cnih3 deletion impairs spatial memory (Fig. 2). To further probe the effects of Cnih3 deletion on memory and cognitive flexibility, we ran an operant sucrose self-administration and spatial reversal learning task (Fig. 4A). Mice were trained to poke a light-cued nose poke port to receive a chocolate-flavored sucrose pellet during acquisition. Time to reach acquisition (Fig. 4B; 3 consecutive sessions where at least 70% of pokes were on the reward-producing port and the session maximum of 30 pellets were obtained), as well as rewards obtained (Fig. 4E) and nose poke discrimination on the first three days (Fig. 4H) and the last day (Fig. 4K) of acquisition were assessed to probe operant learning. Then, mice underwent reversal, where they poked the uncued port to receive sucrose pellets (Fig. 4C-L), and a second reversal, where they poked the cued port to receive rewards (Fig. 4D-J) to probe cognitive flexibility. During acquisition, there were no genotype differences in time to acquire amongst males, but amongst females, Cnih3 KOs were delayed in comparison to their WT counterparts (Fig. 4B; logrank Mantel-Cox test; males: Chi square: 0.01575, p = 0.9001; females: Chi square: 7.028, p = 0.008), which likely contributed to sex by session interactions in rewards earned (Fig. 4E; 3-way ANOVA, F (1.817, 167.1) = 4.917, p = 0.0105) though not discrimination (Fig. 4H; 3-way ANOVA, session x sex x genotype, F (1.949, 179.3) = 3.049, p = 0.0512) during the first 3 days of this phase, and a sex effect in discrimination on the last day (Fig. 4K; 2-way ANOVA, sex, F (1, 92) = 5.088, p = 0.0265). During reversal, we saw no genotype differences in meeting acquisition criteria in either sex (Fig. 4C; logrank Mantel-Cox test; males: Chi square: 0.4682, p = 0.4938; females: Chi square: 1.176, p = 0.2782) or in rewards earned (Fig. 4F; 3-way ANOVA, session x sex x genotype, F (1.623, 149.3) = 1.316, p = 0.2682) or in discrimination (Fig. 4I; 3-way ANOVA, session x sex x genotype, F (1.511, 139.0) = 0.7001, p = 0.4603), on the first three days or on the last day (Fig. 4L; 2-way ANOVA, sex x genotype, F (1, 92) = 0.7059, p = 0.403). Last, during the second reversal, there were no genotype differences in time to acquire in males or females (Fig. 4D; logrank Mantel-Cox test; males: Chi square: 3.107, p = 0.078; females: Chi square: 1.956, p = 0.1619) or in rewards earned (Fig. 4G; 3-way ANOVA, session x sex x genotype, F (1.928, 136.9) = 0.1146, p = 0.885), but there was a sex by genotype interaction for discrimination (Fig. 4J; 3-way ANOVA, sex x genotype, F (1, 71) = 4.601, p = 0.0354) on the first three days of the second reversal. On the last day of the second reversal, Cnih3 deletion produced increased discrimination in males, with no difference in females (Fig. 4M; 2-way ANOVA, sex x genotype, F (1, 71) = 4.927, p = 0.0296; Šídák’s male WT vs KO, p = 0.0007). Together, these data show that Cnih3 deletion impairs initial operant learning in females, but not reversal learning in either sex.

Fig. 4. Cnih3 deletion impairs acquisition in females but not spatial reversal learning during sucrose self-administration.

A Experimental schematic of the sucrose self-administration/spatial reversal learning assay. B, C, and D. B Operant learning or (C). and (D). cognitive flexibility as measured by survival curves depicting the number of sessions to reach the acquisition criteria in (B). acquisition, (C). reversal, and (D). second reversal in males (top) and females (bottom). E, F, and G. E. Operant learning or (F and G). cognitive flexibility as measured by the number of rewards earned on the first three days of (E). acquisition, (F). reversal, and (G). second reversal. H, I., and J. H. Operant learning or I. and J. cognitive flexibility as measured by discrimination of the reward-producing nose poke (% of pokes) on the first three days of (H). acquisition, (I). reversal, and (J). second reversal. K, L, and M. L. Operant learning or (L. and M). cognitive flexibility as measured by discrimination of the reward-producing nose poke (% of pokes) on the last day of (K). acquisition, (L). reversal, and (M) second reversal. Data are reported as mean ± SEM. Logrank Mantel-Cox tests in B-D, Two-way ANOVA (sex x genotype) in K-M. Symbols denote main effects of ^#genotype (^#p < 0.05) or ^&sex (^& p < 0.05, ^&&p < 0.01) and *Sidak’s multiple comparisons (*p < 0.05, ** p < 0.01, *** p < 0.001). Three-way ANOVA (sex x genotype x session) in E-J. Symbols denote effects of ^&sex x session (^&p < 0.05) and *sex x genotype (*p < 0.05). Full statistical reporting is available in Table S1.

PCA indicates that Cnih3 deletion produces moderate differences in operant learning, and may impact cognitive flexibility, during sucrose self-administration

To obtain an unbiased and comprehensive view of Cnih3 involvement on natural reward-seeking, operant learning, and spatial reversal learning, we ran a principal component analysis (PCA) on the data presented in Fig. 4. We used the PCA to identify strong patterns within this dataset to observe which principal components (PCs) contributed the most variance in the dataset based on genotype, with PC1 and PC2 being the most significant contributors. To visually assess similarity between groups, we plotted PC1 scores on the x-axis and PC2 scores on the y-axis for each mouse (each dot corresponds to one mouse) in Fig. 5A and observed only moderate genotype differences with PC1 and PC2 accounting for 40.94% of the variance (Figure S3B). Next, we assessed group differences in the contributors of dataset variance by comparing PC1 (Fig. 5B) and PC2 (Fig. 5C) loadings between sex and genotype. We found that Cnih3 deletion produced a negative PC1 loading, while WT mice produced a positive PC1 loading, suggesting that the variables comprising PC1 are negatively correlated with WT mice and positively correlated with KO mice (Fig. 5B; 2-way ANOVA; genotype: F (1, 71) = 7.246, p = 0.0089). In contrast, female PC2 loadings were more negative than those of males, independent of genotype (Fig. 5C; 2-way ANOVA; sex: F (1, 71) = 13.53, p = 0.0005), suggesting that the variables in PC2 correlate positively in males and negatively in females. Variables included in the analysis were clustered by their intended measure: operant learning (variables from acquisition), cognitive flexibility, and cognitive inflexibility (variables from reversal and second reversal), (see complete list in and direction of variable loading in Figure S3A), and obtained the absolute value (abs) of the loading of each measure for PC1 (Fig. 5D) and PC2 (Fig. 5E). Loadings of 0.3 (dotted line) and above were considered to have a significant contribution (as in [43]; so as to limit drawing conclusions from low-contributing variables), with values approaching 1 indicating a greater contribution. Finally, the summary for each PC was calculated by averaging the PC loadings (abs) for each variable in the corresponding category. Variable clustering within each PC revealed that the main contributors to genotype-based variance in PC1 corresponded to operant learning and cognitive flexibility (Fig. 5D), while variance in PC2 corresponded mainly to cognitive flexibility (Fig. 5E), suggesting that the moderate genotypic differences are in these traits in each PC. Taken together, these data show that there are moderate differences between genotypes in natural reward-seeking and cognition, most identified in operant learning and cognitive flexibility behaviors. Given the lack of differences found in the reversal learning portion of the sucrose self-administration paradigm, it was surprising that the cognitive flexibility variable cluster was identified as a contributor to genotypic variance. Further research is needed to fully assess the role of Cnih3 in reversal learning and cognitive flexibility, but it is clear that Cnih3 deletion produces a delay in operant learning in females only.

Fig. 5. PCA indicates that Cnih3 deletion produces moderate differences in operant learning, and may impact cognitive flexibility, during sucrose self-administration.

A Loadings plot of PC1 and PC2 for each mouse that has completed the experiments described in Fig. 6. B PC1 loadings for each mouse. C PC2 loadings for each mouse. D PC1 loadings of the variance each described behavior contributes to the dataset. The PC1 summary is the average of the PC loading of all components within a defined type of measure. Loadings above 0.3 are considered significant contributors. E PC2 loadings of the variance each described behavior contributes to the dataset. The PC2 summary is the average of the PC loading of all components within a defined type of measure. Loadings above 0.3 are considered significant contributors. Data are reported as mean ± SEM in B-C. Average loading of each measure for the dataset shown in D-E. Two-way ANOVA (sex x genotype) and Sidak’s multiple comparisons in B and C. Symbols denote a ^#main effect of genotype in B (^##p < 0.01)) ^&sex(^&&&p < 0.0001) in C. Full statistical reporting is available in Table S1. Additional data related to this dataset are available in Figure S3.

Cnih3 deletion produces delays in operant learning, and blunts fentanyl intake during IVSA in both sexes and opioid-seeking during extinction and reinstatement in males only

To assess the impact of Cnih3 deletion on opioid self-administration, mice underwent fentanyl IVSA (Fig. 6A). Briefly, mice were implanted with jugular vein catheters, then underwent 25 2-hour sessions of fentanyl IVSA during which pokes in a light-cued nose port triggered an infusion of fentanyl (1ug/kg/infusion). The unlit port produced no rewards, but responses were logged to assess preference for the reward-producing port. Next, mice underwent 21 2 h sessions of extinction in which cues and fentanyl were removed. Last, mice underwent one 2 h session of cue-induced reinstatement, in which cues, but not the fentanyl reward, were reintroduced to assess drug-seeking behavior. On the first day of fentanyl IVSA, Cnih3 deletion reduced exploratory behavior as assessed by the total number of pokes (Fig. 6B; 2-way ANOVA, genotype, F (1, 69) = 6.699, p = 0.0118). Over the 25 days of IVSA, Cnih3 deletion reduced fentanyl intake (Fig. 6C; 3-way ANOVA, session x sex x genotype, F (10.06, 694.0) = 2.200, p = 0.016. Males: 2-way ANOVA, session x genotype, F (6.510, 188.8) = 2.998, p = 0.0065. Females: 2-way ANOVA, session x genotype, F (8.854, 354.1) = 1.951, p = 0.0452) and reduced number of infusions self-administered (Figure S4A; 3-way ANOVA, session x genotype, F (8.634, 594.3) = 2.054, p = 0.034). Comparison of the average number of infusions between the first and last weeks of fentanyl IVSA showed increases between weeks in WT groups but not KOs (Fig. 6D; 3-way ANOVA, session x genotype, F (1, 69) = 4.635, p = 0.0348; Šídák’s: 1st wk:WT male vs. Last wk:WT male p = 0.0001, 1st wk:WT female vs. Last wk:WT female p = 0.0008, Last wk: WT male vs. Last wk: KO male p = <0.0001). On the last day of fentanyl IVSA, we compared the number of active (A) and inactive (I) pokes per group to assess preference for the active port at the end of training), which revealed a lack of preference for the active port only in male Cnih3 KOs (Fig. 6E; 3-way ANOVA, session x sex x genotype, F (1, 68) = 4.044, p = 0.0483; Šídák’s: A:WT male vs. I:WT male p < 0.0001, A:WT female vs. I:WT female p < 0.0001, A:KO female vs. I:KO female p = 0.0049, A:WT male vs. A:KO male p < 0.0001), though there is no difference in discrimination across fentanyl IVSA (Figure S4B; 3-way ANOVA, session x sex x genotype, F (12.46, 859.4) = 0.9995, p = 0.4483). We next assessed time to reach acquisition criteria (2 consecutive sessions where at least 5 infusions are obtained, 70% or more of the pokes are made on the active port, and rewards earned vary by less than 30%), and Cnih3 deletion impaired operant learning in females, but not in males (Fig. 6F; logrank Mantel-Cox test; males: Chi square: 0.01575, p = 0.9001; females: Chi square: 7.028, p = 0.008). During extinction, Cnih3 KO reduced opioid-seeking, indicated by decreased active nose pokes (Fig. 6G; 3-way ANOVA, session x genotype, F (4.815, 332.3) = 3.892, p = 0.0022), and Cnih3 KOs of both sexes showed no preference for the active lever on the first day (Fig. 6H; 3-way ANOVA, session x sex x genotype, F (1, 69) = 5.086, p = 0.0273; Šídák’s: A:KO male vs. I:KO male p = 0.4577, A:KO female vs. I:KO female p = 0.5126). However, we do not see genotype differences across extinction in discrimination (Figure S4C; 3-way ANOVA, session x sex x genotype, F (11.82, 815.3) = 0.6658, p = 0.7828). Comparing the average number of pokes during the first and last weeks of extinction showed that WTs, but not Cnih3 KOs, significantly reduced responding in the absence of reinforcement (Fig. 6I; collapsed due to absence of sex differences; 2-way ANOVA, session x genotype, F (1, 142) = 5.159, p = 0.0246; Šídák’s: 1^st wk WT vs KO p < 0.0001, WT 1^st wk vs last wk p < 0.0001). Lastly, all groups significantly increased active nose pokes during cue-induced reinstatement except for KO males (Fig. 6J; 3-way ANOVA, genotype, F (1, 69) = 7.622, p < 0.0001; Šídák’s: E:WT male vs. R:WT male p = 0.0003, E:WT female vs. R:WT female p < 0.0001, E:KO female vs. R:KO female p = 0.0021). Taken together, Cnih3 deletion blunts fentanyl self-administration in a sex specific manner. It appears that, though both sexes initially differ in similar ways from their WT conspecifics, female KOs discriminate the active poke on the last day of fentanyl IVSA and poke more during cue-induced reinstatement than the last day of extinction, while male KOs do not. This suggests that after acquiring IVSA, female KOs begin to equal the self-administration behavior of WT mice while male KOs do not.

Fig. 6. Cnih3 deletion produces delays in operant learning, and blunts fentanyl intake during IVSA in both sexes and opioid-seeking during extinction and reinstatement in males only.

A Experimental schematic of fentanyl IVSA experiments. B Exploratory behavior as assessed by total pokes on the first day of IVSA. C Opioid intake (ug/kg) across fentanyl IVSA. D Average number of daily fentanyl infusions during the first and last weeks of fentanyl IVSA. E Active (A) and inactive (I) pokes on the last day of fentanyl IVSA. F Operant learning as depicted by survival curves of the number of sessions to reach acquisition criteria in males (left) and females (right). G Opioid seeking as measured by the number of active nose pokes across extinction. H Opioid seeking as measured by the number of active (A) and inactive (I) pokes on the first day of extinction. I Average number of active nose pokes during the first and last weeks of extinction. J Opioid-seeking during reinstatement as measured by the number of active nose pokes during the last day of extinction (E) and cue-induced reinstatement (R). Data are reported as mean ± SEM. Two-way ANOVA (sex x genotype) and Sidak’s multiple comparisons in B and I. Logrank Mantel-Cox tests in F, Symbols denote * significant difference between curves (**p < 0.01). Repeated measures (RM) three-way ANOVA (sex x genotype x time) and Sidak’s multiple comparisons in (C, D, E, G, H, J). Symbols denote ^#main effect of genotype in (B and G) (^#p < 0.05) and an ^& interaction between sex and session in H (^&p < 0.05). Symbols denote *interactions between sex and genotype in J or sex, genotype, and session in (C, D, E, H, I, and J), (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001). Full statistical reporting is available in Table S1, Additional data available in Figure S4.

Cnih3 deletion produces differences in operant learning and fentanyl intake during self-administration

To assess the impact of Cnih3 on opioid IVSA and opioid self-administration, we ran a PCA (as in Fig. 5) on the behaviors presented in Fig. 6 and observed that KO mice cluster tightly together while WT mice have more individual differences, indicating clearer genotype differences in this dataset (Fig. 7A), with PC1 and PC2 accounting for 51.92% of the variance in the dataset (Figure S5A). In PC1 loadings, Cnih3 deletion produced positive loadings in WT mice and negative loadings in KO mice (Fig. 7B; 2-way ANOVA, genotype F (1, 71) = 7.246, p = 0.0089). In PC2, Cnih3 deletion did not significantly impact loading (Fig. 7C; 2-way ANOVA, sex F (1, 71) = 13.53, p = 0.0005). The highest loadings, and therefore main contributors, to the genotype-based variance observed in the dataset in PC1 corresponded to drug-seeking during extinction and post-learning intake (Fig. 7D), while PC2 corresponded to operant learning and post-learning intake (Fig. 7E). Taken together, these data show that there are differences between genotypes in opioid-seeking behavior, which seem to stem from drug-seeking during extinction, post-learning intake, and operant learning. Generally, KO mice have negative PC1 loadings (Figure S5A), exhibiting long acquisition times, low active and inactive pokes on the first day of fentanyl IVSA, as well as low numbers of rewards during weeks 1–5, including the last day, which also showed low inactive pokes. During extinction, KO mice exhibit low active and inactive pokes on the first day, followed by low active pokes during weeks 1, 2, 3, and 4, and low active and inactive pokes during cue-induced reinstatement.

Fig. 7. Cnih3 deletion produces differences in operant learning and fentanyl intake during self-administration.

A Loadings plot of PC1 and PC2 for each mouse that has completed the experiments described in Fig. 6. B PC1 loadings for each mouse. C PC2 loadings for each mouse. D PC1 loadings of the variance each described behavior contributes to the dataset. The PC1 summary is the average of the PC loading of all components within a defined type of measure. Loadings above 0.3 are considered significant contributors. E PC2 loadings of the variance each described behavior contributes to the dataset. The PC2 summary is the average of the PC loading of all components within a defined type of measure. Loadings above 0.3 are considered significant contributors. Data are reported as mean ± SEM in (B-C). Average loading of each measure for the dataset shown in (D-E). Two-way ANOVA (sex x genotype) in (B and C). Symbols denote a ^#main effect of genotype in B (^####p < 0.0001). Full statistical reporting is available in Table S1. Additional data related to this dataset are available in Figure S5.

Discussion

In this study, we investigated the involvement of Cnih3, an AMPAR auxiliary protein, in behavior, a sucrose self-administration and reversal learning task, and fentanyl intravenous self-administration. We employ a mouse model of Cnih3 global deletion in mice to comprehensively assess the impact of Cnih3 on behaviors that may impact opioid use, and opioid use directly. In mice, Cnih3 deletion impaired spatial memory during the novel object recognition task, but not depression-like behavior or anxiety-like behavior. We then probed the effect of Cnih3 deletion on natural reward-seeking and reversal learning using a sucrose self-administration paradigm and found that Cnih3 deletion produced operant learning deficits in acquisition in females only. Last, we used a fentanyl IVSA paradigm to assess the role of Cnih3 in opioid-seeking and demonstrated that Cnih3 deletion blunted operant learning in females only and opioid intake in both sexes.

Cnih3 deletion impairs spatial memory

OUD is a complex disorder that involves several behaviors, including several risk factors that predispose people to opioid misuse- many of which involve alterations in glutamatergic transmission, such as pain [28], and affective disorders such as anxiety [20] and depression [24]. We characterized the role of Cnih3 in a broad host of behaviors that may impact OUD. Despite the critical role of glutamatergic plasticity in these processes, Cnih3 deletion did not impact well-being, motor coordination, thermal nociception, social memory, anxiety-like or depression-like behavior. The glutamatergic functions supporting these behaviors instead likely rely on or are compensated by other AMPAR support proteins, such as TARPs, in the absence of Cnih3. Nevertheless, Cnih3 deletion moderately impaired spatial memory in the novel object recognition task, consistent with our previous findings [10]. Spatial memory is a key part of the formation of the contextual and drug-cue associations that make up opioid-associated memories [12]. Therefore, SNPs previously found in CNIH3 may protect against OUD [9] by disrupting the formation, maintenance, and/or expression of opioid-associated memory. Interestingly, CNIH3 emerged as nominally associated with executive functioning in a large GWAS, suggesting broad correspondence in humans [44].

Cnih3 deletion-based deficits are compounded in opioid use

Cognition and associative learning are key components of maladaptive opioid use, with craving and relapse often triggered by exposure to drug-associated cues and contexts [45]. Given that Cnih3 deletion impaired spatial memory, we suspected that Cnih3 was also important for operant learning, which requires associative learning. To test this, we probed the role of Cnih3 in the acquisition of operant behavior for a natural reward, sucrose, as well as in spatial reversal, a component of cognitive flexibility. This was supported by our finding that female CNIH3 KO mice take longer to acquire sucrose self-administration. Although female KOs appeared to exhibit impaired discrimination during initial acquisition, neither sex appeared impaired in reversal learning. Though our experiments do not assess all aspects of cognition, cognitive flexibility, or spatial memory, these results suggest that Cnih3 moderately impairs operant learning of sucrose self-administration.

Our studies confirm that Cnih3 plays a role in operant learning and opioid intake, as deletion delays acquisition and decreases fentanyl intake in our IVSA model. These mild impairments in cognition that worsen spatial/contextual memory formation affect but do not prohibit opioid self-administration. PCA shows that operant learning and post-learning fentanyl intake were the notable contributors to the genotype-based variance in drug-seeking. The deficits in spatial memory and operant learning in our sucrose self-administration experiments add context to the delays in operant learning seen in our IVSA model, but the contribution of post-learning intake and drug-seeking during extinction cannot be overstated, given that the acquisition of operant learning relies on both the saliency of the reward as well as the ability to associate a cue with a reward. Opioid exposure alone alters AMPARs to favor and stabilize GluA1-containing composition [30, 33], and morphine activation of dopamine neurons in the VTA is dependent on glutamatergic tone [46], suggesting that AMPAR trafficking beyond initial learning is also important for opioid-seeking. We therefore suspect that CNIH3 deletion interrupts AMPAR insertion during the early and late phases of opioid use, leading to both our observed delays in acquisition of the operant behavior and decrease in fentanyl self-administration. As these experiments do not assess potential differences in the rewarding properties of fentanyl or differences in motivation for fentanyl, which may contribute to both associative and post-learning genotype differences, further research is necessary to tease apart the role of Cnih3 in learning and reward processing.

Our studies indicate that Cnih3 deletion produces moderate deficits in cognition and blunts fentanyl self-administration; however, they do not address the question of where in the brain Cnih3 activity is crucial for these behaviors. Our identified deletion-based differences are in spatial memory and operant learning, which are processes mediated by the hippocampus (HPC) and prefrontal cortex (PFC) [47–50]. Both brain regions are also critical for opioid-associated memory [51–54] and exhibit high CNIH3 expression [10, 11]. Previous research from our group showed that Cnih3 deletion alters hippocampal AMPAR subunit composition, leading to impaired LTP, and viral overexpression of Cnih3 in the dorsal HPC showed improved spatial memory [10]. These data suggest that Cnih3-mediated plasticity that underlies both spatial memory and opioid use is likely important in the HPC and PFC.

The regulatory role of Cnih3 may be sex-dependent

Sex differences in our studies increase as the complexity of the behavior increases. The first major sex differences observed are in the sucrose self-administration and spatial reversal learning task, where female mice exhibit delayed operant learning. During fentanyl IVSA, male KOs seemingly perform worse than female KOs, as female KOs favor the active nose port on the last session, while male KOs do not. Additionally, male KOs exhibit blunted cue-induced reinstatement. Given that we do not observe sex differences in fentanyl IVSA in our wild-type mice, but do see impairments predominantly in male KOs, we suspect that Cnih3 function may be sex-dependent.

We do not suspect that the sex differences we see in our knockouts are due to differential levels of Cnih3 expression, as previous research has shown no difference in Cnih3 between male and female mice [10]. We suspect that CNIH3 activity may differ across estrous cycle phase, given that Cnih3 is highly expressed in brain regions with abundant estrogen receptors [11]. Previous work has shown that Cnih3-related spatial memory deficits in female mice are estrous-cycle dependent [10], and that Cnih3 KO mice show larger differences in hippocampal gene expression across the estrous cycle compared to wild-type mice and male mice [55]. Though we did not assess estrous cycle phasing in our experiments, the variability and sub-grouping in some of our female Cnih3 KO data and the previous works cited above, we suspect that Cnih3 function may be modulated by estrogen.

Cnih3 deletion significantly blunts opioid IVSA and may be a potential therapeutic target

Our data suggest CNIH3 may be a potential therapeutic target to alleviate opioid use. Integrating PCA into our behavioral studies enabled us to unbiasedly recognize large-scale patterns in our data. As the complexity of the paradigm and the saliency of the reward increased (e.g., sucrose vs. fentanyl self-administration), Cnih3-deletion-based differences became more evident. This is exemplified by the visual stratification of our PC plots and the larger PC loadings (abs.) affiliated with the included variables in each model. Given that intact AMPAR dynamics are necessary for the identified contributors to genotype-based variance in our datasets (operant learning and post-learning fentanyl use), we suspect that CNIH3 is an important factor in dictating relevant AMPAR dynamics, particularly in those facilitating high-reward states such as opioid use.

Previous research shows that drugs of abuse (including opioids) alter AMPAR subunit composition [13, 34, 36, 56, 57], a process that is mediated by CNIH3 in vitro [16], and that SNPs in CNIH3 are protective against the progression of OUD in humans [9]. Our current findings suggest that global deletion does not significantly impact well-being or behaviors relevant to opioid use, such as anxiety-like and depression-like behavior, but does significantly blunt opioid self-administration in mice. We hypothesize that in instances of impaired CNIH3 function (such as a loss-of-function mutation, low gene expression, or absence), the AMPAR dynamics necessary for drug-induced plasticity that facilitates maladaptive drug use are also impaired, protecting individuals from the development of OUD. This emphasizes CNIH3 as a target that has the potential to alter AMPAR dynamics that facilitate OUD while maintaining necessary AMPAR function. However, the mechanisms by which CNIH3 produces our observed results, as well as the suggestion that CNIH3 functioning may differ by sex, demand further research before contemplating viability as a treatment. Additionally, CNIH3 was identified as a gene relevant to OUD, which dictated the context in which it was assessed in these experiments- it is important to validate whether CNIH3 plays a role in regulating reward-seeking behavior for other drugs of abuse.

Conclusion

OUD continues to strain the US as research struggles to keep up with high relapse and overdose rates. Addressing OUD is difficult due to its nature: predisposition does not always translate to the severity of the disorder, and life experiences, access to medical care, social support, and coexisting conditions can greatly alter OUD expression and development. Genetic tools do not hand us the golden key to unlocking OUD, but how we wield them can greatly sharpen the direction of preclinical research to maximize patient care progression. Unbiased genetic tools, like the GWAS, arm us with a powerful way to identify biomarkers and candidate genes for the prevention and targeted treatment of OUD. GWAS identified that SNPs in CNIH3 were protective against the development of daily opioid use in humans, and here, we have used a global deletion model to assess a potential role for Cnih3 in opioid-related behavior. Importantly, global Cnih3 deletion did not significantly impact baseline functioning or affective states but did produce moderate effects on learning and memory as captured by our behavioral paradigms. Additionally, Cnih3 deletion impaired the onset and progression of opioid use in fentanyl IVSA in a sex specific manner. It is important to acknowledge that our usage of a global KO model may produce potential behavioral and compensatory effects in these studies. However, the results of the GWAS paired with our data, suggest Cnih3 plays a role in opioid-related behavior and, therefore, is a potential player in opioid plasticity.

Materials and methods

Study design

Sample sizes were determined by power analysis. Data inclusion/exclusion criteria were defined prior to experiment onset and described in each subsection. Outliers were defined prior to the study as values ± 2 standard deviations from the mean. Values that met these criteria were excluded and reported (resulting in no exclusion), and no outliers were removed from acquisition data and number of pokes to limit exclusion of potential subgroups and individual differences. Treatments/sides for stimulus presentation were counterbalanced and randomized as described in each subsection. Investigators were blinded to mouse genotype during data collection and analysis.

Mice

The experiments were subdivided into behavioral characterization, sucrose self-administration and reversal learning task, and fentanyl intravenous self-administration, each of which included separate cohorts of mice. Details below. All experiments used adult (8–10-week-old) male and female C57BL6J, WT littermate controls, and Cnih3 knockout mice [10]. Cnih3 KO mice were previously generated from Cnih3^{tm1a(KOMP)Wtsi} BL/6 N mice (Knockout Mouse Project); full details previously published [10]. Briefly, the KO line lacks exon four of Cnih3, truncating the CNIH3 protein while not altering Cnih2 [10, 55]. Genotyping was performed according to the methods used previously [10]. Importantly, we assessed potential effects of genetic background across all assays by examining both the wild-type littermates of Cnih3 KO mice (no deletion) and C57 wild-type mice but found no significant differences between these groups (F = 0.01257-1.799, p = 0.1986-0.9117), indicating a lack of influence. This recapitulates results from a study that determined no significant difference in gene expression in the dorsal hippocampus between Cnih3 littermate controls and C57Bl6J animals [55]. One exception was in the tail suspension test, for which comparisons were therefore made to littermate controls (Fig. 3J-L; Figure S2C-E). Otherwise, wild-type littermates and C57 mice are collapsed and represented as wild-type (WT) throughout.

Behavioral characterization

Mice were singly housed due to equipment limitations in a standard 12:12-hr light cycle and tested during the light phase. The behavioral battery included, in the following order: (i) open field test, (ii) novel object recognition task, (iii) elevated plus maze, (iv) rotarod, (v) nest building, (vi) social interaction test, (vii) hot plate, and (viii) tail suspension test. Mice were habituated to the testing room 1 h before each behavioral test. Behaviors were recorded and analyzed using AnyMaze software unless otherwise stated. Mice for this experiment performed all eight tasks (described below) with at least one “rest day” between tasks. Behavioral apparatuses were cleaned with Cavicide between mice.

(i) and (ii) Open Field Test (OFT) and Novel Object Recognition Task (NORT): The OFT was used to assess exploratory behavior and anxiety-like behavior, and the NORT was used to assess spatial memory. Both took place in a 50 x 50 x 50 cm matte grey box for 3 consecutive days. During each day, mice were gently placed in the center of the box and allowed to freely explore for 10 min, and the time and distance in each compartment were recorded. NORT Day 1 was a habituation phase (also termed the open field test; OFT), during which mice freely explored the empty chamber. Day 2 was a familiarization phase, during which two identical objects (A) were placed in two investigation zones 5 cm from the walls. In addition to the measurements above obtained with AnyMaze, the time spent interacting with each object was hand-scored. On Day 3, the same object A was placed in one investigation zone of the chamber, while a novel object (B) was placed in the other investigation zone. The positions for objects A and B were alternated so that the location of the novel object was counterbalanced. Objects that produced equal investigation times in C57 mice were chosen to avoid object bias [58].

(iii) Elevated Plus Maze (EPM): The EPM was used to assess anxiety-like behavior and consisted of an apparatus with 2 open arms, 2 closed arms (50 cm length x 10 cm width), and a center compartment (10 x 10cm). Each mouse was placed gently in the center compartment facing away from the experimenter and was allowed to freely explore the apparatus for 10 min. Parameters recorded were time and distance in each compartment.

(iv) Rotarod: An accelerating rotarod (Ugo Basile) was used to assess motor coordination and balance [59]. Briefly, mice received a maximum of 5 training sessions where mice spent 120 s walking on the rotarod at a fixed speed of 4 rpm. All mice completed training without falling in five attempts or fewer. One hour after successful completion of the training session, the latency to fall as the rotarod accelerated from 4 to 40 rpm over 5 min was assessed. Five consecutive acceleration trials were performed, with 10 min rest time between each trial.

(v) Nest-building: The nest-building task was used to assess naturalistic behavior and well-being and was carried out in the home cage [60]. Each mouse was given 3 g of untorn nestlet in a clean cage. 24 h later, the intact portions of the nest were weighed to calculate % of the nestlet that was torn.

(vi) Social Interaction Test (SIT): The social interaction test was used to assess social behavior and was conducted in a 60 × 42 × 22 cm box divided into 3 equal chambers. The center chamber was always empty, and the two extreme chambers contained an inverted wired cup that was magnetically secured to the apparatus floor. For this test, mice were allowed to freely explore the apparatus for 10 min on 3 consecutive days. Day 1 was a habituation session in which the mice freely explored the three-compartment chamber with both cups empty. On day 2 (social novelty phase), an unknown age and sex matched C57 mouse was placed under one of the two cups. On day 3 (social preference phase), the mouse from day 2 was placed in the same cup and side of the apparatus as the previous day, while a novel age and sex-matched mouse was placed under the cup on the other side. The side of the novel mouse was counterbalanced between subjects. Time spent in each compartment was assessed with AnyMaze, and time spent investigating each cup was hand-scored.

(vii) Hot Plate: Mice were gently placed in the center of a 54 °C hot plate apparatus (BIOSEB hot and cold plate, BIO-CHP) and observed for signs of thermal nociception (licking/kicking of paws, jumping, etc.). At the first instance of the above behaviors, or once a maximum time of 45 s was reached, mice were removed from the hot plate and placed back in the home cage. Latency (s) to the first response was recorded.

(viii) Tail Suspension Test (TST): The tail suspension test was used to assess depression-like behavior and took place in a matte white PLA 3D printed apparatus (50 cm tall, 23 cm wide). A metal rod was secured to a divot in the top of the box. A 4.5 cm long plastic straw was placed at the base of the tail to prevent climbing, and then mice were suspended by the tail from the metal rod. Time spent mobile and immobile, and the number of bouts of each, were measured over 6 min.

We encountered one cohort of video recording failure during the OFT that resulted in the loss of data from 7 WT and 1 KO males.

Sucrose self-administration and reversal learning task

Mice were group-housed in a reverse 12:12 hr light cycle and tested during the dark cycle. Mice were placed into MedAssociates chambers and learned, without pre-training, to poke a light-cued nose port (counterbalanced between subjects) to receive one chocolate sucrose pellet (TestDiet chocolate sucrose tab 20MG 1818335(5TUT)) in an FR1 schedule of reinforcement during daily 1-hr sessions. Mice could receive a maximum of 30 sucrose pellets, after which the program terminated. Responses on the active and inactive nose ports were recorded (60-sec timeout). Upon reaching acquisition criteria (30 pellets consumed and ≥70% responses on the active nose poke for 3 consecutive sessions), the active and inactive nose ports were switched such that the unlit nose port was reinforced. The same acquisition criteria were used as a measure of cognitive flexibility under these conditions. After re-reaching acquisition criteria, the operant paradigm was changed yet again to match the original conditions such that the lit nose poke was reinforced with the sucrose pellet reward. Time to reach acquisition on this set of parameters was quantified and used as a measure of cognitive flexibility. Mice that did not reach acquisition criteria within 50 sessions, appeared ill, or lost >20% bodyweight were excluded from the analysis (n = 4).

Fentanyl intravenous self-administration (IVSA)

IV surgery: Mice were acclimated to a reverse light cycle for at least one week before surgery. Catheters were prepared ahead of time from a 7 cm long piece of polyurethane tubing (Instech, BTPU-027), and a silicone bulb was placed at 0.9 cm from one end and allowed to dry for at least one hour. Before use, the catheter was dipped in Chlorohexidine to sterilize it, and attached to a catheter base (Instech, VAM1B/25) on the side furthest from the silicone bulb. Mice were anesthetized deeply under isoflurane at 2.5-4% and the skin between the shoulder blades and above the right jugular vein was shaved and scrubbed for surgery. A 1 cm incision was made between the shoulders and above the jugular vein, through which the catheter was passed subcutaneously. The jugular vein was isolated by gross dissection, and a small incision was made in the vein using a 26-G needle. Vein hemostats were used to open the incision and slip the catheter (trimmed if needed) into the jugular vein. Once inside the vein, placement was verified by gently pulling back the syringe plunger and observing for blood in the tubing. Then, one suture was tied below, and another above, the silicone bulb to secure the catheter. The neck incision was closed, and the mouse was flipped over. The catheter port was placed subcutaneously, leaving the connector exposed, and sutured into place. Then, the back incision was closed, post-operative drugs were administered according to IACUC protocol, and mice were monitored during recovery in a clean cage before returning to the colony.

After surgery, mice were given post-op care (8 mg/kg Baytril and 5 mg/kg Carprofen once daily) for at least two days and recovered for at least five days before the onset of behavior. Catheters were kept patent and clean using a heparin/gentamicin solution in saline (100USP/mL heparin,

0.44 mg/mL gentamicin) and were flushed with 0.3 ml of the solution daily before entry into the behavioral chambers. At the end of the experiment, catheter patency was verified by flushing at least 0.3 mL of 5% Evans Blue through the catheter after decapitation. If dye is visualized outside the jugular vein after flush (leakage), the data were excluded.

Self-Administration task

Mice were singly housed in a reverse 12:12 h light cycle and tested during the dark cycle. Fentanyl IVSA took place in operant boxes (Med Associates) equipped with two nose poke ports. A cue light within the active port was illuminated to signal drug availability. Nose pokes in the active port resulted in the delivery of fentanyl (1 μg/kg/infusion, i.v.) followed by an 8 s ‘timeout’ period during which reinforcement was withheld and the cue light was turned off. Each session lasted 2 h. Mice continued daily training for 25 sessions, during which acquisition of the behavior was monitored (criteria: ≥5 infusions, <30% variability in fentanyl intake, and a discrimination index ≥0.7 (DI; active pokes/total pokes) for 2 consecutive days). Next, mice underwent 21 daily 2-h extinction sessions where cue lights were off, and the drug was not available. 24 h after the last extinction session, mice underwent a 2-h test of cue-induced reinstatement where non-reinforced nose pokes in response to the reintroduction of the cue light were measured as a proxy for drug-seeking behavior. Mice that did not reach acquisition criteria within 25 sessions, pulled their catheters, failed the catheter patency test following sacrifice, appeared ill, or lost >20% bodyweight were excluded from the analysis (n = 39).

Statistical analysis

GraphPad Prism version 10.4.2 was used for all statistics on mouse behavior. All the experiments were replicated at least twice to prevent nonspecific day/condition effects. Behavioral data were assessed for normality using D’Agostino and Pearson tests and Shapiro–Wilk tests, and then analyzed using two-way (sex × genotype) and three-way (sex × genotype × session) ANOVAs. Šídák’s multiple comparisons tests were used to probe significant interactions. We assessed acquisition using survival plots and analyzed differences between the curves using logrank Mantel-Cox tests. Principal component analyses (PCA) were used to perform unbiased pattern recognition to identify factors contributing to variance. In this case, PCA was used to assess the contribution of aspects of the sucrose self-administration and reversal learning task, and fentanyl intravenous self-administration experiments to the variance in the data attributed to genotypic differences. Only mice that completed each task per its subdivided experiment were included in the PCA, as missing variables are not tolerated in this analysis. In the case of missing variables (one cohort of sucrose self-administration mice did not complete the second reversal and was therefore not included in the PCA), all variables from those individuals were omitted. In Prism, data were standardized to have a mean of 0 and a standard deviation of 1, and principal components were selected based on parallel analyses, with the percentile level set at 95% and 1000 simulations. Principal components 1 and 2 (PC1, PC2) were selected for further analysis as they accounted for the highest contribution of variance in each dataset. For interpretation, parameter loadings onto principal components were examined, wherein those >0.3 (absolute value) were considered significant (as described in [43]). All data are expressed as the mean ± SEM, and significance was set at p < 0.05, α = 0.05; see statistical details in table S1.

Data and materials availability

All data reported and additional information required to reanalyze the data reported in this paper are available from the lead contact upon request.

Author contributions

Conceptualization: TL, JAM. Data curation: TL, AA, EN. Formal analysis: TL, AA, EN. Funding acquisition: JAM. Investigation: TL, AL, TAA, AP, JJD, AA, EN, JAM. Methodology: TL, AL, TAA, AA, EN, JAM. Supervision: TL, AA, EN, JAM. Validation: TL, AL, TAA, AP, JJD, AA, EN, JAM. Visualization: TL, AA, EN. Writing – original draft: TL and JAM. Writing – review & editing: TL, AL, TAA, AP, JJD, AA, EN, JAM.

Funding

R01DA058613 (JAM). R01DA054900 (JAM).

Competing interests

The authors declare no competing interests.

Ethics approval

All procedures were approved by Washington University in St. Louis under IACUC protocol number 24-0109 in accordance with the National Institutes of Health Guidelines for the Care and Use of Laboratory Animals.

Supplementary information

The online version contains supplementary material available at 10.1038/s41398-026-04054-x.