1 INTRODUCTION

Psychostimulant (methamphetamine, cocaine) use disorders (PUDs) are a serious public health concern. Until the COVID-19 pandemic emerged, the opioid epidemic crisis had begun to plateau. Meanwhile, PUDs have quietly made a resurgence, with increased use and deaths.1, 2 Yet, despite an estimated 40%–50% heritability for PUDs,3-5 genome-wide association studies have identified few loci.6 In one study, a significant GWAS hit for cocaine dependence mapped to FAM53B.6, 7 Notably, an unbiased, quantitative trait locus (QTL) approach in mice identified a trans-expression QTL regulating Fam53b expression that was genetically correlated with variance in cocaine intravenous self-administration (IVSA) in BXD-RI mice, exemplifying cross-species bidirectional translation with discovery genetics in rodents.

Reduced Complexity Crosses exploit the extreme, near-isogenic nature of closely related inbred substrains to rapidly map, pinpoint and validate quantitative trait loci (QTLs) containing causal quantitative trait genes (QTGs) and quantitative trait variants (QTVs) underlying complex trait variation,8, 9 including gene expression and behavior.10-13 Of relevance to the present study, Kumar and colleagues used a mouse Reduced Complexity Cross between C57BL/6 substrains to map a missense variant in Cyfip2 with sensitivity to cocaine-induced velocity and extended these findings to methamphetamine.12 We previously used a similar Reduced Complexity Cross to map and validate Cyfip2 in binge-like eating.11 Also of relevance to the present study, we exploited the reduced complexity of C57BL/6 substrains to identify a functional noncoding single nucleotide deletion in Gabra2 (alpha-2 subunit of the GABA-A receptor) that induced a loss-of-function decrease in transcript and protein expression.13 Correction of this mutation via CRISPR/Cas9 gene editing restored Gabra2 expression at both the transcript and protein levels.13 DBA/2 mouse substrains combined with historical BXD-RI substrains have also been exploited to identify a functional missense variant in trace amine-associated receptor 1 (Taar1) underlying differences in the aversive properties of methamphetamine self-administration, body temperature and toxicity.14-17

Administration of addictive drugs such as opioids and psychostimulants increases dopamine release in forebrain regions, including the dorsal striatum and nucleus accumbens, which contributes to the locomotor stimulant and rewarding properties of drugs of abuse.20, 18, 19 Psychostimulant-induced locomotor activity is a rapid, high-throughput heritable trait that is amenable to QTL mapping in multiple genetic populations20-23 and has a shared genetic basis with other addiction-relevant behavioral traits. As two examples, we mapped and validated genetic factors influencing psychostimulant and opioid-induced locomotor activity, including Csnk1e24 and Hnrnph1.25 Subsequently, we and others have extended the role of these two genes to other complex behavioral models for addiction, including reward as measured via conditioned place preference26, 27 and reinforcement as measured via intravenous and oral self-administration.27, 28

C57BL/6J (B6J) and C57BL/6NJ (B6NJ) are two substrains of C57BL/6, the most commonly used mouse strain in biomedical research, and are 99.9% genetically similar, yet exhibit significant differences in several addiction-associated traits,29 including ethanol consumption,30, 31 nicotine behaviors32 and psychostimulant behaviors.12, 29 Although phenotypic differences between B6 substrains can be quite large, genotypic diversity is extremely small, with only an estimated 10,000 to 20,000 variants (SNPs plus indels) distinguishing the two strains.33-35

In the present study, we used a Reduced Complexity Cross between C57BL/6 substrains to map the genetic basis of sensitivity to the locomotor stimulant properties of methamphetamine, including maximum speed and distance traveled. Following the identification of two historical loci, including one locus for sensitized methamphetamine-induced maximum speed near the Cyfip2 missense mutation that was previously identified for acute and sensitized cocaine velocity12 and a second locus near the functional intronic variant in Gabra2,13 we used a CRISPR/Cas9 gene-edited knockin mouse model with the corrected Gabra2 mutation to validate this functional indel as necessary for enhanced acute stimulant sensitivity that is exhibited in the parental C57BL/6J substrain.12

2 MATERIALS AND METHODS 2.1 C57BL/6J (B6J), C57BL/6NJ (B6NJ) and a B6J x B6NJ-F2 reduced complexity cross (Bryant Lab, BUSM)

All experiments involving mice were approved by the Boston University School of Medicine (BUSM) and University of Tennessee Health Science Center (UTHSC) Institutional Animal Use and Care Committees and were conducted in accordance with the AAALAC Guide for the Use and Care of Laboratory Animals.36 Mice were housed in an AAALAC-accredited temperature- and climate-controlled facilities on a 12 h light/dark cycle (lights on at 0630 h for BUSM and 0600 h for UTHSC). Mice were housed in same-sex groups of two to five mice per cage with standard laboratory chow and water available ad libitum except during testing. All behavioral testing was performed during the light phase of the 12 h light/dark cycle.

C57BL/6J mice (B6J; n = 31; 16 females, 15 males; all 68 days old on Day 1 the 5-day locomotor protocol) and C57BL/NJ mice (B6NJ; n = 32; 16 females, 16 males, all 68 days old on Day 1 of the 5-day locomotor protocol) were purchased from The Jackson Laboratory (Bar Harbor, ME) at 7 weeks of age and were habituated in the vivarium 1 week prior to experimental testing that occurred next door. For QTL mapping, a unidirectional cross was conducted whereby B6J females were crossed to B6NJ males to generate B6J x B6NJ-F1 mice and B6J x B6NJ F1 offspring were intercrossed to generate B6J x B6NJ F2 mice. All mice within a cage were assigned the same treatment.

All mice comprising the parental substrains and F2 offspring for which we present behavioral data had a prior, identical history of naloxone-induced conditioned place aversion as described in our original publication.37 Notably, however, the parental substrain breeders and F1 breeders that were used to generate the phenotyped F2 offspring were not themselves phenotyped and were experimentally naive. F2 females and F2 males were 62–114 days old on Day 1 of the five-day locomotor protocol. Briefly, following initial assessment of preference for the drug-paired side, 24 h later, mice received two alternating injections of naloxone hydrochloride (4 mg/kg, i.p.) and two alternating injections of saline (i.p.), separated by 48 h. Then, 72 and 96 h after the second saline trial, mice were re-assessed for drug-free and state-dependent conditioned place aversion for the naloxone-paired side, respectively. Thus, all mice received a total of three injections of 4 mg/kg naloxone over 9 days. One week following recovery from the test for naloxone-induced conditioned place aversion, mice were tested for methamphetamine stimulant sensitivity in a five-day protocol (two saline injections, three methamphetamine injections) as described below.

2.2 Gabra2 knockin mice

Gene-edited knockin mice were generated by inserting the corrected Gabra2 single intronic nucleotide on the mutant C57BL/6J (B6J) background via CRISPR/Cas9 gene editing as previously described.13 Between 1976 and 1991, this single nucleotide deletion became fixed and exists in many engineered lines, BXD recombinant inbred strains that were generated after 1991, the Collaborative Cross, Diversity Outbred mice and the majority of consomic lines.13 The C57BL/6NJ (B6NJ) substrain contains the deleted nucleotide and thus, inserting the nucleotide into the genome of the B6J background is predicted to modulate Gabra2-related phenotypes in a direction consistent with the B6NJ phenotype. Briefly, a sgRNA was designed that targeted Gabra2 at the intron/exon junction near chromosome 5 at 71,014,638 Mb (mm10). A T7 promoter containing the sgRNA template was used to produce sgRNA and Cas9 mRNA that was then purified, ethanol precipitated and re-suspended in DEPC-treated water. A 121 nucleotide single-stranded DNA repair template oligo with the T insertion in the intron of Gabra2 along with sgRNA and Cas9 mRNA were co-injected into the cytoplasm of B6J one-cell embryos. Offspring of injected embryos were screened for the insertion via PCR amplification of the knockin site. PCR products containing the amplicon were sequenced directly or subcloned into pCR2.1-TOPO (Invitrogen) and sequenced. The male founder (F0) was crossed to female B6J mice to generate F1 progeny. F1 mice were crossed to generate F2 mice. The colony is maintained through heterozygous breeding and all behavioral phenotyping was performed in generations F2 and higher. Potential off-targets were screened using CRISPOR (RRID:SCR_015935)38, 39 as previously described13 and no off-target modifications were detected in the top 15 predicted off-target sites. A total of 8 wild-type females (91–155 days old), 10 wild-type males (99–121 days old), 8 KI females (91–149 days old) and 9 KI males (86–149 days old) were tested. Age ranges reflect the age on Day 1 of the five-day locomotor protocol.

2.3 Drugs

Methamphetamine hydrochloride (MA) (Sigma, St. Louis, MO) was dissolved in sterilized, physiological saline (0.9%) prior to injection (10 ml/kg, i.p.). The dose of MA (2 mg/kg) was chosen based on our prior success in mapping QTLs with this dose24, 25, 40, 41 and based on a previous study that identified C57BL/6 substrain differences in MA-induced locomotor activity.12

2.4 Methamphetamine-induced maximum speed and distance traveled in B6J, B6NJ and B6J x B6NJ-F2 mice (Bryant Lab, BUSM)

The plexiglas apparatus consisted of an open field (40 cm length x 20 cm width x 45 cm tall; Lafayette Instruments, Lafayette, IN) surrounded by a sound-attenuating chamber (Med Associates, St. Albans, VT) that was unlit inside. Behaviors were recorded using a security camera system (Swann Communications, Melbourne, Australia) and then video tracked (Anymaze, Stoelting, Wood Dale, IL). We employed a 5-day locomotor protocol,41 which is an extended version of the 3-day protocol comprising the acute methamphetamine response.25 On Days (D)1 and 2, following habituation to the next-door testing room for a minimum of 30 min, mice were injected with SAL (100 ul/10 g, i.p.). On D3, D4 and D5 mice were injected with methamphetamine (2 mg/kg, 100 ul/10 g, i.p.). Following i.p. injection, mice were immediately placed into the open field and video recorded over 30 min. Each mouse was tested at the same time of day, every day until completion of the 5-day protocol. Locomotor phenotypes, including total distance traveled and maximum speed while mobile, were calculated with AnyMaze.

Data were analyzed using repeated measures ANOVA with strain and sex as factors and day as a categorical repeated measure. Time course analysis for a given day was analyzed in a similar manner but with Time (six, 5-min bins) as the repeated measure. Post hoc analysis was conducted using the family-wide error variance with simple contrasts and Tukey's honestly significant difference to correct for multiple comparisons. Mice were tested during the light phase of the 12/12 h light/dark cycle (lights on at 0630 h) between 0800 and 1300 h. Each mouse was tested at the same time of day, every day, until completion of the 5-day protocol.

2.5 Methamphetamine-induced distance traveled in Gabra2 knockin (KI) mice (Mulligan Lab, UTHSC)

Validation for the role of the Gabra2 functional intronic single nucleotide deletion (see following text) in methamphetamine-induced distance traveled was conducted in the Mulligan Lab at UTHSC. Procedures were similar, although not identical. The regimen was identical (saline on Days 1–2; 2 mg/kg methamphetamine on Days 3–5). The major difference was that a larger open field arena (40 cm x 40 cm x 40 cm) was employed at UTHSC and there was no sound-attenuating chamber. Also, unlike the arenas at BUSM, the UTHSC arenas were open air arenas without sound attenuating chambers. Additionally, mice were tail marked and handled for at least 3 days prior to testing. Prior to injection, the experimenter examined the tail marks and placed each mouse in a separate holding cage with clean bedding for at least 5 min before receiving an injection of saline (100 μl per 30 g body weight, i.p.). i.p.) or methamphetamine (2 mg/kg, i.p.) or saline. All other procedures were the same as BUSM. The larger arena size (1.5-fold larger than the arena used in the Bryant Lab) likely accounts for the overall higher level of locomotor activity in this study compared with the parental strain and F2 studies. Similar to the Bryant Lab, behavior was also recorded with a security camera system and locomotor phenotypes were also calculated in AnyMaze. Mice were also tested during the light phase of the 12 h/12 h light/dark cycle (lights on at 0600 am) between 1000 and 1400 h. Each mouse was tested at the same time of day, every day, until completion of the 5-day protocol.

2.6 Statistical analysis of parental substrains and KI mice

In analyzing summed data across days, we first ran repeated measures (RM) analysis of variance (ANOVA) with Substrain/Genotype and Sex as factors and Day as the repeated measure. We then ran separate RM ANOVAs for females and males with Substrain/Genotype as a factor and Day as the repeated measure. In analyzing time course data post-methamphetamine, similar analyses were conducted, except that “Time” (5-min bins) rather than Day was treated as the repeated measure. Tukey's post hoc or simple contrasts were conducted to identify the source of main effects and interactions.

2.7 DNA collection and genotyping in B6J x B6NJ-F2 mice

DNA was extracted from spleens of F2 mice and prepared for genotyping using a standard salting out protocol. Ninety SNP markers spaced ~30 Mb (~15 cM) apart were genotyped using a custom-designed 96 x 96 Fluidigm SNPtype array (South San Francisco, CA).10, 11 SNPs were called using the Fluidigm SNP Genotyping Analysis Software and SNPtype normalization with the default threshold.

2.8 QTL analysis in B6J x B6NJ-F2 mice

QTL analysis was performed in F2 mice using the R package R/qtl (RRID:SCR_009085) as previously described.11, 29, 42 Quality checking of genotypes and QTL analysis were performed in R (https://www.r-project.org/) using R/bestNormalize (https://github.com/petersonR/bestNormalize) and R/qtl.42 Phenotypes were assessed for normality using the Shapiro–Wilk Test. Because some of the data residuals deviated significantly from normality, we used the orderNorm function to perform Ordered Quantile normalization43 on all phenotypes. QTL analysis was performed using the “scanone” function and Haley-Knott (HK) regression. Permutation analysis (perm = 1000) was used to compute genome-wide suggestive (p < 0.63) and significance thresholds for log of the odds (LOD) scores (p < 0.05). Sex was included as an additive covariate in the QTL model. In a follow-up analysis, Age was also included as a covariate. Peak marker positions were converted from sex-averaged cM to Mb using the JAX Mouse Map Converter (http://cgd.jax.org/mousemapconverter). Percent phenotypic variance explained by each QTL was calculated using the “fitqtl” function.

Power analysis of Day 5 maximum speed and Day 3 distance traveled using the data from 184 F2 mice was conducted using the R package R/qtlDESIGN (RRID:SCR_013424).44 For each behavioral phenotype, we generated plots showing power versus % variance explained for an additively inherited QTL.

2.9 RNA-seq

Striatum was chosen for eQTL analysis for historical reasons and because this brain region is a major local site of drug action where methamphetamine binds to the dopamine transporter (and other monoamine transporters) and vesicular monoamine transporters to cause reverse transport of dopamine into the synapse, thus inducing stimulant, rewarding and reinforcing effects.45 Furthermore, Gabra2-containing GABA-A receptors are concentrated in the striatum46 and are a dominant receptor type.47 Thus, the striatum is a highly relevant brain tissue to ascertain the effects of Gabra2 genetic variation on the transcriptome as they relate to methamphetamine-induced locomotor stimulation.

Striatum punches were collected as described25 for RNA-seq from 23 F2 mice. Details on the prior experimental history of these mice are published.10 Mice were previously trained and tested for place conditioning to oxycodone hydrochloride (1.25 mg/kg, i.p.; a total of 3 injections over 9 days). Six days later, these mice underwent an additional 4 daily injections of oxycodone (20 mg/kg, i.p.) before being tested for antinociception on the hot plate and then the following week, an additional 4 daily injections of the same dose of oxycodone prior to testing for affective withdrawal on the elevated plus maze 16 h later. Brains were harvested 24 h after behavioral assessment of oxycodone withdrawal (~40 h after the final injection of oxycodone).

Brains were rapidly removed and sectioned with a brain matrix to obtain a 3 mm thick section where a 2 mm diameter punch of the striatum was collected. Left and right striatum punches were pooled and immediately placed in RNA later (Life Technologies, Grand Island, NY) and stored for 48 h at 4°C prior to storage in a − 80°C freezer. Total RNA was extracted using the RNeasy kit (Qiagen, Valencia, CA) as described.25 RNA was shipped to the University of Chicago Genomics Core Facility for cDNA library preparation using the Illumina TruSeq (oligo-dT; 100 bp paired-end reads). Libraries were prepared according to Illumina's detailed instructions accompanying the TruSeq® Stranded mRNA LT Kit (Part# RS-122-2101). The purified cDNA was captured on an Illumina flow cell for cluster generation and sample libraries were sequenced at 23 samples per lane over five lanes (technical replicates) according to the manufacturer's protocols on the Illumina HiSeq 4000 machine, yielding an average of 69.4 million paired-end reads per sample. FASTQ files were quality checked via FASTQC and possessed Phred quality scores >30 (i.e., less than 0.1% sequencing error). This data set is publicly available on Gene Expression Omnibus (GEO #119719).

2.10 Cis- and trans-expression QTL (eQTL) analysis

Details on eQTL mapping are published.10 We conducted eQTL analysis on striatal samples from 23 mice (8 F2 females, 15 F2 males; age range = 80–128 days old at the time of tissue harvesting). We aligned FastQ files to the reference genome (mm38) via TopHat (RRID:SCR_013035)48 using the mm38 build and Ensembl Sequence and genome annotation. We used featureCounts (RRID:SCR_012919) to count and align reads. For cis-eQTL analysis, we used the same marker panel that we used in behavioral QTL analysis. We removed lowly expressed exons that did not possess at least 10 reads total across all 115 count files. Because of the low resolution of QTL mapping in an F2 cross, we liberally defined a gene with a cis-eQTL as any gene possessing a genome-wide significant association between expression and a polymorphic marker that was within 70 Mb of a SNP (the largest distance between any two SNPs from the 90-SNP panel). Analysis was conducted using limma (RRID:SCR_010943) with default TMM normalization and VOOM transformation.49, 50 We used limma rather than R/qtl for eQTL analysis which allowed us to account for technical replicates in the analysis based on multiple sequencing of the same samples across sequencing runs. A linear model was employed whereby sample replicates were treated as a random effects repeated measure. The duplicate Correlation() function was used to estimate within-sample correlation, which was then included in the lmFit() function. An ANOVA test was conducted for gene expression that included Sex and Age as covariates and Genotype as a fixed effect. Gene-level tests were conducted using the likelihood Ratio test. A false discovery rate of 5% was employed as the cut-off for statistical significance.51

2.11 Enrichment analysis

Enrichment analysis of genes whose transcripts correlated with Gabra2 transcript levels (r ≤ −0.5 or r ≥ 0.5; p ≤ 0.015) was conducted using the online tool Enrichr (RRID:SCR_001575)52, 53 where we report GO terms for molecular, cellular and biological function.

3 RESULTS 3.1 B6J mice show an increase in methamphetamine-induced maximum speed and locomotor activity compared with B6NJ mice

In examining B6 substrain differences in maximum speed, there were no substrain difference in response to saline injections (10 ml/kg, i.p.) on Days 1 and 2 (p's > 0.05). However, in response to methamphetamine (2 mg/kg, i.p.) on Day 3, Day 4 and Day 5, B6J mice showed increased maximum speed relative to B6NJ mice, regardless of whether the data were collapsed across Sex or analyzed separately in females and males (Figure 1A–C). RM ANOVA of the sex-combined data set across days indicated a main effect of Substrain (F1,59 = 59.96, p = 1.49 × 10−10), Sex (F1,59 = 18.02, p = 7.85 x 10−5), Day (F4,236 = 377.90, p < 2 × 10−16) and a Substrain x Day interaction (F4,236 = 24.51, p < 2 × 10−16). Tukey's post hoc showed a significant increase in maximum speed in B6J versus B6NJ mice on Day(D) 3, D4 and D5 (*all p'sadjusted < 0.0001; Figure 1A). In breaking down the data set by Sex, for females-only, RM ANOVA indicated a main effect of Substrain (F1,30 = 22.29, p = 5.11 × 10−5), Day (F4,120 = 199.44, p < 2 × 10−16) and a Substrain x Day interaction (F4,120 = 13.21, p = 6 × 10−9). Tukey's post hoc test showed a significant difference increase in maximum speed in B6J versus B6NJ females on D3 (*padjusted = 0.0005), D4 and D5 (*p'sadjusted < 0.0001; Figure 1B). For males-only, RM ANOVA indicated a main effect of Substrain (F1,29 = 38.67, p = 8.75 × 10−7), Day (F4,116 = 178.66, p < 2 × 10−16) and a Substrain x Day interaction (F4,116 = 11.67, p = 5.36 × 10−8). Tukey's post hoc showed a significant increase in maximum speed in B6J versus B6NJ males on D1 (padjusted = 0.03), D3, D4 and D5 (*p'sadjusted < 0.0001; Figure 1C).

Maximum speed and distance traveled in response to saline (Days 1–2) and methamphetamine (Days 3–5) in the parental C57BL/6J (B6J) and C57BL/6NJ (B6NJ) substrains. (A) Sex-combined maximum speed (m/s) across training days. There was a significant increase in maximum speed in B6J versus B6NJ mice on Day(D) 3, D4 and D5 (*all p'sadjusted < 0.0001). (B) Maximum speed (m/s) across training days in females. There was a significant difference increase in maximum speed in B6J versus B6NJ females on D3 (*padjusted = 0.0005), D4 and D5 (*p'sadjusted < 0.0001). (C) Maximum speed across training days in males. There was a significant increase in maximum speed in B6J versus B6NJ males on D1 (padjusted = 0.03), D3, D4 and D5 (*p'sadjusted < 0.0001). (D) Sex-combined distance traveled (m) across training days. There was a significant increase in distance traveled in B6J versus B6NJ mice on D3, D4 and D5 (all p'sadjusted < 0.0001). (E) Distance traveled across training days in females. There was a significant increase in distance traveled in B6J versus B6NJ females on D3, D4 and D5 (*all p'sadjusted < 0.0001). (F) Distance traveled across training days in males. There was a significant increase in distance traveled in B6J versus B6NJ males on Day 3, Day 4 and Day 5 (*all p'sadjusted < 0.0001)

In examining B6 substrain differences in distance traveled following saline versus methamphetamine, there were no substrain differences in distance traveled in response to saline on Days 1 and 2 (p's > 0.05). However, in response to methamphetamine, B6J mice showed an increase in distance traveled compared with B6NJ mice on Day 3, Day 4 and Day 5, regardless of whether the data were collapsed across Sex or analyzed separately in females and males (Figure 1D–F). RM ANOVA of the sex-combined data set across days showed a main effect of Substrain (F1,59 = 82.39, p = 8.51 × 10−13), Day (F4,236 = 798.0, p < 2 × 10−16), a Substrain × Day interaction (F4,236 = 67.99, p < 2 × 10−16) and a Sex × Day interaction (F4,236 = 6.46; p = 6.0 × 10−5). Tukey's post hoc showed a significant increase in distance traveled in B6J versus B6NJ mice on D3, D4 and D5 (all p'sadjusted < 0.0001; Figure 1D). In breaking down the data set by Sex, RM ANOVA of females-only indicated a main effect of Substrain (F1,30 = 41.18, p = 4.37 × 10−7), Day (F4,120 = 408,83, p < 2 × 10−16) and a Substrain x Day interaction (F4,120 = 27.05, p = 5.27 × 10–10). Tukey's post hoc showed a significant increase in distance traveled in B6J versus B6NJ females on D3, D4 and D5 (*all p'sadjusted < 0.0001; Figure 1E). For males-only, RM ANOVA indicated a main effect of Substrain (F1,29 = 40.16, p = 6.34 × 10−7), Day (F4,116 = 396.98, p < 2 × 10−16) and a Substrain x Day interaction (F4,116 = 44.59, p < 2 × 10−16). Tukey's post hoc test showed a significant increase in distance traveled in B6J versus B6NJ males on Day 3, Day 4 and Day 5 (*all p'sadjusted < 0.0001; Figure 1F).

To summarize, these results replicate previous B6 substrain differences in the B6 mice from the J lineage show increased stimulant sensitivity compared with B6 mice from the N lineage.12

3.2 Chromosome 11 QTL near the Cyfip2 missense mutation underlying variation in sensitized methamphetamine-induced maximum speed (m/s)

Next, we sought to identify the genetic basis of differential methamphetamine-induced stimulant sensitivity in B6 substrains using an F2 Reduced Complexity Cross.8, 9 We identified a genome-wide significant QTL on chromosome 11 underlying maximum speed while mobile that was specific to methamphetamine treatment and emerged after the second methamphetamine injection on Day 4 (LOD = 3.5; p = 0.039, peak = 21 cM [36 Mb]; Bayes interval: 18–39 cM [31–63 Mb]; 11% of the phenotypic variance explained; Figure 2A) and the third methamphetamine injection on Day 5 (LOD = 4.2, p = 0.009; peak = 21 cM [36 Mb]; Bayes interval: 18–34 cM [31–57 Mb]; 11% of the phenotypic variance explained; Figure 2A). The marker nearest the peak (rs48169870; 18 cM [31 Mb]) was located just proximally to the Cyfip2 missense mutation (rs24064617401; 28 cM [46 Mb]; Figure 2B) previously identified for acute and sensitized cocaine velocity.12 Like the previous finding, the effect plot for maximum speed indicated that the B6J allele was associated with increased methamphetamine-induced maximum speed (Figure 2C). Thus, the locus containing the Cyfip2 missense mutation is associated with behavioral sensitivity to multiple psychostimulants.

Genome-wide significant QTL on chromosome 11 near Cyfip2 underlying variation in sensitized methamphetamine-induced maximum speed. Mice were treated on Day(D) 1 and D2 with saline (i.p.) and on D3, D4 and D5 with methamphetamine and behavioral activity was recorded over 30 min. (A) Genome-wide significant QTL on chromosome 11 for maximum speed following the second methamphetamine injection on D4 and following the third methamphetamine injection on D5. Solid horizontal lines for panels A and B indicate significance thresholds for each phenotype (p < 0.05). (B) Chromosome 11 QTL plot for maximum speed on D1 through D5. (C) Effect plot of maximum speed as a function of Genotype at the peak locus for maximum speed on D1 through D5. J, homozygous for B6J allele; BN, heterozygous; N, homozygous for B6NJ allele

3.3 Chromosome 5 QTL near the Gabra2 intronic deletion underlying variation in acute methamphetamine-induced locomotor activity (total distance traveled; m)

We identified two genome-wide significant QTLs on chromosome 5 that influenced locomotor activity (distance, m). The first QTL was for Day 2 distance traveled (saline) and was localized more proximally on chromosome 5 (peak = 22 cM [41 Mb]; LOD = 3.8; p = 0.037; Bayes: 14–34 cM [29–66 Mb]; 10% of the phenotypic variance explained Figure 3A,B). A complete list of SNPs, indels and SVs for the Day 2 QTL was obtained from the Sanger database (https://www.sanger.ac.uk/)33, 35 and is provided in Table S1. The second QTL was for distance traveled D3 following 2 mg/kg methamphetamine and was more distally localized near the Gabra2 intronic deletion (71 Mb) (peak = 35 cM [67 Mb]; LOD = 5.2; p < 0.001; Bayes interval: 32–47 cM [60–95 Mb]; 14% of the phenotypic variance explained; Figure 3A,B). A similarly localized QTL was also detected on D4 following the second methamphetamine injection (peak = 34 cM [66 Mb]; LOD = 3.6; p = 0.034; Bayes: 16–46 cM [30–93 Mb]; 12% of the phenotypic variance explained; Figure 3A,B). A complete list of SNPs, indels and SVs for the Day 3 (the most robust methamphetamine QTL) was obtained from the Sanger database (https://www.sanger.ac.uk/)33, 35 and is provided in Table S2. The effect plot of total distance traveled for D1 through D5 over 30 min at the peak associated marker (rs29547790; 70.93 Mb) indicated that the B6J allele was associated with increased methamphetamine-induced distance traveled (Figure 3C). In examining the time course for the chromosome 5 QTL for the sex-combined data set, the J allele was consistently associated with increased methamphetamine-induced distance traveled across the six, 5-min time bins (Figure 3D). RM ANOVA indicated a main effect of Genotype (F2,174 = 7.20; p = 0.001), Sex (F1,174 = 3.89; p = 0.05), Time (F5,870 = 485.02; p < 2 × 10−16), a Genotype x Time interaction (F10,870 = 2.39; p = 0.0085) but no interactions with Sex (p's > 0.46). There was a significant increase in methamphetamine-induced distance traveled in mice with the J/J allele relative to the J/N and N/N alleles (*Tukey's padjusted < 0.05 for the three comparisons at each time point; Figure 3D). Because we identified both a Sex x Day interaction and a Substrain x Day interaction in distance traveled in the parental substrains, we broke down and analyzed the time course of the effect plot separately in females and males and found that males clearly drove the QTL effect compared with females (Figure 3E,F). For the females-only data set, RM ANOVA indicated no effect of Genotype (F2,103 = 2.13), an effect of Time (F5,515 = 290.22; p < 2 × 10−16) and no interaction (p = 0.51). Simple contrasts did not identify any significant differences among genotypes at any time point (p's > 0.05). For males-only, there was a main effect of Genotype (F2,71 = 7.49; p = 0.0011), Time (F5,355 = 195.53; p < 2 × 10−16) and a Genotype x Time interaction (F10,355 = 1.85; p = 0.05). Mice with the J/J allele showed a significant increase in methamphetamine-induced distance traveled compared with both J/N and N/N (*Tukey's padjusted < 0.05 for the three comparisons at each time point; Figure 3F).

Genome-wide significant QTL on chromosome 5 near Gabra2 underlying variation in acute methamphetamine-induced distance traveled. (A) Genome-wide significant QTL on chromosome 5 for distance traveled (m) on Day(D) 2 over 30 min following i.p. saline, a second QTL for distance traveled D3 following 2 mg/kg methamphetamine, and a third, similarly localized QTL on D4 following the second methamphetamine injection. Solid horizontal lines for panels A and B indicate significance threshold for each phenotype (p < 0.05). (B) Chromosome 5 QTL plot for distance traveled on D1 through D5. (C) Effect plot of total distance traveled for D1 through D5 over 30 min at the peak associated marker (rs29547790; 70.93 Mb). (D) Time course in 5-min bins of the effect plot for acute methamphetamine-induced distance traveled on D3. * = significant increase in J/J relative J/N and N/N (Tukey's padjusted < 0.05 for the three comparisons at each time point). (E) Time-course in females. (F) Time-course in males. * = significant increase in J/J relative J/N and N/N (Tukey's padjusted < 0.05 for the three comparisons at each time point). J = homozygous for B6J allele; BN = heterozygous; N = homozygous for B6NJ allele. Green, dashed traces denote distance traveled on D3

Power analysis of the 184 F2 mice that we phenotyped in this study indicated that we had ~80% power to detect QTLs explaining at least 10% of the variance in maximum speed and in distance traveled (Figure S1).

Because there was an age range of the F2 mice and because age can affect both baseline and drug-induced locomotor activity, we also tested for Age as a covariate for the 10 activity-related phenotypes; Age was only a significant covariate for D1 and D2 distance and not for any of the eight other phenotypes (ps > 0.4; Table S3). When we re-ran QTL analyses for each of the 10 phenotypes with Age as a covariate (along with Sex), all the existing significant results remained significant except for D2 Distance which was no longer statistically significant (chromosome 5: LOD = 3.1; p = 0.10; Table S3). All the nonsignificant results without Age as a covariate remained nonsignificant (Table S3). Thus, none of our fundamental results/conclusions change, regardless of whether Age is included as a covariate.

3.4 Striatal cis-eQTL analysis identifies Gabra2 as the top transcript associated with rs29547790; the peak chromosome 5 marker linked to methamphetamine-induced distance traveled

Given that the likely causal variant underlying maximum speed induced by methamphetamine for the chromosome 11 locus was a missense variant in Cyfip2,12 we turned our attention to the novel chromosome 5 locus whose peak marker (rs29547790; 70,931,531 bp) is located just proximally to a functional intronic deletion in Gabra2 (71 Mb)13 which codes for the alpha-2 subunit of the GABA-A receptor. In a genome-wide cis-eQTL analysis, we examined eQTLs associated with the peak chromosome 5 marker associated with methamphetamine-induced locomotor activity (rs29547790; 70.93 Mb). The top transcript showing an association with rs29547790 was Gabra2 (chromosome 5, 70.96 Mb; padjusted = 7.24 × 10−27; Table 1; see Table S4 for a complete list of cis-eQTLs). Given this finding and the prior literature implicating Gabra2 in psychostimulant behavioral responses,54 Gabra2 is a top candidate quantitative trait gene underlying the chromosome 5 QTL for methamphetamine-induced locomotor activity.

TABLE 1. Cis-eQTL gene transcripts showing peak marker association in transcript variance with rs29547790 (chromosome 5: 70.9 Mb) (FDR < 0.05) Gene Chr.5 location (Mb) Distance from rs29547790 (Mb) Log2FC (JJ vs. NN) p-value Adjusted p-value Gabra2 70.96 0.026 −0.88 1.66E−30 7.24E−27 Muc3a 137.21 66.28 0.56 2.60E−10 1.14E−07 Atp8a1 67.62 3.08 0.18 1.28E−06 9.57E−05 Ptpn11 121.13 50.20 0.16 2.17E−05 0.000778 Cxcl5 90.76 19.83 −2.41 3.22E−05 0.001042 Fgfr3 33.72 37.19 0.19 0.000123 0.002843 Cux2 121.86 50.92 −0.33 0.000136 0.003049 Ankrd13a 114.77 43.84 0.27 0.0