Normalizing MeCP2 levels through genetic or pharmacological means is sufficient to rescue nearly all behavioral and molecular deficits in a mouse model of MDS (Sztainberg et al., 2015). For example, bolus injection of an antisense oligonucleotide (ASO) targeting MECP2 acutely reduces MeCP2 protein in MDS mice and ameliorates behavioral deficits (Shao et al., 2021b). We reasoned that genes whose expression was corrected after ASO treatment and whose abundance highly correlated with MeCP2 protein levels would be candidate genes to contribute to the behavioral rescue. To this end, we calculated the correlation coefficient between the fold-change of the 100 genes rescued by ASO treatment and the fold-change of MeCP2 protein levels in the hippocampi of MECP2 duplication mice (Shao et al., 2021b; Figure 1A). We found 16 genes whose fold-changes were highly correlated (Spearman’s rho >±0.75) with MeCP2 levels (Figure 1—figure supplement 1A). To further refine these candidates, we flagged genes that were predicted to be loss-of-function intolerant in humans as these genes could have biological consequences if their expression level is mildly altered (Karczewski et al., 2020; Lek et al., 2016). Of the 16 MeCP2-correlated genes, we identified four genes – growth differentiation factor 11 (Gdf11), family with sequence similarity 131 member B (Fam131b), versican (Vcan), and forkhead box protein P1 (Foxp1) with a probability of loss intolerance (pLI) greater than 0.9 (Figure 1—figure supplement 1A).

Gene expression changes in Mecp2 mutant mouse models have been inconsistent in published reports (Sanfeliu et al., 2019). To test if any of the four MeCP2-correlated, loss-intolerant transcripts were robustly sensitive to MeCP2 levels, we queried expression changes in 20 readily available transcriptional profiles generated in MECP2 perturbed rat (Veeraragavan et al., 2016), mouse, and human brain samples (Supplementary file 1). Strikingly, we found that Gdf11 was significantly altered in 19 of the 20 profiles, whereas the other three genes did not consistently exhibit significant difference compared to control (Figure 1B and Figure 1—figure supplement 1B–D). GDF11 levels were downregulated two-fold in human postmortem RTT excitatory neurons, while Gdf11 levels were downregulated in the transcriptomes of Mecp2-null male mice, Mecp2-null heterozygous female mice (Pohodich et al., 2018), and a panel of MECP2 point mutant male mice in multiple brain regions (Jiang et al., 2021; Boxer et al., 2020; Baker et al., 2013); these data suggest Gdf11 is robustly sensitive to MeCP2 function. Additionally, we also found that Gdf11 was significantly upregulated in mouse models overexpressing MECP2 (Figure 1B). We then investigated whether Gdf11 and Mecp2 expression were correlated among individual cell types in the brain. We mined a previous single-nucleus RNA-sequencing experiment in the normal mouse brain and found that the overall Spearman’s correlation between Gdf11 and Mecp2 across all cell types was 0.65 (Saunders et al., 2018). Further, Gdf11 and Mecp2 were highly correlated in neuronal and glial cell types (astrocytes, microglia, neurons, and oligodendrocytes; Spearman’s rho = 0.6–0.92) (Figure 1—figure supplement 1E). Within neurons specifically, we found that Gdf11 and Mecp2 were highly correlated in both excitatory neurons (Spearman’s rho = 0.65) and inhibitory neurons (Spearman’s rho = 0.76) (Figure 1—figure supplement 1F). Given the strong dynamic correlation of Gdf11 with MeCP2 levels upon ASO treatment and with Mecp2 expression in individual brain cell types (Figure 1A and Figure 1—figure supplement 1E and F), these data suggest Gdf11 is also sensitive to MeCP2 levels. Taken together, these results demonstrate Gdf11 is a transcript that is robustly sensitive to both MeCP2 function and protein levels.

To further investigate if MeCP2 transcriptionally regulates the Gdf11 locus, we took advantage of a Gdf11-knockout allele that has a LacZ reporter cassette knocked into the gene (Gdf11^tm2b; Figure 1C). We crossed this allele with both a Mecp2-knockout (Guy et al., 2001) and a MECP2 duplication (MECP2-TG1) (Collins et al., 2004) mouse model to generate a series of Mecp2 or Gdf11 single-mutants and Mecp2; Gdf11 double-mutants and performed qPCR for Mecp2, Gdf11, and LacZ in the cerebellum, a brain region where Gdf11 is highly expressed in the mouse (Lein et al., 2007; Saunders et al., 2018). The expression of Mecp2 was not changed by reduction of Gdf11, but Gdf11 was significantly decreased in Mecp2-knockout males (p<0.05) and increased in MECP2-TG1 animals (p<0.0001). Furthermore, Mecp2-knockout trended towards a further decrease of Gdf11 in the Mecp2^-/y; Gdf11^tm2b/+ double mutants while Gdf11 was restored to wild-type expression in the MECP2-TG1; Gdf11^tm2b/+ double mutants. Last, expression of the LacZ reporter was modulated by MeCP2 levels as well, being significantly downregulated in Mecp2-knockout (p<0.05) and upregulated in MECP2-TG1 animals (p<0.05) (Figure 1C). These results demonstrate that the Gdf11 locus is transcriptionally sensitive to MeCP2 protein levels.

While MeCP2 has a well-described role as a transcriptional repressor and its loss leads to a de-repression of gene expression, MeCP2 loss has been shown to cause a down-regulation of gene expression (Ben-Shachar et al., 2009; Boxer et al., 2020; Chahrour et al., 2008; Chen et al., 2015; Clemens et al., 2020; Lyst et al., 2013; Nan et al., 1998; Nan et al., 1997; Samaco et al., 2012). We next sought to mechanistically understand how loss of MeCP2 leads to down-regulation of Gdf11. We first investigated whether MeCP2 has enriched binding near the Gdf11 locus. We performed Cleavage Under Targets & Release Using Nuclease (CUT&RUN) (Skene and Henikoff, 2017) to profile MeCP2 binding in wild-type, Mecp2-knockout, and MECP2-TG1 mouse hippocampus (Figure 1—figure supplement 2A and B). We next identified peaks of enriched binding using the MACS algorithm (Zhang et al., 2008), which identified a peak upstream of the Gdf11 transcriptional start site (TSS). To further define the characteristics of this region, we mined existing histone acetylation data for marks of open chromatin (H3K27ac), promoter (H3K4me3), and putative enhancers (H3K4me1) (Halder et al., 2016). The peak of MeCP2 signal coincided with a region of high H3K27ac and H3K4me1, but low H3K4me3, suggesting that MeCP2 binds a putative enhancer of Gdf11 (Figure 1—figure supplement 2D). Due to the broad binding pattern of MeCP2 (Chen et al., 2015; Skene et al., 2010), we integrated the signal in a symmetric window straddling the peak to robustly quantify MeCP2 occupancy. We observed a significant (p<0.01) increase of MeCP2 binding within this region upstream of Gdf11 in the MECP2-TG1 hippocampus. Importantly, the MeCP2 binding signal was almost completely absent in the Mecp2-knockout hippocampus in this same region (Figure 1D). Concurrently, we profiled the repressive histone modification H3K27me3 (Figure 1—figure supplement 2C), which has been shown to interact with and be modulated by MeCP2 (Lee et al., 2020). Within the window of MeCP2 binding upstream of Gdf11, we found a significant (p<0.01) increase in H3K27me3 occupancy in Mecp2-knockout tissue and a trend towards a reciprocal decrease in H3K27me3 occupancy in MECP2-TG1 tissue (Figure 1E and Figure 1—figure supplement 2E). The increase in H3K27me3 in Mecp2-knockout tissue corroborates the downregulation we observe in Gdf11 expression in transcriptional profiles collected from Mecp2-deficient models (Figure 1B). These results suggest that MeCP2 binds upstream of Gdf11, at a putative enhancer, and modifies the chromatin landscape around the Gdf11 locus to regulate Gdf11 expression.

Since Gdf11 dysregulation is rescued upon normalization of MECP2 dosage, we sought to determine if normalization of Gdf11 alone can rescue behavioral deficits in MECP2 duplication mice. We crossed the MECP2-TG1 allele with the Gdf11^tm2b allele to normalize Gdf11 expression (Figure 1C). We generated a cohort of wild-type, MECP2-TG1, and MECP2-TG1; Gdf11^tm2b/+ double mutant littermates and behaviorally characterized these mice with an extensive battery of assays starting at 16 weeks of age (Figure 2—figure supplement 1A). Compared to wild-type mice, MECP2-TG1 mice are hypoactive when placed in an open arena, traveling less distance, and exhibiting fewer horizontal and vertical exploratory activities (Figure 2A). Normalizing Gdf11 expression in the double mutant mice resulted in rescue of total distance traveled and horizontal activity. We also observed a partial rescue of vertical exploration and an improvement in anxiety-like behavior in the open arena as assessed by entries into the center (Figure 2A and Figure 2—figure supplement 1B). We did not observe a difference in the ratio of the distance or time spent in the center of the arena (Figure 2—figure supplement 1B). Next, MECP2-TG1 mice experience hypoactivity and anxiety-like behavior compared to wild-type mice when placed in an elevated plus maze apparatus, traveling less distance and entering the open arms fewer times. In contrast, we observed the double mutant animals entered the open arms and crossed the maze arms like wild-type animals, and the double mutant mice traveled significantly further (p<0.01) on the maze compared to MECP2-TG1 mice (Figure 2B). Lastly, MECP2-TG1 mice exhibit improved learning compared to wild-type mice in a shock-tone cued fear paradigm. Specifically, mice are trained to associate a tone with a foot shock and then placed in the same environment without tone (context) or a new environment with a tone (cue). Learning in this assay is measured by the amount of time the mouse freezes, where MECP2-TG1 mice freeze more in both context and cued tests. Normalization of Gdf11 led to wild-type levels of freezing during the training routine, and normalized the cued learning compared to MECP2-TG1 mice (Figure 2—figure supplement 1C). Contextual learning was improved in double mutant mice with no significant difference observed compared to wild-type mice (Figure 2C). Reduction of Gdf11 did not rescue all phenotypes present in MECP2-TG1 mice, as we did not observe a rescue in motor coordination on the rotating rod, change in the time spent in dark during the light-dark assay, nor any changes in sociability. However, we did observe rescue of total activity, as measured by distance traveled, in both light-dark and social assays (Figure 2—figure supplement 1D–G). These data suggest that normalization of aberrant Gdf11 expression improves several behavioral deficits, most notably locomotion, general activity, and hippocampal learning in MECP2-TG1 mice.

The previous results raised the possibility that mouse neurological function is sensitive to changes in Gdf11 dosage. Supporting this evidence, loss of one copy of GDF11 function causes a developmental disorder with neurological features in humans (Ravenscroft et al., 2021). Furthermore, GDF11 is broadly and highly expressed in both the human and mouse brains, suggesting a potentially important role in brain function (Figure 3—figure supplement 1A and B; Li et al., 2017).

To determine if loss of one copy of Gdf11 resulted in neurobehavioral deficits, we generated a cohort of Gdf11^tm2b/+ mice to characterize with a comprehensive battery of tests. We first confirmed that both Gdf11^tm2b/+ and Gdf11^tm2b/tm2b mice exhibited abnormal anterior/posterior skeletal patterning by counting the number of attached and total ribs in neonatal pups. We observed that Gdf11^tm2b/tm2b pups had an abnormal number of 10 attached at 18 total ribs (n=4), while the Gdf11^tm2b/+ pups had an abnormal number of 8 attached and 14 total ribs (n=24), matching the phenotypes previously described (Figure 3—figure supplement 1C; McPherron et al., 1999). These skeletal abnormalities, the observed perinatal lethality of Gdf11^tm2b/tm2b pups (McPherron et al., 1999), and the 50% reduction of Gdf11 in the brains of Gdf11^tm2b/+ mice (Figure 1C) demonstrate that the Gdf11^tm2b allele is a null allele. Interestingly, we also observed the same skeletal patterning defects in MECP2-TG1; Gdf11^tm2b/+ pups, which had an abnormal number of 8 attached and 14 total ribs (n=4), while MECP2-TG1 pups had the same number of 7 attached and 13 total ribs (n=4) as wild-type pups (Figure 3—figure supplement 1C), further suggesting that the regulatory relationship between Gdf11 and Mecp2 is brain specific. Taken together, Gdf11^tm2b heterozygotes can be used to model the effects of losing one copy of Gdf11. Given these results, we generated a cohort of wild-type and Gdf11^tm2b/+ littermates and began general health assessments at weaning (Figure 3—figure supplement 2A). We observed that these mice did not display any overt phenotypes and did not have a change in body weight in either sex through sixteen weeks of age (Figure 3—figure supplement 2B).

We began deeper behavioral characterization of the cohort at sixteen weeks of age. In the open field assay, we observed that Gdf11^tm2b/+ mice are hyperactive as indicated by a significant increase in total distance traveled (p<0.05), a significant increase in horizontal activity counts (p<0.05), and a significant increase in vertical exploratory activity counts (p<0.01) (Figure 3A). Gdf11^tm2b/+ mice explored the center of the arena at a similar rate and ratio as wild-type mice (Figure 3—figure supplement 2C). In the elevated plus maze, Gdf11^tm2b/+ mice also displayed hyperactivity, traveling a significantly further distance (p<0.01) and decreased anxiety-like behavior, entering the open arm a significantly higher number of times (p<0.001; Figure 3B). Gdf11^tm2b/+ mice traveled a significantly further distance (p<0.05) in the light-dark assay while spending significantly more time (p<0.05) in the dark (Figure 3—figure supplement 2D). Furthermore, when placed in a three-chamber apparatus, Gdf11^tm2b/+ mice explored more both during habituation and testing phases (Figure 3C and Figure 3—figure supplement 2E). This manifests with a significantly (p<0.05) decreased time socializing compared to control (Figure 3C). The ratio of time spent investigating the partner mouse over the total investigation time was significantly (p<0.0001) less in Gdf11^tm2b/+ mice (Figure 3—figure supplement 2F), though importantly the Gdf11^tm2b/+ mice are still social and prefer interacting with the partner mouse over the object (p<0.001; Figure 3C). Next, we observed a significant (p<0.0001) reduction in motor coordination and learning as assessed by the rotating rod assay (Figure 3D). Finally, we observed a trend in reduction in learning during the training phase of a shock-tone conditioning paradigm (Figure 3—figure supplement 2G), while observing a significant reduction in contextual learning (p<0.01) and cued learning (p<0.05; Figure 3E). These data collectively suggest that neurological behaviors are sensitive to loss of one copy of Gdf11 in mice.

As Gdf11 has been implicated to naturally decrease in aging in mice (Poggioli et al., 2016), we tested if a reduction in Gdf11 dosage can alter aging by tracking the survival of a cohort of Gdf11^tm2b/+ mice. Strikingly, we found a significant (p<0.001) reduction in survival of Gdf11^tm2b/+ mice compared to wild-type littermates (n=15 wild-type and n=17 Gdf11^tm2b/+), exhibiting a median survival of 1.8 years (Figure 3F). Also, a portion Gdf11^tm2b/+ mice were euthanized due to reaching ethical endpoints including poor body conditioning, self-lesioning, and development of tumors, which suggests that Gdf11 dosage plays an important role in long-term health and survival in mice.

Learning and memory is linked to the generation of adult-born neurons in the subgranular zone (SGZ) of the mouse dentate gyrus (Kempermann and Gage, 2002). Complete loss of Gdf11 in adulthood was shown to increase proliferation in the SGZ, specifically in the number of amplifying progenitors (SOX2 positive), while impairing the overall number of newborn neurons that survive and integrate into the existing hippocampal circuitry (Mayweather et al., 2021). Additionally, neurogenesis is negatively impacted in the olfactory epithelium with loss or inhibition of GDF11 (Wu et al., 2003). To test if loss of one copy of Gdf11 altered overall proliferation in the SGZ, potentially acting as a mechanism for the changes in learning and memory observed in Gdf11^tm2b/+ mice, we labeled newborn, dividing cells by injecting Gdf11^tm2b/+ and control mice with 5-Ethyl-2’-deoxyuridine (EdU). We injected the animals with EdU (10 mg/kg) for 5 consecutive days and fixed the brains at 7 days after the first injection. We performed these injections in sixteen-week-old animals, matching the timepoint when we performed behavioral characterization. We then stained for EdU, SOX2, and GFAP to label proliferating cells (EdU⁺), neural stem cells (SOX2⁺GFAP⁺), and amplifying progenitors (SOX2⁺) (Cope and Gould, 2019; Zhao and Praag, 2020). We did not observe any significant changes in the number or density of these cell populations in Gdf11^tm2b/+ mice using quantitative stereology (Figure 4 and Figure 4—figure supplement 1A–D). Furthermore, no gross changes in brain anatomy or volume of the dentate gyrus were observed in Gdf11^tm2b/+ mice (Figure 4—figure supplement 1E–G). These results indicate that the behavioral deficits caused by loss of one copy of Gdf11 are not due to altered proliferation in the adult mouse SGZ.

Baylor College of Medicine Institutional Animal Care and Use Committee (IACUC, Protocol AN-1013) approved all mouse care and manipulation. Mice were housed in an AAALAS-certified level 3 facility on a 14 hr light cycle. Sperm from Gdf11^{tm2a(EUCOMM)Hmgu} mice was obtained from the MMRRC (Stock ID: 037721-UCD, RRID: MMRRC_037721-UCD). In vitro fertilization was performed with C57BL/6J donor eggs by the Genetic Engineering Mouse Core at Baylor College of Medicine. Gdf11^tm2a mice were maintained on C57BL/6J background. Gdf11^tm2b mice were generated by crossing male Gdf11^tm2a mice with Sox2-Cre female mice obtained from Jackson lab (Stock No: 008454; RRID: IMSR_JAX:008454). Gdf11^tm2b mice were maintained on C57BL/6J background. MECP2-TG1 and Mecp2-knockout mice were previously described (Collins et al., 2004; Guy et al., 2001), maintained on a C57BL/6J background, and available from Jackson lab (Stock Nos: 008679; RRID: IMSR_JAX:008679 and 003890; RRID: IMSR_JAX:003890). Gdf11 and MECP2 double mutant mice were generated by breeding male Gdf11^tm2b/+ mice to female MECP2-TG1 or Mecp2^+/- mice. Mice were monitored daily by veterinary staff. Survival age was taken as the age in which a mouse died of natural causes or when a mouse was euthanized due to poor body condition, self-lesioning, or tumor growth according to the institutional veterinarian guidelines. All procedures to maintain and use these mice were approved by the Institutional Animal Care and Use Committee for Baylor College of Medicine and Affiliates.

Total RNA was isolated from one cerebellar hemisphere using the Qiagen miRNeasy Mini kit (Qiagen #217004). On column DNAse digestion (Qiagen #79254) was performed according to manufacturer’s protocol to remove genomic DNA. Two μg of total RNA was used to synthesized cDNA using the M-MLV reverse transcriptase (Invitrogen #28025013) according to manufacturer’s protocol. qRT-PCR was performed using a CFX96 Real-Time System (Bio-Rad) using PowerUp SYBR Green Master Mix (ThermoFisher #A25741), 0.4 μM forward and reverse primers, and 1:20 dilution of cDNA. The following cycling conditions were used: 95 °C for 5 min, 39 cycles of 95 °C for 11 s, 60 °C for 45 s, plate read, a final melt of 95 °C, and melt curve of 65–95°C at +0.5 °C increments. The specificity of the amplification products was verified using melt-curve analysis. The Ct values were calculated with the Bio-Rad software, and relative gene expression was calculated using the ΔΔCt method using Ppia for normalization. All reactions were performed in technical duplicate with a minimum of three biological replicates. Data are presented as mean ± SEM in main figure.

All behavioral assays were performed during the light period. Mice were habituated to the test room for at least 30 min before each test. Mice were given at least one day to recover between different assays. All testing, data acquisition, and analyses were carried out by an individual blinded to the genotype. The order of assays is described in Figure 2—figure supplement 1 and Figure 3—figure supplement 2 beginning at approximately 16 weeks of age. The cohort details for each group of experiments are listed: (1) for MECP2-TG1; Gdf11^tm2b/+ versus wild-type and MECP2-TG1: n=30 wild-type mice (19 male, 11 female); n=26 MECP2-TG1 mice (14 male, 12 female); and n=22 MECP2-TG1; Gdf11^tm2b/+ mice (16 male, 6 female); and for fear conditioning assay (C), n=17 wild-type mice (9 male, 8 female); n=25 MECP2-TG1 mice (13 male, 12 female); and n=22 MECP2-TG1; Gdf11^tm2b/+ mice (16 male, 6 female) for (2) Gdf11^tm2b/+ versus wild type: n=29 wild-type mice (15 male, 14 female) and n=34 Gdf11^tm2b/+ mice (16 male, 18 female); and for fear conditioning assay, n=28 wild-type mice (16 male, 12 female) and n=24 Gdf11^tm2b/+ mice (9 male, 15 female). Outlier analysis was not performed on the data points, and no data points were removed from any assays.

The lighting in the test room was set to 750 lux, and the background noise level was set to 62 dB with a white noise generator. After habituation, the mice were placed in the center zone of a plus maze that has two open arms. The mice were placed facing one of the open arms and allowed to freely move for 10 min. The movement of the mice was tracked with the ANY-maze software system (Stoelting Inc). The total distance and distance traveled per zone, time spent in each arm zone, distance traveled in each arm zone, and number of zone crossings were recorded and tabulated by the software.

The lighting in the test room was set to 150 lux, and the background noise level was set to 62 dB with a white noise generator. After habituation, mice were placed in an open plexiglass area (40x40 x 30 cm), and their movement and behavior were tracked by laser photobeam breaks. Mice were allowed to freely move for 30 min. The total distance traveled, horizontal or vertical laser beam breaks (activity counts), entries into the center 10x10 cm, and time in the center 10x10 cm were recorded and tabulated by the AccuScan Fusion software (Omnitech Electronics Inc).

The lighting in the test room was set to 150 lux, and the background noise level was set to 62 dB with a white noise generator. After habituation, mice were placed in an plexiglass arena with an open zone (36x20 x 26 cm) and a closed, dark zone made of black plastic (15.5 x. 20x26 cm) with a 10.5 x. 5 cm opening. Mice were allowed to freely move for 10 min. The movement of the mice was tracked by laser photobeam breaks. The total and zone-specific distance traveled, the total or zone-specific horizontal or vertical laser beam breaks (activity counts), entries into the dark zone, and time spent in either zone were recorded and tabulated by the AccuScan Fusion software (Omnitech Electronics Inc).

The lighting in the test room was set to 150 lux, and the background noise level was set to 62 dB with a white noise generator. Age and sex matched C57BL/6J mice were used as novel partner mice, which were housed together and had no overtly aggressive behaviors observed. Two days before the test, the novel partner mice were habituated to the wire cups (3-inch diameter by 4-inch height) for 1 hr/day for 2 days. After habituation, the test mice were placed in a clear plexiglass chamber (24.75x16.75 x 8.75 inch) with two removable partitioned that divide the chamber into three zones. Empty wire cups were placed in the left and right chambers. For the acclimation phase, the mice were placed into the center zone and the partitions removed. The mice were allowed to freely explore the apparatus for 10 min. For the socialization phase, a novel partner mouse was randomly placed into the left or right wire cup. An inanimate object as control was plaed in the wire cup of the opposing chamber. The mice were allowed to explore the apparatus again for 10 minutes. The total distance traveled was tracked by the ANY-maze software system (Stoelting Inc). For all phases, the total amount of time the mice spent rearing, sniffing, or pawing the cup was recorded manually.

The lighting in the test room was at ambient levels, and the background noise level was set to 62 dB with a white noise generator. After habituation, mice were placed on an accelerating rotarod apparatus (Ugo Basile). The rod accelerated from 4 to 40 r.p.m. for 5 min. For cohorts containing the MECP2-TG1 allele (Figure 2 and Figure 2—figure supplement 1), the maximum trial length was 10 min in concordance with previous studies performing the rotarod assay in MECP2-TG1 mice and the observation that MECP2-TG1 mice perform above wild-type mice on this assay (Collins et al., 2004); for all other testing, the maximum trial length was 5 min as a standard starting assay length for an uncharacterized mouse line. Mice were tested four trials each day, with an interval of at least 30 min between trials, for four consecutive days. The time it took for each mouse to fall from the rod (i.e. latency to fall) was recorded.

Animals were habituated in an adjacent room to the testing room with a light level of 150 lux and noise level of 62 dB using a white noise generator. On the training day, mice were brought to the testing groom (ambient light, no noise) and placed into a chamber containing a grid floor that can deliver an electric shock with one mouse per chamber (Med Associates Inc). The chamber was housed within a sound-attenuating box with a digital camera, loudspeaker, and a light. The training paradigm consisted of 2 min of no noise or shock, then a tone for 30 s (5 kHz, 85 dB), ending with a foot shock for 2 s (1 mA). The tone and shock pairing was repeated after 2 min of no noise or shock. The mice were returned to their home cages after training. On the testing day, occurring 24 hr later, two tests are performed – context and cued learning. For the context test, the mice were placed in the exact same chamber with no noise or shock for 5 min. After 1-hr post-context test, the cued learning test was performed. For the cued learning test, the mice were placed in a novel environment for 6 min. The first 3 min with no noise or shock, and the last 3 min with tone only (5 kHz, 85 dB). Movement was recorded on video and freezing, as defined by absence of all movement except respiration, was quantified using the FreezeFrame software (ActiMetrics) using a bout duration of 1 s and movement threshold of 10.