Estrogen-related receptor β deficiency alters body composition and response to restraint stress

Background Estrogen-related receptors (ERRs) are orphan nuclear hormone receptors expressed in metabolically active tissues and modulate numerous homeostatic processes. ERRs do not bind the ligand estrogen, but they are able to bind the estrogen response element (ERE) embedded within the ERR response elements (ERREs) to regulate transcription of genes. Previous work has demonstrated that adult mice lacking Errβ have altered metabolism and meal patterns. To further understand the biological role of Errβ, we characterized the stress response of mice deficient for one or both alleles of Errβ. Results Sox2-Cre:Errβ mice lack Errβ expression in all tissues of the developing embryo. Sox2-Cre:Errβ+/lox heterozygotes were obese, had increased Npy and Agrp gene expression in the arcuate nucleus of the hypothalamus, and secreted more corticosterone in response to stress. In contrast, Sox2-Cre:Errβlox/lox homozygotes were lean and, despite increased Npy and Agrp gene expression, did not secrete more corticosterone in response to stress. Sox2-Cre:Errβ+/lox and Sox2-Cre:Errβlox/lox mice treated with the Errβ and Errγ agonist DY131 demonstrated increased corticotropin-releasing hormone (Crh) expression in the paraventricular nucleus of the hypothalamus, although corticosterone levels were not affected. Nes-Cre:Errβlox/lox mice, which selectively lack Errβ expression in the nervous system, also demonstrated elevated stress response during an acoustic startle response test and decreased expression of both Crh and corticotropin-releasing hormone receptor 2 (Crhr2). Conclusions Loss of Errβ affects body composition, neuropeptide levels, stress hormones, and centrally-modulated startle responses of mice. These results indicate that Errβ alters the function of the hypothalamic-pituitary-adrenocortical axis and indicates a role for Errβ in regulating stress response.


Background
ERRs are nuclear hormone receptors that regulate multiple homeostatic processes throughout life [1]. ERRs were initially identified on the basis of sequence homology to estrogen receptors (ERs) [2]. The homology between Errs and Ers is 36% in the ligand binding domain and 68% in the DNA binding domain. ERRs bind both ERR response elements (ERREs) and the closely related estrogen response elements (EREs) embedded within an ERRE sequence on DNA to modulate transcription of target genes [3][4][5][6][7][8]. Errs activate gene transcription by binding to DNA, either as a monomer, homodimer, or a heterodimer complex, which includes two different Err isoforms [1,6,7,9,10]. While their binding sites are similar to those of Ers, Errs do not bind estradiol and instead activate transcription in a ligandindependent manner, leading to their classification as orphan nuclear receptors. The three different Err genes, α, β and γ, have highly conserved ligand and DNA binding domains and thus may regulate homeostatic processes in a compensatory manner [11].
In mice, Errβ and Errγ are selectively expressed in the brain and multiple peripheral tissues [2,[12][13][14] and share the highest degree of sequence homology [11], suggesting that they may share overlapping functions. Since Errs recognize the same response elements, they are likely to regulate overlapping subsets of target genes [11].
We have previously reported that whole-body or central nervous system-specific deletion of Errβ increases expression of Errγ and ultimately alters body composition, metabolism, meal patterns, and energy expenditure of mice [11]. Further, inhibition of Errβ or Errγ alter metabolic parameters, whole-body energy balance (e.g. body composition, food intake and neuropeptide expression), while deletion of Errβ reciprocally modulates expression of Errγ (and vice versa) suggesting that balanced expression of Errβ and Errγ is important for control of energy balance and food intake [14][15][16][17][18].
Alterations in glucocorticoid signaling and whole-body energy balance positively correlate with one another, with increased glucocorticoid levels resulting in increased body weight [19][20][21]. Errβ suppresses glucocorticoid receptor activity in neuroblastoma and kidney cells in a dose-dependent manner, suggesting that it may also regulate metabolism at least in part through modulation of the hypothalamic-pituitary-adrenal (HPA) axis [22]. The HPA axis is regulated by corticotrophin-releasing hormone (Crh) released from neurosecretory cells of the hypothalamic paraventricular nucleus. Crh stimulates release of adrenocorticotropic hormone (ACTH) from the anterior pituitary, and ACTH, in turn, triggers glucocorticoid secretion from the adrenal gland. Negative feedback from ACTH and glucocorticoid secretion ultimately modulates Crh expression in the paraventricular nucleus via glucocorticoid receptors [23]. Disrupting glucocorticoid feedback loops can alter whole-body energy balance (e.g. body weight). Glucocorticoid excess (Cushing's disease) increases central fat deposition, whereas decreased body weight is associated with glucocorticoid insufficiency (Addison's disease) [19][20][21]. In addition to these effects on metabolism, alterations in the HPA axis can also influence anxiety and stress, which increase Neuropeptide Y (Npy) secretion. Npy further augments obesity susceptibility by inducing food intake and contributing to leptin resistance [23][24][25].
Consequently, we propose that Errβ modulates stress responses. Since Errβ suppresses glucocorticoid receptor activity [22], we hypothesized that the HPA axis may be altered in mice that carry heterozygous or homozygous loss of function mutations of Errβ in all somatic tissues [14,26,27]. The effects of Errβ deficiency on body weight, body composition, neuropeptide levels, stress hormones, and stress responses were examined in Sox2-Cre:Errβ +/lox and Sox2-Cre:Errβ lox/lox mice, in which Errβ expression is disrupted in all somatic tissues. These results indicate that Errβ modulates stress responses, at least in part through central mechanisms.

Results
Errβ gene dosage alters body weight and body composition Sox2-Cre:Errβ +/lox heterozygous mice express one allele of Errβ, resulting in higher levels of Errβ expression relative to Sox2-Cre:Errβ lox/lox homozygous mice. Alterations in energy balance are observed in mice deficient for Errβ in all embryonic tissues (Sox2-Cre:Errβ lox/lox ) [14]. Because Errβ is proposed to modulate energy balance in a dosedependent manner, we characterized Sox2-Cre:Errβ lox/lox and Sox2-Cre:Errβ +/lox mice to determine whether gene dosage altered development of body weight and body composition. We previously showed that Sox2-Cre:Errβ lox/lox mice have decreased body weight and fat mass by nine months of age [8]. Body weight and body composition (fat mass and lean mass) were measured in Sox2-Cre:Errβ lox/lox , Sox2-Cre:Errβ +/lox , and WT mice at three weeks and at nine months of age (Table 1). By three weeks, body composition differences began to emerge between the genotypes: Sox2-Cre: Errβ +/lox mice significantly increased fat mass (fat mass: F 1,8 = 9.32, P = 0.05), while Sox2-Cre:Errβ lox/lox mice trended toward decreased fat mass (fat mass: F 1,10 = 4.95, P = 0.05) compared to WT mice. There was no difference in body weight among the genotypes at three weeks, implying that alterations in body composition arise prior to weight changes in Errβ-deficient mice.
At nine months of age, Sox2-Cre:Errβ +/lox mice had increased fat mass and no change in lean mass relative to WT mice (fat mass: F 1,9 = 35.90, P = 0.002). However, Sox2-Cre:Errβ lox/lox mice demonstrated the opposite trend in body composition, with decreases in both fat and lean mass (fat free mass) relative to WT mice (fat mass: F 1,10 = 46.53, P < =0.0001; lean mass: F 1,10 = 6.21, P = 0.03). Accordingly, body weight increased in Sox2-Cre:Errβ +/lox mice (F 1,9 = 32.31, P = < 0.000001) and decreased in Sox2-Cre:Errβ lox/lox mice (F 1,10 = 32.57, P = 0.0004) relative to WT mice. Given these differences, the Sox2-Cre:Errβ +/lox mice surprisingly had a similar macrostructure of food intake as the Sox2-Cre:Errβ lox/lox [14], relative to WT mice. Specifically, after consuming a meal, the duration of time that the mouse was satiated was decreased (satiety ratio), the total number of pellets consumed was increased, and the duration of time between meals (intermeal interval, IMI) was not changed for Sox2-Cre:Errβ +/lox mice, but IMI was decreased for Sox2-Cre:Errβ lox/lox mice ( Table 1). The difference in IMI between the genotypes may be a compensatory change due to peripheral signals modulated by the increases in both body weight and fat mass observed in the Sox2-Cre:Errβ +/lox mice.

Hypothalamic neuropeptide expression in Errβ mutant mice
In the brain, Errβ is primarily expressed in the hindbrain, whereas Errγ is expressed in both the hindbrain and hypothalamus [14,28,29]. Nuclei of the hindbrain send primary projections to the hypothalamus (e.g., nucleus tractus solitarius to the paraventricular nucleus) and the amygdala, and activity in these nuclei can modulate hypothalamic gene expression [30][31][32]. Furthermore, in the absence of Errβ, Errγ can modulate food intake [14]. Since Sox2-Cre:Errβ +/lox and Sox2-Cre:Errβ lox/lox mice demonstrated alterations in body weight and body composition relative to WT mice, we sought to determine if hypothalamic neuropeptides known to modulate energy balance, Npy and Agrp, were differentially expressed in the brains of these mutants. Brain tissue sections of threeweek-old WT, Sox2-Cre:Errβ +/lox , and Sox2-Cre:Errβ lox/lox mice were hybridized with cRNA probes specific to Npy and Agrp mRNA. Npy ( Figure 1a) and Agrp (Figure 1b) staining were least intense in the hypothalamus of WT brain tissues, more intense in Sox2-Cre:Errβ +/lox brain tissues, and most intense in Sox2-Cre:Errβ lox/lox brain tissues. Expression of Npy and Agrp, as determined by hypothalamic ISH staining, appears to correlate inversely with Errβ expression. Increased staining expression of Npy and Agrp may contribute to the increased fat mass of three-week -old Sox2-Cre:Errβ +/lox mice; conversely, the high levels of Npy and Agrp in Sox2-Cre:Errβ lox/lox mice may be a downstream response to decreased fat mass.
Neural progenitor-specific deletion of Errβ alters acoustic startle response Sox2-Cre:Errβ +/lox and Sox2-Cre:Errβ lox/lox mice demonstrate differences in HPA activation, which may arise from central and/or peripheral mechanisms. In the central nervous system, Errβ expression is restricted to the hindbrain. Nes-Cre:Errβ lox/lox mice lack Errβ in neural progenitor cells, effectively resulting in selective loss of Errβ expression in the hindbrain [14]. Therefore, we investigated the central role of Errβ in modulating stress responses in Nes-Cre:Errβ lox/lox and WT mice using an acoustic startle test. The neuroanatomical and neurochemical basis of the acoustic startle response has been well mapped and involves neurons found in the amygdala, dorsomedial hypothalamus, and brainstem [35][36][37][38][39]. The amygdala elicits behavioral stress responses associated with the acoustic Figure 2 Glucocorticoid levels of wild-type (WT), Sox2-Cre:Errβ +/lox , and Sox2-Cre:Errβ lox/lox mice after restraint stress. a) Baseline, stress, and recovery glucocorticoid levels were measured in serum of WT mice and after treatment with Errβ/Errγ agonist DY131 using a corticosterone radioimmunoassay. b) Baseline, stress, and recovery glucocorticoid levels were measured in serum of Sox2-Cre:Errβ +/lox mice and after treatment with Errβ/Errγ agonist DY131 using a corticosterone radioimmunoassay. c) Baseline, stress, and recovery glucocorticoid levels were measured in serum of Sox2-Cre:Errβ lox/lox mice and after treatment with Errβ/Errγ agonist DY131 using a corticosterone radioimmunoassay. *P < 0.05. startle response and expresses the neuromodulators Crh and Npy [36,40]. Nes-Cre:Errβ lox/lox mice have decreased Npy expression in the hindbrain [14], which may modify neural circuitry activated by physical and psychological stress and, more specifically, the acoustic startle response.
We measured PPI and the acoustic startle response to determine if Nes-Cre:Errβ lox/lox mice had alterations in stress responses that arise from dysfunction of the inhibitory hindbrain circuit associated with PPI or the excitatory circuit associated with the acoustic startle response [41]. The acoustic startle response was measured after delivery of a prepulse intensity signal (0, 74, 78, 82, 86, or 90 dB) followed by the lead interval to a strong auditory stimulus. We observed a greater startle response in Nes-Cre:Errβ lox/lox mice (n = 8, db120; 1081.5 ±150) compared to WT mice (n = 12, db120; 475.8 ± 27) (Figure 4a However, the amplitude of the startle response decreased in Nes-Cre:Errβ lox/lox mice when the intensity of the prepulse tone increased. Crh expression was measured in the hindbrain of Nes-Cre:Errβ lox/lox mice and WT mice. Indeed, Nes-Cre:Errβ lox/lox mice have decreased expression of Crh and Crhr2 relative to WT (Figure 4b, F 1,10 = 6.54, P = 0.03 and Figure 4c, F 1,10 = 6.23, P = 0.03). These results indicate alterations in the excitatory pathway that generates a startle response, but not the inhibitory pathway arising from the pedunculopontine tegmental nucleus associated with PPI [41][42][43]. The increased acoustic startle response in Nes-Cre: Errβ lox/lox mice may thus arise from altered activity of the excitatory pathway involving Crh and Crhr2 expression and the pontine reticular nucleus, bed nucleus of the stria terminalis, amygdala, and hypothalamus [37][38][39][43][44][45][46][47][48]. The hindbrain excitatory pathways, which include catecholaminergic projections to the paraventricular nucleus of the hypothalamus, increase Crh expression in the hypothalamus, suggesting that hindbrain signaling may alter the HPA-axis feedback loop [49].

Discussion
ERRs are involved with energy balance and metabolism [14][15][16][17][18]. Using mice globally deficient for Errβ, we have shown that Errβ modulates body composition, stress signaling, and hypothalamic neuropeptide expression ( Table 2). Errβ gene dosage affected body composition and stress response with increased fat mass and corticosterone levels in Sox2-Cre:Errβ +/lox mice and decreased fat mass and corticosterone levels in Sox2-Cre:Errβ lox/lox mice (Table 1 and Figure 2). Additionally, central nervous system-specific Errβ deletion alters stress associated with the acoustic startle response pathways ( Figure 4).
Hypothalamic expression of Npy and Agrp, orexigenic factors that increase fat mass and food intake [50][51][52], increased in both Sox2-Cre:Errβ +/lox and Sox2-Cre: Errβ lox/lox mice (Figure 1). These results suggest that increased anabolic neuropeptide expression may be due to central or peripheral mechanisms that are activated following global deletion of Errβ. Increased Npy and Agrp expression may be due to differences in leptin levels from adipose mass. Increased fat mass and lean mass were measured in Sox2-Cre:Errβ +/lox mice, although decreased fat mass and lean mass were measured in highly-active Sox2-Cre:Errβ lox/lox mice at nine months of age (Table 1). Expression of Npy and leptin are coordinately regulated, as Npy blunts the effects of leptin and increased leptin levels decrease Npy expression [23,[53][54][55]. Thus, Sox2-Cre:Errβ +/lox mice may consume more food and increase Npy expression and fat mass due to leptin resistance; Sox2-Cre:Errβ lox/lox mice may increase Npy expression to compensate for decreased fat mass arising from increased physical activity Figure 3 Crh expression of wild-type (WT), Sox2-Cre:Errβ +/lox , and Sox2-Cre:Errβ lox/lox mice. Brain tissue of WT, Sox2-Cre:Errβ +/lox , and Sox2-Cre:Errβ lox/lox mice injected with saline (top) or Errβ/Errγ agonist DY131 (bottom) were stained for Crh by ISH (n = 2/genotype).
( Figure 1 and Table 1). In support of this, Nes-Cre:Errβ lox/lox mice have increased lean mass, no change in physical activity and have decreased Npy expression in the hindbrain [14]. Changes in body composition emerged prior to changes in body weight, suggesting that both peripheral and central signals may be altered to regulate the development of increased fat mass ( Table 1).
The opposite phenotypes that are seen in the Sox2-Cre: Errβ +/lox and Sox2-Cre:Errβ lox/lox mice may arise from the ability of Errβ or Errγ to regulate gene transcription as both homodimers and Errβ/Errγ heterodimers [1,6,7,9,10]. Errβ/Errγ heterodimers have been predicted to exist, but to our knowledge it has not been directly detected in vivo [1]. RIP140 is a nuclear receptor corepressor that regulates cellular metabolism [56][57][58]. RIP140 enhanced transcriptional activity for all three mouse Err genes [59]. Mice lacking RIP140 are lean, with increased metabolic rate and insulin sensitivity [58]. Similarly, Sox2-Cre:Errβ lox/lox mice are lean with increased metabolic rate (Table 1 and [14]), and Nes-Cre:Errβ lox/lox have increased lean mass, increased metabolic rate and insulin sensitivity [14]. Since deletion of both Errβ and RIP140 exhibit similar phenotypes, this suggests that increased lean mass relative to fat mass, metabolic rate and insulin sensitivity may arise from both the RIP140 corepressor and Errβ [59].
Crh is expressed in the paraventricular nucleus of the hypothalamus and initiates ACTH release from the pituitary [40,65]. Crh has since been found to be synthesized in extra-hypothalamic sites, where it also acts to modulate stress response and food intake [40,[65][66][67]. ERR family members also modulate stress responses by regulating glucocorticoid receptor activity in muscle and neuroblastoma cell lines [22,33]. Further, Errβ and Crh are expressed in similar regions of the hindbrain [29]. Here we demonstrate that Errβ deletion modulates corticosterone levels after exposure to restraint stress, with increased levels in Sox2-Cre:Errβ +/lox mice and decreased levels in Sox2-Cre: Errβ lox/lox mice relative to WT (Figure 2). Neural connections projecting to the hypothalamus from extrahypothalamic sites, such as the hindbrain, may also regulate hypothalamic Crh release and Crh expression [30,49,[68][69][70].
Biological activity of Crh is inhibited by Crhbp, and Errβ binds to the promoter region of the Crhbp gene [60,71], which contains three ERE half sites [72]. Mice that overexpress Crhbp have increased Crh expression, potentially resulting from a compensatory response aimed at ameliorating disruptions in stress response [73]. Similarly, increased Crh expression was observed when Errβ was reduced (Sox2-Cre:Errβ +/lox ) or eliminated (Sox2-Cre:Errβ +/lox ) in somatic tissue, and Errγ was activated using DY131 ( Figure 3). Therefore, we propose that partial or complete deletion of Errβ may alter Crh expression by modulating transcription of Crhbp or Crhr2, resulting in altered corticosterone secretion. Furthermore, Sox2:Errβ lox/lox mice lack corticosterone secretion after restraint stress (Figure 2), which may result from altered Crhr2 expression (Figure 4c) and changes in negative feedback. Therefore, brain regions that express Crhr2 may show reduced Crh signaling (Figure 4b and 4c), as in the hindbrain [64].
Errβ binds to cis-regulatory regions of the Cckbr gene [60], which is expressed in the hindbrain [29,74] and the corresponding gene maps to a genomic locus of the genome associated with obesity [75]. Cckbr deficient mice (Cckbr −/− ) display a similar phenotype to Sox2-Cre: Errβ +/lox mice, and have increased body weight and food intake, which may arise from changes in Cholecystokinin (Cck) signaling (e.g. satiety), and increased metabolism [74,76]. However, Cckbr −/− mice also have blunted stress responses associated with anxiety-like behavior [77] and increased Npy expression [78], which resembles the phenotype of Sox2-Cre:Errβ lox/lox mice ( Figure 1 and Table 1). Therefore, heterodimers of Errβ alone, or Errβ in combination with ERRγ, may regulate Cckbr transcription, thereby partially accounting for the differences in the phenotypes seen in Sox2-Cre:Errβ +/lox and Sox2-Cre:Errβ lox/lox mice ( Table 2). Differences in developmental compensation arising from Errβ and/or Errγ may also contribute to the phenotype differences in Sox2-Cre: Errβ +/lox and Sox2-Cre:Errβ lox/lox mice.
Nes-Cre:Errβ lox/lox mice show increased Errγ expression relative to WT animals [14], while mice deficient for Errγ show increased Errβ expression [17]. This suggests that homozygous mice have reciprocal patterns of Errβ and Errγ expression, potentially arising from developmental compensation and heterozygous mice may partially lack this compensation, contributing to phenotype differences. The Errβ/Errγ agonist (DY131) increased Crh expression more when Errβ expression was reduced (Sox2:Errβ +/lox mice) than when Errβ expression was absent (Sox2:Errβ lox/lox mice) (Figure 3). These results suggest that the ratio of Errβ to Errγ signaling may contribute to the observed difference in Crh expression, Crhr2 expression and corticosterone secretion in the two genotypes examined.
Our data suggest that central Errβ modulates stress responses, food intake and body weight, although it remains to be determined whether peripheral Errβ also modulates components of the HPA axis and acoustic startle response. Nes-Cre:Errβ lox/lox mice lack Errβ in the hindbrain and have decreased expression of Crh, Crhr2 and Npy [14], suggesting that neuromodulators involved with the acoustic startle response reside in the hindbrain to modulate stress and anxiety. However, other changes in neural circuitry (e.g. altered Cckbr expression) regulating the acoustic startle response in Nes-Cre:Errβ lox/lox mice are likely to exist and remain to be identified.

Conclusions
Mice heterozygous for Errβ deletion have increased fat mass and stress hormone secretion after restraint stress, while those homozygous for Errβ deletion have decreased fat mass and secrete higher baseline levels of stress hormones. These effects may be modulated by components of the HPA axis, such as Crh, Crhbp, Crhr2, Npy or Cckrb. Central Errβ signaling influences stress associated behavior (e.g. the acoustic startle response), possibly through regulation of Npy, Crh and Crhr2 expression in the hindbrain or hypothalamic projections to the amygdala [32,62,63,80]. Since the neural circuitry controlling the acoustic startle response is well-conserved between rodents and humans [36,81], these data suggest that ERRβ or ERRγ may be promising candidates for pharmacological treatment of excessive anxiety or stress levels in humans.

Methods
Animals, housing, food intake, and physical activity measurement Sox2-Cre:Errβ lox/lox , Sox2-Cre:Errβ +/lox , and wild-type (WT) (Errβ lox/lox ) mice were generated as previously described [26]. Briefly, Errβ mice have a conditional allele, with loxP sites flanking exon 2 of the Errβ gene that encodes the DNA binding domain (exon 2) [26]. Expression of cre recombinase will excise the loxP-flanked exon 2 from the Errβ gene. Sox2-Cre deletes Errβ from all embryonic tissues and Nestin-Cre deletes Errβ from developing neural tissue. Sox2-Cre:Errβ lox/lox mice completely lack functional Errβ because both alleles have been removed. Sox2-Cre:Errβ +/lox have one wild-type allele of the Errβ gene, since the other allele has been excised by the loxP sites. These two mouse lines enable us to address possible phenotypic differences due to differences in gene dosage. Wild-type (WT) mice used for these studies were homozygous for the floxed Errβ allele. Mice were maintained on a 12:12 hour light-dark cycle in a temperature-and humidity-regulated vivarium and had ad libitum access to standard laboratory chow (2018, Harlan-Teklad, Harlan Laboratories, Frederick, MD, USA) and water at all times. Different cohorts of mice were analyzed at three weeks and nine months of age. Food intake data and physical activity levels were collected as previously described [14]. Physical activity levels were measured by detecting and counting horizontal beam breaks in a 40 cm × 40 cm × 30 cm plexiglass chamber (Digiscan, Accuscan Instruments, Columbus, OH). All experimental procedures were performed in accordance with the Johns Hopkins University School of Medicine Institutional Animal Care and Use Committee and the National Institutes of Health Guide for the Care and Use of Laboratory Animals.

Restraint stress test, corticosterone radioimmunoassay, and DY131 injections
Baseline blood glucocorticoid levels were measured and mice were placed into a restraining tube (one mouse/tube) for one hour. Upon removal from the restraining tube, blood samples were collected again. Animals were then returned to their housing and blood samples were collected after a one-hour recovery period. Blood was collected in heparin-coated tubes and centrifuged at 3800 rpm for 20 min at 4°C. Corticosterone assays were performed with a radioimmunoassay kit for corticosterone per manufacturer's directions (MP Biomedicals, Solon, OH, USA). DY131 (Tocris, Bristol, BS11, United Kingdom) at a dose of 10 μM/g body weight was injected, and data for meal patterns collected as previously described [14].

Prepulse inhibition (PPI) of acoustic startle response
Startle reactivity and PPI were measured using two startle chambers located inside a sound-attenuating chamber (San Diego Instruments, San Diego, CA, USA). Mice were placed in a Plexiglass tube within the soundproof PPI box for a five-minute acclimation period, which provides exposure to a continuous background noise (70 dB) to elicit an increase in startle amplitude [43]. Mice were then exposed for five minutes without any startle stimulus. The PPI session then began and mice were randomly exposed to the following trials: pulse alone (120 dB), no stimulus, or five prepulse combinations (a 20 ms non-startling prepulse at 74, 78, 82, 86, or 90 dB, followed by an 80 ms startle stimulus at 120 dB). The force from the startle reaction was recorded by an accelerometer with SR-LAB software (San Diego Instruments). Results were analyzed by PPI percentage, which was calculated as: mean startle amplitude on pulse alone ð Þ − mean startle amplitude on prepulse ð Þ = mean startle amplitude on pulse alone ð Þ :

Statistical analysis
All value comparisons were made using one-way ANOVA to identify individual differences between groups, and P < 0.05 was considered significant (Statistica v.8.0, Tulsa, OK, USA).