Developmental age and FAAH genetic variation converge to mediate fear regulation in female mice

Abstract

Anxiety disorders are more prevalent in females than in males yet a majority of basic neuroscience studies are performed in males. Furthermore, anxiety disorders peak in prevalence during adolescence, yet little is known about neurodevelopmental trajectory of fear expression, particularly in females. To examine these factors we fear conditioned juvenile, adolescent, and adult female mice and exposed them to fear extinction and a long-term recall test. For this, we used knock-in mice containing a common human mutation in the gene for fatty acid amide hydrolase (FAAH), the primary catabolic enzyme for the endocannabinoid anandamide (FAAH-IN). This mutation has been shown to impart a low-anxiety phenotype in humans, and rodents relative to their wild-type littermates. We find an impact of the FAAH polymorphism on developmental changes in fear behavior. Specifically, the FAAH polymorphism appears to induce a state of hypervigilance (increased fear) during adolescence. We also used markerless pose estimation software to classify alternative behaviors outside of freezing. These analyses revealed age differences in vigilance to indicators of threat and in the propensity of mice to explore aversive environment, though genotypic differences were minimal. These findings address a gap in the literature regarding developmental patterns of fear learning and memory as well as the mechanistic contributions of the endocannabinoid system in females.

Introduction

Adaptive behavior is defined by the ability to adjust responding based on previous experiences and the immediate environment. In many cases, an organism must evaluate the likelihood of threat in a given environment and use this calculation to inform behavioral selection. While persistently anticipating threat can aid the avoidance of harm, it can also limit the pursuit of other goal directed behaviors without appropriate regulation. A predominant focus on threat is a key characteristic of many anxiety disorders. Anxiety disorders are nearly twice as prevalent in females than in males and are associated with greater comorbidities and burden of illness (McLean et al., 2011). Yet, basic neuroscience studies of sex differences in behavioral, circuit, and molecular mechanisms of fear learning and memory have only begun to scratch the surface (Bangasser & Cuarenta, 2021). A greater understanding of how fear manifests behaviorally as well as the neural mechanisms by which fear can be regulated in females, is necessary to elucidate the pathology underlying anxiety disorders and develop efficient treatments.

The endocannabinoid system is well known for its widespread contributions to cognitive, behavioral, and physiological processes, including the regulation of stress, fear, and anxiety (Lu & Mackie, 2016, 2021; Maldonado et al., 2020; Mechoulam & Parker, 2013; Morena et al., 2016). The endocannabinoid neuromodulatory network primarily includes CB1 and CB2 receptors, inhibitory G-protein-coupled receptors that bind the endogenous ligands anandamide and 2-arachidonoylglycerol. Regulation of endocannabinoid signaling is accomplished by the enzymes fatty acid amide hydrolase (FAAH) and monoacylglyceride lipase, which hydrolyze anandamide and 2-arachidonoylglycerol, respectively.

Across species in adulthood, both naturally occurring and artificially induced fluctuations in endocannabinoid signaling have been shown to dramatically impact behavioral and physiological responses to stress and threat (Maldonado et al., 2020; Petrie et al., 2021). Genetic variation between individuals can lead to dramatic alterations in endocannabinoid signaling. One notable example is a common single nucleotide polymorphism in the human FAAH gene (C385A; rs324420), present as either the more common AC or rare AA genotype in ~38% of individuals of Western European descent though prevalence ranges from ~23-68% have been reported across different backgrounds (Cariaso & Lennon, 2019; Lugo Nevarez, 2020; 1000 Genomes Project Consortium, 2012). The variant allele disrupts FAAH enzymatic activity, resulting in increased levels of anandamide (Chiang, 2004; Sipe et al., 2002). Human carriers of the variant allele exhibit reduced trait anxiety, lower stress reactivity, facilitated fear extinction, and enhanced extinction recall (Dincheva et al., 2015; Gunduz-Cinar et al., 2013; Mayo et al., 2020). Similarly, knock-in mice that express the variant FAAH A allele also exhibit greater reductions in freezing across extinction and decreased anxiety-like behaviors in a series of anxiety tasks (Dincheva et al., 2015; Gee et al., 2016). The majority of studies using mice genetically modified to target the endocannabinoid system use exclusively male subjects, leaving a gap in the literature regarding the impact of the FAAH polymorphism on fear behavior in females.

Moreover, fluctuations in endocannabinoid signaling in early life contribute to the maturation of corticolimbic circuitry, likely setting the stage for affective processes across the lifespan (Meyer et al., 2018). Yet, little is known about the neurodevelopmental contributions of the endocannabinoid system to profiles of fear learning and memory across juvenile and adolescent periods.

The goal of the present study was to investigate the interaction between age and endocannabinoid signaling in female mice using the variant FAAH transgenic mouse line. Mice underwent fear conditioning and extinction during juvenile, adolescent, or adult ages, followed by a long-term fear memory test which took place in adulthood for all groups. A secondary goal of this study was to consider alternative fear-related phenotypes. While freezing is a valuable proxy for fear in rodents, recent work has emphasized that the behavioral manifestations of fear can be much more dynamic (Greiner et al., 2019, 2019; Gruene et al., 2015; Mitchell et al., 2022). Thus, we used markerless pose estimation to quantify pivoting, head elongation, grooming, rearing, and jumping, as measures of exploration or engagement with the environment that are not detectable through analyses restricted to freezing.

Our study adds to a sparse but important body of literature considering fear regulation specifically in females. Our results indicate robust developmental age driven differences and modest impacts of FAAH genetic variation. These findings have the potential to inform treatments of fear-related psychiatric disease tailored to specific age, sex, and genetic background.

Materials & methods

Subjects

Mice homozygous for the FAAH C385A mutation (FAAH^C/A) were generated as previously described (Dincheva et al., 2015). Heterozygous breeding pairs were mated in-house to produce mice homozygous for the FAAH C385A mutation (FAAH-IN) and control mice without the FAAH C385A mutation (wild-type). Female pups were weaned on postnatal day (PND) 21 and group-housed in cohorts of 3-5 per cage. Mice were group-housed throughout the study. Cages were randomly assigned to undergo testing as juveniles (PND 22-27; n = 14 wild-type, 19 FAAH-IN), adolescents (PND 30-36, n = 20 wild-type, 20 FAAH-IN), or adults (PND 68-77, n = 20 wild-type, 21 FAAH-IN). Mice were maintained on a 12-h light/dark cycle at 18-22°C with food (LabDiet, PicoLab Rodent Diet 20) and water ad libitum. All behavioral tests were conducted during the light cycle. Experiments were carried out in accordance with the National Institutes of Health’s Guide for the Care and Use of Laboratory Animals and protocols were approved by the Weill Medical College of Cornell University’s Institutional Animal Care and Use Committee.

Behavioral apparatus

Behavioral procedures were carried out in standard conditioning chambers (Med Associates) as previously described (Meyer et al., 2019).The auditory conditioned stimulus (CS) was a 5 kHz tone played at 80 dB for 30 s, delivered through a speaker mounted 13 cm above the chamber floor. During fear conditioning, the aversive US was a 0.7 mA foot shock (1 s duration) that co-terminated with the tone. Fear conditioning and fear extinction were carried out in separate contexts (Context A and Context B, respectively). Fear recall tests were carried out in Context B. In Context A, ambient light was provided throughout the session through a light box (125 lx; ceiling-mounted 52 cm above the floor). The chamber walls were aluminum (side walls) and clear acrylic (front, back, and top). The chamber floor was a stainless-steel grid. The chamber was scented with peppermint (1/1000 in 70% ethanol). In Context B, ambient light was provided through an LED Stimulus Light (50 lx; 18 cm above the grid floor). The chamber interior was altered using a white acrylic cylindrical contextual insert and a white acrylic floor cover. The chamber was scented with (−)-Limonene, 92 % (1/1000 in 70 % ethanol). All behavioral sessions were recorded and analyzed using VideoFreeze^® software (Med Associates).

Behavioral procedures

A schematic of the behavioral protocol used in this study is presented in Figure 1.

Figure 1. Schematic of Experimental Design.

Juvenile (postnatal day, PND 22-27), adolescent (PND 30-36), and adult (PND 68-77) mice underwent fear conditioning and fear extinction on consecutive days, followed by a long-term fear recall test four weeks later. CS = Conditioned Stimulus.

Fear conditioning

Mice were acclimated to the conditioning chamber (Context A) for two minutes then presented with three tone-shock pairings on a 30 s ITI schedule. Mice remained in the conditioning chamber for one minute after the final tone-shock pairing before being returned to their home cages.

Fear extinction

Extinction took place 24 hours after fear conditioning in a new context (Context B) to isolate cued fear from residual contextual fear. After a two-minute baseline, the mice experienced five presentations of the CS in the absence of the foot shock on a 30 s ITI schedule. Mice remained in the conditioning chamber for one minute after the final tone before being returned to their home cages.

Fear Recall Tests

To provide a measure of extinction retention, mice underwent a long-term fear recall test four weeks after extinction. The test took place in Context B and consisted of a 2 min baseline period followed by a single CS presentation, after which mice remained in the conditioning chamber for one minute after the tone before being returned to their home cages.

Phenotype Analysis

Phenotype Classification

Our examination of phenotypes during fear conditioning, extinction, and recall sessions can be grouped into three categories: fear-related behaviors (freezing, head elongation, pivoting, jumping), environment-engagement (rearing, walking), and self-engagement (grooming). Head elongation and pivoting were interpreted as fear-related phenotypes as they often occurred interspersed within freezing bouts, though may reflect greater deliberation over the potential for threat than bouts in which the mouse is fully immobile (i.e., freezing). Jumping was interpreted as an active fear response possibly indicative of an escape attempt.

DeepLabCut analyses

Conditioning chamber videos were analyzed to quantify grooming, rearing, head elongation, pivoting, walking, and jumping behaviors. Automated detection of these behaviors was achieved using the markerless pose-estimation software DeepLabCut (version 2.2.1), which provides location tracking for individual mouse body parts (Mathis et al., 2018). Specifically, 10 frames were extracted from 27 example videos each, and nose tip, paws, ears, tail base, and four additional midline points extending from the forelimbs to the hindlimbs were manually labeled for each frame. 95% of these frames were then used to train a ResNet-50-based neural network for 95000 training iterations, ending with a train error: 1.66 pixels, test error: 3.44 pixels (image size was 320 by 240 pixels), with a p-cutoff of 0.6.

With the tracking data produced by DeepLabCut, jumping behavior in a given video was detected by analyzing the locations of the tail base and left/right rear paws. For each of these three body parts, the y-axis coordinates were collected for every frame of the video; these y-axis coordinates were scanned for outliers, specifically those points greater than

$$\text{Q}3+1.5\left(\text{Q}3-\text{Q}1\right)$$

where Q3 and Q1 are the upper and lower quartiles respectively. Video frames in which all three body parts contained a y-axis outlier were considered to be a single jump event if the previous frame contained no y-axis outliers for the three body parts. This method reliably detected all jump events when compared against hand-scored ethograms of jumping behavior.

Grooming and rearing behavior was detected using B-SOiD, an openly available behavioral classification software (Hsu & Yttri, 2021). Using tracking data produced by DeepLabCut, B-SOiD trains on example videos by extracting spatiotemporal relationships between body parts and clustering the extracted relationships into separate presumed behaviors; B-SOiD then implements a machine learning classifier to predict behavior category in novel videos based on these spatiotemporal patterns.

Here, B-SOiD was trained on 25 videos, using tracking data for nose tip, front paws and sagittal plane points. Clustering was performed using a size of 0.2-1.0%. Video clips belonging to each cluster were generated and inspected, resulting in identification of four grooming clusters and two rearing clusters. Novel videos were then processed, and hand-scored ethograms of grooming and rearing were compared against B-SOiD predictions to verify that these clusters captured the behaviors of interest. Time spent grooming or rearing was then calculated from the number of frames B-SOiD assigned to the respective behavior clusters.

Following these analyses, patterns of movement in the tail base, forelimb midline, and nose points were used to identify walking, pivoting, and head elongation behaviors in the remaining video frames. Walking behavior was defined as a movement in all three body parts, pivoting as movement in both forelimb midline and nose points, and head elongation as movement only in the nose point. A body part was considered moving if the pixel distance between the location of the body part in a given video frame and the average location of that body part in the preceding five frames exceeded a threshold value derived from distance data in frames that were manually identified to contain body part movement. Visual inspection of processed videos and corresponding frame data verified that threshold values ranging from 1-3 pixels captured body part movement well.

Statistical Analyses

VideoFreeze^® (Med Associates) was set at a motion threshold of 18 Units for automatic scoring of freezing. The percentage time spent freezing was calculated by dividing the cumulative time freezing during the CS by the CS duration (30s). For the main analyses and within-genotype follow-up analyses we use linear regression models as shown in Table 1 and described below. To analyze freezing during fear conditioning, we ran a linear model of percent freezing regressed on age, tone, and genotype. In a subsequent analysis, we ran a linear model of percent freezing regressed on age and tone for wild-type and FAAH-IN mice separately. For fear extinction learning, we ran a linear model of percent freezing regressed on age, extinction day, and genotype. In a follow-up analysis, we ran a linear model of percent freezing regressed on age and extinction day for wild-type and FAAH-IN mice separately. Additionally, to examine the rate of extinction learning across extinction days, we ran a linear model of percent freezing regressed on age, extinction day, and genotype and with an age-extinction day interaction. In a follow-up analysis, we ran a linear model of percent freezing regressed on an age by extinction day interaction for wild-type and FAAH-IN mice separately. To analyze the magnitude of learning across extinction, we ran a linear model of differential freezing regressed on age and genotype. In a follow-up analysis, we ran a linear model of differential freezing on age for wild-type and FAAH-IN mice separately. For freezing to the tone during recall, we ran a linear model of percent freezing regressed on age and genotype. In a follow-up analysis, we ran a linear model of percent freezing regressed on age for wild-type and FAAH-IN mice separately. For each phenotype, we ran linear models of time engaging in the behavior during fear conditioning, extinction, or recall on age for wild-type and FAAH-IN mice separately. Across all analyses, differences were considered significant for P values less than 0.05. Differences were considered trending for P values less than 0.08. Statistical analyses were performed using R.

Table 1. Summary of statistical analyses.

For the main analyses and within-genotype follow-up analyses we use the above linear regression models to estimate effects in which Y captures the outcome variable, β are the estimated coefficients for each independent variable, and ε is the error term.

Test	Outcome Variable (Y)	Equation	Follow-up analyses in wild-type and FAAH separately
Fear Conditioning	Freezing (%)	$Y=\alpha +{\beta }_{1}age+{\beta }_{2}tone+{\beta }_{3}genotype+ϵ$	$Y=\alpha +{\beta }_{1}age+{\beta }_{2}tone+ϵ$
Fear Extinction	Freezing (%)	$Y=\alpha +{\beta }_{1}age+{\beta }_{2}day+{\beta }_{3}genotype+ϵ$	$Y=\alpha +{\beta }_{1}age+{\beta }_{2}day+ϵ$
Learning Rate across Extinction	Freezing (%)	$Y=\alpha +{\beta }_1(age\times day)+{\beta }_{2}genotype+ϵ$	$Y=\alpha +{\beta }_1(agextone)+ϵ$
Differential Freezing	Differential Freezing	$Y=\alpha +{\beta }_{1}age+{\beta }_{2}genotype+ϵ$	$Y=\alpha +{\beta }_{1}age+ϵ$
Extinction Recall	Freezing (%)	$Y=\alpha +{\beta }_{1}age+{\beta }_{2}genotype+ϵ$	$Y=\alpha +{\beta }_{1}age+ϵ$
Phenotypes	Time engaged in behavior	n/a	$Y=\alpha +{\beta }_{1}age+ϵ$

Results

Freezing Behavior

Fear Conditioning

Detailed statistical results are reported in Tables S1 and S2. Mice were exposed to three tone-shock pairings (Figure 2A). Overall, mice increased freezing across the three tones (t = 19.20, p < 0.001). Additionally, across fear conditioning, adult mice froze less than juveniles (t = 4.77, p < 0.001) and adolescents (t = 3.40, p < 0.001), respectively. There were no significant differences in freezing to the tone between juveniles and adolescents (p = 0.125). There was no effect of genotype on fear conditioning (p = 0.434).

Figure 2. Freezing behavior during fear conditioning, extinction, and long-term recall.

Learning curves for female mice during (A) fear conditioning and (B) extinction showing average percentage of time spent freezing during the tone. Regardless of genotype, adults froze less to the tone during fear conditioning. Across extinction, adolescents froze more than both juveniles and adults. Wild-type adolescents froze significantly more than juveniles but only marginally more than adults. FAAH-IN adolescents froze significantly more than both juveniles and adults. (C) Spontaneous recovery across extinction days, shown as percent rebound. Wild-type juveniles showed more spontaneous recovery relative to wild-type adults. Both FAAH-IN juveniles and adolescents showed stronger spontaneous recovery than adults. (D) Differential freezing from early to late trials across extinction showed that there were no significant age or genotype effects on extinction learning across days. (E) Percent freezing to the tone during the four-week long-term recall test. For both wild-type and FAAH-IN mice, juveniles froze significantly less to the tone than adolescents and adults. Top row = wild-type (WT) mice, bottom row = FAAH-IN mice. T = tone; D = day.

Next, we analyzed age differences in fear conditioning within each genotype. For this, we regressed percent freezing on age and tone within each genotype. Wild-type mice increase freezing across the three tones (t = 13.60, p < 0.001). Additionally, across fear conditioning, wild-type adults froze less than juveniles (t = 3.74, p < 0.001) and adolescents (t = 2.79, p = 0.005). There were no significant differences in freezing across conditioning between wild-type juveniles and adolescents (p = 0.226). FAAH-IN mice increased freezing across the three tones (t = 13.50, p < 0.001). Similarly, FAAH-IN adults froze less across conditioning than juveniles (t = 3.01, p = 0.003) and adolescents (t = 2.02, p = 0.043). There were no significant differences in freezing across conditioning between FAAH-IN juveniles and adolescents (p = 0.314). Taken together, these results suggest that in both genotypes, tone-elicited freezing decreases with maturity.

Extinction

For extinction training, we looked at the average cue-elicited freezing across the four days of extinction (Figure 2B). In our first model, we regressed percent freezing on day, age, and genotype. Overall, mice reduced freezing each day (t = 7.40, p < 0.001). Additionally, adolescents froze more than juveniles (t = 4.86, p < 0.001) and adults (t = 4.88, p < 0.001), across extinction training. There were no significant differences in freezing between juveniles and adults (p = 0.794). There was no effect of genotype on freezing across extinction training (p = 0.358). Because adolescents froze more than juveniles and adults, we checked for differences in the rate of learning across extinction. Therefore, in a second model we included an interaction term between age and day. We found that juveniles learned at a similar rate to both adolescents (p = 0.249) and adults (p = 0.413).

Although there was no overall effect of genotype on freezing across extinction, pre-existing questions regarding the impact of genotype on the development of fear regulation led us to a third model in which we regressed percent freezing on age and day within each genotype. Wild-type mice reduced freezing each day across extinction training (t = 5.68, p < 0.001). Freezing across extinction differed significantly between wild-type adolescents and juveniles, with juveniles freezing less than adolescents (t = 2.33, p = 0.020). Only marginal differences in freezing were apparent between adolescents and adults, with adults also freezing somewhat less than adolescents (t = 1.91, p = 0.056). There were no differences in freezing between juveniles and adults (p = 0.552). FAAH-IN mice reduced freezing each day across extinction training (t = 4.88, p < 0.001), with FAAH-IN adolescents freezing more than both juveniles (t = 4.53, p < 0.001) and adults (t = 4.92, p < 0.001). There were no significant differences in freezing between FAAH-IN juveniles and adults (p = 0.783).

To examine age and genotype differences in recovery of fear responding between extinction sessions we analyzed spontaneous recovery across extinction, measured as freezing to the first tone of the extinction day minus freezing to the last tone of the prior extinction day (Figure 2C). For this, we regressed spontaneous recovery by extinction day, age, and genotype. Overall, mice reduced their spontaneous recovery across extinction days (t = 6.02, p = 0.013). There was no effect of genotype on spontaneous recovery (p = 0.838). We observed an age-dependent increase in spontaneous recovery such that juveniles showed significantly stronger spontaneous recovery relative to adolescents (t = 2.10, p = 0.036) and adults (t = 5.04, p < 0.001) and adolescents exhibited significantly stronger spontaneous recovery relative to adults (t = 3.09, p = 0.003). When we subsetted the data by genotype, we found that wild-type mice did not reduce spontaneous recovery across extinction (p = 0.357). Furthermore, while wild-type juveniles (t = 2.59, p = 0.010) exhibited significantly more spontaneous recovery relative to wild-type adults, there were no significant differences between juveniles and adolescents (p = 0.094) or adolescents and adults (p = 0.316). In contrast, FAAH-IN mice showed a significant reduction in spontaneous recovery across extinction (t = 2.64, p = 0.009). Furthermore, both FAAH-IN juvenile (t = 4.61, p < 0.001) and adolescent (t = 3.45, p < 0.001) mice exhibited significantly more spontaneous recovery than adults. There were no significant differences between FAAH-IN juveniles and adolescents (p = 0.232).

We looked at the rate of learning across age within each genotype. To measure this, we subsetted the data by genotype and regressed freezing on an interaction between age and day (i.e., slope, Tables S1 and S2). We found that in both wild-type and FAAH-IN mice, juveniles learned at a similar rate to both adolescents (wild-type: p = 0.598; FAAH-IN: p = 0.269) and adults (wild-type: p = 0.243; FAAH-IN: p = 0.996). Adolescents and adults also learned at a similar rate (wild-type: p = 0.480; FAAH-IN: p = 0.254). This was confirmed by an analysis of differential freezing (the average freezing on day 4 minus the average freezing on day 1 of extinction; Figure 2D). We ran linear models that regressed differential freezing on age and genotype and on age within each genotype separately. Overall, there were no significant age (all p < 0.145) or genotype (p = 0.648) differences in differential freezing. These results suggest that while adolescent mice exhibit elevated freezing during extinction, an effect exacerbated by the presence of the FAAH polymorphism, the rate of fear reduction is comparable across all mice.

Recall

Four weeks following the final day of extinction training, mice were returned to the extinction context to assess the durability of extinction memory (Figure 2E). We regressed freezing to the recall tone on age and genotype. We found that juveniles froze less than adolescents (t = 5.68, p < 0.001) and adults (t = 4.88, p < 0.001), respectively. There were no differences between adolescents and adults (p = 0.382) and no effect of genotype on recall (p = 0.458).

Next, we subsetted the data by genotype and regressed freezing to the recall tone on age. Similar to data collapsed across genotype, we find that in both wild-type and FAAH-IN mice, juveniles froze significantly less than adolescents (wild-type: t = 3.63, p < 0.001; FAAH-IN: t = 4.33, p < 0.001) and adults (wild-type: t = 3.26, p = 0.001; FAAH-IN: t = 3.58, p < 0.001). No differences were found between adolescents and adults of either genotype (wild-type: p = 0.681; FAAH-IN: p = 0.417). These data seem to suggest that juvenile mice of both genotypes exhibit the strongest extinction memory, though an alternative explanation may be that the fear memory formed following juvenile conditioning is more labile, and extinction leads to partial erasure of the fear memory (e.g. Gogolla et al., 2009; Pattwell et al., 2012).

Extended behavioral classification

Another focus of the paper was to examine alternative fear-related phenotypes and how developmental trajectories of these phenotypes are impacted by FAAH genetic variation (Figure 3). For this, we used markerless pose estimation software to detect instances of head elongation, pivoting, rearing, walking, grooming, and jumping during tone presentations in the same mice. Eight mice were excluded from analysis due to errors converting video recordings of behavior for analysis by DeepLabCut and B-SOiD, resulting in the following n’s: juvenile, 13 wild-type, 16 FAAH-IN; adolescent, 18 wild-type, 19 FAAH-IN; adult, 19 wild-type, 21 FAAH-IN. Detailed statistical results are reported in Table S3.

Figure 3. Extended behavioral classification.

Percent time female mice are engaged in rearing, walking, grooming, head elongation, pivoting, freezing, or other behaviors across tones for (A, B) fear conditioning, (C, D) extinction, and (E, F) recall. Although individual phenotype stats are run on total seconds, data is presented as percent time for ease of visualization. WT = wild-type.

Freezing

Results reported in this section are nearly identical to those outlined above, with the exception that tone-elicited freezing data was collapsed across extinctions sessions. In contrast to the other fear-related phenotypes discussed in this section, freezing data was extracted from VideoFreeze and not DeepLabCut analyses. During fear conditioning, wild-type juveniles (t = 3.16, p = 0.002) and adolescents (t = 2.08, p = 0.038) froze more than adults, though did not differ from one another (p = 0.214). FAAH-IN juveniles froze more than adults (t = 2.87, p = 0.004). No other differences were observed for FAAH-IN mice (all p > 0.111).

During extinction, wild-type adolescents froze more than juveniles (t = 2.07, p = 0.039). No other differences were observed for wild-type mice (all p > 0.178). FAAH-IN adolescents froze more than juveniles (t = 3.64, p < 0.001) and adults (t = 4.04, p < 0.001), while juveniles and adults did not differ (p = 0.913).

During the recall session, time spent freezing was lowest in juvenile mice. Wild type juveniles froze less than adolescents (t = 2.48, p = 0.013) and adults (t = 2.63, p = 0.009). FAAH-IN juveniles also froze less than adolescents (t = 2.80, p = 0.005) and adults (t = 2.54, p = 0.011). Adolescent and adult mice freezing did not differ for either genotype (all p > 0.736).

Head elongation

Differences in head elongation were limited to extinction, with FAAH-IN adolescents spending less time engaged in head elongation than juveniles (t = 3.03, p = 0.002) and adults (t = 2.59, p = 0.010). No other differences were observed across sessions for any age or genotype (all p > 0.153).

Pivoting

During fear conditioning, no differences in pivoting behavior were observed between ages of wild-type animals (all p > 0.788). In contrast, FAAH-IN adults spent more time pivoting than juveniles (t = 1.11, p = 0.004), though only marginally more time pivoting than adolescents (t = 1.83, p = 0.067). FAAH-IN juveniles and adolescents did not differ in time spent pivoting (p = 0.260).

During extinction, wild-type adults spent more time pivoting than adolescents (t = 2.29, p = 0.022), though no other differences were observed (all p > 0.199). FAAH-IN adults spent more time pivoting than both juveniles (t = 2.16, p = 0.031) and adolescents (t = 4.02, p < 0.001), but juveniles and adolescents did not differ from each other (p = 0.107).

Conversely, during the recall session, wild-type juveniles spent more time pivoting than adolescents (t = 2.76, p = 0.006) and adults (t = 3.33, p = 0.001). FAAH-IN juveniles also spent more time pivoting than adolescents (t = 3.37, p = 0.001) and adults (t = 3.10, p = 0.002). Adolescents and adults did not differ for either genotype (all p > 0.557).

Rearing

During fear conditioning and extinction, a relatively linear increase with age was observed for rearing behavior, regardless of genotype. During fear conditioning, wild-type juveniles reared less than adolescents (t = 2.19, p = 0.028) and adults (t = 3.86, p < 0.001). Adolescents also spent less time rearing than adults though the effect was marginal (t = 1.79, p = 0.073). FAAH-IN adults reared more than juveniles (t = 4.38, p < 0.001) and adolescents (t = 2.63, p = 0.009). Juveniles reared marginally less than adolescents (t = 1.83, p = 0.067).

During extinction, wild-type juveniles reared less than adults (t = 2.18, p = 0.029), though they did not differ from adolescents (p = 0.290). Wild type adolescents and adults also did not differ (p = 0.221). Compared to FAAH-IN adults, juveniles (t = 4.02, p < 0.001) and adolescents (t = 2.28, p = 0.022) spent less time rearing. Juveniles also spent less time rearing than adolescents though the effect was marginal (t = 1.83, p = 0.068). No differences in rearing were observed during the recall session for any age or genotype (all p > 0.168).

Walking

During fear conditioning, wild-type juveniles spent less time walking than adolescents (t = 2.05, p = 0.040) and adults (t = 3.06, p = 0.002), while no differences in walking were found for adolescents relative to adults (p = 0.279). Similarly, FAAH-IN juveniles walked less than adults (t = 2.32, p = 0.020) and only marginally less than adolescents (t = 1.86, p = 0.063). Again, no differences in walking were found for adolescents relative to adults (p = 0.660).

Interestingly, this pattern was reversed during extinction. Wild-type juveniles spent more time walking than adolescents (t = 3.44, p < 0.001) and adults (t = 4.54, p < 0.001). Similarly, FAAH-IN juveniles walked more than adolescents (t = 2.33, p = 0.019) and adults (t = 2.87, p = 0.004). This pattern persisted for the recall session. Wild-type juveniles walked more than adolescents (t = 2.67, p = 0.008) and adults (t = 2.65, p = 0.008). FAAH-IN juveniles walked more than adolescents (t = 2.84, p = 0.004) and adults (t = 2.74, p = 0.006). No differences were observed between adolescents and adults during extinction or recall for either genotype (all p > 0.246).

Grooming

During fear conditioning, wild-type juveniles engaged in more grooming than other ages, grooming more than adolescent counterparts (t = 2.21, p = 0.027) and marginally more than adults (t = 1.79, p = 0.074). Wild-type juveniles also engaged in 2.41 seconds more grooming than adolescents (t = 2.85, p = 0.004) during extinction, though did not differ from adults (p = 0.157). Wild-type adults also groomed somewhat more than adolescents during extinction though the effect was marginal (t = 1.95, p = 0.051). Grooming did not differ between ages for wild-type mice during recall (all p > 0.160) nor did it differ between FAAH-IN mice during any session (all p > 0.135).

Jumping

During fear conditioning (Figure 4A), FAAH-IN adults jumped marginally more than juveniles (t = 1.94, p = 0.052). During extinction (Figure 4B), FAAH-IN adults jumped significantly more than juveniles (t = 2.22, p = 0.027) and adolescents (t = 2.41, p = 0.016). Jumping did not differ between ages for wild-type mice during fear conditioning (all p > 0.079) or extinction (all p > 0.170). None of the mice from either genotype jumped during the recall tone.

Figure 4. Jumping during fear conditioning and extinction.

Total number of jumps by female mice across tones for (A) fear conditioning and (B) extinction. FAAH-IN adults jumped significantly more than juveniles or adolescents during extinction tones while there were no significant age differences for wild-type (WT) mice.

Discussion

Our study focused exclusively on female mice, aiming to address a gap in the literature regarding developmental patterns of fear learning and memory as well as the mechanistic contributions of the endocannabinoid system in females. Our results revealed a genotypic impact on developmental changes in fear behavior, such that the FAAH polymorphism appears to induce a state of hypervigilance (increased fear) during adolescence. In both genotypes, adolescent females froze more than juvenile counterparts, though elevations in freezing were exacerbated by the FAAH polymorphism, with significant differences between both adolescents and adults only apparent in FAAH-IN mice. Similarly, examination of spontaneous recovery between extinction days revealed that while juveniles of both genotypes show high spontaneous recovery, only in FAAH-IN mice did adolescents exhibit significantly higher spontaneous recovery than adults. FAAH-IN adolescents also exhibited fewer head elongations during extinction, indicating that freezing bouts were robust and relatively uninterrupted. An increased salience of affective cues, particularly those that predict threat, may serve adolescents an evolutionary advantage (Callaghan et al., 2019). Adolescence is characterized by increased exploration of novel environments as individuals transition to a relative state of independence. Thus, elevated freezing during early extinction sessions in FAAH-IN adolescent females may reflect a more adaptive response profile. Such an interpretation would require that fear levels do not remain high, as elevated and inflexible fear responding is a hallmark of dysregulated fear responding (in rodents) and anxiety disorders (in humans). Indeed, FAAH-IN adolescent females reduce freezing across sessions at a comparable rate to adult and juvenile counterparts. Furthermore, fear responding during the long-term recall test is comparable between FAAH-IN female mice trained during adolescence and those trained as adults, suggesting that differences in the strength of the fear memory are transient.

While males were not included in this study, a comparison of our findings with the broader literature indicates sex differences in how the FAAH polymorphism impacts fear regulation. This genetic profile has been associated with reduced anxiety-like phenotypes in the open field arena and elevated plus maze in FAAH-IN adult male mice (Dincheva et al., 2015; Gee et al., 2016). Furthermore, FAAH-IN adult males exhibit a greater reduction in freezing across extinction (Dincheva et al., 2015) and wild-type adult male rats exhibit reduced fear potentiated startle 24 hours following extinction when exposed to a FAAH inhibitor (AM404;(Chhatwal et al., 2005)). In contrast, our findings do not support a genotypically-mediated difference in freezing during extinction. While outside the parameters of our statistical model, in order to directly compare to existing literature, we ran an isolated analysis comparing freezing between adult FAAH-IN and adult wild-type mice each day of extinction (data not shown). Yet, no differences emerged. Additional work directly comparing across sexes will be helpful to better understand the interaction between FAAH genetic variation and sex. Across development, little has been done to examine the impact of the FAAH polymorphism on fear and anxiety phenotypes, although one study reported decreased anxiety-like phenotypes in the elevated plus maze in FAAH-IN adolescent males (Gee et al., 2016), possibly contrasting the fear hypervigilance we observe in FAAH-IN adolescent females. The same study showed no differences observed in juvenile males (Gee et al., 2016).

Although our study revealed an impact of the FAAH polymorphism on fear extinction, no major differences were seen for fear conditioning or recall. In contrast, for both genotypes, robust differences were observed across the three ages. Indeed, adult female mice froze less than adolescents and juveniles during fear conditioning. As discussed below, it is unclear whether this effect derives from differences in learning or differences in behavioral expression. Conversely, female mice trained as juveniles froze less than other groups during the long-term fear recall test, consistent with evidence in males that fear memories formed early in life may be more labile and subject to erasure (Gogolla et al., 2009; Pattwell et al., 2012). To our knowledge, this is the first evidence of a parallel reduction in recall in female mice. Previous work in rats has shown that, relative to juvenile males, juvenile females exhibit greater spontaneous recovery, though a comparison to older rats was not made (Park et al., 2017). Although juvenile rodents have been shown to retain long-term fear memories by PND 24, fear memories in slightly younger pups (P17) are unstable and are quickly forgotten, indicating the phenomenon of infantile amnesia where fear memories are unstable and quickly forgotten (Alberini & Travaglia, 2017; Callaghan & Richardson, 2011; Madsen & Kim, 2016; Pattwell & Bath, 2017). The juvenile mice used in this study were in the age range of PND 22-27, which overlaps with the closing of the infantile amnesia period. Thus, juvenile mice may be largely unsuccessful in recalling the original fear memory, independent of any degree of extinction learning (though some recall does occur, ~25%, suggesting a partially intact fear memory).

Interestingly, freezing did not differ between adolescent and adult groups during the recall test, contrasting evidence in males of poorer retention of extinction that occurred during adolescence relative to adulthood (Baker & Richardson, 2015; Kim et al., 2011; McCallum et al., 2010; Pattwell et al., 2012). One possible explanation is an earlier onset of puberty which may correspond to an earlier maturation of fear regulatory systems in female mice (Gerhard et al., 2021; Piekarski et al., 2017). Although, it is also important to note the timing of our recall test (4 weeks after extinction). While this recapitulates previous work from our laboratory (Gerhard & Meyer, 2021) is substantially longer than the more commonly used 24-hour recall time point. Additional work is needed to elucidate how sex and developmental age interact to influence the magnitude and longevity of extinction retention.

One caveat to consider is the difficulty determining whether differences in freezing between groups reflects differences in fear learning, the salience of the fear cue, or expression of freezing as a phenotype. If freezing manifests differently across mouse development, it is critical to take this into account when comparing behavior between ages, so as not to conflate a difference in learning with a difference in performance. In many cases freezing provides a reliable and robust measure of threat processing in rodents and is easily comparable across studies. Yet, previous work has shed light on sex differences in the behaviors elicited in response to fear stimuli (Greiner et al., 2019; Gruene et al., 2015; Mitchell et al., 2022). We put forth that the same may hold true for developmental age, as levels of exploration, fear responding, and fear regulation have been shown to vary alongside ethological landmarks of development (Callaghan et al., 2019).

Consideration of phenotypes outside of freezing revealed age differences in vigilance to indicators of threat, though genotypic differences were minimal. In an inverse pattern to freezing, which was observed more consistently in younger mice, adult mice exhibited incomplete immobility. We observed the pivoting phenotype in adults of both genotypes, and more frequent jumping in FAAH-IN mice, relative to juveniles and adolescents during fear conditioning and extinction sessions. These findings suggest a more active threat monitoring response is mounted by adult female mice, regardless of genotype. We have observed similar evidence of jumping in adult male mice relative to adolescents (Gerhard & Meyer, 2021). Active threat monitoring is likely related to contextual familiarity, as it appears to decrease alongside time spent in conditioning chambers. In contrast, female mice trained as juveniles appear to engage an active threat monitoring response specific to later behavioral tests, pivoting more than older mice during the recall session. Notably, we did not analyze darting for the mice included in this study, which has previously been reported in female rats (Greiner et al., 2019; Gruene et al., 2015; Mitchell et al., 2022) and mice (Fadok et al., 2017; Hersman et al., 2020; Trott et al., 2022), as there was insufficient evidence of this phenotype to train a neural network. Differences in experimental parameters, such as conditioning chamber size, shock intensity, and number of trials are possible explanations (Mitchell et al., 2022).

Our results also revealed age differences in the propensity of mice to explore aversive environments, although the observed trends differed between the two exploratory phenotypes considered (rearing and walking). More rearing was observed with increased age, particularly in early sessions when the conditioning chambers were less familiar. Conversely, while more walking was observed in developmentally older mice during fear conditioning, mice conditioned as juveniles walked more during extinction and recall. To the extent that walking can be interpreted as a measure of exploration and represent lower fear, our findings indicate strong initial fear learning in juvenile female mice, followed by a robust and long-lasting attenuation of fear during and after extinction. This interpretation is consistent with freezing data discussed above and previous literature suggesting lability of fear memories in juvenile male mice (Gogolla et al., 2009; Pattwell et al., 2012), extending these findings to female juvenile mice with both wild-type and FAAH-IN genotypes.

Conclusion

The developmental ages examined here reflect a timeframe during which profound maturation is seen across corticolimbic circuits, including fine-tuning by the endocannabinoid system (Meyer et al., 2018). This ongoing development likely accounts for age differences in fear learning and memory seen here. However, most of our understanding of the mechanism by which endocannabinoids regulate affective function comes from adults. Studies in adult humans have shown that carriers of the A385 allele exhibit enhanced fear regulation, strengthened frontoamygdala functional connectivity, and reductions in fear-elicited dorsal anterior cingulate cortex activity (Dincheva et al., 2015; Gee et al., 2016; Gunduz-Cinar et al., 2013; Mayo et al., 2020; Zabik et al., 2022). Similar findings have been revealed through adult rodent studies (Dincheva et al., 2015; Gee et al., 2016). By considering dynamics of fear learning and memory across three ages and specifically in females, a population largely underrepresented in basic neuroscience, our findings emphasize possible age and sex differences in how endocannabinoid genetic variation influences affective processes. We hope that this will serve as a launch pad to integrate neurobiologically focused studies in the future.

Finally, these findings emphasize the importance of considering additional phenotypes to bolster claims about fear learning and memory. In extending our analyses to consider alternative behaviors outside of freezing, we found that substantially more time was dedicated to fear-related phenotypes than exploration and grooming during cue presentations. Thus, mice may prioritize monitoring the external environment for threat over exploration and self-focused phenotypes. Still, 18-35% of the cue period was dedicated to phenotypes other than those discussed above. Deep learning analysis of behavior is not without limitation, and algorithms defined to capture a given phenotype may mislabel or entirely miss some behavioral events. Yet, even consideration of five phenotypes in addition to freezing does not appear to fully capture the rich diversity of animal behavior. As software platforms that provide automated analysis of rodent posture and correlation to behavioral repertoires have begun to burgeon in recent years, the future will be ripe for more comprehensive, and perhaps more ethologically valid consideration of how fear is expressed among different groups.

Data Availability Statement

Research data are not shared. Data will be made available upon reasonable request to authors.